Calculus For Engineers, 4th Edition

Calculus for Engineers
Fourth
Edition
Calculus for Engineers
Fourth
Edition
Donald Trim
UNIVERSITY OF MANITOBA
Toronto
Library and Archives Canada Cataloguing in Publication
Trim, Donald W.
Calculus for engineers/Donald Trim.—4th ed.
Includes index.
ISBN-13 978-0-13-157713-8
ISBN-10 0-13-157713-1
1. Calculus—Textbooks. 2. Engineering mathematics—Textbooks. I. Title.
QA303.2.T75 2008 515 C2006-905116-X
Copyright © 2008, 2004, 2001, 1998 Pearson Education Canada, a division of Pearson
Canada Inc., Toronto, Ontario.
Pearson Prentice Hall. All rights reserved. This publication is protected by copyright, and
permission should be obtained from the publisher prior to any prohibited reproduction,
storage in a retrieval system, or transmission in any form or by any means, electronic,
mechanical, photocopying, recording, or likewise. For information regarding permission,
write to the Permissions Department.
ISBN-13 978-0-13-157713-8
ISBN-10 0-13-157713-1
Editor-in-Chief: Gary Bennett

Marketing Manager: Colleen Gauthier
Acquisitions Editor: Michelle Sartor
Developmental Editor: Kimberley Hermans
Production Editor: Marisa D’Andrea
Copy Editor: Gail Marsden
Proofreader: Margaret Bukta
Production Coordinator: Avinash Chandra
Page Layout: PreTeX, Inc.
Art Director: Julia Hall
Interior Design: Geoff Agnew
Cover Design: Maki Ikushima
Cover Image: First Light
12345 12 11 10 09 08
Printed and bound in United States.

In memory of my father, who,
like his son, was a man of few words.
Our love was expressed in other ways.
!
Brief Contents
Preface xvi
Chapter 1 Calculus Preparation 1
Chapter 2 Limits and Continuity 104
Chapter 3 Differentiation 149
Chapter 4 Applications of Differentiation 237
Chapter 5 The Indefinite Integral and the Antiderivative 335
Chapter 6 The Definite Integral 374
Chapter 7 Applications of the Definite Integral 410
Chapter 8 Techniques of Integration 490
Chapter 9 Parametric Equations and Polar Coordinates 541
Chapter 10 Infinite Sequences and Series 587
Chapter 11 Vectors and Three-Dimensional Analytic Geometry 695
Chapter 12 Differential Calculus of Multivariable Functions 799
Chapter 13 Multiple Integrals 896
Chapter 14 Vector Calculus 982
Chapter 15 Differential Equations 1047
Appendix A Mathematical Induction A–1
Appendix B Determinants B–1
Appendix C Complex Numbers C–1
Answers to Even-Numbered Exercises E–1
Index I–1
vii
Contents
Preface xvi
Chapter 1 Calculus Preparation 1

1.1 Introduction 1
1.2 Solving Polynomial Equations 2
1.3 Plane Analytic Geometry and Straight Lines 6
1.4 Conic Sections 14
1.5 Functions and Their Graphs 29
1.6 Inverse Functions 45
1.7 Trigonometry Review 50
1.8 The Inverse Trigonometric Functions 64
1.9 Exponential and Logarithm Review 76
1.10 Hyperbolic Functions 87
1.11 Approximating Solutions to Equations 91
Summary 99
Key Terms 101
Review Exercises 102
Chapter 2 Limits and Continuity 104

Application Preview 104
2.1 Limits 105
2.2 Infinite Limits 116
2.3 Limits at Infinity 119
2.4 Continuity 128
2.5 Heaviside and Dirac-Delta Functions 136
Consulting Project 1 139
Application Preview Revisited 141
2.6 A Mathematical Definition of Limits 143
Summary 147
Key Terms 147
viii
Contents ix
Chapter 3 Differentiation 149

3.1 The Derivative 149
3.2 Rules for Differentiation 159
3.3 Differentiability and Continuity 167
3.4 Product and Quotient Rules 172
3.5 Higher-Order Derivatives 178
3.6 Velocity, Speed, and Acceleration 180
3.7 The Chain Rule and the Extended Power Rule 187
3.8 Implicit Differentiation 192
3.9 Derivatives of the Trigonometric Functions 200
3.10 Derivatives of the Inverse Trigonometric Functions 208
3.11 Derivatives of the Exponential and Logarithm Functions 211
3.12 Logarithmic Differentiation 223
3.13 Derivatives of the Hyperbolic Functions 226
3.14 Rolle’s Theorem and the Mean Value Theorems 229
Summary 232
Key Terms 234
Chapter 4 Applications of Differentiation 237

4.1 Newton’s Iterative Procedure for Solving Equations 237
4.2 Increasing and Decreasing Functions 244
4.3 Relative Maxima and Minima 248
4.4 Concavity and Points of Inflection 256
4.5 Drawing Graphs with Calculus 263
4.6 Analyzing Graphs with Calculus 270
4.7 Absolute Maxima and Minima 274
4.8 Velocity and Acceleration 293
4.9 Related Rates 299
4.10 LCR-Circuits 309
4.11 Indeterminate Forms and L’Hôpital’s Rule 317
4.12 Differentials 327
Summary 332
Key Terms 333
x Contents
Chapter 5 The Indefinite Integral and the Antiderivative 335

5.1 The Reverse Operation of Differentiation 336
5.2 Integrating Velocity and Acceleration 344
5.3 Change of Variable in the Indefinite Integral 349
5.4 Deflection of Beams 352
5.5 An Introduction to Separable Differential Equations 359
Summary 371
Key Terms 372
Chapter 6 The Definite Integral 374

6.1 Sigma Notation 374
6.2 The Need for the Definite Integral 380
6.3 The Definite Integral 384
6.4 The First Fundamental Theorem of Integral Calculus 387
6.5 The Second Fundamental Theorem of Integral Calculus 393
6.6 Average Values 399
6.7 Change of Variable in the Definite Integral 404
Summary 407
Key Terms 408
Chapter 7 Applications of the Definite Integral 410

7.1 Area 410
7.2 Volumes of Solids of Revolution 419
7.3 Lengths of Curves 431
7.4 Work 436
7.5 Energy 443
7.6 Fluid Pressure 447
7.7 Centres of Mass and Centroids 454
7.8 Moments of Inertia 465
7.9 Additional Applications 470
7.10 Improper Integrals 479
Summary 487
Key Terms 488
Contents xi
Chapter 8 Techniques of Integration 490

8.1 Integration Formulas and Substitutions 490
8.2 Integration by Parts 494
8.3 Trigonometric Integrals 500
8.4 Trigonometric Substitutions 507
8.5 Completing the Square and Trigonometric Substitutions 514
8.6 Partial Fractions 517
8.7 Numerical Integration 528
Summary 538
Key Terms 539
Chapter 9 Parametric Equations and Polar Coordinates 541

9.1 Parametric Equations 541
9.2 Polar Coordinates 554
9.3 Curves in Polar Coordinates 557
9.4 Areas in Polar Coordinates 566
9.5 Definitions of Conic Sections 569
9.6 Conic Sections in Polar Coordinates 578
Summary 584
Key Terms 585
Chapter 10 Infinite Sequences and Series 587

10.1 Infinite Sequences of Numbers 588
10.2 Sequences of Functions 598
10.3 Taylor Polynomials, Remainders, and Series 600
10.4 Power Series 611
10.5 Taylor Series Expansions of Functions 620
10.6 Sums of Power Series 634
10.7 Applications of Taylor Series and Taylor’s Remainder Formula 636
10.8 Convergence of Sequences of Numbers 646
10.9 Infinite Series of Numbers 656
10.10 Integral, Comparison, and Limit Comparison Tests 665
xii Contents
10.11 Limit Ratio and Limit Root Tests 672

10.12 Absolute and Conditional Convergence, Alternating Series 677
10.13 Exact and Approximate Values for Sums of Series of Numbers 683
Summary 691
Key Terms 692
Chapter 11 Vectors and Three-Dimensional Analytic Geometry 695

11.1 Rectangular Coordinates in Space 695
11.2 Curves and Surfaces 698
11.3 Vectors 707
11.4 Scalar and Vector Products 724
11.5 Planes and Lines 731
11.6 Geometric Applications of Scalar and Vector Products 735
11.7 Physical Applications of Scalar and Vector Products 741
11.8 Hanging Cables 749
11.9 Differentiation and Integration of Vectors 755
11.10 Parametric and Vector Representations of Curves 761
11.11 Tangent Vectors and Lengths of Curves 765
11.12 Normal Vectors, Curvature, and Radius of Curvature 770
11.13 Displacement, Velocity, and Acceleration 780
Summary 794
Key Terms 796
Chapter 12 Differential Calculus of Multivariable Functions 799

12.1 Multivariable Functions 799
12.2 Limits and Continuity 803
12.3 Partial Derivatives 807
12.4 Gradients 810
12.5 Higher-Order Partial Derivatives 813
12.6 Chain Rules for Partial Derivatives 822
12.8 Directional Derivatives 839
12.9 Tangent Lines and Tangent Planes 844
12.12 Lagrange Multipliers 868
12.13 Least Squares 880
Contents xiii
12.15 Taylor Series for Multivariable Functions 890

Summary 893
Key Terms 894
Chapter 13 Multiple Integrals 896

13.1 Double Integrals and Double Iterated Integrals 896
13.2 Evaluation of Double Integrals by Double Iterated Integrals 900
13.3 Areas and Volumes of Solids of Revolution 906
13.5 Centres of Mass and Moments of Inertia 916
13.6 Surface Area 925
13.7 Double Iterated Integrals in Polar Coordinates 930
13.8 Triple Integrals and Triple Iterated Integrals 940
13.9 Volumes 945
13.11 Triple Iterated Integrals in Cylindrical Coordinates 955
13.12 Triple Iterated Integrals in Spherical Coordinates 959
13.13 General Transformations in Multiple Integrals 965
13.14 Derivatives of Definite Integrals 972
Summary 978
Key Terms 979
Chapter 14 Vector Calculus 982

14.1 Vector Fields 983
14.2 Line Integrals 992
14.3 Line Integrals Involving Vector Functions 998
14.4 Independence of Path 1005
14.5 Energy and Conservative Force Fields 1012
14.6 Green’s Theorem 1016
14.7 Surface Integrals 1022
14.8 Surface Integrals Involving Vector Fields 1027
14.9 The Divergence Theorem 1033
14.10 Stokes’s Theorem 1039
Summary 1043
Key Terms 1045
xiv Contents
Chapter 15 Differential Equations 1047

15.1 Introduction 1047
15.2 Separable Differential Equations 1053
15.3 Linear First-Order Differential Equations 1061
15.4 Second-Order Equations Reducible to Two First-Order Equations 1067
15.5 Newtonian Mechanics 1072
15.6 Linear Differential Equations 1081
15.7 Homogeneous Linear Differential Equations 1084
15.8 Homogeneous Linear Differential Equations with Constant Coefficients 1087
15.9 Nonhomogeneous Linear Differential Equations with Constant Coefficients 1092
15.10 Applications of Linear Differential Equations 1106
Summary 1118
Key Terms 1119
Appendix A Mathematical Induction A–1
Appendix B Determinants B–1
Appendix C Complex Numbers C–1
Answers to Even-Numbered Exercises E–1
Index I–1
About the Author xv
About the Author

Donald W. Trim received his honours degree in mathematics and physics in 1965. After pro-
ceeding to his masters degree, he obtained his doctorate degree with a dissertation in general
relativity. As a graduate student at the University of Waterloo, he discovered that teaching was
his life’s ambition. He quickly became well known for his teaching. In 1971, he was invited to
become the first member of the Department of Applied Mathematics in the Faculty of Science
at the University of Manitoba. In 1976, after only five years at the university, he received one
of the university’s highest awards for teaching excellence based on submissions by faculty and
graduating students. At the University of Manitoba, no individual can receive this award twice
within a ten-year span. Professor Trim won the award in 1988 and again in 2006. In 1997, he
was awarded a 3M teaching fellowship. In addition, the Faculty of Engineering presented him
with a gold replica of the engineer’s iron ring in appreciation for his service.
Professor Trim’s skills in teaching are reflected in the notes and books he has written. These
include Introduction to Applied Mathematics, Introduction to Complex Analysis and Its Appli-
cations, Applied Partial Differential Equations, and this calculus text. All have received most
favourable reviews from students. To bridge the gap between calculus and partial differential
equations, he is currently writing a book on ordinary differential equations.
Preface
A Special Focus
This text has been written especially for students in engineering and the physical sciences, but
students from other disciplines might well find this book appealing. Like every instructor, I
have developed certain views on the best pedagogical approach to calculus. These views are
often independent of the area of study of the student.
It is the choice of examples and exercises that help to focus a text for students in a specific
discipline. Throughout this book, I have chosen to emphasize the physical applications of
calculus. To this end, I have consulted engineers and physicists, and I have searched the literature
carefully for physically meaningful examples and exercises. At the same time, there are many
examples and exercises from the social sciences, business, medicine, and others.
Three categories of applications recur throughout the book to demonstrate the indispensi-
bility of calculus in engineering and the physical sciences:
• Velocity, speed, and acceleration play a major role in many physical systems. In Chapter 3,
velocity and acceleration are introduced as derivatives, and then given a fuller discussion
in Chapter 4. They are treated again from an integration point of view in Chapter 5.
Instead of expecting the use of standard physics formulas, I encourage students to state time
conventions, choose coordinate systems, and develop whatever equations are necessary to
solve a particular problem.
• Resistors, capacitors, and inductors and how they relate to charge and current are introduced
through derivatives in Chapter 3. As calculus unfolds, they are further developed through
antiderivatives.
• Differential equations provide mathematical models for a wide variety of applications.
They are introduced briefly in Chapter 3 so that students become familiar with the ideas
surrounding this important branch of calculus. Separable differential equations are treated
fully in Chapter 5 along with a variety of applications. Other types of differential equations
are discussed in detail in Chapter 15.
Other Features
A careful effort has been made to incorporate features that will enhance applications and facilitate
learning.
• Chapter 1 provides a review of precalculus material (in addition to some non-review topics
that are indispensable in some applications). Each review section begins with a diagnostic
test for students to test their familiarity with the material. Should their test score be less
than adequate, they are advised to study the material in the section, work the exercises, and
retake the test. Even with acceptable test scores, students are advised to at least read all
review sections.
xvi
Preface xvii
• Electronic calculators or software packages such as Maple, Mathematica, MathCad, and

MatLab are now commonplace in many engineering and science courses. And the use of
these devices is becoming more and more prevalent in the workplace. Accordingly, I have
taken advantage of the plotting facilities of calculators and computers in this book. At the
same time, a great disparity exists among calculators and computers with respect to taking
limits, differentiating, and integrating, and I have not therefore written a totally integrated
text that uses these devices to do these operations.
• Numerous Examples with worked-out solutions are given throughout the book.
• Key terms are boldfaced where they are defined in the text, and they are listed near the end
of each chapter.
• Important Definitions, Theorems, and Corollaries are highlighted where they are introduced
in the text.
• A special icon, consisting of an exclamation mark in a triangular sign, is placed in the
margin to alert the reader to pitfalls that commonly entrap students. Such pitfalls may be
inappropriate calculations or misinterpretations of ideas.
• A concise Summary is given near the end of each chapter.

• A list of the main Derivative and Integration Formulas is given on the inside of the back
cover for easy reference. Some Geometric Formulas and Trigonometric Identities can be
found inside the front cover.
• The Exercises at the ends of the sections and the Review Exercises at the ends of the
chapters have been graded into three difficulty levels. Exercises with no asterisk are the
easiest. They are routine problems designed to reinforce fundamentals. Exercises with one
asterisk (∗ ) demand more thought. This grade of exercise contains most of the applications.
Fundamentals should be mastered before tackling applications to problems from other fields.
Exercises with two asterisks (∗∗ ) are the most challenging. They should be attempted only
after most of the single-asterisk problems have been successfully solved.
• Some Exercises require the use of a calculator or computer. This may be due to intensive
calculations with unwieldy numbers or because a detailed plot of a function is needed.
These Exercises are marked with a calculator icon, . I stress the need for students to be
able to draw graphs without the need of technology, but at the same time, I suggest they
check their work with a computer plot. These Exercises are not given a calculator icon
designation.
• Answers to Even-Numbered Exercises, except the challenging two-asterisk ones, are pro-
vided after the Appendices.
• Three Appendices are included near the back of the book: Appendix A on Mathematical
Induction, Appendix B on Determinants, and Appendix C on Complex Numbers.
Approach
For the most part, the approach in this book is intuitive, making free use of geometry and
familiar physical settings to motivate and illustrate concepts. For instance, limits are introduced
intuitively, but an optional section is included to allow exposure to the mathematical definition
of a limit. Derivatives are introduced and defined algebraically as instantaneous rates of change,
and interpreted geometrically as slopes of tangent lines. The definite integral is introduced
through two geometric and two physical problems, but its definition is algebraic.
In Sections 1.3 and 1.4, fundamentals of plane analytic geometry are reviewed in order to
provide a way of visualizing problems. The most difficult part of a solution to a problem is
frequently the initial step. Once started, the solution often unfolds smoothly and easily, but
that first step sometimes seems impossible. One of the best ways to start a problem is with a
diagram. A picture, no matter how rough, is invaluable in giving a “feeling” for what is going
on. It displays the known facts surrounding the problem; it facilitates “seeing” what the problem
xviii Preface
really is and how it relates to the known facts; and it often suggests that all-important first step.
Students should be encouraged to develop the habit of making diagrams at every opportunity
— not just to solve problems, but to understand what calculus is all about.
The main goal of this book is to help students think logically. I do not provide an exhaustive
list of formulas and procedures to memorize so that students can solve problems by rote. Instead,
I provide a few formulas and, I hope, a great deal of insight into the ideas surrounding these
formulas. By working on the examples and problems in this book, students learn how to organize
facts and interpret them mathematically. They learn how to decide exactly what a particular
problem is, and how to produce a step-by-step procedure by which to solve it.
Order of Topics
The order of topics in this book is fairly standard. The main exceptions are the following:
• Conic sections are introduced in Chapter 1 as illustrations of how the form of the equation
of a curve dictates its shape and conversely how the shape of a curve influences its equation.
Detailed discussions of properties of conic sections are delayed until Chapter 9.
• From the outset, we address what it means for x0 to approximate a solution of the equation
f (x) = 0 to a given number of decimal places or with error less than some specified
amount. Approximations are introduced in a geometric way in Chapter 1 through the Zero
Intermediate Value Theorem. Newton’s iterative procedure for approximating solutions to
equations is the first application of differentiation in Chapter 4. Numbers in subsequent
applications need not then be contrived in order to lead to equations with elementary solu-
tions. Throughout, I demand that the accuracy of approximations be clearly identified.
• Trigonometric, exponential, and logarithm functions are reviewed in Chapter 1. Inverse
trigonometric and hyperbolic functions are also introduced in this chapter. Derivatives
of these functions are developed in Chapter 3, making them available for applications of
differentiation in Chapter 4 and integration in Chapters 5–7.
• Limits at infinity and infinite limits are presented in Chapter 2, along with general discus-
sions on limits.
• Differential equations are so important to students in the physical sciences that separa-
ble equations are treated at the earliest opportunity, immediately after the introduction of
indefinite integrals in Chapter 5.
• Applications of the definite integral in Chapter 7 are presented before techniques of integra-
tion in Chapter 8. In this way, students see the diversity of applications of integration and
can appreciate why it is necessary to antidifferentiate more difficult functions. Applications
are covered again in the examples and exercises of Chapter 8.
• It is customary to discuss series of numbers in order to familiarize students with the idea
of convergence before treating Taylor and power series. Series of functions are treated first
to accommodate schools that need Taylor series, but have little use for series of constants.
Such would be the case if students required infinite series, say, only to study series solutions
of differential equations. They would have little use for series of constants. My experience
has shown that because students can so easily visualize partial sums of power series on
graphing calculators and computers, they do not have difficulty beginning with series of
functions. They can see convergence happening.
Treatment of Particular Topics

A subtle difference in how a topic is treated can significantly affect whether a student fully
understands that topic. From my experiences in teaching calculus, I have developed certain
preferences which I have incorporated in this book.
Preface xix
• In the review of logarithm functions in Chapter 1, logarithms are treated as powers and as
the inverse of exponentials. When we study definite integrals with variable upper limits in
Section 6.5, the integral definition of the natural logarithm function is given.
• Relative extrema are discussed in Section 4.3, whereas absolute extrema are discussed in
Section 4.7. Because they are different and are used differently, they are treated in separate
sections. Curve sketching uses relative extrema, while most applied extrema problems
involve absolute extrema.
• The topic of curve sketching has been divided into two separate sections. Section 4.5
contains the standard approach to curve sketching using local extrema, concavity, points of
inflection, limits, etc., but not technology. Section 4.6 begins with a computer or calculator
plot of a function and uses calculus to analyze what we see (or do not see).
• Antiderivatives or indefinite integrals are treated in a chapter unto themselves, namely,
Chapter 5. Relegating them to a section in a chapter on differentiation is insufficient.
Conceptually, integration is simple — backwards differentiation. But this operation is so
important, and it is so much more difficult to perform, only a separate chapter does it justice.
Integration techniques are further explored in Chapter 8.
• Antidifferentiation is approached from an organizational viewpoint. The emphasis is on
learning to organize an integrand into a form in which integration is obvious. Needless
memorization of formulas is discouraged.
• Antiderivatives are used to evaluate definite integrals. This is the first fundamental theorem
of integral calculus in Section 6.4. The fact that a definite integral with a variable upper
limit (Section 6.5) can be regarded as an antiderivative is given secondary importance. At
first, the importance of the antiderivative as a calculational tool for definite integrals is
emphasized.
• In Chapter 6, the definite integral is motivated through four types of problems: area, volume,
work, and fluid flow. At the same time, definite integrals are defined algebraically as limits
of Riemann sums. This approach permits us to use the definition of a definite integral in a
wide variety of applications without fear that it has been associated with any one application
in particular. Definitions of double, triple, line, and surface integrals as limit-summations
in multivariable calculus follow quite naturally.
• I prefer the notation Sin−1 for the inverse sine function. It is so natural once the notation
f −1 (x) has been introduced for inverse functions in general. The first letter is capitalized as
a reminder that this is indeed an inverse trigonometric function and that the −1 superscript
should not be interpreted as a power. In Section 1.8, where inverse trigonometric functions
are introduced for the first time, the arcsin notation is also presented so that students become
familiar with it.
• There is no agreement in the mathematical community on principal values for the inverse
secant and inverse cosecant functions. My choice is based on practical reasons for engineers
and physical scientists. For values outside the first quadrant, third-quadrant angles stated
between −π and −π/2 are used. Third-quadrant angles simplify derivatives of these
functions. In addition, choosing angles between −π and −π/2 matches the usual methods
when working in polar coordinates and with complex variables. When polar angle θ is
restricted to an interval of length 2π , it is most often chosen to be −π < θ < π . This
is especially true in complex analysis where principal values of the argument of a complex
number are always chosen in this interval. Principal values of all multivalued functions
(logarithm, inverse trigonometric, and inverse hyperbolic) use these principal values.
• I have chosen to keep the polar coordinate r nonnegative. Nothing is lost by this convention.
Curves may have slightly different equations depending on which convention is used. But
by choosing r > 0, considerable simplification is achieved when equations of curves in
polar coordinates are rewritten in Cartesian coordinates.
• In Chapter 10, the limit ratio test for convergence of infinite series is applied only to series
with positive terms. It is not used directly as a test for convergence of series with positive
xx Preface
and negative terms. Convergence is discussed one step at a time. First, convergence of
series of positive terms using the comparison, limit comparison, integral, limit ratio, and
limit root tests are discussed. Next, series with positive and negative terms are introduced,
and the above tests as tests for absolute convergence are used.
• Three-dimensional analytic geometry and vectors in Chapter 11 provide the tools for mul-
tivariable calculus. I stress the value of drawing curves and surfaces in space, using the
curve sketching tools learned in Chapters 1–4. Such diagrams are essential to the evalu-
ation of double, triple, line, and surface integrals, and to an appreciation of many of the
ideas of differential calculus. Vectors are handled algebraically and geometrically. Every
algebraic definition is interpreted geometrically, and every geometric definition is followed
by an algebraic equivalent. Differentiation and integration of vectors dependent on a single
parameter lead to discussions on tangent and normal vectors to curves, curvature and arc
length, and three-dimensional kinematics.
• Gradients are useful in many areas of applied mathematics. I introduce gradients in Section
12.4, and apply them to directional derivatives and normal vectors to curves and surfaces
in Sections 12.8 and 12.9. In this way, gradients are not associated with any particular
application.
• Chain rules for composite multivariable functions are endless in variations and applications.
In Chapter 12, I show how to appreciate each term in a chain rule as a contribution of
particular variables to the overall rate of change of a function, and then to create a schematic
diagram to handle the most complicated functional situations.
• In Chapter 13, definitions of double and triple integrals as limit-summations are analo-
gous to the definition of the definite integral presented earlier. The evaluation by double
and triple iterated integrals is geometric. Through representative boxes, rectangles, strips,
and columns, I show how to visualize the summation process and affix appropriate limits
to integrals. There is no algebraic manipulation of inequalities; a thoughtfully prepared
diagram does it all. The geometric approach also helps us visualize integrations in polar,
cylindrical, and spherical coordinates. I demonstrate that many of the applications of the
definite integral in Chapter 7 are handled much more easily with double integrals.
• For simplicity, definitions are kept to a minimum in Chapter 14. One kind of line integral
f (x, y, z)ds is presented and then integrals of the form P dx +Qdy +Rdz are treated as
a special case when f (x, y, z) is the tangential component of some vector function defined
along the curve. They can all be evaluated by substituting from parametric equations of the
curve. Alternatively, it may be expedient to use Green’s or Stokes’s theorems or determine
whether the integral is independent of path. Likewise, one surface integral f (x, y, z)dS
is presented, and then in many applications, f (x, y, z) is treated as the normal component
of some vector function defined on the surface.
• Chapter 15 provides a sound introduction to differential equations. It contains sections on
separable equations, linear first-order equations, and second-order equations easily reduced
to first-order equations. The exercises introduce (first-order) homogeneous and Bernoulli
equations. Considerable space (four sections) is devoted to the important topic of linear
differential equations. Applications to Newtonian mechanics, vibrating mass-spring sys-
tems, and electric circuits are discussed in detail. Other applications are introduced through
examples and exercises.
New in This Edition

In writing this edition I have reviewed my approach to each topic, discussions pertaining to it,
and choice of examples and exercises. The following changes are the most substantial:
• Each chapter contains “Consulting Projects”— problems that might typically arise in the
physical sciences. Solutions require calculus, and are often somewhat complex. They give
Preface xxi
us the opportunity to illustrate how to think through a multistage problem; to organize its
many facets; and provide a step-by-step, logical solution.
• Each section of Chapter 1 that contains review material contains a diagnostic test so that
students can judge the preparation they need for calculus.
• Inverse trigonometric and hyperbolic functions have been moved to Chapter 1.
• It has always been my practice in teaching calculus to tell students that they should attempt
to estimate an answer to a problem so that they’ll know whether their solution to the
problem is reasonable. I have incorporated this idea in this edition of the text (I often call
it “ballparking” the answer).
• I have tried to emphasize the need to “see” and “feel” calculus — its principles, rules, and
applications. This can often be done with a simple picture.
• Many new physical applications are included in the examples and exercises.
Supplements
A supplements package has been carefully prepared to aid instructors and students:
• An Instructor’s Solutions Manual, written by the author, providing complete solutions to
all the exercises in the text.
• An Instructor’s Resource CD-ROM, including the Instructor’s Solutions Manual (in PDF
format), PowerPoints, Image Library, and TestGen. TestGen is a computerized test bank
that allows instructors to edit existing questions, add new questions, and generate tests.
• A Student’s Solutions Manual, providing complete solutions to all the even-numbered
exercises (except those challenging questions marked with two asterisks), is available to
students.
• A Text-Enrichment Website has been created for this text, containing additional material
related to topics covered in the text. Visit this site at www.pearsoned.ca/text/trim to find
material on vector analysis and flux and circulation, as well as translation and rotation of
axes.
Acknowledgments
I am grateful to the many people who offered helpful suggestions for the preparation of this
edition. I would particularly like to thank the following instructors who provided formal reviews:
Dejan Delic, Ryerson University
Steven J. Desjardins, University of Ottawa
Samuel Dube, Carleton University
David L. Harmsworth, University of Waterloo
Kahina Sid Idris, University of Windsor
Rachel Kuske, University of British Columbia
Patricia Nieva, University of Waterloo
Any errors in the text or in the supplements are the responsibility of the author. I would
appreciate having them brought to my attention. Please send them to me care of Acquisitions
Editor, Mathematics, Higher Education Division, Pearson Education Canada, 26 Prince Andrew
Place, Toronto, Ontario, M3C 2T8.
Geometric Formulas
h h
b b
Parallelogram: Triangle:
Area = bh Area = 12 bh
r
θ a
Circle: b
Area = πr 2 Trapezoid:
Circumference = 2πr Area = 12 h(a + b)
Area of sector = 12 r 2θ
Length of arc = rθ
h
h
r r
Right circular cylinder: Right circular cone:

Volume = πr 2h Volume = 13 πr 2 h
Lateral surface area = 2πrh Lateral surface area = πr √r 2 + h2
Sphere:
Volume = 43 πr 3
Area = 4πr 2
Trigonometric Formulas
y x y
sin θ = cos θ = tan θ =
r r x y
r r x
csc θ = sec θ = cot θ =
y x y (x, y)
sin2 θ + cos2 θ = 1
1 + tan2 θ = sec2 θ r
1 + cot 2 θ = csc2 θ θ
x
Cosine law- c2 = a 2 + b2 − 2ab cos C

sin A sin B sin C
Sine law- = =
a b c A
sin(A + B) = sin A cos B + cos A sin B c b
sin(A − B) = sin A cos B − cos A sin B
cos(A + B) = cos A cos B − sin A sin B B C
cos(A − B) = cos A cos B + sin A sin B a
tan A + tan B tan A − tan B

tan(A + B) = tan(A − B) =
1 − tan A tan B 1 + tan A tan B
sin 2A = 2 sin A cos A
cos 2A = cos2 A − sin2 A = 1 − 2 sin2 A = 2 cos2 A − 1
2 tan A
tan 2A =
1 − tan2 A
! " ! "
1 A+B A−B
sin A sin B = [− cos(A + B) + cos(A − B)] sin A − sin B = 2 cos sin
2 ! 2 " ! 2 "
1 A+B A−B
sin A cos B = [sin(A + B) + sin(A − B)] cos A + cos B = 2 cos cos
2 ! 2 " ! 2 "
1 A+B A−B
cos A cos B = [cos(A + B) + cos(A − B)] cos A − cos B = −2 sin sin
2 ! " ! " 2 2
A+B A−B
sin A + sin B = 2 sin cos
2 2
Inverse Inverse
Trigonometric Trigonometric
Function Principal Values Function Principal Values
−1 π π −1
Sin x − ≤y≤ Cot x 0<y<π
2 2
−1 π π −1 π π
Tan x − <y< Csc x −π < y ≤ − , 0<y≤
2 2 2 2
π π
Cos −1 x 0≤y≤π Sec −1 x −π ≤ y < − , 0≤y<
2 2
Integration Formulas
Elementary Forms
! !
! !
u dv = uv − v du un sin au n
un cos au du = − un−1 sin au du
! a a
un+1 ! !
un du = + C, n "= −1 sinn−1 au cos au
n−1
n+1 sinn au du = − + sinn−2 au du
! an n
1 ! !
du = ln |u| + C cosn−1 au sin au n−1
u cosn au du = + cosn−2 au du
! an n
u u a ! !
du = − 2 ln |a + bu| + C tann−1 au
a + bu b b tann au du = − tann−2 au du
a(n − 1)
Trigonometric Integrals
! Integrals Involving Exponentials and Logarithms
sin u du = − cos u + C !
1
! eau du = eau + C
a
cos u du = sin u + C !
1
! a bu du = a bu + C
b ln a
tan u du = ln | sec u| + C !
! ln u du = u ln u − u + C
csc u du = ln | csc u − cot u| + C ! !
un eau n
! un eau du = − un−1 eau du
a a
sec u du = ln | sec u + tan u| + C !
un+1 un+1
! un ln u du = ln u − +C
n+1 (n + 1)2
cot u du = ln | sin u| + C !
eau
! eau sin bu du = 2 (a sin bu − b cos bu) + C
a + b2
sec2 u du = tan u + C !
eau
! eau cos bu du = 2 (a cos bu + b sin bu) + C
a + b2
csc2 u du = − cot u + C
√
! Integrals Involving u2 ± a 2
sec u tan u du = sec u + C ! "u#
1 1
! 2 2
du = Tan−1 +C
a +u a a
csc u cot u du = − csc u + C ! "u#
1 1
! √ du = Sec−1 +C
u 1 u u2 − a 2 a a
sin2 au du = − sin 2au + C ! $
2 4a u$ 2 a2 $
! u2 ± a 2 du = u ± a2 ± ln |u + u2 ± a 2 | + C
u 1 2 2
cos2 au du = + sin 2au + C !
2 4a 1 $
! √ du = ln |u + u2 ± a 2 | + C
1 1 u2 ± a 2
sec3 u du = sec u tan u + ln | sec u + tan u| + C
2 2 √
! Integrals Involving a 2 − u2
sin (a − b)u sin (a + b)u
sin au sin bu du = − + C, a 2 "= b2 !
2(a − b) 2(a + b) 1 "u#
! √ du = Sin−1 +C
sin (a − b)u sin (a + b)u a 2 − u2 a
cos au cos bu du = + + C, a 2 "= b2 % %
2(a − b) 2(a + b) !
! 1 1 %u + a %
du = ln % %+C
cos (a − b)u cos (a + b)u a 2 − u2 2a % u − a %
sin au cos bu du = − − + C, a 2 "= b2
2(a − b) 2(a + b) ! $ "u#
! ! u$ 2 a2
un cos au n a 2 − u2 du = a − u2 + Sin−1 +C
un sin au du = − + un−1 cos au du 2 2 a
a a
Derivative Formulas
d du dv
(u + v) = + d u du
du dx dx a = a u ln a
dx dx
d n du
(u ) = nun−1 d u du
dx dx e = eu
dx dx
d dv du
(uv) = u +v d 1 du
dx dx dx loga u = loga e
dx u dx
du dv
d 1 du
d ! u " v dx − u dx ln u =
= dx u dx
dx v v2
d du
d du sinh u = cosh u
sin u = cos u dx dx
dx dx
d du
d du cosh u = sinh u
cos u = − sin u dx dx
dx dx
d du
d du tanh u = sech 2 u
tan u = sec2 u dx dx
dx dx
d du
d du coth u = −csch 2 u
cot u = − csc2 u dx dx
dx dx
d du
d du sech u = −sech u tanh u
sec u = sec u tan u dx dx
dx dx
d du
d du csch u = −csch u coth u
csc u = − csc u cot u dx dx
dx dx
d 1 du
d 1 du Sinh−1 u = √
Sin−1 u = √ dx 2
u + 1 dx
dx 1 − u2 dx
d 1 du
d −1 du Cosh−1 u = √
Cos−1 u = √ dx u2 − 1 dx
dx 1 − u2 dx
d 1 du
d 1 du Tanh−1 u =
Tan−1 u = dx 1 − u2 dx
dx 1 + u2 dx
d 1 du
d −1 du Coth−1 u =
Cot−1 u = dx 1 − u2 dx
dx 1 + u2 dx
d −1 du
d 1 du Sech−1 u = √
Sec−1 u = √ dx u 1 − u2 dx
dx 2
u u − 1 dx
d −1 du
d −1 du Csch−1 u = √
Csc−1 u = √ dx 2
|u| 1 + u dx
dx u u2 − 1 dx
CHAPTER
1 Calculus Preparation
1.1 Introduction
Over the years, I have learned that the most frequent reason for students failing to achieve
optimum results in calculus courses is inadequate preparation. Students may understand calculus
concepts completely, they may have every formula and every rule memorized perfectly, they
may even have a clear idea of the procedures required to solve problems, but because they
lack skills in algebra, geometry, and trigonometry, they cannot put their knowledge to work.
It is unfortunate that they never get to demonstrate their calculus knowledge because of poor
mathematical preparation. To obtain the best possible grades in calculus, it is essential to have
excellent algebraic skills and a good grasp of the elements of analytic geometry and trigonometry.
In this chapter we give you the opportunity to test your skills and knowledge in these three areas,
and, should it be necessary, the means by which to make improvements. There are also sections
that may contain material unfamiliar to many readers. Topics in these sections are essential to
some of the applications of calculus; your instructor will indicate whether they are required for
your course.
The calculus course at your institution may, or may not, spend time on the review sections
in this chapter. If it does not, your instructor may advise you to review certain sections on your
own. DO IT! You could regret ignoring review material in this chapter in order to get to calculus
in Chapter 2 more quickly. Each review section opens with a diagnostic test to determine your
knowledge of the material in the section. Give yourself the suggested time to take the test, no
longer. You must not only be able to solve the test questions, but you must also be able to do so
reasonably quickly. Do NOT use a calculator unless specifically instructed to do so. Answers
are provided at the end of the section along with marks for each question. Assign yourself
partial marks for partially correct answers, but try to be objective in doing so. It is difficult
to be specific as to what constitutes an acceptable score on the diagnostic tests. Certainly a
score of less than 50% indicates that detailed study of the section is required. A score of more
than 50%, but not much more, would also suggest the need for careful review. Marks in the
80%–100% range indicate a good working knowledge of material in the section, but it could be
beneficial to give the section a quick reading, paying particular attention to parts corresponding
to incorrectly answered test questions. In addition, not only is it helpful to refresh your memory
on concepts learned some time ago, it is also wise to become familiar with the terminology,
notation, and conventions set forth in these sections. To improve your skills on, and knowledge
of, the material in a review section, read the discussions and examples thoroughly, try as many
of the exercises as you can, and then retake the diagnostic test. If you are conscientious in your
work, we are confident that you will do much better the second time. You will be well rewarded
for taking the time to do this; your calculus studies will be so much easier. In fact, the time you
spend on calculus preparation now will more than compensate for extra time that you would
spend on solving calculus problems later. Believe me; I have taught thousands of students just
like you.
1
2 Chapter 1 Calculus Preparation
1.2 Solving Polynomial Equations

Here is the diagnostic test for this section; give yourself 40 minutes to do it. When you have
marked your test using the answers at the end of the section, decide whether a brief reading or
a thorough treatment is needed for material in this section.
DIAGNOSTIC TEST FOR SECTION 1.2

Find all real solutions for each of the following polynomial equations. 9. x 4 − 3x 2 − 4 = 0 10. x 4 + x 3 + 3x 2 − x − 4 = 0
Give multiplicities of any repeated roots.
11. 3x 3 + x 2 + x − 2 = 0 12. 24x 3 + 2x 2 − 27x + 10 = 0
1. 9x + 5 = 0 2. x 2 − 4x − 5 = 0
In questions 13 and 14 you are given the zeros of the polynomial. Write
3. 6x 2 + 7x − 20 = 0 4. x 2 + 3x + 10 = 0 each polynomial in factored form.
5. x 2 + 6x − 4 = 0 6. 36x 2 − 108x + 81 = 0 13. Zeros of x 3 + 6x 2 − x − 30 are x = −5, −3, 2.

√
7. x 3 − 8 = 0 8. x 3 − 4x 2 + 5x − 2 = 0 14. Zeros of 2x 2 + 4x − 10 are x = −1 ± 6.
A real polynomial in x of degree n , where n ≥ 0 is an integer, is an expression of the

form
Pn (x) = an x n + an−1 x n−1 + · · · + a1 x + a0 (1.1)
where an , an−1 , . . . , a0 are real constants with an $= 0. (If an were equal to zero, the
polynomial would not be of degree n .) When Pn (x) is set equal to zero, the resulting equation
Pn (x) = an x n + an−1 x n−1 + · · · + a1 x + a0 = 0 (1.2)
is called a polynomial equation of degree n . Examples are 3x 3 − 2x + 5 = 0 and 2x 2 + 5x −

9 = 0. In this section, we are concerned with the number of solutions of a polynomial equation,
whether solutions are real or complex, and techniques for finding the real solutions. (Complex
solutions are dealt with in Appendix C.) Values of x that satisfy equation 1.2 are called roots
or solutions of the equation. They are also called zeros of the polynomial Pn (x) .
When n = 1, equation 1.2 is called a linear equation (or equation of degree 1),
a 1 x + a0 = 0 . (1.3)
Its only solution is x = −a0 /a1 . For example, the solution of 2x + 5 = 0 is x = −5/2.
Quadratic equations (equations of degree 2) are obtained when n = 2. It is customary in
this case to denote coefficients as follows
ax 2 + bx + c = 0. (1.4)
Quadratic equations can sometimes be solved by factoring, and can always be solved with the
quadratic formula. It is preferable to initially attempt to factor. For the example 6x 2 − 5x − 4 =
0, we write (3x − 4)(2x + 1) = 0, from which the two solutions are x = 4/3 and x = −1/2.
Likewise, for x 2 − 6x + 9 = 0, we factor in the form (x − 3)2 = 0, and the only solution is
x = 3. When the quadratic does not factor easily, we use the quadratic formula
√
−b ± b2 − 4ac
x = . (1.5)
2a
Application of this formula to the equation 2x 2 + 4x − 5 = 0 gives

! √ √
−4 ± 42 − 4(2)(−5) −4 ± 56 −2 ± 14
x = = = .
2 (2 ) 4 2
Quadratic formula 1.5 indicates that quadratic equation 1.4 has two distinct real solutions when
the discriminant b2 − 4ac > 0, two real solutions that are equal when b2 − 4ac = 0, and no
real solutions when b2 − 4ac < 0. (Solutions are complex numbers in the last case.)
The next simplest polynomial equation (after the linear and quadratic) is the cubic equation,
ax 3 + bx 2 + cx + d = 0, (1.6)
and after that the quartic,
ax 4 + bx 3 + cx 2 + dx + e = 0. (1.7)
There are procedures that give exact roots for both of these equations, but they are of so little
practical use in this day of the electronic calculator and personal computer, we omit their
discussions. For such equations, it is often sufficient to use a numerical procedure to approximate
roots to some degree of accuracy (Sections 1.11 and 4.1). On the other hand, when an exact
solution of a polynomial equation can be found, it can be removed from the equation, yielding a
simpler equation to solve for the remaining roots. The process by which this is done is a result
of the factor theorem.
THEOREM 1.1 (Factor Theorem)

x − a is a factor of Pn (x) if and only if Pn (a) = 0.
The factor theorem is very useful in solving polynomial equations. It does not find solutions,
however. What it does is simplify the problem each time a solution is found. To illustrate,
consider the quartic equation
P (x) = x 4 + 2x 3 + x 2 − 2x − 2 = 0. (1.8a)
A moment’s reflection indicates that x = 1 satisfies the equation. The factor theorem then
implies that x − 1 is a factor of the quartic. The remaining cubic factor can be obtained by long
division, synthetic division, or mental long division. The result is
P (x) = x 4 + 2x 3 + x 2 − 2x − 2 = (x − 1)(x 3 + 3x 2 + 4x + 2).
What this means is that equation 1.8a can be replaced by
P (x) = (x − 1)(x 3 + 3x 2 + 4x + 2) = 0. (1.8b)
To find further solutions of quartic equation 1.8a, we need only examine the cubic x 3 + 3x 2 +
4x + 2 in 1.8b for its zeros. Once we notice that a zero is x = −1, we may factor x + 1 from
the cubic and replace 1.8b with
P (x) = (x − 1)(x + 1)(x 2 + 2x + 2) = 0. (1.8c)
The remaining two solutions are complex numbers. Thus quartic equation 1.8a has two real
solutions x = ±1 and two complex solutions.
Once a solution a of a polynomial equation has been found, x − a can be factored from the
polynomial. When a is a fraction, it is recommended that this procedure be modified slightly.
To illustrate, consider the equation
2x 3 − x 2 − 9x + 9 = 0.
A solution of this equation is x = 3/2 (we show how we found this solution shortly). When
we factor x − 3/2 from the cubic, the result is
(x − 3/2)(2x 2 + 2x − 6) = 0,
but the work involves fractions. They can be avoided by factoring 2(x − 3/2) = 2x − 3 from
the cubic instead. Calculations are simpler, and the result is
(2x − 3)(x 2 + x − 3) = 0.
The remaining two solutions can be obtained with the quadratic formula
! √
−1 ± 12 − 4(1)(−3) −1 ± 13
x = = .
2 2
What we are suggesting is that when a rational number r = p/q is a solution of a polynomial
equation (with integer coefficients), it is simpler to factor q(x − p/q) = qx − p from the
polynomial (and in so doing only integer arithmetic is involved).
All real polynomials can be factored into real linear factors and irreducible real quadratic
factors. An irreducible real quadratic factor is one that has complex zeros, such as x 2 + 1
and 2x 2 + x + 6; they are characterized by negative discriminants. Finding the linear factors
of a polynomial goes hand-in-hand with finding real zeros of the polynomial. We have shown
the factorization of three polynomials below.
x 4 + 2x 3 + x 2 − 2x − 2 = (x − 1)(x + 1)(x 2 + 2x + 2), (1.9a)

x 3 − 3x 2 + 3x − 1 = (x − 1)3 , (1.9b)
x 8 + 7x 7 − 86x 5 − 95x 4 + 363x 3 + 486x 2 − 540x − 648 = (x + 3)4 (x − 2)3 (x + 1). (1.9c)
Each linear factor in these polynomials leads to a real zero of the polynomial. For polynomial
1.9a, the real zeros are ±1, and it also has two complex zeros; for polynomial 1.9b, the zeros
are 1, 1, 1; and for 1.9c, the zeros are −3, −3, −3, −3, 2, 2, 2, −1. In the case of 1.9b and
1.9c, there are repetitions. We say that x = 1 is a zero of multiplicity 3 for the polynomial
in 1.9b; the multiplicity corresponds to the number of times the factor x − 1 appears in the
factorization. Each of the zeros in 1.9a is of multiplicity 1. In 1.9c, x = −3 has multiplicity 4,
the zero x = 2 has multiplicity 3, and x = −1 has multiplicity 1. These examples suggest that
the number of zeros of a polynomial, taking multiplicities into account, is equal to the degree
of the polynomial. This is confirmed in the following theorem.
THEOREM 1.2
Every polynomial of degree n ≥ 1 has exactly n zeros (counting multiplicities).
EXAMPLE 1.1
Find all roots and their multiplicities for the polynomial equation
x 3 + x 2 − 16x + 20 = 0.
SOLUTION With a little experimentation, we find that x = 2 is a solution of the equation. It
follows that x − 2 can be factored from the cubic polynomial, and the equation can be written
in the form
0 = (x − 2)(x 2 + 3x − 10) = (x − 2)(x + 5)(x − 2) = (x − 2)2 (x + 5).
Thus, x = 2 is a solution with multiplicity 2, and x = −5 is a solution with multiplicity 1.
The following theorem provides a quick way to locate all rational roots of a polynomial equation
when its coefficients are rational numbers.
THEOREM 1.3 (Rational Root Theorem)

Suppose that r = p/q is a rational root (in lowest terms) of a polynomial equation
an x n + · · · + a1 x + a0 = 0 with integer coefficients, and a0 $= 0. Then, p divides a0
and q divides an .
This is a powerful result. It narrows the field of possible rational solutions of polynomial
equations with integer coefficients to a finite set. We illustrate with two examples.
EXAMPLE 1.2
Find all real solutions of the cubic equation 5x 3 + x 2 + x − 4 = 0.
SOLUTION Since divisors of −4 are ±1, ±2, and ±4, and those of 5 are ±1 and ±5, the
only possible rational solutions are
±1, ±2, ±4, ±1/5, ±2/5, ±4/5.
Trial and error leads to the fact that x = 4/5 satisfies the equation. We now factor 5(x − 4/5) =
5x − 4 from the cubic,
0 = (5x − 4)(x 2 + x + 1).
The remaining two solutions are complex numbers.
EXAMPLE 1.3
Find all real solutions of the quartic equation x 4 − x 3 − 3x 2 + 5x − 2 = 0.
SOLUTION Since divisors of −2 are ±1 and ±2, and those of 1 are ±1, the only possible
rational solutions are the integers ±1, ±2. Since x = 1 is a solution, we factor x − 1 from
the polynomial,
(x − 1)(x 3 − 3x + 2) = 0.
Once again, the only possible rational zeros of the cubic are ±1, ±2. We find that x = 1 is a
zero, and when x − 1 is factored from the cubic,
0 = (x − 1)(x − 1)(x 2 + x − 2) = (x − 1)2 (x + 2)(x − 1) = (x − 1)3 (x + 2).
The only roots of the equation are therefore x = 1 (multiplicity 3) and x = −2.
The list of possible rational roots, as predicted by the rational root theorem, is shortest when
any common factors in coefficients of the equation have been removed. For example, 3x 3 +
2x 2 − 2 = 0 and 9x 3 + 6x 2 − 6 = 0 have the same solutions. The rational root theorem yields
±1, ±2, ±1/3, and ±2/3 as possible solutions of the first equation, and ±1, ±2, ±3, ±6,
±1/3, ±2/3, ±1/9, and ±2/9 as possible solutions of the second equation. Clearly then it is
advantageous to remove the common factor 3 from the second equation.
Given the zeros of a polynomial, it is easy to write the polynomial in factored form. For
example, if the zeros of x 3 + 3x 2 − 6x − 8 are known to be x = −4, −1, 2, the factored form
of the polynomial is (x + 4)(x + 1)(x − 2) . Zeros of 2x 2 + 9x − 5 are x = −5, 1/2; its
factored form is 2(x − 1/2)(x + 5) or (2x − 1)(x + 5) . Finally, the zeros of 2x 2 + 2x − 4 are
x = −2, 1; its factored form is 2(x − 1)(x + 2) . Watch for the number preceding the product
of factors. It can always be determined by examining the coefficient of the highest power of x .
EXERCISES 1.2
In Exercises 1–30 find all real solutions of the polynomial equation. ∗ 23. 3x 4 + x 3 + 5x 2 = 0
Include multiplicities when they are greater than one.
∗ 24. 6x 3 + x 2 + 19x − 20 = 0
1. 3x − 2 = 0 2. 14x + 5 = 0
∗ 25. x 5 + 5x 4 − 9x − 45 = 0
2 2
3. x + 2x − 3 = 0 4. 12x + 11x − 5 = 0
∗ 26. x 5 − 15x 4 + 85x 3 − 225x 2 + 274x − 120 = 0
5. 2x 2 + 5x + 10 = 0 6. −4x 2 + 10x + 9 = 0 ∗ 27. 4x 4 + 4x 3 + 17x 2 + 16x + 4 = 0
7. x 2 + 8x + 16 = 0 8. 4x 2 − 36x + 81 = 0 ∗ 28. 25x 4 − 120x 3 + 109x 2 − 36x + 4 = 0
9. 2x 2 + 5x − 10 = 0 10. 4x 2 − 8x + 9 = 0 ∗ 29. x 5 + 9x 4 + 47x 3 + 125x 2 + 18x − 200 = 0
11. x 3 − 3x 2 + 3x − 1 = 0 12. 8x 3 + 12x 2 + 6x + 1 = 0 ∗ 30. x 6 + 16x 4 − 81x 2 − 1296 = 0
13. x 3 − 2x 2 + 5x − 10 = 0 In Exercises 31–36 you are given a polynomial and its zeros. Write the
polynomial in factored form.
14. x 3 + 4x 2 + 12x + 9 = 0
31. 2x 2 + 8x − 10, x = −5, 1
3 2
∗ 15. x + 12x + 48x + 64 = 0 √
32. 2x 2 − 3x − 7, x = (3 ± 65)/4
∗ 16. x 4 + 7x 3 + 9x 2 − 21x − 36 = 0 33. x 3 − 11x 2 + 4x + 60, x = −2, 3, 10
∗ 17. x 4 − 16 = 0 34. 24x 3 + 22x 2 − 27x + 5, x = −5/3, 1/4, 1/2
35. x 4 − 2x 3 + 2x − 1, x = −1, x = 1 (muliplicity 3)

∗ 18. 2x 4 + 9x 3 − 6x 2 − 8x − 15 = 0
36. 16x 4 − 8x 2 + 1, x = ±1/2 each of multiplicity 2
∗ 19. 6x 4 + x 3 + 53x 2 + 9x − 9 = 0
37. Find a polynomial that has only zeros x = −1/3, 4/5, 3, each
∗ 20. 12x 4 + 19x 2 + 5 = 0 with multiplicity 1 and x = 4 with multiplicity 3. Is it unique?
38. Prove the following corollary to the rational root theorem, called
∗ 21. x 3 − 23x 2 − 21x − 72 = 0 the “integer root theorem”: If r is a rational zero of a polynomial
Pn (x) = x n + an−1 x n−1 + · · · + a0 with integer coefficients, where
∗ 22. x 4 − 4x 3 − 44x 2 + 96x + 576 = 0 a0 $= 0, then r is an integer that divides a0 .
ANSWERS TO DIAGNOSTIC TEST FOR SECTION 1.2
1. −5/9 (1 mark) 8. 1 with multiplicity 2, 2 (3 marks)

2. −1, 5 (2 marks) 9. ±2 (3 marks)
3. −5/2, 4/3 (2 marks) 10. ±1, (3 marks)
4. No real solutions (2 marks) 11. 2/3 (3 marks)
√
5. −3 ± 13 (2 marks) 12. −5/4, 1/2, 2/3 (4 marks)
6. 3/2 with multiplicity 2 (3 marks) 13. (x − 2)(x + 3)(x + 5) (2 marks)

√ √
7. 2 (2 marks) 14. 2(x + 1 − 6)(x + 1 + 6) (3 marks)
1.3 Plane Analytic Geometry and Straight Lines

Here is the diagnostic test for this section; give yourself 30 minutes to do it. When you have
marked your test using the answers at the end of the section, decide whether a brief reading or
a thorough treatment is needed for material in this section.
1. Find the distance between the points (−1, 3) and (2, −6) and the 6. Find the equation of the line through the point (1, 0) parallel to the
coordinates of the midpoint of the line segment that joins them. line 2x − 4y = 5.
2. Find the slope of the line 3x + 10y = 14. 7. Find the equation of the line through the point (4, −2) perpendicular
to the line 3x + 2y = 9.
3. Find the x - and y -intercepts of the line 5x − 2y = 11.
8. Find the point of intersection of the lines 2x − 3y = 10 and x +
4. Find, in general form, the equation of the line through the point 5y = 6.
(3, −5) with slope −2. 9. Find the distance from the point (1, 2) to the line 3x − 4y = 5.
5. Determine whether the lines 2x − 3y = 4 and 5x − 8y = 13 are 10. Draw the three lines x + 2y = 4, x = 2, and x = 3y − 5 on the
parallel, perpendicular, or neither. same set of axes.
FIGURE 1.1 The coordi-

nates of a point A sound knowledge of analytic geometry, a union of algebra and geometry, is essential
y to the study of calculus. On the one hand, it provides a way to describe geometric objects
algebraically; on the other hand, it permits a geometric visualization of algebraic statements.
P (x, y) Our approach to calculus is visual; we draw pictures at every possible opportunity, by hand,
y by graphing calculator, and by computer. We draw pictures to introduce ideas, to illustrate
concepts, to reinforce principles, and to solve problems. We want you to see and feel calculus
x x
in all its aspects. Analytic geometry is the basis for many of these pictures. In this section, we
review the fundamentals of plane analytic geometry, paying particular attention to straight lines.
In Section 1.4 we review circles, parabolas, ellipses, and hyperbolas.
Points in the plane are identified by an ordered pair (x, y) of real numbers called their
Cartesian coordinates (Figure 1.1). The x -coordinate of a point P is its perpendicular distance
FIGURE 1.2 Signs of from the y -axis, and the y -coordinate is its perpendicular distance from the x -axis.
coordinates in the four quadrants The axes divide the plane into four parts which, beginning at the upper right and proceeding
y counterclockwise, are called, respectively, the first, second, third, and fourth quadrants. The
axes themselves are not considered part of any quadrant. Points in the first quadrant have x - and
Second First y -coordinates that are both positive; points in the second quadrant have a negative x -coordinate
quadrant quadrant and a positive y -coordinate; and so on (Figure 1.2).
(−, +) (+, +)
If P and Q are any two points with coordinates (x1 , y1 ) and (x2 , y2 ) , respectively
(Figure 1.3), then by the Pythagorean relation for the right-angled triangle P QS ,
x
Third Fourth
quadrant quadrant &P Q&2 = &P S&2 + &QS&2 ,
(−, −) (+, −)
where &P Q& denotes the length of the line segment joining P and Q , and &P S& and &QS&
denote the lengths of the related line segments. But †
FIGURE 1.3 Length of the &P S& = |x2 − x1 | and &QS& = |y2 − y1 |,
line segment joining two arbitrary
points and therefore ‡
y
&P Q&2 = |x2 − x1 |2 + |y2 − y1 |2 or
Q (x2 , y2 ) !
y2
&P Q& = (x2 − x1 )2 + (y2 − y1 )2 . (1.10)
y1
P (x1, y1) S (x2 , y1) † The vertical lines around |x − x | and |y − y | denote absolute values. They operate on whatever is between
2 1 2 1
them to produce a nonnegative result according to |x| = x if x ≥ 0 and |x| = −x if x < 0. For example,
x1 x2 x |4| = 4 and | − 3| = 3. We discuss absolute values from a functional point of view in Section 1.5.
‡ √x always denotes the positive square root of x , so that √4 = 2.
FIGURE 1.4 Length of line joining FIGURE 1.5 Midpoint of line segment
two specific points
y
y
P (−3, 4) Q (x 2 , y2 )
R (x, y)
U (x 2 , y )
P (x1, y1) S
T (x, y1)
x
Q (2, −1) x
Equation 1.10 expresses the length of a line segment joining points P and Q in terms of their
coordinates (x1 , y1 ) and (x2 , y2 ) . For example, in Figure 1.4
! √
&P Q& = (2 + 3)2 + (−1 − 4)2 = 5 2.
Suppose that R(x, y) in Figure 1.5 is the midpoint of the line segment joining P (x1 , y1 )
and Q(x2 , y2 ) . Since triangles P TR and RUQ are congruent, it follows that &RU & = &P T & ,
and for P and Q as shown in Figure 1.5, this condition becomes x2 − x
= x − x1 . Consequently,
x1 + x 2
x = . (1.11a)
2
Similarly, equality of &RT & and &QU & gives

y1 + y 2
y = . (1.11b)
2
Derivation of these results assumed that x2 > x1 and y2 > y1 , but the same formulas are valid
for any x1 , x2 , y1 , and y2 whatsoever. Coordinates of the midpoint of a line segment, therefore,
are averages of coordinates of the ends of the line segment.
An equation such as y = x 2 specifies a relationship between numbers represented by the
letter x and those represented by the letter y : y must always be the square of x . Algebraically,
we speak of pairs of values x and y that satisfy this equation. It is customary to write these pairs
in the form (x, y) , a few simple ones being (0, 0) , (1, 1) , (−1, 1) , (2, 4) , and (−2, 4) . If we
interpret each pair of solutions (x, y) as coordinates of a point, we find that all such points
lie on the curve in Figure 1.6. This curve, then, is a geometric visualization of solution pairs
FIGURE 1.6 Graph of of the equation y = x 2 . Every pair of values x and y that satisfies y = x 2 is represented by
y = x2 a point on the curve. Conversely, the coordinates (x, y) of every point on the curve provide a
pair of numbers that satisfies y = x 2 . Algebraic solutions, then, are represented geometrically
y as points on a curve; points on the curve provide algebraic solutions.
35 If x and y are the coordinates of a point in the plane, then the point is on the curve in
30 Figure 1.6 if and only if x and y satisfy the equation y = x 2 . In other words, this equation
25
completely characterizes all points on the curve. We therefore call y = x 2 the equation of
20
the curve. Thus the equation of a curve is an equation that the coordinates of every point on
15
10
the curve satisfy and at the same time is an equation that no point off the curve satisfies. In
5 the remainder of this section we discuss straight lines and their equations. In Section 1.4, we
show that parabolas, circles, ellipses, and hyperbolas can be recognized by the distinctive forms
1 2 3 4 5 6 x of their equations. Once again note the algebraic-geometric interplay; the form of an equation
dictates the shape of the curve, and, conversely, the shape of a curve determines the form of its
equation.
The Straight Line

The slope of the line through the points (x1 , y1 ) and (x2 , y2 ) in Figure 1.7a is defined as the
quotient
y2 − y 1
m= . (1.12)
x 2 − x1
The difference y2 − y1 is called the rise because it represents the vertical distance between the
points, and x2 − x1 the run, the horizontal distance between the points. It is easy to show using
similar triangles that m is independent of the two points chosen on the line; that is, no matter
which two points we choose on the line to evaluate m , the result is always the same. The four
numbers x1 , x2 , y1 , and y2 vary, but the ratio (y2 − y1 )/(x2 − x1 ) remains unchanged. A
horizontal line (Figure 1.7b) has slope zero (since y2 − y1 = 0), whereas the slope of a vertical
line is undefined (since x2 − x1 = 0). In Figure 1.7c, line l1 , which leans to the right, has
positive slope, and line l2 , which leans to the left, has negative slope.
FIGURE 1.7a Slope of a FIGURE 1.7b Slopes of FIGURE 1.7c Lines with
line through two points horizontal and vertical lines positive and negative slopes
y y y
l2 l1
(x2, y2 )
y2 − y 1
x x x
(x1, y1) x2 − x 1
FIGURE 1.8 Equation of

line through two given points
In Figure 1.8 we have shown the line through the points (1, 2) and (3, 5) . Its slope is
(5 − 2)/(3 − 1) = 3/2. To find the equation for this line, we let (x, y) be the coordinates of
y any other point on the line. Since the slope of the line must also be given by (y − 2)/(x − 1) ,
(x, y) it follows that
(3, 5) y−2 3
= ,
x−1 2
3 and when this equation is simplified,
2 (1, 2)
1 2y = 3x + 1.
1 2 3 x
If (x, y) are the coordinates of any point not on this line, then they do not satisfy this equation
because the slope (y−2)/(x−1) joining (x, y) to (1, 2) is not equal to 3/2. Thus, 2y = 3x+1
is the equation for the straight line through (1, 2) and (3, 5) .
FIGURE 1.9 Equation of a
line through given point with given We can use this procedure to find the equation for the straight line through any point (x1 , y1 )
slope with any slope m (Figure 1.9). If (x, y) are the coordinates of any other point on the line, then
the slope of the line is (y − y1 )/(x − x1 ) ; therefore,
y
y − y1
=m or y − y1 = m(x − x1 ). (1.13)
(x, y) x − x1
(x1, y1) This is called the point-slope formula for the equation of a straight line; it uses the slope m of
the line and a point (x1 , y1 ) on the line to determine the equation of the line.
x Other formulas for the equation of a line are available when different characteristics of the
Slope m line are given. They are listed below. The point-slope formula is included for completeness.
Characteristics
Form of Equation Name of Equation Determining the Line
y − y1 = m(x − x1 ) Point-slope formula m is the slope;
(x1 , y1 ) is a point on the line
y − y1 x − x1 Two-point formula Two points (x1 , y1 ) and
=
y2 − y1 x2 − x1 (x2 , y2 ) on the line
y = mx + b Slope y -intercept formula m is the slope;
b is the y -intercept
y = m(x − a) Slope x -intercept formula m is the slope;
a is the x -intercept
x y Two-intercept formula a and b are x - and y -intercepts
+ =1
a b
Point-slope formula 1.13 encompasses all of these. The characteristics in the right column
determine the slope of a line and a point on it; therefore, with minimal calculations, the point-
slope formula can be used in all situations. For example, if we know that the slope of a line
is m and the y -intercept is b , then a point on the line is (0, b) . Hence, equation 1.13 gives
y − b = m(x − 0) , or y = mx + b , the slope y -intercept formula.
There is one other equation that is worth mentioning. Vertical lines, which do not have
slopes, cannot be represented in form 1.13; they have form x = k , for some constant k .
However, all lines can, for various values of the constants A , B , and C , be represented in the
form
Ax + By + C = 0. (1.14)
This is often called the general equation of a line.
EXAMPLE 1.4
Find the general equation of the line through the points (−1, 1) and (2, 3) .
SOLUTION Since the slope of the line is (3 − 1)/(2 + 1) = 2/3, we can use the point
(−1, 1) and the slope 2/3 in point-slope formula 1.13 to give
2
y−1 = (x + 1) or 2x − 3y + 5 = 0.
3
The same result is obtained if the point (2, 3) is used in place of (−1, 1) . The two-point formula
could also be used.
EXAMPLE 1.5
FIGURE 1.10 Graphs of

Find equations for the lines in Figure 1.10.
4y = 20 − 5x and x = −1 SOLUTION Since the slope of l1 is −5/4 and its y -intercept is 5, we use formula 1.13 with
y the point (0, 5) ,
5
6 y − 5 = − (x − 0) or 4y = 20 − 5x.
4
5
4
The equation of l2 is clearly x = −1.
l2 3 l1
2
1
−1 1 2 3 4 5 6 x
We do not recommend the use of the slope y -intercept formula y = mx + b for finding the
equation of a line; the point-slope formula y − y1 = m(x − x1 ) is more versatile. The slope
y -intercept formula is, however, useful in finding the slope of a line with given equation. For
instance, to find the slope of the line 2x + 3y = 9, we express it in slope y -intercept form
y = −2x/3 + 3. It follows that the coefficient of x , namely, −2/3, is the slope of the line.
To find the point of intersection of two straight lines — say, 3y = 2x + 5 and 4y + x =
FIGURE 1.11 Intersection
point of two lines
14 (Figure 1.11) — we find the point whose coordinates (x, y) satisfy both equations. If we
solve each equation for x and equate the expressions, we obtain (3y − 5)/2 = 14 − 4y ,
y which immediately yields y = 3. Either of the original equations then gives x = 2, and the
3y = 2x + 5 point of intersection of the lines is (2, 3) . Note once again the algebraic–geometric interplay.
Geometrically, (2, 3) is the point of intersection of two straight lines. Algebraically, the two
numbers constitute the solution of the equations 3y = 2x + 5 and 4y + x = 14.
Of particular importance to the study of lines are the concepts of parallelism and perpen-
4y + x = 14 dicularity.
x
DEFINITION 1.1
Two distinct lines are said to be parallel if they have no point of intersection.
EXAMPLE 1.6
Verify that the lines with equations 2x − y = 4 and 4x − 2y = 7 are parallel.
SOLUTION If we solve each of these equations for y and equate the resulting expressions,
we obtain
7
2x − 4 = 2x − ,
2
an obvious impossibility. Consequently, the lines do not intersect.
Geometrically, the following is clear.
THEOREM 1.4
Two distinct lines are parallel if and only if they have the same slope.
For example, if we write the equations of the lines in Example 1.6 in the form y = 2x − 4
and y = 2x − 7/2, coefficients of the x -terms identify the slopes of the lines. Since each has
slope 2, the lines are parallel.
DEFINITION 1.2
Two lines are said to be perpendicular if they intersect at right angles.
The following theorem gives a test for perpendicularity of straight lines in terms of slopes.
THEOREM 1.5
Two lines with nonzero slopes m1 and m2 are perpendicular if and only if
1
m1 = − or m1 m2 = −1. (1.15)
m2
EXAMPLE 1.7
Find the equation of the straight line that passes through the point (2, 4) and is perpendicular
to the line 3x + y = 5.
SOLUTION Since the slope of the given line y = −3x + 5 is −3, the required line has slope
1/3. Using point-slope formula 1.13, we find that the equation of the required line is
1
y−4 = (x − 2) or x − 3y = −10.
3
There is a useful formula for finding the (shortest) distance, d , from a point (x1 , y1 ) to a line
with equation Ax + By + C = 0 (Figure 1.12). It is
|Ax1 + By1 + C|
d = √ . (1.16)
A2 + B 2
For example, the distance from the point (−2, −1) to the line 3x + 2y = 2 in Figure 1.13 is
|3(−2) + 2(−1) − 2| 10
√ = √ .
32 + 22 13
FIGURE 1.12 Shortest distance from a point to a line FIGURE 1.13 Distance from (−2, −1) to 3x + 2y = 2
y (x1, y1) y
d
Ax + By + C = 0
x
(−2, −1)
x 3x + 2y = 2
EXERCISES 1.3
In Exercises 1–4 find the distance between the points. 16. Through the point (−1, −2) and crossing the y -axis at 4
1. (1, 3) , (3, 4) 2. (−2, 1) , (4, −2) 17. Crossing the x - and y -axes at 1 and −3, respectively
3. (−1, −2) , (−3, −8) 4. (3, 2) , (−4, −1) 18. Through the origin and the midpoint of the line segment joining
5.–8. Find the midpoint of the line segment joining the points in (3, 4) and (−7, 8)
Exercises 1–4.
In Exercises 9–18 draw the line described and find its equation. In Exercises 19–26 determine whether the lines are perpendicular, par-
allel, or neither.
9. Through the points (1, 2) and (−3, 4)
19. y = −x + 4, y = x + 6
10. Through the points (3, −6) and (5, −6)
11. Through the point (−2, −3) with slope 3 20. x + 3y = 4, 2x + 6y = 7
12. Through the point (1, −3/2) with slope −1/2
21. x = 3y + 4, y = x/3 − 2
13. The y -axis
22. 2x + 3y = 1, 3x − 2y = 5
14. The x -axis
15. Through the point (4, 3) and crossing the x -axis at −2 23. y = 3x + 2, y = −x/2 + 1
24. x − y = 5, 2x + 3y = 4 ∗ 48. Repeat Exercise 47 for the relationship between temperatures in

degrees Fahrenheit, TF , and in kelvin, TK .
25. x = 0, y = 5
∗ 49. Consider a metal bar of length L0 at temperature T0 . If the bar is
26. x + y + 2 = 0, 3x − y = 4 heated or cooled, its length changes. The amount is described by the
coefficient of linear expansion α . It is the change in length per unit
In Exercises 27–32 find the point of intersection of the lines. length per degree Celsius.
(a) Show that the length L of the bar at temperature T is given
27. x + y = 0, x − 2y = −3 by the formula L = L0 [1 + α(T − T0 )]. Draw its graph.
28. x = 1, y = 2 (b) If steel railroad rails 10 m long are laid with their ad-
jacent ends 3 mm apart at a temperature of 20◦ C, at
29. 3x + 4y = 6, x − 6y = 3 what temperature will their ends be in contact? For steel,
α = 1.17 × 10−5 / ◦ C.
30. y = 2x + 6, x = y + 4
∗ 50. A median of a triangle is a line segment drawn from a vertex to the
31. x/2 + y/3 = 1, 2x − y/4 = 15 midpoint of the opposite side. Find equations for the three medians of
32. 14x − 2y = 5, 3x + 10y = 12 the triangle with vertices (1, 1) , (3, 5) , and (0, 4) . Show that all three
medians intersect in a point called the centroid of the triangle.
In Exercises 33–38 find the (shortest) distance from the point to the line. ∗ 51. If (x1 , y1 ) , (x2 , y2 ) , (x3 , y3 ) , and (x4 , y4 ) are vertices of any
quadrilateral, show that the line segments joining the midpoints of ad-
33. (3, 4) to x + y = 1 34. (1, −3) to x + 2y = 3 jacent sides form a parallelogram.
35. (5, 1) to x − y = 4 36. (3, 1) to y = −x ∗ 52. Find the equation of the perpendicular bisector of the line segment
joining (−1, 2) and (3, −4) . (The perpendicular bisector is the line
37. (6, −2) to x = −1 38. (4, −2) to 15x − 2y + 3 = 0 that cuts the line segment in half and is perpendicular to it.)
∗ 53. Find coordinates of the point that is equidistant from the three
In Exercises 39–46 find the equation of the line described. points (1, 2) , (−1, 4) , and (−3, 1) .
∗ 39. Parallel to x + 2y = 15, and through the point of intersection of ∗ 54. Prove Theorem 1.5.
2x − y = 5 and x + y = 4
∗ 55. Generalize the result of equations 1.11a and b to prove that if a
∗ 40. Perpendicular to x − y = 4, and through the point of intersection point R divides the length P Q so that
of 2x + 3y = 3 and x − y = 4
length P R r1
∗ 41. Parallel to the line through (1, 2) and (−3, 0) , and through the = ,
length RQ r2
point (5, 6)
where r1 and r2 are positive integers, then the coordinates of R are
∗ 42. Perpendicular to the line through (−3, 4) and (1, −2) , and
through the point (−3, −2) r1 x2 + r2 x1 r1 y2 + r2 y1
x = and y = .
∗ 43. Crossing through the first quadrant to form an isosceles triangle r1 + r2 r1 + r2
with area 8 square units ∗ 56. Is a line parallel to itself?
∗ 44. Through (3, 5) and crossing through the first quadrant to form a ∗∗ 57. Prove that in any triangle the sum of the squares of the lengths of
triangle with area 30 square units the medians is equal to three-fourths of the sum of the squares of the
lengths of the sides. (A median of a triangle is a line segment drawn
∗ 45. Has slope 2, and that part in the second quadrant has length 3
from one vertex to the midpoint of the opposite side.)
∗ 46. Parallel to the x -axis, below the point of intersection of the lines
∗∗ 58. Let P be any point inside an equilateral triangle (figure below).
x = y and x + y = 4, and forms with these lines a triangle with area
Show that the sum of the distances of P from the three sides is always
9 square units
equal to the height h of the triangle.
∗ 47. (a) What is the formula by which we convert temperature TF in
degrees Fahrenheit to temperature TC in degrees Celsius?
(b) What is the formula by which we convert temperature TC in
degrees Celsius to temperature TF in degrees Fahrenheit?
h
(c) Can we interpret the formulas in parts (a) and (b) as equa- P
tions for straight lines in a TF TC -plane? Draw both lines.
(d) At what temperature are TF and TC numerically equal?

√
1. 3 10, (1/2, −3/2) (2 marks) 10. (5 marks)
2. −3/10 (1 mark) y
3. 11/5, −11/2 (2 marks) x=2
4. 2x + y − 1 = 0 (2 marks) x + 2y = 4 x = 3y − 5
5. Neither (2 marks)
6. x − 2y = 1 (3 marks) 2
7. 2x − 3y = 14 (3 marks) 5/3
8. (68/13, 2/13) (3 marks) −5 2 4 x
9. 2 (2 marks)
1.4 Conic Sections

Here is the diagnostic test for this section. Give yourself 60 minutes to do it.

In questions 1–10 identify the curve as a straight line, parabola, circle, 14. Find equations for the asymptotes of the hyperbola x 2 − 3y 2 = 4.
ellipse, hyperbola, or none of these.
15. Find equations for the asymptotes of the hyperbola 4x 2 − 2y 2 =
2
1. 3x − y = 4 2. x = y − 2y + 3 4x − 10y .
4
3. y = x + 3 4. 3x 2 + y 2 = 4 16. Find the point(s) of intersection of the curves x + 2y = 5 and
2
5. y − 2x = x 2 2 2
6. x + 4y + 5 = 0 x 2 + 2y 2 = 9.
2 2
7. x + y − 2y = 16 8. 3x + y 2 = 3 17. Find the highest point on the parabola y = −x 2 + 6x + 4.
9. x 2 + 2y 2 + x = 2y 10. x 2 − x + y 2 + y = 0 18. Find the points where the parabola x = y 2 − 5y − 6 crosses the
11. Which, if any, of the points (1, 0) , (0, 1) , and (2, −1) are on the x -axis and the y -axis.
circle 3x 2 + 3y 2 + 2y = 6x + 5?
19. Find the equation of a parabola that passes through the points
12. Find the centre and radius of the circle x 2 + y 2 + 2x − 4y = 25. (−1, 1) , (0, 3) , and (1, 1) .
13. Find the centre of the ellipse 2x 2 + 4x + 6y 2 + 9y = 36. What 20. Find the equation of the ellipse with centre at the origin that passes
are its axes of symmetry? through the points (−4, 0) and (3, 4) .
Parabolas, circles, ellipses, and hyperbolas arise in a multitude of applications. We do

not give a complete development of these curves together with their many properties here; we
show only how the form of the equation for each conic section relates to its shape. Detailed
discussions of conic sections in terms of focus and directrix are given in Sections 9.5 and 9.6.
The Parabola
When the y -coordinate of a point (x, y) on a curve is related to its x -coordinate by an equation
of the form
y = ax 2 + bx + c, (1.17)
where a , b , and c are constants (with a $= 0), the curve is called a parabola. The simplest of all
parabolas is y = x 2 (Figure 1.6). For every point (x, y) to the right of the y -axis on this curve,
there is a point equidistant to the left of the y -axis which has the same y -coordinate; that is, the
point (−x, y) is also on the curve. Putting it another way, that part of the parabola to the left of
the y -axis is the image in the y -axis (thought of as a mirror) of that part to the right of the y -axis.
Such a curve is said to be symmetric about the y-axis. It happens whenever the equation of a
curve is unchanged when each x therein is replaced by −x . In other words, we have the test: A
curve is symmetric about the y-axis if its equation remains unchanged when x is replaced by −x .
FIGURE 1.14 Parabola
The parabola y = 9 − x 2 is shown in Figure 1.14. It is symmetric about the y -axis;
opening downward
replacing x by −x leaves the equation unchanged. The parabola is said to open downward,
y whereas the parabola y = x 2 in Figure 1.6 opens upward. The sign of the coefficient of x 2
dictates which way a parabola opens (positive for upward and negative for downward).
The parabola y = 2x 2 + 4x − 6 is plotted in Figure 1.15. It is not symmetric about the
10
y = 9 − x2 y -axis; its equation changes when x is replaced by −x . The parabola appears to be symmetric
5 about the line x = −1 with lowest point (−1, −8) . To prove that this is indeed the case, we
rewrite the equation as
−2 −1 x
1 2 3 y = 2(x 2 + 2x) − 6 = 2(x + 1)2 − 8.
This form clearly indicates that the smallest value for y is −8, and it occurs when x = −1.
It also indicates that the parabola is symmetric about x = −1. To see this we replace x with
−1 ± a , where a > 0 is any number whatsoever. The result is y = 2(−1 ± a + 1)2 − 8
= 2a 2 − 8. This shows that points on the parabola with x -coordinates smaller than −1 and
FIGURE 1.15 Parabola larger than −1 by the same amount a have the same y -coordinate, namely 2a 2 − 8.
opening upward The technique used above to rewrite y = 2x 2 + 4x − 6 as 2(x + 1)2 − 8 is called comp-
y leting the square. If we apply the same technique to the general parabola y = ax 2 + bx + c ,
y = 2x 2 + 4x − 6 we obtain
" # " # " #
40 b 2 b 2 b2
y =a x + x +c = a x+ + c− . (1.18)
30 a 2a 4a
20
This form for the equation of the parabola shows the following:
10
1. When a > 0, the parabola opens upward and has a minimum at the point
−4 1234 x " #
−10 b b2
− ,c − .
2a 4a
2. When a < 0, the parabola opens downward and has a maximum at the point
" #
b b2
− ,c − .
2a 4a
3. The parabola is symmetric about the line x = −b/(2a) (Figure 1.16); that is, for every
point P on one side of the line, there is a point Q on the other side that is the mirror image
of P in the line.
FIGURE 1.16 Symmetry 4. The parabola crosses the x -axis when
of general parabola " #2 " #
b b2
y 0 = y = a x+ + c− .
b 2a 4a
x=−
2a
To solve this equation for x , we first write
" #
b 2 b2 c
Q P x+ = − ,
2a 4a 2 a
x
b b2 and then take square roots of each side,
− ,c−
2a 4a $
b b2 c
x+ =± 2
− .
2a 4a a
FIGURE 1.17a Parabola that FIGURE 1.17b Parabola that FIGURE 1.17c Parabola that
intersects x -axis in two points touches x -axis at one point does not intersect x -axis
y y y
x x x
b
−
−b − b2 − 4ac −b + b2 − 4ac 2a
2a 2a
Finally, then,
$ √
b b2 c −b ± b2 − 4ac
x =− ± − = . (1.19)
2a 4a 2 a 2a
This is quadratic formula 1.5 that we encountered in Section 1.2, but now in a geometric
setting. It determines points where the parabola y = ax 2 + bx + c crosses the x -axis.
When b2 −4ac > 0, the parabola crosses the x -axis twice (Figure 1.17a); when b2 −4ac =
0, it touches the x -axis at one point (Figure 1.17b); and when b2 − 4ac < 0, the parabola
has no points in common with the x -axis (Figure 1.17c).
When x and y in equation 1.17 are interchanged, the resulting parabola,
x = ay 2 + by + c, (1.20)
opens to the right or left rather than up or down. Figures 1.18 and 1.19 show the parabolas
x = y 2 + 1 and x = −y 2 + 4y − 4, respectively. The parabola x = y 2 + 1 is symmetric
about the x -axis; any point (x, y) on the parabola is symmetric with the corresponding point
(x, −y) . In general, a curve is symmetric about the x-axis if its equation remains unchanged
when y is replaced by −y .
FIGURE 1.18 Parabola opening to the right FIGURE 1.19 Parabola opening to the left
y y
4
3
x = y2 + 1 x = −y 2 + 4y − 4
2
5
1
4
5 10 15 20 x 3
−1
2
−2
−10 −8 −6 1
−3
−4 −4 −2 x
EXAMPLE 1.8
Find equations for the parabolas in Figures 1.20 and 1.21.
SOLUTION The fact that the parabola in Figure 1.20 is symmetric about the x -axis means
that b in equation 1.20 must vanish; that is, its equation must be of the form x = ay 2 + c . Since
the points (2, 0) and (0, 3) are on the parabola, their coordinates must satisfy the equation of
the parabola,
2 = a(0)2 + c, 0 = a(3)2 + c.
FIGURE 1.20 Parabola through three special points FIGURE 1.21 Parabola through any three points
y y
3 (4, 6)
(−1, 3) (0, 2)
2 x
x
−3
These imply that c = 2 and a = −c/9 = −2/9. Thus, the equation of the parabola is
x = −2y 2 /9 + 2.
The parabola in Figure 1.21 has no special attributes that we can utilize (such as the position
of the line of symmetry in Figure 1.20). We therefore use the facts that its equation must be
of the form y = ax 2 + bx + c , and the three points (−1, 3) , (0, 2) , and (4, 6) are on the
parabola. Substitution of these coordinates into the equation gives
3 = a(−1)2 + b(−1) + c, 2 = a(0)2 + b(0) + c, 6 = a(4)2 + b(4) + c.
The second equation yields c = 2, and when this is substituted into the other two equations,
a − b = 1, 16a + 4b = 4.
These can be solved for a = 2/5 and b = −3/5; therefore, the required equation is
y = 2x 2 /5 − 3x/5 + 2.
Suppose we had not given you the pictures in this example, but only asked for the equation
of a parabola passing through the points. There would have been an additional parabola through
the points (−1, 3) , (0, 2), and (4, 6) , one opening to the right; its equation is x = (2y 2 −
13y + 18)/3. Only a parabola opening to the left can be found through the points (0, 3) ,
(0, −3), and (2, 0) .
EXAMPLE 1.9
When a shell is fired from the artillery gun in Figure 1.22, it follows a parabolic path
4.905
y =− x 2 + x tan θ,
v2 cos2 θ
where v is the muzzle velocity of the shell and θ is the angle at which the shell is fired. Find
the range R of the shell and the maximum height attained by the shell.
FIGURE 1.22 Trajectory of an artillery shell
Gun
R x
SOLUTION We can find R by setting y = 0 and solving for x :

" #
−4.905
0 = x + tan θ x.
v 2 cos2 θ
One solution is x = 0, corresponding to the firing position of the shell. The other solution gives
the range of the shell,
tan θ v 2 cos2 θ v 2 sin θ cos θ

x =R = = .
4.905 4.905
Maximum height of the shell is attained when x = R/2, in which case
% &2 % &
−4.905 v 2 sin θ cos θ v 2 sin θ cos θ
y = + tan θ
v 2 cos2 θ 2(4.905) 2(4.905)
−v 2 sin2 θ v 2 sin2 θ v 2 sin2 θ

= + = .
19.62 9.81 19.62
In many of the examples and exercises of the book, we ask you to draw and/or plot curves.
To plot a curve, you are to use a graphing calculator or computer. To draw a curve, you are to do
so without these devices. Sometimes, as we shall see, a drawing is more informative than a plot.
EXERCISES 1.4A
In Exercises 1–12 draw the parabola. Use a calculator or computer to 16. y 17. y
plot the parabola as a check.
1. y = 2x 2 − 1 2. y = −x 2 + 4x − 3
3. y = x 2 − 2x + 1 4. 3x = 4y 2 − 1
2
5. x = y 2 + 2y
−1 3 x 1 x
6. 2y = −x 2 + 3x + 4 −1
7. x + y 2 = 1
8. 2y 2 + x = 3y + 5 In Exercises 18–23 find all points of intersection for the curves. In

each case draw or plot the curves.
9. y = 4x 2 + 5x + 10
18. y = 1 − x 2 , y = x + 1
10. x = 10y 2 19. y + 2x = 0, y = 1 + x 2
11. y = −x 2 + 6x − 9 20. y = 2x − x 2 − 6, 25 + x = 5y
21. x = y(y − 1) , 2y = 2x + 1
12. x = −(4 + y)2
22. x = −y 2 + 1, x = y 2 + 2y − 3
13. Find x - and y -intercepts for the parabolas (a) y = x 2 − 2x − 5 23. y = 6x 2 − 2, y = x 2 + x + 1
and (b) x = 4y 2 − 8y + 4.
∗ 24. For what angle θ is the range of the artillery shell in Example 1.9
largest?
In Exercises 14–17 determine the equation for each parabola shown. ∗ 25. The cable of the suspension bridge in the following figure hangs in
the shape of a parabola. The towers are 200 m apart and extend 50 m
14. y 15. y above the roadway. If the cable is 10 m above the roadway at its lowest
(2, 3) 4 point, find the length of the supporting rods 30 m from the towers.
Tower
Supporting Cable
1
rod
x 2 x 50
30
10 Roadway
200
∗ 26. Find points of intersection for the parabolas y = (x − 2)2 and (a) Use the facts that resistances at temperatures 0◦ C, 100◦ C,
5x = y 2 + 4. and 700◦ C are, respectively, 10.000 # , 13.946 # , and
24.172 # to determine R0 , a , and b .
∗ 27. Find the height of the parabolic arch in the figure below.
(b) Plot a graph of the function on the interval 0 ≤ T ≤ 700.
(c) At what temperature is the resistance 20 # ?
3 ∗ 30. The parabola y = x 2 − 1 in the following figure represents the

base of a wall perpendicular to the xy -plane ( x and y measured in
4 metres). A rope is attached to a stake at position (3, 4) , pulled tight,
wrapped around that part of the base of the wall containing the vertex
of the parabola, and tied to the vertex. At what point does it meet the
5 wall?
∗ 28. Determine the equation of a parabola of type 1.17 passing through
the points (1, 2) , (−3, 10) , and (3, 4) . y
∗ 29. Resistance R in ohms in a platinum resistance thermometer is −1 Wall
related to temperature T in degrees Celsius by the equation
(3, 4)
R = R0 (1 + aT + bT 2 ), 1
−1
x
where R0 , a , and b are constants. y= x2 −1
The Circle
When the x - and y -coordinates of points on a curve are related by an equation of the form
(x − h)2 + (y − k)2 = r 2 , (1.21)
where h , k , and r > 0 are constants, the curve is called a circle. It takes but a quick recollection
of distance formula 1.10 to convince ourselves that this definition of a circle coincides with our
FIGURE 1.23 Circle with intuitive idea of a circle. If we write equation 1.21 in the form
centre (h, k) and radius r
!
y (x − h)2 + (y − k)2 = r,
the left side is the distance from the point (x, y) to the point (h, k) . Equation 1.21 therefore
describes all points (x, y) at a fixed distance r from (h, k) , a circle centred at (h, k) with
r radius r (Figure 1.23). For example, the radius of the circle in Figure 1.24 is equal to 2, and its
(x, y) (h, k)
equation is therefore
(x + 1)2 + (y − 2)2 = 4.
x
When equation 1.21 is expanded, we have
x 2 − 2hx + h2 + y 2 − 2ky + k 2 = r 2 or
FIGURE 1.24 Circle with
centre (−1, 2) and radius 2
x 2 + y 2 − 2hx − 2ky + h2 + k 2 − r 2 = 0.
y
This shows that the equation of a circle may be given in another form, namely,
(−1, 2) x 2 + y 2 + f x + gy + e = 0, (1.22)
(1, 2)
where e , f , and g are constants. Given this equation, the centre and the radius can be identified
by reversing the expansion and completing the squares of x 2 + f x and y 2 + gy . For instance,
−1 x if x 2 + y 2 + 2x − 3y − 5 = 0, then
0 = (x + 1)2 + (y − 3/2)2 − 5 − 1 − 9/4 = (x + 1)2 + (y − 3/2)2 − 33/4.

√
The centre of the circle is therefore (−1, 3/2) and its radius is 33/2.
When the centre of a circle is the origin (0, 0) , equation 1.21 simplifies to
x2 + y2 = r 2. (1.23)
Be careful to use equation 1.21, not equation 1.23, when the centre of the circle is not the origin.
It is a common error to use equation 1.23.
EXAMPLE 1.10
Figure 1.25 shows an arc of a circle. Find the equation for the circle.
FIGURE 1.25 Equation of SOLUTION From the symmetry of the figure, we see that the centre of the circle is on the
a circle passing through 3 points
y -axis. Its equation must be of the form
y
3 x 2 + (y − k)2 = r 2 .
2
1 Because (4, 0) is a point on the circle, these coordinates must satisfy its equation; that is,
−3 −2 −1 1 2 3 4
−4 x
16 + (0 − k)2 = r 2 or 16 + k 2 = r 2 .
−2
Similarly, since (0, −1) is on the circle,
(−1 − k)2 = r 2 or 1 + 2k + k 2 = r 2 .
If we subtract these two equations, we obtain 2k − 15 = 0, from which we see that k = 15/2.
Consequently, r 2 = 16 + k 2 = 16 + 225/4 = 289/4, and the equation of the circle is
x 2 + (y − 15/2)2 = 289/4.
EXAMPLE 1.11
The straight lines in Figure 1.26a represent a welded frame where AOB is a right angle. The
circle represents a wheel rotating on a pin through O perpendicular to the plane of the frame. If
the minimum clearance between circle and side AB must be 10 cm, find the maximum radius
of the wheel. Locate the point on the circle closest to AB .
FIGURE 1.26a Wheel spinning on welded frame FIGURE 1.26b Reorientation of wheel on frame
y
1
A Q (x, y)
O
P (a, b)
2m r
1m B
O 2 x
A B
SOLUTION Suppose we rotate and flip the figure and establish the coordinate system in
Figure 1.26b. Let the maximum radius of the circle be r and let P (a, b) be the point on the
circle closest to AB . Intuitively, the shortest distance &P Q& between wheel and AB occurs
when line OP Q is perpendicular to AB . Since the slope of the line AB is −1/2, its equation is
1
y − 1 = − (x − 0) *⇒ x + 2y − 2 = 0.
2
According to formula 1.16, the distance from the origin to line AB is
|(0) + 2(0) − 2| 2
&OQ& = √ = √ .
2
1 +2 2 5
√
For the length of √P Q to be 1/10 m, it follows that r + 1/10 = 2/ 5, and the radius of the
wheel is r = (2/ 5 − 1/10) m. To locate P , we note that the slope of OP is 2 (the negative
of the reciprocal of the slope of AB ). It follows that b/a = 2. Furthermore, because P is on
the circle (whose equation is x 2 + y 2 = r 2 ), we must have a 2 + b2 = r 2 . When we substitute
b = 2a into this equation, we obtain
r 2r
a 2 + 4a 2 = r 2 *⇒ a = √ and b = √ .
5 5
EXERCISES 1.4B
In Exercises 1–10 draw the circle. Use a calculator or computer to plot 13. y 14. y
the circle as a check. 3
x
1. x 2 + y 2 = 50
(1, 4)
2. (x + 5)2 + (y − 2)2 = 6 −3
2 2
2
3. x + 2x + y = 15 (3, 2)
4. x 2 + y 2 − 4y + 1 = 0 1 x
5. x 2 − 2x + y 2 − 2y + 1 = 0 15. y 16. y
2 2
6. 2x + 2y + 6x = 25 (2, 7)
2 2
7. 3x + 3y + 4x − 2y = 6 (−5, 6)
8. x 2 + 4x + y 2 − 2y = 5 (1, 3)
2 2
9. x + y − 2x − 4y + 5 = 0 (5, 1)
10. x 2 + y 2 + 6x + 3y + 20 = 0 3 x x
(2, −2)
In Exercises 11–16 find an equation for the circle.
11. y 12. y
∗ 17. A ladder of length L rests vertically against a wall. If the lower
2 end of the ladder is moved along level ground away from the wall while
the top of the ladder remains in contact with the wall, find an equation
for the curve followed by the midpoint of the ladder.
2 x 2 x ∗ 18. Find the equation of a circle that passes through the points (3, 4)
and (1, −10) , and has its centre (a) on the line 2x + 3y + 16 = 0 and
(b) on the line x + 7y + 19 = 0.
In Exercises 19–22 find points of intersection for the curves. radius (see Exercise 24) and, (b) taking the equation of the circle in
form 1.21 and requiring A , B , and C to be on the circle.
∗ 19. x 2 + 2x + y 2 = 4, y = 3x + 2
∗ 28. Prove that the three altitudes of the triangle in Exercise 27 intersect
∗ 20. x 2 + y 2 − 4y + 1 = 0, 2x + y = 1 in a point called the orthocentre of the triangle.
2 2 2
∗ 21. x + y = 9, y = 3x + 4 ∗∗ 29. Show that if a line Ax + By + C = 0 and a circle (x − h)2
∗ 22. (x + 3)2 + y 2 = 25, y 2 = 16(x + 1) + (y − k)2 = r 2 do not intersect, then the shortest distance between
them is the smaller of the two numbers
∗ 23. Show that every equation of the form 1.22 represents a circle, a √
point, or nothing at all. |(Ah + Bk + C) ± r A2 + B 2 |
√ .
∗ 24. Prove that the perpendicular bisector of a chord of a circle always A2 + B 2
passes through the centre of the circle.
∗∗ 30. The incircle of a triangle is that circle which lies interior to the
∗ 25. Two lights are 100 m apart, one at the origin, and the other at point triangle but touches all three sides. The centre of the incircle is called
(100, 0) in the xy -plane. The light at the origin is 10 times as bright the incentre. Show that the incentre (x, y) of the triangle with vertices
as the other light. Find, and draw, all points in the xy -plane at which (0, 0) , (2, 0) , and (0, 1) must satisfy the equations
the amount of light received from both sources is the same. Assume
that the amount of light received at a point is directly proportional to |x + 2y − 2|
the brightness of the source and inversely proportional to the square of |x| = |y| = √ .
5
the distance from the source.
∗ 26. Two loudspeakers are 20 m apart. One is at the origin, and the Solve these equations for the incentre, and explain why there are four
other is at the point (0, 20) in the xy -plane. The speaker at the origin points that satisfy these equations.
is only 70% as loud as the other. Find, and draw, all points in the xy - ∗∗ 31. Loudspeakers at points (x , y ) and (x , y ) emit sounds with
1 1 2 2
plane at which the amount of sound received from both speakers is the intensities I1 and I2 , respectively. The amount of sound received at a
same. Assume that the amount of sound received at a point is directly point (x, y) from either speaker is directly proportional to the intensity
proportional to the loudness of the speaker and inversely proportional of the sound at the speaker and inversely proportional to the square of
to the square of the distance from the speaker. the distance from the point to the source. Show that all points in the
∗ 27. The circumcircle for a triangle is that circle which passes through xy -plane at which the amount of sound received from both sources
all three of its vertices. Find the circumcircle for the triangle with is equal lie on a circle with centre on the line through (x1 , y1 ) and
vertices A(1, 1) , B(−3, 3) , and C(2, 4) by (a) finding its centre and (x2 , y2 ) .
The Ellipse
The set of points whose coordinates (x, y) satisfy an equation of the form
x2 y2
+ = 1, (1.24)
a2 b2
where a and b are positive constants, is said to constitute an ellipse. Since this equation is
so similar to equation 1.23, and is exactly the same when a = b = r , it is not unreasonable
to expect that the shape of this curve might be similar to a circle, especially when values of a
and b are close together. This is indeed the case, as Figure 1.27a and b illustrate. The ellipse
is symmetric about the x - and y -axes, and this is consistent with the fact that equation 1.24
remains unchanged when x and y are replaced by −x and −y . The ellipse is elongated in the
x -direction when a > b (Figure 1.27a), and when b > a it is elongated in the y -direction
(Figure 1.27b). The point of intersection of the lines of symmetry of an ellipse is called the
centre of the ellipse. For equation 1.24 the centre is the origin since the x - and y -axes are the
lines of symmetry.
FIGURE 1.27a Ellipse elongated in x -direction FIGURE 1.27b Ellipse elongated in y -direction
y y
b
b
−a a x
−a a x
−b
−b
EXAMPLE 1.12
Find the equation of the ellipse that has its centre at the origin, the x - and y -axes as axes of
symmetry, and passes through the points (4, 1) and (−2, 3) .
SOLUTION If we substitute coordinates of the points into equation 1.24 (since they are both
on the ellipse),
42 12 16 1
+ =1 *⇒ + = 1,
a2 b2 a2 b2
(−2)2 32 4 9
+ =1 *⇒ + = 1.
a2 b2 a2 b2
When the first equation is multiplied by −9 and added to the second, the result is
140
− = −8.
a2
Thus, a 2 = 35/2, and when this is substituted into the first equation,
32 1 35
+ =1 *⇒ b2 = .
35 b2 3
The equation required is therefore 2x 2 /35 + 3y 2 /35 = 1, or 2x 2 + 3y 2 = 35.
When equation 1.24 is changed to
(x − h)2 (y − k)2
+ = 1, (1.25)
a2 b2
where h and k are constants, the curve is still an ellipse; its shape remains the same. Just as
a change from equation 1.23 to 1.21 for a circle moves the centre of the circle from (0, 0) to
(h, k) , equation 1.25 moves the centre of the ellipse to (h, k) . Lines x = h and y = k are
the new lines of symmetry (Figure 1.28), and a and b are the distances between the centre and
where the ellipse crosses the lines of symmetry.
FIGURE 1.28 Ellipse with centre (h, k)
(h − a, k) (h, k + b)
(h, k)
y=k
(h + a, k)
(h, k − b)
x=h
x
EXAMPLE 1.13
Find the centre of the ellipse 16x 2 + 25y 2 − 160x + 50y = 1175, and draw the ellipse.
SOLUTION When we complete squares on the x - and y -terms,
(x − 5)2 (y + 1)2
16(x − 5)2 + 25(y + 1)2 = 1600 or + = 1.
100 64
The centre of the ellipse is (5, −1) ; it cuts the lines y = −1 and x = 5 at distances of 10 and
8 units from the centre, respectively (Figure 1.29).
FIGURE 1.29 An ellipse graphed from its equation
(5, 7)
x
(−5, −1) (5, −1) (15, −1)
(5, −9)
EXERCISES 1.4C
In Exercises 1–8 draw the ellipse. Use a calculator or computer to plot ∗ 10. Find the width of the elliptic arch in the figure below.
the ellipse as a check.
1 3
1. x 2 /25 + y 2 /36 = 1 2
5 2
2. 7x 2 + 3y 2 = 16
3. 9x 2 + 289y 2 = 2601
In Exercises 11–16 find all points of intersection for the curves. In
4. 3x 2 + 6y 2 = 21
each case draw the curves.
5. x 2 + 16y 2 = 2 11. x 2 + 4y 2 = 4, y = x
6. x 2 + 2x + 4y 2 − 16y + 13 = 0 12. 16x 2 + 9y 2 = 144, y = x + 3

√ √
∗ 13. 9x 2 − 18x + 4y 2 = 27, 2y = 3x + 3
2 2 √ √
7. 9x + y − 18x − 6y = 26 2 2
∗ 14. 9x − 18x + 4y = 27, 2y = − 3x + 5 3
8. x 2 + 4x + 2y 2 + 16y + 32 = 0 ∗ 15. x 2 + 4y 2 − 8y = 0, y = x 2
∗ 16. x 2 + 4y 2 = 4, y = x 2 − 4
9. Find the equation of an ellipse that passes through the points (−2, 4) ∗∗ 17. Show that every point on the ladder in Exercise 17 of Section 1.4B
and (3, 1) . follows an ellipse.
The Hyperbola
Changing one sign in the equation of an ellipse leads to a curve with totally different character-
istics. The set of points whose coordinates (x, y) satisfy an equation of the form
x2 y2
− =1 or (1.26a)
a2 b2
y2 x2
− = 1, (1.26b)
b2 a2
where a and b are positive constants, is called a hyperbola. Hyperbola 1.26a crosses the x -axis
at x = ±a , but does not cross the y -axis. Since the equation remains unchanged when x is
replaced by −x and y is replaced by −y , the hyperbola is symmetric about both the x -axis
and the y -axis. It follows that if we draw that part of the hyperbola in the first quadrant, we
can obtain its second, third, and fourth quadrant points by reflection. By taking positive square
roots of
y2 x2
= − 1,
b2 a2
we obtain
$
FIGURE 1.30a One- y x2
quarter of the hyperbola = − 1,
b a2
x 2 /a 2 − y 2 /b2 = 1
from which
y b
y= x b! 2
a y = x − a2.
a
This equation describes the top half of the hyperbola. Because a is a fixed constant, we can say
that for large values of x , values of a 2 are insignificant compared to values of x 2 , and values
of y are approximately equal to bx/a . This means that for large values of x , the hyperbola is
a x
very close to the line y = bx/a . We have shown these facts in Figure 1.30a. The complete
hyperbola, obtained by reflecting Figure 1.30a in the axes, is shown in Figure 1.30b. The lines
FIGURE 1.30b Entire
hyperbola using its symmetry
y = ±bx/a that the hyperbola approaches for large positive and negative values of x are called
asymptotes of the hyperbola.
b y b Hyperbola 1.26b is shown in Figure 1.31.
y=− x y= x
a a
y2 x2
FIGURE 1.31 The hyperbola 2
− 2 =1
b a
−a
a x y
b b
x 2 y2 y=− x b y= x
− =1 a a
a2 b2
x
−b
y2 x2
2 − 2 =1
b a
EXAMPLE 1.14
√
Find the equation of a hyperbola that cuts the y -axis at y = ±5 and has lines y = ±x/ 3 as
asymptotes.
SOLUTION Since the hyperbola crosses the y -axis at y = ±5, we write its equation in form
1.26b with b = 5,
y2 x2
− = 1.
25 a2
When we solve this equation for y in terms of x , the result is
5!
y =± x 2 + a2.
a
√ √
Since asymptotes of this hyperbola are ±5x/a , it follows that 5/a = 1/ 3, or a = 5 3.
The equation of the hyperbola is therefore
y2 x2
− = 1.
25 75
When x and y in equations 1.26 are replaced by x − h and y − k , the resulting equations
(x − h)2 (y − k)2
− =1 and (1.27a)
a2 b2
(y − k)2 (x − h)2
− =1 (1.27b)
b2 a2
still describe hyperbolas. They are shown in Figures 1.32 and 1.33. These are the hyperbolas
of Figures 1.30 and 1.31 shifted so that the asymptotes intersect at the point (h, k) . Equations
of the asymptotes are y = k ± b(x − h)/a , and the lines x = h and y = k are now axes of
symmetry.
FIGURE 1.32 Hyperbola with centre at (h, k) FIGURE 1.33 Hyperbola with centre at (h, k)
y y b
y=k+ (x − h)
b a
y=k+ (x − h) b
a y=k− (x − h)
a
b
y=k− (x − h)
a
(h, k + b)
(h − a, k) (h + a, k)
y=k
(h, k)
(h, k)
(h, k − b)
x x
x=h
EXAMPLE 1.15
Find asymptotes for the hyperbola x 2 − y 2 + 4x + 10y = 5.
SOLUTION If we complete squares on x - and y -terms, we obtain

FIGURE 1.34 A hyper-
bola graphed from its equation (y − 5)2 (x + 2)2
(x + 2)2 − (y − 5)2 = −16 or − = 1.
y 16 16
The axes of symmetry of the hyperbola are x = −2 and y = 5, intersecting at the point
(−2, 5) . When we solve for y in terms of x , the result is
(−2, 9)
(−2, 5) !
y = 5± (x + 2)2 + 16.
(−2, 1)
−2 x Therefore, the asymptotes are y = 5 ± (x + 2) . The hyperbola is shown in Figure 1.34.
y = 5 + (x + 2) y = 5 − (x + 2)
Hyperbolas are but one of many curves that have asymptotes. For instance, in Figure 1.35,
the curve y = (x 3 − 3x 2 + 1)/(x 2 + 1) is asymptotic to the line y = x − 3. The curve
2
y = 2e−x /100 is asymptotic to the x -axis (Figure 1.36). Limits (discussed in Chapter 2)
provide a unifying structure for all types of asymptotes.
FIGURE 1.35 Asymptote of nonhyperbolic curve FIGURE 1.36 Illustration of horizontal asymptote
5 y 2
y
1.5 2 /100
y = 2e−x
−7.5 −5 −2.5 2.5 5 7.5 x
1
x 3 − 3x 2 + 1
y= −5 y=x−3
x2 + 1
0.5
−7.5
−10
−20 −10 10 20 x
EXAMPLE 1.16
To the right of the right branch of the hyperbola x 2 − 4y 2 = 5 in Figure 1.37 is a swamp. A
pipeline is to originate from point (15, 10) to meet with a pipeline running north from point
(−1, −100) along the line x = −1. The pipeline must meet the north-south line as far down
the line x = −1 as possible. Given that the pipeline from (15, 10) should be straight, determine
where it should meet the north-south pipeline.
FIGURE 1.37 Best line along which to build a pipeline
y
P1(15, 10)
x 2 − 4y 2 = 5
Swamp
−10 −5 5 10 x
P2 (−1, y) −5
SOLUTION The pipeline from P1 (15, 10) should meet the line x = −1 at P2 (−1, y) so
that line P1 P2 just touches the right half of the hyperbola. If we let m be the slope of P1 P2 ,
then using point-slope formula 1.13, the equation of line P1 P2 is
y − 10 = m(x − 15).
To find the required position of P2 , line P1 P2 must intersect the hyperbola in exactly one point
(most lines intersect in two points, or not at all). Points of intersection are found by solving
y − 10 = m(x − 15) and x 2 − 4y 2 = 5.
Substituting from the first equation into the second gives
x 2 − 4[m(x − 15) + 10]2 = 5.

If we expand the second term on the left,
x 2 − 4[m2 (x − 15)2 + 20m(x − 15) + 100] = 5.
This can be rearranged into the form
(1 − 4m2 )x 2 + (120m2 − 80m)x + (−900m2 + 1200m − 405) = 0.
Given a value for m , solutions for x of this quadratic equation are x -coordinates of points of in-
tersection of the line with slope m and the hyperbola. We want only one solution. Consequently,
the discriminant must be equal to zero,
(120m2 − 80m)2 − 4(1 − 4m2 )(−900m2 + 1200m − 405) = 0.
When terms are expanded, this equation reduces to
0 = 176m2 − 240m + 81 = (4m − 3)(44m − 27),
and therefore solutions are m = 3/4 and m = 27/44. Thus, lines through (15, 10) with slopes
3/4 and 27/44 touch the hyperbola at only one point. The line with smaller slope 27/44 touches
the left branch of the hyperbola. The line with larger slope 3/4 touches the right branch; this is
the line we want. Its equation is y − 10 = (3/4)(x − 15) , and it cuts the line x = −1 when
y = 10 + (3/4)(−1 − 15) = −2. Thus, the pipeline should meet the north-south pipeline at
the point (−1, −2) .
We shall find a much easier solution to this problem when we have studied some calculus,
but the solution above shows that it can be done without calculus, albeit not easily.
EXERCISES 1.4D
In Exercises 1–10 draw the hyperbola. Use a calculator or computer to ∗ 11. Find the equation of a hyperbola that passes through the point
plot the hyperbola as a check. (1, 2) and has asymptotes y = ±4x .
1. y 2 − x 2 = 1 2. x 2 − y 2 = 1
In Exercises 12–18 find all points of intersection for the curves. In
3. x 2 − y 2 /16 = 1 4. 25y 2 − 4x 2 = 100 each case draw the curves.
5. y 2 = 10(2 + x 2 ) 6. 3x 2 − 4y 2 = 25 12. x 2 − 2y 2 = 1, x = 2y
13. 9y 2 − 4x 2 = 36, y = x
7. x 2 − 6x − 4y 2 − 24y = 11
14. 9y 2 − 4x 2 = 36, x = 3y
2 2
8. 9x − 16y − 18x − 64y = 91 15. 3x 2 − y 2 = 3, 2x + y = 1
∗ 16. x 2 − 2x − y 2 = 0, x = y 2
9. 4y 2 − 5x 2 + 8y − 10x = 21
∗ 17. x 2 − 2x − y 2 = 0, x = −y 2
10. x 2 + 2x − 16y 2 + 64y = 79 ∗ 18. 9(x − 1)2 − 4(y − 1)2 = 36, 27x = 5(y − 1)2

√
1. Straight line (1 mark) 12. (−1, 2) , 30 (3 marks)
2. Parabola (1 mark) 13. (−1, −3/4) , x = −1, y = −3/4 (4 marks)
3. None of these (1 mark) √
14. y = ±x/ 3 (2 marks)
4. Ellipse (1 mark)
√
5. Hyperbola (1 mark) 15. 5/2 ± 2(x − 1/2) (3 marks)
6. None of these (1 mark) 16. (4, −3) , (19/3, −2/3) (4 marks)
7. Circle (1 mark) 17. (3, 13) (3 marks)
8. Parabola (1 mark)
18. (−6, 0) , (0, 6) , (0, −1) (3 marks)
9. Ellipse (1 mark)
10. Circle (1 mark) 19. y = 3 − 2x 2 (3 marks)
11. (0, 1) (3 marks) 20. 16x 2 + 7y 2 = 256 (4 marks)
1.5 Functions and Their Graphs


In questions 1 and 2 find the largest possible domain for the function. 16. f (x) = |x 3 − 8| 17. f (x) = x 3 − 4x
√ 18. The graph of a function f (x) is shown to the left below. Draw a
1. f (x) = 4 − x2 2. f (x) = (x + 3)/(x 2 − 2x − 4)
graph of the function f (−x) on the axes to the right.
y y
In questions 3 and 4 find the range of the function.
√
3. f (x) = − 4 + x 2 4. f (x) = 3 + 2|x|
(1,1) (2,1)
In questions 5–8 determine whether the function is even, odd, or neither
even nor odd. x x
4 2 3 5
5. f (x) = x − 2x − 5 6. f (x) = x + 5x − x 19. If the graph of f (x) is that in question 18, draw a graph of f (x+2)
4
7. f (x) = x − 2x 8. f (x) = (x + 1)/(x − 3) 2 on the left set of axes below.
9. Find the even and odd parts of the function f (x) = x/(x + 1) . y y
2
10. Does the equation x = y − 2y + 1 define y as a function of x ?
11. What is a rational function?
In questions 12–17 draw the graph of the function.

√ x x
12. f (x) = 8 − |x| 13. f (x) = 9 − x2
√ 20. If the graph of f (x) is that in question 18, draw a graph of f (2x)
14. f (x) = −2 x 15. f (x) = −|1 − x 2 | − (x 2 − 1) on the right set of axes above.
Most quantities that we encounter in everyday life are dependent on many, many other
quantities. For example, think of what might be affecting room temperature as you read this
sentence — thermostat setting; outside temperature; wind conditions; insulation of the walls,
ceiling, and floors; and perhaps other factors that you can think of. Functional notation allows
interdependences of such quantities to be represented in a very simple way.
When one quantity depends on a second quantity, we say that the first quantity is a function
of the second. For example, the volume V of a sphere depends on its radius r ; in particular,
V = 4π r 3 /3. We say that V is a function of r . When an object is dropped, the distance d
(metres) that it falls in time t (seconds) is given by the formula d = 4.905t 2 . We say that d is
a function of t . Mathematically, we have the following definition.
DEFINITION 1.3
A quantity y is said to be a function of a quantity x if there exists a rule by which we can
associate exactly one value of y with each value of x . The rule that associates the value
of y with each value of x is called the function.
If we denote the rule or function in this definition by the letter f , then the value that f
assigns to x is denoted by f (x) , and we write
y = f (x). (1.28)
In our first example above, we write V = f (r) = 4π r 3 /3, and the function f is the
operation of cubing a number and then multiplying the result by 4π/3. For d = f (t) =
4.905t 2 , the function f is the operation of squaring a number and multiplying the result by
4.905.
We call x in equation 1.28 the independent variable because values of x are substituted
into the function, and y the dependent variable because its values depend on the assigned
values of x . The domain of a function is the set of all specified (real) values for the independent
variable. It is an essential part of a function and should always be specified. Whenever the
domain of a function is not mentioned, we assume that it consists of all possible values for
which f (x) is a real number.
As the independent variable x takes on values in the domain, a set of values of the dependent
variable is obtained. This set is called the range of the function. For the function V = f (r)
= 4π r 3 /3, which represents the volume of a sphere, the largest possible domain is r > 0, and
the corresponding range is V > 0. Note that mathematically the function f (r) is defined for
negative as well as positive values of r , and r = 0; it is because of our interpretation of r as
the radius of a sphere that we restrict r > 0. The function d = f (t) = 4.905t 2 represents
the distance fallen in time t by an object that is dropped at time t = 0. If it is dropped from a
height of 20 m, then it is clear that √
the range of this function is 0 ≤ d ≤ 20. The domain that
gives rise to this range is 0 ≤ t ≤ 20/4.905.
EXAMPLE 1.17
Find the largest possible domain for the function
'
8 + 2x − x 2
f (x) = .
x+1
SOLUTION We begin by factoring the quadratic in the numerator

$
(2 + x)(4 − x)
f (x) = ,
x+1
and note that f (x) is defined whenever (2 + x)(4 − x)/(x + 1) ≥ 0. To determine when this
is true, we examine, in tabular form, signs of the individual factors. The first line in Table 1.1
indicates that x + 2 is positive for x > −2, and is negative for x < −2. The second and third
lines show similar results for 4 − x and x + 1. The last line of the table counts the number of
negative signs in the lines above it for the intervals x < −2, −2 < x < −1, −1 < x < 4,
and x > 4 to arrive at the sign of (2 + x)(4 − x)/(x + 1) . When we note that division points
x = −2 and x = 4 are acceptable, but x = −1 is not (it gives division by 0), the largest
domain of the function consists of all values of x in the intervals x ≤ −2 and −1 < x ≤ 4.
TABLE 1.1
−3 −2 −1 0 1 2 3 4 5 x
x+2 − +
4−x + −
x+1 − +
(2 + x)(4 − x)/(x + 1) + − + −
EXAMPLE 1.18
Does the equation x − 4 − y 2 = 0 define y as a function of x for x ≥ 4?
SOLUTION For any x > 4, the equation has two solutions for y :
√
y = ± x − 4.
Since the equation does not define exactly one value of y for each value of x , it does not define
y as a function of x .
If we add to the equation in Example 1.18 an additional restriction such as y ≥ 0, then y is

defined as a function of x , namely,
√
y = x − 4.
Note that x − 4 − y 2 = 0 does, however, define x as a function of y :
x = y 2 + 4.
In this case, y is the independent variable and x is the dependent variable. In other words,
whenever an equation (such as x − 4 − y 2 = 0) is to be regarded as defining a function, it must
be made clear which variable is to be considered as independent and which as dependent. For
example, the equation x − 4 − y = 0 defines y as a function of x , and x as a function of y .
In the study of calculus and its applications we are interested in the behaviour of functions;
that is, for certain values of the independent variable, what can we say about the dependent
variable? The simplest and most revealing method for displaying characteristics of a function
is a graph. To obtain the graph of a function f (x) we use a plane coordinatized with Cartesian
coordinates x and y (in short, the Cartesian xy -plane). The graph of the function f (x)
is defined to be the curve with equation y = f (x) , where x is limited to the domain of
the function. For example, the graph of the function f (x) = x 3 − 27x + 1 is shown in
Figure 1.38.
FIGURE 1.38 Graph of a function
y
y = x 3 − 27x + 1
40
20
−6 −4 −2 2 4 6 x
−20
−40
This curve is a pictorial or geometric representation of the function f (x) = x 3 − 27x + 1.

The value of the function for a given x is visually displayed as the y -coordinate of that point on
the curve with x -coordinate equal to the given x . This graph clearly illustrates properties of the
function that may not be so readily obvious from the algebraic definition f (x) = x 3 − 27x + 1.
For instance,
1. For negative values of x , the largest value of f (x) is f (−3) = 55; for positive values of
x , the smallest value of f (x) is f (3) = −53.
2. As x increases, values of f (x) increase for x < −3 and x > 3, and values of f (x)
decrease for −3 < x < 3.
3. f (x) is equal to zero for three values of x — one a little less than −5, one a little larger
than 0, and one a little larger than 5.
Graphing calculators and computers have become indispensable tools for modern scientists;
one reason for this is the ability of these electronic devices to graph functions so quickly. We
hasten to point out, however, that machine-generated graphs may sometimes be misleading; care
must be taken not to make rash assumptions based on machine output. We illustrate with some
examples below. In addition, because calculators and computers have finite screen resolution,
it may not always be perfectly clear how to interpret every aspect of a machine-generated
graph. Mathematical analysis may be required to confirm or deny what is being suggested by
the graph. This is very important! We use machine-generated graphs extensively, but we are
always prepared to corroborate what we see, or don’t see, with rigorous mathematical analysis.
Recall that when we ask you to plot a graph, we intend for you to use a graphing calculator or
computer. When we ask you to draw a graph, we expect you to do so without these devices.
What follows are some examples of computer-generated graphs that are misleading.
EXAMPLE 1.19
A very powerful software package once gave the graph for the function f (x) = x + sin (2π x)
on the interval 0 ≤ x ≤ 24 in Figure 1.39. (It doesn’t in newer versions of the package.) How
do you feel about what you see?
FIGURE 1.39 Incorrect graph produced by computer program
20
15
10
5 10 15 20 x
SOLUTION The graph appears to be the straight line y = x , implying that therefore
sin (2π x) is always zero. This is ridiculous; it is a fluke, resulting from the choice of sam-
pling points taken by the software package in plotting the graph. Choosing the plot interval to
be 0 ≤ x ≤ 25 gave Figure 1.40. This is still misleading. The plot in Figure 1.41 gives a true
indication of the nature of the function. Examples like this are rare, but they can occur. Use
your common sense.
FIGURE 1.40 Another incorrect FIGURE 1.41 Correct graph produced

graph produced by a computer program by a computer program
y y
25 20
20
15
15
10
10
5 5
5 10 15 20 25 x 5 10 15 20 x
We encountered absolute values in Section 1.3. The absolute value function, denoted by |x| ,
is defined as (
−x, x < 0,
|x| = (1.29)
x, x ≥ 0.
It defines the size or magnitude of its argument without regard for sign. The graph of the function
is composed of two straight lines with slopes ±1 meeting at right angles at the origin. This does
not appear to be the case in the computer-generated plot of Figure 1.42. Why?
FIGURE 1.42 Graph of |x|
1.5
0.5
−2 −1 1 2 x
EXAMPLE 1.20
The most up-to-date version of the software package that produced the graph in Figure 1.39
yields Figure 1.43a when asked to plot the function f (x) = x 1/3 on the interval −8 ≤ x ≤ 8.
It is also accompanied by warning messages to the effect that values of f (x) are not real for
negative values of x . Define input for the computer so that a proper graph is generated.
FIGURE 1.43a Incorrect FIGURE 1.43b Correct

graph of x 1/3 graph of x 1/3
y y
2 2
1.5
1.5 1
0.5
1
−8 −6 −4 −2 2 4 6 8x
−0.5
0.5
−1
−1.5
2 4 6 8 x −2
SOLUTION To fully explain what the computer is doing, we need complex numbers from
Appendix C. To be brief, every nonzero real number has three cube roots, one of which is
real and two of which are complex. The computer is programmed to yield the real cube root
of positive real numbers (81/3 = 2, for instance). Unless directed otherwise, however, the
computer, and maybe your calculator, produces a complex number when asked for the cube
root of a negative real number such as −8. It does not yield −2, which is also a cube root
of −8, unless specifically told to do so. We can instruct the computer to do this by redefining
f (x) = x 1/3 as
(
−|x|1/3 , x < 0,
f (x) =
x 1/3 , x ≥ 0.
A plot of this function is shown in Figure 1.43b.
Quite often the argument of the absolute value function is a function of x rather than x
itself. Such is the case in the following example.
EXAMPLE 1.21
A computer-generated graph of the function f (x) = |x 3 + 5| is shown in Figure 1.44. It
appears to touch the x -axis somewhere between x = −2 and x = −1, but we cannot be
positive of this graphically. In addition, we cannot be sure whether the graph is rounded or
whether it comes to a sharp point at this same value of x . Perform some elementary graphing
to answer these questions.
FIGURE 1.44 Graph of |x 3 + 5|
y
12
10
8
6
4
2
−2 −1 1 2 x
SOLUTION To draw the curve y = |x 3 + 5| , we begin with y = x 3 in Figure 1.45a. Next

we draw the curve y = x 3 + 5 by adding 5 to every ordinate in Figure 1.45a. The result in
Figure 1.45b is the curve in Figure 1.45a shifted upward 5 units. It crosses the x -axis at −51/3 .
The last step is to take the absolute value of every ordinate on the curve y = x 3 + 5. This
FIGURE 1.45a FIGURE 1.45b FIGURE 1.45c

Building the graph of |x 3 + 5|
y y y
y= x3 y = x3 + 5 y = |x 3 + 5|
−51/3
x x −51/3 x
changes no ordinate that is already positive, but changes the sign of any ordinate that is negative.
The result and final drawing is shown in Figure 1.45c. Clearly then, the graph touches the x -axis
at x = −51/3 , and there is a sharp point on the graph at (−51/3 , 0) .
We might get the impression from Figures 1.42 and 1.45c that absolute values lead to curves
with sharp points. This is not always the case. When absolute values are placed around the
function x 3 in Figure 1.46a, the resulting function |x 3 | is shown in Figure 1.46b. There is no
sharp point at the origin.
FIGURE 1.46a FIGURE 1.46b

Taking absolute values does not always lead to sharp points on a curve
y y
1 1
y = x3 y = |x 3|
0.5 0.5
−0.5
−1 0.5 1 x −1 −0.5 0.5 1x
−0.5 −0.5
−1 −1

Graphs to illustrate that taking absolute values sometimes leads to sharp points on a curve
y y
10 10 y = |2x 3 − 11x 2 + 16x − 7|

y = 2x 3 − 11x 2 + 16x − 7
5 5
1 2 3 4 x 1 2 3 4 x
−5 −5
−10 −10
The absolute value of the function 2x 3 − 11x 2 + 16x − 7 in Figure 1.47a leads to the
graph in Figure 1.47b for |2x 3 − 11x 2 + 16x − 7| . It has a sharp point at the position between
x = 3 and x = 4 where it touches the x -axis, but not at x = 1.
EXAMPLE 1.22
√
FIGURE 1.48 The top Plot a graph of the function f (x) = 2 4 − x 2 (using a calculator or computer). Does the
half of an ellipse graph look familiar? Confirm what it suggests.
y SOLUTION The domain of the function is −2 ≤ x ≤ 2.√Its graph in Figure 1.48 appears to be
the top half of an ellipse. To confirm this, we square y = 2 4 − x 2 , obtaining y 2 = 4(4 −x 2 ) ,
4 or 4x 2 + y 2 = 16. √ This is indeed the equation of an ellipse. Only the top half of the ellipse is
y = 2 4 − x2 defined by y = 2 4 − x 2 since the square root requires y to be nonnegative.
−2 2 x
EXAMPLE 1.23
√
Repeat Example 1.22 with f (x) = 4 − x2 .
SOLUTION
√ The graph in Figure 1.49a once again suggests the top half of an ellipse. Squaring
y = 4 − x 2 leads to x 2 + y 2 = 4, a circle, not an ellipse. The reason the graph does not
appear to be semicircular is that the computer has chosen different scales for the x - and y -axes.
If we instruct the computer to use equal scales, the graph in Figure 1.49b does indeed look like
a semicircle.
FIGURE 1.49a A semicircle that does not look like a semicircle FIGURE 1.49b A semicircle that does look like a semicircle
y y
2 2
y= 4− x2
y= 4 − x2
1.5 1.5
1 1
0.5 0.5
−2 −1 1 2 x −2 −1 1 2 x
EXAMPLE 1.24
√ √
Plot a graph of the function f (x) = x + 3x 2 − 4, x ≥ 2/ 3. Does it appear to have an
asymptote? Confirm this mathematically.
SOLUTION The plot in Figure 1.50a appears to be fairly straight, suggesting an asymptote.
To confirm this, we note that for large x , the 4 is insignificant compared to 3x 2 , and there-
fore
√ √ √
f (x) ≈ x + 3x 2 = x + 3x = (1 + 3)x.
√
Thus, the line y = (1 + 3)x is an asymptote for the graph. It is shown along with the graph
of the function in Figure 1.50b.
FIGURE 1.50a Graph that suggests an asymptote FIGURE 1.50b Graph showing asymptote
y _
y y = (1 + √3)x
8
8
6
y = x + 3x 2 − 4 6
4 y = x + 3x 2 − 4
4
2
2
0
1 2 3 4 x 0
1 2 3 4 x
EXAMPLE 1.25
When a ball is dropped from the top of a building 20 m high at time t = 0, the distance d (in
metres) that it falls in time t is given by the function d = f (t) = 4.905t 2 . Draw its graph.
FIGURE 1.51 Distance
fallen by a ball dropped from a SOLUTION In this example independent and dependent variables are denoted by letters t and
height of 20 m d , which suggest their physical meaning — time and distance — rather than the generic labels
d x and y . With the axes labelled correspondingly as the t -axis and d -axis, we draw that part
of the parabola d = 4.905t 2 in Figure 1.51. The remainder of the parabola has no physical
20 significance in the context of this problem.
d = 4.905t 2
10
20
4.905 Even and Odd Functions
1 2 t The parabola y = x 2 + 1 in Figure 1.52 is the graph of the function x 2 + 1; it is symmetric about
the y -axis. The curve in Figure 1.53 is the graph of the function (x 4 − 2x 2 + 5)/(x 2 + 6) ; it is
also symmetric about the y -axis. These are examples of a special class of functions identified
in the following definition.
DEFINITION 1.4
A function f (x) is said to be an even function if for each x in its domain
f (−x) = f (x); (1.30a)
it is said to be an odd function if for each x in its domain
f (−x) = −f (x). (1.30b)
The first of these implies that the equation y = f (x) for the graph of an even function is
unchanged when x is replaced by −x ; therefore, the graph of an even function is symmetric
about the y -axis. As a result, f (x) = x 2 + 1 and f (x) = (x 4 − 2x 2 + 5)/(x 2 + 6) are even
functions.
Equation 1.30b implies that if (x, y) is any point on the graph of an odd function, so too is
the point (−x, −y) . This is illustrated by the graph of the odd function f (x) = 2x 5 − 3x 3 +x
FIGURE 1.52 An even function FIGURE 1.53 An even function
y y
y = x2 + 1 3.5 x 4 − 2x 2 + 5
y=
8 3 x2 + 6
2.5
6
2
4 1.5
1
2 0.5
−2 −1 0 1 2 x
−2 −1 1 2 x
in Figure 1.54. Another way to describe the graph of an odd function is to note that either half
( x < 0 or x > 0) is the result of two reflections of the other half, first in the y -axis and then
in the x -axis. Alternatively, either half of the graph is a result of rotating the other half by π
radians (one-half a revolution) around the origin.
FIGURE 1.54 An odd function
y
0.75
y = 2x 5 − 3x 3 + x 0.5
0.25
−1
−2 1 2 x
−0.25
−0.5
−0.75
EXAMPLE 1.26
Which of the following functions are even, odd, or neither even nor odd?
!
FIGURE 1.55a Half the
(a) f (x) = |x| (b) f (x) = x 5 − x (c) f (x) = x 2 + x
graph of an even function
Draw a graph of each function.
y SOLUTION
y= x (a) Since
! !
f (−x) = |− x| = |x| = f (x),
√
this function is even. Its graph, the curve y =
√ |x| , is symmetric about the y -axis.
x When x > 0, this equation becomes y = x , the half-parabola in Figure 1.55a.
The complete graph of the function is shown in Figure 1.55b.
(b) Since
FIGURE 1.55b The full f (−x) = (−x)5 − (−x) = −x 5 + x = −(x 5 − x) = −f (x),
graph from symmetry
y
this function is odd. Its graph is shown in Figure 1.56. The right half of the graph
in Figure 1.56 was drawn by analyzing signs of f (x) in the factored form f (x) =
y = |x|
x(x 2 + 1)(x − 1)(x + 1) . The left half is the result of a half revolution of the right
half about the origin.
(c) The function f (x) = x 2 + x is neither even nor odd. Its graph is a parabola that
x opens upward, crossing the x -axis at x = 0 and x = −1 (Figure 1.57).
FIGURE 1.56 An odd function x 5 − x FIGURE 1.57 A function that is neither even nor odd
y y
y = x2 + x
y = x5 − x
−1
1 x
−1 x
We have seen even functions, odd functions, and functions that are neither even nor odd. Can
functions be both even and odd? Such a function would have to satisfy both 1.30a and 1.30b,
and hence
f (x) = −f (x).
But this implies that f (x) = 0, and this therefore is the only even and odd function.
It is often useful to divide a function into what are called its even and odd parts. For example,
if f (x) is defined for all real x , we may write that
% & % &
f (x) + f (−x) f (x) − f (−x)
f (x) = + . (1.31)
2 2
It is straightforward to show that [f (x)+f (−x)]/2 is an even function, and [f (x)−f (−x)]/2
is an odd function. In other words, we have written f (x) as the sum of an even function and
an odd function; they are called the even and odd parts of f (x) . If we denote them by
f (x) + f (−x) f (x) − f (−x)
fe (x) = and fo (x) = , (1.32a)
2 2
then we have
f (x) = fe (x) + fo (x). (1.32b)
EXAMPLE 1.27
x+1
Write the functions (a) f (x) = x 3 − 2x 2 + x + 5 and (b) g(x) = in terms of their
x−2
even and odd parts.
SOLUTION Solution
(a) Clearly, the even and odd parts of f (x) are fe (x) = −2x 2 + 5 and fo (x) = x 3 +x ,
respectively. Thus,
f (x) = (−2x 2 + 5) + (x 3 + x).
(b) The even and odd parts of g(x) are
% & % &
1 x+1 −x + 1 x2 + 2 1 x+1 −x + 1 3x
ge (x) = + = 2 , go (x) = − = 2 .
2 x−2 −x − 2 x −4 2 x−2 −x − 2 x −4
Thus,
x2 + 2 3x
g(x) = + 2 .
x2 − 4 x −4
Polynomials and Rational Functions

Two of the most important classes of functions are polynomials and rational functions. A
polynomial of degree n is a function of the form
f (x) = an x n + an−1 x n−1 + · · · + a2 x 2 + a1 x + a0 , (1.33)
where n is a nonnegative integer, and a0 , a1 , . . . , an are real numbers ( an $= 0). They are
defined for all values of x . Graphs of linear and quadratic polynomials are straight lines and
parabolas. The cubic polynomial x 3 + 12x 2 − 40x + 6 and the quartic x 4 + 10x 3 + 6x 2 − 64x + 5
are plotted in Figures 1.58 and 1.59. Figure 1.58 has a low point just to the right of x = 1 and
a high point just to the left of x = −9. We could approximate these points more and more
closely by reducing the plot interval. Calculus provides an analytic way to determine them.
FIGURE 1.58 A cubic polynomial FIGURE 1.59 A quartic polynomial
y y
y = x 3 + 12x 2 − 40x + 6
600
600
400
y = x 4 + 10x 3+ 6x 2 − 64x + 5 400
200
−15 200
−10 −5 5 x −8 −4
−200 −6 −2 2 4 x
−200
−400
A rational function R(x) is defined as the quotient of two polynomials P (x) and Q(x) :
P (x)
R(x) = . (1.34)
Q(x)
Rational functions are undefined at points where Q(x) = 0. For example, the rational function
x 3 + 3x 2 − 12x + 1
f (x) =
x 2 − 2x − 3
is undefined at x = −1 and x = 3. The computer-generated graph on the interval −2 ≤ x ≤ 4

in Figure 1.60a indicates that function values become very large positively and negatively near
x = −1 and x = 3. We investigate this behaviour in detail in Chapter 2. The graph may or
may not be accompanied by messages indicating that division by 0 is encountered at x = −1
and x = 3, depending on the plot interval specified to the plotting device. A plot on the larger
interval −10 ≤ x ≤ 10 in Figure 1.60b suggests that the graph has an asymptote. We shall see
why in Chapter 2.
FIGURE 1.60a Unbounded behaviour of a FIGURE 1.60b Suggested asymptote of a

rational function rational function
y y
100 x 3+ 3x 2 −12x + 1 40
y=
x 2 − 2x − 3
50 20
−10
−2 −1 1 2 3 4 x −5 5 10 x
−50
−20
−100
Every function f (x) can be represented pictorially by its graph, the curve with equation
y = f (x) . But what about the reverse situation? Does every curve in the xy -plane represent
a function f (x) ? The curves in Figure 1.61, which both extend between x = a and x = b ,
illustrate that the answer is no. The curve in Figure 1.61a represents a function, whereas the
curve in Figure 1.61b does not, because for values of x between a and c there are two possible
values of y . In other words, a curve represents a function f(x) if every vertical line that intersects
the curve does so at exactly one point.
FIGURE 1.61a Curve that is the graph FIGURE 1.61b Curve that is not the
of a function graph of a function
y y
FIGURE 1.62 Translated
circles
y
x 2 + y2 = r 2
r (x − c)2 + y 2 = r 2 a b x x
a c b
r c x
c
Translation of Curves
In Section 1.4 we saw how graphs of conic sections can be shifted in the xy -plane, and how
these shifts, or translations as they are called, are reflected in the equations of the curves. This
principle applies to all curves, not just conic sections. When every x in the equation of a curve
FIGURE 1.63 Translated is replaced by x − c , where c is a constant, the curve is shifted c units to the right. When every
parabolas x is replaced by x + c , the curve is shifted c units to the left. For example, when x in the
equation x 2 + y 2 = r 2 is replaced by x − c , the centre of the resulting circle is shifted from the
y
origin to the point (c, 0) (Figure 1.62). When each x in the parabola y = x 2 + x is replaced
y = (x + c)2 + (x + c) y = x 2 + x by x + c ,
= (x + c)(x + c + 1) = x(x + 1)
y = (x + c)2 + (x + c) = (x + c)(x + c + 1),
the parabola is shifted c units to the left (Figure 1.63).
Vertical shifts result when y is replaced by y ± c in the equation of a curve. For instance,
−c − 1 −c −1 x
the curve x 2 − (y + 1)2 = 1 is the hyperbola x 2 − y 2 = 1 in Figure 1.64a shifted downward
c 1 unit (Figure 1.64b).
FIGURE 1.64a The hyperbola x 2 − y 2 = 1 FIGURE 1.64b The hyperbola x 2 − y 2 = 1 translated
y y
y = −x y=x y = −x − 1 y=x−1
x2 − y2 = 1
x
(−1, −1) (1, −1)
−1 1 x
x 2 − ( y + 1)2 = 1
EXAMPLE 1.28
Describe the relationship between the curves |x| + |y| = 1 and |x| + |y − a| = 1, where
a > 0 is a constant. Draw both curves.
SOLUTION Since |x| + |y − a| = 1 can be obtained from |x| + |y| = 1 by replacing y by
y − a , the first curve is the second shifted a units upward. The curve |x| + |y| = 1 is easily
drawn without an electronic device. Since this equation remains unchanged when x is replaced
by −x and y is replaced by −y , the curve |x| + |y| = 1 is symmetric about both the x -axis
and the y -axis. This means that we can concentrate our efforts on drawing the graph in the first
quadrant, where the equation reduces to x + y = 1. The segment of this straight line in the
first quadrant is shown in Figure 1.65a. To obtain |x| + |y| = 1 (Figure 1.65b), we reflect
this curve in the axes. Finally, |x| + |y − a| = 1 may be obtained by shifting |x| + |y| = 1
upward a units (Figure 1.65c).
FIGURE 1.65a One-quarter of FIGURE 1.65b |x| + |y| = 1 FIGURE 1.65c |x|+|y −a| = 1
|x| + |y| = 1 from symmetry by translation
y y y
1 1 a+1 |x| + |y − a| = 1
x+y=1 |x| + |y| = 1
(−1, a) (1, a)
1 x −1 1 x
a−1
−1
EXERCISES 1.5
In Exercises 1–8 find the largest possible domain of the function. 25. Verify that when an odd function f (x) is defined at x = 0, its
value must be f (0) = 0.
√ 1
1. f (x) = 9 − x2 2. f (x) =
x−2 26. Prove each of the following:
1 1
3. f (x) = √ 4. f (x) = √ (a) The product of two even functions or two odd functions is
x x2 + 4 x x2 − 4
' an even function.
1 x2 − 4
∗ 5. f (x) = √ ∗ 6. f (x) = (b) The product of an even and an odd function is an odd func-
x 4 − x2 9 − x2
tion.
√ √
∗ 7. f (x) = 4 −9 + 6x − x 2 ∗ 8. f (x) = x 3 − x 2
In Exercises 9–18 determine algebraically whether the function is even, In Exercises 27–56 first draw, then plot, a graph of the function.
odd, or neither even nor odd. Plot each function to confirm your con-
clusion geometrically. √ √
27. f (x) = 5 − x2 28. f (x) = − 5 − x 2
9. f (x) = 1 + x 2 + 2x 4 10. f (x) = x 5 − x √ √
29. f (x) = − x 2 − 5 30. f (x) = x2 − 5
11. f (x) = 12x 2 + 2x 12. f (x) = x 1/5 √ √
31. f (x) = 5 − 4x 2 32. f (x) = − 5 − 4x 2
x−1
13. f (x) = 14. f (x) = x(x 2 + x)
x+1 33. f (x) = |x| + 2x 34. f (x) = |x| − 2x
x|x| √ √
15. f (x) = 16. f (x) = x 3 + x 35. f (x) = − −x 36. f (x) = x 1/3
3 + x2
√ 37. f (x) = x 2/3 38. f (x) = 3x 3/2
x2 + 1 x2 + x4 + 1
∗ 17. f (x) = ∗ 18. f (x) = √
1 − 2x 4 x 6 + 3x 2 ∗ 39. f (x) = 3|x|3/2 ∗ 40. f (x) = x x + 1
√ √
∗ 41. f (x) = x x 2 − 1 ∗ 42. f (x) = −x 4 − 9x 2
In Exercises 19–24 find the even and odd parts of the function, if they √ √
exist. ∗ 43. f (x) = x 2 4 − x ∗ 44. f (x) = x 2 x 2 − 4
x−2 √
19. f (x) = x 3 + 3x 2 − 2x 20. f (x) = ∗ 45. f (x) = x 2 4 − 9x 2 ∗ 46. f (x) = |x 2 − x − 12|
x+5 √ √
x3 ∗ 47. f (x) = 2x − x 2 ∗ 48. f (x) = 2x − 4x 2
21. f (x) = |x| 22. f (x) = √ √
x2 +3 ∗ 49. f (x) = 4x 2 − 2x ∗ 50. f (x) = 2x − x 2 − 4
2x √
23. f (x) = 24. f (x) = x−1 √
3 + 5x ∗ 51. f (x) = x + 2 + x
√ √ !
∗ 52. f (x) = 9 − 4x 2 + 4x 2 − 9 ∗ 73. x 2 − (y + 5)2 = 1 ∗ 74. x = − 64 + 9(y + 5)2
∗ 53. f (x) = x 2 + |x| − 2 ∗ 75. y = (x + 1)4 − (x + 1)2 ∗ 76. |x| + |y + 2| = 1
! ) )
∗ 54. f (x) = (x 2 − 4)2 ∗ 77. y = )3 − |x + 2|) − 1 ∗ 78. |x| − |y| = 1
! √
∗ 55. f (x) = 2 − 1 + x ∗ 79. x = 4 − |y|
! ∗ 80. (y − 1)2 = (x − 1)2 [1 − (x − 1)2 ]
∗ 56. f (x) = (x 2 − 1)2 − (x 2 − 1)
57. What condition ensures that a curve in the xy -plane represents a ∗ 81. An electronic signal is defined by
function x = f (y) ?  0, t < 0,


 t, 0 ≤ t ≤ 1,
A curve is translated vertically when each y in its equation is replaced
s(t) =

 (3 − t)/2, 1 ≤ t ≤ 3,
by y + c ; it is shifted horizontally when each x is replaced by x + c . 
0, t > 3,
If each x in the equation of a curve is replaced by cx , where c > 1 is
a constant, the curve is compressed by a factor of c in the x -direction. where t is time. Draw graphs of the time-shifted signals s(t − 2) and
When 0 < c < 1, the curve is stretched by a factor of 1/c in the s(t + 1) .
x -direction. When each y is replaced by cy , the curve is compressed
or stretched in the y -direction. Illustrate this by drawing and plotting ∗ 82. The graph of an electronic signal s(t) is shown in the figure below,
the pair of curves in Exercises 58–63 . where t is time.
(a) Find an algebraic representation for s(t) .
58. y = x 2 , y = x 2 /9 = (x/3)2 (b) Draw graphs of the time-shifted signals s(t + 1/2) and
59. y = x, 5y = x s(t − 3) .
60. x 2 + y 2 = 1, x 2 + 16y 2 = 1 (c) What are algebraic representations for the graphs in part
(b)?
61. x 2 − y 2 = 1, x 2 − 4y 2 = 1
∗ 62. |x| + |y| = 1, |x| + |y/2| = 1
4
∗ 63. x 2 − (y + 5)2 = 1, x 2 − (y/2 + 5)2 = 1 Straight line
Parabola
When each x is replaced by −x in the equation of a curve, the curve −1 1 2 3 4
is reflected in the y -axis. When each y is replaced by −y , the curve is
reflected in the x -axis. Illustrate this by drawing and plotting the pair ∗ 83. A cherry orchard has 255 trees, each of which produces an average
of curves in Exercises 64–67 . of 25 baskets of cherries. For each additional tree planted, the yield per
3 2 3 2
tree decreases by one-twelfth of a basket. If x represents the number
64. y = x − 3x , y = −x − 3x of extra trees (beyond 255) and Y the total yield, find Y as a function
65. (x − 2)2 + y 2 = 4, (x + 2)2 + y 2 = 4 of x , and draw its graph. How many more trees should be planted for
maximum yield?
66. x = y 2 − 2y, x = y 2 + 2y
√ √ ∗ 84. A rectangle with sides parallel to the axes is inscribed inside the
67. x = y, x = −y ellipse b2 x 2 + a 2 y 2 = a 2 b2 (figure below). Find a formula for the
area A of the rectangle in terms of x . Draw a graph of this function.
∗ 68. The floor function is defined by f (x) = -x. = greatest integer y
that does not exceed x .
(a) Draw a graph of the floor function. b 2x 2 + a 2 y 2 = a 2b 2
b
(b) If first-class postage is 51 cents for each 50 g, or fraction (x, y)
thereof, up to and including 500 g, draw a graph of this cost
y
function.
(c) Express the cost function in part (b) in terms of the floor x a x
function.
In Exercises 69–80 draw the curve. Indicate whether the curve defines
∗ 85. A man 2 m tall walks along the edge of a straight road that is 5 m
y as a function of x .
wide. On the other edge of the road stands a street light 10 m high. Find
2
∗ 69. y = x − x 6 2
∗ 70. y = (x + 1) − 2(x + 1) a functional relationship for the length of the man’s shadow in terms of
! his distance from the point on his side of the road directly across from
2 2
∗ 71. (x − 2) + y = 4 ∗ 72. y = 4 − (x − 2)2 the light. Plot a graph of this function.
∗ 86. When two substances, A and B , are brought together, a chemical ∗ 90. A box measuring 1 m on each side is attached to a rope as shown in
reaction takes place to form a new substance, C . It requires 2 L of A the figure below. The rope passes over a pulley 10 m from the ground
for each litre of B to produce 3 L of C . The rate R at which A and B and a truck pulls on the other end in a horizontal direction along a line
react to form C is proportional to the product of the amounts of A and 1 m above the ground. We denote positions of the truck and the bottom
B present at that instant. If the original amounts of A and B are 20 L of the box by x and y , respectively. Find y as a function of x if the
and 40 L, respectively, and if x represents the amount of C present in truck starts at position x = 5 m and stops when the top of the box
the reaction at any given time, find a formula for R as a function of x . touches the pulley. Assume that the length of rope between truck and
Draw a graph of this function, and determine when the reaction rate is box is 25 m. Draw a graph of the function.
highest.
∗ 87. In classical physics and engineering, the mass m of an object is
constant, independent of how fast it is moving. In special relativity,
however, m is given by the formula
1
m0 10 y
m= ! ,
1 − (v 2 /c2 )
where m0 is the mass of the object when it is not moving, v is the speed
of the mass, and c is a constant (the speed of light). Draw a graph of x 1
this function, and draw any conclusions that you feel are suggested.
∗ 88. A square plate 4 m on each side is slowly submerged in a large
tank of water. One diagonal is kept vertical and lowered at a rate
of 0.5 m/s, entering the water at time t = 0. If A is the area of the In Exercises 91–94 draw a graph of the function where -x. is the floor
submerged portion of the surface (one side only) at time t until complete function of Exercise 68.
submersion occurs, find A as a function of t and draw its graph.
∗ 91. f (x) = -2x. ∗ 92. f (x) = x + -x.
∗ 89. Because of construction, no passing is permitted on a 10-km stretch
of highway. If cars travel at v km/h along this stretch, a safe distance
∗ 93. f (x) = x-x. ∗ 94. f (x) = -x + -x..
between them must be maintained, and this distance increases as v ∗ 95. Is -f (x) + g(x). = -f (x). + -g(x). ?
increases. In particular, the highway traffic commission has determined
that for speeds over 50 km/h, the distance in metres between cars should ∗ 96. If a thermal nuclear reactor is built in the shape of a right circular
be at least cylinder of radius r and height h , then neutron diffusion theory requires
3v 2 r and h to satisfy an equation of the form
d = .
500
If it is supposed that everyone maintains the safe distance and the same a2 b2
+ = 1,
constant speed v through the stretch, find the number q of cars leaving r2 h2
the “bottleneck” per hour as a function of speed v . Draw a graph
of this function for 50 ≤ v ≤ 100, and determine the speed that where a and b are positive constants. Draw a graph of the function
maximizes q . r = f (h) defined by this equation for h ≥ 2b .

1. −2 ≤ x ≤ 2 (1 mark) 11. The quotient of two polynomials (1 mark)
√
2. All reals except x = 1 ± 5 (2 marks)
3. All reals ≤ −4 (2 marks) 12. (2 marks)
y
4. All reals ≥ 3 (2 marks)
5. Even (1 mark) 8
6. Odd (1 mark)
7. Neither (1 mark)
8. Neither (1 mark)
9. −x 2 /(1 − x 2 ) , x/(1 − x 2 ) (3 marks) −8 8 x
10. No (1 mark)
13. (2 marks) 17. (3 marks)

y y
3
Semi-circle
−2 2 x
−3 3 x
14. (2 marks) 18. (2 marks)

y y
x (−2, 1) (1, 1)
(1, −2)
Half a parabola x
19. (2 marks)
15. (3 marks) y
y
−1 1 x
(−3, 1) 1
y = 2 − 2x2
−2 x
16. (3 marks)
y 20. (2 marks)
y
8
(−1/2, 1) (1, 1)
x
2 x
1.6 Inverse Functions

This material is likely to be new for many students. It does not therefore have a diagnostic test
associated with it. If you have previously studied inverse functions, it would still be advisable to
at least read the section to ensure that you are familiar with the vocabulary, notation, and theory.
When we speak of a function y = f (x) , there is a unique y associated with each x ; that
is, given a value of x in the domain of f (x) , the function associates one — and only one —
value of y in the range. However, it may happen, and quite often does, that a value of y in
the range of the function may be associated with more than one value of x . For example, each
y > 0 in the range of √ the function y = f (x) = x 2 (Figure 1.66) is associated with two values
of x , namely, x = ± y .
Some functions have the property that each value of y in the range arises from only one x
in the domain. For instance, given any value y in the range of the function y = f (x) = x 3
in Figure 1.67, there is a unique x such that y = x 3 , namely, x = y 1/3 . Such a function
is said to be one-to-one. Formally, we say that a function f (x) is one-to-one if for any two
distinct values x1 and x2 in the domain of f (x) , it follows that f (x1 ) $= f (x2 ) . To repeat,
a one-to-one function f (x) has the property that given any y in its range, there is one — and
only one — x in its domain for which y = f (x) . We can therefore define a function that maps
values in the range of f (x) onto values in the domain, a function that maps y onto x if x is
mapped by f (x) onto y . We call this function the inverse function of f (x) . It reverses the
action of f (x) . For the function f (x) = x 3 in Figure 1.67, the inverse function is the function
that takes cube roots of real numbers.
FIGURE 1.66 A function that is not one-to-one FIGURE 1.67 A function that is one-to-one
y y
y = g (x) = x 2
y
y
x = y1/ 3 x
x=− y x= y x
y = f (x) = x 3
Unless there is a good reason to do otherwise, it is our custom to denote the independent
variable of a function by the letter x and the dependent variable by y . For the inverse function
of a function y = f (x) , there is a natural tendency to denote the independent variable by y and
the dependent variable by x, since the inverse function maps the range of f (x) onto its domain.
When discussing general properties of inverse functions, as we do in this section, it is usually
better to maintain our usual practice of using x as independent variable and y as dependent
variable even for inverse functions. This may not be advisable in applications when letters for
independent and dependent variables represent physical or geometric quantities.
When a function f (x) has an inverse function, we adopt the notation f −1 (x) to represent
the inverse function. For example, when f (x) = x 3 , the inverse function is f −1 (x) = x 1/3 .
Be careful with this notation. Do not interpret the “ −1” as a power and write f −1 (x) as
1/f (x) . This is not correct. The notation f −1 represents a function, just as tan represents the
√
tangent function and represents the positive square root function.
The inverse function f −1 (x) of a function f (x) “undoes” what f (x) “does”; it reverses
the effect of f (x) . For example, the function f (x) = x 2 ,√x ≥ 0 in Figure 1.68 is one-to-one;
it squares nonnegative numbers. Its inverse is f −1 (x) = x , the positive square root function
(Figure 1.69). Squaring a positive number x and then taking the positive square root of the
inverse of g(x) = x 2 , x ≤ 0 (Figure 1.70), is the
result returns the original x . Similarly, the √
−1
negative square root function g (x) = − x (Figure 1.71).
√
FIGURE 1.68 The function x 2 , x ≥ 0 FIGURE 1.69 The inverse function x
y y
y = f −1 (x) = x
y = f (x) = x2
x
x
√
FIGURE 1.70 The function x 2 , x ≤ 0 FIGURE 1.71 The inverse function − x
y y
y = g (x) = x 2 x
y= g−1 (x) =− x
The fact that f −1 (x) undoes f (x) can be stated algebraically as follows. For each x in
the domain of f (x) ,
. /
f −1 f (x) = x. (1.35)
This is the defining relation for an inverse function. We understand that the domain of f −1 (x)
is the same as the range of f (x) , so that f −1 (x) operates on all outputs of f (x) .
EXAMPLE 1.29
What is the inverse function of f (x) = (x + 1)/(x − 2) ?
SOLUTION To find the algebraic definition of f −1 (x) , we solve y = (x + 1)/(x − 2) for
x in terms of y (you will see why in a moment). First we cross-multiply,
x + 1 = y(x − 2);
then group terms in x ,

x(y − 1) = 2y + 1;
and finally, divide by y − 1,
2y + 1
x = .
y−1
What have we accomplished by solving for x in terms of y ? If we take an x in the domain of
f (x) , then f (x) produces y = (x + 1)/(x − 2) . Now this same x and y satisfy the equation
x = (2y + 1)/(y − 1) because this equation is simply a rearrangement of y = (x + 1)/(x − 2) .
Consequently, if we substitute y into the right side of the equation,
2y + 1
x = ,
y−1
we obtain the original x . This equation must therefore define the inverse function of f (x) . In
other words, if we denote the independent variable by x , the inverse function is
2x + 1
f −1 (x) = .
x−1
Example 1.29 has illustrated that to find the inverse of a function y = f (x) , the equation should
be solved for x in terms of y , and then variables should be renamed. If the equation does not
have a unique solution for x in terms of y , then f (x) does not have an inverse function. Such
is the case for the function g(x) = x 2 in Figure 1.66. Solving y = g(x) = x 2 for x gives
√
two solutions x = ± y .
So far our discussion of inverse functions has been algebraic. The geometry of inverse
functions is most revealing. You may have noticed a relationship between the graphs of f (x)
and f −1 (x) in Figures 1.68 and 1.69, and of g(x) and g −1 (x) in Figures 1.70 and 1.71. The
FIGURE 1.72 Graph of inverse function is the mirror image of the function in the line y = x ; that is, graphs of inverse
x 2 and its mirror image in y = x
pairs are symmetric about the line y = x . These two examples are not mere coincidence,
y graphs of inverse functions are always mirror images of each other in the line y = x .
y=x
y = x2 This suggests that a very simple way to graphically determine the inverse of a function
f (x) is to take its mirror image in the line y = x . Note, too, that if the mirror image does not
represent a function, then no inverse function for f (x) exists. For example, the mirror image of
(1, 1)
y = x 2 , −∞ < x < ∞ , in the line y = x is shown in Figure 1.72 and does not represent a
x function. We conclude as before that this function does not have an inverse.
If y = f −1 (x) is the mirror image of y = f (x) in the line y = x , then y = f (x) is the
Mirror image of mirror image of y = f −1 (x) . This means that f (x) is the inverse of f −1 (x) , and that f (x)
y = x 2 in y = x
undoes what f −1 (x) does:
. /
f f −1 (x) = x. (1.36)
Geometrically, a function f (x) has an inverse if the reflection of its graph y = f (x) in
the line y = x represents a function. But we know that a curve represents a function if every
FIGURE 1.73 An increasing vertical line that intersects it does so at exactly one point. Furthermore, a vertical line intersects
function the reflected curve at exactly one point only if its horizontal reflection intersects y = f (x) at
y exactly one point. These two facts enable us to state that a function f (x) has an inverse function
Increasing if — and only if — every horizontal line that intersects it does so at exactly one point. This is
function a geometric interpretation of a function being one-to-one. See, for example, the functions in
Figures 1.68 and 1.70. Horizontal lines that intersect the curves do so at exactly one point. The
function in Figure 1.66, which has no inverse, is intersected in two points by every horizontal
line y = c > 0.
x
When the graph of a function always moves upward and to the right on an interval I (Figure
1.73), the function is said to be increasing on the interval. We shall discuss increasing functions
in Section 4.2. What is clear is that such a function passes the horizontal line test (and is therefore
one-to-one). Likewise, a function whose graph moves downward and to the right on an interval
I (Figure 1.74) is said to be decreasing on I ; it is one-to-one. A function that is either increasing
FIGURE 1.74 A decreasing
on an interval I or decreasing on I is said to be a strictly monotonic function on I . What we
function
have shown is the following result.
y
Decreasing
function
THEOREM 1.6
A function that is strictly monotonic on an interval has an inverse function on that interval.
x
It is important to realize that being strictly monotonic is a sufficient condition for existence
of an inverse function; that is, if a function is strictly monotonic, then it has an inverse. It is not,
however, a necessary condition. The function in Figure 1.75a is not strictly monotonic on the
interval −1 ≤ x ≤ 1, but it does have the inverse function in Figure 1.75b.

A function that is not strictly monotonic, and its inverse
y y
2 2
1 1
x −1 1 2 x
−1 1 2
−1 −1
EXAMPLE 1.30
Does the function f (x) = x 2 − 2x have an inverse? Does it have an inverse when its domain
is restricted to the interval x ≥ 1, and to the interval x ≤ 1?
SOLUTION The graph of f (x) in Figure 1.76 indicates that f (x) does not have an inverse.
When restricted to the interval x ≥ 1, the function is one-to-one, and does have an inverse. To
find it, we solve y = x 2 − 2x for x in terms of y . The quadratic formula applied to
x 2 − 2x − y = 0
gives
√
2± 4 + 4y !
x = = 1± y + 1.
2
√
Since x must be greater than or equal to 1, we must choose x = 1 + y + 1. Hence, the
inverse of f (x) = x 2 − 2x, x ≥ 1 is
√
f −1 (x) = 1 + x + 1.
FIGURE 1.76 A function that does not have an inverse
y = f (x) = x 2 − 2x
2 x
Similarly, when restricted to the interval x ≤ 1, f (x) has the inverse

√
f −1 (x) = 1 − x + 1.
This example illustrates that when a function is not strictly monotonic on an interval I , the
interval can usually be subdivided into subintervals on which the function is strictly monotonic,
and on each such subinterval the function has an inverse.
EXERCISES 1.6
In Exercises 1–14 determine graphically whether the function has an 9. f (x) = x + |x|
inverse. Find each inverse function. √
10. f (x) = 1 − x2
√
1. f (x) = 2x + 3 2. f (x) = x + 1
11. f (x) = x + 2x 2 + 2,
4
x≤0
x+5 2
3. f (x) = x 2 + x 4. f (x) = 12. f (x) = x − 2x + 4, x≥1
2x + 4 " #3
x+2
5. f (x) = 1/x 6. f (x) = 3x 3 + 2 13. f (x) =
√ x−2
7. f (x) = 4 − x2, 0≤x≤2 x
14. f (x) =
8. f (x) = 2x + |x| 3 + x2
In Exercises 15–20 show that the function does not have an inverse. function x = f −1 (r) . In this function x depends on r , indicating that
Subdivide its domain of definition into subintervals on which the func- if the price of the object is set at r , then the market will demand x of
tion has an inverse, and find the inverse function on each subinterval. them per week. This function is therefore called the demand function
a
(right figure). Find demand functions if (a) r = + c, a , b,
∗ 15. f (x) = x 4 ∗ 16. f (x) = 1/x 4 x+b
a
∗ 17. f (x) = x 2 + 2x + 3 ∗ 18. f (x) = x 4 + 4x 2 + 2 and c positive constants; (b) r = + c , a , b , and c positive
x2 + b
x2 x4 constants. Draw the demand function x = f −1 (r) and given function
∗ 19. f (x) = ∗ 20. f (x) = r = f (x) in each case.
x2 + 4 x2 + 4
∗ 21. Give an example of a function that is defined for all x , but does
not have an inverse on any interval whatsoever. ∗ 24. Show that the demand function
∗ 22. Give an example of a function that is defined for all x , has an
inverse on the interval 0 ≤ x ≤ 1, but does not have an inverse on any
other interval. x = f (r) = 4a 3 − 3ar 2 + r 3 , 0 < r < 2a
∗ 23. If a manufacturing firm sells x objects of a certain commodity per
week, it sells them at a price of r per object, and r depends on x ,
r = f (x) . In economic theory this function is usually considered
decreasing, as shown in the left figure below; hence it has an inverse has an inverse function r = f −1 (x) . What is the domain of f −1 (x) ?
Draw its graph.
r x
∗ 25. Find the inverse function for
x2
f (x) = , −1 < x ≤ 0 .
x r (1 + x)2
1.7 Trigonometry Review

Here is the diagnostic test for this section; give yourself 60 minutes to do it. Use a calculator
only in problems that specify their use.
In all questions, angles are in radian measure unless indicated otherwise. In questions 5–8 draw a graph of the function.
1. Express the angles 135◦ and −270◦ in radian measure. 5. f (x) = 3 sin 2x
2. Express the angles 2π/3 and −9π/4 in degree measure. 6. f (x) = 2 cos (x − π/3)
3. Evaluate the following quantities:
7. f (x) = tan (x + π/2)
(a) sin (π/3)
8. f (x) = 4 cos (2x + π/2)
(b) cot (−π/4)
(c) cos (5π/3)
(d) csc (−3π/4) . In questions 9–11 find all solutions to the equation.
4. Find the unspecified angles and the length of the third side of the √
triangle below. You will need a calculator for this question. 9. sin x = 3/2
10. cos x = −1/2
2.4 11. tan 2x = 1
12. Express the function f (x) = 2 sin 2x − 3 cos 2x in the form
0.63 A sin (2x + φ) where A is positive and φ is an angle in the interval
3.5 0 < φ < 2π . You may use a calculator.
Many physical systems exhibit an oscillatory nature: vibrations of a plucked guitar string,
motion of a pendulum, alternating electric currents, fluctuations in room temperature as the
thermostat continually engages and disengages the furnace on a cold day, and the rise and
fall of tides and waves. Magnitudes of these oscillations are best described by the sine and
cosine functions of trigonometry. In addition, rates at which these oscillations occur can be
represented by derivatives of these functions from calculus. In this section, we briefly review
the trigonometric functions and their properties, placing special emphasis on those aspects that
FIGURE 1.77 Definition are most useful in calculus.
of a radian In trigonometry, angles are usually measured in degrees, radians, or mils; in calculus, angles
are always measured in radians. In preparation, then, for the calculus of trigonometric functions,
r
we work completely in radian measure. By definition, a radian is that angle subtended at the
centre of a circle of radius r by an arc of equal length r (Figure 1.77). When an arc has length s
1 radian (Figure 1.78), the number of units of radius in this arc is s/r ; this is called the radian measure
r of the angle θ subtended at the centre of the circle
s
θ = . (1.37)
r
In particular, if s contains π units of radius ( s = π r ), then s represents one-half the circum-

ference of the complete circle, and θ = π r/r = π radians. In degree measure, this angle is
180◦ ; hence we can state that π radians is equivalent to 180◦ . This statement enables us to
FIGURE 1.78 Relationship
convert angles expressed in degrees to radian measure, and vice versa. If the degree measure of
between arc length and angle at centre
of circle
an angle is φ , then it is π φ/180 radians; conversely, if an angle measures θ radians, then its
degree measure is 180θ/π . For example, the radian measure for φ = 45◦ is 45π/180 = π/4
s radians.
Elementary trigonometry is concerned with relationships among angles and lengths and, in
particular, angles and sides of triangles. This naturally restricts angles to the range 0 ≤ θ ≤ π .
For the purposes of calculus, however, we need to talk about angle θ , where θ is any real number
r — positive, negative, or zero. To do this we first define what we mean by the standard position
of an angle. If θ > 0, we draw a line segment OP through the origin O of the xy -plane
(Figure 1.79) in such a way that the positive x -axis must rotate counterclockwise through an
angle θ to coincide with OP . In other words, we now regard an angle as rotation, rotation
of the positive x -axis to some terminal position. If 0 < θ < π/2, then OP lies in the first
quadrant; and if π/2 < θ < π , then OP is in the second quadrant. But if θ > π , we have a
geometric representation of θ also. For example, angles 7π/4 and 9π/4 are shown in Figure
FIGURE 1.79 Standard
position of an angle
1.80. When θ < 0, we regard θ as a clockwise rotation (Figure 1.81).
y
P FIGURE 1.80 Positive an- FIGURE 1.81 Negative
gles as counterclockwise rotations angle as clockwise rotations
y y
P
O x
4
O
x x
4 −
4
P P
For any angle θ , the angles θ + 2nπ , where n is an integer, have the same terminal position
of OP as θ . They are different angles, however, because the positive x -axis must encircle the
origin one or more times before reaching the terminal position in the case of θ + 2nπ .
With angles represented as rotations, it is easy to define the six trigonometric functions. If
FIGURE 1.82 Definitions
of trigonometric functions
θ is an angle in standard position (Figure 1.82) and (x, y) are the coordinates of P , we define
y y x y
sin θ = , cos θ = , tan θ = ,
P (x, y) r r x
r r x (1.38)
csc θ = , sec θ = , cot θ = ,
r y x y
y
!
O wherever these ratios are defined, and where r = x 2 + y 2 is always assumed positive. Since
x x x is positive in the first and fourth quadrants, and y is positive in the first and second, signs of
the trigonometric functions in the various quadrants are as shown in Figure 1.83. Furthermore,
since r $= 0, sin θ and cos θ are defined for all θ , whereas tan θ and sec θ are not defined for
x = 0, and csc θ and cot θ do not exist when y = 0.
FIGURE 1.83 Signs of Definitions 1.38 indicate that
trigonometric functions in four
quadrants 1 1 1
csc θ = , sec θ = , cot θ = . (1.39)
y sin θ cos θ tan θ
Sine and cosecant All trigonometric In addition, the fact that r 2 = x 2 + y 2 leads to the identity
positive; all functions positive
others negative
sin2 θ + cos2 θ = 1, (1.40a)
Tangent and Cosine and secant x
cotangent positive; positive; all which, in turn, implies that
all others negative others negative
1 + tan2 θ = sec2 θ, (1.40b)
1 + cot 2 θ = csc2 θ. (1.40c)
Table 1.2 contains values of the trigonometric functions for the most commonly used angles.
You should commit at least values for sine, cosine, and tangent to memory.
TABLE 1.2
sin x cos x tan x csc x sec x cot x

0 0 1 0 undefined 1 undefined
√ √ √ √
π/6 1/2
√ 3/2
√ 1/ 3
√2 2/ 3
√ 3
π/4 1/ 2 1/ 2
√ √1 √2 2 1
√
π/3 3/2 1/2 3 2/ 3 2 1/ 3
π/2 1 0 undefined 1 undefined 0
FIGURE 1.84 Cosine law
Identity 1.40a is simply a restatement of the Pythagorean relation ( r 2 = x 2 + y 2 ), which allows
y
P us to express the hypotenuse of a right-angled triangle in terms of the other two sides. If the
triangle is not right-angled (Figure 1.84), it is not possible to express one side, c , in terms of the
b c other two sides, a and b alone, but it is possible to express c in terms of a and b and the angle
θ between them. The coordinates of P and Q in Figure 1.84 are (b cos θ, b sin θ ) and (a, 0) ;
therefore, by distance formula 1.10,
a Q x
c2 = &P Q&2 = (b cos θ − a)2 + (b sin θ )2
= b2 (cos2 θ + sin2 θ ) + a 2 − 2ab cos θ.
Since sin2 θ + cos2 θ = 1, we have
c2 = a 2 + b2 − 2ab cos θ. (1.41)
This result, called the cosine law, generalizes the Pythagorean relation to triangles that are not
right-angled. It reduces to the Pythagorean relation c2 = a 2 + b2 when θ = π/2.
FIGURE 1.85 Sine law

If A , B , and C are the angles in the triangle of Figure 1.85, and a , b , and c are the lengths
of the opposite sides, then by drawing altitudes of the triangle we can show that
c A
b sin A sin B sin C
= = . (1.42)
B C a b c
a
This result, known as the sine law, is also useful in many problems. The sine and cosine laws can
be used to find angles and lengths of sides of triangles as illustrated in the following examples.
EXAMPLE 1.31
Find length a and angles θ and φ in the triangle of Figure 1.86.
FIGURE 1.86 Identifying SOLUTION We can calculate a with the cosine law,
angles and sides of a triangle
a 2 = 22 + 32 − 2(2)(3) cos(1.2).
1.2
2 This gives a = 2.9. The sine law can now be used to find θ ,
3
θ sin θ sin 1.2 3 sin 1.2

= *⇒ sin θ = .
φ 3 2 .9 2 .9
a
The inverse sine button on a calculator gives θ = 1.3. (We will deal with inverse trigonometric
functions in detail in Section 1.8.) Finally, φ = π − 1.2 − 1.30 = 0.64.
EXAMPLE 1.32
Distances across water are often much greater than they appear from land. We would like to
calculate the distance from the straight shoreline to the island in Figure 1.87a. What we could
do is take two points B and C on the shore some distance apart, 1 km say, and measure the
angles ABC and ACB as shown in Figure 1.87b. Use this information to calculate how far
the island is from shore.

Distance from shore to an island
Island A Island A
1.1 0.8 1.1 0.8

Shoreline B 1 Shoreline C B D C
1
SOLUTION We require the length of AD in Figure 1.87c. Angle BAC = π − 0.8 − 1.1 =
1.242. We use the sine law to find b ,
sin 1.1 sin 1.242 sin 1.1

= *⇒ b = = 0.942.
b 1 sin 1.242
We can now use triangle ADC to calculate that the length of AD is b sin 0.8 = (0.942)(0.8) =
0.75 km.
A large number of identities are satisfied by the trigonometric functions. They can all be derived
FIGURE 1.88 Proof of
compound angle formula from the following compound-angle formulas for sines and cosines:
y sin (A + B) = sin A cos B + cos A sin B, (1.43a)
P sin (A − B) = sin A cos B − cos A sin B, (1.43b)

Q
A cos (A + B) = cos A cos B − sin A sin B, (1.43c)
B
cos (A − B) = cos A cos B + sin A sin B. (1.43d)
O x
Let us prove one of these, say identity 1.43d, where A and B are the angles in Figure 1.88.
If P and Q lie on the circle x 2 + y 2 = r 2 , then their coordinates are (r cos A, r sin A)
and (r cos B, r sin B) , respectively. According to formula 1.10, the length of P Q is given
by
&P Q&2 = (r cos A − r cos B)2 + (r sin A − r sin B)2

= r 2 (cos2 A + sin2 A) + r 2 (cos2 B + sin2 B)
− 2r 2 (cos A cos B + sin A sin B)
= 2r 2 − 2r 2 (cos A cos B + sin A sin B).
But, according to cosine law 1.41,
&P Q&2 = &OP &2 + &OQ&2 − 2&OP &&OQ& cos (A − B)

= r 2 + r 2 − 2r 2 cos (A − B)
= 2r 2 − 2r 2 cos (A − B).
Comparison of these two expressions for &P Q&2 immediately implies identity 1.43d.
By expressing tan (A + B) as sin (A + B)/ cos (A + B) and using identities 1.43a and
1.43c, we find a compound-angle formula for the tangent function,
tan A + tan B
tan (A + B) = , (1.44a)
1 − tan A tan B
and similarly,
tan A − tan B
tan (A − B) = . (1.44b)
1 + tan A tan B
By setting A = B in 1.43a,c and 1.44a, we obtain the double-angle formulas,
sin 2A = 2 sin A cos A, (1.45)
cos 2A = cos2 A − sin2 A, (1.46a)
= 2 cos2 A − 1, (1.46b)
= 1 − 2 sin2 A, (1.46c)
2 tan A
tan 2A = . (1.47)
1 − tan2 A
EXAMPLE 1.33
Use a trigonometric identity to find the cosine of the angle θ/2 if it is known that the cosine of
θ is 0.3.
SOLUTION If we set A = θ/2 in double-angle formula 1.46b, and solve for cos2 (θ/2) , we
obtain
1 + cos θ 1 + 0 .3 1.3
cos θ = 2 cos2 (θ/2) − 1 *⇒ cos2 (θ/2) = = = .
2 2 2
Hence,
$
1.3
cos (θ/2) = ± = ±0.81.
2
Without further information about θ or θ/2, we cannot decide which sign to choose. For
instance, θ = 1.266 is an angle whose cosine is 0.3. The cosine of half this angle is 0.81. On
the other hand, θ = 5.0 is also an angle whose cosine is 0.3, but the cosine of half this angle
is −0.81.
When pairs of compound-angle formulas 1.43 are added or subtracted, the product
formulas result. For example, subtracting 1.43c from 1.43d gives
1
sin A sin B = [− cos (A + B) + cos (A − B)]. (1.48a)
2
The other product formulas are
1
sin A cos B = [sin (A + B) + sin (A − B)], (1.48b)
2
1
cos A cos B = [cos (A + B) + cos (A − B)]. (1.48c)
2
By setting X = A + B and Y = A − B in 1.48, we obtain the sum and difference

formulas,
" # " #
X+Y X−Y
sin X + sin Y = 2 sin cos , (1.49a)
2 2
" # " #
X+Y X−Y
sin X − sin Y = 2 cos sin , (1.49b)
2 2
" # " #
X+Y X−Y
cos X + cos Y = 2 cos cos , (1.49c)
2 2
" # " #
X+Y X−Y
cos X − cos Y = −2 sin sin . (1.49d)
2 2
The following example is typical of problems in calculus when trigonometric functions are
involved.
EXAMPLE 1.34
Write the expression cos4 θ in terms of cos 2θ and cos 4θ .
SOLUTION If we replace A by θ in double-angle formula 1.46b,
cos 2θ = 2 cos2 θ − 1.
It follows that
1 + cos 2θ
cos2 θ = .
2
Consequently,
1 1
cos4 θ = (cos2 θ )2 = (1 + cos 2θ )2 = (1 + 2 cos 2θ + cos2 2θ ).
4 4
But if we now replace θ by 2θ in the identity cos2 θ = (1 + cos 2θ )/2, we obtain
1 + cos 4θ
cos2 2θ = .
2
Thus,
" #
4 1 1 + cos 4θ 1
cos θ = 1 + 2 cos 2θ + = (3 + 4 cos 2θ + cos 4θ ).
4 2 8
By regarding arguments of the trigonometric functions as angles, we have been stressing ge-
ometric properties of these functions. In particular, identities 1.43–1.49 have been based on
definitions of the trigonometric functions as functions of angles. What is important about a
function — be it a trigonometric function or any other kind of function — is that there is a
number associated with each value of the independent variable. How we arrive at this number
is irrelevant. As far as properties of the function are concerned, only its values are taken into
account. Thus, when we discuss properties of a trigonometric function, what is important is not
that its argument can be regarded as an angle or that its values can be defined as ratios of sides
of a triangle, but that we know its values. Indeed, it is sometimes unwise to regard arguments
of trigonometric functions as angles. Consider, for example, the motion of a mass m suspended
from a spring with spring constant k (Figure 1.89). If x = 0 is the position at which the mass
FIGURE 1.89 Displace- would hang motionless, then when vertical oscillations are initiated, the position of m as a
ment of mass in vibrating system function of time t is always of the form
0$ 1 0$ 1
k k
x = A cos t + B sin t ,
m m
k where A and B are constants. Clearly, there are no angles associated √ with the motion of
x m , and it is therefore unnatural to attempt to interpret the argument ( k/m)t as a physical
angle.
m x=0 Henceforth, we consider the trigonometric functions as those of a real variable. If it is
convenient to regard the argument as an angle, then we do so, but only if it is convenient. To
emphasize this, we replace θ with our usual generic label for the independent variable of a
function, namely, x . With this change, the trigonometric functions are sin x , cos x , tan x ,
csc x , sec x , and cot x . Their graphs are shown in Figures 1.90.
These graphs illustrate that sin x , csc x , tan x , and cot x are odd functions, and cos x and
sec x are even. Trigonometric functions are periodic. A function f (x) is said to be periodic if
there exists a number T such that for all x in its domain of definition,
f (x + T ) = f (x). (1.50)
The smallest such positive number T is called the period of f (x) . Clearly, then, sin x , cos x ,
csc x , and sec x are periodic with period 2π , whereas tan x and cot x have period π .
FIGURE 1.90a Sine function FIGURE 1.90b Cosine function FIGURE 1.90c Tangent function
y y y
y = tan x
y = sin x 1 y = cos x
1 −π
x x x
− − − −
2 2 2 2
FIGURE 1.90d Cotangent func- FIGURE 1.90e Secant function FIGURE 1.90f Cosecant function
tion
y y
y y = sec x y = csc x
y = cot x
1 − 1
− x x
− 2 −1 2
x 2 2 2
− 2
2 −1
When two periodic functions are added together, the resulting

√ function may, or may not,
be periodic. For example, the function
√ f (x) = 3 sin 2 x + 3 3 cos 2x is the addition of two
π -periodic functions 3 sin 2x and 3 3 cos 2x . It also has period π . Function f (x) = 2 sin 3x
has period 2π/3; function g(x) = −3 cos 2x has period π . Their sum 2 sin 3x − 3 cos 2x
in Figure 1.91 has period 2π ; it is the smallest √ interval in which both functions reproduce
themselves. The√ function f (x) = sin x + 2 sin 2x in Figure 1.92 is not periodic even though
sin x and 2 sin 2x are both periodic.
FIGURE 1.91 Graph of 2 sin 3x − 3 cos 2x FIGURE 1.92 Graph of a nonperiodic function
y y
3
4
2
25 x
−10 −5 5 10 x
−2 −3
−4
Seldom is the argument of a trigonometric function just x in applications; it is usually a

function of x .
EXAMPLE 1.35
Draw the graph of the function f (x) = cos (x − π/3) .
SOLUTION The graph of this function is that of cos x translated π/3 units to the right
(Figure 1.93).
FIGURE 1.93 Translated graph of cosine function
1 y = cos(x − /3)
x
3 3
−1
EXAMPLE 1.36
When a mass vibrates on the end of a spring as in Figure 1.94, and there is no friction with the
surface, or air resistance, the position of the mass (in centimetres) relative to its position when
the spring is unstretched takes the form
√
x(t) = 3 sin 2t + 3 3 cos 2t.
Numbers would be different, but x(t) would be a combination of a sine function and a cosine
function with the same argument. We have chosen simple numbers here so that unnecessarily
complicated calculations do not obscure the significance of the discussion. Plot a graph of x(t)
for t ≥ 0, and then draw a graph. What advantages are derived from the drawing as opposed to
the plot?
FIGURE 1.94 Vibrating mass-spring FIGURE 1.95 Computer plot for displacement
system of mass in vibrating system
x
6
Spring
4
Mass 2
2 4 6 8 10 t
x −2
x=0
−4
−6
SOLUTION A plot of the function is shown in Figure 1.95.

To draw a graph of x(t) , we first write it in the form x(t) = A sin (2t + φ) , where A and
φ are constants. When we equate this to the given expression for x(t) , and expand sin (2t + φ)
with compound-angle formula 1.43a, we obtain
√
3 sin 2t + 3 3 cos 2t = A sin (2t + φ) = A[sin 2t cos φ + cos 2t sin φ ].
This equation will be true for all t if we choose A and φ to satisfy

√
3 3 = A sin φ, 3 = A cos φ.
To solve these for A and φ , we square each equation and add the results,
27 + 9 = A2 sin2 φ + A2 cos2 φ = A2 .
This implies that A = ±6. If we choose A = 6 ( A = −6 works equally well), then

√
3 3 = 6 sin φ, 3 = 6 cos φ.
These equations are satisfied by φ = π/3 (there are other angles also), and therefore x(t) can
be expressed in the form
x(t) = 6 sin (2t + π/3) = 6 sin [2(t + π/6)].

The function is most easily graphed by shifting the graph of f (t) = sin 2t in Figure 1.96a to
the left by π/6 units and changing the scale on the x -axis. The result is shown in Figure 1.96b.
With x(t) expressed in the form 6 sin (2t + π/3) , it is clear that the function has period π
and that oscillations take place between x = ±6. These are facts that we could surmise from
FIGURE 1.96a Sine function with period π FIGURE 1.96b Hand-drawn graph of displacement
of mass in vibrating system
x x = sin 2t x
1 6
t t
2 2 2 3 6 6
−1 −6
the plot in Figure 1.95, but evidence would not be conclusive. Were we to need values of t
at which the function has value 3, say, it is definitely advantageous to have x(t) in the form
6 sin (2t + π/3) . We can visualize possibilities as t -coordinates of points of intersection of the
curve in Figure 1.96b with the horizontal line x = 3. There is an infinity of values, and to find
them algebraically, we must solve 3 = 6 sin (2t + π/3) , or sin (2t + π/3) = 1/2. We can
think of 2t + π/3 as an angle whose sine is 1/2, one possibility being π/6. But there are many
other angles with a sine equal to 1/2. Because the sine function is 2π -periodic, each of the
angles π/6 + 2nπ , where n is an integer, also has sine equal to 1/2. Furthermore, the sine of
5π/6 is also equal to 1/2, and when multiplies of 2π are added, each of the angles 5π/6 + 2nπ
has sine equal to 1/2. In other words, all angles that have a sine equal to 1/2 are
π 5π
+ 2nπ and + 2nπ,
6 6
where n is an integer. In other words, we can set
π

 + 2nπ
π 6
2t + =
3  5π + 2nπ.

6
Consequently,
 π 
  π
 − + 2nπ  − + nπ
2t = 6 *⇒ t = 12
 π  π
 + 2nπ  + nπ.
2 4
Because we are only interested in positive values of t , we must choose n ≥ 1 when combined
with −π/12, and n ≥ 0 when combined with π/4. The smallest positive value of t is π/4;
the second smallest is 11π/12.
You begin to appreciate the elegance of this solution and the simplicity of the result when
you compare the magnitude of the problem
√ were you to attempt to find, using a calculator or
computer, all solutions of 3 sin 2t + 3 3 cos 2t = 3. Think about it.
The curve in Figure 1.96b is an example of a general sine function,
f (x) = A sin (ωx + φ), (1.51)
where A and ω are positive constants and φ is also constant. The graph of the general sine
function is shown in Figure 1.97. The number A , which represents half the range of the function,
is called the amplitude of the oscillations. The period is 2π/ω , and −φ/ω is called the phase
shift.
FIGURE 1.97 General sine function
y
A
−A
Example 1.36 is an illustration of the following very useful result. A function of the form
f (x) = B sin ωx +C cos ωx can always be expressed in form 1.51 for a general sine function.
The amplitude is given by the formula
!
A= B 2 + C2. (1.52)
Function B sin ωx + C cos ωx can also be expressed in the form A cos (ωx + ψ) , and A is
once again given by the formula in equation 1.52.
EXAMPLE 1.37
FIGURE 1.98 Schematic The emf device in the LC -circuit of Figure 1.98 produces a constant voltage of V volts. If
for LC -circuit the switch is closed at time t = 0, and there is no initial charge on the capacitor, the charge
C thereafter is given by
" #
V t
Q(t) = 1 − cos √ , t ≥ 0.
C LC
V L
Draw a graph of this function.
√
S SOLUTION We √ begin by drawing a graph of cos (t/ LC) . It is a standard cosine function
with period 2π LC (Figure 1.99a). The graph of Q(t) in Figure 1.99b is then obtained by
turning Figure 1.99a upside down, shifting it upward 1 unit, and changing the scale on the
Q -axis.
FIGURE 1.99a Cosine function needed for charge on capacitor FIGURE 1.99b Charge on capacitor
1 Q
2V/C
−1
t
It is worthwhile pointing out here, as we are sure you realize, that electronic devices cannot
plot curves containing unspecified parameters. This is another reason why we must develop our
graphing skills and not rely totally on graphing calculators and computers.
EXAMPLE 1.38
Find all solutions for each of the following equations:
1
(a) sin x = √
2
√
3
(b) cos 2x = −
2 √
(c) tan (3x + 1) = − 3
SOLUTION
√
(a) One solution of the equation sin x = 1/ 2 is π/4. This√is not the only solution,
however; there are many angles with a sine equal to 1/ 2. Since sin x is √ 2π -
periodic, the angles 2nπ + π/4, for n any integer, all have sine √ equal to 1 / 2.
Because sin (π − x) = sin x , it follows that sin (3π/4) = 1/ 2, and therefore
3π/4 is another solution. When multiples of 2π are added to this angle, 2nπ + 3π/4
are also solutions. Thus, the complete set of solutions is
π 3π
2nπ + , 2nπ + ,
FIGURE 1.100 Simplified
4 4
representation of solutions to a where n is an integer. Figure 1.100 suggests that this set of numbers can be repre-
trigonometric equation sented more compactly as an initial rotation of π/2, plus or minus π/4, and possible
y multiples of 2π ; that is,
" #
π π 4n + 1 π
3π/4 x = ± + 2nπ = π± .
2 4 2 4
√
π/4 (b) One solution of the equation cos 2x = − 3/2 for 2x is 5π/6. But there are others.
Since the cosine function is even, all solutions are given by
x
5π
2x = ± + 2nπ (where n is an integer).
6
Consequently,
5π
x =± + nπ.
12
√
(c) One solution of the equation tan (3x + 1) = − 3 for 3x + 1 is 3x + 1 = 2π/3.
Since the tangent function is π -periodic, all solutions can be expressed in the form
2π (3n + 2)π
3x + 1 = + nπ = , (where n is an integer).
3 3
Consequently,
(3n + 2)π 1
x = − .
9 3
EXERCISES 1.7
In Exercises 1–10 express the angle in radians. the smaller building to the bottom of the taller building is 3/5 radians.
How tall are the buildings?
1. 30◦ 2. 60◦ 24. If the angle of elevation of the sun is 0.80 radian, and the length
◦ ◦ of the shadow of a flagpole is 20 metres, how high is the pole?
3. 135 4. −90
5. −300◦ 6. 765◦ For each of the triangles in Exercises 25–28 use the cosine law and/or
◦ ◦ the sine law to find the lengths of all three sides and the measures of all
7. 72 8. −128
the interior angles.
9. 321◦ 10. −213◦ 25. 26.
2
5
3
In Exercises 11–20 express the angle in degrees.
11. π/3 12. −5π/4 3

4
13. 3π/2 14. 8π 27. 28.
15. −5π/6 16. 1 5 6

4
17. −3 18. 2.5
19. −3.6 20. 11 7
21. What angle is subtended at the centre of a circle of radius 4 by an 29. Use compound-angle formulas 1.43 to prove identities 1.44.
arc of length (a) 2, (b) 7, and (c) 3.2?
30. Verify double-angle formulas 1.45, 1.46, and 1.47.
22. The angle of elevation from a transit to the top of a building is 1.30 31. Use compound-angle formulas 1.43 to prove product formulas
radians (left figure below). If the transit is 2 metres above the ground, 1.48.
and the distance from the building to the transit is 30 metres, how high 32. Show that the sum and difference formulas 1.49 can be obtained
is the building? from 1.48.
In Exercises 33–54 draw a graph of the function.
33. f (x) = 3 sin x 34. f (x) = sin 2x

35. f (x) = 3 sin 2x 36. f (x) = sin (x + π/4)
1.30 11/10
3/5 37. f (x) = 3 sin (x + π/4) 38. f (x) = sin (2x + π/4)
2m 39. f (x) = 3 sin (2x + π/4) 40. f (x) = 4 cos (x/3)
30 m 100 m
41. f (x) = 2 sin (x/2 − π ) 42. f (x) = 5 cos (π/2 − 3x)
43. f (x) = sec 2x 44. f (x) = tan 3x
23. Two buildings are 100 metres apart (right figure above). The angle
of elevation from the top of the smaller building to the top of the taller 45. f (x) = csc (x − π/3) 46. f (x) = cot (x + π/4)
2
√
building is 11/10 radians. The angle of depression from the top of ∗ 47. f (x) = tan x ∗ 48. f (x) = 1 − cos2 x
√
∗ 49. f (x) = 1 + tan2 x ∗ 50. f (x) = 5 − 2 sec x ∗ 71. Amplitude modulation is the process of multiplying a low-
frequency signal by a high-frequency sinusoid (as occurred in Exercise
∗ 51. f (x) = 4 + 2 tan x ∗ 52. f (x) = tan |x|
70). It is the technique used to broadcast AM radio signals. The AM
∗ 53. f (x) = −| cot 2x| ∗ 54. f (x) = 3 csc (x/2) signal is a product of the form x(t) = v(t) cos (2πf t) , where the
frequency f is much higher than any frequency in v(t) . The cosine
∗ 55. When a javelin is released from height h above the ground with term is the carrier signal and v(t) is the voice or music signal to be
speed v at angle θ with the horizontal (figure below), the horizontal transmitted.
distance R that it travels is given by the formula (a) Plot a graph of x(t) when f = 700 Hz and v(t) = 5 +
2 cos (40π t) .
0 $ 1
v 2 cos θ 2gh (b) What are minimum and maximum amplitudes of the mod-
2
R = sin θ + sin θ + 2 , ulated signal?
g v
In Exercises 72–77 express each function as a general sine function,
where g = 9.81 is the acceleration due to gravity. identifying its amplitude, period, and phase shift. Draw a graph of each
function.
y ∗ 72. f (x) = 3 sin 3x + 3 cos 3x
∗ 73. f (x) = 2 sin 4x − 2 cos 4x
√
∗ 74. f (x) = −2 sin x + 2 3 cos x
h √
x ∗ 75. f (x) = −2 sin 5x − 2 3 cos 5x
R
∗ 76. f (x) = sin x cos x ∗ 77. f (x) = sin2 2x − cos2 2x
(a) What is R for thrower A , who releases the javelin with
speed v = 20 m/s at angle θ = π/4 and height 2 m?
(b) What is R for thrower B , who, being slightly taller, releases In Exercises 78–82 verify the identity.
the javelin with the same speed and angle, but at height
∗ 78. cos 3x = 4 cos3 x − 3 cos x
2.1 m?
∗ 79. sin 4x = 8 cos3 x sin x − 4 cos x sin x
(c) With what speed must thrower A release the javelin if he
is to achieve the same R as thrower B , assuming that θ = 3 tan x − tan3 x
∗ 80. tan 3x =
π/4 and h = 2 m? 1 − 3 tan2 x
2x 3 sin x
∗ 81. tan =
In Exercises 56–69 find all solutions of the equation. All solutions 2 1 + cos x
involve the standard angles in Table 1.2. For equations with solutions 1 + tan x 2 π3
∗ 82. = tan x +
involving other angles, we require the inverse trigonometric functions 1 − tan x 4
of Section 1.8.
√
56. sin x = 3/2 57. cos x = 2 ∗ 83. Show that a function f (x) = A cos ωx + B sin ωx can always
√ be written in the form
58. cos x = −1/2 59. cot x = 3 !
f (x) = A2 + B 2 sin (ωx + φ),
2
60. sin x = cos x 61. 2 cos x = 1 where φ is defined by the equations
√ √
62. cos 2x = −1/ 2 63. tan 3x = − 3 A B
√ sin φ = √ and cos φ = √ .
64. 2 sin 3x − 2 = −1 65. sec 4x = − 2 A2 + B2 A2 + B2
∗ 66. sin 2x = sin x ∗ 67. sin2 x − sin x − 2 = 0 ∗ 84. In Exercise 83 can we replace the two equations defining φ with
2
∗ 68. 3 cot x − 1 = 0 ∗ 69. sin x + cos x = 1 the single equation tan φ = A/B ?
∗ 70. When two musical instruments play notes with nearly identical
frequencies, beat notes result. Suppose, for example, that the signals In Exercises 85–88 find all solutions of the equation in the interval
produced by the instruments are at 180 Hz and 220 Hz, say cos (360π t) 0 ≤ x < 2.
and cos (440πt) .
∗ 85. sin 4x = cos 2x
(a) Show that the combined signal can be expressed in the form
∗ 86. cos x + cos 3x = 0
x(t) = 2 cos (40π t) cos (400π t) .
∗ 87. sin 2x + cos 3x = sin 4x
(b) Plot, on the same axes, graphs of ±2 cos (40π t) and x(t) √
for 0 ≤ t ≤ 0.8. The amplitude of cos (400π t) is ∗ 88. sin x + cos x = 3 sin x cos x
modulated by 2 cos (40π t) . You would hear the signal
cos (400π t) fade in and out as its amplitude rises and falls. ∗ 89. Verify that if A , B , and C are the angles of a triangle, then
This phenomenon is called beating of tones in music. Mu-
tan A + tan B + tan C = tan A tan B tan C.
sicians use it to tune two instruments to the same pitch. (See
also Exercise 71.) Hint: Expand tan (A + B + C) in terms of tan A , tan B , and tan C .

1. 3π/4, −3π/2 (2 marks) 7. (2 marks)
◦ ◦
y
2. 120 , −405 (2 marks)
√ √
3. (a) 3/2 (b) −1 (c) 1/2 (d) − 2 (4 marks)
4. 0.74 radians, 1.8 radians, 2.1 (6 marks)
5. (2 marks)
−π −_
π _
π π x
y 2 2
3
8. (3 marks)
y
−π −_
π _
π π x 4
2 2
−3
−π − _
π _
π π x
6. (3 marks) 2 2
y
2 −4
1
9. π/3 + 2nπ , 2π/3 + 2nπ , n an integer (3 marks)
−2π −π − _
π _
π π 2π x 10. 2π/3 + 2nπ , −2π/3 + 2nπ , n an integer (3 marks)
3 3
11. π/8 + nπ , n an integer (3 marks)
−2 √
12. 13 sin (2x + 5.3) (3 marks)
1.8 The Inverse Trigonometric Functions

This material is likely new for many students. It does not therefore have a diagnostic test
associated with it.
Functions that involve a finite number of additions, subtractions, multiplications, divisions,
and roots are called algebraic functions. More specifically, a function y = f (x) is said to be
algebraic if, for all x in its domain, it satisfies an equation of the form
P0 (x)y n + P1 (x)y n−1 + · · · + Pn−1 (x)y + Pn (x) = 0, (1.53)
where P0 (x) , . . . , Pn (x) are polynomials in x , and n is a positive integer. A polynomial P (x)
is itself algebraic since it satisfies y − P (x) = 0; that is, it satisfies 1.53 with n = 1,
P0 (x) = 1, and P1 (x) = −P (x) . Rational functions P (x)/Q(x) are also algebraic [ n = 1,
P0 (x) = Q(x) , and P1 (x) = −P (x) ]. The function f (x) = x 1/3 is algebraic since it
satisfies y 3 − x = 0. The equation y 3 + y = x defines y as a function of x (a graph of the
curve would illustrate that it satisfies the vertical line test). We cannot find the function in the
form y = f (x) , but according to equation 1.53, the function so defined is algebraic.
A function that is not algebraic is called a transcendental function. The trigonomet-
ric functions and the exponential and logarithm functions (to be reviewed in Section 1.9) are
transcendental. In this section, we consider the inverse trigonometric functions; they are also
transcendental.
In Section 1.6 we discussed inverse functions, what it means for one function to be the
inverse of another. We learned that, algebraically, a function has an inverse only if it is one-
to-one, or geometrically, if its graph passes the horizontal line test. Since the trigonometric
functions do not satisfy these conditions, they do not have inverses. But we also learned that it
is usually possible to restrict the domain of a function that is not one-to-one in such a way that
an inverse function can be defined. We do this for the trigonometric functions in this section.
The function f (x) = sin x , defined for all real x , does not have an inverse; it is not one-
to-one; its graph fails the horizontal line test. By restricting the domain of sin x , however, the
function can be made one-to-one, and this can be done in many ways. In particular, that part of
sin x on the interval −π/2 ≤ x ≤ π/2 is one-to-one, and therefore has an inverse function.
This function, denoted by
y = Sin −1 x, (1.54)
and called the inverse sine function, is shown in Figure 1.101. The range of the function is
π π
− ≤ Sin −1 x ≤ . (1.55)
2 2
The values that fall in this range are called principal values of the inverse sine function. They
have resulted from our restriction of the domain of sin x to this same interval.
An equivalent way to derive the inverse sine function is as follows. The reflection of the
graph of y = sin x in the line y = x does not represent a function, but by restricting the
range of values of the reflected curve, we can produce a single-valued function (Figure 1.102).
Once again this can be done in many ways, and when we do so by restricting the y -values to
the interval [−π/2, π/2], the resulting function is called the inverse sine function Sin −1 x .
Note very carefully that Sin −1 x is not the inverse function of f (x) = sin x , because the sine
function has no inverse. It is, however, the inverse of the sine function restricted to the domain
−π/2 ≤ x ≤ π/2. It is perhaps then a misnomer to call the function Sin −1 x the inverse sine
function, but this has become the accepted terminology.
We now know what the inverse sine function looks like graphically, but what does it mean
to say that y = Sin −1 x ? Certainly, given any value of x , we can push a few buttons on an
electronic calculator and find Sin −1 x for that x . To use inverse trigonometric functions in
practice, we must have a feeling for what they do. To obtain this insight, we note that if (x, y)
is a point on the curve y = Sin −1 x , then (y, x) is a point on the sine curve; that is,
y = Sin −1 x only if x = sin y. (1.56)
FIGURE 1.101 Inverse of sin x found by restricting the FIGURE 1.102 Inverse of sin x found by reflecting its
domain of the function graph in line y = x and restricting the range of the reflection
y y
y=x
y= Sin−1 x
1 y = sin x
y = Sin−1 x
2
−1
−1 1 x
1 x
−1 y = sin x
Reflection of
y = sin x
In the latter equation y is an angle and x is the sine of that angle. Thus, when we see y =
Sin −1 x , we may read “ y equals inverse sine x ,” but we should think “ y is an angle whose sine
is x .” For instance, if x = 1/2, then y = Sin −1 (1/2) means “ y is an angle whose sine is
1/2.” Clearly, the angle whose sine is 1/2 is π/6, and we write y = Sin −1 (1/2) = π/6.
Remember that we must choose the principal value π/6; there are many angles that have a
sine equal to 1/2, but the inverse sine function demands that we choose that angle in the range
−π/2 ≤ y ≤ π/2, and state it with a number in this interval. For example, 2π + π/6 may
represent the same angle with the positive x -axis as π/6, but the number 2π + π/6 does not
lie between −π/2 and π/2. In summary, the function sin x regards x as an angle and assigns
to x the sine of the angle; the inverse sine function Sin −1 x regards x as the sine of an angle
and assigns to x the angle with that sine.
Another notation that is commonly used for the inverse sine function is arcsin x . One reason
the notation arcsin x is preferable to Sin −1 x is the possible misinterpretation of Sin −1 x .
Sometimes students regard the “ −1” as a power and write Sin −1 x as 1/ sin x . This is not
correct, a fact that you were warned about in Section 1.6. Sin −1 is the name of a function,
√
just as sin is the name of the sine function, and is the notation for the positive square root
function. The capital S in Sin −1 should also warn you that this is the inverse sine function.
EXAMPLE 1.39
Simplify each of the following expressions:
√
(a) Sin−1 (− 3/2) (b) Sin−4 1 (1) √ 5 (c) Sin−1 (3)
(d) Sin−1 (3/5) + Sin−1 (4/5) (e) sin Sin−1 ( 3/2)
SOLUTION
√ √
(a) Sin −1 (− 3/2) asks for the angle whose sine is equal to − 3/2. Clearly,
√
Sin −1 (− 3/2) = −π/3.
FIGURE 1.103 Tri- (b) Sin −1 (1) = π/2.

angle to fit the statement φ = (c) Sin −1 (3) is not defined since the domain of Sin −1 x is −1 ≤ x ≤ 1.
Sin −1 (3/5)
(d) If φ = Sin −1 (3/5) , then φ is illustrated in the triangle in Figure 1.103. Since the
third side must have length 4, it follows that Sin −1 (4/5) = π/2 − φ , and
5 " # " # 2π 3
2 3 3 4 π
Sin −1 + Sin −1 =φ+ −φ = .
5 5 2 2
4 √ 5 √
(e) sin Sin −1 ( 3/2) = sin (π/3) = 3/2.
EXAMPLE 1.40
Find the values of x for which the following are valid:
. /
(a) sin Sin −1 x = x (b) Sin −1 (sin x) = x
SOLUTION These two equations express the fact that the sine function and the inverse sine
function are inverses, provided that we are careful about domains:
. /
(a) The equation sin Sin −1 x = x is valid for −1 ≤ x ≤ 1. Given an x in this
interval, Sin −1.x finds that
/ angle (in the principal value range) which has x as its
sine. Then sin Sin −1 x takes the sine of this angle. Naturally, it returns the original
number x .
(b) The function Sin −1 (sin x) is defined for all x , but only on the domain −π/2 ≤
x ≤ π/2 is it equal to x .
Our analysis of the inverse sine function is now complete. We could give a similar discussion
for each of the other five trigonometric functions. Instead, we give an abbreviated version for
the inverse cosine function and tabulate results for the remaining four functions.
The reflection of the graph of the function f (x) = cos x is shown in Figure 1.104, and it
does not represent a function. If we restrict the y -values on the reflected curve to
0 ≤ y ≤ π, (1.57)
then we do obtain a function called the inverse cosine function, denoted by
y = Cos −1 x. (1.58)
The values in 1.57 are the principal values of the inverse cosine function. Note again that
Cos −1 x is not the inverse function of cos x , but of f (x) = cos x, 0 ≤ x ≤ π .
FIGURE 1.104 Inverse cosine function from reflection of cosine graph in line y = x
y=x
y = Cos−1 x 2
2
−1 1 x
y = cos x
When we write y = Cos −1 x , we read this as “ y equals inverse cosine x ” but it means
that “ y is an angle whose cosine is x ” for if y = Cos −1 x , then x = cos y .
The remaining four inverse trigonometric functions, along with those of inverse sine and
cosine, are shown in Figure 1.105. Black and blue curves represent reflections in the line y = x
of the trigonometric functions tan x , cot x , csc x , and sec x . Blue curves represent princi-
pal values of the inverse trigonometric functions Tan −1 x , Cot −1 x , Csc −1 x , and Sec −1 x .
Principal values of the six inverse trigonometric functions are listed in Table 1.3.
FIGURE 1.105a Graph of FIGURE 1.105b Graph of FIGURE 1.105c Graph of

Sin −1 x Cos −1 x Tan −1 x
y y y
π π y = Tan−1 x
2
y = Cos−1 x π
π π
−1 1 x 2 2
y = Sin−1 x x
π
− −1 1 x −
π
2 2
−π
FIGURE 1.105d Graph of FIGURE 1.105e Graph of FIGURE 1.105f Graph of

Cot −1 x Csc −1 x Sec −1 x
y y y
y = Cot −1 x y = Csc−1 x y = Sec−1 x
π π π
π π π
2 2 2
1 1
x −1 x −1 x
π π π
− − −
2 2 2
−π −π −π
TABLE 1.3
Inverse Trigonometric Function Principal Values

π π
Sin −1 x − ≤y≤
2 2
π π
Tan −1 x − <y<
2 2
−1
Cos x 0≤y≤π
Cot −1 x 0<y<π
π π
Csc −1 x −π < y ≤ − , 0 < y ≤
2 2
π π
Sec −1 x −π ≤ y < − , 0 ≤ y <
2 2
It would be reasonable to ask why principal values of Sec −1 x were not chosen as 0 ≤ y <
π/2, π/2 < y ≤ π . Had they been chosen in this way, they would have been very similar
to those of Cos −1 x and Cot −1 x . Likewise, why are the principal values of Csc −1 x not
−π/2 ≤ y < 0, 0 < y ≤ π/2? The answer is that they could have been selected in this
way, and some authors do indeed make this choice; it is simply a matter of preference. Each
choice does, however, create corresponding changes in later work. Specifically, in the exercises
of Section 3.10, if principal values of Sec −1 x and Csc −1 x are selected in this alternative way,
then derivatives of these functions are modified correspondingly.
In the remainder of this section we solve problems that make use of inverse trigonometric
functions.
EXAMPLE 1.41
Find all solutions of the following equations:
(a) sin x = 0.4
(b) 3 cos 2x = −0.21
(c) 5 tan (3x − 1) = 4
SOLUTION
(a) One solution of this equation is
x = arcsin(0.4) = 0.412 radians (to three decimal places).
A second solution is π − 0.412, and when we add multiples of 2π , we obtain the

complete set of solutions
FIGURE 1.106 Simplified 2nπ + 0.412, 2nπ + (π − 0.412),

representation of solutions to a
trigonometric equation
where n is an integer. Following the lead of Example 1.38, Figure 1.106 suggests
y that this set of numbers can be represented more compactly as an initial rotation of
π/2, plus or minus π/2 − 0.412 = 1.159, and possible multiples of 2π ; that is,
1.159 1.159
" #
π 4n + 1
± 1.159 + 2nπ = π ± 1.159.
2 2
0.412 0.412
x
Often in problems like this, x = arcsin(0.4) = 0.412 is the only solution given.
We can see the reasoning behind this conclusion. There is only one principal value of
the inverse sine function, and therefore one solution to the equation. But the equation
sin x = 0.4 says nothing about the inverse sine function. We have introduced it
simply as a convenience. When we obtain the solution 0.412, we must ask whether
there are other solutions to the original equation. For the equation sin x = 0.4, there
are other solutions. In some problems, the principal value is the only acceptable
solution. Remember, then, if we introduce an inverse trigonometric function into a
problem along with its corresponding principal values, we must ask whether there
are other possibilities besides the principal values.
(b) Since cos 2x = −0.07, one solution of this equation for 2x is
2x = Cos −1 (−0.07) = 1.6409 radians (to four decimal places).
But there are others. Since the cosine function is even, all solutions are given by
2x = ±1.6409 + 2nπ (where n is an integer).
Consequently,
±1.6409 + 2nπ
x = = nπ ± 0.820 (to three decimal places).
2
(c) Since tan (3x − 1) = 0.8, one solution of this equation for 3x − 1 is
3x − 1 = Tan −1 (0.8) = 0.6747 radians (to four decimal places).
Since the tangent function is π -periodic, all solutions can be expressed in the form
3x − 1 = 0.6747 + nπ (where n is an integer).
Consequently,
1 + 0.6747 + nπ nπ
x = = 0.558 + (to three decimal places).
3 3
EXAMPLE 1.42
Find all solutions of the equation
cos2 x + 3 cos x = 2.
SOLUTION The quadratic formula applied to the equation
(cos x)2 + 3(cos x) − 2 = 0
yields
√ √
−3 ± 9+8 −3 ± 17
cos x = = .
2 2
Since cos x can √
only take on values in the interval −1 ≤ cos x ≤ 1, the possibility that
cos x = (−3 − 17)/2 must be rejected, leaving
√
−3 + 17
cos x = .
2
√
From the inverse cosine solution x = Cos −1 [( 17 − 3)/2)] = 0.975 radians, we obtain all
solutions
2nπ ± 0.975,
where n is an integer.
EXAMPLE 1.43
Find all times when the mass in Example 1.36 is 2.5 centimetres to the right of its equilibrium
position.
SOLUTION Since the position of the mass is given by x(t) = 6 sin [2(t + π/6)], it is
2.5 centimetres to the right of its equilibrium position when
2.5 = 6 sin [2(t + π/6)].
One solution of this equation for 2(t + π/6) is Sin −1 (2.5/6) = Sin −1 (5/12) . All solutions
of the equation are given by
2 π3 π 6π 7
2 t + = ± − Sin −1 (5/12) + 2nπ
6 2 2
where n is an integer. We can now solve this equation for t :
% &
π π π 1 −1
t+ = ± − Sin (5/12) + nπ
6 4 4 2
% &
π π 1
t = ± − Sin−1 (5/12) + nπ
12 4 2
" #
12n + 1
t = π ± 0.571
12
Since t must be positive, acceptable solutions are
" #
 12n + 1

 π + 0.571, where n ≥ 0
12 #
t = "

 12n + 1
 π − 0.571, where n > 0.
12
EXAMPLE 1.44
Find all solutions of the equation
1
tan (cos x) = √ .
3
FIGURE 1.107 Simplified √
solutions of a trigonometric SOLUTION If√we set y = cos x , then tan y = 1/ 3, and one solution of this equation is
equation y = Tan −1 (1/ 3) = π/6. But there are many other solutions for y , namely
y π
y = + nπ,
6
where n is an integer. But y = cos x , and cos x must take on values in the interval −1 ≤
1.02 cos x ≤ 1. There is only one possibility for n , namely n = 0; hence,
π
x cos x = .
6
1.02
From the solution x = Cos −1 (π/6) = 1.02 and Figure 1.107, we obtain
x = ±1.02 + 2nπ,
where n is an integer, as the complete set of solutions.
EXAMPLE 1.45
Find constants A > 0 and 0 < φ < π so that the function f (x) = 3 cos ωx − 4 sin ωx ,
where ω > 0 is a constant, can be expressed in the form A sin (ωx + φ) for all real x .
SOLUTION If we expand A sin (ωx + φ) by means of compound-angle formula 1.43a, and
equate it to f (x) , we have
A[sin ωx cos φ + cos ωx sin φ ] = 3 cos ωx − 4 sin ωx.
This equation will be valid for all x if we can find values of A and φ so that
A cos φ = −4 and A sin φ = 3.
When we square and add these equations, the result is
A2 cos2 φ + A2 sin2 φ = A2 = (−4)2 + (3)2 = 25.
Consequently, A = 5, and
4 3
cos φ = − and sin φ = .
5 5
The only angle in the range 0 < φ < π satisfying these equations is φ = arccos(−4/5) =
2.50 radians. Notice that arcsin(3/5) does not give this angle. Thus, f (x) can be expressed in
the form
f (x) = 5 sin (ωx + 2.50).
Perpendicularity and parallelism deal with lines that make a right angle at their point of inter-
section or that make no angle since parallel lines do not intersect. Lines that intersect usually do
so at angles other than π/2 radians. In order to determine the angle at which two lines intersect,
we first define the inclination of a line.
DEFINITION 1.5
The inclination of a line l is the angle of rotation φ (0 ≤ φ < π ) from the positive
x -direction to the line.
Line y = x + 3 in Figure 1.108 has inclination π/4 radians, and line y = −x + 4 has
inclination 3π/4 radians.
FIGURE 1.108 Inclinations of lines with slopes ±1
y=x+3 y = −x + 4
When the slope m of a line is positive, as shown in Figure 1.109a, it is clear that φ must
be in the interval 0 < φ < π/2, and m and φ are related by tan φ = m . When the slope of l
is negative as in Figure 1.109b, we use identity 1.44b to write
tan π − tan φ
m = − tan (π − φ) = − = tan φ.
1 + tan π tan φ
FIGURE 1.109a Inclina- FIGURE 1.109b Inclina-

tion of line with positive slope tion of line with negative slope
y y
l l
x x
Thus, whenever the slope m of a line is defined, the inclination is related to m by the equation
tan φ = m. (1.59)
In some sense this equation is true even when m is not defined. This occurs for vertical lines,
which have no slope. For a vertical line, φ = π/2 and tan φ is undefined. Thus, equation
1.59 is also valid for vertical lines from the point of view that neither side of the equation is
defined. Notice that it is not correct for us to write φ = Tan −1 m, since the principal values
of the inverse tangent function ( −π/2 < Tan −1 m < π/2) do not coincide with the specified
values for inclination (0 ≤ φ < π ).
EXAMPLE 1.46
What are the inclinations of the lines 2x − 3y = 4 and 2x + 3y = 4?
SOLUTION From y = 2x/3 − 4/3, the slope of the first line is 2/3. The inclination of this
line is φ = Tan −1 (2/3) = 0.588 radians. From y = −2x/3 + 4/3, the slope of the second
line is −2/3. The inclination of this line is φ = π + Tan −1 (−2/3) = 2.55 radians.
When two lines l1 and l2 with nonzero slopes m1 and m2 intersect (Figure 1.110), the angle θ
(0 < θ < π ) between the lines is given by the equation
φ 1 = θ + φ2 or θ = φ1 − φ2 .
By applying the tangent function to both sides of this equation and using identity 1.44b, we can
express θ in terms of m1 and m2 ,
FIGURE 1.110 Angle
between intersecting lines tan φ1 − tan φ2 m1 − m2
tan θ = tan (φ1 − φ2 ) = = .
1 + tan φ1 tan φ2 1 + m1 m2
y
This equation determines θ when φ1 > φ2 . When φ1 < φ2 , the equation is replaced by
m2 − m1
tan θ = .
l2 l1 1 + m1 m 2
In both cases we may write ) )
x ) m1 − m 2 )
tan θ = )) ),
)
1+m m 1 2
for the acute angle between the lines. It follows then that the acute angle between two lines
with slopes m1 and m2 is ) )
) m 1 − m2 )
θ = Tan −1 )) ). (1.60)
1 + m1 m 2 )
EXAMPLE 1.47
Find the angle between the lines 2x − 3y = 4 and x + 4y = 6.
SOLUTION Since slopes of these lines are 2/3 and −1/4, it follows that
) )
) 2/3 − (−1/4) )
−1 )
θ = Tan )
) 1 + (2/3)(−1/4) ) = 0.833 radians.
EXERCISES 1.8
In Exercises 1–16 evaluate the expression (if it has a value). 23. 4 sin2 x − 2 cos2 x = 1 24. 4 sin2 x + 2 cos2 x = 1
1. Tan −1 (−1/3) 2. Sin −1 (1/4) 25. cos2 x − 3 cos x + 1 = 0 26. sin2 x − 3 sin x − 5 = 0
√ √
3. Sec −1 ( 3) 4. Csc −1 (−2/ 3)
5. Cot −1 (1) 6. Cos −1 (3/2) In Exercises 27–30 draw a graph of the function.
7. Sin −1 (π/2) 8. Tan −1 (−1) . /2

√ / 27. f (x) = 2 + Csc −1 x
. −1
. /
9. sin Tan 3 10. tan Sin −1 3 √ √
28. f (x) = Tan −1 x + Sec −1 x
−1 −1
11. Sin [tan (1/6)] 12. Tan [sin (1/6)]
4 5 29. f (x) = Sin −1 (x − 3)
13. sec Cos −1 (1/2) 14. Sin −1 [sin (3π/4)]
4 √ 5 4 . √ /5 30. f (x) = Sin −1 x + Csc −1 x
15. sin Sin −1 (1/ 2) 16. Sin −1 cos Sec −1 (− 2)
In Exercises 17–26 find all solutions of the equation. In Exercises 31–36 find the inclination of the line.
17. sin x = 1/3 18. tan x = −1.2 31. x − y + 1 = 0 32. x + 2y = 3
19. cos 2x = 1/3 20. cot 4x + 1 = −1.2 33. 3x − 2y = 1 34. y − 3x = 4

21. 2 sin (1 − x) = 1.4 22. 3 tan 3x + 2 = −1.2 35. x = 4 36. y = 2
In Exercises 37–44 determine whether the lines are perpendicular, par- line so that the resulting signal is x(t) = f (t) + g(t) + h(t) . Express
allel, or neither. In the last case determine the angle between the lines. x(t) in the form A sin (ωt + φ) for appropriate values of A > 0 and
−π < φ < π .
37. y = −x + 4, y = x + 6
38. x + 3y = 4, 2x + 6y = 7 ∗ 63. Repeat Exercise 62 but express x(t) in the form A cos (ωt + φ) .
39. x = 3y + 4, y = x/3 − 2
∗ 64. A crank of length R with slider C is rotating clockwise about O
40. 2x + 3y = 1, 3x − 2y = 5 as shown below. The slider moves in a slotted lever hinged at A at a
41. y = 3x + 2, y = −x/2 + 1 distance L from O . Find angle θ as a function of angle φ .
42. x − y = 5, 2x + 3y = 4
Slider
43. x = 0, y = 5
φ C
44. x + y + 2 = 0, 3x − y = 4
O R
∗ 45. If φ is the angle formed by AB and AO in the figure below, find L
φ as a function of θ . θ
A
y
B ∗ 65. The angle of elevation of the top of a tower from A is φ , and
l L the angle from B at a distance d from A is θ (figure below). Find a
formula for θ in terms of φ .
O A x
x
h
In Exercises 46–51 find all solutions of the equation.
∗ 46. sin x tan2 x − 3 + tan2 x − 3 sin x = 0 θ φ

B A
∗ 47. sin x + cos x = 1
√ . / d
∗ 48. sec (sin x) = − 2 ∗ 49. cos Sin −1 x = 1/2
√ ∗ 66. A pendulum consists of a mass m suspended from a string of length
∗ 50. sec (tan x) = − 2 L (figure below). At time t = 0, the mass is pulled through a small
∗ 51. Cos −1 [tan (x 2 + 4)] = 2π − 5 angle θ0 — to the right when θ0 > 0, and to the left when θ0 < 0
— and given an initial speed v0 to the right. Its subsequent angular
∗ 52. Draw graphs of the following functions: (a) f (x) = sin (Sin −1 x) ;
displacements are given by
(b) f (x) = Sin −1 (sin x) .
∗ 53. Draw graphs of the following functions: (a) f (x) = cos (Cos −1 x) ; v0
(b) f (x) = Cos −1 (cos x) . θ = θ(t) = θ0 cos ωt + sin ωt, t ≥ 0,
ωL
∗ 54. Express the function f (x) = 4 sin 2x + cos 2x in the form
R sin (2x + φ) , where R > 0 and 0 < φ < π . √
where ω = 9.81/L , provided that any resistance due to the air is
∗ 55. Express the function f (x) = −2 sin 3x + 4 cos 3x in the form neglected. Show that θ(t) can be expressed in the form
R cos (3x + φ) , where R > 0 and 0 < φ < π .
∗ 56. Repeat Exercise 54 for f (x) = −2 sin 2x + 4 cos 2x . '
v02
∗ 57. Repeat Exercise 55 for f (x) = −4 sin 3x + 5 cos 3x . θ(t) = θ02 + sin (ωt + φ),
ω2 L2
∗ 58. Two electric signals f (t) = 4 cos (ωt + 2π/3) and g(t) =
3 sin (ωt + π/3) are fed into the same line so that the resulting signal
is x(t) = f (t) + g(t) . Express x(t) in the form A sin (ωt + φ) for where φ = Tan −1 (ωLθ0 /v0 ) .
appropriate values of A > 0 and −π < φ < π .
∗ 59. Repeat Exercise 58 but express x(t) in the form A cos (ωt + φ) .
∗ 60. Two electric signals f (t) = 2 sin (ωt + 4) and g(t) =
3 sin (ωt + 1) are fed into the same line so that the resulting signal
is x(t) = f (t) + g(t) . Express x(t) in the form A cos (ωt + φ) for L
appropriate values of A > 0 and −π < φ < π .
∗ 61. Repeat Exercise 60 but express x(t) in the form A sin (ωt + φ) .
∗ 62. Three electric signals f (t) = 5 cos (ωt + 3π/2) , g(t) =
4 cos (ωt + π/3) , and h(t) = 2 sin (ωt + π/4) are fed into the same m
∗ 67. What changes, if any, must be made in Exercise 66 if the initial ∗ 72. Repeat Exercise 71 if 5 cos ωt = A cos(ωt + 1) + 5 sin(ωt + φ) .
speed v0 is to the left rather than the right?
∗ 68. A mass m is suspended from a spring with constant k (figure ∗ 73. Show that
below). If at time t = 0, the mass is given an initial displacement y0
and an upward speed v0 , its subsequent displacements are given by  " #

 −1 1
v0 
 Sin , x ≥ 1,
y = y(t) = y0 cos ωt + sin ωt, t ≥ 0, −1
x
ω Csc x = " #

 −1 1
√ 
 −π − Sin , x ≤ −1.
where ω = k/m , provided that any resistance due to the air is x
neglected. Show that y(t) can be expressed in the form
'
v2 ∗ 74. Show that
y(t) = y02 + 02 sin (ωt + φ),
ω 8 √
−1 −Cos−1 1 − x 2 , −1 ≤ x < 0,
where φ = Tan −1
(ωy0 /v0 ) . Sin x = √
Cos−1 1 − x 2 , 0 ≤ x ≤ 1.
∗ 75. Prove that

k  " #

 −1 1
y 
 Cos , x ≥ 1,
y = 0 at −1
x
Sec x = " #
equilibrium m 
 1

 −Cos−1 , x ≤ −1.
x
∗ 69. What changes, if any, must be made in Exercise 68 if the initial
speed v0 is downward rather than upward?
∗ 76. Prove that
∗ 70. An inductance L , resistance R , and capacitance C are con-
nected in series with a generator producing an oscillatory voltage  " #
E = E0 cos ωt , t ≥ 0 (figure below). If L , C , R , E0 > 0, and 
 −1 1

 Tan , x > 0,
ω > 0 are all constants, the steady-state current I in the circuit is −1
x
Cot x = " #
E0 
 1

 π + Tan−1 , x < 0.
I = ' " #2 x
1
R 2 + ωL −
ωC
% " # & ∗ 77. Verify that
1
• R cos ωt + ωL − sin ωt .
8
ωC √
−1 Tan−1 x 2 − 1, x ≥ 1,
Express I in the form Sec x = √
−1 2
−π + Tan x − 1, x ≤ −1.
I = A cos (ωt − φ),
where A > 0 and −π/2 ≤ φ ≤ π/2. ∗ 78. Verify that
8 √
L
Cot−1 x 2 − 1, x ≥ 1,
Csc −1 x = √
−1 2
C E −π + Cot x − 1, x ≤ −1.
R
∗ 79. Verify that if 0 ≤ x < 1, then
∗ 71. It is known that 5 cos ωt = A cos (ωt − π/6) + 5 cos (ωt + φ) $

for all t , where ω is a fixed constant. Find, exactly, all possible values 1+x
2 Tan −1 = π − Cos −1 x.
for A > 0 and φ . 1−x
1.9 Exponential and Logarithm Review


1. Evaluate the following quantities: (a) log2 32 (b) log3 (1/27) In questions 5–8 draw a graph of the function.
(c) 10log10 2x (d) e2ln 4
5. f (x) = e3x 6. f (x) = log10 (x − 1)
In questions 2–4 find all solutions of the equation.
7. f (x) = 4−x 8. f (x) = ex + e−x
2x+1
2. log10 3x = 2 3. e =4
4. ln x − ln (x − 1) = 1 9. If y = 4e2x−1 , find x in terms of y .
In elementary algebra we learned the basic rules for products and quotients of powers:
a b a c = a b+c ; (1.61a)
ab
= a b−c ; (1.61b)
ac
. b /c
a = a bc ; (1.61c)
where a > 0, and b and c are real constants. These rules are used to develop the exponential
function
f (x) = a x , (1.62)
that is, a raised to the exponent x for variable x .
We concentrate on the case when a > 1. The meaning of a x when x is an integer is clear,
and when x = 1/n , where n > 0 is an integer, a x = a 1/n is the nth root of a . When x is a
positive rational number n/m ( n and m positive integers), 1.61c implies that
. /1/m . /n
a x = a n/m = a n or a x = a n/m = a 1/m ;
that is, a n/m is the mth root of a to the integer power n , or a n/m is the integer power n of the
mth root of a . When x is a negative rational, we write a x = 1/a −x , where −x is positive.
These results lead to the points and the graph of y = a x in Figure 1.111.
FIGURE 1.111 The exponential function
y
y = ax
(a > 1)
(3, a 3)
−1, 1 (2, a 2 )
a
−2, 1 1 (1, a)
a2
−3 −2 −1 1 2 3 4 5 x
There is a difficulty with the definition of a x and its graph in Figure 1.111. We are not
prepared to resolve the problem now, but we would be remiss in not pointing it out. How do
√
we define a x when x is an irrational number? For instance, what is the value of a 2 or a π ?
If a x is undefined whenever x is irrational, the graph in Figure 1.111 is misleading. Although
we have joined the points with the smoothest possible curve, there is actually an infinite number
of values of x (the irrational numbers) at which there is, as yet, no dot on the curve. We
require limits from Chapter 2 to deal with this problem, and we therefore set it aside until
Section 2.4.
EXAMPLE 1.48
Plot graphs of the exponential functions 2x and 3x on the same axes.
SOLUTION Graphs of these functions are shown in Figure 1.112. Notice that both curves
pass through the point (0, 1) . More generally, a 0 = 1 for any a . When x > 0, the graph of
3x is higher than that of 2x , whereas the opposite is true for x < 0.
FIGURE 1.112 Two exponential functions
y
y = 3x
y = 2x
y = 2x
y = 3x
x
EXAMPLE 1.49
Draw graphs of the functions x 4 (a power function) and 2x (an exponential function) on the
same axes.
SOLUTION Graphs of these functions are shown in Figure 1.113, but no attempt has been
made to use a scale on either the x - or y -axis. Notice here that for x > 16, 2x > x 4 . This is
always the situation for power and exponential functions. Given any exponential function a x
( a > 1), and any power function x n ( n > 1), there always exists a value of x , say X , such
that when x > X , we have a x > x n . In short, exponential functions grow more rapidly than
power functions for large values of x .
FIGURE 1.113 Comparison of graphs of power and exponential functions
y = x4
y = 2x
1 16 x
In terms of the exponential function, rules 1.61 take the form
a x1 a x2 = a x1 +x2 , (1.63a)
x1
a
= a x1 −x2 , (1.63b)
a x2
. x1 /x2
a = a x1 x2 . (1.63c)
The exponential function is one-to-one; its graph passes the horizontal line test. As a result,
it has an inverse; we call it the logarithm function to base a , denoted by loga x . As an inverse,
the logarithm function reverses the action of the exponential function. For instance, working
with base 10, since 103 = 1000, it follows that log10 1000 = 3; since 10−2 = 0.01, we
have log10 0.01 = −2. We can work with logarithms to any base, but it is customary to use
bases greater than 1. With base 3, say, we can write that log3 81 = 4 since 34 = 81, and
that log3 (1/9) = −2 since 3−2 = 1/9. These examples demonstrate that the logarithm of a
number is a power. To say that y is the logarithm of x to base a is to say that x = a y . In
general, the logarithm of x to base a is the power to which a must be raised in order to produce
x . Algebraically, we write this as
y = loga x only if x = ay . (1.64)
When the first of these is substituted into the second, the result is
a loga x = x. (1.65a)
When the second is substituted into the first,
y = loga (a y ),
or since y is arbitrary, we may replace it with x ,
loga a x = x. (1.65b)
The second of these is valid for all x , but the first is true only for x > 0. These equations
simply express the fact that exponential and logarithm functions are inverses of each other; do
one, then the other, and you are back where you started (see equations 1.35 and 1.36).
A graph of the logarithm function f (x) = loga x can be obtained by reflecting the graph
of y = a x in the line y = x (Figure 1.114). It passes through the point (1, 0) for any a . In
other words, loga 1 = 0. As values of x get closer and closer to 0, their logarithms become
very large negative numbers. We cannot take the logarithms of 0 or negative numbers.
FIGURE 1.114 Logarithm function
y = log a x
1 x
Corresponding to rules 1.63 for the exponential function are the following rules for loga-
rithms:
loga (x1 x2 ) = loga x1 + loga x2 , (1.66a)

" #
x1
loga = loga x1 − loga x2 , (1.66b)
x2
. x /
loga x1 2 = x2 loga x1 . (1.66c)
To prove 1.66a, say, we set z = loga (x1 x2 ) , in which case
a z = x1 x 2
. /. /
= a loga x1 a loga x2 (using 1.65a)
= a loga x1 +loga x2 . (using 1.63a)
Thus,
loga x1 + loga x2 = z = loga (x1 x2 ).
We leave proofs of 1.66b and 1.66c to the exercises.
EXAMPLE 1.50
Simplify the following expressions:
2 . /
(a) 3log3 (x )
(b) 10−4 log10 x (c) loga a −x+3 (d) log2 8 + log3 (1/27)
SOLUTION
(a) Identity 1.65a implies that
2
3log3 (x )
= x2.
(b) Since −4 intervenes between the logarithm and exponential operations, we cannot
use 1.65a immediately. The −4 can be relocated, however, with 1.66c:
−4 ) 1
10−4 log10 x = 10log10 (x = (if x > 0).
x4
(c) Identity 1.65b gives
. /
loga a −x+3 = −x + 3.
(d) Since log2 8 = 3 and log3 (1/27) = −3,
log2 8 + log3 (1/27) = 3 − 3 = 0.

EXAMPLE 1.51
Solve the following equations:
(a) log5 x = −3 (b) log10 x + log10 (x + 1) = 0 (c) 10x − 12 + 10−x = 0
SOLUTION
(a) By means of equation 1.64,
1
x = 5−3 = .
125
(b) Since log10 x + log10 (x + 1) = log10 [x(x + 1)], we can write
0 = log10 [x(x + 1)].
If we now take exponentials to base 10,
100 = 10log10 [x(x+1)] or 1 = x(x + 1).
This quadratic equation has solutions

√ √
−1 ± 1+4 −1 ± 5
x = = .
2 2
Since √
x must be positive (the original equation demands this), the only solution is
x = ( 5 − 1)/2.
(c) If we multiply the equation by 10x , the result is
0 = 102x − 12(10x ) + 1 = (10x )2 − 12(10x ) + 1.
But this is a quadratic equation in 10x , so that

√
x 12 ± 144 − 4 √
10 = = 6± 35.
2
Finally, we have
2 √ 3
x = log10 6 ± 35 .
EXAMPLE 1.52
The ear hears by detecting pressure variations of impinging sound waves. The loudness of
the sound is related to the intensity of the sound wave, which is measured in watts per square
metre (energy transmitted by the sound wave per unit time per unit area). The lowest intensity
detectable by the ear is normally taken as I0 = 10−12 W/m 2 at a frequency of 1000 Hz; it is
called the audible sound threshold. By comparison, the intensity of sound from a jet engine
is about 104 W/m 2 ; it is 1016 times that of the audible sound threshold. Because the range of
intensities to which the ear is sensitive is so large, dealing directly with intensities is cumbersome.
Logarithms provide a way to reduce this enormous range to a manageable size. As a number
increases by a factor of 10, its logarithm increases by 1. For example, the difference between the
logarithms of 10 and 100 is 2 − 1 = 1. If a number increases by a factor of 1014 , its logarithm
increases by 14. This range is deemed to be a little too compact; it is expanded by a factor of
10 in the following definition. The loudness of a sound is said to be L decibels if
" #
I
L = 10 log10 ,
I0
where I is the intensity of the sound and I0 is the intensity of sound at the audible threshold.
Use this definition to answer the following questions.
(a) What is the loudness of sound at the audible sound threshold?

(b) Express the intensity I of a sound in terms of I0 and its decibel reading L .
(c) If decibel readings for a voice, a car, and a jet engine are 70, 100, and 160, respectively,
what are the corresponding intensities of the sound waves relative to I0 ?
(d) If the pain threshold for sound has an intensity 1014 times I0 , what is its decibel
reading?
(e) If the intensity I1 of one sound is 10 times the intensity I2 of a second sound, how
do their decibel readings compare?
SOLUTION
(a) The decibel reading for the audible sound threshold is L = 10 log10 (I0 /I0 ) = 0.
(b) When we take both sides as exponents of powers of 10, and use properties 1.66c and
1.65a, we obtain
" #10
L 10 log10 (I /I0 ) log10 (I /I0 )10 I
10 = 10 = 10 = .
I0
If we take 10th roots of both sides, we have

. /1/10 I
10L = or I = I0 10L/10 .
I0
(c) Since the decibel level of the normal voice is 70, its intensity is
I = I0 1070/10 = 107 I0 .
Similarly, the intensities of a car and a jet are 1010 and 1016 times I0 .
(d) For an intensity of 1014 I0 , the decibel reading is
L = 10 log10 (1014 ) = 10(14) = 140.

(e) If L1 and L2 are the decibel readings for sounds with intensities I1 and I2 , then
L1 = 10 log10 (I1 /I0 ) and L2 = 10 log10 (I2 /I0 ).
When we subtract these readings, we obtain
L1 − L2 = 10 log10 (I1 /I0 ) − 10 log10 (I2 /I0 )

= 10[log10 (I1 /I0 ) − log10 (I2 /I0 )]
" #
I1 /I0
= 10 log10 (using 1.66b)
I2 /I0
" #
I1
= 10 log10
I2
" #
10I2
= 10 log10 (since I1 = 10I2 )
I2
= 10 log10 10
= 10.
Thus, when the intensity of one sound is 10 times that of another, their decibel readings
differ by 10.
It is sometimes necessary to change from one base of logarithms to another. If we take logarithms
to base b on both sides of identity 1.65a, we obtain immediately
logb x = (loga x)(logb a). (1.67)
This equation defines logb a as the conversion factor from logarithms to base a to logarithms
to base b .
Before the discovery of calculus, the base of logarithms was invariably chosen to be 10.
TABLE 1.4
Such logarithms are called common logarithms; they correspond to the exponential function
. /
1 n 10x . Another base for exponentials and logarithms, however, that is much more convenient in
n 1+ n most applications is a number, denoted by the letter e , and defined in a variety of ways. One
1 2.000 000 way is to consider the numbers
" #n
3 2.370 370 1
1+ (1.68)
5 2.488 320 n
10 2.593 742 for ever-increasing values of n . The numbers in Table 1.4 are steadily increasing but getting
100 2.704 814
closer together. They suggest that for larger and larger values of n , the function (1 + 1/n)n is
indeed getting closer to some number — to 12 decimal places this number is
1 000 2.716 924
10 000 2.718 146 e ≈ 2.718281828459.
100 000 2.718 255
1 000 000 2.718 282 This number e is irrational, with a nonterminating, nonrepeating decimal expansion. Why
this number is so convenient as a base for logarithms and exponentials is shown in Section 3.11.
For now, let us rewrite some of the more important formulas of this section with a set equal to
e . The exponential function to base e is ex , and equations 1.63 in terms of ex read
ex1 ex2 = ex1 +x2 , (1.69a)

x1
e
= ex1 −x2 , (1.69b)
e x2
. x1 /x2
e = e x1 x2 . (1.69c)
Logarithms to base e are usually given the notation ln x rather than loge x , and are called
natural logarithms:
ln x = loge x. (1.70)
In terms of ln x , rules 1.66 are
ln (x1 x2 ) = ln x1 + ln x2 , (1.71a)
" #
x1
ln = ln x1 − ln x2 , (1.71b)
x2
. x /
ln x1 2 = x2 ln x1 . (1.71c)
Identities 1.65 become
x = eln x , x > 0, (1.72a)
x = ln (ex ). (1.72b)
Graphs of ex and ln x in Figures 1.115 and 1.116 have the same shape as those of a x in
Figure 1.111 and loga x in Figure 1.114. Only the “steepness” of the curves is affected by a
change of base.
FIGURE 1.115 Graph of ex FIGURE 1.116 Graph of ln x
y y
y = ex y = ln x
1 x
x
Neither of the functions ex nor ln x is even or odd, but functions derived from them may
be even or odd. This is illustrated in the following example.
EXAMPLE 1.53
2
Is the function f (x) = e−ax , where a > 0 is a constant, even or odd? Draw its graph.
SOLUTION Since
2 2
f (−x) = e−a(−x) = e−ax = f (x),
2
the function e−ax is even. Graphs for a = 1, 2, and 3 are plotted in Figures 1.117a–c. The
value of a controls the spread of the curve. For any value of a , the graph passes through the
point (0, 1) and is asymptotic to the x -axis (Figure 1.117d). This curve is very important in
statistics. It is called the bell curve or normal distribution.
2 2
FIGURE 1.117a Graph of e−x FIGURE 1.117b Graph of e−2x
y y
1 1
y = e−x
2
0.8 0.8
y = e−2x
2
0.6 0.6
0.4 0.4
0.2 0.2
−2 −1 1 2x −2 −1 1 2 x
2 2
FIGURE 1.117c Graph of e−3x FIGURE 1.117d Graph of e−ax
y y
1
0.8
y = e−ax
2
y = e−3x
2
0.6
0.4
0.2
x
−2 −1 1 2 x
You will be making graphs of functions at every turn in this book. We remind you that when
we ask you to plot a graph, you are to use a graphing calculator or computer. When we ask you
to draw a graph, you are to do so without these devices. In the preceding example, we plotted
2 2
graphs of f (x) = e−ax for a = 1, 2, and 3, and then drew y = e−ax by hand.
EXAMPLE 1.54
Draw a graph of the function f (x) = ln (x + a) , where a > 0 is a constant.
SOLUTION By translating the graph of f (x) = ln x in Figure 1.118a to the left by a units,
we obtain the graph of f (x) = ln (x + a) in Figure 1.118b.
FIGURE 1.118a Graph of ln x FIGURE 1.118b Graph of ln x translated a units to the left
y y
y = ln (x + a)
y = ln x
ln a
1 x −a 1−a x
EXAMPLE 1.55
The emf device in the RC -circuit of Figure 1.119 produces a constant voltage of V volts. If
the switch is closed at time t = 0, and there is no initial charge on the capacitor, the charge
thereafter is given by
Q(t) = CV [1 − e−t/(RC) ], t ≥ 0.
Draw a graph of this function.
FIGURE 1.119 Schematic
SOLUTION A graph of et/(RC) (Figure 1.120a) has the same shape as that in Figure 1.115;
for RC-circuit
the constant RC affects only the steepness of the curve. The graph of e−t/(RC) in Figure 1.120b
C is that of Figure 1.120a reflected in the vertical axis. The graph of Q(t) in Figure 1.120c is
then obtained by turning Figure 1.120b upside down, shifting it upward 1 unit, changing the
scale on the vertical axis by a factor CV , and retaining only that part of the curve t ≥ 0. It is
asymptotic to the line Q = CV .
V R
FIGURE 1.120a Graph of et/(RC) FIGURE 1.120b Graph of e−t/(RC)
y y
S
y = e−t / (RC)
y = e t / (RC)
1 1
t t
FIGURE 1.120c Charge on capacitor
Q
CV
Q = CV [1 − e−t / (RC)]
EXERCISES 1.9
In Exercises 1–13 find all values of x satisfying the equation. divide the N people into groups of x people, pool their blood, and test
for the virus. If the blood is disease-free for a group, the individuals
1. log10 (2 + x) = −1 need not be tested separately. If the blood test is positive, each of the
2. 103x = 5 x people in the group is tested separately. It is shown in probability
3. log10 (x 2 + 2x + 1) = 1 theory that when the group size is x , then on the average, the expected
total number of tests needed to test the N people completely is
4. ln(x 2 + 2x + 10) = 1
% &
2
5. 105−x = 100 1
y = N 1 − (0.99)x + .
2
6. 101−x = 100
x
∗ 7. log10 (x − 3) + log10 x = 1 Plot this function for N = 100 to determine the group size that mini-
∗ 8. log10 (3 − x) + log10 x = 1 mizes y .
∗ 9. log10 [x(x − 3)] = 1
∗ 29. Show that f (x) = (1 − e−1/x )/(1 + e−1/x ) is an odd function.
∗ 10. 2 log10 x + log10 (x − 1) = 2
∗ 30. Is there a difference between the graphs of the functions f (x) =
∗ 11. loga x + loga (x + 2) = 2 loga (x 2 ) and g(x) = 2 loga x ?
∗ 12. loga [x(x + 2)] = 2
% " # & ∗ 31. (a) In the early period of reforestation, the percentage increase
x+3 of timber each year is almost constant. If the original
∗ 13. log10 log10 + 4 = −1
200x amount A0 of a certain timber increases 3.5% the first year
and 3.5% each year thereafter, find an expression for the
amount of timber after t years.
In Exercises 14–27 draw a graph of the function. As a check, use a
(b) How long does it take for timber of this type to double?
calculator or computer to plot a graph.
2
∗ 32. A new car costs $20 000. In any year it depreciates to 75% of its
∗ 14. f (x) = e−x ∗ 15. f (x) = ln (cos x) value at the beginning of that year. What is the value of the car after t
∗ 16. f (x) = loga |x 2 − 1| ∗ 17. f (x) = a loga (2x+1) years?
∗ 18. f (x) = log10 (4x) ∗ 19. f (x) = ln (1 − x) ∗ 33. If the effective height of the earth’s atmosphere (in metres) is the
2 2 solution of the equation
∗ 20. f (x) = ln (1 − x ) ∗ 21. f (x) = ln (x − 1)
∗ 22. f (x) = 10x+2 ∗ 23. f (x) = e2−x 10−6 = (1 − 2.08 × 10−6 y)56 ,
x2 −x 2
∗ 24. f (x) = e ∗ 25. f (x) = e sin x
find y .
−x 2 2 x
∗ 26. f (x) = e cos x ∗ 27. f (x) = x e , −1 ≤ x ≤ 1
∗ 34. Show that if y is the logarithm of x to base a , then −y is the
logarithm of x to base 1/a .
∗ 28. A large number N of people are to have their blood tested to
determine whether they have been infected by a virus. One way is to ∗ 35. Prove 1.66b and 1.66c.
test all N people individually, resulting in N tests. An alternative is to ∗ 36. Is identity 1.66a valid for all x1 and x2 ?
∗ 37. The magnitude of an earthquake is measured in much the same (a) The time constant τ for the circuit is the length of time for
way as noise level. An earthquake of minimal size is taken as having the current to become i0 /e . Find τ .
value 0 on the Richter scale. Any other earthquake of intensity I is
(b) Show that if i is the current at any time t , then the current
said to have magnitude R on the Richter scale if
at time t + τ is i/e .
" #
I
R = log10 ,
I0
In Exercises 41–44 find all values of x satisfying the equation.
where I0 is the intensity of the minimal earthquake being used as ref-
erence. ∗ 41. 3a 2x + 3a −2x = 10
(a) Express the intensity of an earthquake in terms of I0 and
its reading on the Richter scale. ∗ 42. 2x + 4x = 8x
(b) What are readings on the Richter scale of earthquakes that
have intensities 1.20 × 106 and 6.20 × 104 times I0 ? ∗ 43. 3x+4 = 7x−1
∗ 38. (a) If P dollars is invested at i % compounded n times per
∗ 44. logx 2 = log2x 8
year, show that the accumulated value after t years is
" #nt 2 √ 3
i
A = P 1+ . ∗ 45. Show that f (x) = ln x + x 2 + 1 is an odd function. Plot
100n
its graph to confirm this geometrically.
(b) How long does it take to double an investment if interest is
8% compounded semiannually? 46. Repair costs on the car in Exercise 32 are estimated at $50 the
(c) Calculate the maximum possible value of A if i is fixed but first year, increasing by 20% each year thereafter. Set up a function
the number of times that interest is compounded is unlim- C(t) that represents the average yearly cost of repairs associated with
ited; that is, calculate what happens to A as n gets larger owning the car for t years.
and larger and larger. This method of calculating interest
is called continuously compounded interest. ∗ 47. A straight-wire conductor has length 2L and circular cross-section
(d) What is the accumulated value of a $1000 investment after of radius R . If the wire carries current i > 0, then the magnitude of
10 years at 6% compounded continuously? Compare this the vector potential at a distance r from the centre of the wire is
to the accumulated value if interest is calculated only once
each year. % " # &

 µ0 i 4L2 r2
∗ 39. If, in the circuit shown below, V0 is the voltage across the capacitor 
 4π ln 1 + − 1 + , 0≤r ≤R
R2 R2
at time t = 0 when the switch is closed, the voltage thereafter is f (r) = " 2
#

V = V0 e−t/(RC) .  µ0 i ln 1 + 4L ,

r > R,
4π r2
S where µ0 is a positive constant.

C R (a) Draw a graph of this function.
(b) Find the radius r > R for which f (r) = f (0) .
(a) The time constant τ for the circuit is the length of time for
the voltage on the capacitor to become V0 /e . Find τ .
In Exercises 48–50 solve the given equation for y in terms of x .
(b) Show that if V is the voltage at any time t , then the voltage
at time t + τ is V /e .
e 2 x − e −2 x
∗ 40. If, in the circuit shown below, i0 is the current in the inductor at time ∗ 48. y =
2
t = 0 when the switch is closed, the current thereafter is i = i0 e−Rt/L .
ex + e−x
∗ 49. y =
S 2
L R
ex − e−x
∗ 50. y =
ex + e−x
1. (a) 5 (b) −3 (c) 2x (d) 16 (4 marks) 7. (2 marks)

y
2. e2 /3 (2 marks)
3. (ln 4 − 1)/2 (2 marks) 1
4. e/(e − 1) (3 marks)
x
5. (2 marks)
y
8. (3 marks)
y
1
2
6. (2 marks)
y
x
1 2 x
9. (1/2)[1 + ln (y/4)] (3 marks)
1.10 Hyperbolic Functions

This material is likely new for most students. It does not therefore have a diagnostic test
associated with it.
Certain combinations of the exponential function occur so often in physical applications
that they are given special names. Specifically, half the difference of ex and e−x is defined
as the hyperbolic sine function and half their sum is the hyperbolic cosine function. These
functions are denoted as follows:
ex − e−x ex + e−x
sinh x = and cosh x = . (1.73)
2 2
According to equation 1.32, they are the odd and even parts of ex .
The names of these hyperbolic functions and their notations bear a striking resemblance
to those for the trigonometric functions, and there are reasons for this. First, the hyperbolic
functions sinh x and cosh x are related to the curve x 2 − y 2 = 1 (see Figure 1.123), called the
unit hyperbola, in much the same way as the trigonometric functions sin x and cos x are related
to the unit circle x 2 + y 2 = 1. We will point out one of these similarities in Example 1.56.
Second, for each identity satisfied by the trigonometric functions, there is a corresponding
identity satisfied by the hyperbolic functions — not the same identity, but one very similar. For
example, using equations 1.73, we have

" #2 " #2
2 2 ex + e−x ex − e−x
(cosh x) − (sinh x) = −
2 2
1 4. / . /5
= e 2 x + 2 + e −2 x − e 2 x − 2 + e −2 x
4
= 1.
Thus the hyperbolic sine and cosine functions satisfy the identity
cosh 2 x − sinh 2 x = 1, (1.74)
which is reminiscent of the identity cos2 x + sin2 x = 1 for the trigonometric functions.
Just as four other trigonometric functions are defined in terms of sin x and cos x , four
corresponding hyperbolic functions are defined as follows:
sinh x ex − e−x cosh x ex + e−x

tanh x = = , coth x = = ,
cosh x ex + e−x sinh x ex − e−x
(1.75)
1 2 1 2
sech x = = , csch x = = .
cosh x ex + e−x sinh x ex − e−x
These definitions and 1.74 immediately imply that
1 − tanh 2 x = sech 2 x, (1.76a)
coth 2 x − 1 = csch 2 x, (1.76b)
analogous to 1 + tan2 x = sec2 x and 1 + cot 2 x = csc2 x , respectively.
FIGURE 1.121a Graph of cosh x FIGURE 1.121b Graph of sinh x FIGURE 1.121c Graph of tanh x
y y y
1
y = sinh x y = tanh x
y = cosh x x
1 x
−1
x
FIGURE 1.121d Graph of coth x FIGURE 1.121e Graph of sech x FIGURE 1.121f Graph of csch x
y y y
y = coth x 1 y = sech x
1 y = csch x
x
x x
−1
Graphs of the six hyperbolic functions are shown in Figures 1.121. The functions cosh x
and sech x are even; the other four are odd.
Most trigonometric identities can be derived from the compound-angle formulas for
sin (A ± B) and cos (A ± B) . It is easy to verify similar formulas for the hyperbolic functions:
sinh (A ± B) = sinh A cosh B ± cosh A sinh B, (1.77a)
cosh (A ± B) = cosh A cosh B ± sinh A sinh B. (1.77b)
For example, equations 1.73 give

" #" #
eA + e−A eB + e−B
cosh Acosh B − sinh Asinh B =
2 2
" −A # " #
eA − e eB − e−B
−
2 2
1 4. /
= eA+B + eA−B + eB−A + e−A−B
4
. /5
− eA+B − eA−B − eB−A + e−A−B
14 5
= eA−B + e−(A−B)
2
= cosh (A − B).
With these formulas, we can derive hyperbolic identities analogous to trigonometric identities
1.44–1.49:
tanh A ± tanh B
tanh (A ± B) = , (1.77c)
1 ± tanh A tanh B
sinh 2A = 2 sinh A cosh A, (1.77d)
2 2
cosh 2A = cosh A + sinh A (1.77e)
2
= 2 cosh A − 1 (1.77f)
= 1 + 2 sinh 2 A, (1.77g)
2 tanh A
tanh 2A = , (1.77h)
1 + tanh 2 A
1 1
sinh Asinh B = cosh (A + B) − cosh (A − B), (1.77i)
2 2
1 1
sinh Acosh B = sinh (A + B) + sinh (A − B), (1.77j)
2 2
1 1
cosh Acosh B = cosh (A + B) + cosh (A − B), (1.77k)
2 2
" # " #
A+B A−B
sinh A + sinh B = 2 sinh cosh , (1.77l)
2 2
" # " #
A+B A−B
sinh A − sinh B = 2 cosh sinh , (1.77m)
2 2
" # " #
A+B A−B
cosh A + cosh B = 2 cosh cosh , (1.77n)
2 2
" # " #
A+B A−B
cosh A − cosh B = 2 sinh sinh . (1.77o)
2 2
In Example 1.56, we illustrate a geometric parallel between the trigonometric sine and cosine
functions and the hyperbolic sine and cosine functions.
EXAMPLE 1.56
Show that:
(a) every point (x, y) on the unit circle x 2 + y 2 = 1 can be expressed in the form
x = cos t , y = sin t for some real number t in the interval 0 ≤ t < 2π ;
(b) every point (x, y) on the right half of the unit hyperbola x 2 − y 2 = 1 can be
expressed in the form x = cosh t , y = sinh t for some real number t .
SOLUTION
(a) If t is the angle in Figure 1.122, then clearly the coordinates of P are x = cos t and
y = sin t . As angle t ranges from 0 to 2π , P traces the circle exactly once.
(b) A sketch of the unit hyperbola x 2 − y 2 = 1 is shown in Figure 1.123. If x = cosh t
and y = sinh t are coordinates of a point P in the plane, where t is some real number,
then identity 1.74 implies that x 2 − y 2 = cosh 2 t − sinh 2 t = 1. In other words, P
is on the unit hyperbola. Furthermore, since x = cosh t is always positive, P must
be on the right half of the hyperbola. Finally, because the range of x = cosh t is
x ≥ 1 in Figure 1.121a, and the range of y = sinh t is −∞ < y < ∞ in Figure
1.121b, it follows that every point on the right half of the hyperbola can be obtained
from some value of t . Note that t is not the angle formed by the positive x -axis and
the line joining the origin to (x, y) .
FIGURE 1.122 Relationship between unit circle and FIGURE 1.123 Relationship between unit hyperbola and
trigonometric functions hyperbolic functions
y y
1 P (x, y)
P (x, y)
1
y t>0
t
x 1 x O 1 Q x
t<0
x2 + y2 = 1
x2 − y2 = 1
EXERCISES 1.10
In Exercises 1–10 evaluate the expression (if it has a value). 9. e−2 cosh e 10. sinh [Cot −1 (−3π/10)]
11. Verify the results in identities 1.77c–o.
1. 3 cosh 1 2. sinh (π/2)
∗ 12. Vertical vibrations of the beam in the figure below involve the
√ −1 function
3. tanh 1 − sin 3 4. Sin (sech 10)
y = f (x) = A cos kx + B sin kx + C cosh kx + D sinh kx
5. Cos −1 (2 csch 1) 6. coth (sinh 5)
√ where A , B , C , and D are constants such that C = −A , D = −B ,
7. ln |sinh (−3)| 8. sech [sec (π/3)] and A and B must satisfy the equations
Sech −1 x .
(c) Show that
A(cos kL − cosh kL) + B(sin kL − sinh kL) = 0, 2 3
!
Sinh−1 x = ln x + x2 + 1 ;
A(cos kL + cosh kL) + B(sin kL + sinh kL) = 0.
2 ! 3
Cosh−1 x = ln x + x2 − 1 ;
Eliminate A and B between these equations to show that k must satisfy
the condition " #
1 1+x
tan kL = tanh kL. Tanh−1 x = ln , |x| < 1.
2 1−x
∗∗ 14. When an object of mass m falls from rest at time t = 0 under the
y influence of gravity and an air resistance proportional to the square of
L
velocity, its velocity v as a function of time is defined by the equation
' "√ #
1 β mg/β − v βt
x ln √ =− ,
2 mg mg/β + v m
where β > 0 is a constant and g > 0 is the acceleration due to gravity.

∗ 13. Each hyperbolic function has associated with it an inverse hyper-
bolic function. (a) Show that when this equation is solved for v in terms of t ,
the result is
(a) Draw the inverse functions Sinh −1 x , Tanh −1 x , Coth −1 x , 0$ 1
and Csch −1 x for sinh x , tanh x , coth x , and csch x .
$
mg βg
v(t) = tanh t .
(b) Why do cosh x and sech x not have inverse functions? It is β m
customary to associate functions Cosh −1 x and Sech −1 x
with cosh x and sech x by restricting their domains to (b) Determine the limit of v for large t , called the limiting
nonnegative numbers. Draw graphs of Cosh −1 x and velocity.
1.11 Approximating Solutions to Equations

This material is likely new for most students. It does not therefore have a diagnostic test
associated with it.
In writing this book we have assumed that you have access to a graphing calculator and/or
a computer with a mathematical software package. Every software package has one or more
equation-solving commands; most graphing calculators have an equation-solving routine. In
other words, you have a device that solves equations either exactly or approximately. In this
section we discuss some useful concepts related to accuracies of approximations. Suppose, for
example, that we are to find all solutions of the equation
2x − 4 = cos (x 2 − 7x + 10). (1.78a)
We can express the equation in the equivalent form
f (x) = 2x − 4 − cos (x 2 − 7x + 10) = 0, (1.78b)
where solutions are now visualized as x -intercepts of the graph of the function f (x) . The graph
in Figure 1.124 indicates that the only solution of the equation is between x = 2 and x = 3.
FIGURE 1.124 Graphical solutions of 2x − 4 − cos(x 2 − 7x + 10) = 0
y
4
−1 1 2 3 4x
−2
−4
−6
Suppose we use our calculator or computer to find this root and the result is x = 2.323 762 277.
The number of digits depends on the device used and/or how it is programmed. For the purposes
of our discussion here, the number of digits is immaterial. Naturally, if 2.323 762 277 constitutes
the full display of a calculator, we might be skeptical of the accuracy of the last digit, perhaps
even the second-to-last digit. If this is computer output, we might be unsure of how the machine
arrived at the last digit. Did it round, and if so, what are its rules for rounding, or did it simply
truncate after the 10th digit? There is a very simple way to verify the accuracy of this, or any
other, approximate root of an equation. The following theorem provides it.
THEOREM 1.7 (Intermediate Value Theorem)

If a function f (x) is continuous on the closed interval a ≤ x ≤ b , and k is any number
between f (a) and f (b) , then there exists at least one value of c between a and b such
that f (c) = k .
A function f (x) is a continuous function on an interval a ≤ x ≤ b if its graph can

be traced between a and b without lifting pencil from paper. A more mathematical definition
of continuous will be given in Section 2.4. For our present purposes, the geometric version
suffices.
FIGURE 1.125a The intermediate value theorem FIGURE 1.125b The zero intermediate value theorem
y y
(b, f (b))
f(b) b, f(b)
k
f(a) (a, f(a))
a c b x
(a, f (a))
a c c c b x
Figure 1.125a illustrates Theorem 1.7 geometrically. Because the graph of f (x) can be
traced from the point (a, f (a)) to the point (b, f (b)) without lifting pencil from paper, it
follows that the graph must cross the horizontal line y = k at least once, and this is a value for
c . This particular figure shows three such values. Theorem 1.7 is what mathematicians call an
existence theorem. It states that a number c exists that satisfies f (c) = k for given k , but does
not provide a way to find c . We shall encounter other existence theorems in calculus.
What is important for our present purposes is the following corollary.
COROLLARY 1.7.1 (The Zero Intermediate Value Theorem)

If f (a)f (b) < 0 for a function f (x) that is continuous on a ≤ x ≤ b , then there exists
at least one number c between a and b for which f (c) = 0.
The condition f (a)f (b) < 0 requires that one of f (a) and f (b) be positive and the
other be negative. [We have shown f (a) < 0 and f (b) > 0 in Figure 1.125b.] The choice of
k = 0 in Theorem 1.7 gives this corollary.
Without continuity of f (x) , we cannot be sure, in general, whether there are solutions to
the equation f (x) = 0 when f (a)f (b) < 0. The function in Figure 1.126a has what is
called a discontinuity at x = d , and there are no solutions of f (x) = 0. The function f (x)
in Figure 1.126b also has a discontinuity at x = d , but there are two solutions of f (x) = 0
between a and b .
There are two common ways to discuss the accuracy of an approximation to the solution of
an equation, and simple as the zero intermediate value theorem is, it handles both situations.

Functions not satisfying the conditions of the zero intermediate value theorem
y y
a d b x a d b x
Approximations Rounded to a Specified Number of

Decimal Places
We say that x is an approximation to the root α of an equation
f (x) = 0, (1.79)
correctly rounded to k decimal places, if x has k decimal places, and α rounds to the same k
decimal places. For example, the approximation x = 2.323 762 277 rounded to four decimal
places is x = 2.3238. How can we verify that the root of equation 1.78b rounds to these same
four decimal places? We evaluate f (x) at x = 2.323 75 and x = 2.323 85,
f (2.323 75) = −0.000 047, f (2.323 85) = 0.000 33.
Because one of these values is positive and the other is negative, the zero intermediate value
theorem implies that the solution of equation 1.78b must lie between 2.323 75 and 2.323 85. But
every number between 2.323 75 and 2.323 85 rounds to 2.3238. In other words, x = 2.3238
is a solution of equation 1.78b correctly rounded to four decimal places.
In general, we can say that x is an approximation to a root of the equation f (x) = 0,
correctly rounded to k decimal places, if x has exactly k digits after the decimal, and
" # " #
10−k 10k
f x− f x+ < 0. (1.80)
2 2
Maximum Possible Error

We are often asked to find an approximation to the solution of an equation such as 1.78b, and
be sure that the error is less than some given value * , say * = 0.0001 or * = 0.000 000 1.
The smaller the value of * , the more accurate must be the approximation. To illustrate, suppose
an approximation to the solution of 1.78b is required with error less than * = 0.000 01. We
could verify that x = 2.323 762 277 has error less than * = 0.000 01, but there is little point
in carrying an approximation with nine decimal places when an error of 0.000 01 is concerned
with the fifth decimal place. We suspect that if we round the approximation to five decimal
places, the result x = 2.323 76 has error less than 0.000 01. To verify this, we evaluate
f (2.323 75) = −0.000 047 and f (2.323 77) = 0.000 029.
The fact that these values have opposite signs guarantees that the root is between 2.323 75
and 2.323 77. Since the difference between these numbers is 0.000 02 and our approximation
2.323 76 is halfway between them, it follows that the error in 2.323 76 must be less than 0.000 01.
In general, we can say that x is an approximation to a root of the equation f(x) = 0, with
error less than ! , if
f (x − *)f (x + *) < 0. (1.81)
There is the potential to use calculators or computers unwisely here. Avoid operating
calculators and computers at or near their limits. For example, suppose that the solution x =
2.323 762 277 of equation 1.78b constitutes the full display of a calculator. It would be unwise
to attempt to verify that this solution has error less than 10−9 using the same calculator. To do
so would require f (2.323 762 276) and f (2.323 762 278) . These values are very, very close
to zero. How could we be certain of their positivity and negativity when we are asking the
calculator to perform very sensitive calculations with numbers at the limits of its capabilities?
EXAMPLE 1.57
Find an approximation to the smallest root of the equation
2x 3 e−x + 5x 2 − 1 = 0
correctly rounded to six decimal places.
FIGURE 1.127 Graphical solution of 2x 3 e−x + 5x 2 − 1 = 0
−1.5 −1 −0.5 0.5 1 1.5 x
−1
−2
SOLUTION The graph of f (x) = 2x 3 e−x + 5x 2 − 1 in Figure 1.127 shows three solutions.
Our computer gives x = −0.766 051 059 as an approximation to the smallest root. To verify
that x = −0.766 051 is an approximation, correctly rounded to six decimal places, we calculate
f (−0.766 051 5) = −8.1 × 10−7 , f (−0.766 050 5) = 1.0 × 10−6 .
The fact that these values have opposite signs confirms the six-decimal-place accuracy of
x = −0.766 051.
EXAMPLE 1.58
Find an approximation to the largest root of the equation
x + 6 sin x = 0
with error less than 10−8 .
FIGURE 1.128 Graphical solutions of x + 6 sin x = 0
10
−15 −10 −5 5 10 x
−10
−20
SOLUTION The graph of f (x) = x + 6 sin x in Figure 1.128 shows five solutions. Our
computer yields x = 5.225 963 530 as an approximation to the largest root. For an approxima-
tion with error less than 10−8 , we take x = 5.225 963 53. For verification, we evaluate
f (5.225 963 52) = −4.1 × 10−8 and f (5.225 963 54) = 3.8 × 10−8 .
When we have an approximation to the solution of an equation correctly rounded to k decimal

places, we can say that we have an approximation with error no greater than 10−k /2 (compare
equations 1.80 and 1.81).
Knowing an approximation with maximum possible error, however, does not guarantee a
predictable number of correctly rounded decimal places. Let us illustrate. Suppose that we
have used our calculator or computer to approximate the root of an equation f (x) = 0 and the
result is 3.115 00. Suppose further that we know that the error is less than 10−5 . Can we give
an approximation correctly rounded to two decimal places? No! The fact that the error is less
than 10−5 allows us to say that the solution satisfies
3.115 00 − 10−5 < x < 3.115 00 + 10−5 *⇒ 3.114 99 < x < 3.115 01.
Since left and right sides of the latter inequality round to 3.11 and 3.12, we do not know the
approximation correctly rounded to two decimal places. However, we do know an approximation
correctly rounded to three and four decimal places, namely, 3.115 and 3.1150.
As a second example, suppose we know that an approximation to the root of f (x) = 0 is

3.435 with error less than 10−3 . How many decimal places can we guarantee? We can say that
the solution satisfies
3.435 − 0.001 < x < 3.435 + 0.001 *⇒ 3.434 < x < 3.436.
Since the numbers in the right inequality do not both round to the same two decimal places, we
can guarantee only one correctly rounded decimal, namely 3.4.
To emphasize this point again, we cannot make generalizations to the effect that a maximum
possible error of 10−k , say, guarantees any number of decimal places. In every specific example,
we will be able to determine how many decimal places are possible, but general statements
covering all situations are not possible.
EXAMPLE 1.59
Find points of intersection of the curves
y = x 3 − 3x 2 + 2x + 5, y = 6 − 5x 2 − 3x 4 .
Give coordinates correctly rounded to four decimal places.
FIGURE 1.129a Intersection points of two FIGURE 1.129b Intersection points of

curves with poor choice of x -range curves with better choice of x -range
y 6 y
100
50 4
−2
−6 −4 2 4 6 x 2
−50
−100 −2 −1 1 2 x
−2
−150
−200 −4
−250 −6
SOLUTION Graphs of the curves in Figure 1.129a indicate that whatever points of intersection
there are, they are in the interval −2 ≤ x ≤ 2. Plots on this interval in Figure 1.129b indicate
two points of intersection. To find x -coordinates of the points of intersection, we solve
x 3 − 3x 2 + 2x + 5 = 6 − 5x 2 − 3x 4 *⇒ 3x 4 + x 3 + 2x 2 + 2x − 1 = 0.
Solutions with four decimal places are −0.8924 and 0.3422. To confirm that all four decimal
places are correct, we calculate the following values of f (x) = 3x 4 + x 3 + 2x 2 + 2x − 1:
f (−0.892 45) = 3.1 × 10−4 , f (−0.892 35) = −4.6 × 10−4 ,
f (0.342 15) = −4.0 × 10−4 , f (0.342 25) = 2.1 × 10−5 .

Intersection points on the curves corresponding to x = −0.8924 and x = 0.3422 can be found
by substituting these values into the equations for the curves. If we substitute x = −0.8924
into y = x 3 − 3x 2 + 2x + 5, we obtain y = 0.115 379, whereas in y = 6 − 5x 2 − 3x 4 , we
get y = 0.115 459. These numbers do not agree to four decimal places. What this points out is
that the result of calculations with numbers accurate to four decimal places is unlikely to yield
numbers accurate to four decimal places; accuracy will be lost. How much depends on the nature
and the number of calculations. A similar situation arises with x = 0.3422; the equation of one
curve gives y = 5.373 17 and the other gives y = 5.373 36. They do not agree to four decimal
places. What we should have done is carry more decimal places in intermediate calculations.
For instance, we could carry six decimal places with x = −0.892 410 and x = 0.342 245.
Verification that x = −0.8924 and x = 0.3422 are correct to four decimal places is the same.
But using x = −0.892 410 and x = 0.342 245 leads to corresponding y -values that agree to
four decimal places no matter which equation y = x 3 − 3x 2 + 2x + 5 or y = 6 − 5x 2 − 3x 4
is used, namely y = 0.1153 and y = 5.3732.
EXAMPLE 1.60
At time t = 0, a 5 # resistor, a 2 H inductor, and a 0.01 F capacitor are connected with a
generator producing an alternating voltage of 10 sin 5t , where t ≥ 0 is time in seconds (Figure
1.130). The current i in the circuit thereafter is i(t) = f (t) + g(t) , where
9 0 √ 1 √ 0 √ 1:
18 5 31t 12 31 5 31t
f (t) = −e−5t/4 cos + sin
5 4 165 4
and
4 2
g(t) = cos 5t + sin 5t
5 5
FIGURE 1.130 Schematic for LCR-circuit
0.01 F
10 sin 5t 5Ω
S
2H
are called the transient and steady-state parts of the current, respectively. Plot graphs of f (t) ,
g(t) , and i(t) , and explain why the names for f (t) and g(t) are appropriate. Determine the
smallest time (correctly rounded to three decimal places) at which the magnitude of the current
in the circuit is 1 A.
SOLUTION Plots of f (t) and g(t) are shown in Figures 1.131a and b; their sum is plotted in
Figure 1.131c. The exponential factor e−5t/4 causes values of f (t) to approach zero within a
few seconds. In other words, f (t) is significant only for small t ; it dies off quickly, and therefore
√
the adjective transient is appropriate. The function g(t) is periodic with amplitude 2/ 5. It
remains for all time, and once f (t) becomes insignificant, current i(t) essentially becomes
g(t) . In other words, the current eventually settles down to g(t) , and hence the terminology
steady state for g(t) is appropriate.
FIGURE 1.131a Transient part of current FIGURE 1.131b Steady-state part of current
2 f (t) 0.75 g (t)

0.5
1
0.25
1 3 4 t −0.25 1 2 3 4 t
2
−0.5
−1
−0.75
FIGURE 1.131c Addition of transient and steady-state parts of current
i
2
1 2 3 4 t
−1
−2
Figure 1.131c makes it clear that there are nine times at which the magnitude of the current
is 1 A, four times when the current is positive and five times when it is negative. The smallest
is near t = 0.15 s. It satisfies the equation i(t) = −1, or, writing the equation in our standard
form 1.79, we have
9 0 √ 1 √ 0 √ 1:
−5t/4 18 5 31t 12 31 5 31t
0 = h(t) = −e cos + sin
5 4 165 4
4 2
+ cos 5t + sin 5t + 1.
5 5
When solving equations of this complexity for one of several roots, calculators and computers
usually require the equation and a reasonable approximation to the root, the closer the better.
Our computer returned the root t = 0.146 206 when given t = 0.15 as initial approximation.
To verify that t = 0.146 is correctly rounded to three decimal places, we calculate that
h(0.1455) = −0.012 and h(0.1465) = 0.0052.
EXERCISES 1.11
In Exercises 1–16 use a calculator or computer to find approximations Verify the accuracy of each root. Make a plot in order to determine the
to all roots of the equation accurate to six decimal places. Verify the number of roots.
accuracy of each root. Make a plot in order to determine the number
of roots. 17. x 3 − 5x − 1 = 0, 10−3
18. x 4 − x 3 + 2x 2 + 6x = 0, 10−4
1. x 2 + 3x + 1 = 0 2. x 2 − x − 4 = 0
x
3. x 3 + x − 3 = 0 4. x 3 − x 2 + x − 22 = 0 19. = x 2 + 2, 10−5
x+1
5. x 3 − 5x 2 − x + 4 = 0 6. x 5 + x − 1 = 0 20. (x + 1)2 = x 3 − 4x, 10−3
x+1 ∗ 21. (x + 1)2 = 5 sin 4x, 10−3
7. x 4 + 3x 2 − 7 = 0 8. = x2 + 1
x−2
∗ 22. cos2 x = x 2 − 1, 10−4
2
∗ 9. x − 10 sin x = 0 ∗ 10. sec x = ∗ 23. x + (ln x)2 = 0, 10−3
1 + x4
3x −4
∗ 24. e x
+ e = 4, 10
∗ 11. (x + 1)2 = sin 4x ∗ 12. (x + 1)2 = 5 sin 4x
∗ 13. x + 4 ln x = 0 ∗ 14. x ln x = 6
In Exercises 25–28 find all points of intersection for the curves accurate
∗ 15. ex + e−x = 10x ∗ 16. x 2 − 4e−2x = 0
to four decimal places.
25. y = x 3 , y =x+5
In Exercises 17–24 use a calculator or computer to find approximations
to all roots of the equation with error no greater than that specified. 26. y = (x + 1)2 , y = x 3 − 4x
Summary 99
27. y = x 4 − 20, y = x 3 − 2x 2 ∗ 32. Planck’s law for the energy density E of blackbody radiation at
x 2 1000◦ K states that
28. y = , y =x +2
x+1
kλ−5
∗ 29. When the beam in the figure below vibrates vertically, there are E = E(λ) = ,
ec/λ−1
certain frequencies of vibration, called natural frequencies. They are
solutions of the equation
where k > 0 is a constant and c = 0.000 143 86. This function is
ex − e−x shown in the figure below. The value of λ at which E is a maximum
tan x =
ex + e−x must satisfy the equation
divided by 20π . Find the two smallest frequencies correct to four
decimal places.
(5λ − c)ec/λ − 5λ = 0.
10 m
Find this value of λ correct to seven decimal places.
E Maximum E
∗ 30. A stone of mass 100 g is thrown vertically upward with speed

20 m/s. Air exerts a resistive force on the stone proportional to its
speed, and has magnitude 0.1 N when the speed of the stone is 10 m/s.
It can be shown that the height y above the projection point attained
by the stone is given by
. /
y = −98.1t + 1181 1 − e−t/10 m,
where t is time (measured in seconds with t = 0 at the instant of ∗ 33. Let f (x) be a continuous function with domain and range both
projection). equal to the interval [a, b]. Show that there is at least one value of x
(a) The time taken for the stone to return to its projection point in a ≤ x ≤ b for which f (x) = x .
can be obtained by setting y = 0 and solving the equation
for t . Do so (correct to two decimal places).
∗ 34. Use the zero intermediate value theorem to prove that at any given
(b) When air resistance is neglected, the formula for y is time there is a pair of points directly opposite each other on the equator
y = 20t − 4.905t 2 m. of the earth that have exactly the same temperature. Hint: Take the
equator to be the circle x 2 + y 2 = r 2 . Let f (x) be the temperature
What is the elapsed time in this case from the instant the
on the upper semicircle and g(x) be the temperature on the lower
stone is projected until it returns to the projection point?
semicircle. Consider the function F (x) = f (x) − g(−x) .
∗ 31. A uniform hydro cable P = 80 m long with mass per unit length
ρ = 0.5 kg/m is hung from two supports at the same level L = 70 m
apart (figure below). The tension T in the cable at its lowest point must ∗ 35. A marathoner runs the 26-odd miles from point A to point B
satisfy the equation starting at 7:00 a.m. Saturday morning. Starting at 7:00 a.m. Sunday
ρgP morning she runs the course again, but this time from point B to point
= eρgL/(2T ) − e−ρgL/(2T ) , A . Prove that there is a point on the course that she passed at exactly
T the same time on both days.
where g = 9.81. If we set z = ρg/(2T ) , then z must satisfy
2P z = eLz − e−Lz . ∗∗ 36. (a) Use the zero intermediate value theorem to prove that when
Solve this equation for z and hence find T correct to one decimal place. the domain of a continuous function is an interval, so also
is its range. Hint: Use the idea that a set S of points on the
70 m y -axis constitutes an interval if for any two points c and d
in S , the points c < y < d are all in S .
80 m
(b) If the domain is an open interval, is the range an open in-
terval?
SUMMARY
In this chapter we have reviewed basic concepts from algebra, analytic geometry, and trigonom-
etry, and introduced material that is essential to many of the applications of calculus. To find
real solutions of polynomial equations with integer coefficients, we use the rational root theorem
to narrow the field of possibilities and the factor theorem to remove roots from the equation as
they are found.
Analytic geometry is a combination of geometry and algebra. Algebraic equations are used
to describe geometric curves and curves are the geometric representation of equations. The form
of an equation dictates the shape of the curve and, conversely, the shape of a curve influences
its equation. To illustrate this fact, we discussed straight lines, circles, parabolas, ellipses, and
hyperbolas. The most common forms for equations of these curves are as follows:

 y − y1 = m(x − x1 ) Point-slope

 x − x1 y − y1


x −x = y −y
 Two-point

 2 1 2 1
Straight line y = mx + b Slope y -intercept

 y = m(x − a) Slope x -intercept

 x y



 + = 1 Two-intercept
a b
Ax + By + C = 0 General
8
y = ax 2 + bx + c Vertical axis of symmetry
Parabola
x = ay 2 + by + c Horizontal axis of symmetry
8
x 2 + y 2 + f x + gy + e = 0
Circle
(x − h)2 + (y − k)2 = r 2
 2
 x y2

 2 + 2 =1
a b
Ellipse

 (x − h) 2
(y − k)2
 + =1
a 2 b2
 2

 x y2

 − =1

 a2 b2



 2 2
y − x
 =1
 2
b a2
Hyperbola

 (x − h)2 (y − k)2

 − =1

 a2 b2




 (y − k)2
 (x − h)2
 − =1
b2 a2
Basic to all mathematics is the concept of a function, a rule that assigns to each number x
in a domain, a unique number y in the range. A function is simply another way of saying “a
quantity y depends on x .” The notation y = f (x) for a function immediately suggests that
a function can be represented geometrically by a curve — the curve with equation y = f (x)
— and we call this curve the graph of the function. Polynomials are functions of the form
an x n + · · · + a1 x + a0 , where n is a nonnegative integer and coefficients an , . . . , a0 are
constants. Rational functions are quotients of polynomials. Some functions are even, some are
odd, most are neither even nor odd, and only f (x) ≡ 0 is both even and odd.
Some of the most important functions in mathematics have inverses, in particular, the
exponential and logarithmic functions, the trigonometric
. functions,
/ and the hyperbolic functions.
A function f −1 (x) is the inverse of f (x) if f −1 f (x) = x for each x in the domain of
f (x) . The inverse function f −1 (x) reverses the action of f (x) . The graph of f −1 (x) is the
mirror image of the graph of f (x) in the line y = x . Increasing functions have inverses, as
do decreasing functions. The domain of a function that does not have an inverse can usually be
subdivided into subdomains on which the function does have inverses.
Trigonometric functions play a prominent role in many areas of applied mathematics. Par-
ticularly important are descriptions of oscillatory systems by the sine and cosine functions. The
sine, cosecant, tangent, and cotangent are odd functions, while cosine and secant are even. All are
Key Terms 101
periodic; sine, cosecant, cosine, and secant have period 2π , and tangent and cotangent have pe-
riod π . Trigonometric functions satisfy many identities. Recognizing when these identities can
be used to advantage to simplify expressions, or write them in alternative forms, is a huge asset.
The trigonometric functions do not have inverses, but their domains can be restricted so as to
create inverse functions; these domains turn out to be the principal values of the associated inverse
function. The inverse trigonometric functions reverse the roles of trigonometric functions. A
trigonometric function such as the sine function associates a value called sin x with a real number
(angle) x . The corresponding inverse function, Sin −1 x , regards x as the sine of an angle, and
yields the angle in the principal value range with sine equal to x .
Exponential and logarithmic functions are also important in applications; they are inverses
of each other. These functions are not periodic, nor are they even or odd. The exponential
function a x raises a to power x . As its inverse, the logarithm function loga x does the reverse.
It determines the power that a must be raised to produce x .
Hyperbolic functions are special combinations of exponential functions that arise suffi-
ciently often in applications to warrant special consideration. They satisfy identities very similar
to the trigonometric functions. Each hyperbolic function has an associated inverse hyperbolic
function.
In approximating solutions to equations, it is always necessary to indicate the accuracy of
the approximation. This can be done by correctly rounding the approximation to a specified
number of decimal places, or determining an approximation with maximum error. In both cases,
the zero intermediate value theorem is instrumental in verifying the accuracy.
KEY TERMS
In reviewing this chapter, you should be able to define or discuss the following key terms:
Polynomial Polynomial equation
Roots Solutions (zeros)
Linear equation Quadratic equation
Quadratic formula Discriminant
Irreducible real quadratic factor Multiplicity
Cartesian coordinates Quadrants
Length of a line segment Absolute values
Midpoint of a line segment Coordinates
Equation of a curve Slope of a line
Rise Run
Point-slope formula General equation of a line
Parallel lines Perpendicular lines
Parabola Symmetric about the x -axis and y -axis
Circle Ellipse
Hyperbola Asymptotes
Function Independent variable
Dependent variable Domain
Range Graph of a function
Absolute value function Even function
Odd function Even and odd parts of a function
Rational function Translation
One-to-one Inverse function
Strictly monotonic function Radian
Standard position of an angle Cosine law
Sine law Compound-angle formulas
Double-angle formulas Product formulas
Sum and difference formulas Periodic
Period General sine function
Amplitude Phase shift
Inclination of a line Algebraic function
Transcendental function Inverse trigonometric functions

Principal values Acute angle between two lines with slopes
Exponential functions Logarithm functions
Natural logarithms Hyperbolic functions
Intermediate value theorem Continuous function
Correctly rounded to k decimal places Zero intermediate value theorem
Approximation with maximum error Transient
Steady state
REVIEW
EXERCISES
In Exercises 1–4 find all real solutions of the polynomial equation, 31. 2x 2 + 20x + 38 = 3y 2 + 12y
giving multiplicities for any repeated roots.
1. x 3 − x 2 − 4 = 0 2. 2x 3 − 9x 2 + 27 = 0 32. 2y 2 − x = 3x 2 − y
3. 2x 4 − x 3 − 9x 2 + 13x − 5 = 0
∗ 4. 36x 4 + 12x 3 − 179x 2 − 30x + 225 = 0
In Exercises 33–34 find the distance from the point to the line.
In Exercises 5–6 find the distance between the points and also the mid- 33. (1, −3), y = 2x + 3 34. (−2, −5), x = 4 − 3y
point of the line segment joining the points.
5. (−1, 3), (4 , 2 ) 6. (2, 1), (−3, −4)
In Exercises 35–58 draw the curve. Then use a calculator or computer
In Exercises 7–10 find the equation for the line described. to plot the curve as a check.
7. Parallel to the line x − 2y = 4 and through the point (2, 3) 35. y = 2x 2 + 3 36. x 2 = 4 − y 2
8. Perpendicular to the line joining (−2, 1) to the origin and through
the midpoint of the line segment joining (1, 3) and (−1, 5) 37. y = x 3 − 1 38. |y| = |x|
∗ 9. Perpendicular to the line x =!4y − 11 and through the point of
intersection of this line and x = y 2 + 9 39. 4x 2 + y 2 = 0 40. x 2 + 3y 2 = 6
∗ 10. Joining the points of intersection of the curves y = x 2 and 5x =

41. 2y 2 − x 2 = 3 42. x 2 − 2x − y 2 + 4y = 1
6 − y2
43. y = sin 3x 44. y = cos (2x + π/2)
In Exercises 11–20 find the largest possible domain for the function.
√ √ 45. y = cos (2x − π/4) 46. y = 2 sin (3x + π/2)
11. f (x) = x2 + 5 12. f (x) = x2 − 5
√
1 x+4 ∗ 47. x 2 − 4y + 2 = 4x − 2y 2 ∗ 48. y = −x 2 + 4x + 4
13. f (x) = 14. f (x) = 3
x 2 + 3x + 2 x + 2x 2 + x
√
15. f (x) = (x 3 − 8)1/3 16. f (x) = x 3/2 ∗ 49. y = |x| + |x − 1| ∗ 50. y = |x − 1| − 1
√ 1
∗ 17. f (x) = x 2 + 4x − 6 ∗ 18. f (x) = √ ∗ 51. x = tan y ∗ 52. y = 2 ln (3x + 4)
2 x 2 + 4x − 5
$ $
2x + 1 1 ∗ 53. x = e−y ∗ 54. y = sin |x|
∗ 19. f (x) = +2 ∗ 20. f (x) = x−
x−3 x √
∗ 55. |y| = | sin x| ∗ 56. y = sin 2x
In Exercises 21–32 identify the curve as a straight line, parabola, circle,
∗ 57. y = sinh (2x − 1) ∗ 58. y = 4 tanh 3x
ellipse, hyperbola, or none of these.
21. x + 2y = 4 22. x = y 2 − 2y + 3
23. y = x 3 + 3 24. x 2 + 2y 2 = 4 In Exercises 59–60 find the angle between the lines.
2 2 2 2
25. y − x = x 26. x + y + 5 = 0
∗ 59. x + 2y = 4, y = 3x − 2
27. x 2 − 2x + y 2 = 16 28. x + y 2 = 3
29. x 2 + 2y 2 + y = 2x 30. x 2 − x + y 2 + y = 0 ∗ 60. x = 4y + 2, 2x + 3y = 5
In Exercises 61–64 give an example of a function y = f (x) with the

indicated properties.
61. The range of the function consists of one number only. Spring (k = 16)
∗ 62. The algebraic formula defining the function cannot be extended
beyond −1 ≤ x ≤ 2. M=1
∗ 63. The domain of the function consists of all reals except x = ±1.
x
64. The domain of the function is x ≤ 0 and the range is y ≥ 1. x=0
∗ 83. A lighthouse is 6 km offshore and a cabin on the straight shoreline

In Exercises 65–68 show that the function does not have an inverse. is 9 km from the point on the shore nearest the lighthouse (figure below).
Subdivide its domain of definition into subintervals on which the func-
tion has an inverse, and find the inverse function on each subinterval.
∗ 65. f (x) = x 2 − 4x + 3 ∗ 66. f (x) = x 4 − 8x 2 Lighthouse

'
x2 + 2 x2
∗ 67. f (x) = ∗ 68. f (x) =
x2 + 3 x+1 6
∗ 69. Express f (x) = cos 2x − sin 2x in the form f (x) =

A sin (2x + φ) . Use this to draw a graph of the function and find
the second smallest positive value of x for which f (x) = 0. Shoreline x
∗ 70. Is the function f (x) = 2 sin 2x − 3 cos 3x periodic? If so, what Cabin
is its period? 9
Show that if a man rows at 3 km/h and walks at 5 km/h, and he beaches
In Exercises 71–80 find all solutions of the equation. the boat at distance x from the near point on the shore, then the total
travel time from lighthouse to cabin is
∗ 71. cos2 x + 5 cos x − 6 = 0
√
∗ 72. 4 sin 2x = 1 x 2 + 36 9−x
2
t = f (x) = + , 0 ≤ x ≤ 9.
∗ 73. csc (x + 1) = 3 3 5
∗ 74. Tan −1 (3x + 2) = 5 − 2π Plot a graph of this function.
∗ 75. cos 2x = sin x
∗ 76. ln (sin x) + ln (1 + sin x) = ln 3 − ln 2
. / In Exercises 84–85 find all solutions of the equation correctly rounded
∗ 77. 3 Sin −1 ex+2 = 2
to three decimal places.
. /
∗ 78. 3 sin ex+2 = 2
84. x 3 − 2x 2 + 4x − 5 = 0
∗ 79. tan (x cosh 2) = 1/4
∗ 80. sinh x = 4 ∗ 85. x 2 − 1 = sin x
∗ 81. Draw graphs of the following functions:
. /
(a) f (x) = tan Tan −1 x
In Exercises 86–87 find all solutions of the equation with error less
(b) f (x) = Tan −1 (tan x)
than 10−4 .
∗ 82. When the mass in the figure below is pulled 5 cm to the right of
the position ( x = 0) it would occupy were the spring unstretched, and 86. x 3 + 12x 2 + 4x − 5 = 0
given speed 2 m/s to the right, its position thereafter is given by
∗ 87. x 2 − 1 = 24 sin x
1 1
x(t) = cos 4t + sin 4t.
20 2
∗ 88. Prove that the diagonals of a rhombus intersect at right angles.
Find when the mass passes through x = 0 for the first time. (A rhombus is a parallelogram with all sides of equal length.)
CHAPTER
2 Limits and Continuity
With a solid foundation of fundamentals in Chapter 1, you are well-prepared to study calculus.
We hope that you have been conscientious in your review. The better your algebraic skills and
the more familiar you are with analytic geometry and trigonometry, the easier calculus will be.
In Chapter 1, we placed tremendous emphasis on graphing. This was by design. The
most difficult part of the solution to many problems is frequently the initial step. Once started,
the solution often unfolds smoothly and easily, but that first step sometimes seems impossible.
One of the best ways to start a problem is with a diagram. A picture, no matter how rough, is
invaluable in giving you a “feeling” for what is going on. It displays the known facts surrounding
the problem; it permits you to “see” what the problem really is and how it relates to the known
facts; and it often suggests that all-important first step. We encourage you to develop the habit
of making diagrams at every opportunity — not just to solve problems, but to understand what
calculus is all about. We want you to see and feel calculus in all its aspects.
We introduce each remaining chapter of the book with an Application Preview, a problem
from one of the engineering disciplines, the solution of which requires material to be introduced
in that chapter. The solution of the problem is identified as the Application Review Revisited at
the appropriate place in the chapter. Here is the Application Preview for this chapter.
Application Preview The figure on the left below shows a complicated electrical network containing capacitors,
inductors, resistors, and a source of electric voltage E . Electrical engineers are interested in the
induced current in various parts of the network when the source is turned on and off very quickly.
E R1 R2 C1 E
4 E4(t)
3 E3(t)
L2 R3 L2
2 E2(t)
R4 R5
1 E1(t)
C2 R6
a a+1 t
Function E1 (t) in the figure on the right represents 1 V of potential being turned on at time
t = a , and turned off again one second later. This is not a very short period of time. Graph
E2 (t) in the same figure represents 2 V turned on for one-half of a second; E3 (t) is 3 V for
one-third of a second; and E4 (t) is 4 V for one-quarter of a second. Were we to continue this
process indefinitely, the source would apply ever-increasing voltages over ever-decreasing time
intervals, but the product of the voltage and the length of the time interval is always unity. The
ultimate result of this process is what is called an instantaneous application of 1 V to the circuit.
THE PROBLEM How do we represent the result of this process as a mathematical function,
and how do we operate with this function in equations? (For the answer, see Dirac-delta functions
in Section 2.5 on page 141.)
104
2.1 Limits 105
Can you see the problem? As a gets closer and closer to zero, the graph of E is zero
everywhere, except at t = a , where it becomes “infinite.” In what sense is this a function? Get
the feeling that this cannot be a function as we now understand functions.
The concept of a limit is crucial to calculus, for the two basic operations in calculus are
differentiation and integration, each of which is defined in terms of a limit. For this reason
you must have a clear understanding of limits from the beginning. In Sections 2.1–2.4 we give
an intuitive discussion of limits of functions; in Section 2.6 we show how these ideas can be
formalized mathematically.
2.1 Limits
The functions in Figure 2.1 all have value 3 at x = 1, but behaviours of the functions close to
x = 1 are totally different. As x gets closer and closer to 1 in Figure 2.1a, function values get
closer and closer to 3. In Figure 2.1b, function values get closer and closer to 3 if x approaches
1 through numbers smaller than 1, but they get closer and closer to 2 if x approaches 1 through
numbers larger than 1. Function values approach 2 as x gets closer and closer to 1 in Figure
2.1c whether x approaches 1 through numbers smaller than 1 or larger than 1. In this section
we emphasize the distinction between the value of a function at a point, and the values of the
function as we approach the point. We can see the distinction graphically; we now want to
express it algebraically.

The value of a function at x = 1 need not equal its limit as x approaches 1
y y y
3 3 3
1 x 1 x 1 x
The value of the function f (x) = x 2 − 4x + 5 at x = 2 is f (2) = 1. A completely

different consideration is contained in the question,“What number do values of f (x) = x 2 −
4x + 5 get closer and closer to as x gets closer and closer to 2?” Table 2.1 shows that as x gets
closer and closer to 2 , values of x 2 − 4x + 5 get closer and closer to 1 .
TABLE 2.1
x f (x) = x 2 − 4x + 5 x
1.9 1.01 2.1
1.99 1.000 1 2.01
1.999 1.000 001 2.001
1.9999 1.000 000 01 2.0001
106 Chapter 2 Limits and Continuity
FIGURE 2.2 Limit of x 2 − 4x + 5 is 1 as x approaches 2
y
y = x 2 − 4x + 5
2 x
Likewise, the graph of the function (Figure 2.2) clearly shows that values of f (x) approach
1 as x approaches 2. This statement is not precise enough for our purposes. For instance, the
graph also indicates that as x gets closer and closer to 2, values of f (x) get closer and closer to
0. They do not get very close to 0, but nonetheless, values of f (x) do get closer and closer to 0
as x gets closer and closer to 2. In fact, we can make this statement for any number less than 1.
To distinguish 1 from all numbers less than 1, we say that x 2 − 4x + 5 can be made arbitrarily
close to 1 by choosing x sufficiently close to 2. We can make values of x 2 − 4x + 5 within 0.1
of 1 by choosing values of x sufficiently close to 2; we can make values of x 2 − 4x + 5 within
0.01 of 1 by choosing values of x even closer to 2 ; we can make values of x 2 − 4x + 5 within
0.001 of 1 by choosing values of x yet even closer to 2, and so on. For numbers less than 1,
we cannot do this. For instance, it is not true that x 2 − 4x + 5 can be made arbitrarily close
to 0. The closest the function gets to 0 is 1 unit when x = 2. In calculus we say that the limit
of x 2 − 4x + 5 as x approaches 2 is 1 to represent the more lengthy statement “ x 2 − 4x + 5
can be made arbitrarily close to 1 by choosing x sufficiently close to 2.” In addition, we have a
notation to represent both statements:
lim (x 2 − 4x + 5) = 1. (2.1)
x→2
This notation is read “the limit of (the function) x 2 − 4x + 5 as x approaches 2 is equal to 1,”
and this stands for the statement “ x 2 − 4x + 5 can be made arbitrarily close to 1 by choosing
x sufficiently close to 2.”
We emphasize that the limit in 2.1 is not concerned with the value of x 2 − 4x + 5 at x = 2.
It is concerned with the number that x 2 − 4x + 5 approaches as x approaches 2. These numbers
are not always the same. For example, the limit as x approaches 1 of the function in Figure 2.1c
is 2, whereas the value of the function at x = 1 is 3.
Generally, we say that a function f (x) has limit L as x approaches a , and write
lim f (x) = L (2.2)

x→a
if f (x) can be made arbitrarily close to L by choosing x sufficiently close to a . Sometimes

it is more convenient to write f (x) → L as x → a to mean that f (x) approaches L as x
approaches a . This is especially so in the middle of a paragraph, as opposed to a displayed
equation such as 2.2.
The value of the function f (x) at x = a is irrelevant to the limit of f (x) as x approaches
a . In Figure 2.3a, they are the same; the value of the function at x = a is L , the same as the
limit as x → a . In Figure 2.3b, the value of the function f (a) at x = a is different from the
limit L as x → a . Finally, in Figure 2.3c, the function has no value at x = a , but the limit as
x → a is L .
2.1 Limits 107

Figures to illustrate that the value of a function at x = a may be different from its limit as x → a .
y y y
f(a)
L L L
a x a x a x
EXAMPLE 2.1
Evaluate lim (x 2 + 2x + 5) .
x→1
SOLUTION As x gets closer and closer to 1, values of x 2 + 2x + 5 get arbitrarily close to

8; therefore, we write
lim (x 2 + 2x + 5) = 8.
x→1
This is corroborated by the graph of the function in Figure 2.4.
FIGURE 2.4 Limit of x 2 + 2x + 5 is 8 as x approaches 1
8 y = x 2 + 2x + 5
−1 1 x
To calculate the limit of a function f (x) as x approaches a , we evaluate f (x) at values of x

that get closer and closer to a . For limits of complicated functions such as
x 2 (3 − x)
lim ,
x→5 x 3 + x
it would be tedious to evaluate x 2 (3 − x)/(x 3 + x) at many values of x approaching 5. The

following theorem provides a much easier method.
THEOREM 2.1
If limx→a f (x) = F and limx→a g(x) = G , then
(i) lim [f (x) + g(x)] = F + G. (2.3a)

x→a
(ii) lim [f (x) − g(x)] = F − G. (2.3b)

x→a
(iii) lim [cf (x)] = cF, when c is a constant. (2.3c)

x→a
(iv) lim [f (x)g(x)] = F G. (2.3d)

x→a
f (x) F
(v) lim = , provided that G #= 0. (2.3e)
x→a g(x) G
What this theorem says is that a limit such as limx→5 [x 2 (3 − x)/(x 3 + x)] can be broken
down into smaller problems and reassembled later. For instance, since
lim x 2 = 25, lim (3 − x) = −2, lim x 3 = 125, lim x = 5,

x→5 x→5 x→5 x→5
we may write
x 2 (3 − x) 25(−2) −50 5
lim 3
= = =− .
x→5 x + x 125 + 5 130 13
Although the results of Theorem 2.1 may seem evident, to prove them mathematically is not
a simple task. In fact, because we have not yet given a precise definition for limits, a proof
is impossible at this time. When we give definitions for limits in Section 2.6, it will then be
possible to prove the theorem (see Exercises 31–35 in Section 2.6).
EXAMPLE 2.2
x+2
Evaluate lim .
x→−2 x 2 + 9
SOLUTION Since limx→−2 (x + 2) = 0 and limx→−2 (x 2 + 9) = 13, part (v) of Theorem

2.1 gives
x+2 0
lim = = 0.
x→−2 x 2 + 9 13
EXAMPLE 2.3
x 2 (1 − x 3 )
Evaluate lim .
x→−1 2x 2 + x + 1
SOLUTION Using Theorem 2.1, we can write
x 2 (1 − x 3 ) (1)(2)
lim = = 1.
x→−1 2x 2 + x + 1 2 + (−1) + 1
Be sure that you understand how we obtained the expression
(1)(2)
.
2 + (−1) + 1
2.1 Limits 109
In particular, we did not set x = −1 in x 2 (1 − x 3 )/(2x 2 + x + 1) . Indeed, this is not permitted

because to evaluate a limit as x approaches −1, we are not to set x = −1; we are to let x
get closer and closer to −1. What we did do is take limits of x 2 , 1 − x 3 , 2x 2 , and x as x
approaches −1, and then use Theorem 2.1.
The following example illustrates what can happen if we substitute x = a into f (x) in the
evaluation of limx→a f (x) .
EXAMPLE 2.4
x2 − 9
Evaluate lim .
x→3 x − 3
SOLUTION Because limx→3 (x − 3) = 0, we cannot use Theorem 2.1. Nor can we set
x = 3 in (x 2 − 9)/(x − 3) because it is inherent in the limiting procedure that we do not
put x = 3. Besides, if we did, we would obtain the meaningless expression 0/0. Figure 2.5a
contains a typical graph of the function (x 2 − 9)/(x − 3) on the interval −3 ≤ x ≤ 6, using
a calculator or computer. It shows no anomaly in the behaviour of the function at x = 3. The
graph may, however, be accompanied by a message indicating that the function is undefined at
x = 3, as indeed it is. But the fact that the function is undefined at x = 3 does not concern
us here; we are interested in values of the function near x = 3, not at x = 3, and the graph
indicates that as x → 3, values of the function approach 6. To verify this algebraically, we
factor x 2 − 9 into (x − 3)(x + 3) and divide out a factor of x − 3 from numerator and
denominator:
x2 − 9 (x − 3)(x + 3)
lim = lim = lim (x + 3) = 6.
x→3 x − 3 x→3 x−3 x→3
Dividing by the factor x − 3 would not be permissible if x − 3 were equal to 0, that is, if x
were equal to 3. But once again this cannot happen, because in the limiting operation we let x
get closer and closer to 3, but do not set x = 3.
FIGURE 2.5a Computer FIGURE 2.5b Hand-

graph of (x 2 − 9)/(x − 3) drawn graph of (x 2 − 9)/(x − 3)
y y
x2 − 9
y=
8 x−3
6
6
x2 − 9 3
4 y= −3
x−3
3 x
2
−2 2 4 x
Note in this example that although the limit is 6, there is no value of x for which the function
(x 2 − 9)/(x − 3) is ever equal to 6. The graph of the function is a straight line with the point
at x = 3 removed. We have shown this with an open circle in Figure 2.5b.
EXAMPLE 2.5
−2x + 3x 2
Evaluate lim .
x→0 4x − x 2
SOLUTION The function is undefined at x = 0, but
−2x + 3x 2 x(−2 + 3x) 3x − 2 −2 1

lim 2
= lim = lim = =− .
x→0 4x − x x→0 x(4 − x) x→0 4 − x 4 2
Figure 2.6a shows a computer-generated graph of the function on the interval −2 ≤ x ≤ 2.

We have redrawn the graph in Figure 2.6b with a hole at x = 0.
FIGURE 2.6a Computer FIGURE 2.6b Redrawn

graph of (−2x + 3x 2 )/(4x − x 2 ) graph of (−2x + 3x 2 )/(4x − x 2 )
y y
2 2
−2x + 3x 2 −2x + 3x 2
y= y=
4x − x 2 1 4x − x 2 1
−2 −1 1 2 x −2 −1 1 2x
−1 −1
EXAMPLE 2.6
√
1+x−1
Evaluate lim √ .
x→0 x
SOLUTION Since the limit of the denominator as x approaches 0 is 0, we cannot immediately

use Theorem 2.1. The function is undefined for x ≤ 0; its graph on the interval 0.001 ≤ x ≤
1 in Figure 2.7 suggests, although not conclusively, that the limit is 0. To verify this we
rationalize the numerator,
√ that is, rid the numerator of the square root by multiplying numerator
and denominator by 1 + x + 1:
√ !√ √ "
1+x−1 1+x−1 1+x+1
lim √ = lim √ √
x→0 x x→0 x 1+x+1
x
= lim √ #√ $
x→0 x 1+x+1
√
x
= lim √
x→0 1+x+1
= 0.
2.1 Limits 111
√ √
FIGURE 2.7 Suggested limit of ( 1 + x − 1)/ x as x approaches 0
0.3
0.2
1+x−1
y=
0.1 x
0.2 0.4 0.6 0.8 x
The following is a second example of this type.
EXAMPLE 2.7
√ √
( x + 3 − 2)( x − 1 + 3)
Evaluate lim .
x→1 x−1
SOLUTION Once again we cannot use Theorem 2.1; the denominator has limit 0 as x → 1.
The graph of the function for 1.001 ≤ x ≤ 2 in Figure 2.8 may suggest a limit, but certainly
√ evidence is far from conclusive. Following the lead of Example 2.6, we rationalize
the √ the term
x + 3 − 2 in the numerator by multiplying numerator and denominator by x + 3 + 2,
√ √ √ √ √
( x + 3 − 2)( x − 1 + 3) ( x + 3 − 2)( x − 1 + 3)( x + 3 + 2)
lim = lim √
x→1 x−1 x→1 (x − 1)( x + 3 + 2)
√
(x + 3 − 4)( x − 1 + 3)
= lim √
x→1 (x − 1)( x + 3 + 2)
√
x−1+3
= lim √
x→1 x+3+2
3
= .
4
√ √
FIGURE 2.8 Suggested limit of ( x + 3 − 2)( x − 1 + 3)/(x − 1) as x approaches 1
y
0.8
0.6
( x + 3 − 2)( x − 1 + 3)
0.4 y=
x−1
0.2
0
1.2 1.4 1.6 1.8 x
EXAMPLE 2.8
sin 2x
Evaluate lim .
sin x
x→0
SOLUTION Once again we cannot immediately use Theorem 2.1 since the limit of the de-
nominator is zero. But using the double-angle formula sin 2x = 2 sin x cos x , we find that
sin 2x 2 sin x cos x
lim = lim = lim (2 cos x) = 2.
x→0 sin x x→0 sin x x→0
You may feel that we are overemphasizing limits in which both the numerator and denominator
are approaching zero (Examples 2.4–2.8). We stress this type of limit because when we use the
definition of a derivative in the next chapter, we always encounter this situation.
EXAMPLE 2.9
Do the functions
! " ! "
1 2 1
f (x) = sin and g(x) = x sin
x x
have limits as x approaches 0?
SOLUTION These limits are more difficult to find. To get a feeling for the behaviour of the
function sin (1/x) near x = 0, we plot its graph on the interval −0.01 ≤ x ≤ 0.01 (Figure
2.9a); it is a washout. The graph for −0.1 ≤ x ≤ 0.1 in Figure 2.9b is more instructive.
It shows that the function oscillates back and forth between ±1 more and more rapidly as x
approaches 0. As a result, the limit of sin(1/x) does not exist as x approaches 0.
FIGURE 2.9a sin(1/x) for −0.01 ≤ x ≤ 0.01 FIGURE 2.9b sin(1/x) for −0.1 ≤ x ≤ 0.1
y y
1 1
0.5 0.5
−0.01 −0.005 0.005 0.01 x −0.1 −0.05 0.05 0.1 x
−0.5 −0.5
−1 −1
The function g(x) = x 2 sin (1/x) has exactly the same number of oscillations as f (x)
= sin (1/x) , but the oscillations become smaller and smaller as x approaches 0 (Figure 2.10a).
In other words, limx→0 g(x) = 0.
FIGURE 2.10a Suggested limit of x 2 sin(1/x) as x approaches 0 FIGURE 2.10b x 2 sin(1/x) and ±x 2 for squeeze theorem
y y
y= x2 sin (1/x) 0.004 y = x2 sin (1/x)
0.004 y = x2
−0.1 0.01 x −0.1 0.01 x
−0.004 −0.004 y = −x2

2.1 Limits 113
We can confirm the geometric conclusion that lim x→0 g(x) = 0 in this example using the
following theorem.
THEOREM 2.2 (Squeeze or Sandwich Theorem)

Suppose that functions f (x), g(x) , and h(x) satisfy the following two properties:
1. f (x) ≤ g(x) ≤ h(x) for all x in some open interval containing x = a ;
2. lim f (x) = L = lim h(x) .
x→a x→a
Then lim g(x) = L also.

x→a
With Figure 2.11 we can see this result geometrically. The graph of g(x) is always between
FIGURE 2.11 Illustration
of the squeeze theorem
those of f (x) and h(x) in an open interval around x = a (condition 1). Since graphs of f (x)
and h(x) come together at x = a (condition 2), so also must the graph of g(x) .
y For g(x) = x 2 sin (1/x) in Example 2.9, we know that −1 ≤ sin (1/x) ≤ 1 for all x . If
y = h(x) we multiply all terms by x 2 , we obtain
L y = g(x)
! "
1
−x 2 ≤ x 2 sin ≤ x2.
y = f(x) x
a x This shows that the graph of x 2 sin (1/x) is between those of −x 2 and x 2 (Figure 2.10b). Since
limx→0 (−x 2 ) = limx→0 x 2 = 0, the squeeze theorem gives limx→0 x 2 sin (1/x) = 0.
EXAMPLE 2.10
! "
3
Use the squeeze theorem to evaluate lim x cos , if it exists.
x→0 x
SOLUTION The function cos (3/x) , like sin (1/x), oscillates violently as x approaches zero.
But we know that −1 ≤ cos (3/x) ≤ 1, and multiplication by |x| gives
! "
3
−|x| ≤ |x| cos ≤ |x|.
x
Since −|x| is always less than or equal to zero and |x| is always greater than or equal to zero,
we can write that ! "
3
−|x| ≤ x cos ≤ |x|.
x
Since limx→0 (−|x|) = limx→0 |x| = 0, the squeeze theorem requires that limx→0 x cos (3/x) =
0 also.
Here is a good question for you. Why in the last example did we multiply all parts of the
inequality −1 ≤ cos (3/x) ≤ 1 by |x| rather than x ?
One-Sided Limits
When we write L = limx→a f (x) , we mean that f (x) gets arbitrarily close to L as x gets
closer and closer to a . But how is x to approach a ? Does x approach a through numbers larger
than a , or does it approach a through numbers smaller than a ? Or does x jump back and forth
between numbers larger than a and numbers smaller than a , gradually getting closer and closer
to a ? We have not previously mentioned “mode” of approach simply because it would have
made no difference to our discussion. In each of the preceding examples, all possible modes of
approach lead to the same limit. In particular, Table 2.1 and Figure 2.2 illustrate that 1 is the
limit of f (x) = x 2 − 4x + 5 as x approaches 2 whether x approaches 2 through numbers
larger than 2 or through numbers smaller than 2.
Approaching a number a either through numbers larger than a or through numbers smaller
than a are two modes of approach that will be very important; therefore, we give them special
notations:
lim f (x) lim f (x)
x→a − x→a +
indicates that x approaches a indicates that x approaches a
through numbers smaller than a through numbers larger than a
(often called a left-hand limit (often called a right-hand limit
since x approaches a along since x approaches a along
the x -axis from the left of a ) the x -axis from the right of a )
Example 2.6 should, in fact, be designated a right-hand limit,

√
1+x−1
lim √ = 0,
x→0+ x
√
since the presence of x in the denominator demands that x be positive. Similarly, the limit
of Example 2.7 should only be right-handed.
Do not interpret the − and + in a − and a + as approaching a through negative and positive
numbers. This is the case only when a = 0. For instance, when a = 5, 5 is approached through
positive numbers whether it is approached from the left or from the right.
A natural question to ask is: What should we conclude if for a function f (x)
lim f (x) #= lim f (x)?

x→a + x→a −
Our entire discussion has suggested (and indeed it can be proved; see Exercise 20 in Section
2.6) that if a function has a limit as x approaches a , then it has only one such limit; that is, the
limit must be the same for every possible method of approach. Consequently, if we arrive at two
different results depending on the mode of approach, then we conclude that the function does
not have a limit. This situation is illustrated in Figure 2.1b and again in the following example.
EXAMPLE 2.11
|x|
Evaluate, if possible, lim .
x→0 x
SOLUTION If x < 0, then |x| = −x , and

FIGURE 2.12 Illustration
of no limit as x approaches 0
|x| −x
y lim = lim− = −1.
|x| x→0− x x→0 x
y=
x
1 If x > 0, then |x| = x , and
|x| x
x lim = lim+ = 1.
x→0+ x x→0 x
−1 The function has a right-hand limit and a left-hand limit at x = 0, but because they are not
the same, limx→0 (|x|/x) does not exist. The graph of f (x) = |x|/x in Figure 2.12 clearly
illustrates the situation.
2.1 Limits 115
EXAMPLE 2.12
The weight W of an object depends on its distance d from the centre of the earth. If d is less
than the radius R of the earth, then W is directly proportional to d ; and if d is greater than or
equal to R , then W is inversely proportional to d 2 . If the weight of the object on the earth’s
surface is W0 , find a formula for W as a function for d and draw its graph.
FIGURE 2.13 Weight of SOLUTION When d < R , we may write that W = kd ; and when d ≥ R , W = !/d 2 ,
an object as a function of distance where k and ! are constants of proportionality; that is,
from centre of earth %
kd, 0 ≤ d < R
W !
W =
W0 , R ≥ d.
d2
Since the weight of the object is W0 on the surface of the earth when d = R , it follows that
W0 = !/R 2 , from which ! = W0 R 2 . It now remains to find k . If we physically moved the
R d object from below the surface of the earth to the surface, it would slowly gain weight. Its weight
would approach W0 , that on the surface of the earth. In other words, the limit of W as r → R −
must be W0 ; that is,
W0
W0 = lim− kd = kR '⇒ k = .
r→R R
Thus,

Wd
 0 , 0<d <R

W = R .
2
 W0 R , d ≥ R

d2
A graph is shown in Figure 2.13.
EXERCISES 2.1
In Exercises 1–41 find the indicated limit, if it exists. x 3 − 6x 2 + 11x − 6 12x + 5
19. lim 20. lim
x→0 x 2 − 3x + 2 x→−1 x2 − 2x + 1
x2 − 5 x3 + 8
1. lim 2. lim * √
x→7 x + 2 x→−2 x + 5 2−x 1 − x2
2 2
21. lim 22. lim
x + 3x + 2 x + 3x x→1 2+x x→5 3x + 2
3. lim 4. lim
x→−5x 2 + 25 x→0 3x 2 − 2x
tan x sin x
2x − 3 2x − 4 23. lim 24. lim
5. lim 2 6. lim x→0 sin x x→π/4 tan x
x→3 x − 5
+ x→2− 3x + 2
x 4 + 5x 3 x 2 + 2x + 4 sin 4x sin 6x
7. lim 8. lim ∗ 25. lim ∗ 26. lim
x→0 3x 4 − x 3
− x→2 + x−3 x→0 sin 2x x→0+ sin 3x
x2 − 4 x2 − 9 sin 2x x−2
9. lim 10. lim ∗ 27. lim ∗ 28. lim √ √
x→2 x − 2 x→3+ x−3 x→0+ tan x x→2 x− 2
x 2 − 25 x 2 − 2x − 3
11. lim 12. lim √ √
x→5− x−5 x→3 3−x 1−x− 1+x |x 2 − 25|
∗ 29. lim ∗ 30. lim
x 2 − 4x + 4 x 3 − 6x 2 + 12x − 8 x→0 x x→5+ x 2 − 25
13. lim 14. lim
x→2 x−2 x→2 x 2 − 4x + 4
|x 2 − 25| |x 2 − 25|
x 3 − 6x 2 + 11x − 6 x 3 − 6x 2 + 11x − 6 ∗ 31. lim ∗ 32. lim
15. lim 16. lim x→5− x 2 − 25 x→5 x 2 − 25
x→1 x 2 − 3x + 2 x→2 x 2 − 3x + 2
√ √ √
x 3 − 6x 2 + 11x − 6 x 3 − 6x 2 + 11x − 6 x+2− 2 1− x2 + 1
17. lim 18. lim ∗ 33. lim √ ∗ 34. lim
x→3+ x 2 − 3x + 2 x→3− x 2 − 3x + 2 x→0+ x x→0 2x 2
√ √ √ √
x+2 1+x− 1−x x+h− x
∗ 35. lim √ √ ∗ 36. lim ∗ 56. Evaluate lim .
x→−2 −x − 2 x→0 x h→0 h
√ √
x + 3 − −x − 1 1 − e−1/x
∗ 37. lim √ 57. Plot a graph of the function f (x) = . What does the
x→−2+ x+2 1 + e−1/x
x plot suggest for right-hand and left-hand limits of f (x) as x approaches
∗ 38. lim √
x→0 x+4−2 0? What is f (0) ?
√ √
1+x− 1−x ∗ 58. At the present time it is impossible for us to calculate algebraically
∗ 39. lim √ √
x→0 2+x− 2−x sin x − x
√ √ lim . What can we do?
x + 1 − 2x + 1 x→0 x3
∗ 40. lim √ √
x→0 3x + 4 − 2x + 4 (a) One suggestion might be to use a calculator or computer to
√
x+3−2 evaluate the function (sin x − x)/x 3 for various values of
∗ 41. lim x that approach 0. Try this with x = 10−n , n = 1, . . . , 7,
x→1 x−1
and make any conclusion that you feel is justified. How
would you feel about the use of the calculator if you knew
In Exercises 42–49 assume that a > 0 is a constant and calculate the that the value of the limit were −1/6?
limit, if it exists.
(b) Another suggestion might be to plot the function on smaller
x 2 − a2 and smaller intervals around x = 0. Do this on the intervals
42. lim
x→a x − a −0.1 ≤ x ≤ 0.1, −0.01 ≤ x ≤ 0.01, −0.001 ≤ x ≤
x 3 − a3 0.001, and −0.0001 ≤ x ≤ 0.0001. What happens?
43. lim
x→a x − a f (x) − g(x)
x+a sin 2ax ∗ 59. Assume the existence of lim . If limx→a f (x) =
44. lim 2 ∗ 45. lim x→a x−a
x→−a x + ax − x − a x→0 sin ax L , find limx→a g(x) .
√ √ √ √
x− a x+a− a
∗ 46. lim ∗ 47. lim √ ∗ 60. If f (x) is an even function and limx→a f (x) = L , find, if pos-
x→a x−a x→0+ x sible, limx→−a f (x) .
√ √ √ √
a+x− a−x x 2 + a 2 − 2x 2 + a 2
∗ 48. lim ∗ 49. lim √ √ ∗ 61. If f (x) is an even function and limx→a + f (x) = L , find, if
x→0 x x→0 3x 2 + 4 − 2x 2 + 4
possible, limx→−a − f (x) .
∗ 50. Plot a graph of f (x) = (1/x) sin x on the interval −π ≤ x ≤ π .
What does it suggest for the limit of the function as x → 0? (This will ∗ 62. If f (x) is an even function and limx→a + f (x) = L , find, if
be confirmed in Section 3.9.) possible, limx→−a + f (x) .
∗ 63.–65. Repeat Exercises 60–62 for an odd function f (x) .

In Exercises 51 and 52 use the squeeze theorem to discuss the limits.
! " ! " ∗ 66. If limx→0 f (x) = F , what do you conclude about limx→0 f (x)
1 3
51. lim x sin 52. lim x 4 cos sin(1/x) ? Hint: See Example 2.9.
x→0 x x→0 x
∗ 53. Does the floor function *x+ of Exercise 68 in Section 1.5 have a ∗ 67. If a , b , c , and d are constants, find the following limits, if they
limit, a right-hand limit, or a left-hand limit as x approaches integer exist:
values?
a + ce−1/x
∗ 54. Prove or disprove the following statement: If f (x) < g(x) for (a) lim
all x #= a , then lim f (x) < lim g(x). x→0+ b + de−1/x
x→a x→a
(x + h)n − x n a + ce−1/x
∗ 55. If n is a positive integer, evaluate lim . Hint: Use (b) lim
h→0 h x→0− b + de−1/x
either the binomial theorem or the result that
a + ce−1/x
n n n−1 n−2 n−2 n−1 (c) lim
a − b = (a − b)(a +a b + · · · + ab +b ). x→0 b + de−1/x
2.2 Infinite Limits

Functions do not always have limits. Sometimes this is due to the fact that right-hand and left-
hand limits are not identical; sometimes it is a result of erratic oscillations. Examples of both of
these situations were discussed in Section 2.1 (see Examples 2.9–2.11). Nonexistence of a limit
2.2 Infinite Limits 117
may also be due to excessively large values of the function. For instance, consider the function
f (x) = 1/(x − 2)2 , which is not defined at x = 2. Does it have a limit as x approaches 2?
Table 2.2 indicates that as x approaches 2, values of 1/(x − 2)2 become very large; in fact,
values of the function can be made arbitrarily large by choosing x sufficiently close to 2. Thus,
the function does not have a limit as x approaches 2. We express this symbolically in the form
1
lim = ∞.
x→2 (x − 2)2
The symbol ∞ represents what mathematicians call infinity. Infinity is not a number; it is
simply a symbol that we find convenient to represent various ideas. In the equation above it
states that the limit does not exist , and indicates that the reason it does not exist is that values of
the function become arbitrarily large as x approaches 2.
The graph of f (x) in Figure 2.14 further illustrates this point. We say that the line x = 2
is a vertical asymptote for the graph.
TABLE 2.2 FIGURE 2.14 Unbounded behaviour of 1/(x − 2)2 near x = 2
y
x 1/(x − 2)2 x 1
y=
1.9 102 2.1 (x − 2)2
1.99 104 2.01
1.999 106 2.001
1.9999 108 2.0001 2 x
1.99999 1010 2.00001
EXAMPLE 2.13
1
Evaluate lim , if it exists.
FIGURE 2.15 Unbounded x→1 x−1
behaviour of 1/(x − 1) near x = 1
SOLUTION In this example we consider right- and left-hand limits as x approaches 1. We
y find that
1
y= 1 1
x−1 lim =∞ and lim = −∞,
x→1 x − 1
+ x→1 x − 1
−
the latter meaning that as x approaches 1 from the left, the function takes on arbitrarily “large”
negative values. Either one of these expressions is sufficient to conclude that the function
1 x 1/(x − 1) does not have a limit as x approaches 1. The function is shown in Figure 2.15; the
line x = 1 is a vertical asymptote.
EXAMPLE 2.14
Discuss left- and right-hand limits of the function
x2 − 9
f (x) =
x2 + x − 2
as x → 1 and x → −2.
FIGURE 2.16 Unbounded behaviour of (x 2 − 9)/(x 2 + x − 2) near x = −2 and x = 1
y
60
x2 − 9
y=
40 x2 +x−2
20
−3 −2 −1 1 2 3 x
−20
−40
−60
SOLUTION The function is undefined at x = 1 and x = −2. (Why?) Its graph is shown
in Figure 2.16. To confirm algebraically what we see geometrically, it is advantageous to factor
numerator and denominator of f (x) as much as possible,
(x + 3)(x − 3)
f (x) = .
(x + 2)(x − 1)
Consider now the (right-hand) limit as x → 1+ . Each of the four factors has a limit as x → 1+ :
x + 3 → 4, x − 3 → −2 , x + 2 → 3, x − 1 → 0.
Were we to combine them according to Theorem 2.1, we would write
4(−2)
.
3(0)
In spite of the fact that this is not correct — part (v) of the theorem does not allow a 0 in the
denominator — this expression lets us see what is happening to f (x) as x → 1+ . It states
that as x → 1+ , the numerator of f (x) approaches −8 and the denominator approaches 0.
But this means that f (x) must be taking on larger and larger values as x → 1+ . Are these
values positive or negative? The numerator is clearly negative, but the sign of the denominator
depends on whether the 0 is approached through positive or negative numbers. Recalling that
0 arose from the fact that x − 1 → 0 as x → 1+ , we can be more specific; x − 1 must
approach 0 through positive numbers since x > 1 for x → 1+ . We indicate this by writing
x − 1 → 0+ as x → 1+ . The fraction displayed above is therefore replaced by
4(−2)
,
3(0+ )
and it is clearly negative. In other words,
x2 − 9 (x + 3)(x − 3)
lim 2
= lim+ = −∞,
x→1+ x +x−2 x→1 (x + 2)(x − 1)
as indeed Figure 2.16 indicates. Similarly, as x → 1− , the fraction
4(−2)
,
3(0− )
which shows limits of the four factors of f (x) , indicates that
(x + 3)(x − 3)
lim = ∞.
x→1− (x + 2)(x − 1)
In the following, the fractions on the right yield results as x → −2+ and x → −2− :
(x + 3)(x − 3) (1)(−5)
lim = ∞,
x→−2+ (x + 2)(x − 1) (0+ )(−3)
(x + 3)(x − 3) (1)(−5)
lim = −∞.
x→−2− (x + 2)(x − 1) (0− )(−3)
In the last limit of this example, we did not write
(x + 3)(x − 3) (1)(−5)
lim = − = ∞.
(x + 2)(x − 1)
x→2− (0 )(−3)
(1)(−5)
To include the fraction − as part of a mathematical equation is not acceptable; division
(0 )(−3)
by 0 is impossible. Place the fraction to the right of the limit to aid in its evaluation, but do not
include it in the equation.
EXERCISES 2.2
1
In Exercises 1–24 evaluate the limit, if it exists. ∗ 21. lim ln (4x) ∗ 22. lim
x→0+ x→1 ln |x − 1|
1 1 1/x
1. lim 2. lim ∗ 23. lim e ∗ 24. lim e1/|x|
x→2+ x−2 x→2− x−2 x→0 x→0
1 1
3. lim 4. lim
x→2 x−2 x→2+ (x − 2)2 In Exercises 25–28 assume that a > 0 is a constant and calculate the
limit, if it exists.
1 1
5. lim 6. lim
x→2− (x − 2)2 x→2 (x − 2)2 x−a |x − a|
∗ 25. lim ∗ 26. lim
5x 6x 2 + 7x − 5
x 2 − 2ax + a 2
x→a + x→a x 2 − 2ax + a 2
7. lim 8. lim √ √
x→1 (x − 1)3 x→1/2 2x − 1 a+x− a
∗ 27. lim ∗ 28. lim e1/(|x|−a)
2x + 3 x−2 x→0− x2 x→−a
9. lim 10. lim
x→1 x 2 − 2x + 1 x→2 x 2 − 4x + 4 ∗ 29. It is not clear whether the limit lim x 2 ln x exists due to the fact
x→0+
∗ 11. lim csc x 12. lim sec (x − π/4) that limx→0+ x 2 = 0 and limx→0+ ln x = −∞ . It depends on which
x→0 x→π/4
term is more dominant in the product, x 2 or ln x .
∗ 13. lim sec (x − π/4) 14. lim cot x (a) Calculate x 2 , ln x , and x 2 ln x for x = 10−n , n =
x→3π/4 x→0+
1, . . . , 10, and use this information to decide on a value
∗ 15. lim tan x ∗ 16. lim tan x for the limit.
x→π/2+ x→π/2−
√ (b) Plot graphs of x 2 ln x near x = 0 to confirm your calcula-
x 2 − 2x + 1 1+x−1 tion in part (a).
∗ 17. lim 3 ∗ 18. lim
x→1 x − 3x 2 + 3x − 1 x2
x→0
∗ 30. Repeat Exercise 29 for the limit lim x 10 e1/x , but pick your own
x→0+
2x |4 − x|
∗ 19. lim √ ∗ 20. lim 2 values of x at which to evaluate x 10 and e1/x .
x→0 1− x2 + 1 x→4 x − 8x + 16
2.3 Limits at Infinity

In many applications we are concerned with the behaviour of a function as its independent
variable takes on very large values, positively or negatively. For instance, consider finding, if
possible, a number that the function f (x) = (2x 2 + 3)/(x 2 + 4) gets closer and closer to as x
becomes larger and larger and larger. The graph of f (x) on the interval 0 ≤ x ≤ 100 in Figure
2.17 suggests that function values are approaching 2 for large x . To confirm this algebraically,
we divide the numerator and denominator of f (x) by x 2 ,
3
2x 2 + 3 2+
f (x) = = x2 .
x2 +4 4
1+
x2
For very large x , the terms 3/x 2 and 4/x 2 are very close to zero, and therefore f (x) is
approximately equal to 2. Indeed, f (x) can be made arbitrarily close to 2 by choosing x
sufficiently large. In calculus we express this fact by saying that the limit of (2x 2 + 3)/(x 2 + 4)
as x approaches infinity is 2, and we write
2x 2 + 3
lim = 2.
x→∞ x2 + 4
Once again we stress that ∞ is not a number. The notation x → ∞ simply means “as x gets
larger and larger and larger.” We say that the line y = 2 is a horizontal asymptote for the
graph of the function f (x) = (2x 2 + 3)/(x 2 + 4) .
We can also find limits of functions as x takes on arbitrarily large negative numbers, denoted
by x → −∞ .
FIGURE 2.17 Limit of (2x 2 + 3)/(x 2 + 4) for large x
y
2
1.8
1.6
1.4 2x 2 + 3
y=
x2 + 4
1.2
20 40 60 80 100 x
0.8
We can also find limits of functions as x takes on arbitrarily “large” negative numbers,
denoted by x → −∞ .
EXAMPLE 2.15
5x 4 − 3x + 5
Evaluate lim , if it exists.
x→−∞ x 4 − 2x 2 + 5
SOLUTION Division of numerator and denominator by x 4 leads to
3 5
5− +
lim x3 x 4 = 5.
x→−∞ 2 5
1− +
x2 x4
FIGURE 2.18 Limits of (5x 4 − 3x + 5)/(x 4 − 2x 2 + 5) for large positive and negative x
y 5x 4 − 3x + 5
y=
x 4 − 2x 2 + 5
5
−4 −2 2 4 x
If the limit in Example 2.15 is taken as x → ∞ , the same result is obtained. The graph of
this function is shown in Figure 2.18. The line y = 5 is a horizontal asymptote.
In general, we say that a line y = L is a horizontal asymptote for the graph of a function
f (x) if either, or both, of the following situations exist:
lim f (x) = L or lim f (x) = L.
x→−∞ x→∞
In Example 2.15, both of these conditions are satisfied for L = 5.
EXAMPLE 2.16
Evaluate the following limits, if they exist:
2x 3 − 4 2x 2 − 14 2x 4 − 14
(a) lim (b) lim (c) lim
x→∞ x3 + x2 + 2 x→−∞ 3x 3 + 5x x→∞ 3x 3 + 5x
SOLUTION
(a) To obtain this limit we divide numerator and denominator of the fraction by x 3 :
4
2x 3 − 4 2−
lim = lim x3 = 2.
x→∞ x3 + x2 +2 x→∞ 1 2
1+ +
x x3
(b) Once again we divide numerator and denominator by x 3 :
2 14
2x 2 − 14 −
lim = lim x x3 .
x→−∞ 3x 3 + 5x x→−∞ 5
3+
x2
Since the numerator approaches 0 and the denominator approaches 3, we conclude
that
2x 2 − 14
lim = 0.
x→−∞ 3x 3 + 5x
We could also have obtained this limit by dividing numerator and denominator by x 2
instead of x 3 ,
14
2x 2 − 14 2−
lim = lim x2 .
x→−∞ 3x 3 + 5x x→−∞ 5
3x +
x
Now the numerator approaches 2, but since the denominator becomes very large, the
fraction once again approaches zero.
(c) Division by x 3 in this case gives
14
2x 4 − 14 2x −
lim = lim x3 .
x→∞ 3x 3 + 5x x→∞ 5
3+
x2
Since the numerator becomes arbitrarily large as x → ∞ and the denominator
approaches 3, it follows that
2x 4 − 14
lim = ∞.
x→∞ 3x 3 + 5x
EXAMPLE 2.17
Draw a graph of the function
1
f (x) = sin x, x ≥ π.
x
Does it have a horizontal asymptote?
SOLUTION We draw the graph in Figure 2.19 by making the oscillations of sin x become
smaller and smaller as x gets larger and larger. The graph indicates that the positive x -axis is a
horizontal asymptote. This is confirmed by the fact that
1
lim sin x = 0.
x→∞ x
We could have reasoned this out as follows. The function sin x does not have a limit as x → ∞ ,
but as x → ∞ its values oscillate back and forth between ±1. Since these values are multiplied
by 1/x , which is getting smaller and smaller, the product (1/x) sin x must be getting closer
and closer to 0. We could also use the squeeze theorem to arrive at the same limit.
Notice that the graph actually crosses the asymptote an infinite number of times.
FIGURE 2.19 Limit of x −1 sin x for large x
y
1
y= sin x
x
Figures 2.17, 2.18, and 2.19 indicate that the graph of a function y = f (x) can approach
an asymptote y = L in three ways: from above (Figure 2.18), from below (Figure 2.17),
and oscillating about the asymptote, gradually getting closer and closer to it (Figure 2.19). A
computer-generated graph may or may not always make it clear which situation prevails.
EXAMPLE 2.18
Plot a graph of the function f (x) = (3x − 6)/(x 2 + 5) . Indicate any horizontal asymptotes
and determine how the graph approaches these asymptotes.
FIGURE 2.20 Horizontal asymptote of (3x − 6)/(x 2 + 5)
y
0.25
−100 −50 50 100 x

−0.25
−0.5
−0.75
3x − 6
−1 y=
x2 + 5
−1.25
−1.5
SOLUTION We begin by plotting a graph of the function on the domain −100 ≤ x ≤ 100
(Figure 2.20). It suggests that y = 0 is a horizontal asymptote, and that the graph approaches
y = 0 from above when x → ∞ and from below when x → −∞ . We can confirm this
algebraically in various ways. To verify the situation as x → ∞ , we calculate
6
3x − 6 3−
lim = lim x = 0.
x→∞ x2 + 5 x→∞ 5
x+
x
This confirms that y = 0 is indeed a horizontal asymptote as x → ∞ . To decide how the
asymptote is approached, we note that for large positive x , both 3 − 6/x and x + 5/x are
positive, and therefore f (x) must approach 0 through positive numbers. We indicate this by
writing
3x − 6
lim 2 = 0+ .
x→∞ x + 5
We also could have reasoned as follows. The graph of the function crosses the asymptote y = 0
when
3x − 6
= 0,
x2 + 5
and the only solution of this equation is x = 2. Combine this with the graph in Figure 2.20,
and we conclude that f (x) > 0 for all x > 2, and hence the graph must approach y = 0 from
above as x → ∞ .
Similar reasoning shows that the graph approaches y = 0 from below as x → −∞ .
In the following example, the function is more complicated. It discusses an alternative way
for determining how graphs of rational functions with horizontal asymptotes approach these
asymptotes.
EXAMPLE 2.19
Find vertical and horizontal asymptotes for the graph of the function f (x) = (x 2 − 16)/
(x 2 + x − 6) . Determine how the graph approaches horizontal asymptotes.
SOLUTION The plot on the interval −10 ≤ x ≤ 10 in Figure 2.21a suggests vertical
asymptotes at x = 2 and x = −3. These are confirmed with the following limits:
(x + 4)(x − 4) 6(−2)
lim = −∞,
x→2+ (x + 3)(x − 2) (5)(0+ )
(x + 4)(x − 4) 6(−2)
lim = ∞,
x→2− (x + 3)(x − 2) (5)(0− )
(x + 4)(x − 4) (1)(−7)
lim = ∞,
x→−3+ (x + 3)(x − 2) (0+ )(−5)
(x + 4)(x − 4) (1)(−7)
lim = −∞.
x→−3− (x + 3)(x − 2) (0− )(−5)
The plot in Figure 2.21b suggests that y = 1 is a horizontal asymptote as x → ∞ . To
verify this, we calculate
16
x 2 − 16 1−
lim 2 = lim x2 = 1.
x→∞ x + x − 6 x→∞ 1 6
1+ −
x x2
FIGURE 2.21a Vertical asymptotes of FIGURE 2.21b Suggested horizontal

(x 2 − 16)/(x 2 + x − 6) asymptote of (x 2 − 16)/(x 2 + x − 6) as x → ∞
y y
1.0
15
x 2 − 16 0.8
10 y=
x2 + x − 6
5 0.6
x 2 − 16
y=
−10 −5 5 10 x 0.4 x2 + x − 6
−5
0.2
−10
−15 0 20 40 60 80 100 x
FIGURE 2.21c Suggested horizontal asymptote of (x 2 − 16)/(x 2 + x − 6) as x → −∞
1.0
0.8
x 2 − 16 0.6
y=
x2 + x − 6
0.4
0.2
−30 −25 −20 −15 −10 −5 0 x
To show whether the graph approaches this horizontal asymptote from above or below, we use
long division to express f (x) in the form
x 2 − 16 x + 10
f (x) = 2
= 1− 2 .
x +x−6 x +x−6
Because (x + 10)/(x 2 + x − 6) is positive for large x , it follows that f (x) is less than 1 for
large x , and the graph approaches y = 1 from below as x → ∞ .
The plot in Figure 2.21c indicates that y = 1 is also a horizontal asymptote as x → −∞ .

Since (x + 10)/(x 2 + x − 6) is negative for “large” negative x , it follows that f (x) is greater
than 1 for such x , and the graph approaches y = 1 from above.
Previous graphs in this section have had at most one horizontal asymptote. The following
example has two.
EXAMPLE 2.20
√
2x 2 + 4
Evaluate lim , if it exists. What is the limit as x → −∞ ?
x→∞ x+5
SOLUTION When we divide numerator and denominator by x , and take the x inside the
square root as x 2 ,
* *
√ 1+ 2 2x 2 + 4 4
2x + 4 2+ √
2x 2 +4 x2 x2
lim = lim x = lim = lim = 2.
x→∞ x+5 x→∞ 1 x→∞ 5 x→∞ 5
(x + 5) 1+ 1+
x x x
In evaluating the limit as x → −∞ , we must be extra careful:
√
√ 2x 2 + 4
2x 2 +4 x
lim = lim .
x→−∞ x+5 x→−∞ 5
1+
x
It is not correct in this case to take x inside the square root as
√ *
2x 2 + 4 2x 2 + 4
=
x x2
since for negative x , the expression √

on the left is negative and that on the right is positive. In
this case, we should replace x by − x 2 , and write
√ √ *
2x 2 + 4 2x 2 + 4 2x 2 + 4
= √ =− .
x − x2 x2
Hence,
* *
√ 2x 2 + 4 4
− − 2+ √
2x 2 + 4 x2 x2
lim = lim = lim = − 2.
x→−∞ x+5 x→−∞ 5 x→−∞ 5
1+ 1+
x x
The graph of
√this function (Figure
√ 2.22) confirms these limits; there are two horizontal asymp-
totes, y = 2 and y = − 2.
√
FIGURE 2.22 y= 2x 2 + 4/(x + 5) illustrating two horizontal asymptotes
y
2x 2 + 4
10 y=
x+5
5
−30 −20 −10 10 20 x

−5
−10
−15
Graphs of functions have horizontal asymptotes if either, or both, of the conditions limx→±∞
f (x) = L hold. What this means is that for large positive or negative values of x , the function is
approximately equal to L . The larger the value of x , the better the approximation. Hyperbolas
also have asymptotes, but they are not usually horizontal. For example, asymptotes of the
hyperbola x 2 − y 2 /9 = 1 in Figure 2.23 are y = ±3x . These are often called oblique (or
slanted) asymptotes.
y2 FIGURE 2.24 Vertical and oblique asymptotes of a rational function

FIGURE 2.23 Oblique asymptotes of hyperbola x 2 − =1
9 y
y
30
y = −3x 15 y = 3x 2x 3 + 4x 2 − x + 1
20 y=
10 2 − x − x2
10
5
−10 −5 5 10 x
−4 −2 2 4 x −10
−5 −20
−10 −30
−15
Graphs of functions can also have oblique asymptotes. In Figure 1.35, y = x − 3 is an

oblique asymptote for the graph of f (x) = (x 3 − 3x 2 + 1)/(x 2 + 1) . We can confirm this
algebraically if we divide x 2 + 1 into x 3 − 3x 2 + 1. The result is
x 3 − 3x 2 + 1 4−x
f (x) = 2
= x−3+ 2 .
x +1 x +1
As x → ±∞ , the term (4 − x)/(x 2 + 1) → 0, and therefore f (x) → x − 3. For large

values of x , the function f (x) can be approximated more and more closely by x − 3. This
representation for f (x) can be used to show that the depiction of the asymptote in Figure 1.35
is incorrect. For large positive x , the term (4 − x)/(x 2 + 1) < 0, so that f (x) < x − 3 and
the graph should approach the asymptote from below, not above.
Rational functions P (x)/Q(x) always have oblique asymptotes when the degree of poly-
nomial P (x) is exactly one more than the degree of polynomial Q(x) . A second example is
the function f (x) = (2x 3 + 4x 2 − x + 1)/(2 − x − x 2 ) in Figure 2.24. It has an oblique
asymptote y = −2x − 2 identified from
2 x 3 + 4x 2 − x + 1 x+5
f (x) = = −2 x − 2 + .
2−x− x2 2 − x − x2
It also has two vertical asymptotes, one at x = −2 and the other at x = 1.
Functions other than rational functions can also have oblique asymptotes. The graph of
f (x) = x + 2 + + 5e−x in Figure 2.25 is asymptotic to y = x + 2. The straightness of the
graph of f (x) = (2x 4 + 4x 2 − x + 1)/(x 2 + x − 2) for large positive x in Figure 2.26
√
suggests an asymptote. It is 2(x − 1/2) , but to show this algebraically is difficult.
FIGURE 2.25 Oblique FIGURE 2.26 Oblique

asymptote of x + 2 + 5e−x asymptote of a root function
y y
14 2x 4 + 4x 2 − x + 1
25 y=
12 x2 + x − 2
20
10
8 15
6 10
4 y = x + 2 + 5e−x
5
2
0 2 4 6 8 10 x 5 10 15 20 x
EXERCISES 2.3
1 1
In Exercises 1–38 evaluate the limit, if it exists. ∗ 21. lim cos x ∗ 22. lim cos x
x→∞ 2x x→−∞ 2x
x+1 1−x
1. lim 2. lim sin 4x sin2 x
x→∞ 2x − 1 x→∞ 3 + 2x ∗ 23. lim ∗ 24. lim
x→∞ x2 x→∞ x
x2 + 1 1 − 4x 3
3. lim 4. lim 1
x→∞ 2x 3 + 5 x→∞ 3 + 2x − x 2 ∗ 25. lim tan x ∗ 26. lim tan x
2 3 2 x→−∞ x
x→∞
2+x−x x − 2x
5. lim 6. lim #+ $ #+ $
x→−∞ 3 + 4x 2 x→−∞ 3x 3 + 4x 2 ∗ 27. lim x2 + 1 − x ∗ 28. lim x2 + 4 − x
x→∞ x→∞
x 3 − 2x 2 + x + 1 x 3 − 2x 2 + x + 1
7. lim 8. lim #+ $ #+ $
x→−∞ x 4 + 3x x→−∞ x2 − x + 1 ∗ 29. lim 2x 2 + 1 − x ∗ 30. lim 2x 2 + 1 − x
√ x→∞ x→−∞
x2 + 1 3x − 1 √ √
∗ 9. lim ∗ 10. lim √ 3x 2 + 2 4x 2 + 7
x→∞ 2x + 1 x→∞ 5 + 4x 2 ∗ 31. lim ∗ 32. lim
√ √ x→∞ x+4 x→∞ 2x + 3
1 − 2x 2 1 − 2x √ √
∗ 11. lim ∗ 12. lim 3x 2 + 2 4x 2 + 7
x→−∞ x+2 x→−∞ x + 2
∗ 33. lim ∗ 34. lim
* √ x→−∞ x+4 x→−∞ 2x + 3
2+x 3+x
∗ 13. lim ∗ 14. lim √ #+ + $
x→∞ x−2 x→∞ x ∗ 35. lim x2 + 4 − x2 − 1
! " x→∞
1
∗ 15. lim (x 2 − x 3 ) ∗ 16. lim x + #√
3
√
3
$
x→∞ x→∞ x ∗ 36. lim 1+x− x
x→∞
x x2 #+ $
∗ 17. lim √ ∗ 18. lim √ ∗ 37. lim x2 + x − x
x→∞ x+5 x→−∞ 3−x x→∞
x 3x #+ $
∗ 19. lim √
3
∗ 20. lim √
3 ∗ 38. lim x2 + x − x
x→−∞ 4+ x3 x→∞ 2+ 4x 3 x→−∞
√
In Exercises 39–42 assume that a > 0 is a constant and calculate the 3x − 1 5x 2 + 7
∗ 45. f (x) = √ ∗ 46. f (x) =
limit, if it exists. 5 + 2x 2 2x + 3
x 2 + ax − 2 x 1 − 4x 3 3x 3 + 2x − 1
∗ 39. lim ∗ 40. lim √ ∗ 47. f (x) = ∗ 48. f (x) =
x→∞ ax 2 + 5 x→∞ 2
ax + 3x + 2 3 + 2x − x 2 1 − 3x + x 2
√ √
+ ax 2 + 7
∗ 41. lim ( x 2 + ax − x) ∗ 42. lim ax 2 + bx + c
x→∞ x→−∞ x − 3a ∗ 49. What is the value of lim , where a > 0, and
x→−∞ dx + e
b , c , d , and e are constants?
∗∗ 50. What conditions on the constants a , b , c , d , e , and f will ensure

In Exercises 43–48 identify all horizontal, vertical, and oblique asymp-
that
totes for the graph of the function. Determine whether the graph ap-
proaches horizontal and oblique asymptotes from above or below. #+ + $
lim ax 2 + bx + c − dx 2 + ex + f
x→∞
2−x x+3
∗ 43. f (x) = ∗ 44. f (x) =
3 + 4x 2x − 5 exists? What is the value of the limit in this case?
2.4 Continuity
We have noticed that sometimes the limit of a function f (x) as x → a is the same as the value
f (a) of the function at x = a . This property is described in the following definition. This is
the situation at x = 1 in Figure 2.1a, but not the case in Figure 2.1b or 2.1c.
DEFINITION 2.1
A function f (x) is said to be continuous at x = a if it satisfies three conditions:
1. f (x) is defined at x = a ;
2. limx→a f (x) exists;
3. The value of the function in 1 and its limit in 2 are the same.
All three conditions can be combined by writing the single equation
lim f (x) = f (a). (2.4a)

x→a
Equivalent to this, and sometimes more useful, is the equation
lim f (a + h) = f (a). (2.4b)

h→0
(See Exercise 46.)
The graph in Figure 2.27 illustrates a function that is discontinuous (i.e., not continuous) at
x = a , b , c , d , and e . For instance, at x = a , conditions 1 and 2 are satisfied but condition 3 is
violated; at x = b , condition 2 is satisfied but conditions 1 and 3 are not. Figure 2.27 suggests
that discontinuities of a function are characterized geometrically by separations in its graph. This
is indeed true, and it is often a very informative way to illustrate the nature of a discontinuity.
2.4 Continuity 129
FIGURE 2.27 Various types of discontinuities for a function
a b c d e x
When a function is defined on a closed interval a ≤ x ≤ b , Definition 2.1 must be rephrased

in terms of right- and left-hand limits for continuity at x = a and x = b . Specifically, we say
that f (x) is continuous from the right at x = a if limx→a + f (x) = f (a) , and that f (x) is
continuous from the left at x = b if limx→b− f (x) = f (b) .
A function is said to be continuous on an interval if it is continuous at each point of that
interval. In the event that the interval is closed, a ≤ x ≤ b , continuity at x = a and x = b
is interpreted as continuity from the right and left, respectively. Geometrically, a function is
continuous on an interval if a pencil can trace its graph completely without being lifted from the
page.
EXAMPLE 2.21
Draw graphs of the following functions, indicating any discontinuities:
x 2 − 16 1 |x 2 − 25|
(a) f (x) = (b) f (x) = (c) f (x) =
x−4 (x − 4)2 x 2 − 25
FIGURE 2.28a A function with a hole FIGURE 2.28b Computer plots do not show holes
y y
x2 − 16 8
y=
x−4
6
8
x 2 − 16
4 4 y=
x−4
−4
4 x 2
−4 −2 2 4 6 x
SOLUTION
(a) Since f (x) is undefined at x = 4, the function is discontinuous there. For x #= 4,
(x + 4)(x − 4)
f (x) = = x + 4.
x−4
Consequently, the graph of f (x) is a straight line with a hole at x = 4 (Figure
2.28a). The computer-generated graph in Figure 2.28b does not display the dis-
continuity at x = 4; depending on the plot interval, the graph may, or may not,
be accompanied by an error message about division by 0 at x = 4. Note that
limx→4 [(x 2 − 16)/(x − 4)] exists and is equal to 8.
FIGURE 2.29 A function may (b) Since 1/(x − 4)2 is undefined at x = 4, the function is discontinuous there. The
be unbounded near a discontinuity
limits
y
1 1 1 1
1/4 lim = ∞, lim− = ∞, lim = 0, lim =0
1 x→4+ (x − 4 )2 x→4 (x − 4 )2 x→∞ (x − 4 )2 x→−∞ (x − 4)2
y=
(x − 4)2
are displayed in Figure 2.29.
4 x (c) The function f (x) = |x 2 − 25|/(x 2 − 25) is undefined at x = ±5, and is therefore
discontinuous at these values. When −5 < x < 5,
25 − x 2
f (x) = = −1,
x 2 − 25
and when |x| > 5,

x 2 − 25
f (x) = = 1.
x 2 − 25
The graph is shown in Figure 2.30a; Figure 2.30b is a computer version. Neither of
the following limits exists:
|x 2 − 25| |x 2 − 25|
lim or lim ,
x→5 x 2 − 25 x→−5 x 2 − 25
although right- and left-hand limits exist as x → 5 and x → −5.
FIGURE 2.30a A function with different left-hand and FIGURE 2.30b Computer plots do not show that a function
right-hand limits may be undefined
y y
1.0 1.0
0.5 | x 2 − 25| 0.5 |x 2 − 25|

y= y=
x2 − 25 x 2 − 25
−10 −5 5 10 x −10 −5 5 10 x
−0.5 −0.5
−1.0 −1.0
We now mention some general properties of continuous functions. According to the following
theorem, when continuous functions are added, subtracted, multiplied, and divided, the result
is a continuous function.
THEOREM 2.3
If functions f (x) and g(x) are continuous at x = a , then so also are the functions
f (x) ± g(x) , f (x)g(x) , and f (x)/g(x) [provided that g(a) #= 0 in the case of
division].
2.4 Continuity 131
This is easily established with Theorem 2.1. For instance, to verify that f (x) + g(x) is
continuous, we note that since limx→a f (x) = f (a) and limx→a g(x) = g(a) ,
lim [f (x) + g(x)] = lim f (x) + lim g(x) (by Theorem 2.1)
x→a x→a x→a
= f (a) + g(a) [by continuity of f (x) and g(x)].
COROLLARY 2.3.1
If functions f (x) and g(x) are continuous on an interval, then so are the functions
f (x) ± g(x) , f (x)g(x) , and f (x)/g(x) [provided that g(x) never vanishes in the
interval in the case of division].
It is an immediate consequence of this corollary that polynomials are continuous for all real
numbers. Rational functions P (x)/Q(x) , where P and Q are polynomials, are continuous
on intervals in which Q(x) #= 0. The trigonometric functions sin x and cos x are continuous
for all x ; sec x and tan x have discontinuities at x = (2n + 1)π/2, where n is an integer; and
csc x and cot x have discontinuities at x = nπ (see Figures 1.90). The inverse trigonometric
functions are continuous wherever they are defined.
From our graphs of exponential and logarithm functions in Section 1.9, we might also be
led to conclude that these functions are continuous. But recall that there is a problem defining
exponential functions for irrational exponents. We know what 10x means when x is a rational
number, say, x = n/m . It means the mth root of√10n when n is an integer, √ or it means the n
th
th 1/m 2 2
power of the m root 10 . But what does 10 mean? We define 10 as limx→√2 10 , x
√
where x approaches 2, but√ does so only through rational numbers. For instance, from the
infinite decimal expansion for 2 =√1.414 213 562 . . . , we could create the following sequence
of rational numbers that approach 2 ever more closely as we include more digits after the
decimal:
1.4, 1.41, 1.414, 1.414 2, 1.414 21, 1.414 213, 1.414 213 5, . . . .
√
2
We define 10 to be that number approached by the following sequence of numbers:
101.4 , 101.41 , 101.414 , 101.414 2 , 101.414 21 , 101.414 213 , 101.414 213 5 , . . . .
Notice that each term in this sequence is 10 raised to a rational number. It can be shown
that these numbers do indeed approach a limit, and that this limit is independent of the particular
√ √
2
sequence of rationals√used to approach 2. With this value for 10 , the function 10x is
continuous at x = 2. In this way, exponential functions a x can be made continuous for
all reals. According to Theorem 2.3, the hyperbolic functions are all continuous except for the
hyperbolic cotangent and cosecant, which are discontinuous when their arguments are zero. The
following theorem implies that logarithm functions loga x , which are inverses of exponential
functions, are continuous for x > 0.
THEOREM 2.4
When a function f (x) is continuous on an interval I , and has an inverse function on I ,
then f −1 (x) is continuous on the range of f (x) .
Discontinuities are often characterized according to “size.” A discontinuity of a function

f (x) at x = a is said to be a removable discontinuity if limx→a f (x) exists, and either f (a)
is undefined or f (a) is defined and not equal to the limit. The former is illustrated at x = b in
Figure 2.27 and the latter at x = a in the same figure. A removable discontinuity can be removed
from a function by defining or redefining the function at the discontinuity as its limiting value
as the discontinuity is approached. For example in Figure 2.28a of Example 2.21, we would
define the value of f (x) at the removable discontinuity x = 4 as f (4) = limx→4 f (x) = 8.
This new function, which differs from f (x) only at x = 4, is now continuous for all x .
A function f (x) is said to have a (finite) jump discontinuity at x = a if right-hand and
left-hand limits exist as x approaches a , but these limits are different,
lim f (x) − lim f (x) = finite, nonzero number.

x→a + x→a −
Such is the case at the discontinuities in Figures 2.12 and 2.30a. Jump discontinuities are
not removable; the function cannot be defined or redefined only at the discontinuity to create a
continuous function.
A function f (x) is said to have an infinite discontinuity at x = a if limx→a + f (x) =
±∞ or limx→a − f (x) = ±∞ or both. Examples can be found in Figures 2.14, 2.15, and
2.16.
Not all discontinuities can be classified according to size. The function f (x) = sin (1/x)
in Figure 2.9 is discontinuous at x = 0 due to violent oscillations. This discontinuity is not
removable, nor is it a jump discontinuity or an infinite discontinuity.
EXAMPLE 2.22
The floor function (also called the greatest integer function) f (x) = *x+ of Exercise 68 in
Section 1.5 has many applications. Characterize its discontinuities.
FIGURE 2.31 The floor SOLUTION The graph of the function in Figure 2.31 indicates that it has jump discontinuities
function at integer values of x .
y
2
1 Besides adding, subtracting, multiplying, and dividing functions, functions
√ can also be com-
posed, or substituted one into another. For instance, when f (x) = x 2 + x and g(x) = ex ,
−3 −2 −1 1 2 3 x
the function obtained by replacing each x in f (x) by g(x) is called the composition of f and
−2 g,
# $ +
−3 f g(x) = e2x + ex .
Composition is often denoted by
# $
(f ◦ g)(x) = f g(x) .
Were we required to find the limit of this function as x → 1, say, we would probably write
nonchalantly that
+ +
lim e2x + ex = e2 + e.
x→1
Effectively, we have interchanged the operation of taking square roots and that of taking limits;
that is, without thinking we have written
+ * +
# $
lim e2x + ex = lim e2x + ex = e2 + e.
x→1 x→1
This is correct according to the following theorem, because the square root function is continuous.
THEOREM 2.5
If limx→a g(x) = L , and f (y) is a function that is continuous at y = L , then
# $ , -
lim f g(x) = f lim g(x) = f (L). (2.5)
x→a x→a
2.4 Continuity 133
An immediate consequence of this result is that the composition of a continuous function

with a continuous function yields a continuous function. We state this as a corollary.
COROLLARY 2.5.1
# $
If g(x) is continuous at x = a , and f (y) is continuous at g(a) , then f g(x) is
continuous at x = a .
EXAMPLE 2.23
! "
x2 − 2
Evaluate lim sin .
x→2 2x − 3
SOLUTION Since the sine function is continuous for all values of its argument, we may take
the limit operation inside the function and write
! " ! "
x2 − 2 x2 − 2
lim sin = sin lim = sin 2.
x→2 2x − 3 x→2 2x − 3
We have used computer plots in this section to illustrate various kinds of limits. We wanted you
to see limits graphically and get a visual feeling for them. In the next two examples we illustrate
how limits can be used to draw graphs of rational functions in the absence of technology.
EXAMPLE 2.24
x 2 − x + 12
Use limits to draw a graph of the rational function f (x) = .
x 2 + 4x − 5
SOLUTION Because x 2 + 4x − 5 = (x − 1)(x + 5) , the function is undefined at x = 1 and
x = −5. In order to discover the nature of the graph near these discontinuities, we calculate
left- and right-hand limits for each value of x ,
. / . /
x 2 − x + 12 42 x 2 − x + 12 42
lim =∞ , lim = −∞ ,
x→−5− (x − 1)(x + 5) (−6)(0− ) x→−5+ (x − 1)(x + 5) (−6)(0+ )
. / . /
x 2 − x + 12 12 x 2 − x + 12 12
lim = −∞ , lim =∞ .
x→1− (x − 1)(x + 5) (0− )(6) x→1+ (x − 1)(x + 5) (0+ )(6)
We use long division and limits as x → ±∞ to find horizontal asymptotes,
! "
x 2 − x + 12 17 − 5x
lim = = 1+ ,
lim 1+
x→−∞ (x − 1)(x + 5) x 2 + 4x − 5
x→−∞
! "
x 2 − x + 12 17 − 5x
lim = lim 1 + 2 = 1− .
x→∞ (x − 1)(x + 5) x→∞ x + 4x − 5
These limits are shown in Figure 2.32a. To finish the graph as in Figure 2.32b, we join the parts
smoothly, add the y -intercept at −12/5, and note that the graph has no x -intercepts. When we
have learned how to take derivatives in Chapter 3, we will be able to locate the precise positions
of the high point of the graph between the vertical asymptotes and the low point to the right of
x = 1. Notice that the graph crosses the horizontal asymptote once. We can locate this position
by solving
x 2 − x + 12
= 1.
x 2 + 4x − 5
The solution is x = 17/5.

x 2 − x + 12
Using limits to draw the graph of f (x) = 2
x + 4x − 5
y y
1 1
−5 1 x −5 1 x
−12/5
EXAMPLE 2.25
x 4 − 3x 3
Use limits to draw a graph of the rational function f (x) = .
x 3 + 2x 2 + x + 2
SOLUTION To discover points of discontinuity of the graph we factor the denominator, and
for limit calculations, we also factor the numerator,
x 3 (x − 3)
f (x) = .
(x + 2)(x 2 + 1)
The function is undefined at x = −2, and we therefore calculate
. /
x 3 (x − 3) (−8)(−5)
lim = −∞
x→−2− (x + 2)(x 2 + 1) (0− )(5)
and
. /
x 3 (x − 3) (−8)(−5)
lim =∞
x→−2+ (x + 2)(x 2 + 1) (0+ )(5)
Since the degree of the numerator is one more than that of the denominator, we have a slanted
asymptote that can be obtained by long division,
9x 2 + 3x + 10
f (x) = x − 5 + .
x 3 + 2x 2 + x + 2
Because (9x 2 + 3x + 10)/(x 3 + 2x 2 + x + 2) is positive for large x , the graph approaches

the slanted asymptote y = x − 5 from above as x → ∞ . It approaches the asymptote from
below as x → −∞ . These facts are shown in Figure 2.33a along with a y -intercept of zero,
which is also an x -intercept, and an additional x -intercept of 3. We join these parts of the graph
smoothly as shown in Figure 2.33b. We have flattened the graph at the origin because of the
x 3 -factor in the numerator (just as the curve y = x 3 is flat at the origin).
2.4 Continuity 135

x 4 − 3x 3
Using limits to draw graph of f (x) = 3
x + 2x 2 + x + 2
y y
−2 3 x −2 3 x
y=x−5 y=x−5
The following table indicates when rational functions have horizontal or oblique asymptotes.
TABLE 2.3
P (x)
Horizontal and Oblique Asymptotes for Rational Functions
Qx
If degree P < degree Q , graph has horizontal asymptote y = 0.
coefficient of highest power of P (x)

If degree P = degree Q , graph has horizontal asymptote y =
coefficient of highest power of Q(x)
If degree P = degree Q + 1, graph has oblique asymptote.
If degree P > degree Q + 1, graph has neither horizontal nor oblique asymptote.
EXERCISES 2.4
In Exercises 1–30 plot a graph of the function indicating any discon- 11. f (x) = sec 2x
tinuities. Classify each discontinuity as a removable discontinuity, a
12. f (x) = sin (1/x)
jump discontinuity, or an infinite discontinuity, if possible.
13. f (x) = x 2 sin (1/x)
2
1 16 − x x + 12
1. f (x) = 2. f (x) = 14. f (x) = , −5 ≤ x ≤ 4
x+2 x+4 x2 − 9
12 x 2 + 2x
3. f (x) = |x 2 − 5| 4. f (x) = 15. f (x) = , −3 ≤ x ≤ 6
x2 + 2 x2 − 9
12 12 x 3 − 27 x
5. f (x) = 6. f (x) = 16. f (x) = 17. f (x) =
x2 + 2x x2 + 2x + 2 |x − 3| x2 − 1
3 + 2x − x 2 x 3 + x 2 − 2x 3x + 2 x 2 − 3x + 2
7. f (x) = 8. f (x) = 18. f (x) = 19. f (x) =
x+1 x2 − x x2 −x−2 x 2 + 4x − 5
3x 2 − 6x x 2 + 3x + 2
x 3 − 2x 2 + 5x − 10 20. f (x) = 21. f (x) =
9. f (x) = x2 − 6x − 7 x+4
x−2
2
x − 2x + 4 1
22. f (x) = 23. f (x) =
10. f (x) = tan x x−1 x3 − 4x
2x 3 − 2 1 ∗ 45. Illustrate graphically that the function f (x) = x −1 sin x has a

24. f (x) = 25. f (x) =
x3
+ 5x 2 x4 + 3x 2 removable discontinuity at x = 0.
|3x + 1| 1 ∗ 46. Verify that the condition in equation 2.4b is equivalent to that in
26. f (x) = ∗ 27. f (x) = √
x+5 x−1 equation 2.4a.
1 1 47. (a) The signum function (or sign function), denoted by sgn, is
∗ 28. f (x) = √ ∗ 29. f (x) = √ defined by
5+x 5−x
* 
x−3  −1, x < 0
∗ 30. f (x) = sgn x = 0, x =0
x+2 
1, x > 0.
Draw its graph, indicating any discontinuities.
In Exercises 31–42 use limits to draw a graph of the function. Use a
plot to check the accuracy of your graph. (b) Draw a graph of the function f (x) = sgn(x + 1) − sgn
(x − 1) , indicating any discontinuities.
x−2 2x − 1 ∗ 48. (a) Draw a graph of the function f (x) = *10x+/10. Where
31. f (x) = 32. f (x) =
x+3 3 − 4x is the function discontinuous?
x+2 1−x (b) Prove that f (x) truncates positive numbers after the first
∗ 33. f (x) = 2 ∗ 34. f (x) = decimal.
x − 4x − 5 2x 2 + 5x − 3
∗ 49. What function truncates negative numbers after two decimals?
x2 + x + 2 x 2 − 3x − 4
∗ 35. f (x) = ∗ 36. f (x) = ∗ 50. (a) Draw a graph of the function f (x) = *x + 1/2+ . Where
x 2 − 6x + 9 3x 2 − 14x − 5
is the function discontinuous?
x2 − x − 2 3x 2 − 2x + 5 (b) Prove that f (x) rounds positive numbers to the nearest
∗ 37. f (x) = ∗ 38. f (x) =
3x + 1 1−x integer.
x 3
x − x 2 + 2x − 8
3 ∗ 51. What function rounds positive numbers to:
∗ 39. f (x) = ∗ 40. f (x) =
x3 − 1 x 3 − 3x 2 + 3x − 1 (a) the nearest tenth;
2
x − 5x + 6 3
x − 64 (b) the nearest hundredth;
∗ 41. f (x) = ∗ 42. f (x) = (c) 10−n , where n is a positive integer?
x 3 − 64 x 2 − 5x + 6
∗ 43. Is the function ∗∗ 52. Determine points of continuity, if any, for the function
0 0
x sin (1/x), x #= 0 1, x is a rational number,
f (x) = f (x) =
0, x =0 0, x is an irrational number.
continuous at x = 0? ∗∗ 53. Determine points of continuity, if any, for the function

∗ 44. If x = f (t) represents the displacement function of a particle 0
moving along the x -axis, can f (t) have discontinuities? Explain with x, x is a rational number,
f (x) =
graphs. 0, x is an irrational number.
2.5 Heaviside and Dirac-Delta Functions

One of the simplest, but at the same time most useful, functions in engineering and physics is
the Heaviside unit step function. It is defined as follows with the graph in Figure 2.34:
0
0, x <0
h(x) = (2.6)
1, x > 0.
We have drawn small circles at x = 0 to indicate that the function does not have a value
there; it has a jump discontinuity. Closely related to this function is that in Figure 2.35; the jump
from value 0 to value 1 takes place at x = a rather than at x = 0. Since this simply shifts the
graph in Figure 2.34 by a units to the right, it is customary to denote this function by h(x − a) .
Algebraically, we have
0
0, x <a
h(x − a) = (2.7)
1, x > a .
For example, consider the Heaviside function h(x − 4) . To evaluate this function at x = 6,
say, we have two choices. Because 6 > 4, equation 2.7 implies that the value of the function
at x = 6 is 1. Alternatively, we can substitute x = 6 into h(x − 4) to get h(6 − 4) = h(2) ,
and this is equal to 1 by equation 2.6.
FIGURE 2.34 Heaviside unit step function h(x) FIGURE 2.35 Heaviside unit step function h(x − a)
y y
y = h(x) y = h(x − a)
1 1
x a x
The product of h(x − a) with any other function f (x) results in a function f (x)h(x − a)
whose values are equal to those of f (x) for x > a , but whose values are 0 for x < a . For
example, if f (x) = x 2 + 2 (Figure 2.36a), a graph of f (x)h(x + 1) = (x 2 + 2)h(x + 1) is
as shown in Figure 2.36b.
Think of h(x − a) as a switch that turns values of a function f (x) on for x > a . For
the function f (x) in Figure 2.37a, f (x)h(x − a) is shown in Figure 2.37b. Values are 0 for
x < a and those of f (x) for x > a .
FIGURE 2.36a The parabola x 2 + 2 FIGURE 2.36b The parabola x 2 + 2 multiplied by h(x + 1)
y y
5 5
4 4
3 3
2 y= x2 + 2 2
y = (x 2 + 2) h(x + 1)
1 1
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2x −1 −0.5 0 0.5 1 1.5 x
FIGURE 2.37a Any function f (x) FIGURE 2.37b Function f (x) multiplied by h(x − a)
y y
y = f (x)
3 3
y = f (x) h(x − a)
2 2
1 1
a
−1 −0.5 0.5 1 1.5 2 2.5 x
−1 −0.5 0.5 1 1.5 2 2.5 x
−1 −1
−2 −2
−3 −3
The function h(x − a) − h(x − b) for b > a is shown in Figure 2.38. Algebraically, we
have 
 0, x < a
h(x − a) − h(x − b) = 1, a <x b .
When multiplying the function f (x) in Figure 2.37a, it yields the function f (x)[h(x − a)
− h(x − b)] with graph in Figure 2.39. Think of h(x − a) − h(x − b) as an on–off switch.
It turns values of other functions on for x > a and off again for x > b .
FIGURE 2.38 Graph of h(x − a) − h(x − b) FIGURE 2.39 Graph of f (x) multiplied by h(x −a)−h(x −b)
y y
y = h(x − a) − h(x − b)
3 y = f (x) [ h(x − a) − h(x − b) ]
2
1
1
−1 a 0.5 1 1.5 b 2.5 x

−1
a b x −2
−3
−4
FIGURE 2.40 A function with different definitions on various intervals
y
4 y = f (x)
−2 2 4 x
−2
−4
−6
Heaviside functions provide a convenient representation for functions that have different
definitions on different intervals. Such functions are said to be piecewise defined. For instance,
the function in Figure 2.40 is defined as follows:

 3x + 3, x < 0
f (x) = 4 − x 2 , 0 < x < 2

2, x > 2.
It can be represented algebraically in the form
f (x) = (3x + 3)[1 − h(x)] + (4 − x 2 )[h(x) − h(x − 2)] + 2h(x − 2)

= 3x + 3 + (−3x − 3 + 4 − x 2 )h(x) + (x 2 − 4 + 2)h(x − 2)
= 3x + 3 + (1 − 3x − x 2 )h(x) + (x 2 − 2)h(x − 2).
There are many physical examples in which Heaviside functions are very useful. Consider
a mass m attached to a spring as shown in Figure 2.41. Motion is initiated at time t = 0 by
pulling the mass away from its equilibrium position and releasing it. During the subsequent
motion, various forces could act on the mass, including the spring, friction with the surface on
which it slides, air resistance, and others. Suppose that among the others, a force to the left
with magnitude 5 N is applied for 3 s beginning at time t = 1 s. This force can be represented
algebraically as
F (t) = −5[h(t − 1) − h(t − 4)].
The generator in Figure 2.42, were it operational from time t = 0, would produce an
oscillating voltage A sin ωt , where A and ω are positive constants. If the generator is indeed
turned on at time t = 0, by closing the switch, and off again after 10 s by opening the switch,
the voltage applied to the circuit can be expressed in the form
E(t) = A sin ωt [h(t) − h(t − 10)].
The beam in Figure 2.43 is made more rigid by attaching a second beam over the middle
half. This extra support creates additional loading on the original beam. If the support has mass
per unit length m , then its weight per unit length is −9.81m , and the extra loading per unit
length can be expressed algebraically as
F (x) = −9.81m[h(x − L/4) − h(x − 3L/4)].
FIGURE 2.41 Schematic for vibrating mass-spring system FIGURE 2.42 Schematic for LCR circuit
A sin ωt R
m
x
x=0 S
L
FIGURE 2.43 Schematic for loaded beam
y
L
L/2
Extra
support
x
Beam
Throughout the text, we will be hired as consulting engineers to tackle projects. Information
will be given to us that may not always be sufficient to finish our work. We may have to make
justifiable assumptions. Here is Project 1.
Consulting Project 1
The national space association is building a two-stage rocket wherein the rocket burns
fuel during the first stage, jettisons the thrusters, burns fuel in the second stage, achieves
orbit, and turns the remaining engines off. The initial mass of the rocket is M0 kilograms,
of which M1 kilograms is the first stage. This consists of the mass of the thrusters and
fuel. Thrusters burn fuel at a constant rate of r1 kilograms per second for t1 seconds. After
the t1 seconds, the thrusters are jettisoned, and the remaining engines ignite and burn fuel
at a constant rate of r2 kilograms per second for the next t2 minutes. At this time the
rocket achieves orbit and engines are shut down. For enormous numbers of calculations,
the association wants a single formula for the mass of the rocket at all times during take
off and after reaching orbit. Our task is to find one.
SOLUTION We quickly realize that there will be different formulas for the mass of the
rocket during the various stages. Heaviside functions are ideal for representing piecewise
defined functions in a single formula. The mass of the rocket during lift-off is
M0 − r1 t, 0 < t < t1 .
After t1 seconds, the thrusters are jettisoned, and the remaining engines ignite and burn
fuel at a constant rate of r2 kilograms per second for the next 60t2 seconds. The mass of
the rocket during this time interval is
M0 − M1 − r2 (t − t1 ), t1 < t < t1 + 60t2 .
Finally, when the rocket achieves orbit and the engines are shut down, the mass of the
rocket is
M0 − M1 − 60t2 r2 , t > t1 + 60t2 .
We now use Heaviside functions to bring these expressions into a single formula for M(t) ,
the mass of the rocket at time t :
M(t) = (M0 − r1 t)[h(t) − h(t − t1 )]

+ [M0 − M1 − r2 (t − t1 )][h(t − t1 ) − h(t − t1 − 60t2 )]
+ (M0 − M1 − 60t2 r2 )h(t − t1 − 60t2 ).
By recombining this expression into terms involving h(t) , h(t −t1 ) , and h(t −t1 − 60t2 ) ,
we can also write
M(t) = (M0 − r1 t)h(t) − [M1 − r2 t1 + (r2 − r1 )t ]h(t − t1 )

+ r2 (t − t1 − 60t2 )h(t − t1 − 60t2 ).
A graph of this function is shown in Figure 2.44. It has a jump discontinuity at t = t1

when thrusters are jettisoned, and a removable discontinuity at t = t1 + 60t2 when
engines are turned off. We could remove the discontinuity at t = t1 + 60t2 by defining
M(t1 + 60t2 ) = M0 − M1 − 60t2 r2 .
FIGURE 2.44 Function representing mass of multi-stage rocket
M(t)
M0
M0 − r1t1
M 0 − M1
M0 − M1 − 60t2r2
t1 t1 + 60t2 t
Application Preview In the Application Preview we introduced the problem of describing an instantaneously
Revisited applied voltage to an electrical network. The same situation would arise were we to consider
the mathematical representation of striking the mass in Figure 2.41 with a hammer. We would
also find the same situation if a point-load were to replace the extra support on the beam in
Figure 2.43, or be in addition to the extra support.
To find a mathematical representation for the instantaneously applied voltage, the force
due to an impinging hammer, and a point-load on a beam, we begin with the function (b −
a)−1 [h(x − a) − h(x − b)] in Figure 2.45a. What is important to notice here is that the area
of the rectangle formed by drawing vertical lines at x = a and x = b is 1. If b is replaced by
a + $ so that $ is the width of the nonzero portion of the graph, then the area of the rectangle is
also 1 if the height of the function is 1/$ (Figure 2.45b). The function describing this graph is
1
[h(x − a) − h(x − a − $)].
$
This is often called a unit pulse function.
FIGURE 2.45a Unit pulse function (b − a)−1 [h(x − a) − h(x − b)] FIGURE 2.45b Unit pulse function 1
$
[h(x − a) − h(x − a − $)]
y y
1 1
b−a
a b x x
a
Suppose we make $ in Figure 2.45b smaller and smaller, so that a + $ → a and the height
of the horizontal line moves upward (in order that the area under the curve will always be 1).
A few smaller values of $ are shown in Figure 2.46. If we take the limit as $ → 0, we obtain
what is called the Dirac-delta function. It is denoted by
1
δ(x − a) = lim [h(x − a) − h(x − a − $)]. (2.9)
$→0 $
FIGURE 2.46 Unit pulse functions leading to Dirac-delta function
y
3
x
a a+ a+
3 2
Its value is zero for every x #= a , and somehow its value at x = a is “equal to infinity.” Clearly,
this is not a function as we understand functions, since functions cannot have infinite values. It
is called a generalized function. Generalized functions have operational properties that make
them very useful in handling point sources in engineering and physics. We will introduce these
properties at appropriate places in the text.
Point sources arise in most areas of engineering and physics. For instance, suppose we
were to strike the left side of the vibrating mass in Figure 2.41 with a hammer at time t = t0 .
This would be equivalent to applying a force over a very short time interval. If it were deemed
that striking the mass is equivalent not to a 10 N force for 1 s, nor a 20 N force for 1/2 s, nor a
40 N force for 1/4 s, but to the limit of this sequence of forces, we would say that the mass has
been struck with a force of 10 N at t = t0 , and express the force in the form
F (t) = 10 δ(t − t0 ).
If we were to take a sharp object, place it at the centre of the beam in Figure 2.43, and push
downward with a force of 200 N, we would represent this force as
F (x) = −200 δ(x − L/2).
It is a point force of 200 N at x = L/2.

A unit voltage at time t = t0 for the circuit in Figure 2.42 is δ(t − t0 ) . It is equivalent to
applying a voltage not equivalent to 1 V for 1 s, nor 2 V for 1/2 s, nor 4 V for 1/4 s, but a voltage
equivalent to the limit of such voltages.
We shall learn how to manipulate Dirac-delta functions in these applications as calculus
unfolds throughout the text.
EXERCISES 2.5
In Exercises 1–6 express the piecewise defined function in terms of In Exercises 7–11 the force described acts on the mass on the end of
Heaviside functions. Draw a graph of each function. the spring in Figure 2.41. Represent the force in terms of Heaviside
% functions or Dirac-delta functions, as appropriate.
1, x<0
∗ 1. f (x) = 2 − x, 0<x<2 7. A force of F N to the left for the first T seconds after t = 0
2, x>2 8. A periodic force F (t) = 10 sin 4t that is turned on for two periods
0 beginning at time t = 1
2
1−x , x<0
∗ 2. f (x) =
x2, x>0 9. An instantaneous force of 50 N to the left at time t = 4
1 ∗ 10. A force of 100 N that begins at time t = 10 and increases linearly
sin x, x<0
∗ 3. f (x) = in size to 200 N over the next 50 seconds
2 + 2 cos x, x>0
 ∗ 11. Every 10 seconds beginning at t = 0 s and ending at t = 60 s,
 0, x < −1
 the mass is stuck with a hammer with a force of 60 N to the right
x − 2, −1 < x < 3
∗ 4. f (x) =
 x − 4, 3 < x < 5

0, x>5 In Exercises 12–15 the load described is applied to the beam in Fig-
 ure 2.43 without the extra support. Represent the force in terms of
 0, x<0
 Heaviside functions or Dirac-delta functions, as appropriate.
x, 0<x<1
∗ 5. f (x) =
 1 − x, 1 < x < 2
 12. An extra support of length L/2 and mass m over the left half of
0, x>2 the beam
 0, x<0
 13. A downward force of F > 0 N concentrated at x = L/3

 sin x, 0<x<π

 14. An upward force of F1 > 0 N concentrated at x = x1 and a
2 sin (x − π ), π < x < 2π
∗ 6. f (x) = downward force of F2 > 0 N concentrated at x = x2

 3 sin (x − 2 π ), 2π < x < 3π


 4 sin (x − 3π ), 3π < x < 4π 15. A total mass m evenly distributed over the last two-thirds of the
0, x > 4π beam
∗ 16. The function h(x − a) − h(x − b) is an on-off switch. What ∗ 25. Express the step function shown below in terms of Heaviside func-
function would be an on-off-on switch? tions.
∗ 17. How would you represent a switch that turns on at time t = 0, off
y
at t = 1, on again at t = 2, off again at t = 3, and so on?
∗ 18. Is h(x − a)h(x − b) = h(x − b) , when b > a ? 4
3
In Exercises 19–24 draw a graph of the function. 2
∗ 19. f (x) = h(2x − 1) ∗ 20. f (x) = h(3 − 2x) 1

2
∗ 21. f (x) = h(x − 1) ∗ 22. f (x) = 5 h(4 − x 2 )
1 2 3 4 5 x
∗ 23. f (x) = x 2 h(x 2 − 2x − 3)
∗ 24. f (x) = (5 − x) h(2 − x 3 )
2.6 A Mathematical Definition of Limits

Our work on limits has been intuitive, but some later topics require a precise definition of
limits. To obtain such a definition, we begin with our intuitive statement of a limit and make a
succession of paraphrases, each of which is one step closer to the definition. We do this because
the definition of a limit is at first sight quite overwhelming, and we wish to show that it can be
obtained by a fairly straightforward sequence of steps. We hope that this will be a convincing
argument that the definition of a limit does indeed describe in mathematical terms our intuitive
concept of a limit. Our intuitive statement is:
• A function f (x) has limit L as x approaches a if f (x) can be made arbitrarily close to
L by choosing x sufficiently close to a .
Next we paraphrase “ f (x) can be made arbitrarily close to L .”
• A function f (x) has limit L as x approaches a if the difference |f (x) − L| can be made
arbitrarily close to zero by choosing x sufficiently close to a .
Now we take the important step — make “ |f (x) − L| can be made arbitrarily close
to zero” mathematical.
• A function f (x) has limit L as x approaches a if given any real number $ > 0, no matter
how small, we can make the difference |f (x) − L| less than $ by choosing x sufficiently
close to a .
Penultimately, we paraphrase “by choosing x sufficiently close to a .”
• A function f (x) has limit L as x approaches a if given any $ > 0, we can make
|f (x) − L| < $ by choosing |x − a| sufficiently close to zero.
Finally, by making “choosing |x−a| sufficiently close to zero” precise, we arrive at the definition
of a limit.
DEFINITION 2.2
A function f (x) has limit L as x approaches a if given any $ > 0, we can find a δ > 0
such that
|f (x) − L| < $
whenever 0 < |x − a| < δ .
Notice that by requiring 0 < |x − a| , this definition states explicitly that as far as limits
are concerned, the value of f (x) at x = a is irrelevant. In taking limits we consider values of
x closer and closer to a , but we do not consider the value of f (x) at x = a . This definition
states in precise terms our intuitive idea of a limit: that f (x) can be made arbitrarily close to L
by choosing x sufficiently close to a . Perhaps you will get a better feeling for this definition if
we interpret it graphically. Figure 2.47a indicates a function that has limit L as x approaches a .
Let us illustrate what must be done to verify algebraically that limx→a f (x) = L . We suppose
that we are given a value $ > 0, which we should envisage as being very small, although we
are never told exactly what it is. We must show that x can be restricted to 0 < |x − a| < δ
so that |f (x) − L| < $ . The latter inequality is equivalent to −$ < f (x) − L < $ or
L − $ < f (x) < L + $ , and this describes a horizontal band of width 2$ around the line
y = L (shaded in Figure 2.47b). What Definition 2.2 requires is that we find an interval of
width 2δ around x = a , as |x − a| < δ is equivalent to a − δ < x < a + δ , such that
whenever x is in this interval, the values of f (x) are all within the shaded horizontal band
around y = L . Such an interval is shown in Figure 2.47c for the given $ . Now Definition 2.2
requires us to verify that the δ -interval can always be found no matter how small $ is chosen
to be. This is always possible for the function illustrated in Figure 2.47, and it is clear that the
smaller the given value of $ , the smaller δ will have to be chosen. For instance, for the value of
$ in Figure 2.47d, δ is smaller than that in Figure 2.47c. In other words, the value of δ depends
on the value of $ . Herein lies the difficulty in using Definition 2.2. In order to ensure that δ can
be found for any given value of $ , we usually determine precisely how δ depends functionally
on $ . We illustrate with two examples.
FIGURE 2.47a Graphs to illustrate the mathematical definition of limit FIGURE 2.47b
y y
L L
a x a x
FIGURE 2.47c FIGURE 2.47d
y y
L L
a x a x
EXAMPLE 2.26
Use Definition 2.2 to prove that lim (2x + 4) = 10.
x→3
SOLUTION It is true that based on Theorem 2.1, this is quite obvious, but we are required to
use Definition 2.2. We must show that given any $ > 0, we can choose x sufficiently close to
3 that
|(2x + 4) − 10| < $.
To do so, we rewrite the left side of the inequality with all x ’s in the combination x − 3,
|(2x + 4) − 10| = |2x − 6| = 2|x − 3|.
We must now choose x so that

2|x − 3| < $.
But this will be true if |x − 3| < $/2. In other words, if we choose x to satisfy 0 < |x − 3| <
$/2, then
,$ -
|(2x + 4) − 10| = 2|x − 3| < 2 = $.
2
The verification is now complete. We have shown that we can make 2x + 4 as close to 10 as
we want (within $ ) by choosing x sufficiently close to 3 (within δ = $/2).
The following example is more complicated, but a manageable one.
EXAMPLE 2.27
Use Definition 2.2 to prove that lim (x 2 + 5) = 9.
x→2
SOLUTION We must show that given any $ > 0, we can choose x sufficiently close to 2 that
|(x 2 + 5) − 9| < $ or |x 2 − 4| < $.
Once again we rewrite the left side of the inequality with all x ’s in the combination x − 2:
|x 2 − 4| = |(x − 2)2 + 4x − 8| = |(x − 2)2 + 4(x − 2)|.
We must now choose x so that
|(x − 2)2 + 4(x − 2)| < $.
Now, all real numbers a and b satisfy the inequality
|a + b| ≤ |a| + |b|. (2.10)
With a replaced by (x − 2)2 and b replaced by 4(x − 2) , we can say that
|(x − 2)2 + 4(x − 2)| ≤ |x − 2|2 + 4|x − 2|.
As a result, we consider finding x so that
FIGURE 2.48 The parabola |x − 2|2 + 4|x − 2| < $.

z 2 + 4z − $
y If we set z = |x − 2| and consider the parabola Q(z) = z2 + 4z − $ in Figure 2.48, we will
y = Q (z)
be able to see what to do. The parabola crosses the z -axis when
z 2 + 4z − $ = 0 ,
a quadratic with solutions

√
−4 ± 16 + 4$ √
z z= = −2 ± 4 + $.
2
√
The graph shows that whenever 0 < z < −2 + 4 + $ , then
z 2 + 4z − $ < 0 .
√
Since z = |x − 2| , we can say that if 0 < |x − 2| < 4 + $ − 2, then
|x − 2|2 + 4|x − 2| < $,
and therefore
|(x − 2)2 + 4(x − 2)| < $.
2
Verification is now complete. We have shown that we can make √ x + 5 as close to 9 as we want
(within $ ) by choosing x sufficiently close to 2 (within δ = 4 + $ − 2).
These examples have indicated the difference between “evaluation of” and “verification of” a
limit. Evaluation is calculational; it comes first. Verification is much more difficult; it requires
a clear understanding of Definition 2.2. Results in succeeding chapters rely heavily on our
intuitive understanding of limits and the ability to calculate limits, but use of Definition 2.2 is
kept to a minimum.
EXERCISES 2.6
In Exercises 1–9 use Definition 2.2 to verify the limit. ∗ 23. lim (x + 5) = ∞ ∗ 24. lim (5 − x 2 ) = −∞
x→∞ x→∞
∗ 1. lim (x + 5) = 6 ∗ 2. lim (2x − 3) = 1
x→1 x→2 x+2 x+2
∗ 25. lim =1 ∗ 26. lim =1
x→∞ x−1 x→−∞ x−1
∗ 3. lim (x 2 + 3) = 3 ∗ 4. lim (x 2 + 4) = 5
x→0 x→1
∗ 27. lim (5 − x) = ∞ ∗ 28. lim (3 + x − x 2 ) = −∞
2
∗ 5. lim (3 − x ) = −1 ∗ 6. lim (x 2 − 7x) = −12 x→−∞ x→−∞
x→−2 x→3
∗ 7. lim (x − 3x + 4) = 8 ∗ 8. lim (x 2 + 3x + 5) = 9
2 ∗ 29. Prove that if limx→a f (x) = L > 0, then there exists an open
x→−1 x→1 interval I containing a in which f (x) > 0 except possibly at x = a .
x+2
∗ 9. lim =4
x→2 x−1 ∗ 30. Does the function
0
x sin (1/x), x #= 1/(nπ )
In Exercises 10–19 give a mathematical definition for each statement. f (x) =
1, x = 1/(nπ ),
∗ 10. lim f (x) = L ∗ 11. lim f (x) = L
x→a + x→a − where n is an integer, have a limit as x → 0?
∗ 12. lim f (x) = L ∗ 13. lim f (x) = L
x→∞ x→−∞
In Exercises 31–35 we use Definition 2.2 to prove Theorem 2.1. In
∗ 14. lim f (x) = ∞ ∗ 15. lim f (x) = −∞
x→a x→a each exercise assume that limx→a f (x) = F and limx→a g(x) = G .
∗ 16. lim f (x) = ∞ ∗ 17. lim f (x) = −∞ ∗ 31. Show that given any $ > 0, there exist numbers δ1 > 0 and
x→∞ x→∞
δ2 > 0 such that
∗ 18. lim f (x) = ∞ ∗ 19. lim f (x) = −∞
x→−∞ x→−∞
|f (x) − F | < $/2 whenever 0 < |x − a| < δ1 and
∗ 20. Use Definition 2.2 to prove that a function f (x) cannot have two |g(x) − G| < $/2 whenever 0 < |x − a| < δ2 .
limits as x approaches a .
Use these results along with identity 2.10 to prove part (i) of Theorem
In Exercises 21–28 use the appropriate definition from Exercises 10– 2.1.
19 to verify the limit. ∗ 32. Use a proof similar to that in Exercise 31 to verify part (ii) of
1 −1 Theorem 2.1.
∗ 21. lim =∞ ∗ 22. lim = −∞
x→1 (x − 1)2 x→−2 (x + 2)2 ∗ 33. Verify part (iii) of Theorem 2.1.
Key Terms 147
∗∗ 34. (a) Verify that ∗∗ 35. (a) Verify that when G #= 0,

2 2
|f (x)g(x) − F G| ≤ |f (x)||g(x) − G| + |G||f (x) − F |. 2 f (x) F 22 |f (x) − F | |F ||G − g(x)|
2
2 g(x) − G 2 ≤ |g(x)|
+
|G||g(x)|
.
(b) Show that given any $ > 0, there exist numbers δ1 > 0,
δ2 > 0, and δ3 > 0 such that (b) Show that given any $ > 0, there exist numbers δ1 > 0,
δ2 > 0, and δ3 > 0 such that
|f (x)| < |F | + 1 whenever 0 < |x − a| < δ1 ,
|G|
$ |g(x)| > whenever 0 < |x − a| < δ1 ,
|g(x) − G| < whenever 0 < |x − a| < δ2 , 2
2(|F | + 1)
$|G|
and |f (x) − F | < whenever 0 < |x − a| < δ2 , and
4
$ $|G|2
|f (x) − F | < whenever 0 < |x − a| < δ3 . |g(x) − G| < whenever 0 < |x − a| < δ3 .
2|G| + 1 4(|F | + 1)
(c) Use these results to prove part (iv) of Theorem 2.1. (c) Now prove part (v) of Theorem 2.1.
SUMMARY
In Section 2.1 we introduced limits of functions. For the most part our discussion was in-
tuitive, beginning with the statement “limx→a f (x) = L if f (x) can be made arbitrarily
close to L by choosing x sufficiently close to a .” This idea was then extended to include the
following:
Right-hand limits: lim f (x);

x→a +
Left-hand limits: lim f (x);

x→a −
Limits at infinity: lim f (x), lim f (x);

x→∞ x→−∞
Infinite limits: lim f (x) = ∞, lim f (x) = −∞.

x→a x→a
Keep in mind that the term infinite limits is somewhat of a misnomer since in both situations
the limit does not exist.
A function f (x) is continuous at a point x = a if
lim f (x) = f (a).

x→a
To be continuous at x = a , a function must be defined at x = a and have a limit as x → a , and

these numbers must be the same. The function is continuous on an interval if it is continuous
at each point of that interval. Geometrically, this means that one must be able to trace its
graph completely without lifting pencil from page. Most discontinuities can be characterized
as removable, finite, or infinite.
We illustrated that discontinuous functions such as the Heaviside and Dirac-delta functions
model many physical situations.
In Section 2.6 we developed the mathematical definition of a limit.
KEY TERMS
Limit Squeeze or sandwich theorem

Left-hand limit Right-hand limit
Infinity Vertical asymptote
Limits at infinity Horizontal asymptote

Oblique asymptotes Continuous function
Continuity from the right Continuity from the left
Continuity on an interval Removable discontinuity
(Finite) jump discontinuity Infinite discontinuity
Composition Heaviside unit step function
Piecewise defined function Unit pulse function
Dirac-delta function
REVIEW
EXERCISES
In Exercises 1–20 evaluate the limit, if it exists. In Exercises 21–32 draw a graph of the function, indicating any dis-
continuities. Determine whether discontinuities are removable, jump,
or infinite.
x 2 − 2x x2 − 1 1 x
1. lim 2. lim 21. f (x) = 22. f (x) =
x→1 x + 5 x→−1 x + 1 x−2 x−2
x2 x 2 − 36
x 2 + 4x + 4 x+5 23. f (x) = 24. f (x) =
3. lim 4. lim x−2 x−6
x→−2 x+3 x→∞ x − 3 2 2
x+1 2x + 12
25. f (x) = 26. f (x) = 22 2
x 2 + 3x + 2 5 − x3 x−1 x − 12
5. lim 6. lim
x→−∞ 2x 2 − 5 x→−∞ 3 + 4x 3 |x + 1| x+1
∗ 27. f (x) = ∗ 28. f (x) =
x−1 |x − 1|
3x 3 + 2x − 5 4 − 3x + x 2 2x
7. lim 8. lim ∗ 29. f (x) =
x→∞ x 2 + 5x x→∞ 3 + 5x 3 x2 − 3x − 4
2x 2
x 2 − 2x x 2 − 4x + 4 ∗ 30. f (x) =
9. lim 2 10. lim x2 − 3x − 4
x→2+ x + 2x x→2− x−2
x − 3x 2 + 3x − 1
3
∗ 31. f (x) =
x 2 + 2x x 2 + 5x x−1
11. lim 12. lim
x→0 3x − 2x 2 x→1 (x − 1)2 x 3 − 3x 2 + 3x − 1
∗ 32. f (x) =
x 2 − 2x + 1
√ √
x−1 x−1
13. lim 14. lim
x→1 x x→1 x − 1 In Exercises 33–34 express the piecewise defined function in terms of
Heaviside functions.
(2 − 4x)3 cos 5x %
15. lim ∗ 16. lim x2, x<0
x→1/2 x(2x − 1)2 x→∞ x 33. f (x) = x, 0<x<4
5 − 2x, x > 4
√ %
3x 2 + 4 3 + x3, x < −1
∗ 17. lim x sin x ∗ 18. lim
x→−∞ x→−∞ 2x + 5 34. f (x) = x 2 + 2, −1 < x < 2
4, x>2
√
3x 2 + 4 #√ √ $ ∗ 35. For what values of x is the function f (x) = *x 2 + discontinuous?
∗ 19. lim ∗ 20. lim 2x + 1 − 3x − 1
x→∞ 2x + 5 x→∞ Draw its graph.
CHAPTER
3 Differentiation
Application Preview The container in the figure below represents a chemical reactor in which a chemical is either
created or broken down. The chemical enters in the form of a solution at one concentration and
leaves the reactor at a different concentration.
THE PROBLEM Given the flow rates at which solution enters and leaves the reactor, and
Incoming Exiting the concentration of chemical entering the reactor, find the concentration at which chemical
chemical chemical leaves the reactor. Can you see that there is not enough information to solve the problem? Some
information must be given, or assumed, about the rate at which chemical is formed or broken
down.
Chemical reaction Chapters 1 and 2 have prepared the way for calculus. The functions and curves in these
chapters yield a wealth of examples for our discussions, and limits from Chapter 2 provide the
tool by which calculus is developed. Calculus has two major components, differentiation and
integration. In this chapter we study differentiation. We learn what a derivative is algebraically
and geometrically, and develop some of its properties; we learn how to differentiate polynomi-
als and rational, trigonometric, exponential, logarithm, and hyperbolic functions; and we see
glimpses of the applications that are to follow in Chapter 4.
3.1 The Derivative

Very few quantities in real life remain constant; most are in a state of change. For example, room
temperature, the speed of a car, and the angle of elevation of the sun are three commonplace
quantities that are constantly changing. A few more technical ones are current in a transmission
line, barometric pressure, moisture content of the soil, and stress in vertical members of tall
buildings during hurricanes. Rates at which these quantities change are called derivatives. To
study the concept of a rate of change more thoroughly, we consider two commonplace physical
situations.
Displacement and Speed

Calculus plays a central role in discussions that trace the motion of particles and objects moving
under the influence of forces. Without calculus, analysis of a simple system such as a mass on
the end of a spring would be difficult; planning lift-off, space travel, and re-entry of spacecraft
would be impossible. We introduce basic ideas connected with motion here, and with each new
development of calculus, we take the analysis a little further.
Suppose a particle is at some position on the x -axis at time t = 0, and for t > 0, various
forces act on it resulting in motion that is confined to the x -axis (Figure 3.1).
149
150 Chapter 3 Differentiation
FIGURE 3.1 Displacement of particle moving along the x -axis
Particle
x
0
x
The position of the particle can be identified by its x -coordinate, and x can always be
regarded as a function of time t . Suppose, for example, that
x = f (t) = t 3 − 27t 2 + 168t + 20, t ≥ 0,
where x is measured in metres and t in seconds. This is called the displacement function;
it gives a complete history of the particle’s motion. It tells us not only where the particle is at
any given time, it also contains information about the velocity, speed, and acceleration of the
particle. We begin by plotting a graph of the displacement function (Figure 3.2). Ordinates of
this curve represent horizontal displacements of the particle relative to x = 0. For example,
the height of the displacement graph at time t = 0 is 20. This means that the particle begins
motion 20 m to the right of the origin. At t = 4 s, it is 324 m to the right of the origin, and at
t = 14 s, it is 176 m to the left of the origin. Whenever ordinates are positive, the particle is
to the right of the origin. Whenever ordinates are negative, the particle is to the left of x = 0.
Figure 3.2 indicates that the particle is to the right of x = 0 for the first ten seconds and then
again after t ≈ 17 s. Between these times, it is to the left of the origin.
FIGURE 3.2 Displacement graph for motion along the x -axis
x
x = t 3 − 27t 2 + 168t + 20
300
200
100
5 10 15 t
−100
−200
For 0 < t < 4, values of x are getting larger. This means that the particle is moving to
the right. For 4 < t < 14, the particle moves to the left (values of x are getting smaller), and
for t > 14, it moves to the right once again. This means that the particle stops and reverses
direction at t = 4 s and at t = 14 s.
Everyone has an intuitive idea about speed, when something moves quickly and when it
moves slowly. To discuss speed of the above particle quantitatively, we need derivatives, but
even now we can derive qualitative information about how fast the particle is moving. The more
distance the particle covers in a given time interval, the faster it is moving. In Table 3.1, we
have tabulated the distance that the particle travels each second for the first twenty seconds.
The first two entries indicate that the particle travels 142 m in the first second, but only 94 m
from t = 1 s to t = 2 s. The particle is therefore travelling faster during the first second of its
motion than during the next second. The remainder of the table suggests that the particle slows
down until, as we know, it comes to a stop at t = 4 s. It then speeds up until somewhere around
t = 9 s when it starts to slow down again, and comes to a stop at t = 14 s. It then picks up
speed thereafter.
TABLE 3.1
time (s) 1 2 3 4 5 6 7 8 9 10
distance (m) 142 94 52 16 14 38 56 68 74 74
time (s) 11 12 13 14 15 16 17 18 19 20
distance (m) 68 56 38 32 30 52 94 142 196 256
We can see this graphically as well as in tabular form. The steeper the displacement graph,
the larger the speed, the flatter the graph, the smaller the speed. This means that speed is greatest
when the graph rises or falls quickly. At the beginning of the particle’s motion (for small t > 0),
the graph is relatively steep, and therefore the particle is moving quickly. Steepness decreases
as we head toward t = 4 s, and therefore the particle is slowing down. At t = 4 s, the particle’s
speed is zero (it reverses direction). The particle then picks up speed, moving now to the left
until steepness is greatest somewhere around t = 9 s. It then starts to slow down, coming to a
stop at t = 14 s. It then moves to the right thereafter picking up speed (as steepness increases).
Again, all of this is quite qualitative, but when we learn about derivatives, we can be much more
specific and we can add information about the velocity and acceleration of the particle. We
begin with velocity.
Figure 3.3 is an enlargement of that part of the graph in Figure 3.2 for the first four seconds.
At time t = 1 s, the particle is 162 m to the right of the origin. At t = 4 s, it is at position
x = 324 m. During these three seconds, the particle has moved 324 − 162 or 162 m; its
displacement at t = 4 s relative to its position at t = 1 s is 162 m to the right. When we divide
the displacement by time taken to travel it, we obtain 162/3 = 54. This is called the average
velocity of the particle during the time interval and has units of metres per second. During
the time interval 1 ≤ t ≤ 3, the displacement of the particle is 146 m; therefore, its average
velocity in this time interval is 146/2 or 73 m/s. The average velocity during 1 ≤ t ≤ 2 is
94 m/s. What we are doing is taking average velocities over shorter and shorter time intervals,
all beginning at t = 1 s. If we continue the process indefinitely, we are in effect taking the
limit of average velocities as the length of the time interval starting at t = 1 approaches 0. This
will be called the instantaneous velocity of the particle at t = 1 s. Let us introduce notation to
describe the limiting process, and do so at an arbitrary time t0 rather than at t = 1.
FIGURE 3.3 Enlargement of graph in Figure 3.2

x
300 (4, 324)
(3, 308)
250 (2, 256)
200
150 (1, 162)
100
50
t
1 2 3 4
We let !t represent a small interval of time, often called an increment of time. The
function value f (t0 ) is the position (or displacement) of the particle at time t0 , f (t0 + !t) is
its position at time t0 + !t , and f (t0 + !t) − f (t0 ) is the difference in these displacements.
It represents the displacement of the particle at time t0 + !t relative to its position at time t0 .
It may or may not represent the distance travelled by the particle during the time interval !t .
(Can you explain why?) The quotient
f (t0 + !t) − f (t0 )

!t
is the average velocity of the particle during the time interval between t0 and t0 + !t . The limit
of this quotient as !t → 0 is called the instantaneous velocity of the particle at time t0 ; it is
denoted by
f (t0 + !t) − f (t0 )
v(t0 ) = lim . (3.1)
!t→0 !t
It represents how fast the particle is moving at time t0 ; it is the instantaneous rate of change of
displacement with respect to time. For example, the instantaneous velocity at t = 1 is
f (1 + !t) − f (1)
v(1) = lim
!t→0 !t
(1 + !t)3 − 27(1 + !t)2 + 168(1 + !t) + 20 − (1 − 27 + 168 + 20)
= lim .
!t→0 !t
Simplification of the numerator gives
117(!t) − 24(!t)2 + (!t)3

v(1) = lim ,
!t→0 !t
and when we divide numerator and denominator by !t ,
v(1) = lim [117 − 24(!t) + (!t)2 ] = 117.

!t→0
The particle is travelling 117 m/s at time t = 1 s.

A similar calculation at t = 5 s leads to an instantaneous velocity v(5) = −27 m/s. The
negative sign indicates that the particle is moving to the left. The positive velocity v(1) = 117
means that the particle is moving to the right at t = 1 s. We shall have much more to say about
velocity in Section 3.6 when we relate velocity and speed, and we also introduce acceleration.
For now, we simply want you to appreciate that instantaneous velocity is a rate of change; it is
the instantaneous rate of change of displacement with respect to time.
FIGURE 3.4 Cylinder Rate of Rainfall

catching rainfall
For our second situation in which to introduce rates of change, consider an open cylindrical
container placed outside a house. The depth D (measured in millimetres) of water in the
container during a rainstorm (Figure 3.4) is a function of time t (in hours), say D = f (t) . A
graph of this function might look like that in Figure 3.5. There is no water in the container at
time t = 0 when the storm begins. As the rain falls, the depth of water increases, and finally,
when the rain stops at time t = t , the depth of water remains at a constant level D thereafter.
Intuitively, rain is falling fastest when this curve is steepest, somewhere around the time indi-
cated as t̃ . When the curve is flat (just after t = 0 and just before t = t ), very little rain is falling.
We want to be more specific; we want to be able to say exactly how fast the rain is falling at any
FIGURE 3.5 Depth of rainwater in cylinder as a function of time

D
D
D
t t t
given time t0 . We proceed as we did in defining velocity. If we let !t be a small increment

of time, then the difference f (t0 + !t) − f (t0 ) is the number of millimetres of rain that falls
during the time interval !t after t0 . The quotient
f (t0 + !t) − f (t0 )

!t
with units of millimetres per hour is called the average rate of rainfall during the time interval.
The limit
f (t0 + !t) − f (t0 )
lim
!t→0 !t
is called the instantaneous rate of rainfall at time t0 . It is the instantaneous rate of change of
depth of water in the cylinder with respect to time, at time t0 . If its value were 5, say, then were
rain to fall at this rate over an extended period of time, the depth of water in the container would
increase by 5 mm every hour.
The rates of change in these two situations are typical of rates of change that occur in a
multitude of applications — applications from such diverse fields as engineering, physics, eco-
nomics, psychology, and medicine, to name a few. We now reformulate them in a mathematical
framework that allows us to introduce them in every area of applied mathematics. Suppose a
function f (x) is defined for all x , and x = a and x = a + h are two values of x ( a is taking
the place of t0 and h is replacing !t in the previous situations). The quotient
f (a + h) − f (a)
h
is called the average rate of change of f (x) with respect to x in the interval between a and
a + h . The limit of the quotient as h approaches zero,
f (a + h) − f (a)
lim ,
h→0 h
is called the instantaneous rate of change of f (x) with respect to x at x = a . It is also called
the derivative of f (x) at x = a .
DEFINITION 3.1
The derivative of a function f (x) with respect to x at a point x = a , denoted by f & (a) ,
is defined as
f (a + h) − f (a)
f & (a) = lim , (3.2a)
h→0 h
provided that the limit exists.
The following limit is equivalent to that in equation 3.2a but it avoids the introduction of h :
f (x) − f (a)
f & (a) = lim . (3.2b)
x→a x−a
The operation of taking the derivative of a function is called differentiation. We say that
we differentiate the function when we find its derivative. It would now be appropriate to use
Definition 3.1 to calculate derivatives of various functions at various points. Before doing so,
however, we feel that it is important to discuss the geometric interpretation of the derivative. It
will be prevalent in many applications.
Tangent Lines and the Geometric Interpretation of

the Derivative
FIGURE 3.6 Tangent line Algebraically, the derivative of a function is its instantaneous rate of change. Geometrically,
at one point may intersect curve at derivatives are intimately connected to tangent lines to curves, and most students of calculus
another point have an intuitive idea of what it means for a line to be tangent to a curve. Often, it is the idea
y
of touching. A line is tangent to a curve if it touches the curve at exactly one point. For curves
such as circles, ellipses, and parabolas, this notion is adequate, but in general it is unsatisfactory.
For instance, in Figure 3.6 we have drawn what would look like the tangent line to the curve
P (1, 1) y = x 3 at the point P (1, 1) . But this line intersects the curve again at the point Q(−2, −8) .
The tangent line at P does not touch the curve at precisely one point; it intersects the curve at
y = x3 x a second point.
You might reply that tangency is a local concept; whether the tangent line at (1, 1) in
Figure 3.6 intersects the curve at another point some distance from (1, 1) is irrelevant. True,
Q(−2, −8)
and perhaps we could remedy the situation by defining the tangent line at (1, 1) as the line that
touches the curve at (1, 1) and does not cross it there. Unfortunately, this definition does not
always work either. For instance, what is the tangent line to y = x 3 at (0, 0) (Figure 3.7)?
FIGURE 3.7 Tangent line The only reasonable line is the x -axis, but notice that the x -axis crosses the curve at (0, 0) . To
to y = x 3 at (0, 0) crosses curve the right of (0, 0) , the curve is above the tangent line, and to the left of (0, 0) , it is below the
y
tangent line. This happens quite often, as we shall see later.
y = x3
How then are we to define the tangent line to a curve? The idea that a tangent line touches
a curve is good, but it needs to be phrased properly. Consider defining what is meant by the
tangent line at the point P on the curve y = f (x) in Figure 3.8. When P is joined to another
point Q1 on the curve by a straight line l1 , certainly l1 is not the tangent line to y = f (x) at
x P . If we join P to a point Q2 on y = f (x) closer to P than Q1 , then l2 is not the tangent line
Tangent line at P either, but it is closer to it than l1 . A point Q3 even closer to P yields a line l3 that is even
at (0, 0) closer to the tangent line than l2 . Repeating this process over and over again leads to a set of
lines l1 , l2 , l3 , . . . , which get closer and closer to what we feel is the tangent line to y = f (x)
at P . We therefore define the tangent line to y = f (x) at P as the limiting position of these
lines as points Q1 , Q2 , Q3 , . . . get arbitrarily close to P .
FIGURE 3.8 Lines approaching tangent line to a curve
y = f (x)
P l1
Q1
Q2 l2
Q3
l3
x
The line in Figure 3.6 satisfies this definition; it is the limiting position of lines joining P
to other points on the curve which approach P . It is irrelevant whether this line intersects the
curve again at some distance from P . What is the tangent line to the curve y = x 3 at the origin
(0, 0) ? According to the above definition, the limiting position of lines joining (0, 0) to other
points on the curve is the x -axis; that is, the x -axis is tangent to y = x 3 at (0, 0) (Figure 3.7).
As mentioned earlier, the tangent line actually crosses from one side of the curve to the other at
(0, 0) . To the left of x = 0, the tangent line is above the curve, whereas to the right of x = 0,
it is below the curve.
Now that we understand what it means for a line to be tangent to a curve, let us make
the connection with derivatives. Suppose P (a, f (a)) is a point on the curve y = f (x) in
Figure 3.9. If Q(a + h, f (a + h)) is another point on the curve, then the quotient
f (a + h) − f (a)
h
in equation 3.2a is the slope of the line joining P and Q . As h → 0, point Q moves along the
curve toward P , and the line joining P and Q moves toward the tangent line at P . It follows
that the limit in equation 3.2a, the derivative f & (a) , is the slope of the tangent line to the curve
y = f (x) at x = a (Figure 3.10).
FIGURE 3.9 Lines used to find tangent line to a curve FIGURE 3.10 Tangent line to a curve has slope
f & (a)
y
y = f (x) y
y = f (x)
P (a, f (a))
P (a, f (a))
f (a + h) − f (a) Q (a + h, f (a + h))
slope =
h
x f &(a) = slope of the tangent line
x
To summarize, algebraically the derivative f & (a) of a function f (x) at x = a is its

instantaneous rate of change; geometrically, it is the slope of the tangent line to the graph of
f (x) at the point (a, f (a)) .
Let us now use equation 3.2a to calculate derivatives of some simple functions.
EXAMPLE 3.1
FIGURE 3.11 Tangent Find f & (1) if f (x) = x 2 .
2
line to parabola y = x at (1, 1)
SOLUTION According to equation 3.2a,
y
f (1 + h) − f (1) (1 + h)2 − 1
f & (1) = lim = lim
Tangent h→0 h h→0 h
line has
y = x2
slope 2
(1 + 2 h + h2 ) − 1 2 h + h2
= lim = lim = lim (2 + h) = 2.
h→0 h h→0 h h→0
(1, 1)
Algebraically, the instantaneous rate of change of f (x) = x 2 when x = 1 is equal to 2.
Geometrically, the slope of the tangent line to the curve y = x 2 at the point (1, 1) is 2
(Figure 3.11).
x
EXAMPLE 3.2
Find f & (2) if f (x) = x 3 − 12x .
FIGURE 3.12 Tangent
line may have slope equal to 0 SOLUTION Using equation 3.2a,
y f (2 + h) − f (2) [(2 + h)3 − 12(2 + h)] − (8 − 24)
f & (2) = lim = lim
h→0 h h→0 h
y = x 3 − 12x 6h2 + h3
= lim = lim (6h + h2 ) = 0.
h→0 h h→0
2√3
x This result is substantiated in Figure 3.12, where we see that the tangent line to y = x 3 − 12x
is horizontal (has zero slope) at x = 2.
(2, −16)
In Examples 3.1 and 3.2 we required the derivative for the function at only one value of a ,
and therefore set a in equation 3.2a equal to this value. An alternative, which is far more
advantageous, especially in an example where the derivative is required at a number of points, is
to evaluate f & (a) and then set a to its desired value, or values, later. For instance, in Example
3.1 we calculate that
f (a + h) − f (a) (a + h)2 − a 2
f & (a) = lim = lim
h→0 h h→0 h
2ah + h2
= lim = lim (2a + h) = 2a.
h→0 h h→0
This is the derivative of f (x) = x 2 at any value x = a . For a = 1, we obtain f & (1)
= 2(1) = 2. But it is also easy to calculate f & (a) at other values of a . For example,
f & (0) = 0, f & (−1) = −2, and f & (4) = 8.
We now carry this idea to its logical conclusion. The derivative of f (x) at a is denoted by
f & (a) . But what is a ? It is a specific value of x at which to calculate the derivative, and it can
be any value of x . Why not simply drop references to a , and talk about the derivative of f (x)
at values of x ? Following this suggestion, we denote by f & (x) the derivative of the function
f (x) at any value of x . With this notation, equation 3.2a is replaced by
f (x + h) − f (x)
f & (x) = lim (3.3)
h→0 h
for the derivative of f (x) at x . We call f & (x) the derivative function.
When a function is represented by the letter y , as in y = f (x) , another common notation
dy
for the derivative is . Be careful when using this notation. Do not interpret it as a quotient: dy
dx
and dx do not have separate meanings; it is one symbol representing an accumulation of all the
operations in equation 3.3. It is not therefore to be read as “ dy divided by dx .” Typographically,
dy
it is easier to print dy/dx rather than , and we will take this liberty whenever it is convenient
dx
to do so. But remember, dy/dx for the moment is not a quotient, it is one symbol representing
the limit operation in equation 3.3. We will change this in Section 4.12.
Sometimes it is more convenient to use parts of each of these notations and write
d
f (x).
dx
In this form we understand that d/dx means to differentiate with respect to x whatever follows
it, in this case f (x) . Let us use these new notations in calculating two more derivatives.
EXAMPLE 3.3
Find dy/dx if y = f (x) = (x − 1)/(x + 2) .
SOLUTION Using equation 3.3,

! "
dy f (x + h) − f (x) 1 x+h−1 x−1
= lim = lim − .
dx h→0 h h→0 h x+h+2 x+2
If we bring the terms in parentheses to a common denominator, the result is

# $
dy 1 (x + h − 1)(x + 2) − (x + h + 2)(x − 1)
= lim .
dx h→0 h (x + h + 2)(x + 2)
When we simplify the numerator, we find

# $
dy 1 3h
= lim
dx h→0 h (x + h + 2)(x + 2)
3 3
= lim = .
h→0 (x + h + 2)(x + 2) (x + 2)2
EXAMPLE 3.4
Find dv/dt if v = f (t) = 1/t .
SOLUTION In terms of variables v and t , equation 3.3 takes the form
! "
dv f (t + h) − f (t) 1 1 1
= lim = lim −
dt h→0 h h→0 h t +h t
t − (t + h) −1 1
= lim = lim = − 2.
h→0 t (t + h)h h→0 t (t + h) t
In Section 1.8 we proved that the inclination φ of a line l , where 0 ≤ φ < π , is related to
its slope m by the equation tan φ = m (equation 1.59). When l is the tangent line to a curve
y = f (x) at point (x0 , y0 ) (Figure 3.13), its slope is f & (x0 ) . Hence, the inclination φ of the
tangent line to a curve at a point (x0 , y0 ) is given by the equation
tan φ = f & (x0 ). (3.4)
FIGURE 3.13 Relating slope and inclination of a tangent line to a curve
y
Tangent line
at (x0, y0)
(x0, y0)
Slope = f &(x0)
φ
x
EXAMPLE 3.5
Find the inclination of the tangent line to the curve y = (x − 1)/(x + 2) at the point (4, 1/2) .
SOLUTION In Example 3.3, we calculated the slope of the tangent line to this curve at any
point as
dy 3
= .
dx (x + 2)2
According to equation 3.4, the inclination φ of the tangent line at (4, 1/2) is given by
3
tan φ = ⇒ φ = 0.083 radian.
(4 + 2 )2
In applications, independent and dependent variables of functions y = f (x) represent physical

quantities and have units associated with them. Units for the derivative f & (x) are units of y
divided by units of x . We see this from equation 3.3, where units of the quotient [f (x + h)
− f (x)]/ h are clearly units of y divided by units of x , and the limit of this quotient as h → 0
does not alter these units. For example, if x measures length in metres and y measures mass in
kilograms, the units of dy/dx are kilograms per metre.
EXERCISES 3.1
In Exercises 1–10 use equation 3.3 to find f & (x) . In Exercises 24–27 find the inclination of the tangent line to the curve
at the given point.
1. f (x) = x + 2 2. f (x) = 3x 2 + 5
24. y = x 2 at (1, 1)
3. f (x) = 1 + 2x − x 2 4. f (x) = x 3 + 2x 2
x+4 25. y = x 3 − 6x at (2, −4)
5. f (x) = x 4 + 4x − 12 6. f (x) =
x−5 ∗ 26. y = 1/x 2 at (2, 1/4)
x2 + 2
∗ 7. f (x) = ∗ 8. f (x) = x 2 (x + 2) ∗ 27. y = 1/(x + 1) at (0, 1)
x+3
3x − 2 x2 − x + 1
∗ 9. f (x) = ∗ 10. f (x) = In Exercises 28–32 find f & (x) .
4−x x2 + x + 1
√
∗ 28. f (x) = x 8 ∗ 29. f (x) = x+1
In Exercises 11–14 find the specified rate of change.
1 1
11. The rate of change of the circumference C of a circle with respect ∗ 30. f (x) = ∗ 31. f (x) = √
(x − 2)4 x−3
to its radius r
√
12. The rate of change of the area A of a circle with respect to its 32. f (x) = x x + 1
radius r
13. The rate of change of the area A of a sphere with respect to its
radius r In Exercises 33–34 find the specified rate of change.
14. The rate of change of the volume V of a sphere with respect to its ∗ 33. The rate of change of the radius r of a circle with respect to its
radius r area A
∗ 34. The rate of change of the volume V of a sphere with respect to its
Answer Exercises 15–19 by drawing a graph of the function. Do not area A
calculate the required derivative.
35. Find the angle between the tangent lines to the curves y = x 2 and
15. What is f & (x) if f (x) = 2x − 4? x = y 2 at their point of intersection in the first quadrant.
16. What is f & (x) if f (x) = mx + b , where m and b are constants? ∗ 36. Find f & (x) if f (x) = |x| .
17. What is f & (0) if f (x) = x 2 ? ∗ 37. A sphere of radius R has a uniform charge distribution of ρ
18. What is f & (1) if f (x) = (x − 1)2 ? coulombs per cubic metre. The electrostatic potential V at a distance
r from the centre of the sphere is defined by
19. What is f & (0) if f (x) = x 1/3 ?
 ρ
 2 2
 6%0 (3R − r ), 0 ≤ r ≤ R ,

In Exercises 20–23 find the equation of the tangent line to the curve at V = f (r) =
the given point. 
 R3ρ
 , r > R,
3%0 r
2 2
20. y = x + 3 at (1, 4) 21. y = 3 − 2x − x at (4, −21)
2 where %0 is a constant. Draw a graph of this function. Does f (r)
∗ 22. y = 1/x at (2, 1/4)
appear to have a derivative at the surface of the sphere? That is, does
∗ 23. y = (x + 1)/(x + 2) at (0, 1/2) f & (R) exist? Prove your conjecture using Definition 3.1.
∗ 38. Repeat Exercise 37 for the magnitude E of the electrostatic field: ∗∗ 39. Find f & (x) if f (x) = x 1/3 .
 ρr
 0 ≤ r ≤ R,
 3%0 ,

E = f (r) = ∗∗ 40. Let f (x) be a function with the property that f (x + z) =

 ρR 3 f (x)f (z) for all x and z , and be such that f (0) = f & (0) = 1.
 , r > R.
3%0 r 2 Prove that f & (x) = f (x) for all x .
3.2 Rules for Differentiation

Since calculus plays a key role in many branches of applied science, we need to differentiate many
types of functions: polynomial, rational, trigonometric, exponential, and logarithm functions, to
name a few. To use equation 3.3 each time would be extremely laborious. Fortunately, however,
we can develop a number of rules for taking derivatives that eliminate the necessity of using the
definition each time. We state each of these formulas as a theorem.
THEOREM 3.1
If f (x) = c , where c is a constant, then f & (x) = 0.
PROOF By equation 3.3,
f (x + h) − f (x) c−c
f & (x) = lim = lim = 0.
h→0 h h→0 h
In short, the derivative of a constant function is zero;
d
(c) = 0. (3.5)
dx
THEOREM 3.2
If f (x) = x , then f & (x) = 1.
PROOF With equation 3.3,
f (x + h) − f (x) (x + h) − x h
f & (x) = lim = lim = lim = 1.
FIGURE 3.14 Derivatives h→0 h h→0 h h→0 h
of functions y = c and y = x
In short,
y
d
y=x (x) = 1. (3.6)
dx
c y=c
With the graphs of the functions f (x) = c and f (x) = x in Figure 3.14, we can see the
results of equations 3.5 and 3.6. The tangent line to y = c always has slope zero, whereas the
tangent line to y = x always has slope equal to 1.
x
THEOREM 3.3
If f (x) = x n , where n is a positive integer, then f & (x) = nx n−1 .
PROOF Equation 3.3 gives
f (x + h) − f (x) (x + h)n − x n
f & (x) = lim = lim .
h→0 h h→0 h
If we expand (x + h)n by means of the binomial theorem (an alternative proof is given in
Exercise 35), we have
# $
& 1 n n−1 n(n − 1) n−2 2 n−1 n n
f (x) = lim x + nx h + x h + · · · + nxh +h −x .
h→0 h 2
The first and last terms in brackets cancel, and dividing h into the remaining terms gives
# $
n−1 n(n − 1) n−2 n−1
&
f (x) = lim nx + x h + ··· + h = nx n−1 .
h→0 2
In short,
d n
(x ) = nx n−1 . (3.7)
dx
This is called the power rule for differentiation. Although we have proved the power rule
only for n a positive integer, it is in fact true for every real number n . We will assume that
equation 3.7 can be used for any real number n , and will prove this more general result in Section
3.11. For example,
d 3 d ) 1/3 * 1
(x ) = 3x 2 and x = x − 2 /3 .
dx dx 3
THEOREM 3.4
If g(x) = cf (x) , where c is a constant, and f (x) has a derivative, then
g & (x) = cf & (x). (3.8a)
PROOF By equation 3.3 and Theorem 2.1,
g(x + h) − g(x) cf (x + h) − cf (x)

g & (x) = lim = lim
h→0 h h→0 h
f (x + h) − f (x)
= c lim = cf & (x).
h→0 h
Thus, for y = f (x) , we may write that
d dy
(cy) = c . (3.8b)
dx dx
THEOREM 3.5
If p(x) = f (x) + g(x) , where f (x) and g(x) have derivatives, then
p & (x) = f & (x) + g & (x). (3.9a)

PROOF Equation 3.3 gives
p(x + h) − p(x)
p & (x) = lim
h→0 h
[f (x + h) + g(x + h)] − [f (x) + g(x)]

= lim
h→0 h
# $
f (x + h) − f (x) g(x + h) − g(x)
= lim +
h→0 h h
= f & (x) + g & (x).
In short, if we set u = f (x) and v = g(x) ,
d du dv
(u + v) = + , (3.9b)
dx dx dx
or, in words, the derivative of a sum is the sum of the derivatives.
We now use these formulas to calculate derivatives in the following examples.
EXAMPLE 3.6
Find dy/dx if
x 4 − 6x 2
(a) y = x4 (b) y = 3x 6 − x −2 (c) y =
3x 3
SOLUTION
(a) By power rule 3.7,
dy
= 4x 3 .
dx
(b) Equation 3.9 allows us to differentiate each term separately; by equations 3.8 and 3.7
it follows that
dy
= 3(6x 5 ) − (−2x −3 ) = 18x 5 + 2x −3 .
dx
(c) If we write y in the form y = (x/3) − 2x −1 , we can proceed as in part (b):
dy 1 1 2
= (1) − 2(−x −2 ) = + 2 .
dx 3 3 x
EXAMPLE 3.7
If f (x) = 3x 4 − 2, evaluate f & (1) .
SOLUTION Since f & (x) = 12x 3 , it follows that f & (1) = 12. Geometrically, 12 is the slope
of the tangent line to the curve y = 3x 4 − 2 at the point (1, 1) in Figure 3.15. Algebraically,
the result implies that at x = 1, y changes 12 times as fast as x .
FIGURE 3.15 Tangent line to y = 3x 4 − 2 at (1, 1)
y = 3x 4 − 2
(1, 1)
−2
EXAMPLE 3.8
Find the equation of the tangent line to the curve y = f (x) = x 3 + 5x at the point (1, 6) .
SOLUTION Since f & (x) = 3x 2 + 5, the slope of the tangent line to the curve at (1, 6) is
f & (1) = 3(1)2 + 5 = 8.
Using point-slope formula 1.13 for a straight line, we obtain for the equation of the tangent line
FIGURE 3.16 Normal at (1, 6)
line to a curve is perpendicular to
y − 6 = 8(x − 1) or 8x − y = 2 .
tangent line
y
Tangent line
The line through (1, 6) perpendicular to the tangent line in Figure 3.16 is called the normal
(1, 6)
line to the curve at (1, 6) . Since two lines are perpendicular only if their slopes are negative
Normal line reciprocals (see equation 1.15), the normal line at (1, 6) must have slope −1/8. The equation
of the normal line to y = x 3 + 5x at (1, 6) is therefore
x
1
y − 6 = − (x − 1) or x + 8y = 49.
8
y = x 3 + 5x
EXAMPLE 3.9
Find, accurate to four decimal places, points on the curve y = x 4 − 4x 3 − x 2 + x at which
the slope of the tangent line is −1.
SOLUTION It is always wise to have an idea about how many solutions to expect for a
problem and approximate values for them. (We might call this “ball-parking” the answer.) In
this problem, a plot of the curve in Figure 3.17 should provide this information. However, we
must be careful in trying to estimate where the slope of the graph is −1. Because scales are
different on the axes, a line with slope −1 is not inclined at π/4 radians with respect to the
negative x -axis. The line in the figure has slope −1 for reference. The graph suggests three
points at which the tangent line is parallel to this line: one just to the left of x = 0, one just to
the right of x = 0, and one near x = 3.
FIGURE 3.17 Finding points on a curve where tangent line has given slope
40 y
Line with slope = −1
30
20
10
x
−2 −1 1 2 3 4
−10
−20
−30
To confirm this, we set the slope of the tangent line to the curve equal to −1:
dy
−1 = = 4x 3 − 12x 2 − 2x + 1 *⇒ 2(2x 3 − 6x 2 − x + 1) = 0.
dx
Six-digit solutions of this equation are −0.459 261, 0.350 168, and 3.109 09. To confirm
−0.4593, 0.3502, and 3.1091 as solutions correct to four decimal places, we calculate g(x) =
2x 3 − 6x 2 − x + 1 at the following values:
g(−0.459 35) = −5.1 × 10−4 , g(−0.459 25) = 6.5 × 10−5 ,
g(0.350 15) = 8.0 × 10−5 , g(0.350 25) = −3.7 × 10−4 ,
g(3.109 05) = −9 × 10−4 , g(3.109 15) = 1.1 × 10−3 .
The zero intermediate value theorem confirms the four-decimal-place approximations. Us-
ing the six-decimal-place approximations, and rounding results to four decimal places, cor-
responding points on the curve are (−0.4593, −0.2382) , (0.3502, 0.0708) , and (3.1091,
−33.3326) .
Increment Notation
In Section 3.1, we have used the letter h to represent a small change in x when calculating
derivatives. An alternative notation that is sometimes more suggestive was used in the introduc-
tory examples of that section. A small change in x , called an increment in x , is often denoted
by !x . It is pronounced “delta x ” since ! is the capital letter delta in the Greek alphabet.
When x is given an increment !x in the function y = f (x) , the corresponding change or
increment in y is denoted by !y . It is equal to
!y = f (x + !x) − f (x). (3.10)
With this notation, equation 3.3 takes the form

!y
f & (x) = lim . (3.11a)
!x→0 !x
The notation dy/dx for the derivative fits very nicely with increment notation,
dy !y
= lim ; (3.11b)
dx !x→0 !x
the derivative of y with respect to x is the limit of the change in y divided by the change in x
as the change in x approaches zero. We use this notation in the following example.
EXAMPLE 3.10
Use equation 3.11 to calculate the derivative of y = f (x) = 3x 2 − 2x , and check your answer
using the differentiation rules discussed earlier in this section.
SOLUTION Since
!y = f (x + !x) − f (x)
= [3(x + !x)2 − 2(x + !x)] − [3x 2 − 2x ]
= 3[x 2 + 2x !x + (!x)2 ] − 2x − 2!x − 3x 2 + 2x
= !x(6x − 2 + 3!x),
equation 3.11 gives
dy !y !x(6x − 2 + 3!x)
= lim = lim = lim (6x − 2 + 3!x) = 6x − 2.
dx !x→0 !x !x→0 !x !x→0
Power rule 3.7, and rules 3.8 and 3.9 for differentiation of 3x 2 − 2x , also yield 6x − 2.
Figure 3.18a shows parts of two transmission lines, one straight, and the other in the shape
of a parabola as it circumnavigates a lake. Numbers represent lengths in kilometres but
no attempt has been made to adhere to a scale. The two transmission lines are to be joined
by a third that should be as short as possible. We are to find its length and where it should
join the existing lines.
FIGURE 3.18a Schematic FIGURE 3.18b

for finding shortest distance be-
y
tween two transmission lines
Parabolic transmission line y = 30x2
Q(c, d)
30 −100
1 x
P(a, b)
100 50 x + 2y = −100
40 −50
Straight transmission line Town Town
SOLUTION We begin by choosing a coordinate system to facilitate analysis. Since

equations for straight lines are straightforward in any coordinate system, we choose axes
to simplify the equation of the parabola (Figure 3.18b). In this coordinate system, the
equation of the parabola is y = 30x 2 and the equation of the line is x + 2y + 100 = 0.
Our problem now is to find the shortest distance between the line and the parabola (which
must be a straight line distance) and the points on these curves at which the line segment
should be drawn. Let the required points on the straight line and parabola have coordinates
P (a, b) and Q(c, d) , respectively. This is an important step in our analysis. When
unknowns are required, identify them, in this case give names to the coordinates of the
unknown points. In addition, do not designate either one of the points by coordinates
The shortest transmission line joining the existing lines is approximately 44.7 km long.
Since the length+of the line segment joining P to the y -intercept of the straight trans-
mission line is (8003/400)2 + (−50 + 31 997/800)2 ≈ 22.4, the new transmission
line should begin 62.4 km west of the town and be perpendicular to the existing straight
transmission line.
EXERCISES 3.2
In Exercises 1–20 find f & (x) . signal x(t) = cos (1000π t 2 + 100π t) , where t is time in seconds.
(a) Plot x(t) for 0 ≤ t ≤ 0.1. Notice how the frequency
1. f (x) = 2x 2 − 3 2. f (x) = 3x 3 + 4x + 5 increases as t increases.
3. f (x) = 10x 2 − 3x 4. f (x) = 4x 5 − 10x 3 + 3x (b) The frequency in hertz of the signal, at any given time, is
defined as the derivative of the phase 1000π t 2 + 100π t at
5. f (x) = 1/x 2 6. f (x) = 2/x 3 that time, divided by 2π . Find the frequencies of the signal
1 3 at t = 0 and t = 0.1.
7. f (x) = 5x 4 − 3x 3 + 1/x 8. f (x) = − +
2x 2 x4 ∗ 29. The general formula for a chirp signal (see Exercise 28) is x(t) =
1 1 cos (αt 2 + βt + φ) , where α > 0, β > 0, and φ are constants.
9. f (x) = x 10 − 10. f (x) = 5x 4 + The derivative of αt 2 + βt + φ , divided by 2π , is the instantaneous
x 10 4x 5
frequency of the signal. If the signal begins at time t1 and ends at time
1 √ t2 , what is the difference in initial and final frequencies of the signal?
11. f (x) = 5x −4 + 12. f (x) = x
4 x −5 ∗ 30. At what point(s) on the curve y = x 3 + x 2 − 22x + 20 does the
3 2 1 tangent line pass through the origin?
13. f (x) = +√ 14. f (x) = + x 3/ 2
x2 x x 3/ 2 ∗ 31. At what point(s) on the parabola y = x 2 does the normal line pass
15. f (x) = 2x 1/ 3
− 3x 2 /3
16. f (x) = π x π through the point (2, 5) ? Can you suggest an application of this result?
2 2 ∗ 32. Show√that the sum of√the x - and y -intercepts of the tangent line to
√
17. f (x) = (x + 2) 18. f (x) = (4x 6 − x 2 )/x 5 the curve x + y = a is always equal to a .
19. f (x) = x 5/3 − x 2/3 + 3 20. f (x) = (2x + 5)3 ∗ 33. Show that the line segment cut from the tangent line at a point P
on the curve y = 1/x by the coordinate axes is bisected by P .
∗ 34. A hill is best described by a parabola containing the three points in
In Exercises 21–24 find equations for the tangent and normal lines to the figure below, all measurements in metres. A transmitter 30 m high
the curve at the point indicated. In each case, draw the curve and lines. stands at the point (−120, 0) . What is the closest point to the base
of the hill on the positive x -axis that a receiver can detect the signal
21. y = x 2 − 2x + 5 at (2, 5) unobstructed by the hill?
√
22. y = x + 5 at (4, 7) y
Transmitter
23. y = 2x 3 − 3x 2 − 12x at (2, −20) 20
√ 30 Hill
∗ 24. x = y + 1 at (3, 8)
x
−120 −100 100
4 3 2
∗ 25. Find the points on the curve y = x /4 − 2x /3 − 19x /2 + 22x
at which the slope of the tangent line is 2. ∗ 35. Give an alternative derivation of power rule 3.7 based on the iden-
tity
∗ 26. Show that the x -intercept of the tangent line at any point (x0 , y0 )
on the parabola y = ax 2 bisects that part of the x -axis between x = 0 a n − bn = (a − b)(a n−1 + a n−2 b + · · · + abn−2 + bn−1 ).
and x = x0 .
∗ 36. Find all pairs of points on the curves y = x 2 and y = −x 2 + 2x
3 2
∗ 27. Draw a graph of the function x = f (t) = t − 8t . Find the − 2 that share a common tangent line.
value(s) of t at which the tangent line to this curve is parallel to the line
∗ 37. Show that the tangent lines at any two points P and Q on the
x = 6t − 3.
parabola y = ax 2 + bx + c intersect at a point that is on the vertical
∗ 28. A chirp signal in acoustics is a signal whose frequency changes line halfway between P and Q (provided that neither P nor Q is at
linearly from a low value to a high value. For example, consider the the vertex of the parabola).
d when a and b are constants, and n > 1 is an integer.

∗ 38. Find a formula for |x|n when n > 1 is an integer.
dx
∗∗ 39. Prove that
d ∗∗ 40. Find the two points on the curve y = x(1 + 2x − x 3 ) that share
(ax + b)n = an(ax + b)n−1
dx a common tangent line.
3.3 Differentiability and Continuity

Many functions fail to have a derivative at isolated points. For example, consider the function
FIGURE 3.20 Function
f (x) = |x| in Figure 3.20. It is clear that for x > 0, f & (x) = 1, and for x < 0, f & (x) = −1.
f (x) = |x| has no derivative at At x = 0, however, there is a problem. If f (x) is to have a derivative at x = 0, it must be
x=0 given by
y f (0 + h) − f (0) |h|
lim = lim .
y = |x| h→0 h h→0 h
But this limit does not exist since
|h| |h|
lim = −1 and lim = 1.
x h→0− h h→0+ h
Consequently, f (x) = |x| does not have a derivative at x = 0.

The same conclusion can be drawn at any point at which the graph of a function takes an
abrupt change in direction. Such a point is often called a corner. As a result, the function in
Figure 3.21 does not have a derivative at x = a , b , c , or d . It is also true that a function cannot
have a derivative at a point where the function is discontinuous (see, e.g., the discontinuities in
Figure 2.27). This result is an immediate consequence of the following theorem.
FIGURE 3.21 Points
where a continuous function has
no derivative
THEOREM 3.6
If a function has a derivative at x = a , then the function is continuous at x = a .
y
PROOF To prove this theorem we show that existence of f & (a) implies that limx→a f (x) =
f (a) , the condition that defines continuity of f (x) at x = a (Definition 2.1 in Section 2.4).
We show that limx→a [f (x) − f (a)] = 0:
# $ # $
a b c d x f (x) − f (a) f (x) − f (a) , -
lim [f (x) − f (a)] = lim · (x − a) = lim lim (x − a) ,
x→a x→a x−a x→a x−a x→a
provided both limits on the right exist. Obviously, the second limit has value zero, and the first
is definition 3.2b for f & (a) , which we have assumed exists. Thus,
lim [f (x) − f (a)] = f & (a) · 0 = 0.

x→a
The following result is logically equivalent to Theorem 3.6; it is called the contrapositive
of the theorem. Although equivalent to Theorem 3.6, we shall designate it as a corollary.
COROLLARY 3.6.1
If f (x) is discontinuous at x = a , then f & (a) does not exist.
Students are often heard to say that having a derivative is equivalent to having a tangent line
(because derivatives are slopes of tangent lines); that is, a function f (x) has derivative f & (a)
at x = a if and only if the graph of f (x) has a tangent line at x = a . This is not quite true.
For example, consider the function f (x) = x 1/3 (Figure 3.22). At (0, 0) , the tangent line to
the graph is the y -axis (use the definition of tangent line in Section 3.1 to convince yourself of
this). The derivative of f (x) is f & (x) = (1/3)x −2/3 ; it does not exist at x = 0. Thus, we
have a tangent line, but no derivative. The reason is that the tangent line is a vertical line, and
vertical lines do not have slopes. What can we say, then? Two things:
) *
1. If f & (a) exists, then y = f (x) has a tangent line at a, f (a) with slope f & (a) .
) *
2. If f & (a) does not exist, then y = f (x) either has a vertical tangent line at a, f (a) or
does not have a tangent line when x = a .
FIGURE 3.22 Tangent line to y = x 1/3 at (0, 0) is the y -axis
y = x1/ 3
Tangent line
at (0, 0)
We introduced this section by showing that the derivative of |x| is 1 when x > 0; it is −1
when x < 0; and it does not exist at x = 0. These can be combined into the simple formula
d |x|
|x| = . (3.12)
dx x
It is straightforward to generalize equation 3.12 and obtain the derivative of |f (x)| at any point
at which f (x) ,= 0 and f & (x) exists. When f (x) > 0, we may write
d d
|f (x)| = f (x) = f & (x).
dx dx
On the other hand, when f (x) < 0, we have
d d
|f (x)| = [−f (x)] = −f & (x).
dx dx
Both of these results are contained in the one equation
d |f (x)| &
|f (x)| = f (x). (3.13)
dx f (x)
This is the derivative of |f (x)| at any point at which f (x) ,= 0. The exceptional case when
f (x) = 0 is discussed in Exercise 28.
Right- and Left-Hand Derivatives

When a function f (x) is defined on a closed interval b ≤ x ≤ c , equation 3.3 can be used to
calculate f & (x) only at points in the open interval b < x < c . For instance, it is impossible to
evaluate
f (b + h) − f (b)
f & (b) = lim
h→0 h
since f (x) is not defined for x < b and therefore f (b + h) is not defined for h < 0. When
a function f (x) is defined only to the right of a point, we define a right-hand derivative at the
point; and when f (x) is defined only to the left of a point, we define its left-hand derivative.
DEFINITION 3.2
The right-hand derivative of f (x) with respect to x is defined as
f (x + h) − f (x)
f+& (x) = lim+ , (3.14a)
h→0 h
provided that the limit exists. The left-hand derivative of f (x) is
f (x + h) − f (x)
f−& (x) = lim− , (3.14b)
h→0 h
if the limit exists.
The left-hand derivative at a point is not confined to the situation where a function is defined
only to the left of the point; nor is the right-hand derivative restricted to the situation where the
function is defined only to the right of the point. We may consider left- and right-hand derivatives
at any point, as well as a “full” derivative. Obviously, when a function has a derivative at a point
x , its right- and left-hand derivatives both exist at x and are equal to f & (x) . It is possible,
however, for a function to have both a left- and a right-hand derivative at a point but not a
derivative. An example of this is the absolute value function f (x) = |x| at x = 0 (Figure
3.20). Its right-hand derivative at x = 0 is equal to 1 and its left-hand derivative there is −1.
When a function has a derivative at x = a , we say that it is differentiable at x = a . When
it has a derivative at every point in some interval, we say that it is differentiable on that interval.
In the event that the interval is closed, b ≤ x ≤ c , we understand that derivatives at x = b and
x = c mean right- and left-hand derivatives, respectively.
EXAMPLE 3.11
What is the derivative of the Heaviside function h(x − a) introduced in Section 2.5?
SOLUTION The tangent line to the graph of the function (Figure 3.23) is horizontal at every
point except x = a , where the function is undefined. In other words, h& (x − a) = 0 except at
the discontinuity x = a . Right- and left-hand derivatives do not exist at x = a either.
FIGURE 3.23 Tangent line to y = h(x − a) is horizontal except at x = a
y = h (x − a)
1
a x
By examining the definition of the derivative of h(x−a) , a very useful result for applications
emerges. According to equation 3.3,
h(x + !x − a) − h(x − a)
h& (x − a) = lim .
!x→0 !x
If we change variables in this limit by setting % = −!x ,
h(x − % − a) − h(x − a) h(x − a) − h(x − a − %)
h& (x − a) = lim = lim .
−%→0 −% %→ 0 %
But according to equation 2.9, this is the definition of the Dirac-delta function δ(x − a) . In
other words, we may write that h& (x − a) = δ(x − a) .
EXAMPLE 3.12
Is the function .
g(x) = x 2 sin (1/x), x ,= 0
0, x =0
differentiable at x = 0?
SOLUTION We encountered the function x 2 sin (1/x) in Example 2.9 of Section 2.1 and
drew its graph (except for the point at x = 0) in Figure 2.10a. According to equation 3.3, the
derivative of g(x) at x = 0 is
g(0 + h) − g(0)
g & (0) = lim ,
h→0 h
provided that the limit exists. When we substitute from the definition of g(x) for g(h) and
g(0) , this limit takes the form
! "
& h2 sin (1/ h) − 0 1
g (0) = lim = lim h sin =0
h→0 h h→ 0 h
(see Exercise 51 in Section 2.1).
Angle Between Intersecting Curves

The angle θ between two curves that intersect at a point (x0 , y0 ) (Figure 3.24) is defined
as the angle between the tangent lines to the curves at (x0 , y0 ) . This can be calculated using
formula 1.60 once slopes of the tangent lines are known.
In the event that θ = π/2 (Figure 3.25), the curves are said to be orthogonal or perpen-
dicular at (x0 , y0 ) . Should θ = 0, the curves are said to be tangent at (x0 , y0 ) (Figure 3.26).
FIGURE 3.24 Angle FIGURE 3.25 Orthogonal FIGURE 3.26 Tangent

between curves that intersect curves at (x0 , y0 ) curves at (x0 , y0 )
Tangent y y
y
lines to
curves
Tangent
line to
(x0, y0) both
(x0, y0)
(x0, y0) curves at
(x0, y0)
x x
x
EXAMPLE 3.13
Find the angle between the line x + 2y = 5 and the curve y = x 3 + 31 at their point of
intersection.
FIGURE 3.27 Finding angle between curves at their point of intersection
y
31 y = x3 + 31
5/2 x + 2y = 5
5 x
SOLUTION A quick diagram (Figure 3.27) allows us to ballpark the location of the point of
intersection of the curves. To find it, we solve the equation of the line for y = (5 − x)/2 and
equate it to x 3 + 31,
5−x
x 3 + 31 = *⇒ 2x 3 + x + 57 = 0.
2
According to the rational root theorem of Section 1.2, the only possible rational solutions of this
equation are
1 3 19 57
±1, ±3, ±19, ±57, ± , ± , ± , ± .
2 2 2 2
Figure 3.27 makes it clear that we should only try −3. It is indeed a solution, and the only one.
The point of intersection of the curves is therefore (−3, 4) . The slope of the line is −1/2, and
to find the slope of the cubic, we calculate that dy/dx = 3x 2 . The slope at (−3, 4) is 27.
Using formula 1.60, the acute angle between the curves at their point of intersection is
/ /
/ 27 − (−1/2) /
θ = Tan −1 / /
/ 1 + 27(−1/2) / = 1.1 radians.
It looks much larger than this. Why?
EXERCISES 3.3
In Exercises 1–6 determine whether the function has a right-hand In Exercises 10–13 determine whether the statement is true or false.
derivative, a left-hand derivative, and a derivative at the given value of x .
10. If a function f (x) has
) a derivative
* at x = a , then its graph has a
tangent line at the point a, f (a) .
1. f (x) = |x − 5| at x = 5
) *
11. If a function f (x) has a tangent line at a point a, f (a) , then it
2. f (x) = x 3/2 at x = 0 has a derivative at x = a .
) *
3. f (x) = |x − 5|3 at x = 5 12. If a function does not have a tangent line at a, f (a) , then it does
not have a derivative at x = a .
4. f (x) = sgn x at x = 0 (See Exercise 47 in Section 2.4.) 13. If a function f (x) does not) have a*derivative at x = a , then it
does not have a tangent line at a, f (a) .
5. f (x) = (x 2 − 1)/(x − 1) at x = 1
∗ 6. f (x) = -x. at x = 1 (See Exercise 68 in Section 1.5.) In Exercises 14–16 show algebraically that f & (0) does not exist. Draw
a graph of the function.
∗ 7. Does it make any difference in Example 3.11 if we define h(a) = 0?
∗ 14. f (x) = x 1/3
∗ 8. Does it make any difference in Example 3.11 if we define h(a) = 1?
∗ 15. f (x) = x 2/3
∗ 9. Does it make any difference in Exercise 4 if sgn x does not have a
value at x = 0? ∗ 16. f (x) = x 1/4
(x, y) . Letters x and y already identify variable points on the straight line and parabola;
to identify fixed points with the same letters would lead to confusion. Now the big step.
Of the infinite number of line segments joining points P on the straight transmission
line to points Q on the parabolic transmission line, which is shortest? Our intuition tells
us that the shortest line segment is the one that is simultaneously perpendicular to both
transmission lines. Recognizing this, we could attempt to ballpark positions of P and
Q by plotting the existing transmission lines with equal scales on the axes (otherwise,
lengths and angles are not what they seem). We have done this in Figure 3.19 and drawn
a line that seems perpendicular to both transmission lines. It intersects the parabola very
close to the origin so that coordinates of Q should both be close to zero. Coordinates of
P would appear to be close to (−20, −40) .
Using slopes we can now set up equations for the unknown coordinates. Since the
FIGURE 3.19 Scale dia-
gram of transmission lines
slope of line segment P Q is (d − b)/(c − a) , and that of x + 2y + 100 = 0 is −1/2,
perpendicularity requires
y d−b
= 2.
−20 −10 x c−a
−10
The slope of the tangent line to the parabola at any value of x is 60x , so that the slope
−20 of the tangent line at Q is 60c . Since the slope of the normal line is −1/(60c) , and this
−30 must be parallel to line segment P Q , we must have
−40
−50
d−b 1
=− .
c−a 60c
We now have two equations in the four unknown coordinates. Two more equations can
be obtained by using the fact that P is on the line x + 2y + 100 = 0 and Q is the on the
parabola y = 30x 2 ,
a + 2b + 100 = 0, d = 30c2 .
When we equate right sides of the first two equations, we obtain
1 1
2 = − *⇒ c =− .
60c 120
This now implies that
! "2
1 1
d = 30 − = .
120 480
We substitute these values for c and d into the first equation,
! "
1 1 3
−b = 2 − − 2a *⇒ 2a − b = .
480 120 160
If we double this equation and add it to a + 2b = −100, we obtain
3 8003 8003 3 31 997
5a = − 100 *⇒ a =− *⇒ b =− − =− .
80 400 200 160 800
Thus, the required point on the straight transmission line is (−8003/400, −31 997/800) ≈
(−20, −40) and the point on the parabolic line is (−1/120, 1/480) . These agree with
our predictions for positions of P and Q . The length of the line segment joining these
points is
0! "2 ! "2
1 8003 1 31 997
/P Q/ = − + + + ≈ 44.7.
120 400 480 800
In Exercises 17–18 find the angle (or angles) between the curves at their ∗ 29. If limx→∞ f (x) = L (so that y = L is a horizontal asymptote
point (or points) of intersection. for the graph of the function), is it necessary that limx→∞ f & (x) = 0?
∗ 30. The Green’s function for displacements of a taut string with con-
∗ 17. y = x 2 , x+y =2 ∗ 18. y = x 2 , y = 1 − x2
stant tension τ and length L , and ends fixed at x = 0 and x = L on
the x -axis is
In Exercises 19–20 determine whether the curves are orthogonal. 1
G(x; X) = [x(L − X)h(X − x) + X(L − x)h(x − X)],
19. x − 2y + 1 = 0, y =2−x 2 Lτ
20. y = 3 − x 2 , 4y − 7 = x 2 where h(x − X) is the Heaviside function in Section 2.5. Think of

G(x; X) as a function of x that also depends on X where X can have
any value between 0 and L . It is the displacement at position x in the
In Exercises 21–22 show that the curves are tangent at the indicated string if a unit force in the positive y -direction is applied to the string
point. at position X .
(a) Draw a graph of G(x; L/2) . Is it symmetric about
21. y = x − 2x 2 , y = x 3 + 2x at (−1, −3) x = L/2?
22. y = x 3 , y = x 2 + x − 1 at (1, 1) (b) Draw a graph of G(x; X) when L/2 < X < L .
∗ 23. (a) The equation x + 2y = C1 , where C1 is a constant, rep- (c) Show algebraically that G(x; X) is continuous for 0 ≤
resents a one-parameter family of curves. For each value x ≤ L , except for a removable discontinuity at x = X .
of C1 , the parameter, a different curve in the family is ob- (d) Show that dG/dx is continuous for all x ,= X , and has a
tained. Draw curves corresponding to C1 = −2, −1, 0, 1, jump discontinuity at x = X of −1/τ .
2.
∗ 31. Are f+& (a) and lim f & (x) the same? Illustrate with h(x − a) .
(b) Draw a few curves from the one-parameter family 2x−y = x→a +
C2 on the graph in part (a). ∗ 32. Is the function
(c) Two families of curves are said to be orthogonal trajecto- 1
ries if every curve from one family intersects every curve x sin (1/x), x ,= 0
f (x) =
from the other family orthogonally. Are the families in parts 0, x =0
(a) and (b) orthogonal trajectories? differentiable at x = 0?
∗ 24. For what value of k are the one-parameter families 2x − 3y = C1 ∗ 33. For what values of the real number n is the function
and x + ky = C2 orthogonal trajectories? 1 n
x sin (1/x), x ,= 0
f (x) =
∗ 25. Draw a graph of the function f & (x) when f (x) = -x. is the floor 0, x =0
function (Example 2.22 in Section 2.4).
differentiable at x = 0?
∗ 26. Is the function f (x) = x|x| differentiable at x = 0? ∗∗ 34. Is the function
∗ 27. Find f & (x) if f (x) = |x| + |x − 1| . Draw graphs of f (x) and .
f & (x) . f (x) = x 2 , x a rational number
0, x an irrational number
∗ 28. If f (x) is a differentiable function, does |f (x)| have a derivative
at points where f (x) = 0? Hint: Draw some pictures. differentiable at x = 0?
3.4 Product and Quotient Rules

In this section we add two more formulas to those of Section 3.2 for calculating derivatives. The
first is a rule for differentiating a function that is the product of two other functions.
THEOREM 3.7
If p(x) = f (x)g(x) , where f (x) and g(x) are differentiable, then
p & (x) = f (x)g & (x) + f & (x)g(x). (3.15a)

p(x + h) − p(x)
p & (x) = lim
h→0 h
f (x + h)g(x + h) − f (x)g(x)
= lim .
h→0 h
To organize this quotient further, we add and subtract the quantity f (x + h)g(x) in the numer-
ator:
. 2
& [f (x + h)g(x + h) − f (x + h)g(x)] + [f (x + h)g(x) − f (x)g(x)]
p (x) = lim
h→0 h
# $
g(x + h) − g(x) f (x + h) − f (x)
= lim f (x + h) + g(x)
h→0 h h
= f (x)g & (x) + g(x)f & (x).
In taking the limit of the first term, we have used the fact that lim h→0 f (x + h) = f (x) ,
which follows from continuity of f (x) (see Theorem 3.6 and Exercise 46 in Section 2.4).
This result is called the product rule for differentiation. If we set u = f (x) and v = g(x) ,
then the product rule may also be expressed in the form
d dv du
(uv) = u +v . (3.15b)
dx dx dx
For a change, we use increment notation to prove the quotient rule. Use of h in place of
!x works equally well.
THEOREM 3.8
If p(x) = f (x)/g(x) , where f (x) and g(x) are differentiable, then
g(x)f & (x) − f (x)g & (x)

p & (x) = . (3.16a)
[g(x)]2
PROOF Using equation 3.11 yields
p(x + !x) − p(x)

p & (x) = lim
!x→0 !x
# $
1 f (x + !x) f (x)
= lim −
!x→0 !x g(x + !x) g(x)
f (x + !x)g(x) − g(x + !x)f (x)

= lim .
!x→0 !xg(x)g(x + !x)
To simplify this limit, we add and subtract f (x)g(x) in the numerator:
[f (x + !x)g(x) − f (x)g(x)] − [g(x + !x)f (x) − f (x)g(x)]

p & (x) = lim
!x→0 !xg(x)g(x + !x)
# . $ # $2
1 f (x + !x) − f (x) g(x + !x) − g(x)
= lim g(x) − f (x)
!x→0 g(x)g(x + !x) !x !x
. 2
1 f (x + !x) − f (x) g(x + !x) − g(x)
= g(x) lim − f (x) lim
[g(x)]2 !x→0 !x !x→0 !x
g(x)f & (x) − f (x)g & (x)

= .
[g(x)]2
If we set u = f (x) and v = g(x) , then quotient rule 3.16a can be expressed in the form
du dv
d 3u4 v −u
= dx 2 dx . (3.16b)
dx v v
It makes no difference which term in product rule 3.15 is written first; it does make a
difference in the quotient rule. Do not interchange the terms in 3.16.
EXAMPLE 3.14
For the following two functions, find f & (x) in simplified form:
√
2 4 2 x
(a) f (x) = (x + 2)(x + 5x + 1) (b) f (x) = 2
3x − 2
SOLUTION
(a) With product rule 3.15,
d 4 d
f & (x) = (x 2 + 2) (x + 5x 2 + 1) + (x 4 + 5x 2 + 1) (x 2 + 2)
dx dx
= (x 2 + 2)(4x 3 + 10x) + (x 4 + 5x 2 + 1)(2x)
= 6x 5 + 28x 3 + 22x.
(b) With quotient rule 3.16,

! "
2
1 √
(3x − 2) √ − x(6x)
2 x
f & (x) =
(3x 2 − 2)2
3x 2 − 2 − 12x 2
√
2 x
=
(3x 2 − 2)2
9x 2 + 2
=− √ .
2 x(3x 2 − 2)2
EXAMPLE 3.15
Draw a graph of the function y = f (x) = (x − 1)/(x + 2) . Calculate dy/dx and show
qualitatively that the graph agrees with your calculation.
FIGURE 3.28 Relationship between the derivative of a function and its graph’s shape
−2 1 x
−1
SOLUTION Limits were used to draw the graph in Figure 3.28. Using quotient rule 3.16, we
find that
dy (x + 2)(1) − (x − 1)(1) 3
= = .
dx (x + 2)2 (x + 2)2
The sketch in Figure 3.28 and dy/dx agree that:
(a) The slope of the curve is always positive.
(b) The slope becomes larger and larger as x approaches −2, either from the left or the
right; that is,
lim f & (x) = ∞.
x→−2
(c) The slope approaches zero as x approaches ±∞ ; that is,
lim f & (x) = 0.

x→±∞
Statements (a), (b), and (c) in Example 3.15 were arrived at by examining dy/dx = 3/(x + 2)2 .
They could also have been realized by drawing a graph of the derivative function (Figure 3.29).
Ordinates of this graph are slopes in Figure 3.28. For example, at x = 1, the height of the curve
in Figure 3.29 is 1/3. At x = 1 in Figure 3.28, the slope of the tangent line is 1/3. Notice
how clear statements (a), (b), and (c) are from the graph of f & (x) in Figure 3.29. The slope is
always positive, limx→−2 f & (x) = ∞ , and limx→±∞ f & (x) = 0.
FIGURE 3.29 The derivative function
−2 x
In Chapter 4 when we apply derivatives to geometric and physical problems, we need to

know when the derivative exists and when it does not exist, and when it is positive, negative,
and zero. A graph of the derivative function is an excellent way to discover and visualize these
properties.
EXAMPLE 3.16
Plot a graph of the function f (x) = (x 2 + 4x − 1)/(x 3 + 2) on the interval −10 ≤ x ≤ 10.
Find, to five decimal places, points where the tangent line to the graph is horizontal.
FIGURE 3.30 Plot to indicate where the tangent line is horizontal

y
3
x 2 + 4x − 1
2 y=
x3 + 2
1
x
−10 −5 5 10
−1
−2
SOLUTION A graph is shown in Figure 3.30. The graph has a vertical asymptote at x =
−21/3 . Obviously, there is a point near x = 1 at which the tangent line is horizontal. In addition,
because the graph crosses the x -axis near x = −5, and y = 0 is a horizontal asymptote as
x → −∞ , there must be at least one point to the left of x = −5 at which the tangent line is
horizontal. Whether there is more than one such point is not clear in Figure 3.30. The plot of
(x 3 + 2)(2x + 4) − (x 2 + 4x − 1)(3x 2 ) −x 4 − 8x 3 + 3x 2 + 4x + 8
f & (x) = =
(x 3 + 2)2 (x 3 + 2)2
in Figure 3.31a crosses the x -axis near x = 1, thus confirming the point in Figure 3.30 where
the tangent line is horizontal. It does not make clear, however, the number of points to the left of
x = −5 at which the graph of f & (x) crosses the x -axis. The graph of f & (x) in Figure 3.31b
clearly indicates one, and only one, point to the left of x = −5 at which f & (x) = 0. To find
the two points then where f & (x) = 0, we set
−x 4 − 8x 3 + 3x 2 + 4x + 8 = 0.
Our computer gives two (real) solutions of this equation, −8.316 793 and 1.238 656 0. The
zero intermediate value theorem of Section 1.11 guarantees that −8.316 79 and 1.238 66 are
solutions, correct to five decimal places, when we calculate
f & (−8.316 795) = −3.1 × 10−9 , f & (−8.316 785) = 1.5 × 10−8 ,
f & (1.238 655) = 2.2 × 10−6 , f & (1.238 665) = −1.9 × 10−5 .

Derivative function plots to determine where the derivative is zero
10 0.006
−x 4 − 8x 3 + 3x 2 + 4x + 8
8 f &(x) = 0.004
(x 3 + 2)2
& −x 4 − 8x 3 + 3x 2 + 4x + 8
6 f (x) =
(x 3 + 2)2 0.002
4
x
−50 −40 −30 −20 −10
2
− 0.002
x
−10 −5 5 10
EXAMPLE 3.17
Show that when f (x) is a function, differentiable for all x , and h(x − a) is the Heaviside unit
step function of Section 2.5,
d
[f (x)h(x − a)] = f & (x)h(x − a) when x ,= a.
dx
SOLUTION The product rule gives
d
[f (x)h(x − a)] = f & (x)h(x − a) + f (x)h& (x − a).
dx
Because h& (x − a) = 0 at all x , except x = a where the derivative does not exist, the required
result now follows. If we wish a result that includes x = a , we could use Example 3.11 to write
d
[f (x)h(x − a)] = f & (x)h(x − a) + f (x)δ(x − a).
dx
EXERCISES 3.4
In Exercises 1–16 find f & (x) in simplified form. ∗ 20. Find a rule for the derivative of the product of three functions f (x) ,
2
g(x) , and h(x) .
1. f (x) = (x + 2)(x + 3)
2. f (x) = (2 − x 2 )(x 2 + 4x + 2)
In Exercises 21–23 find the angle (or angles) between the curves at
x x2 their point (or points) of intersection.
3. f (x) = 4. f (x) = 2
3x + 2 4x − 5
x2 x3 ∗ 21. y = x 3 , y = 2/(1 + x 2 )
5. f (x) = 6. f (x) =
2x − 1 4x 2 + 1 ∗ 22. y = 2x + 2, y = x 2 /(x − 1)
√
√ x ∗ 23. y = 5 − x 2 , y = 3x/(x − 1)
7. f (x) = x(x + 1) 8. f (x) =
3x + 2
2x 2 − 5 x+5 ∗ 24. Find all points on the curve y = (5 − x)/(6 + x) at which the
9. f (x) = 10. f (x) = tangent line passes through the origin.
3x + 4 2x 2 − 1
∗ 25. (a) A manufacturer’s profit from the sale of x kilograms of a
x(x + 1) x 2 + 2x + 3 commodity per week is given by
11. f (x) = 12. f (x) = 2
1 − 3x x − 5x + 1
1 3x − 200
13. f (x) = P (x) = .
x3− 3x 2 + 2x + 5 x + 400
x 3 + 3x 2 + 3x + 10 Plot a graph of this function.
∗ 14. f (x) =
(x + 1)3 (b) The average profit per kilogram when x kilograms are sold
1/ 3
√ is given by
x x + 2x
∗ 15. f (x) = √ ∗ 16. f (x) = √
1− x x−4 P (x)
17. Find equations for the tangent and normal lines to the curve y =
p(x) = .
x
(x + 3)/(x − 4) at the point (1, −4/3) . Plot the curve and lines.
Plot a graph of this function.
∗ 18. (a) Plot a graph of the function f (x) = x 2 /(x 2 + x − 2) .
(c) If a point (x, P ) on the total profit curve is joined to the
(b) Find f & (x) and show that it agrees with the plot.
origin, the slope of this line is the average profit p(x) for
∗ 19. If the total cost of producing x items of a commodity is given by that x . Use this idea to find the sales level for highest
the equation ! " average profit.
x+b
C(x) = ax , ∗ 26. (a) Find f & (x) if f (x) = h(x + 1)(x 2 + x) . Draw graphs
x+c
of f (x) and f & (x) .
where a , b , and c are constants, show that the marginal cost C & (x) is (b) Does the derivative exist at x = −1 if we define f (−1) =
# $
c(b − c) 0?
a 1+ .
(x + c)2 ∗ 27. Repeat Exercise 26 if f (x) = h(x + 1)(x + 1)2 .
3.5 Higher-Order Derivatives

When y = f (x) is a function of x , its derivative f & (x) is also a function of x . We can therefore
take the derivative of the derivative to get what is called the second derivative of the function.
This can be repeated over and over again. For instance, if y = f (x) = x 3 + 1/x , then
dy 1
f & (x) = = 3x 2 − 2 .
dx x
We denote the second derivative of y with respect to x by d 2 y/dx 2 or f && (x) :
! "
&& d 2y d dy 2
f (x) = = = 6x + .
dx 2 dx dx x3
Similarly,
d 3y 6
f &&& (x) = 3
= 6 − 4.
dx x
This is called the third derivative or the derivative of order three. Clearly, we can continue the
differentiation process indefinitely to produce derivatives of any positive integer order whatso-
ever.
EXAMPLE 3.18
How many derivatives does f (x) = x 8/3 have at x = 0?
SOLUTION Since
! "! " ! "! "! "
8 5/3 8 5 2 /3 8 5 2
&
f (x) = x , &&
f (x) = x , &&&
f (x) = x −1/3 ,
3 3 3 3 3 3
and f &&& (0) is not defined, f (x) has only a first and a second derivative at x = 0.
EXAMPLE 3.19
Find a formula for the second derivative of a product, d 2 (uv)/dx 2 if u = f (x) and v = g(x) .
SOLUTION Since
d du dv
(uv) = v +u ,
dx dx dx
then
! "
d2 d du dv
(uv) = v +u
dx 2 dx dx dx
d 2u dv du d 2v du dv
=v 2
+ + u 2
+
dx dx dx dx dx dx
d 2u du dv d 2v
=v + 2 + u .
dx 2 dx dx dx 2
This formula is very handy. It is worth memorizing. See Exercise 14 for extensions to higher
order derivatives of products.
3.5 Higher-Order Derivatives 179
EXERCISES 3.5
In Exercises 1–10 find the derivative indicated. d 2n 2
∗ 15. Evaluate (x − 1)n for n a positive integer.
dx 2n
1. f && (x) if f (x) = x 3 + 5x 4
2. f &&& (x) if f (x) = x 3 − 3x 2 + 2x + 1 ∗ 16. The Green’s function for the deflection of a diving board of length
3. f && (2) if f (x) = (x + 1)(x 3 + 3x + 2) L as it bends under its own weight and any other loading is
4. f &&& (1) if f (x) = x 4 − 3x 2 + 1/x

√ 1 x3 Xx 2
5. f && (x) if f (x) = (x + 1)/ x G(x; X) = (x − X)3 h(x − X) − +
6EI 6EI 2EI
6. f &&& (t) if f (t) = t 3 − 1/t 3
7. d 9 y/dx 9 if y = x 10 where h(x − X) is the Heaviside function of Section 2.5, and E and
√ I are constants depending on the material of the board and its cross-
∗ 8. f && (u) if f (u) = u/(u + 1) section. Think of G(x; X) as a function of x that also depends on X
∗ 9. d 2 t/dx 2 if t = x/(2x − 6) where X can take on any value between 0 and L . It is the deflection at
√ position x if the only force acting on the diving board is a unit force in
∗ 10. f && (x) if f (x) = x/( x + 1) the positive y -direction at position X . It also assumes that the board
∗ 11. Steady-state temperature T in a region bounded by two concentric is massless.
spheres of radii a and b (where a X for any X .
where r is the radial distance from the common centre of the spheres.
(a) Verify that for any constants c and d , the function (c) Verify that G(x; X) , dG/dx , and d 2 G/dx 2 are all con-
d tinuous for 0 ≤ x ≤ L , except for a removable disconti-
T = f (r) = c + nuity at x = X .
r
satisfies the equation. (d) Verify that d 3 G/dx 3 is continuous for all x ,= X , and has
(b) If temperatures on the two spheres are maintained at con- a jump discontinuity at x = X of (EI )−1 .
stant values Ta and Tb , calculate c and d in terms of a , b ,
Ta , and Tb .
∗ 17. The Green’s function for a beam of length L that has its ends
∗ 12. If heat is generated at the centre of the sphere in Exercise 11 at a clamped horizontally at x = 0 and x = L is
constant rate, the equation satisfied by T (r) is
! "
d dT 1
r2 = kr 2 , G(x; X) = (x − X)3 h(x − X)
dr dr 6EI
where k is a constant. x3
+ (−L3 + 3LX2 − 2X3 )
(a) Verify that 6EI L3
kr 2 c x2
T (r) = + +d + (X3 − 2LX2 + L2 X).
6 r 2EI L2
satisfies the equation for any constants c and d .
(b) If temperatures on the two spheres are maintained at con- (See Exercise 16 for a description of E and I and the interpretation of
stant values Ta and Tb , calculate c and d in terms of a , b , G(x; X) .)
Ta , Tb , and k .
(a) Draw a graph of G(x; X) when X = L/2. Is its tangent
∗ 13. Find constants a , b , c , and d in order that the function y = line horizontal at x = 0 and x = L ? Is it symmetric
f (x) = ax 3 + bx 2 + cx + d has its first derivative equal to 4 when about x = L/2? Does it have a removable discontinuity
x = 1 and y = 0, and its second derivative equal to 5 when x = 2 at x = L/2?
and y = 4.
∗ 14. (a) Find a formula for d 3 (uv)/dx 3 if u = f (x) and v = (b) Verify that G(x; X) , dG/dx , and d 2 G/dx 2 are all con-
g(x) . tinuous for 0 ≤ x ≤ L , except for a removable disconti-
(b) On the basis of part (a) and Example 3.19, could you haz- nuity at x = X .
ard a guess at a formula for d 4 (uv)/dx 4 (i.e., do you see
the pattern emerging)? Exercise 18 asks you to verify the (c) Verify that d 3 G/dx 3 is continuous for all x ,= X , and has
correct formula. a jump discontinuity at x = X of (EI )−1 .
! "
∗∗ 18. If n is a positive integer, and u and v are functions of x , show by n
where d 0 u/dx 0 = u , and are the binomial coefficients
mathematical induction that r
! "
n n!
= .
r r ! (n − r)!
5n ! " ! r " ! n−r "
dn n d u d v
(uv) = , (If you are not familiar with sigma notation, see Section 6.1. Mathe-
dx n r=0
r dx r dx n−r
matical induction is discussed in Appendix A.)
3.6 Velocity, Speed, and Acceleration

Most applications of differentiation are dealt with in Chapter 4, but velocity, speed, and accel-
eration are so important in engineering and the physical sciences that it is important to discuss
them as soon as possible. In Section 3.1, we used the position function
x(t) = t 3 − 27t 2 + 168t + 20, t ≥ 0,
to define velocity as the instantaneous rate of change of displacement with respect to time. (Un-
less we indicate otherwise, velocity means instantaneous velocity, rather than average velocity.)
We can now say that velocity is the derivative of displacement,
dx
v(t) = = 3t 2 − 54t + 168 m/s.
dt
In terms of the graph of the displacement function (Figure 3.32), velocity is the slope of the
tangent line. For instance, v(3) = 33 m/s. The particle is moving to the right at 33 m/s; the
slope of the tangent line at (3, 308) is 33. (See the triangle on the tangent line and remember
that the axes have different scales.) After 5 s, v(5) = −27 m/s. The particle is moving to
the left at 27 m/s; the slope of the tangent line is −27 at (5, 310) . By factoring v(t) in the
form
v(t) = 3(t − 4)(t − 14),
we see that velocity is zero at t = 4 and t = 14. This is consistent with horizontal tangent
lines at (4, 324) and 14, −176) .
FIGURE 3.32 Displacement function illustrates the velocity of object
x
500 x = t 3 − 27t 2 + 168t + 20
2
400
66
300
(3, 308)
200
100
5 10 15 20 t
−100
The particle is moving to the right whenever its velocity is positive, to the left when its
velocity is negative. Graphically, it moves to the right when the slope of the tangent line is
positive (0 ≤ t < 4 and t > 14), and to the left when the slope of the tangent line is negative
(4 < t < 14).
Speed is defined to be the magnitude of velocity,
speed = |v(t)|. (3.17)

It represents how fast the particle is moving without regard for direction. For instance, at time
t = 0, velocity and speed are both 168 m/s, but at t = 5, velocity is −33 m/s and speed
is 33 m/s. Do not confuse velocity and speed; they are not interchangeable. Velocity can be
positive or negative depending on direction of motion; speed is never negative.
EXAMPLE 3.20
Figure 3.33 shows the graph of the displacement function x(t) for a particle moving along
the x -axis during the time interval 0 ≤ t ≤ 10. Use the graph to answer the following
questions:
(a) Is the particle moving to the left or right at t = 7 s?
(b) How many times does the particle stop moving?
(c) Is the velocity at t = 0.5 s greater or smaller than at t = 3.5 s?
(d) Is the speed at t = 0.5 s greater or smaller than at t = 3.5 s?
(e) Estimate the average velocity of the particle over the interval.
(f) At what times would you say that the velocity is greatest and smallest?
(g) At what times would you say that the speed is greatest and smallest?
FIGURE 3.33 Displacement function for x(t)
200 x
150
100
50
2 4 6 8 10 t
−50
−100
SOLUTION
(a) Since the slope of the tangent line at t = 7 s is negative, the velocity is negative, and
the particle is moving to the left.
(b) The particle stops moving when its velocity is zero, and this occurs when the tangent
line to the graph is horizontal. This happens three times.
(c) At t = 0.5 s, the velocity is negative; at t = 3.5 s, it is positive. Thus, it is smaller
at t = 0.5 s.
(d) The tangent line at t = 0.5 s is much steeper than it is at t = 3.5 s. Since speed
measures steepness (slope without regard for sign), speed is greater at 0.5 s.
(e) Average velocity is the difference between initial and final displacements divided by
the length of the time interval. If we estimate x(0) = 200 and x(10) = 120,
120 − 200
average velocity ≈ = −8 m/s.
10
(f) Since the slope of the graph appears to be greatest at t = 10 s, the velocity must be
greatest then. The slope and velocity appear to be smallest at t = 0 s.
(g) Because speed is the magnitude of velocity, it is never negative. Speed is zero if
velocity is zero. Since this happens three times, speed is smallest (and has value 0) at
the three times when the tangent line is horizontal. Speed is greatest when steepness
of the graph is greatest. This is either at t = 0 s or t = 10 s. It is difficult to tell
from the graph, but we favour a steeper tangent line at t = 0 s.
EXAMPLE 3.21
You are now told that the displacement function for the graph in Figure 3.33 is
521t 2 1243t 399

x(t) = t 4 − 20t 3 + − + .
4 4 2
Answer the following questions:
(a) What are the speed and velocity at t = 1 s?
(b) Is the velocity equal to zero at t = 2 s?
(c) What is the average velocity of the particle for 0 ≤ t ≤ 10?
(d) At what time is speed greatest?
(e) To three decimal places, at what times is the speed of the particle equal to 20 m/s?
FIGURE 3.34a Graph of velocity FIGURE 3.34b Graph of speed
v v
521t 1243
200 v = 4t 3 − 60t 2 + − 200 521t 1243
2 4 Speed = 4t 3 − 60t 2 + −
2 4
100 100
2 4 6 8 10 t 2 4 6 8 10 t
−100 −100
−200 −200
−300 −300
SOLUTION The velocity of the particle is
dx 521t 1243
v(t) = = 4t 3 − 60t 2 + − m/s.
dt 2 4
The above questions will be answered algebraically, but it is helpful to visualize responses by
plotting the velocity and speed functions. We have done this in Figure 3.34. Figure 3.34b
reflects that part of the graph in Figure 3.34a below the t -axis, in the t -axis.
(a) At t = 1 s, velocity is
521 1243 425

v(1) = 4(1)3 − 60(1)2 + − =− m/s.
2 4 4
The speed is 425/4 m/s.
(b) Geometrically, the velocity is zero at t = 2 s if the tangent line in Figure 3.33
is horizontal at t = 2. It looks close. Equivalently, does the graph in Figure
3.34a cross the t -axis at t = 2? Again, the decision is not clear. Since v(2)
= 4(2)3 − 60(2)2 + (521/2)(2) − 1243/4 = 2.25 m/s, the velocity is not zero at
t = 2 s.
(c) Since displacements at t = 0 s and t = 10 s are 399/2 m and 117 m,
117 − 399/2
average velocity = = −8.25 m/s.
10
(d) Figure 3.34b indicates that speed is greatest at t = 0 s or t = 10 s, favouring t = 0 s.
Because v(0) = −1243/4 m/s and v(10) = 1117/4 m/s, it follows that speed is
greatest at t = 0 s [see also part (g) in Example 3.20].
(e) Solutions can be visualized as times when the height of the graph in Figure 3.34b
is 20. Clearly, there are six such times. Alternatively, solutions are times when the
ordinate in Figure 3.34a is ±20, the same six times. Solutions of
521t 1243
20 = v(t) = 4t 3 − 60t 2 + −
2 4
are 2.306 95, 4.240 43, and 8.452 62. The zero intermediate value theorem can be
used to confirm three-decimal-place solutions 2.307 s, 4.240 s, and 8.453 s. Three-
decimal-place solutions of v(t) = −20 are 1.718 s, 5.300 s, and 7.982 s.
The velocity of a particle moving along the x -axis is the derivative of its displacement function
x(t) with respect to t ,
dx
v(t) = . (3.18)
dt
The (instantaneous) acceleration of the particle is defined as the rate of change of velocity
with respect to time:
dv d 2x
a(t) = = . (3.19)
dt dt 2
In actual fact, x(t) , v(t) , and a(t) are the x -components of the displacement, velocity, and
acceleration vectors, respectively. In the absence of a complete discussion of vectors, we omit
the terms vector and component, and simply call x(t) , v(t) , and a(t) displacement, velocity,
and acceleration. When distance is measured in metres and time in seconds, velocity is measured
in metres per second (m/s). Since acceleration is the time derivative of velocity, its units must be
units of velocity divided by units of time [i.e., metres per second per second (m/s/s)]. Usually,
we shorten this by saying “metres per second squared” and write m/s 2 .
For example, in Figure 3.2 the acceleration of the particle with displacement function
x(t) = t 3 − 27t 2 + 168t + 20 is
d
a(t) = (3t 2 − 54t + 168) = 6t − 54 m/s2 .
dt
The graphical interpretation of acceleration requires the concept of “concavity.” This is discussed
in Section 4.4.
An industrialist is having a problem with the design and manufacture of a cam. A cam
is a machine part that rotates about an axis to cause periodic movement in another part,
called a follower. The plate cam in Figure 3.35a rotates about an axis through the origin
O and perpendicular to the plate. The follower moves up and down along the y -axis as
point A on its end remains in contact with the cam. Suppose Figure 3.35b represents the
displacement y = f (θ ) of the follower, from its lowest position (1 cm above O ), as a
function of angle θ through which the cam has rotated.
FIGURE 3.35a Plate cam
Follower
A Cam
O x
FIGURE 3.35b Displacement of follower in plate cam
3/4
θ1 2π π 4π 2π − θ1 2π θ
3 3
The follower rises 2 cm during a rotation of 2π/3 radians. Because velocity of

the follower is proportional to the slope of the graph, velocity increases in the interval
0 ≤ θ ≤ θ1 , when the follower rises 3/4 cm, and decreases in the interval θ1 ≤ θ ≤
2π/3 when the follower rises the final 5/4 cm. The follower is stationary in the interval
2π/3 ≤ θ ≤ π . It then retraces its path back to zero displacement above its minimum
position in a similar fashion. All curves are parabolas; the first has a minimum at (0, 0)
and the second has a maximum at (2π/3, 2) . The industrialist has two problems for
us. One is to ensure that the follower moves smoothly; that is, its motion is not “jerky.”
Secondly, he wants us to give him a scale diagram of the cam itself. He has not specified
angle θ1 .
SOLUTION We begin be using our knowledge of equations for parabolas. Equations
of the first two parabolas must be of the form
.
aθ 2 , 0 ≤ θ ≤ θ1 ,
f (θ ) =
A(θ − 2π/3)2 + 2, θ1 ≤ θ ≤ 2π/3,
as shown here where a and A are constants. For (θ1 , 0.75) to be a point on both parabolas,
0.75 = aθ12 , 0.75 = A(θ1 − 2π/3)2 + 2.
Now comes the most difficult consideration of the project, interpretation of the follower
moving smoothly. This means that there can be no sudden changes in its velocity. Since
velocity is related to slope of the curve, we ask where sudden changes in slope could occur.
Slope changes gradually along parabolas, and there is no difficulty in the transition from
parabola to horizontal straight line at (0, 0) and (2π/3, 2) . The only questionable point
is where the above two parabolas join. We must ensure that slopes of the two parabolas
match at the point (θ1 , 0.75) . Otherwise, there will be a jump in the velocity there. For
the left-hand derivative of aθ 2 and the right-hand derivative of A(θ − 2π/3)2 + 2 to be
equal at (θ1 , 0.75) , we must have
2aθ1 = 2A(θ1 − 2π/3).

We now have three equations in three unknowns, a , A , and θ1 . If we solve the first two
equations for a and A in terms of θ1 , and substitute into the last equation,
! " # $
0.75 0.75 − 2
2 θ1 = 2 (θ1 − 2π/3).
θ12 (θ1 − 2π/3)2
The solution of this equation is θ1 = π/4. This gives a = 12/π 2 and A = −36/(5π 2 ) .
The industrialist therefore has no choice for angle θ1 . Smooth motion of the follower
requires θ1 = π/4. The equation of the displacement curve for the follower above its
minimum position is therefore
6
12θ 2 /π 2 , 0 ≤ θ ≤ π/4
y = f (θ ) = −36(θ − 2π/3)2 /(5π 2 ) + 2, π/4 ≤ θ ≤ 2π/3
2, 2π/3 < θ ≤ π .
Now consider the shape of the cam itself. When we add unity to f (θ ) , we obtain the
distance of the end A of the follower above O . This represents the distance from the
centre of the cam to its outer edge as a function of angle through which it has rotated,
call it g(θ ) = 1 + f (θ ) (Figure 3.36a). To obtain a scale diagram of the cam, we need
the equation of this curve. If (x, y) is a point on the curve, then trigonometry indicates
that x = g(θ ) cos θ and y = g(θ ) sin θ . Unfortunately, it is not possible for us to
eliminate θ between these equations and find an equation in x and y only. On the other
hand, electronic devices have programs to plot curves given in this form, and we will deal
with them at length in Chapter 9 as parametrically defined curves. The resulting plot and
required cam shape is shown in Figure 3.36b.
FIGURE 3.36a Schematic for shape of FIGURE 3.36b Actual shape of cam
cam
y
y 3
(x, y)
g(θ) = 1 + f(θ)
2
O x 1
−3 −2 −1 1 x
−1
−2
−3
EXERCISES 3.6
1. The figure at right shows the graph of the displacement function
x(t) of a particle moving along the x -axis during the time interval x
0 ≤ t ≤ 6. Answer the following questions. 6
(a) Is the particle to the left or right of the origin at times t = 1 4
and t = 4?
(b) Is the particle moving to the right or to the left at times 2
t = 1/2 and t = 3?
(c) How many times does the particle change direction? 1 2 3 4 5 6 t
(d) Is the velocity greater at t = 7/2 or at t = 9/2? −2
(e) Is the speed greater at t = 7/2 or at t = 9/2?
2. You are given that the displacement function in Exercise 1 is where x is measured in metres and t ≥ 0 is time in seconds.
Determine algebraically: (a) position, velocity, speed, and acceleration
t4 32t 3 25t 2 251t at t = 3 s; (b) when the object is instantaneously at rest; (c) when
x(t) = − + − .
6 15 3 30 acceleration vanishes; (d) times when the object is moving to the right
and left; (e) if and when velocity is 1 m/s; (f) if and when speed is
Confirm each answer in Exercise 1 algebraically. 1 m/s; (g) if and when velocity is 20 m/s; and (h) if and when speed is
3. Repeat Exercise 1 for the graph below. 20 m/s.
∗ 18. Repeat Exercise 17 if
x
6
x(t) = t 3 − 9t 2 + 15t − 2.
4 ∗ 19. Can the position curve x = x(t) of a realistic particle moving

along the x -axis be represented by a function x(t) that has a disconti-
nuity? Explain.
2
∗ 20. When an object moves with constant acceleration a along the x -
axis, its position as a function of time t must be of the form
1 2 3 4 5 6 t 1
x = x(t) = at 2 + bt + c,
−2 2
where b and c are constants.
4. You are given that the displacement function in Exercise 3 is (a) If the object is at positions x1 and x2 at times t1 and t2 , what
is its average velocity over the time interval t1 ≤ t ≤ t2 ?
14t 3 101t 2 132t
x(t) = − + + 2. (b) At what time in this time interval is the instantaneous ve-
45 45 45
locity equal to the average velocity? Where is the object at
Confirm each answer in Exercise 3 algebraically. this time? Is it at the midpoint of the interval between x1
and x2 , is it closer to x1 , or is it closer to x2 ? (Assume in
5. When x(t) is the displacement function for a particle moving along
this part of the problem that a > 0 and that the velocity of
the x -axis, the third derivative x &&& (t) is called jerk. Why is this name
the object is positive at time t1 .)
appropriate? Find jerk for the displacement function in Exercise 2.
∗ 21. Find f (θ) in the consulting project of this section when rotation
6. Find jerk for the displacement function in Exercise 4. (See Exercise
2π/3 is replaced by θ2 , and displacements 3/4 and 2 are replaced by
5 for the definition of jerk.)
y1 and y2 , respectively. In the process, show that y1 /y2 = θ1 /θ2 , and
that this implies that the point (θ1 , y1 ) is on the line joining (0, 0) and
In Exercises 7–11 find the velocity and acceleration of an object that (θ2 , y2 ) .
moves along the x -axis with the given position function. Assume that ∗ 22. The figure below represents a portion of the displacement f (θ) of
x is in metres and t is in seconds. Draw graphs of x(t) , v(t) , and a(t) , a follower in a cam mechanism. It consists of a straight line portion
and examine the graphs from the point of view that ordinates on v(t) between A and B and two parabolic portions OA and BC that are
represent slopes on x(t) , and ordinates on a(t) are slopes on v(t) . horizontal at O and C . Given are the rises y1 , y2 , and y3 , and angle
Draw a graph of speed as a function of t also. θ3 . It is required that the slope of the curve be continuous at θ1 and θ2 .
By taking f (θ) in the form
7. x(t) = 2t + 5, t ≥5 8. x(t) = t 2 − 7t + 6, t ≥0
 2
2
9. x(t) = t + 5t + 10, t ≥1  aθ , 0 ≤ θ ≤ θ1
y(θ) = mθ + b, θ1 < θ < θ2
10. x(t) = 4t − t 3 , t ≥0 
A(θ − θ3 )2 + y3 , θ2 ≤ θ ≤ θ3
11. x(t) = 1/t , t ≥1
show that
Repeat the calculations for the above five exercises for Exercises 12–16 2y1 θ3 (y1 + y2 )θ3
but plot graphs instead of drawing them.
θ1 = , θ2 = ,
y1 − y2 + 2y3 y1 − y2 + 2y3
12. x(t) = −2t 3 + 2t 2 + 16t − 1, t ≥0 and find a , m , b , and A .
13. x(t) = t 3 − 9t 2 + 15t + 3, t ≥2 y C
14. x(t) = t + 4/t, t ≥1 y3
B
y2
15. x(t) = (t − 4)/t 2 , t ≥2
√
16. x(t) = (t − 1)2 t, t ≥1
A
y1
∗ 17. An object moving along the x -axis has position given by
x(t) = t 3 − 9t 2 + 24t + 1, O θ1 θ2 θ3 θ
3.7 The Chain Rule and the Extended Power Rule

With the differentiation rules in Sections 3.2 and 3.4, we can differentiate any polynomial
whatsoever. For instance, if f (x) = x 3 − 3x 2 + 2x + 1, then f & (x) = 3x 2 − 6x + 2. But
consider the polynomial f (x) = (2x 2 − 3)8 , which has been conveniently factored for us. To
find its derivative, we could expand (2x 2 − 3)8 by the binomial theorem, say, then differentiate,
and finally, simplify. We could also consider using the product rule over and over and over
again. Thus, in spite of the fact that the rules of Sections 3.2 and 3.4 permit differentiation of
(2x 2 − 3)8 , they are not convenient to use. Even more unpleasant would be differentiation
of the rational function g(x) = 1/(3x 2 + 8)12 . In this section, we obtain results that enable
us to differentiate quickly much wider classes of functions, which include (2x 2 − 3)8 and
1/(3x 2 + 8)12 .
Suppose that y is defined as a function of u by
u
y = f (u) = ,
u+1
and u , in turn, is defined as a function of x by
√
x
u = g(x) = .
x+2
These equations imply that y is a function of x ; indeed, y is the composition of f and g ,
√
x
) *
y = f g(x) = √x + 2 .
x
+1
x+2
After some algebraic simplification we find that
√
x
y = √ ,
x+x+2
and we can therefore calculate
! " ! "
√ 1 √ 1
( x + x + 2) √ − x √ +1
dy 2 x 2 x
= √ .
dx ( x + x + 2 )2
This can be reduced to
dy 2−x
= √ √ .
dx 2 x( x + x + 2)2
But notice that if we differentiate the original functions, we obtain
dy (u + 1)(1) − u(1) 1
= = and
du (u + 1)2 (u + 1)2
! "
1 √
(x + 2) √ − x(1)
du 2 x 2−x
= = √ .
dx (x + 2)2 2 x(x + 2)2
The product of these derivatives is
dy du 2−x
= √
du dx 2 x(x + 2)2 (u + 1)2
2−x 2−x
= ! √ "2 = √ √ .
√ x 2 x( x + x + 2)2
2 x(x + 2)2 +1
x+2
Consequently, for this example, we can write
dy dy du
= .
dx du dx
) *
According to the following theorem, the derivative of a composite function y = f g(x) can
always be calculated by the above formula.
THEOREM 3.9 (Chain Rule)

If y = f (u) and u = g(x)
) are
* differentiable functions, then the derivative of the
composite function y = f g(x) is
dy dy du
= . (3.20a)
dx du dx
Increment notation is particularly useful in proving the chain rule.

PROOF By equation 3.11, the derivative of the composite function is
) * ) *
dy !y f g(x + !x) − f g(x)
= lim = lim .
dx !x→0 !x !x→0 !x
Now a change !x in x produces a change g(x + !x) − g(x) in u . If we denote this change
by !u , then it in turn produces the change f (u + !u) − f (u) in y . We may write
dy f (u + !u) − f (u)
= lim .
dx !x→0 !x
Since f (u) is differentiable, its derivative exists and is defined by
f (u + !u) − f (u)
f & (u) = lim .
!u→0 !u
An equivalent way to express the fact that this limit is f & (u) is to say that
f (u + !u) − f (u)
= f & (u) + %,
!u
where % must satisfy the condition lim!u→0 % = 0. We may write
f (u + !u) − f (u) = [f & (u) + % ] !u,
and if we substitute this into the second expression for dy/dx above, we obtain
. 2
dy [f & (u) + % ] !u !u
= lim = lim [f & (u) + % ] .
dx !x→0 !x !x→0 !x
But
!u g(x + !x) − g(x) du
lim = lim = .
!x→0 !x !x→0 !x dx
Furthermore, since g(x) is differentiable, it is continuous (Theorem 3.6), and this implies that
!u → 0 as !x → 0. Consequently, lim!x→0 % = lim!u→0 % = 0, and these results give
dy du dy du
= f & (u) = .
dx dx du dx
This result is called the chain rule for the derivative of a composite function. It expresses
the derivative of a composite function as the product of the derivatives of the functions in the
composition. From the point of view of rates of change, the chain rule seems quite reasonable.
It states that if a variable is defined in terms of a second variable, which is in turn defined in
terms of a third variable, then the rate of change of the first variable with respect to the third is
the rate of change of the first with respect to the second multiplied by the rate of change of the
second with respect to the third. For example, if car A travels twice as fast as car B , and car B
travels three times as fast as car C , then car A travels six times as fast as car C .
It is essential to understand the difference between the derivatives dy/dx and dy/du in
equation 3.20a. The second, dy/du , is the derivative of y regarded as a function of u , the given
function
) *y = f (u) . On the other hand, dy/dx is the derivative of the composite function
f g(x) .
The) chain*rule can also be expressed in terms of the circle notation for composite functions.
With f g(x) denoted by (f ◦ g)(x) , equation 3.20a takes the form
) *
(f ◦ g)& (x) = f & g(x) g & (x), (3.20b)
or
(f ◦ g)& (x) = (f & ◦ g)(x) g & (x). (3.20c)
EXAMPLE 3.22
√
Find dy/dt at t = 4 when y = x 2 − x and x = t/(t + 1) .
SOLUTION By the chain rule
# √ $
dy dy dx (t + 1)(1/2)t −1/2 − t(1)
= = (2x − 1) .
dt dx dt (t + 1)2
√ dy //
When t = 4, we find x = 4/(4 + 1) = 2/5. We use the notation / to represent
dt t=4
dy/dt evaluated at t = 4,
7 √ 8
dy // (4 + 1)(1/2)(4)−1/2 − 4 3
/ = [2(2/5) − 1] = .
dt t=4 (4 + 1)2 500
The Extended Power Rule

When the function f (u) in Theorem 3.9 is a power function, the chain rule gives what we
call the extended power rule, often considered the most important differentiation formula of
calculus.
COROLLARY 3.9.1
When u = g(x) is differentiable,
d n du
u = nun−1 . (3.21)
dx dx
It is essentially power rule 3.7 in Section 3.2 with an extra factor du/dx to account for the
fact that what is under the power ( u ) is not just x ; it is a function of x . In the special case that u
is equal to x , equation 3.21 reduces to 3.7. Power rule 3.7 was verified only for n a nonnegative
integer, but we have been using it for any real number n . We shall use its generalization 3.21 for
any real number n also in spite of the fact that, in effect, it has only been verified for nonnegative
integers. In Section 3.11 we provide the justification.
It is important not to confuse rules 3.7 and 3.21. Rule 3.7 can be used only for x n ; if
anything other than x is raised to a power, formula 3.21 should be used. The most common
error in using equation 3.21 is to forget the du/dx . With equation 3.21 it is a simple matter to
differentiate the functions in the first paragraph of this section:
d d
(2x 2 − 3)8 = 8(2x 2 − 3)7 (2x 2 − 3) = 8(2x 2 − 3)7 (4x) = 32x(2x 2 − 3)7
dx dx
and
# $
d 1 d d
2 12
= (3x 2 + 8)−12 = −12(3x 2 + 8)−13 (3x 2 + 8)
dx (3x + 8) dx dx
−12 −72x
= (6x) = .
(3x 2+ 8) 13 (3x 2 + 8)13
Notice that in neither of these examples do you see the letter u , although rule 3.21 is stated in
terms of u ’s. We could have introduced u , in the first example, say, by setting u = 2x 2 − 3,
and proceeded as follows: With u = 2x 2 − 3,
d d 8 d 8 du
(2x 2 − 3)8 = u = (u ) = 8u7 (4x) = 32xu7 = 32x(2x 2 − 3)7 .
dx dx du dx
But this is unnecessary; with an understanding of equation 3.21, we should proceed directly
to the derivatives without defining an intermediate variable. With a little practice, the writing
should be shortened even more. For example, calculation of the derivative of (2x 2 − 3)8 should
appear as
bring down the power lower power by one
d (2x2 − 3)8 = 8(2x2 − 3)7(4x) = 32x(2x2 − 3)7

dx
same expression under power derivative of expression under power
EXAMPLE 3.23
A balloon always remains spherical when it is being filled. If the radius of the balloon is
increasing at a rate of 2 mm/s when the radius is 10 cm, how fast is the volume changing at this
instant?
SOLUTION Because the rate of change of the radius r of the balloon is 2 mm/s, we can say
that the derivative
dr
= 0.2 cm/s,
dt
when r = 10 cm. Since the volume of the balloon is V = (4/3)π r 3 , and r is a function of
time t , we may differentiate this equation with respect to t ,
! "
dV 4 2 dr dr
= π 3r = 4π r 2 .
dt 3 dt dt
When r = 10,
dV
= 4π(10)2 (0.2) = 80π.
dt
Thus, the volume is increasing at 80π cm 3 /s.
EXERCISES 3.7
In Exercises 1–8 use the chain rule to find dy/dx . x 2 (x 3 + 3)2
∗ 27. f (x) =
(x − 2)(x + 5)2
1 :
1. y = t 2 + , t = x2 + 1 √
t ∗ 28. f (x) = x 1 + x 1 + x
u √
2. y = , u= x+1 1 +
u+1 ∗ 29. y = t 2 + 3 , t = 4 − x 2
t
1
3. y = (u2 + 1)(u + 1), u= √ 1
x−4 ∗ 30. y = (2s − s 2 )1/3 , s=
s x2
+5
4. y = , s = x 2 − 2x + 3 v2 +
s2 −2 ∗ 31. y = 3 , v = x x2 − 1
√ x v −1
5. y = (v 2 + v)( v + 1), v= √
x2 − 1 u x−1
t +3 x−2 ∗ 32. y = , u=
6. y = , t= u+5 x
t −4 x+1 +
4 3 2
2
∗ 33. y = u (u − 2u) , u = x − 2x 2
t +3 :
7. y = , t = (3x + 2)(x 2 + 4x) √ x2 + 1
t −4 ∗ 34. y = t + t + t, t = 2
) √ * x+1 x −1
8. y = u2 1 + u , u = ! 2 "3
x − x2 v +1 1
∗ 35. y = 3
, v= 3
1−v x + 3x 2 + 2
√
In Exercises 9–36 find dy/dx , where y = f (x) . k
√ ∗ 36. y = , k = x(x 2 + 5)5
1 + k + k2
∗ 9. f (x) = x(x 3 + 3)4 ∗ 10. f (x) = x x + 1
x
∗ 11. f (x) = x 2 (2x + 1)2 ∗ 12. f (x) = √ ∗ 37. If y = f (u) , u = g(s) , and s = h(x) , show that
2x + 1
(2x − 1)2 dy dy du ds
∗ 13. f (x) = (x + 2)2 (x 2 + 3)∗ 14. f (x) = = .
3x + 5 dx du ds dx
3x + 5 ∗ 38. When an electrostatic charge q = 5 × 10−6 C is at a distance r
∗ 15. f (x) = ∗ 16. f (x) = x 3 (2 − 5x 2 )1/3
(2x − 1)2 metres from a stationary charge Q = 3 × 10−6 C, the magnitude of
x3 the force of repulsion of Q on q is
∗ 17. f (x) =
(2 − 5x 2 )1/3 Qq
F = newtons,
∗ 18. f (x) = (x + 1)2 (3x + 1)3 4π %0 r 2
√
x 1/ 3 2 − 3x
∗ 19. f (x) = √ ∗ 20. f (x) = where %0 = 8.85 × 10−12 . If q is moved directly away from Q at
1− x x2 2 m/s, how fast is F changing when r = 2 m?
! 3 "4 9
x −1 4 2−x ∗ 39. When a mass m of 5 kg is r metres from the centre of the earth,
∗ 21. f (x) = ∗ 22. f (x) = the magnitude of the force of attraction of the earth on m is
2x 3 + 1 2+x
∗ 23. f (x) = (x 3 − 2x 2 )3 (x 4 − 2x)5 GmM
+ F = newtons,
∗ 24. f (x) = (x + 5)4 1 + x 3 r2
√
x 1 − x2 where M is the mass of the earth in kilograms and G = 6.67 × 10−11 .
∗ 25. f (x) = If m is falling at 100 km/h when it is 5 km above the surface of the
(3 + x)1/3
+ earth, how fast is F changing? The mean density of the earth is 5.52 ×
∗ 26. f (x) = x(x + 5)4 1 + x 3 103 kg/m 3 , and its mean radius is 6370 km.
∗ 40. Show that if an equation can be solved for both y in terms of x In Exercises 55–58 show that the families of curves are orthogonal
[y = f (x)] and x in terms of y [x = g(y)], then trajectories (see Exercise 23 in Section 3.3). Draw both families of
curves.
dy 1
= .
dx dx ∗ 55. y = mx, x2 + y2 = r 2
dy
∗ 56. y = ax 2 , x 2 + 2y 2 = c 2
∗ 41. Prove that the derivative of an odd function is an even function,
and that the derivative of an even function is an odd function. Recall ∗ 57. x 2 − y 2 = C1 , xy = C2
Definition 1.4 in Section 1.5.
∗ 58. 2x 2 + 3y 2 = C 2 , y 2 = ax 3
∗ 42. Determine whether the following very simple proof of the chain
rule has a flaw. If !u denotes the change in u = g(x) resulting from
∗ 59. If y = f (u) and u = g(x) , use the chain rule to show that
a change !x in x , and !y is the change in y = f (u) resulting from
!u , then ! "2
! "! " d 2y d 2 u dy d 2y du
dy !y !y !u !y !u = + .
= lim = lim = lim lim dx 2 dx 2 du du2 dx
dx !x→0 !x !x→0 !u !x !x→0 !u !x→0 !x
! "! "
!y !u ∗ 60. Use the result of Exercise 59 to find d 2 y/dx 2 at x = 1 when
= lim lim
!u→0 !u !x→0 !x
2
dy du y = (u + 1)3 , u = 3x − .
= . x2
du dx
∗ 61. Find the rate of change of y = f (x) = x 9 + x 6 with respect to
x3 .
In Exercises 43–46 find d 2 y/dx 2 .
√
x ∗ 62. Find the rate of change of y = f (x) = 1 − x 2 with respect to
∗ 43. y = v 2 + v, v= x/(x + 1) .
x+1
1 √ ∗ 63. If y = f (u) and u = g(x) , show that
∗ 44. y = (u + 1)3 − , u = x + x+1
u
√ ! "3
∗ 45. y = t − 1, t = (x + x 2 )2 d 3y d 3 u dy d 2 y d 2 u du d 3y du
√ = + 3 + .
s x dx 3 dx 3 du du2 dx 2 dx du3 dx
∗ 46. y = , s= √
s+6 1+ x
∗ 64. At what point(s) on the hyperbola x 2 − 16y 2 = 16 do tangent
lines pass through the point (2, 3) ?
In Exercises 47–54 assume that f (u) is a differentiable function of
u . Find the derivative of the given function with respect to x in as ∗ 65. Generalize the result of equation 3.13 to find a formula for
simplified a form as possible. Then set f (u) = u3 − 2u and simplify
further. d
|f (x)|n for n > 1 an integer.
2 dx
∗ 47. f (2x + 3) ∗ 48. [f (3 − 4x)]
∗ 49. f (1 − x 2 ) ∗ 50. f (x + 1/x)
) * + ∗ 66. The curve x 2 y 2 = (x + 1)2 (4 − x 2 ) is called a conchoid of
∗ 51. f f (x) ∗ 52. 3 − 4[f (1 − 3x)]2 Nicomedes. Plot the curve and find points where its tangent line is
f (−x) ) * horizontal.
∗ 53. ∗ 54. f x − f (x)
3 + 2f (x 2 ) ∗∗ 67. Repeat Exercise 34 in Section 3.2 if the hill is the arc of a circle.
3.8 Implicit Differentiation

We say that y is defined explicitly as a function of x if the dependence of y on x is given in
the form
y = f (x). (3.22)
Examples are y = x 2 , y = 3x + sin x , and y = 1/(x + 1) . In each case, the dependent
variable stands alone on the left side of the equation. The differentiation rules in Theorems 3.1
to 3.5, 3.7, and 3.8 are applicable to explicitly defined functions.
An equation in x and y may define y as a function of x even when it is not in form 3.22.
Equations for which y has not been separated out are often written in the generic form
F (x, y) = 0. (3.23)
The notation F (x, y) is used to denote an expression that depends on two variables x and
3
y . For example, the volume of a right circular cylinder depends on its radius r and height h ;
FIGURE 3.37 For y −
in particular, V = π r 2 h . In such a case we write that V = F (r, h) = π r 2 h . Examples of
3y − 2x = 0, y is not a function
of x
equations of form 3.23 are y − x 3 = 0 and y 3 − 3y − 2x = 0. Equation 3.23 is said to define
y implicitly as a function of x in some domain D , if for each x in D , there is one, and only
y
one, value of y for which (x, y) satisfies the equation. The equation y − x 3 = 0 defines y
√3
implicitly as a function of x for all (real) x . The explicit definition of the function is y = x 3 .
The equation y 3 − 3y − 2x = 0 does not define y as a function of x . For each x in the interval
−1 < x < 1, there are three solutions of the equation for y . For x < −1 and x > 1, the
(−1, 1) equation has only one value of y corresponding to each x . This is most easily seen from the plot
of the curve in Figure 3.37. We shall have more to say about the equation y 3 − 3y − 2x = 0
later in this section.
x Each of the following equations defines y as a function of x , and does so implicitly:
(1, −1)
y + x 2 − 2x + 5 = 0;
− √3
2y + x 2 + y 2 − 25 = 0, y ≥ −1;
y3 − 3y − 2x = 0
x + y 5 + x 2 + y = 0.
It is easy to solve the first equation for the explicit definition of the function,
y = −x 2 + 2x − 5,
and find its derivative,

dy
= −2 x + 2 .
dx
We can also obtain an explicit definition of the function defined by the second equation, but not
so simply. If we write
y 2 + 2y + (x 2 − 25) = 0,
and use quadratic formula 1.19, we obtain
+
−2 ± 4 − 4(x 2 − 25) +
y = = −1 ± 26 − x 2 .
2
Since y is required to be greater than or equal to −1, we must choose
+
y = −1 + 26 − x 2 .
This is the explicit definition of the function, and it is now easy to find the derivative of y with
respect to x :
dy 1 −x
= √ (−2x) = √ .
dx 2 26 − x 2 26 − x 2
The third equation is quite different, for it is impossible to solve this equation for an explicit
definition of the function. Does this mean that it is also impossible to obtain dy/dx ? The
answer is no! To see this, we differentiate both sides of the equation with respect to x , keeping
in mind that y is a function of x . Only the term in y 5 presents any difficulty, but its derivative
can be calculated using extended power rule 3.21,
dy dy
1 + 5y 4 + 2x + = 0.
dx dx
We can now solve this equation for dy/dx by grouping the two terms in dy/dx on one side
of the equation, and transposing the remaining two terms:
dy
(5y 4 + 1) = −1 − 2x.
dx
Division by 5y 4 + 1 gives the required derivative,
dy 2x + 1
=− 4 .
dx 5y + 1
This process of differentiating an equation that implicitly defines a function is called implicit
differentiation. It could also have been used in each of the first two examples. If y + x 2 −
2x + 5 = 0 is differentiated with respect to x , we have
dy
+ 2x − 2 = 0,
dx
from which, as before,
dy
= −2 x + 2 .
dx
When 2y + x 2 + y 2 − 25 = 0 is differentiated with respect to x , we find that
dy dy
2 + 2x + 2y = 0,
dx dx
from which
dy
(2 + 2y) = −2x,
dx
and therefore
dy −x
= .
dx y+1
Although this result
√ appears different from the previous expression for dy/dx , when we recall
that y = −1 + 26 − x 2 , we find that
dy −x −x
= √ = √ .
dx −1 + 26 − x 2 + 1 26 − x 2
These examples illustrate that if implicit differentiation is used to obtain a derivative, then
the result may depend on y as well as x . Naturally, if we require the derivative at a certain
value of x , then the y -value used is determined by the original equation defining y implicitly
as a function of x .
EXAMPLE 3.24
Assuming that y is defined implicitly as a function of x by the equation
x 3 y 3 + x 2 y + 2x = 12,
find dy/dx when x = 1.

SOLUTION When we differentiate both sides of the equation with respect to x , using the
product rule on the first two terms, we find
dy dy
3x 2 y 3 + 3x 3 y 2 + 2xy + x 2 + 2 = 0.
dx dx
Thus,
dy
(3x 3 y 2 + x 2 ) = −(2 + 2xy + 3x 2 y 3 ) and
dx
dy 2 + 2xy + 3x 2 y 3
=− .
dx 3x 3 y 2 + x 2
When x = 1 is substituted into the given equation defining y as a function of x , the result is
0 = y 3 + y − 10 = (y − 2)(y 2 + 2y + 5),
and the only solution of this equation is y = 2. We now substitute x = 1 and y = 2 into the
formula for dy/dx to calculate the derivative at x = 1,
dy // 2 + 2(1)(2) + 3(1)2 (2)3 30

/ =− 3 2 2
=− .
dx x=1 3(1) (2) + (1) 13
Implicit differentiation can also be used to find second- and higher-order derivatives of functions
that are defined implicitly. Calculations can be messy, but the principles are the same. For
example, when the equation x 5 + y 3 + y 2 = 1 defines y implicitly as a function of x , it is
straightforward to calculate that
dy −5x 4
= .
dx 3y 2 + 2y
To find the second derivative d 2 y/dx 2 , we differentiate both sides of this equation with respect
to x . We use the quotient rule on the right, and when differentiating the denominator we once
again keep in mind that y is a function of x . The result is
! "
2 3 4 dy dy
(3y + 2y)(−20x ) + 5x 6y +2
d 2y dx dx
= .
dx 2 (3y 2 + 2y)2
We now replace dy/dx by its expression in terms of x and y ,

" !
3 2 −5x 4 4
−20x (3y + 2y) + 5x (6y + 2)
d 2y 3y 2 + 2y
= ,
dx 2 (3y 2 + 2y)2
and bring the two terms in the numerator to a common denominator:
−20x 3 (3y 2 + 2y)2 − 25x 8 (6y + 2)

d 2y (3y 2 + 2y)
=
dx 2 (3y 2 + 2y)2
20x 3 (3y 2 + 2y)2 + 25x 8 (6y + 2)
=− .
(3y 2 + 2y)3
It would not be a pleasant task to proceed to the third derivative of y with respect to x .
In the above examples we seem to have adopted the principle that both sides of an equation
can be differentiated with respect to the same variable. This is not always true. For example, if
we differentiate both sides of the equation 4x = 2x , we obtain the ludicrous result that 4 = 2.
Obviously, then, this equation cannot be differentiated with respect to x .
Possibly the reason that differentiation fails in the above example is that the equation
FIGURE 3.38 Ship sailing
northeast
contains only one variable, namely, x . Perhaps a more reasonable question might be: Can every
equation containing two variables be differentiated? To answer this question we consider the
y
situation of a ship heading northeast from a pier. Let us choose east as the positive x -direction
and north as the positive y -direction, both originating from the pier (Figure 3.38). Since the x -
and y -coordinates of the ship change at the same rate, it follows that the rate of change of the
Ship (x, y) y -coordinate of the ship with respect to its x -coordinate when the ship is 2 km from the pier is
equal to 1. We can get this result by noting that the path of the ship is the line y = x , and the
2
√2 derivative of this gives dy/dx = 1. But consider the √following argument.
Pier
When the ship is 2 km from the pier, x = y = 2, and therefore
√2 x
x 2 + y 2 = 4.
If we differentiate this equation with respect to x , we find that
dy dy x
2x + 2y =0 or =− ,
dx dx y
√
and with x = y = 2,
dy
= −1.
dx
Differentiation of the equation x 2 + y 2 = 4, which contains two variables, has led to an
erroneous result.
These two examples have certainly illustrated that not all equations can be differentiated.
What then distinguishes an equation that can be differentiated from one that cannot? Recall
that differentiation is a limiting process. Evaluate whatever is to be differentiated, say f (x) ,
at x + h , subtract its value at x , divide by h , and take the limit as h → 0. If both sides of
an equation are to be differentiated with respect to a variable, then both sides must be equal for
a continuous range of values of that variable. The equation 4x = 2x cannot be differentiated
because it is true only for x = 0. The equation x 2 + y 2 = 4 cannot be differentiated in the ship
example because √ in that example it is valid only when the ship is 2 km from the pier (i.e., only
when x = y = 2). The equation y = x can be differentiated because it is valid at every
point along the path of the ship. This is an extremely important principle, and we will return to
it many times. To emphasize it once again, we may differentiate an equation with respect to a
variable only if the equation is valid for a continuous range of values of that variable.
Each of the three equations at the beginning of this section was given as defining y as a
function of x , and as such defines the function for some domain of values for x . Differentiation
of these equations was therefore acceptable according to the principle stated above.
When we are told that equation 3.23 defines y as a function of x , implicit differentiation
leads to the derivative dy/dx . But how can we tell whether an equation F (x, y) = 0 defines y
implicitly as a function of x ? Additionally, given that F (x, y) = 0 does define y as a function
of x , how do we know that the expression for dy/dx obtained by implicit differentiation is a
valid representation of the derivative of the function? After all, functions do not always have
derivatives at all points in their domains.
Answers to these questions are intimately related. To see how, suppose that an equation
F (x, y) = 0 in x and y is satisfied by a point (x0 , y0 ) , and when the equation is differentiated
implicitly it leads to a quotient for dy/dx ,
dy P (x, y)
= .
dx Q(x, y)
It can be shown that if Q(x0 , y0 ) ,= 0, then the equation F (x, y) = 0 defines y implicitly
as a function of x for some open interval containing x0 , and the derivative of this function
at x0 is P (x0 , y0 )/Q(x0 , y0 ) . Proofs are usually given in advanced books on mathematical
analysis. When Q(x0 , y0 ) = 0, two possibilities exist. First, the equation F (x, y) = 0 might
not define y as a function of x in an interval around x0 . Second, the equation might define
a function, but the function does not have a derivative at x0 . To illustrate, consider first the
equation y 3 − 3y − 2x = 0, which we introduced earlier in this section. The curve defined
by this equation is shown in Figure 3.37, and it clearly illustrates that the equation does not
define y as a function of x . It does define the three functions of x in Figure 3.39. Suppose we
differentiate the equation with respect to x ,
dy dy
3y 2 −3 − 2 = 0,
dx dx
and solve for dy/dx ,
dy 2
= .
dx 3(y 2 − 1)
This derivative is obviously undefined at the points (−1, 1) and (1, −1) , and these are precisely
the points that separate the original curve into three parts. It is impossible to find a portion of
the curve around either of these points that defines y as a function of x . At any other point on
the curve, the formula dy/dx = (2/3)(y 2 − 1)−1 is a valid representation for the derivative
for whichever function of Figure 3.39 contains the point.
FIGURE 3.39 Division of a curve into parts each of which represents a function
y y y
√3
(−1, 1) (−1, 1)
x x x
FIGURE 3.40 Graph with
vertical tangent line at (0, 0) (1, −1) (1, −1)
y
−√3
x = y3
Consider now the equation x = y 3 , shown graphically in Figure 3.40. Implicit differenti-
ation leads to
dy 1
x = .
dx 3y 2
Clearly, y is a function of x for all x , but dy/dx is undefined at (0, 0) since the tangent line
is vertical at this point.
EXAMPLE 3.25
The curve y 2 = x 2 − x 4 is called a lemniscate. It is plotted in Figure 3.41. Use implicit
FIGURE 3.41 Curve with
differentiation to find dy/dx and discuss what happens when the point (0, 0) is substituted into
no tangent line at (0, 0)
the result.
y
SOLUTION Implicit differentiation of y 2 = x 2 − x 4 gives
dy
2y = 2 x − 4x 3 ,
dx
y2 = x 2 − x 4
from which
dy x(1 − 2x 2 )
−1 1 x = .
dx y
This result is undefined at (0, 0) , and Figure 3.41 indicates why. The equation does not define
y as a function of x around x = 0.
EXAMPLE 3.26
Solve Example 1.16 using derivatives.
SOLUTION To the right of the right branch of the hyperbola x 2 − 4y 2 = 5 in Figure 3.42
is a swamp. We are required to find the point P2 at which a straight pipeline from P1 (15, 10)
meets a pipeline along the line x = −1 as far down the line as possible. The required point
P2 occurs when line P1 P2 is tangent to the hyperbola. Let the point of tangency be Q(a, b) .
Differentiation of x 2 − 4y 2 = 5 with respect to x gives
dy dy x
2x − 8y =0 *⇒ = .
dx dx 4y
The slope of the tangent line at Q(a, b) is therefore a/(4b) . Since the slope of P1 Q is
(b − 10)/(a − 15) , it follows that
b − 10 a
= .
a − 15 4b
FIGURE 3.42 Best line along which to build a pipeline
y P1(15, 10)
x 2 − 4y 2 = 5
Swamp
Q (a, b)
x
−10 −5 5 10
P2
−5
x = −1
Cross-multiplication leads to
a 2 − 4b2 = 15a − 40b.
Since Q(a, b) is on the hyperbola, it also follows that
a 2 − 4b2 = 5.
When we subtract these equations,
1
0 = 15a − 40b − 5 *⇒ a = (1 + 8b).
3
Substitution of this into a 2 − 4b2 = 5 gives
1
(1 + 8b)2 − 4b2 = 5 *⇒ 1 + 16b + 64b2 − 36b2 = 45.
9
Thus,
0 = 28b2 + 16b − 44 = 4(7b + 11)(b − 1),
and solutions are b = 1 and b = −11/7. The second solution gives a point on the left branch
of the hyperbola at which a tangent line passes through (15, 10) . The solution b = 1 gives
the point (3, 1) on the right branch of the hyperbola. The slope of the tangent line at (3, 1)
is 3/4, and therefore its equation is y − 1 = (3/4)(x − 3) . It cuts the line x = −1 when
y = 1 + (3/4)(−1 − 3) = −2. Thus, P2 has coordinates (−1, −2) .
EXERCISES 3.8
In Exercises 1–10 find dy/dx wherever y is defined as a function of x . Calculate elasticity for the function defined by each of the following
equations:
4 3
1. y + y = 4x
! "
2. x 4 + y 2 + y 3 = 1 x+1 400y + 200
(a) f (x) = x (b) x =
3. xy + 2x = 4y + 2 2 x+2 3−y
4. 2x 3 − 3xy 4 + 5xy − 10 = 0 ∗ 29. Show that the elasticity of a function is equal to 1 if and only if the
5. x + xy 5 + x 2 y 3 = 3 tangent line to its graph passes through the origin.
2
6. (x + y) = 2x ∗ 30. Find that point P (a, b) on the first-quadrant part of the ellipse
7. x(x − y) − 4y 3 = 2x + 5 2x 2 + 3y 2 = 14 at which the tangent line at P is perpendicular to the
√ line joining P and (2, 5) .
∗ 8. x + y + y 2 = 12x 2 + y
+ ∗ 31. Show that the equation of the tangent line to the hyperbola
∗ 9. 1 + xy − xy = 15 b2 x 2 − a 2 y 2 = a 2 b2 at the point (x0 , y0 ) is b2 xx0 − a 2 yy0 = a 2 b2 .
x y
∗ 10. − =4 ∗ 32. Prove that for any circle (x − h)2 + (y − k)2 = r 2 ,
x+y x
/ /
11. Find equations for the tangent and normal lines to the curve xy 2 +
/
/ d 2 y/dx 2 /
/ 1
/ [1 + (dy/dx)2 ]3/2 / = r .
y 3 = 2 at the point (1, 1) .
∗ 33. A solution passes through a conical filter 24 cm deep and 16 cm

In Exercises 12–15 find d 2 y/dx 2 wherever y is defined as a function across the top into a cylindrical vessel of diameter 12 cm. Find an
of x . equation relating the depth h of solution in the filter and depth H of
∗ 12. x 2 + y 3 + y = 1 ∗ 13. 2x 2 − y 3 = 4 − xy solution in the cylinder. What is the rate of change of h with respect to
H?
∗ 14. y 2 + 2y = 5x ∗ 15. (x + y)2 = x
∗ 34. If x objects are sold at a price of r(x) per object, the total revenue
∗ 16. Find dy/dx at x = 1 when x 3 y + xy 3 = 2. is R(x) = xr(x) . Find the marginal revenue R & (x) if price is defined
implicitly by the equation
In Exercises 17–20 find dy/dx and d 2 y/dx 2 . x = 4a 3 − 3ar 2 + r 3 ,

∗ 17. (x + y)2 = x 2 + y 2 ∗ 18. x 2 y 3 + 2x + 4y = 5
+ where a > 0 is a constant, and 0 < x < 4a 3 .
∗ 19. xy 2 − 3x 2 y = x + 1 ∗ 20. x = y 1 − y 2
∗ 35. The general polynomial of degree n is
√
∗ 21. Find dy/dx when x = 0 if 1 − xy + 3y = 4.
a0 + a1 x + a2 x 2 + · · · + an x n ,
2 3
∗ 22. Find dy/dx when x = 1 if x y + xy = 2.
∗ 23. Find dy/dx and d 2 y/dx 2 when x = 2 if y 5 + (x − 2)y = 1. where a0 , a1 , . . . , an are constants. Show that two polynomials of
degree n ,
∗ 24. Find dy/dx and d 2 y/dx 2 when y = 1 if x 2 + 2xy + 3y 2 = 2.
∗ 25. Find point(s) on the curve xy 2 + x 2 y = 16 at which the slope of a0 + a1 x + a2 x 2 + · · · + an x n = b0 + b1 x + b2 x 2 + · · · + bn x n ,
the tangent line is equal to zero.
∗ 26. Find point(s) on the curve x 2 + y 2/3 = 2 at which the second can be equal for all x if and only if a0 = b0 , a1 = b1 , . . . , an = bn .
derivative is equal to zero.
∗ 36. (a) Find f & (0) if y = f (x) is defined implicitly as a function
∗ 27. If a thermal nuclear reactor is built in the shape of a right circular of x by
cylinder of radius r and height h , then, according to neutron diffusion
theory, r and h must satisfy the equation +
! "2 3 4 x 1 + 2y = x 2 − y.
2.4048 π 2
+ = k = constant.
r h (b) Show that by squaring the equation in part (a), we obtain
Find dr/dh .
x 2 + 4x 2 y = x 4 + y 2 .
∗ 28. The elasticity of a function y = f (x) is defined as
Ey x dy Differentiate this equation with respect to x to find f & (0) .
= .
Ex y dx Do you have any difficulties? Explain.
∗ 37.–40. Use implicit differentiation to redo Exercises 55–58 in Sec- ∗ 48. Verify power rule 3.7 in the case that n is a rational number.
tion 3.7.
∗ 49. Find points on the curve xy 2 + x 2 y = 2 at which the slope of the
∗ 41. (a) Use implicit differentiation to find dy/dx if y 2 =
tangent line is equal to 1.
x 2 − 4x 4 .
(b) Can you calculate dy/dx at x = 0 using the result of part ∗ 50. Show that the families of curves y 2 = x 3 /(a − x) and
(a)? Draw the curve y 2 = x 2 − 4x 4 in order to explain (x 2 + y 2 )2 = b(2x 2 + y 2 ) are orthogonal trajectories.
this difficulty.
√ √ ∗ 51. Let (x, y) be any point on the curve x 2/3 + y 2/3 = a 2/3 , where
∗ 42. (a) Find dy/dx if x + y = 1, where 0 ≤ x ≤ 1, defines
a > 0 is a constant. Find a formula for the length of that part of the
y implicitly as a function of x .
√ √ tangent line at (x, y) between the coordinate axes.
(b) Draw the curve |x| + |y| = 1.
∗ 52. The curve described by the equation x 3 + y 3 = 3axy , where
∗ 43. What is dy/dx if x 2 + 4x + y 2 + 6y + 15 = 0? a > 0 is a constant, is called the folium of Descartes (figure below).
√ √ Find points where the slope of the tangent line is equal to −1.
∗ 44. Find dy/dx if y = u/ u2 − 1, and x 2 u2 + u2 − 1 = 4
defines u implicitly as a function of x . y
∗ 45. Given that the equations x 3 + y3 = 3axy
y 4 + yv 3 = 3 and x 2 v + 3xv 2 = 2x 3 y + 1
define y as a function of v , and v as a function of x , find dy/dx in x

terms of x , y , and v .
∗ 46. Show that any function defined implicitly by the equation
x2
− x = C,
y3 ∗ 53. Use implicit differentiation to solve Exercise 67 in Section 3.7.
where C is a constant, satisfies the equation ∗∗ 54. (a) When a point (x, y) moves in the xy -plane so that the
product of its distances from the points (a, 0) and (−a, 0)
dy is always equal to a constant which we denote by c2 ( c >
3x 2 + y 4 = 2xy.
dx a ), the curve that it follows is called the ovals of Cassini.
Verify that the equation for these ovals is (x 2 +y 2 +a 2 )2 =
∗ 47. Show that any function defined implicitly by the equation c 4 + 4a 2 x 2 .
√
2x 2 − 3y = Cx 2 y 3 , (b) Show that when c < 2a , there are six points on the√ ovals
at which the tangent line is horizontal, but when c ≥ 2a ,
where C is a constant, satisfies the equation there are only two such points.
dy ∗∗ 55. Show that the ellipse b2 x 2 + a 2 y 2 = a 2 b2 and the hyperbola

(xy − x 3 ) + y 2 = 0.
dx d 2 x 2 − c2 y 2 = c2 d 2 intersect orthogonally if a 2 − b2 = c2 + d 2 .
3.9 Derivatives of the Trigonometric Functions

The trigonometric functions, their properties, and the identities that they satisfy were discussed
in Section 1.7. We emphasize an important convention in calculus. Arguments of trigonometric
functions are always measured in radians, never degrees. With this in mind, we prove the
following theorem.
THEOREM 3.10
sin θ
lim =1 (3.24)
θ →0 θ
This is strongly suggested by the machine-generated plot of (sin θ )/θ in Figure 3.43. There
should be an open circle at the point (0, 1) since the function is undefined at θ = 0.
sin θ
FIGURE 3.43 Plot suggesting that lim =1
θ→0 θ
y
1
y=
0.5
−10 −5 5 10
−0.5
−1
PROOF If θ is a positive acute angle as shown in Figure 3.44, then
FIGURE 3.44 Proof that area of triangle BOP < area of sector BOA < area of triangle OBT .
sin θ
lim
θ→0 θ
=1 With θ expressed in radians, the area of a sector of a circle is (r 2 θ )/2. Consequently,
Tangent line 1 1 1
y to circle at B /BP //OP / < (1)2 θ < /OB//BT /.
2 2 2
When we multiply each term in these inequalities by 2 and express lengths of the line segments
B in terms of θ ,
1 (sin θ )(cos θ ) < θ < (1)(tan θ ).
θ T
O P A x Division by sin θ gives
θ 1
cos θ < < ,
sin θ cos θ
Circle of
radius 1 and when each term is inverted, the inequality signs are reversed:
1 sin θ
> > cos θ.
cos θ θ
We now take limits as θ → 0+ . Since cos θ → 1 and 1/ cos θ → 1, the squeeze theorem of
Section 2.1 implies that
sin θ
lim = 1.
θ →0 + θ
Since (sin θ )/θ is an even function (Figure 3.43), it follows that
sin θ
lim = 1.
θ →0 − θ
Since left- and right-hand limits of (sin θ )/θ are both equal to 1, the full limit as θ → 0 is
also 1,
sin θ
lim = 1.
θ →0 θ
With this result we can now find the derivative of the sine function. Derivatives of the other
five trigonometric functions follow easily.
THEOREM 3.11
d
sin x = cos x (3.25)
dx
PROOF If we set f (x) = sin x in equation 3.3 and use the trigonometric identity
! " ! "
A+B A−B
sin A − sin B = 2 cos sin ,
2 2
we obtain
f (x + h) − f (x)
f & (x) = lim
h→0 h
sin (x + h) − sin x
= lim
h→0 h
# ! " ! "$
1 2x + h h
= lim 2 cos sin
h→0 h 2 2
#! " $
h sin (h/2)
= lim cos x +
h→0 2 h/2
! "
h sin (h/2)
= lim cos x + lim
h→0 2 h/2→0 h/2
= cos x.
It is a simple application of the chain rule to prove the following corollary.
COROLLARY 3.11.1
If u = f (x) is a differentiable function,
d du
sin u = cos u . (3.26)
dx dx
EXAMPLE 3.27
Find derivatives for the following functions:
(a) f (x) = sin 3x (b) f (x) = sin3 4x
SOLUTION
(a) According to formula 3.26,
d
f & (x) = cos 3x (3x) = 3 cos 3x.
dx
(b) For this function we must use extended power rule 3.21 before equation 3.26,
d d
f & (x) = 3 sin2 4x sin 4x = 3 sin2 4x cos 4x (4x) = 12 sin2 4x cos 4x.
dx dx
Since the cosine function can be expressed in terms of the sine function, it is straightforward to
find its derivative.
THEOREM 3.12
d
cos x = − sin x (3.27)
dx
PROOF Since cos x can always be expressed in the form cos x = sin (π/2 − x) , it follows
that
d d 3π 4 3π 4
cos x = sin − x = − cos − x = − sin x.
dx dx 2 2
COROLLARY 3.12.1
If u = f (x) is a differentiable function,
d du
cos u = − sin u . (3.28)
dx dx
Derivatives of the other trigonometric functions are obtained by expressing them in terms
of the sine and cosine functions. For the tangent function, we use the chain rule and the quotient
rule to calculate
! "
d d sin u du cos u(cos u) − sin u(− sin u) du 1 du
tan u = = 2
= .
dx du cos u dx cos u dx cos2 u dx
Consequently,
d du
tan u = sec2 u . (3.29)
dx dx
A similar calculation gives
d du
cot u = − csc2 u . (3.30)
dx dx
For the secant function, we obtain
! "
d d 1 du du 1 sin u du
sec u = = (−1)(cos u)−2 (− sin u) =
dx du cos u dx dx cos u cos u dx
or
d du
sec u = sec u tan u . (3.31)
dx dx
Similarly,
d du
csc u = − csc u cot u . (3.32)
dx dx
Notice the relationship between derivatives of the cofunctions — cosine is the cofunction
of sine, cotangent of tangent, and cosecant of secant — and corresponding derivatives for the
functions. Each function is replaced by its cofunction and a negative sign is added.
EXAMPLE 3.28
Find dy/dx if y is defined as a function of x in each of the following:
(a) y = sin 2x (b) y = 4 sec (2x 3 + 5)
2
(c) y = tan 4x (d) y = 2 csc 3x 2 + 5x sin x
2
(e) x tan y + y sin x = 5
SOLUTION
dy d
(a) = (cos 2x) (2x) = 2 cos 2x
dx dx
dy d
(b) = 4 sec (2x 3 + 5) tan (2x 3 + 5) (2x 3 + 5)
dx dx
= 24x 2 sec (2x 3 + 5) tan (2x 3 + 5)
dy d
(c) = 2 tan 4x tan 4x = 2 tan 4x(4 sec2 4x) = 8 tan 4x sec2 4x
dx dx
dy d
(d) = −2 csc 3x 2 cot 3x 2 (3x 2 ) + 5 sin x + 5x cos x
dx dx
= −12x csc 3x 2 cot 3x 2 + 5 sin x + 5x cos x
(e) If we differentiate the equation (implicitly) with respect to x , we obtain
dy dy
2x tan y + x 2 sec2 y + sin x + y cos x = 0.
dx dx
When we solve this equation for dy/dx , the result is
dy 2x tan y + y cos x
=− 2 2 .
dx x sec y + sin x
EXAMPLE 3.29
In the mechanism of Figure 3.45, crank OA rotates at one thousand revolutions per minute
(rpm). Find an expression for the velocity of the piston in terms of R and θ .
FIGURE 3.45 Velocity of SOLUTION If the crank starts with θ = 0 at time t = 0, then x and y coordinates of the
piston in crank mechanism moving end of the crank are x = R cos θ and y = R sin θ . The position of the piston relative
to O is x + L = R cos θ + L . The velocity of the piston is therefore
A
R dθ
θ v = −R sin θ .
O
dt
Since the crank rotates at 1000 rpm,
dθ 2π(1000) 100π
L = = radians per second.
dt 60 3
Thus,
100π
v =− R sin θ m/s,
3
provided R is in metres.
EXAMPLE 3.30
The 1 kg mass in Figure 3.46 is pulled 10 cm to the right of its position when the spring is un-
stretched (called the equilibrium position), and given velocity 3 m/s to the left. Its displacement
from the equilibrium position thereafter is given by
1 3
x(t) = cos 20t − sin 20t m, t ≥ 0.
10 20
Find all times when the velocity is zero and all times when the acceleration is zero.
FIGURE 3.46 Velocity and acceleration of oscillating mass
k = 400
m=1
x
x=0
FIGURE 3.47 Simple harmonic motion of oscillating mass
x
0.15
0.1
0.05
t
0.5 1 1.5 2
−0.05
−0.1
−0.15
SOLUTION The displacement function in Figure 3.47 indicates that the mass oscillates back
and forth about the equilibrium position. This is called simple harmonic motion. It is a direct
result of the fact that the mass moves along a frictionless surface and no account has been taken
of air resistance.
To determine times at which velocity and acceleration vanish, it is advantageous to follow
the lead of Example 1.36 in Section 1.7 and express x(t) in the form
1 3
cos 20t − sin 20t = A sin (20t + φ).
10 20
When we expand the right side with a compound-angle formula, we obtain
1 3
cos 20t − sin 20t = A(sin 20t cos φ + cos 20t sin φ).
10 20
This equation will be valid for all t ≥ 0 if we choose A and φ to satisfy
1 3
= A sin φ, − = A cos φ.
10 20
Squaring and adding these equations eliminates φ :
√
1 9 2 13
+ =A *⇒ A=± .
100 400 20
If we choose to use the positive value, then
2 3
sin φ = √ and cos φ = − √ .
13 13
When we let φ = 2.55 be the smallest positive solution of these equations, we express x(t) in
the form √
13
x(t) = sin (20t + φ) m.
20
Velocity and acceleration are

√ √
v(t) = 13 cos (20t + φ) m/s, a(t) = −20 13 sin (20t + φ) m/s2 .
The velocity of the mass is zero when the mass is at the points farthest from the equilibrium
position. It occurs when
√ π
0 = 13 cos (20t + φ) *⇒ 20t + φ = + nπ,
2
where n is an integer. Thus,
π nπ φ
t = + − ,
40 20 20
and these are positive for n ≥ 1. The acceleration vanishes when
√
0 = −20 13 sin (20t + φ) *⇒ 20t + φ = nπ,
where n is an integer. Thus,

nπ φ
t = − ,
20 20
and this is positive for n ≥ 1. Notice that acceleration is zero when the mass passes through
the equilibrium position. This is to be expected because at x = 0, the spring has no stretch and
therefore exerts no force on the mass.
EXERCISES 3.9
In Exercises 1–30 find dy/dx . ∗ 24. y = sin4 x 2 − cos4 x 2
1. y = 2 sin 3x 2. y = cos x − 4 sin 5x ∗ 25. y = u3 sec u, u = x tan (x + 1)

√ √
∗ 26. y = 3 − sec v, v = tan x
3. y = sin2 x 4. y = tan−3 3x
+
∗ 27. y = t 2 + 1, t = sin (sin x)
5. y = sec4 10x 6. y = csc (4 − 2x) +
∗ 28. y = (1 + sec3 u)1/3 , u = 1 + cos x 2
7. y = sin2 (3 − 2x 2 ) 8. y = x cot x 2
sin2 x
∗ 29. y =
1 + cos3 x
sin 2x x sin x
9. y = 10. y =
cos 5x x+1 1 + tan3 (3x 2 − 4)
∗ 30. y =
x 2 sin x
11. y = sin3 x + cos x 12. y = sin 2x cos 2x
√ ) * 1/ 4 In Exercises 31–32 find d 2 y/dx 2 .

13. y = sin 3x ∗ 14. y = 1 + tan3 x
∗ 31. sin y = x 2 + y ∗ 32. tan y = x + xy
∗ 15. 2 sin y + 3 cos x = 1 ∗ 16. x cos y − y cos x = 3
In Exercises 33–38 evaluate the limit, if it exists.
∗ 17. 4 sin2 x − 3 cos3 y = 1 ∗ 18. tan (x + y) = y
tan x 1 − cos x
33. lim ∗ 34. lim
x→0 x x→0 x
∗ 19. x + sec xy = 5 ∗ 20. x 3 y + tan2 y = 3x
sin 2x sin (2/x)
∗ 35. lim ∗ 36. lim
3
∗ 21. x = y csc y 3
∗ 22. y = cos (tan x) x→0 x x→∞ sin (1/x)
(x + 1)2 sin x cos x

3 2 ∗ 37. lim ∗ 38. lim
∗ 23. y = x − x cos x + 2x sin x + 2 cos x x→0 3x 3 x→π/2 (x − π/2)2
∗ 39. Draw graphs of the functions f (x) = | sin x| and g(x) = sin |x| . ∗ 43. The two identical cranks in the figure below rotate at constant
Where do these functions fail to be differentiable? angular speeds ω1 and ω2 radians per second, driving the attachment
vertically. Find the angular speed dθ/dt of the slotted bar in terms of
∗ 40. Does limx→0 g & (x) exist for the function g(x) in Exercise 32 of α , β , ω1 , and ω2 .
Section 3.3?
∗ 41. The angular displacement of the pendulum in the following figure θ

at time t is given by
θ = f (t) = A cos (ωt + φ), t ≥ 0,
α β
R R
ω1 ω2
3R
∗ 44. Find all positive values of x for which the derivative of the function
f (x) = cos (x + 1/x) is equal to zero.
where A , ω , and φ are constants. Verify that θ = f (t) satisfies the
equation
∗ 45. An elliptic cam rotates counterclockwise at 600 rpm (figure below).
d 2θ Find a formula for the velocity of the follower if the ellipse has major
+ ω 2 θ = 0. and minor axes of lengths 2a and 2b .
dt 2
∗ 42. When the mass m in the figure below moves vertically on the end
of the spring, its displacement y must satisfy the equation
d 2y
m + ky = 0,
dt 2
where k > 0 is the constant of elasticity for the spring. Verify that the
function ∗∗ 46. Find a formula for the following derivative, simplified as much as
possible,
;9 < ;9 <
k k d
y = f (t) = A sin t + B cos t | sin x|n , where n ≥ 1 is an integer.
m m dx
∗∗ 47. Show that the function

satisfies this equation for any constants A and B .
.
2
f (x) = x sin (1/x), x ,= 0
0, x =0
has a derivative at x = 0, but f & (x) is not continuous at x = 0.
∗∗ 48. Any cross-section of the reflector in a car headlight is in the form

of a parabola, y 2 = 4x , 0 ≤ x ≤ 2, with the bulb at the point
y (1, 0) called the focus of the parabola. It is a principle of optics that
all light rays from the bulb are reflected by the mirror so that the angle
between the incident ray and the normal to the mirror is equal to the
m angle between the reflected ray and the normal. Show that all rays are
y = 0 at
equilibrium reflected parallel to the x -axis.
3.10 Derivatives of the Inverse Trigonometric Functions

Derivatives of inverse trigonometric functions are most easily obtained with implicit differenti-
ation. We begin with the inverse sine function.
THEOREM 3.13
d 1
Sin −1 x = √ . (3.33)
dx 1 − x2
PROOF If we set y = Sin −1 x , then x = sin y , and we can differentiate implicitly with
respect to x :
dy
1 = cos y .
dx
Thus,
dy 1
= .
dx cos y
To express cos y in terms of x we can proceed in two ways:
(i) From the trigonometric identity sin2 y + cos2 y = 1, we obtain
: +
cos y = ± 1 − sin2 y = ± 1 − x 2 .
We know that y = Sin −1 x and the principal values of Sin −1 x are −π/2 ≤ y ≤
π/2. Therefore, y is an angle in either the first or fourth quadrant. It follows that
cos y ≥ 0, and we must choose
+
cos y = 1 − x2.
FIGURE 3.48 Triangle to
fit equation x = sin y
(ii) We use the triangle in Figure 3.48 to replace the trigonometric identity.
√ The triangle
is designed to fit the equation x = sin y . The third side is then 1 − x 2 . It follows
that +
1
x cos y = 1 − x2.
y To ensure that cos y is indeed positive, we resort once again to the fact that y =
Sin −1 x and principal values are −π/2 ≤ y ≤ π/2. Finally, then, we have
dy 1
= √ .
dx 1 − x2
It is a straightforward application of the chain rule to obtain the following corollary.
COROLLARY 3.13.1
If u(x) is a differentiable function, then
d 1 du
Sin −1 u = √ . (3.34)
dx 1−u 2 dx
3.10 Derivatives of the Inverse Trigonometric Functions 209
Next we obtain the derivative of the inverse cosine function.
THEOREM 3.14
d −1
Cos −1 x = √ (3.35)
dx 1 − x2
PROOF If we set y = Cos −1 x , then x = cos y , and we can differentiate with respect to x :
dy
1 = − sin y ,
dx
FIGURE 3.49 Triangle to
fit equation x = cos y and solve for
dy −1
= .
dx sin y
1 √
1 − x2 The triangle in Figure 3.49, obtained from x = cos y , yields sin y = 1 − x 2 . Since the
principal values of y = Cos −1 x are 0 ≤ y ≤ π , it follows that sin y is indeed nonnegative,
y and therefore
x
dy −1
= √ .
dx 1 − x2
COROLLARY 3.14.1
If u(x) is differentiable, then
d −1 du
Cos −1 u = √ . (3.36)
dx 1 − u2 dx
Derivatives for the other inverse trigonometric functions can be derived in a similar way.
We list them below and include derivatives of Sin −1 u and Cos −1 u for completeness:
d 1 du
Sin−1 u = √ ; (3.37a)
dx 2
1 − u dx
d −1 du
Cos−1 u = √ ; (3.37b)
dx 1 − u2 dx
d 1 du
Tan−1 u = ; (3.37c)
dx 1 + u2 dx
d −1 du
Cot−1 u = ; (3.37d)
dx 1 + u2 dx
d 1 du
Sec−1 u = √ ; (3.37e)
dx 2
u u − 1 dx
d −1 du
Csc−1 u = √ . (3.37f)
dx u u2 − 1 dx
Note that the derivative of an inverse cofunction is the negative of the derivative of the cor-
responding inverse function; that is, derivatives of Cos −1 u , Cot −1 u , and Csc −1 u are the
negatives of the derivatives of Sin −1 u , Tan −1 u , and Sec −1 u .
EXAMPLE 3.31
= >2
(a) y = Sec−1 (x 2 ) (b) y = 2x Cot−1 (3x) (c) y = Sin−1 (x 3 )
Sin−1 x
(d) y = (e) Cos−1 (xy) + x 2 y + 5 = 0
Cos−1 x
SOLUTION
dy 1 d 2 2x 2
(a) = + (x ) = √ = √
dx x 2 (x 2 )2 − 1 dx 2
x x −14 x x4 − 1
! "
dy −1 6x
(b) = 2 Cot −1 (3x) + 2x 2
(3) = 2 Cot −1 (3x) −
dx 1 + 9x 1 + 9x 2
dy d 1 6x 2 Sin −1 (x 3 )
(c) = 2 Sin −1 (x 3 ) Sin −1 (x 3 ) = 2 Sin −1 (x 3 ) √ (3x 2 ) = √
dx dx 1 − x6 1 − x6
! " ! "
1 −1
Cos −1 x √ − Sin −1 x √
dy 1−x 2 1 − x2 Cos −1 x + Sin −1 x
(d) = ) −1 *2 =√ ) *2
dx Cos x 1 − x 2 Cos −1 x
(e) Differentiation with respect to x gives
! "
−1 dy dy
+ y+x + 2xy + x 2 = 0.
1 − x2y2 dx dx
Thus,
; <
dy −x 2 y
+ +x = + − 2xy,
dx 1− x2y2 1 − x2y2
from which
+ + +
dy y − 2xy 1 − x 2 y 2 1 − x2y2 y − 2xy 1 − x 2 y 2
= + + = + .
dx 1 − x2y2 −x + x 2 1 − x 2 y 2 −x + x 2 1 − x 2 y 2
EXAMPLE 3.32
Find the derivative of the function f (x) = Sin −1 x + Cos −1 x . What is your conclusion?
SOLUTION The derivative of the function is
1 −1
f & (x) = √ +√ = 0.
1− x2 1 − x2
Since this result is valid for the entire domain −1 ≤ x ≤ 1 of f (x) , it follows that the function
f (x) must be a constant,
f (x) = C = a constant.
Since f (0) = Sin −1 (0) + Cos −1 (0) = 0 + π/2 = π/2, it follows that C = π/2.
Consequently,
π
Sin −1 x + Cos −1 x = .
2
EXERCISES 3.10
3x 4 √
In Exercises 1–30 y is defined as a function of x . Find dy/dx in as 1 −1 x2 − 9
22. y = Csc −
simplified a form as possible. 3 3 x2
3x 4 +
1. y = Cos −1 (2x + 3) 23. y = x Cos −1 − 4 − x2
2
√
2. y = Cot −1 (x 2 + 2) Csc −1 (3x) 9x 2 − 1
24. y = −
x x
3. y = Csc −1 (3 − 4x) +
2 −1
25. y = x Sec x− −1 x2
4. y = Tan −1 (2 − x 2 ) ) * +
2
26. y = x Cos −1 x − 2x − 2 1 − x 2
5. y = Sec −1 (3 − 2x 2 )
27. y = (1 + 9x 2 ) Cot −1 (3x) + 3x
6. y = x Csc −1 (x 2 + 5) + ! "
2 −1 x − 2
∗ 28. y = (x − 2) 4x − x + 4 Sin
7. y = (x 2 + 2) Sin −1 (2x) 2
√ +
8. y = Tan −1 x + 2 ∗ 29. y = (2x 2 − 1) Sin −1 x + x 1 − x 2
+ ; √ <
9. y = Sin −1 1 − x 2 −1 2x
∗ 30. y = Tan √
+ 1 + x4
10. y = Cot −1 x 2 − 1
= >2 ∗ 31. Evaluate the derivative of
11. y = Tan −1 (x 2 ) +
f (x) = Sec −1 x + Cot −1 x 2 − 1.
12. y = x 2 Sec −1 x
) * What is your conclusion?
13. y = tan 3 Sin −1 x ∗ 32. (a) Show that if the principal values of f (x) = Sec −1 x are
chosen as 0 ≤ y ≤ π , y ,= π/2, then the derivative of
14. y = Cot −1 [(1 + x)/(1 − x)]
the function is
15. y = Csc −1 (1/x) d 1
Sec −1 x = √ .
−1
16. y = Sin [(1 − x)/(1 + x)] dx |x| x 2 − 1
17. y = Tan −1 (u2 + 1/u), u = tan (x 2 + 4) (b) What is the derivative of f (x) = Csc −1 x if its principal
√ values are chosen as −π/2 ≤ y ≤ π/2, y ,= 0?
18. y = t Cos −1 t, t = 1 − x2
∗ 33. Find the angle between the curves y = Sin −1 x and y = Cos −1 x
2 −1
19. y sin x + y = Tan x at their point of intersection.
20. Sin −1 (xy) = 5x + 2y

∗ 34. Verify the results in equations 3.37c–f.
+ ∗ 35. If the crank in Exercise 64 of Section 1.8 rotates at 300 rpm, find
21. y = x 2 − 1 − Sec −1 x an expression for dθ/dt .
3.11 Derivatives of the Exponential and Logarithm

Functions
Exponential functions abound in applied mathematics, and much of this is due to the fact that
the derivative of an exponential function always returns the exponential function. We obtain
derivatives of exponential and logarithm functions in this section, but many of the applications
that rely on these derivatives are delayed until Section 5.5. Exponential and logarithm functions
and their properties were discussed in Section 1.9. We assume that the reader is familiar with this
material. To differentiate logarithm and exponential functions, we must return to the definition
of the derivative (see equation 3.3 in Section 3.1). We begin with the natural logarithm function
ln x .
THEOREM 3.15
The derivative of the logarithm function ln x is
d 1
ln x = . (3.38)
dx x
d ln (x + h) − ln x
ln x = lim
dx h→0 h
# ! "$
1 x+h
= lim ln [since ln b − ln c = ln (b/c)]
h→0 h x
# ! "$
1 h
= lim ln 1 +
h→0 h x
# ! "$
1 x h
= lim ln 1 +
x h→0 h x
! "x/ h
1 h
= lim ln 1 + (since c ln b = ln bc ).
x h→0 x
Since the logarithm function is continuous, we may interchange the limit and logarithm opera-
tions (see Theorem 2.5 in Section 2.4),
7 ! " 8
d 1 h x/ h
ln x = ln lim 1 + .
dx x h→0 x
But if we set v = x/ h , then
! "x/ h ! "x/ h ! "v ! "v

h h 1 1
lim 1+ = lim 1+ = lim 1+ = lim 1+ =e
h→0 x h/x→0 x 1/v→0 v v→∞ v
(see expression 1.68), and therefore
d 1
ln x = .
dx x
The chain rule gives the following corollary.
COROLLARY 3.15.1
If u(x) is a differentiable function, then
d 1 du
ln u = . (3.39)
dx u dx
In the rare instance that we might need to differentiate the logarithm function to base a ,= e ,
namely, loga x , we rewrite the change of base formula 1.67 with a replacing b and e replacing
x,
loga x = (ln x)(loga e).
Differentiation with respect to x gives the next corollary.
COROLLARY 3.15.2
d 1
loga x = loga e (3.40)
dx x
The chain rule gives the next result.
COROLLARY 3.15.3
If u(x) is a differentiable function of x , then
d 1 du
loga u = loga e. (3.41)
dx u dx
We now use implicit differentiation to find the derivatives of exponential functions.
THEOREM 3.16
d x
e = ex (3.42)
dx
PROOF If we set y = ex , then x = ln y , and we can differentiate (implicitly) with respect

to x using equation 3.39:
1 dy
1 = .
y dx
Thus,
dy
= y = ex .
dx
The exponential function ex is therefore its own derivative. In fact, for any constant C
whatsoever, the function Cex differentiates to give itself. In Chapter 5 we shall see that Cex is
the only function that is its own derivative. The chain rule gives the derivative of eu when u is
a function of x .
COROLLARY 3.16.1
For a differentiable function u(x) ,
d u du
e = eu . (3.43)
dx dx
We can find the derivative of exponential functions a x with bases other then e by writing
d x d , ln (a x ) - d x ln a
a = e = e = ex ln a (ln a) = a x ln a.
dx dx dx
We state this as a corollary.
COROLLARY 3.16.2
d x
a = a x ln a (3.44)
dx
The chain rule now gives
COROLLARY 3.16.3
For a differentiable function u(x) ,
d u du
a = au ln a. (3.45)
dx dx
EXAMPLE 3.33
(a) y = 23x (b) y = log10 (3x 2 + 4) (c) y = x 2 e −2 x
√
(d) y = (ln x)/x (e) y = 1 + e 2x
SOLUTION
(a) Using equation 3.45 yields
dy d
= 23x (3x) ln 2 = (3 ln 2)23x .
dx dx
(b) With equation 3.41, we have
dy 1 d 6x
= 2
(3x 2 + 4) log10 e = 2
log10 e.
dx 3x + 4 dx 3x + 4
(c) The product rule and equation 3.43 give
dy
= 2xe−2x + x 2 e−2x (−2) = 2x(1 − x)e−2x .
dx
(d) The quotient rule and equation 3.38 yield
dy x(1/x) − ln x 1 − ln x
= 2
= .
dx x x2
(e) With extended power rule 3.21 and 3.43, we obtain
dy 1 d 2 e 2x e 2x
= √ (1 + e 2 x ) = √ = √ .
dx 2 1 + e2x dx 2 1 + e 2x 1 + e 2x
EXAMPLE 3.34
Find values of x for which the first derivative of the function f (x) = x 2 ln x is equal to zero,
and values of x for which the second derivative is equal to zero.
FIGURE 3.50 Graph of y = x 2 ln x to illustrate where tangent line is horizontal
0.6
y = x 2 ln x
0.4
0.2
0.5 1 1.5 x
−0.2
SOLUTION A graph of the function is shown in Figure 3.50. It suggests that the tangent line
is horizontal at a value of x near 1/2. To find it we set
! "
& 2 1
0 = f (x) = 2x ln x + x = x(2 ln x + 1).
x
Since the function is undefined at x = 0, we must have
1
2 ln x + 1 = 0, or ln x = − .
2
√
Thus, x = e−1/2 = 1/ e .
The second derivative is equal to zero when
! "
&& 2
0 = f (x) = (2 ln x + 1) + x = 2 ln x + 3.
x
The only solution of this equation is x = e−3/2 . The geometric significance of the point on the
graph at which f && (x) = 0 will be discussed in Section 4.4. Can you see it?
In Section 3.2 we proved power rule 3.7 only in the case that n is a nonnegative integer. It is
now easy to prove it for all real n , at least when x > 0. To do this we write x n in the form
x n = en ln x (see equation 1.72a). Formulas 3.43 and 3.38 give
d n d n ln x d 3n4
x = e = en ln x (n ln x) = x n = nx n−1 .
dx dx dx x
A discussion of the power rule in the case that x ≤ 0 depends on the value of n (see Exer-
cise 49).
EXAMPLE 3.35
If air resistance proportional to velocity and friction with the surface on which the mass slides
are taken into account in the mass-spring system of Example 3.30, the displacement function is
of the form
7 √ √ 8
−t/2 1 1599t 1 1599t 1
x(t) = e cos − sin + ,
10 2 8 2 4000
from time t = 0 when motion begins until the mass comes to rest for the first time. Find this
time.
FIGURE 3.51 Displacement of mass taking friction and air resistance into account
0.1
0.1 0.2 0.3 0.4 t
−0.1
SOLUTION A plot of x(t) is shown in Figure 3.51, but it can only be used until the tangent
line is horizontal for the first time. This is for t ≈ 0.1. To find it more accurately, we set the
velocity equal to 0,
7 √ √ 8
1 −t/2 1 1599t 1 1599t
0 = − e cos − sin
2 10 2 8 2
7 √ √ √ √ 8
−t/2 1599 1599t 1599 1599t
+ e − sin − cos .
20 2 16 2
Of the infinity of possible solutions to this equation, only the smallest positive one is acceptable
here. To find it we simplify the equation to
7 √ √ 8
1 −t/2
√ 1599t √ 1599t
0 = − e (4 + 5 1599) cos + (4 1599 − 5) sin .
80 2 2
This equation implies that

√ √
1599t 4 + 5 1599
tan = √ .
2 5 − 4 1599
√ √
The smallest positive angle with tangent equal to (4 + 5 1599)/(5 − 4 1599) is
√ ; √ <
1599 t −1 4 + 5 1599
= Tan √ = 2.2205 radians.
2 5 − 4 1599
Consequently, the time when the mass stops moving to the left is
2
t = √ (2.2205) = 0.111 s.
1599
Application Preview The container in Figure 3.52 represents a chemical reactor in which a chemical is either created
Revisited or broken down. The chemical enters in the form of a solution at one concentration and leaves
the reactor at a different concentration. We assume that flow rates in and out are the same,
say q cubic metres per second, and therefore the volume V of solution in the reactor remains
constant. Let us assume that Ci is the concentration in kilograms per cubic metre of the chemical
entering the reactor beginning at time t = 0, and there is initially no chemical in the reactor.
As the chemical is created (or broken down) in the reactor, its concentration varies throughout
the reactor so that concentration depends not only on time but also on position. This makes
the problem of predicting concentration of chemical leaving the reactor far more difficult than
we can handle at this time. Suppose we add a mixer that is so efficient that concentration of
chemical is the same at every point in the reactor. Concentration then depends only on time t ,
denoted by C(t) , and this is also the concentration at which chemical leaves the reactor. Our
problem then is to find C(t) . Now, concentration of the chemical in the reactor is the amount
A(t) of chemical divided by V , the volume of solution. We find it easier to work with A(t)
than C(t) .
FIGURE 3.52 Schematic for chemical reactor
q q
The derivative dA/dt is the rate of change of the amount of chemical in the reactor at any
given time. This must be equal to the rate at which chemical enters less the rate at which it
leaves, plus (or minus) the rate at which it is created (broken down),
dA 1 ? 1 ? 1 ?
rate at which rate at which rate at which chemical
= − + .
dt chemical enters chemical leaves is created or broken down
The rate at which chemical enters is qCi . The rate at which it leaves is q(A/V ) . For many
chemical reactions it has been shown experimentally that the rate at which chemical is being
created or broken down at any given time is proportional to how much chemical is present at
that time. Thus, the rate at which it is created or broken down is kA , where k is a constant
( k > 0 for chemical formation, k < 0 for chemical breakdown). When we substitute these
expressions into the above equation, we obtain
dA qA 3 q4
= qCi − + kA = qCi + k − A.
dt V V
This is the equation that the amount of chemical A(t) must satisfy at any given time; it is known
as a differential equation. A differential equation is an equation that contains one or more
derivatives of a function, and the objective is to solve the equation for functions that satisfy the
equation. For the equation above, we must find functions A(t) whose first derivatives are equal
to k − q/V times themselves plus a constant qCi . Differential equations are discussed in detail
in Section 5.5 and Chapter 15.
But we also know that the initial amount of chemical in the reactor is zero so that the
function A(t) must also satisfy the condition A(0) = 0. In other words, A(t) must satisfy
dA 3 q4
= k− A + qCi , A(0) = 0.
dt V
This is known as an initial-value problem. It consists of the differential equation and an initial
condition for the unknown function A(t) . We shall find that there are many functions satisfying
the differential equation, but only one of them also satisfies the initial condition.
Since k − q/V is a constant, equation 3.43 indicates that the derivative of the function
e(k−q/V )t is k − q/V times itself. Furthermore, if we multiply e(k−q/V )t by any constant D
whatsoever, derivatives of the functions De(k−q/V )t are k − q/V times themselves. This must
be a part of A(t) . The other part is simple in form but harder to discover. If we express the
differential equation in the form
dA
− (k − q/V )A = qCi ,
dt
we need a function that yields qCi when substituted into the left side of the equation. The
constant function −qCi /(k − q/V ) does it. Putting these together, we suspect that A(t) must
be of the form
qCi
A(t) = De(k−q/V )t − .
k − q/V
It is straightforward to check that the derivative of this function does indeed satisfy the required
differential equation for any constant D . It remains only to choose D so that A(t) satisfies the
initial condition A(0) = 0. This requires
qCi qCi
0 = D− *⇒ D = .
k − q/V k − q/V
Consequently, the amount of chemical exiting the reactor at any given time is
qCi qCi qCi V = >

A(t) = e(k−q/V )t − = 1 − e(k−q/V )t .
k − q/V k − q/V q − kV
The concentration of chemical exiting the reactor is
qCi = >
C(t) = 1 − e(k−q/V )t .
q − kV
The analysis in the above Application Peview Revisited will help us in our next, more
challenging consultation.
Figure 3.53 is a schematic for a heat tank. Water enters at constant temperature 10◦ C, the
water is heated, and it then leaves at higher temperature. The tank is perfectly insulated so
that no heat can escape from its sides, and therefore all heat supplied by the heater raises
the temperature of the water. When the tank is full, and this is always the case, the mass
of water is 100 kg. Water enters the tank at a rate of 3/100 kg/s, and leaves at the same
rate. The heater adds energy to the water at the rate of 2000 joules per second (J/s). We
are asked to determine whether the temperature of the water in the tank just keeps rising
and rising, or whether it somehow levels off after a long time.
SOLUTION Temperature of the water in the tank depends on both time and position
in the tank. To remove spatial dependence (otherwise the problem will be impossible to
solve), we add a mixer assumed so efficient that temperature of the water is the same at
every point in the tank. Temperature then depends only on time t , denoted by T (t) , and
this is also the temperature at which water leaves the tank. If we can find a formula for
T (t) , we will know what happens to the system.
FIGURE 3.53 Schematic for heat tank
Cool liquid
Heater
Perfect insulation
Mixer Hot liquid
Where do we start? We want temperature, temperature is due to heat, and heat is a

form of energy. We have energy entering the tank in water at temperature 10◦ , energy
being added by the heater, energy raising the temperature of the water, and energy leaving
in water at temperature T (t) . (For simplicity, we will ignore energy associated with
the mixer. It will have minimal effect anyway.) Since we can calculate rates for these
energies, and the tank is perfectly insulated, we must be able to establish an energy
balance equation for the system. We feel that it should take the form
 
6 @ 6 @ 6 @  rate at which 
rate at which rate at which rate at which  
energy is used to raise
energy enters tank + energy is added = energy leaves tank + .
in cool water by heater in warmer water

 temperature of  
water in tank
The first three rates are easily calculated; the fourth is more difficult. The second term on
the left is 2000. For the remaining terms, we must introduce the specific heat cp = 4190
J/kg · C of water. It is the energy required to raise the temperature of 1 kg of water by 1◦ C.
(To raise 5 kg of water 10◦ C, say, requires (5)(10)(4190) = 209 500 J.) Since 3/100 kg
of water at temperature 10◦ C enter the tank each second, the rate at which energy enters
the tank due to this water is (3/100)(10)(4190) = 1257 J/s. (This is the energy required
to raise 3/100 kg of water from 0◦ C to 10◦ C.) In a similar way, the rate at which energy
leaves the tank in the water at temperature T is (3/100)(T )(4190) = 1257T /10 J/s.
This leaves only the last term on the right side of the energy balance equation. The rate of
change of temperature of the 100 kg of water in the tank is dT /dt ; that is, the temperature
changes dT /dt degrees each second. It follows that the rate at which energy is used to
raise the temperature of this water is (100)(4190)(dT /dt) .
When these rates are substituted into the energy balance equation, the result is
1257T dT
1257 + 2000 = + 419 000 .
10 dt
This is another example of a differential equation; find the function T (t) that satisfies the
equation. We also need to specify the initial temperature of the water in the tank. We are
told that when the heater is turned on, temperature of the water in the tank is the same as
the incoming water, namely 10◦ C. Let us choose this to be time t = 0. The initial-value
problem for T (t) is then
dT 3T 3257
=− + , T (0) = 10.
dt 10 000 419 000
Equation 3.43 suggests that to account for the term −3T /10 000 on the right side
of the differential equation, e−3t/10 000 should be involved. Indeed, the derivative of the
function De−3t/10 000 is −3/10 000 times itself for any constant D whatsoever. This
must be a part of T (t) . The other part is simple in form but harder to discover. If we
express the differential equation in the form
dT 3T 3257
+ = ,
dt 10 000 419 000
we need a function that yields 3257/419 000 when substituted into the left side of the equa-
3257 10 000 32 570
tion. The constant function · = does it. Putting these together,
419 000 3 1257
we suspect that temperature of the water in the tank must be of the form
32 570
T (t) = De−3t/10 000 + .
1257
It is straightforward to check that the derivative of this function does indeed satisfy the
required differential equation for any constant D . It remains only to choose D so that
T (t) satisfies the initial condition T (0) = 10. This requires
32 570 −20 000
10 = D + *⇒ D = .
1257 1257
Consequently, temperature of the water at any time t is
32 570 20 000
D(t) = − e−3t/10 000 .
1257 1257
The graph of T (t) is shown in Figure 3.54. It begins at T = 10 and has a horizontal
asymptote. This is so because
32 570
lim T (t) = ≈ 26.
t→∞ 1257
In other words, the temperature of the water does not rise indefinitely; it levels off at 26 ◦ C.
FIGURE 3.54 Temperature of water leaving heat tank
T
30
25
20
15
10
5
0 5000 10 000 15 000 20 000 t
EXERCISES 3.11
In Exercises 1–30 y is defined as a function of x . Find dy/dx in as 3. y = log10 (2x + 1) 4. y = e1−2x
simplified a form as possible.
5. y = xe2x 6. y = x ln x
2x 2 2 ln x
1. y = 3 2. y = ln (3x + 1) 7. y = e 8. y = log10 (3 − 4x)
9. y = ln (sin x) 10. y = ln (3 cos x) ∗ 32. If heat is generated at a constant rate in the copper wire of Exercise
11. y = x ln (x + 1) 2
12. y = x + x e 3 4x 31 (perhaps because of electric current), the differential equation is
replaced by ! "
1−x
e d dT
13. y = 14. y = sin (e2x ) r = k,
1−x dr dr
−2 x
15. y = ln (ln x) 16. y = e sin 3x
2 4x
where k is a constant.
∗ 17. y = ln (x e )
(a) Verify that the function
ex − e−x
∗ 18. y = x
e + e−x T (r) = kr + c ln r + d
∗ 19. y = e−x ln x
satisfies the differential equation for any constants c and d .
∗ 20. y = x [sin (ln x) − cos (ln x)]
(b) If temperature on inner and outer edges of the insulation
∗ 21. y = esin u , u = e1/x
r = a and r = b are constants Ta and Tb , find c and d
∗ 22. y = ln (cos v), v = sin2 x in terms of a , b , Ta , Tb , and k .
∗ 23. ln (x + y) = x 2 y ∗ 33. (a) Two substances A and B react to form a third substance C
y 2
∗ 24. xe + x ln y + y sin x = 0 in such a way that 1 g of A reacts with 1 g of B to produce
2 g of C . If 10 g of A and 15 g of B are brought together
∗ 25. y = ln (sec x + tan x) at time t = 0, the number of grams of C in the mixture as
+ 3 + 4
∗ 26. y = x x 2 + 1 − ln x + x 2 + 1 a function of time t is
3 + 4
60(1 − e−10t )
∗ 27. y = ln x + 4 + 8x + x 2 x(t) = .
3 − 2e−10t
1
∗ 28. y = x − ln (1 + 5e4x ) Verify that x(t) satisfies the (differential) equation
4
xy 2
∗ 29. e = (x + y) dx
1/x 1/y 1 1 = (20 − x)(30 − x).
∗ 30. e +e = + dt
y x
∗ 31. The figure below shows a long cylindrical cable. Copper wire (b) Plot x(t) . What is the limit of x(t) for large t ? Is this
runs down the centre of the cable and insulation covers the wire. If r reasonable?
measures radial distance from the centre of the cable, then steady-state ∗ 34. (a) Two substances A and B react to form a third substance C
temperature T in the insulation is a function of r that must satisfy the in such a way that 1 g of A reacts with 2 g of B to produce
differential equation ! " 3 g of C . If 10 g of A and 30 g of B are brought together
d dT at time t = 0, the number of grams of C in the mixture as
r = 0.
dr dr a function of time t is
(a) Verify that the function 90(1 − e−15t )

x(t) = .
3 − 2e−15t
T (r) = c ln r + d
Verify that x(t) satisfies the (differential) equation
satisfies the differential equation for any constants c and d .
(b) If temperature on inner and outer edges of the insulation dx
r = a and r = b are constants Ta and Tb , find c and d = (30 − x)(45 − x).
dt
in terms of a , b , Ta , and Tb .
(b) Plot x(t) . What is the limit of x(t) for large t ? Is this
reasonable?
∗ 35. (a) What is the differential equation satisfied by temperature
of the water in the tank of Figure 3.53 if the rate at which
water enters and leaves the tank is 100/(t + 1) kg/s?
Insulation (b) Assuming that temperature of the water in the tank is 10 ◦ C
at time t = 0, verify that
a
4190t + 4189 t +1
b
T (t) = + .
419(t + 1) 419
Plot a graph of T (t) .

∗ 36. (a) What is the differential equation satisfied by temperature of of solution is removed from the tank each second. It is assumed that
the water in the tank of Figure 3.53 if the amount of energy the mixer is so efficient that the concentration of salt is the same at all
supplied by the heater for the first 10 min is q = 20t J/s, points in the tank, and this is the concentration at which solution leaves
0 ≤ t ≤ 600? the tank. If S(t) represents the number of kilograms of salt in the tank
(b) Assuming that temperature of the water in the tank is 10◦ C at any given time, then dS/dt is the rate of change of the amount of
at time t = 0, verify that salt in the tank. It must be equal to the rate at which salt enters less the
rate at which salt leaves:
200t 1 962 290 2 000 000
T (t) = − + e−3t/10 000 , 0 ≤ t ≤ 600.
1257 3771 3771 dS 1 ? 1 ?
rate at which rate at which
= − .
Plot a graph of T (t) . dt salt enters tank salt leaves tank
∗ 37. (a) What is the differential equation satisfied by temperature of
the water in the tank of Figure 3.53 if the temperature of the The rate at which salt enters the tank is (1/20)(4) = 1/5 kg/s. The
incoming water is a function of time, T0 = 10e−t , t ≥ 0? rate at which salt leaves is (S/10 000)(4) = S/2500 kg/s. Thus,
(b) Assuming that temperature of the water in the tank is 10◦ C
at time t = 0, verify that
dS 1 S
20 000 74 240 000 30 = − , S(0) = 100.
T (t) = − e −3t/10 000
− −t
e . dt 5 2500
1257 12 566 229 9997
Plot a graph of T (t) . Verify that the solution of this differential equation and initial condition
is
∗ 38. A spring with constant 5 N/m is attached to a fixed wall on one end
and a 1 kg mass m on the other (figure below). We choose a coordinate S(t) = 100(5 − 4e−t/2500 ) kg.
system with x positive to the right and with x = 0 at the centre of mass
of m when the spring is in the unstretched position. At time t = 0, m
is pulled 1 m to the left of x = 0 and given a speed of 3 m/s to the right. Plot its graph. Is it what you would expect?
During the subsequent motion, a frictional force equal in newtons to
∗ 40. (a) Show that if solution in Exercise 39 is removed at 2 L/s
twice the velocity of the mass acts on m . It can be shown that if x(t)
is the position of m , then x(t) must satisfy the differential equation rather than 4 L/s, the differential equation for S(t) is
d 2x dx
+2 + 5x = 0 dS 1 2S
dt 2 dt = − , S(0) = 100.
dt 5 10 000 + 2t
&
and the initial conditions x(0) = −1 and x (0) = 3. Verify that
x(t) = e−t (sin 2t−cos 2t) satisfies the equation and initial conditions.
(b) Verify that the solution is
2
Fixed t 2 × 106
wall S(t) = 500 + − .
10 5000 + t
m
Plot a graph of S(t) . Does it have an asymptote?
x
x=0 ∗ 41. (a) Show that if solution in Exercise 39 is removed at 8 L/s
∗ 39. The figure below shows a tank containing 10 000 L of water in rather than 4 L/s, the differential equation for S(t) is
which is dissolved 100 kg of salt.
4 L/s Mixer 4 L/s dS 1 8S

= − , S(0) = 100.
dt 5 10 000 − 4t
(b) Verify that the solution is
3t t2
S(t) = 100 + − .
25 15 625
For how long is the solution valid? Plot a graph of S(t) .

Starting at time t = 0, a solution containing 1 kg of salt for each
20 L of solution is added to the tank at 4 L/s. Simultaneously, 4 L (c) What is the maximum amount of salt in the tank?
∗ 42. (a) A beer vat contains 2000 L of beer, 4% of which is alcohol. ∗ 45. Verify that the function
Beer with 8% alcohol is added to the vat at 2 L/s, and well-
stirred mixture is removed at the same rate. Show that if # ! " $
x x−1
A(t) represents the number of litres of alcohol in the vat, y = f (x) = Ax + B ln +1
then 2 x+1
dA 4 A satisfies the differential equation

= − , A(0) = 80.
dt 25 1000
d 2y dy
(b) Verify that the solution of this differential equation is (1 − x 2 ) 2
− 2x + 2y = 0
dx dx
A(t) = 160 − 80e−t/1000 .
Plot a graph of A(t) . for any constants A and B .
(c) When is the beer in the vat 5% alcohol? ∗ 46. The equation x = ey − e−y defines y implicitly as a function of
∗ 43. Two long, parallel rectangular loops lying in the same plane have x . Find dy/dx in terms of x in two ways:
lengths l and L and widths w and W , respectively (figure below). If
the loops do not overlap, and the distance between the near sides is s , (a) Solve first for an explicit definition for y = f (x) .
the mutual inductance between the loops is
(b) Differentiate implicitly with respect to x .
 
µ0 l  s+W  ∗ 47. (a) Show that when a liquid enters the tank in Figure 3.53 at
M = ln 
 ! "
, ṁ kilograms per second, the mass of liquid in the tank is
2π W
s 1+ M kilograms, the temperature of liquid entering the tank
s+w is T0 ◦ C, the specific heat of the liquid is cp , and the heater
where µ0 > 0 is a constant. Show that the derivative of M as a adds energy at the rate of q joules per second, the differen-
function of s is negative for s > 0. tial equation governing temperature T (t) of liquid in the
tank is
dT
ṁcp T0 + q = ṁcp T + Mcp .
dt
l s L
(b) Verify that the solution of this differential equation is

w ! "
W q ) *
T (t) = T0 e−ṁt/M + T0 + 1 − e−ṁt/M
∗ 44. Two parallel wires carrying current i (figure below) create a mag- cp ṁ
netic field. The flux , of this field through the loop of dimensions h
and w is # $ when the temperature of the liquid in the tank is T0 at time
µ0 hi R(r + w)
,= ln , t = 0, and ṁ , T0 , and q are all constants.
2π r(R + w)
where µ0 > 0 is a constant. Draw a graph of , , first as a function of 48. Find d 2 y/dx 2 if y is defined implicitly as a function of x by
h and then as a function of w .
3y 4
i i ln (x 2 + y 2 ) = 2 Tan −1 .
x
r h ∗∗ 49. (a) The function f (x) = x n is not defined for x < 0 if n is

irrational, that is, is not rational. Verify power rule 3.7 for
x < 0 if n is rational and x n is defined.
w
R (b) Discuss the derivative of f (x) = x n at x = 0.
3.12 Logarithmic Differentiation

Compare the following three functions: x n , a x , and x x . The first, in which the exponent is
constant and the base is variable, is called a power function. The second, in which the base is
constant and the exponent is variable, is called an exponential function. We have differentiation
rules for power functions and exponential functions, namely
d n d x
x = nx n−1 and a = a x ln a,
dx dx
respectively. The third function x x , which we consider only for x > 0, is neither a power
nor an exponential function; therefore, we cannot use either of the formulas above to find its
derivative. Instead, we set y = f (x) = x x , and take natural logarithms,
ln y = x ln x.
Implicit differentiation with respect to x now gives
1 dy x
= ln x + = ln x + 1.
y dx x
Consequently,
dy d x
= y(ln x + 1) or x = x x (ln x + 1).
dx dx
This process of taking logarithms and then differentiating is called logarithmic differentiation.
EXAMPLE 3.36
Find the derivative of the function f (x) = x sin x when x > 0.
SOLUTION If we set y = x sin x , then
ln y = sin x ln x.
Differentiation with respect to x gives
1 dy 1
= cos x ln x + sin x,
y dx x
and this can be solved for dy/dx ,

! "
dy sin x 1
=x cos x ln x + sin x .
dx x
Logarithmic differentiation can also be used to differentiate complicated products or quotients,

as in the following example.
EXAMPLE 3.37
Find the derivative of the function
x 3 (x 2 + 1)2/3
y = f (x) =
sin3 x
on the interval 0 < x < π . Extend the result to other values of x .

SOLUTION When 0 < x < π , we take natural logarithms of both sides of the definition
for y :
2
ln y = 3 ln x + ln (x 2 + 1) − 3 ln (sin x).
3
Differentiation with respect to x gives
1 dy 3 2 2x 3
= + − cos x,
y dx x 3 x2 + 1 sin x
and therefore
# $ # $
dy 3 4x x 3 (x 2 + 1)2/3 3 4x
=y + − 3 cot x = + − 3 cot x .
dx x 3(x 2 + 1) sin3 x x 3(x 2 + 1)
When x is not in the interval 0 < x < π , this derivation may not be valid. For instance,
when x < 0, it is not acceptable to write ln x , and when x is in the interval π < x < 2π ,
the term ln (sin x) is not defined. These difficulties are easily overcome by first taking absolute
values,
|x|3 (x 2 + 1)2/3
|y| = .
| sin x|3
Logarithms now give
2
ln |y| = 3 ln |x| + ln (x 2 + 1) − 3 ln | sin x|.
3
To differentiate this equation, we use equation 3.13, which states that when f (x) is differen-
tiable, and f (x) ,= 0,
d |f (x)| &
|f (x)| = f (x).
dx f (x)
When this is combined with formula 3.39, we obtain
d 1 d 1 |f (x)| & f & (x)

ln |f (x)| = |f (x)| = f (x) = . (3.46)
dx |f (x)| dx |f (x)| f (x) f (x)
Application of this result to the equation in ln |y| gives
1 dy 3 2 2x cos x
= + −3 .
y dx x 3 x2 +1 sin x
This equation is identical to that obtained for dy/dx when 0 < x < π , but its derivation here
shows that it is valid even when x is not in the interval 0 < x < π.
EXERCISES 3.12
In Exercises 1–24 use logarithmic differentiation to find f & (x) . ex

∗ 17. f (x) = x 2 ln x ∗ 18. f (x) =
ln (x − 1)
∗ 1. f (x) = x −x , x>0
∗ 19. f (x) = (x 3 + 3)3 (x 2 − 2x)
4 cos x
∗ 2. f (x) = x , x>0 √
x(1 − x 2 )
∗ 3. f (x) = x 4x , x>0 ∗ 20. f (x) = √
1 + x2
∗ 4. f (x) = (sin x)x , 0 < x < π
! "x x2 − 1
1 ∗ 21. f (x) = √
∗ 5. f (x) = 1 + , x>0 x 1 − 4 tan2 x
x +
! "x 2 ∗ 22. f (x) = x 3 (x 2 − 4x) 1 + x 3
1
∗ 6. f (x) = 1 + , x>0 sin3 3x
x ∗ 23. f (x) =
! "1/x tan5 2x
1 sin 2x sec 5x
∗ 7. f (x) = , x>0 ∗ 24. f (x) =
x (1 − 2 cot x)3
! "3/x
2 ∗ 25. If u(x) is positive for all x , find a formula for the derivative of uu
∗ 8. f (x) = , x>0
x with respect to x .
∗ 9. f (x) = (sin x)sin x , 0<x<π ∗ 26. If a company sells a certain commodity at price r , the market
demands
∗ 10. f (x) = (ln x)ln x , x>1 x = r a e−b(r+c)
∗ 11. f (x) = (x 2 + 3x 4 )3 (x 2 + 5)4 items per week, where r > a/b , and a , b , and c are positive constants.
√
x(1 + 2x 2 ) (a) Show that the demand increases as the price decreases.
∗ 12. f (x) = √
1 + x2 (b) Calculate the elasticity of demand defined by
√
∗ 13. f (x) = x 3 1 − sin x ∗ 14. f (x) = (x 2 + 3x)3 (x 2 + 5)4
Er x dr
= .
∗ 15. f (x) = x 2 e4x ∗ 16. f (x) = x 3/2 e−2x Ex r dx
3.13 Derivatives of the Hyperbolic Functions

Since the hyperbolic sine and cosine functions are defined in terms of the exponential function,
for which we know the derivative, and the remaining hyperbolic functions are defined in terms
of the hyperbolic sine and cosine, it follows that calculation of the derivatives of the hyperbolic
functions should be straightforward. Indeed, if u(x) is a differentiable function of x , then
d du
sinh u = cosh u , (3.47a)
dx dx
d du
cosh u = sinh u , (3.47b)
dx dx
d du
tanh u = sech 2 u , (3.47c)
dx dx
d du
coth u = −csch 2 u , (3.47d)
dx dx
d du
sech u = −sech u tanh u , (3.47e)
dx dx
d du
csch u = −csch u coth u . (3.47f)
dx dx
3.13 Derivatives of the Hyperbolic Functions 227
EXAMPLE 3.38
Find dy/dx if y is defined as a function of x by:
(a) y = sech (3x 2 ) (b) y = tanh ()1 − 4x) *
(c) y = cos 2x sinh 2x (d) y = cosh Tan −1 x 2
SOLUTION
dy d
(a) = −sech (3x 2 ) tanh (3x 2 ) (3x 2 ) = −6x sech (3x 2 ) tanh (3x 2 )
dx dx
dy d
(b) = sech 2 (1 − 4x) (1 − 4x) = −4 sech 2 (1 − 4x)
dx dx
dy
(c) = cos 2x(2 cosh 2x) + (−2 sin 2x) sinh 2x
dx
= 2(cos 2x cosh 2x − sin 2x sinh 2x)
dy ) *d ) * 2x
(d) = sinh Tan−1 x 2 Tan−1 x 2 = sinh Tan−1 x 2
dx dx 1 + x4
EXAMPLE 3.39
When a uniform cable is hung between two fixed supports (Figure 3.55), the shape of the curve
FIGURE 3.55 Shape of y = f (x) must satisfy the differential equation
hanging cable 0
2
! "2
y d y ρg dy
B = 1+ ,
dx 2 H dx
A where ρ is the mass per unit length of the cable, g > 0 is the acceleration due to gravity, and
H > 0 is the tension in the cable at its lowest point. Verify that a solution of the equation is
H 3 ρgx 4
x y = f (x) = cosh + C,
ρg H
where C is a constant.
SOLUTION In Exercises 42 of Section 8.4 we derive this solution. For now we simply wish
to verify that the hyperbolic cosine is indeed a solution. The first derivative of the function is
dy 3 ρgx 4
= sinh ,
dx H
and therefore its second derivative is
d 2y ρg 3 ρgx 4
= cosh .
dx 2 H H
On the other hand,
0 ! "2 9 9
ρg dy ρg 3 ρgx 4 ρg 3 ρgx 4 ρg 3 ρgx 4
1+ = 1 + sinh 2 = cosh 2 = cosh .
H dx H H H H H H
= >
Thus, the function y = H /(ρg) cosh (ρgx/H )+C does indeed satisfy the given differential
equation.
Example 3.39 shows that the many telephone and hydro wires crisscrossing the country hang in
the form of hyperbolic cosines. Engineers often call this curve a catenary.
EXAMPLE 3.40
In studying wave guides, the electrical engineer often encounters the differential equation
d 2y
− ky = 0,
dx 2
√ √
where k > 0 is a constant. Verify that y = f (x) = A cosh kx + B sinh kx is a solution
for any constants A and B .
SOLUTION The first derivative of y = f (x) is
dy √ √ √ √
= k A sinh k x + k B cosh k x,
dx
and therefore
d 2y √ √ √ √
2
= kA cosh k x + kB sinh k x = k(A cosh k x + B sinh k x).
dx
Thus,
d 2y
= ky.
dx 2
EXERCISES 3.13
In Exercises 1–10 y is defined as a function of x . Find dy/dx in as where k > 0 is a constant.
simplified a form as possible. (a) Verify that a solution is
1. y = csch (2x + 3) y = f (x) = A cos kx + B sin kx + C cosh kx + D sinh kx
2. y = x sinh (x/2) for any constants A , B , C , and D .
√ (b) If the left end ( x = 0) is fastened horizontally and the
3. y = 1 − sech x right end ( x = L ) is pinned, then f (x) must satisfy the
conditions
4. y = tanh (ln x)
f (0) = f & (0) = f (L) = f && (L) = 0.
5. cosh (x + y) = 2x
+ Show that these restrictions imply that C = −A ,
6. y + coth x = 1+y D = −B , and A and B must satisfy the equations
7. y = u cosh u, u = ex + e−x A(cos kL − cosh kL) + B(sin kL − sinh kL) = 0,
8. y = tan (cosh t), t = cos (tanh x) A(cos kL + cosh kL) + B(sin kL + sinh kL) = 0.
9. y = Tan −1 (sinh x) (c) Eliminate A and B between these equations to show that
k must satisfy the condition
√
10. y = ln tanh 2x
tan kL = tanh kL.
11. Verify the differentiation formulas 3.47. y
L
∗ 12. To analyze vertical vibrations of the beam in the figure below, we
must solve the differential equation
d 4y x
− k 4 y = 0,
dx 4
∗ 13. Each hyperbolic function has associated with it an inverse hyper-

bolic function. (See Exercise 13 in Section 1.10.) Obtain the following d 1
Tanh−1 x = , |x| < 1;
derivatives of these functions: dx 1 − x2
d 1
Coth−1 x = , |x| > 1;
dx 1 − x2
d 1 d −1
Sinh−1 x = √ ; Sech−1 x = √ ;
dx 2
x +1 dx x 1 − x2
d 1 d −1
Cosh−1 x = √ ; Csch−1 x = √ .
dx x2 − 1 dx |x| 1 + x 2
3.14 Rolle’s Theorem and the Mean Value Theorems

Certain results in calculus are immediately seen to be important. For example, the power, prod-
uct, and quotient rules that eliminate the necessity of using equation 3.3 to calculate derivatives
are clearly indispensable. Even the algebraic and geometric interpretations of the derivative
itself are recognized as useful. Through various examples and exercises of this chapter, we have
hinted at the variety and quantity of applications of the derivative. These will be dealt with
at length in Chapter 4. Other results in calculus, especially those of a theoretical nature, are
regarded as less important, or even unimportant, often because it is not obvious how they will
be used. In this section we consider three very important theorems, without which we would
encounter serious difficulty in treating many of the topics in the remainder of this book. The
first theorem is needed to prove the second, and the second leads immediately to the third.
THEOREM 3.17 (Rolle’s Theorem)

Suppose a function f (x) satisfies the following three properties:
1. f (x) is continuous for a ≤ x ≤ b ;
2. f & (x) exists for a < x < b ;
3. f (a) = f (b) .
Then, there exists at least one point c in the open interval a < c < b at which f & (c) = 0.
FIGURE 3.56 Rolle’s

For the function in Figure 3.56 there are two possible choices for c . Geometrically, Rolle’s
theorem theorem seems quite evident. Begin with two points, say P and Q in Figure 3.57, which have
the same y -coordinate. Now try to join these points by a curve that never has a horizontal
y
f (a) = f (b)
tangent line while satisfying the following two conditions:
(a) Do not lift the pencil from the page — continuity of f (x) ;
f (a) f (b) (b) The curve must have a tangent line at all points, and this tangent line must not be
a b x vertical — f & (x) exists at all points.
It is impossible; therefore, a point c where f & (c) = 0 must exist.
FIGURE 3.57 Two level
To verify this theorem directly requires Theorem 4.2 in Section 4.7. Since the latter result
points for Rolle’s theorem is quoted without proof, it seems as reasonable to accept Rolle’s theorem on obvious geometric
grounds as to base a proof on a theorem which itself is stated without proof. But for those who
y
would like to see a proof based on Theorem 4.2 in Section 4.7, see Exercise 10 in that section.
Like the intermediate value theorem in Section 1.11, Rolle’s theorem is an existence theo-
P Q rem; it stipulates the existence of c , but does not provide a way to find it.
Rolle’s theorem can be used to prove the following result.
x
THEOREM 3.18 (Cauchy’s Generalized Mean Value Theorem)

Suppose functions f (x) and g(x) satisfy the following three properties:
1. f (x) and g(x) are continuous for a ≤ x ≤ b ;
2. f & (x) and g & (x) exist for a < x < b ;
3. g & (x) =
, 0 for a < x < b .
Then, there exists at least one point c in the open interval a < c < b for which
f (b) − f (a) f & (c)

= & . (3.48)
g(b) − g(a) g (c)
PROOF First note that g(b)−g(a) cannot equal zero. If it did, then g(a) would equal g(b) ,
and Rolle’s theorem applied to g(x) on the interval a ≤ x ≤ b would imply the existence
of a point c at which g & (c) = 0, contrary to the given assumption. To prove the theorem, we
construct a function h(x) to satisfy the conditions of Rolle’s theorem. Specifically, we consider
f (b) − f (a)
h(x) = f (x) − f (a) − [g(x) − g(a)].
g(b) − g(a)
Since f (x) and g(x) are continuous for a ≤ x ≤ b , so too is h(x) . In addition,
f (b) − f (a) &

h& (x) = f & (x) − g (x);
g(b) − g(a)
therefore, h& (x) exists for a < x < b . Finally, since h(a) = h(b) = 0, we may conclude
from Rolle’s theorem that there exists a number c such that a < c < b , and
f (b) − f (a) &

0 = h& (c) = f & (c) − g (c);
g(b) − g(a)
that is,
f & (c) f (b) − f (a)

= .
g & (c) g(b) − g(a)
The next theorem states an important special case of this result that occurs when g(x) = x .
THEOREM 3.19 (Mean Value Theorem)

Suppose a function f (x) satisfies the following two properties:
1. f (x) is continuous for a ≤ x ≤ b ;
2. f & (x) exists for a < x < b ;
Then, there exists at least one point c in the open interval a < c < b for which
f (b) − f (a)
f & (c) = . (3.49)
b−a
FIGURE 3.58 Mean value From a geometric point of view equation 3.49 seems as obvious as Rolle’s theorem. Figure
theorem 3.58 illustrates
) that
* the) quotient*[f (b) − f (a)]/(b − a) is the slope of the line l joining the
y points a, f (a) and b, f (b) on the graph y = f (x) . The mean value theorem states that
there is at least one point c between a and b at which the tangent line is parallel to l . In Figure
(b, f (b)) 3.58 there are clearly two such points. Algebraically, equation 3.49 states that at some point
l between a and b , the instantaneous rate of change of the function f (x) is equal to its average
rate of change over the interval a ≤ x ≤ b . Similar interpretations of Theorem 3.18 are given
in Section 9.1.
(a, f (a))
a c c b x
EXAMPLE 3.41
Find all values of c satisfying the mean value theorem for the function f (x) = x 3 − 4x on the
interval −1 ≤ x ≤ 3.
SOLUTION Since f (x) is differentiable, and therefore continuous, at each point in
−1 ≤ x ≤ 3, we can indeed apply the mean value theorem and claim the existence of at
least one number c in −1 < x < 3 such that
f (3) − f (−1)
f & (c) = ;
3 − (−1)
that is,
15 − 3
3c 2 − 4 =
= 3.
4
√ √ √
Consequently, c = ± 7/3. Since − 7/3 < −1, c = 7/3 is the only value of c in the
interval −1 < x < 3 (see Figure 3.59).
FIGURE 3.59 Mean value theorem for x 3 − 4x on −1 ≤ x ≤ 3
(3, 15)
y = x 3 − 4x
(−1, 3) 7
3
−2 1 2 x
EXAMPLE 3.42
A traffic plane (Figure 3.60) measures the time that it takes a car to travel between points A and
B as 15 s, and radios this information to a patrol car. What is the maximum speed at which the
police officer can claim that the car was travelling between A and B ?
SOLUTION The average speed of the car between A and B is 500/15 = 100/3 m/s. Ac-
cording to the mean value theorem, the instantaneous speed of the car must also have been
100/3 m/s at least once. This is the maximum speed attributable to the car between A and B . It
may have been travelling much faster at some points, but from the information given, no speed
greater than 100/3 m/s can be claimed by the officer.
FIGURE 3.60 Traffic plane measuring speed of car
A B
500 m
EXERCISES 3.14
In Exercises 1–14 decide whether the mean value theorem can be ap- In Exercises 15–18 decide whether Cauchy’s generalized mean value
plied to the function on the interval. If it cannot, explain why not. If it theorem can be applied to the functions on the interval. If it cannot,
can, find all values of c in the interval that satisfy equation 3.49. explain why not. If it can, find all values of c in the interval that satisfy
equation 3.48.
∗ 1. f (x) = x 2 + 2x, −3 ≤ x ≤ 2 ∗ 15. f (x) = x 2 , g(x) = x, 1≤x≤2
∗ 2. f (x) = 4 + 3x − 2x 2 , 1≤x≤3 ∗ 16. f (x) = x + 1, g(x) = |x|3/2 , −1 ≤ x ≤ 1

2 3
∗ 17. f (x) = x + 3x − 1, g(x) = x + 5x + 4, 0≤x≤2
∗ 3. f (x) = x + 5, 2≤x≤3
∗ 18. f (x) = x/(x + 1), g(x) = x/(x − 1), −3 ≤ x ≤ −2
∗ 4. f (x) = |x|, −1 ≤ x ≤ 1
∗ 19. Show that if |f & (x)| ≤ M on a ≤ x ≤ b , then
∗ 5. f (x) = |x|, 0≤x≤1
|f (b) − f (a)| ≤ M(b − a).
∗ 6. f (x) = x 3 + 2x 2 − x − 2, −3 ≤ x ≤ 2
∗ 20. Show that the value c that satisfies the mean value theorem for any
3 2
∗ 7. f (x) = x + 2x − x − 2, −1 ≤ x ≤ 2 quadratic function f (x) = dx 2 + ex + g on any interval a ≤ x ≤ b
whatsoever is c = (a + b)/2.
∗ 8. f (x) = (x + 2)/(x − 1), 2≤x≤4 ∗ 21. Use the mean value theorem to show that
∗ 9. f (x) = (x + 1)/(x + 2), −3 ≤ x ≤ 2 | sin a − sin b| ≤ |a − b|
∗ 10. f (x) = x 2 /(x + 3), −2 ≤ x ≤ 3 for all real a and b . Is the same inequality valid for the cosine function?
∗ 22. Let f (x) and g(x) be two functions that are differentiable at each
∗ 11. f (x) = sin x, 0 ≤ x ≤ 2π
point of the interval a ≤ x ≤ b . Prove that if f (a) = g(a) and
f (b) = g(b) , then there exists c in the open interval a < x < b for
∗ 12. f (x) = ln (2x + 1), 0≤x≤2
which f & (c) = g & (c) .
∗ 13. f (x) = e−x , −1 ≤ x ≤ 1 ∗ 23. Verify that for a cubic polynomial f (x) = dx 3 + ex 2 + gx + h
defined on any interval a ≤ x ≤ b , the values of c that satisfy equation
∗ 14. f (x) = sec x, 0≤x≤π 3.49 are equidistant from x = −e/(3d) .
SUMMARY
In this chapter we defined the derivative of a function y = f (x) as
dy f (x + h) − f (x)
= f & (x) = lim .
dx h→0 h
Summary 233
Algebraically, it is the instantaneous rate of change of y with respect to x ; geometrically, it

is the slope of the tangent line to the graph of f (x) . To eliminate the necessity of using this
definition over and over again, we derived the sum, product, quotient, and power rules:
d du dv
(u + v) = + ,
dx dx dx
d dv du
(uv) = u +v ,
dx dx dx
du dv
d 3u4 v −u
= dx 2 dx ,
dx v v
d n du
(u ) = nun−1 .
dx dx
These four simple rules are fundamental to all calculus.
When a function y = f (x) is defined implicitly by some equation F (x, y) = 0, we use
implicit differentiation to find its derivative. We differentiate each term in the equation with
respect to x , and then solve the resulting equation for dy/dx . We pointed out that care must be
taken in differentiating equations. An equation can be differentiated with respect to a variable
only if it is valid for a continuous range of values of that variable. ) *
The chain rule defines the derivative of a composite function y = f g(x) as the product
of the derivatives of y = f (u) and u = g(x) :
dy dy du
= .
dx du dx
These rules and techniques form the basis for the rest of differential calculus. When they
are combined with the derivatives of the trigonometric, inverse trigonometric, exponential,
logarithm, and hyperbolic functions, we are well prepared to handle those applications of calculus
that involve differentiation. Derivative formulas for these trancendental functions are listed
below:
d du d du
sin u = cos u , cos u = − sin u ,
dx dx dx dx
d du d du
tan u = sec2 u , cot u = − csc2 u ,
dx dx dx dx
d du d du
sec u = sec u tan u , csc u = − csc u cot u ,
dx dx dx dx
d 1 du d 1 du
loga u = loga e , ln u = ,
dx u dx dx u dx
d u du d u du
a = au ln a , e = eu .
dx dx dx dx
d 1 du d −1 du
Sin −1 u = √ , Cos −1 u = √ ,
dx 2
1 − u dx dx 1 − u2 dx
d 1 du d −1 du
Tan −1 u = , Cot −1 u = ,
dx 1 + u2 dx dx 1 + u2 dx
d 1 du d −1 du
Sec −1 u = √ , Csc −1 u = √ ,
dx u u2 − 1 dx dx u u2 − 1 dx
d du d du
sinh u = cosh u , cosh u = sinh u ,
dx dx dx dx
d du d du
tanh u = sech 2 u , coth u = −csch 2 u ,
dx dx dx dx
d du d du
sech u = −sech u tanh u , csch u = −csch u coth u .
dx dx dx dx
Velocity and acceleration have finally been given formal definitions. Velocity is the deriva-
tive of displacement, and acceleration is the derivative of velocity, or the second derivative of
displacement:
dx dv d 2x
v(t) = , a(t) = = .
dt dt dt 2
We completed the chapter by using Rolle’s theorem to prove two mean value theorems.
When f (x) and g(x) are continuous for a ≤ x ≤ b and differentiable for a < x < b ,
Cauchy’s generalized mean value theorem guarantees the existence of at least one point c
between a and b such that
f (b) − f (a) f & (c)
= & ,
g(b) − g(a) g (c)
provided also that g & (x) ,= 0 for a < x < b . When g(x) = x , we obtain as a special case,
the mean value theorem
f (b) − f (a)
f & (c) = .
b−a
KEY TERMS
Displacement function Average velocity

Increment Instantaneous velocity
Average rate of change Instantaneous rate of change
Derivative Differentiation
Tangent line Derivative function
Power rule Normal line
Right-hand derivative Left-hand derivative
Differentiable Angle between intersecting curves
Orthogonal curves Product rule
Quotient rule Second derivative
Third derivative Velocity
Speed Instantaneous acceleratio
Chain rule Extended power rule
Explicit definition of a function Implicit definition of a function
Implicit differentiation Simple harmonic motion
Differential equation Initial-value problem
Energy balance equation Power function
Exponential function Logarithmic differentiation
Rolle’s theorem Chaucy’s generalized mean value theorem
Mean value theorem
REVIEW
EXERCISES
= >
In Exercises 1–58 assume that y is defined as a function of x and find 39. y = e2x cosh 2x 40. ln Tan −1 (x + y) = 1/10
dy/dx in as simplified a form as possible. 0
+ 4+y
1 1 ∗ 41. x = 1 + x cot y 2 ∗ 42. x =
1. y = x 3 + 2. y = 3x 2 + 2x + 4−y
x2 x
0 √
1 1 1/ 3 2 5/ 3 4 + x2 x2 1 − x
3. y = 2x − +√ 4. y = x − x ∗ 43. y = ∗ 44. y =
3x 2 x 3 4 − x2 x+5
9
5. y = x(x 2 + 5)4 6. y = (x 2 + 2)2 (x 3 − 3)3 :
√ x y
∗ 45. y = 7− 7− x ∗ 46. =
3x 2
3x − 2 x+y x−y
7. y = 8. y = 9
x3 −5 x+5 4+t
∗ 47. y = , t = tan x
x 2 + 2x + 2 4x 4−t
9. y = 2 10. y =
x + 2x − 1 x 2 + 5x − 2 y 2 − 2y
∗ 48. x =
x y y 3 + 4y + 6
11. xy + 3y 3 = x + 1 12. + =x
y x x 3 − 6x 2 + 12x − 8
√ ∗ 49. y =
x 2 − 4x + 4
13. x 2 y 2 − 3y sin x = 14 14. x 2 y + y 1 + x = 3
x−2
15. y = tan3 (3x + 2) 16. y = sec2 (1 − 4x)
∗ 50. y = √ √
x− 2
sin 2x ∗ 51. y = x 2x ∗ 52. y = (cos x)x , 0 < x < π/2
17. y = 18. y = sec (tan 2x)
cos 3x
ex
∗ 53. y = ∗ 54. y = log10 (log10 x)
19. y = x 2 cos x 2 20. y = sin2 x cos2 x ex +1
21. y = u2 − 2u, u = (1 + 2x)5/3 ∗ 55. y = ex ln x ∗ 56. x = ey + e−y
∗ 57. xyexy = 1 ∗ 58. x 2 y + ln (x + y) = x + 2

22. y = t + cos 2t, t = x − cos 2x
+ +
23. y = 1 − t 3, t= 1 + x2 In Exercises 59–62 find equations for the tangent and normal lines to
the curve at the point indicated.
+
24. y = v cos2 v, v= 1 − x2 1
59. y = x 3 + 3x − 2 at (1, 2) 60. y = at (0, 1/5)
: x+5
√
25. y = 1+ 1+x 26. x = e2y x 2 + 3x
61. y = cos 2x at (π/2, −1) 62. y = at (1, −4/3)
2x − 5
2xy
27. = x2 + 2 28. y = (x 2 + 1) ln (x 2 − 1)
3x + 4
In Exercises 63–65 find d 2 y/dx 2 assuming that y is defined implicitly
29. x sin y + 2xy = 4 30. 5 cos (x − y) = 1 as a function of x .
31. y = Sin −1 (2 − 3x) 32. y = 3 sinh (x 2 ) ∗ 63. x 2 − y 2 + 2(x − y) = y 3

! " ∗ 64. (x − y)2 = 3xy
Cos −1 x −1 1
33. y = 34. y = Tan +x
Sin −1 x x ∗ 65. sin (x + y) = x
35. y = ecosh x 36. sinh y = sin x

∗ 66. Draw a graph of f (x) = sin x 2 . Is it periodic?
! "
1 ∗ 67. Find all points on the curve y = x 3 + x 2 at which the tangent line
37. Sec −1 (x + y) = xy 38. y = x Csc −1
x2 passes through the origin.
∗ 68. Find that point on the curve x = y 2 − 4 at which the normal line ∗ 72. The curve defined by the equation (x 2 + y 2 )2 = x 2 − y 2 and
passes through the point (−6, 7) . What application could be made of shown below is called a lemniscate. Find the four points at which the
this result? tangent line is horizontal.
∗ 69. (a) How many functions with domain −1 ≤ x ≤ 1 are defined
implicitly by the equation x 2 + y 2 = 1?
(b) How many continuous functions with domain −1 ≤ x ≤ 1 y (x 2 + y2)2 = x 2 − y2
are defined implicitly by the equation?
∗ 70. What is the rate of change of the area A of an equilateral triangle
with respect to its side length L ?
∗ 71. In a heated house, the temperature varies as the thermostat contin- −1 1 x
ually engages and disengages the furnace. Suppose that at the thermo-
stat, the temperature T in degrees Celsius over a four-hour time interval
0 ≤ t ≤ 4 is given by
T = f (t) = 20 + 3 sin (4π t − π/2). ∗ 73. Find all points c in the interval 3 ≤ x ≤ 6 that satisfy equation
(a) Draw a graph of f (t) . 3.49 when f (x) = x 3 + 3x − 2.
(b) How many times is the furnace on during the four-hour
period? ∗ 74. Find all points c in the interval −1 ≤ x ≤ 1 that satisfy equation
(c) What is the maximum time rate of change of temperature? 3.48 when f (x) = 3x 2 − 2x + 4 and g(x) = x 3 + 2x .
CHAPTER
4 Applications of Differentiation
Application Preview Slider-cranks, such as that shown below, transform rotary motion to back-and-forth motion along
a straight line, and vice versa. Rod AB , of length r , is pinned at A , and rod BC , of length L ,
is pinned to rod AB at B . As AB rotates around A , C is confined to move along a horizontal
line segment between points D and E , called the stroke of the mechanism. Rotary motion of B
is transformed to straight-line motion of C along DE . This crank is said to be offset because
the extension of line segment DE does not pass through the centre of the circle A ; it is offset
by an amount e .
B
r L
A C
e
E D
THE PROBLEM Find the offset that maximizes the stroke. (For the solution, see Example
4.28 on page 284.)
Our discussions in Chapter 3 hinted at some of the applications of the derivative; in this
chapter we deal with them in detail. In Section 4.1 we show that derivatives provide one of
the most powerful methods for approximating solutions to equations. Sections 4.2–4.7 are
devoted to the topic of optimization, its theory and applications. Sections 4.7–4.10, with their
wealth of applied problems, show the power of calculus in applied mathematics. In Section 4.7
we illustrate the simplicity that calculus brings to solving applied maxima-minima problems;
in Section 4.8 we develop a deeper understanding for the already familiar notions of velocity
and acceleration; in Section 4.9 we use the interpretation of the derivative as a rate of change to
investigate interdependences of related quantities in a wide variety of applications; and in Section
4.10 we apply derivatives to LCR -circuits. The calculation of many otherwise intractable limits
becomes relatively straightforward with L’Hôpital’s rule in Section 4.11. In Section 4.12 we
discuss differentials, quantities essential to the topic of integration, which begins in Chapter 5.
4.1 Newton’s Iterative Procedure for Solving Equations

In almost every area of applied mathematics, it is necessary to solve equations. When the
problem is to solve one equation in one unknown, the equation can be expressed in the form
f (x) = 0, (4.1)
where f (x) is usually a differentiable function of x . It might be a polynomial, a trigonometric

function, an exponential or a logarithm function, or a complicated combination of these. The
equation may have one solution or many solutions; these solutions are also called roots of the
equation or zeros of the function. Few equations can be solved by formula. Even when f (x) is
a polynomial, the only simple formula is quadratic formula 1.5, which solves the equation for
second-degree polynomials. Functions other than polynomials rarely have formulas for their
zeros.
237
238 Chapter 4 Applications of Differentiation
Since we can seldom find exact solutions of 4.1, we consider approximating the solutions.
Sophisticated calculators have routines for approximating solutions to equations; computer
software packages have one or more commands for doing this. Unfortunately, with calculators
and computers, you really have no idea what is going on inside them. You supply an equation
FIGURE 4.1 Volume of a and out comes a number. In this section we develop one of the most powerful techniques for
half-filled hemispherical tank approximating solutions to equations. The method is calculus-based and it, or a modification
of it, is used by many calculators and computers to solve equations. If your machine uses this
technique you will now understand what it is doing.
5 To introduce the technique, consider the problem of finding the depth of water that half fills
(by volume) the hemispherical tank in Figure 4.1. It can be shown (see Exercise 21 in Section
7.4) that when the water is x metres deep, the volume of water in the tank is given by the formula
π
(15x 2 − x 3 ).
x 3
Since the volume of one-half the tank is one-fourth that of a sphere of radius 5, it follows
that the tank is half full when x satisfies the equation
! "
1 4 3 π
π(5) = (15x 2 − x 3 ),
4 3 3
and this equation simplifies to
f (x) = x 3 − 15x 2 + 125 = 0.

How do we solve this equation when we know intuitively that the solution we want is somewhere
around x = 3? A sketch of f (x) between x = 3 and x = 3.5 is shown in Figure 4.2.
The curvature has been exaggerated in order to more clearly depict the following geometric
construction. If we set x1 = 3, then x1 is an approximation to the solution of the equation
— not a good approximation,
# but $an approximation nonetheless. Suppose we draw the tangent
line to y = f (x) at x1 , f (x1 ) . If x2 is the point of intersection of this tangent line with
the x -axis, it is clear that x2 is a better approximation
# than $ x1 to the solution of the equation.
If we draw the tangent line to y = f (x) at x2 , f (x2 ) , its intersection point x3 with the
x -axis is an even better approximation. Continuation of this process leads to a succession of
numbers x1 , x2 , x3 , . . . , each of which is closer to the solution of the equation f (x) = 0
than the preceding numbers. This procedure for finding better and better approximations to
the solution of an equation is called Newton’s iterative procedure (or the Newton–Raphson
iterative procedure). We say that the numbers x1 , x2 , x3 , . . . converge to the root of the equation.
FIGURE 4.2 Geometric interpretation of Newton’s iterative procedure
y (x1, f (x1)) = (3, 17)
Tangent line at (x1, f (x1))
x4
x3
x1 x2
3 3.5 x
Tangent line at
(x2, f (x2))
(x2, f (x2))
(3.5, −15.9)
What we need now is an algebraic formula by which to calculate # the$ approximations

x2 , x3 , x4 , . . . . The equation of the tangent line to y = f (x) at x1 , f (x1 ) is
y − f (x1 ) = f " (x1 )(x − x1 ).
To find the point of intersection of this tangent line with the x -axis, we set y = 0,
−f (x1 ) = f " (x1 )(x − x1 ),
and solve for x ,

f (x1 )
x = x1 − .
f " (x1 )
# $
But the point of intersection of the tangent line at x1 , f (x1 ) with the x -axis is the second
approximation x2 ; that is,
f (x1 )
x 2 = x1 − " .
f (x1 )
To find x3 we repeat this procedure with x2 replacing x1 ; the result is
f (x2 )
x 3 = x2 − .
f " (x2 )
As we repeat this process over and over again, the following formula for the (n + 1)th approx-
imation xn+1 in terms of the nth approximation xn emerges:
f (xn )
xn+1 = xn − . (4.2)
f " (xn )
This formula defines each approximation in Newton’s iterative procedure in terms of its prede-
cessor.
Let us use this procedure to approximate the solution of f (x) = x 3 − 15x 2 + 125 = 0 in
the example above. The derivative of f (x) is f " (x) = 3x 2 − 30x , and formula 4.2 becomes
xn3 − 15xn2 + 125

xn+1 = xn − .
3xn2 − 30xn
Calculation of the next four approximations beginning with x1 = 3 gives
x13 − 15x12 + 125 33 − 15(3)2 + 125

x 2 = x1 − = 3 − = 3.269 84;
3x12 − 30x1 3(3)2 − 30(3)
x23 − 15x22 + 125

x 3 = x2 − = 3.263 52;
3x22 − 30x2
x33 − 15x32 + 125

x 4 = x3 − = 3.263 518 2;
3x32 − 30x3
x43 − 15x42 + 125

x 5 = x4 − = 3.263 518 2.
3x42 − 30x4
Newton’s iterative procedure has therefore produced 3.263 518 as an approximate solution to
the equation x 3 − 15x 2 + 125 = 0. In spite of the fact that we have written six decimals in
this final answer, our analysis in no way guarantees this degree of accuracy; we have simply
judged on the basis of x4 = x5 that 3.263 518 might be accurate to six decimal places. The
zero intermediate value theorem (of Section 1.11) confirms this when we calculate
f (3.263 517 5) = 4.8 × 10−5 and f (3.263 518 5) = −1.8 × 10−5 .

EXAMPLE 4.1
Use Newton’s iterative procedure to find the only positive root of the equation
3x 4 + 15x 3 − 125x − 1500 = 0
accurate to five decimal places. (This equation will be encountered in Exercise 48 of Sec-
tion 4.7.)
FIGURE 4.3 Initial approximation for Newton’s iterative procedure
y
400
200
3.5 4 4.5 5 x
−200
−400
−600
y = 3x 4 + 15x 3 − 125x − 1500
−800
−1000
SOLUTION The plot of f (x) = 3x 4 + 15x 3 − 125x − 1500 in Figure 4.3 shows that the
solution is between x = 4 and x = 5. The approximations predicted by Newton’s method are
defined by
f (xn ) 3xn4 + 15xn3 − 125xn − 1500

xn+1 = xn − = x n − .
f " (xn ) 12xn3 + 45xn2 − 125
Suppose we choose x1 = 4 as the initial approximation; this is closer to the root than x = 5.
We find that
x2 = 4.199 56, x3 = 4.187 268, x4 = 4.187 218 7, x5 = 4.187 218 7.
Since
f (4.187 215) = −5.7 × 10−3 and f (4.187 225) = 9.7 × 10−3 ,
it follows that the root is x = 4.187 22, accurate to five decimal places.
In the following example we again use Newton’s iterative procedure to approximate the solution
of an equation but specify the required accuracy as maximum error.
EXAMPLE 4.2
Use Newton’s method to find the smallest root of the cubic equation x 3 − 3x + 1 = 0 with
error less than 10−5 .
y
3
2 y = x 3 − 3x + 1
−2 −1 1 2 x
−1
SOLUTION The graph in Figure 4.4 indicates that the smallest root is just to the right of
x = −2. We take x1 = −2 and use Newton’s iterative procedure to define approximations:
f (xn ) x 3 − 3xn + 1
xn+1 = xn − "
= xn − n 2 .
f (xn ) 3xn − 3
The next four approximations are
x2 = −1.888 89, x3 = −1.879 45, x4 = −1.879 385, x5 = −1.879 385 2.
For an approximation with error less than 10−5 , we suggest −1.879 39. The zero intermediate
value theorem confirms this when we calculate
f (−1.879 40) = −1.1 × 10−4 and f (−1.879 38) = 4.0 × 10−5 .
The value of a technique for approximating solutions of an equation depends on two factors,
applicability and rate of convergence. Newton’s method scores well in both categories. The
method can be applied to any equation 4.1, provided that f (x) is differentiable in an interval
FIGURE 4.5 Initial ap- containing the root. It converges to a solution x = a , provided that f " (a) $= 0 and x1 is
proximation for Newton’s iterative chosen sufficiently close to a (see Exercise 68 in Section 10.1). The condition f " (a) $= 0 is
procedure must be sufficiently included so that the denominator in equation 4.2 does not approach 0 as xn → a . However,
close to root f (xn ) usually approaches 0 faster than f " (xn ) so that even when f " (a) = 0, Newton’s method
y may be successful. It is impossible to indicate how close x1 must be to a in order to guarantee
convergence to a . In some examples x1 can be any number whatsoever, but we are also aware
of examples where |x1 − a| must be less than 0.01 (see Example 4.3). Provided that x1 is
sufficiently close to the root, convergence of the approximations to the root is rapid. This makes
Newton’s method valuable from the point of view of the second criterion, rate of convergence.
x1 = 0 When x1 is not sufficiently close to a , Newton’s iterative procedure may not converge, or
may converge to a solution other than expected. For instance, if we attempt to approximate
the largest root in Example 4.2, and inadvertently choose x1 = 1, we cannot find the second
approximation: algebraically because 3x12 − 3 = 0, and geometrically because at x1 = 1, the
x tangent line is horizontal, and does not intersect the x -axis. The function f (x) in Figure 4.5
1 1 x2
2
has zeros near x = 1/2 and x = 1. If we attempt to find the smaller zero using an initial ap-
proximation x1 = 0, we find the second approximation x2 to be larger than 1. Further iterations
then converge to the zero near x = 1, not the zero near x = 1/2. We conclude, therefore, that
the initial approximation in Newton’s iterative procedure is most important. A poor choice for
x1 may lead to numbers that either converge to the wrong root or do not converge at all.
EXAMPLE 4.3
A uniform hydro cable P = 80 m long with mass per unit length ρ = 0.5 kg/m is hung from
FIGURE 4.6 Finding two supports at the same level L = 70 m apart (Figure 4.6). The tension T in the cable at its
tension in hydro cable at its lowest lowest point must satisfy the equation
point
70 m ρgP
= eρgL/(2T ) − e−ρgL/(2T ) ,
T
80 m
where g = 9.81 m/s2 . If we set z = ρg/(2T ) , then z must satisfy
2P z = eLz − e−Lz .
Find an approximation to the solution of this equation for z that yields T correct to one decimal
place.
SOLUTION When we substitute P = 80 and L = 70, the equation for z becomes
160z = e70z − e−70z .
The exponential function e70z grows very rapidly for z > 0, and e−70z approaches zero very
quickly. It follows that the solution of the equation must be quite small. If we plot f (z) =
e70z − e−70z − 160z on the interval 0 ≤ z ≤ 0.02 (Figure 4.7), we capture the solution. With
z1 = 0.013, Newton’s iterative procedure defines further approximations by
e70zn − e−70zn − 160zn

z1 = 0.013, zn+1 = zn − .
70e70zn + 70e−70zn − 160
To get an idea of how accurate z should be in order to give T correct to one decimal place, we
note that for z = 0.013, tension is T ≈ 189. Thus, we should determine z to around five or
six figures. Iteration of Newton’s procedure gives
z2 = 0.012 957 3, z3 = 0.012 957 0, z4 = 0.012 957 0.
With z approximated by 0.012 957 0, tension is T = 189.3 N. We could verify that this is
accurate to one decimal place by defining g(T ) = eρgL/(2T ) − e−ρgL/(2T ) − ρgP /T and
evaluating g(189.25) and g(189.35) . The first is positive and the second is negative.
f (z) = e70z − e−70z − 160z

0.1
0.05
0.005 0.01 0.015 z

−0.05
−0.1
EXERCISES 4.1
In Exercises 1–16 use Newton’s iterative procedure to find approxima- m times per year are given by the formula
tions to all roots of the equation accurate to six decimal places. In each
 
case, make a plot in order to obtain an initial approximation to each root.
 i/(100m) 
1. x 2 + 3x + 1 = 0 2. x 2 − x − 4 = 0  
M =P ( )−mn  .
 i 
3. x 3 + x − 3 = 0 4. x 3 − x 2 + x − 22 = 0 1− 1+
100m
5. x 3 − 5x 2 − x + 4 = 0 6. x 5 + x − 1 = 0
x+1 (a) What are monthly payments if P = 100 000, i = 5, and
7. x 4 + 3x 2 − 7 = 0 8. = x2 + 1 n = 25?
x−2
2 (b) What interest rate would yield monthly payments of $500
9. x − 10 sin x = 0 10. sec x = for P = 100 000 and n = 25?
1 + x4
11. (x + 1)2 = sin 4x 12. (x + 1)2 = 5 sin 4x ∗ 32. When the beam in the figure below vibrates vertically, there are
certain frequencies of vibration called natural frequencies. They are
13. x + 4 ln x = 0 14. x ln x = 6 solutions of the equation
15. ex + e−x = 10x 16. x 2 − 4e−2x = 0
ex − e−x
tan x =
ex + e−x
In Exercises 17–24 use Newton’s iterative procedure to find approx-
imations to all roots of the equation with error no greater than that divided by 20π . Find the two smallest frequencies correct to four
specified. decimal places.
17. x 3 − 5x − 1 = 0, 10−3
10 m
18. x 4 − x 3 + 2x 2 + 6x = 0, 10−4
x
19. = x 2 + 2, 10−5
x+1
20. (x + 1)2 = x 3 − 4x, 10−3
21. (x + 1)2 = 5 sin 4x, 10−3 ∗ 33. A stone of mass 100 g is thrown vertically upward with speed
2 2 −4
20 m/s. Air exerts a resistive force on the stone proportional to its
22. cos x = x − 1, 10 speed, and has magnitude 0.1 N when the speed of the stone is 10 m/s.
23. x + (ln x)2 = 0, 10−3 It can be shown that the height y above the projection point attained
by the stone is given by
24. e3x + ex = 4, 10−4
# $
y = −98.1t + 1181 1 − e−t/10 m,
In Exercises 25–28 find all points of intersection for the curves accurate
to four decimal places. where t is time (measured in seconds with t = 0 at the instant of
projection).
∗ 25. y = x 3 , y =x+5
(a) The time taken for the stone to return to its projection point
∗ 26. y = (x + 1)2 , y = x 3 − 4x can be obtained by setting y = 0 and solving the equation
for t . Do so (correct to two decimal places).
∗ 27. y = x 4 − 20, y = x 3 − 2x 2
(b) When air resistance is neglected, the formula for y is
x 2
∗ 28. y = , y =x +2
x+1 y = 20t − 4.905t 2 m.
∗ 29. Show algebraically and geometrically that Newton’s method never
gives the solution of the equation f (x) = x 1/3 = 0, for any initial What is the elapsed time in this case from the instant the
approximation whatsoever. stone is projected until it returns to the projection point?
∗ 30. Show algebraically and geometrically that Newton’s method al- ∗ 34. Planck’s law for the energy density E of blackbody radiation at
ways gives the solution of the equation f (x) = x 7/5 = 0, for any 1000 ◦ K states that
initial approximation whatsoever.
∗ 31. Suppose you mortgage your house for P dollars. To repay the kλ−5
E = E(λ) = ,
loan at an interest rate of i %, amortized over n years, payments made ec/λ−1
where k > 0 is a constant and c = 0.000 143 86. This function is (ii) Rise time, Tr , defined as the time for the graph to rise from
shown in the figure below. The value of λ at which E is a maximum 0.1 to 0.9;
must satisfy the equation (iii) Settling time, Ts , defined as the time to reach and remain
c/λ within the interval 0.95 ≤ f (t) ≤ 1.05.
(5λ − c)e − 5λ = 0.
Find √these times, correct to two decimal places, if f (t) = 1 +
Find this value of λ correct to seven decimal places. e−2.5 11t sin (20t − π/2) .
E
Maximum E
1.25
1.0
0.75
0.5
0.25
0.2 0.4 0.6 t

∗ 35. The speed of response of an oscillatory system whose graph y =
f (t) is shown below is often determined by one of the following three ∗ 36. Suppose that a cubic polynomial P (x) has three distinct real zeros.
time constants: Show that when Newton’s method uses an initial approximation that is
(i) Delay time, Td , defined as the time required for the graph equal to the average of two of the zeros, then the first iteration always
to reach 0.5; yields the third zero.
4.2 Increasing and Decreasing Functions

Many mathematical concepts have their origin in intuitive ideas. In this section we analyze
the intuitive idea of one quantity “getting larger” and another “getting smaller.” To describe
what it means for a quantity to be increasing (getting larger) or decreasing (getting smaller),
we first suppose that the quantity is represented by some function f (x) of a variable x . The
mathematical definition for f (x) to be increasing or decreasing is as follows.
DEFINITION 4.1
A function f (x) is said to be increasing on an interval I if for all x1 > x2 in I ,
f (x1 ) > f (x2 ). (4.3a)
A function f (x) is said to be decreasing on I if for all x1 > x2 in I ,
f (x1 ) < f (x2 ). (4.3b)
The continuous function in Figure 4.8 is increasing on the intervals
a ≤ x ≤ b, c ≤ x ≤ d, e ≤ x ≤ f,
and decreasing on the intervals
b ≤ x ≤ c, d ≤ x ≤ e.
When a function has points of discontinuity, such as in Figure 4.9, the situation is somewhat
more complicated. This function is increasing on the intervals
a ≤ x < b, d ≤ x < e, e < x ≤ f,

and decreasing for
b < x < d.
Pay special attention to whether endpoints of each interval are included.

Figures 4.8 and 4.9 indicate that the sign of f " (x) determines whether a function is increas-
ing or decreasing on an interval. The following test describes the situation.
FIGURE 4.8 Intervals on which a continuous FIGURE 4.9 Intervals on which a discontinu-
function is increasing and decreasing ous function is increasing and decreasing
y y
a b c d e f x a b c d e f x
Increasing and Decreasing Test

(i) A function f (x) is increasing on an interval I if on I ,
f " (x) ≥ 0 (4.4a)
and is equal to zero at only a finite number of points.
(ii) A function f (x) is decreasing on an interval I if on I ,
f " (x) ≤ 0 (4.4b)
FIGURE 4.10 f (x) is
Figures 4.10 and 4.11 illustrate why we permit f " (x) to vanish at only a finite number of points
not increasing on a ≤ x ≤ b if
and do not, therefore, allow it to vanish on an interval. In Figure 4.10, f " (x) is equal to zero on
f " (x) = 0
the interval a ≤ x ≤ b , and certainly, f (x) is not increasing on any interval that contains these
y points. The function f (x) in Figure 4.11 has f " (0) = 0 = f " (1) , and yet f (x) is increasing
on the interval a ≤ x ≤ b .
Conditions 4.4 are sufficient to guarantee that a function is increasing or decreasing on an
interval; they are not necessary. For example, the function f (x) = x − sin x in Figure 4.12 has
derivative equal to 0 at the infinity of values x = 2nπ , where n is an integer, yet the function is
increasing on the interval −∞ < x < ∞ . What we are saying is that tests more general than
x 4.4 can be formulated, but we feel that the extra complexity is not worth the gain. For a proof
a b
of 4.4, see Exercise 51.
FIGURE 4.12 Even if f " (x) = 0 at an infinity of points, f (x) may be increasing
FIGURE 4.11 f (x) is
increasing on a ≤ x ≤ b even if y
f " (x) = 0 at a finite number of
10
points y = x − sin x
y 5
−15 −10 −5 5 10 x
−5
−10
a 1 b x −15
Intervals on which a function is increasing and decreasing are separated by points where
the derivative is equal to zero or does not exist. Keep this in mind in the following example.
EXAMPLE 4.4
Find intervals on which the following functions are increasing and decreasing:
x2 − 9
(a) f (x) = 2x 3 + 3x 2 − 5x + 4 (b) f (x) =
x2 + x − 2
SOLUTION
(a) As mentioned above, the key to finding intervals on which the derivative of a function
is positive and negative are points where the derivative is either zero or nonexistent.
Since polynomials have derivatives everywhere, we investigate where the derivative
is equal to zero, √ √
" 2 −6 ± 36 + 120 −3 ± 39
0 = f (x) = 6x + 6x − 5 -⇒ x = = .
12 6
To determine intervals on which the function is increasing and decreasing we can
proceed in a number of ways. Firstly, if we have a plot of the function √ (Figure
4.13), it is clear
√ that f (x) is increasing on the intervals
√ x ≤ (− 3 − 39√)/6 and
x ≥ (−3 + 39)/6, and decreasing on (−3 − 39)/6 ≤ x ≤ (−3 + 39)/6.
Secondly, if we do not have a graph of the function, we could visualize that a√ graph of
f " (x) (Figure 4.14) is a parabola that crosses the x -axis at the points (−3 ± 39)/6.
Notice that we said visualize this graph; we have drawn it, but it would be necessary
"
only to mentally visualize √ it. Since the parabola opens √ upward, f (x) ≥ 0 on
"
the intervals
√ x ≤ (− 3 − 39 )/
√ 6 and x ≥ (− 3 + 39 )/ 6, and f (x) ≤ 0 on
(−3 − 39)/6 ≤ x ≤ (−3 + 39)/6.
FIGURE 4.13 Graph to determine when FIGURE 4.14 Graph to determine the
function is increasing and decreasing sign of f " (x)
y f'(x)
20
y = 2x 3 + 3x 2 − 5x + 4
15
10
5
(−3 − √39)/6 (−3 + √39)/6 x
−3 −2 −1 1 2 x
−5
Finally, we could construct a sign table for f " (x) as introduced in Table 1.1 of
Section 1.5 √ It also shows that f " (x) ≥ 0 on the
(see figure 4.15 below). √ √ intervals
x ≤ (−3 − √39)/6 and x ≥ (−3 + 39)/6, and f " (x) ≤ 0 on (−3 − 39)/6 ≤
x ≤ (−3 + 39)/6.
FIGURE 4.15
(−3 − √39)/6 (−3 + √39)/6

y − (−3 − √39)/6 − + x
y − (−3 + √39)/6 − +
f'(x) + − +
(b) The graph of f (x) in Figure 4.16a indicates discontinuities at x = −2 and x = 1

and the point between x = −2 and x = 0 where f " (x) = 0 separates intervals on
which the function is increasing and decreasing. Whether there are other points far
out on the x -axis where f " (x) = 0 is not clear. To find all points where f " (x) = 0,
we set
(x 2 + x − 2)(2x) − (x 2 − 9)(2x + 1) x 2 + 14x + 9

0 = f " (x) = = .
(x 2 + x − 2)2 (x 2 + x − 2)2
This implies that x 2 +
√14x + 9 = 0, and there are two points where f (x) = 0,
"
namely x = −7 ± 2 10. Figure 4.16b is an exaggerated version of the graph of

"
f (x) to the left
√ of x = −2; it is not a computer plot. It illustrates that f (x) = 0 at
x = −7 − 2 10 and that the graph is asymptotic to the line y = 1 (from above).
We can now say that the function is increasing on the intervals
√ √
−∞ < x ≤ −7 − 2 10, −7 + 2 10 ≤ x < 1, 1 < x < ∞,
and decreasing for
√ √
−7 − 2 10 ≤ x < −2, −2 < x ≤ −7 + 2 10.
Points of discontinuity may also separate intervals on which a function is increasing and decreasing
y y
40 x2 − 9
y=
x2 + x − 2 1
20
x x
−3 −2 −1 1 2 −7 − 2√10 −3
−20
−40
−60
EXERCISES 4.2
In Exercises 1–26 determine intervals on which the function is increas- x x2 + 4
∗ 17. f (x) = ∗ 18. f (x) =
ing and decreasing. 2−x x2 − 1
1. f (x) = 2x − 3 2. f (x) = 4 − 5x x3
∗ 19. f (x) = ∗ 20. f (x) = |x 2 − 1| + 1
x+1
3. f (x) = x 2 − 3x + 4 4. f (x) = −2x 2 + 5x
5. f (x) = 3x 2 + 6x − 2 6. f (x) = 5 + 2x − 4x 2 ∗ 21. f (x) = xe−x ∗ 22. f (x) = x 2 e−x

7. f (x) = 2x 3 − 18x 2 + 48x + 1
∗ 23. f (x) = ln (x 2 + 5) ∗ 24. f (x) = x ln x
3 2
8. f (x) = x + 6x + 12x + 5
9. f (x) = 4x 3 − 18x 2 + 1 x2 − 9 x 3 + 2x 2 − x − 2
∗ 25. f (x) = ∗ 26. f (x) =
x−3 2 − x − x2
10. f (x) = 4 − 18x − 9x 2 − 2x 3
11. f (x) = 3x 4 + 4x 3 − 24x + 2 ∗ 27. If the price of a certain car is set at r , then the market demands x
4 3 2
cars per year, where
12. f (x) = 3x − 4x + 24x − 48x
13. f (x) = x 4 − 4x 3 − 8x 2 + 48x + 24 x = 4a 3 − 3ar 2 + r 3 ,

14. f (x) = x 5 − 5x + 2
where a > 0 is a constant, and 0 < r < 2a . Show that the price
1 2 1 function r = f (x) defined implicitly by this equation is a decreasing
∗ 15. f (x) = x + ∗ 16. f (x) = x +
x x2 function.
Figure (a) below contains the graph of a function f (x) and figure (b), ∗ 40. Show algebraically that the equation x 23 + 3x 15 + 4x + 1 = 0
the graph of f " (x) . Note that f " (x) ≥ 0 when f (x) is increasing has exactly one (real) solution.
and f " (x) ≤ 0 when f (x) is decreasing. In addition, the corner in
∗ 41. Show that the equation ax 5 + bx 3 + c = 0, where a , b , and c
f (x) at x = 2 is reflected in the discontinuity in f " (x) . In Exercises are constants such that ab > 0, has exactly one (real) solution.
28–35 draw similar graphs for the function and its derivative.
∗ 42. Show that the equation x n +ax − 1 = 0, where a > 0 and n ≥ 2
are constants, has exactly one positive root.
y y
4 ∗ 43. Repeat Exercise 42 for the equation x n + x n−1 − a = 0.
3 Graph of f (x) Graph of f "(x)
3 ∗ 44. Show that sin x < x for all x > 0. Hint: Calculate f (0) and
2
2 f " (x) for f (x) = x − sin x .
1 1
∗ 45. Show that cos x > 1 − x 2 /2 for all x > 0. Hint: See the
1 2 3 4 x 1 2 3 4 x technique of Exercise 44.
(a) (b) ∗ 46. Use the result of Exercise 45 to prove that for x > 0,
x3
sin x > x − .
6
∗ 28. f (x) = x 2 + 2x ∗ 29. f (x) = x 4 − x 2

∗ 47. Use the result of Exercise 46 to prove that for x > 0,
x−1 x−4
∗ 30. f (x) = ∗ 31. f (x) = x2 x4
x+2 x2 cos x < 1 − + .
2 24
|x|
∗ 32. f (x) = ∗ 33. f (x) = |x 2 − 4| ∗ 48. Use the technique of Exercise 44 to verify that for x > 0,
x
2
|x − 4| 1 3x
∗ 34. f (x) = ∗ 35. f (x) = sin 3x √ > 1− .
x−2 1 + 3x 2
∗ 49. If f (x) and g(x) are differentiable and increasing on an interval

I , is f (x)g(x) increasing on I ?
In Exercises 36–39 find (accurate to four decimal places) intervals on ∗ 50. If positive functions f (x) and g(x) are differentiable and increas-
which the function is increasing and decreasing. ing on an interval I , is f (x)g(x) increasing on I ?
∗ 51. Verify test 4.4.
∗ 36. f (x) = x 4 + 2x 2 − 6x + 5 ∗ 52. A number x0 is called a fixed point of a function f (x) if f (x0 ) =
x0 . How many fixed points can a function have if f " (x) < 1 for all
∗ 37. f (x) = 3x 4 − 20x 3 − 24x 2 + 48x x?
∗∗ 53. Prove that when 0 < a .
∗ 39. f (x) = tan x − x(x + 2), −π/2 < x < π/2 tan a a
4.3 Relative Maxima and Minima

One of the most important applications of calculus is in the field of optimization, the study of
maxima and minima. In this section we begin discussions of this topic, which continue through
to Section 4.7. Fundamental to discussions on optimization are critical points.
DEFINITION 4.2
A critical point of a function is a point in the domain of the function at which the first
derivative either is equal to zero or does not exist.
Specifically, x = c is a critical point for f (x) if f " (c) = 0 or f " (c) does not exist, but in
the latter case, f (c) must exist. With the interpretation of the derivative as the slope of a tangent
line, we can state that corresponding to a critical point of a function, the graph of the function
has a horizontal tangent line, a vertical tangent line, or no tangent line at all. For example, the
eight points a through h on the x -axis in Figure 4.17 are all critical. At a , b , c , and d , the
tangent line is horizontal; at e and f , the tangent line is vertical; and at g and h , there is no
tangent line.
Often overlooked by students, but very important in this definition, is that a function must
be defined at a critical point; it must have a value, and therefore, there must be a point on the
graph of the function at a critical point. For instance, if the dot in Figure 4.17 at the discontinuity
x = h is replaced by an open circle, then x = h is no longer a critical point.
FIGURE 4.17 The critical points of a function
a b c d e f g h x
When the domain of a function f (x) is a closed interval a ≤ x ≤ b , the endpoint x = a

is said to be critical if the right-hand derivative of f (x) at x = a is equal to zero or does not
FIGURE 4.18 Critical
exist. Likewise, x = b is critical if the left-hand derivative vanishes or does not exist there.
points at end points when function
defined on closed interval
This is consistent with the definition of differentiability in Section 3.3. The function in Figure
4.18 has domain 1 ≤ x ≤ 3. Both endpoints are critical — x = 1 because f+" (1) does not
y
exist, and x = 3 because f−" (3) = 0.
The function in Figure 4.17 is discontinuous at x = h , and we know that a function cannot
y = f (x) have a derivative at a point of discontinuity (Theorem 3.6). This does not mean that every point
of discontinuity of a function is critical. Remember, a function must be defined at a point for
that point to be critical. Thus, points of discontinuity are critical only if the function is defined
x at the point. In the remainder of this section we consider only critical points at which a function
1 2 3
is continuous and which are not endpoints of its domain of definition.
EXAMPLE 4.5
Find critical points for the following functions:
(a) f (x) = x 3 − 7x 2 + 11x + 6
x2
(b) f (x) =
x3 − 1
(c) f (x) = x ln x
(d) f (x) = |x|
SOLUTION
(a) For critical points we first solve
0 = f " (x) = 3x 2 − 14x + 11 = (3x − 11)(x − 1),
and obtain x = 11/3 and x = 1. These are the only critical points since there are
no points where f " (x) does not exist.
(b) For critical points we calculate
(x 3 − 1)(2x) − x 2 (3x 2 ) −x(x 3 + 2)

f " (x) = = .
(x 3 − 1)2 (x 3 − 1)2
Clearly, f " (x) = 0 when x = 0 and when x 3 + 2 = 0, which implies that

x = −21/3 . The derivative does not exist when x 3 − 1 = 0 (i.e., when x = 1).
But f (x) is not defined at x = 1 either, and therefore x = 1 is not a critical point.
There are only two critical points: x = 0 and x = −21/3 .
(c) For critical points we first solve
x
0 = f " (x) = ln x + = ln x + 1.
x
The only solution of this equation is x = 1/e . Since f " (x) exists for all x in the
domain of f (x) , namely, x > 0, the function has no other critical points.
(d) The only critical point of f (x) = |x| is x = 0. Its derivative is equal to 1 when
x > 0, and to −1 when x < 0, but does not exist at x = 0.
At the critical points b and f in Figure 4.19, the graph of the function has “high” points. They
are described in the following definition.
DEFINITION 4.3
A function f (x) is said to have a relative (or local) maximum f (x0 ) at x = x0 if there
exists an open interval I containing x0 such that for all x in I ,
f (x) ≤ f (x0 ). (4.5)
Since such intervals can be drawn around x = b and x = f , relative maxima occur at
these points. At a relative or local maximum, the graph of the function is highest relative to
nearby points.
Critical points d and h in Figure 4.19, where the graph has “low” points, are described in
a similar definition.
FIGURE 4.19 Nature of a graph at critical points
a b c d e f g h i x
DEFINITION 4.4
A function f (x) is said to have a relative (or local) minimum f (x0 ) at x = x0 if there
exists an open interval I containing x0 such that for all x in I ,
f (x) ≥ f (x0 ). (4.6)

Relative minima therefore occur at x = d and x = h in Figure 4.19.

The critical points x = a and x = c , where f " (x) = 0 and x = e and x = g , where
"
f (x) does not exist, will be discussed in Section 4.4. At x = i , the graph takes an abrupt
change in direction. The function has a left- and a right-hand derivative, but f " (i) does not
exist. In Section 3.3 we called the corresponding point on the curve a corner. Corners can
sometimes be relative extrema ( x = h yields a corner and a relative minimum).
Relative maxima and minima represent high and low points on the graph of a function
relative to points near them. It is not coincidence in Figure 4.19 that the two relative maxima
and the two relative minima occur at critical points. According to the following theorem, this is
always the case.
THEOREM 4.1
Relative maxima and relative minima of a function must occur at critical points of the
function.
FIGURE 4.20 Proof that PROOF To verify this we prove that at any point at which the derivative f " (x) of a function
relative extrema occur at critical
f (x) exists and is not zero, it is impossible for f (x) to have a relative extremum, that is, a
points
relative maximum or a relative minimum. Suppose that at some point x = a , the derivative
y f " (a) exists and is positive (Figure 4.20):
f (a + h) − f (a)
"
0 < f " (a) = lim .
f (a) = slope > 0 h→0 h
According to the proof developed in Exercise 29 of Section 2.6, there exists an open interval
f (a + h) I : b < x < c around x = a in which
f (a)
f (a + h) − f (a)
x > 0.
b a a+h c h
This implies that when h > 0 (and a + h is in I ), f (a + h) − f (a) must also be positive,
and therefore f (a + h) > f (a) . But when h < 0 (and a + h is in I ), f (a + h) − f (a)
must be negative, and therefore f (a + h) < f (a) . There is an interval b < x < a in which
f (x) < f (a) , and an interval a < x < c in which f (x) > f (a) . Thus, x = a cannot yield
a relative extremum.
A similar proof holds when f " (a) < 0.
Although every relative extremum of a function must occur at a critical point, not all critical
points give relative extrema. For continuous functions, there is a simple test to determine
whether a critical point gives a relative maximum or a relative minimum. To understand this
test, consider the critical points in Figure 4.19 as you read the statements.
First-Derivative Test for Relative Extrema of

Continuous Functions
(i) If f " (x) (slope of a graph) changes from a positive quantity to a negative quantity
as x increases through a critical point at which f (x) is continuous, the critical point
yields a relative maximum for f (x) .
(ii) If f " (x) changes from a negative quantity to a positive quantity as x increases through
a critical point at which f (x) is continuous, the critical point yields a relative mini-
mum for f (x) .
Various possibilities can occur if f " (x) does not change sign as x increases through a critical
point. For instance, if f " (x) is positive (or negative) on both sides of a critical point [such as
is the case for the critical point x = 0 of f (x) = x 3 ], then the critical point cannot yield a
relative maximum or minimum. It might also happen that f " (x) is both positive and negative
in every interval around a critical point, no matter how small the interval. This is investigated
in Exercises 76–78.
EXAMPLE 4.6
Find relative maxima and relative minima for the following functions:
(a) f (x) = 2x 3 − 9x 2 − 23x + 6
(b) f (x) = 3.01x − sin (3x + 1), 0≤x≤5
5/3 2 /3
(c) f (x) = x −x
(d) f (x) = 5
SOLUTION
(a) The graph of the function in Figure 4.21a indicates a relative maximum near x = −1
and a relative minimum near x = 4. To locate them precisely, we find critical points
by solving
√ √
" 2 18 ± 324 + 24 · 23 9± 219
0 = f (x) = 6x − 18x − 23 -⇒ x = = .
12 6
These are the only critical points, as there are no points at which f " (x)√is undefined.
Figure 4.21a indicates that a relative
√ maximum occurs at x = (9 − 219)/6 and
a relative minimum at x = (9 + 219)/6. This can be confirmed algebraically or
with the graph of f " (x) in Figure 4.21b. It shows that as x increases through
√ √ (9 −
219)/6, f " (x) changes from positive
# √to negative;
$ therefore, x = ( 9 − 219)/6
yields a relative maximum of f (9 − 219)/6 ≈ 18.02. Since f " (x) changes
√
from negative to positive as x #increases√ through $ (9 + 219)/6, this value of x
yields a relative minimum of f (9 + 219)/6 ≈ −102.0.
(b) The graph in Figure 4.22 does not make it clear whether the function has relative
extrema; the tangent line may become horizontal at or near x = 2 and x = 4, but
we cannot be sure. Critical points of the function are given by
3.01
0 = f " (x) = 3.01 − 3 cos (3x + 1) -⇒ cos (3x + 1) = .
3
Since 3.01/3 > 1, this equation has no solutions. Hence, f (x) has no critical points
and there can be no relative maxima or minima.
FIGURE 4.21a Relative extrema FIGURE 4.21b Sign changes of f " (x) at critical points
y y
200 y = 2x 3 − 9x 2 − 23x + 6 30
f "(x) = 6x 2 − 18x − 23
100 20
10
−4 −2 2 4 6 8 x
−100 −2 −1 1 2 3 4 5 x
−10
−200 −20
−300 −30
FIGURE 4.22 Plot may not be conclusive evidence of FIGURE 4.23 Graph showing a relative minimum and
existence of relative extrema suggesting a relative maximum
y y
15
0.5 y = x 5/3 − x 2/3
10
y = 3.01x − sin (3x + 1) x
5 −1 1 2
−0.5
−2 2 4 6 x
−5 −1
(c) The graph of the function in Figure 4.23 makes it clear that a relative minimum occurs
just to the left of x = 1/2. It also suggests that f (0) = 0 is a relative maximum.
Confirmation is provided by critical points, found by first solving
5 2 5x − 2
0 = f " (x) = x 2/3 − x −1/3 = .
3 3 3x 1/3
Clearly, x = 2/5 is a critical point. The relative minimum at this point is f (2/5) ≈
−0.33. In addition, because f " (0) does not exist, but f (0) = 0, it follows that
x = 0 is also a critical point. There are two reasons, each sufficient by itself to
conclude that f (0) = 0 is a relative maximum. First, f (x) is continuous at x = 0,
and f " (x) changes from a positive quantity to a negative quantity as x increases
through 0. Second, since f (x) = x 2/3 (x − 1) is negative for all x < 1, except
at x = 0 where f (0) = 0, there must be a relative maximum at x = 0. It is also
interesting to note that
lim f " (x) = ∞ and lim f " (x) = −∞.
x→0− x→0+
This means that the graph has a very sharp point at (0, 0) .
(d) The graph of this function is a horizontal straight line. Every value of x is critical
and at each value of x , the function has a relative maximum and a relative minimum
of 5.
EXAMPLE 4.7
The equation x 2 y + y 3 = 8 defines y implicitly as a function of x . Find all critical points of
the function and determine whether they yield relative maxima or minima for the function.
SOLUTION When we differentiate the equation implicitly with respect to x , we obtain
dy dy dy 2xy
2xy + x 2 + 3y 2 =0 -⇒ =− 2 .
dx dx dx x + 3y 2
The derivative is defined for all x and y except when both are simultaneously zero, a point that
does not satisfy the original equation. Consequently we consider when the derivative vanishes.
For this to happen, either x or y must be zero. The original equation does not permit y to be
zero, and when x = 0, the only solution of the equation is y = 2. Hence, x = 0 is the only
critical point of the implicitly-defined function. Since the function must be continuous at x = 0
(Theorem 3.6), values of y are close to 2 when values of x are close to zero. It follows that the
derivative changes from a positive quantity to a negative quantity as x increases through zero,
and the function has a relative maximum at x = 2.
EXAMPLE 4.8
The Beattie-Bridgeman equation of state for an ideal gas relates pressure P and volume V
according to
( )! ( )"
RT CT b A - a.
P = 2 1− 3 V +B 1− − 2 1− ,
V V V V V
where R = 0.082 06 is the universal gas constant, T is absolute temperature of the gas, and
a , A , b , B , and C are constants. For air, a = 0.019 31, A = 1.3012, b = −0.0011,
B = 0.046 11, and C = 4.34 × 104 . For T = 300, plot a graph of P as a function of V on
the interval 300 ≤ V ≤ 400. Verify that to one-decimal-place accuracy, V = 373.5 gives a
relative maximum.
FIGURE 4.24 Relative maximum of Beattie–Bridgeman equation of state
P
0.0495
0.049
0.0485
0.048
0.0475
320 340 360 380 400 V

0.0465
SOLUTION The plot of the function in Figure 4.24 confirms a relative maximum near V =
370. To verify that 373.5 is the critical point yielding this maximum, we calculate
( )! ( )"
" 2 5CT b
P (V ) = RT − 3 + V +B 1−
V V6 V
( )! " ( )
1 CT Bb 2 3a
+ RT − 1 + − A − + .
V2 V5 V2 V3 V4
Since P " (373.45) = 2.29 × 10−8 and P " (373.55) = −1.19 × 10−7 , it follows (by the zero
intermediate value theorem) that, to one decimal place, the critical point is 373.5.
EXERCISES 4.3
In Exercises 1–44 find all critical points of the function and determine 9. f (x) = x 3 /3 − x 2 /2 − 2x
algebraically with the first-derivative test which critical points give rel- 10. f (x) = x 1/3
ative maxima and relative minima. Use a plot to confirm your findings.
∗ 11. f (x) = sin2 x
1. f (x) = x 2 − 2x + 6 x3
∗ 12. f (x) =
2. f (x) = 2x 3 + 15x 2 + 24x + 1 x4+1
∗ 13. f (x) = 3x 4 − 16x 3 + 18x 2 + 2
3. f (x) = x 4 − 4x 3 /3 + 2x 2 − 24x 1
x+1 ∗ 14. f (x) = x +
4. f (x) = (x − 1)5 5. f (x) = 2 x
x +8 ∗ 15. f (x) = 2x 3 − 15x 2 + 6x + 4
x2 + 1 x ∗ 16. f (x) = |x| + x
6. f (x) = 7. f (x) = √
x−1 1−x ∗ 17. f (x) = (x − 1)2/3
8. f (x) = x 3 − 6x 2 + 12x + 9
(x − 1)3
∗ 18. f (x) =
(x + 1)4
∗ 19. f (x) = (x + 2)3 (x − 4)3 defined implicitly by the equation 2x 2 − y 3 + xy = 4.

∗ 20. f (x) = x + 2 sin x /
∗ 54. The equation y = x 2 1 − y 2 defines y implicitly as a function
of x . Find critical points of the function at which the first derivative
∗ 21. f (x) = x 1/5 + x
vanishes by (a) using implicit differentiation, and (b) finding the explicit
25x 2 definition of the function.
∗ 22. f (x) = x 2 +
(x − 2)2 ∗ 55. An N -wave solution of the Burgers equation in fluid dynamics is
( ) of the form
x+8 / 2 x
∗ 23. f (x) = x + 100 f (x) = 2
,
x 1 + bex /a
1 + x + x2 + x3 where a > 0 and b > 0 are constants. Prove that f (x) has exactly
∗ 24. f (x) = two critical points, one the negative of the other.
1 + x3
x2 ∗ 56. (a) If f (x) = x 2 is defined only for 0 ≤ x ≤ 1, are x = 0
∗ 25. f (x) = x 5/4 − x 1/4 ∗ 26. f (x) = 2 and x = 1 critical points?
x −4
(b) Does the function have a relative minimum of f (0) = 0
x2 (2x − 1)(x − 8) and a relative maximum of f (1) = 1?
∗ 27. f (x) = 3 ∗ 28. f (x) =
x −1 (x − 1)(x − 4)
∗ 29. f (x) = sin2 x cos x, 0 ≤ x ≤ 2π In Exercises 57–64 determine whether the statement is true or false.
∗ 30. f (x) = sin x + cos x ∗ 57. Points of discontinuity of a function are critical if, and only if, the
∗ 31. f (x) = 2 csc x − cot x, 0 < x < π/2 function is defined at the point of discontinuity.
∗ 32. f (x) = csc x + 8 sec x, 0 < x < π/2 ∗ 58. When a function is defined only on the interval a ≤ x ≤ b , the
ends x = a and x = b must yield relative maxima or minima for the
tan x function.
∗ 33. f (x) =
x
∗ 59. A function is discontinuous at a point if, and only if, it has no
∗ 34. f (x) = x + sin2 x, 0 < x < 2π derivative at the point.
∗ 35. f (x) = e1/x ∗ 36. f (x) = x ln x ∗ 60. A function can have a relative maximum and a relative minimum
at the same point.
∗ 37. f (x) = x 2 ln x ∗ 38. f (x) = xe−2x
∗ 61. If the derivative of a function changes sign when passing through
2
∗ 39. f (x) = xe−x ∗ 40. f (x) = x 2 e3x a point, the point must yield a relative extremum for the function.
∗ 41. f (x) = x 3 − Tan −1 x ∗ 42. f (x) = Cos −1 (2x) − 5x 2 ∗ 62. If a function has two relative maxima, it must have a relative min-
imum between them.
∗ 43. f (x) = h(x − a) (See equation 2.7.)
∗ 63. It is possible for every point in the domain of a function to be
∗ 44. f (x) = /x0 (See Exercise 68 in Section 1.5.) critical.
∗ 64. On an interval of finite length, a nonconstant function can have
In Exercises 45–50 y is defined implicitly as a function of x . Find all only a finite number of critical points at which its derivative is equal to
critical points of the function at which its derivative is equal to zero, zero.
and classify each as yielding a relative maximum or minimum.
∗ 65. Verify that the function f (x) = x 3 + cos x has two critical points,
45. x 4 + y 3 + y 5 = 1 46. x 2 + y 3 + y = 4 one of which is x = 0. Find the other critical point correct to three
decimal places.
47. x 3 y + xy 3 = 2 ∗ 48. y 4 + xy 3 = 1
∗ 66. The gas equation of Dieterici relating pressure P and volume V
∗ 49. x 4 y + y 5 = 32 ∗ 50. x 2 y 4 + y 3 = 1, y≥0
is
P (V − b) = RT e−a/(RT V ) ,
In Exercises 51–52 find all critical points at which the first derivative
vanishes for any function, with y as dependent variable, defined im- where R is the universal gas constant, T is absolute temperature, and
plicitly by the equation. a > 0 and b > 0 are constants.
(a) Verify that the function P (V ) has two critical points pro-
51. x 2 + 2xy + 3y 2 = 2 ∗ 52. x 4 y + y 5 = 4x vided T < a/(4bR) .
∗ 53. Find, accurate to four decimals, all critical points at which the (b) Verify that when T = Tc = a/(4bR) , there is one critical
first derivative vanishes for any function, with y as dependent variable, point at which P = a/(4b2 e2 ) .
∗ 67. (a) The Dirichlet function in digital signal processing is

y
sin (ωL/2)
fL (ω) = ,
L sin (ω/2)
where L is a constant. Plot it for L = 10 on the interval x

−10 ≤ ω ≤ 10. Use the plot to aid in answering the
following questions.
(b) Is f10 (0) defined? What does the graph suggest for the
∗ 75. The equation (x 2 + y 2 )2 = x 2 y describes a bifolium (figure
limit of fL (ω) as ω → 0?
below). Find the points on the curve farthest from the origin.
(c) Is fL (ω) even, odd, or neither even nor odd?
(d) Is f10 (ω) periodic? What is its period? y
(e) For what smallest positive value of ω does f10 (ω) have its
smallest positive relative maximum?
(f) What are the zeros of fL (ω) ? x
∗ 68. (a) Plot a graph of the function f (x) = e−x sin x , x ≥ 0.
∗ 76. (a) Verify that x = 0 is a critical point for the function
(b) For what values of x does f (x) have relative extrema?
0
x sin (1/x), x $= 0
f (x) =
0, x =0
In Exercises 69–72 find all critical points of the function correct to four
decimal places. Determine whether each critical point yields a relative because f " (0) does not exist.
maximum or a relative minimum.
(b) Use a plot to show that f " (x) does not change sign as x
4 2 increases through x = 0.
∗ 69. f (x) = x + 6x + 4x + 1
(c) Does x = 0 yield a relative maximum or minimum for
∗ 70. f (x) = x 4 − 10x 2 − 4x + 5 f (x) ?
∗ 71. f (x) = x 3 − 2 cos x
∗ 77. Repeat Exercise 76 for the function
x2 − 4 0
∗ 72. f (x) = 2 |x sin (1/x)|, x $= 0
(x − 5x + 4)2 f (x) =
0, x = 0.
∗ 73. The distance from the origin/to any point (x, y) on the curve
C : y = f (x) is given by D = x 2 + y 2 . Show that if P (x0 , y0 ) ∗ 78. Repeat Exercise 76 for the function
is a point on C for which D has a relative extremum, then the line OP 0
is perpendicular to the tangent line to C at P . Assume that f (x) is −|x sin (1/x)|, x $= 0
f (x) =
differentiable, and that the curve does not pass through the origin. 0, x = 0.
∗ 74. The equation (x 2 + y 2 + x)2 = x 2 + y 2 describes a cardioid
(shown in the following figure). Find the maximum y -coordinate for ∗∗ 79. Show that the function f (x) = x 3 + px + q has three distinct
points on the curve. zeros if and only if 4p 3 + 27q 2 < 0.
4.4 Concavity and Points of Inflection

In the first three sections of this chapter we concentrated on the first derivative of a function. In
this section we turn our attention to the second derivative.
Consider the function in Figure 4.25. If we draw tangent lines to the graph at the five points
c1 , c2 , c3 , c4 , and c5 , it is clear that the slope is greater at c5 than it is at c4 , greater at c4 than
at c3 , and so on. In fact, given any two points x1 and x2 in the interval a < x x1 , the slope at x2 is greater than at x1 ,
f " (x2 ) > f " (x1 ).
What we are saying is that the function f " (x) is increasing on the interval a < x < b .
FIGURE 4.25 f " (x) increasing on FIGURE 4.26 f " (x) decreasing on FIGURE 4.27 f " (x) changes
a<x<b b<x<c from increasing to decreasing at x = b
y y y
a b c x a b c x
a c1 c2 c3 c 4 c5 b x
The first derivative of the function f (x) in Figure 4.26 is decreasing on the interval b <
x < c . In Figure 4.27, we have pieced together the functions in Figures 4.25 and 4.26 to form
a function that has f " (x) increasing on a < x ≤ b and decreasing on b ≤ x < c . We
assume that the left-hand derivative f−" (b) in Figure 4.25 and the right-hand derivative f+" (b)
in Figure 4.26 are the same, in which case f " (b) exists in Figure 4.27. We give names to
intervals on which the first derivative of a function is increasing and decreasing, and points that
separate such intervals in the following definition.
DEFINITION 4.5
The graph of a function f (x) is said to be concave upward on an interval I if f " (x) is
increasing on I , and concave downward on I if f " (x) is decreasing on I . Points on a
graph that separate intervals of opposite concavity are called points of inflection.
The graph in Figure 4.28 is concave upward on the intervals
b ≤ x ≤ c, d ≤ x < e, e < x ≤ f,
and concave downward for
a ≤ x ≤ b, c ≤ x ≤ d, f ≤ x < g, g < x ≤ h.
The points x = e and x = g are not included in these intervals because f " (x) is not defined at
these points. Points on the curve corresponding to x = b , c , d , and f are points of inflection;
those at x = e and g are not. They do not separate intervals of opposite concavity.
The tangent line has a peculiar property at a point of inflection. To see it, draw the tangent
line at the four points of inflection in Figure 4.28. The tangent line actually crosses from one
side of the curve to the other (something many readers think a tangent line should not do).
FIGURE 4.28 Intervals

on which graph is concave upward
and concave downward
y
a b c d e f g h x
FIGURE 4.29 Horizontal The critical points x = a and x = c in Figure 4.19, where f " (x) = 0, and the critical
point of inflection points x = e and x = g , where f " (x) does not exist, do not yield relative extrema for f (x) .
y Critical points of these types are illustrated again in Figures 4.29 and 4.30. The tangent line at
y = f (x) x = a is horizontal in Figure 4.29 [ f " (a) = 0]. Since the graph is concave downward to the
(a, f(a)) left of x = a and concave upward to the right, the point (a, f (a)) is a point of inflection. We
call it a horizontal point of inflection. The point (a, f (a)) in Figure 4.30 is also a point of
inflection. As the tangent line at this point is vertical, we call it a vertical point of inflection.
In Section 4.2 we stated that a function is increasing on an interval if its first derivative
a x is greater than or equal to zero on that interval, and equal to zero at only a finite number of
FIGURE 4.30 Vertical
points; it is decreasing if its first derivative is less than or equal to zero. Furthermore, points
point of inflection that separate intervals on which the function is increasing and decreasing are points at which
the first derivative is either equal to zero (and changes sign) or does not exist. We can use these
y ideas to locate analogously points of inflection and intervals on which the graph of a function is
y = f(x) concave upward and concave downward.
(a, f(a))
Test for Concavity

a x (i) The graph of a function f (x) is concave upward on an interval I if on I ,
f "" (x) ≥ 0 (4.7a)

(ii) The graph of a function f (x) is concave downward on an interval I if on I ,
f "" (x) ≤ 0 (4.7b)

(iii) Points of inflection occur where
f "" (x) = 0 (4.7c)
and f "" (x) changes sign, or perhaps also where f "" (x) does not exist.
Consequently, just as the first derivative of a function is used to determine its relative extrema
and intervals on which it is increasing and decreasing, the second derivative is used to find
points of inflection and intervals on which the graph of the function is concave upward and
concave downward. This parallelism between the first and second derivatives is reiterated in the
following table.
First Derivative Second Derivative
1. A relative extremum occurs at a point 1. A point of inflection occurs at a point
where f " (x) = 0 and f " (x) changes sign where f "" (x) = 0 and f "" (x) changes sign
as x increases through the point. as x increases through the point.
2. A relative extremum may also occur at 2. A point of inflection may also occur at
a point where f " (x) does not exist. a point where f "" (x) does not exist.
3. f (x) is increasing on an interval if 3. f (x) is concave upward on an interval
f " (x) ≥ 0; f (x) is decreasing on an if f "" (x) ≥ 0; f (x) is concave downward
interval if f " (x) ≤ 0. f " (x) may be equal on an interval if f "" (x) ≤ 0. f "" (x) may
to zero at only a finite number of points in be equal to zero at only a finite number of
either case. points in either case.
These discussions lead to what is often called the second-derivative test for determining
whether a critical point at which f " (x) = 0 yields a relative maximum or a relative minimum.
Second-Derivative Test for Relative Extrema

Suppose f "" (x) is continuous on an open interval containing a critical point x0 of a function
f (x) at which f " (x0 ) = 0. Then:
(i) If f "" (x0 ) > 0, then x = x0 yields a relative minimum for f (x) .
(ii) If f "" (x0 ) < 0, then x = x0 yields a relative maximum for f (x) .
(iii) If f "" (x0 ) = 0, then no conclusion can be made.
In (i), f " (x0 ) = 0 implies that x = x0 is a critical point with a horizontal tangent line;
f "" (x0 ) > 0 implies that around x = x0 the curve is concave upward, and x = x0 therefore
yields a relative minimum. Similarly, f "" (x0 ) < 0 at a critical point in (ii) implies that the
curve is concave downward and therefore has a relative maximum.
To show that the test fails in case (iii), consider three functions f (x) = x 4 , f (x) = −x 4 ,
and f (x) = x 3 in Figures 4.31, 4.32, and 4.33 respectively. For each, f " (0) = f "" (0) = 0;
yet the first has a relative minimum, the second a relative maximum, and the third a horizontal
point of inflection at x = 0.
FIGURE 4.31 A relative FIGURE 4.32 A relative FIGURE 4.33 Horizontal

minimum at x = 0 maximum at x = 0 point of inflection at x = 0
y y y
y = x4
x y = x3
x y = −x 4 x
The second-derivative test very quickly determines the nature of a critical point x0 when
f " (x0 ) = 0 and f "" (x0 ) $= 0. For instance, the second derivative of the function f (x) =
2x 3# − 9x√2
− 23x $+ 6 in Example 4.6(a) of Section √ 4.3 is f "" (x) = 12x − 18. Since
""
f (9 − 219)/6 < 0, the critical point x = (9 − 219)/6 yields a relative maximum,
# √ $ √
and because f "" (9 + 219)/6 > 0, x = (9 + 219)/6 gives a relative minimum.
A more refined test to determine the nature of a critical point x0 when f " (x0 ) = f "" (x0 ) = 0
is discussed in Exercise 35.
EXAMPLE 4.9
For each of the following functions, find all points of inflection and intervals on which the graph
of the function is concave upward and concave downward.
2
(a) f (x) = x 3 − 6x 2 + 16x + 1 (b) f (x) = x 6/5 + x 1/5
3
SOLUTION
(a) The graph of the function in Figure 4.34 indicates that there is one point of inflection.
To find it, we solve
d
0 = f "" (x) = (2x 2 − 12x + 16) = 4x − 12,
dx
and obtain x = 3. Since f "" (x) changes sign as x passes through 3, x = 3 gives the
point of inflection (3, 13) . The graph y = f (x) is concave downward for x ≤ 3
[since f "" (x) ≤ 0] and concave upward for x ≥ 3 [ f "" (x) ≥ 0]. Since f "" (x) is
defined for all x , there can be no other points of inflection.
FIGURE 4.34 Intervals on which y = f (x) is concave upward and downward
y
18 2 3
y= x − 6x 2 + 16x + 1
3
16
14
12
10
1 2 3 4 5 6 x

Intervals on which y = f (x) is concave upward and downward
y y
y = x 6/5 + x1/5 y= x 6/5 + x1/5 1.5

6
1.0
4 0.5
2 −1 −0.5 0.5 1 x
−0.5
−4 −2 2 4 x
(b) The plot of the function in Figure 4.35a suggests a point of inflection between x = 0
and x = 1; the curve seems to be concave downward for small positive x and concave
upward for larger x . The graph on the smaller interval in Figure 4.35b does not make
this any clearer, but it suggests that (0, 0) is a point of inflection. To confirm these
suspicions, we first solve
( )
d 6 1/5 1 − 4 /5 6 4 6x − 4
0 = f (x) = ""
x + x = x − 4 /5 − x −9/5 =
dx 5 5 25 25 25x 9/5
and find x = 2/3. Since f "" (x) changes sign as x passes through 2/3, x = 2/3
gives a point of inflection with f (2/3) ≈ 1.5.
We also note that f " (x) and f "" (x) do not exist at x = 0, although f (0) does ( = 0). Since
f "" (x) changes sign as x passes through 0, (0, 0) must also be a point of inflection. The fact
that ( )
6 1/5 1 6x + 1
lim f " (x) = lim x + x −4/5 = lim =∞
x→0 x→0 5 5 x→0 5x 4/5
indicates that (0, 0) is a vertical point of inflection. With the points of inflection now in place,
we can say that the graph is concave upward for x < 0 and x ≥ 2/3, and concave downward
for 0 < x ≤ 2/3.
EXAMPLE 4.10
The Van der Waals equation for an ideal gas relates pressure P and specific volume V by
RT a
P = − 2,
V −b V
where T is absolute temperature of the gas, R is the universal gas constant, and a and b are
positive constants that depend on the gas. Suppose pressure and volume measurements are taken
for a specific gas at various temperatures, and the results are plotted as in Figure 4.36. Each
curve corresponds to a fixed value of T . One of the curves has a horizontal point of inflection.
FIGURE 4.36 Plots of
It can be used to determine values of a and b for this gas. Suppose the temperature of the
Van der Waals equation for various
temperatures
gas for this curve is Tc , and Vc is the volume, which gives the horizontal point of inflection.
Find expressions for a and b in terms of Tc and Pc = P (Vc ) , called critical temperature and
P pressure.
dP RT 2a d 2P 2RT 6a
SOLUTION Since =− + , and = − , a horizontal
Tc dV (V − b)2 V 3 dV 2 (V − b)3 V 4
Pc
point of inflection occurs for T = Tc and V = Vc if
b Vc V
RTc 2a 2RTc 6a
0 = − + 3, 0 = − 4.
(Vc − b)2 Vc (Vc − b)3 Vc
To find a and b , we solve each equation for a and equate results:
RTc Vc3 RTc Vc4 Vc

= -⇒ 3(Vc − b) = 2Vc -⇒ b = .
2(Vc − b)2 3(Vc − b)3 3
Substitution of this into the first equation gives
2a RTc 9RTc 9RTc Vc

= = -⇒ a = .
V3c (Vc − Vc /3) 2 4Vc2 8
We now have a and b in terms of Tc and Vc . To replace Vc with Pc , we use Van der Waals
equation to write
RTc a
Pc = − 2.
Vc − b Vc
When we substitute the expressions for a and b into this equation we obtain
RTc 9RTc 3RTc 3RTc

Pc = − = -⇒ Vc = .
V c − Vc /3 8Vc 8Vc 8Pc
( )
9RTc 3RTc 27R 2 Tc2 RTc
This then gives a = = and b = .
8 8Pc 64Pc 8Pc
EXERCISES 4.4
In Exercises 1–14 determine where the graph of the function is concave ∗ 27. f "" (x) ≥ 0 on I and f (x) is not concave upward on I .
upward, is concave downward, and has points of inflection.
∗ 28. f (x) is decreasing on I , f (x) is continuous at x0 , but f " (x) is
1. f (x) = x 3 − 3x 2 − 3x + 5 discontinuous at x0 .
2. f (x) = 3x 4 + 4x 3 − 24x + 2 ∗ 29. f (x) is increasing on a ≤ x ≤ x0 , increasing on x0 < x ≤ b ,

but not increasing on I .
1
∗ 3. f (x) = x 2 +
x2 ∗ 30. f (x) is concave downward on a ≤ x < x0 , concave downward
2 on x0 < x ≤ b , but not concave downward on I .
x +4
∗ 4. f (x) =
x2 − 1 ∗ 31. f (x) is increasing on a ≤ x ≤ x0 , concave downward on a ≤
x < x0 , decreasing on x0 ≤ x ≤ b , and concave upward on x0 <
∗ 5. f (x) = x + cos x, |x| < 2π x ≤ b.
∗ 6. f (x) = x 2 − sin x
∗ 32. If f (x) and g(x) are twice differentiable and concave upward on
∗ 7. f (x) = x 2 − 2 sin x an interval I , is f (x)g(x) concave upward on I ?
∗ 8. f (x) = x 2 − 4 sin x, |x| < 2π ∗ 33. If the functions in Exercise 32 are also increasing on I , is
f (x)g(x) concave upward on I ?
∗ 9. f (x) = x ln x ∗ 10. f (x) = x 2 ln x
∗ 34. Repeat Example 4.10 for Dieterici’s equation in Exercise 66 of
∗ 11. f (x) = e1/x ∗ 12. f (x) = xe−2x Section 4.3
∗ 13. f (x) = x 2 e3x ∗ 14. f (x) = x 2 − e−x ∗ 35. The second-derivative test fails to classify a critical point x0 of a
function f (x) as yielding a relative maximum, a relative minimum,
or a horizontal point of inflection if f " (x0 ) = f "" (x0 ) = 0. The
In Exercises 15–22 use the second-derivative test to determine whether following test can be used in such cases. Suppose f (x) has derivatives
critical points where f " (x) = 0 yield relative maxima or relative of all orders in an open interval around x0 , and the first n derivatives
minima. all vanish at x0 , but the (n + 1)th derivative at x0 is not zero. If we
denote the nth derivative of f (x) at x0 by f (n) (x0 ) , these conditions
15. f (x) = x 3 − 3x 2 − 3x + 5 are
1
16. f (x) = x + 0 = f " (x0 ) = f "" (x0 ) = · · · = f (n) (x0 ), f (n+1) (x0 ) $= 0.
x
17. f (x) = 3x 4 − 16x 3 + 18x 2 + 2 Then:
(i) If n is even, f (x) has a horizontal point of inflection at x0 .
18. f (x) = x 5/4 − x 1/4
(ii) If n is odd and f (n+1) (x0 ) > 0, f (x) has a relative mini-
mum at x0 .
∗ 19. f (x) = x ln x ∗ 20. f (x) = x 2 ln x
(iii) If n is odd and f (n+1) (x0 ) < 0, f (x) has a relative max-
2x 2 −2 x imum at x0 .
∗ 21. f (x) = xe ∗ 22. f (x) = x e
A proof of this result requires the use of material from Chapter 10 and is
∗ 23. Is the graph of a function concave upward, concave downward, therefore delayed until that time (see Exercise 16 in Section 10.3). Note
both, or neither on an interval I if on I its second derivative is always that the second-derivative test is the special case when n = 1. Use this
equal to zero? test to determine whether critical points of the following functions yield
relative maxima, relative minima, or horizontal points of inflection:
∗ 24. Prove that the curve y = 2 cos x passes through all points of √
(a) f (x) = (x 2 − 1)3 ; (b) f (x) = x 2 1 − x .
inflection of the curve y = x sin x .
∗ 25. Show that every cubic polynomial has exactly one point of inflec- ∗ 36. Show that the points of inflection of f (x) = (k − x)/(x 2 + k 2 ) ,
tion on its graph. where k is a constant, all lie on a straight line.
∗∗ 37. Prove that if a cubic polynomial has both a relative maximum and
Draw the graph of a function f (x) that is defined everywhere on the a relative minimum, then the point of inflection between these extrema
interval I : a ≤ x ≤ b , and that possesses the properties in Exercises is the midpoint of the line segment joining them.
26–31. In each case assume that a < x0 < b .
∗∗ 38. Show that the graph of the function f (x) = sin (x − sin x) has
∗ 26. f (x) is increasing on I and discontinuous at x0 . an infinite number of horizontal points of inflection.
∗∗ 39. Show that the equation f (x) = x n+1 − bn x + abn = 0, where (b) Show that f "" (0) does not exist. Hint: See Exercise 47 in
a > 0, b > 0, and n ≥ 1 are constants, has exactly two distinct Section 3.9.
nb (c) Is the point (0, 0) a relative maximum, a relative minimum,
positive solutions if and only if a < .
(n + 1)(n+1)/n a horizontal point of inflection, or none of these?
∗∗ 40. (a) In Example 3.12 of Section 3.3, the derivative of the func- ∗∗ 41. Show analytically that the equation
tion
 ( )
 1 2 sin θ = θ(1 + cos θ)
x 2 sin , x $= 0
f (x) = x
 has no solution in the interval 0 < θ < π .
0, x =0
was shown to be zero at x = 0 [i.e., f " (0) = 0]. It follows
that the graph of the function has a horizontal tangent line
at (0, 0) . Draw the graph and the tangent line.
4.5 Drawing Graphs with Calculus

In Sections 4.3 and 4.4, we used plots of functions to help you see the geometric interpretations
of critical points, relative extrema, concavity, and points of inflection. In this section we use
critical points, relative extrema, increasing and decreasing functions, concavity, and points of
inflection to draw (or sketch) graphs of functions. We assume no access to graphing calculators
or computers for the first four examples.
EXAMPLE 4.11
Use calculus to draw a graph of the function f (x) = (x − 4)/x 2 .
SOLUTION With an x -intercept equal to 4, and the limits
lim f (x) = −∞, lim f (x) = −∞, lim f (x) = 0− , lim f (x) = 0+ ,
x→0− x→0+ x→−∞ x→∞
we begin our sketch as shown in Figure 4.37a. This information would lead us to suspect that
the graph should be completed as shown in Figure 4.37b. To verify this we find critical points
for f (x) :
x 2 (1) − (x − 4)(2x) x(8 − x) 8−x

0 = f " (x) = 4
= 4
= .
x x x3
Clearly, x = 8 is the only critical point of the function, and since the first derivative changes
from a positive quantity to a negative quantity as x increases through x = 8, there is a relative
maximum of f (8) = 1/16. Our sketch is now as shown in Figure 4.37c.
Figure 4.37c makes it clear that there is a point of inflection to the right of x = 8, which
we could pinpoint with f "" (x) :
x 3 (−1) − (8 − x)(3x 2 ) 2(x − 12)

f "" (x) = 6
= .
x x4
Since f "" (12) = 0 and f "" (x) changes sign as x passes through 12, (12, 1/18) is the point of
inflection. Figure 4.37d contains the final graph.
FIGURE 4.37a Steps in drawing graph of FIGURE 4.37b

x−4
f (x) = y
x2
y 4 x
4 x
y y
1/16 1/16
4 8 x 4 8 12 x
EXAMPLE 4.12
Use calculus to draw a graph of the function f (x) = x 5/3 − x 2/3 .
SOLUTION We first note that the three points (0, 0) , (1, 0) , and (−1, −2) are on the graph.
Next we add the facts that
lim f (x) = −∞ and lim f (x) = ∞,

x→−∞ x→∞
as shown in Figure 4.38a. We cannot be sure that the concavity is as indicated, but this will be
verified shortly. To find critical points of f (x) , we first solve
5 2 5x − 2
0 = f " (x) = x 2/3 − x −1/3 = .
3 3 3x 1/3
The only solution is x = 2/5, and because f " (x) changes from negative to positive as x
increases through 2/5, we have a relative minimum of f (2/5) = (2/5)5/3 − (2/5)2/3 ≈
−0.33. Since f " (x) does not exist at x = 0, this is also a critical point, and we calculate that
lim f " (x) = ∞ and lim f " (x) = −∞.
x→0− x→0+
This information, along with the relative minimum, is shown in Figure 4.38b. We now join
these parts smoothly to produce the sketch in Figure 4.38c. To verify that the concavity is as
indicated we calculate
( )( ) ( )( )
5 2 −1/3 2 1 2(5x + 1)
""
f (x) = x − − x − 4 /3 = .
3 3 3 3 9x 4/3
Since f "" (x) ≤ 0 for x ≤ −1/5, the graph is concave downward on this interval. It is concave
upward for −1/5 ≤ x < 0 and x > 0 since f "" (x) ≥ 0 on these intervals. A point of
inflection occurs at x = −1/5 where f "" (x) = 0 and changes sign, but not at x = 0 where
f "" (x) does not exist. The final sketch is in Figure 4.38d.

f (x) = x 5/3 − x 2/3
y
y
1
1 x
(0.4, −0.33)
x
(−1, −2)
(−1, −2)
y y
1 1
x x
(0.4, −0.33) (−0.2, −0.41)
y = x5/3 − x2/3
(−1, −2) (−1, −2)
EXAMPLE 4.13
Use calculus to draw a graph of the function f (x) = x + sin x .
SOLUTION We can add ordinates of the graphs of y = x and y = sin x in Figure 4.39a.
This creates oscillations around y = x . They might look like Figure 4.39b, c, or d. To decide
which, we find critical points,
0 = f " (x) = 1 + cos x.
The solutions of this equation are x = (2n + 1)π , where n is an integer. Since f " (x) does
not change sign as x passes through these points, they cannot yield relative maxima or minima.
They must give horizontal points of inflection. The correct graph is Figure 4.39c.
FIGURE 4.39a Possibilities for graph of FIGURE 4.39b

f (x) = x + sin x y
y
y=x
−2π 2π
−π π 3π −2π −π π 2π 3π x
x
y = sin x
y y
−2π −π π 2π 3π x −2π −π π 2π 3π x
EXAMPLE 4.14
Use calculus to draw a graph of the function f (x) = (x 2 + 1)/(x − 1) .
SOLUTION The function is discontinuous at x = 1. We calculate that
x2 + 1 x2 + 1
lim =∞ and lim = −∞.
x→1+ x−1 x→1− x−1
2
If we divide x + 1 by x − 1, we write f (x) in the form
2
f (x) = x + 1 + .
x−1
This shows that y = x + 1 is an oblique asymptote for the graph (Figure 4.40a). Clearly we
should find critical points by solving
2 (x − 1)2 − 2
0 = f " (x) = 1 − = .
(x − 1)2 (x − 1)2
√
This implies that (x − 1)2 = 2, from which x = 1 ± 2. Because√f " (x) changes from
a positive quantity to a negative
√ quantity
√ as x increases
√ through 1 − 2, there is a relative
maximum √ at x = 1 −
√ 2 of f ( 1 − 2 ) =
√ 2 − 2 2. Similarly, a relative minimum of
f (1 + 2) = 2 + 2 2 occurs at x = 1 + 2. The final graph is shown in Figure 4.40b.

x2 + 1
f (x) = y
x−1
(1 + √2, 2 + 2√2)
y
1
−1
−1 1 x
1
−1 (1 − √2, 2 − 2√2)
−1 1 x
Electronic devices cannot plot curves that contain unspecified parameters. For example, we can
2
use a graphing calculator or computer to plot y = e−ax for any given value of a , but we cannot
plot the curve without specifying a value for a . What is appropriate is to conjecture the shape
2
of y = e−ax from plots with various values of a , and verify the conjecture with calculus. Let
us illustrate with two examples.
EXAMPLE 4.15
When two substances A and B are brought together at time t = 0, they react to form substance
C in such a way that 1 g of A reacts with 1 g of B to form 2 g of C. If initial amounts of A and
B are A0 and B0 (where A0 > B0 ), then the number of grams of C at any time t is given by
2A0 B0 (1 − e−kt )
C(t) = ,
A0 − B0 e−kt
where k > 0 is a constant. Draw a graph of C(t) .
SOLUTION Let us get an idea of the shape of the graph by plotting the function with values
A0 = 20, B0 = 15, and k = 0.1. After experimenting with various domains, we arrive at
Figure 4.41 as the most informative.
600(1 − e−0.1t ) 2A0 B0 (1 − e−kt )

FIGURE 4.41 Plot of C(t) = FIGURE 4.42 Graph of C(t) =
20 − 15e−0.1t A0 − B0 e−kt
C C
30 2B0
20
10
5 10 15 20 t t
The graph has no relative extrema, no points of inflection, and is asymptotic to the line
C = 30. We now determine whether these features remain true for all values of A0 > B0 and
k . Critical points of C(t) are given by
! "
(A0 − B0 e−kt )(ke−kt ) − (1 − e−kt )(B0 ke−kt )
0 = C " (t) = 2A0 B0
(A0 − B0 e−kt )2
2A0 B0 (A0 − B0 )ke−kt
= .
(A0 − B0 e−kt )2
There are no solutions and C " (t) always exists, so that C(t) has no relative maxima and minima.
For points of inflection, we consider
! "
"" (A0 − B0 e−kt )2 (−ke−kt ) − e−kt (2)(A0 − B0 e−kt )(B0 ke−kt )
0 = C (t) = 2A0 B0 (A0 − B0 )k
(A0 − B0 e−kt )4
2A0 B0 (A0 − B0 )k 2 e−kt (B0 + A0 e−kt )
= .
(A0 − B0 e−kt )3
Once again there are no solutions, and C "" (t) always exists. Hence the graph of C(t) has no
points of inflection.
The graph begins at (0, 0) with slope 2A0 B0 k/(A0 − B0 ) , and is asymptotic to the line
2A0 B0 (1 − e−kt )
C = lim = 2 B0 .
t→∞ A0 − B0 e−kt
The graph is shown in Figure 4.42.
EXAMPLE 4.16
When a drug is injected into the blood at time t = 0, it is sometimes assumed in biomedical
engineering that the concentration of the drug in a nearby organ is given by a function of the
form
f (t) = k(e−at − e−bt ),
where k , a , and b are positive constants with b > a . Draw a graph of this function indicating
any relative extrema and points of inflection.
FIGURE 4.43 Plot of f (t) = e−t − e−2t
f(t)
0.2
2 4 t
SOLUTION To get an idea of the shape of the graph we plot the function with values a = 1,
b = 2, and k = 1 (Figure 4.43). To determine whether this shape remains constant for all
values of a and b ( k has no effect on shape of the graph), we begin by finding critical points of
the function,
0 = f " (t) = k(−ae−at + be−bt ).
Thus,
b
ae−at = be−bt -⇒ e(b−a)t = ,
a
the solution of which is t = (b − a)−1 ln (b/a) . To test this for maximum or minimum, and
find points of inflection, we use
f "" (t) = k(a 2 e−at − b2 e−bt ).

It is not obvious whether
4 a b
5
f "" [(b − a)−1 ln (b/a)] = k a 2 e− b−a ln (b/a) − b2 e− b−a ln (b/a)
is positive or negative. If we note, however, that at the critical point, ae−at = be−bt , and
substitute this into f "" (t) , we can see that
f "" [(b − a)−1 ln (b/a)] = k [a 2 e−at − b(ae−at )] = kae−at (a − b).

This is clearly negative (since b > a ), and therefore f (t) has a relative maximum at t =
(b − a)−1 ln (b/a) . For points of inflection we set
( )
2 −at 2 −bt 2 b
0 = a e −b e -⇒ t = ln .
b−a a
To confirm that this value of t gives a point of inflection we should check that f "" (t) changes
sign as t passes through this value. This is not obvious. We can, however, conclude that a point
of inflection must occur here if we examine the information in Figure 4.44a. The graph begins
at the origin and achieves a relative maximum at t = (b − a)−1 ln (b/a) where the graph
is concave downward. It is also asymptotic to the positive t -axis since limt→∞ f (t) = 0+ .
Hence, there must be a change in concavity to the right of the maximum, and this can only occur
at t = 2(b − a)−1 ln (b/a) . The final graph is shown in Figure 4.44b.
FIGURE 4.44a Partial graph of k(e−at − e−bt ) FIGURE 4.44b Full graph of k(e−at − e−bt )
a y a y
k (b − a) a b−a k (b − a) a b−a
b b b b
2a
k (b 2 − a 2 ) a b−a
b b
t t
1 b 1 b 2 b
ln ln ln
b−a a b−a a b−a a
EXERCISES 4.5
In Exercises 1–22 find all relative maxima and minima for the function In Exercises 23–34 find all relative maxima and minima for the function.
and points of inflection on its graph. Use this information, and whatever Use this information, and whatever else is appropriate, to draw a graph
else is appropriate, to draw a graph of the function. of the function.
∗ 23. f (x) = x/(x 2 + 3)

x3 x2
1. f (x) = − − 2x ∗ 24. f (x) = x 4 + 10x 3 + 6x 2 − 64x + 5
3 2
x2 − 8
∗ 25. f (x) = x 4 − 2x 3 + 2x ∗ 26. f (x) =
3
2. f (x) = x − 6x + 12x + 9 2 x−5
x2
4 3 2 ∗ 27. f (x) = ∗ 28. f (x) = x 5/4 − x 1/4
3. f (x) = 3x − 16x + 18x + 2 x2 −4
(2x − 1)(x − 8)
4. f (x) = 2x 3 − 15x 2 + 6x + 4 ∗ 29. f (x) = |x 2 − 9| + 2 ∗ 30. f (x) =
(x − 1)(x − 4)
x ∗ 31. f (x) = sin2 x cos x, 0 ≤ x ≤ 2π
∗ 5. f (x) = x − 3x 1/3 ∗ 6. f (x) =
x2 +4 ∗ 32. f (x) = sin x + cos x
x
∗ 7. f (x) = (x − 2)3 (x + 2) ∗ 8. f (x) = x 2/3 (8 − x) ∗ 33. f (x) = √
1−x
( )
(x + 2)2 x+8 / 2
∗ 9. f (x) = ∗ 10. f (x) = 2x 3/2 − 9x + 12x 1/2 ∗ 34. f (x) = x + 100
x3 x
∗ 35. An emf device producing constant voltage V and with constant in-
x 3 + 16 x2 + x + 1 ternal resistance r maintains current i through a circuit with resistance
∗ 11. f (x) = ∗ 12. f (x) =
x x R in the figure below, where
1 x3 V = i(r + R).
∗ 13. f (x) = x + ∗ 14. f (x) =
x x2 −4
The power (work per unit time) necessary to maintain this current in r
2x 2 x2 + 1 and R is given by
∗ 15. f (x) = ∗ 16. f (x) =
x 2 − 8x + 12 x2 − 1
P = i 2 (r + R) = i 2 r + i 2 R.
3
(x − 1)
∗ 17. f (x) = ∗ 18. f (x) = (x − 1)2/3 (a) If we define PR = i 2 R and Pr = i 2 r as the power dissi-
(x + 1)4
pated in R and r , respectively, draw PR and Pr as functions
of R .
∗ 19. f (x) = (x + 2)3 (x − 4)3
(b) Draw a graph of P (R) = PR (R) + Pr (R) .
∗ 20. f (x) = x + 2 sin x
r
E
−1 −1
∗ 21. f (x) = Sin x + Cos x
/ R
∗ 22. f (x) = x Sin −1 x + 1 − x2
∗ 36. The rate of photosynthesis P in a leaf depends on the intensity of chemical reactor reaches a stable steady-state condition, a function that
light I on the leaf according to is encountered is
MI 1
P = − R, f (x) = , x > 0,
I +K b + ea/x
where M > R and K are all positive constants. Draw a graph of this where a and b are positive constants.
function.
(a) To get an idea of what this function looks like, plot a graph
∗ 37. The radial probability density function for the ground state of the when b = 4 and a = 10 on the interval 0 < x ≤ 20.
hydrogen atom is Find its point of inflection.
( )
4r 2 (b) Now show that f (x) = (b + ea/x )−1 , x > 0 has the
P (r) = e−2r/a , r ≥ 0, same shape. Verify that it has no relative extrema, but it has
a2
exactly one point of inflection.
where a > 0 is a constant. Draw a graph of the function identifying ∗ 42. In the kinetic theory of gases, Maxwell’s speed distribution law
its relative maximum and points of inflection. defines the probability P that a molecule of gas moves with speed v as
∗ 38. The radial probability density function for the second state of the
( ) 3/ 2
hydrogen atom is M 2 /(2RT )
( )- P (v) = 4π v 2 e−Mv , v ≥ 0,
2π RT
r2 r .2 −r/a
P (r) = 2− e , r ≥ 0,
8a 3 a where T , the temperature of the gas (in kelvin), M , the molar mass of
the gas, and R , the gas constant, are all constants. We shall have more
where a > 0 is a constant. Draw a graph of the function identifying
to say about this function when we know how to integrate. For now,
its relative extrema.
we simply wish to graph the function.
∗ 39. The Weibull distribution for the probability that a certain wind
speed v occurs is (a) Plot P (v) for oxygen at 50 K and 300 K using M = 0.0320
and R = 8.31. Do they have the same shape? The fact
k - v .k−1 −(v/a)k that most of the 300 K curve is higher than the 50 K curve
P (v) = e , means that oxygen molecules are more likely to move faster
a a
at higher temperatures than at lower temperatures.
where a > 1 and 1 < k < 2 are constants. Draw a graph of this
(b) For any gas at any temperature, find the value of v that gives
function identifying its relative extrema.
the relative maximum and the values of v that give points
∗ 40. In quantum mechanics, the probability that an energy level E will of inflection.
be occupied is ∗ 43. The function
1
P (E) = (E−Ef )/(kT )
,
e +1 1 2 /(2σ 2 )
f (x) = √ e−(x−µ) ,
where Ef is the Fermi energy, k is the Boltzmann constant, and T is 2π σ
temperature. Draw a graph of the function identifying relative extrema
and points of inflection. where σ > 0 and µ are constants is called the normal probability
density function. Draw its graph.
∗ 41. In Example 3.35 in Section 3.11 we discussed chemical formation
in a chemical reactor. What we ignored there was the fact that temper- ∗ 44. Draw a graph of the function
ature usually varies in the reactor, and the rate at which the chemical
is formed or broken down depends on temperature. In other words, 1 + x + x2 + x3
f (x) = .
the situation is much more complicated than we presented. When a 1 + x3
4.6 Analyzing Graphs with Calculus

In Section 4.5 we drew graphs of functions without the use of calculators or computers. In
this section we take a different approach; we use technology to plot the graph of a function
and then use calculus to analyze what we see or do not see on the plot. Remember that we
have illustrated on several occasions that electronic output can sometimes be inconclusive and
even misleading (see for instance Example 1.19). Electronic devices can never replace sound
mathematical analysis. Whenever there is a question about electronic output, we use calculus
to find the answer.
EXAMPLE 4.17
Plot a graph of the function f (x) = (x − 4)/x 2 , and then use calculus to pinpoint significant
information concerning the graph.
SOLUTION A plot of the function on the interval −10 ≤ x ≤ 10 in Figure 4.45a does not
show a lot. It certainly indicates that the y -axis is a vertical asymptote, and this is confirmed
by the fact that limx→0 f (x) = −∞ . The plot also suggests that the x -axis is a horizontal
asymptote, confirmed by
lim f (x) = 0− , lim f (x) = 0+ .

x→−∞ x→∞
The fact that the graph approaches y = 0 from above as x → ∞ means that the graph must
cross the x -axis at least once. It does so exactly once since f (x) = 0 only when x = 4. To
get a better visualization of the graph for x > 0, we now plot it on the interval 1 ≤ x ≤ 10, at
the same time restricting the y -values to −1 ≤ y ≤ 1 (Figure 4.45b). There must be a relative
maximum to the right of x = 4 (perhaps more than one). To find it (or them), we solve
x 2 (1) − (x − 4)(2x) 8−x

0 = f " (x) = = .
x4 x3
FIGURE 4.45a f (x) = (x − 4)/x 2 on −10 ≤ x ≤ 10 FIGURE 4.45b f (x) = (x − 4)/x 2 on 1 ≤ x ≤ 10
y 1.00 y
−10 −5 5 10 x 0.75
−20 0.50
0.25
−40
−60 2 4 6 8 10 x
−0.25
−80 −0.50
−0.75
−100
−1.00
FIGURE 4.45c f (x) = (x − 4)/x 2 on −20 ≤ x ≤ 20

y
0.2
−20 −10 10 20 x
−0.2
−0.4
−0.6
−0.8
−1.0
The only solution is x = 8. This yields a relative maximum. With the graph concave downward
at x = 8, and concave upward for large x (recall that y = 0 is a horizontal asymptote), there
is a point of inflection to the right of x = 8. To find it we solve
x 3 (−1) − (8 − x)(3x 2 ) 2(x − 12)

0 = f "" (x) = 6
= .
x x4
Since f "" (12) = 0 and f "" (x) changes sign as x passes through 12, there is a point of inflection
at (12, 1/18) . By restricting y -values even more and plotting on the interval −20 ≤ x ≤ 20,
we obtain the plot in Figure 4.45c; it shows all of the foregoing information.
EXAMPLE 4.18
Repeat Example 4.17 for the function f (x) = x 5/3 − x 2/3 .
SOLUTION The software package in our computer does not automatically plot a graph of the
function f (x) = x 5/3 − x 2/3 for negative values of x . (It interprets x 5/3 and x 2/3 for negative
x as complex numbers.) Your calculator or computer may do the same. To rectify this, define
f (x) = −(−x)5/3 − (−x)2/3 for x < 0. A plot then looks like that in Figure 4.46 on the
interval −1 ≤ x ≤ 2.
A relative minimum is indicated just to the left of x = 1/2. To locate it, we solve
5 2 5x − 2
0 = f " (x) = x 2/3 − x −1/3 = .
3 3 3x 1/3
The only solution is x = 2/5, and because f " (x) changes from negative to positive as x
increases through 2/5, we do indeed have a relative minimum of f (2/5) = (2/5)5/3 −
(2/5)2/3 = −0.33. The derivative f " (x) does not exist at x = 0, but f (0) = 0. We calculate
that
5x − 2
lim f " (x) = lim =∞ and lim f " (x) = −∞.
x→0− x→0− 3x 1/3 x→0+
This means that there is a very sharp point at (0, 0) where the function has a relative maximum.
Its tangent line is vertical at this point. It is difficult to get a sense of the concavity of the graph
for x < 0, it appears so straight. To assess this we consider
( )( ) ( )( )
5 2 −1/3 2 1 2(5x + 1)
""
f (x) = x − − x − 4 /3 = .
3 3 3 3 9x 4/3
Since f "" (x) changes sign as x passes through −1/5, we have a point of inflection (−1/5, −0.41) .
The graph is concave downward for x ≤ −1/5 and concave upward for −1/5 ≤ x < 0. It is
also concave upward for 0 < x < ∞ , so that (0, 0) is not a point of inflection.
FIGURE 4.46 Plot of f (x) = x 5/3 − x 2/3
y
0.4
0.2 y = x 5/3 − x 2/3
−1 −0.5 0.5 1 1.5 2x

−0.2
−0.4
−0.6
EXAMPLE 4.19
Repeat Example 4.17 for the function f (x) = x + sin x .
SOLUTION A plot of f (x) = x + sin x is shown in Figure 4.47. It is not clear whether it
has relative extrema. Critical points will provide the answer. We solve 0 = f " (x) = 1 + cos x
for x = (2n + 1)π , where n is an integer. Because f " (x) does not change sign as x passes
through these points, they cannot yield relative maxima or minima. They must give horizontal
points of inflection.
FIGURE 4.47 Plot of f (x) = x + sin x
7.5
5
y = x + sin x
2.5
−10 −5 5 10 x
−2.5
−5
−7.5
EXERCISES 4.6
In Exercises 1–20 plot a graph of the function. Use derivatives to (x − 1)3
∗ 17. f (x) = ∗ 18. f (x) = (x − 1)2/3
discuss any significant features of the plot. (x + 1)4
x3 x2 ∗ 19. f (x) = (x + 2)3 (x − 4)3

1. f (x) = − − 2x
3 2
∗ 20. f (x) = x + 2 sin x
2. f (x) = x 3 − 6x 2 + 12x + 9
3. f (x) = 3x 4 − 16x 3 + 18x 2 + 2 In Exercises 21–32 plot a graph of the function. Identify all relative
extrema.
4. f (x) = 2x 3 − 15x 2 + 6x + 4
∗ 21. f (x) = x/(x 2 + 3)
∗ 5. f (x) = x − 3x 1/3
∗ 22. f (x) = x 4 + 10x 3 + 6x 2 − 64x + 5
x
∗ 6. f (x) = ∗ 7. f (x) = (x − 2)3 (x + 2) ∗ 23. f (x) = x 4 − 2x 3 + 2x
x2 +4
(x + 2)2 x2 − 8
∗ 8. f (x) = x 2/3 (8 − x) ∗ 9. f (x) = ∗ 24. f (x) =
x3 x−5
∗ 10. f (x) = 2x 3/2 − 9x + 12x 1/2 x2

∗ 25. f (x) = ∗ 26. f (x) = x 5/4 − x 1/4
x2 −4
x 3 + 16 x2 + x + 1
∗ 11. f (x) = ∗ 12. f (x) = (2x − 1)(x − 8)
x x ∗ 27. f (x) = |x 2 − 9| + 2 ∗ 28. f (x) =
(x − 1)(x − 4)
1 x3
∗ 13. f (x) = x + ∗ 14. f (x) =
x x2 − 4 ∗ 29. f (x) = sin2 x cos x, 0 ≤ x ≤ 2π
2x 2 x2 + 1 x
∗ 15. f (x) = ∗ 16. f (x) = ∗ 30. f (x) = sin x + cos x ∗ 31. f (x) = √
x2 − 8x + 12 x2 − 1 1−x
( )
x+8 / 2 (b) The average production cost per kilogram when x kilo-
∗ 32. f (x) = x + 100
x grams are produced is given by c(x) = C(x)/x . Plot a
∗ 33. (a) Plot a graph of the function f (x) = x 8 −4x 6 −8x 5 +40x 3 . graph of this function.
(b) Find all critical points and classify them as yielding relative (c) Show that the output at which the average cost is least sat-
maxima, relative minima, or horizontal points of inflection. isfies the equation
∗ 34. (a) A company produces x kilograms of a commodity per day
at a total cost of (x + 300)2 (x 2 − 18 000) − 60 000x 2 = 0.
( )
x2 x + 100 Do this in two ways: (i) by finding critical points for c(x) ;
C(x) = + 60, 1 ≤ x ≤ 200.
(ii) by noting that
300 x + 300 # the average
$ cost is the slope of the line
joining a point x, C(x) to the origin; hence minimum
Plot a graph of the function. Show that it is always concave average cost occurs when the tangent line to the C(x) graph
upward. passes through the origin.
4.7 Absolute Maxima and Minima

Seldom in applications of maxima and minima theory do we hear questions such as: What are
the relative maxima of this quantity, or what are the relative minima of that quantity? More
likely it is: What is the biggest of these, the smallest of those, the best way to do this, the
cheapest way to do that? A different kind of extremum is involved — considering all points in
the domain of the function, what is the largest (or smallest) value of the function? We define
this type of extremum as follows.
DEFINITION 4.6
The absolute maximum (or global maximum) of a function f (x) on an interval I is
f (x0 ) if x0 is in I and if for all x in I ,
f (x) ≤ f (x0 ); (4.8a)
f (x0 ) is said to be the absolute minimum (or global minimum) of f (x) on I if for all
x in I ,
f (x) ≥ f (x0 ). (4.8b)
For the function in Figure 4.48, the absolute maximum of f (x) on the interval a ≤ x ≤ b
is f (c) , and the absolute minimum is f (d) . For the function in Figure 4.49, the absolute
maximum on a ≤ x ≤ b is f (b) , and the absolute minimum is f (a) . For the function in
Figure 4.50, the absolute maximum on a ≤ x ≤ b is f (b) , and the absolute minimum is
f (c) . Note that we speak of absolute maxima and minima (absolute extrema) of a function
only on some specified interval; that is, we do not ask for the absolute maximum or minimum
of a function f (x) without specifying the interval I .
Each of the functions in Figures 4.48–4.50 is continuous on a closed interval a ≤ x ≤ b .
The following theorem asserts that every such function has absolute extrema. For a proof of this
result, the interested reader should consult books on advanced analysis.
FIGURE 4.48 FIGURE 4.49 FIGURE 4.50

Absolute maxima and minima of continuous functions on closed intervals
y y y
a c d b x a b x ac b x
THEOREM 4.2
A function that is continuous on a closed interval must attain an absolute maximum and
an absolute minimum on that interval.
The conditions of this theorem are sufficient to guarantee existence of absolute extrema;
that is, if a function is continuous on a closed interval, then it must have absolute extrema on
FIGURE 4.51 Absolute that interval. However, the conditions are not necessary. If they are not met, the function may
maximum and minimum of a or may not have absolute extrema. For instance, if the function in Figure 4.48 is confined to
discontinuous function the open interval a < x < b , it still attains its absolute extrema at x = c and x = d . On
y the other hand, the function in Figure 4.49 does not have absolute extrema on the open interval
a < x < b . The function in Figure 4.51 is not continuous on the closed interval a ≤ x ≤ d ;
it has no absolute maximum on this interval, but it does have absolute minimum f (a) . This
function is not continuous on d ≤ x ≤ b , but it has absolute maximum f (d) and absolute
minimum f (b) .
a c d b x Absolute extrema of the functions in Figures 4.48–4.50 always occur either at a critical
point of the function or at an end of the interval a ≤ x ≤ b . This result is always true whether
the function is continuous or not, whether the interval is closed or not. If a function has absolute
extrema, they occur at critical points or the ends of the interval (if there are ends).
THEOREM 4.3
If a function has absolute extrema on an interval, then they occur either at critical points
or at the ends of the interval.
A plot of a function on an interval normally makes it clear where absolute extrema occur,
provided of course that they exist. It is then a matter of evaluating the function at the appropriate
critical point(s) or endpoint(s). When a plot of the function is unavailable (as would be the case
if the function contained unspecified parameters), the following procedure yields its absolute
extrema.
Finding Absolute Extrema for a Continuous Function

f (x) on a Closed Interval a ≤ x ≤ b
(i) Find all critical points x1 , x2 , . . . , xn of f (x) in a < x < b .
(ii) Evaluate
f (a), f (x1 ), f (x2 ), ..., f (xn ), f (b).
(iii) The absolute maximum of f (x) on a ≤ x ≤ b is the largest of the numbers in (ii);
the absolute minimum is the smallest of these numbers.
Note that it is not necessary to classify the critical points of f (x) as yielding relative maxima,
relative minima, horizontal points of inflection, vertical points of inflection, corners, or none of
these. We need only evaluate f (x) at x = a , x = b and at its critical points. The largest and
smallest of these numbers must be the absolute extrema.
When the function has discontinuities or the interval is not of finite length, Theorem 4.2
cannot be used. Careful consideration of the function at discontinuities and limits as x → ±∞
may be required.
EXAMPLE 4.20
Find the absolute maximum and minimum of the function
x 3 − 2x 2 + x + 20
f (x) =
x2 + 5
on the interval 0 ≤ x ≤ 6, if they exist.
FIGURE 4.52 Graph to illustrate absolute extrema of a function
y
4
x 3 − 2x 2 + x + 20
3.75 y=
x2 + 5
3.5
3.25
3
1 2 3 4 5 6x
2.75
2.5
2.25
SOLUTION Since f (x) is continuous on the closed interval 0 ≤ x ≤ 6, the function must
have absolute extrema. The plot in Figure 4.52 makes it clear that the absolute minimum is at
the critical point to the left of x = 3. The absolute maximum appears to be at x = 6, but there
is a relative maximum near x = 0 that should be investigated. For critical points we solve
(x 2 + 5)(3x 2 − 4x + 1) − (x 3 − 2x 2 + x + 20)(2x)
0 = f " (x) =
(x 2 + 5)2
x 4 + 14x 2 − 60x + 5
= .
(x 2 + 5)2
To four decimal places, solutions of this equation are x = 0.0850 and x = 2.7188. The
absolute minimum of the function is f (2.7188) = 2.262, and because f (0.0850) = 4.008
and f (6) = 4.146, the absolute maximum is 4.146.
Without the graph in Figure 4.52, we would evaluate
f (0 ) = 4 , f (0.0850) = 4.008, f (2.7188) = 2.262, f (6) = 4.146.
The absolute minimum of the function is 2.262 and the absolute maximum is 4.146.
EXAMPLE 4.21
Find the absolute maximum and minimum for the function
2x 2 + 3x
f (x) =
x2 + 4
on the intervals (a) 5 ≤ x < ∞ and (b) 1 ≤ x < ∞ , if they exist.
SOLUTION
(a) The graph in Figure 4.53a appears to be asymptotic to the line y = 2 and have a
relative maximum at or near x = 6. The asymptote is confirmed by
3
2x 2 + 3x 2+
lim = lim x = 2.
x→∞ x2 + 4 x→∞ 4
1+
x2
For critical points, we solve
(x 2 + 4)(4x + 3) − (2x 2 + 3x)(2x) −(x − 6)(3x + 2)

0 = f " (x) = 2 2
= .
(x + 4) (x 2 + 4)2
The only critical point in 5 ≤ x < ∞ is indeed x = 6. Consequently, the

absolute maximum of the function is f (6) = 9/4. Because f (5) = 65/29 and
limx→∞ f (x) = 2+ , the function does not have an absolute minimum on the interval
5 ≤ x < ∞.
(b) The graph on the interval 1 ≤ x < ∞ in Figure 4.53b indicates that the absolute
maximum is still f (6) = 9/4, and the function now has an absolute minimum of
f (1) = 1.
FIGURE 4.53a f (x) = (2x 2 + 3x)/(x 2 + 4) on 5 ≤ x < ∞ FIGURE 4.53b f (x) = (2x 2 + 3x)/(x 2 + 4) on 1 ≤ x < ∞
y y
2.5 2.5
2 2
1.5 1.5
2x 2 + 3x 2x 2 + 3x
y= 2 y=
1 x +4 1 x2 + 4
0.5 0.5
x x
0 5 10 15 20 25 30 0 5 10 15 20 25 30
We now consider applied maxima–minima problems.
EXAMPLE 4.22
A rectangular field is to be fenced on three sides with 1000 m of fencing (the fourth side being
a straight river’s edge). Find the dimensions of the field in order that the area be as large as
possible.
FIGURE 4.54 Fencing a SOLUTION Since the area of the field is to be maximized, we first define a function repre-
rectangular field senting this area. The area of a field of width w and length l (Figure 4.54) is
l
A = lw.
w w
This represents the area of a field with arbitrary length l and arbitrary width w . But there is only
River's edge 1000 m of fencing available for the three sides; therefore, l and w must satisfy the equation
2w + l = 1000.
With this equation we can express A completely in terms of w (or l ):
A(w) = w(1000 − 2w) = 1000w − 2w 2 .
To maximize the area of the field, we must therefore maximize the function A(w) . But what
are the values of w under consideration? Clearly, w cannot be negative, and in order to satisfy
the restriction 2w + l = 1000, w cannot exceed 500. The physical problem has now been
modelled mathematically. Find the absolute maximum of the (continuous) function A(w) on
the (closed) interval 0 ≤ w ≤ 500.
The graph of this function in Figure 4.55 clearly indicates that area is maximized by the
critical point at or near x = 250. To find the critical point, we solve
0 = A" (w) = 1000 − 4w;

the only solution is w = 250. The largest possible area is obtained when the width of the field
is 250 m and its length is 500 m.
Without the graph in Figure 4.55, we would evaluate
A(0) = 0, A(250) = 125 000, A(500) = 0,
and draw the same conclusion.
FIGURE 4.55 Area function for rectangular field
A
120 000
100 000
80 000
A = w(1000 − 2w)
60 000
40 000
20 000
w
100 200 300 400 500
EXAMPLE 4.23
Pop cans that must hold 300 mL are made in the shape of right circular cylinders. Find the
dimensions of the can that minimize its surface area.
FIGURE 4.56a Diagram of pop can FIGURE 4.56b Parts of pop can
2r Side
2πr
Top Bottom
h r r h
SOLUTION The surface area of the can (Figure 4.56a) consists of a circular top, an identical
bottom, and a rectangular piece formed into the side of the can. The total area of these three
pieces is
A = 2π r 2 + 2π rh,
where we measure r and h in centimetres (see Figure 4.56b). Since each can holds 300 mL of
pop, it follows that
300
π r 2 h = 300 or h= 2
.
πr
This equation can be used to express A completely in terms of r :
( )
300
A(r) = 2π r 2 + 2π r .
πr2
For what values of r is A(r) defined? The radius of the can must be positive and therefore r > 0.
How large can r be? In effect, it can be as large as desired. The height of the cylinder can always
be chosen sufficiently small when r is large to satisfy the volume condition π r 2 h = 300. The
can may not be aesthetically pleasing for large r and small h , but no mathematical difficulties
occur in this situation. To solve the problem we must therefore minimize
600
A(r) = 2π r 2 + , r > 0.
r
FIGURE 4.57 Area function for pop can
A
2500
2000 600
r
1500
1000
500
2 4 6 8 10 r
The graph of the function in Figure 4.57, together with the facts that
lim A(r) = ∞ and lim A(r) = ∞,

r→0+ r→∞
allow us to ballpark the answers. Area is minimized at the critical point near r = 3.5. To find
this critical point, we solve
600
0 = A" (r) = 4π r − or 4π r 3 − 600 = 0.
r2
The only solution is r = (150/π )1/3 . Thus, minimum surface area occurs when the radius
of the can is (150/π )1/3 cm and its height, obtained from the fact that h = 300/(π r 2 ) , is
2(150/π )1/3 cm.
Without the graph in Figure 4.57, we would arrive at the same conclusion by noting that
A(r) is always positive and
# $
lim A(r) = ∞, A (150)/π )1/3 = a finite number, lim A(r) = ∞.
r→0+ r→∞
EXAMPLE 4.24
FIGURE 4.58 Water The rate of discharge Q from a circular pipe is

flowing in a circular pipe 6
5/2 (θ − sin θ )3
Q = K(2R) ,
θ
where K > 0 is a constant, R is the radius of the pipe, and θ is the angle subtended at the centre
of the pipe by the wetted perimeter (Figure 4.58). Find the angle for which Q is a maximum
accurate to three decimal places.
R R
θ SOLUTION We first note that Q is maximized when the function f (θ ) = (θ − sin θ )3 /θ
under the radical is maximized. The plot of f (θ ) in Figure 4.59 indicates that the maximum
occurs for the critical point near θ = 5 radians. To find this value, we set the derivative of f (θ )
equal to zero,
Wetted perimeter θ (3)(θ − sin θ )2 (1 − cos θ ) − (θ − sin θ )3 (1)
0 = 2
.
θ
FIGURE 4.59 Discharge

in a circular pipe
f(θ)
40
20
1 2 3 4 5 6 θ
This implies that
0 = 3θ (1 − cos θ ) − (θ − sin θ ) = 2θ − 3θ cos θ + sin θ = g(θ ).
Newton’s iterative procedure for finding approximate solutions defines the sequence
g(θn ) 2θn − 3θn cos θn + sin θn

θ1 = 5, θn+1 = θn − "
= θn − .
g (θn ) 2 + 3θn sin θn − 2 cos θn
Iteration gives θ2 = 5.369 55, θ3 = 5.378 49, and θ4 = 5.378 51. Since g(5.3785) =
1.1 × 10−4 and g(5.3795) = −1.1 × 10−2 , we can say that to three decimal places, θ = 5.379.
The plot in Figure 4.59 is essential in this problem. Without it, we would not have known that
f (θ ) had only one critical point, and an approximation to it.
On the basis of these three examples, we suggest the following steps in solving applied maxima–
minima problems:
1. Sketch a diagram illustrating the situation, if appropriate.
2. Identify the quantity that is to be maximized or minimized, choose a letter to represent it,
and find an expression for this quantity.
3. If necessary, use information in the problem to rewrite the expression in 2 as a function of
only one variable.
4. Determine the domain of the function in 3.
5. Maximize or minimize the function in 3 on the domain in 4.
6. Interpret the maximum or minimum values in terms of the original problem.
Step 2 is crucial. Do not consider subsidiary information in the problem until the quantity to
be maximized or minimized is clearly identified, labelled, and an expression found for it. Only
then should other information be considered. For instance, in Example 4.22, the restriction
2w + l = 1000 for the length of fencing available was not introduced until the expression
A = lw had been identified. Likewise, in Example 4.23, volume condition π r 2 h = 300 was
introduced after surface area A = 2π r 2 + 2π rh .
We now take a look at three additional examples.
EXAMPLE 4.25
A lighthouse is 6 km offshore and a cabin on the straight shoreline is 9 km from the point on the
shore nearest the lighthouse. If a man rows at a rate of 3 km/h, and walks at a rate of 5 km/h,
where should he beach his boat in order to get from the lighthouse to the cabin as quickly as
possible? Repeat the problem in the case that the cabin is 4 km along the shoreline.
FIGURE 4.60 Fastest path SOLUTION Figure 4.60 illustrates the path followed by the man when he beaches the boat x
from lighthouse to cabin kilometres from the point on land closest to the lighthouse. His travel time t for this complete
journey is the sum of his time t1 in the water and his time t2 on land. Since speeds√
are constant,
Lighthouse each of these times may be calculated by dividing distance by speed; that is, t1 = x 2 + 36/3
and t2 = (9 − x)/5. To minimize t for some value of x between 0 and 9, we minimize the
6 function √
x 2 + 36 9−x
t (x) = t1 + t2 = + , 0 ≤ x ≤ 9.
x 3 5
Cabin The graph in Figure 4.61 indicates that t (x) has one critical point and the function has its
9 absolute minimum at this critical point. To find it, we solve
x 1 /
0 = t " (x) = √ − or 5x = 3 x 2 + 36.
2
3 x + 36 5
Thus,
25x 2 = 9(x 2 + 36),
from which we accept only the positive solution x = 9/2. The boat should therefore be beached
4.5 km from the cabin. Without Figure 4.61, we would evaluate
√ √
9 4.52 + 36 9 − 4 .5 81 + 36
t (0 ) = 2 + = 3.8, t (4.5) = + = 3.4, t (9) = ≈ 3.6.
5 3 5 3
FIGURE 4.61 Time function when cabin is FIGURE 4.62 Time function when cabin is
9 km down the shoreline 4 km down the shoreline
t t
4
3.5
3.5
3
3
2.5
2 2.5
x 2 + 36 9−x
1.5 t= + 2
3 5 x 2 + 36 4−x
1 1.5 t= +
0.5 3 5
1
x 0.5
0 2 4 6 8
0 1 2 3 4 x
When the cabin is 4 km along the shoreline,

√
x 2 + 36 4−x
t (x) = + , 0 ≤ x ≤ 4.
3 5
The graph in Figure 4.62 indicates that f (x) does not have any critical points in the interval
0 ≤ x ≤ 4. This is confirmed by the fact that critical points are again given by the equation
x 1
√ − = 0,
3 x 2 + 36 5
but the solution x = 9/2 must be rejected since it does not fall in the interval 0 ≤ x ≤ 4.
Travel time is therefore minimized by heading directly to shore or directly to the cabin, and
Figure 4.62 suggests the cabin. Verification is provided by
√ √ √
36 4 14 16 + 36 2 13 14
t (0 ) = + = , t (4 ) = = < .
3 5 5 3 3 5
EXAMPLE 4.26
A rectangular beam is to be cut from a circular log. Naturally, it is desirable for the beam to be
as strong as possible. From our experience we know that the beam should be deeper than it is
wide. For instance, place a standard 2"" × 4"" × 8" piece of lumber between two supports 2 m
apart, and sit on it. It will probably support you if you sit on the narrow edge, but not if you sit
on the wide edge. In other words, the 2 × 4 is much stronger when its depth is more than its
width. Experimental evidence in structural engineering has shown that the strength of a beam
is proportional to the product of its width and the square of its depth. Find the dimension of the
strongest beam that can be cut from a circular log of radius R .
FIGURE 4.63a Rectangular beam cut from circular log FIGURE 4.63b Width and depth related to radius of log
w/ 2
Log
d/2
R
d d
w w
SOLUTION For a beam with width w and depth d as shown in Figure 4.63a, its strength is
given by
S = kwd 2 ,
where k > 0 is a constant. If we join the centre of the log to one of the corners of the beam
(Figure 4.63b), then from the right-angled triangle
- w .2 ( )2
d
+ = R2 -⇒ d 2 = 4R 2 − w 2 .
2 2
Thus,
S = kw(4R 2 − w 2 ).
This function must be maximized on the interval 0 ≤ w ≤ 2R . We begin by finding its
critical point(s),
dS 2R
0 = = k(4R 2 − 3w 2 ) -⇒ w = √ .
dw 3
We now calculate
( )
2R
S(0) = 0, S √ > 0, S(2R) = 0.
3
√ /
Consequently, the beam is strongest when it is 2R/ 3 centimetres wide and 4 R 2 − 4 R 2 /3 =
√ √
2 2R/ 3 centimetres deep.
EXAMPLE 4.27
A wall of a building is to be braced by a square beam 10 cm by 10 cm that must pass over a parallel
wall 5 m high and 2 m from the building. The beam is to sit flush against the building and on the
floor (Figure 4.64). The top of the 5-m wall can be cut at any angle to accommodate flush contact
with the beam, but the wall must not be shortened. Find the length of the shortest such beam.
SOLUTION The length of beam from A to B is
1 1
D = 1EF 1 + 1AC1 + 1BG1 = x sec θ + cot θ + tan θ.
10 10
Since x − 2 = 5 cot θ ,
1 π
D = f (θ ) = (2 + 5 cot θ ) sec θ + (cot θ + tan θ ), 0 < θ < .
10 2
The graph of this function in Figure 4.65 shows that the absolute minimum occurs at the critical
point near θ = 1. To find it we solve
1
0 = f " (θ ) = sec θ tan θ (2 + 5 cot θ ) + sec θ (−5 csc2 θ ) + (− csc2 θ + sec2 θ )
10
( ) ( )
sin θ 5 cos θ 5 1 1 1
= 2+ − + − .
cos2 θ sin θ sin2 θ cos θ 10 cos2 θ sin2 θ
Multiplication by 10 sin2 θ cos2 θ gives

( )
3 5 cos θ
0 = 10 sin θ 2+ − 50 cos θ + sin2 θ − cos2 θ
sin θ
= 20 sin3 θ + 50 cos θ sin2 θ − 50 cos θ + sin2 θ − (1 − sin2 θ )

= 20 sin3 θ + 50 cos θ (1 − cos2 θ ) − 50 cos θ + 2 sin2 θ − 1
= 20 sin3 θ − 50 cos3 θ + 2 sin2 θ − 1.
FIGURE 4.64 Beam passing over intervening wall FIGURE 4.65 Length of beam
y D
40
B
G 35
F 30
+ 1
25 10
20
10 m
1
Building 15
5 Beam 10
5
Wall
C
x 0 0.2 0.4 0.6 0.8 1 1.2 1.4
2 E A
x
To solve this equation we use Newton’s iterative procedure. Approximations are defined by
20 sin3 θn − 50 cos3 θn + 2 sin2 θn − 1

θn+1 = θn − .
60 sin2 θn cos θn + 150 cos2 θn sin θn + 4 sin θn cos θn
If we use θ1 = 1 as the first approximation, the next three approximations are θ2 = 0.9278,
θ3 = 0.9314, and θ4 = 0.9314. Consequently, the shortest beam is f (0.9314) = 9.8 m.
EXAMPLE 4.28
In the Application Preview, we posed the problem of finding the offset that maximizes the stroke
Application Preview in the crank shown in Figure 4.66 if the offset must not exceed the radius of the circle.
Revisited
SOLUTION To find a formula for the length s of the stroke of the offset crank, we first draw
the crank (Figure 4.67a) in what is called its outer dead
/ position. Point C coincides with D
Maximizing
and A , B , and C are collinear. The length of F D is (L + r)2 − e2 .
FIGURE 4.66 /
stroke of offset crank Figure 4.67b shows the crank in its inner dead position. The length of F E is (L − r)2 − e2 .
B Consequently, the length of the stroke is
r L
A / /
e C s = (L + r)2 − e2 − (L − r)2 − e2 .
E D
FIGURE 4.67a Outer dead position of offset FIGURE 4.67b Inner dead position of offset
crank crank
A r B r A
B L−r
e L C e C
F E D F E D
We must maximize the function s(e) on the interval 0 ≤ e ≤ r . For critical points of the
function, we solve
−e −e
0 = s " (e) = / −/
(L + r)2 − e2 (L − r)2 − e2
7 8
1 1
=e / −/ .
(L − r)2 − e2 (L + r)2 − e2
/ /
This implies that (L − r)2 − e2 = (L + r)2 − e2 , which, upon squaring, gives
L2 − 2Lr + r 2 − e2 = L2 + 2Lr + r 2 − e2 -⇒ 4Lr = 0,
an impossibility. Thus, s(e) has no critical points; its maximum must occur at either e = 0 or
e = r . We therefore calculate
s(0) = (L + r) − (L − r) = 2r,
/ / / /
s(r) = (L + r)2 − r 2 − (L − r)2 − r 2 = L2 + 2Lr − L2 − 2Lr.
It is not apparent which of the two of these is larger, but

/ it can be shown algebraically
/ that 2r is
the smaller (try it). Alternatively, we note that since (L − r) − e < (L + r) − e2 , it
2 2 2
follows that
1 1
/ > / ,
(L − r)2 − e2 (L + r)2 − e2
and therefore s " (e) > 0. The function s(e) must therefore be increasing and its maximum
value must be s(r) when e = r .
Figure 4.68 shows soil to depth D covering a horizontal layer of rock. A geological
engineer sets up a source at A to emit a signal into the soil. The signal penetrates the
soil to the rock below. Some of this signal is reflected back into the soil, some penetrates
the rock, and some travels in the surface of the rock. Each part of the rock acts as an
emitter sending the signal back into the soil, some of which reaches the receiver at B .
This happens in many ways and the receiver records an accumulation of many signals.
Our problem as consultants is to use the signals to determine depth D .
FIGURE 4.68 Sending signals into the earth and receiving them
A (Emitter) Surface B (Receiver)
D Soil
Rock
SOLUTION The first question to ask is: Which of the massive accumulation of inter-
fering signals received at B from various points on the rock can be used? It seems that
the only signal distinguishable from the rest is the one that arrives at B first. Let us
concentrate on it and see what we can determine.
We let v1 and v2 represent how fast the signal travels in the soil and in the rock. It is
known that v2 > v1 , that v1 is measurable, but v2 is unknown. Next question: What is
the path followed by the signal that arrives at B first? In Figure 4.69a, we have shown a
signal following path ACEB , making acute angles θ and φ with the surface at emitter
and receiver. This cannot be the fastest path. If E is moved to the right, the signal travels
less distance in the soil where it is slow and further in the rock where it is fast. In other
words, angle φ should be made larger. But it should not be made larger than θ , else we
would make the same argument about θ . We conclude that the fastest path must occur
when angles θ and φ are equal as shown in Figure 4.69b. The problem is now to determine
angle θ for which travel time is smallest.
FIGURE 4.69a Determination of depth FIGURE 4.69b Paths for signals emit-
of soil above bedrock ted at A and received at B
s s
A Surface of earth B A Surface of earth B
θ φ θ θ
(Soil)
D (Soil) D
speed = v1
C (Rock) E C (Rock) speed = v2 E
The time for the signal to travel from A to B in Figure 4.69b is
2D s − 2D cot θ
t = csc θ + .
v1 v2
The smallest possible value for θ occurs when C and E become coincident, in which case
θ = Tan −1 (2D/s) . Thus, the domain of function t (θ ) is Tan −1 (2D/s) ≤ θ ≤ π/2.
Critical points of t (θ ) are given by
dt 2D 2D
0 = =− csc θ cot θ + csc2 θ
dθ v1 v2
−2D cos θ 2D
= 2
+
v1 sin θ v2 sin2 θ
−2 D
= (v2 cos θ − v1 ).
v1 v2 sin2 θ
Thus, the only critical point is θ = Cos −1 (v1 /v2 ) . It is physically clear that end points
of the domain Tan −1 (2D/s) ≤ θ ≤ π/2 cannot minimize t (θ ) . The critical point
must do so. (It would be a good exercise/ for you to prove this mathematically.) When
cos θ = v1 /v2 , it follows that sin θ = 1 − (v1 /v2 )2 , in which case minimum time is
9 :
2D 1 1 2Dv1 /v2
t = / + s−/
v1 1 − (v1 /v2 )2 v2 1 − (v1 /v2 )2
! "
2D v12 s
= / 1− 2 +
v1 1 − (v1 /v2 )2 v2 v2
; ( )2
2D v1 s
= 1− + .
v1 v2 v2
Now that we have a formula for the time that the fastest signal takes to arrive at
B , what do we do with it? Quantities s , v1 , and t in this formula are measurable, and
therefore known. What are not known are D and v2 . If the receiver is moved to another
location, however, a new distance and minimum time, say S and T , are obtained. They
satisfy
; ( )2
2D v1 S
T = 1− + ,
v1 v2 v2
and these last two equations could be solved for D and v2 . To reduce error due to
experimental measurements, we might suggest to the geological engineer that he take n
measurements resulting in n pairs of values, say si and t i , and plot them on a graph
of t against s (Figure 4.70). They look almost collinear, as they should. If we set
FIGURE 4.70 Determination of v2 and D from line fitting experimental data
t
(sn , tn)
(s3, t3)
(s2 , t2)
(s1, t1)
s
/
b = (2D/v1 ) 1 − (v1 /v2 )2 and m = 1/v2 , then
t = ms + b,
the equation of a straight line. By measuring the slope and t -intercept of the best-fitting
line to the points, it is then possible to solve the equations
; ( )2
2D v1 1
b = 1− , m= ,
v1 v2 v2
for v2 and D . A mathematical way of finding the best-fitting line (instead of just eyeballing
it) is discussed in Section 12.13.
EXERCISES 4.7
In Exercises 1–6 find absolute extrema, if they exist, for the function ∗ 12. An agronomist wishes to fence eight rectangular plots for exper-
on the interval. Do so without plotting a graph of the function. imentation as shown in the figure below. If each plot must contain
9000 m 2 , find the minimum amount of fencing that can be used.
1. f (x) = x 3 − x 2 − 5x + 4, −2 ≤ x ≤ 3
x−4
2. f (x) = , 0 ≤ x ≤ 10
x+1
1 1
∗ 3. f (x) = x + , ≤x≤5
x 2
∗ 4. f (x) = x − 2 sin x, 0 ≤ x ≤ 4π ∗ 13. A closed box is to have length equal to three times its width and
√ total surface area of 30 m 2 . Find the dimensions that produce maximum
∗ 5. f (x) = x x + 1, −1 ≤ x ≤ 1 volume.
12 ∗ 14. A square box with open top is to have a volume of 6000 L. Find
∗ 6. f (x) = , x<0 the dimensions of the box that minimize the amount of material used.
x2 + 2x + 2
∗ 15. A cherry orchard has 255 trees, each of which produces on the
average 25 baskets of cherries. For each additional tree planted in the
In Exercises 7–9 find absolute extrema, if they exist, for the function same area, the yield per tree decreases by one-twelfth of a basket. How
on the interval. Use a plot of the graph of the function as an aid. many more trees will produce a maximum crop?
x+1 x ∗ 16. A high school football field is being designed to accommodate
7. f (x) = , x>1 8. f (x) = , x>0 a 400-m track. The track runs along the length of the field and has
x−1 x2 + 3
semicircles beyond the end zones with diameters equal to the width of
(2x − 1)(x − 8) the field. If the end zones must be 10 m deep, find the dimensions of
9. f (x) = , x ≤ −2
(x − 1)(x − 4) the field that maximize playing area (not including the end zones).
∗ 10. Use Theorem 4.2 to prove Theorem 3.17 in Section 3.14. ∗ 17. When a manufacturing company sells x objects per month, it sets
∗ 11. One end of a uniform beam of length L is built into a wall, and the the price r of each object at
other end is simply supported (figure below). If the beam has constant
mass per unit length m , its deflection y from the horizontal at a distance x2
r(x) = 100 − .
x from the built-in end is given by 10 000
(48EI )y = −mg(2x 4 − 5Lx 3 + 3L2 x 2 ), The total cost C of producing the x objects per month is
where E and I are constants depending on the material and cross- x2

C(x) = + 2x + 20.
section of the beam, and g > 0 is a constant. How far from the built-in 10
end does maximum deflection occur?
Find the number of objects the company should sell per month in order
y to realize maximum profits.
L ∗ 18. The base of an isosceles triangle, which is not one of the equal
sides, has length b , and its altitude has length a . Find the area of the
x
largest rectangle that can be placed inside the triangle if one of the sides
of the rectangle must lie on the base of the triangle.
∗ 19. A manufacturer builds cylindrical metal cans that hold 1000 cm 3 . produces per year, x is the number of employees, and y is the daily
There is no waste involved in cutting material for the curved surface of operating budget. Annual operating costs amount to $20 000 per em-
the can. However, each circular end piece is cut from a square piece of ployee plus the operating budget of $365 y . If the manufacturer wishes
metal, leaving four waste pieces. Find the height and radius of the can to produce 1000 automobiles per year at minimum cost, how many
that uses the least amount of metal, including all waste materials. employees should it hire?
∗ 20. Sides AB and AC of an isosceles triangle have equal length. The ∗ 29. The mass flow rate of gas through a nozzle is given by the function
base BC has length 2a , as does the altitude AD from A to BC . Find <
the height of a point P on AD at which the sum of the distances AP , = 9( ) ( )(γ +1)/γ :
BP , and CP is a minimum.
= 2γ
> p 2/γ p
Q(p) = A p0 ρ0 − ,
γ −1 p0 p0
∗ 21. Two poles are driven into the ground 3 m apart. One pole protrudes
2 m above the ground and the other pole 1 m above the ground. A single
0 ≤ p ≤ p0 ,
piece of rope is attached to the top of one pole, passed through a loop
on the ground, pulled taut, and attached to the top of the other pole. where p is the discharge pressure of the gas from the nozzle. The
Where should the loop be placed in order that the rope be as short as discharge cross-sectional area of the nozzle A , the stagnation pressure
possible? p0 , the stagnation density ρ0 , and γ > 1 are all constants. The
discharge pressure for which Q is a maximum is called the critical
∗ 22. In designing pages for a book, a publisher decides that the rectan-
pressure pc . Find pc in terms of p0 and γ .
gular printed region on each page must have area 150 cm 2 . If the page
must have 2.5-cm margins on each side and 3.75-cm margins at top and ∗ 30. A printer contracts to print 200 000 copies of a membership card.
bottom, find the dimensions of the page of smallest possible area. It costs $10 per hour to run the press, and the press produces 1000
2 2 impressions per hour. The printer is free to choose the number of set
∗ 23. Find the points on the hyperbola y − x = 9 closest to (4, 0) .
types per impression to a limit of 40; each set type costs $2. If x set
∗ 24. Find the point on the parabola y = x 2 closest to (−2, 5) . types are chosen, each impression yields x cards. How many set types
should be used?
∗ 25. A light source is to be placed directly above the centre of a circular
area of radius r (figure below). The illumination at any point on the
∗ 31. At noon a ship S1 is 20 km north of ship S2 . If S1 sails south
at 6 km/h, and S2 east at 8 km/h, find when the two ships are closest
edge of the circle is directly proportional to the cosine of the angle
together.
θ and inversely proportional to the square of the distance d from the
source. Find the height h above the circle at which illumination on the ∗ 32. A military courier is located on a desert 6 km from a point P ,
edge of the table is maximized. which is the point on a long, straight road nearest him. He is ordered
to report to a point Q on the road. If we assume that he can travel at a
Source rate of 14 km/h on the desert and a rate of 50 km/h on the road, find the
point where he should reach the road in order to get to Q in the least
possible time when (a) Q is 3 km from P and (b) Q is 1 km from P .
h d
∗ 33. Find the area of the largest rectangle that can be inscribed inside a
r circle of radius r .
∗ 34. Among all rectangles that can be inscribed inside the ellipse b2 x 2 +
∗ 26. Among all line segments that stretch from points on the positive x - a 2 y 2 = a 2 b2 and have sides parallel to the axes, find the one with
axis to points on the positive y -axis and pass through the point (2, 5) , largest area.
find that one that makes with the positive x - and y -axes the triangle
with least possible area. ∗ 35. Among all rectangles that can be inscribed inside the ellipse b2 x 2 +
a 2 y 2 = a 2 b2 and have sides parallel to the axes, find the one with
∗ 27. A long piece of metal 1 m wide is to be bent in two places to form a
largest perimeter.
spillway so that its cross-section is an isosceles trapezoid (figure below).
Find the angle θ at which the bends should be formed in order to obtain ∗ 36. Find the area of the largest rectangle that has one side on the x -axis
maximum possible flow along the spillway if the lengths of AB , BC , 2
and two vertices on the curve y = e−x .
and CD are all 1/3 m.
∗ 37. Find the volume of the largest right circular cylinder that can be
A D inscribed in a sphere of radius r .
∗ 38. Find the volume of the largest right circular cylinder that can be
inscribed inside a right circular cone of radius r and height h .
B C
∗ 39. Two beams are to be cut from the larger pieces of wood left over
∗ 28. An automobile manufacturer has a Cobb–Douglas production when the strongest beam is cut from the log in Example 4.26. What
function q = x 2/5 y 3/5 , where q is the number of automobiles it are the dimensions of the strongest such beams?
∗ 40. When a shotputter projects a shot from height h above the ground ∗ 46. The illumination at a point is inversely proportional to the square
(figure below), at speed v , its range R is given by the formula of the distance from the light source and directly proportional to the
intensity of the source. If two light sources of intensities I1 and I2
7 6 8
v 2 cos θ 2gh are a distance d apart, at what point on the line segment joining the
2
R = sin θ + sin θ + , sources is the sum of their illuminations a minimum — relative to all
g v2 other points on the line segment?
∗ 47. A window is in the form of a rectangle surmounted by a semicircle
where θ is the angle of projection with the horizontal. Find the angle
with diameter equal to the width of the window. If the rectangle is
θ that maximizes R . What is the angle when v = 13.7 m/s and of clear glass while the semicircle is of coloured glass that transmits
h = 2.25 m? only half as much light per unit area as the clear glass, and if the
y total perimeter is fixed, find the proportions of the rectangular and
semicircular part of the window that admit the most light.
∗ 48. A company wishes to construct a storage tank in the form of a
h rectangular parallelepiped with a square horizontal cross-section. The
x volume of the tank must be 100 m 3 .
R
∗ 41. Two corridors, one 3 m wide and the other 6 m wide, meet at right (a) If material for the sides and top costs $1.25/m2 , and ma-
angles. Find the length of the longest beam that can be transported terial for the bottom costs $4.75/m2 , find the dimensions
horizontally around the corner. Ignore the dimensions of the beam. that minimize material costs.
(b) Repeat part (a) if the 12 edges must be welded at a cost of
∗ 42. Repeat Exercise 41 taking into account that the beam has square $7.50/m of weld.
cross-section 1/3 m on each side.
∗ 49. When an unloaded die is thrown, there is a probability of 1/6 that
∗ 43. A bee’s cell is always constructed in the shape of a regular hexag-
it will come up “two.” If the die is loaded, on the other hand, the
onal cylinder open on one end and a trihedral apex at the other (figure
probability that a “two” will appear is not 1/6, but is some number p
below). It can be shown that the total area of the nine faces is given by
between zero and one. To find p we could roll the die a large number
√ of times, say n , and count the number of times “two” appears, say, m .
3
A = 6xy + x 2 ( 3 csc θ − cot θ). It seems reasonable that an estimate for p is m/n . Mathematicians
2 define a likelihood function, which for the present situation turns out
to be
Find the angle that minimizes A . n!
f (x) = x m (1 − x)n−m .
m! (n − m)!
The value of x that maximizes f (x) on the interval 0 ≤ x ≤ 1 is
called the maximum likelihood estimate of p . Show that this estimate
is m/n .
∗ 50. An underground pipeline is to be constructed between two cities,
y A and B (figure below). An analysis of the substructure indicates that
construction costs per kilometre in regions I ( y > 0) and II ( y < 0)
are c1 and c2 , respectively. Show that the total construction cost is
minimized when x is chosen so that c1 sin θ1 = c2 sin θ2 .
x
∗ 44. Find the absolute extrema of the function f (x) = x/(x 2 + c) on y

the interval 0 ≤ x ≤ c , if they exist. Treat c > 1 as a given constant.
A (0, y1 )
∗ 45. (a) A rental company buys a new machine for p dollars, which
it then rents to customers. If the company keeps the machine I
for t years (before replacing it), the average replacement
cost per year for the t years is p/t . During these t years, (x, 0)
the company must make repairs on the machine, the number x
n depending on t as given by n(t) = t α /β , where α > 1 II
and β > 0 are constants. If r is the average cost per repair,
then the average maintenance cost per year over the life of B(x2, y2 )
the machine is nr/t . The total yearly expense associated
with the machine if it is kept for t years is therefore
∗ 51. A submarine is sailing on the surface due east at a rate of s kilo-
p nr metres per hour. It is to pass 1 km north of a point of land on an island
C(t) = + . at midnight. Soldiers on the island wishing to escape the enemy plan
t t to intercept the submarine by rowing a rubber raft in a straight-line
course at a rate of v kilometres per hour ( v < s ). What is the last
Find the optimum time at which to replace the machine.
instant that they can leave the island and expect to make contact with
(b) Discuss the cases where α = 1 and 0 < α < 1. the submarine?
∗ 52. A packing company wishes to form the 1 m by 2 m piece of card-

board in figure (a) below into a box as shown in figure (b). Cuts are to Home 3 km 1
be made along solid lines and folds along dotted lines, and two sides
are to be taped together as shown. If the outer flaps on top and bottom
must meet in the centre but the inner flaps need not, find the dimensions 5
of the box holding the most volume. How far apart will the inner flaps
be?
∗ 56. A Boeing 727-200 jet transport has wing platform area A =

1 A h B 1600 ft 2 and gross weight w = 150 000 lbs of force. At cruising
l w l w speed v (in miles per hour), the thrust F of the engines is given in
pounds force by
( )
1 2 4w 2
(a) F = ρAv 0.000182 + ,
2 6.5πρ 2 A2 v 4
Bottom is same as top
where ρ is the density of the atmosphere. Optimum speed occurs when
the ratio of thrust to speed is a minimum.
(a) Find the optimum speed of the jet at sea level where ρ =
h 0.0238 slugs per cubic foot.
B A (b) Find the optimum speed at 30 000 ft when density of the
atmosphere is about 0.375 that at sea level.
w
l ∗ 57. When a force F is applied to the object of mass m in the figure
Tape
(b) below, three other forces act on m : the force of gravity mg ( g > 0)
directly downward, a reactional force of the supporting surface, and a
horizontal, frictional force opposing F . The least force F that will
overcome friction and produce motion is given by
∗ 53. The cost of fuel per hour for running a ship varies directly as the
cube of the speed, and is B dollars per hour when the speed is b µmg
kilometres per hour. There are also fixed costs of A dollars per hour. F = ,
cos θ + µ sin θ
Find the most economical speed at which to make a trip.
where µ is a constant called the coefficient of static friction. Find the
angle θ for which F is minimal.
∗ 54. A paper drinking cup in the form of a right circular cone can be
F
made from a circular piece of paper by removing a sector and joining
edges OA and OB , as shown in the figure below. If the radius of the
circle is R , what choice of θ yields a cup of maximum volume?
m
∗ 58. Repeat Exercise 20 if the length of the base of the triangle is 2b

R A instead of 2a (but the altitude is still 2a ).
O ∗ 59. A trucking company wants to determine the highway speed to

recommend to its drivers in order that company costs are kept to a
B minimum. Taken into account will be only hourly wage w (in dollars
and assumed constant) of drivers, and gas consumption. The company
has data to support the hypothesis that for speeds v between 80 km/h
and 100 km/h, the number of kilometres per litre used by the trucks is
a linear function f (v) = a − bv (where b > 0). Find a formula for
the recommended speed in terms of a , b , and w , and p , the price per
∗ 55. The road information sign in the following figure specifies distance litre for gas.
to the next city with markings 1 m high. Bottoms of markings are 5 m
above road level. If the average motorist’s eye level is 1.5 m above ∗ 60. Verify formula 1.16 by minimizing the distance function from a
ground level, at what distance from the sign do the letters appear tallest? point to a line.
∗ 61. A window is in the form of a rectangle surmounted by a semicircle quadrilateral that can be so formed.
with diameter equal to the width of the rectangle. The rectangle is
of clear glass costing a dollars per unit area, while the semicircle is 1m
of coloured glass costing b dollars per unit area. The coloured glass
transmits only a fraction p (0 < p < 1) as much light per unit area 4m
as the clear glass. In addition, the curved portion of the window is
surmounted by a special frame at a cost of c dollars per unit length. 3m
If the total cost of the window must not exceed A dollars, find the
dimensions of the window that admit the most light. 2m
2 2
∗ 62. Find
√ the point on the ellipse 4x + 9y = 36 closest to ∗∗ 68. The lake in the figure below is basically circular with radius r . It
(4, 13 5/6) . has a narrow strip of beach all round and beyond the beach is bush.
∗ 63. The frame for a kite is to be made from six pieces of wood as shown You are walking from point P to point Q, both of which are on the
in the following figure. The four outside pieces have predetermined extension of a diameter of the lake each distance s from the centre of
lengths a and b ( b > a ). Yet to be cut are the two diagonal pieces. the lake. You can walk twice as fast on the beach as in the bush. Design
How long should they be in order to make the area of the kite as large your travel path to get from P to Q as quickly as possible.
as possible?
Lake
r
a a P Q
s s
∗∗ 69. A rectangle has width w and length L . Find the area of the largest
of all rectangles that have sides passing through the corners of the given
b b rectangle.
φ ∗∗ 70. There are n red stakes securely driven into the ground in a straight
line. A blue stake is to be added to the line, and each red stake is to
be joined to the blue one by a string. Where should the blue stake be
placed in order that the total length of all strings be as small as possible?
∗ 64. Suppose that the line Ax + By + C = 0 intersects the parabola
y = ax 2 + bx + c at points P and Q (figure below). Find the point R ∗∗ 71. When blood flows through a vein or artery, it encounters resistance
on the parabola between P and Q that maximizes the area of triangle due to friction with the walls of the blood vessel and the viscosity of
P QR . the blood itself. Poiseuille’s law for laminar blood flow states that for a
circular vessel, resistance R to blood flow is proportional to the length
y L of the blood vessel and inversely proportional to the fourth power of
2
its radius r :
y = ax + bx + c L
R = k 4,
Q r
Ax + By + C = 0
where k is a constant. The figure below shows a blood vessel of radius
P r1 from A to B and a branching vessel from D to C of radius r2 < r1 .
R 2r 2
x C
√ L2
∗∗ 65. Tangent lines to the curve y = 1 − x make triangles with the
positive x - and y -axes. Find the area of the smallest such triangle.
∗∗ 66. A farmer has a square plot of unirrigated land s metres by s metres. A E
His profit is a dollars per square metre. He wishes to install an irrigation 2r1 B
D
system that consists of a rotating arm that pivots at the centre of the
L1
field. The cost for irrigation is c dollars per square metre. Profit on
The resistance encountered by the blood in flowing from A to C is
irrigated land is b dollars per square metre. What should be the length
given by
of the rotating arm for maximum profit? Assume that b > a + c ,
L1 L2
otherwise it makes no sense to irrigate. R =k +k 4,
r14 r2
∗∗ 67. Four rods of lengths 1 m, 4 m, 2 m, and 3 m are hinged together
as shown below to form a quadrilateral. Find the area of the largest where L1 and L2 are the lengths of AD and DC , respectively.
(a) If B is assumed to the right of E , show that R can be

expressed in terms of θ as
k k
R = f (θ) = 4
(X − Y cot θ) + 4 Y csc θ,
r1 r2 i
where X and Y are the lengths of AE and CE , respec-

tively.
(b) Show that f (θ) has only one critical point θ in the range
0 < θ < π/2, and θ is defined by
r24 ∗∗ 74. A rope with a ring at one end is passed through two fixed rings at
cos θ = .
r14 the same level (figure below). The end of the rope without the ring is
then passed through the ring at the other end, and a mass m is attached
(c) Verify that θ yields a relative minimum for f (θ) . to it. If the rope moves so as to maximize the distance from m to the
(d) Show that line through the fixed rings, find angle θ .
; ( )8 Ring Ring
kX kY r2
f (θ ) = 4 + 4 1− < f (π/2).
r1 r2 r1
(e) Does θ provide an absolute minimum for f (θ) ?

∗∗ 72. A right circular cone has radius R and height H . A right circular Ring
cylinder is inscribed inside the cone so that its upper edge is on the cone m
(figure below). Find the radius of the cylinder in order that its surface
area (including top, bottom, and side) be as large as possible.
∗∗ 75. If the fencing in Example 4.22 is to form the arc of a circle, what
is the maximum possible area?
∗∗ 76. The upper left corner of a piece of paper a units wide and b units
long ( b > a ) is folded to the right edge as shown in the following
figure. Calculate length x in order that the length y of the fold be a
minimum.
2R
∗∗ 73. The triangle in the following figure represents cross-sections of a
prism (the length of the prism being perpendicular to the page). A ray x
of light (in the plane of the page) is incident on the prism at angle i
relative to the normal to the prism. The light is refracted at the two
y
faces of the prism and leaves at angle of deviation ψ relative to the
incident direction. Snell’s law relates angles of incident and refracted
light at each of the faces. It states that b
n sin α = sin i and n sin β = sin φ,
where n is the index of refraction of the material of the prism.

(a) Find ψ as an explicit function of i .
(b) Show that the angle of deviation ψ is a minimum when
i = φ . Prove that if ψm is the minimum angle of deviation,
then a
sin [(ψm + γ )/2]
n= .
sin (γ /2)
(c) This equation can be used to determine n experimentally. ∗∗ 77. The line y = mx + c intersects the ellipse b2 x 2 + a 2 y 2 = a 2 b2
Angle of incidence i is varied until ψm is achieved and in two points P and Q . Find the point R on the ellipse in order that
measured. With γ also known, n can be calculated. triangle P QR has maximum possible area.
∗∗ 78. A sheet of metal a metres wide and L metres long is to be bent

into a trough as shown in the following figure. End pieces are also to be
attached. If edge AB must be the arc of a circle, determine the radius
of the circle in order that the trough hold the biggest possible volume.
L
A B
4.8 Velocity and Acceleration

Kinematics is the study of motion — relationships among the displacement, velocity, and ac-
celeration of a body, particularly as they pertain to forces acting on the body. In this section we
bring together for a final discussion everything we have learned about displacement, velocity,
and acceleration, but do so mainly in a one-dimensional setting, and only from a differentiation
point of view; given the position of a particle moving along a straight line, find its velocity and
acceleration and use these quantities to describe the motion of the particle. In Section 5.2 we
reverse these operations; beginning with the acceleration, we find velocity and position. This
process is essential to physics and engineering, where acceleration of a particle is determined
by the forces acting on it.
When the position of a particle moving along the x -axis is known as a function of time t ,
say, x(t) , its instantaneous velocity is the derivative of x(t) with respect to t ,
dx
v(t) = . (4.9)
dt
The instantaneous acceleration of the particle is defined as the rate of change of velocity with
respect to time:
dv d 2x
a(t) = = . (4.10)
dt dt 2
FIGURE 4.71 Displacement function
500 x = t 3 − 27t 2 + 168t + 20

400
(4, 324)
300
200
100 (9, 74)
10 15 20 t
−100
(14, −176)
In Section 3.6 we discussed the motion of a particle moving along the x -axis with position
function
x(t) = t 3 − 27t 2 + 168t + 20, t ≥ 0,
where x is measured in metres and t in seconds. We now add a final touch to the discussion.
The best way to describe the motion of the particle is with a graph of the displacement function
(Figure 4.71). With calculus we can show how geometric properties of the graph reflect important
features about the velocity and acceleration of the particle. The velocity and acceleration of the
particle are
dx
v(t) = = 3t 2 − 54t + 168 = 3(t − 4)(t − 14) m/s,
dt
d 2x
a(t) = = 6t − 54 = 6(t − 9) m/s2 .
dt 2
Ignoring the physical interpretations of dx/dt and d 2 x/dt 2 as velocity and acceleration for
the moment, and concentrating only on the fact that they are the first and second derivatives of
the function x(t) , we immediately find that x(t) has a relative maximum of x(4) = 324 and a
relative minimum of x(14) = −176 at the critical points t = 4 and t = 14. The graph has a
point of inflection at (9, 74) .
Let us now discuss what the graph tells us about the motion of the particle. Ordinates
represent horizontal distances of the particle from the origin x = 0. When an ordinate is
positive, the particle is that distance to the right of the origin; when an ordinate is negative,
the particle is that distance to the left of the origin. For instance, at time t = 0, we calculate
x = 20, and therefore the particle begins 20 m to the right of x = 0. At time t = 4, it is 324 m
to the right of x = 0, and at t = 14, it is 176 m to the left of the origin.
The slope of the graph represents velocity of the particle. When slope is positive, namely
in the intervals 0 < t < 4 and t > 14, the particle is moving to the right along the x -axis;
for 4 < t < 14, velocity is negative, indicating that the particle is moving to the left. At times
t = 4 and t = 14, the particle is instantaneously at rest.
The concavity of the graph reflects the sign of the acceleration. For 0 < t < 9, the
graph is concave downward, its slope is decreasing. Physically, this means that it has a negative
acceleration; that is, its velocity is decreasing. For t > 9, the graph is concave upward, its slope
is increasing. Physically, acceleration is positive; that is, velocity is increasing. At the point
of inflection (9, 74) , the acceleration changes sign. Notice that the acceleration is not zero at
t = 4 and t = 14. In spite of the fact that the velocity is zero at these times, the acceleration
does not vanish. You might ask yourself what feature of the graph would reflect coincident zeros
for velocity and acceleration (see Exercise 16).
Speed is the magnitude of velocity. It represents how fast the particle is moving without
regard for direction. For instance, at time t = 0, the velocity and speed are both 168 m/s,
whereas at time t = 10, velocity is −72 m/s and speed is 72 m/s. Geometrically, speed is
represented by the slope of the graph without regard for sign.
With the ideas above in mind, let us detail the history of the particle’s motion. At time
t = 0, it begins 20 m to the right of the origin, moving to the right with velocity 168 m/s. Since
the acceleration is negative (to the left), the particle is slowing down (both velocity and speed
are decreasing), until at time t = 4 s, it comes to an instantaneous stop 324 m to the right of
the origin. Because the acceleration continues to be negative, the particle moves to the left, its
velocity decreasing, but its speed increasing. At time t = 9 s, when the particle is 74 m to
the right of the origin, the acceleration changes sign. At this instant, the velocity has attained a
(relative) minimum value, but speed is a (relative) maximum. With acceleration to the right for
t > 9 s, the particle continues to move left, but slows down, its velocity increasing until at time
t = 14 s, when it once again comes to a stop 176 m to the left of the origin. For time t > 14 s,
it moves to the right, picking up speed, and passes through the origin just before t = 17 s.
Further analysis of the interdependences of displacement, velocity, and acceleration is
contained in the following example.
EXAMPLE 4.29
The position of a particle moving along the x -axis is given by the function
x(t) = 3t 4 − 32t 3 + 114t 2 − 144t + 40, 0 ≤ t ≤ 5,

where x is measured in metres and t in seconds. Answer the following questions concerning
its motion:
(a) What are its velocity and speed at t = 1/2 s?
(b) When is its acceleration increasing?
(c) Is the velocity increasing or decreasing at t = 2 s?
(d) Is the particle speeding up or slowing down at t = 2 s?
(e) What is the maximum velocity of the particle in the time interval 0 ≤ t ≤ 2?
(f) What is the maximum distance the particle ever attains from the origin?
SOLUTION We use the graph of the displacement function in Figure 4.72 to suggest answers
when possible, and calculus to confirm them, when necessary. The velocity and acceleration
are
v(t) = 12t 3 − 96t 2 + 228t − 144 = 12(t − 1)(t − 3)(t − 4) m/s,

a(t) = 12(3t 2 − 16t + 19) m/s2 .
FIGURE 4.72 Displacement function
40
30 x = 3t 4 − 32t 3 + 114t 2 − 144t + 40
20
10
1 2 3 4 5 t
−10
−20
(a) The velocity is v(1/2) = −105/2 m/s. The speed is 105/2 m/s.
(b) The graph does not tell us when acceleration is increasing. Algebraically, acceleration
is increasing when its derivative is nonnegative. Since da/dt = 12(6t − 16) =
24(3t − 8) , acceleration is increasing for 8/3 ≤ t ≤ 5.
(c) The graph appears to be concave downward at t = 2, implying that velocity is
decreasing. Confirmation is provided by the fact that a(2) = −12 m/s 2 .
(d) The graph seems to be becoming less steep around t = 2, indicating that the particle
is slowing down. Since v(2) = 24 m/s, the particle is moving to the right. But
according to part (c), its acceleration is to the left at this time. With velocity and
acceleration in opposite directions, the particle is slowing down.
(e) Maximum velocity on 0 ≤ t ≤ 2 occurs at the point of inflection to the left of t = 2.
To find it, we set
0 = a(t) = 12(3t 2 − 16t + 19).

√ √
Of the two solutions (8 ± 7)/3, # only√(8 − $ 7)/3 is in the interval 0 ≤ t ≤ 2.
Maximum velocity is therefore v (8 − 7)/3 = 25.35 m/s.
(f) Maximum distance from the origin is represented by the point on the graph farthest
from the t -axis. This is at t = 5 s, for which x(5) = 45 m.
EXAMPLE 4.30
The mass M in Figure 4.73 is pulled x0 metres to the right of its position where the spring is
unstretched, and then it is released. If at time t = t0 , during the subsequent oscillations, it is
struck with a force of F newtons to the right, its position x(t) must satisfy
d 2x
M + kx = F δ(t − t0 ), x(0) = x0 , x " (0 ) = 0 ,
dt 2
where k > 0 is the spring constant, and δ(t − t0 ) is the Dirac-delta function of Section 2.5.
This assumes that air resistance due to motion is negligible, as is friction with the surface on
which oscillations take place.
(a) Verify that
6 6
k F k
x(t) = x0 cos t+√ sin (t − t0 ) h(t − t0 ),
M kM M
where h(t − t0 ) is the Heaviside unit step function, satisfies these conditions at every
t $= t0 .
(b) Show that the velocity of the mass changes by F /M metres per second as a result of
being struck at time t0 .
FIGURE 4.73 Mass vibrating on end of spring
k
M
x
x=0
SOLUTION
(a) Since the derivative of h(t − t0 ) = 0 for any t $= t0 , we find that for t $= t0 ,
6 6 6
" k k F k
x (t) = − x0 sin t+ cos (t − t0 ) h(t − t0 )
M M M M
and
6 √ 6
"" kx0 k F k k
x (t) = − cos t − 3/2 sin (t − t0 ) h(t − t0 ).
M M M M
Thus, for t $= t0 ,
9 6 √ 6 :
d 2x kx0 k F k k
M 2 + kx = M − cos t − 3/2 sin (t − t0 ) h(t − t0 )
dt M M M M
9 6 6 :
k F k
+ k x0 cos t+√ sin (t − t0 ) h(t − t0 )
M kM M
= 0;
that is, x(t) satisfies the differential equation Mx "" + kx = F δ(t − t0 ) for t $= t0 .
In addition, x(0) = x0 , and
9 6 6 6 :
k k F k
x " (0 ) = − x0 sin t+ cos (t − t0 ) h(t − t0 ) = 0.
M M M M
t=0
(b) The change in velocity at t = t0 is

9 6 6 6 :
k k F k
lim x " (t) − lim x " (t) = lim − x0 sin t+ cos (t − t0 )
t→t0+ t→t0− t→t0+ M M M M
9 6 6 :
k k
− lim− − x0 sin t
t→t0 M M
6 6 6 6
k k F k k
=− x0 sin t0 + + x0 sin t0
M M M M M
F
= .
M
EXAMPLE 4.31
Rod AB in the offset slider-crank of Figure 4.74 rotates counterclockwise with constant angu-
lar speed ω about A . End C of the follower BC is confined to straight-line motion along a
horizontal line between D and E . Find expressions for the velocity and acceleration of slider
C.
FIGURE 4.74 Offset slider crank
B
r L
θ
A H
e φ C
F G E D
x
SOLUTION Suppose we let x be the distance from F to C . Then,
x = 1F G1 + 1GC1 = 1AH 1 + 1GC1 = r cos θ + L cos φ.
Now angles θ and φ are not independent; they are related by the offset equation
e = 1BG1 − 1BH 1 = L sin φ − r sin θ.
Although these equations have been developed on the basis of Figure 4.74, which shows θ as
an acute angle, it can be shown that they are valid for any value of θ whatsoever. We could use
the offset equation to express x completely in terms of θ (see Exercise 24), but it is simpler to
work with both θ and φ . Differentiation of the expression for x with respect to time t gives the
velocity of C ,
dx dθ dφ dφ
v = = −r sin θ − L sin φ = −ωr sin θ − L sin φ .
dt dt dt dt
Differentiation of the offset equation relates dφ/dt and ω ,
dφ dθ dφ dφ ωr cos θ
0 = L cos φ − r cos θ = L cos φ − ωr cos θ -⇒ = .
dt dt dt dt L cos φ
Thus, ( )
ωr cos θ
v = −ωr sin θ − L sin φ
L cos φ
( )
sin θ cos φ + cos θ sin φ −ωr sin (θ + φ)
= −ωr = .
cos φ cos φ
A second differentiation gives the acceleration of the slider,
 ( ) 
dθ dφ dφ
cos φ cos (θ + φ) + − sin (θ + φ)(− sin φ)
dv  dt dt dt 
a = = −ωr 



dt cos2 φ
! ( ) ( )"
−ωr ωr cos θ ωr cos θ
= cos φ cos (θ + φ) ω + + sin (θ + φ) sin φ
cos2 φ L cos φ L cos φ
−ω2 r ? @
= 3
cos φ cos (θ + φ)(L cos φ + r cos θ ) + r sin (θ + φ) sin φ cos θ
L cos φ
−ω2 r ? @
= 3
L cos2 φ cos (θ + φ) + r cos2 θ .
L cos φ
EXERCISES 4.8
In Exercises 1–10 find the velocity and acceleration of an object that ∗ 13. A particle moves along the x -axis in such a way that its position
moves along the x -axis with the given position function. In each ex- as a function of time t is given by
ercise, discuss the motion, including in your discussion a graph of the
function x(t) . Assume that x is measured in metres and t in seconds. x(t) = t sin t, 0 ≤ t ≤ 9.
∗ 1. x(t) = 2t + 5, t ≥5
(a) Plot a graph of x(t) .
2
∗ 2. x(t) = t − 7t + 6, t ≥0
(b) Use Newton’s iterative procedure to find when velocity is
∗ 3. x(t) = t 2 + 5t + 10, t ≥1 zero.
∗ 4. x(t) = −2t 3 + 2t 2 + 16t − 1, t ≥0 (c) Use Newton’s iterative procedure to find when acceleration
3 2
∗ 5. x(t) = t − 9t + 15t + 3, t ≥2 is equal to 1.
∗ 6. x(t) = 3 cos 4t, t ≥0 ∗ 14. An object moving along the x -axis has position function given by
∗ 7. x(t) = 1/t, t ≥1
x(t) = 3t 4 − 16t 3 + 18t 2 + 2, 0 ≤ t ≤ 4,
∗ 8. x(t) = t + 4/t, t ≥1
∗ 9. x(t) = (t − 4)/t 2 , t ≥ 2 where x is measured in metres and t in seconds. Determine (a) on
√ what intervals the velocity is increasing and decreasing, (b) on what
∗ 10. x(t) = (t − 1)2 t, t ≥ 1
intervals the speed is increasing and decreasing, (c) when the velocity
∗ 11. An object moving along the x -axis has position given by is a maximum and a minimum, (d) when the speed is a maximum and
a minimum, (e) the maximum distance from the origin achieved by the
x(t) = t 3 − 9t 2 + 24t + 1, 0 ≤ t ≤ 6, particle, and (f) the maximum distance from the point x = 5 achieved
by the particle.
where x is measured in metres and t ≥ 0 is time in seconds. De-
termine (a) whether speed is increasing or decreasing at t = 1 s, (b) ∗ 15. Repeat Exercise 14 for the position function
maximum and minimum velocity, (c) maximum and minimum speed,
(d) maximum and minimum acceleration, and (e) when acceleration is 14t 3 101t 2 132t
increasing. x(t) = − + + 2, 0 ≤ t ≤ 6.
45 45 45
∗ 12. Repeat Exercise 11 if
∗ 16. What feature on the displacement graph would indicate a time
x(t) = 2 − 15t + 9t 2 − t 3 , 0 ≤ t ≤ 6. when the velocity and acceleration are simultaneously zero?
∗ 17. Are critical points for the velocity function the same as those for (c) Estimate the length of the stroke from the graph in part (b).
the speed function? Explain, using graphs. Check this against the formula in Example 4.28.
∗ 18. What is the maximum speed of the particle in Example 4.29 over (d) Assuming that AB rotates with constant angular speed ω ,
the time interval 0 ≤ t ≤ 2? differentiate the function in part (a) to find the velocity of
∗ 19. In many velocity and acceleration problems it is more convenient the slider in terms of θ . Verify that your result is consistent
to express acceleration in terms of a derivative with respect to position, with the velocity formula in Example 4.31.
as opposed to a derivative with respect to time. Show that acceleration (e) Plot the velocity function in part (d) for 0 ≤ θ ≤ 2π
can be written in the form using the values of r , L , and e in part (b) if AB rotates
one revolution each second. Does the velocity appear to be
dv zero when the position graph in part (b) is at its highest and
a =v .
dx lowest points? Estimate maximum and minimum velocities
from the graph.
In Exercises 20–23 assume that a particle moves along the x -axis in ∗∗ 25. A landing approach is to be shaped generally as shown in the figure
such a way that its position, velocity, and acceleration are continuous below. The following conditions are imposed on the approach pattern:
functions on the interval a ≤ t ≤ b . Discuss the validity of each
(a) Altitude must be h metres when descent commences.
statement.
(b) Smooth touchdown must occur at x = 0.
∗ 20. When position has a relative maximum, so does the absolute value
of the distance from the origin to the particle. (c) Constant horizontal speed U metres per second must be
maintained throughout.
∗ 21. When position has a relative minimum, so does the absolute value
of the distance from the origin to the particle. (d) At no time must vertical acceleration in absolute value ex-
ceed some fixed positive constant k .
∗ 22. When velocity has a relative minimum, so does speed.
(e) The glide path must be a cubic polynomial.
∗ 23. When velocity has a relative maximum, so does speed.
Find when descent should commence.
∗ 24. (a) Use the offset equation e = L sin φ − r sin θ to show that
position x of the slider in Example 4.31 can be expressed
in the form y
/ U
x = r cos θ + L2 − (e + r sin θ)2 .
h
(b) Plot this function for 0 ≤ θ ≤ 2π when L = 9 cm,
r = 2 cm, and e = 1 cm. x
4.9 Related Rates

Many interesting and practical problems involving rates of change are commonly referred to as
related rate problems. In these problems, two or more quantities are related to each other and
rates at which some of them change are known. It is required to find rates at which the others
change. Related rate problems deal almost exclusively with rates of change of quantities with
respect to time. To solve these problems we first consider three examples. These will suggest
the general procedures by which all related rate problems can be analyzed. We shall then discuss
two somewhat more complicated problems.
EXAMPLE 4.32
A man 2 m tall walks directly away from a streetlight that is 8 m high at the rate of 3/2 m/s.
How fast is the length of his shadow changing?
FIGURE 4.75 Shadow of
man walking away from lightpost SOLUTION When x denotes the distance between the man and the lightpost (Figure 4.75),
the fact that he walks directly away from the light at 3/2 m/s means that x is changing at a rate
of 3/2 m/s; that is, dx/dt = 3/2 m/s. If s represents the length of the man’s shadow, then we
ds are searching for ds/dt . Similar triangles in Figure 4.75 enable us to relate s and x ,
8 =?
dt
dx = 3 x+s 8
dt 2 2 = ,
s 2
x s
and this equation can be solved for s in terms of x :

x
s = .
3
Now, s and x are each functions of time t ,
s = f (t) and x = g(t),

although we have not calculated the exact form of these functions. Indeed, the essence of the
related rate problem is to find ds/dt without ever knowing f (t) explicitly. To do this we note
that since the equation s = x/3 is valid at any time t when the man is walking away from the
light, we may differentiate with respect to t to obtain
ds 1 dx
= .
dt 3 dt
This equation relates the known rate dx/dt = 3/2 with the unknown rate ds/dt . It follows
that ( )( )
ds 1 3 1
= = ,
dt 3 2 2
and the man’s shadow is therefore getting longer at the rate of 1/2 m/s.
Knowing dx/dt in this example, we have calculated ds/dt , and have done so without finding
s explicitly as a function of time t . This is the essence of a related rate problem. Since dx/dt
is a constant value, it is quite easy to find s as a function of t , and hence ds/dt . Indeed, if
we choose time t = 0 when the man starts to walk away from the streetlight, then his distance
from the light at any given time is x = 3t/2 m. Combine this with the fact that s = x/3 and
we may write
( )
1 3t t
s = = m.
3 2 2
With this explicit formula for s , it is clear that ds/dt = 1/2. What is important to realize
is that the solution in this paragraph is possible only because the man walks at a constant rate.
Were his speed not constant, it might be impossible to find s explicitly in terms of t . The next
example illustrates this point in that the given rate is known only at one instant in time.
EXAMPLE 4.33
A ladder leaning against a house (Figure 4.76) is prevented from moving by a young child.
Suddenly, something distracts the child and she releases the ladder. The ladder begins slipping
down the wall of the house, picking up speed as it falls. If the top end of the ladder is moving
FIGURE 4.76 Ladder
at 1 m/s when the lower end is 15 m from the house, how fast is the foot of the ladder moving
sliding down a wall
away from the house at this instant?
dy
= −1 SOLUTION Figure 4.76 indicates that when y denotes the height of the top of the ladder
dt
above the ground, then dy/dt = −1 m/s when x = 15 m (the negative sign because y is
decreasing). We emphasize here that dy/dt = −1 only when x = 15. What is required is
dx/dt when x = 15 m. Because the triangle in the figure is right-angled, we may write
y 20
x 2 + y 2 = 202 ,
and this equation is valid at any time during which the ladder is slipping. If we differentiate
x with respect to time t , using extended power rule 3.21,
dx dx dy
= ? at x = 15
dt 2x + 2y = 0.
dt dt
√ √
When x = 15, we calculate that y = 400 − 225 = 5 7, and therefore at this instant
dx √
15 + 5 7(−1) = 0.
dt
√
This yields dx/dt = 7/3, and we can√say that when the foot of the ladder is 15 m from the
wall, it is moving away from the wall at 7/3 m/s.
EXAMPLE 4.34
A tank in the form of a right circular cone with altitude 6 m and base radius 3 m (Figure 4.77) is
FIGURE 4.77 Tank being
being filled with water at a rate of 4000 L/min. How fast is the surface of the water rising when
filled with water
the depth is 3 m?
6
SOLUTION Figure 4.77, which illustrates a cross-section of the tank, indicates that when the
r dD = ? depth of water in the tank is D , the volume V of water is
dt
6
at D = 3 1
dV = 4 D V = π r 2 D.
dt 3
Of the three variables V , r , and D in this equation, we are concerned only with V and D , since
dV /dt is given and dD/dt is what we want. This suggests that we eliminate r using similar
triangles. Since r/D = 3/6, we have r = D/2; therefore,
( )2
1 D 1
V = π D = π D3.
3 2 12
Because this result is valid for all time t during the filling process, we can differentiate with
respect to t , once again using extended power rule 3.21:
dV 1 dD
= π D2 .
dt 4 dt
Since dV /dt = 4 m 3 /min (converted from the rate of 4000 L/min since the litre is not an
acceptable unit of measure for volume), we find that when D = 3,
1 dD
4 = π(3)2 ,
4 dt
from which dD/dt = 16/(9π ) . The surface is therefore rising at a rate of 16/(9π ) m/min.
These examples illustrate the following general procedure for solving related rate problems:
1. Sketch a diagram illustrating all given information, especially given rates of change and
desired rates of change. Do not draw the diagram at the instant in question; draw it slightly
before or slightly after.
2. Find an equation valid for all time (in some interval about the instant in question) that
involves only variables whose rates of change are given or required.
3. Differentiate the equation in step 2 and solve for the required rate.
Steps 1 and 3 are usually quite straightforward; step 2, on the other hand, may tax your ingenuity.
To find the equation in the appropriate variables, it may be necessary to introduce and substitute
for additional variables. Finding these substitutions requires you to analyze the problem very
closely.
Be careful not to substitute numerical data that represent the instant at which the derivative
is required before differentiation has taken place. Numerical data must be substituted after
differentiation. For instance, in Example 4.34, radius r of the surface of the water when D = 3
is 3/2. If we substitute this into V = π r 2 D/3, we obtain a function V = π(3/2)2 D/3 =
3π D/4, which is valid only when D = 3. It cannot therefore be differentiated; only equations
that are valid for a range of values of t can be differentiated with respect to t .
We now apply this procedure to two further examples. The first is an extension of Exam-
ple 4.32.
EXAMPLE 4.35
A man 2 m tall walks along the edge of a straight road 10 m wide. On the other edge of the road
stands a streetlight 8 m high. If the man walks at 3/2 m/s, how fast is his shadow lengthening
when he is 10 m from the point directly opposite the light?
FIGURE 4.78 Shadow of man walking away from streetlight
y
10
s
x ds = ? at x = 10
dx = 3 dt
dt 2
SOLUTION First we draw Figure 4.78, wherein the man’s speed is represented as the time
rate of change of his distance x from the point on his side of the road directly opposite the light.
What is required is the rate of change ds/dt of the length of his shadow when x = 10. To find
an equation relating x and s , we first use similar (vertical) triangles to write
y+s 8
= = 4,
s 2
from which y = 3s , an equation that relates s to y , rather than s to x . However, since
y 2 = x 2 + 100,
we substitute to obtain
9s 2 = x 2 + 100.
The derivative of this equation with respect to time t gives
ds dx
18s
= 2x .
dt dt
√ √
When x = 10, we obtain s = 100 + 100/3 = 10 2/3, and at this instant,
7 √ 8 ( )
10 2 ds 3
18 = 2(10) .
3 dt 2
Thus,
√
ds 30 2
= √ = ,
dt 60 2 4
√
and the man’s shadow is therefore lengthening at the rate of 2/4 m/s.
It is worthwhile noting
√ in this example that were we to solve for ds/dt before substituting
x = 10 and s = 10 2/3, then we would have
ds x dx
= .
dt 9s dt
√
Since dx/dt is always equal to 3/2, and s = x 2 + 100/3, we can write
( )
ds x 3 x
= √ = √ ,
dt 3 x + 100 2
2 2
2 x + 100
a general formula for ds/dt . The limit of this rate as x becomes very large is
ds x 1
lim = lim √ = .
x→∞ dt x→∞ 2 x 2 + 100 2
But for very large x , the man essentially walks directly away from the light, and this answer, as
we might expect, is identical to that in Example 4.32.
EXAMPLE 4.36
One end of a rope is tied to a box. The other end is passed over a pulley 5 m above the floor
and tied at a level 1 m above the floor to the back of a truck. If the rope is taut and the truck
moves at 1/2 m/s, how fast is the box rising when the truck is 3 m from the plumbline through
the pulley?
FIGURE 4.79 Mass at- SOLUTION In Figure 4.79, we have represented the speed of the truck as the rate of change
tached to a truck passed over a of length x , dx/dt = 1/2 m/s. What is required is dy/dt when x = 3. To find an equation
pulley relating x and y , we first use the fact that the length z of rope between pulley and truck is the
hypotenuse of a right-angled triangle with sides of lengths x and 4,
Length of rope = L
z2 = x 2 + 16.
z
5
dy For an equation relating y and z , we note that the length of the rope, call it L , remains constant,
y dt = ? at x = 3 and is equal to the sum of z and 5 − y ,
1
L = z + (5 − y).
x
dx 1 These two equations can be combined into
=
dt 2
(L − 5 + y)2 = x 2 + 16,
and differentiation with respect to time t now gives
dy dx
2(L − 5 + y) = 2x .
dt dt
When x = 3, we may write that
(L − 5 + y)2 = 9 + 16 = 25.
We could solve this equation for y (in terms of L ), but it is really not y that is needed to obtain
dy/dt from the preceding equation. It is L − 5 + y , and this is clearly equal to 5. Thus, when
x = 3, we have
( )
dy 1
2(5) = 2(3) ;
dt 2
that is, dy/dt = 3/10, and the box is rising at a rate of 3/10 m/s.
The reader should compare this example with the problem in Exercise 2. They may appear
similar, but are really quite different.
Mechanical engineers have a question concerning the mechanism in Figure 4.80. Rod
OB , of length l , rotates counterclockwise in the xy -plane around the origin at ω rev-
olutions per second. Rod AB , attached to OB , is such that A is confined to sliding
horizontally along the x -axis. For unrestricted motion, the length L of AB is greater
than twice l . The engineers wish to know the maximum speed attained by slider A .
FIGURE 4.80 Speed of slider in a two-bar mechanism
B
I L
θ
O A x
SOLUTION Slider A moves back and forth along the x -axis, repeating its motion for
each revolution of B . To determine the maximum speed of A , we need only consider
its motion as B moves from (l, 0) to (−l, 0) along the upper semicircle. We therefore
take angle θ in the interval 0 ≤ θ ≤ π . The cosine law applied to triangle OAB gives
L2 = l 2 + x 2 − 2lx cos θ . If this is differentiated with respect to time,
dx dx dθ
0 = 0 + 2x − 2l cos θ + 2lx sin θ .
dt dt dt
When we set dθ/dt = 2π ω , and solve for dx/dt ,
dx 2lx sin θ (2π ω) 2π ωlx sin θ

= = m/s.
dt 2l cos θ − 2x l cos θ − x
Two things are worth noticing. Velocity is zero when sin θ = 0, and this is when θ = 0
and θ = π , when B is on the x -axis (as we would expect). The denominator l cos θ − x
can never vanish since x/ l is always greater than unity.
To find maximum speed of the follower, we should determine minimum velocity since
A is moving left when 0 ≤ θ ≤ π . Since this occurs when acceleration is zero, we set
d 2x 2π ωl sin θ dx 2π ωlx cos θ dθ

0 = 2
= +
dt l cos θ − x dt l cos θ − x dt
( )
2π ωlx sin θ dθ dx
− −l sin θ −
(l cos θ − x)2 dt dt
2π ωl sin θ dx
= (l cos θ − x + x)
(l cos θ − x)2 dt
2π ωlx dθ
+ [cos θ (l cos θ − x) + l sin2 θ ]
(l cos θ − x)2 dt
( )
2π ωl 2 sin θ cos θ 2π ωlx sin θ 4π 2 ω2 lx(l − x cos θ )
= +
(l cos θ − x)2 l cos θ − x (l cos θ − x)2
4π 2 ω2 l 3 x sin2 θ cos θ + 4π 2 ω2 lx(l − x cos θ )(l cos θ − x)
= m/s2 .
(l cos θ − x)3
We now set the numerator equal to zero, at the same time removing the factor
4π 2 ω2 lx ,
0 = l 2 sin2 θ cos θ + (l − x cos θ )(l cos θ − x)
= l 2 (1 − cos2 θ ) cos θ + l 2 cos θ − lx − lx cos2 θ + x 2 cos θ

= −l 2 cos3 θ − lx cos2 θ + (2l 2 + x 2 ) cos θ − lx.
This equation must be combined with L2 = l 2 + x 2 − 2lx cos θ to yield x and θ . If we

set y = cos θ , then we must solve the following nonlinear equations for x and y ,
l 2 y 3 + lxy 2 − (2l 2 + x 2 )y + lx = 0, L2 = l 2 + x 2 − 2lxy.
Normal procedure would be to solve one of these equations for x in terms of y , or y

in terms of x , substitute into the other equation, and thereby obtain one equation in one
unknown. Unfortunately, none of these possibilities seems appealing. If we set y = ax
in each of the equations, we obtain
l 2 (a 3 x 3 ) + lx(a 2 x 2 ) − (2l 2 + x 2 )(ax) + lx = 0, L2 = l 2 + x 2 − 2lx(ax).
When we cancel an x from the first equation, then both contain only x 2 ’s,
(l 2 a 3 + la 2 − a)x 2 = 2l 2 a − l, (1 − 2la)x 2 = L2 − l 2 .
When we solve each of these for x 2 , and equate results, we obtain
2l 2 a − l L2 − l 2
= .
l 2 a 3 + la 2 − a 1 − 2la
When we cross multiply,
0 = L2 (l 2 a 3 + la 2 − a) − l 2 (l 2 a 3 + la 2 − a) + l − 4l 2 a + 4l 3 a 2
= (l 2 L2 − l 4 )a 3 + (lL2 + 3l 3 )a 2 − (L2 + 3l 2 )a + l.
This cubic equation must be solved for a (once l and L are specified). When this is done,
x 2 = (L2 − l 2 )/(1 − 2la) gives the position of maximum speed, and angle θ is given
by θ = Cos −1 (ax) . For example, if L = 0.6 m and l = 0.2 m, the equation for a
reduces to
0.0128a 3 + 0.096a 2 − 0.48a + 0.2 = 0.
Of the three solutions −11.0287, 0.461 987, and 3.0667 of this equation, only a =
0.461 987 is acceptable ( a cannot be negative and the largest root leads to a negative value
for x 2 ). With this value of a , we find x = 0.626 529 and θ = 1.277 15. Maximum
speed of the slider is therefore
A A A A
A dx A A A
A A = A 2π ω(0.2)(0.626 529) sin 1.277 15 A = 1.3253ω.
A dt A A 0.2 cos 1.277 15 − 0.626 529 A
EXERCISES 4.9
∗ 1. A convertible is travelling along a straight highway at 100 km/h. A If the water level is rising at 1 cm/min when the depth is 1 m at the deep
child in the car accidentally releases a helium-filled balloon, which then end, at what rate is water being pumped into the pool?
rises vertically at 10 m/s. How fast are the child and balloon separating
∗ 8. Boyle’s law for a perfect gas states that the pressure exerted by the
4 s after the balloon is released?
gas on its containing vessel is inversely proportional to the volume
∗ 2. A rope passes over a pulley and one end is attached to a cart as occupied by the gas. If when the volume is 10 L and the pressure is
shown in the figure below. If the rope is pulled vertically downward at 50 N/m 2 , the volume is increasing at 1/2 L/s, find the rate of change
2 m/s, how fast is the cart moving when s = 6 m? of the pressure of the gas.
∗ 9. A woman driving 100 km/h along a straight highway notes that the
shadow of a cloud is keeping pace with her. What can she conclude
about the speed of the cloud?
5m
∗ 10. A fisherman is trolling at a rate of 2 m/s with his lure 100 m behind
the boat and on the surface. Suddenly a fish strikes and dives vertically
1m at a rate of 3 m/s. If the fisherman permits the line to run freely and it
s always remains straight, how fast is the line being played out when the
reel is 50 m from its position at the time of the strike?
∗ 3. A light is on the ground 20 m from a building. A man 2 m tall walks
from the light directly toward the building at 3 m/s. How fast is the ∗ 11. Air expands adiabatically in accordance with the law P V 7/5 =
length of his shadow on the building changing when he is 8 m from the constant. If at a given time, the volume V is 100 L and the pressure P
building? is 40 N/cm 2 , at what rate is the pressure changing when the volume is
decreasing at 1 L/s?
∗ 4. A funnel in the shape of a right circular cone is 15 cm across the
top and 30 cm deep. A liquid is flowing in at the rate of 80 mL/s and ∗ 12. Sand is poured into a right circular cylinder of radius 1/2 m along
flowing out at 15 mL/s. At what rate is the surface of the liquid rising its axis (figure below). Once sand completely covers the bottom, a right
when the liquid fills the funnel to a depth of 20 cm? circular cone is formed on the top.
∗ 5. A water tank is in the form of a right circular cylinder of diameter (a) If 0.02 m 3 of sand enters the container every minute, how
3 m and height 3 m on top of a right circular cone of diameter 3 m fast is the top of the sand pile rising?
and height 1 m. If water is being drawn from the bottom at the rate of (b) How fast is the sand rising along the side of the cylinder?
1 L/min, how fast is the water level falling when (a) it is 1 m from the
top of the tank and (b) it is 3.5 m from the top of the tank? Sand
∗ 6. A point P moves along the curve y = x 2 + x + 4, where x and
y are measured in metres. Its x -coordinate decreases at 2 m/s. If the
perpendicular from P to the x -axis intersects this axis at point Q ,
how fast is the area of the triangle with vertices P , Q , and the origin
changing when the x -coordinate of P is 2 m?
∗ 7. Water is being pumped into a swimming pool which is 10 m wide,
20 m long, 1 m deep at the shallow end, and 3 m deep at the deep end. 1
∗ 13. A balloon has the shape of a right circular cylinder of radius r and
length l with a hemisphere at each end of radius r . The balloon is
being filled at a rate of 10 mL/s in such a way that l increases twice as 5 C
fast as r . Find the rate of change of r when r = 8 cm and l = 20 cm.
B
∗ 14. An oval racetrack has a straight stretch 100 m long and two semi- 10
circles, each of radius 50 m (figure below). Car 1, on the infield, moves
along the x -axis from O to B . It accelerates from rest at O , attains a
A
speed of 10 m/s at C , and maintains this speed along CB . Car 2 travels
along the quarter oval ADEB . It is at D when Car 1 is at C . Between ∗ 20. Let P (x, y) be a point on the first-quadrant portion of the hyper-
D and B , Car 2 maintains the same rate of change of its x -coordinate bola x 2 − y 2 = 1. Let R be the foot of the perpendicular from P to
as does Car 1. the x -axis, and Q(x ∗ , 0) be the x -intercept of the normal line to the
(a) Find a formula for the rate of change of the y -coordinate hyperbola at P .
of Car 2 between D and B . (a) Show that x ∗ = 2x .
(b) How fast is the y -coordinate of Car 2 changing when it is (b) If P moves along the hyperbola so that its x -coordinate is
at point E ? decreasing at 3 units per unit time, how fast is the area of
(c) If the cars collide at B , which car suffers the most damage? triangle QP R changing when x = 4?
∗ 21. A solution passes from a conical filter 24 cm deep and 16 cm across
y
the top into a cylindrical container of diameter 12 cm. When the depth
of solution in the filter is 12 cm, its level is falling at the rate of 1 cm/min.
100 How fast is the level of solution rising in the cylinder at this instant?
∗ 22. A light is at the top of a pole 25 m high and a ball is dropped at the
same height from a point 10 m from the light. How fast is the shadow
Car 1 B of the ball moving along the ground 1 s later? The distance fallen by
100 O the ball t seconds after it has been dropped is d = 4.905t 2 metres.
C x
∗ 23. A point moves along the parabola y = x 2 − 3x ( x and y measured
Car 2
in metres) in such a way that its x -coordinate changes at the rate of
E
2 m/s. How fast is its distance from the point (1, 2) changing when it
A D
25 is at (4, 4) ?
50
∗ 24. Repeat Exercise 23 given that the parabola is replaced by the curve
∗ 15. A ship is 1 km north of a pier and is travelling N30 ◦ E at 3 km/h. A (x + y)2 = 16x .
second ship is 3/4 km east of the pier and is travelling east at 7 km/h.
How fast are the ships separating? ∗ 25. The volume of wood in the trunk of a tree is sometimes calculated
by considering it as a frustrum of a right circular cone (figure below).
∗ 16. The circle in the figure below represents a long-playing record
which is rotating clockwise at 100/3 rpm. A bug is walking away (a) Verify that the volume of the trunk is
from the centre of the record directly toward point P on the rim of the
1
record at 1 cm/s. When the bug is at position R , 10 cm from O , angle V = π h(R 2 + rR + r 2 ).
θ is π/4 radians. Find the rate at which the distance from the bug to 3
the fixed point Q is changing when the bug is at R .
(b) Suppose that at the present time the radii of the top and
bottom are r = 10 and R = 50 cm, and the height is
1m h = 30 m. If the tree continues to grow so that ratios r/R
O Q and r/ h always remain the same as they are now, and R
increases at a rate of 1/2 cm/year, how fast will the volume
R P be changing in 2 years?
∗ 17. Eight skaters form a “whip.” Show that the seventh person on the
whip travels twice as fast as the fourth person. 2r
∗ 18. A particle moves counterclockwise around a circle of radius 5 cm
centred at the origin making 4 revolutions each second. How fast is the
particle moving away from the point with coordinates (5, 6) when it is
at position (−3, 4) ? h
∗ 19. Two people ( A and B in the following figure) walk along opposite
sides of a road 10 m wide. A walks to the right at 1 m/s, and B walks
to the left at 2 m/s. A third person, C , walks along a sidewalk 5 m from
the road in such a way that B is always on the line joining A and C .
Find the speed of C . 2R
∗ 26. Sand is poured into a right circular cone of radius 2 m and height
3 m along its axis (figure below). The sand forms two cones of equal
height h , one inverted on top of the other.
(a) If 0.02 m 3 of sand enters the container every minute, how
fast is the top of the pile rising when it is just level with the
top of the container?
(b) How fast is the sand rising along the side of the container
at this instant?
h
R
3 R
h ∗ 31. A hemispherical tank of radius 3 m has a light on its upper edge
(figure below). A stone falls vertically along the axis of symmetry of
∗ 27. If to the mechanism in Figure 4.80 we add a rod AC (figure below), the tank, and when it is 1 m from the bottom of the tank, it is falling at
where C is confined to sliding vertically, find the velocity of C in terms 2 m/s. How fast is its shadow moving along the surface of the tank at
of x , θ , and y . this instant?
y C
B Light
l L y
O A x Stone
x
k
Shadow of stone
∗ 28. In the figure below, a plane flies due north at 200 km/h at constant
altitude 1 km. A car travels due east on a straight highway at 100 km/h.
At the moment the plane crosses over the highway, the car is 2 km east ∗ 32. If the sides of a triangle have lengths a , b , and c , its area is given
of the point on the road directly below the plane. How fast are the plane by
/
and car separating 1 min after this? A= s(s − a)(s − b)(s − c),
Plane where s = (a + b + c)/2 is one-half its perimeter. If the length of

each side increases at a rate of 1 cm/min, how fast is A changing when
1 km a = 3 cm, b = 4 cm, and c = 5 cm?
∗ 33. (a) If θ is the angle formed by the minute and hour hands of a
Car clock, what is the time rate of change of θ (in radians per
North minute)?
East (b) If the lengths of the hands on the clock are 10 cm and 7.5 cm,
find the rate at which their tips approach each other at 3:00
∗ 29. The infield of a baseball diamond is a square with distances be-
(i.e., find dz/dt in the figure below).
tween bases being 27 m (approximately). The hitter hits a ground ball
to the third baseman, accelerates quickly, and attains a speed of 6 m/s (c) Repeat part (b) but replace time 3:00 with 8:05.
as she runs to first base. The third baseman catches the ball at a point
2 m from the bag on the line betwen second and third base, and throws
the ball to the first baseman at 35 m/s. If the ball is halfway to first base 12
when the batter is three-quarters of the way to first base, how fast is the 11 1
distance between them changing at this instant? 10 10 z 2
∗ 30. In the following figure the boy’s feet make 1 revolution per second 9 3
around a sprocket of radius R metres. The chain travels around a 7:5
sprocket of radius r metres on the back wheel, which itself has radius 8 4
R metres. If a stone embedded in the tire becomes dislodged, how fast 7 5
is it travelling when it leaves the tire? Assume that the rear wheel has 6
been placed on a stand so that the bicycle is stationary.
4.10 LCR -Circuits 309
∗∗ 34. A runner moves counterclockwise around the track in the figure

below at a rate of 4 m/s. A camera at the centre of the track is placed
on a swivel so that it can follow the runner. Find the rate at which the
camera turns when (a) the runner is at A and (b) the runner is at B .
A B 1
z 2
30 m
Camera 6 x
y 10
Semicircle 100 m Semicircle
8
∗∗ 35. A man 2 m tall walks along the edge of a straight road 10 m wide 10
(the figure to the right). On the other edge of the road stands a streetlight
8 m high. A building runs parallel to the road and 1 m from it. If the
man walks away from the light at 2 m/s, how fast is the height of the
shadow on the wall changing when he is 10 m from the point on the
road directly opposite the light?
4.10 LCR-Circuits
FIGURE 4.81 Circuit Modern electronic equipment contains a vast array of devices; many find their origin in three
containing capacitor and voltmeter fundamental elements — capacitors, resistors, and inductors. How these elements relate to
Capacitor one another and how they affect voltages, charges, and currents in electric circuits can be fully
understood with calculus. We begin with capacitors.
A capacitor is a device that stores equal amounts of positive charge and negative charge in
such a way that the charges cannot neutralize one another. Suppose Q > 0 is the amount of
positive charge, and therefore −Q is the negative charge. Separation of these charges creates a
potential difference V at the terminals of the capacitor. It can be measured by placing a voltmeter
across the terminals as shown in Figure 4.81. The size of V depends on how charges are stored
Voltmeter in the capacitor; different configurations lead to different potential differences. When we divide
Q by the potential difference V that it produces, we obtain what is called the capacitance C
of the capacitor,
FIGURE 4.82 Circuit Q
C = . (4.11)
containing capacitor and battery V
C = 10−6 F
It has units of coulombs per volt, called farads (F). The higher the capacitance, the more charge
that can be stored per volt of potential difference.
If a 9 V battery is connected to a 10−6 F capacitor (Figure 4.82) and the switch is closed, the
S battery creates a flow of charge in the circuit until the capacitor is charged with Q = 9(10−6 )
coulombs. The rate at which charge flows is called current, denoted by the letter i . It is
measured in amperes (A); 1 A is a flow of 1 C of charge per second. For a simple circuit like
that in Figure 4.82, we can think of i as the rate of change of charge Q on the capacitor, and
9V
therefore
dQ
FIGURE 4.83 Direction
i = . (4.12)
dt
of current flow related to potential
difference
When there is no capacitor in a circuit, or even when there is, we can think of charge Q flowing
past any specific point in the circuit. How much charge flows past this point per unit time is
Potential represented by i .
+ difference − A resistor is an electronic device that retards the flow of charge in a circuit. When a
=V
potential difference V is maintained between the terminals A and B of the resistor in Figure
4.83, where the + and − signs indicate that B is at higher potential than A , positive charge
B Resistor A flows from B to A . If the rate of flow (current) is i amperes, the ratio
V
Current R = (4.13)
i
is called the resistance of the resistor. It is measured in volts per ampere, called ohms ( - ).
The higher the resistance, the smaller the current generated by a given voltage, or the larger the
voltage required to produce a given current. For example, to maintain a current of 2 A through
a 3 - resistor requires 6 V, but to maintain the same current through a 30 - resistor requires
60 V.
The third fundamental circuit element is the inductor. Voltage across a capacitor is related
to charge; voltage across a resistor is related to current (the rate of change of charge); voltage
across an inductor is related to the rate of change of current. In other words, an inductor reacts
FIGURE 4.84 Circuit for to changes in current. If V is the voltage across the terminals of an inductor and current i is
Kirchhoff’s loop rule
changing, then the inductance of the inductor is defined as
C
V V
i L= = 2 . (4.14)
R di/dt d Q/dt 2
+
V
−
It has units of volts per ampere per second, called henries (H). When inductance is large, a small
L rate of change of current produces a large voltage across the terminals of the inductor.
Charge flows through a resistor or any other electric device when a potential difference is
S created between its terminals. A device that creates and maintains potential difference is an
emf device (emf is short for electromotive force). Examples are batteries, electric generators,
solar cells, and thermopiles. An emf device is shown in Figure 4.84 in a circuit including a
capacitor with capacitance C , a resistor with resistance R , an inductor with inductance L , and
a switch S . This is called an LCR-circuit. When the switch S is closed, potential difference
across the terminals of the emf device causes charge to flow in the circuit creating a current
i . Kirchhoff’s loop rule for electric circuits implies that the sum of the potential differences
across the capacitor, resistor, and inductor must be equal to output potential of the emf device.
Using equations 4.11, 4.13, and 4.14, we obtain
di Q
L + Ri + = V. (4.15a)
dt C
If i is replaced by dQ/dt , we have
d 2Q dQ Q
L 2
+R + = V, (4.15b)
dt dt C
and if this equation is differentiated with respect to t , we also have
d 2i di i dV
L 2
+R + = . (4.15c)
dt dt C dt
These are differential equations that must be solved for Q (4.15b) or i (4.15c) when V is
given as a function of t .
EXAMPLE 4.37
At time t = 0, a 6 - resistor, a 1 H inductor, and a 0.04 F capacitor are connected with a

generator producing a voltage of 10 sin 5t , where t ≥ 0 is time in seconds, by closing the
switch (Figure 4.85).
FIGURE 4.85 Current in an LCR -circuit
S
0.04 F
10 sin 5t 6-
1H
(a) What differential equation must current in the circuit satisfy for t > 0?
(b) Verify that
5 1
i(t) = sin 5t − e−3t (5 cos 4t + 10 sin 4t)
3 3
satisfies the equation in part (a).
(c) Plot a graph of i(t) and explain the significance of each term.
SOLUTION
(a) With the particular values for L , C , R , and V in Figure 4.85, equation 4.15c becomes
d 2i di
+ 6 + 25i = 50 cos 5t.
dt 2 dt
(b) Since
di 25
= cos 5t + e−3t (5 cos 4t + 10 sin 4t)
dt 3
1
− e−3t (−20 sin 4t + 40 cos 4t)
3
25 1
= cos 5t + e−3t (−25 cos 4t + 50 sin 4t)
3 3
and
d 2i 125
2
=− sin 5t − e−3t (−25 cos 4t + 50 sin 4t)
dt 3
1
+ e−3t (100 sin 4t + 200 cos 4t)
3
125 1
=− sin 5t + e−3t (275 cos 4t − 50 sin 4t),
3 3
we find that
d 2i di 125 1
+ 6 + 25i = − sin 5t + e−3t (275 cos 4t − 50 sin 4t)
dt 2 dt 3 3
+ 50 cos 5t + 2e−3t (−25 cos 4t + 50 sin 4t)
125 25
+ sin 5t − e−3t (5 cos 4t + 10 sin 4t)
3 3
= 50 cos 5t.
Thus, i(t) does indeed satisfy i "" + 6i " + 25i = 50 cos 5t .
FIGURE 4.86 Current in an LCR-circuit
i
1.5
1
0.5
1 2 3 4 5 6t
−0.5
−1
−1.5
(c) The plot of i(t) in Figure 4.86 is composed of two functions, (5/3) sin 5t and
−(1/3)e−3t (5 cos 4t + 10 sin 4t) . Just after the switch is closed, both parts con-
tribute significantly to i(t) . Within a few seconds, however, the exponential factor
e−3t causes the second term of i(t) to become negligible. This is called the transient
part of the current; it persists for a very short time interval. The term (5/3) sin 5t
remains for all time, and once the transient part of the current becomes insignificant,
i(t) is essentially (5/3) sin 5t . This is called the steady-state part of the current.
EXAMPLE 4.38
When an inductor and capacitor are connected to an emf device (Figure 4.87), the charge on the
capacitor must satisfy equation 4.15b with R = 0,
d 2Q Q
L 2
+ = V.
dt C
(a) Verify that if the emf device is a battery producing constant voltage V , beginning at
time t = 0, and the capacitor has no initial charge, then the function
! ( )"
t
Q(t) = CV 1 − cos √
LC
satisfies this equation, and the conditions Q(0) = 0 and i(0) = 0.

(b) Draw a graph of Q(t) and interpret it in terms of charge on the capacitor and current
in the circuit.
FIGURE 4.87 Charge on capacitor in LC -circuit
V S
L
SOLUTION ! ( )"
dQ 1 t
(a) Since = CV √ sin √ , it follows that
dt LC LC
! ( )" ! ( )"
d 2Q Q CV t CV t
L 2 + =L cos √ + 1 − cos √ = V.
dt C LC LC C LC
Clearly, Q(0) = 0 and i(0) = Q" (0) = 0 also.
√
(b) To graph Q(t) , we first draw − cos (t/ LC) in Figure 4.88a. Shifting this curve
upward 1 unit and changing the scale on the vertical axis gives the graph of Q(t) in
Figure 4.88b.
When the circuit is closed at t = 0, there is no charge on the capacitor or current
in the circuit. The battery immediately begins charging the capacitor, but because
there is little charge on the capacitor, the voltage across its terminals is small. The
remainder of the voltage (making up V which is constant for all time) is across the
inductor. This is consistent with the fact that concavity is relatively large here and
voltage across the inductor is proportional to the second
√ derivative of Q , and the
second derivative is positive. As time approaches π LC/2, charge on the capacitor
approaches CV , what would normally
√ be its capacity if the inductor were not a part
of the circuit. At time t = π LC/2, voltage across the capacitor is V , and that
across the inductor vanishes. There is a point of inflection at which d 2 Q/dt 2 = 0.
Current is now at a maximum in the circuit and the capacitor continues to accumulate
charge. The voltage across the capacitor exceeds V , and therefore that across the
inductor is negative. This agrees with the√ fact that the curve is concave downward
and therefore d 2 Q/dt 2 < 0. At t = π LC , charge has reached a maximum value
of 2CV , voltage across the capacitor is 2V , and that across the inductor is −V .
Charge now begins to flow in the reverse direction √ and current (slope) is negative.
The capacitor discharges completely at t = 2π LC , and the cycle repeats.
√
FIGURE 4.88a Graph of − cos(t/ LC) FIGURE 4.88b Graph of charge on capacitor
1 Q
2CV
−1
EXAMPLE 4.39
The emf device in the RC -circuit of Figure 4.89 produces a constant voltage of V volts. If the
switch is closed at time t = 0 and then opened again at t = t0 , charge on the capacitor must
satisfy
dQ Q
R + = V [1 − h(t − t0 )], t > 0,
dt C
where h(t − t0 ) is the Heaviside unit step function of Section 2.5.
FIGURE 4.89 Charge on capacitor in RC -circuit
V S
(a) Verify that
Q(t) = CV [1 − e−t/(RC) ] − CV [1 − e−(t−t0 )/(RC) ]h(t − t0 )
satisfies this equation for all t $= t0 .

(b) Draw a graph of Q(t) . What is the initial charge on the capacitor?
(c) Are charge on the capacitor and current in the circuit continuous?
SOLUTION
(a) Since the derivative of the Heaviside function is zero for every t $= t0 ,
( ) ( )
dQ 1 −t/(RC) 1
= CV e − CV e−(t−t0 )/(RC) h(t − t0 )
dt RC RC
V ? −t/(RC) @
= e − e−(t−t0 )/(RC) h(t − t0 ) .
R
Hence,
dQ Q ? @ ? @
R + = V e−t/(RC) − e−(t−t0 )/(RC) h(t − t0 ) + V 1 − e−t/(RC)
dt C
? @
− V 1 − e−(t−t0 )/(RC) h(t − t0 )
= V [1 − h(t − t0 )].
(b) To graph Q(t) , we write it in the form

0
1 − e−t/(RC) , 0 ≤ t < t0
Q(t) = CV −(t−t0 )/(RC) −t/(RC)
e −e , t > t0
0
1 − e−t/(RC) , 0 ≤ t < t0
= CV ? @
− 1 − et0 /(RC) e−t/(RC) , t > t0 .
For 0 ≤ t < t0 , we first draw e−t/(RC) as in Figure 4.90a, turn it upside down, shift
it vertically one unit, and change the scale on the vertical axis (Figure 4.90b). As
t → t0− , the graph approaches CV [1 − e−t0 /(RC) ].
For t > t0 , the graph declines exponentially. As t → t0+ , it approaches
CV [1 − e−t0 /(RC) ], and it is asymptotic to the t -axis (Figure 4.90c). Combining
Figures 4.90b and c gives the final graph in Figure 4.90d. The capacitor has no initial
charge.
FIGURE 4.90a e−t/(RC) for 0 ≤ t ≤ t0 FIGURE 4.90b Charge on capacitor for 0 ≤ t < t0
Q
1
CV [1 − e −t0 /(RC)]
t0 t
t0 t
FIGURE 4.90c Charge on capacitor for t > t0 FIGURE 4.90d Charge on capacitor
Q Q
CV [1 − e −t0 /(RC)] CV [1 − e −t0 /(RC)]
t0 t t0 t
(c) If we define the value of the charge on the capacitor at t0 to be CV [1 − e−t0 /(RC) ],
it is continuous. Since current in the circuit is the slope of the curve, it is undefined
at t0 and therefore current is discontinuous at time t0 . It suddenly reverses direction
when the emf device is disconnected.
EXERCISES 4.10
∗ 1. If the switch in the RC -circuit shown to the right is closed at time (c) Draw a graph of the function in part (b) when Q0 = 0.
t = 0, equation 4.15b for charge on the capacitor becomes
dQ Q C
R + = V, t > 0.
dt C
(a) Verify that when V is constant, the function
Q(t) = De−t/(RC) + CV
S
V
satisfies the equation for any constant D .
(b) If Q0 is the charge on the capacitor when the switch is
closed, show that
Q(t) = CV [1 − e−t/(RC) ] + Q0 e−t/(RC) . R

∗ 2. The current i in the RC -circuit of Exercise 1 must satisfy the dif- (a) Verify that when V = A sin ωt , where A and ω are con-
ferential equation stants, the function
di i dV
R + = . t t A/ω
dt C dt Q(t) = D cos √ + E sin √ − sin ωt
LC LC 1
If V = V0 sin ωt , where V0 and ω are constants, verify that a solution ωL −
is ωC
V0
i = f (t) = Ae−t/(RC) + sin (ωt − φ), satisfies the equation for any constants D and E .
Z
(b) If Q0 is the charge on the capacitor when the switch is
where A is any constant whatsoever, and closed, show that
6 √
1 1 t A LC t A/ω
Z = R2 + , tan φ = − .
Q(t) = Q cos √ + sin √ − sin ωt.
ω2 C 2 ωCR 0
LC 1 LC 1
ωL − ωL −
∗ 3. If the switch in the LR -circuit below is closed at time t = 0, ωC ωC
equation 4.15a for current in the circuit becomes
C
di
L + Ri = V , t > 0.
dt
(a) Verify that when V is constant, the function
i(t) = De−Rt/L + V /R V S
satisfies the equation for any constant D .

(b) Using the fact that i(0) = 0, determine D , and draw a
graph of i(t) .
L
L
∗ 6. If the voltage source labelled V in the circuit of Exercise 5 is sud-
denly short-circuited by the switch labelled S , the current i in the circuit
thereafter must satisfy the equation
d 2i 1
V S L + i = 0,
dt 2 C
where L and C are the sizes of the inductor and capacitor, respectively,
and t is time. Verify that
( ) ( )
R t t
i = f (t) = A cos √ + B sin √
∗ 4. The current i in the LR -circuit in the circuit of Exercise 3 must LC LC
satisfy the differential equation
satisfies this equation for any constants A and B whatsoever.
di ∗ 7. (a) If V = A sin ωt , where A and ω are constants, in Exercise
L + Ri = V .
dt 1, and there is no charge on the capacitor at time t = 0,
verify that
If V is as in Exercise 2, verify that a solution is
CA ωRC 2 A
−Rt/L V0 Q(t) = sin ωt + [e−t/(RC) − cos ωt ]
i = f (t) = Ae + sin (ωt − φ), 1 + ω2 R 2 C 2 1 + ω2 R 2 C 2
Z
satisfies RQ" + Q/C = V .
where A is any constant, and
(b) Show that Q(t) can be expressed in the form
/ ωL
Z = R 2 + ω2 L2 , tan φ = . ωRC 2 A −t/(RC) A/ω
R Q(t) = e + cos (ωt − φ),
1 + ω2 R 2 C 2 Z
∗ 5. If the switch in the following LC -circuit is closed at time t = 0,
equation 4.15b for charge on the capacitor becomes where
6
d 2Q Q 1 1
L + = V, t > 0. Z = R2 + and tan φ = − .
dt 2 C ω2 C 2 ωCR
∗ 8. (a) If V = A sin ωt , where A and ω are constants, in Exercise ∗ 10. An inductor L , a resistor R , and a capacitor C are connected with
3, verify that a generator, producing an oscillatory voltage V = V0 cos ωt , for t ≥ 0
ωLA RA (figure below). If L , C , R , V0 , and ω are all constants, steady-state
i(t) = (e−Rt/L − cos ωt) + 2 sin ωt current i(t) in the circuit is given by
R2
+ω L 2 2 R + ω2 L2
satisfies Li " + Ri = V .
(b) Show that i(t) can be expressed in the form ! ( ) "
V0 1
ωLA A i(t) = # $2 R cos ωt + ωL − sin ωt .
1
i(t) = e−Rt/L + sin (ωt − φ), R 2 + ωL − ωC
ωC
R 2 + ω2 L2 Z
where
/ ωL
Z = R 2 + ω2 L2 and tan φ = . Find the value of ω that makes the amplitude of the current a maximum.
R Do this in two ways:
∗ 9. The current in the RCL -circuit below must satisfy equation 4.15c.
If V = A sin ωt , where A and ω are constants, verify that a solution
is (a) Express i(t) in the form
−Rt/(2L) A
i(t) = e (D cos νt + E sin νt) + sin (ωt − φ),
Z
C V0
i(t) = cos (ωt − φ),
Z
S
in which case amplitude V0 /Z must be maximized.
R
(b) Find critical points for i(t) as given.
L
where D and E are any constants whatsoever, and C
6
1 R2
ν = − ,
LC 4L2
; S
( )2
1 R
Z = R 2 + ωL − ,
ωC
1
ωL −
tan φ = ωC .
R L
4.11 Indeterminate Forms and L’Hôpital’s Rule

Derivatives are instantaneous rates of change, defined as limits of average rates of change. As a
result, it was necessary to discuss limits in Chapter 2 prior to the introduction of derivatives in
Section 3.1. Now that we have derivatives, it may seem quite surprising that they can be used
to evaluate many limits.
The Indeterminate Form 0/0

If we let x approach zero in numerator and denominator of the limit
√
1+x−1
lim √ , (4.16)
x→0+ x
we find that
#√ $ √
lim 1+x−1 = 0 and lim x = 0.
x→0+ x→0 +
We say that limit 4.16 is of the indeterminate form 0/0. Similarly, the limit
x 3 − x 2 − 8x + 12
lim (4.17)
x→2 x 2 − 4x + 4
is of the indeterminate form 0/0 since both numerator and denominator approach zero as x
approaches 2.
In Section 2.1 we evaluated limit 4.16 by rationalizing the numerator (see Example 2.6).
Factoring numerator and denominator in limit 4.17 gives
x 3 − x 2 − 8x + 12 (x − 2)2 (x + 3)
lim = lim = lim (x + 3) = 5.
x→2 x 2 − 4x + 4 x→2 (x − 2)2 x→2
These two examples illustrate the “trickery” to which we resorted in Chapter 2 in order to
evaluate limits. With Cauchy’s generalized mean value theorem from Section 3.14, however,
we can prove a result called L’Hôpital’s rule, which makes evaluation of many limits of the
indeterminate form 0/0 quite simple.
THEOREM 4.4 (L’Hôpital’s rule)

Suppose functions f (x) and g(x) satisfy the following conditions:
1. f (x) and g(x) are differentiable in an open interval I except possibly at the point
x = a in I ;
2. g " (x) $= 0 in I except possibly at x = a ;
3. lim f (x) = 0 = lim g(x) ;
x→a x→a
f " (x)
4. lim " = L.
x→a g (x)
Then,
f (x)
lim = L.
x→a g(x)
PROOF We define two functions F (x) and G(x) that are identical to f (x) and g(x) on I
but have value zero at x = a :
0 0
f (x), x $= a g(x), x $= a
F (x) = G(x) =
0, x = a, 0, x = a.
Since F " (x) = f " (x) and G" (x) = g " (x) for x in I except possibly at x = a , F (x) and
G(x) are therefore differentiable on I except possibly at x = a . In addition, differentiability
of a function implies continuity of the function (Theorem 3.6), so that F (x) and G(x) must
certainly be continuous on I except possibly at x = a . But
lim F (x) = lim f (x) = 0 = F (a),

x→a x→a
and the same is true for G(x) ; hence, F (x) and G(x) are continuous for all x in I . Con-
sequently, F (x) and G(x) are identical to f (x) and g(x) in every respect, except that they
have been assigned a value at x = a to guarantee their continuity there. This extra condition
permits us to apply Cauchy’s generalized mean value theorem to F (x) and G(x) on the interval
between a and x , as long as x is in I . There exists a number c between a and x such that
F (x) − F (a) F " (c)

= " ,
G(x) − G(a) G (c)
or since F (a) = G(a) = 0,

F (x) F " (c)
= " .
G(x) G (c)
Since x and c are points in I , we can also write that F (x) = f (x) , F " (c) = f " (c) ,
G(x) = g(x) , G" (c) = g " (c) , and therefore
f (x) F (x) F " (c) f " (c)

= = " = " .
g(x) G(x) G (c) g (c)
If we now let x approach a , then c must also approach a since it is always between a and x .
Consequently,
f (x) f " (c) f " (x)
lim = lim " = lim " ,
x→a g(x) c→a g (c) x→a g (x)
and if
f " (x)
L = lim ,
x→a g " (x)
it follows that
f (x)
lim = L.
x→a g(x)
This theorem is also valid if L is replaced by ∞ or −∞ . The only difference in the proof
is to make the same change in the last sentence.
Theorem 4.4 is also valid if x → a is replaced by either a right-hand limit, x → a + , or
a left-hand limit, x → a − . The only difference in these cases is that interval I is replaced by
open intervals a < x < b and b < x < a , respectively, and the proofs are almost identical.
In addition, the following theorem indicates that x → a can be replaced by x → ∞ (or
x → −∞ ).
THEOREM 4.5 (L’Hôpital’s rule)

Suppose functions f (x) and g(x) satisfy the following conditions:
1. f (x) and g(x) are differentiable for some interval x > b > 0;
2. g " (x) $= 0 for x > b > 0;
3. lim f (x) = 0 = lim g(x) ;
x→∞ x→∞
f " (x)
4. lim " =L (or ±∞ ).
x→∞ g (x)
Then,
f (x)
lim =L (or ±∞ ).
x→∞ g(x)
In other words, L’Hôpital’s rule applies to any type of limit that yields the indeterminate
form 0/0 (be it x → a , x → a + , x → a − , x → ∞ , or x → −∞ ). A common error
when using L’Hôpital’s rule is to differentiate f (x)/g(x) with the quotient rule and then take
the limit of the resulting derivative. L’Hôpital’s rule calls for the limit of f " (x)/g " (x) ; f (x)
and g(x) are differentiated separately.
If we use L’Hôpital’s rule on limit 4.16, we find that
1
√ √ √
1+x−1 2 1+x x
lim √ = lim+ = lim+ √ = 0.
x→0+ x x→0 1 x→0 1+x
√
2 x
For limit 4.17, we have
x 3 − x 2 − 8x + 12 3x 2 − 2x − 8
lim = lim ,
x→2 x 2 − 4x + 4 x→2 2x − 4
which is still a limit of the indeterminate form 0/0. Note that this is a conditional equation; that
is, it says that the limit on the left is equal to the limit on the right, provided that the limit on the
right exists. If we apply L’Hôpital’s rule a second time, to the limit on the right, we obtain
x 3 − x 2 − 8x + 12 6x − 2
lim 2
= lim = 5.
x→2 x − 4x + 4 x→ 2 2
EXAMPLE 4.40
Evaluate the following limits:
3+x tan x
(a) lim √ √ (b) lim
x→−3 3 − −x x→0 x
x−4 x 3 − 4x 2 + 9x − 36
(c) lim (d) lim
x→4 x 2 − 8x + 16 x→4 x2 + 5
SOLUTION
(a) Since we have the indeterminate form 0/0, we use L’Hôpital’s rule to write
3+x 1 √ √
lim √ √ = lim = lim 2 −x = 2 3.
x→−3 3 − −x x→−3 1 x→−3
√
2 −x
(b) If we use L’Hôpital’s rule, we have
tan x sec2 x
lim = lim = 1.
x→0 x x→0 1
(c) By L’Hôpital’s rule,
x−4 1
lim = lim .
x→4 x 2 − 8x + 16 x→4 2x − 8
Since
1 1
lim =∞ and lim = −∞,
x→4+ 2x − 8 x→4− 2x − 8
we conclude that
x−4 x−4
lim =∞ and lim = −∞.
x→4+ x2 − 8x + 16 x→4− x2 − 8x + 16
(d) This limit is not of the indeterminate form 0/0 since limx→4 (x 2 + 5) = 21; thus
we cannot use L’Hôpital’s rule. Since limx→4 (x 3 − 4x 2 + 9x − 36) = 0,
x 3 − 4x 2 + 9x − 36
lim = 0.
x→4 x2 + 5
Had we used L’Hôpital’s rule in part (d) of this example, we would have obtained an incorrect
answer:
x 3 − 4x 2 + 9x − 36 3x 2 − 8x + 9 25
lim 2
= lim = .
x→4 x +5 x→4 2x 8
In other words, L’Hôpital’s rule is not to be used indiscriminately; it must be used only on the
indeterminate forms for which it is designed.
The Indeterminate Form ∞/∞

The limit √
x−1 1+
lim
x→∞ 2x + 5
is said to be of the indeterminate form ∞/∞ since numerator and denominator become
increasingly large as x → ∞ . Theorems 4.4 and 4.5 for L’Hôpital’s rule can be adapted to this
indeterminate form also; hence we calculate that
1
√ √
1+ x−1 2 x−1 1
lim = lim = lim √ = 0.
x→∞ 2x + 5 x→∞ 2 x→∞ 4 x − 1
EXAMPLE 4.41
Evaluate the following limits, if they exist:
√
x2 2x 2 + 3x + 2
(a) lim x (b) lim
x→∞ e x→−∞ 1−x
SOLUTION
(a) Since this limit exhibits the indeterminate form ∞/∞ , we use L’Hôpital’s rule to
write
x2 2x
lim x = lim x .
x→∞ e x→∞ e
Since this limit is still of the form ∞/∞ , we use L’Hôpital’s rule again:
x2 2
lim= lim x = 0.
x→∞ ex x→∞ e
xn
The same result would occur for any positive power n on x ; that is, lim = 0.
x→∞ ex
What this shows is that exponential functions grow more rapidly for large x than
power functions.
(b) By L’Hôpital’s rule,
4x + 3
√ √
2x 2+ 3x + 2 2 2x 2 + 3x + 2 −(4x + 3)
lim = lim = lim √ .
x→−∞ 1−x x→−∞ −1 x→−∞ 2 2x 2 + 3x + 2
This limit is also of the indeterminate form ∞/∞ . Further applications of L’Hôpital’s
rule do not lead to a simpler form for the limit. Thus, L’Hôpital’s rule does not prove
advantageous on this limit. It is better to divide numerator and denominator by x :
√ 6
√ 2x 2 + 3x + 2 3 2
− 2+ +
lim
2x 2 + 3x + 2
= lim x = lim
x x 2 = √2.
x→−∞ 1−x x→−∞ 1 x→−∞ 1
−1 −1
x x
EXAMPLE 4.42
Show that L’Hôpital’s rule cannot be used to evaluate
x − cos x
lim .
x→∞ x
What is the value of the limit?
SOLUTION The limit is of the indeterminate form ∞/∞ . If we apply L’Hôpital’s rule we
obtain
x − cos x 1 + sin x
lim = lim .
x→∞ x x→∞ 1
But this limit does not exist, and therefore L’Hôpital’s rule has failed. But we do not need the
rule since division of numerator and denominator by x gives
x − cos x - cos x .
lim = lim 1 − = 1.
x→∞ x x→∞ x
The Indeterminate Form 0 · ∞

The limits
lim xe−2x and lim x 2 ln x
x→∞ x→0+
are said to be of the indeterminate form 0 · ∞ . L’Hôpital’s rule can again be used if we first
rearrange the limits into one of the forms 0/0 or ∞/∞ :
x 1
lim xe−2x = lim = lim = 0;
x→∞ x→∞ e 2x x→∞ 2e2x
1 ( 2)
2 ln x x x
lim x ln x = lim = lim+ = lim+ − = 0.
x→0 + x→0 + 1 x→0 − 2 x→0 2
x2 x3
Note that had we converted the second limit into the 0/0 form, we would have had
x2 2x
lim x 2 ln x = lim = lim+ = lim+ −2x 2 (ln x)2 .
x→0+ x→0+ 1 x→0 −1 x→0
ln x x(ln x) 2
Although this is correct, the limit on the right is more difficult to evaluate than the original. In
other words, we must be judicious in converting a limit from the 0 · ∞ indeterminate form to
either 0/0 or ∞/∞ .
EXAMPLE 4.43
Evaluate the following limits if they exist:
(a) lim (x − π/2) sec x (b) lim xe1/x

x→π/2 x→0+
SOLUTION
x − π/2 1
(a) lim (x − π/2) sec x = lim = lim = −1
x→π/2 x→π/2 cos x x→π/2 − sin x
e1/x e1/x (−1/x 2 )

(b) lim xe1/x = lim+ = lim+ = lim+ e1/x = ∞
x→0+ x→0 1/x x→0 −1/x 2 x→0
The Indeterminate Forms 00 , 1∞ , ∞0 , and ∞ − ∞

Various other indeterminate forms arise in the evaluation of limits, and many of these can be
reduced to the 0/0 and ∞/∞ forms by introducing logarithms. In particular, the limits
( )x 2
1
lim x x , lim 1+ , lim (sec x)cos x , and lim (sec x − tan x) (4.18)
x→0+ x→∞ x x→π/2− x→π/2
are said to display the indeterminate forms 00 , 1∞ , ∞0 , and ∞ − ∞ , respectively. To evaluate

limx→0+ x x , we set
L = lim x x
x→0+
and take natural logarithms of both sides,

( )
x
ln L = ln lim x .
x→0+
As the logarithm function is continuous, we may interchange the limit and logarithm operations
(see Theorem 2.5),
ln x
ln L = lim (ln x x ) = lim x ln x = lim .
x→0+ x→0+ x→0+ 1
x
We are now in a position to use L’Hôpital’s rule:
ln L = lim x = lim (−x) = 0.

x→0+ −1 x→0+
x 2
Exponentiation of both sides of ln L = 0 now gives L = e0 = 1; that is,
lim x x = 1.
x→0+
For the second limit in 4.18 we again set

( )x 2
1
L = lim 1+
x→∞ x
and take natural logarithms:
9 ( )x 2 : 9 ( )x 2 :
1 1
ln L = ln lim 1+ = lim ln 1 +
x→∞ x x→∞ x
 ( )
x+1
! ( )" ln
1  x 
= lim x 2 ln 1 + = lim 

.

x→∞ x x→∞ 1
x 2
By L’Hôpital’s rule, we have
 ( )
x −1
 x + 1 x2  x2
ln L = lim   = lim = ∞.
x→∞  −2  x→∞ 2(x + 1)
x3
Consequently,
( )x 2
1
L = lim 1+ = ∞.
x→∞ x
In the third limit of 4.18, we set
L= lim (sec x)cos x ,

x→π/2−
in which case
! "
ln L = ln lim
−
(sec x)cos x = lim [cos x ln (sec x)]
x→π/2 x→π/2−
 1 
! " sec x tan x
ln (sec x)  
= lim = lim −  sec x 
x→π/2− sec x x→π/2 sec x tan x
= lim cos x = 0.
x→π/2−
Thus,
L= lim (sec x)cos x = e0 = 1.
x→π/2−
Finally, the last limit in 4.18 is evaluated by rewriting it in the 0/0 form,
( )
1 − sin x − cos x
lim (sec x − tan x) = lim = lim = 0.
x→π/2 x→π/2 cos x x→π/2 − sin x
EXAMPLE 4.44
Plot a graph of the function f (x) = x 2 ln x . Find limits of f (x) and f " (x) as x → 0+ .
Where is the point of inflection on the graph?
FIGURE 4.91 Plot of x 2 ln x
y
0.8
0.6 y = x 2 ln x
0.4
0.2
0.5 1 1.5 2 x
−0.2
SOLUTION The plot in Figure 4.91 suggests that f (x) and f " (x) both approach 0 as x →
0+ . To confirm this we use L’Hôpital’s rule to calculate
1
ln x x
lim f (x) = lim x 2 ln x = lim = lim+ = lim+ (−2x 2 ) = 0−
x→0+ x→0+ x→0+ 1 x→0 2 x→0
−
x2 x3
and
( ) 1
x2 ln x x
lim f " (x) = lim 2x ln x + = 2 lim+ = 2 lim+ = 2 lim+ (−x) = 0− .
x→0+ x→0+ x x→0 1 x→0 1 x→0
−
x x2
For the point of inflection between x = 0 and x = 1/2, we solve
2x
0 = f "" (x) = 2 ln x + + 1 = 2 ln x + 3.
x
The only solution is x = e−3/2 . Since f "" (x) changes sign as x passes through e−3/2 , there is
a point of inflection at (e−3/2 , −3e−3 /2) .
EXERCISES 4.11
In Exercises 1–42 evaluate the limit, if it exists. (1 − 1/x)3 sin (1/x)
11. lim
√ 12. lim
x→1+ x−1 x→∞ 1/x 2
2
x + 3x 2
x −9 √ √ √
1. lim 2. lim x−3 5+x− 5−x
x→0 x 3 + 5x 2 x→3 x − 3 13. lim √ 14. lim
x→9− 9−x x→0 x
x 3 + 3x − 2 2x 2 + 3x x − sin x x n − an
3. lim 4. lim 15. lim 16. lim
x→−∞ x 2 + 5x + 1 x→∞ 5x 3 + 4 x→0 x3 x→a x−a
2
x − 10x + 25 1 (1 − cos x) 2
tan x
x→5 x 3 − 125 x→1 (x − 1)2 x→0 3x 2 x→0 x
√ # √ $3/2
x2 + 1 sin x sin 3x 1− 2−x
x→∞ 2x + 5 x→−∞ x x→0 tan 2x x→1 x−1
√ √
sin (2/x) cos x x + 1 − 2x + 1
9. lim 10. lim ∗ 21. lim √ √
x→∞ sin (1/x) x→π/2 (x − π/2)2 x→0 3x + 4 − 2x + 4
(1 − cos x)2 ∗ 58. Planck’s law for the energy density ψ of blackbody radiation states
∗ 22. lim
x→0 3x 4 that
( ) kλ−5
1 (x − 2)10 ψ = ψ(λ) = ,
∗ 23. lim x sin ∗ 24. lim #√ √ $10 ec/λ − 1
x→∞ x x→2 x− 2
( ) where k and c are positive constants and λ is the wavelength of the
4 2 radiation.
∗ 25. lim − ∗ 26. lim xex
x→0 x 2 1 − cos x x→∞ (a) Show that lim ψ(λ) = 0 and lim ψ(λ) = 0.
( ) λ→0+ λ→∞
4
∗ 27. lim x 2 e−4x ∗ 28. lim x sin (b) Show that ψ(λ) has one critical point that must satisfy
x→∞ x→−∞ x the equation (5λ − c)ec/λ = 5λ . Find the critical point
∗ 29. lim x cot x ∗ 30. lim csc x(1 − cos x) accurate to seven decimal places when c = 0.000 143 86.
x→0 x→0
(c) Draw a graph of the function ψ(λ) when c = 0.000 143 86
∗ 31. lim (sin x) x
∗ 32. lim x sin x and k = 1.
x→0+ x→0+
( )x ∗ 59. The following limit arises in the calculation of the electric field
x+5 intensity for a half-wave antenna: lim f (θ) , where
∗ 33. lim ∗ 34. lim (1 + x)cot x θ →0
x→∞ x+3 x→0
0 B
∗ 35. lim x 1/x ∗ 36. lim | ln x|sin x sin [π/2(cos θ − 1)] sin [π/2(cos θ + 1)]
x→∞ x→0+ f (θ) = sin θ + .
cos θ − 1 cos θ + 1
1/x
∗ 37. lim xe ∗ 38. lim (tan x − csc x) Evaluate this limit by first showing that f (θ) can be written in the form
x→0+ x→0
( ) -π .
x 1 2 cos cos θ
∗ 39. lim (csc x − cot x) ∗ 40. lim − 2
x→0 x→1 ln x x ln x f (θ) = ,
( ) ( ) sin θ
x 1 1 1
∗ 41. lim − ∗ 42. lim − and then using L’Hôpital’s rule.
x→1 x − 1 ln x x→0 x 2 sin2 x
∗ 60. Find all values of a , b , and c for which
In Exercises 43–54 draw a graph of the function. eax − bx − cos (x + cx 2 )
lim = 5.
∗ 43. f (x) = xe−2x ∗ 44. f (x) = x 2 e3x x→0 2x 3 + 5x 2
∗ 45. f (x) = xe−x

2
∗ 46. f (x) = e1/x ∗ 61. The maximum flow rate of gas through a nozzle is governed by the
function ( )(x+1)/(x−1)
ln x 2
∗ 47. f (x) = ∗ 48. f (x) = x 2 ln x f (x) = x .
x x+1
x2 (a) Plot a graph of f (x) on the interval 0 ≤ x ≤ 20. Did you
∗ 49. f (x) = xe1/x ∗ 50. f (x) =
ln x get any error messages? Should you have? Show that the
∗ 51. f (x) = x x , x>0 ∗ 52. f (x) = x 10 e−x function is discontinuous at x = 1, but limx→1 f (x) =
1/e .
∗ 53. f (x) = e−x ln x, x>0
(b) Plot the function on the interval 0 ≤ x ≤ 200. Does the
∗ 54. f (x) = 2 csc x − cot x, 0 < x < π/2 graph appear to have a horizontal asymptote? Show that
( ) limx→∞ f (x) = 2.
x + a cx
∗ 55. Evaluate lim for any constants a , b , and c . ∗∗ 62. (a) Sketch a graph of the function
x→∞ x + b
∗ 56. The indeterminate forms 00 , 1∞ , and ∞0 are often evaluated by 0 2

e−1/x , x =
$ 0
introducing logarithms. Show that the limit lim (x − ln x) can be f (x) =
x→∞ 0, x = 0.
evaluated by introducing exponentials.
∗ 57. When an electrostatic field E is applied to a gaseous or liquid polar (b) Show that for every positive integer n ,
dielectric, a net dipole moment P per unit volume is set up, where 2
e−1/x
lim = 0.
eE + e−E 1 x→0 x n
P (E) = − .
eE − e−E E
(c) Prove by mathematical induction that f (n) (0) = 0, where
Show that limE→0+ P (E) = 0. f (n) (0) is the nth derivative of f (x) evaluated at x = 0.
4.12 Differentials
In Section 3.1 we pointed out that the notation dy/dx for the derivative of a function y = f (x)
should not be considered a quotient. Beginning in Chapter 5, however, it is essential that we be
able to do this, and therefore in this section we define “differentials” dx and dy so that dy/dx
can be regarded as a quotient.
When we use the notation
dy f (x + /x) − f (x)
= lim
dx /x→ 0 /x
for the derivative of a function y = f (x) , we call /x an increment in x . It represents a

change in the value of the independent variable from some value x to another value x + /x .
This change can be positive or negative depending on whether we want x + /x to be larger
or smaller than x . When the independent variable changes from x to x + /x , the dependent
variable changes by an amount /y , where
/y = f (x + /x) − f (x). (4.19)
In other words, /y is the change in y resulting from the change /x in x . For the function in
Figure 4.92, /y is positive when /x is positive, and /y is negative when /x is negative.
FIGURE 4.92a dy and /y for dx > 0 FIGURE 4.92b dy and /y for dx < 0
y y
y = f (x) y = f (x)
f (x +/ x)
/y dy
f (x) f (x)
/y dy
f (x + /x)
dx
dx
x x +/x = x + dx x x + /x x x
/x /x
For example, when y = x 2 − 2x , the change /y in y when x is changed from 3 to 3.2 is
/y = [(3.2)2 − 2(3.2)] − [32 − 2(3)] = 0.84.
The function increases by 0.84 when x increases from 3 to 3.2.

For purposes of integration, a topic that begins in Chapter 5 and continues in every chap-
ter thereafter, an alternative notation for an increment in the independent variable x is more
suggestive.
DEFINITION 4.7
An increment /x in the independent variable x is denoted by
dx = /x, (4.20)
and when written as dx , it is called the differential of x .

The differential dx is synonymous with the increment /x ; it represents a change in x

(in most applications a very small change). The differential of the dependent variable is not
synonymous with /y .
DEFINITION 4.8
The differential of y = f (x) , corresponding to the differential dx in x , is denoted by
dy and is defined by
dy = f " (x) dx. (4.21)
The difference between /y and dy is most easily seen in Figures 4.92. We know that /y
is the exact change in the function y = f (x) when x is changed by an amount /x or dx .
It is the difference in the height of the curve at x and x + /x = x + dx . Now the slope of
the tangent line to the graph at the point (x, y) is f " (x) . Definition 4.8 indicates that dy can
be interpreted as the difference in the height of this tangent line at x and at x + dx . In other
words, dy is the change in y corresponding to the change dx in x if we follow the tangent line
to y = f (x) at (x, y) rather than the curve itself.
Figures 4.92 also suggest that when dx is very small (close to zero), dy is approximately
equal to /y ; that is,
dy ≈ /y when dx ≈ 0.
This is illustrated in the following numerical example.
EXAMPLE 4.45
√
Find /y and dy for the function y = f (x) = x 2 + 1 when x = 2 and dx = 0.1.
SOLUTION According to equation 4.19, the change in y as x increases from 2 to 2.1 is
/ /
/y = f (2.1) − f (2) = (2.1)2 + 1 − 22 + 1 = 0.089 87.
√
Since f " (x) = x/ x 2 + 1, the differential of y for x = 2 and dx = 0.1 is
2
dy = f " (2)(0.1) = √ (0.1) = 0.089 44.
22 + 1
The difference between dy and /y is, therefore, 0.000 43, a difference of 43 parts in 8987.
Before the invention of electronic calculators, differentials were used to approximate a function
near points at which it was easily evaluated. For example, imagine trying to evaluate the function
f (x) = x 1/3 at x = 126 without a calculator. We could use differentials to approximate 1261/3
as follows:
1261/3 = f (126) = f (125) + /y ≈ f (125) + dy

1 1 376
= 5 + f " (5)(1) = 5 + = 5+ = .
3(125)2/3 75 75
With the advent of the electronic calculator, problems of this type are archaic. On the other
hand, differentials are indispensable when we examine changes in a function without specifying
values for the independent variable. Very prominent in this context are relative and percentage
changes.
When a quantity y undergoes a change /y , then its relative change is defined as

/y
, (4.22)
y
and its percentage change is given by
/y
100 . (4.23)
y
Sometimes relative and percentage changes are more important than actual changes. To
illustrate this, consider the function V = 4π r 3 /3, which represents the volume of a sphere. If
the radius of the sphere is increased from 0.10 m to 0.11 m, then the change in the volume of
the sphere is
4 4
/V = π(0.11)3 − π(0.10)3 = 4.4π × 10−4 m 3 .
3 3
This is not a very large quantity, but in relation to the original size of the sphere, we have a
relative change of
/V 4.4π × 10−4
= = 0.33
V 4π(0.10)3 /3
and a percentage change of
/V
100 = 33%.
V
Suppose the same increase of 0.01 m is applied to a sphere with radius 100 m. The change
in the volume is
4 4
/V = π(100.01)3 − π(100)3 = 4.0π × 102 m 3 .
3 3
This is quite a large change in volume, but the relative change is
/V 4.0π × 102
= = 3.0 × 10−4 ,
V 4π(100)3 /3
and the percentage change is
/V
100 = 0.03%.
V
Although the change 400π in V when r = 100 is much larger than the change 0.000 44π
when r = 0.1, the relative and percentage changes are much smaller when r = 100. We see,
then, that in certain cases it may be relative and percentage changes that are significant rather
than actual changes.
In the example above, when r = 100 m, the change dr = 0.01 is certainly small compared
to r . Therefore, we should be able to use the differential dV = V " (r) dr = 4π r 2 dr to
approximate /V . With r = 100 and dr = 0.01, we have
dV = 4π(100)2 (0.01) = 4.0π × 102 m 3 .
(To two significant figures, dV is equal to /V .) Thus, relative change in V is approximately

equal to dV /V and percentage change 100dV /V . So, for small changes in an independent
variable, the differential of the dependent variable may be used in place of its increment in the
calculation of relative and percentage changes; that is, equations 4.22 and 4.23 can be replaced
by
dy dy
and 100 .
y y
We do this in the following example.
EXAMPLE 4.46
When a pendulum swings, the frequency (number of cycles per second) of its oscillations is
given by
6
g
f = h(l) = 2π ,
l
where l is the length of the pendulum and g > 0 is the acceleration due to gravity. If the length
of the pendulum is increased by 41 %, calculate the approximate percentage change in f .
SOLUTION The approximate change in f is given by
( )
" 1 −3/2
√ √ dl
df = h (l) dl = 2π g − l dl = −π g 3/2 ;
2 l
hence the approximate percentage change in f is

( √ ) ( 1/2 ) ( )
df −π g dl l 1 dl
100 = 100 √ = − 100 .
f l 3/2 2π g 2 l
But because l increases by 41 %, it follows that 100(dl/ l) = 1/4, and
df 1
100 =− .
f 8
Therefore, the frequency changes by − 18 %, the negative sign indicating that because l increases,
f decreases.
FIGURE 4.93 Differen-
tials do not approximate changes at
critical points
The differential dy cannot always be used as an approximation for the actual change /y in a
y function y = f (x) . Sometimes it cannot be used even when dx is very close to zero. For
example, if f (x) = 2x 3 + 9x 2 − 24x + 6, then
dy = f " (x) dx = (6x 2 + 18x − 24) dx.
If x is changed from 1 to 1.01, then the approximate change in y as predicted by the differential
1 x is
dy = (6 + 18 − 24)(0.01) = 0.
In fact, for any dx whatsoever, we find that dy = 0. Geometrically speaking, we can see
why. Since f " (1) = 0, x = 1 is a critical point of f (x) (Figure 4.93); therefore, dy , which
is the tangent line approximation to /y , will always be zero. We cannot use differentials to
(1, −7)
approximate function changes at critical points.
The differential for the function y = f (x) = x 100 is
dy = 100x 99 dx.
If x is changed by 1% from x = 1 to x = 1.01, then dx = 0.01 and
dy = 100(1)99 (0.01) = 1.0.
The actual change in y is

/y = (1.01)100 − 1100 = 1.7.
We would hardly regard this dy as a very good approximation for /y even though the change
in x is only 1%.
The latter example raises the question “How small, in general, must dx be in order that
dy be a reasonable approximation for /y ?” This is not a simple question to answer. We
will discuss approximations in more detail in Chapter 10, and then be able to answer a more
important question: How good an approximation to /y is dy ? After all, it is not much use to
say that dy ≈ /y if we cannot say to how many decimal places the approximation is accurate.
Suffice it to say now that use of the differential dy to approximate /y is to be regarded with
some reservation. This is not to say that differentials are useless. We will see in Chapters 5–7
that differentials are indispensable to the topic of integration.
We make one last comment before leaving this section. If equation 4.21 is divided by
differential dx , then
dy
= f " (x).
dx
Now, the left side of this equation is the differential of y divided by the differential of x . The
quotient of differentials dy and dx is equal to the derivative f " (x) . The entity dy/dx can
henceforth be regarded either as “the derivative of y with respect to x ” or as “ dy divided by
dx ,” whichever is appropriate for the discussion at hand.
EXERCISES 4.12
In Exercises 1–10 find dy in terms of x and dx . sidered a perfect sphere (radius 6.37 × 106 m), then this law predicts a
gravitational attraction of 9.81m newtons on a mass m on its surface.
x+1 Use differentials to determine the height above the surface of the earth
1. y = x 2 + 3x − 2 2. y =
x−1 at which the gravitational attraction decreases to 9.80m newtons.
/
3. y = x 2 − 2x 4. y = sin (x 2 + 2) − cos x ∗ 17. When a force F is applied to the object of mass m in the figure
/
5. y = x 1/3 − x 5/3 6. y = x 3 3 − 4x 2 below, three other forces act on m : the force of gravity directly down-
ward, a reactional force of the supporting surface, and a horizontal
x 3 − 3x 2 + 3x + 5
7. y = x 2 sin x 8. y =
x 2 − 2x + 1 F
C 2
√ x (x − 2)
9. y = 1 + 1 − x 10. y = 3
x + 5x
m
∗ 11. The momentum M and kinetic energy K of a mass m moving
1
with speed v are given by M = mv and K = mv 2 . If v is changed
2 frictional force opposing motion. The least force that will overcome
by 1%, what are approximate percentage changes in M and K ?
friction and produce motion is given by
∗ 12. The magnitude of the gravitational force of attraction between two
point masses m and M is given by 9.81µm
F = ,
GmM cos θ + µ sin θ
F = ,
r2
where µ is a constant called the coefficient of static friction. Use
where G > 0 is a constant and r is the distance between the masses. differentials to calculate the approximate percentage change in F if θ
If r changes by 2%, by how much does F change approximately? is increased by 2% from an angle of π/4 radians.
∗ 13. According to Example 1.9, the range of a shell fired from an ar-
tillery gun with velocity v at angle θ is given by R = (v 2 sin 2θ )/9.81. ∗ 18. The volume of a right circular cylinder is V = π r 2 h , where r is
Use differentials to find the approximate percentage change in R if θ the radius and h is the height. Use differentials to show the following:
is increased by 1% from an angle of π/3 radians. (a) If h can be measured exactly, but r is subject to an error of
∗ 14. According to Example 1.9, the maximum height attained by a shell a %, the error in V is 2a %.
fired from an artillery gun with velocity v at angle θ is given by H =
(b) If r can be measured exactly, but h is subject to an error of
(v 2 sin2 θ )/19.62. Use differentials to find the approximate percentage
b %, the error in V is b %.
change in H if θ is increased by 2% from an angle of π/3 radians.
(c) What is the maximum percentage error in V if r is subject
∗ 15. Under adiabatic expansion, a gas obeys the law P V 7/5 = a constant,
to an error of a % and h is subject to an error of b %?
where P is pressure and V is volume. If the pressure is increased by
2%, find the approximate percentage change in the volume. ∗ 19. Use differentials to show that if y = x n , where n is a nonzero
∗ 16. The magnitude of the gravitational force of attraction between two constant, and x is subject to an error of a %, then the resulting error in
point masses m and M is defined in Exercise 12. If the earth is con- y is na %.
∗ 20. Suppose that z = x n y m , where n and m are nonzero constants. ∗ 22. A prism can be used to measure the index of refraction n of the
material in the prism. According to Exercise 73 in Section 4.7, n is
(a) If x is subject to an error of a %, but y is subject to no error, given by the formula
what is the resulting error in z ?
sin [(ψm + γ )/2]
(b) If y is subject to an error of b %, but x is subject to no error, n= .
sin (γ /2)
what is the resulting error in z ?
Use differentials to find the approximate percentage error in n if the
(c) What is the maximum percentage error in z if x and y are measurement of ψm can be out by 1% when ψm = π/6 and γ = π/3.
subject to errors of a % and b %, respectively? Assume that γ is known exactly.
∗ 23. Repeat Exercise 22 if ψm is known precisely but the measurement
∗ 21. Repeat Exercise 20 if z = x n /y m . of γ can be out by 1%.
SUMMARY
In this chapter we discussed a number of applications of differentiation, the first of which was
Newton’s iterative procedure for approximating the roots of equations. It is perhaps the most
popular of all approximation methods, because of its speed, simplicity, and accuracy.
In Section 4.3 we defined a critical point of a function as a point in its domain where its first
derivative either vanishes or does not exist. Geometrically, this corresponds to a point where the
graph of the function has a horizontal tangent line, a vertical tangent line, or no tangent line at
all. The first derivative test indicates if critical points yield relative maxima or relative minima.
A function is increasing (or decreasing) on an interval if its graph slopes upward to the right
(respectively, left), and this is characterized by a nonnegative (respectively, nonpositive) deriva-
tive. It is concave upward (or downward) if its slope is increasing (respectively, decreasing),
and consequently if its second derivative is nonnegative (respectively, nonpositive). Points that
separate intervals of opposite concavity are called points of inflection.
In Section 4.7 we illustrated that many applied extrema problems require absolute extrema
rather than relative extrema. Absolute extrema of a continuous function on a closed interval
must occur at either critical points or the ends of the interval. This fact implies that to find the
absolute extrema of a continuous function f (x) on a closed interval a ≤ x ≤ b , we evaluate
f (x) at its critical points between a and b and at a and b . The largest and smallest of these
values are the absolute extrema of f (x) on a ≤ x ≤ b . Plotting f (x) can also prove valuable.
When an object moves along a straight line, its velocity and acceleration are the first and
second derivatives, respectively, of its displacement with respect to time. In other words, if we
observe straight-line motion of an object, and record its position as a function of time, then we
can calculate the velocity and acceleration of that object at any instant.
Changes in a number of interrelated quantities usually produce changes in the others —
sometimes small, sometimes large. How the rates of change of these variables relate to each
other was the subject of Section 4.9. Related rate problems made us acutely aware of the
importance of differentiating an equation with respect to a variable only if the equation is valid
for a continuous range of values of that variable.
Potentials across resistors and inductors are expressed in terms of derivatives. Potentials
across capacitors, resistors, and inductors are respectively,
Q dQ di d 2Q
V = , V = iR = R , V =L =L 2.
C dt dt dt
Cauchy’s generalized mean value theorem enabled us to develop L’Hôpital’s rule in Section
4.11 for evaluation of various indeterminate forms such as 0/0, ∞/∞ , 0 · ∞ , 00 , 1∞ , ∞0 ,
and ∞ − ∞ .
When variables change by small amounts, corresponding changes in related variables can
often be approximated by differentials. Particularly important in error analyses are relative and
percentage changes.
KEY TERMS
Newton’s iterative approach Increasing function

Decreasing function Critical point
Relative (or local) maximum Relative (or local) minimum
First-derivative test Concave upward
Concave downward Points of inflection
Horizontal point of inflection Vertical point of inflection
Second-derivative test Absolute (or global) maximum
Absolute (or global) minimum Instantaneous velocity
Instantaneous acceleration Speed
Related rate problems Capacitor
Capacitance Current
Resistor Resistance
Inductor Emf force
Kirchhoff’s loop rule Indeterminate forms
L’Hôpital’s rule Differential
Relative change Percentage change
REVIEW
EXERCISES
∗ 1. (a) Prove that the area of the isosceles triangle in the figure ∗ 5. Of all pairs of positive numbers that multiply to some given constant
below is c > 0, find that pair which has the smallest sum.
l 2 sin θ
A= . ∗ 6. Solve Exercise 5 for the largest sum.
2
∗ 7. Two sides of the triangle in the figure below maintain constant
(b) If the angle θ is increasing at 1/2 radian per minute, but l lengths of 3 cm and 4 cm, but the length l of the third side decreases at
remains constant, how fast is the area of the triangle chang- the rate of 1 cm/min. How fast is angle θ changing when l is 4 cm?
ing? What does your answer predict when θ = 0, π/2,
and π ?
(c) When does A change most rapidly and most slowly; that 4 3
is, when is |dA/dt| largest and smallest?
In Exercises 8–19 evaluate the limit, if it exists.
l l 3x 2 + 2x 3 sin 3x
8. lim 9. lim
x→0 3x 3 − 2x 2 x→∞ 2x
2
x − 16 sin 3x
10. lim 11. lim
x→4 x−4 x→0 2x
2
√
sin x x−2
∗ 2. Draw a graph of each of the following functions, indicating all rel- 12. lim ∗ 13. lim √ √
ative maxima and minima and points of inflection.
x→−∞ 2x x→2 + x− 2
2 − 3x
∗ 14. lim x e ∗ 15. lim x 2x
3 2 x 2 − 2x + 4 x→∞ x→0+
(a) f (x) = 4x + x − 2x + 1 (b) f (x) = 2
x − 2x + 1 sin 2x
∗ 16. lim x 4 ln x ∗ 17. lim
∗ 3. Of all pairs of positive numbers that add to some given constant x→0+ x→0 tan 3x
( )x
c > 0, find that pair which has the largest product. x+1
18. lim ∗ 19. lim xex
∗ 4. Solve Exercise 3 for the smallest product. x→∞ x−1 x→−∞
∗ 20. Use Newton’s method to find all critical points for the following
functions accurate to six decimal places:
x3 + 1 c
(a) f (x) = x 4 + 3x 2 − 2x + 5 (b) f (x) = a
3x 3 + 5x + 1
∗ 21. An object moves along the x -axis with its position defined as a
function of time t by
b
44 ∗ 28. A football team presently sells tickets at prices of $8, $9, and
x = x(t) = t 4 − t 3 + 62t 2 − 84t, t ≥ 0.
3 $10 per seat, depending on the position of the seat. At these prices
it averages sales of 10 000 at $10, 20 000 at $9, and 30 000 at $8.
Plot a graph of this function, indicating times when the velocity and The team wishes to raise the price of each ticket by the same amount,
acceleration of the object are equal to zero. but feels that for every dollar the price is raised, 10% fewer tickets of
each type will be sold. What price increase per ticket will maximize
∗ 22. If the graph in the figure below represents the position x(t) of an
revenue?
object moving along the x -axis, what could physically cause the corner
at time t0 ? ∗ 29. Repeat Exercise 28 given that the team takes into account the fact
that profit from concession sales for each person at the game is 50 cents.
x
∗ 30. If a particle moves away from the origin along the positive x -axis
with a constant speed of 10 m/s, how fast is its distance from the curve
y = x 2 changing when it is at x = 3 m?
x = x (t)
∗ 31. Each evening a cow in a pasture returns to its barn at point B (figure
below). But it always does so by first walking to the river for a drink.
If the cow walks at 2 km/h and stops to drink for 2 min, what is the
minimum time it takes for the cow to get from the pasture to the barn?
What is the minimum time if the cow walks twice as fast?
t0 t
1
∗ 23. If an object moves along the x -axis with constant acceleration, is km Barn
1 2
it possible for a graph of its position function x(t) to have a point of km
4
inflection? B
1 km
∗ 24. An open box is formed from a square piece of cardboard ( l units
long on each side) by cutting out a square at each corner and folding Pasture
up the sides. What is the maximum possible volume for the box? 3 km
4
∗ 25. (a) If the average speed of a car for a trip is 80 km/h, must it
at some time have had an instantaneous speed of 80 km/h?
Explain. River
(b) Must the car at some instant have had a speed of 83 km/h?
∗ 32. A farmer has 100 ha to plant in corn and potatoes. Undamaged
∗ 26. Draw graphs of the following functions, indicating all relative max-
corn yields p dollars per hectare and potatoes q dollars per hectare. For
ima and minima and points of inflection:
each crop, the loss due to disease and pests per unit hectare is directly
proportional to the area planted. If the farmer plants x hectares of
x3 corn, the loss due to disease and pests is equal to ax per hectare. The
(a) f (x) = (b) f (x) = x 2 + sin2 x
x2 −1 total loss of corn is therefore ax 2 hectares. Similarly, the total loss of
potatoes is by 2 if the area planted in potatoes is y hectares. Find the
∗ 27. If at some instant of time sides a and b of the triangle in the areas that should be planted in corn and potatoes in order to minimize
following figure form a right angle, and if these sides are increasing at monetary loss. Substitute sample values for a , b , p , and q to see
equal rates, does it remain a right-angled triangle? whether your results look reasonable.
CHAPTER
5 The Indefinite Integral and the
Antiderivative
Application Preview Suppose all vehicles in a single lane of traffic on a highway have the same speed v . Let l be
the average length of the vehicles (figure below), and let d be the distance between vehicles
(assumed uniform).
A l d
The rate r at which traffic flows is defined to be the number of vehicles passing a fixed point
A per unit time. This is equal to the inverse of the time taken for one vehicle to pass A , including
the distance between vehicles. At speed v , this time is (l + d)/v , so that r = v/(l + d) .
Naturally a traffic engineer would like to move traffic along as quickly as possible, and this could
be accomplished by increasing v and decreasing d . But safety is an important consideration,
and d is not, or should not be, independent of v ; it should be increased as v is increased.
Experimental measurements have suggested that the shortest stopping distance is 52 m when a
vehicle is travelling 22 m/s, and 96 m for 31 m/s.
THE PROBLEM Find a reasonable dependence for d as a function of v and use it to find
a suggested speed v for maximum flow rate r if cars maintain a safe driving distance. Take
l = 4 m for the average length of a vehicle. (See Example 5.7 on page 346 for the solution.)
In Chapter 3 we introduced the derivative and ways to differentiate functions defined explic-
itly and implicitly. We then discussed various applications of calculus, including velocity and
acceleration, related rates, maxima and minima, Newton’s iterative procedure, and L’Hôpital’s
rule. In this chapter, we reverse the differentiation process. Instead of giving you a function and
asking for its derivative, we give you the derivative and ask you to find the function. This process
of backwards differentiation or antidifferentiation has such diverse applications that antidiffer-
entiation is as important to calculus as differentiation. In many problems we find ourselves
differentiating at one stage and antidifferentiating at another.
Antidifferentiation is a much more difficult process than differentiation. All but one of the
rules for differentiation were developed in Chapter 3; the one remaining rule is in Section 9.1.
In contrast to this, the list of formulas and techniques for finding antiderivatives is endless. In
this chapter we introduce the three simplest but most important techniques; in Chapter 8 we
discuss many others.
335
336 Chapter 5 The Indefinite Integral and the Antiderivative
5.1 The Reverse Operation of Differentiation

In our discussions on velocity and acceleration in Section 4.8, we showed that when an ob-
ject moves along the x -axis with its position described by the function x(t) , its velocity and
acceleration are the first and second derivatives of x(t) with respect to time t :
dx d 2x
v(t) = and a(t) = .
dt dt 2
For example, if x(t) = t 3 + 3t 2 , then
v(t) = 3t 2 + 6t and a(t) = 6t + 6.
When engineers and physicists study the motions of objects, a more common type of
problem is to determine the position of an object; that is, the position is not given. What they
might know, however, is the acceleration of the object (perhaps through Newton’s second law,
which states that acceleration is proportional to the resultant force on the object). So the question
we must now ask is: If we know the acceleration a(t) of an object as a function of time, can
we obtain its velocity and position by reversing the differentiations? In the example above, if
we know that
dv
a(t) = = 6t + 6,
dt
can we find the function v(t) that differentiates to give 6t + 6? We know that to arrive at the
terms 6t and 6 after differentiation, v(t) might have contained the terms 3t 2 and 6t . In other
words, one possible velocity function that differentiates to give a(t) = 6t + 6 is v(t) = 3t 2
+ 6t . It is not, however, the only one; v(t) = 3t 2 + 6t + 10 and v(t) = 3t 2 + 6t − 22 also
have derivative 6t + 6. In fact, for any constant C whatsoever, the derivative of
v(t) = 3t 2 + 6t + C
is a(t) = 6t + 6. In Theorem 5.1 we shall show that this velocity function represents all
functions that have 6t + 6 as their derivative. We shall also demonstrate how to evaluate the
constant C .
For the present, let us set C = 0 so that
dx
v(t) = = 3t 2 + 6t.
dt
We now ask what position function x(t) differentiates to give 3t 2 + 6t . One possibility is
x(t) = t 3 + 3t 2 , and for the same reason as above, so is
x(t) = t 3 + 3t 2 + D
for any constant D whatsoever.

Thus, by reversing the differentiation operation in this example we have proceeded from
the acceleration a(t) = 6t + 6 to possible velocity functions v(t) , and then to possible position
functions x(t) . This process of antidifferentiation has applications far beyond velocity and
acceleration problems, and we shall see many of them as we progress through this book. We
begin our formal study of antidifferentiation with the following definition.
DEFINITION 5.1
A function F (x) is called an antiderivative of f (x) on an interval I if on I ,
F " (x) = f (x). (5.1)

For example, since

d 4
x = 4x 3 ,
dx
we say that x 4 is an antiderivative of 4x 3 for all x . But for any constant C , the function
x 4 + C is also an antiderivative of 4x 3 . The following theorem indicates that these are the only
antiderivatives of 4x 3 .
THEOREM 5.1
If F (x) is an antiderivative of f (x) on an interval I , then every antiderivative of f (x)
on I is of the form
F (x) + C, where C is a constant.
PROOF Suppose that F (x) and G(x) are two antiderivatives of f (x) on I . If we define a
function D(x) = G(x) − F (x) , then on I
D " (x) = G" (x) − F " (x) = f (x) − f (x) = 0.

If x1 and x2 are any two points in the interval I , then certainly D " (x) = 0 on the interval
x1 ≤ x ≤ x2 . But differentiability of D(x) on x1 ≤ x ≤ x2 implies continuity of D(x)
thereon also (Theorem 3.6). We may therefore apply the mean value theorem (Theorem 3.19)
to D(x) on the interval x1 ≤ x ≤ x2 , and conclude that there exists a number c between x1
and x2 such that
D(x2 ) − D(x1 )
D " (c) = .
x 2 − x1
But because D " (c) = 0, it follows that D(x1 ) = D(x2 ) . Since x1 and x2 are any two points
in I , we conclude that D(x) must have the same value at every point in I ; that is, on I ,
D(x) = G(x) − F (x) = C,

where C is a constant. Consequently,
G(x) = F (x) + C.
Because of this theorem, if we find one antiderivative F (x) of f (x) by any means what-
soever, then we have found every antiderivative of a function f (x) , since every antiderivative
can be written as F (x) plus a constant C . Thus, every antiderivative of a function f (x) is of
the form F (x) + C , where F (x) is any one antiderivative. We call F (x) + C the indefinite
integral of f (x) . The operation of taking the indefinite integral is denoted by
!
f (x) dx = F (x) + C, (5.2)
where the differential dx indicates that integration is with respect to x . We call f (x) the
integrand of the indefinite integral. For example, we write
! !
x3 1 1
x 2 dx = +C and dx = − + C.
3 x3 2x 2
Distinguish between an antiderivative of a function f (x) and the indefinite integral of
f (x) . Both reverse differentiation. An antiderivative of f (x) is a function that differentiates
to f (x) ; the indefinite integral of f (x) is all functions that differentate to f (x) , it adds an
arbitrary constant to any antiderivative.
The following theorem is fundamental to the calculation of antiderivatives and indefinite
integrals. Its proof is a straightforward exercise in differentiation and the use of Definition 5.1.
THEOREM 5.2
If f (x) and g(x) have antiderivatives on an interval I , then on I :
! ! !
(i) [f (x) + g(x)] dx = f (x) dx + g(x) dx; (5.3a)
! !
(ii) kf (x) dx = k f (x) dx, k a constant. (5.3b)
For example, to evaluate the indefinite integral of 2x 3 − 4x , we write

! ! !
3 3
(2x − 4x) dx = 2x dx + −4x dx [by part (i) of Theorem 5.2]
! !
3
=2 x dx − 4 x dx [by part (ii) of Theorem 5.2)]
" # " #
x4 x2
=2 −4 +C
4 2
x4
= − 2x 2 + C.
2
Every differentiation formula developed in Chapter 3 can be expressed as an integration
formula. In fact, equations 5.3 are integral counterparts of equations 3.9 and 3.8. Some of the
differentiation formulas for trigonometric, inverse trigonometric, exponential, logarithm, and
hyperbolic functions are restated below as integration formulas.
!
sin x dx = − cos x + C, (5.4a)
!
cos x dx = sin x + C, (5.4b)
!
sec2 x dx = tan x + C, (5.4c)
!
sec x tan x dx = sec x + C, (5.4d)
!
csc2 x dx = − cot x + C, (5.4e)
!
csc x cot x dx = − csc x + C, (5.4f)
!
1
√ dx = Sin−1 x + C, (5.4g)
1 − x2
!
1
dx = Tan−1 x + C, (5.4h)
1 + x2
!
1
√ dx = Sec−1 x + C, (5.4i)
x x2 − 1
!
ex dx = ex + C, (5.4j)
!
a x dx = a x loga e + C, (5.4k)
!
1
dx = ln |x| + C, (5.4l)
x
!
cosh x dx = sinh x + C, (5.4m)
!
sinh x dx = cosh x + C, (5.4n)
!
sech 2 x dx = tanh x + C, (5.4o)
!
csch 2 x dx = −coth x + C, (5.4p)
!
sech x tanh x dx = −sech x + C, (5.4q)
!
csch x coth x dx = −csch x + C. (5.4r)
Equation 5.4l follows immediately from equation 3.46 with f (x) = x .

Perhaps the most important integration formula is the counterpart of power rule 3.7. Since
d n+1
x = (n + 1)x n ,
dx
it follows that !
(n + 1)x n dx = x n+1 + C.
With property 5.3b, the n + 1 can be removed from the integral and taken to the other side of
the equation: !
1
x n dx = x n+1 + C, n %= −1, (5.5)
n+1
where C = C/(n + 1) . As indicated, this result is valid provided that n %= −1, but formula
5.4l takes care of this exceptional case.
EXAMPLE 5.1
Evaluate ! " #
1
2x + dx.
x2
SOLUTION Using Theorem 5.2 and formula 5.5, we find that
! " # ! ! " #
1 1 x2 1 1
2x + dx = 2 x dx + dx = 2 − + C = x2 − + C.
x2 x2 2 x x
Where is x 2 − 1/x + C , the indefinite integral of 2x + 1/x 2 ? It is valid on the intervals x < 0
and x > 0, but not for all x because the function is not defined at x = 0. To make this clear,
we should write 
 1
! " #  x 2 − + C1 , x < 0
1 x
2x + 2 dx =
x  x 2 − 1 + C2 , x > 0

x
where the constants C1 and C2 need not be the same. For brevity we often write
! " #
1 1
2x + dx = x 2 − + C,
x2 x
thereby suppressing the complete description of the indefinite integral. When the context de-
mands that we distinguish various intervals on which the indefinite integral is defined, we shall
be careful to give the extended version.
EXAMPLE 5.2
Find a curve that passes through the point (1, 5) and whose tangent line at each point (x, y)
has slope 5x 4 − 3x 2 + 2.
SOLUTION If y = f (x) is the equation of the curve, then
dy
= 5x 4 − 3x 2 + 2.
dx
If we take indefinite integrals of both sides of this equation with respect to x , we obtain
! !
dy
dx = (5x 4 − 3x 2 + 2) dx or
dx
y = x 5 − x 3 + 2x + C.
Since (1, 5) is a point on the curve, its coordinates must satisfy the equation of the curve:
5 = 15 − 13 + 2(1) + C.
Thus C = 3, and the required curve is y = x 5 − x 3 + 2x + 3.
In taking the indefinite integral of each side of the equation
dy
= 5x 4 − 3x 2 + 2
dx
in the example above, we added an arbitrary constant C to the right-hand side. You might
question why we did not add a constant to the left-hand side. Had we done so, the result would
have been
y + D = x 5 − x 3 + 2x + E.
If we had then written
y = x 5 − x 3 + 2x + (E − D)
and defined C = E − D , we would have obtained exactly the same result. Hence, nothing is
gained by adding an arbitrary constant to both sides; a constant on one side is sufficient.
The problem in Example 5.2 was geometric: Find the equation of a curve satisfying certain
properties. We quickly recast it as the problem of finding the function y = f (x) that satisfies
the equation
dy
= 5x 4 − 3x 2 + 2,
dx
subject to the additional condition that f (1) = 5. Once again this is a differential equation.
We discuss differential equations briefly in Section 5.5 and study them in detail in Chapter 15.
When we solve a differential equation with no subsidiary conditions, say,
dy
= 4x 2 + 7x,
dx
we do not get a function, but rather a one-parameter family of functions. For this differential
equation, we obtain
4 3 7 2
y = x + x + C,
3 2
FIGURE 5.1 One-parameter a one-parameter family of cubic polynomials, C being the parameter. Geometrically, we have
family of solutions of a differential
the one-parameter family of curves in Figure 5.1. Parameter C represents a vertical shift of
equation
one curve relative to another. Note that if a vertical line is drawn at any position x to intersect
y these curves, then at the points of intersection, every curve has exactly the same slope, namely
4x 2 + 7x . For example, the slope of each cubic at x = 0 is zero. If an extra condition is added
C=1
to the differential equation, such as to demand that y be equal to 4 when x = 2, then
C=0
4 7
4 = (2)3 + (2)2 + C,
2 3 2
1 or, C = −62/3. This condition singles out one particular function from the family, namely
x y = 4x 3 /3 + 7x 2 /2 − 62/3. Geometrically, it determines that curve in the family which passes
C = −1 through the point (2, 4) .
−1
In this chapter, we discuss three basic ways to find antiderivatives. First, some antideriva-
tives are obvious, and the better you are at differentiation, the more obvious they will be. For
example, you should have no trouble recognizing that
! !
7x 6 1 −1
7x 5 dx = +C and dx = + C.
6 x3 2x 2
Our second method results from the answer to the following question: How do we check
that a function F (x) is an antiderivative of f (x) ? We differentiate F (x) , of course. This
simple fact suggests an approach to slightly more complex problems, say,
!
(2x + 3)5 dx.
We might reason that in order to have 2x + 3 raised to power 5 after differentiation, we had
2x + 3 to power 6 before differentiation; that is, a reasonable proposal for an antiderivative is
(2x + 3)6 . Differentiation of this function gives
d
(2x + 3)6 = 6(2x + 3)5 (2) = 12(2x + 3)5 ,
dx
and we see that (2x + 3)6 is not a correct antiderivative. It has produced (2x + 3)5 , as required,
but it has also given an undesirable factor of 12. We therefore adjust our original proposal by
multiplying it by 1/12; that is, the correct indefinite integral is
!
1
(2x + 3)5 dx = (2x + 3)6 + C.
12
This is what we call adjusting constants: we propose an antiderivative that is within a mul-
tiplicative constant of being correct. Adjusting the original proposal by the inverse of this
constant gives the correct antiderivative. But remember, our initial proposal must be within a
multiplicative constant. We cannot adjust x ’s. Let us illustrate with two very similar problems,
! !
1 1
dx and dx.
(5x + 2 )5 (5x 2 + 2 )5
For the first problem we propose as an antiderivative (5x + 2)−4 . Differentiation gives
( )
d 1 −4 −20
= (5) = .
dx (5x + 2)4 (5x + 2)5 (5x + 2)5
Since we are out by a factor of −20, we adjust our original proposal with −1/20,
!
1 −1
dx = + C.
(5x + 2 )5 20(5x + 2)4
It might seem as logical to propose (5x 2 + 2)−4 as an antiderivative for the second problem,
but differentiation yields
( )
d 1 −4 −40x
2 4
= 2 5
(10x) = .
dx (5x + 2) (5x + 2) (5x 2 + 2)5
This time the discrepancy is −40x . We cannot adjust x ’s. The original proposal must be
abandoned. This indefinite integral is quite difficult; it will have to wait for the more powerful
techniques of Chapter 8. To emphasize once again, do not try to adjust x ’s, only constants.
EXAMPLE 5.3
Evaluate the following indefinite integrals:
! ! !
x
(a) cos 3x dx (b) e−2x dx (c) dx
3x 2−4
SOLUTION
(a) To obtain cos 3x after differentiation, we propose sin 3x as an antiderivative. Since
d
sin 3x = 3 cos 3x,
dx
it is necessary to adjust with 1/3,
!
1
cos 3x dx = sin 3x + C.
3
−2 x
(b) With an initial proposal of e for an antiderivative based on the fact that the deriva-
tive of an exponential function always returns the same exponential, we calculate
d −2 x
e = −2 e − 2 x .
dx
Consequently, we adjust with −1/2,
!
1
e−2x dx = − e−2x + C.
2
(c) Since
d 1
ln (3x 2 − 4) = (6x),
dx 3x 2 − 4
the required indefinite integral is
!
x 1
dx = ln |3x 2 − 4| + C.
3x 2−4 6
We have inserted absolute values as suggested by formula 5.4l.
The third technique for finding antiderivatives is discussed in Section 5.3.

EXAMPLE 5.4
Find continuous indefinite integrals for the Heaviside unit step function h(x − a) introduced
in Section 2.5.
SOLUTION Because the function has two values (see Figure 2.35), we subdivide the integra-
tion into two parts. For x < a , h(x − a) = 0, and the indefinite integral is a constant. For
x > a , h(x − a) = 1, and the indefinite integral is x plus a constant. Thus,
! *
C, x <a
h(x − a) dx =
x + D, x > a .
For continuity at x = a , we require C = a + D &⇒ D = C − a . Consequently, continuous

indefinite integrals are
! *
FIGURE 5.2 Ramp function C, x <a
h(x − a) dx =
y (x − a) + C, x > a .
Slope = 1
For future applications, it is convenient to express these in the form
!
C h(x − a) dx = (x − a)h(x − a) + C,
a x where, for continuity at x = a , we understand that the value of the indefinite integral at x = a
is C . A graph of this function is shown in Figure 5.2. It is called a ramp function.
EXERCISES 5.1
In Exercises 1–20 evaluate the indefinite integral. In Exercises 21–24 find the curve y = f (x) that passes through the
! ! given point and whose slope at each point (x, y) is defined by the
1. (x 3 − 2x) dx 2. (x 4 + 3x 2 + 5x) dx derivative indicated.
! ! 21. dy/dx = x 2 − 3x + 2, (2, 1)

3 2
3. (2x − 3x + 6x + 6) dx 4. sin x dx 22. dy/dx = 2x + 4x,3
(0, 5)
! ! 4
23. dy/dx = −2x + 3x + 6, 2
(1, 0)
√
5. 3 cos x dx 6. x dx 7
24. dy/dx = 2 − 4x + 8x , (1, 1)
! " # ! " #
1 1 2 ∗ 25. Find the equation of the curve that has a second derivative equal
7. x 10 − dx 8. − dx to 6x 2 and passes through the points (0, 2) and (−1, 3) .
x3 x2 x4
! ! " # ∗ 26. Find a function f (x) that has a relative maximum f (2) = 3 and
+ 3/ 2 , 1 1
9. x − x 2/7 dx 10. + √ dx has a second derivative equal to −5x .
x2 2 x
! " # ! " # ∗ 27. Is it possible to find a function f (x) that has a relative minimum
4 1/ 3 1 3 f (2) = 3 and has a second derivative equal to −5x ?
11. + 2x dx 12. − + 3x dx
x 3/ 2 2x 2
! !
1 + √ , In Exercises 28–67 use adjusting constants to evaluate the indefinite
13. dx 14. 2 x + 3x 3/2 − 5x 5/2 dx
xπ integral.
! ! ! !
√ √
15. x 2 (x 2 − 3) dx 16. x(x + 1) dx 28. x + 2 dx 29. (x + 5)3/2 dx
! " # ! ! !
x−2 √ 1
17. dx 18. x 2 (1 + x 2 )2 dx 30. 2 − x dx 31. √ dx
x3 4x + 3
! ! ! !
(x − 1)2
19. (x 2 + 1)3 dx 20. √ dx 32. (2x − 3)3/2 dx 33. (3x + 1)5 dx
x
! ! ! !
1 1 1
34. (1 − 2x)7 dx 35. dx ∗ 60. √ dx ∗ 61. dx
(x + 4)2 1− 4x 2 1 + 9x 2
! ! ! !
1 1 3x
36. dx 37. x(x 2 + 1)3 dx ∗ 62. √ dx ∗ 63. dx
(1 + 3x)6 x 3x 2 −1 1 + 5x 2
! ! ! !
x
38. x 2 (2 + 3x 3 )7 dx 39. dx ∗ 64. cosh 4x dx ∗ 65. x sinh 3x 2 dx
(2 + x 2 )2
! ! ! !
40. cos 2x dx ∗ 41. cos2 x sin x dx ∗ 66. sech 2x tanh 2x dx ∗ 67. x 2 csch 2 4x 3 dx
! !
∗ 42. 3 sin 2x cos 2x dx ∗ 43. sec 12x tan 12x dx
! ! In Exercises 68–73 find a one-parameter family of functions satisfying
∗ 44. csc2 4x dx ∗ 45. e4x dx the differential equation.
! ! dy 1 dy √
−x 2 e3/x ∗ 68. = x3 − 2 ∗ 69. = 3 − 4x
∗ 46. xe dx ∗ 47. dx dx x dx
x2
! ! dy 1 dy
1 ∗ 70. = ∗ 71. = x 2 (2x 3 + 4)4
∗ 48. e4x−3 dx ∗ 49. dx dx (3x + 5)3/2 dx
3x + 2
! ! dy x3 dy
2 x ∗ 72. = ∗ 73. = sin x(1 + cos2 x)
∗ 50. dx ∗ 51. dx dx (2 + 3x 4 )2 dx
7 − 5x 1 − x2
! ! ∗ 74. Find a function y = f (x) that satisfies the differential equation
3x 2
∗ 52. dx ∗ 53. 2x dx
1 − 4x 3 dy 1
! ! x
= 2
e dx x
∗ 54. 32x dx ∗ 55. dx
ex + 1
! ! and passes through the two points (1, 1) and (−1, −2) .
4 cos x
∗ 56. sin x(1 + cos x) dx ∗ 57. dx ∗ 75. (a) Find the indefinite integral of the signum function of Exer-
sin3 x cise 47 in Section 2.4.
! !
sec2 x (b) Prove that sgn x cannot have an antiderivative on any in-
∗ 58. e2x (1 + e2x )3 dx ∗ 59. dx
tan2 x terval containing x = 0.
5.2 Integrating Velocity and Acceleration

In Section 4.8 we discussed relationships among position, velocity, and acceleration from the
viewpoint of derivatives. We now consider these relationships through antiderivatives, an ap-
proach providing a far more practical viewpoint when it comes to applications.
For motion along the x -axis, velocity is the derivative of position with respect to time:
v(t) = dx/dt . We can say, therefore, that the indefinite integral of velocity represents every
possible position function with this velocity:
!
x(t) = v(t) dt. (5.6)
Similarly, as acceleration is the derivative of velocity, the indefinite integral of acceleration

represents every possible velocity function,
!
v(t) = a(t) dt. (5.7)
Thus, given the acceleration of an object moving along a straight line, we can antidifferentiate
to find its velocity and antidifferentiate again for its position. Since each antidifferentiation
introduces an arbitrary constant, additional information must be specified in order to evaluate
these constants.
EXAMPLE 5.5
The car in Figure 5.3 accelerates from rest when the light turns green. Initially, the acceleration
is 10 m/s 2 , but it decreases linearly, reaching zero after 10 s. Find the velocity and position of
FIGURE 5.3 Car accelerat-
ing from rest
the car during this time interval.
SOLUTION Let us set up the coordinate system in Figure 5.3 and choose time t = 0 when
the car pulls away from the light. Given this time convention, the acceleration of the car as a
function of time t (see Figure 5.4) is
a(t) = 10 − t, 0 ≤ t ≤ 10.
If v(t) denotes the velocity of the car during the time interval, then
x=0 x dv
t=0 = 10 − t,
dt
and integration gives
t2
v(t) = 10t − + C.
2
By our time convention, the velocity of the car is zero at time t = 0; that is, v(0) = 0, and
FIGURE 5.4 Car’s acceler-
from this we obtain
ation function
(0 )2
a 0 = 10(0) − + C.
2
Consequently, C = 0, and
t2
10 v(t) = 10t − .
2
This is the velocity of the car during the time interval 0 ≤ t ≤ 10, measured in metres per
second.
Since the position of the car with respect to its original position at time t = 0 is denoted
10 t by x ,
dx t2
v = = 10t − .
dt 2
Thus,
t3
x(t) = 5t 2 − + D.
6
Since x(0) = 0, the constant D must also be zero, and the position of the car indicated by its
distance in metres from the stoplight, for 0 ≤ t ≤ 10, is
t3
x(t) = 5t 2 − .
6
EXAMPLE 5.6
A stone is thrown vertically upward over the edge of a cliff at 25 m/s. When does it hit the base
of the cliff if the cliff is 100 m high?
SOLUTION Let us measure y as positive upward, taking y = 100 and t = 0 at the point
and instant of projection (Figure 5.5). A law of physics states that when an object near the
earth’s surface is acted on by gravity alone, it experiences an acceleration whose magnitude is
9.81 m/s 2 . If a denotes the acceleration of the stone and v its velocity, then
dv
a = = −9.81
dt
FIGURE 5.5 Stone thrown

( a is negative since it is in the negative y -direction). To obtain v we integrate with respect to t :
upward over edge of cliff
v(t) = −9.81t + C.
t=0
y = 100
v = 25
By our time convention, v(0) = 25, so that 25 = −9.81(0) + C . Thus, C = 25, and
v(t) = −9.81t + 25.

y
We now have the velocity of the stone at any given instant. To find the position of the stone we
set v = dy/dt ,
y=0
dy
= −9.81t + 25,
dt
and integrate once again:
y(t) = −4.905t 2 + 25t + D.
Since y = 100 when t = 0, it follows that 100 = −4.905(0)2 + 25(0)+D . Hence, D = 100,
and
y(t) = −4.905t 2 + 25t + 100.
We have found the equation that tells us exactly where the stone is at any given time. To
determine when the stone strikes the base of the cliff, we set y(t) = 0; that is,
0 = −4.905t 2 + 25t + 100,
a quadratic equation with solutions

- √
−25 ± 252 − 4(−4.905)(100) 25 ± 2587
t = = = 7.7 or − 2.6.
−9.81 9.81
Since the negative root must be rejected, we find that the stone strikes the base of the cliff after
7.7 s.
An important point to note in Examples 5.5 and 5.6 is that the coordinate system and time con-
vention were specified immediately; that is, we decided, and did so at the start of the problem,
where to place the origin of our coordinate system, which direction to choose as positive, and
when to choose t = 0. Only then were we able to specify the correct sign for acceleration. Fur-
thermore, we determined constants from antidifferentiations using initial conditions expressed
in terms of our coordinate system and time convention. Throughout the solutions, we were
careful to refer everything to our choice of coordinates and time. Remember, then, to specify
clearly the coordinate system and time convention at the beginning of a problem.
We should also note that in each of the examples we integrated with respect to time only
those equations that were valid for a range of values of time. It is a common error to integrate
equations that are only valid at one instant. For instance, students reason in Example 5.6 that
the initial velocity is 25, v = dy/dt = 25, and hence y(t) = 25t + C . This is incorrect,
because the equation dy/dt = 25 is valid only at time t = 0, and cannot be integrated.
EXAMPLE 5.7
Find the speed that maximizes flow rate in the Application Preview and draw any conclusions
Application Preview that you feel are justified.
Revisited
SOLUTION Flow rate is, at the moment, a function of speed v and distance d between
vehicles, r = v/(l + d) . To maximize r , we must express it as a function of one variable.
As was suggested, v and d are related, d should increase when v increases. We use the
fact that d should be the stopping distance for cars travelling at speed v to find a functional
relationship between v and d . The stopping distance of a vehicle is composed of two parts,
distance dT that the vehicle travels during the time it takes the driver to get his foot from
the accelerator to the brake, and dB , the distance travelled while the brake is applied. If T
represents the reaction time for the driver to apply the brake, then dT = vT , where v is
the speed of the vehicle before braking. If a is the acceleration of the vehicle during the
braking period, its velocity is V = at + C . If we choose t = 0 when the brake is applied,
then V = v at t = 0, and this implies that V = at + v . Distance travelled during the
braking period is x = at 2 /2 + vt + D . If we choose x = 0 at t = 0, then D = 0, and
x = at 2 /2 + vt . The vehicle stops when 0 = V = at + v &⇒ t = −v/a , and therefore
dB = a(−v/a)2 /2 + v(−v/a) = −v 2 /(2a) . The stopping distance for a vehicle when
reaction time is T and acceleration during braking is a is
v2
d = dT + dB = vT − .
2a
We can evaluate T and a using the fact that d = 52 m when v = 22 m/s, and d = 96 m when
v = 31 m/s. These give
222 312
52 = 22T − , 96 = 31T − ,
2a 2a
the solution of which is a = −6.138 m/s 2 and T = 0.5715 s. The flow function is
v v
r(v) = = .
l+d l + vT − v 2 /(2a)
Its graph is shown in Figure 5.6 for l = 4. We see that it has one critical point, which is given
by
" # .
v2 v/
l + vT − (1) − v T −
2a a
0 = r " (v) = " # .
2 2
v
l + vT −
2a
When we set the numerator to zero, and multiply by 2a ,
FIGURE 5.6 Flow func- √

tion for cars on highway
0 = 2al + 2avT − v 2 − 2avT + 2v 2 = 2al + v 2 &⇒ v = −2al.
r √
0.6 The speed for maximum flow rate is −2(−6.138)(4) = 7.0 m/s, which is about 25 km/h.
0.4
Obviously highway speed limits are not set with safety in mind. On the other hand, safe
distances between cars at typical highway speeds of say 100 km/h, or 27.8 m/s, would be
0.2 27.8(0.5715) + (27.8)2 /12.276 = 78.8 m, and we know that drivers do not maintain any such
distance.
2 4 6 8 v
EXERCISES 5.2
In Exercises 1–8 we have defined acceleration a(t) of an object moving ∗ 15. You are standing on a bridge 25 m above a river. If you wish to
along the x -axis during some time interval and specified the initial drop a stone onto a piece of floating wood, how soon before the wood
conditions x(0) and v(0) . Find the velocity v(t) and position x(t) of reaches the appropriate spot should you drop the stone?
the object as functions of time.
∗ 16. You are standing at the base of a building and wish to throw a ball
1. a(t) = t + 2, 0 ≤ t ≤ 3; v(0) = 0, x(0) = 0 to a friend on the roof 20 m above you. With what minimum speed
must you throw the ball?
2. a(t) = 6 − 2t, 0 ≤ t ≤ 3; v(0) = 5, x(0) = 0
3. a(t) = 6 − 2t 0 ≤ t ≤ 4; v(0) = 5, x(0) = 0 ∗ 17. A car is travelling at 20 m/s when the brakes are applied. What
2
constant deceleration must the car experience if it is to stop before
4. a(t) = 120t − 12t , 0 ≤ t ≤ 10; v(0) = 0, x(0) = 4 striking a tree that is 50 m from the car at the instant the brakes are
5. a(t) = t 2 + 1, 0 ≤ t ≤ 5; v(0) = −1, x(0) = 1 applied? Assume that the car travels in a straight line.
2
6. a(t) = t + 5t + 4, 0 ≤ t ≤ 15; v(0) = −2, x(0) = −3 ∗ 18. The position of an object moving along the x -axis is given by
7. a(t) = cos t, t ≥ 0; v(0) = 0, x(0) = 0
x(t) = t 3 − 6t 2 + 9t − 20, t ≥ 0.
8. a(t) = 3 sin t, t ≥ 0; v(0) = 1, x(0) = 4
∗ 9. The velocity of an object moving along the x -axis is given in metres Draw graphs of the position, velocity, and acceleration functions for
per second by this motion. Pay special attention to the fact that v(t) represents the
v(t) = 3t 2 − 9t + 6 slope of x = x(t) and a(t) is the slope of v = v(t) .
where t is time in seconds. If the object starts from position x = 1 m ∗ 19. Repeat Exercise 18 given that the acceleration of the object is
at time t = 0, answer each of the following questions:
(a) What is the acceleration of the object at t = 5 s? a(t) = 6t − 30, t ≥ 0,
(b) What is the position of the object at t = 2 s?
(c) Is the object speeding up or slowing down at t = 5/4 s? and v(0) = −33 and x(0) = 400.
(d) What is the closest the object ever comes to the origin? ∗ 20. You are called on as an expert to testify at a traffic hearing. The
∗ 10. A particle moving along the x -axis has acceleration question concerns the speed of a car that made an emergency stop with
brakes locked and wheels sliding. The skid mark on the road measured
a(t) = 6t − 2 9 m. Assuming that the deceleration of the car was constant and could
not exceed the acceleration due to gravity of a freely falling body (and
in metres per second per second for time t ≥ 0. this is indeed a reasonable assumption), what can you say about the
(a) If the particle starts at the point x = 1 moving to the left speed of the car before the brakes were applied? Are you testifying for
with speed 3 m/s, find its position as a function of time t . the prosecution or the defence?
(b) At what time does the particle have zero velocity (if any)? ∗ 21. When the brakes of an automobile are applied, they produce a
∗ 11. The acceleration of a particle moving along the x -axis is given in constant deceleration of 5 m/s 2 .
metres per second per second by (a) What is the distance, from the point of application of brakes,
required to stop a car travelling at 100 km/h?
a(t) = 6t − 15,
(b) Repeat part (a) for 50 km/h.
where t ≥ 0 is time in seconds. (c) What is the ratio of these distances?
(a) If the velocity of the particle at t = 2 s is 6 m/s, what is its
(d) Repeat parts (a), (b), and (c) given that the reaction time of
velocity at t = 1 s?
the driver to get her foot from accelerator to brake is 3/4 s,
(b) If the particle is 10 m to the right of the origin at time t = 0, and distances are calculated taking this reaction time into
what is its position as a function of time t ? account.
(c) What is the closest the particle ever comes to the origin?
∗ 22. A stone is dropped into a well and the sound of the stone striking
∗ 12. A car is sitting at rest at a stoplight. When the light turns green at the water is heard 3.1 s later. If the speed of sound is 340 m/s, how
time t = 0, the driver immediately presses the accelerator, imparting deep is the surface of the water in the well?
an acceleration of a(t) = (3 − t/5) m/s 2 to the car for 10 s.
∗ 23. Two trains, one travelling at 100 km/h and the other at 60 km/h, are
(a) Where is the car after the 10 s? headed toward each other along a straight, level track. When they are
(b) If the driver applies the brakes at t = 10 s, and the car 2 km apart, each engineer sees the other’s train and locks his wheels.
experiences a constant deceleration of 2 m/s 2 , where and
when does the car come to a stop? (a) If the deceleration of each train has magnitude 1/4 m/s 2 ,
determine whether a collision occurs.
∗ 13. Find how far a plane will move when landing if in t seconds af-
ter touching the ground, its speed in metres per second is given by (b) Repeat part (a) given that the deceleration is caused by the
180 − 18t . wheels being reversed rather than locked.
∗ 14. A stone is thrown directly upward with an initial speed of 10 m/s. (c) Illustrate graphically the difference between the situations
How high will it rise? in parts (a) and (b).
∗ 24. A steel bearing is dropped from the roof of a building. An observer ∗ 28. What speed maximizes the flow rate in Example 5.7 if cars are
standing in front of a window 1 m high notes that the bearing takes 1/8 s required to maintain only a fraction k (0 < k < 1) of the safe distance?
to fall from the top to the bottom of the window. The bearing continues
to fall, makes a completely elastic collision with a horizontal sidewalk, ∗ 29. Two stones are thrown vertically upward over the edge of a bot-
and reappears at the bottom of the window 2 s after passing it on the tomless abyss, the second stone t0 units of time after the first. The first
way down. After a completely elastic collision, the bearing will have stone has an initial speed of v0" , and the second an initial speed of v0"" .
the same speed at a point going up as it had going down. How tall is
the building? (a) Show that if the stones are ever to pass each other during
∗ 25. A construction elevator without a ceiling is ascending with constant their motions, two conditions must be satisfied:
speed 10 m/s. A girl on the elevator throws a ball directly upward from
a height of 2 m above the elevator floor just as the elevator floor is gt0 > v0" − v0"" and v0" > gt0 /2,
28 m above the ground. The initial speed of the ball with respect to the
elevator is 20 m/s.
where g > 0 is the acceleration due to gravity.
(a) What is the maximum height attained by the ball?
(b) How long does it take for the ball to return to the elevator (b) Show that the first condition is equivalent to the requirement
floor? that stone 1 must begin its downward trajectory before stone
2.
∗ 26. Two stones are thrown vertically upward one second apart over
the edge of the cliff in Example 5.6. The first is thrown at 25 m/s, the (c) Show that the second condition is equivalent to the require-
second at 20 m/s. Determine if and when they ever pass each other. ment that stone 1 must not pass its original projection point
∗ 27. In the theory of special relativity, Newton’s second law ( F = ma ) before the projection of stone 2.
is replaced by
0 1 ∗ 30. Speed bumps are to be placed on a straight stretch of road in order to
d v ensure that traffic speed does not exceed 10 m/s. The question concerns
F = m0 - , their spacing. Suppose vehicle speed is reduced to 2.5 m/s at bumps.
dt 1 − (v 2 /c2 ) Vehicles accelerate uniformly away from a bump at 3 m/s 2 , and then
decelerate uniformly toward the next bump at 7 m/s 2 . Find the distance
where F is the applied force, m0 the mass of the particle measured at
between bumps.
rest, v its speed, and c the speed of light — a constant. Show that if
we set a = dv/dt , then
∗∗ 31. It takes time T to drive your car a distance D along a straight
m0 a highway. You do so by accelerating uniformly from rest, attaining
F = . maximum speed V , which you maintain for some length of time, and
[1 − (v 2 /c2 )]3/2
then decelerating uniformly to a stop. How long do you maintain speed
Explain the difference between this law and Newton’s second law. V if the magnitudes of the acceleration and deceleration are the same?
5.3 Change of Variable in the Indefinite Integral

In Section 5.1 we suggested two methods for evaluating indefinite integrals — recognition and
adjusting constants. In this section we show how a change of variable can often replace a
complex integration problem with a simpler one.
Consider the indefinite integral
!
√
x 2x + 1 dx.
What is annoying about this integrand is the sum of two

√ terms 2x + √1 under a square root. This
can be changed by setting u = 2x + 1. As a result, 2x + 1 = u , and the x in front of the
square root is equal to (u − 1)/2. Now the differential dx is used to indicate integration in the
problem with respect to x . But surely there
√ must be another reason why √ we have chosen the
differential to denote this. If we regard x 2x + 1 dx as a product of x 2x + 1 and dx , then
perhaps we should obtain an expression for dx in terms of du . To do this we note that since
u = 2x + 1, then
du
= 2.
dx
As derivatives can be regarded as quotients of differentials, we can rewrite this equation in the
form
du
dx = .
2
If we make all these substitutions into the indefinite integral
!
√
x 2x + 1 dx,
the result is an integration problem in the variable u , which is easy to evaluate:
! " # ! " #
u − 1 √ du 1 1 2 5/2 2
u = (u3/2 − u1/2 ) du = u − u3/2 + C.
2 2 4 4 5 3
If u is now replaced by 2x + 1, the result is
1 1
(2x + 1)5/2 − (2x + 1)3/2 + C.
10 6
√
Differentiation of this function quickly indicates that its derivative is indeed x 2x + 1 and,
therefore,
!
√ 1 1
x 2x + 1 dx = (2x + 1)5/2 − (2x + 1)3/2 + C.
10 6
In this example, the substitution u = 2x + 1 replaces a complex integration in x with a simple
one in u . Once the problem in u is solved, replacement of u ’s with x ’s gives the solution to the
original indefinite integral. This method is generally applicable and is justified in the following
theorem.
THEOREM 5.3
Suppose the change of variable u = h(x) ⇔ x = g(u) replaces the indefinite integral
! !
+ ,
f (x) dx with f g(u) g " (u) du,
where g(u) is differentiable on some interval,

+ , and f (x) is continuous on the range of
g(u) . If F (u) is an antiderivative of f g(u) g " (u) , then
!
+ ,
f (x) dx = F h(x) + C.
EXAMPLE 5.8

! ! !
√ x
sin3 x cos2 x dx
5
(a) 2x + 4 dx (b) √ dx (c)
x+1
SOLUTION
(a) We could adjust constants in this case. With an initial guess of (2x + 4)6/5 , we
calculate
d 6
(2x + 4)6/5 = (2x + 4)1/5 (2).
dx 5
Thus,
!
5
(2x + 4)1/5 dx = (2x + 4)6/5 + C.
12
Alternatively, if we set u = 2x + 4, then du = 2dx , and
! ! " #
1/5 1/5 du 1 5 6/5 5
(2 x + 4 ) dx = u = u +C = (2x + 4)6/5 + C.
2 2 6 12
(b) If we set u = x + 1, then du = dx , and
! ! !
x u−1 2
√ dx = √ du = (u1/2 − u−1/2 ) du = u3/2 − 2u1/2 + C
x+1 u 3
2
= (x + 1)3/2 − 2(x + 1)1/2 + C.
3
√
A different substitution is also possible. If we set u = x + 1, then
1
du = √ dx and
2 x+1
! ! " #
x 2 u3
√ dx = (u − 1)(2du) = 2 −u +C
x+1 3
2
= (x + 1)3/2 − 2(x + 1)1/2 + C.
3
(c) If we set u = cos x , then du = − sin x dx and
! !
sin3 x cos2 x dx = sin2 x cos2 x sin x dx
!
= (1 − cos2 x) cos2 x sin x dx
! !
= (1 − u2 )u2 (−du) = (u4 − u2 ) du
u5 u3 1 1
= − +C = cos5 x − cos3 x + C.
5 3 5 3
EXERCISES 5.3
! !
In Exercises 1–30 evaluate the indefinite integral. e 2x ln x
∗ 27. dx ∗ 28. dx
! ! e 2x + 1 x
√ ! !
1. (5x + 14)9 dx 2. 1 − 2x dx 1 x
∗ 29. dx ∗ 30. dx
x ln x (x 2 + 1)[ln (x 2 + 1)]2
! ! !
1 5
3. dy 4. dx ∗ 31. Consider the integral sin3 x cos3 x dx .
(3y − 12)1/4 (5 − 42x)1/4
! ! (a) Evaluate it by making the substitution u = sin x .
x (b) Evaluate it by making the substitution u = cos x .
5. x 2 (3x 3 + 10)4 dx 6. dx
(x 2 + 4)2 (c) Verify that these answers are the same.
3
! ! !
4 x2 x2
7. sin x cos x dx 8. dx ∗ 32. Evaluate dx .
(x − 2)4 1+x
! !
√ x
9. z 1 − 3z dz 10. √ dx In Exercises 33–36 use the suggested change of variable to evaluate
2x + 3
the indefinite integral.
! √ !
1+ x - ! √
11. √ dx 12. s 3 s 2 + 5 ds 4x − x 2
x ∗ 33. dx; set u = 2/x
x3
! ! ! √
√ x − x2
13. sin2 x cos3 x dx 14. 1 − cos x sin x dx ∗ 34. dx; set u = 1/x
x4
!
! ! x
x3 2
- ∗ 35. dx; set u2 = (5 + x)/(1 − x)
15. dx 16. y y − 4 dy (5 − 4x − x 2 )3/2
(3 − x 2 )3 !
1
! √ ! ∗ 36. √ dx ;
(1 + u)1/2 8 3 6
3(1 − x2) − (5 + 4x) 1 − x 2
17. √ du 18. x (3x − 5) dx
u
set u2 = (1 − x)/(1 + x)
! 1/ 4 ! √
1+z x+1
19. √ dz 20. dx ∗ 37. Show that the substitution u − x = x 2 + x + 4 replaces the
z (x 2 + 2x + 2)1/3 integral ! -
! ! x 2 + x + 4 dx
(x − 1)(x + 2) cos3 x
21. √ dx ∗ 22. dx
x (3 − 4 sin x)4 with the integral of a rational function of u .
! ! 2 √
√ 3
√ ∗ 38. Show that the substitution (x + 1)u = 4 + 3x − x 2 replaces
∗ 23. 1 + sin 4t cos 4t dt ∗ 24. 1 + x dx the integral !
1
! ! √ dx
4 + 3x − x 2
∗ 25. tan2 x sec2 x dx ∗ 26. tan x sec2 x dx
with the integral of a rational function of u .
5.4 Deflection of Beams

When a beam that might otherwise be horizontal is subjected to loads, it bends. By analyzing
FIGURE 5.7 Deflection of
beam under loading
internal forces and moments, it can be shown that the shape y(x) of a uniform beam with
constant cross-section (Figure 5.7) is governed by the equation
y
d 4y F (x)
= , (5.8)
dx 4 EI
x
Beam where E is a constant called Young’s modulus of elasticity (depending on the material of
the beam), and I is also a constant (the moment of inertia of the cross-section of the beam).
FIGURE 5.8 Beam bend- Quantity F (x) is the load placed on the beam; it is the vertical force per unit length in the
ing under its own weight x -direction, placed at position x , including the weight of the beam itself. For example, if a
y beam has mass 100 kg and length 10 m (Figure 5.8), then the load due to its weight is a constant
F (x) = −9.81(100/10) = −98.1 N/m at every point of the beam.
Mass = 100 kg Suppose a block with mass 50 kg, uniform in cross-section, and length 5 m is placed on the
left half of the beam in Figure 5.8 (see Figure 5.9). It adds an additional load of 9.81(10) =
x 98.1 N/m over the interval 0 < x < 5. The total load can be represented in terms of Heaviside
10 m
unit step functions as
F (x) = −98.1 − 98.1[h(x) − h(x − 5)].

FIGURE 5.9 Beam bend-
ing under its own weight and an [Recall that the function h(x) − h(x − 5) turns the function ( −98.1 in this case) in front of
additional load
it on at x = 0 and off at x = 5.] Because the beam extends from x = 0 to x = 10, and
y h(x) = 1 thereon, we can write that
Mass = 50 kg
F (x) = −196.2 + 98.1 h(x − 5), 0 < x < 10.
5 10
x Accompanying equation 5.8 will be four boundary conditions defining the type of support
Mass = 100 kg
(if any) at each end of the beam. Three types of supports are common. We discuss them at the
left end of the beam, but they also occur at the right end.
FIGURE 5.10 Simple
support at x = 0 1. Simple Support
y The end of a beam is simply supported when it cannot move vertically but is free to rotate.
Visualize a horizontal pin perpendicular to the xy -plane passing through a hole in the end of
the beam at x = 0 (Figure 5.10). The pin is fixed, but the end of the beam can rotate on the
x
pin. In this case, y(x) must satisfy the boundary conditions
Pin
y(0) = y "" (0) = 0. (5.9a)
FIGURE 5.11 Built-in
support at x = 0 2. Built-in End
y If the end x = 0 of the beam is permanently fixed in a horizontal position (Figure 5.11),
y(x) satisfies
x y(0) = y " (0) = 0. (5.9b)
3. Free Support
FIGURE 5.12 Free end or
no support at x = 0
If the end x = 0 of the beam is not supported (Figure 5.12), y(x) satisfies
y y "" (0) = y """ (0) = 0. (5.9c)
When two boundary conditions at each end of a beam accompany differential equation 5.8,
x
we have what is called a boundary-value problem.
EXAMPLE 5.9
A uniform beam with mass 100 kg and length 10 m has both ends built in horizontally. Find the
deflection curve for the beam.
SOLUTION The boundary-value problem for deflection y(x) is
d 4y 98.1
=− , 0 < x < 10,
dx 4 EI
y(0) = y " (0) = 0 = y(10) = y " (10).
Four integrations of the differential equation give
" #
1 98.1x 4 3 2
y(x) = − + Ax + Bx + Cx + D ,
EI 24
where A , B , C , and D are constants. The boundary conditions require these constants to
satisfy
0 = EIy(0) = D,
0 = EIy " (0) = C,
98.1(10)4
0 = EIy(10) = − + A(10)3 + B(10)2 + C(10) + D,
24
98.1(10)3
0 = EIy " (10) = − + 3A(10)2 + 2B(10) + C.
6
These yield A = 327/4 and B = −1635/4, and therefore the curve of deflection for the beam
is
" #
1 327x 4 327x 3 1635x 2 327
y(x) = − + − = (−x 4 + 20x 3 − 100x 2 ).
EI 80 4 4 80EI
Maximum deflection of the beam should occur at its midpoint. To confirm this, we find critical
points,
327 327
0 = y " (x) = (−4x 3 + 60x 2 − 200x) = − x(x − 10)(x − 5).
80EI 20EI
Solutions are x = 0 and x = 10 (because each end is fixed horizontally) and x = 5. Maximum
deflection is y(5) = −40 875/(16EI ) . For a beam such that the product EI = 106 , the
deflection at x = 5 is y(5) = −2.55 × 10−3 m, that is, 2.55 mm.
This problem was relatively straightforward due to the fact that the load function F (x) = −98.1
is continuous. Discontinuous load functions lead to more complicated calculations.
EXAMPLE 5.10
The end x = 0 of the beam in Figure 5.9 is horizontally built-in, and the right end is free, just
FIGURE 5.13 Deflections
of a diving board
like a diving board (Figure 5.13). Find the curve of deflection.
y
SOLUTION The boundary-value problem for deflections is
5 10
x d 4y 1
= [−196.2 + 98.1 h(x − 5)], 0 < x < 10,
dx 4 EI
y(0) = y " (0) = 0, y "" (10) = y """ (10) = 0.
Integration of the differential equation four times on the intervals 0 < x < 5 and 5 < x < 10
gives
4
1 −8.175x 4 + Ax 3 + Bx 2 + Cx + D, 0 < x < 5
y(x) =
EI −4.0875x 4 + P x 3 + Qx 2 + Rx + S, 5 < x < 10.
To evaluate the eight constants, we impose the four boundary conditions, and also demand that
y(x) and its first three derivatives be continuous at x = 5. This means that left- and right-hand
limits of EIy(x) , EIy " (x) , EIy "" (x) , and EIy """ (x) are the same at x = 5.
EIy(0) = 0 &⇒ D = 0,
EIy " (0) = 0 &⇒ C = 0,
EIy "" (10) = 0 &⇒ −49.05(10)2 + 6P (10) + 2Q = 0,
EIy """ (10) = 0 &⇒ −98.1(10) + 6P = 0.
lim EIy(x) = lim EIy(x) &⇒
x→5− x→5+
−8.175(5)4 + 125A + 25B + 5C + D = −4.0875(5)4 + 125P + 25Q + 5R + S,

lim EIy " (x) = lim EIy " (x) &⇒
x→5− x→5+
−32.7(5)3 + 75A + 10B + C = −16.35(5)3 + 75P + 10Q + R,

lim EIy "" (x) = lim EIy "" (x) &⇒
x→5− x→5+
−98.1(5)2 + 30A + 2B = −49.05(5)2 + 30P + 2Q,

lim EIy """ (x) = lim EIy """ (x) &⇒
x→5− x→5+
−196.2(5) + 6A = −98.1(5) + 6P .
Solutions of these equations are
A = 245.25, B = −3065.625, C = 0, D = 0,
P = 163.5, Q = −2452.5, R = −2043.75, S = 2554.6875.
The function describing deflections of the beam is, therefore,
4
1 −8.175x 4 + 245.25x 3 − 3065.625x 2 , 0 ≤ x ≤ 5
y(x) = 4 3 2
EI −4.0875x + 163.5x − 2452.5x − 2043.75x + 2554.6875, 5 < x ≤ 10.
A graph of this function for EI = 106 is shown in Figure 5.14.
FIGURE 5.14 Deflections of a diving board
y
2 4 6 8 10
x
−0.02
−0.04
−0.06
−0.08
−0.1
−0.12
−0.14
Fortunately there are easier ways to solve this problem. One such is to use what are called
Laplace transforms. You will learn them in advanced calculus courses. Even at this stage we
can simplify calculations considerably. Difficulties arose when we integrated the differential
equation separately on the intervals 0 < x < 5 and 5 < x < 10. This introduced four
additional constants of integration that were evaluated by demanding that y(x) , y " (x) , y "" (x) ,
and y """ (x) be continuous at x = 5. Suppose we demand from the beginning that y(x) and its
first three derivatives be continuous for the length of the beam. What this means is that we need
continuous antiderivatives of h(x − a) . We did this in Example 5.4. The continuous indefinite
integral of h(x − a) is
!
h(x − a) dx = (x − a)h(x − a) + C, (5.10a)
provided we understand that the value is C at x = a . The following indefinite integrals are
also continuous:
!
1
(x − a)h(x − a) dx = (x − a)2 h(x − a) + C, (5.10b)
2
!
1
(x − a)2 h(x − a) dx = (x − a)3 h(x − a) + C, (5.10c)
3
!
1
(x − a)3 h(x − a) dx = (x − a)4 h(x − a) + C, (5.10d)
4
provided again that each indefinite integral is given value C at x = a . In general,

!
1
(x − a)n h(x − a) dx = (x − a)n+1 h(x − a) + C. (5.11)
n+1
With these, the solution to Example 5.10 is far easier. Four integrations of
d 4y 1
= [−196.2 + 98.1h(x − 5)]
dx 4 EI
using formulas 5.10 give
( )
1 196.2x 4 98.1
y(x) = − + (x − 5)4 h(x − 5) + Ax 3 + Bx 2 + Cx + D .
EI 24 24
The boundary conditions require that
0 = EIy(0) = D,
0 = EIy " (0) = C,
196.2(10)2 98.1(5)2
0 = EIy "" (10) = − + + 6A(10) + 2B,
2 2
0 = EIy """ (10) = −196.2(10) + 98.1(5) + 6A.
These can be solved for A = 245.25 and B = −3065.625. The deflected curve is
196.2x 4 98.1
y(x) = − + (x − 5)4 h(x − 5) + 245.25x 3 − 3065.625x 2 .
24 24
This is equivalent to the function y(x) found in Example 5.10, but its derivation and final form
are unmistakably simpler.
In Section 2.5 we suggested that the Dirac-delta function is used to model point sources in
engineering and physics. We use it here to model a point force applied to a beam. The load due
to a point force of magnitude F newtons applied vertically downward at a point x0 on a beam
is represented mathematically by −F δ(x − x0 ) . Suppose it is applied to a beam of length L ,
and that F is so large that the weight of the beam is negligible by comparison. In this case, the
differential equation describing deflections is
d 4y F
=− δ(x − x0 ).
dx 4 EI
If the beam is horizontally fixed at x = 0 and simply supported at x = L , deflections must

also satisfy y(0) = y " (0) = 0 = y(L) = y "" (L) . According to Example 3.11 in Section 3.3,
h" (x − a) = δ(x − a) . Consequently, we can write that
!
δ(x − a) dx = h(x − a) + C. (5.12)
Using this result, integration of the differential equation gives
d 3y 1
3
= [−F h(x − x0 ) + C ].
dx EI
Three additional integrations using formulas 5.10 yield
( )
1 F 3 3 2
y(x) = − (x − x0 ) h(x − x0 ) + Cx + Ax + Bx + D ,
EI 6
where we have absorbed a factor of 6 into C . The boundary conditions require that
0 = EIy(0) = D,
0 = EIy " (0) = B,
F (L − x0 )3
0 = EIy(L) = − + CL3 + AL2 + BL + D,
6
0 = EIy "" (L) = −F (L − x0 ) + 6CL + 2A.
Values of A and C are
F x0 (L − x0 )(x0 − 2L) F (L − x0 )(2L2 + 2Lx0 − x02 )

A= , C = .
4 L2 12L3
Thus,
(
1 F F (L − x0 )(2L2 + 2Lx0 − x02 )x 3
y(x) = − (x − x0 )3 h(x − x0 ) +
EI 6 12L3
)
F x0 (L − x0 )(x0 − 2L)x 2
+ .
4 L2
A beam in a building is horizontally fixed at both ends. The size and distribution of
load on the beam determine its deflection at various points; the more load at a particular
point on the beam, the greater the deflection at that point. We are asked to determine
the difference between deflections at the midpoint of the beam when a given total load is
evenly distributed along the beam as opposed to when it is concentrated at the midpoint
of the beam.
SOLUTION Let us suppose that the length of the beam is L and the total load is F > 0.
When this load is evenly distributed along the beam, differential equation 5.8 becomes
d 4y F
=− ,
dx 4 EI L
subject to the boundary conditions
y(0) = y " (0) = 0 = y(L) = y " (L).
Four integrations of the differential equation give

" #
1 F x4
y(x) = − + Ax 3 + Bx 2 + Cx + D .
EI 24L
The boundary conditions require
0 = EIy(0) = D,
0 = EIy " (0) = C,
F L3
0 = EIy(L) = − + AL3 + BL2 + CL + D,
24
F L2
0 = EIy " (L) = − + 3AL2 + 2BL + C.
6
These give A = F /12 and B = −F L/24, so that
" #
1 F x4 F x3 F Lx 2
y(x) = − + − .
EI 24L 12 24
Deflection at the midpoint of the beam is y(L/2) = −F L3 /(384EI ) . When F is

concentrated at the midpoint of the beam differential equation 5.8 becomes
d 4y F
4
=− δ(x − L/2).
dx EI
Four integrations give
( )
1 F
y(x) = − (x − L/2)3 h(x − L/2) + Ax 3 + Bx 2 + Cx + D .
EI 6
The boundary conditions require
0 = EIy(0) = D,
0 = EIy " (0) = C,

F (L − L/2)3
0 = EIy(L) = − + AL3 + BL2 + CL + D,
6
F (L − L/2)2
0 = EIy " (L) = − + 3AL2 + 2BL + C.
2
These give A = F /12 and B = −F L/16, so that
( )
1 F 3 F x3 F Lx 2
y(x) = − (x − L/2) h(x − L/2) + − .
EI 6 12 16
Deflection at the midpoint is y(L/2) = −F L3 /(192EI ) . The ratio of concentrated

deflection to distributed deflection at the midpoint is
−F L3 /(192EI )
= 2.
−F L3 /(384EI )
EXERCISES 5.4
1. Find deflections of a uniform beam with mass m and length L when ∗ 11. (a) A simply-supported beam of length 4 m is to carry a to-
both ends are simply supported. tal uniform load of 1000 N (including its weight), and
2. Repeat Exercise 1 if both ends are fixed horizontally. a concentrated load of 1500 N at its centre. Assuming
EI = 106 , use Exercise 10 to find deflections of the beam.
3. Repeat Exercise 1 if the left end x = 0 is fixed horizontally and the
right end is free. (b) If safety codes do not permit deflections to exceed 1/360 of
4. Repeat Exercise 1 if the left end is fixed horizontally and the right span, is the beam acceptable?
end is simply supported.
∗ 12. Find deflections of the beam in Figure 5.9 if the 50 kg block is on
5. A concentrated force of F newtons is applied vertically downward the right half of the beam. Is the deflection at the right end larger or
at the midpoint of a uniform beam of length L . Both ends of the beam smaller than in Figure 5.9?
are built in horizontally. If F is so large that the weight of the beam is
negligible in comparison, find deflections of the beam. ∗ 13. Find deflections of the beam in Figure 5.9 if the 50 kg block is
6. Repeat Exercise 5 if the left end of the beam is built in horizontally centred on the beam. How does the deflection at the right end compare
and the right end is free. Is the beam straight for x > L/2? to that in Figure 5.9 and Exercise 12?
7. Repeat Exercise 5 if both ends of the beam are simply supported. ∗ 14. A concentrated force F is applied vertically downward at the right
∗ 8. Repeat Exercise 5 if the mass m of the beam is taken into account. end of a uniform beam of length L . The left end of the beam is built
in horizontally and the right end is free. If the weight of the beam
∗ 9. Repeat Exercise 6 if the mass m of the beam is taken into account. is negligible in comparison to F , find deflections of the beam. Hint:
∗ 10. Repeat Exercise 7 if the mass m of the beam is taken into account. Place F at a point x0 to the left of x = L , solve for deflections, and
take the limit as x0 → L− .
5.5 An Introduction to Separable Differential Equations

A differential equation is an equation that contains derivatives of some unknown function; the
equation must be solved for all functions that satisfy it. Because differential equations arise
in so many engineering problems, both elementary and advanced, we discuss what are called
separable differential equations here. A full treatment of differential equations is taken up in
Chapter 15.
When a differential equation for y as a function of x contains dy/dx , but no higher order
derivatives, it is called a first-order differential equation. The vast majority of such differential
equations can be expressed in the form
dy
= F (x, y); (5.13)
dx
that is, they can be solved for dy/dx in terms of x and y . Examples are
dy dy
(1 − y) = 3x 2 and + 2xy = 4x
dx dx
Each of these can be solved for dy/dx ,
dy 3x 2 dy
= and = 4x − 2xy.
dx 1−y dx
Differential equation 5.13 is said to be separable if it can be expressed in the form
dy M(x)
= ; (5.14)
dx N (y)
that is, dy/dx is a function of x divided by a function of y . Both of the above examples are
separable. The first is already in this form; the second can be put so,
dy 2x
= 4x − 2xy = 2x(2 − y) = .
dx 1
2−y
What is equivalent to equation 5.14 is to say that 5.13 is separable if it can be expressed in the
form
N (y) dy = M(x) dx. (5.15)
When a differential equation is written in this way, it is said to be separated — separated in the
sense that x - and y -variables appear on opposite sides of the equation. For a separated equation
we can write therefore that
dy
N (y) = M(x), (5.16)
dx
and if we integrate both sides with respect to x , we have
! !
dy
N (y) dx = M(x) dx + C. (5.17)
dx
Cancellation of differentials on the left leads to the solutions
! !
N (y) dy = M(x) dx + C. (5.18)
What we mean by saying that 5.18 represents solutions for 5.14 is that any function defined
implicitly by 5.18 is a solution of 5.14. A word of warning is appropriate here. Looking at
equation 5.18 in isolation, it might appear that we have integrated the left side of equation 5.15
with respect to y and the right side with respect to x . This is not true. We do not differentiate
one side of an equation with respect to one variable and the other side with respect to a different
variable. Why then would we expect to be able to integrate different sides of an equation with
respect to different variables? What we did was rewrite 5.15 in form 5.16 and integrate both sides
of this equation with respect to x . Cancellations of differentials led to 5.18. Thus, although we
did not integrate the left side of equation 5.15 with respect to y and the right side with respect
to x to get equation 5.18, we can now interpret 5.18 in this way.
For instance, in the first example above, we write
(1 − y)dy = 3x 2 dx.
According to equation 5.18, solutions are defined implicitly by

! !
y2
(1 − y) dy = 3x 2 dx &⇒ y− = x 3 + C.
2
We can find explicit solutions by solving the equation for y in terms of x . Multiplying by −2
expresses the equation as a quadratic in y :
y 2 − 2y + 2(x 3 + C) = 0.
Therefore -
2± 4 − 8(x 3 + C) -
y = = 1± 1 − 2(x 3 + C).
2
Explicit solutions of the differential equation are therefore
- -
y(x) = 1 + 1 − 2(x 3 + C) and y(x) = 1 − 1 − 2(x 3 + C),
provided expressions on the right are indeed functions of x . Once C is determined, this will
be true only for certain values of x . For example, suppose we require the solution of the
differential
- equation that satisfies the extra condition y(0) = 3. The second function y(x) =
1 − 1 − 2(x 3 + C) cannot satisfy this condition because y cannot be greater than 1. If we
substitute x = 0 and y = 3 into the other function,
√
3 = 1+ 1 − 2C,
and this requires C = −3/2. Thus, the solution of the differential equation for which y(0) = 3
is - -
y(x) = 1 + 1 − 2(x 3 − 3/2) = 1 + 4 − 2x 3 .
Since 4 − 2x 3 must be nonnegative for this function to be defined, the solution is valid only on
the interval x ≤ 21/3 . In fact, because the derivative of this function is not defined at x = 21/3 ,
we should consider only x < 21/3 .
We now consider four problems that give rise to separable differential equations.
EXAMPLE 5.11
An ore sample contains, along with various impurities, an amount A0 of radioactive material,
say, uranium. Disintegrations gradually reduce this amount of uranium. Experiments have led
to the following law of radioactive disintegration: The time rate of change of the amount of
radioactive material is proportional at any instant to the amount of radioactive material present
at that time. Find the amount of uranium in the sample as a function of time.
SOLUTION If we let A(t) be the amount of uranium in the sample at any time t , then the
law of radioactive disintegration states that
dA
= kA, (5.19)
dt
where k is a constant. Since A is decreasing, dA/dt and hence k must be negative. If we
choose t = 0 when the amount of uranium is A0 , then the differential equation must be solved
for A(t) subject to the initial condition A(0) = A0 . The differential equation is separable:
1
dA = k dt,
A
and solutions are therefore defined implicitly by

! !
1
dA = k dt &⇒ ln |A| = kt + C.
A
We can omit the absolute values since A is always positive, in which case the initial condition
requires ln A0 = 0 + C . Thus,
ln A = kt + ln A0 .
To find A(t) explicitly, we take exponentials on both sides of the equation:
A = ekt+ln A0 = ekt eln A0 = A0 ekt .
The amount of uranium therefore decreases exponentially in time. To find k we need to know
A at one additional time. For example, if we know that one ten-millionth of 1% of the original
amount of uranium decays in 6.5 years, then
0.999 999 999 A0 = A0 e13k/2 .
If we solve this for k , we obtain
2
k = ln (0.999 999 999) = −1.54 × 10−10 ,
13
and therefore
−10 t
A(t) = A0 e−1.54×10 .
The law of radioactive disintegration has an important application in the dating of once-living
plants and animals. All living tissue contains two isotopes of carbon: C 14 (carbon-14), which is
radioactive, and C 12 (carbon-12), which is stable. In living tissue, the ratio of the amount of C 14
to that of C 12 is 1/10 000 for all fragments of the tissue. When the tissue dies, however, the ratio
changes due to the fact that no more carbon is produced, and the original C 14 present decays
radioactively into an element other than C 12 . Thus, as the dead tissue ages, the ratio of C 14 to
C 12 decreases, and by measuring this ratio, it is possible to predict how long ago the tissue was
alive. Suppose, for example, the present ratio of C 14 to C 12 in a specimen is 1/100 000; that is,
one-tenth that for a living tissue. Then 90% of the original amount of C 14 in the specimen has
disintegrated.
If we let A(t) be the amount of C 14 in the specimen at time t , taking A = A0 to be the
amount present in the living specimen at its death ( t = 0), then A = A0 ekt . To determine k , we
use the fact that the half-life of C 14 is approximately 5550 years. (The half-life of a radioactive
element is the time required for one-half an original sample of the material to disintegrate.) For
carbon-14, this means that A is equal to A0 /2 when t = 5550; that is,
A0
= A0 e5550k .
2
When we divide by A0 and take natural logarithms,
1
k =− ln 2 = −0.000 125.
5550
Consequently, the amount of C 14 in the specimen of dead tissue at any time t is given by
A = A0 e−0.000 125t .
If T is the present time, when the amount of C 14 is known to be 10% of its original amount,
then at this time
0.1A0 = A0 e−0.000 125T .
The solution of this equation is
ln 10
T = = 18 400,
0.000 125
and we conclude that the tissue died about 18 400 years ago.
EXAMPLE 5.12
When a hot (or cold) object is placed in an environment that has a different temperature, the
object cools down (or heats up). For example, when a hot cup of coffee is placed on a table
it cools down due to colder room temperature. Suppose, for example, that the cup of coffee is
initially at temperature 95◦ C and the room stays at constant temperature 20◦ C. Newton’s law
of cooling states that the rate of change of the temperature of the coffee is proportional to the
difference between temperature of the coffee and that of the room. Find the temperature of the
coffee as a function of time.
SOLUTION If T (t) denotes temperature of the coffee, then Newton’s law of cooling can be
stated algebraically as
dT
= k(T − 20), (5.20)
dt
where k < 0 is a constant. It is negative because T − 20 > 0 and dT /dt < 0. According
to equation 5.20, the coffee cools quickly at first because T − 20 is large, but more and more
slowly as T approaches 20. Differential equation 5.20 for T (t) is separable:
1
dT = k dt.
T − 20
Consequently, solutions are defined implicitly by
! !
1
dT = k dt &⇒ ln |T − 20| = kt + C.
T − 20
We can drop absolute values since temperature of the coffee is never less than 20◦ C. Exponen-
tiating both sides of this equation gives
T − 20 = ekt+C = eC ekt &⇒ T = 20 + Dekt ,
where D = eC . If we choose t = 0 when temperature of the coffee is 95◦ C, then 95 =
20 + D &⇒ D = 75. Thus, temperature of the coffee is
T (t) = 20 + 75ekt .
To find k we need to know temperature of the coffee at one other time. For example, if we knew
that temperature dropped to 50◦ C in 5 min, then
1
50 = 20 + 75e5k &⇒ k = ln (2/5) = −0.183.
5
Coffee temperature is
T (t) = 20 + 75e−0.183t .
We can also express T (t) in the form
" #t/5
t/5 ] 2
T (t) = 20 + 75e(t/5) ln (2/5) = 20 + 75eln [(2/5) = 20 + 75 .
5
The plot of these functions in Figure 5.15 indicates that coffee temperature never reaches 20◦ C;
the graph is asymptotic to the line T = 20. However, temperature of the coffee is within 5◦ of
20◦ C in about 15 min, and within 1◦ in about 24 min.
FIGURE 5.15 Temperature of coffee
T
100
80
60
40
20
0 5 10 15 20 25 t
Admittedly, Newton’s law of cooling is an approximation to what really happens physically. It

assumes that there is no temperature variation within the cup of coffee, and that room temperature
remains constant. Idealistic as this is, it is a good starting position for further discussions.
EXAMPLE 5.13
Figure 5.16 shows a liquid container with a hole in its bottom. It is interesting, and perhaps
surprising, to find that the speed at which liquid exits through the hole is independent of the
shape of the container; it depends only on the depth of liquid. We show this here, and then
discover a technique for finding the depth of liquid in the container as a function of time.
FIGURE 5.16 Container with hole through which liquid escapes
Surface of liquid
Thin layer of liquid

y
y=0
Hole
If we consider a thin surface layer of the liquid, then during a small interval of time, the depth
of liquid in the container drops by an amount equal to the thickness of the layer. Simultaneously,
a volume of liquid equal to that in the layer exits through the hole and, in so doing, causes the
gravitational potential energy of the layer, due to its elevated position relative to the hole, to
be converted into kinetic energy of the liquid passing through the hole. Suppose we let v be
the speed at which the liquid leaves the container when the depth of liquid is y . The potential
energy of the layer relative to the bottom of the container is
(mass of layer)(g)(y),
where g = 9.81. The kinetic energy of an equal amount of liquid as it leaves the container is
1
(mass of layer)v 2 .
2
When we equate these energies, and cancel the mass of the layer, we obtain
1
v 2 = gy.
2
In other words, the speed at which liquid exits through the hole is
-
v = 2gy. (5.21)
This result is known as Torricelli’s law. It results from an idealized situation in which all
potential energy is converted into kinetic energy. Experience suggests that exit speed depends
on other factors as well — the size of the hole, for one. Water leaves more slowly through
√ a
small hole than through a large one. It is often assumed that v is somewhat less than 2gy ,
and equation 5.21 is replaced by
-
v = c 2gy, (5.22)
where 0 < c < 1 is a constant called the discharge coefficient. We call 5.22 the modified
Torricelli law. We use it to find a differential equation satisfied by depth of liquid in the
container. Suppose the area of the surface of the liquid is a function of depth y denoted by
A(y) , and V (t) is the volume of liquid in the container at any time t . Since depth of liquid
changes at rate dy/dt , the rate at which the volume of liquid in the container changes is
dV dy
= A(y) .
dt dt
Both dV /dt and dy/dt are negative since V and y are decreasing. On the other hand, the rate
at which liquid exits through the hole is the product av of the area a of the hole and exit speed
v . It follows then that
dy -
A(y) = −av = −ac 2gy. (5.23)
dt
Once the shape of the container is specified, then A(y) is known, and this equation becomes a
differential equation for y(t) . Solve for depth when the container is a right circular cylinder of
radius r and height h with vertical axis. Determine how long it takes a full tank to empty.
SOLUTION When the container is a right circular cylinder with radius r , then A(y) = π r 2 ,
in which case differential equation 5.23 becomes
√
2 dy
- 1 2gac
πr = −ac 2gy &⇒ √ dy = − dt,
dt y πr2
a separated differential equation. Solutions are defined implicitly by
! ! √ √
1 2gac √ 2gac
√ dy = − dt &⇒ 2 y = − t + D.
y πr2 πr2
√
If the cylinder is originally full and liquid exits starting at time t = 0, then 2 h = D , and
√ " √ #
√ 2gac √ √ gact 2
2 y = − t + 2 h &⇒ y = h− √ .
πr2 2π r 2
It is now a simple matter to determine how long the cylinder takes to empty. Setting y(t) = 0
and solving for t gives
3
πr2 2h
t = .
ac g
This was a particularly simple example in that the cross-sectional area A(y) of the container
was constant. Containers with variable cross-sections are discussed in the exercises.
EXAMPLE 5.14
A tank originally contains 1000 L of water, in which 5 kg of salt has been dissolved.
(a) If a brine mixture containing 2 kg of salt for each 100 L of solution is poured into the
tank at 10 mL/s, find the amount of salt in the tank as a function of time.
(b) If at the same time brine is being added, the mixture in the tank is being drawn off at
10 mL/s, find the amount of salt in the tank as a function of time. Assume that the
mixture is stirred constantly.
SOLUTION
(a) Suppose we let S(t) represent the number of grams of salt in the tank at time t . If
we choose time t = 0 at the instant the brine mixture begins entering the original
solution, then S(0) = 5000. Since 10 mL of mixture enters the tank each second,
and each millilitre contains 0.02 g of salt, it follows that 0.2 g of salt enters the tank
each second. Consequently, after t seconds, 0.2t grams of salt has been added to the
tank, and the total amount of salt in the solution is, in grams,
S(t) = 5000 + 0.2t.

(b) Once again we let S(t) represent the number of grams of salt in the tank at time t .
Its derivative dS/dt , the rate of change of S(t) , is the difference between how fast
salt is being added to the tank in the brine and how fast salt is being removed as the
mixture is removed,
dS * 5 * 5
= rate salt added − rate salt leaves .
dt
As in part (a), salt is being added at the constant rate of 0.2 g/s. The rate at which salt
leaves the tank, on the other hand, is not constant; it depends on the concentration of
salt in the tank. Since the tank always contains 106 mL of solution, the concentration
of salt in the solution at time t is, in grams per millilitre, S/106 , where S = S(t) is
the amount of salt in the tank at that time. As solution is being drawn off at the rate
of 10 mL/s, the rate at which salt leaves the tank is, in grams per second,
S S
(10) = .
106 105
Consequently,
dS 1 S 20 000 − S
= − 5 = .
dt 5 10 100 000
To find S(t) we must solve this differential equation subject to the condition that
S(0) = 5000. The differential equation is separable:
1 1
dS = dt.
20 000 − S 100 000
Solutions are defined implicitly by
t
− ln |20 000 − S| = + C.
100 000
Let us solve for S before evaluating the constant of integration. Multiplication by

−1 and exponentiation give
|20 000 − S| = e−C−t/100 000 &⇒ 20 000 − S = ±e−C e−t/100 000 .

Because ±e−C is an unknown constant, we simplify matters by setting D = ±e−C ,

in which case
S(t) = 20 000 − De−t/100 000 .
FIGURE 5.17 Amount of
salt in tank The initial condition S(0) = 5000 requires 5000 = 20 000 − D &⇒ D = 15 000,
S and therefore the number of grams of salt in the tank is
20 000
S(t) = 20 000 − 15 000e−t/100 000 .
10 000
A graph of this function is shown in Figure 5.17. It is asymptotic to the line S =
20 000. After a very long time, the amount of salt in the tank levels off so that its
200 000 400 000 t concentration is 20 000/1 000 000 = 0.02 g/mL. This, as might be expected, is the
concentration of the incoming solution.
This project concerns the rate of evaporation of water from a hemispherical tank, open on
the top. All vertical cross sections of the tank are semicircles with radius r metres, one
of which is shown in Figure 5.18. The tank is originally full of water and the problem is
to determine how long it takes for the water to completely evaporate.
FIGURE 5.18 Evaporation

of water from a hemispherical
tank
y
−r r
y x
(x, y)
h x
SOLUTION We must make some kind of assumption about the rate at which water
evaporates. Since evaporation takes place at the surface of the water, it seems reasonable
to assume that evaporation is proportional to the surface area of the water at any given
time. If V (t) denotes the volume of water in the tank at time t , evaporation is represented
by the derivative dV /dt , the time rate of change of the volume of water in the tank. If
A(t) is the surface area of the water at time t , then we have
dV
= kA,
dt
where k < 0 is the constant of proportionality. It is negative because A is positive and
dV /dt is negative. Essentially our problem is to solve this differential equation for V (t)
and determine when V = 0. An immediate difficulty is that the equation contains three
variables, t , A , and V ; one of which must be eliminated.
Certainly, t must remain, so either A or V must go. It would seem that we need a
functional relationship between V and A . It is easy to see that the area of the surface of
the water at any instant is π x 2 , where x is the radius of the surface. What can we say
about the volume V of a spherical segment? In many references we find the following
formula for such a volume,
π h2
(3r − h),
V =
3
where r is the radius of the sphere and h is the depth of water. (You may recall that we
used this formula in Section 4.1. We will verify it in Section 7.2.) With h = r + y , we
can write that
π π
V = (r + y)2 [3r − (r + y)] = (2r 3 + 3r 2 − y 3 ).
3 3
Since the equation of the semicircle is x 2 + y 2 = r 2 , we may also express A in terms of

y,
A = π(r 2 − y 2 ).
These two equations determine V in terms of A , but to find exactly how would be difficult.
Instead, notice that if we substitute both of them into the differential equation, we obtain
" #
π 2 dy 2 dy
3r − 3y = kπ(r 2 − y 2 ).
3 dt dt
When we simplify this equation, the result is
dy
= k.
dt
How simple. We now have a differential equation for y as a function of t . Integration
gives y = kt + C . If we assume that evaporation begins at time t = 0 when y = 0, we
must set C = 0, and therefore y = kt . Water has completely evaporated when y = −r ,
and this occurs at time t = −r/k . This is the required time. It is known explicitly when
k is known. For instance, if measurements indicate that the water level in the tank drops
by 1% of the radius in 4 days, then
1 r
− r = 4k &⇒ k =− .
100 400
It follows then that the tank empties in
" #
400
−r − = 400 days.
r
EXERCISES 5.5
1. Bacteria in a culture increase at a rate proportional to the number 6. Suppose the amount of a drug injected into the body decreases at a
present. If the original number increases by 25% in 2 h, when will it rate proportional to the amount still present. If a dose decreases by 5%
double? in the first hour, when will it decrease to one-half its original amount?
2. If the number of bacteria in a culture doubles in 3 h, when will it 7. A sugar cube 1 cm on each side is dropped into a cup of coffee.
triple? If the sugar dissolves in such a way that the cube always remains a
3. If one-half of a sample of radioactive substance decays in 15 days, cube, compare the times for the cube to completely dissolve under the
how long does it take for 90% of the sample to decay? following conditions:
4. If 10% of a sample of radioactive material decays in 3 s, what is its (a) Dissolving occurs at a rate proportional to the surface area
half-life? of the remaining cube; and
5. After 4 half-lives of a radioactive substance, what percentage of the (b) Dissolving occurs at a rate proportional to the amount of
original amount remains? sugar remaining.
8. Solve Exercise 7 if the sugar is not in the form of a cube, but rather ∗ 20. The water trough in the figure below is 4 m long. Its cross-section
in free form from the sugar bowl. Assume that the sugar consists of n is an isosceles triangle with a half-metre base and a half-metre altitude.
2
spherical particles each of radius r0 cm. Water leaks out through a hole of area√1 cm in the bottom with speed
in metres per second given by v = gD/2, where D (in metres) is
9. An analysis of a sample of fossil remains shows that it contains only
the depth of water in the trough, and g > 0 is the acceleration due to
1.51% of the original C 14 in the living creature. When did the creature
gravity. This is the modified Torricelli law 5.22 with c = 1/2. Find
die?
how long a full trough takes to empty.
10. If a fossilized creature died 100 000 years ago, what percentage of
the original C 14 remains? 1
11. The amount of a drug such as penicillin injected into the body is 2
used up at a rate proportional to the amount still present. If a dose 4
decreases by 5% in the first hour, when does it decrease to one-half its
original amount?
∗ 12. Glucose is administered intravenously to the bloodstream at a con-
stant rate of R units per unit time. As the glucose is added, it is
converted by the body into other substances at a rate proportional to the
1
amount of glucose in the blood at that time. Show that the amount of
2
glucose in the blood as a function of time t is given by
∗ 21. A spring of negligible mass and elasticity constant k > 0 is at-
R tached to a wall at one end and a mass M at the other (figure below).
C(t) = (1 − e−kt ) + C0 e−kt ,
k The mass is free to slide horizontally along a frictionless surface. If
x = 0 is taken as the position of M when the spring is unstretched and
where k is a constant and C0 is the amount at time t = 0 when
M is set into motion, the differential equation describing the position
the intravenous feeding is initiated. Draw a graph of this function for
x(t) of M is
C0 < R/k and C0 > R/k . d 2x k
= − x.
∗ 13. Prove that if a quantity decreases at a rate proportional to its present dt 2 M
amount, and if its percentage decrease in some interval of time is i %,
If motion is initiated by imparting a speed v0 in the positive x -direction
then its percentage decrease in any interval of time of the same length
to M at position x = 0, find the velocity of M as a function of position.
is i %.
Hint: Express d 2 x/dt 2 as follows, and then separate the differential
∗ 14. Water at temperature 90◦ C is placed in a room at constant tempera- equation
ture 20◦ C. Newton’s law of cooling states that the time rate of change of d 2x dv dv dx dv
the temperature T of the water is proportional to the difference between 2
= = =v .
dt dt dx dt dx
T and the temperature of the environment:
dT k
= k(T − 20),
dt M
◦
where k is a constant. If the water cools to 60 C in 40 min, find T as
x
a function of t . x=0
∗ 15. A thermometer reading 25◦ C is taken outside where the tempera- ∗ 22. When a mass m falls under the influence of gravity alone, it expe-
ture is −20◦ C. If the reading drops to 0◦ C in 4 min, when will it read riences an acceleration d 2 r/dt 2 described by
−19◦ C?
∗ 16. A boy lives 6 km from school. He decides to walk to school at d 2r GmM
m =− 2 ,
a speed that is always proportional to the square of his distance from dt 2 r
the school. If he is half-way to school after one hour, find his distance
where M is the mass of the earth, G is a positive constant, and r is
from school at any time. How long does it take him to reach school?
the distance from m to the centre of the earth (figure below). If m is
∗ 17. A tank has 100 L of solution containing 4 kg of sugar. A mixture dropped from a height h above the surface of the earth, find its velocity
with 10 g of sugar per litre of solution is added at 200 mL/min. At when it strikes the earth. What is the maximum attainable speed of m ?
the same time, 200 mL of well-stirred mixture is removed each minute.
Find the amount of sugar in the tank as a function of time. r
∗ 18. A tank in the form of an inverted right-circular cone of height H m
and radius R is filled with water. Water escapes through a hole of h
cross-sectional area a at the vertex. Use the modified Torricelli law Earth
5.22 to find a formula for the time the tank takes to empty.
∗ 19. A container in the form of an inverted right-circular cone of radius R
4 cm and height 10 cm is full of water. Water evaporates from the
surface at a rate proportional to the area of the surface. If the water
level drops 1 cm in the first 5 days, how long will it take for the water
to evaporate completely?
∗ 23. Pieces of ice produced by a refrigerator are in the form of half disks where E and I are constants depending on the cross-section and ma-
(figure below with dimensions in centimetres). As the ice melts, the terial of the board, A is the weight supported by the boy at x = 0, and
ratio of its radius to its thickness remains constant. If the rate of change g > 0 is the acceleration due to gravity.
of the volume of the piece is proportional to its surface area, and its
radius is 1/2 cm after 10 min, when will it completely melt? y
Semicircle
x
1
L L
4 2 2
∗ 24. When a mothball is exposed to the air, it slowly evaporates. The (a) Solve this equation along with the conditions f (0) =
mothball remains spherical at all times, and evaporation is proportional f (L/2) = 0 to find f (x) as a function of x , E , I , m , g ,
to the surface area of the mothball at any given time. If its radius after and A .
one year is half its original radius, when will it disappear completely? (b) Find a condition that permits determination of A .
∗ 25. As a spherical raindrop falls through a cloud, its mass increases ∗ 28. The steady-state temperature T in a region bounded by two con-
at a rate proportional to the product of its surface area and its velocity. centric spheres of radii 1 m and 2 m must satisfy the differential equation
Assuming that the raindrop begins with zero radius, find its radius as a
function of distance fallen through the cloud. d 2T 2 dT
2
+ = 0,
dr r dr
∗ 26. The figure below shows an open cylinder that is always kept full
by the tap. A hole is to be drilled in the side of the cylinder at a point where r is the radial distance from the common centre of the spheres.
where the stream of water will hit the ground as far from the cylinder as If temperatures on the inner and outer spheres are maintained at 10◦ C
possible. The higher the hole, the more time the water has to reach the and 20◦ C respectively, find the temperature distribution T (r) between
ground, and therefore the farther from the cylinder it will reach. On the the spheres.
other hand, the lower the hole, the greater the speed that the water exits
through the hole. Use the modified Torricelli law to find the optimum ∗ 29. A room with volume 100 m 3 initially contains 0.1% carbon diox-
position taking both factors into account. ide. Beginning at time t = 0, fresher air containing 0.05% carbon
dioxide flows into the room at 5 m 3 /min. The well-mixed air in the
room flows out at the same rate. Find the amount of carbon dioxide
FIGURE 5.19 Water escaping from hole in cylinder in the room as a function of time. What is the limit of the function as
t → ∞?
Tap y ∗ 30. It is sometimes assumed that the density ρ of the atmosphere is
related to height h above sea level according to the differential equation
dρ ρ 2−δ
Hole =− ,
dh kδ
H where δ > 1 and k > 0 are constants.
(a) Show that if ρ0 is density at sea level, then
h
" #
h δ−1
ρ δ−1 = − + ρ0δ−1 .
x k δ
(b) If air pressure P and density are related by the equation

∗ 27. Three boys of the same height carry a board of length L and uni- P = kρ δ , prove that
form mass per unit length m horizontally as shown in the figure that
1−1/δ −1/δ
follows. The weight of the board causes it to bend, and its shape is P 1−1/δ = P0 − [(1 − 1/δ)ρ0 P0 ]h,
the same on either side of the middle boy. Between the first two boys,
for 0 ≤ x ≤ L/2, the displacement of the board from the horizontal where P0 is air pressure at sea level.
y = f (x) must satisfy the differential equation (c) Show that the effective height of the atmosphere is
d 2y mg 2 δP0
(EI ) = Ax − x , .
dx 2 2 (δ − 1)ρ0
Summary 371
∗ 31. A certain chemical dissolves in water at a rate proportional to the ∗ 32. Two substances A and B react to form a third substance C in such
product of the amount of undissolved chemical and the difference be- a way that 2 g of A react with 1 g of B to produce 3 g of C . The rate
tween concentration in a saturated solution and the existing concentra- at which C is formed is proportional to the amounts of A and B still
tion in the solution. A saturated solution contains 25 g of chemical in present in the mixture. Find the amount of C present in the mixture
100 mL of solution. If 50 g of chemical is added to 200 mL of water, as a function of time when the original amounts of A and B brought
find a formula for the amount of chemical dissolved as a function of together at time t = 0 are 20 g and 30 g, respectively.
time. Draw its graph.
SUMMARY
The indefinite integral of a function f (x) is a family of functions, each of which has f (x) as
its first derivative. If we can find one antiderivative of f (x) , then the indefinite integral is that
antiderivative plus an arbitrary constant. According to Theorem 5.2, the indefinite integral of a
sum of two functions is the sum of their indefinite integrals, and multiplicative constants may
be bypassed when finding indefinite integrals:
! ! ! ! !
[f (x) + g(x)] dx = f (x) dx + g(x) dx; cf (x) dx = c f (x) dx.
The most important integration formula is for powers of x ,

!  1
n x n+1 + C, n %= −1
x dx = n+1

ln |x| + C, n = −1.
Other integration formulas that arise from differentiations of trigonometric, inverse trigonomet-
ric, exponential, logarithmic, and hyperbolic functions are listed in equations 5.4.
In this chapter we have studied three ways to evaluate indefinite integrals. (Others will
follow in Chapter 8.) First, due to our expertise in differentiation, some antiderivatives are
immediately recognizable. Second, sometimes an antiderivative can be guessed to within a
multiplicative constant, and this constant can then be adjusted. Third, a change of variable can
often replace a complex integration problem with a simpler one.
Integration plays a fundamental role in kinematics. Since velocity is the derivative of
position, and acceleration is the derivative of velocity, it follows that position is the indefinite
integral of velocity, and velocity is the indefinite integral of acceleration,
! !
v(t) = a(t) dt and x(t) = v(t) dt.
Many physical systems are modelled by differential equations, and the solution of a dif-
ferential equation usually involves one or more integrations. The differential equation for the
deflection of a beam requires four integrations, resulting in four arbitrary constants that are
determined by two boundary conditions at each end of the beam. The Dirac-delta function is
most effective in representing point forces on beams.
First order differential equations are separable if they can be expressed in the form N (y) dy =
M(x) dx . Solutions are then defined implicitly by
! !
N (y) dy = M(x) dx.
KEY TERMS
Antiderivative and indefinite integral Integrand

Ramp function Change of variable
Boundary conditions Simply supported
Boundary-value problem Differential equation
First-order differential equation Separable differential equation
REVIEW
EXERCISES
In Exercises 1–28 evaluate the indefinite integral. T and the temperature of the environment. If the water cools to 50◦ C
! ! " # in 10 min, find T as a function of t .
3 2 1 1
1. (3x − 4x + 5) dx 2. + 2x − dx ∗ 31. Find the curve y = f (x) for which f "" (x) = x 2 + 1, and that
x5 x3 passes through the point (1, 1) with slope 4.
! ! " #
1 √
3. (2x 2 − 3x + 7x 6 ) dx 4. − 2 x dx ∗ 32. Find the curve y = f (x) for which f "" (x) = 12x 2 , and that
2
x passes through the two points (1, 4) and (−1, −3) .
! !
√ ∗ 33. Find the curve y = f (x) for which f "" (x) = 24x 2 + 6x , and
5. x − 2 dx 6. x(1 + 3x 2 )4 dx
that is tangent to the line y = 4x + 4 at (1, 8) .
! " # ! " # ∗ 34. A boy lives 6 km from school. He decides to walk to school at a
√ 1 x2 + 5
7. x−√ dx 8. √ dx speed that is always proportional to the square root of his distance from
x x
the school. If he is halfway to school after 1 h, find his distance from
! ! "√ #
1 x 15 school at any time. How long does it take him to reach school?
9. dx 10. − √ dx
(x + 5)4 x2 x ∗ 35. If a ball is thrown vertically upward with a speed of 30 m/s, how
! ! - high will it rise?
11. sin 3x dx 12. x 1 − x 2 dx ∗ 36. A stone is thrown vertically downward over the edge of a bridge
! ! 50 m above a river. If the stone strikes the water in 2.2 s, what was its
13. x cos x 2 dx 14. x 2 (1 − 2x 2 )2 dx initial speed?
! !
√ x
15. x 1 + x dx 16. √ dx In Exercises 37–44 evaluate the indefinite integral.
2−x ! !
! ! 1 x
1 √ ∗ 37. - √ dx ∗ 38. √ dx
17. dx 18. (2 + x)2 dx 1+ x 1+x+1
(1 + x)2 ! !
! ! sin x
1 4 ∗ 39. √ dx ∗ 40. x 8 (3 − 2x 3 )6 dx
19. √ √ dx 20. sin x cos x dx 4 + 3 cos x
x(2 + x)2
! ! ! !
(2 + x)4
21. e 3 − 5x
dx 22. xe −4 x 2
dx ∗ 41. dx ∗ 42. sin3 x cos3 x dx
x6
! ! ! !
ex − 1 1 1
23. dx 24. dx ∗ 43. √ dx ∗ 44. tan x dx
e 2x 5x ln x x 1 + 3 ln x
! ! ∗ 45. A graph of the acceleration a(t) of an object is shown in the figure
x 3
25. √ dx 26. dx below. Find its velocity v(t) and position x(t) in the time interval
1− 4x 4 1+ 7x 2
0 ≤ t ≤ 15 if v(0) = 0 = x(0) , and draw graphs of each.
! !
27. x cosh 5x 2 dx 28. sech 2 5x dx a
Parabola
29. If the number of bacteria in a culture triples in 3 days, when will (5, 2)
it quadruple its original number? 2 Straight line
◦
30. Water at temperature 70 C is placed outside where temperature is 1
−20◦ C. Newton’s law of cooling states that the time rate of change of
the temperature T of the water is proportional to the difference between 5 10 15 t
∗ 46. Find the equation of the curve that passes through the point (1, 1) ∗ 49. Find deflections of the beam in Figure 5.9 if the mass of the block
such that the slope of the tangent line at any point (x, y) is half the is M kg and M is so large that the mass of the beam can be neglected
square of the slope of the line from the origin to (x, y) . in comparison. Is the beam straight for x > 5? Would you expect it
to be?
∗ 47. The equation y = x 3 + C describes a family of cubics where C
represents the distance of its horizontal point of inflection above the ∗ 50. Repeat Exercise 49 if the mass is on the right half of the beam. Is
x -axis. Find the equation of the curve that passes through the point the beam straight for x < 5? Would you expect it to be?
(1, 1) and intersects each of these curves at right angles. ∗ 51. Repeat Exercise 49 if the mass is on the middle half of the beam.
∗ 48. A container in the form of an inverted right-circular cone of radius Are the portions of the beam 0 < x < 5/2 and 15/2 < x < 10
6 cm and height 15 cm is full of water. Water evaporates from the straight? Would you expect them to be?
surface at a rate proportional to the area of the surface. If the water
level drops 1 cm in the first 6 days, how long will it take for half the
water to evaporate?
CHAPTER
6 The Definite Integral
Application Preview Currents in circuits are often monitored by a controller. The controller takes action whenever the
current strays significantly from its expected value. The current i in the left figure below differs
from its steady-state value significantly at time t0 , but does so for a very short time interval, so
short perhaps, that it might be deemed acceptable.
i i
Steady-state Steady-state
current current
t0 t t0 t
On the other hand, the current in the right figure moves only half as far from the steady-state
value as in the left figure, but does so for an extended period of time. This might be deemed
unacceptable.
THE PROBLEM Devise a way to distinguish mathematically, for purposes of the controller,
between very short, but very abnormal behaviour of a function, and long-term, but less dramatic
changes from the norm. (See Example 6.15 on page page 401 for a solution.)
There are two aspects to calculus: differentiation and integration. We dealt with differ-
entiation and its applications in Chapters 3 and 4. Integration in Chapter 5 was synonymous
with antidifferentiation and the indefinite integral. In this chapter we investigate a new type of
integral called the definite integral. Before doing that, we introduce sigma notation, a compact
notation for sums of terms, particularly useful for definite integrals. In Section 6.2 we discuss
four problems that motivate the concept of the definite integral. In subsequent sections we de-
velop the definition for the integral and discuss various ways to evaluate it. Section 6.6 presents
an application of the definite integral — finding the average value of a function — and Chapter
7 develops a multitude of physical applications. The definite integral is, by definition, very
different from the indefinite integral, yet the two are intimately related through the fundamental
theorems of integral calculus (Sections 6.4 and 6.5).
6.1 Sigma Notation

One of the most important notations in calculus, sigma notation, is used to represent a sum of
terms, all of which are similar in form. For example, the six terms in the sum
1 2 3 4 5 6
+ + + + +
1+ 22 1+ 32 1+ 42 1+ 52 1+ 62 1 + 72
374
are all formed in the same way: Each is an integer divided by 1 plus the square of the next
integer. If k represents an integer, we can say that every term has the form k/[1 + (k + 1)2 ];
the first term is obtained by setting k = 1, the second by setting k = 2, and so on, until k = 6.
As k/[1 + (k + 1)2 ] represents each and every term in the sum, we can describe the sum in
words by saying, “Assign k in k/[1 + (k + 1)2 ] the integer values between 1 and 6, inclusively,
and add the resulting numbers together.” The notation used to represent this statement is
6
! k
.
1 + (k + 1)2
k=1
The ! symbol is the Greek capital letter sigma, which in this case means “sum.” Summed are
expressions of the form k/[1 + (k + 1)2 ], and the “ k = 1” and “6” indicate that every integer
from 1 to 6 is substituted into k/[1 + (k + 1)2 ]. We call k/[1 + (k + 1)2 ] the general term
of the sum, since it represents each and every term therein. The letter k is called the index of
summation or variable of summation, and 1 and 6 are called the limits of summation. Any
letter may be used to represent the index of summation; most commonly used are i , j , k , l , m ,
and n . Here is another example:
15
!
n3 = 53 + 63 + 73 + 83 + 93 + 103 + 113 + 123 + 133 + 143 + 153 .
n=5
In summing a large number of terms, it is quite cumbersome to write them all down. One
way around this difficulty is to write the first few terms to indicate the pattern by which the terms
are formed, three dots, and the last term. For example, to indicate the sum of the cubes of the
positive integers less than or equal to 100, we write
13 + 23 + 33 + 43 + 53 + · · · + 1003 ,
where the dots indicate that all numbers between 53 and 1003 are to be filled in according to the
same pattern suggested. Obviously, it would be preferable to express the sum in sigma notation,
100
!
k3,
k=1
which is compact and leaves no doubt as to the pattern by which terms are formed.
EXAMPLE 6.1
Write each of the following sums in sigma notation:
1 4 9 16 169
(a) + + + + ··· +
2·3 3·4 4·5 5·6 14 · 15
16 32 64 128 4096
(b) √ + √ + √ + √ + ··· + √
2 3 4 5 10
SOLUTION
(a) To determine the pattern by which terms are formed, it is often advantageous to write
values of the variable of summation above the terms. If we use i as the variable with
i = 1 corresponding to the first term, we write
i=1 i=2 i=3 i=4 i=?
1 4 9 16 169
+ + + +···+ .
2·3 3·4 4·5 5·6 14 · 15
376 Chapter 6 The Definite Integral
The general term is i 2 /[(i + 1)(i + 2)], and
13
!
1 4 9 16 169 i2
+ + + + ··· + = .
2·3 3·4 4·5 5·6 14 · 15 (i + 1)(i + 2)
i=1
(b) If n is chosen as the index of summation and n = 2 to correspond to the first term,
n=2 n=3 n=4 n=5 n=? 10

16 32 64 128 4096 ! 2n+2
√ + √ + √ + √ +···+ √ = √ .
2 3 4 5 10 n=2
n
The representations for the sums in Example 6.1 in terms of sigma notation are not unique. In
fact, there is an infinite number of representations for each sum. Consider, for example, the sum
represented by
!9
2i+3
√ .
i=1
i+1
If we write out some of the terms in the summation, we find that
!9
2i+3 16 32 64 128 4096
√ = √ + √ + √ + √ + ··· + √ ,
i=1
i+1 2 3 4 5 10
the same sum as that in Example 6.1(b). This sum can also be represented by
6
! 23
!
2 j +6 2m−11
√ and √ .
j =−2
j +4 m=15
m − 13
We can transform any one of these representations into any other by making a change of variable
of summation. For example, if in the summation of Example 6.1(b) we set i = n − 1, then
n = i + 1, and
2n+2 2i+3
√ = √ .
n i+1
For the limits, we find that i = 1 when n = 2, and i = 9 when n = 10. It follows that
!10 !9
2n+2 2i+3
√ = √ .
n=2
n i=1
i+1
Similarly, the changes j = n − 4 and m = n + 9 transform
!10
2n+2
√
n=2
n
into
6
! 23
!
2 j +6 2m−11
√ and √ .
j =−2
j +4 m=15
m − 13
EXAMPLE 6.2
Change each of the following summations into representations that are initiated with the integer 1:
26
! 102
!
i 2 /3 j 2 + 2j + 5
(a) (b)
i2 + i + 1 sin (j + 5)
i=4 j =−3
SOLUTION
(a) To initiate the summation at 1, we want n = 1 when i = 4, so we set n = i − 3.
Then i = n + 3, and by substitution we have
26
! !23 !23
i 2 /3 (n + 3)2/3 (n + 3)2/3
= = .
i2 + i + 1 n=1
(n + 3)2 + (n + 3) + 1 n=1
n2 + 7n + 13
i=4
(b) In this case, we set n = j + 4. Then j = n − 4, and

102
! 106
! 106
!
j 2 + 2j + 5 (n − 4)2 + 2(n − 4) + 5 n2 − 6n + 13
= = .
sin (j + 5) sin (n + 1) sin (n + 1)
j =−3 n=1 n=1
If we examine the results of Example 6.2 and the summations immediately preceding this
example, we soon come to realize that there is a very simple way to change variables. To
illustrate, consider once again
!10
2n+2
√ .
n=2
n
Should we wish to initiate the summation with 1 rather than 2, we lower both limits by 1. To
compensate, we replace each n in the general term by n + 1; the result is
!9 !9
2(n+1)+2 2n+3
√ = √ .
n=1
n+1 n=1
n+1
Similarly, for simplicity in the summation

10
! (n + 4)2
,
n=1
en+4
it would be advisable to lower each n in the general term by 4. This can be done provided that
we raise each limit by 4:
10
! 14
!
(n + 4)2 n2
= .
n=1
en+4 n=5
en
Every summation represented in sigma notation is of the form
n
!
f (i), (6.1)
i=m
where m and n are integers ( n > m ), and f (i) is some function of the index of summation
i . In Example 6.2(a), f (i) = i 2/3 /(i 2 + i + 1) , m = 4, and n = 26; in Example 6.1(a),
f (i) = i 2 /[(i + 1)(i + 2)], m = 1, and n = 13. The following properties of sigma notation
are easily proved by writing out each summation.
THEOREM 6.1
If f (i) and g(i) are functions of i , and m and n are positive integers such that n > m ,
then
n
! n
! n
!
[f (i) + g(i)] = f (i) + g(i); (6.2a)
i=m i=m i=m
n
! n
!
cf (i) = c f (i) (6.2b)
i=m i=m
if c is a constant independent of i .
Compare Theorem 6.1 with Theorem 5.2 in Section 5.1; notice the similarities between
properties of summations in sigma notation and those of indefinite integrals.
We emphasize that sigma notation is simply a concise symbolism used to represent a sum
of terms; it does not evaluate the sum. In the following discussion, we develop formulas for
sums that prove useful
"n in future work.
Summation i=1 i represents the sum of the first n positive integers:
n
!
i = 1 + 2 + 3 + 4 + · · · + (n − 1) + n.
i=1
If we write the terms on the right in reverse order, we have

n
!
i = n + (n − 1) + (n − 2) + · · · + 4 + 3 + 2 + 1.
i=1
Addition of these two equations gives us

n
!
2 i = (n + 1) + (n + 1) + (n + 1) + · · · + (n + 1) + (n + 1) = n(n + 1).
i=1
Consequently,
n
! n(n + 1)
i = . (6.3)
2
i=1
This result can be used to develop formulas for the sums of the squares, cubes, and so on, of the
positive integers. To find the sum of the squares of the first n positive integers, we note that by
expansion and simplification
i 3 − (i − 1)3 = 3i 2 − 3i + 1
for any integer i whatsoever. It follows that

n
! n
!
[i 3 − (i − 1)3 ] = (3i 2 − 3i + 1).
i=1 i=1
But if we write the left-hand side in full, we find

n
!
[i 3 − (i − 1)3 ] = [13 − 03 ] + [23 − 13 ] + [33 − 23 ] + · · · + [n3 − (n − 1)3 ].
i=1
Most of these terms cancel one another, leaving only n3 ; that is,
n
!
[i 3 − (i − 1)3 ] = n3 .
i=1
Thus,
n
!
3
n = (3i 2 − 3i + 1)
i=1
! n n
! n
!
=3 i2 − 3 i+ 1 (using Theorem 6.1)
i=1 i=1 i=1
n
! n(n + 1)
=3 i2 − 3 +n (using formula 6.3).
2
i=1
"n 2
We can solve this equation for i=1 i :
n
! # $
2 1 3 3n(n + 1) n(n + 1)(2n + 1)
i = n + −n = . (6.4)
3 2 6
i=1
A similar procedure beginning with the identity i 4 − (i − 1)4 = 4i 3 − 6i 2 + 4i − 1 yields

n
! n2 (n + 1)2
i3 = . (6.5)
4
i=1
These results can also be established independently of one another by mathematical induction.
(See Appendix A for proofs of formulas 6.3 and 6.4 using this technique.)
EXERCISES 6.1
tan 1 tan 2 tan 3 tan 4 tan 225
In Exercises 1–10 express the sum in sigma notation. Initiate the sum- 8. + + + + ··· +
mation with the integer 1. 2 1 + 22 1 + 32 1 + 42 1 + 2252
1 1 1
9. 43 + 52 + 6 + 1 + + + ··· +
1. 2 · 3 + 3 · 4 + 4 · 5 + 5 · 6 + · · · + 99 · 100 8 92 2518
10. 0.9 + 0.99 + 0.999 + · · · + 0.999 999 999

1 2 3 4 5 10
2. + + + + + ··· +
2 4 8 16 32 1024
In Exercises 11–15 verify by a change of variable of summation that
the two summations are identical.
16 17 18 199
3. + + + ··· + 24
! 27
!
14 + 15 15 + 16 16 + 17 197 + 198 n2 i 2 − 6i + 9
11.
n=1
2n + 1 2i − 5
i=4
√ √ √ √ √ √
4. 1 + 2+ 3+2+ 5+ 6+ 7+ 8 + 3 + · · · + 121 101 99
! 3k − k 2 ! 2 − m − m2
12. √ √
k=2
k+5 m=0
7+m
1 1 1 1
5. 1 + + + + ··· +
2 2·3 2·3·4 2 · 3 · 4 · 5 · · · · · 16 20 16
! 2n ! 2j
13. (−1)n 16(−1)j
n2 + 1 j 2 + 8j + 17
6. −2 + 3 − 4 + 5 − 6 + 7 − 8 + · · · − 1020 n=5 j =1
37
! 39
!
33i 33m
2·3 6·7 10 · 11 14 · 15 414 · 415 14.
7. + + + + ··· + i! 729(m − 2)!
1·4 5·8 9 · 12 13 · 16 413 · 416 i=0 m=2
225
! 220
!
1 1 ∗ 29. Prove formula 6.5.
15. n
% n
&% n &
r=15
r 2 − 10r n=10
n2 − 25 ! ! !
∗ 30. Is [f (i)g(i)] equal to f (i) g(i) ?
i=1 i=1 i=1
In Exercises 16–25 use Theorem 6.1 and formulas 6.3–6.5 to evaluate ∗ 31. A finite geometric series is a sum of terms of the form
the sum.
a + ar + ar 2 + ar 3 + · · · + ar n−1 .
12
! 21
!
16. (3n + 2) 17. (2j 2 + 3j ) There is a first term a , and every term thereafter is obtained by multi-
n=1 j =1 plying the preceding term by r (called the common ratio).
n
! 29
! (a) If Sn represents this sum, express Sn in sigma notation.
18. (4m − 2)2 19. (k 3 − 3k 2 )
a(1 − r n )
m=1 k=2 (b) Prove that Sn = .
25 n
1−r
! !
20. (n + 5)(n − 4) 21. i(i − 3)2
n=1 i=1
Use the formula in Exercise 31 to sum the finite geometric series in
24
! 17
!
2 3 2 Exercises 32–35.
∗ 22. (n − 5) ∗ 23. (i − 3i )
n=10 i=7 1 1 1 1 1
∗ 32. + + + + ··· +
n
! 2n
! 8 16 32 64 1 048 576
∗ 24. (k + 3)(k + 4) ∗ 25. (i 2 + 2i − 3) 1 1 1 1 1
k=5
∗ 33. 1 − + − + − ··· −
i=n 3 9 27 81 19 683
∗ 34. 40(0.99) + 40(0.99)2 + 40(0.99)3 + · · · + 40(0.99)15
∗ 26. Verify Theorem 6.1. √
n ∗ 35. 0.99 + 0.99 + (0.99)3/2 + (0.99)2 + · · · + (0.99)10
! 1 1 1 1 ' n '
∗ 27. Find a formula for . Hint: = − . '! ' ! n
k(k + 1) k(k + 1) k+1 ' '
k=1
k ∗ 36. Prove that ' f (i)' ≤ |f (i)| .
' '
i=1 i=1
∗ 28. If f (x) is a function of x , defined for all x , simplify the sum
∗∗ 37. Express the following summation in sigma notation:
n
!
[f (i) − f (i − 1)]. 1 1 1 1 1 1 1 1
1+ − − + + − − + ··· + .
i=1 2 4 8 16 32 64 128 4096
6.2 The Need for the Definite Integral

In this section we consider four inherently different problems: one on area, one on volume, one
on blood flow, and one on work; but we shall see that a common method of solution exists for all
four of them. This common theme leads to the definition of the definite integral in Section 6.3.
Problem 1
We first consider the problem of finding the area A in Figure 6.1a. At present, we have formulas
for areas of very few geometric shapes — rectangles, triangles, polygons, and circles, and the
FIGURE 6.1a Area under a curve y = f (x) FIGURE 6.1b Approximation of area by rectangles
y y
y = f (x) y = f (x)
A1 A2
A Ai
An
a b x x1 x2 xi−1 xi xn−1 x
a = x0 b = xn
shape in Figure 6.1a is not one of them. We can, however, find an approximation to the area by
constructing rectangles in the following way. Between a and b pick n − 1 points x1 , x2 , x3 ,
. . . , xn−1 on the x -axis such that
a = x0 < x1 < x2 < · · · < xn−1 < xn = b.
Draw vertical lines through each of these n − 1 points to intersect the curve y = f (x) and
form rectangles, as shown in Figure 6.1b.
If Ai denotes the area of the i th such rectangle ( i = 1, . . . , n ), then
Ai = (height of rectangle)(width of rectangle)

= f (xi )(xi − xi−1 ).
The sum of these n rectangular areas is an approximation to the required area; that is, A is
approximately equal to
n
! n
!
Ai = f (xi )(xi − xi−1 ).
i=1 i=1
If we let the number of these rectangles get larger and larger (and at the same time require each
to have smaller and smaller width that eventually approaches zero), the approximation appears
to get better and better. In fact, we expect that
n
! n
!
A = lim Ai = lim f (xi )(xi − xi−1 ). (6.6)
n→∞ n→∞
i=1 i=1
Problem 2
If the area in Figure 6.1a is rotated around the x -axis, it traces out a volume V . This is
certainly not a standard shape for which we have a volume formula, and we therefore consider
finding an approximation to V . We take the rectangles in Figure 6.1b (which approximate the
area) and rotate them around the x -axis (Figure 6.2). Each Ai traces out a disc of volume Vi
( i = 1, . . . , n ), where
Vi = (surface area of disc)(thickness of disc)

= π (radius of disc)2 (thickness of disc)
= π [f (xi )]2 (xi − xi−1 ).
FIGURE 6.2 Approximation of volume by discs
y = f (x)
f (xi)
a x1
xi−1
xi
Vi b
xn−1 x
An approximation to the required volume is then

n
! n
!
Vi = π [f (xi )]2 (xi − xi−1 ).
i=1 i=1
Again we feel intuitively that as the number of discs becomes larger and larger (and the width
of each approaches zero), the approximation becomes better and better, and
n
! n
!
V = lim Vi = lim π [f (xi )]2 (xi − xi−1 ). (6.7)
n→∞ n→∞
i=1 i=1
FIGURE 6.3 Blood flow

through an artery Problem 3
Cross-section of
When blood flows through a vein or artery, it encounters resistance due to friction with the walls
blood vessel
of the blood vessel and due to the viscosity of the blood itself. As a result, the velocity of the
blood is not constant across a cross-section of the vessel; blood flows more quickly near the
centre of the vessel than near its walls. It has been shown that for laminar blood flow in a vessel
ri−1
of circular cross-section (Figure 6.3), the velocity of blood is given by
r2 ri rn = R
r1 v = v(r) = c(R 2 − r 2 ), 0 ≤ r ≤ R,
where c > 0 is a constant, R is the radius of the blood vessel, and r is radial distance measured
from the centre of the vessel. We wish to find the rate of blood flow through the vessel, that is,
the volume of blood flowing through the cross-section per unit time.
If v were constant over the cross-section, then flow per unit time would be the product of
v and the cross-sectional area. Unfortunately, this is not the case, but we can still use the idea
that flow is velocity multiplied by area. We divide the cross-section into rings with radii
0 = r0 < r1 < r2 < · · · < rn−1 < rn = R.
Over the i th ring, the variation in v is small and v can be approximated by v(ri ) . The flow
through the i th ring can therefore be approximated by
Fi = (area of ring)(velocity at outer radius of ring)

= (π ri2 − π ri−
2
1 )v(ri ).
An approximation to the required flow F is the sum of these Fi :

n
! n
!
Fi = (π ri2 − π ri−
2
1 )v(ri ),
i=1 i=1
and it seems reasonable that as the number of rings increases, so does the accuracy of the
approximation; that is, we anticipate that
n
! n
!
F = lim Fi = lim (π ri2 − π ri−
2
1 )v(ri ). (6.8)
n→∞ n→∞
i=1 i=1
Problem 4
A spring is fixed horizontally into a wall at one end, and the other end is free. Consider finding
the work to stretch the spring 3 cm by pulling on its free end (Figure 6.4a).
FIGURE 6.4a Initial and final positions of stretched spring FIGURE 6.4b Calculating work to stretch spring
Unstretched
position
3
0 = x0 3 = xn
x
Stretched x1 xi
position x2 xi−1
Let us choose an x -axis positive to the right with x = 0 at the position of the free end of
the spring when it is in the unstretched position (Figure 6.4b). In order to calculate the work
to stretch the spring, we must know something about the forces involved. It has been shown
experimentally that the force F that must be exerted on the free end of the spring in order to
maintain a stretch x in the spring is proportional to x :
F = F (x) = kx,
where k > 0 is a constant. This then is the force that will perform the work. The basic
definition of work W done by a constant force F acting along a straight-line segment of length
d is W = F d . Unfortunately, our force F (x) is not constant; it depends on x , and we cannot
therefore simply multiply force by distance. What we can find, however, is an approximation
to the required work by dividing the length between x = 0 and x = 3 into n subintervals by
n − 1 points x1 , x2 , . . . , xn−1 such that
0 = x0 < x1 < x2 < · · · < xn−1 < xn = 3.
When the spring is stretched between xi−1 and xi , the force necessary to maintain this stretch
does not vary greatly and can be approximated by F (xi ) . It follows then that the work necessary
to stretch the spring from xi−1 to xi is approximately equal to
Wi = F (xi )(xi − xi−1 ).
As a result, an approximation to the total work required to pull the free end of the spring from
x = 0 to x = 3 is
n
! n
!
Wi = F (xi )(xi − xi−1 ).
i=1 i=1
Once again we expect that as n becomes indefinitely large, this approximation approaches the
required work W , and
n
! n
!
W = lim Wi = lim F (xi )(xi − xi−1 ). (6.9)
n→∞ n→∞
i=1 i=1
Each of these four problems on area, volume, blood flow, and work has been tackled in the
same way, and the method can be described qualitatively as follows.
The quantity to be calculated, say W , cannot be obtained for the object G given because
no formula exists. As a result, n smaller objects, say Gi , are constructed. The Gi are chosen
in such a way that the quantity W can be calculated, exactly or approximately, for each Gi , say
Wi . Then an approximation for W is
n
!
Wi .
i=1
If the number of Gi is increased indefinitely, this approximation becomes more and more
accurate and
n
!
W = lim Wi .
n→∞
i=1
It is this limit-summation process that we discuss throughout the remainder of the chapter. We
begin in Section 6.3 with a mathematical description of the process, and by doing so, we obtain a
unified approach to the whole idea. We then discover that there is a very simple way to calculate
these limits. At that point we will be ready to use the technique in a multitude of applications,
including the four problems in this section.
6.3 The Definite Integral

The four problems of Section 6.2 have a common theme: the limit of a summation. By means
of a summation we approximated some quantity (area, volume, blood flow, work), and the limit
led, at least intuitively, to an exact value for the quantity. In this section, we investigate the
mathematics of the limit summation — but only its mathematics. We concentrate here on what
a definite integral is, and how to evaluate it; interpretation of the definite integral as area, volume,
work, and so on, is made in Chapter 7.
To define the definite integral of a function f (x) on an interval a ≤ x ≤ b (Figure 6.5),
we divide the interval into n subintervals by any n − 1 points:
a = x0 < x1 < x2 < · · · < xi < · · · < xn−1 < xn = b.
FIGURE 6.5 Defining the definite integral of f (x) from x = a to x = b
y = f (x)
f (x *1) f (x *i ) f (x *n )
x *1 x*i x*n
a = x0 x1 xi b = xn x
xi −1 xn−1
Next we choose in each subinterval xi−1 ≤ x ≤ xi any point xi∗ whatsoever, and evaluate
f (xi∗ ) .
We now form the sum
n
!
f (x1∗ )(x1 − x0 ) + f (x2∗ )(x2 − x1 ) + · · · + f (xn∗ )(xn − xn−1 ) = f (xi∗ )(xi − xi−1 )
i=1
n
!
= f (xi∗ ) #xi ,
i=1
where we have set #xi = xi − xi−1 . We denote by '#xi ' the length of the longest of the n
subintervals,
'#xi ' = max |#xi |.
i=1,...,n
It is often called the norm of the particular partition of a ≤ x ≤ b into the subintervals
#xi . With this notation, we are ready to define the definite integral of f (x) . It is the limit
6.3 The Definite Integral 385
of the summation above as the number of subintervals becomes increasingly large and every
subinterval shrinks to a point. An easier way to say this is to take the limit as the norm of the
partition approaches zero. In other words, we define the definite integral of f (x) with respect
to x from x = a to x = b as
( b n
!
f (x) dx = lim f (xi∗ ) #xi , (6.10)
a '#xi '→0
i=1
provided that the limit exists. If the limit exists, but is dependent on the choice of subdivision
#xi or star points xi∗ , then the definite integral is of little use. We stipulate, therefore, that in
order for the definite integral to exist, the limit of the sum in equation 6.10 must be independent
of the manner of subdivision of the interval a ≤ x ≤ b and choice of star points in the
subintervals. At first sight this requirement might seem rather severe, since we must now check
that all subdivisions and all choices of star points lead to the same limit before concluding
that the definite integral exists. Fortunately, however, the following theorem indicates that for
continuous functions, this is unnecessary. A proof of this theorem can be found in advanced
books on mathematical analysis.
THEOREM 6.2
If a function f (x) is continuous on a finite interval a ≤ x ≤ b , then the definite integral
of f (x) with respect to x from x = a to x = b exists.
For a continuous function, the definite integral exists, and any choice of subdivision and star
points leads to its correct value through the limiting process. We call f (x) on the left-hand side of
"n the integrand, and a and b the lower and upper limits of integration, respectively.
equation 6.10
The sum i=1 f (xi∗ ) #xi is called a Riemann sum, and because of this, definite integral 6.10
is also called the Riemann integral. The integral was named after German mathematician
G. F. B. Riemann (1826–1866), who introduced the notion of the definite integral as a sum.
EXAMPLE 6.3
Evaluate the definite integral
( 1
x 2 dx.
0
FIGURE 6.6 Definite SOLUTION Since f (x) = x 2 is continuous on the interval 0 ≤ x ≤ 1, the definite integral
integral of f (x) = x 2 from x = 0 exists, and we may choose any subdivision and star points in its evaluation. The simplest
to x = 1 partition is into n equal subintervals of length 1/n by the points (Figure 6.6)
y
i
xi = , i = 0, . . . , n.
(1, 1) n
y = x2 We choose for star points the right end of each subinterval; that is, in xi−1 ≤ x ≤ xi , we choose
xi∗ = xi = i/n . Then, by equation 6.10, we have
( 1 n
! n
!
2
n−1 x dx = lim f (xi∗ ) #xi = lim (xi∗ )2 #xi .
0 '#xi '→0 '#xi '→0
n i=1 i=1
1 2 3 1 x
i Since all subintervals have equal length #xi = 1/n , the norm of the partition is '#xi ' = 1/n ,
n n n xi =
i−1 n and taking the limit as '#xi ' → 0 is tantamount to letting n → ∞ . Thus,
xi −1 =
n ( n ) *2 ) * n
1 ! i 1 1 !
2
x dx = lim = lim i2.
0 n→∞ n n n→∞ n3
i=1 i=1
If we now use formula 6.4 for the sum of the squares of the first n positive integers, we obtain
( 1
1 n(n + 1)(2n + 1) 1
x 2 dx = lim = .
0 n→∞ n3 6 3
This example illustrates that even for an elementary function such as f (x) = x 2 , evaluation of
the definite integral by equation 6.10 is quite laborious. In fact, had we not known formula 6.4 for
the sum of the squares of the integers, we would not have been able to complete the calculation.
Imagine the magnitude of the problem were the integrand equal to f (x) = x(x + 1)−2/3 . In
other words, if definite integrals are to be at all useful, we must find a simpler way to evaluate
them. This we do in Section 6.4, but in order to stress the definite integral as a limit summation,
we consider one more example.
EXAMPLE 6.4
Evaluate the definite integral
( 1
(5x − 2) dx.
−1
FIGURE 6.7 Definite SOLUTION Since f (x) = 5x − 2 is continuous on the interval −1 ≤ x ≤ 1, the definite
integral of f (x) = 5x − 2 from integral exists, and we may choose any partition and star points in its evaluation. For n equal
x = −1 to x = 1 subdivisions of length 2/n , we use points (Figure 6.7)
y 2i
f(x) = 5x − 2
xi = −1 + , i = 0, . . . , n.
n
xn−2 If we choose the right end of each subinterval as a star point, that is, xi∗ = xi = −1 + 2i/n ,
x0 x2 xn−1 then equation 6.10 gives
x1 xn
( 1 n
! n
!
−1 4−n 1 x
(5x − 2) dx = lim f (xi∗ ) #xi = lim (5xi∗ − 2) #xi .
2−n n −2 n −2 −1 '#xi '→0 '#xi '→0
i=1 i=1
n n−4 n
n Once again all subintervals have equal length #xi = 2/n , and therefore we may replace
'#xi ' → 0 with n → ∞ ,
( 1 n # )
! * $) * n #
! $
2i 2 20i − 14n
(5x − 2) dx = lim 5 −1 + −2 = lim .
−1 n→∞ n n n→∞ n2
i=1 i=1
We can break the summation into two parts, and take constants outside each summation to obtain
( % n n
& % n n
&
1 ! 20i ! 14 20 ! 14 !
(5x − 2) dx = lim − = lim i− 1 .
−1 n→∞ n2 n n→∞ n2 n
i=1 i=1 i=1 i=1
With formula 6.3 for the sum of the first n positive integers,
( 1 # $ # $
20 n(n + 1) 14 10 − 4n
(5x − 2) dx = lim − n = lim = −4 .
−1 n→∞ n2 2 n n→∞ n
Note that in Example 6.3 the value of the definite integral is positive, and in Example 6.4 it is
negative. The value of the definite integral in Exercise 8 is zero. In other words, the value of a
definite integral can be positive, negative, or zero, depending on the limits and the integrand.
EXERCISES 6.3
In Exercises 1–8 use equation 6.10 to evaluate the definite integral. ∗∗ 13. Use the formula
( 1 ( 2
(n + 1)θ nθ
∗ 1. x dx ∗ 2. 3x dx n
! cos sin
0 0 cos iθ = 2 2
θ
( 1 ( 2
i=1 sin
2
∗ 3. (3x + 2) dx ∗ 4. x 3 dx
0 0 ( π/2
( 2 ( 0 to evaluate cos x dx .
∗ 5. (x 2 + 2x) dx ∗ 6. (−x + 1) dx 0
1 −1
( b
( 1 ( 1 ∗∗ 14. In this exercise we evaluate x k dx for b > a > 0 and any
∗ 7. x 2 dx ∗ 8. x 3 dx a
−1 −1 k (= −1.
( 1 (a) Let h = (b/a)1/n and subdivide the interval a ≤ x ≤ b
∗ 9. Evaluate the definite integral x 15 dx . into n subintervals by the points
−1
( 1
∗ 10. x0 = a, x1 = ah, x2 = ah2 , ...,
(a) Consider the definite integral 2x dx . Show that when
0
the interval 0 ≤ x ≤ 1 is subdivided into n equal subin- xi = ahi , ..., xn = ahn = b.
tervals, and star points are chosen as right-hand endpoints
in each subinterval, equation 6.10 leads to
Show that with the choice of xi∗ = xi in the i th subinterval
( 1 n
! xi−1 ≤ x ≤ xi , equation 6.10 gives
1
2x dx = lim 2i/n .
0 n→∞ n i=1 ( %) * n &
b
k k+1 h − 1 ! k+1 i
x dx = a lim (h ) .
(b) Use the formula in Exercise 31(b) of Section 6.1 to express a n→∞ h i=1
the summation in closed form,
( 1
21/n (b) Use the result of Exercise 31 in Section 6.1 to write the
2x dx = lim .
0 n→∞ n(21/n − 1) summation in closed form, and hence show that
(c) Use L’Hôpital’s rule to evaluate this limit, and hence find ) *k/n %) *1/n &
the value of the definite integral. b b
( −1
b a a
( 3 x k dx = (bk+1 − a k+1 ) lim ) *(k+1)/n .
a n→∞ b
∗ 11. Use the technique of Exercise 10 to evaluate ex dx . −1
1 a
∗∗ 12. Use the formula

(c) Use L’Hôpital’s rule to evaluate this limit, and hence obtain
(n + 1)θ nθ
n
! sin sin (
sin iθ = 2 2 b
bk+1 − a k+1
θ x k dx = .
i=1 sin a k+1
2
( π
to evaluate sin x dx . ∗∗ 15. Show that the definite integral of the function in Exercise 52 of
0 Section 2.4 does not exist on any interval a ≤ x ≤ b whatsoever.
6.4 The First Fundamental Theorem of Integral Calculus

In Section 6.3 we demonstrated how to evaluate definite integrals using definition 6.10. Inte-
grands x 2 and 5x − 2 in Examples 6.3 and 6.4 are very simple polynomials, as are the integrands
in questions 1–9 of Exercises 6.3, but in spite of this, calculations were frequently laborious, and
invariably required summation formulas from Section 6.1. We promised a very simple technique
that would replace these calculations, and this is the substance of the first fundamental theorem
of integral calculus.
THEOREM 6.3 (First Fundamental Theorem of Integral

Calculus)
If f (x) is continuous on the interval a ≤ x ≤ b , and F (x) is an antiderivative of f (x)
thereon, then
( b
f (x) dx = F (b) − F (a). (6.11)
a
PROOF Since f (x) is continuous on a ≤ x ≤ b , the definite integral of f (x) from x = a

to x = b exists and is defined by equation 6.10, where we are at liberty to choose the #xi and
xi∗ in any way whatsoever. For any choice of #xi , a convenient choice for the xi∗ can be found
by applying the mean value theorem n times to F (x) , once on each subinterval xi−1 ≤ x ≤ xi .
This is possible since F ) (x) = f (x) is continuous for a ≤ x ≤ b . The mean value theorem
states that for each subinterval, there exists at least one point ci between xi−1 and xi such that
F (xi ) − F (xi−1 )
= F ) (ci ), i = 1, . . . , n.
xi − xi−1
But F ) (ci ) = f (ci ) , so that
F (xi ) − F (xi−1 ) = f (ci ) #xi , i = 1, . . . , n.
If we now choose xi∗ = ci , then
f (xi∗ )#xi = F (xi ) − F (xi−1 ),
and equation 6.10 gives

( b n
! n
!
f (x) dx = lim f (xi∗ ) #xi = lim [F (xi ) − F (xi−1 )].
a '#xi '→0 '#xi '→0
i=1 i=1
When we write out all terms in the summation, we find that many cancellations take place:
( b
f (x) dx = lim {[F (x1 ) − F (x0 )] + [F (x2 ) − F (x1 )] + · · · + [F (xn ) − F (xn−1 )]}
a '#xi '→0
= lim {F (xn ) − F (x0 )}

'#xi '→0
= lim {F (b) − F (a)}

'#xi '→0
= F (b) − F (a).
If we introduce the notation

{F (x)}ba
to represent the difference F (b) − F (a) , then Theorem 6.3 can be expressed in the form
( b +( ,b
f (x) dx = f (x) dx . (6.12)
a a
This is a fantastic result. No longer is it necessary to consider limits of summations in order to

evaluate definite integrals. We simply find an antiderivative of the integrand, substitute x = b
and x = a , and subtract. For instance, to evaluate the definite integral in Example 6.3, we
easily write
( 1 + ,1
2 x3 1 1
x dx = = −0 = .
0 3 0 3 3
Note that had we used the indefinite integral x 3 /3 + C for x 2 , we would have had
( 1 + ,1 + ,
2 x3 1 1
x dx = +C = +C −C = .
0 3 0 3 3
Because the arbitrary constant always vanishes in the evaluation of definite integrals, we need
not use the indefinite integral in this context; any antiderivative will do.
Similarly, for Example 6.4, we obtain
( 1 + ,1 + , + ,
5x 2 5 5
(5x − 2) dx = − 2x = −2 − +2 = −4 .
−1 2 −1 2 2
EXAMPLE 6.5
Evaluate the following definite integrals:
( 2 ( 4
2
√
(a) (3x − x + 4) dx (b) x + 4 dx
1 −2
SOLUTION
( 2 + ,2 + ,
x2 1 19
(a) (3x 2 − x + 4) dx = x3 − + 4x = {8 − 2 + 8} − 1 − + 4 =
1 2 1 2 2
( 4 + ,4 √
√ 2 2 2 28 2
(b) x + 4 dx = (x + 4)3/2 = (8)3/2 − (2)3/2 =
−2 3 −2 3 3 3
EXAMPLE 6.6
Evaluate
( 3 ) *
1 3
+ 3x dx.
1 x2
SOLUTION Since −1/x + 3x 4 /4 is an antiderivative for 1/x 2 + 3x 3 for 1 ≤ x ≤ 3,
( 3 ) * + ,3 + , + ,
1 3 1 3x 4 1 243 3 182
+ 3x dx = − + = − + − −1 + = .
1 x2 x 4 1 3 4 4 3
EXAMPLE 6.7
( 1
1
Can the definite integral dx be evaluated with Theorem 6.3?
−1 x2
SOLUTION No! Theorem 6.3 requires the integrand 1/x 2 to be continuous on the interval
−1 ≤ x ≤ 1, and this is not the case. The function is discontinuous at x = 0.
There is a difficulty with Theorem 6.3. It is subtle, but important. The theorem states that to
evaluate definite integrals of continuous functions, we use antiderivatives. But how do we know
that continuous functions have antiderivatives? We don’t yet. This fact will be established in
Section 6.5 when we verify the second fundamental theorem. You might ask why the second
fundamental theorem is not proved first. Would it not be more logical first to establish existence
of antiderivatives, and then use this fact to prove Theorem 6.3? From a logic point of view, the
answer is yes. However, from a practical point of view, Theorem 6.3 is so useful we want to
give it every possible emphasis. To prove the second fundamental theorem first would detract
from the importance and simplicity of Theorem 6.3.
Before we move on to the second fundamental theorem, we present the following theorems,
which describe properties of the definite integral.
THEOREM 6.4
If f (x) is continuous on a ≤ x ≤ b , then:
( a ( b
(i) f (x) dx = − f (x) dx; (6.13a)
b a
( b ( c ( b
(ii) f (x) dx = f (x) dx + f (x) dx. (6.13b)
a a c
THEOREM 6.5
If f (x) and g(x) are continuous on a ≤ x ≤ b , then:
( b ( b ( b
(i) [f (x) + g(x)] dx = f (x) dx + g(x) dx; (6.14a)
a a a
( b ( b
(ii) kf (x) dx = k f (x) dx, (6.14b)
a a
when k is a constant.
Properties 6.14 are analogous to 5.3 for indefinite integrals. Theorems 6.4 and 6.5 can be
proved using either Theorem 6.3 or equation 6.10.
We can also establish the following property.
THEOREM 6.6
When f (x) is continuous on a ≤ x ≤ b and m ≤ f (x) ≤ M on this interval,
( b
m(b − a) ≤ f (x) dx ≤ M(b − a). (6.15)
a
PROOF By equation 6.10, we can write

( b n
!
f (x) dx = lim f (xi∗ ) #xi
a '#xi '→0
i=1
n
!
≤ lim M #xi
'#xi '→0
i=1
n
!
=M lim #xi
'#xi '→0
i=1
=M lim [(x1 − a) + (x2 − x1 ) + (x3 − x2 ) + · · · + (b − xn−1 )]

'#xi '→0
=M lim (b − a)
'#xi '→0
= M(b − a).
A similar proof establishes the inequality involving m .
EXAMPLE 6.8
Use Theorem 6.6 to find a maximum possible value for
( 4
sin x 2
dx.
1 1 + x2
SOLUTION Clearly, sin x 2 ≤ 1 for all x , and on the interval 1 ≤ x ≤ 4. The largest value
of 1/(1 + x 2 ) is 1/2. Consequently, (sin x 2 )/(1 + x 2 ) ≤ 1/2 for 1 ≤ x ≤ 4 and, by Theorem
6.6,
( 4
sin x 2 1 3
2
dx ≤ (4 − 1) = .
1 1 + x 2 2
Velocity and Speed Revisited Once Again

Suppose that v(t) = 3t 2 − 6t − 105 represents the velocity (in metres per second) of a particle
moving along the x -axis beginning at time t = 0. We can easily calculate the definite integral
of v(t) between any two times, say t = 0 and t = 12. By doing so, we get our first glimpse
of definite integrals at work in applied problems.
( 12 ( 12 - .12
v(t) dt = (3t 2 − 6t − 105) dt = t 3 − 3t 2 − 105t = 36.
0 0 0
Realizing that integration is a limit summation, and what is being added are products of velocities
v(t) multiplied by small time increments dt , we interpret 36 as the displacement of the particle
at time t = 12 s relative to its displacement at t = 0 s. Although we do not have enough
information to determine where the particle is at any given time, we can say that at t = 12 s, it
is 36 m to the right of where it is at t = 0 s. In general, when v(t) is the velocity of a particle
moving along the x -axis,
( b
v(t) dt (6.16)
a
is the displacement of the particle at time t = b relative to its displacement at time t = a . If

the definite integral is positive, then at time t = b the particle is to the right of its position at
t = a ; and if the integral is negative, then at t = b , the particle is to the left of its position at
t = a.
Speed is the magnitude of velocity. If we integrate speed between the same limits, we get
a different result,
( 12 ( 12
|v(t)| dt = |3t 2 − 6t − 105| dt.
0 0
Since 3t 2 − 6t − 105 < 0 for 0 ≤ t < 7, and 3t 2 − 6t − 105 > 0 for 7 < t ≤ 12, we divide
the integration into two parts,
( 12 ( 7 ( 12
|v(t)| dt = −(3t 2 − 6t − 105) dt + (3t 2 − 6t − 105) dt
0 0 7
- .7 - .12
= −t 3 + 3t 2 + 105t + t 3 − 3t 2 − 105t
0 7
= 1114.
Since this integral adds products of speed |v(t)| and time increments dt , which we interpret as
distance travelled, 1114 m must be the distance travelled by the particle between t = 0 s and
t = 12 s. In general,
( b
|v(t)| dt (6.17)
a
is the distance travelled between times t = a and t = b .
These ideas are reinforced by graphs of velocity and speed in Figures 6.8. Because velocity
is negative for 0 < t < 7, the particle is moving to the left at these times. Products v(t) dt
are negative during this time interval, and therefore contribute negatively to the definite integral
of v(t) . The particle stops at t = 7 s, and then moves to the right from t = 7 s to t = 12 s.
Products v(t) dt then contribute positively to the definite integral, with the ultimate result being
36 m. At time t = 12 s, the particle is 36 m to the right of its position at t = 0 s. On the other
hand, speed is always positive, and therefore products |v(t)| dt always contribute positively to
the integral of speed.
FIGURE 6.8a Velocity function v(t) = 3t 2 − 6t − 105 FIGURE 6.8b Speed function |v(t)| = |3t 2 − 6t − 105|
v |v|
250 250
200 200
150
100 150
50 100
2 4 6 8 10 12 t 50
−50
−100
2 4 6 8 10 12 t
EXERCISES 6.4
( 1 ( −1
In Exercises 1–40 evaluate the definite integral. 1
3. (4x 3 + 2x) dx 4. dx
−1 −3 x2
( 4 ( 3 ( 2 ) * ( π/2
3 2 2 3
1. (x + 3) dx 2. (x − 2x + 3) dx 5. x + dx 6. sin x dx
3 1 4 x3 0
( 1 ( −2 ) * ( 3 ( 1
1 (x + 1)2
7. (x 2 − 1 − x 4 ) dx 8. − 2x dx 29. dx 30. 34x dx
−1 −1 x2 1 x 0
( 5 ( 4
( ( 31. |x| dx 32. x|x + 1| dx
2 1
4 2 2 0 0
9. (x + 3x + 2) dx 10. x(x + 1) dx ( (
5 1
1 0
∗ 33. |x| dx ∗ 34. x|x + 1| dx
−5 −2
( 1 ( 2π ( 1/ 2 ( 1
1 1
11. x 2 (x 2 + 1)2 dx 12. cos 2x dx ∗ 35. √ dx ∗ 36. dx
0 0 − 1/ 2 1 − x2 −1 1 + x2
( 3 ( 1
1
( 3 2 ( 1 ∗ 37. √ dx ∗ 38. cosh 2x dx
x +3 x x2 − 1
13. dx 14. (x 2.2 − x π ) dx 2 0
1 x2 0 ( 1 ( 1/ 2
2 1
∗ 39. csch x dx ∗ 40. dx
−1 0 1 + 4x 2
( 1 ( 4 2 2
(x − 1)
15. x 2 (x 3 − x) dx 16. dx
−1 3 x2
In Exercises 41–46 v(t) represents the velocity of a particle moving
along the x -axis. Calculate the definite integrals of v(t) and |v(t)|
( 2 ) * ( 3 between the two times shown. Interpret each number and plot (or
√ 1
17. x−√ dx 18. (x − 1)3 dx draw) a graph of the velocity function to corroborate results.
1 x −2
∗ 41. v(t) = 3t 2 − 6t − 105, t = 0 to t = 9
( 4 ( π/4 ∗ 42. v(t) = 3t 2 − 6t − 105, t = 0 to t = 5
(x 2 − 1)(x 2 + 1)
19. dx 20. 3 cos x dx
2 x2 0 ∗ 43. v(t) = −t 2 + 3t − 2, t = 0 to t = 2
∗ 44. v(t) = −t 2 + 3t − 2, t = 1 to t = 2
( (
π/4 π
∗ 45. v(t) = t 3 − 3t 2 + 2t , t = 0 to t = 2
21. sec2 x dx 22. sin x cos x dx
0 π/2 ∗ 46. v(t) = t 3 − 3t 2 + 2t , t = 0 to t = 3
( π/3 ( π/4 In Exercises 47–52 use Theorem 6.6 to find maximum and minimum
23. csc2 3x dx 24. sec x tan x dx values for the integral.
π/6 −π/4
( π/4 ( π/2
sin x sin x
( ( ∗ 47. dx ∗ 48. dx
2 2 0 1+ x2 0 1+x
x x
25. 2 dx 26. e dx ( π ( π/2
0 −1 sin x sin 2x
∗ 49. dx ∗ 50. dx
0 2 + x2 π/4 10 + x 2
( 1 ( −2 ( 1 ( 3/
1
27. e3x dx 28. dx ∗ 51. (1 + 4x 4 ) cos (x 2 ) dx ∗ 52. 4 + x 3 dx
0 −3 x 0 1
6.5 The Second Fundamental Theorem of Integral Calculus

The first fundamental theorem of integral calculus in Section 6.4 allows us to use antiderivatives
to evaluate definite integrals of continuous functions. We now show that continuous functions
always have antiderivatives.
When f (t) is continuous for a ≤ t ≤ b , its definite integral
( b
f (t) dt
a
is a number. If b is changed but a is kept fixed, the value of the definite integral changes; for
each value of b , there is a new value for the definite integral. In other words, the value of the
definite integral is a function of its upper limit. Suppose we replace b by x , and denote the
resulting function by F (x) ,
( x
F (x) = f (t) dt.
a
We now show that the derivative of F (x) is f (x) ; that is, F (x) is an antiderivative of f (x) .
THEOREM 6.7 (The Second Fundamental Theorem of Integral

Calculus)
When f (x) is continuous for a ≤ x ≤ b , the function
( x
F (x) = f (t) dt (6.18)
a
is differentiable for a ≤ x ≤ b , and F ) (x) = f (x) .
PROOF If x is any point in the open interval a < x < b , then h can always be chosen
sufficiently small that x + h is also in the interval a < x < b . By equation 3.3, the derivative
of F (x) at this x is defined as
F (x + h) − F (x)
F ) (x) = lim
h→0 h
#( x+h ( x $
1
= lim f (t) dt − f (t) dt
h→0 h a a
#( x+h ( a $
1
= lim f (t) dt + f (t) dt (by property 6.13a)
h→0 h a x
( x+h
1
= lim f (t) dt (by property 6.13b).
h→0 h x
According to property 6.15,

( x+h
mh ≤ f (t) dt ≤ Mh,
x
where m and M are the minimum and maximum of f (x) on the interval between x and x + h .
Division by h gives
(
1 x+h
m≤ f (t) dt ≤ M.
h x
Consider what happens as we let h → 0. The limit of the middle term is F ) (x) . Furthermore,
the numbers m and M must approach one another, and in the limit must both be equal to f (x) ;
they are minimum and maximum values of f (x) on the interval between x and x + h , and h
is approaching zero. We conclude therefore that
( x+h
)
F (x) = lim f (t) dt = f (x).
h→0 x
This argument can also be used to establish that F (x) has a right-hand derivative f (a) at
x = a , and a left-hand derivative f (b) at x = b .
According to Theorem 6.3, we evaluate the definite integral of a continuous function over
the interval a ≤ x ≤ b by calculating the difference between values of any antiderivative of
f (x) at x = b and at x = a . Theorem 6.7 establishes the fact that continuous functions
have antiderivatives. It does not, however, yield an antiderivative of f (x) in a form useful for
evaluation of a definite integral of f (x) . If we were to use the antiderivative of equation 6.18
in equation 6.11, we would obtain
( b ( b ( a ( b
f (x) dx = f (t) dt − f (t) dt = f (t) dt.
a a a a
This is certainly true, but not very helpful. To use equation 6.11 to evaluate the definite integral of
a function f (x) , we need an antiderivative of f (x) written in terms of functions that we already
know. Some integrands have easily computed antiderivatives such as those in Chapter 5; others
have antiderivatives that are somewhat more complicated and require the integration techniques
of Chapter 8 to express them in terms of well-known functions. There are some functions,
however, that do not have antiderivatives that can be expressed as the sum of a finite number of
2
well-known functions. A simple example is e−x . Theorem 6.7 guarantees that this function
has an antiderivative; it just cannot be expressed as a finite sum of well-known functions.
Symbolically, we may write the result of Theorem 6.7 in the form
( x
d
f (t) dt = f (x). (6.19)
dx a
We use this in the following four examples.
EXAMPLE 6.9
( x /
d
Evaluate 1 − t 2 dt .
dx 0
SOLUTION According to 6.19,

( x / /
d
1 − t 2 dt = 1 − x2.
dx 0
This is valid for −1 ≤ x ≤ 1.
EXAMPLE 6.10
( 2x 2
d sin t
Evaluate dt .
dx 1 1 + t2
SOLUTION For this problem we set u = 2x 2 , and invoke the chain rule,
( 2x 2 ( u # ( u $
d sin t d sin t d sin t du
2
dt = dt = dt
dx 1 1+t dx 1 1+ t2 du 1 1+ t2 dx
sin u 4x sin (2x 2 )
= 2
(4x) = .
1+u 1 + 4x 4
EXAMPLE 6.11
( 5
d
Evaluate (1 + t 3 )2/3 dt .
dx x
SOLUTION We can solve this problem by reversing the limits on the integral, which according
to equation 6.13a, introduces a negative sign. Thus,
( 5 ( x
d 3 2 /3 d
(1 + t ) dt = − (1 + t 3 )2/3 dt = −(1 + x 3 )2/3 .
dx x dx 5
In the following example, the variable x appears in both limits.
EXAMPLE 6.12
( 2x
d
Evaluate cos (2t 3 + 1) dt .
dx x2
SOLUTION Since the integrand is continuous for all real numbers, property 6.13b permits us
to write
( 2x ( a ( 2x
cos (2t 3 + 1) dt = cos (2t 3 + 1) dt + cos (2t 3 + 1) dt
x2 x2 a
for any real number a whatsoever. To find the derivative of the second integral on the right,
we set v = 2x and use the chain rule (as in Example 6.10), and for the derivative of the first
integral on the right, we reverse the limits (as in Example 6.11) and then use the chain rule with
u = x2 :
( 2x ( x2 ( 2x
d 3 d 3 d
cos (2t + 1) dt = − cos (2t + 1) dt + cos (2t 3 + 1) dt
dx x2 dx a dx a
# ( u $ # ( v $
d 3 du d 3 dv
= − cos (2t + 1) dt + cos (2t + 1) dt
du a dx dv a dx
= −2x cos (2u3 + 1) + 2 cos (2v 3 + 1)
= −2x cos (2x 6 + 1) + 2 cos (16x 3 + 1).
The Natural Logarithm Function

In Section 1.9 we reviewed properties of exponential and logarithm functions. First came the
exponential function a x , and the logarithm function loga x is its inverse function. The logarithm
of x to base a , loga x , is a power, the power to raise a in order to get x ; that is,
y = loga x if x = ay .
Based on this definition, it was straightforward to derive properties of the logarithm function
(equations 1.66) based on corresponding properties (equations 1.63) for the exponential function.
The natural logarithm function ln x uses base e , where e is the limit of (1 + 1/n)n as n
approaches infinity. We saw the advantage of ln x , as opposed to loga x , for differentiation in
Section 3.11; ln x avoids an extra constant.
The natural logarithm function can be introduced independently of the exponential function
using a definite integral. For x > 0 we define
( x
1
ln x = dt, (6.20)
1 t
and with this definition we can derive all properties of the logarithm function. We begin with its
graph. Since 1/t is positive for t > 0, it follows that ln x > 0 for x > 1, and property 6.13a
implies that it is negative for 0 < x < 1. According to equation 6.19, the derivative of ln x is
( x
d d 1 1
ln x = dt = . (6.21)
dx dx 1 t x
Because 1/x > 0 for x > 0, it follows that the derivative of ln x is positive, and therefore it is
an increasing function. The second derivative is negative,
d2 1
2
ln x = − 2 ,
dx x
so that the graph is concave downward. The graph must therefore look like that in Figure 6.9.
What is not clear at this point is that the graph is asymptotic to the negative y -axis, although
this might seem reasonable on the basis that 1/t becomes infinite as t → 0+ .
FIGURE 6.9 Graph of ln x
y
1
1 2 3 4 5 x
−1
−2
−3
−4
When u(x) is a differentiable function of x , equation 6.21 and the chain rule give
d 1 du
ln u = . (6.22)
dx u dx
By making specific choices for u(x) , we can derive properties 1.71 for the natural logarithm
function. First, for u = x1 x , where x1 > 0 is a fixed number, 6.22 gives
d 1 1
ln (x1 x) = x1 = . (6.23)
dx x1 x x
Equations 6.21 and 6.23 show that ln x and ln (x1 x) have the same derivative for all x > 0.
Theorem 5.1 implies that these functions can differ by at most a constant; that is,
ln (x1 x) = ln x + C.
If we set x = 1, and note that 6.20 implies that ln 1 = 0, we obtain ln x1 = 0 + C .

Consequently,
ln (x1 x) = ln x + ln x1 ,
and for x = x2 , we obtain identity 1.71a,
ln (x1 x2 ) = ln x1 + ln x2 . (6.24)
Differentiation formula 6.22 implies that for x > 0,

) * ) *
d 1 1 1 1
ln = − 2 =− .
dx x 1/x x x
When this is added to 6.21,
# ) *$
d 1 1 1
ln x + ln = − = 0.
dx x x x
Because this is valid for x > 0, Theorem 5.1 implies that
) *
1
ln x + ln = C.
x
Substitution of x = 1 gives C = 0, and therefore
) *
1
ln = − ln x. (6.25)
x
We can now verify property 1.71b,
) * # ) *$
x1 1
ln = ln x1
x2 x2
) *
1
= ln x1 + ln (by 6.24)
x2
= ln x1 − ln x2 (by 6.25).
For fixed x2 > 0, equation 6.22 gives
d 1 x2 d d
ln (x x2 ) = x x2 x x2 −1 = = x2 ln x = (x2 ln x).
dx x 2 x dx dx
Since this is valid for all x > 0, Theorem 5.1 once again gives
ln (x x2 ) = x2 ln x + C.
If we set x = 1, we obtain C = 0, and for x = x1 ,
ln (x1 x2 ) = x2 ln x1 . (6.26)
With the logarithm function defined by 6.20, there is a natural way to introduce the number
e . We define e as the number whose natural logarithm is 1:
( e
1
ln e = 1 *⇒ dt = 1. (6.27)
1 t
The resulting exponential function ex would satisfy properties 1.69 since all exponential func-
tions satisfy properties of this type. Furthermore, it follows from 6.26 and 6.27 that
ln (ex ) = x ln e = x. (6.28)
This is property 1.72b. We can also establish 1.72a by setting y = eln x . When we take
logarithms of both sides, and use 6.27 and 6.28,
ln y = ln (eln x ) = ln x(ln e) = ln x. (6.29)
The graph in Figure 6.9 indicates that for any given y -value on the curve, there is only one value
of x that gives that y . In other words, if ln y = ln x as in 6.29, it follows that y = x . Since
y = eln x , we have x = eln x .
EXERCISES 6.5
( sin x ( sin x
In Exercises 1–20 differentiate the definite integral with respect to x . 1
13. cos (t 2 ) dt 14. √ dt
( ( 0 cos x t +1
x x
2 1 ( √
2 x ( 2 x
√
1. (3t + t) dt 2. √ dt √ √
0 1 t2 + 1 15. t dt 16. √
t dt
0 x
( 2 ( −1
2 3 ( x2 (
3. sin (t ) dt 4. t cos t dt 2
x x 17. t 2 e4t dt 18. ln (t 2 + 1) dt
( ( 1 x
3x 2x √ ( 2x ( 3x
5. (2t − t 4 )2 dt 6. t + 1 dt 2
0 1 19. t ln t dt 20. e−4t dt
x −2 x
( 3x 2 ( 5x+4 / ∗ 21. Verify that when a(x) and b(x) are differentiable functions of x ,
7. sin (3t + 4) dt 8. t 3 + 1 dt ( b(x)
4 −2 d db da
( ( ) * f (t) dt = f [b(x)] − f [a(x)] .
2x √ 4x+4 dx a(x) dx dx
3 1
9. (3 t − 2t) dt 10. t −√ dt
x 4x t
Is equation 6.19 a special case of this result?
( x ( −2 x 2
11. tan (3t + 1) dt 12. sec (1 − t) dt 22.–28. Use the result of Exercise 21 to redo Exercises 8, 10, 12, 14,
−2 x −x 2 16, 18, and 20.
6.6 Average Values

The average value of two numbers c and d is defined as (c + d)/2. The average value of a
set of n numbers y1 , y2 , . . . , yn is (y1 + y2 + · · · + yn )/n . We would like to extend this idea
to define the average value of a function f (x) over an interval a ≤ x ≤ b . By beginning with
some simple functions we can see how to do this.
The function f (x) in Figure 6.10a is equal to 1 for the first third of the interval shown, 2
for the second third, and 3 for the last third. We would expect its average value over the interval
0 ≤ x ≤ 6 to be 2. The function in Figure 6.10b takes on the same function values, namely, 1,
2, and 3, but not on the same subintervals. The fact that it has value 2 for 2 ≤ x ≤ 5 and value
3 for 5 ≤ x ≤ 6 suggests that its average value should be somewhat less than the average value
of 2 for the function in Figure 6.10a. The average value of the function in Figure 6.10c should be
even less than that in Figure 6.10b. These three functions indicate that two factors are important
when considering average values of functions: values that the function takes on, and lengths of
the intervals on which they take these values. Perhaps what should be done to calculate average
values for these functions is to add together the products obtained by multiplying each of the
function values by the length of the interval in which it has this value, and then divide this sum
by the length of the overall interval.
For the function in Figure 6.10a, this yields an average value of
1
[1(2 − 0) + 2(4 − 2) + 3(6 − 4)] = 2;
6−0

Average values of three piecewise constant functions
y y y
3 3 3
2 2 2
1 1 1
x 2 4 6 x 2 4 6 x
2 4 6
for the function in Figure 6.10b,
1 11
[1(2 − 0) + 2(5 − 2) + 3(6 − 5)] = ;
6−0 6
and for the function in Figure 6.10c,
1 3
[1(4 − 0) + 2(5 − 4) + 3(6 − 5)] = .
6−0 2
This procedure is applicable only to a function whose domain can be subdivided into a finite
number of subintervals inside each of which the function has a constant value. Such a function
is said to be piecewise constant. What shall we do for functions that are not piecewise constant?
By rephrasing the procedure above, it will become obvious. The same three average values are
obtained if we adopt the following approach [illustrated for the function f (x) in Figure 6.10c].
Divide the interval 0 ≤ x ≤ 6 into three subintervals 0 < x ≤ 4, 4 < x ≤ 5, and 5 < x ≤ 6.
Pick a point in each subinterval; call the points x1∗ , x2∗ , and x3∗ . Evaluate f (x) at each point
and multiply by the length of the subinterval in which the point is found. Add these results, and
divide by the length of the interval to obtain the average value,
1 1 3
[f (x1∗ )(4 − 0) + f (x2∗ )(5 − 4) + f (x3∗ )(6 − 5)] = [1(4) + 2(1) + 3(1)] = .
6 6 2
But this is the procedure used to define the definite integral of a function; it lacks the limit
because the function is piecewise constant. In other words, for a function f (x) that has a value
at every point in the interval a ≤ x ≤ b (but is not necessarily piecewise constant), we define
its average value as
( b n
!
1 1
average value = f (x) dx = lim f (xi∗ ) #xi . (6.30)
b−a a b−a '#xi '→0
i=1
EXAMPLE 6.13
What is the average value of the function f (x) = x 2 on the interval 0 ≤ x ≤ 2?
SOLUTION By equation 6.30,
( 2 + ,2 ) *
1 2 1 x3 1 8 4
average value = x dx = = = .
2 0 2 3 0 2 3 3
EXAMPLE 6.14
Find the average value of f (x) = sin x on the intervals (a) 0 ≤ x ≤ π/2, (b) 0 ≤ x ≤ π ,
and (c) 0 ≤ x ≤ 2π .
FIGURE 6.11 Average SOLUTION We calculate average values on these intervals as:
value of sin x
y
( π/2
1 2 π/2 2 2
1 (a) sin x dx = {− cos x}0 = (0 + 1) =
y = sin x π/2 0 π π π
( π
1 1 1 2
(b) sin x dx = {− cos x}π0 = (1 + 1) =
x π 0 π π π
( 2π
1 1 1
−1 (c) sin x dx = {− cos x}20π = (−1 + 1) = 0
2π 0 2π 2π
The graph of sin x in Figure 6.11 also suggests that the average values in parts (a) and (b) should
be the same, and that the average value in part (c) should be zero.
EXAMPLE 6.15
In the Application Preview we questioned how a controller might monitor current in a circuit.
Application Preview One way would be for it to constantly measure the average value of the current over some fixed
Revisited interval of time. Suppose the current in the circuit is as shown in Figure 6.12, and the controller
constantly measures the average value of i(t) over the previous 0.1 s. If the average value ever
reaches 150% of steady-state current, the controller takes corrective action. At what time, if
any, does the controller react?
FIGURE 6.12 Abnormal SOLUTION If we denote the average value of i(t) over the 0.1 s before time t by i(t) , then
current in electric circuit
(
i i = 10 + 200(t − 1.9) 1 t
i(t) = i(t) dt.
30 0 .1 t−0.1
i = 10 − 200(t − 2.1)
20
For t ≤ 1.9, i(t) is always equal to 10 and the controller takes no action. For 1.9 < t ≤ 2,
10 i(t) is increasing and so also is i(t) :
1.9 2 2.1 t +( 1.9 ( t ,

i(t) = 10 10 dt + [10 + 200(t − 1.9)] dt
t−0.1 1.9
01 21.9 1 2t 3
= 10 10t t−0.1
+ 100t 2 − 370t 1.9
= 20(50t 2 − 190t + 181).
This reaches 150% of steady-state current in the interval 1.9 < t ≤ 2 if
20(50t 2 − 190t + 181) = 15.
Solutions of this quadratic equation are t = 1.83 and t = 1.97. The first is unacceptable, and
therefore the controller reacts at 1.97 s.
Theorem 6.7 can be used to establish the next theorem.

THEOREM 6.8 (Mean Value Theorem for Definite Integrals)

If f (x) is continuous for a ≤ x ≤ b , then there exists at least one number c between a
and b such that
( b
f (x) dx = (b − a)f (c). (6.31)
a
PROOF By Theorem 6.7, the function

( x
F (x) = f (t) dt
a
is continuous for a ≤ x ≤ b , and has derivative f (x) for a < x < b . Mean value theorem
3.19 applied to F (x) guarantees at least one number c between a and b such that
F (b) − F (a) = (b − a)F ) (c).
By substitution,
( b ( a
f (t) dt − f (t) dt = (b − a)f (c).
a a
Since the second integral vanishes, we have
( b
f (x) dx = (b − a)f (c).
a
By writing equation 6.31 in the form
( b
1
f (c) = f (x) dx,
b−a a
Theorem 6.8 states that the function must take on its average value at least once in the interval.
EXAMPLE 6.16
Find all values of c satisfying Theorem 6.8 for the function f (x) = x 2 on the interval 0 ≤
x ≤ 1.
SOLUTION Substituting a = 0 and b = 1 in equation 6.31 gives
( 1 + ,1
2 2 x3 1 1
(1 − 0)c = x dx = = −0 = .
0 3 0 3 3
√
Of the two solutions c√= ±1/ 3 for this equation, only the positive one is between 0 and
1. Thus, only c = 1/ 3 satisfies Theorem 6.8 for the function f (x) = x 2 on the interval
0 ≤ x ≤ 1.
EXAMPLE 6.17
Find all values of c satisfying Theorem 6.8 for the function f (x) = sin x on the interval
π/4 ≤ x ≤ 3π/4.
SOLUTION Substituting a = π/4 and b = 3π/4 in equation 6.31 gives
) * ( 3π/4
3π π 3π/4 1 1 √
− sin c = sin x dx = {− cos x}π/4 = √ + √ = 2.
4 4 π/4 2 2
Thus,
√
2 2
sin c = .
π
√
There are two√ angles between π/4 and 3π/4 with √ a sine equal to 2 2/π , namely
c = Sin −1 (2 2/π ) = 1.12 and c = π − Sin −1 (2 2/π ) = 2.02.
EXERCISES 6.6
In Exercises 1–20 find the average value of the function over the inter- ∗ 20. f (x) = ,x-, 0 ≤ x ≤ 3.5 (See Exercise 68 in Section 1.5.)
val.
∗ 21. If a particle moving along the x -axis is at position x1 at time t1
1. f (x) = x 2 − 2x, 0≤x≤2 and at position x2 at time t2 , its average velocity over this time interval
is
2. f (x) = x 3 − x, −1 ≤ x ≤ 1 x2 − x1
.
t2 − t1
3. f (x) = x 3 − x, 0≤x≤1
Verify that this is the same as the average of the velocity function over
4. f (x) = x 4 , 1≤x≤2 the time interval t1 ≤ t ≤ t2 .
√ ∗ 22. The velocity v of blood flowing through a circular vein or artery
5. f (x) = x + 1, 0≤x≤1
of radius R at a distance r from the centre of the blood vessel is
√
6. f (x) = x + 1, −1 ≤ x ≤ 1
v(r) = c(R 2 − r 2 ).
4
7. f (x) = x − 1, 0≤x≤1
(See Problem 3 in Section 6.2.) What is the average value of v(r) with
8. f (x) = x 4 − 1, 0≤x≤2 respect to r ?
9. f (x) = cos x, −π/2 ≤ x ≤ π/2

In Exercises 23–30 find all values of c satisfying equation 6.31 for the
10. f (x) = cos x, 0 ≤ x ≤ π/2 function f (x) on the specified interval.
11. f (x) = |x|, −2 ≤ x ≤ 2 23. f (x) = 2x − x 2 , 0≤x≤2
12. f (x) = |x|, 0≤x≤2

24. f (x) = x 3 − 8x, −2 ≤ x ≤ 2
2
∗ 13. f (x) = |x − 4|, 0≤x≤3
25. f (x) = cos x, 0 ≤ x ≤ π/2
2
∗ 14. f (x) = |x − 4|, −3 ≤ x ≤ 3
26. f (x) = cos x, 0≤x≤π
∗ 15. f (x) = sgn x, −1 ≤ x ≤ 1 (See Exercise 47 in Section 2.4.) √
27. f (x) = x + 1, 1≤x≤3
16. f (x) = sgn x, −1 ≤ x ≤ 3 (See Exercise 47 in Section 2.4.)
∗ 28. f (x) = x 2 (x + 1), 0≤x≤1
∗ 17. f (x) = h(x − 1), 0 ≤ x ≤ 2 (See Section 2.5.)
√
∗ 18. f (x) = h(x − 4), 0 ≤ x ≤ 2 (See Section 2.5.) ∗ 29. f (x) = x x 2 + 1, 0≤x≤2
∗ 19. f (x) = ,x-, 0 ≤ x ≤ 3 (See Exercise 68 in Section 1.5.) ∗ 30. f (x) = 1/x 2 + 1/x 3 , 1≤x≤2
In many applications involving erratic functions, such as the daily price ages f (x) and lags behind it.
of gold, or the values of stocks and bonds, it is advantageous to define
a moving average. It is the average of a function over an interval of ∗ 31. f (x) = sin x, L = π ∗ 32. f (x) = sin x, L = 2π
fixed length L , but of variable position. The length L moving average ∗ 33. f (x) = x , L = 1 2
∗ 34. f (x) = x 3 , L = 2
of a function f (x) at x is
∗ 35. f (x) = |x|, L = 1
+
( 1 − |x|, |x| ≤ 1
1 x ∗ 36. f (x) = , L=1
f (x) = f (t) dt. 0, |x| > 1
L x−L
∗ 37. f (x) = h(x − 1), L = 1, where h(x − 1) is the Heaviside
function
In Exercises 31–38 calculate the moving average of f (x) for the given ∗ 38. f (x) = h(x − a) − h(x − b), L = b − a , where b > a > 0
length L . Plot, or draw, f (x) and f (x) to illustrate how f (x) aver- and h(x − a) is the Heaviside function
6.7 Change of Variable in the Definite Integral

The first fundamental theorem of integral calculus indicates that to evaluate the definite integral
( −1
x
√ dx
−4 x+5
√
we should first find an antiderivative for x/ x + 5. To do this we set u = x + 5, in which
case du = dx , and
( ( (
x u−5
√ dx = √ du = (u1/2 − 5u−1/2 ) du
x+5 u
2 2
= u3/2 − 10u1/2 + C = (x + 5)3/2 − 10(x + 5)1/2 + C.
3 3
Consequently,
( −1 + ,−1
x 2 3/2 1/2
√ dx = (x + 5) − 10(x + 5)
−4 x+5 3 −4
# $ # $
2 3/2 1/2 2 3/2 1/2
= (4 ) − 10(4) − (1) − 10(1)
3 3
16
=− .
3
An alternative approach, which usually turns out to be less work, is to make the change of
variable u = x + 5 directly in the definite integral. In this case we again replace
x u−5
√ dx by √ du.
x+5 u
In addition, we replace the limits x = −4 and x = −1 by those values of u that correspond

to these values of x , namely, u = 1 and u = 4, respectively. We then obtain
( −1 ( 4 ( 4
x u−5
√ dx = √ du = (u1/2 − 5u−1/2 ) du
−4 x+5 1 u 1
+ ,4
2 3/2 1/2
= u − 10u
3 1
6.7 Change of Variable in the Definite Integral 405
# $ # $
2 2 3/2
= (4)3/2 − 10(4)1/2 − (1) − 10(1)1/2
3 3
16
=− .
3
That this method is generally acceptable is stated in the following theorem.
THEOREM 6.9
Suppose f (x) is continuous on a ≤ x ≤ b , and we set x = g(u) , where a = g(α)
and b = g(β) . Then
( b ( β
f (x) dx = f (g(u)) g ) (u) du, (6.32)
a α
if g ) (u) is continuous on α ≤ u ≤ β , and if when u is between α and β , g(u) is

between a and b .
EXAMPLE 6.18
( 4
√
x
Evaluate √ 4
dx .
2 ( x − 1)
√ √
SOLUTION If we set u = x − 1, then du = [1/(2 x)] dx , and
( 4
√ ( 1 ( 1 ) *
x u+1 1 2 1
√ dx = √ 2 (u + 1) du = 2 √ + + du
2 ( x − 1)4 2 −1 u4 2 −1 u2 u3 u4
+ ,1
1 1 1
=2 − − 2 − 3 √
u u 3u 2 −1
# $ # $
1 1 1 1
= −2 1 + 1 + +2 √ + √ + √ = 21.2.
3 2−1 ( 2 − 1)2 3( 2 − 1)3
What follows is a discussion leading to the well-known formula for the area of a circle. Although
you should be able to follow the calculations, it is highly likely that you will not understand the
source of some of the ideas. Do not be alarmed; in time, we will deal with all aspects of the
derivation in detail. For now, simply take the example as an illustration of the power of integral
calculus, and an indication of things to come.
FIGURE 6.13 Area of a
circle
When we study applications of definite integrals in Chapter 7, we shall see that the area of
the circle in Figure 6.13 is given by the definite integral
y
r x2 + y2 = r 2 ( r /
A=4 r 2 − x 2 dx.
0
The integral itself actually gives the first quadrant area; the 4 provides a quadrupling
√ factor
r x for the other three quadrants. It is not possible to guess an antiderivative for r 2 − x 2 . In
Section 8.4 we learn that an appropriate change of variable is to set x = r sin θ . The integrand
becomes
/ / / √
r 2 − x 2 = r 2 − r 2 sin2 θ = r 1 − sin2 θ = r cos2 θ = r| cos θ |,
√
absolute values being necessary to ensure positivity for cos2 θ . Since dx = r cos θ dθ , we
have ( θ2
A=4 r| cos θ | r cos θ dθ,
θ1
where θ1 and θ2 are values of θ corresponding to x = 0 and x = r . When x = 0,
the equation x = r sin θ requires that 0 = r sin θ . We choose θ1 = 0 as the solu-
tion of this equation. (There are others, and in Section 8.4 we shall see why this choice
is made.) Similarly, when x = r , we choose the solution θ2 = π/2 of r = r sin θ .
Then ( π/2
A=4 r| cos θ | r cos θ dθ.
0
For θ in the interval 0 ≤ θ ≤ π/2, cos θ is positive, and absolute values may be dropped,
( π/2
2
A = 4r cos2 θ dθ.
0
To find an antiderivative for cos2 θ , we solve double-angle formula 1.46b for cos2 θ ,
1 + cos 2θ
cos 2θ = 2 cos2 θ − 1 *⇒ cos2 θ = .
2
Thus,
( π/2 ) * ( π/2
1 + cos 2θ
A = 4r 2 dθ = 2r 2 (1 + cos 2θ ) dθ
0 2 0
+ ,π/2
1
= 2r 2 θ + sin 2θ = π r 2,
2 0
the formula for the area of a circle. Be reminded that you are not expected to solve other prob-
lems like this yet. In time, yes, but not now. What we hope is that the example gives you a
glimpse of the power of integral calculus and the importance of the tools that we have developed
in Chapters 5 and 6.
EXERCISES 6.7
4
( 95 (
In Exercises 1–22 evaluate the definite integral. √ 1
x2
( ( ∗ 15. 1 + x dx ∗ 16. dx
2 1 √ 4 − 1/ 2 1+x
1. x(3x 2 − 2)4 dx 2. z 1 − z dz ( 1 ( 1' '
1 0 |x| ' x '
∗ 17. dx ∗ 18. ' '
( 0 ( π/3 (x + 2)3 ' (x + 2)3 ' dx
x −1 −1
3. √ dx 4. cos5 x sin x dx ( 1 ( 2
−1 x +3 π/4 3 (ln x)2
( 3 / ( ∗ 19. x 2 ex dx ∗ 20. dx
6 x
x 0 1
5. x 3 9 − x 2 dx 6. √ dx ( 4 ( π/4
1 −5 x 2 − 12 1 sec2 x
4 ∗ 21. dx ∗ 22. √ dx
( 5 / ( 1 2 x ln x −π/4 4 + 3 tan x
2 x2
7. y y − 4 dy 8. dx
4 1/2 1+x
/ √ ∗ 23. Show that if f (x) is an odd function, then
( 4 ( 1
1+ u x+1 ( a
9. √ du 10. dx
1 u 2
−2 (x + 2x + 2)
1/ 3 f (x) dx = 0;
( ( π/6 −a
4
x2 √
11. dx 12. 2 + 3 sin x cos x dx and that if f (x) is an even function,
3 (x − 2)4 0 ( (
a a
( π/2 ( 4
sin3 x (x + 1)(x − 1) f (x) dx = 2 f (x) dx.
13. dx 14. √ dx −a 0
π/4 (1 + cos x)4 1 x
Summary 407
( −1
√
∗ 24. Show algebraically that if f (x) is a continuous function with pe- x 2 − 6x 1
∗ 26. dx, u=
riod p , then −6 x4 x
( a+p ( p ( 0
x 1−x
f (x) dx = f (x) dx. ∗ 27. dx, u2 =
a 0 −4 (5 − 4x − x 2 )3/2 5+x
∗∗ 28. Show that

In Exercises 25–27 use the suggested substitution to evaluate the defi- ( π ( 2π
2
nite integral. cos [(π/2) cos θ ] π 1 − cos φ
dθ = dφ.
0 sin θ 2 0 φ(2π − φ)
( 3
1 1
∗ 25. √ dx, u= √ This integral is used in calculating radiated power from a half-wave
1 x 3/ 2 4−x x antenna.
SUMMARY
The definite integral of a continuous function f (x) from x = a to x = b is a number, one
that depends on the function f (x) and the limits a and b . We have defined the definite integral
by subdividing the interval a ≤ x ≤ b into n parts by n + 1 points a = x0 < x1 < · · · <
xn−1 < xn = b , and choosing a point xi∗ in each subinterval xi−1 ≤ x ≤ xi . The definite
integral is then the limit of the summation
( b n
!
f (x) dx = lim f (xi∗ ) #xi ,
a '#xi '→0
i=1
where #xi = xi − xi−1 . Since all calculations in this limit take place on the x -axis, we regard
the definite integral as an integration along the x -axis from x = a to x = b .
The first fundamental theorem of integral calculus allows us to calculate definite integrals
of continuous functions using antiderivatives,
( b +( ,b
f (x) dx = f (x) dx .
a a
This presupposes that a continuous function has an antiderivative, as verified in Section 6.5. It
was shown that when the definite integral is given a variable upper limit x , the resulting function
( x
F (x) = f (t) dt
a
is an antiderivative of f (x) ; that is, F ) (x) = f (x) .

The average value of a function f (x) over an interval a ≤ x ≤ b is defined as
( b
1
f (x) dx.
b−a a
The mean value theorem for definite integrals guarantees the existence of at least one number
c between a and b , at which f (x) takes on its average value,
( b
1
f (c) = f (x) dx.
b−a a
Evaluation of a definite integral with a complex integrand can sometimes be simplified with
an appropriate change of variable.
KEY TERMS
Sigma notation General term

Index of summation or variable of summation Limits of summation
Finite geometric series Norm
Definite integral Lower limit of integration
Upper limit of integration Riemann sum
Riemann integral First fundamental theorem of integral calculus
Second fundamental theorem Average value
of integral calculus
Piecewise constant Mean value theorem for definite integrals
Change of variable
REVIEW
EXERCISES
In Exercises 1–20 evaluate the definite integral. ∗ 22. Prove that each of the following answers is incorrect, but do so
without evaluating the definite integral. Think about what the definite
( 3 ( 1
2 2 4 integral represents.
1. (x + 3x − 2) dx 2. (x − x ) dx
0 −1 ( 4 ( −2
1
( 1 ( 2
(a) (x 2 + 3x) dx = −3 (b) dx = 5
3 2 0 −3 x
3. (x − 3x) dx 4. (x − 2x) dx
−1 0
( 2 ( −2 In Exercises 23–26 find the average value of the function on the inter-
2 1 val.
5. (x + 1) dx 6. dx
1 −3 x2 √
23. f (x) = x + 4, 0≤x≤1
( 9) * (
√1 π
7. √ − x dx 8. cos x dx 24. f (x) = 1/x 2 − x, −2 ≤ x ≤ −1
4 x 0
( ( √
1 2 ∗ 25. f (x) = x x + 1, 0≤x≤1
2 2 2
9. x(x + 1) dx 10. x (x + 3) dx
−1 1 ∗ 26. f (x) = cos3 x sin2 x, 0 ≤ x ≤ π/2
( 3 ( 5 /
√
11. x + 1 dx 12. x x 2 − 1 dx
0 1 In Exercises 27–32 differentiate the integral with respect to x .
( 4 )√ * ( 0
( x / ( −3
x+1 √
13. √ dx 14. x x + 1 dx 27. t t 3 + 1 dt 28. t 2 (t + 1)3 dt
1 x −1 1 x
( 2 ( −2 ( x2 / (
x2 + 1 √ 4
15. dx 16. x 2 2 − x dx 29. t 2 + 1 dt 30. t cos t dt
1 (x + 1)4 −4 1 2x
( π/4 ( 3 ( ( x2
cos x 1−x
1
17. dx 18. x(1 + 2x 2 )4 dx 31. dt 32. sin2 t dt
0 (1 + sin x)2 2 2x+3 t2 + 1 −x 2
( 8 ( 4
(1 + x 1/3 )2
∗ 19. dx ∗ 20. |x + 2| dx In Exercises 33–34 v(t) represents the velocity of a particle moving
1 x 2 /3 −4
along the x -axis. Calculate the definite integrals of v(t) and |v(t)|
between the two times shown. Interpret each number physically and
draw a graph of v(t) to corroborate results.
∗ 21. Use equation 6.10 to evaluate the following definite integrals:
( 2 ( 3 ∗ 33. v(t) = t 3 − 6t 2 + 11t − 6, t = 0 to t = 1
(a) (x − 5) dx (b) (x 2 + 3) dx
0 0 ∗ 34. v(t) = t 3 − 6t 2 + 11t − 6, t = 1 to t = 3
( 2 ( 5
In Exercises 35–42 evaluate the definite integral. 6x 2 + 8x + 2
∗ 39. x 2 (4 − x 3 )5 dx ∗ 40. √ dx
( 1 ( π/6
−1 1 x 3 + 2x 2 + x
x3 cos3 x
∗ 35. dx ∗ 36. √ dx
0 (x 2 + 1)3/2 0 1 + sin x
( 1 / ( 2
' ' ( 2 ( 1
' x ' x − 25 1
∗ 37. x 3 1 − x 2 dx ∗ 38. '√ ' ∗ 41. √ dx ∗ 42. √ √ dx
' 3 + x ' dx 1 x−5 0 2+x+ x
−1 −1
CHAPTER
7 Applications of the Definite
Integral
Application Preview The figure on the left below shows gas in the cylinder of a steam engine, diesel engine, or internal
combustion engine. It requires work to move the piston to the left and compress the gas in the
cylinder. On the other hand, if the gas expands, it does work in moving the piston to the right.
P
Cylinder B C
Piston
Gas
D
A
V
The figure on the right above is the Rankine cycle for an idealized steam engine; it represents
the relationship between pressure P and volume V of gas in the cylinder during one cycle. Water
at low temperature and pressure (point A ) is heated at constant volume (along path AB ). Along
BC the water is converted to steam and expands slightly, and the expansion continues along
CD . To complete the cycle, the steam is cooled and condensed to water along DA .
THE PROBLEM Determine the output of the steam engine during one cycle. (See page
440 for the solution.)
In Chapter 6 we defined the definite integral of a function f (x) as the limit of a summation
! b n
"
f (x) dx = lim f (xi∗ ) !xi .
a !!xi !→0
i=1
The limit summation on the right defines the value of the definite integral on the left. In
this chapter we think of this equation in the reverse direction in order to evaluate geometric and
physical quantities. Each quantity is expressed as a limit summation of the form in this equation.
The limit summation can immediately be interpreted as a definite integral. The definite integral
can then be evaluated by means of an antiderivative of the integrand.
7.1 Area
We have formulas for the area of very few shapes — squares, rectangles, triangles, and polygons
of any shape, since they can be divided into rectangles and triangles. Consider finding the area
in Figure 7.1a. Each of us has intuitive ideas about this area. In this section we take these
intuitive ideas of what the area ought to be, and make a precise mathematical definition of what
the area is.
410
7.1 Area 411
FIGURE 7.1a Area under curve y = f (x) FIGURE 7.1b Approximation of area under curve y = f (x)
y y
y = f (x)
y = f (x)
A1 A2
A Ai An
a b x x1 x2 xi−1 xi xn−1 x
a = x0 b = xn
Recall from Problem 1 of Section 6.2 that the area in Figure 7.1a can be approximated by
rectangles. Specifically, we partition the interval a ≤ x ≤ b into n parts by points
a = x0 < x1 < x2 < · · · < xn−1 < xn = b,

and construct rectangles as shown in Figure 7.1b. The area Ai of the i th rectangle is
Ai = f (xi )(xi − xi−1 ) = f (xi ) !xi ,

and an approximation to the required area A is therefore
n
" n
"
Ai = f (xi ) !xi .
i=1 i=1
If we let the number of rectangles get larger and larger, and at the same time require each
rectangle to have smaller and smaller width that eventually approaches zero, we feel that better
and better approximations can be obtained. In fact, if we take the limit as the norm !!xi !
of the partition approaches zero, we should get A . Since we have no formal definition for the
area of odd-shaped figures, we take this opportunity to make our own. We define the area A in
Figure 7.1a as
n
"
A= lim f (xi ) !xi . (7.1)
!!xi !→0
i=1
Area A then has been defined as the limit of a sum of rectangular areas.
The right side of equation 7.1 is strikingly similar to the definition of the definite integral
of f (x) from x = a to x = b (equation 6.10):
! b n
"
f (x) dx = lim f (xi∗ ) !xi .
a !!xi !→0
i=1
∗
The only difference is the absence of ’s in equation 7.1. But when we recall that in equation
6.10, xi∗ may be chosen as any point in the subinterval xi−1 ≤ x ≤ xi , we see that by choosing
xi∗ = xi ,
! b n
"
f (x) dx = lim f (xi )!xi .
a !!xi !→0
i=1
It follows that area A of Figure 7.1a may be calculated by means of the definite integral
! b
A= f (x) dx. (7.2)
a
412 Chapter 7 Applications of the Definite Integral
It is important to realize that equation 7.2 does not imply that a definite integral should always
be thought of as an area; on the contrary, we will find that definite integrals can represent many
quantities. What we have said is that the area in Figure 7.1a is defined by the limit in equation
7.1. But this limit may also be interpreted as the definite integral of f (x) with respect to x
from x = a to x = b ; hence, the area may be calculated by the definite integral in equation
7.2. Since definite integrals can be evaluated using antiderivatives, it seems that we have a very
simple way to find areas.
It is simple to extend this result to the problem of finding the area in Figure 7.2. With our
FIGURE 7.2 Area between
two curves and two vertical lines
interpretation of equation 7.2, we can state that the area under the curve y = f (x) , above the
x -axis, and between the vertical lines x = a and x = b is given by
y
y = f (x) ! b
A2 = f (x) dx.
a
Since the area under the curve y = g(x) is given similarly by

! b
y = g (x) A1 = g(x) dx,
a b x a
it follows that the required area is

! b ! b ! b
A = A2 − A 1 = f (x) dx − g(x) dx = [f (x) − g(x)] dx. (7.3)
a a a
We now have two formulas for finding areas: equation 7.2 for the area under a curve, above
the x -axis, and between two vertical lines; and equation 7.3 for the area between two curves
and two vertical lines. Note that 7.2 is a special case of 7.3. At this point we could solve
a number of area problems using these two results, but to do so would strongly suggest that
FIGURE 7.3 Integral for integration should be approached from a “formula” point of view; and if there is any point of
area under a curve view that we wish to adopt, it is completely the opposite. By the end of this chapter we hope
y to have developed a sufficiently clear understanding of the limit-summation process that use of
integration in situations other than those discussed here will be straightforward.
y = f (x) To illustrate how we arrive at the correct definite integral for area problems without memo-
rizing the formulas in either equation 7.2 or 7.3, consider again finding the area in Figure 7.1a.
We draw a rectangle of width dx at position x as shown in Figure 7.3. The area of this rectangle
f (x) is
f (x) dx.
a b x
x
dx We visualize this rectangle as a representative for a large number of such rectangles between a
and b . To find the required area we add together all such rectangular areas and take the limit as
their widths approach zero. But this is the concept of the definite integral, so that we write for
the limit-summation process
! b
FIGURE 7.4 Integral for
area between two curves and two
A= f (x) dx.
a
vertical lines
Limits a and b identify x -positions of first and last rectangles, respectively.
y
y = f (x) Similarly, for the area in Figure 7.4 we draw a rectangle of width dx and length f (x)−g(x)
and, therefore, of area
f (x) [f (x) − g(x)] dx.
y = g (x) To add areas of all such rectangles between a and b and, at the same time, to take the limit as
their widths approach zero, we once again use the definite integral,
g (x) ! b
a b x
x A= [f (x) − g(x)] dx.
dx a
7.1 Area 413
For area problems, then, we start with the area of a representative rectangle and proceed
to the required area by summation with the definite integral. We express this symbolically as
follows:
area of vertical rectangle
# $% &
width of
vertical
x -position of length of vertical rectangle rectangle
last rectangle #
' $% (& #$%&
! y -coordinate y -coordinate
A= of upper end − of lower end dx . (7.4)
of rectangle of rectangle
x -position of
first rectangle
EXAMPLE 7.1
Find the area enclosed by the curves y = x 2 and y = x 3 .
FIGURE 7.5a Area FIGURE 7.5b Estimate of area bounded by y = x 2

2 3
bounded by y = x and y = x and y = x 3
y y
(1, 1)
1 (1,1)
y= x2
y = x3
x x
dx x
SOLUTION The area of the representative rectangle in Figure 7.5a is
(x 2 − x 3 ) dx;
hence,
! 1 ) *1
x3 x4 1
A= (x 2 − x 3 ) dx = − = .
0 3 4 0 12
It is easy to check whether 1/12 is a reasonable answer. This area is contained inside the square
of area one in Figure 7.5b. It is reasonable that the area bounded by the curves is one-twelfth
that of the square. We could have used this argument to ballpark the answer before making any
calculations.
In the next two examples, expressions for areas of representative rectangles vary within the region
specified. In such cases, we set up different integrals corresponding to different representative
rectangles.
EXAMPLE 7.2
Find the area bounded by the x -axis and the curve y = x 3 − x .
SOLUTION Areas of rectangles between x = −1 and x = 0 (Figure 7.6) are
[(x 3 − x) − 0] dx,
FIGURE 7.6 Area bounded by the curves

y = x 3 − x and y = 0
y
y = x3 − x
x
−1 dx
dx x 1 x
whereas between x = 0 and x = 1 areas are
[0 − (x 3 − x)] dx.
Consequently,
! 0 ! 1
A= (x 3 − x) dx + (−x 3 + x) dx
−1 0
) 4 2
*0 ) *1
x x x4 x2
= − + − +
4 2 −1 4 2 0
+ , + ,
1 1 1 1
=− − + − +
4 2 4 2
1
= .
2
We could have saved ourselves some calculations in this example by noting that because of the
symmetry of the diagram ( x 3 − x is an odd function), the two areas are identical. Hence, we
could find the left (or right) area and double it,
! 0 ) *0 + ,
x4 x2 1 1 1
A=2 (x 3 − x) dx = 2 − = −2 − = .
−1 4 2 −1 4 2 2
EXAMPLE 7.3
Find the area of the triangle with edges y = x , y = −x/2, and y = 5x − 44.
SOLUTION Areas of representative rectangles to the left and right of x = 8 (Figure 7.7a)
are, respectively,
[x − (−x/2)] dx and [x − (5x − 44)] dx;
therefore,
! 8 ! 11
A= [x − (−x/2)] dx + [x − (5x − 44)] dx
0 8
! 8 ! 11
3
= x dx + (44 − 4x) dx
2 0 8
7.1 Area 415
) *8
3 x2 - .11
= + 44x − 2x 2 8
2 2 0
= 48 + (484 − 242) − (352 − 128)

= 66.
Figure 7.7b provides a quick check. The required area would seem to be somewhere between a
third and half that of the 15 × 11 rectangle with area 165 square units.
FIGURE 7.7a Area FIGURE 7.7b Estimate of

bounded by y = x , y = −x/2, area of triangle
and y = 5x − 44
y
(11, 11)
y
(11, 11)
dx
y=x
x
(8, −4)
dx
y = 5x − 44 x
x
y=−
2 (8, −4)
We see from the examples above that the length of a representative rectangle in equation 7.4 as
“upper y minus lower y ” is valid whether the rectangle is in the first quadrant (Figure 7.5a),
the second and fourth quadrants (Figure 7.6), or partially in the first and partially in the fourth
(Figure 7.7a). In fact, it is valid for rectangles in all quadrants. Remember this; we use it in
many applications.
FIGURE 7.8 Using horizontal rectangles to find area
x = g ( y) x = f ( y)
dy
y
a
x
There is nothing special about vertical rectangles. Sometimes it is more convenient to

subdivide an area into horizontal rectangles. For example, to find the area in Figure 7.8, we
draw a representative rectangle at position y of width dy . Its length is f (y) − g(y) , and
therefore its area is
[f (y) − g(y)] dy.
Adding over all rectangles gives

! b
A= [f (y) − g(y)] dy. (7.5)
a
Corresponding to equation 7.4, we could write that for horizontal rectangles,
area of horizontal rectangle

# $% &
width of
horizontal
y -position of length of horizontal rectangle rectangle
# $% (& #$%&
! last rectangle ' x -coordinate x -coordinate
A= of right end − of left end dy. (7.6)
of rectangle of rectangle
y -position of
first rectangle
EXAMPLE 7.4
√ √
Find the area bounded by the curves y = x + 14, x = y , and y = 0.
FIGURE 7.9 Area bounded SOLUTION Subdivision of the region (Figure 7.9) into vertical rectangles results in two
√ √
by y = x + 14, x = y , and integrations: one to the left and the other to the right of the y -axis. On the other hand, throughout
y=0
the required region, the area of a horizontal rectangle is
y
√
[ y − (y 2 − 14)] dy,
y = x + 14 and therefore
x2 (2, 4)
x1 ! ) *4
x= y 4 /√ 0 2 1 16 64
dy y A= y − y 2 + 14 dy = y 3/2 3
− y + 14y = − + 56 = 40.
0 3 3 0 3 3
−14 x
In choosing between horizontal and vertical rectangles, consider two objectives:
1. Minimize the number of integrations.

2. Obtain simple definite integrals.
For instance, by choosing one type of rectangle we may obtain only one definite integral, but
it may be very difficult to evaluate. If the other type of rectangle leads to two simple definite
integrals, then it would be wise to choose the two simple integrals.
EXAMPLE 7.5
Find the area enclosed by the curves
x x2
y = √ , y = , y = −x 2 , x = 6, x ≥ 0.
2
x − 16 15
SOLUTION Examination of Figure 7.10 indicates that horizontal rectangles necessitate

√ three
integrals. In addition, we would have to solve equations y = x 2 /15 and y = x/ x 2 − 16
7.1 Area 417
for x in terms of y . We therefore opt for vertical rectangles.
! 5 1 2 ! 6 1 2
x2 2 x 2
A= − (−x ) dx + √ − (−x ) dx
0 15 5 x 2 − 16
! 5 ! 6 1 2
16 x
= √ x 2 dx + + x 2 dx
15
0 5
2
x − 16
) 3 *5 )3 *6
16 x 2
x3
= + x − 16 +
15 3 0 3 5
+ , 4√ 5 + ,
16 125 125
= + 20 + 72 − 3+
15 3 3
= 76.2.
FIGURE 7.10 Area bounded by four curves
x
y=
y x2 − 16
x2 5
y= (5, )
3
15
5 6 x
dx
y = −x 2
dx
x=6
The mean value theorem for definite integrals (Theorem 6.8) states that there is a number c
between a and b such that
! b
f (x) dx = (b − a)f (c).
a
It has a very simple interpretation in terms of area when f (x) ≥ 0 for a ≤ x ≤ b . Since
f (x) ≥ 0 for a ≤ x ≤ b , the definite integral may be interpreted as the area under the curve
y = f (x) , above the x -axis, and between vertical lines at x = a and x = b (Figure 7.11). The
right side is the area of the rectangle of width b − a and height f (c) shaded in Figure 7.11. The
mean value theorem guarantees at least one point c between a and b for which the rectangular
area is equal to the area under the curve. For the curve in Figure 7.11, there is exactly one such
point c ; for the curve in Figure 7.12, there are three choices for c .
FIGURE 7.11 Area interpretation of FIGURE 7.12 Three points satisfy the
mean value theorem for definite integrals mean value theorem for this function
y y
f (c)
f (c)
a c b x a c c c b x
EXERCISES 7.1
In Exercises 1–16 find the area of the region bounded by the curves. (d) Repeat part (b) with eight rectangles all of equal width,
denoting the sum of the areas of the eight rectangles by A8 .
1. y 2 = 4x, x 2 = 4y 2. y = x 3 + 8, y = 4x + 8 Show on the graph the extra precision of A8 over A4 .
3. yx 2 = 4, y = 5 − x 2 4. x = y(y − 2), x + y = 12 This discussion is continued in Exercise 48.
∗ 18. Repeat Exercise 17 with the area bounded by the curves y =
5. x = 4y − 4y 2 , y = x − 3, y = 1, y = 0
x 3 + 1, x = 1, x = 3, and y = 0.
6. y = e3x , x = 1, x = 2, y = −x
7. 12y = 7 − x 2 , y = 1/(2x) In Exercises 19–22 set up (but do not evaluate) definite integral(s) for
√ the area of the region bounded by the curves.
8. y = x + 4, y = (x + 4)2 /8 3
19. x = 1/ 4 − y 2 , 4x = −y 2 , y = −1, y = 1
9. y = (x − 1)5 , x = 0, y = 0
10. y = sin x (0 ≤ x ≤ π ), y = 0 20. x 2 + y 2 = 4, x 2 + y 2 = 4x (interior to both)
11. y = x 5 − x, y = 0 21. x 2 + y 2 = 4, x 2 + y 2 = 6x (interior to both)
12. x + y = 1, x + y = 5, y = 2x + 1, y = 2x + 6 22. x 2 + y 2 = 16, x = y 2 (smaller area)
13. x = sec2 y, y = 0, y = π/4, x = 0

∗ 23. Find the area of the region bounded by y 2 = 4ax and x 2 = 4ay ,
14. xy = e, y = x 2 , y = 2 (smaller area) where a > 0 is a constant.
15. y = e2 (2 − x 2 ), y = e2x , y = e−2x , y ≥ 1

In Exercises 24–35 find the area of the region bounded by the curves.
16. x = |y| + 1, x + (y − 1)2 = 4, y = 0 (above the x -axis)
√
∗ 24. y = x/ x + 3, x = 1, x = 6, y = −x 2
∗ 17. (a) Find the area A of the region bounded by the curves y =
∗ 25. x = y 2 + 2, x = −(y − 4)2 , y = −x + 4, y = 0
16 − x 2 , y = 0, x = 1, and x = 3. In the remainder of
this problem we approximate A by rectangles and show that √
∗ 26. y = x 3 − x, x + y + 1 = 0, x = y + 1
the accuracy of the approximation increases as the number 6 6
of rectangles increases. To do this you will need a large 6 x 6
6
∗ 27. y = 6 6 , y = 0, x = −1, x = 1
graph of the function f (x) = 16 − x 2 on the interval (x − 2)3 6
0 ≤ x ≤ 3.
2
(b) On the graph draw two rectangles of equal width (1 unit) and ∗ 28. x = 2ye−y , y = x
with heights determined in the same way as in Figure 7.1b.
If A2 denotes the sum of the areas of these two rectangles, ∗ 29. y = sin3 x, y = 1/8, 0 ≤ x ≤ 2π
what is A2 ? What is the error in the approximation of A
by A2 ? Illustrate this error on the graph. ∗ 30. y = ln x 2 , y = 1 − x 2 , y = 1
(c) Repeat part (b) with four rectangles all of width one-half, ∗ 31. |x|1/2 + |y|1/2 = 1
denoting the sum of the areas of the four rectangles by A4 .
Show on the graph the extra precision of A4 over A2 . ∗ 32. y 2 = x 2 (4 − x 2 )
∗ 33. y 2 = x 4 (9 + x) ∗ 49. Show that the curves

∗ 34. y 2 = x 2 (x 2 − 4), x = 5 x3
y = and 204y = 13x 2 − 1
2
∗ 35. (2x − y) = x , x = 4 3 x4 + 16
bound three regions. Find the area of the largest region.
∗ 36. The tangent line at a point on the first quadrant part of the parabola
∗ 50. Let P be a point on the cubic curve y = f (x) = ax 3 . Let the
y = 2 − x 2 makes a triangle with the positive x - and y -axes. Find
tangent line at P intersect y = f (x) again at Q , and let A be the area
the point for which the area of the triangle is smallest.
of the region bounded by y = f (x) and the line P Q . Let B be the
∗ 37. Repeat Exercise 36 for the curve y = 2 − x 4 . area of the region defined in the same way by starting with Q instead
of P . Show that B is 16 times as large as A .
∗ 51. The parabola y = ax 2 and the circle x 2 +(y −r)2 = r 2 intersect
In Exercises 38–45 it is necessary to use a calculator or computer to
in two points as shown in the figure below. Find the value of a that
find points of intersection of the curves. Find the area of the region
makes the area inside the parabola and below the horizontal line through
bounded by the curves (to three decimal places).
the points of intersection as large as possible.
∗ 38. y = x 3 + 3x 2 + 2x + 1, x = 0, y = 0 y
3 2
∗ 39. y = x − 4x, y = 2 − x − x y = ax2
4 2
∗ 40. y = x − 5x + 5, y = 0
∗ 41. y = ex , y = 2 − x 2 x2 + (y−r)2 = r2
∗ 42. y = cos x, 4y = x + 2
2
∗ 43. y = , y = x 3 + 3x − 1, x = 0
x+2 x
3 2
∗ 44. y = x − 3x + 4x − 2, x = 4 − y 2 ∗∗ 52. Prove that the result in Exercise 50 is valid for any cubic y =
√ f (x) = ax 3 + bx 2 + cx + d .
∗ 45. x = y 3 − y 2 − 2y, x = 2y + 1
∗∗ 53. Show that the area of the region in the figure below is
! xQ
∗ 46. For what values of m do the curves 1
[f (x) − mx − b][1 + mf * (x)] dx,
x 1 + m2 xP
y = and y = mx
3x 2 + 1 where xP and xQ are x -coordinates of P and Q .
bound a region with finite area? Find the area.
y y = f(x)
√ Q
∗ 47. Find a point (a, b) on the curve y = x/ x 2 + 1 such that the
region bounded by this curve, the x -axis, and the line x = a has area P y = mx + b
equal to twice that of the region bounded by the curve, the y -axis, and
the line y = b .
x
∗ 48. If 2n ( n = 1, 2, . . . ) rectangles (all of equal width) are drawn in
Exercise 17 to approximate A , and A2n denotes the sum of the areas ∗∗ 54. Suppose that the horizontal line y = h intersects the parabola
of these rectangles, show that y = ax 2 +bx +c ( a > 0) in two points P and Q . Show that the area
1 + ,2 of the region so bounded is two-thirds of the length of P Q multiplied
1 1 3 1 by the distance from the vertex of the parabola to the horizontal line.
A2n = A − n−2 + + ,
2 6 2n−3 22n−3 ∗∗ 55. A circular pasture has radius R . A cow is tied to a stake at the
edge of the pasture. What length of rope permits the cow to graze on
and hence that lim A2n = A .
n→∞ half the pasture?
7.2 Volumes of Solids of Revolution

In Section 6.2 we discussed the idea of rotating flat surfaces around coplanar lines to produce
volumes of solids of revolution. To find the volume generated when the region in Figure 7.1a is
revolved about the x -axis, we again approximate the region by n rectangles as in Figure 7.1b. If
each of these rectangles is rotated around the x -axis, then n discs are formed. Since the radius
FIGURE 7.13 Approximating volume of solid of revolution by discs
y = f (x)
f (xi)
a
xi − 1 xi b
Vi xn − 1 x
of the i th disc is f (xi ) (Figure 7.13), its volume is given by
Vi = π [f (xi )]2 (xi − xi−1 ) = π [f (xi )]2 !xi .
An approximation to the required volume V is therefore
n
" n
"
Vi = π [f (xi )]2 !xi .
i=1 i=1
If we let the number of rectangles get larger and larger, and at the same time require the widths to
get smaller and smaller, it seems reasonable that we will obtain better and better approximations.
Furthermore, if we take the limit as the norm of the partition approaches zero, we should get
what we think is V . Since we have no formal definition for such volumes, we make our own.
We define
n
"
V = lim π [f (xi )]2 !xi . (7.7)
!!xi !→0
i=1
But we can interpret the right side of this definition as the definite integral of π [f (x)]2 with
respect to x from x = a to x = b ; that is,
! b n
"
π [f (x)]2 dx = lim π [f (xi )]2 !xi .
a !!xi !→0
i=1
Consequently, we can calculate the volume of the solid of revolution generated by rotating the
region in Figure 7.1a around the x -axis, by means of the definite integral
! b
V = π [f (x)]2 dx. (7.8)
a
The volume of the solid of revolution has been defined by the limit in 7.7, but for evaluation of
this limit we use the definite integral in 7.8.
FIGURE 7.14 Integral for FIGURE 7.15 Integral for

volume when area under a curve volume when area between two
is rotated around x -axis curves is rotated around x -axis
y y
y = f (x) y = f (x)
a a
x y = g (x)
b b
dx x x
To avoid memorizing 7.8 as a formula, we use the technique introduced in Section 7.1.
For the volume obtained by rotating the region in Figure 7.1a about the x -axis, we construct a
rectangle of width dx at position x , as shown in either Figure 7.3 or Figure 7.14. When this
rectangle is rotated around the x -axis, the volume of the disc generated is
π [f (x)]2 dx,
where dx is the thickness of the disc and π [f (x)]2 is the area of its flat surface. This disc is
pictured as representing a large number of such discs between a and b . We find the required
volume by adding volumes of all such discs and taking the limit as their widths approach zero.
But this is the concept of the definite integral, so that we write for the limit-summation process
! b
V = π [f (x)]2 dx,
a
where limits x = a and x = b identify x -positions of first and last discs, respectively. This is
called the disc method for finding the volume of a solid of revolution.
A slightly more general problem is that of finding the volume of the solid of revolution
generated by rotating a region bounded by two curves and two vertical lines as shown in Figure
7.15 around the x -axis. If a representative rectangle of width dx at position x is rotated around
the x -axis, the volume formed is a washer. Since the outer and inner radii of this washer are
f (x) and g(x) , respectively, its volume is
{π [f (x)]2 − π [g(x)]2 } dx.
To add volumes of all such washers between a and b , and at the same time take limits as their
widths approach zero, we use the definite integral
! b
V = {π [f (x)]2 − π [g(x)]2 } dx. (7.9)
a
EXAMPLE 7.6
Prove that the volume of a sphere of radius r is 4π r 3 /3.
FIGURE 7.16 Volume of SOLUTION A sphere of radius r is formed when the semicircle x 2 + y 2 ≤ r 2 (y ≥ 0)
a sphere is rotated around the x -axis. The volume of the representative disc generated by rotating the
y rectangle in Figure 7.16 around the x -axis is
x 2 + y2 = r 2
πy 2 dx = π(r 2 − x 2 ) dx.
y
Since the volume formed by the left quarter circle is the same as for the right quarter circle, we
calculate the volume generated by the right quarter and double the result:
−r dx r x
! ) *r
r
2 2 2 x3 4
V =2 π(r − x ) dx = 2π r x − = π r 3.
0 3 0 3
EXAMPLE 7.7
FIGURE 7.17 Volume of Prove that the volume of a right circular cone of base radius r and height h is π r 2 h/3.
a right circular cone
SOLUTION The cone can be generated by rotating the triangle in Figure 7.17 around the
y
rx
x -axis. Volume of the representative disc formed by rotating the rectangle shown is
y=
h 4 rx 52
πy 2 dx = π dx;
x=h h
r
y hence,
! ) *h
h
π r 2x2 πr2 x3 1
x x V = dx = = π r 2 h.
dx 0 h2 h2 3 0 3
h
EXAMPLE 7.8
Find the volume of the solid of revolution obtained by rotating the region bounded by the curves
y = 1 − x 2 and y = 4 − 4x 2 around (a) the x -axis and (b) the line y = −1.
SOLUTION
(a) Let us ballpark the answer before we begin. If the rectangle in Figure 7.18a is rotated
around the x -axis, it produces a cylinder with radius 4 and height 2, and therefore of
FIGURE 7.18a Estimation of volume FIGURE 7.18b Volume when area is

of solid of revolution rotated around the x -axis
y y
4
y=4− 4x 2 4
y = 4 − 4x2
y = 1 − x2
y = 1 − x2 y2
1
1 y1
−1 1 x
−1 1x
volume π(4)2 (2) = 32π cubic units. We might estimate that the required volume
would be about half of this. Let us find out. When the rectangle in Figure 7.18b is
rotated around the x -axis, the volume of the washer formed is
(πy22 − πy12 ) dx = π [(4 − 4x 2 )2 − (1 − x 2 )2 ] dx = 15π(1 − 2x 2 + x 4 ) dx.

Because of the symmetry of the region, we rotate only the right half and double the
result:
! 1 ) *1
2 4 2x 3 x5
V =2 15π(1 − 2x + x ) dx = 30π x − + = 16π,
0 3 5 0
FIGURE 7.19 Volume exactly our estimate.

when area is rotated around the (b) When the rectangle in Figure 7.19 is rotated around the line y = −1, the inner
line y = −1
and outer radii of the washer are r1 and r2 , respectively. Now r1 is a length in the
y y -direction, and in Section 7.1 we learned that to calculate lengths in the y -direction,
y=4− 4x 2 4 we take upper y minus lower y . Hence, r1 = (1 − x 2 ) − (−1) = 2 − x 2 . Similarly,
r2 = (4 − 4x 2 ) − (−1) = 5 − 4x 2 . The volume of the washer is therefore
y = 1 − x2
(π r22 − π r12 ) dx = [π(5 − 4x 2 )2 − π(2 − x 2 )2 ] dx = 3π(7 − 12x 2 + 5x 4 ) dx.
r2
1 Consequently,
−1 r1 1 x ! 1 - .1
y = −1
V =2 3π(7 − 12x 2 + 5x 4 ) dx = 6π 7x − 4x 3 + x 5 0
= 24π.
0
We can also rotate horizontal rectangles around vertical lines to produce discs and washers as
illustrated in the next two examples.
EXAMPLE 7.9
Find the volume of the solid of revolution when the region enclosed by the curves y = ln x ,
y = 0, y = 1, and x = 0 is revolved around the y -axis.
SOLUTION When the rectangle in Figure 7.20 is rotated around the y -axis, the volume of
the disc formed is
π x 2 dy = π(ey )2 dy = π e2y dy.
The required volume is therefore
! 1 ) *1
2y 1 2y π
V = π e dy = π e = (e2 − 1).
0 2 0 2
FIGURE 7.20 Volume when area is rotated around the y -axis
1 y = ln x or x = e y
x
dy
1 x
EXAMPLE 7.10
Find the volume of the solid of revolution obtained by rotating the region enclosed by the curves
y = x 2 − 1 and y = 0 around the line x = 5.
FIGURE 7.21 Volume SOLUTION When the horizontal rectangle in Figure 7.21 is rotated around the line x = 5,
when area is rotated around the the volume of the washer formed is
line x = 5
y (π r22 − π r12 ) dy = π [(5 − x2 )2 − (5 − x1 )2 ] dy.
−1 x2 x1 1 Since x1 and x 2 − 1, we solve this √

√ x2 are x -coordinates of points on the curve y = √ equation
r1
x for x = ± y + 1. Since x1 > 0 and x2 < 0, we set x1 = y + 1 and x2 = − y + 1.
r2 Thus, the volume of the washer can be expressed as
dy
−1 14
y = x2 − 1 3 52 4 3 52 2 3
x=5 π 5+ y+1 − 5− y+1 dy = 20π y + 1 dy.
The required volume is therefore

! 0 ) *0
3 2 3/2 40π
V = 20π y + 1 dy = 20π (y + 1) = .
−1 3 −1 3
Washers in Example 7.10 are not as straightforward as in previous examples. For some problems,
washers are totally inappropriate. Consider, for example, rotating the region in Figure 7.22
around the x -axis. Each of the rectangles shown yields a washer with a volume formula different
from the others. As a result, use of washers requires six definite integrals and only one of these
is easy to set up. Determination of this volume seems to lend itself to the use of horizontal rather
than vertical rectangles.
FIGURE 7.22 Volume for

which washers are inappropriate
y
b
x = g ( y)
x = f ( y)
a
x
To see that this is indeed true, we divide the interval a ≤ y ≤ b into n parts by the points
a = y0 < y1 < y2 < · · · < yn−1 < yn = b.
In each subinterval yi−1 ≤ y ≤ yi , we find the midpoint
yi−1 + yi
yi∗ =
2
FIGURE 7.23 Approximation of volume of

solid of revolution by cylindrical shells
y
y*n
yn = b
yn−1
y*i
yi x = g ( y)
yi−1
y*1 x = f ( y)
y1
y0 = a
and construct a rectangle of length f (yi∗ )−g(yi∗ ) and width yi −yi−1 , as shown in Figure 7.23.
When this i th rectangle is rotated around the x -axis, a cylindrical shell is formed. Since the
length of the shell is f (yi∗ ) − g(yi∗ ) , and its inner and outer radii are yi−1 and yi , respectively,
its volume is
(πyi2 − πyi−
2 ∗ ∗ ∗ ∗
1 )[f (yi ) − g(yi )] = π(yi + yi−1 )(yi − yi−1 )[f (yi ) − g(yi )]
= 2πyi∗ [f (yi∗ ) − g(yi∗ )] !yi ,
where !yi = yi − yi−1 . When we add the volumes of all such shells, we obtain an approxi-
mation to the required volume V :
n
"
2πyi∗ [f (yi∗ ) − g(yi∗ )] !yi .
i=1
If we let the number of rectangles get larger and larger, and at the same time require the widths
to get smaller and smaller, it seems reasonable to expect that approximations will get better and
better. As the norm !!yi ! approaches zero, the limit should yield what we think is V . We
therefore define the required volume as
n
"
V = lim 2πyi∗ [f (yi∗ ) − g(yi∗ )]!yi . (7.10)
!!yi !→0
i=1
But the right side of this definition is the definite integral of 2πy [f (y) − g(y)] with respect to
y from y = a to y = b ; that is,
! b n
"
2πy [f (y) − g(y)] dy = lim 2πyi∗ [f (yi∗ ) − g(yi∗ )]!yi .
a !!yi !→0
i=1
Consequently, the volume of the solid of revolution generated by rotating the region in Fig-
ure 7.22 around the x -axis is given by the definite integral
! b
V = 2πy [f (y) − g(y)] dy. (7.11)
a
In practice we develop the integral in 7.11 by drawing a rectangle of width dy at position

y as shown in Figure 7.24a. When this rectangle is rotated around the x -axis, the volume of the
cylindrical shell generated is approximately
2πy [f (y) − g(y)] dy.
We obtain this by picturing the shell as being cut along the rectangle and opened up into a slab
with dimensions dy , f (y) − g(y) , and 2πy as in Figure 7.24b. Thickness dy of the shell
corresponds to the thickness of the slab; length f (y) − g(y) of the shell corresponds to that
of the slab; and inner circumference 2πy of the shell corresponds to the width of the slab. If
we now add the volumes of all such cylindrical shells, and at the same time take the limit as
their widths approach zero, we obtain the required volume. But this limit summation defines
the definite integral
! b
2πy [f (y) − g(y)] dy.
a
It follows that
! b
V = 2πy [f (y) − g(y)] dy.
a
The method described is called the cylindrical shell method for finding the volume of a solid
of revolution.
FIGURE 7.24a Using differentials to set FIGURE 7.24b Volume of cylindrical

up integral for cylindrical shells shell of thickness dy
y
f ( y) − g ( y)
b
x = g ( y)
dy
dy
x = f ( y)
a
y
EXAMPLE 7.11
Find the volume of the solid of revolution when the region enclosed by x = 2y − y 2 and the
y -axis is revolved around the x -axis.
FIGURE 7.25a Volume when area is FIGURE 7.25b Estimate of volume of

rotated around the x -axis solid of revolution
y y
2
2
x = 2y − y 2
(1, 1)
x
dy
y 1 x
x
SOLUTION The volume of the cylindrical shell formed by rotating the rectangle in Figure 7.25a
around the x -axis is approximately
(2πy)(x) dy = 2πy(2y − y 2 ) dy = 2π(2y 2 − y 3 ) dy.
Hence,
! 2 ) *2 + ,
2 3 2y 3 y4 16 8π
V = 2π(2y − y ) dy = 2π − = 2π −4 = .
0 3 4 0 3 3
We can check that this is reasonable by rotating the rectangle in Figure 7.25b around the x -axis.
The cylinder so generated has volume π(2)2 (1) = 4π . That the above volume is two-thirds of
this is acceptable.
EXAMPLE 7.12
Use cylindrical shells to calculate the volume in Example 7.10.
SOLUTION When we rotate the rectangle in Figure 7.26 around the line x = 5, the volume
of the cylindrical shell is approximately
2π(5 − x)(0 − y) dx = 2π(5 − x)(1 − x 2 ) dx = 2π(x 3 − 5x 2 − x + 5) dx.
Total volume is therefore
! 1 ) *1
3 2 x4 5x 3 x2
V = 2π(x − 5x − x + 5) dx = 2π − − + 5x
−1 4 3 2 −1
+ , + ,
1 5 1 1 5 1 40π
= 2π − − + 5 − 2π + − −5 = .
4 3 2 4 3 2 3
This is the result obtained by the washer method in Example 7.10.
FIGURE 7.26 Volume when area is rotated around the line x = 5
y x=5
dx
−1 1
y x
y = x2 − 1
−1
Even though the region in this example is symmetric about the y -axis, the volume generated
by the right half is less than that generated by the left. For this reason we cannot integrate
from x = 0 to x = 1 and double the result. If rotation were performed about the x -axis, we
could indeed integrate over either half of the interval and double. Finally, for rotation about the
y -axis, we would integrate over only one-half of the interval and neglect the other half in order
to eliminate duplications.
The volume in Example 7.12 was also calculated with washers in Example 7.10. The
volume in Example 7.11 could also be done with washers, but not so easily. Try it. Washers
cannot be used in the following example.
EXAMPLE 7.13
Figure 7.27 shows a plot of the function f (x) = x −1 sin x on the interval π/2 ≤ x ≤ π .
Find the volume of the solid of revolution when the area bounded by this curve and the lines
x = π/2 and y = 0 is rotated around the y -axis.
FIGURE 7.27 Volume SOLUTION When we rotate the rectangle shown around the y -axis, the volume of the repre-
when area is rotated around the sentative cylindrical shell is approximately
y -axis + ,
y
sin x
2π(x) dx = 2π sin x dx.
0.6 x
y=_
1
x sin x
0.4 The total volume is therefore
! π 7 8π
0.2 V = 2π sin x dx = 2π − cos x = 2π.
π/2 π/2
π π x
2
Equations 7.9 and 7.11 provide two methods for calculating volumes of solids of revolution:
washers and shells. Where applicable, both methods give the same results; as they should. We
illustrate this for the region bounded by the curves in Figure 7.28. As indicated in the figure, each
curve defines y as a function of x , and x as a function of y . When this region is rotated around
the x -axis, cylindrical shells can be used to calculate the volume of the solid of revolution,
! d ! d ! d
V = 2πy [q(y) − p(y)] dy = 2π yq(y) dy − 2π yp(y) dy. (7.12)
c c c
FIGURE 7.28 Demonstration that washers and cylindrical shells give the same volume
d
y = f (x)
or
x = p ( y)
y = g (x)
or
c x = q ( y)
a b x
Suppose we make the change of variable x = q(y) in the first integral. When this equation
is solved for y in terms of x , the result is y = g(x) , and therefore dy = g * (x) dx . Because
x = a when y = c and x = b when y = d , we obtain
! d ! b
yq(y) dy = g(x)xg * (x) dx.
c a
Now the product rule for differentiation gives
d - .
x [g(x)]2 = 2xg(x)g * (x) + [g(x)]2 ,
dx
and therefore
1 d 1
g(x)xg * (x) = {x [g(x)]2 } − [g(x)]2 .
2 dx 2
So
! d ! b ) *
1 d - 2
. 1 2
yq(y) dy = x [g(x)] − [g(x)] dx
c a 2 dx 2
) *b ! b
1 2 1
= x [g(x)] − [g(x)]2 dx
2 a 2 a
! b
1 1 1
= b[g(b)]2 − a [g(a)]2 − [g(x)]2 dx
2 2 2 a
! b
1 1 1
= bd 2 − ac2 − [g(x)]2 dx.
2 2 2 a
Similarly, the change of variable x = p(y) on the second integral in 7.12 leads to
! d ! b
1 2 1 2 1
yp(y) dy = bd − ac − [f (x)]2 dx.
c 2 2 2 a
Substitution of these into 7.12 gives
) ! b * ) ! b *
1 1 1 1 1 1
V = 2π bd 2 − ac2 − [g(x)]2 dx − 2π bd 2 − ac2 − [f (x)]2 dx
2 2 2 a 2 2 2 a
! b ! b
= −π [g(x)]2 dx + π [f (x)]2 dx
a a
! b - .
= π [f (x)]2 − π [g(x)]2 dx.
a
But this is the integral obtained when the washer method is used to find the volume.
EXERCISES 7.2
In Exercises 1–12 use the disc or washer method to find the volume of ∗ 31. y = (x + 1)1/4 , y = −(x + 1)2 , x = 0, about x = 0
the solid of revolution obtained by rotating the region bounded by the
∗ 32. y = x 4 − 3, y = 0, about y = −1
curves about the line.
1. x 2 + y 2 = 36, about y = 0
In Exercises 33–36 a calculator or computer is needed to find the points
2. y 2 = 5 − x, x = 0, about x = 0 of intersection of the curves. Find the volume of the solid of revolution
when the region bounded by the curves is rotated about the line, correct
3. y = x 2 + 4, y = 2x 2 , about y = 0
to three decimal places.
2
4. x − y = 16, x = 20, about x = 0 √
∗ 33. y = x 3 − x, y = x, about x = 0
5. x − 1 = y 2 , x = 5, about x = 1 −2 x 2
∗ 34. y = e , y =4−x , about y = −1
6. x + y + 1 = 0, 2y = x − 2, y = 0, about y = 0
1
7. y = 4x 2 − 4x, y = x 3 , about y = −2 ∗ 35. y = √ , y = 16 − x 2 , about y = −1
x−1
8. x = 2y − y 2 − 2, x = −5, about x = 0 √
∗ 36. y = 4 − x, y = x 3 + 1, y = 0, about y = 0
2
9. y = 5 − x, x = 0, about x = 6
10. y = x − 2x, y = 2x − x 2 ,
2
about y = 2 ∗ 37. A tapered rod of length L has circular cross-sections. If the radii
of its ends are a and b , what is the volume of the rod?
11. y = csc x, y = 0, x = π/4, x = 3π/4, about y = 0
∗ 38. During one revolution an airplane propeller displaces an amount
12. y = ln (x + 1), y = 1, x = 0, about x = 0 of air that can be calculated as a volume of a solid of revolution. If
the region yielding the volume is that bounded by the curves 64x =
y(y−4) and 64x = y(4 −y) , and is rotated about the x -axis, calculate
In Exercises 13–24 use the cylindrical shell method to find the volume the volume of air displaced.
of the solid of revolution obtained by rotating the region bounded by
the curves about the line. ∗ 39. (a) If a sphere of radius r is sliced a distance h from its centre,
show that the volume of the smaller piece is
13. y = 1 − x 3 , x = 0, y = 0, about x = 0
√ π
14. y = − 4 − x, x = 0, y = 0, about y = 0
V = (r − h)2 (2r + h).
3
15. y = (x − 1)2 , y = 1,about x = 0 (b) Use the result in part (a) to find the ratio of h to r in order
√ that the smaller piece will have volume equal to one-third
16. x + y = 4, y = 2 x − 1, y = 0, about y = 0
of the sphere. Give your answer to four decimal places.
17. y = 3x − x 2 , y = x 2 − 3x , about x = 4
∗ 40. An embankment is to be built around the circular wading pool in
18. y = 2 − |x|, y = 0, about y = −1 figure (a) below. Figure (b) shows a cross-section of the embankment.
√
19. x = y 3 , y = 2 − x, y = 0, about y = 1 (a) Find a cubic polynomial y = ax 3 + bx 2 + cx + d to fit
the three points (0, 0) , (4, 2) , and (6, 0) .
20. y = x 2 , y = −x 2 , x = −1, about x = −1
√ (b) Determine the amount of fill required to build the embank-
21. y = − 9 − x, x = 0, y = 0, about y = 0 ment.
22. y = x, xy = 9, x + y = 10, (x ≥ y) , about x = 0
6m
23. y = 0, (x + 1)y = sin x, 0 ≤ x ≤ 2π , about x = −1
24. y = 10 − x 2 , x 2 y = 9, about y = 0
10 m
In Exercises 25–32 use the most appropriate method to find the volume
of the solid of revolution obtained by rotating the region bounded by
the curves about the line.
(a)
∗ 25. (x 2 + 1)2 y = 4, y = 1, about x = 0
∗ 26. y = (x − 1)2 − 4, 5y = 12x, x = 0, (x ≥ 0) , about y
x=0
∗ 27. x 2 − y 2 = 5, 9y = x 2 + 9, 9y + x 2 + 9 = 0, (−3 ≤ x ≤ 3) ,
about x = 0 2m
∗ 28. y = |x 2 − 1|, x = −2, x = 2, y = −1, about y = −1 x

4m
∗ 29. y = x 2 − 2, y = 0, about y = −1
3 6m
∗ 30. x = 4 + 12y 2 , x − 20y = 24, y = 0, about y = 0 (b)
∗∗ 41. Find the volume of the donut obtained by rotating the circle in the
figure below about the y -axis. y
L
H
a x
a−b a a+b x
∗∗ 44. A right circular cone of height H and base radius r has its vertex
at the centre of a sphere of radius R ( R < H ). Find that part of the
volume of the sphere inside the cone.
∗∗ 45. Devise a way to calculate the volume of the solid of revolution
when the region in the first quadrant bounded by
∗∗ 42. A cylindrical hole is bored through the centre of a sphere, the length y = 1 − x2, x = 0, y =0
of the hole being L . Show that no matter what the radius of the sphere,
the volume of the sphere that remains is always the same and equal to is rotated about the line y = x + 1. Hint: Distance formula 1.16.
the volume of a sphere of diameter L .
∗∗ 46. A sphere of ice cream is to be placed in a cone of height 1 unit
∗∗ 43. Water half fills a cylindrical pail of radius a and height L . When (figure below). What radius of the sphere gives the most volume of ice
the pail is rotated about its axis of symmetry with angular speed ω cream inside the cone (as opposed to above the cone) for a cone with
(figure following), the surface of the water assumes a parabolic shape, base angle 2θ ?
the cross-section of which is given by
ω2 x 2
y =H + ,
2g
1
where g > 0 is the acceleration due to gravity and H is a constant.
Find the speed ω , in terms of L , a , and g , at which water spills over 2θ
the top.
7.3 Lengths of Curves

Formula 1.10 defines the length of the straight-line segment joining two points (x1 , y1 ) and
(x2 , y2 ) . The formula s = rθ gives the length of the arc of a circle of radius r subtended by
an angle θ at the centre of the circle. In this section we derive a result that will theoretically
enable us to find the length of any curve.
FIGURE 7.29a Finding the length FIGURE 7.29b Approximating the length of a
of a curve curve by straight line segments
y y
Pn−1
Pn−2
y = f (x)
B (b, f (b)) B = Pn
P3
P2
P1
(xi, yi) Pi (xi , yi )
A (a, f (a)) A = P0 Pi−1(xi−1, yi −1)

x xi − 1 xi x
/ Consider / the length

0 finding 0 of the curve C : y = f (x) in Figure 7.29a joining points
A a, f (a) and B b, f (b) . To find its length L we begin by approximating C with a series
of straight-line segments. Specifically, we choose n − 1 consecutive points on C between A

and B ,
A = P0 , P1 , P2 , . . . , Pn−1 , Pn = B,
and join each Pi−1 to Pi ( i = 1, . . . , n ) by means of a straight-line segment, as in Figure 7.29b.
If coordinates of Pi are denoted by (xi , yi ) , then the length of the line segment joining Pi−1
and Pi is
3
!Pi−1 Pi ! = (xi − xi−1 )2 + (yi − yi−1 )2 .
For a large number of these line segments, it is reasonable to approximate L by the sum of the
lengths of the segments:
n
" n 3
"
!Pi−1 Pi ! = (xi − xi−1 )2 + (yi − yi−1 )2 .
i=1 i=1
In fact, as we increase n and at the same time decrease the length of each segment, we expect
the approximation to become more and more accurate. We therefore define
n
" n 3
"
L= lim !Pi−1 Pi ! = lim (xi − xi−1 )2 + (yi − yi−1 )2 , (7.13a)
!!xi !→0 !!xi !→0
i=1 i=1
where !xi = xi − xi−1 . When we note that yi−1 = f (xi−1 ) and yi = f (xi ) , we have
n 3
"
L= lim (!xi )2 + [f (xi ) − f (xi−1 )]2
!!xi !→0
i=1
9 1 22
n
" f (xi ) − f (xi−1 )
= lim 1+ !xi . (7.13b)
!!xi !→0 !xi
i=1
In order for 7.13b to be a useful definition from the point of view of calculation, we must find
a convenient way to evaluate the limit summation. To do this we assume that f * (x) exists at
every point in the interval a ≤ x ≤ b , and apply the mean value theorem to f (x) on each
subinterval xi−1 ≤ x ≤ xi (Figure 7.30). The theorem guarantees the existence of at least one
point xi∗ between xi−1 and xi such that
f (xi ) − f (xi−1 ) f (xi ) − f (xi−1 )

f * (xi∗ ) = = .
xi − xi−1 !xi
Consequently, using these points xi∗ , we can express the length of C in the form
n :
"
L= lim 1 + [f * (xi∗ )]2 !xi . (7.14)
!!xi !→0
i=1
FIGURE 7.30 Mean value theorem applied to i th line segment
y = f (x) B = Pn
Pi
Pi−1
A = P0
xi−1 x*i xi x
But
3 the right side of this equation is the definition of the definite integral of the function
1 + [f * (x)]2 with respect to x from x = a to x = b :
! b 3 n :
"
1 + [f * (x)]2 dx = lim 1 + [f * (xi∗ )]2 !xi .
a !!xi !→0
i=1
Thus we can calculate the length of C by the definite integral

9 + ,2
! b
dy
L= 1+ dx. (7.15)
a dx
In this derivation we assumed that f * (x) was defined at each point in the interval a ≤
x ≤ b , necessary for the mean value theorem to apply. But in order to guarantee
3 existence of
the definite integral in 7.15, Theorem 6.2 requires continuity of the integrand 1 + (dy/dx)2 .
Consequently, to ensure that the length of a curve can be calculated by means of 7.15, we assume
that f (x) has a continuous first derivative on the interval a ≤ x ≤ b . When we study improper
integrals in Section 7.10, we shall be able to weaken the continuity requirement.
EXAMPLE 7.14
√
Find the length of the curve y = x 3/2 from (1, 1) to (2, 2 2) (Figure 7.31).
FIGURE 7.31 Length of SOLUTION Since dy/dx = (3/2)x 1/2 , equation 7.15 gives
curve y = x 3/2 from x = 0 to 9
x=2 ! 2 + ,2 ! 2
;
3 9
y L= 1+ x 1/2 dx = 1+ x dx
y= x 3/2 1 2 1 4
(2, 2√2 ) < + ,3/2 =2 '+ ,3/2 + ,3/2 (
8 9 8 11 13 1
= 1+ x = − = (223/2 − 133/2 ).
27 4 27 2 4 27
1
(1, 1)
x
The result in equation 7.15 is clearly useful when the equation of the curve is expressed in the
form y = f (x) . If the curve is defined in the form x = g(y) (Figure 7.32), a similar analysis
gives
9 + ,2
! d
dx
FIGURE 7.32 Length L= 1+ dy. (7.16)
c dy
of curve with equation in form
x = g(y)
For example, we may write the equation of the curve in Example 7.14 in the form x = y 2/3 ,
y in which case dx/dy = (2/3)y −1/3 , and
B ( g(d), d)
√ 9 + ,2 √ ;
x = g( y)
! 2 2 ! 2 2
2 4
L= 1+ y −1/3 dy = 1+ y −2/3 dy
1 3 1 9
! √ 9 ! √ 3
2 2 2 2
A ( g(c), c) 4 + 9y 2/3 4 + 9y 2/3
= dy = dy.
1 9y 2/3 1 3y 1/3
x
If we set u = 4 + 9y 2/3 , then du = 6y −1/3 dy , and
! 22 ) *22
1 √ du 1 2 3/2 1
L= u = u = (223/2 − 133/2 ).
3 13 6 18 3 13 27
FIGURE 7.33 Using differentials to set up integral for length of a curve
y
B
y = f (x)
C
(x, y) dy
A dx
a x x + dx b x
Integrals 7.15 and 7.16 can both be interpreted geometrically. This interpretation will be useful
when we study parametric equations of curves in Section 9.1 and line integrals in Chapter 14.
Figure 7.33 shows the curve C of Figure 7.29a. At a point (x, y) on C , we draw the tangent
line to the curve. If dx is a short length along the x -axis at position x , the length of that part
of the tangent line between vertical lines at x and x + dx is given by
3
(dx)2 + (dy)2 .
This length along the tangent line closely approximates the length of the curve between the
same two vertical lines, the approximation being more accurate the shorter the length dx . If we
picture a large number of these tangential line segments between a and b , we can find the total
length of C by adding together all such lengths and taking the limit as each approaches zero.
But this process is represented by the definite integral, and we therefore write
! x=b 3 ! y=d 3
L= (dx)2 + (dy)2 or L= (dx)2 + (dy)2 . (7.17)
x=a y=c
The first integral is chosen when it is more convenient to integrate with respect to x , in which
case dx is taken outside the square root. We get
9 + ,2
! b
dy
L= 1+ dx,
a dx
which is equation 7.15. The second corresponds to a more convenient integral with respect to
y , in which case dy is taken outside the square root. We obtain
9 + ,2
! d
dx
L= 1+ dy,
c dy
which is equation 7.16.
EXAMPLE 7.15
Find the length of the curve 24xy = x 4 + 48 between (2, 4/3) and (3, 43/24) .
SOLUTION It is easier to express y in terms of x (rather than x in terms of y ),
x 4 + 48 x3 2
y = = + .
24x 24 x
We now use the fact that small lengths along the curve are approximated by
9 + ,2 9 + ,2
3 dy x2 2
FIGURE 7.34 Length of (dx)2 + (dy)2 = 1+ dx = 1+ − dx
dx 8 x2
curve 24xy = x 4 + 48 from x = 2
to x = 3 ; ;
x4 1 4 1
y = 1+ − + dx = (x 8 + 32x 4 + 256) dx
64 2 x4 64x 4
1 3 x 4 + 16
x3 2 = (x 4 + 16)2 dx = dx.
y= + 8x 2 8x 2
24 x
43 The total length of the curve is therefore (Figure 7.34)

(3, )
24
4 dy ! ! + , ) *3
(2, ) 3
x 4 + 16 1 3
16 1 x3 16 9
3 2
dx L= 2
dx = x + 2 dx = − = .
2 8x 8 2 x 8 3 x 2 8
x x + dx x
EXERCISES 7.3
√ √
In Exercises 1–10 find the length of the curve. 14. x 2 − y 2 = 1 from (2, − 3) to (3, 2 2)
√
1. 8x 2 = y 3 from (1, 2) to (2 2, 4) 15. y = sin x from (0, 0) to (π, 0)
√ √ 16. y = ln (cos x) between the lines x = 0 and x = π/4
2. 3y = 2(x 2 + 1)3/2 from (−2, 10 5/3) to (−1, 4 2/3)
17. y = ln x from (1, 0) to any other point on the curve
3. x = 2(y − 2)3/2 from (0, 2) to (2, 3)
18. 8y 2 = x 2 (1 − x 2 ) (complete length)
4. y = (x − 1)3/2 from (2, 1) to (10, 27)
19. x = y 2 − 2y from (0, 0) to (0, 2)
x3 1
5. y = + from (1, 7/12) to (2, 13/6)
4 3x 20. x 2 /4 + y 2 /9 = 1 (complete length)
6. y = (ex + e−x )/2 between the lines x = 0 and x = 1 √
∗ 21. Find the length of the curve 3x = y(y − 3) from (−2/3, 1) to
4
x 1 (2/3, 4) .
∗ 7. y = + from (2,129/32) to (1/2, 33/64)
4 8x 2
∗ 22. Find the length of the curve x 2/3 + y 2/3 = 1.
∗ 8. 36xy = x 4 + 108 from (3, 7/4) to (4, 91/36)
∗ 23. If n is any number other than 1 or −1, and 0 < a < b , find the
y7 1 length of the curve
∗ 9. x = + between the lines y = −1 and y = −2
20 7y 5
x n+1 1
5 3 y = +
∗ 10. y = x /5 + 1/(12x ) from (1, 17/60) to (2,3077/480) n+1 4(n − 1)x n−1
between x = a and x = b .
In Exercises 11–20 set up (but do not evaluate) a definite integral for
the length of the curve. ∗ 24. If n is any number greater than 1/2, and 0 < a < b , find the
length of the curve
11. y = x 2 from (0, 0) to (1, 1)
x 2n+1 1
12. y = 3x 2 − 4x from (1, −1) to (2, 4) y = +
4(2n − 1) (2n + 1)x 2n−1
√
13. x 2 − y 2 = 1 from (1, 0) to (2, 3) between x = a and x = b .
7.4 Work
When a body moves a distance d along a straight line l under the action of a constant force F ,
which acts in the same direction as the motion (Figure 7.35), the work done on the body by F
is defined as
W = F d. (7.18)
FIGURE 7.35 Definition of work done FIGURE 7.36 Approximating work

by a constant force along a straight line done by a variable force
Direction x1* x2* x*i

of
motion d a = x 0 x1 x2 xi − 1 xi xn − 1 b = x n x
Direction l
of
force
Forces often vary either in magnitude or direction, or both, and when such forces act on
the body, calculation of the work done is not as simple as using equation 7.18. In this section
we use definite integrals to calculate work done by forces that always act along the direction of
motion, but do not have constant magnitude. Specifically, let us consider a particle that moves
along the x -axis from x = a to x = b , where b > a , under the action of some force. Suppose
the force always acts in the positive x -direction, but its size, which we denote by F (x) , is not
constant along the x -axis. To find the work done by the force as the particle moves from x = a
to x = b , we cannot simply multiply force by distance. [Where would we evaluate F (x) ?]
Instead, we divide the interval a ≤ x ≤ b into n subintervals by points (Figure 7.36):
a = x0 < x1 < x2 < · · · < xn−1 < xn = b.
In each subinterval xi−1 ≤ x ≤ xi , we choose a point xi∗ . If lengths of subintervals are

small and F (x) is continuous, then F (x) does not vary greatly over any given subinterval. In
such a situation, we may approximate F (x) by a constant force F (xi∗ ) on each subinterval
xi−1 ≤ x ≤ xi . It follows that work done by the force over the i th subinterval is approximately
F (xi∗ )(xi − xi−1 ) = F (xi∗ ) !xi .
Furthermore, an approximation to the total work done by the force as the particle moves from
x = a to x = b is
n
"
F (xi∗ ) !xi .
i=1
As n becomes larger and each !xi approaches zero, this approximation becomes better, and
we therefore define
n
"
W = lim F (xi∗ ) !xi . (7.19)
!!xi !→0
i=1
But this limit may also be interpreted as the definite integral of F (x) with respect to x from
x = a to x = b . Consequently, W may be calculated with the definite integral
! b
W = F (x) dx. (7.20)
a
It is simple enough to interpret parts of the integral in equation 7.20, and in so doing we begin to
feel the definite integral at work. The integrand F (x) is the force at position x , and dx is a small
distance along the x -axis. The product F (x) dx is therefore interpreted as the (approximate)
7.4 Work 437
work done by the force along dx . The definite integral then adds over all dx ’s, beginning at
x = a and ending at x = b , to give total work.
What is important about equation 7.20 is not the particular form of the definite integral, but
the fact that work done by a force can be evaluated by means of a definite integral. As we solve
work problems in this section, we find that the form of the definite integral varies considerably
from problem to problem, but the underlying fact remains that each problem is solved with a
definite integral.
With the exception of work problems that involve emptying tanks (Example 7.18), we
recommend that you always make a diagram illustrating the physical setup at some intermediate
stage between start and finish. Determine forces at this position in order to set up the work
integral.
In the derivation of 7.20 we assumed continuity of F (x) , and this guarantees existence of
the definite integral.
EXAMPLE 7.16
Find the work necessary to expand a spring from a stretch of 5 cm to a stretch of 15 cm if a force
of 200 N stretches it 10 cm.
SOLUTION Let the spring be stretched in the positive x -direction, and let x = 0 correspond
to the free end of the spring in the unstretched position in the top half of Figure 7.37. In the
bottom half of this figure we have shown the spring stretched to an intermediate position.
FIGURE 7.37 Work to stretch a spring
Unstretched position
dx
0 0.05 x 0.15 x
Stretched an amount x
According to Hooke’s law, when the spring is stretched an amount x , the restoring force in
the spring is proportional to x :
Fs = −kx,
where k > 0 is a constant. The negative sign indicates that the force is in the negative x -
direction. Since Fs = −200 N when x = 0.10 m,
−200 = −k(0.10),
and it follows that k = 2000 N/m. The force required to counteract the restoring force of the
spring when it is stretched an amount x is therefore
F (x) = 2000x.
Work done by this force in stretching the spring from position x a further amount dx is approx-
imately
2000x dx,
and hence total work to increase the stretch from 5 cm to 15 cm is
! 0.15 ) *0.15
x2
W = 2000x dx = 2000 = 20 J.
0.05 2 0.05
We could check on whether this answer is reasonable as follows. The minimum force during
stretching is at the beginning when the stretch is 5 cm where F = 2000(0.05) = 100 N.
Maximum force is at 15 cm where F = 2000(0.15) = 300 N. Work should be between
100(0.1) = 10 J and 300(0.1) = 30 J. In essence, we are applying Theorem 6.6 from Section
6.4.
EXAMPLE 7.17
A cable hangs vertically from the top of a building so that a length of 100 m, having a mass of
300 kg, is hanging from the edge of the roof. What work is required to lift the entire cable to
the top of the building?
SOLUTION When the cable has been lifted to an intermediate point where its lower end
is y metres above its original position (Figure 7.38), the length of cable still hanging is
100 − y m. Since each metre of cable has mass 3 kg, the mass of 100 − y m of cable is
3(100 − y) kg. It follows that the force that must be exerted to overcome gravity on this much
cable is 9.81(3)(100 − y) N. The work that this force does in raising the end of the cable an
additional amount dy is approximately
9.81(3)(100 − y) dy,
and total work to raise the entire cable is therefore
! 100 ) *100
y2
W = 9.81(3)(100 − y) dy = 29.43 100y − = 147 150 J.
0 2 0
FIGURE 7.38 Work to raise a cable
Cable y = 100
Building dy
y
(Initial position of end of cable)
y=0
EXAMPLE 7.18
A tank in the form of an inverted right circular cone of depth 10 m and radius 4 m is full of
water. Find the work required to pump the water to a level 1 m above the top of the tank.
SOLUTION A cross-section of the tank is shown in Figure 7.39. Suppose we approximate

water in the tank with circular discs formed by rotating the rectangles shown around the y -axis.
7.4 Work 439
FIGURE 7.39 Work to empty a tank
y
4
1 (4, 10)
dy
x
5
10 y= x
2
y
The force of gravity on the representative disc at position y is its volume π x 2 dy , multiplied
by the density of water (1000 kg/m 3 ), multiplied by the acceleration due to gravity ( −9.81),
−9.81(1000)π x 2 dy.
It is negative because gravity acts in the negative y -direction. The work done by an equal and
opposite force in lifting the disc to a level 1 m above the top of the tank (a distance 11 − y ) is
(11 − y)9.81(1000)π x 2 dy.
The total work to empty the tank is therefore

! 10
W = (11 − y)9810π x 2 dy.
0
To express x in terms of y , we note that x and y are coordinates of points on the straight line
through the origin and the point (4, 10) . Since the equation of this line is y = 5x/2, we obtain
! 10 + ,2 ! 10
2y
W = 9810π (11 − y) dy = 1569.6π (11y 2 − y 3 ) dy
0 5 0
) *10
11y 3 y4
= 1569.6π − = 5.75 × 106 J.
3 4 0
Two observations are noteworthy:

1. In Examples 7.16 and 7.17, the differential represents distance moved and the integrand
represents force, a direct application of the discussion leading to equation 7.20. In Example
FIGURE 7.40 Work to 7.18, the differential is part of the force, and the distance moved is part of the integrand.
empty a tank This is why we suggested earlier that it is not advisable to use equation 7.20 as a formula.
y The integrand is not always force and the differential is not always distance moved.
2. In each of these examples we set up the coordinate system; it was not given. We can use
4 1 any coordinate system whatsoever; but once we have chosen our coordinates, we must refer
dy y x everything to that system. For instance, were we to use the coordinate system in Figure
7.40 for Example 7.18, the definite integral would be
! 0 1 22
10 x 2
W = (1 − y)9810π (10 + y) dy.
−10 5
−10 Evaluation of this definite integral once again leads to W = 5.75 × 106 J.
Heat Engines
In the Application Preview, we introduced the problem of converting heat to mechanical energy.
Application Preview With the laws of thermodynamics, we can study this conversion in heat engines. Figure 7.41
Revisited shows gas in a cylinder closed on one end with a piston on the other. The force on the piston face
due to the pressure of the gas is F = P A , where P is the pressure of the gas in the cylinder
and A is the area of the face of the piston. If the piston is moved to the right a small amount
dx by the pressure of the gas in the cylinder, the work done is F dx . As the piston face moves
from position x1 to position x2 , the total work is
! x2 ! x2
W = F dx = P A dx.
x1 x1
FIGURE 7.41 Work done

But A dx is the change in the volume of gas in the cylinder due to the change dx in x . If
by gas in moving piston we denote this change by dV , and let V1 and V2 represent the volumes of gas in the cylinder
at positions x1 and x2 , respectively, then
Cylinder
! V2
Gas Piston
W = P dV . (7.21)
V1
x
x=0 dx
Positive values of this integral correspond to the situation described above when the gas is
expanding; negative values correspond to decreasing volume. Although we developed formula
7.21 on the basis of the cylinder and piston in Figure 7.41, it represents the work done by
expanding or compressing gases for any shape. In order to evaluate the definite integral, it is
necessary to specify P as a function of V .
FIGURE 7.42 Work done
Figure 7.42 shows the states through which the gas in a heat engine might pass during one
by heat engine for one cycle complete cycle. Beginning at point B and proceeding along C2 to A , the volume of the gas
in the engine is increasing while its pressure is decreasing. The gas is doing positive work, the
P B amount given by the integral in 7.21, where P is defined in terms of V by the equation for C2 .
As the gas returns to state B along C1 , integral 7.21 defines the work done on the gas. The
difference of these integrals represents the net work done by the gas during one cycle of the
C2 engine. In Section 14.3 we interpret the integrations as line integrals, and continue discussions
C1 from that point of view. For now, notice that subtraction of the integrals represents the area
A enclosed by C1 and C2 . In other words, we can calculate the output of the engine by calculating
V the area enclosed by the curves.
The Rankine cycle represents an idealized steam engine; the pressure and volume of steam
during a complete cycle might be those in Figure 7.43. Water at low temperature and pressure
(point A ) is heated at constant volume (along path AB ). Along BC the water is converted to
steam and expands slightly, and the expansion continues along CD . To complete the cycle, the
steam is cooled and condensed to water along DA . If the gas is expanded adiabatically along
CD (temperature is held constant), then P and V are related by P V γ = k , where γ > 0 and
k > 0 are constants ( γ = 1.4 for air). Using point C in Figure 7.43, k = 105 (0.02)1.4 =
4.2 × 102 . To determine the output of this particular engine, we calculate the area enclosed by
FIGURE 7.43 Rankine cycle for steam engine
5
P B(0.01, 10 )
105 C(0.02, 105)
5 × 104
104 D(0.104, 104)

A(0.01, 104)
V
0.02 0.04 0.06 0.08 0.10
7.4 Work 441
the curves: ! 0.104

W = (0.02 − 0.01)(105 − 104 ) + (kV −1.4 − 104 ) dV
0.02
) − 0 .4 *0.104
kV
= 900 + − 104 V = 2.5 × 103 J.
− 0 .4 0.02
EXERCISES 7.4
1. A spring requires a 10 N force to stretch it 3 cm. Find the work to 10. In Exercise 9, find the work to empty the tank through an outlet
increase the stretch of the spring from 5 cm to 7 cm. 2 m above the top of the tank.
2. Find the work to increase the stretch of the spring in Exercise 1 from
7 cm to 9 cm. ∗ 11. The ends of a trough are isosceles triangles with width 2 m and
depth 1 m. The trough is 5 m long and it is full of water. How much
3. A cage of mass M kilograms is to be lifted from the bottom of a
work is required to lift all of the water to the top of the trough?
mine shaft h metres deep. If the mass of the cable used to hoist the
cage is m kilograms per metre, find the work done.
∗ 12. How much work is required to lift the water in the trough of Exer-
4. A uniform cable of length 50 m and mass 100 kg hangs vertically cise 11 to a height of 2 m above the top of the trough?
from the top of a building 100 m high. How much work is required to
get 10 m of the cable on top of the building? ∗ 13. A rectangular swimming pool full of water is 25 m long and 10 m
5. A 2 m chain of mass 20 kg lies on the floor. If friction between floor wide. The depth is 3 m for the first 10 m of length, then decreases
and chain is ignored, how much work is required to lift one end of the linearly to 1 m at the shallow end. How much work is required to lower
chain 2 m straight up? Assume that the suspended portion of the chain the level of the water in the pool by 1/2 m?
makes a right angle with the portion on the floor.
∗ 14. How much work is required to empty the pool of Exercise 13 over
6. How much work is required to lift the end of the chain in Exercise
its edge?
5 only 1 m?
7. How much work is required to lift the end of the chain in Exercise ∗ 15. The force of repulsion between two point charges of like sign, one
5 a distance of 4 m? of size q and the other of size Q , has magnitude
8. A 5 m chain of mass 15 kg hangs vertically. It is required to lift
the lower end of the chain 5 m so that it is level with the upper end. qQ
Calculate the work done using each of the coordinate systems in the F = ,
4π &0 r 2
figure below.
where &0 is a constant and r is the distance between the charges. If Q
y=0
is placed at the origin, and q is moved along the x -axis from x = 2 to
y x = 5, find the work done by the electrostatic force.
∗ 16. Two positive charges q1 and q2 are placed at positions x = 5 and

y x = −2 on the x -axis. A third positive charge q3 is moved along the
y x -axis from x = 1 to x = −1. Find the work done by the electrostatic
y=0 y=0
forces of q1 and q2 on q3 (see Exercise 15).
9. A tank filled with water has the form of a paraboloid of revolution ∗ 17. If the chain in Exercise 5 is stretched out straight on the floor, and
with vertical axis (figure below). If the depth of the tank is 12 m and if the coefficient of friction between floor and chain is 0.01, what work
the diameter of the top is 8 m, find the work in pumping the water to is required to raise one end of the chain 2 m? The force of friction on
the top of the tank. that part of the chain on the floor is 0.01 times the weight of the chain
on the floor. Assume, unrealistically, that the suspended portion of the
y chain makes a right angle with the portion on the floor.
8
∗ 18. Two similar springs, each 1 m long in an unstretched position, have
spring constant k newtons per metre. The springs are joined together at
P (figure below), and their free ends are fastened to two posts 4 m apart.
What work is done in moving the midpoint P a distance b metres to
12
Parabolic the right?
cross-section
in xy-plane P
4
x
∗ 19. If the force of a crossbow (in newtons) is proportional to the draw a constant. The force needed to move the piston is F = P A , where
(in metres), and it is 200 N at a full draw of 50 cm, what work is required A is the area of the face of the piston. If the radius of the cylinder is 3
to fully draw the crossbow? units, find the work done in moving the piston from a point 10 units to
∗ 20. A bucket of water with mass 100 kg is on the ground attached to a point 5 units from the closed end.
one end of a cable with mass per unit length 5 kg/m. The other end ∗ 25. A town’s water is supplied from the water tower shown below.
of the cable is attached to a windlass 100 m above the bucket. If the The tower is a cylinder of length 5 m and radius 3 m capped on top
bucket is raised at a constant speed, water runs out through a hole in the and bottom by hemispheres of radius 3 m. The water is pumped from
bottom at a constant rate to the extent that the bucket would have mass a well 50 m below ground. If the diameter of the pipe from pump to
80 kg when it reaches the top. To further complicate matters, a pigeon tank is 10 cm, how much work is required to fill the tank initially?
of mass 2 kg lands on the bucket when it is 50 m above the ground. He
immediately begins taking a bath, splashing water over the side of the Hemisphere
bucket at the rate of 1 kg/m. Find the work done by the windlass in
raising the bucket 100 m. 6m
5m
∗ 21. A hemispherical tank with diametral plane at the top has radius
5 m. Hemisphere
(a) Show that if the depth of oil in the tank is h metres when
the tank is one-half full by volume, then h must satisfy the 20 m
equation Ground
10 cm
50 m
h3 − 15h2 + 125 = 0. Submersible pump
Find h to four decimal places. Water
(b) Find the work to empty the half-full tank if the oil has
density 750 kg/m 3 , and the outlet is at the top of the tank.
In Exercises 26–30 calculate the work done by a gas taken through the
∗ 22. Newton’s universal law of gravitation states that the force of at- cycle shown in the figure.
traction between two point masses m and M has magnitude
∗ 26. The Stirling cycle in the figure below consists of two isothermal
GmM processes and two constant-volume processes.
F = ,
r2
P C(0.2, 100 000)
100 000
where r is the distance between the masses and G = 6.67 × 10−11
N · m 2 /kg 2 is a constant.
(a) If M represents the mass of the earth and we regard it as a PV = k1
60 000
point mass concentrated at its centre, show that Newton’s B(0.2, 60 000) D(0.6, 100 000/3)
universal law of gravitation at the earth’s surface reduces to
F = mg , where g = 9.82 m/s 2 . Assume for the calcu- PV = k2
20 000
lation of M that the earth is a sphere with radius 6370 km
and mean density 5.52 × 103 kg/m 3 . A(0.6, 20 000)
(b) Use the original law F = GmM/r 2 with the earth re- 0.2 0.4 0.6 V
garded as a point mass to calculate the work required to
∗ 27. The Ericsson cycle in the figure below consists of two isothermal
lift a mass of 10 kg from the earth’s surface to a height of
processes and two constant-pressure processes.
10 km.
(c) Calculate the work in part (b) using the constant gravita- P B(0.01, 100 000)
tional force F = mg in part (a). Is there a significant 100 000
difference? C(0.02, 100 000)
∗ 23. A right circular cylinder with horizontal axis has radius r metres
and length h metres. If it is full of oil with density ρ kilograms per
50 000 PV = k1
cubic metre, how much work is required to empty the tank through an
outlet at its top? PV = k2
∗ 24. A gas is confined in a cylinder, closed on one end with a piston on D(2, 10 000)
the other. If the temperature is held constant, then P V = C , where A(0.1, 10 000)
P is the pressure of the gas in the cylinder, V is its volume, and C is 0.5 1.0 1.5 2.0 V
7.5 Energy 443
∗ 28. The Otto cycle for the internal combustion engine in the figure ∗ 31. A gas is confined in a cylinder, closed on one end with a piston
below consists of two adiabatic processes and two constant-volume on the other. If the gas expands adiabatically, then the pressure P and
processes. volume V of the gas obey the law
1 000 000
P C(0.0002, 1 040 000) P V 7/5 = C,
where C is a constant. The force on the piston face due to the pressure
of the gas is given by F = P A , where A is the area of the face of
600 000 B(0.0002, PV1.4 = k1
the piston. Show that if the piston is moved so as to reduce the volume
700 000) occupied by the gas from V0 to V0 /2, then the work done by the piston
PV1.4 = k2 D(0.0008, 150 000) is
200 000 5 / 2 /5 0 − 2 /5
A(0.0008, 100 000) 2 − 1 CV0 .
2
0.0002 0.0008 V ∗ 32. A drop of liquid with initial mass M kilograms falls from rest
under gravity and evaporates uniformly, losing mass m kilograms each
second. Neglecting air resistance, what work is done by gravity on the
∗ 29. The Diesel cycle in the figure below consists of two adiabatic drop before it completely evaporates?
processes, a constant-pressure process and a constant-volume process.
∗∗ 33. A diving bell of mass 10 000 kg is attached to a chain of mass
P B(0.0002, 2 300 000) 5 kg/m, and the bell sits on the bottom of the ocean 100 m below the
surface (figure below). How much work is required to lift the bell to
2 000 000 C(0.0005 75, 2 300 000)
deck level 5 m above the surface? Take into account the fact that when
the bell and chain are below the surface, they weigh less than when
1 500 000
above. The apparent loss in weight is equal to the weight of water
PV1.4 = k1 displaced by the bell and chain. Assume that the bell is a perfect cube,
1 000 000
D(0.002, 400 000) 2 m on each side; therefore, when completely submerged, it displaces
1.4
500 000 PV = k2 8 m 3 of water. Assume also that the chain displaces 1 L of water per
A(0.002, 90 000) metre of length.
0.0005 0.001 0.0015 0.002 V
∗ 30. The Brayton cycle for a gas turbine in the figure below consists of
two adiabatic processes and two constant-pressure processes.
P B(0.0001, 1 040 000) 3m

1000 000 C(0.006, 1 040 000) 5m
PV1.4 = k1
600 000
100 m
PV1.4 = k2
D(0.0024, 150 000) 2m
200 000
A(0.0004, 150 000)
0.0005 0.001 0.0015 0.002 V
7.5 Energy
Many engineering systems are analyzed on the basis of energy, and there are many kinds of
energy — potential, kinetic, thermal, and electrical, to name a few. Often one kind of energy is
transformed into another, or others, as a system evolves. We consider a number of examples of
this here.
In lifting a mass vertically, work is done against gravity. We say that the mass gains
gravitational potential energy as a result of the lift, the amount being equal to the work done.
For instance, when a mass m is lifted an amount h > 0 against gravity in Figure 7.44a, the
force is constant (provided that h is relatively small), and the work done is W = mgh , where
g = 9.81. We say that the mass has gained gravitational potential energy as a result of the lift,
GPE = mgh. (7.22)


Lifting a mass against gravity results in a gain in gravitational potential energy
Final Final
position P2 position P2
h h
Initial Initial
position P1 position P1
h0
P0
We also say that the difference in gravitational potential energies between initial and final
positions P1 and P2 is mgh . In Figure 7.44b, we have shown the same situation, but initial
position P1 is at height h0 above some other reference point denoted by P0 . At position P1 ,
the mass has GPE mgh0 relative to P0 ; at position P2 , it has GPE mg(h + h0 ) relative to P0 .
The difference in these GPEs is still mg(h + h0 ) − mgh0 = mgh .
We have mentioned gains in GPE, differences in GPE, and GPE at one point relative to
another, and this is what is important. Often, however, it is convenient to talk about GPE at a
point, rather than the difference in GPE between points. This can be accomplished by choosing
some specific point as having zero GPE and saying that the GPE at any other point is the
difference in GPE between the two points. For instance, if GPE is set equal to zero at P0 in
Figure 7.44b, then GPE is mgh0 at P1 and mg(h + h0 ) at P2 . The GPE at any point below
P0 is negative.
If m is allowed to drop from position P2 , it loses GPE as it falls. It picks up speed, and
energy is associated with the speed. This is called kinetic energy; it is defined by
1
mv 2 ,
KE = (7.23)
2
where v is the speed of m . Provided that air drag on m is ignored, whatever GPE m loses, it
gains in KE. To put it another way, the sum of its GPE and KE is always the same:
1
GPE + KE = C = mgh + mv 2 . (7.24)
2
This is called conservation of energy. Energy simply changes from gravitational potential to
kinetic during the fall. We apply conservation of energy in the following example.
EXAMPLE 7.19
A stone is thrown vertically upward with speed 20 m/s at the edge of a 50-m cliff. How fast is
it travelling when it strikes the base of the cliff?
SOLUTION Let us take the base of the cliff (Figure 7.45) as our place of zero gravitational
potential energy. When the stone is initially thrown upward, its GPE is mg(50) and its KE is
m(20)2 /2. If we substitute these into 7.24, we obtain
C = 50mg + 200m.
Substituting this back into 7.24 and replacing h by y gives
1
mgy + mv 2 = 50mg + 200m +⇒ 2gy + v 2 = 100g + 400.
2
This equation relates speed and height. At the bottom of the cliff y = 0, in which case
3
v 2 = 100g + 400 +⇒ v = 100g + 400 = 37.2.
The stone therefore strikes the bottom of the cliff at 37.2 m/s.
7.5 Energy 445
FIGURE 7.45 Schematic for stone thrown upward over the edge of a cliff
20 m/s
Cliff
50 m
y=0 (GPE = 0)
We can also solve the problem using the technique from Section 5.2. Since acceleration of
the stone is a = dv/dt = −9.81,
v = −9.81t + C.
(Here v is velocity rather than speed.) If we choose t = 0 when the stone is thrown upward,
then v(0) = 20, and this implies that C = 20. Hence,
dy
= v = −9.81t + 20.
dt
A second integration gives
y = −4.905t 2 + 20t + D.
Since y(0) = 50, it follows that D = 50, and
y = −4.905t 2 + 20t + 50.
The stone strikes the bottom of the cliff when

√
2 −20 ± 400 − 4(−4.905)(50)
0 = y(t) = −4.905t + 20t + 50 +⇒ t = .
−9.81
The positive solution is t = 5.827 s, and at this time
v = −9.81(5.827) + 20 = −37.2 m/s.
The energy solution related speed v to position y ; the solution using integration relates velocity
to time t . Which is preferable depends on the problem.
EXAMPLE 7.20
Show that when a spring is stretched (or compressed) an amount X , the potential energy stored
in the spring is kX 2 /2. If a mass m is attached to the end of the spring, the spring is stretched an
amount X , and the mass is then released, what is its speed at the instant the spring is unstretched?
SOLUTION The force necessary to maintain stretch x in the spring of Figure 7.46a is kx ,
where k is the spring constant. The work to stretch the spring X is
! ) *X
X
kx 2 1
W = kx dx = = kX 2 .
0 2 0 2
This is potential energy stored in the spring.

FIGURE 7.46a Calculation of potential FIGURE 7.46b Conservation of energy for

energy stored in a spring vibrating mass-spring system
x X x X
x=0 dx x=0
when spring when spring
unstretched unstretched
Suppose now that the mass is attached to the spring, the spring is stretched X , and the mass
is released. During the subsequent motion to the left (Figure 7.46b), potential energy stored in
the spring turns into kinetic energy of the mass (provided there is no friction between m and the
surface along which it slides, and there is no air drag on m ). Conservation of energy equation
7.24 still holds, but mgh is replaced by kx 2 /2:
1 1
C = kx 2 + mv 2 .
2 2
Initially, the stretch is X and the mass is not moving, so that
1
C = kX 2 .
2
Hence,
1 1 1 k(X 2 − x 2 )
kX 2 = kx 2 + mv 2 +⇒ v2 = .
2 2 2 m
When the spring is unstretched, x = 0, in which case
;
k
v = X.
m
This is the speed of m each time it passes through x = 0.
EXERCISES 7.5
1. If the stretch of a spring is doubled, by what factor does the potential ∗ 3. When the crossbow of Exercise 19 in Section 7.4 is fired, the poten-
energy in the spring change? tial energy stored in the draw is transformed into kinetic energy of the
∗ 2. (a) Mass m in the figure below is pulled a distance x0 to the arrow. If the mass of the arrow is 20 g, what is the speed of the arrow
right of its equilibrium position. It is then given speed v0 as it leaves the crossbow?
to the right. Find an equation relating speed of the mass ∗ 4. The force of repulsion between point charges is defined in Exercise
and stretch (or compression) in the spring thereafter. 15 of Section 7.4. If Q is placed at the origin, and q is moved along
the x -axis from x = r to x = r/2, what is the gain in electrostatic
energy?
k ∗ 5. (a) A 50-m chain with mass 100 kg hangs from the top of a
m building 50 m high so that its lower end just touches the
ground. What gravitational potential energy is stored in the
(b) Use the equation in part (a) to find the maximum stretch the chain? (Take ground level as zero potential.)
spring experiences.
(c) Use the equation in part (a) to find the maximum speed (b) Use energy considerations to determine the work to lift the
experienced by m . chain to the top of the building.
∗ 6. (a) If the mass m in the figure below is slowly allowed to com- is µmg , where 0 < µ < 1 is a constant and g = 9.81.
press the spring, what is the ultimate compression? Use energy considerations to show that when the mass is at
(b) If the mass is dropped, and strikes the spring with speed position x , its speed is given by the equation
v0 , what is the maximum compression experienced by the
spring? kx02 = kx 2 + mv 2 + 2µmg(x0 − x).
(b) Use the equation in part (a) to determine where the mass
m comes to an instantaneous stop for the first time.
(c) What is the limit of the expression in part (b) as µ → 0?
Is this to be expected?
(d) What is the limit of the expression in part (b) as µ → 1?
Is this to be expected?
k
∗ 10. Two springs with constants k1 and k2 are joined together, and then
one end is fastened to a wall (figure below). It is shown in physics
that when a horizontal force is applied to the free end, the ratio of the
stretches s1 and s2 in the two springs is inversely proportional to the
ratio of their spring constants:
∗ 7. A particle is moved along the x -axis by a force F (x) in the x -
direction. Use Newton’s second law to show that if F (x) is the total s1 k2
= .
force on the particle, then the work done by F (x) is equal to the change s2 k1
in kinetic energy of the particle.
Show that if the free end is moved so as to produce a total stretch in the
∗ 8. (a) Newton’s universal law of gravitation is stated in Exercise
springs of L , the potential energy stored in the springs is
22 of Section 7.4. Suppose M represents the mass of the
earth, which we regard as a sphere with radius 6370 km and k1 k2 L2
mean density 5.52 × 103 kg/m 3 . Calculate the gravitational .
potential energy gained by a 10-kg mass lifted from the 2(k1 + k2 )
earth’s surface to a height of 100 km. Treat the earth as a
point mass concentrated at its centre.
(b) If the mass is dropped from this height, with what speed k1 k2
does it strike the earth?
∗ 9. (a) Mass m in the figure for Exercise 2 is pulled a distance x0
to the right of its equilibrium position and then released.
As it moves to the left, friction between m and the surface
on which it slides retards the motion. The force of friction
7.6 Fluid Pressure

When an object is immersed in a fluid, it is acted on by fluid forces. These forces are independent
of the object, and are therefore a property of the fluid itself. They always act perpendicularly
FIGURE 7.47 Determina-
to the surface of the submerged object. We use the concept of pressure to describe these fluid
tion of pressure in a fluid
forces. We define pressure at a point in a fluid as the magnitude of the force per unit area that
y would act on a surface at that point in the fluid. Because pressure is the magnitude of the fluid
Surface of fluid force at a point, it is therefore a positive quantity. Experience suggests that pressure depends on
x two factors: depth below the surface of the fluid and the type of fluid itself. In order to discover
the precise dependence, we consider a small horizontal disc of the fluid (Figure 7.47).
y Suppose we denote by P (y) the functional dependence of pressure P on depth y . Then
Area A
pressure at the bottom of the disc is P (y) and pressure at the top of the disc is P (y + !y) . If
the fluid is stationary, then the sum of all vertical forces on the disc must be zero. There are three
vertical forces acting on the disc: fluid forces on its top and bottom faces, and gravity. Since
!y pressure P (y) , which is force per unit area, is the same at all points on the bottom of the disc,
it follows that the force on the bottom of the disc must be AP (y) , that is, the product of the
area A of the bottom of the disc and pressure at points on the bottom of the disc. Similarly, the
fluid force on the top of the disc is −AP (y + !y) ; it is negative because it is in the negative
y -direction. Finally, if ρ is the density of the fluid (mass per unit volume), the force of gravity
on the disc is −9.81ρ(A !y) . Since the sum of these three forces must be zero, we set
AP (y) − AP (y + !y) − 9.81ρ(A !y) = 0.
Rearrangement of this equation yields
P (y + !y) − P (y)
= −9.81ρ,
!y
and if we take limits of both sides as !y → 0, we obtain
dP
= −9.81ρ.
dy
This differential equation for P (y) is immediately integrable:
P (y) = −9.81ρy + C.
Since fluid pressure at the surface of the fluid is equal to zero [ P (0) = 0], C must be equal to
zero; hence,
P = −9.81ρy. (7.25)
Since −y is a measure of depth d below the surface of the fluid, we have shown that
P = 9.81ρd, (7.26)
where d is always taken as positive.

An illuminating interpretation of formula 7.26 is suggested by Figure 7.48. Above a point
at depth d below the surface of a fluid, we consider a column of fluid of unit cross-sectional
area. The weight of this column of fluid is its volume multiplied by 9.81ρ :
W = 9.81ρV .
But V is the product of the length of the column, d , and the (unit) cross-sectional area; that is,
V = (1)d = d , and hence
W = 9.81ρd. (7.27)
A comparison of equations 7.26 and 7.27 suggests that pressure at any point is precisely the
weight of a column of fluid of unit cross-sectional area above that point.
We now consider the problem of determining total force on one side of a flat plate
(Figure 7.49), immersed vertically in a fluid of density ρ . If the area of the plate is subdi-
vided into horizontal rectangles of width dy , then pressure at each point of this rectangle is
FIGURE 7.48 Pressure is FIGURE 7.49 Force on

weight of a column of fluid flat surface submerged in a fluid
Surface of fluid y
Surface of fluid
x = f ( y) x
b
y
dy
x1
d
x = g ( y)
a
x2
approximately P = −9.81ρy . This is an approximation because slight variations in pressure do

occur over vertical displacements within the rectangle. It follows that force on the representative
rectangle is approximately equal to its area multiplied by −9.81ρy ,
−9.81ρy(x2 − x1 ) dy = −9.81ρy [f (y) − g(y)] dy.
Total force on the plate is found by adding forces on all such rectangles and taking the limit as
their widths approach zero. We obtain the required force, therefore, as
! b
F = −9.81ρy [f (y) − g(y)] dy. (7.28)
a
Once again we do not suggest that equation 7.28 be memorized because it is associated with
the choice of coordinates in Figure 7.49. For a different coordinate system, the definite integral
would be correspondingly different (see Example 7.21). What is important is the procedure:
Subdivide the surface of the plate into horizontal rectangles, find the force on a representative
rectangle, and finally, add over all rectangles with a definite integral.
Note that vertical rectangles cannot be used without further discussion since it is not evident
how to calculate the force on such a rectangle. Consideration of vertical rectangles is given in
Exercise 8.
EXAMPLE 7.21
The vertical face of a dam is parabolic with breadth 100 m and height 50 m. Find the total force
due to fluid pressure on the face.
SOLUTION If we set up the coordinate system in Figure 7.50, we see that the edge of the
dam has an equation of the form y = kx 2 . Since (50, 50) is a point on this curve, it follows
that k = 1/50, and y = x 2 /50. Area of the representative rectangle is 2x dy , and it is at
depth 50 − y below the surface of the water. Since density of water is 1000 kg/m 3 , force on
the representative rectangle is approximately equal to
3
(9.81)(1000)(50 − y)2x dy = 19 620(50 − y) 50y dy.
Total force on the dam must therefore be

! 50 √ √ ! 50
1/2
F = 19 620(50 − y)5 2y dy = 98 100 2 (50y 1/2 − y 3/2 ) dy
0 0
) *50
√ 100y 3/2 2y 5/2
= 98 100 2 − = 6.54 × 108 N.
3 5 0
FIGURE 7.50 Fluid force on a dam
(50, 50)
x
dy
y y = kx 2
x
EXAMPLE 7.22
A tank in the form of a right circular cylinder of radius 2 m and length 10 m lies on its side. If it
is half-filled with oil of density ρ kilograms per cubic metre, find the force on each end of the
FIGURE 7.51 Fluid force
on end of half-filled cylindrical
tank.
tank
y SOLUTION Force on the representative rectangle in Figure 7.51 is approximately equal to
the pressure 9.81ρ(−y) multiplied by area 2x dy of the rectangle
x2 + y2 = 4
3
9.81ρ(−y)2x dy = −19.62ρy 4 − y 2 dy.
dy y x
Total force on each end of the tank is therefore
x
! 0 3 ) *0
1
F = −19.62ρy 4 − y 2 dy = 19.62ρ (4 − y 2 )3/2 = 52.32ρ N.
−2 3 −2
Marine engineers are building buoys made of wood and concrete. Both are cylinders with
diameter 25 cm. The length of the concrete cylinder is 90 cm. Densities of the wood and
concrete cylinders are 580 kg/m 3 and 2900 kg/m 3 , respectively. They have asked us to
determine the length of the wood part of the buoy if there is to be 1 m above water.
SOLUTION To solve this problem, we need Archimedes’ principle. It is discussed in
Exercise 18 and states that the buoyant force on an object when immersed or partially
immersed in a fluid is equal to the weight of the fluid displaced by the object. Suppose we
denote the length of the wood cylinder by L (Figure 7.52). For the buoy to float in this
position, the magnitude of the force of gravity on the buoy must be equal to the buoyant
force. The force of gravity on the buoy is
' + ,2 ( ' + ,2 + ,(
25 25 9
580g π (L) + 2900g π .
200 200 10
The buoyant force is equal to the weight of the water displaced by the buoy
FIGURE 7.52 Determin- ' + (
ing the length of buoy
,2
25
1000g π (L − 1 + 0.9) .
200
Wood 100 cm
Surface of
L
water When we equate these two expressions and cancel πg(25/200)2 from each term, we
25 cm obtain
+ , + ,
9 1
580L + 2900 = 1000 L − +⇒ L = 6.45.
Concrete 90 cm 10 10
The wood cylinder should be 6.45 m long.

Figure 7.53a shows an inclined-tube reservoir manometer, used to measure pressure vari-
ations on the surface of the gauge liquid in the reservoir. The reservoir is a cylinder with
diameter D and the tube is a cylinder with diameter d . When pressures on liquid surfaces
are the same, the surfaces are at the same level as shown. When extra pressure is exerted
on surface A , the surface falls an amount H (Figure 7.53b), and liquid is pushed into
the tube causing surface B to rise an amount h (both relative to their positions in Figure
7.53a). Our problem is to find an expression for the extra pressure in terms of length L
and to determine design parameters that can be modified to make the manometer more
effective.
FIGURE 7.53a Analysis

of parameters in an inclined-tube
reservoir manometer
D d
Surface A
Surface B
Gauge liquid
FIGURE 7.53b
L Surface B
Surface A h
k
H θ
Q
SOLUTION Suppose we let the extra pressure on surface A be denoted by !P . To

find a formula for !P in terms of h and H , we consider pressure at point Q . From the
point of view of the reservoir, the pressure at Q is !P + ρgk , where k is the depth of
liquid in the reservoir and ρ is its density. From the point of view of the tube, the pressure
at Q is ρg(k + H + h) . Since these must be the same
!P + ρgk = ρg(k + H + h) +⇒ !P = ρg(H + h).
Because the volume of liquid in the manometer has not changed, we can say that
+ ,2 + ,2 + ,2
D d d
π H =π L +⇒ H = L.
2 2 D
In addition, h = L sin θ , so that

'+ ,2 ( '+ ,2 (
d d
!P = ρg L + L sin θ = ρgL + sin θ .
D D
The effectiveness of the manometer is in using L to predict !P . Most desirable is for

L to be large for small changes in pressure. The larger L , the less likely errors in its
measurement will affect the calculation of !P . When we write
!P
L= '+ ,2 (,
d
ρg + sin θ
D
we can see that L increases as ρ , d/D , and sin θ all decrease. In other words, the gauge
liquid should have small density, the tube diameter should be small relative to the reservoir
diameter, and angle θ should be as small as possible. This is theoretically speaking of
course.
Let us discuss each of these factors briefly. The manometer will be measuring the
pressure of a fluid (gas) on top of surface A . It follows that the liquid should be immiscible
with the fluid. It should also develop a reasonable meniscus so that length L can be
measured satisfactorily, and it should suffer minimal loss due to evaporation. Hydrocarbon
liquids turn out to be most suitable with lowest densities around 80% that of water. Use
of such liquids increases effectiveness of the manometer by 25% compared to water.
The ratio d/D should be minimized mathematically. There are limits to how small
d can be and how large D should be. Tube diameter, in practice, should exceed 6 mm in
order to avoid excessive capillary effects. The situation d = D and θ = π/2 corresponds
to a U-tube manometer. For given !P ,
+ ,
!P !P
L= = 0 .5 .
ρg(1 + 1) ρg
Consider the situation when d = 6 mm, D = 60 mm say, and θ = π/4. Then

+ ,
!P !P
L= √ = 1.39 .
ρg(1/100 + 1/ 2) ρg
Finally, angle θ should be minimized. A practical lower limit is π/18 radians; below this
angle, the meniscus becomes indistinct and it is difficult to get an accurate reading for L .
EXERCISES 7.6
1. A tropical fish tank has length 1 m, width 0.5 m, and depth 0.5 m. ∗ 3. The vertical end of a water trough is an isosceles triangle with width
Find the force due to water pressure on each of the sides and bottom 2 m and depth 1 m. Find the force of the water on each end when the
when the tank is full. trough is half full by volume.
∗ 4. A square plate, 2 m on each side, has one diagonal vertical. If it is

one-half submerged in water, what is the force due to water pressure
2. The vertical surface of a dam exposed to the water of a lake has the
on each side of the plate?
shape shown below. Find the force of the water on the face of the dam.
y ∗ 5. A cylindrical oil tank of radius r and height h has its axis vertical.
If the density of the oil is ρ , find the force on the bottom of the tank
300 m when it is full.
x4 ∗ 6. A rectangular swimming pool full of water is 25 m long and 10 m

y= wide. The depth is 3 m for the first 10 m at the deep end, decreasing
45 × 106
linearly to 1 m at the shallow end. Find the force due to the weight of
x
the water on each of the sides and ends of the pool.
∗ 7. The vertical face of a dam across a river has the shape of a parabola
36 m across the top and 9 m deep at the centre. What is the force that y Fluid
the river exerts on the dam if the water is 0.5 m from the top? surface
Semicircle 1m x
∗ 8. Show that the force due to fluid pressure on the vertical rectangle in
the figure below is
9.81ρ
F = h(y12 − y22 ), 3m 4m
2
where ρ is the density of the fluid.
y
Semicircle
Surface of fluid
y2 x
∗ 12. In Exercise 6, find the force due to water pressure on each part of
the bottom of the pool.
y1
h ∗ 13. The bow of a landing barge (figure below) consists of a rectangu-
lar flat plate A metres wide and B metres long. When the barge is
stationary, this plate makes an angle of π/6 radians with the surface of
the water. Find the maximum force of the water on the bow.
∗ 9. A flat plate in the shape of a trapezoid is submerged vertically in

a fluid with density ρ . The plate has two parallel vertical sides of
lengths 6 and 8 and a third side of length 5 that is perpendicular to the
parallel sides and at a depth of 1 below the surface of the fluid (figure B
below). Find the force due to fluid pressure on each side of the plate,
using both horizontal and vertical rectangles (see Exercise 8 for vertical
rectangles). A
Surface of fluid
1 ∗ 14. A water tank is to be built in the form of a rectangular box with all
5 six sides welded along their joins (figure below). Sides ABCD and
EF GH are to be square, and all sides except ABCD are supported
6
from the outside. If the maximum force that ABCD can withstand is
8
20 000 N, what is the largest cross-section that can be built, assuming
that at some stage the tank will be full?
E F
∗ 10. The base of a triangular plate, of length a , lies in the surface of a

fluid of density ρ . The third vertex of the triangle is at depth b below A
the surface (figure below). Show that the force due to fluid pressure on B
H
each side of the plate is 9.81ρab2 /6, no matter what the shape of the G
triangle.
a Fluid D C
surface
∗ 15. A cylindrical oil tank of radius r and length h has its axis hori-
b zontal. If the density of the oil is ρ , find the force on each end of the
tank when it is half full.
∗ 16. A right circular cylinder of radius r and height h is immersed in

a fluid of density ρ with its axis vertical. Show that the buoyant force
∗ 11. Set up (but do not evaluate) a definite integral(s) to find the force due on the cylinder due to the pressure of the fluid is equal to the weight
to water pressure on each side of the flat vertical plate in the following of the fluid displaced. This is known as Archimedes’ principle, and is
figure. valid for an object of any shape.
∗ 17. The lower half of a cubical tank 2 m on each side is occupied by

water, and the upper half by oil (density 0.90 g/cm 3 ). Vacuum
(a) What is the force on each side of the tank due to the pressure
of the water and the oil?
h
(b) If the oil and water are stirred to create a uniform mixture,
does the force on each side change? If not, explain why
not. If so, by how much does it increase or decrease?
∗ 18. Archimedes’ principle states that the buoyant force on an object A B
when immersed or partially immersed in a fluid is equal to the weight
of the fluid displaced by the object. Mercury
(a) Show that if an object floats partially submerged in water,
the percentage of the volume of the object above water is ∗ 23. (a) A block of wood (density 0.40 g/cm 3 ) is cubical (0.25 m
on each side). If it floats in water, how deep is its lowest
ρw − ρo point below the surface? Refer to Archimedes’ principle in
100 ,
ρw Exercise 18.
where ρo and ρw are the densities of the object and water, (b) Repeat part (a) given that the block is a sphere of radius
respectively. 0.25 m.
(b) If the densities of ice and water are 915 kg/m 3 and
1000 kg/m 3 , respectively, show that only 8.5% of the vol- ∗ 24. A tank is to be built in the form of a right circular cylinder with
ume of an iceberg is above water. horizontal axis. The ends are to be joined to the cylindrical side by a
continuous weld. One end of the tank and the cylindrical side are sup-
∗ 19. A wood pole with radius 10 cm and length 3 m is to be used as a ported from the outside. The remaining unsupported end can withstand
buoy. It has density 500 kg/m 3 . To make it float vertically a concrete a total force of 40 000 N less 1000 N for each metre of weld on that
cylinder with radius 10 cm and length 30 cm is attached to one end. If end. What is the maximum radius for the tank if it is to hold a fluid
the density of the concrete is 3000 kg/m 3 , how much of the buoy will with density 1.019 × 103 kg/m 3 ?
be above water?
∗ 20. What should be the length of the concrete attachment in Exercise ∗ 25. Find the ratio L/R such that the forces due to fluid pressure on
19 if exactly 1 m of buoy is to be above water? the rectangular and semicircular parts of the plate in the figure below
are equal.
∗ 21. A square log 20 cm by 20 cm and length 2 m has been floating
in water for a number of years. Water gradually permeated the log so Fluid surface
that the density of the log is no longer constant. It varies linearly with
depth beginning with the density of water at the edge deepest in water
to 500 kg/m 3 at its top edge. Determine the height of log protruding L
from the water.
∗ 22. If a full tube of mercury is inverted in a large container of mercury, R
the level of mercury in the tube will fall, but it will stabilize at a point
higher than that in the container (following figure). This is due to the
fact that air pressure acts on the surface of the mercury in the container
but not on the surface of the mercury in the tube. The extra column of
mercury, of height h , creates a force at A that counteracts the atmo- ∗ 26. (a) A spherical shell with inner radius 1 m and outer radius 2 m
spheric pressure transmitted through the mercury in the container to the
has density 2 kg/m 3 . It is cut in half by a plane through its
tube so that the total pressure at A is equal to the total pressure at B .
centre. The flat edge of one of the halves is placed carefully
(a) Show that if the density of mercury is 13.6 g/mL, then the on the surface of a large container of water. How far does
atmospheric pressure at the surface of the mercury in the the shell sink?
container is 1.33 × 105 h N/m 2 , provided that h is measured
in metres. (b) Repeat part (a) if a tiny hole is drilled through the shell at
(b) If h is measured as 761 mm, find the atmospheric pressure. its upper most point.
7.7 Centres of Mass and Centroids

Everyone is acquainted with the action of a teeter-totter or seesaw. Two children of unequal
masses can pass many hours rocking, provided that the child with greater mass sits closer to the
fulcrum. In this section we discuss the mathematics of the seesaw. This requires a definition
of moments of masses, and moments lead to the idea of the centre of mass of distributions of
masses, lumped or continuous.
To discuss the mathematics of a seesaw, we consider in Figure 7.54 a uniform seesaw of

length 2L balanced at its centre, with a child of mass m at one end. If a second child of equal
mass is placed at the other end, the ideal seesaw situation is created. If, however, the mass of
the second child is M > m , then this child must be placed somewhat closer to the fulcrum. To
find the exact position, we must determine what might be called the rocking power of a mass.
A little experimentation shows that when M = 2m , M must be placed halfway between the
end and the fulcrum; when M = 3m , M must be placed a distance L/3 from the fulcrum; and
in general, when M = am , M must be placed L/a from the fulcrum. Now the rocking power
of the child of mass m is constant, and for each mass M = am we have found an equal and
opposite rocking power if M is placed at L/a .
FIGURE 7.54 Moments of children on a seesaw
L L
M m
Clearly, then, rocking power depends on both mass and distance from the fulcrum. A
little thought shows that the mathematical quantity that remains constant for the various masses
M = am is the product of M and distance to the fulcrum; in each case, this product is
(am)(L/a) = mL , the same product as for the child of mass m . It would appear, then, that
rocking power should be defined as the product of mass and distance. We do this in the following
definition, and at the same time give rocking power a new name.
DEFINITION 7.1
The first moment of a point mass m about a point P is the product md , where d is the
directed distance from P to m .
If directed distances to the right of point P in Figure 7.55 are chosen as positive and
distances to the left are negative, then d1 is positive and d2 is negative. Mass m1 has a positive
first moment m1 d1 about P , and m2 has a negative first moment m2 d2 . In Figure 7.56, we have
placed five children of masses m1 , m2 , m3 , m4 , and m5 on the same seesaw. A sixth child of
mass m6 is to be placed somewhere on the seesaw so that all six children form the ideal seesaw.
FIGURE 7.55 Definition of moment of mass about a point on a line
d2 d1
m2 P m1
FIGURE 7.56 Determination of positions of children for ideal seesaw
L L
3L
x
4 L L
2 4
m1 m2 m3 m4 m6 m5
To find the appropriate position for the sixth child, we let x be the directed distance from
the fulcrum to the point where this child should be placed. The total first moment of all six
children about the fulcrum, choosing distances to the right as positive and to the left as negative,
is
m1 (−L) + m2 (−3L/4) + m3 (−L/2) + m4 (L/4) + m5 (L) + m6 (x).
We regard this as the resultant first moment of all six children attempting to turn the seesaw
— clockwise if the moment is positive, counterclockwise if the moment is negative. Balance
occurs if this resultant first moment is zero:
3 1 1
0 = −m1 L − m2 L − m3 L + m4 L + m5 L + m6 x.
4 2 4
We may solve this equation for the position of m6 :
L
x = (4m1 + 3m2 + 2m3 − m4 − 4m5 ).
4 m6
FIGURE 7.57 Placing the fulcrum for an ideal seesaw
x6
x5
x4
x3
x2
x1
m1 m2 m3 m4 m5 m6
x
x=0 x
We now turn this problem around and place the six children at distances from the left end
as shown in Figure 7.57. If the mass of the seesaw itself is neglected, where should the fulcrum
be placed in order to create the ideal seesaw? (See Exercise 16 for the case when the mass of
the seesaw is not neglected.) To solve this problem, we let the distance that the fulcrum should
be placed from the left end be represented by x . In order for balance to occur, the total first
moment of all six children about the fulcrum must vanish; hence,
0 = m1 (x1 − x) + m2 (x2 − x) + m3 (x3 − x) + m4 (x4 − x) + m5 (x5 − x) + m6 (x6 − x).
The solution of this equation is

6
m1 x1 + m2 x2 + m3 x3 + m4 x4 + m 5 x5 + m6 x6 1 "
x = = m i xi , (7.29)
FIGURE 7.58 Moment of m 1 + m2 + m3 + m4 + m5 + m6 M
a mass about a line i=1
y
>6
l where M = i=1 mi is the total mass of all six children. This point x at which the fulcrum
creates a balancing position is called the centre of mass for the six children. It is a point where
masses to the right are balanced by masses to the left.
d In the remainder of this section we extend the idea of a centre of mass of point masses along
a line (the seesaw) to the centre of mass of a distribution of point masses in a plane, and then
m to the centre of mass of a continuous distribution of mass. Our first step is to define the first
x moment of a point mass in a plane about a line in the plane (Figure 7.58).
DEFINITION 7.2
The first moment of a mass m about a line l is md , where d is the directed distance from
l to m .
Once again directed distances are used in calculating first moments, and therefore distances
on one side of the line must be chosen as positive and distances on the other side as negative.
For vertical and horizontal lines, there is a natural convention for doing this. Distances to the
right of a vertical line are chosen as positive, and distances to the left are negative. Distances
upward from a horizontal line are positive, and distances downward are negative. In particular,
when a mass m is located at position (x, y) in the xy -plane, its first moments about the x - and
y -axes are my and mx , respectively.
First moments of a system of n point masses m1 , m2 , . . . , mn located at points (x1 , y1 ) ,
(x2 , y2 ), . . . , (xn , yn ) , respectively (Figure 7.59a) about the x - and y -axes are defined as the
sums of the first moments of the individual masses about these lines:
n
"
first moment of system about x -axis = m i yi , (7.30a)
i=1
n
"
first moment of system about y -axis = m i xi . (7.30b)
i=1
FIGURE 7.59a System of n point FIGURE 7.59b Centre of mass of a

masses embedded in a plate system of n point masses
y y
m1
m1(x1, y1)
m2
m2(x2, y2) mn
mn(xn , yn) m3 y=y
m3(x3, y3)
(x, y)
Plate
mt x
mt(xt , yt) x x=x
What is the physical meaning of these first moments? Do they, for instance, play the same
role that first moments did for the seesaw? To see this we imagine that each point mass is
embedded in a thin plastic plate in the xy -plane. The plate itself is massless and extends to
include all n masses mi . Picture now that the plate is horizontal, and a sharp edge is placed
along the y -axis. Does the plate rotate about this edge, or does it balance? It is clear that the
plate balances if the first moment of the system about the y -axis is equal to zero, and rotates
otherwise, the direction depending on whether the first moment is positive or negative. Similarly,
the plate balances on a sharp edge placed along the x -axis if the first moment of the system about
the x -axis vanishes. In general, the plate balances along any straight edge if the first moment
of the system about that edge vanishes.
We defined the centre of mass for a distribution of children on the seesaw as the point at
which to place the fulcrum in order to obtain balance. Analogously, we define the centre of
mass (x, y) of the distribution of point masses in Figure 7.59a as the position to place a sharp
point in order to obtain balance. Remember that the plastic itself is massless and only the point
masses mi can create moments. Balance occurs at a point (x, y) if balance occurs about every
straight line through (x, y) . In particular, balance must occur about the lines x = x and y = y
parallel to the y - and x -axes as in Figure 7.59b. Since balance occurs about x = x if the total
first moment of the system about this line vanishes, we obtain the condition that
0 = m1 (x1 − x) + m2 (x2 − x) + · · · + mn (xn − x),
which can be solved for x ,

n
1 "
x = m i xi , (7.31)
M
i=1
>n
where M = i=1 mi is the total mass of the system. Note that this equation is identical to
7.29. Similarly, for balance about y = y , we find that y must be
n
1 "
y = m i yi . (7.32)
M
i=1
We have obtained a unique point (x, y) based on conditions of balance about the lines x = x
and y = y . Does this necessarily imply that balance occurs about every straight line through
(x, y) ? The answer is yes (see Exercise 39).
Every planar point mass distribution has a centre of mass (x, y) defined by equations 7.31
and 7.32. Our derivation has shown that the first moment of the system about any line through
(x, y) must be equal to zero. This point is significant in another way. If a particle of mass M
(the total mass of the system) is located at the centre of mass (x, y) , its first moment about the
y -axis is Mx . But from equation 7.31, we have
n
"
Mx = m i xi , (7.33)
i=1
and we conclude that the first moment of this fictitious particle M about the y -axis is exactly
the same as the first moment of the system about the y -axis. Similarly, the first moment of M
about the x -axis is My , and from 7.32,
n
"
My = m i yi . (7.34)
i=1
>n of mass of a system of point masses mi is a point at which a single particle of

Thus, the centre
mass M = i=1 mi has the same first moments about the x - and y -axes as the system. It can
be shown further (Exercise 39) that the first moment of M about any line is the same as the first
moment of the system about that line.
In summary, we defined the centre of mass of a system of point masses as a balance point.
We found as a result that the centre of mass is a point at which a single particle of mass equal to
the total mass of the system has the same first moment about any line as the system itself. This
argument is reversible. Were we to define the centre of mass as a point to place the mass of the
system for equivalent first moments, it would be a balance point. In other words, we have two
equivalent definitions of the centre of mass of a system of point masses — a balance point or an
equivalent point for first moments.
We now make the transition from a discrete system of particles to a continuous distribution
FIGURE 7.60 Centre of
mass of a thin plate
of mass in the form of a thin plate of constant mass per unit area ρ (Figure 7.60). In order to
find the mass of the plate, we proceed in exactly the same way that we did for areas. We divide
y the plate into vertical rectangles, the mass in a representative rectangle of width dx at position
y = f (x)
x being
ρ [f (x) − g(x)] dx.
dx To find the total mass of the plate, we add over all such rectangles of ever-diminishing widths:
y2 ! b
M = ρ [f (x) − g(x)] dx. (7.35)
a
y1 y = g (x)
Based on our discussion for systems of point masses, we define the centre of mass of a
a x b x
continuous distribution of mass as that point (x, y) where a particle of mass M has the same
first moments about the x - and y -axes as the distribution. In algebraic terms, we note that
Mx is the first moment about the y -axis of a particle of mass M at (x, y) . To this we must
equate the first moment of the original distribution about the y -axis. Now each point in the
representative rectangle in Figure 7.60 is approximately the same distance x from the y -axis
— approximately, because the rectangle does have finite, though very small, width. The first
moment, then, of this rectangle about the y -axis is approximately
xρ [f (x) − g(x)] dx.
We find the first moment of the entire plate about the y -axis by adding first moments of all such
rectangles and taking the limit as their widths approach zero. But once again this is the process
defined by the definite integral, and we obtain, therefore, for the first moment of the plate about
the y -axis,
! b
xρ [f (x) − g(x)] dx.
a
Consequently,
! b
Mx = xρ [f (x) − g(x)] dx, (7.36)
a
and this equation can be solved for x once M and the integral on the right have been evalu-
ated. Equation 7.36 represents for continuous distributions what equation 7.33 does for discrete
distributions.
To find y we must equate the product My to the first moment of the plate about the x -axis.
If we consider the representative rectangle in Figure 7.60 we see that not all points therein are
the same distance from the x -axis. To circumvent this problem, we consider all of the mass of
the rectangle to be concentrated at its centre of mass. Since the centre of mass is the midpoint
of the rectangle — a point distant [f (x) + g(x)]/2 from the x -axis — it follows that the first
moment of this rectangle about the x -axis is
1
[f (x) + g(x)]ρ [f (x) − g(x)] dx.
2
The total first moment of the plate about the x -axis is the definite integral of this expression,
and we set
! b
1
My = [f (x) + g(x)]ρ [f (x) − g(x)] dx. (7.37)
a 2
This equation is used to evaluate y .
Equations 7.36 and 7.37 can be memorized as formulas for x and y , but it is easier to
perform the foregoing operations mentally and arrive at these equations. Besides, for various
shapes of plates, we might use horizontal rectangles or combinations of horizontal and vertical
rectangles, and in such cases 7.36 and 7.37 would have to be modified.
On the basis of definitions 7.36 and 7.37 for (x, y) , we can show that the first moment of
M at (x, y) about any line is the same as the first moment of the plate about that same line.
This implies that the plate balances along any line through (x, y) and therefore at (x, y) .
EXAMPLE 7.23
Find the centre of mass of a thin plate of constant mass per unit area ρ if its edges are defined
by the curves
FIGURE 7.61 Centre of y = 2 − x2, y = 0, x = 0, x ≥ 0.
mass of plate bounded by y =
2 − x 2 , y = 0, and x = 0 SOLUTION Using vertical rectangles (Figure 7.61), we find that
y ! √ ! √ ) *√2 √
2 2
2 x3 4 2ρ
2 M = ρy dx = ρ (2 − x ) dx = ρ 2x − = .
0 0 3 0 3
y = 2 − x2 If (x, y) is the centre of mass of the plate, then Mx is the first moment of the single particle
y of mass M about the y -axis. This must be equated to the first moment of the plate about the
y -axis. Since xρ(2 − x 2 ) dx is approximately the first moment of the rectangle in Figure 7.61
about the y -axis, the following integral gives the first moment of the plate about the y -axis:
x 2 x
! √
2 ! √
2 ) *√2
dx 2 3 2 x4
xρ(2 − x ) dx = ρ (2x − x ) dx = ρ x − = ρ.
0 0 4 0
Hence we set Mx = ρ , and solve this equation for x :

ρ 3 3
x = =ρ √ = √ .
M 4 2ρ 4 2
To find y , we calculate the first moment of the plate about the x -axis. Since the centre of mass
of the rectangle in Figure 7.61 is y/2 units above the x -axis, it follows that the first moment of
this rectangle about the x -axis is (y/2)ρy dx . When we integrate this to find the first moment
of the plate about the x -axis, and equate it to My , the result is
! √ ! √ ! √
2 2 2
y ρ 2 2 ρ
My = ρy dx = (2 − x ) dx = (4 − 4x 2 + x 4 ) dx
0 2 2 0 2 0
) *√2 √
ρ 4x 3 x5 16 2ρ
= 4x − + = .
2 3 5 0 15
Thus, √
16 2ρ 3 4
y = √ = .
15 4 2ρ 5
EXAMPLE 7.24
Find the centre of mass of a thin plate of constant mass per unit area ρ if its edges are defined
by the curves
y = 2x − x 2 , y = x 2 − 4.
FIGURE 7.62 Centre of SOLUTION Using vertical rectangles (Figure 7.62) yields
mass of plate bounded by y = ! 2 ! 2
2x − x 2 and y = x 2 − 4 M = ρ(y2 − y1 ) dx = ρ [(2x − x 2 ) − (x 2 − 4)] dx
y −1 −1
! 2 ) *2
y = 2x − x2 2x 3
=ρ (4 + 2x − 2x 2 ) dx = ρ 4x + x 2 − = 9ρ.
y2 2 −1 3 −1
x If (x, y) is the centre of mass of the plate, then first moments of the plate and M about the
x
y -axis give
y1 ! 2 ! 2
Mx = xρ(y2 − y1 ) dx = ρ (4x + 2x 2 − 2x 3 ) dx
y = x2 − 4 −1 −1
(−1, −3)
) *
4 2
2x 3 x 9ρ
−4 = ρ 2x 2 + − = .
dx 3 2 −1 2
Thus,
9ρ 1 1
x = = .
2 9ρ 2
To find y , we use moments about the x -axis to write
! 2 ! 2
1 ρ
My = (y1 + y2 )ρ(y2 − y1 ) dx = (y22 − y12 ) dx
−1 2 2 −1
! 2 ! 2
ρ
= [(2x − x 2 )2 − (x 2 − 4)2 ] dx = 2ρ (−x 3 + 3x 2 − 4) dx
2 −1 −1
) *2
x4 27ρ
= 2ρ − + x 3 − 4x =− .
4 −1 2
Consequently,
27ρ 1 3
y =− =− .
2 9ρ 2
Looking at Figure 7.62, it would appear to balance at the point (1/2, −3/2).
EXAMPLE 7.25
Find first moments of a thin plate of constant mass per unit area ρ about the lines (a) y = 0,
(b) y = −2, and (c) x = −2, if its edges are defined by the curves
x = |y|3 , x = 2 − y2.
FIGURE 7.63 First mo- SOLUTION

ments of plate about lines (a) Since the mass is distributed symmetrically about the x -axis (Figure 7.63), the first
moment about y = 0 is zero.
y x = 2 − y2
x = −2 (b) Since the centre of mass of the plate is on the x -axis, its first moment about the line
x = y3 y = −2 is
x1 (1, 1)
(2)(mass of plate) = 2(2)(mass of plate above the x -axis)
x2
! 1 ! 1
2 x
=4 ρ(x2 − x1 ) dy = 4ρ (2 − y 2 − y 3 ) dy
(1, −1) 0 0
) 3 4
*1
y = −2 x = −y 3 y y
= 4ρ 2 y − −
3 4 0
17ρ
= .
3
(c) The first moment of the plate about the line x = −2 is twice the first moment of
its upper half. Since the x -coordinate of the centre of mass of the representative
rectangle in Figure 7.63 is (x1 + x2 )/2, the distance from the line x = −2 to the
centre of mass of the rectangle is 2 + (x1 + x2 )/2. The first moment of the horizontal
rectangle about x = −2 is therefore
+ ,
x 1 + x2
2+ ρ(x2 − x1 ) dy,
2
and the first moment of the plate is
! 1+ , ! 1
x1 + x2
2 2+ ρ(x2 − x1 ) dy = ρ [4(x2 − x1 ) + (x22 − x12 )] dy
0 2 0
! 1 ! 1
= 4ρ (x2 − x1 ) dy + ρ [(2 − y 2 )2 − (y 3 )2 ] dy
0 0
! 1
17ρ
= +ρ (4 − 4y 2 + y 4 − y 6 ) dy
3 0
) *1
17ρ 4y 3 y5 y7
= + ρ 4y − + −
3 3 5 7 0
881ρ
= .
105
It has become apparent through our discussions and examples that in calculating the centre of
mass of a thin plate with constant mass per unit area ρ , ρ is really unnecessary. As a constant
it is taken out of each integration and cancels in the final division. The location of the centre of
mass depends only on the geometric shape of the plate, and for this reason we could replace all
references to mass by area. In particular, the mass of the plate M can be replaced by its area A ,
first moments (of mass) Mx and My can be replaced by first moments (of area) Ax and Ay ,
and equations 7.36 and 7.37 then take the form
! b
Ax = x [f (x) − g(x)] dx, (7.38)
a
! b
1
Ay = [f (x) + g(x)][f (x) − g(x)] dx. (7.39)
a 2
It is customary when using first moments of area to call (x, y) the centroid of the area rather
than the centre of mass of the plate, simply because all references to mass have been deleted.
We emphasize, however, that the statements in this paragraph apply only when mass per unit
area is constant.
Figure 7.64a shows an automatic valve consisting of a plate L metres wide and H metres
high that pivots about a horizontal axis through point A . Water creates pressure on parts
of the valve above and below A . If the force on that part of the valve below A is greater
than on that part above A , the valve remains closed. If the force is greater on that part
above A , the valve opens to release water from left to right. Design specifications require
the valve to open when the depth of water is D . Our problem is to locate the position of
A for this to happen.

Determination of pivot position in an automatic valve
Water Water
D Valve D Valve
A H A H
h dy
y y
y
SOLUTION First we should point out that it is not the forces of the water on the top and
bottom parts of the valve that determine whether the valve opens or closes; it is moments
of these forces about the pivotal axis through A . With this in mind let us consider a small
horizontal strip of width dy and length L on the face of the valve (Figure 7.64b). The
force due to water pressure on it is ρg(D − y)L dy when water depth is D . If the
required position of the pivotal point A is h metres above the bottom of the valve, then
the moment of this force about A is (h − y)ρg(D − y)L dy . Moments will be positive
for areas below A and negative for areas above A . The total moment on the valve is
! H ! H
(h − y)ρg(D − y)L dy = ρgL [hD − (h + D)y + y 2 ] dy
0 0
) *H
y2 y3
= ρgL hDy − (h + D) +
2 3 0
1 2
H2 H3
= ρgL hDH − (h + D) + .
2 3
The valve is on the verge of opening when this moment vanishes,

1 2
H2 H3
0 = ρgL hDH − (h + D) + .
2 3
The solution of this equation is h = (3DH − 2H 2 )/(6D − 3H ) .
EXERCISES 7.7
In Exercises 1–5 find the centre of mass of the thin plate with constant 8. y = x 3 , x = y 3
mass per unit area.
9. y = x, y = 2x, 2y = x + 3
1. 2.
y y 10. x = y 2 − 2y, x + y = 12
Semiparabolic
√
Parabolic a a 11. y = 2 − x, x + y = 2
12. x = 4y − 4y 2 , y = x − 3, y = 1, y = 0
h h
13. x 3 y = 8, y = 9 − x 3
x x
∗ 14. The edges of a thin plate with constant mass per unit area ρ are
3. 4. defined by the curves y = |x|1/2 , y = x + 2, and y = 2 − x . Find
y y its first moment about the line x = −5.
a ∗ 15. Show that the centroids of regions A and B in the figure below
have coordinates
+ , + ,
n+1 n+1
h xA = a, y A = b,
n+2 4n + 2
r
+ , + ,
n+1 n+1
x x xB = a, y B = b.
Parabolic spandrel Quarter-circle 2n + 4 2n + 1
5.
y
y c>0
n>1
r (a, b)
B
Semicircle r x
y = cx n A
In Exercises 6–13 find the centroid of the region bounded by the curves. x
6. y = x 2 − 1, y = −x 2 − 2x − 1
∗ 16. Find the centre of mass of the seesaw in Figure 7.57 if the mass of
√ the seesaw is not neglected. Assume that it has uniform mass per unit
7. y = |x|, y = 2 − x 2 length ρ and length 2L .
In Exercises 17–19 find the centroid of the region bounded by the ∗ 29.
curves. e/ y e
2 /2
√ d
∗ 17. x = y + 2, y = x, y = 0
f
∗ 18. y + x 2 = 0, x = y + 2, x + y + 2 = 0,
2 I-section
y = 2 (above y + x 2 = 0 ) f
√ c
∗ 19. y = 2 − x, 15y = x 2 − 4 2
∗ 20. Find the centre of mass of the thin plate in the figure below if it
has constant mass per unit area. b
2 a/ a/ x
2 2
1 ∗ 30.
3 y
1
f
Parabolic 2 Parabolic e
c
5 2 d/ Z-section
2
x
d/ c
2 2
b
In Exercises 21–25, a plate with constant mass per unit area ρ is
a
bounded by the given curves. Find its first moment about the suggested ∗ 31.
line. y
∗ 2
21. x + 2 = y , y = x about x + y = 1 d d
∗ 22. y = 2x , y = −2x , y = 4 − 2x 2 , y ≥ 0 about y = x Channel
∗ 23. y = x 2 − 2x , x + y = 12 about 3x + y = 1 b b
∗ 24. x = y 3 , x + y = 2, y = 0 about x + y + 1 = 0 c
∗ 25. x = y 2 − 2y , x = 2y − y 2 about x + 2y = 4 a/ a/ x
2 2
∗ 26. If a region A can be subdivided into n subregions Ai ( i =
1, . . . , n ) such that the centroid of each Ai is (x i , y i ) , show that
the centroid (x, y) of A is given by In Exercises 32–35 find coordinates of the centroid of the region accu-
rate to three decimal places.
n n
1 " 1 "
x = Ai x i , y = Ai y i . ∗ 32. y = x 3 + 3x 2 + 2x + 1, x = 0, y = 0
A A
i=1 i=1 ∗ 33. y = x 4 − 5x 2 + 5, y = 0 (above the x -axis)
√
∗ 34. x = y 3 − y 2 − 2y,√x = 2y + 1
In Exercises 27–31 use the technique suggested in Exercise 26 to find 3
∗ 35. y = x − x, y = x
the centroid of the region.
∗ 27. ∗ 36. Prove the following theorem of Pappus: If a plane region is re-
y volved about a coplanar axis not crossing the region, the volume gen-
L-section erated is equal to the product of the area of the region and the circum-
ference of the circle described by the centroid of the region.
c ∗ 37. Use the result of Exercise 36 to find the volume of the donut in
b Exercise 41 of Section 7.2.
∗ 38. Use the result of Exercise 36 to find the volume in Exercise 42 of
d Section 7.2.
x ∗ 39. (a) Show that the first moment of the system of point masses in
a
Figure 7.59b about any> line is the same as the first moment
∗ 28. of a point mass M =
n
y i=1 mi at (x, y) about that line.
d/ d/ Hint: Use formula 1.16 for the distance from a point to a
2 2
c line.
(b) Does it follow that the first moment of the system about any
b line through (x, y) is zero?
2 T-section
a
∗ 40. A thin flat plate of area A is immersed vertically in a fluid of
b
2 density ρ . Show that the total force due to fluid pressure on each side
of the plate is equal to the product of 9.81ρ , A , and the depth of the
x centroid of the plate below the surface of the fluid.
∗∗ 41. The dam in the figure below is 4 m high, 10 m wide, and a metres ∗∗ 42. The gate AB in the figure below is 525 mm wide and is held in
thick. It is made of concrete with density 2400 kg/m 3 . Determine the its closed position by a vertical cable and by a 525 mm hinge located
minimum value of a if the dam is not to overturn about point A when along its top edge B . For a depth d = 1.8 m of water, determine
d = 4 m. the minimum tension in the cable required to prevent the gate from
opening.
a
Dam Cable
450 mm
4m Water
d Water
1.8 m B
600 mm A
A B Gate
300 mm
(a) Assume that a seal exists at B , so that no water pressure is ∗∗ 43. Show that the centroid of a triangle with vertices (x1 , y1 ) , (x2 , y2 ) ,
present under the dam. and (x3 , y3 ) is
x1 + x2 + x3 y1 + y2 + y3
x = , y = .
(b) Assume that no seal exists at B , so that full hydrostatic 3 3
pressure is present under the dam from A to B .
7.8 Moments of Inertia

Newton’s second law F = ma is fundamental to the study of translational motion of bodies.
For rotational motion of bodies, its counterpart is τ = I α , where τ is torque, α is angular
acceleration, and I is the moment of inertia of the body. The kinetic energy of a body of mass m
moving with velocity v is mv 2 /2. The kinetic energy of a body rotating with angular velocity
ω is I ω2 /2. Thus, for rotational motion, there is a quantity called the moment of inertia of a
body that is analogous to mass in translational equations. In this section we define and calculate
moments of inertia.
To define moments of inertia of bodies, we begin with the moment of inertia of a point
mass.
DEFINITION 7.3
The moment of inertia or second moment of a point mass m about a line l (Figure 7.58)
is the product md 2 , where d is the directed distance from l to m .
In particular, if m is at position (x, y) in the xy -plane, its moments of inertia about the x -
and y -axes are my 2 and mx 2 , respectively. For a system of n particles of masses m1 , m2 , . . . ,
mn located at points (x1 , y1 ) , (x2 , y2 ) , . . . , (xn , yn ) as in Figure 7.59a, moments of inertia
of the system about the x - and y -axes are sums of the moments of inertia of the particles about
the x - and y -axes:
n
"
moment of inertia about x -axis = mi yi2 , (7.40a)
i=1
n
"
moment of inertia about y -axis = mi xi2 . (7.40b)
i=1
The transition from the discrete case to a continuous distribution in the form of a thin plate
with constant mass per unit area ρ is not always so simple as for first moments. First consider
the moment of inertia of the plate in Figure 7.60 in Section 7.7 about the y -axis. The mass of
the representative rectangle is
ρ [f (x) − g(x)] dx,
and each point of the rectangle is approximately the same distance x from the y -axis. The
moment of inertia, then, of this rectangle about the y -axis is approximately
x 2 ρ [f (x) − g(x)] dx.
The moment of inertia of the plate about the y -axis is found by adding moments of inertia of
all such rectangles and taking the limit as their widths approach zero. But again this process
defines a definite integral, and therefore the moment of inertia of the plate in Figure 7.60 about
the y -axis is
! b
I = x 2 ρ [f (x) − g(x)] dx. (7.41)
a
EXAMPLE 7.26
Find the moment of inertia about the y -axis of a thin plate with constant mass per unit area ρ
if its edges are defined by the curves
√
y = x 3, y = 2 − x, x = 0.
FIGURE 7.65 Moment of SOLUTION Since the moment of inertia of the vertical rectangle in Figure 7.65 is
inertia of plate about y -axis
/√ 0
y x 2 ρ(y2 − y1 ) dx = ρx 2 2 − x − x 3 dx,
y = √2 − x
√2 the moment of inertia of the plate is
(1, 1)
! 1 ! 1 ! 1
2
/√ 3
0 2
√
y2
y = x3 I = ρx 2−x−x dx = ρ x 2 − x dx − ρ x 5 dx.
0 0 0
y1
2 x
In the first integral we set u = 2 − x , in which case du = −dx , and
x
dx
! 1 ) *1 ! 2
x6 ρ
I =ρ (2 − u)2 u1/2 (−du) − ρ =ρ (4u1/2 − 4u3/2 + u5/2 ) du −
2 6 0 1 6
) *2 √
8u3/2 8u5/2 2u7/2 ρ 256 2 − 319
=ρ − + − = ρ.
3 5 7 1 6 210
EXAMPLE 7.27
Find the moment of inertia about the line y = −1 of a thin plate of constant mass per unit area
ρ if its edges are defined by the curves
x = y2, x = 2y.
FIGURE 7.66 Moment of inertia of plate about the line y = −1
x = y2
x2 (4, 2)
x1 x = 2y
y = −1
SOLUTION Since the directed distance from the line y = −1 to all points in the horizontal
rectangle in Figure 7.66 is approximately y + 1, the moment of inertia of the rectangle about the
line y = −1 is approximately (y + 1)2 ρ(x2 − x1 ) dy . It follows that the moment of inertia
of the plate is
! 2 ! 2
2
I = (y + 1) ρ(x2 − x1 ) dy = ρ (y + 1)2 (2y − y 2 ) dy
0 0
! 2 ) 5 *2
y 28ρ
=ρ (−y 4 + 3y 2 + 2y) dy = ρ − + y 3 + y 2 = .
0 5 0 5
In Examples 7.26 and 7.27, and in the discussion leading to equation 7.41, we chose rectangles
FIGURE 7.67 Moment of
inertia of rectangle about x -axis
that had lengths parallel to the line about which we required the moment of inertia. This is
not just coincidence; the use of perpendicular rectangles is more complicated. Consider, for
y instance, finding the moment of inertia about the x -axis of the plate in Example 7.26. To use
h the vertical rectangles in Figure 7.65, we first require the moment of inertia of such a rectangle
about the x -axis. The mass of the rectangle, as in Example 7.26, must be multiplied by the
dy square of the distance from the x -axis to the rectangle. Unfortunately, different points in the
y2 rectangle are at different distances. One suggestion might be to concentrate all of the mass of
the rectangle at its centre of mass and use the distance from the x -axis to the centre of mass.
y
y1 This is incorrect . The centre of mass is a point at which mass can be concentrated if we are
discussing first moments. We are discussing second moments. (In Exercise 12 we show that
x the moment of inertia of the rectangle in Figure 7.67 cannot be obtained by concentrating its
mass at its centre of mass.) What are we to do then? To use this type of rectangle, we must first
develop a formula for its moment of inertia. To obtain this formula, let us consider the moment
of inertia about the x -axis of the rectangle of width h and length y2 − y1 in Figure 7.67. If we
subdivide this rectangle into smaller rectangles of width dy , the moment of inertia of the tiny
rectangle about the x -axis is approximately
y 2 ρh dy.
The moment of inertia of the long, vertical rectangle can be obtained by adding over all the tiny
rectangles as their widths dy approach zero,
! y2 ! y2
ρh
y 2 ρh dy = ρh y 2 dy = (y23 − y13 ). (7.42)
y1 y1 3
We can use this formula to state that the moment of inertia about the x -axis of the vertical
rectangle in Figure 7.65 is
ρ ρ
(y23 − y13 ) dx =
[(2 − x)3/2 − x 9 ] dx.
3 3
The moment of inertia of the plate about the x -axis is therefore
! 1 ) *1 √
ρ 3/2 9 ρ 2 5/2 x 10 (16 2 − 5)ρ
I = [(2 − x) − x ] dx = − (2 − x) − = .
0 3 3 5 10 0 30
The alternative procedure for this problem is to use horizontal rectangles that are parallel
to the x -axis and obtain two definite integrals:
! ! √
1 2
I = y 2 ρy 1/3 dy + y 2 ρ(2 − y 2 ) dy.
0 1
In summary, we have two methods for determining moments of inertia of thin plates:
1. Choose rectangles parallel to the line about which the moment of inertia is required, in
which case only the basic idea of mass times distance squared is needed.
2. Choose rectangles perpendicular to the line about which the moment of inertia is required,
in which case formula 7.42, or a similar formula, is needed.
As for finding centres of mass, we could, in the special case of uniform mass distributions, drop
all references to mass and talk about second moments of area about a line.
EXAMPLE 7.28
Find the moment of inertia about the line x = 2 of a plate with mass per unit area ρ if its edges
FIGURE 7.68 Moment of
are defined by the curves
inertia of plate about line x = 2
y x = 2y − y 2 , x = y 2 − 2y.
2 SOLUTION We first divide the plate into horizontal rectangles of width dy , and then subdivide
x = y2 − 2y x = 2y − y2
this rectangle into smaller rectangles of width dx (Figure 7.68). Since the directed distance
dy from the line x = 2 to the tiny rectangle is x − 2, the moment of inertia of the tiny rectangle
dx about x = 2 is
x1 x2 x=2 (x − 2)2 ρ dy dx.
It follows that the moment of inertia of the long horizontal rectangle about x = 2 is
x ! x2 ) *x2
2 1 3 ρ
(x − 2) ρ dy dx = ρ dy (x − 2) = [(x2 − 2)3 − (x1 − 2)3 ] dy.
x1 3 x1 3
The moment of inertia of the entire plate is now
! 2
ρ
I = [(x2 − 2)3 − (x1 − 2)3 ] dy
0 3
! 2
ρ
= [(2y − y 2 − 2)3 − (y 2 − 2y − 2)3 ] dy
3 0
! 2
2ρ
= (24y − 12y 2 + 8y 3 − 12y 4 + 6y 5 − y 6 ) dy
3 0
) *2
2ρ 12y 5 y7
= 12y 2 − 4y 3 + 2y 4 − + y6 −
3 5 7 0
1184ρ
= .
105
EXERCISES 7.8
In Exercises 1–10 the curves define a thin plate with constant mass per ∗ 15.
unit area ρ . Find its moment of inertia about the line. y
1. y = x 2 , y = x 3 , about the y -axis a

d
2. y = x, y = 2x + 4, y = 0, about the x -axis
3. y = x 2 , 2y = x 2 + 4, about y = 0 c
x
d
4. y = x 2 − 4, y = 2x − x 2 , about x = −2 c Z-section
3 d
5. xy y 2 + 12 = 1, x = 0, y = 1, y = 1/2, about y = 0
a
6. y = |x|1/3 , y = 2 − |x|1/3 , about x = 0
7. x + y = 2, y − x = 2, y = 0, about x = −2
∗ 16.
2 2 y
8. x = 1 − y , x = y − 1, about x = −1 c
2
9. x = 1 − y , x = y − 1, 2
about y = 1 d
10. x = y 2 , x + y = 2, about y = 3 I-section
a
∗ 11. (a) If IAx and IBx represent the second moments of area about x
the x -axis of the regions in the figure of Exercise 15 in
Section 7.7, show that IBx = 3nIAx . d
(b) Show that if IAy and IBy represent the second moments of
area about the y -axis, then nIAy = 3IBy .
d
12. What is the product of the mass of the rectangle in Figure 7.67 and b
the square of the distance from the x -axis to the centre of mass of the
rectangle? Is it equal to the expression in equation 7.42?
∗ 17.
y
In Exercises 13–17 find second moments of area of the section about c
(a) the x -axis and (b) the y -axis. Channel
∗ 13. b
y d
x
c a
L-section
∗ 18. The radius of gyration r of a thin plate with constant mass per unit
area about a line is defined by I = Mr 2 , where M is the mass of
b the plate and I is its moment of inertia about that line. Find radii of
gyration about the x - and y -axes for the plate with edges y = 2x 3 ,
y + x 3 = 0, and 2y = x + 3. Explain the physical significance of r .
c
x
a ∗ 19. Show that the kinetic energy of a long-playing record is equal to
one-half the product of the moment of inertia of the record about a line
∗ 14. through its centre and perpendicular to its face, and the square of its
y
angular speed.
d
∗ 20. Prove the parallel axis theorem: The moment of inertia of a thin
c
plate (with constant mass per unit area) with respect to any coplanar
line is equal to the moment of inertia with respect to the parallel line
through the centre of mass plus the mass multiplied by the square of
e T-section the distance between the lines.
∗ 21. If a line x = x̃ is drawn through the region in Figure 7.60, what

x integral represents the second moment of area about this line? What
a should be the value of x̃ for the smallest possible second moment?
∗ 22. The polar moment of inertia of a point mass m at (x, y) is defined In Exercises 23–25 the curves define a plate with mass per unit area
as the product of m and the square of its distance from the origin, equal to 2. Find its moment of inertia about the line accurate to three
J0 = m(x 2 + y 2 ) . For the thin plate (with constant mass per unit area) decimal places.
in the figure below, let Ix and Iy be its moments of inertia about the
x - and y -axes. Show that J0 = Ix + Iy . ∗ 23. y = 1 − x 2 , x = y 2 , about y = 0
√
∗ 24. y = x 3 − x , y = x , about x = 0
y √
∗ 25. y = x 3 − x, y = x , about y = 0
∗∗ 26. Find the second moment of area of a rectangle about its diagonal.
7.9 Additional Applications

Volumes by Slicing
If we can represent the area of parallel cross-sections of a volume as a function of one variable,
we can use a definite integral to calculate the volume. In particular, when we use the disc or
washer method to determine the volume of a solid of revolution, parallel cross-sections are
circles. In the following example, parallel cross-sections are squares.
EXAMPLE 7.29
A uniformly tapered rod of length 2 m has square cross-sections. If the areas of its ends are
4 cm 2 and 16 cm 2 as in Figure 7.69a, what is the volume of the rod?
FIGURE 7.69a Tapered rod with square cross-sections FIGURE 7.69b Volume of tapered rod with square cross-
sections
Area = 16 cm2
A (x)
Area = 4 cm2
x=2
x
x=0 x
2 dx
SOLUTION If we define an x -coordinate perpendicular to the square cross-sections as in

Figure 7.69b, then the side lengths of the cross-sections at x = 0 and x = 2 are 0.02 m and
0.04 m, respectively. Since the rod is uniformly tapered, the side length of the cross-section at
x is + ,
0.04 − 0.02 2+x
0.02 + x = 0.02 + 0.01x = .
2 100
The area of the cross section at x is therefore
A(x) = 10−4 (2 + x)2 .
If we construct at x a slab of cross-sectional area A(x) and width dx , the volume of the slab is
A(x) dx = 10−4 (2 + x)2 dx.

To obtain the volume of the rod, we add volumes of all such slabs between x = 0 and x = 2,
and take the limit as their widths approach zero:
! 2 ! 2
V = A(x) dx = 10−4 (2 + x)2 dx
0 0
) *2
(2 + x)3 10−4 56
= 10−4 = (64 − 8) = × 10−4 m3 .
3 0 3 3
Area of a Surface of Revolution

If a curve in the xy -plane is rotated about the x - or y -axis (or a line parallel to the x - or y -axis)
in order to produce a surface, we can calculate the area of this surface using lengths along curves,
discussed in Section 7.3.
EXAMPLE 7.30
If that part of the parabola y = x 2 between x = 0 and x = 1 is rotated around the y -axis, find
the area of the surface of revolution traced out by the curve (Figure 7.70).
FIGURE 7.70 Area of surface of revolution
(1, 1)
y = x2
(x, y) dy
dx
x x
SOLUTION We approximate length along the parabola corresponding to a change dx in x

by the tangential straight-line length:
9 + ,2
3 dy 3 3
(dx)2 + (dy)2 = 1+ dx = 1 + (2x)2 dx = 1 + 4x 2 dx.
dx
If this straight-line segment is rotated around the y -axis, each point follows a circular path of ra-
dius approximately equal to x ; therefore, the area traced out by the line segment is approximately
equal to
43 5
(2π x) 1 + 4x 2 dx .
We find total surface area by adding all such areas, and taking the limit as widths dx approach
zero:
! ) *1 √
1 3 (1 + 4x 2 )3/2 (5 5 − 1)π
A= 2π x 1 + 4x 2 dx = 2π = .
0 12 0 6
EXAMPLE 7.31
Find the area of the surface of revolution traced out by rotating that part of the curve y = x 3
between x = 1 and x = 2 about the x -axis.
SOLUTION We approximate length along the cubic corresponding to a change dx in x by

the tangential straight-line length:
9 + ,2
3 dy 3 3
(dx)2 + (dy)2 = 1+ dx = 1 + (3x 2 )2 dx = 1 + 9x 4 dx.
dx
If this straight-line segment is rotated about the x -axis (Figure 7.71), each point follows a circular
path of radius approximately equal to y ; therefore, the area traced out by the line segment is
approximately equal to
43 5
(2πy) 1 + 9x 4 dx .
By adding over all such areas, we obtain the area of the surface,
! 2 3 ! 2 3
A= 2πy 1 + 9x 4 dx = 2π x 3 1 + 9x 4 dx
1 1
) *
3/2 2
(1 + 9x 4 ) (1453/2 − 103/2 )π
= 2π = .
54 1 27
FIGURE 7.71 Area of surface of revolution
y
(2, 8)
y = x3
(x, y)
(1, 1)
Rates of Flow
In Problem 3 of Section 6.2 we considered laminar blood flow in a circular vessel. Specifically,
velocity of blood through the cross section in Figure 7.72a is a function of radial distance r from
the centre of the vessel:
v(r) = c(R 2 − r 2 ), 0 ≤ r ≤ R,
where c > 0 is a constant and R is the radius of the vessel. If we construct, at radius r , a thin
ring of width dr as in Figure 7.72b, then the area of this ring is approximately (2π r) dr . Since
v does not vary greatly over this ring, the amount of blood flowing through the ring per unit
time is approximately v(r) multiplied by the area of the ring:
v(r)(2π r dr).
We can find the total flow through the blood vessel by adding flows through all rings and taking
the limit as widths dr of the rings approach zero:
! ! ) *R
R R
2 2 R2r 2 r4 π cR 4
F = v(r)2π r dr = 2π rc(R − r ) dr = 2π c − = .
0 0 2 4 0 2
FIGURE 7.72a Circular blood ves- FIGURE 7.72b Rate of blood flow
sel of radius R through a circular blood vessel
r R R
r
dr
Elongation of Rods
FIGURE 7.73 Hooke’s
Hooke’s law states that forces exerted by springs are proportional to stretch and compression.
law for force in stretched spring
In particular, suppose x measures distance to the right in Figure 7.73 and x = 0 corresponds to
F the right end of the spring when the spring is unstretched and uncompressed. Hooke’s law then
states that a force F applied to the spring as shown causes stretch
x
x=0 F
when spring x = ,
unstretched
k
where k is the spring constant.
Figure 7.74 shows a rod of natural length L . When a force F is applied to the right end of
the rod, the rod acts like a spring in that the rod stretches. It stretches only minutely even when
F is large; but it does stretch. It is shown in the area of strength of materials that the amount of
stretch is related to F by the equation
FL
x = , (7.43)
AE
where A is the (constant) cross-sectional area of the rod and E is Young’s modulus of elasticity,
a constant that depends on the material of the rod. In effect, AE/L plays the role of spring
constant k .
FIGURE 7.74 Stretch in rod when force is applied to one end
F
L
x
x=0
Now suppose the rod is turned vertically and hung from its top end (Figure 7.75). Due to
FIGURE 7.75 Stretch in
rod hung vertically
its weight, the rod stretches, and using 7.43, we can calculate how much. If we consider a small
length dy at position y , the force on each cross section in this element is approximately the
same, and equal to the weight of that part of the rod below it,
ρg(L − y)A,
y=0
where ρ is the density of the material in the rod. According to 7.43, the element dy stretches
by
y ρg(L − y)A dy ρg(L − y) dy
= .
AE E
L Total stretch in the rod is therefore
dy ! ) *L
L
ρg(L − y) ρg y2 ρgL2
dy = Ly − = .
0 E E 2 0 2E
This may be somewhat surprising in that stretch does not depend on the cross-sectional area
of the rod. This is explained by the fact that the weight of rod below any cross-section is
proportional to cross-sectional area A , but stretch in 7.43 is inversely proportional to A . The
two factors compensate. To get an idea of the magnitude of this stretch, suppose that L = 2 m,
ρ = 8000 kg/m 3 , and E = 2 × 1011 N/m 2 . Then
ρgL2 8000(9.81)(2)2
= = 7.85 × 10−7 m.
2E 2(2 × 1011 )
Suppose now that cross-sectional area A of the rod is not constant; the rod is tapered, say,
with circular cross-sections (Figure 7.76). Radius of the large end is r and the rod tapers to a
point at y = L . The force on cross-sections of the element of width dy at position y is again
the weight of rod below it, namely,
1
π x 2 (L − y)ρg,
3
FIGURE 7.76 Elongation of tapered rod hung vertically
r
r x
y
x
L dy
L y
where x is the radius of the rod at position y . Since the equation of the side of the rod is
−L 4 y5
y = (x − r) +⇒ x = r 1− ,
r L
the weight below y is
1 4 y 52 π r 2 ρg
πr2 1 − (L − y)ρg = (L − y)3 .
3 L 3L2
The element dy therefore stretches:
π r 2 ρg 3 dy r 2 ρg(L − y)3 dy ρg(L − y)

2
(L − y) 2
= 2 2 2
= dy.
3L πx E 3L Er (1 − y/L) 3E
This is one-third the stretch for the nontapered rod, and hence the total stretch of the tapered rod
is ρgL2 /(6E) .
Green’s Functions
Green’s functions are widely used to solve engineering problems, especially in the presence
of quantities represented by Dirac-delta functions (see Section 2.5). We illustrate with the
following problem for static deflections of a taut string of negligible mass, constant tension τ ,
and length L , with ends fixed at x = 0 and x = L on the x -axis:
d 2y
−τ = F (x), 0 < x < L, (7.44a)
dx 2
y(0) = y(L) = 0. (7.44b)
Quantity F (x) is the load per unit x -length on the string. To find y(x) we would integrate the
differential equation twice and use the end conditions to evaluate the constants of integration.
For instance, if F (x) = kx(x − L) , where k < 0 is a constant, integration gives
+ ,
k Lx 3 x4
y(x) = − + Cx + D.
τ 6 12
The end conditions require

+ ,
k L4 L4
0 = y(0) = D, 0 = y(L) = − + CL + D.
τ 6 12
These give D = 0 and C = −kL3 (12τ ) , and therefore

+ ,
k Lx 3 x4 kL3 x kx
y(x) = − − = (2Lx 2 − x 3 − L3 ).
τ 6 12 12τ 12τ
Green’s functions provide an alternative way to solve such problems. The Green’s function
for this problem is
1
G(x; X) = [x(L − X) h(X − x) + X(L − x) h(x − X)], (7.45)
Lτ
where h(x − X) is the Heaviside function (Section 2.5). Think of X as a parameter that can
take on any value between 0 and L . Given x and X , G(x; X) represents the deflection in the
string at position x if a unit force in the positive y -direction is applied at position X . Given
that the load on the string is F (x) , the deflection at any point x is given by the definite integral
! L
y(x) = G(x; X)F (X) dX. (7.46)
0
We reason as follows: If G(x; X) is the deflection at x due to a unit force at X , then

G(x; X)F (X) is the deflection at x due to load F (X) dX on that part of the string dX
and X . Integration from 0 to L gives the deflection at x due to the entire load on the string.
When F (x) = kx(x − L) , as above,
! L
1
y(x) = [x(L − X) h(X − x) + X(L − x) h(x − X)]kX(X − L) dX.
0 Lτ
It is always necessary to subdivide the integration into two parts, one from 0 to x , and the other
from x to L . Since h(X − x) = 0 when 0 < X < x and h(x − X) = 0 when x < X < L ,
we obtain
! x ! L
k k
y(x) = X(L − x)X(X − L) dX + x(L − X)X(X − L) dX
Lτ 0 Lτ x
) *x ) *L
k(L − x) X4 LX 3 kx 2LX 3 X4 L2 X 2
= − + − −
Lτ 4 3 0 Lτ 3 4 2 x
kx
= (2Lx 2 − x 3 − L3 ).
12τ
This is the same solution as was obtained by solving the differential equation and evaluating
constants. The differential equation and end conditions are built into the Green’s function. The
definite integral in 7.46 takes care of the loading F (x) .
If the string is subjected to a concentrated load of F newtons attached at x = L/3,
then F (x) = −F δ(x − L/3) , where δ(x − L/3) is the Dirac-delta function. It is a very
complicated process to solve the differential equation for such a load. Use of the Green’s
function is particularly simple. The deflection is again given by the integral in equation 7.46,
where we use the property in Exercise 34 of Section 7.10:
! L
1
y(x) = [x(L − X) h(X − x) + X(L − x) h(x − X)](−F )δ(X − L/3) dX
0 Lτ
1 + , + , + ,2
−F L L L L
= x L− h − x + (L − x) h x −
Lτ 3 3 3 3
1 + , + ,2
−F L L
= 2x h − x + (L − x) h x −
3τ 3 3
 2F x

− , 0 < x < L/3
= 3τ

 − F (L − x), L/3 < x < L.
3τ
The graph is shown in Figure 7.77 where we have filled in the removable discontinuity at
x = L/3.
FIGURE 7.77 Deflection of string due to a point load at x = L/3
L
_ L x
3
− 2FL
9τ
EXERCISES 7.9
1. Verify that the surface area of a sphere of radius r is 4π r 2 . ∗ 9. Find the area of the surface of revolution formed by rotating about
the x -axis that part of the curve 24xy = x 4 + 48 between x = 1 and
2. Find the area of the curved surface of a right circular cone of radius x = 2.
r and height h .
∗ 10. Find the area of the surface of revolution generated by rotating the
3. Calculate the rate of flow of blood through a circular
√ vessel of radius curve 8y 2 = x 2 (1 − x 2 ) about the x -axis.
R if the velocity profile is (a) v = f (r) = cR R 2 − r 2 and (b)
v = f (r) = (c/R 2 )(R 2 − r 2 )2 . ∗ 11. Find the area of the surface of revolution generated by rotating the
loop of the curve 9y 2 = x(3 − x)2 about (a) the y -axis and (b) the
4. Find the volume of the pyramid in the figure below. x -axis.
∗ 12. The base of a solid is the circle x 2 + y 2 = r 2 , and every plane sec-
tion perpendicular to the x -axis is an isosceles triangle (figure below).
Find the volume of the solid.
h r
b
b
r
r y
5. The amount of water consumed by a community varies throughout
x
the day, peaking, naturally, around meal hours. During the 6-h period
between 12:00 noon ( t = 0) and 6:00 p.m. ( t = 6), we find that the
∗ 13. The end of the rod at x = 0 in the figure below is rigidly fixed. If
number of cubic metres of water consumed per hour at time t is given
a force with magnitude F is applied to the right end, how long is the
by the function
rod?
f (t) = 5000 + 21.65t 2 − 249.7t 3 + 97.52t 4 − 9.680t 5 .
Find the total consumption during this 6-h period.
∗ 6. The number of bees per unit area at a distance x from a hive is given
by F
600 000 x
ρ(x) = (R 3 + 2R 2 x − Rx 2 − 2x 3 ), 0 ≤ x ≤ R, L
31π R 5
where R is the maximum distance travelled by the bees. (natural length)
(a) What is the number of bees in the colony?
(b) How many bees are within a distance R/2 of the hive? ∗ 14. In the figure below, a mass M is placed on a vertical rod. If the
length of the compressed rod is L , what is its length if M is removed
∗ 7. If the radius of the blood vessel in Figure 7.72 is reduced to R/2 and the rod is turned horizontally? Assume that M is so large that the
because of arteriosclerosis, the velocity profile is weight of the rod can be neglected in comparison.
2 2
v(r) = c(R − 4r ), 0 ≤ r ≤ R/2.
What percentage of the normal flow ( π cR 4 /2) gets through the hard- M
ened vessel?
∗ 8. A tree trunk of diameter 50 cm (figure below) has a wedge cut from
it by two planes. The lower plane is perpendicular to the axis of the L
trunk, and together the planes make an angle π/3 radians, meeting
along a diameter of the circular cross-section of the trunk. Find the
volume of the wedge.
∗ 15. What is the answer to Exercise 14 if the weight of the rod is taken
into account?
∗ 16. Suppose the rod in Exercise 13 is turned vertically so that its top
end is fixed, and the force F pulls vertically downward on the lower
end. How long is the rod if the weight of the rod is also taken into
3 account?
∗ 17. What happens when you attempt to find the stretch of the rod in
Figure 7.76 if the pointed end is at y = 0 and the larger end is at
50 y = L?
∗ 18. Suppose a mass M is attached to the lower end of the tapered rod where E and I are material constants, and F (x) is the loading. The
in Figure 7.76. It is so large that the mass of the rod is negligible by Green’s function for this problem is
comparison. What happens when you attempt to find the stretch in the
rod? 1 x3 Xx 2
G(x; X) = (x − X)3 h(x − X) − + .
∗ 19. A tapered rod of length L has square cross-sections. The squares 6EI 6EI 2EI
on the ends have dimensions a and b ( b > a ). The rod is placed in
a vertical position with the larger end below the smaller end, and the (a) Use formula 7.46 to find deflections of the beam when
smaller end fixed in position (figure below). What is the length of the F (x) = k , where k < 0 is a constant (perhaps the weight
rod in this position? of the beam itself).
(b) Where is y(x) a minimum?
a
L
x
L
∗ 27. (a) Repeat Exercise 26 if the loading F (x) = −F δ(x −L/2)

b is due to a concentrated load F at x = L/2, and the weight
∗ 20. What is the length of the rod in Exercise 19 if the rod is turned of the board is negligible in comparison. Use the property
upside down? in Exercise 34 of Section 7.10.
∗ 21. Repeat Exercise 20 if a mass M is distributed over the bottom of (b) Draw a graph of the diving board. Is any part of it straight?
the rod. ∗ 28. When both ends of a beam are clamped horizontally, deflections
∗ 22. Repeat Exercise 19 if a mass M is distributed over the bottom of must satisfy
the rod. d 4y
EI = F (x),
∗ 23. Electrons are fired from an electron gun at a target (figure below). dx 4
The probability that an electron strikes the target in a ring of unit area
at a distance x from the centre of the target is given by y(0) = y * (0) = 0 = y(L) = y * (L).
5 The Green’s function for this problem is

p = f (x) = (R 3 − x 3 ), 0 ≤ x ≤ R.
3π R 5 1
What percentage of a cascade of electrons hits within a distance r from G(x; X) = (x − X)3 h(x − X)
6EI
the centre of the target?
x3
+ (−L3 + 3LX2 − 2X3 )
Target 6EI L3
x2
+ (X3 − 2LX2 + L2 X).
x 2EI L2
Electron gun (a) Use formula 7.46 to find deflections of the beam when
F (x) = k , where k < 0 is a constant (perhaps the weight
R of the beam itself).
(b) Verify that y(x) has a minimum where it should be ex-
∗ 24. (a) Find the solution of formula 7.44 for deflections of the string pected.
if F (x) = k , where k < 0 is a constant.
∗ 29. Repeat Exercise 28 if the loading F (x) = −F δ(x − L/2) is due
(b) Draw a graph of the deflected string. Is it symmetric about
to a concentrated load F at x = L/2 and the weight of the beam is
x = L/2 with minimum at x = L/2? Is this to be
negligible in comparison. Use the property in Exercise 34 of Section
expected?
7.10.
∗ 25. Find the solution for formula 7.44 if the string is subjected to the ∗ 30. The vertical wall of a rectangular container filled with water has
constant loading of Exercise 24 and a concentrated load of F newtons at
a vertical rectangular slit with height h and width w . The upper edge
the centre of the string. Use the property in Exercise 34 of Section 7.10.
of the slit is H units √
below the surface of the water. Use the modified
∗ 26. Static deflections y(x) of a diving board of length L and fixed end Torricelli law v = c 2gh from Section 5.5 to show that the volume
x = 0 (figure below) must satisfy rate at which water runs out of the slit, assuming that the container is
4 kept full, is √
d y 2 2gcw
EI = F (x), 0 < x < L, [(H + h)3/2 − H 3/2 ].
dx 4 3
y(0) = y * (0) = 0 = y ** (L) = y *** (L),
∗ 31. Two right circular cylinders, each of radius r , have axes that inter- does it take for the water to evaporate completely? Note that Exercise
sect at right angles. Find the volume common to the two cylinders. 19 in Section 5.5 is this same problem given a container that is a right
∗ 32. What is the volume of air trapped in the attic of a house if the roof circular cone. Now we ask you to repeat the problem with no knowledge
has shape shown in the figure below? The peak of the roof is 2 m above of the shape of the container.
the base.
15 m
12 m
10
10 m ∗∗ 36. An appliance retailer must be concerned with her inventory costs.

For example, given that she sells N refrigerators per year, she must
decide whether to order the year’s supply at one time, in which case
∗ 33. A certain population is being started at time t = 0. Once born, the a large storage area would be necessary, or to make periodic orders
probability that an individual lives to an age t is p(t) . For time t > 0, throughout the year, perhaps of N/12 at the first of each month. In the
the birth rate is r(t) individuals per unit time. Show that at time T the latter case she would incur costs due to paperwork, delivery charges,
number of individuals in the population is and so on. Let us suppose that the retailer decides to order in equal-lot
sizes, x , at equally spaced intervals, N/x times per year.
! T
N (T ) = p(T − t)r(t) dt. (a) If each order has fixed costs of F dollars, plus f dollars for
0 each refrigerator, what are the total yearly ordering costs?
(b) Between successive deliveries, the retailer’s stock dwindles
∗ 34. (a) An orange is spherical with radius r . Suppose it is cut into
from x to 0. If she assumes that the number of refrigerators
slices of equal width. If the thickness of each slice is t , find
in stock decreases linearly in time, and the yearly stocking
the area of peel on each slice if thickness of peel is h .
cost per refrigerator is p dollars, what are the yearly stock-
(b) Find the volume of peel in each slice. ing costs?
∗∗ 35. The depth of water in the container in the figure below is originally (c) If the retailer’s total yearly inventory costs are her ordering
10 cm. Water evaporates from the surface at a rate proportional to the costs in part (a) plus her stocking costs in part (b), what
area of the surface. If the water level drops 1 cm in 5 days, how long value of x minimizes inventory costs?
7.10 Improper Integrals

According to the first fundamental theorem of integral calculus (Theorem 6.3), definite integrals
are evaluated with antiderivatives,
! b )! *b
f (x) dx = f (x) dx ,
a a
provided that f (x) is continuous on a ≤ x ≤ b . In this section we investigate what to do

when f (x) is not continuous on a ≤ x ≤ b , or when either a or b is infinite.
Consider whether a reasonable meaning can be given to the integrals
! ∞ ! ∞
1 1
dx and √ dx.
1 x2 1 x
Both integrals are improper in the sense that their upper limits are not finite, and we therefore
call them improper integrals. If b > 1, there is no difficulty with evaluation and interpretation
of
! b ! b
1 1
2
dx and √ dx.
1 x 1 x
Clearly,
! b ) *b
1 1 1
dx = − = 1− and
1 x2 x 1 b
! b √
1 - √ .b
√ dx = 2 x 1 = 2 b − 2.
1 x
We can interpret the first integral as the area under the curve y = 1/x 2 , above the x -axis, and
FIGURE 7.78 Area under
curve y = 1/x 2
between the vertical lines x = 1 and x = b (Figure √ 7.78). The second integral has exactly the
same interpretation but uses the curve y = 1/ x in place of y = 1/x 2 (Figure 7.79).
y In Table 7.1 we have listed values of these definite integrals corresponding to various values
of b . It is clear that the two integrals display completely different characteristics. As b is made
very large, the integral of 1/x 2 is always less than 1, but gets closer to 1 as b increases. In other
1 words,
y=
x2 ! b
1
lim 2
dx = 1.
b→∞ 1 x
x
Geometrically, the area in Figure 7.78 is always less than 1 but approaches 1 as b approaches
1 b
infinity. √
Contrast this with the integral of 1/ x , which becomes indefinitely large as b increases:
FIGURE 7.79 Area under ! b
√ 1
curve y = 1/ x lim √ dx = ∞;
b→∞ 1 x
y
that is, the area in Figure 7.79 can be made as large as desired by choosing b sufficiently large.
TABLE 7.1
1
y=
x ! b ! b
1 1
b dx √ dx
1 x2 1 x
1 b x 100 0.99 18
10 000 0.999 9 198
1 000 000 0.999 999 1 998
100 000 000 0.999 999 99 19 998
On the basis of these calculations, we would√like to say that the improper integral of 1/x 2 has
value 1, whereas the improper integral of 1/ x has no value. In other words, if these improper
integrals are to have values, they should be defined by the limits
! ∞ ! b ! ∞ ! b
1 1 1 1
dx = lim dx and √ dx = lim √ dx.
1 x2 b→∞ 1 x2 1 x b→∞ 1 x
In the second case the limit does not exist, and we interpret this to mean that the improper
integral does not exist or has no value.
In general, then, an improper integral with an infinite upper limit is defined in terms of
limits as follows.
DEFINITION 7.4
If f (x) is continuous for x ≥ a , we define
! ∞ ! b
f (x) dx = lim f (x) dx, (7.47)
a b→∞ a

If the limit exists, we say that the improper integral is equal to the limit, or that the improper
integral converges to the limit. If the limit does not exist, we say that the improper integral has
no value, or that it diverges.
EXAMPLE 7.32
Determine whether the following improper integrals converge or diverge:
! ∞ ! ∞ ! ∞
1 x2 2x 2 + 6
(a) dx (b) √ dx (c) dx
3 (x − 2)3 0 1 + x3 1 3x 2 + 5
SOLUTION
(a) Using equation 7.47 gives us
! ∞ ! b ) *b
1 1 −1
dx = lim dx = lim
3 (x − 2)3 b→∞ 3 (x − 2)3 b→∞ 2(x − 2)2 3
1 2
1 1 1
= lim − = ,
b→∞ 2 2(b − 2 )2 2
and the improper integral therefore converges to 1/2.

(b) Once again equation 7.47 gives
! ∞ ! ) 3 *b
x2 b
x2 2
√ dx = lim √ dx = lim 1 + x3
0 1 + x3 b→∞ 0 1 + x3 b→∞ 3 0
1 3 2
2 2
= lim 1 + b3 − = ∞,
b→∞ 3 3
and the improper integral therefore diverges.

FIGURE 7.80 Area under (c) Though we cannot at present find an antiderivative for f (x) = (2x 2 + 6)/(3x 2 + 5) ,
curve y = (2x 2 + 6)/(3x 2 + 5) we can solve this problem by interpreting the improper integral as area. The graph
of f (x) in Figure 7.80 indicates that the curve is asymptotic to the line y = 2/3.
y
Clearly, then, the improper integral
2x 2 + 6
y= !
3x 2 + 5 ∞
2x 2 + 6
2 dx
3 1 3x 2 + 5
1 x can have no value since the area under the curve is larger than the area of a rectangle
of width 2/3 and infinite length.
We can define improper integrals with infinite lower limits in exactly the same way as for
improper integrals with infinite upper limits. Specifically, if f (x) is continuous for x ≤ b , we
define
! b ! b
f (x) dx = lim f (x) dx, (7.48)
−∞ a→−∞ a
provided that the limit exists. For example,

! 5 ! 5 ) *5
1 1 −1
dx = lim dx = lim
−∞ (x − 6)5 a→−∞ a (x − 6)5 a→−∞ 4(x − 6)4 a
1 2
1 1 1
= lim − + =− .
a→−∞ 4 4(a − 6)4 4
When f (x) is continuous for all x , we define

! ∞ ! c ! b
f (x) dx = lim f (x) dx + lim f (x) dx, (7.49)
−∞ a→−∞ a b→∞ c
provided that both limits exist. The number c is arbitrary; existence of the limits is independent
of c (see Exercise 33). For example, if we choose c = 0, then
! ∞ ! 0 ! b
x x x
dx = lim dx + lim dx
−∞ (1 + x 2 )2 a→−∞ a (1 + x )2 2 b→∞ 0 (1 + x 2 )2
) *0 ) *b
−1 −1
= lim + lim
a→−∞ 2(1 + x 2 ) b→∞ 2(1 + x 2 ) 0
a
1 2 1 2
1 1 1 1
= lim − + lim −
a→−∞ 2(1 + a 2 ) 2 b→∞ 2 2 (1 + b 2 )
1 1
=− + = 0.
2 2
EXAMPLE 7.33
(a) Is it possible to paint the area bounded by the curves y = 1/x , x = 1, and y = 0?
(b) Rotate the area around the x -axis to form what is sometimes called Gabriel’s horn.
Does it have finite volume?
SOLUTION
FIGURE 7.81 Painting (a) Painting is possible if the area of the region bounded by the curves is finite (Fig-
area under y = 1/x ure 7.81). Since
y ! ∞ ! b
1 1
dx = lim dx = lim {ln |x|}b1 = lim ln b = ∞,
1 x b→∞ 1 x b→∞ b→∞
1 1
y=x
the area is not finite, and cannot therefore be painted.
y
(b) We use discs to determine whether Gabriel’s horn has finite volume:
1 b x
! ! ) * 4
dx ∞
2
b
π −π b π5
πy dx = lim dx = lim = lim π − = π.
1 b→∞ 1 x2 b→∞ x 1 b→∞ b
The volume is finite. Consider this now. The horn can be filled with paint, but the
flat cross-sectional area inside the horn cannot be painted. How can this be?
EXAMPLE 7.34
Find the escape velocity of a projectile from the earth’s surface.
FIGURE 7.82 Escape
velocity of projectile from earth’s SOLUTION Suppose the projectile is fired from the earth’s surface with speed v in the x -
gravitational field direction (Figure 7.82). If the projectile is to escape the earth’s gravitational pull, its initial
kinetic energy must be greater than or equal to the work done against gravity as the projectile
R travels from the earth’s surface ( x = R ) to a point where it is free of gravity ( x = ∞ ). When
the projectile is a distance x from the centre of the earth, the force of attraction on it is
x x + dx x
x=0
GMm
Earth
F (x) = − ,
x2
where G > 0 is a constant, m is the mass of the projectile, and M is the mass of the earth. This
is known as Newton’s universal law of gravitation. The work done against gravity in travelling
from x = R to x = ∞ is
! ∞ ! b
GmM GmM
W = 2
dx = lim dx
R x b→∞ R x2
) * 1 2
GmM b GmM GmM GmM
= lim − = lim − = .
b→∞ x R b→∞ R b R
Since the initial kinetic energy of the projectile is mv 2 /2, it escapes the gravitational pull of the
earth if
1 2 GmM
mv ≥ ,
2 R
that is, if
;
2GM
v ≥ .
R
√
Hence 2GM/R is the escape velocity (notice that it is independent of the mass of the
projectile). If we take the mean radius of the earth as 6370 km, its mean density as ρ =
5.52 × 103 kg/m 3 , and G = 6.67 × 10−11 , we obtain v = 11.2 km/s. (Compare this with a
308 Winchester that fires a 150-grain bullet with a muzzle velocity of only 0.81 km/s.)
A second type of improper integral occurs when the integrand is discontinuous at a point or
points in the interval of integration. For example, integrands of the improper integrals
! 5 ! 5
1 1
√ dx and dx
1 x−1 1 (x − 1)4
each have infinite discontinuities at x = 1. If c is a number between 1 and 5, it is easy to
calculate ! 5
1 7 √ 85 √
√ dx = 2 x − 1 = 4 − 2 c − 1 and
c x−1 c
! 5 ) *5
1 −1 1 1
dx = = − ,
c (x − 1)4 3(x − 1)3 c 3(c − 1)3 192
and one possible interpretation of these definite integrals is the areas in Figures 7.83 and 7.84.
If we let c approach 1 from the right in the first integral, we find that
! 5 4 5
1 √
lim √ dx = lim+ 4 − 2 c − 1 = 4;
c→1+ c x−1 c→1
√
FIGURE 7.83 Area under curve y = 1/ x − 1 FIGURE 7.84 Area under curve y = 1/(x − 1)4
y y
1 1
y= y=
x−1 (x − 1)4
1 1
2 256
1 c 5 x 1 c 5 x
and in the second integral,

! 5 1 2
1 1 1
lim dx = lim+ − = ∞.
c→1+ c (x − 1)4 c→1 3(c − 1)3 192
√
It would seem reasonable then to define the area under the curve y = 1/ x − 1, above
the x -axis, and between the vertical lines x = 1 and x = 5 as 4 square units. On the other
hand, the area bounded by y = 1/(x − 1)4 , y = 0, x = 5, and x = c becomes increasingly
large as x = c moves closer to x = 1, and we cannot therefore assign a value to this area when
c = 1. In other words, if we define
! 5 ! 5
1 1
√ dx = lim+ √ dx and
1 x−1 c→1 c x−1
! 5 ! 5
1 1
dx = lim+ dx,
1 (x − 1)4 c→1 c (x − 1)4
then the first improper integral has a value of 4, whereas the second has no value.
These improper integrals are examples of the general situation described in the following
definition and illustrated in Figures 7.85.

Improper integrals for functions with infinite discontinuities
y y y
a b x a b x a d b x
DEFINITION 7.5
If f (x) is continuous at every point in the interval a ≤ x ≤ b except at x = a , then
! b ! b
f (x) dx = lim+ f (x) dx, (7.50)
a c→a c
provided that the limit exists (Figure 7.85a). If f (x) is continuous at every point in the
interval a ≤ x ≤ b except x = b , then
! b ! c
f (x) dx = lim− f (x) dx, (7.51)
a c→b a
provided that the limit exists (Figure 7.85b). If f (x) is continuous at every point in the
interval a ≤ x ≤ b except x = d where a < d < b , then
! b ! c ! b
f (x) dx = lim− f (x) dx + lim+ f (x) dx, (7.52)
a c→d a c→d c
provided that both limits exist (Figure 7.85c).

EXAMPLE 7.35
Determine whether the following improper integrals converge or diverge:
! 2 ! 1 ! 5
1 1 x
(a) dx (b) dx (c) √ dx
−2 (x − 2 )3 −1 x4 1 x−1
FIGURE 7.86 Improper SOLUTION
integral of f (x) = 1/(x − 2)3
(a) The graph of the integrand (Figure 7.86) shows a discontinuity at x = 2. Hence,
y
! 2 ! c ) *c
1 1 −1
dx = lim− dx = lim−
−2 c −2 (x − 2)3 c→2 −2 (x − 2)3 c→2 2(x − 2)2 −2
1 2 x 1 2
1 − −1 1
y= 8 = lim− + = −∞.
(x − 2) 3
c→2 2(c − 2)2 32
The improper integral therefore diverges.

(b) Since the discontinuity is interior to the interval of integration (Figure 7.87), we use
FIGURE 7.87 Improper equation 7.52:
integral of f (x) = 1/x 4
! 1 ! c ! 1
1 1 1
y dx = lim− dx + lim+ dx
−1 x4 c→0 −1 x4 c→0 c x4
1 ) *c ) *1
y=
x4 −1 −1
= lim− + lim+
c→0 3x 3 −1 c→0 3x 3 c
1 1 2 1 2
1 1 1 1
−1 1 x = lim− − − 3 + lim+ − .
c c c→0 3 3c c→0 3c3 3
Since neither of these limits exists, the improper integral diverges.

(c) Since the integrand is discontinuous only at x = 1 (Figure 7.88),
FIGURE 7.88
√Improper
integral of f (x) = x/ x − 1
! 5 ! 5
x
x x
y y= √ dx = lim+ √ dx.
x−1 1 x−1 c→1 c x−1
If we set u = x − 1, then du = dx , and

! ! !
x u+1
√ dx = √ du = (u1/2 + u−1/2 ) du
x−1 u
1 c 5 x
2 2 √
= u3/2 + 2u1/2 + C = (x − 1)3/2 + 2 x − 1 + C.
3 3
Thus,
! 5 ) *5
x 2 3/2
√
√ dx = lim+ (x − 1) + 2 x − 1
1 x−1 c→1 3 c
1 2
28 2 3/2
√ 28
= lim+ − (c − 1) − 2 c − 1 = ,
c→1 3 3 3
and the improper integral converges.

FIGURE 7.89 Improper

integral of f (x) = 1/(x 2 − 1)
Various types of improper integrals may occur in the same problem. For example, Figure 7.89
shows that the integral of f (x) = 1/(x 2 − 1) from x = −3 to x = ∞ involves the use of
y five limits:
! ∞ ! c ! 0 ! c
1 1 1 1
dx = lim dx + lim + dx + lim− dx
−3 −1 1 3 x −3 x2 − 1 c→−1− −3 x2 − 1 c→−1 c x2 − 1 c→1 0 x2 − 1
−1
! 10 ! b
1 1
+ lim+ dx + lim dx.
c→1 c x2 −1 b→∞ 10 x2 −1
FIGURE 7.90 Improper If any one of these limits fails to exist, then the improper integral diverges.
integral of function with jump If an integrand has a jump discontinuity, such as the function f (x) in Figure 7.90, then the
discontinuity improper integral of f (x) from x = a to x = b is defined in terms of two limits,
y
! b ! c ! b
f (x) dx = lim− f (x) dx + lim+ f (x) dx,
a c→d a c→d c
but there is no question in this case that the improper integral converges.
a d b x
EXERCISES 7.10
In Exercises 1–18 determine whether the improper integral converges ∗ 20. (a) Is it possible to assign a number to represent the area
or diverges. Find the value for each convergent integral. bounded by the curves y = 1 −x −1/4 , y = 1, and x = 1?
! !
∞
1 ∞
1 (b) If the region in part (a) is rotated about the line y = 1, is it
1. dx 2. dx possible to assign a number to represent its volume?
3 (x + 4)2 3 (x + 4)1/3
! −4 ! −4
x x ∗ 21. Repeat Exercise 20 if y = 1 −x −1/4 is replaced by y = 1 −x −2/3 .
3. √ dx 4. dx
−∞
2
x −2 −∞ (x 2 − 2)4
! ∞ 10 3 ! ∞
∗ 22. Repeat Exercise 20 if y = 1 − x −1/4 is replaced by y = 1 − x −3 .
10 x x3
5. dx 6. dx ∗ 23. A function f (x) qualifies as a probability density function (pdf)
−∞ (x 4 + 5)2 −∞ (x + 5)1/4
4
! ! on the interval
! x ≥ 0 if it satisfies two conditions: f (x) ≥ 0 for
1 1 ∞
1 1
7. dx 8. √ dx x ≥ 0 and f (x) dx = 1.
0 (1 − x)5/3 0 1−x 0
! 3 !
∞ 5
x (a) Show that each of the following functions qualifies as a pdf:
9. x x 2 − 1 dx 10. √ dx
1 2 x2 − 4
! ! 6x 2x
1
x ∞
1 f (x) = ; f (x) =
11. dx 12. dx (1 + 3x 2 )2 (1 + x)3
−1 (1 − x 2 )2 −∞ x2
! ∞ ! π/2 (b) If a variable x has pdf f (x) defined for x ≥ 0, then
1 x
13. √ dx 14. dx the probability that x lies in an interval I is the definite
0 x −∞ (x 2 − 4)2 integral of f (x) over I . In particular, the probability that
! !
∞ ∞ x is greater than or equal to a is
15. cos x dx 16. sin x dx
4 −∞ ! ∞
! ∞ ! 3 3 P (x ≥ a) = f (x) dx.
x x
∗ 17. √ dx ∗ 18. √ dx a
0 x+3 2 x2 − 4
Calculate P (x ≥ 3) for each pdf in part (a).
∗ 19.
! ∞If f (x) is a continuous odd function, is it necessarily true that ! ∞
1
f (x) dx = 0? ∗ 24. Verify that dx converges if p > 1 and diverges if p ≤ 1.
−∞ 1 xp
Summary 487
∗ 25. The force of repulsion between two point charges of like sign, one ∗ 34. One of the most important functions in physics and engineering is
of size q and the other of size unity, has magnitude the Dirac-delta function δ(x − a) (introduced in Section 2.5 as the
limit of the unit pulse function shown below):
q
F = ,
4π &0 r 2 δ(x − a) = lim P& (x − a).
&→0
where &0 is a constant and r is the distance between the charges. The
potential V at any point P due to charge q is defined as the work For a continuous function f (x) ,
required to bring the unit charge to P from infinity along the straight
! ∞ ! ∞ C D
line joining q and P . Find a formula for V .
f (x)δ(x − a) dx = f (x) lim P& (x − a) dx.
∗ 26. Verify that if f (x) is continuous on a ≤ x ≤ b except for a finite −∞ −∞ &→0
discontinuity at d (Figure 7.90), then the improper integral of f (x)
from x = a to x = b must converge. Assuming that the order of integration and the process of taking the
limit as & approaches zero can be interchanged, that is,
∗ 27. Find the length of the loop of the curve 9y 2 = x(3 − x)2 .
! ∞ C D ! ∞
∗ 28. Is equation 7.49 equivalent to the following equation?
f (x) lim P& (x − a) dx = lim f (x)P& (x − a) dx,
! ∞ ! a −∞ &→0 &→0 −∞
f (x) dx = lim f (x) dx
a→∞ −a
−∞ show that ! ∞
f (x)δ(x − a) dx = f (a).
In Exercises 29–32 determine whether the improper integral converges −∞
or diverges. Hint: Compare each integral to a known convergent or This is the fundamental operational property that makes the Dirac-delta
divergent integral. function so valuable.
! ∞ ! 3
x2 x3 y
∗ 29. √ dx ∗ 30. 3 2
dx
2 x2 − 1 1 (27 − x )
! 1 ! −2 √
x2 −x
∗ 31. √ dx ∗ 32. 2 2
dx 1
0 1 − x2 −∞ (x + 5)
∗ 33. Show that existence of the limits in equation 7.49 is independent

of the choice of c . a x
SUMMARY
In this chapter we used definite integrals in a wide variety of applications. With differentials
and representative elements we were able to avoid memorization of formulas. To summarize
the use of representative elements in various applications, consider the region R in Figure 7.91.
To find the area of R , we draw rectangles of length f (x) − g(x) and width dx . Areas
[f (x) − g(x)] dx of all such rectangles are then added to find the total area:
! b
[f (x) − g(x)] dx.
a
FIGURE 7.91
y
y = f (x)
y = g (x)
a b x
dx
If R is rotated around the x -axis to form a solid of revolution, the rectangle generates a washer
with volume {π [f (x)]2 − π [g(x)]2 } dx , and therefore the total volume is
! b - .
π [f (x)]2 − π [g(x)]2 dx.
a
If R is rotated about the y -axis, the rectangle traces out a cylindrical shell of volume 2π x [f (x)−
g(x)] dx , and the total volume is
! b
2π x [f (x) − g(x)] dx.
a
First moments of this rectangle about the x -and y -axes are

1
[f (x) + g(x)][f (x) − g(x)] dx and x [f (x) − g(x)] dx,
2
and these lead to the definite integrals
! b ! b
1- 2 2
.
[f (x)] − [g(x)] dx and x [f (x) − g(x)] dx
a 2 a
for first moments of area of R about the x - and y -axes. Second moments of area of region R
about the x - and y -axes are
! b ! b
1- .
[f (x)]3 − [g(x)]3 dx and x 2 [f (x) − g(x)] dx,
a 3 a
where in the first integral it was necessary to use formula 7.42 for the second moment of the
rectangle.
The key, then, to use of the definite integral is to divide the region into smaller elements,
calculate the quantity required for a representative element, and then use a definite integral to
add over all elements.
An important point to keep in mind is that for lengths in the y -direction, we always take
“upper y minus lower y ,” and for lengths in the x -direction, we use “larger x minus smaller
x .” This way, many sign errors can be avoided when calculating lengths. For instance, note that
this rule is used to find the length of a rectangle (in areas, volumes, fluid pressure), the radius
of a circle (in volumes, surface area), depth (fluid pressure), and some distances (work).
An improper integral is a definite integral with an infinite limit and/or a point of discontinuity
in the interval of integration. The value of an improper integral is defined by first calculating
it over a finite interval in which the integrand is continuous, and then letting the limit on the
integral approach either infinity or the point of discontinuity.
KEY TERMS
Area Representative rectangle

Volumes of solids of revolution Disc method
Washer method Cylindrical shell method
Length of a curve Work
Energy Kinetic energy
Conservation of energy Fluid pressure
First moment of mass Centre of mass
Moment of inertia or second moment Volumes by slicing
Area of a surface of revolution Green’s function
Improper integrals
REVIEW
EXERCISES
In Exercises 1–5 calculate for the region bounded by the curves: (a)
its area; (b) the volumes of the solids of revolution obtained by rotating Surface of water
the region about the x - and y -axes; (c) its centroid; and (d) its second
moments of area about the x - and y -axes.
2m
50 cm
1. y = 9 − x 2 , y = 0 2. y = x 3 , y = 2 − x, x = 0
3. x = y 2 − 1, x = 4 − 4y 2 4. 2y = x, y + 1 = x, y = 0
∗ 5. 2x + y = 2, y − 2x = 2, x = y + 1, y + x + 1 = 0
30 cm 70 cm 80 cm
In Exercises 6–10 find the volume of the solid of revolution obtained

by rotating the region bounded by the curves about the line. 18. Find the surface area of the volume of the solid of revolution ob-
tained by rotating the region bounded by the curves 2x + y = 2,
6. y = |x| + 2, y = 3, about y = 1 y − 2x = 2, and y = 0 about the x -axis.
∗ 19. A tank in the form of an inverted right circular cone of depth 10 m
7. x = y 4 , x = 4, about x = 4 and radius 4 m is full of water. How much work is required to lower the
water level in the tank by 2 m by pumping water out through an outlet
8. y = x 3 − x, y = 0, about y = 0
1 m above the top of the tank?
9. yx 2 = 1, x = 1, x = 2, y = 0, about y = −1 ∗ 20. Show that if it takes W units of work to stretch a spring from
√ equilibrium to a certain stretch, then it requires 3W more units of work
∗ 10. y = x + 2, y = 4 − 2x, y = 0, about y = 1 to double that stretch.
In Exercises 21–28 determine whether the improper integral converges

In Exercises 11–15 find the first and second moments of area of the or diverges. Evaluate each convergent integral.
region bounded by the curves about the line.
! ∞ ! 3
1 1
2 2 ∗ 21. dx ∗ 22. √ dx
11. y = 4x , y = 2 + 2x , about y = −1 1 x 3/ 2 0 3−x
√ √ √ ! 0 ! 2
12. 2x = y, 2x = y − 2, x = 0, about x = 1 1 x
∗ 23. dx ∗ 24. √ dx
−1 (x + 1)2 −2 4 − x2
13. 3y = 2x + 6, 2y + x = 4, y = 0, about y = −2 ! !
∞ −3
1 1
14. 3y = 2x + 6, 2y + x = 4, y = 0, about y = 2 ∗ 25. 3
dx ∗ 26. √ dx
−∞ (x + 3) −∞ −x
! ∞ 3 !
∗ 15. y = x 2 − 2x (y ≤ 0), y = 2x, y = 4 − 2x , about x = 1 ∞
x
∗ 27. x x 2 + 4 dx ∗ 28. dx
−6 −∞ (x 2 − 1)2
16. A water tank in the form of a right circular cylinder with radius
1 m and height 3 m has its axis vertical. If it is half full, how much ∗ 29. A uniformly tapered rod of length 1 m has circular cross-sections,
work is required to empty the tank through an outlet at the top of the the radii of its ends being 1 cm and 2 cm. Find the volume of the rod by
tank? (a) using the fact that every cross-section is circular and (b) considering
∗ 17. A car runs off a bridge into a river and submerges. If the tops of the rod as a volume of a solid of revolution.
its front and rear side windows are 2 m below the surface of the water ∗ 30. In Exercise 16, how much work is required to empty the tank if its
(see figure), what is the force due to water pressure on each window? axis is horizontal and the outlet is at the top of the tank?
CHAPTER
8 Techniques of Integration
Application Preview Shown below are a tractor and trailer in two positions. Both tractor and trailer begin on the
x -axis (position A ). The tractor then turns to the right and moves so that the pivot point where
tractor and trailer are coupled together follows a curve whose equation is Y = g(X) . Tractor
and trailer are shown at some time later in position B .
y
Path followed by pivot point
Tractor Path followed by point

between rear wheels
Position B
Pivot point
x
Position A
THE PROBLEM Given that the pivot point moves along the curve Y = g(X) , find the
equation of the curve followed by the point midway between the rear wheels. (See Example 8.13
on page 511 for the solution.)
In Chapter 5, we discussed three techniques for evaluating antiderivatives. First, since
every differentiation formula can be restated as an antidifferentiation formula, it follows that the
more competent we are at differentiation, the more likely we are to recognize an antiderivative.
Our second technique was called “adjusting constants.” It is applicable to sufficiently simple
integrands that we can immediately guess the antiderivative to within a multiplicative constant.
It is then a matter of adjusting the constant. The third technique was to change the variable of
integration. By a suitable transformation, replace a complicated antiderivative with a simpler
one.
In this chapter we develop additional techniques applicable to more complex integration
problems. At the same time we take the opportunity to review the applications of definite
integrals in Chapter 7.
8.1 Integration Formulas and Substitutions

Every differentiation formula can be restated as an antidifferentiation formula. A partial list of
frequently encountered formulas is as follows:
!
un+1
un du = + C, n != −1; (8.1a)
n+1
!
1
du = ln |u| + C; (8.1b)
u
490
!
eu du = eu + C; (8.1c)
!
a u du = a u loga e + C; (8.1d)
!
cos u du = sin u + C; (8.1e)
!
sin u du = − cos u + C; (8.1f)
!
sec2 u du = tan u + C; (8.1g)
!
csc2 u du = − cot u + C; (8.1h)
!
sec u tan u du = sec u + C; (8.1i)
!
csc u cot u du = − csc u + C; (8.1j)
!
tan u du = ln | sec u| + C = − ln | cos u| + C; (8.1k)
!
cot u du = − ln | csc u| + C = ln | sin u| + C; (8.1l)
!
sec u du = ln | sec u + tan u| + C; (8.1m)
!
csc u du = ln | csc u − cot u| + C; (8.1n)
!
1
√ du = Sin−1 u + C; (8.1o)
1 − u2
!
1
du = Tan−1 u + C; (8.1p)
1 + u2
!
1
√ du = Sec−1 u + C; (8.1q)
u u2 − 1
!
cosh u du = sinh u + C; (8.1r)
!
sinh u du = cosh u + C; (8.1s)
!
sech 2 u du = tanh u + C; (8.1t)
!
csch 2 u du = −coth u + C; (8.1u)
!
sech u tanh u du = −sech u + C; (8.1v)
!
csch u coth u du = −csch u + C. (8.1w)
All of these except Formulas 8.1k, l, m, n are restatements of differentiation formulas from
Chapter 3. Formulas 8.1k and 8.1m are verified in Example 8.1 below. Formulas 8.1l and 8.1n
are similar.
492 Chapter 8 Techniques of Integration
EXAMPLE 8.1
Verify formulas 8.1k and 8.1m.
SOLUTION To verify 8.1k we express tan u in terms of sin u and cos u :
! !
sin u
tan u du = du = − ln | cos u| + C = ln | sec u| + C.
cos u
If integration to ln | cos u| is not obvious, try the substitution v = cos u .

To obtain the antiderivative of sec u requires a subtle trick. We first multiply numerator
and denominator by sec u + tan u :
! ! !
sec u + tan u sec2 u + sec u tan u
sec u du = sec u du = du.
sec u + tan u sec u + tan u
If we now set v = sec u + tan u , then dv = (sec u tan u + sec2 u) du , and

! !
1
sec u du = dv = ln |v| + C = ln | sec u + tan u| + C.
v
We could list many other integration formulas, but one of the purposes of this chapter is to
develop integration techniques that eliminate the need for excessive memorization of formulas.
Our emphasis is on development of antiderivatives rather than memorization of formulas, and
we prefer to keep the list of formulas short.
In Chapter 5 we demonstrated how a change of variable could sometimes replace a complex
integration problem with a simpler one. When the integrand contains x 1/n , where n is a positive
integer, it is often useful to set u = x 1/n , or, equivalently, x = un . We illustrate in the following
example.
EXAMPLE 8.2
! √ ! ! √
2− x x 1/3 x
(a) √ dx (b) dx (c) dx
2+ x 1 + x 2 /3 1 + x 1/3
SOLUTION
√
(a) If we set u = x or x = u2 , then dx = 2u du , and
! √ ! !
2− x 2−u 2 u − u2
√ dx = (2u du) = 2 du.
2+ x 2+u 2+u
Long division of −u2 + 2u by u + 2 immediately gives

! √ ! " #
2− x 8
√ dx = 2 −u + 4 − du
2+ x u+2
" 2 #
u
= 2 − + 4u − 8 ln |u + 2| + C
2
√ $ √ %
= −x + 8 x − 16 ln 2 + x + C.
(b) If we set u = x 1/3 or x = u3 , then dx = 3u2 du , and

! ! !
x 1/3 u u3
dx = (3u2 du) = 3 du.
1 + x 2 /3 1 + u2 1 + u2
Division of u3 by u2 + 1 leads to
! ! " #
x 1/3 u
dx = 3 u− du
1 + x 2 /3 u2 + 1
" 2 #
u 1 2
=3 − ln |u + 1| + C
2 2
3 3
= x 2 /3 − ln (x 2/3 + 1) + C.
2 2
(c) The purpose of the substitution u = x 1/n is to rid the integrand of fractional powers.
In the presence of both x 1/2 and x 1/3 , we set u = x 1/6 or x = u6 . Then dx =
6u5 du , and
! √ !
x u3
1 3
dx = (6u5 du) (and by long division)
1+x / 1 + u2
! " #
6 4 2 1
=6 u −u +u −1+ du
1 + u2
" 7 #
u u5 u3 −1
=6 − + − u + Tan u + C
7 5 3
6 6
= x 7/6 − x 5/6 + 2x 1/2 − 6x 1/6 + 6 Tan−1 (x 1/6 ) + C.
7 5
EXERCISES 8.1 ! !
sin θ x+3
In Exercises 1–22 evaluate the indefinite integral. 13. dθ 14. √ dx
cos θ − 1 2x + 4
! ! ! !
x2 2
ex
1. dx 2. xe−2x dx 15. dx 16. sin3 2x cos 2x dx
5 − 3x 3 1 + e 2x
! ! ! !
x ex x3
3. dx 4. dx 17. x 5 (2x 2 − 5)4 dx 18. dx
(x 2 + 2)1/3 1 + ex (x + 5)2
! !
! ! & √
4t + 8 19. z2 3 − z dz 20. tan 3x dx
5. dt 6. x 2 1 − 3x 3 dx
t 2 + 4t + 5
! ! √
! ! (x − 3)2/3 x
x 2 21. dx 22. dx
7. (x + 1)(x 2 + 2x)1/3 dx 8. dx (x − 3)2/3 + 1 1 + x 1/ 4
(1 + x 3 )3
! √ ! √ ∗ 23. Find the length of the curve y = ln (cos x) from (0, 0) to

x 1− x (π/4, −(ln 2)/2) .
9. √ dx 10. √ dx
1− x x
∗ 24. Find the area of the region bounded by the curves
! !
x+2 x2 + 2 x2 + 1
11. dx 12. dx y = , x + 3y = 7.
x+1 x2 + 1 x+1
√
∗ 25. (a) Find the centroid of the region bounded by the curves y = ∗∗ 31. Use the substitution u − x = x 2 + 3x + 4 to evaluate
(3 + x)3/2 , x = 1, and y = 0.
!
(b) What is its second moment of area about the line x = 1? 1
dx.
∗ 26. If f (x) is a continuous even function, prove that (x 2 + 3x + 4)3/2
! a ! a
√
f (x) dx = 2 f (x) dx; ∗∗ 32. Show ! that the substitution u − x = x 2 + bx + c replaces the
−a 0 &
integral x 2 + bx + c dx with the integral of a rational function
and if f (x) is a continuous odd function,
of u .
! a
∗∗ 33. Show that when the quadratic c + bx − x 2 factors as c + bx −
f (x) dx = 0.
−a
x 2 = (p + x)(q − x) , then either
√ √ of the substitutions (p + x)u =
c + bx − x 2 or (q − x)u = c + bx − x 2 replaces the integral
∗ 27. (a) Show that the function f (x) = λe−λx , where λ > 0 is a
constant, qualifies as a probability density function on the !
1
interval x ≥ 0. See Exercise 23 in Section 7.10 for the √ dx
definition of a pdf. c + bx − x 2
(b) If the probability that x ≥ 3 is 0.5, what is λ ?
with the integral of a rational function of u .
∗ 28. A paratrooper and his parachute fall from rest with a combined !
mass of 100 kg. At any instant during descent, the parachute has an air ∗∗ 34. If a , b , and n are constants, evaluate x(a + bx)n dx .
resistance force acting on it that in newtons is equal to one-half its speed
(in metres per second). Assuming that the paratrooper falls vertically ∗∗ 35. The function
downward and that the parachute is already open when the jump takes
place, the differential equation describing his motion is 1 2 2 /(2σ )
f (x) = √ e−(x−µ) , −∞ < x < ∞,
dv 2π σ
200 = 1962 − v,
dt
where µ and σ are constants, is called the normal probability density
where v is velocity and t is time. function. Consequently,
(a) Find v as a function of t .
! ∞
(b) If x measures vertical distance fallen, then v = dx/dt .
f (x) dx = 1.
Use this to find the position of the paratrooper as a function −∞
of time.
! 4π √ If the expected value of x is defined as
∗ 29. Evaluate 1 + cos x dx . Hint: cos x = 2 cos2 (x/2) − 1.
0
∗ 30. Evaluate ! ! ∞
1 xf (x) dx,
dx
x(3 + 2x n ) −∞
for any n > 0. Hint: Find A and B in order that

show that µ is the expected value.
n−1
! ∞
1 A Bx
= + . ∗∗ 36. Evaluate, if possible, e−|a−x| e−|x| dx , where a is a constant.
x(3 + 2x n ) x 3 + 2x n −∞
8.2 Integration by Parts

One of the most powerful techniques for finding antiderivatives is that called integration by
parts. It results from the product rule for differentiation:
d dv du
(uv) = u +v .
dx dx dx
If we take antiderivatives of both sides of this equation with respect to x , we obtain
! ! !
d dv du
(uv) dx = u dx + v dx.
dx dx dx
Since the antiderivative on the left is uv , this equation may be rewritten in the form
! !
u dv = uv − v du. (8.2)
Equation 8.2 is called the integration-by-parts formula. It says that given an unknown integral
(the left side of 8.2), if we arbitrarily divide the function after the integral sign into two parts —
one called u and the other called dv — then we can set the unknown integral equal to the right
side of 8.2. What this does is replace the integral
! !
u dv with v du.
The problem is to choose u and dv so that the new integral is simpler than the original. For
many examples, it is advantageous to let dv be the most complicated part of the integrand (plus
differential) that we can integrate mentally. The example
!
xex dx
will clarify these ideas. To use integration by parts on this problem we must define u and dv
in such a way that u dv = xex dx . There are four possibilities:
u = 1, dv = xex dx;
u = x, dv = ex dx;
u = ex , dv = x dx;
u = xex , dv = dx.
Our rule suggests the choice
u = x, dv = ex dx,
in which case
du = dx, v = ex .
Formula 8.2 then gives
! !
x x
xe dx = xe − ex dx.
We have therefore replaced integration of xex with integration of ex , a definite simplification.

The final solution is therefore
!
xex dx = xex − ex + C.
This example illustrates that integration by parts replaces an integral with two others: one
mental integration (from dv to v ) and a second integration that is hopefully simpler than the
original. We should also note that in the evaluation of the mental integration we did not include
a constant of integration. For example, in the integration from dv = ex dx to v = ex , we did
not write v = ex + D . Had we done so, the solution would have proceeded as follows:
! !
xex dx = x(ex + D) − (ex + D) dx
= x(ex + D) − ex − Dx + C
= xex − ex + C.
Constant D has therefore disappeared from the eventual solution. This always occurs, and
consequently, we do not include an arbitrary constant in the mental integration for v .
EXAMPLE 8.3
Evaluate the following integrals:
! ! !
2
(a) ln x dx (b) x cos x dx (c) ex sin x dx
SOLUTION
(a) Integration by parts leaves no choice for u and dv in this integral; they must be
u = ln x, dv = dx,
in which case
1
du = dx, v = x.
x
Then we have
! !
1
ln x dx = x ln x − x dx = x ln x − x + C.
x
(b) If we set
u = x2, dv = cos x dx,
then
du = 2x dx, v = sin x, and
! !
x 2 cos x dx = x 2 sin x − 2x sin x dx.
Integration by parts has therefore reduced the power on x from 2 to 1. To eliminate

the x in front of the trigonometric function completely, we perform integration by
parts once again, this time with
u = x, dv = sin x dx.
Then
du = dx, v = − cos x, and
! " ! #
2 2
x cos x dx = x sin x − 2 −x cos x − − cos x dx
= x 2 sin x + 2x cos x − 2 sin x + C.

(c) If we set
u = ex , dv = sin x dx,
then
du = ex dx, v = − cos x, and
! !
ex sin x dx = −ex cos x − −ex cos x dx.
Integration by parts appears to have led nowhere; the integration of ex cos x is essen-
tially as difficult as that of ex sin x . If, however, we persevere and integrate by parts
once again with
u = ex , dv = cos x dx,
then
du = ex dx, v = sin x, and
! !
ex sin x dx = −ex cos x + ex sin x − ex sin x dx.
This equation can now be solved for the unknown integral. By transposing the last
term on the right to the left, we find that
!
2 ex sin x dx = ex (sin x − cos x),
and therefore,
!
1
ex sin x dx = ex (sin x − cos x) + C,
2
to which we have added the arbitrary constant C .
The technique in part (c) of this example is often called integration by reproduction. Two
applications of integration by parts enabled us to reproduce the unknown integral and then solve
the equation for it.
EXAMPLE 8.4
!
x3
Evaluate √ dx .
9 + x2
Solution Methods
(i) If we set u = 9 + x 2 , then du = 2x dx , and
! ! ! !
x3 x2 u − 9 du 1
√ dx = √ (x dx) = √ = (u1/2 − 9u−1/2 ) du
9 + x2 9 + x2 u 2 2
" # &
1 2 1
= u3/2 − 18u1/2 +C = (9 + x 2 )3/2 − 9 9 + x 2 + C.
2 3 3
√ ' √ (
(ii) If we set u = 9 + x 2 , then du = x/ 9 + x 2 dx , and
! ! " # !
x3 2 x u3
√ dx = x √ dx = (u2 − 9) du = − 9u + C
9 + x2 9 + x2 3
1 &
= (9 + x 2 )3/2 − 9 9 + x 2 + C.
3
(iii) If we set
x
u = x2, dv = √ dx,
9 + x2
then &
du = 2x dx, v = 9 + x2,
and
! & ! &
x3 2
√ dx = x 9+ x2 − 2x 9 + x 2 dx
9 + x2
& ) *
2 1 2 3/2
=x 9+ x2 −2 (9 + x ) +C
3
& 2
= x 2 9 + x 2 − (9 + x 2 )3/2 + C.
3
This example illustrates that more than one technique may be effective on an indefinite integral,
and the answer may appear different depending on the technique used. According to Theorem
5.1, these solutions can only differ by an additive constant. To illustrate this in Example 8.4,
note that the solution obtained by method (iii) can be written as
& ) *
2 2 2 1& + ,
9+ x2 x − (9 + x ) + C = 9 + x 2 −18 + x 2 + C
3 3
1& + ,
= 9 + x 2 −27 + (9 + x 2 ) + C
3
& 1
= −9 9 + x 2 + (9 + x 2 )3/2 + C,
3
which is the solution obtained by the other two methods. It is also worth noting that the three
methods work very well provided that the power on x in the numerator of the integrand is odd.
When it is even, they fail miserably. In such cases, trigonometric substitutions (Section 8.4)
come to the rescue (see Example 8.12).
Integration-by-parts formula 8.2 applies to indefinite integrals. For definite integrals, it is
replaced by
! b - .b ! b
u dv = uv − v du. (8.3)
a a a
EXAMPLE 8.5
! 4
x
Evaluate √ dx .
0 x+8
SOLUTION If we set
1
u = x, dv = √ dx,
x+8
then √
du = dx, v = 2 x + 8, and
! 4 - √ .4 ! 4
x √
√ dx = 2x x + 8 − 2 x + 8 dx
0 x+8 0 0
/ 04
√ 2 3/2
= 8 12 − 2 (x + 8)
3 0
16 $ √ √ %
= 4 2−3 3 .
3
EXERCISES 8.2
! !
In Exercises 1–18 evaluate the indefinite integral. x x2
9. √ dx 10. √ dx
! ! 2+x 2+x
! !
1. x sin x dx 2. x 2 e2x dx x
11. √ dx 12. (x − 1)2 ln x dx
! ! 2 + x2
√ ! !
3. x 4 ln x dx 4. x ln (2x) dx
13. ex cos x dx 14. Tan −1 x dx
! !
√ ! !
5. z sec2 (z/3) dz 6. x 3 − x dx
15. cos (ln x) dx 16. e2x cos 3x dx
! !
√
7. Sin −1 x dx 8. x 2 x + 5 dx ! !
x3
17. √ dx 18. ln (x 2 + 4) dx
5 + 3x 2
1
∗ 19. Consider the evaluation of x 5 ex dx . Were we to use integration In Exercises 27–30 find the Laplace transform of the function.
by parts, the answer would eventually be of the form
! ∗ 27. f (t) = e3t ∗ 28. f (t) = t 2
x 5 ex dx = Ax 5 ex +Bx 4 ex +Cx 3 ex +Dx 2 ex +Exex +F ex +G. ∗ 29. f (t) = sin t ∗ 30. f (t) = te−t
∗ 31. Show that the function
Differentiate this equation, and thereby obtain A , B , C , D , E , and F
(with no integration). x α−1 e−x/β
f (x) =
∗ 20. The face of a dam is shown in the figure below, where the curve %(α)β α
has equation
for positive constants α and β qualifies as a probability density function
k|x| 1 on the interval x ≥ 0. See Exercise 23 in Section 7.10 for the definition
y =e − 1, k = ln 201. of a pdf and Exercise 26 for the definition of %(α) .
100
Find the total force on the dam due to water pressure when the water
level on the dam is 100 m. In the study of signals and communications in electrical engineering,
the Fourier transform of a function f (t) is
y
! ∞
100 m F (ω) = f (t)e−iωt dt,
−∞
200 m where i is the complex number whose square is −1; that is, i 2 = −1.
In Exercises 32–35 find the Fourier transform of the function f (t) .
x /
1, |t| < L/2
−1
∗ 21. If we set u = x and dv = dx , then du = −x dx and −2 ∗ 32. f (t) = L > 0 a constant
0, |t| > L/2
v = x , and 
! " # ! " # !  0, |t| > T
1 1 1 1 ∗ 33. f (t) = 1 + t/T , −T ≤ t ≤ 0 T > 0 a constant
dx = x − x − 2 dx = 1 + dx. 
x x x x 1 − t/T , 0 ≤ t ≤ T
1
Subtraction of x −1 dx from each side of this equation now gives ∗ 34. f (t) = e−a|t| , a > 0 a constant
0 = 1. What is wrong with the argument? 
 0, t < a
∗ 35. f (t) = 1, a < t b
of engineering. We are constantly evaluating integrals of the form
! L ! L
nπx nπx In Exercises 36–38 evaluate the indefinite integral.
f (x) sin dx and f (x) cos dx,
−L L −L L ! !
√
∗ 36. Tan −1 x dx ∗ 37. x 2 cos2 x dx
where L > 0 is a fixed constant and n is a positive integer. In Exercises
22–25 evaluate these integrals for the given function f (x) . !
∗ 38. xex sin x dx
∗ 22. f (x) = x ∗ 23. f (x) = x 2
∗ 24. f (x) = 1 − 2x ∗ 25. f (x) = 2x 2 − 3x ∗∗ 39. Use the!substitution x = sin2 θ and integration by parts to find a
1
x n (1 − x)m dx .
formula for
∗ 26. The gamma function %(n) for n > 0 is defined by the improper 0
integral ! ∞ ∗∗ 40. Verify that for n ≥ 1 an integer,
%(n) = x n−1 e−x dx.
0
! (n/2)
5 (−1)r n!
x n cos x dx = sin x x n−2r
Show that when n is a positive integer, %(n) = (n − 1)!. (n − 2 r)!
r =0
((n−1)/2)
The Laplace transform of a function f (t) is the function F (s) defined
5 (−1)r n!
+ cos x x n−2r−1 + C,
by the improper integral (n − 2r − 1)!
r =0
! ∞
F (s) = e−st f (t) dt. where (n/2) indicates the floor function of Example 2.22 in
0 Section 2.4.
8.3 Trigonometric Integrals

Differentiation formulas for trigonometric functions suggest three pairings of these functions:
sine and cosine, tangent and secant, and cotangent and cosecant. The derivative of either function
in a given pair involves at most functions of that pair. Note that integration formulas 8.1e–n also
substantiate this pairing. Each indefinite integral gives functions in the same pairing. Because
of this, a first step in every integral involving trigonometric functions is to rewrite the integrand
so that each term involves only one pair. For example, to evaluate
!
tan x + sec x cot 2 x
dx,
sin3 x
we would first rewrite it in the form
! " #
csc3 x cos x
+ dx.
cot x sin5 x
The first term contains cosecants and cotangents, and the second has sines and cosines. We now
consider integrals involving each pair.
Integrals Involving Sine and Cosine

We frequently encounter integrals of the form
!
cosn x sinm x dx,
where m and n may or may not be integers. The key to evaluation of this type of integral is recog-
nition of the fact that when either n = 1 or m = 1, the integrations become straightforward.
To illustrate this, consider differentiation of sinn+1 x by power rule 3.21,
d
sinn+1 x = (n + 1) sinn x cos x.
dx
This means that differentiation of a power of sin x leads to a power of sin x one lower than the
original, multiplied by cos x . In order to reverse the procedure and antidifferentiate a power of
sin x , we need a factor of cos x . It then becomes a matter of raising the power on sin x by one
and dividing by the new power; that is, we can say
!
1
sinn x cos x dx = sinn+1 x + C, n != −1. (8.4a)
n+1
Similarly,
!
1
cosn x sin x dx = − cosn+1 x + C, n != −1. (8.4b)
n+1
Formulas 8.1l and 8.1k contain the n = −1 cases for these integrals. Of course, a substitution
can always be made to evaluate integrals 8.4a and b. For example, if in 8.4a we set u = sin x ,
then du = cos x dx , and
! !
un+1 sinn+1 x
sinn x cos x dx = un du = +C = + C.
n+1 n+1
Also useful are the formulas
!
1
sin nx dx = − cos nx + C and (8.4c)
n
!
1
cos nx dx = sin nx + C. (8.4d)
n
Trigonometric identities are used to write integrands in the forms contained in equations
8.4. Particularly helpful are
sin2 x + cos2 x = 1, (8.5a)
sin 2x = 2 sin x cos x, (8.5b)

2
cos 2x = 2 cos x − 1, (8.5c)
2
cos 2x = 1 − 2 sin x. (8.5d)
EXAMPLE 8.6
Evaluate
! the following integrals:! !
3 3 5
(a) sin x dx (b) sin x cos x dx (c) sin 2x cos6 2x dx
! √ ! !
cos3 3x
(d) sin x cos3 x dx (e) 2
dx (f) sin2 x dx
! ! sin 3x
(g) cos4 x dx (h) sin2 x cos2 x dx
SOLUTION
(a) With identity 8.5a, we may write
! ! !
sin3 x dx = sin x(1 − cos2 x) dx = (sin x − cos2 x sin x) dx,
and the second term is in form 8.4b. Thus,

!
1
sin3 x dx = − cos x + cos3 x + C.
3
(b) Once again we use 8.5a to rewrite the integrand as two easily integrated terms,
! !
3 5
sin x cos x dx = sin x(1 − cos2 x) cos5 x dx
!
= (cos5 x sin x − cos7 x sin x) dx
1 1
= − cos6 x + cos8 x + C.
6 8
(c) This integrand is already in a convenient form for integration. Adjusting constants
leads to !
1
sin 2x cos6 2x dx = − cos7 2x + C.
14
(d) Using 8.5a, we obtain
! √ !
sin x cos3 x dx = sin1/2 x cos x (1 − sin2 x) dx
!
= (sin1/2 x cos x − sin5/2 x cos x) dx
2 2
= sin3/2 x − sin7/2 x + C.
3 7
(e) Again we use 8.5a,

! ! ! " #
cos3 3x cos 3x (1 − sin2 3x) cos 3x
dx = dx = − cos 3x dx
sin2 3x sin2 3x sin2 3x
−1 1
= − sin 3x + C.
3 sin 3x 3
In these five examples, at least one of the powers on the sine and cosine functions
was odd, and in each case we used identity 8.5a. In the next three examples, where
all powers are even, identities 8.5b–d are useful.
(f) Double-angle formula 8.5d can be rearranged as sin2 x = (1 − cos 2x)/2, and
therefore ! !
1
sin2 x dx = (1 − cos 2x) dx.
2
Formula 8.4d can now be used on the second term,
! " #
1 1
sin2 x dx = x − sin 2x + C.
2 2
(g) With double-angle formula 8.5c rewritten in the form cos2 x = (1 + cos 2x)/2, we
obtain
! ! !
1 1
cos4 x dx = (1 + cos 2x)2 dx = (1 + 2 cos 2x + cos2 2x) dx.
4 4
When x is replaced by 2x in 8.5c, the result is cos 4x = 2 cos2 2x − 1. This can be

rearranged as cos2 2x = (1 + cos 4x)/2, and therefore
! ! " #
1 1 + cos 4x
cos4 x dx = 1 + 2 cos 2x + dx.
4 2
We now use 8.4d on the two cosine terms,

! " #
4 1 3x 1
cos x dx = + sin 2x + sin 4x + C.
4 2 8
(h) In this integral we use identities 8.5b and d,

! ! !
2 2 2 1
sin x cos x dx = (sin x cos x) dx = sin2 2x dx
4
! " # " #
1 1 − cos 4x 1 1
= dx = x − sin 4x + C.
4 2 8 4
EXAMPLE 8.7
!
Evaluate sin 5x cos 2x dx .
SOLUTION Unlike Example 8.6, in which all trigonometric functions had the same argument,
this integrand is the product of trigonometric functions with different arguments. Our first step
is to rewrite the integrand so that we do not have such a product. Note that a sum is obviously
integrable, but not a product. The product formula
1
sin A cos B = [sin (A + B) + sin (A − B)]
2
can be used to advantage here. We can rewrite the integrand in the form
! !
1
sin 5x cos 2x dx = (sin 7x + sin 3x) dx,
2
and now integration is easily handled by 8.4c,

! " #
1 1 1 1 1
sin 5x cos 2x dx = − cos 7x − cos 3x + C = − cos 7x − cos 3x + C.
2 7 3 14 6
Two-integration by parts (and reproduction) can also be used to evaluate this integral, but the
above method is much simpler.
Integrals Involving Tangent and Secant

For integrals involving tangent and secant, we have two alternatives: Rewrite the integrand in
terms of sines and cosines, or express the integrand in terms of easily integrated combinations
of tangent and secant. It is usually more fruitful to investigate the second possibility before
expressing the integrand completely in terms of sines and cosines. In this regard we note that
combinations of tangent and secant that can be integrated immediately are
!
tann+1 x
tann x sec2 x dx = + C, n != −1; (8.6a)
n+1
!
secn x
secn x tan x dx = + C, n != 0. (8.6b)
n
Once again these results are suggested by differentiations of powers of tan x and sec x . Alter-
natively, these integrals can be evaluated with substitutions: u = tan x for 8.6a and u = sec x
for 8.6b.
To rearrange integrands into these integrable combinations, we use the identity
1 + tan2 x = sec2 x. (8.7)
EXAMPLE 8.8
! ! !
(a) tan2 x dx (b) tan3 x dx (c) sec4 x dx
! ! !
4 2 3 3 sec4 x
(d) tan 3x sec 3x dx (e) tan x sec x dx (f) dx
tan2 x
SOLUTION
! !
2
(a) tan x dx = (sec2 x − 1) dx = tan x − x + C
! !
3
(b) tan x dx = tan x (sec2 x − 1) dx
!
= (sec2 x tan x − tan x) dx
1 1
= sec2 x + ln | cos x| + C, or tan2 x + ln | cos x| + C
2 2
! !
4
(c) sec x dx = sec2 x (1 + tan2 x) dx
!
= (sec2 x + tan2 x sec2 x) dx
1
= tan x + tan3 x + C
3
!
1
(d) tan4 3x sec2 3x dx = tan5 3x + C
15
! !
(e) tan3 x sec3 x dx = tan x (sec2 x − 1) sec3 x dx
!
= (sec5 x tan x − sec3 x tan x) dx
1 1
= sec5 x − sec3 x + C
5 3
! !
sec4 x sec2 x (1 + tan2 x)
(f) dx = dx
tan2 x tan2 x
! " #
sec2 x
= + sec2 x dx
tan2 x
1
=− + tan x + C
tan x
EXAMPLE 8.9
!
Evaluate sec3 x dx .
SOLUTION We use integration by parts and reproduction. By setting
u = sec x, dv = sec2 dx,

then
du = sec x tan x dx, v = tan x,
and formula 8.2 gives
! !
3
sec x dx = sec x tan x − sec x tan2 x dx.
Since tan2 x = sec2 x − 1, we can write

! !
sec3 x dx = sec x tan x − sec x (sec2 x − 1) dx
!
= sec x tan x − sec3 x dx + ln | sec x + tan x|.
We now solve for

!
1
sec3 x dx = (sec x tan x + ln | sec x + tan x|) + C.
2
Integrands in the following example cannot be transformed by means of identity 8.7 into forms
8.6, and we therefore express them in terms of sines and cosines.
EXAMPLE 8.10
! !
tan2 2x 1
(a) dx (b) dx
sec3 2x sec x tan2 x
SOLUTION
! !
tan2 2x 1
(a) dx = sin2 2x cos 2x dx = sin3 2x + C
sec3 2x 6
! ! !
1 cos3 x cos x (1 − sin2 x)
(b) dx = dx = dx
sec x tan2 x sin2 x sin2 x
! ' (
cos x 1
= 2
− cos x dx = − − sin x + C
sin x sin x
Integrals Involving Cotangent and Cosecant

Since derivatives of the cotangent and cosecant pair
d d
cot x = − csc2 x, csc x = − csc x cot x
dx dx
are analogous to those of the tangent and secant pair
d d
tan x = sec2 x, sec x = sec x tan x,
dx dx
a discussion parallel to that concerning the tangent and secant could be made here. Aside from
the fact that cotangent replaces tangent and cosecant replaces secant, there are also sign changes.
Examples are given in the exercises.
EXERCISES 8.3
In Exercises 1–20 evaluate the indefinite integral. φ2 are constants, is P = V i . Show that the average power supplied
! ! to the resistor is Pav = (Vm im /2) cos (φ1 − φ2 ) .
cos x
1. cos3 x sin x dx 2. dx ∗ 33. The alternating current in a power line is given by
sin3 x
! !
I = f (t) = A cos ωt + B sin ωt,
3. tan5 x sec2 x dx 4. csc3 x cot x dx
! ! where A , B , and ω are constants, and t is time. The root-mean-square
3
√ 4 (rms) current Irms is defined by
5. cos (x + 2) dx 6. tan x sec x dx
! T
! ! 1
1 6 (Irms )2 = I 2 dt,
7. 4
dt 8. sec 3x tan 3x dx T 0
sin t
! ! where the interval 0 ≤ t ≤ T represents any
tan3 x sec2 x √ number of complete
9. cos2 x dx 10. 2
dx oscillations of the current. Show that Irms is 1/ 2 times the amplitude
sin x
of the current.
! ! 2
csc θ
11. sin3 y cos2 y dy 12. dθ
cot 2 θ The dc value of a periodic waveform f (t) with period p in the theory
! ! of signals is defined as
sin θ sec2 x
13. dθ 14. √ dx
1 + cos θ 1 + tan x ! p/2
1
! ! Fdc = f (t) dt.
3 + 4 csc2 x p −p/2
15. cos θ sin 2θ dθ 16. dx
cot 2 x
! ! In Exercises 34–37 find the dc value for the waveform.
17. sin5 x cos5 x dx 18. sin4 x dx ∗ 34. f (t) of Exercise 33
! ! ∗ 35. f (t) = A + B cos ωt , A , B , ω constants
tan3 x csc4 x
19. dx 20. dx
sec4 x cot 3 x ∗ 36. f (t) = sin2 ωt , ω constant
∗ 37. f (t) = t , 0 < t < 1, f (t + 1) = f (t)
21. Find the area of the region between the x -axis and y = sin x for
0 ≤ x ≤ π. !
∗∗ 38. Evaluate(1 + sin θ + sin2 θ + sin3 θ + · · ·) dθ .
∗ 22. Find the volume of the solid of revolution obtained by rotating the
region bounded by y = tan x , y = 0, and x = π/4 about the line
∗∗ 39. If n is a positive even integer, show that
y = −1.
! π& n/2−1 "
! 5 n/2 − 1# tan2r+1 x
∗ 23. Evaluate 1 − sin2 x dx . secn x dx = + C.
0
r=0
r 2r + 1
∗∗ 40. A set of functions f1 (x) , f2 (x) , . . . is said to be orthonormal over

In Exercises 24–31 evaluate the indefinite integral. an interval a ≤ x ≤ b if
! ! ! b /
cos3 θ 0, m != n
∗ 24. cot 4 z dz ∗ 25. dθ fn (x)fm (x) dx =
3 + sin θ 1, m = n.
a
! !
cos4 θ
∗ 26. dθ ∗ 27. sin4 x cos2 x dx Show that the set of functions
1 + sin θ
! ! 1 1 1
√ , √ sin x, √ cos x,
∗ 28. cos 6x cos 2x dx ∗ 29. cos2 2x sin 3x dx 2π π π
! ! 1 1 1
1 √ sin 2x, √ cos 2x, √ sin 3x, ...
∗ 30. dx ∗ 31. sec5 x dx π π π
sin x cos2 x
∗ 32. The power required to maintain current i(t) = im cos (ωt + φ1 ) , is orthonormal over the interval 0 ≤ x ≤ 2π . Orthonormality of
where im > 0 and φ1 are constants, through a resistor that has voltage these functions is the basis for Fourier series, a topic of fundamental
across its terminals V (t) = Vm cos (ωt + φ2 ) , where Vm > 0 and importance in all branches of engineering.
8.4 Trigonometric Substitutions

Physical and geometric problems arising√ from circles,
√ ellipses, and hyperbolas
√ often give rise to
integrals involving the square roots a 2 − b2 x 2 , a 2 + b2 x 2 , and b2 x 2 − a 2 . Trigono-
metric substitutions replace them with integrals involving trigonometric
√ functions.
Consider first two integrals that involve square roots of the form a 2 − b2 x 2 :
! !
1 1
(i) √ dx and (ii) √ dx.
1− x2 4 − 9x 2
Integral (i) is evident from differentiation of the inverse trigonometric functions in Section 3.10,
!
1
√ dx = Sin −1 x + C.
1 − x2
(See also formula 8.1o) We now show that a substitution can be made√on this integral that is
applicable to much more difficult problems
√ involving the square root a 2 −√b2 x 2 , including
(ii). If instead of the square root 1 − x √ 2 in (i), we had the square root 1 − sin2 θ , we
could immediately simplify the latter root to cos2 θ = | cos θ | . This prompts us to make the
substitution x = sin θ , from which we have dx = cos θ dθ . Integral (i) then becomes
! !
1 1
√ dx = cos θ dθ.
1− x2 | cos θ |
To eliminate the absolute values, we need to know whether cos θ is positive or negative. But
this means that we must know the possible values of θ . There is a problem. The equation
x = sin θ does not really define θ ; for any given x , we do not have a unique θ , but an infinite
number of possibilities. We must therefore restrict the values of θ , and we do this by specifying
that θ = Sin −1 x . In other words, although we have used the equation x = sin θ to change
variables, and will continue to do so, it is really the equation θ = Sin −1 x that properly defines
the change. Since θ so defined must lie in the interval −π/2 ≤ θ ≤ π/2 (the principal values
of the inverse sine function), it follows that cos θ ≥ 0, and absolute values may be dropped.
Consequently,
! ! !
1 cos θ
√ dx = dθ = dθ = θ + C = Sin −1 x + C.
1− x2 | cos θ |
√
little use since the square root 4 − 9 sin2 θ
For integral (ii), the substitution x = sin θ is of √
does
√ not simplify. If, √ however, the square root were 4 − 4 sin2 θ , it would immediately reduce
to 4 − 4 sin2 θ = 4 cos2 θ = 2| cos θ | . We can obtain this result if the substitution is mod-
ified to x = (2/3) sin θ [or, more properly, θ = Sin −1 (3x/2) ]. Since dx = (2/3) cos θ dθ ,
we have ! ! !
1 (2/3) cos θ 2 cos θ
√ dx = √ dθ = dθ .
4 − 9x 2 4 − 4 sin2 θ 3 2| cos θ |
Once again absolute values may be dropped because −π/2 ≤ θ ≤ π/2,
! ! " #
1 1 θ 1 3x
√ dx = dθ = +C = Sin −1 + C.
4 − 9x 2 3 3 3 2
√
For integrals containing square roots of the form a 2 + b2 x 2 , consider the pair of integrals
! !
1 1
(iii) dx (iv) dx.
1 + x2 (3 + 5x 2 )3/2
Integral (iii) obviously has answer Tan −1 x + C . Alternatively, if we had 1 + tan2 θ instead of
1 + x 2 , the two terms in the denominator would simplify to the one term sec2 θ . We therefore
substitute x = tan θ or, more properly, θ = Tan −1 x , from which we have dx = sec2 θ dθ .
With this substitution, we obtain
! ! !
1 sec2 θ
dx = dθ = dθ = θ + C = Tan −1 x + C.
1 + x2 sec2 θ
The substitution x = tan θ in (iv) yields (3 + 5x 2 )3/2 = (3 + 5 tan2 θ )3/2 . For simplifi-
cation, we need a 3 in front of the tan2 θ and not a 5. This can be accomplished by modification
of our substitution to 6
3
x = tan θ,
5
from which 6
3
dx = sec2 θ dθ.
5
Then
! ! 6 !
1 1 3 1 sec2 θ
dx = sec2 θ dθ = dθ√
(3 + 5x 2 )3/2 (3 + 3 tan2 θ )3/2 5 3 5 (sec2 θ )3/2
! !
1 1 1 1
= √ dθ = √ cos θ dθ = √ sin θ + C.
3 5 sec θ 3 5 3 5
To express
√ √ sin θ in terms of x , we draw the triangle in Figure√ 8.1 to fit the change of variable
FIGURE 8.1 Triangle to fit
x=
√
3/5 tan θ
5 x/ 3 = tan θ . Since the hypotenuse of the triangle is 3 + 5x 2 , we obtain
! √
1 1 5x x
2 3 2
dx = √ √ +C = √ + C.
(3 + 5x ) /
3 5 3 + 5x 2 3 3 + 5x 2
5x
√
For integrals containing square roots of the form b2 x 2 − a 2 , consider
3 !
1
√ dx.
2
x −4
√ √
If we had 4 sec2 θ − 4 instead of x 2 − 4, an immediate simplification would occur. We
therefore substitute x = 2 sec θ , from which dx = 2 sec θ tan θ dθ . Once again the real
substitution is θ = Sec −1 (x/2) , in which case θ is an angle in either the first or third quadrant.
With this change,
& & √
x2 − 4 = 4 sec2 θ − 4 = 4 tan2 θ = 2| tan θ | = 2 tan θ
and
! ! !
1 2 sec θ tan θ
√ dx = dθ = sec θ dθ = ln | sec θ + tan θ | + C.
x2 − 4 2 tan θ
To express tan θ in terms of x , we draw

√ the triangle in Figure 8.2 to fit the change of variable
x = 2 sec θ x/2 = sec θ . Since the third side is x 2 − 4, we have
! 7 √ 7
1 7x x 2 − 4 77 7 & 7
7 7 2 − 47 + D,
√ dx = ln 7 + 7 + C = ln 7 x + x 7
x x2 − 4 72 2 7
where D = C − ln 2.
We have illustrated by examples that trigonometric
√ substitutions
√ can be useful
√ in the eval-
2 uation of integrals involving the square roots a 2 − b2 x 2 , a 2 + b2 x 2 , and b2 x 2 − a 2 .
Essentially, the method replaces terms under the square root by a perfect square and thereby rids
the integrand of the square root. To obtain the trigonometric substitution appropriate to each
square root, we suggest the following procedure. We have been working with the trigonometric
identities
1 − sin2 θ = cos2 θ,
1 + tan2 θ = sec2 θ,
sec2 θ − 1 = tan2 θ.
√
To determine the trigonometric
√ substitution appropriate to, say, a 2 − b2 x 2 , we mentally set
a = b = 1 and obtain 1 − x 2 . We note that 1√ −x 2 resembles 1 − sin2 θ , and immediately set
x = sin θ . To simplify the original square root a 2 − b2 x 2 , we then modify this substitution
to
) " #*
a −1 bx
x = sin θ or θ = Sin .
b a
√
Similarly, for the square root a 2 + b2 x 2 , we initially set x = tan θ , having mentally set
a = b = 1, and noted that 1 + x 2 resembles 1 + tan2 θ . We then modify the substitution to
) " #*
a −1 bx
x = tan θ or θ = Tan .
b a
√
Finally, for the square root b2 x 2 − a 2 , we initially set x = sec θ and then modify to
) " #*
a −1 bx
x = sec θ or θ = Sec .
b a
√ √ √
The square roots a 2 − b2 x 2 , a 2 + b2 x 2 , and b2 x 2 − a 2 are replaced, respectively,
by
√ √ √
a 2 − a 2 sin2 θ a 2 + a 2 tan2 θ a 2 sec2 θ − a 2
√ √ √
= a 2 cos2 θ = a 2 sec2 θ = a 2 tan2 θ
= a| cos θ | = a| sec θ | = a| tan θ |
= a cos θ ; = a sec θ ; = a tan θ.

In each case it is our choice of principal values for the inverse trigonometric functions that
enables us to neglect absolute values.
We now consider more complex examples.
EXAMPLE 8.11
! ! &
1
(a) dx (b) x 2 3 − x 2 dx
x 2 (9x 2 + 2)
SOLUTION
$√ % $√ %
(a) If we set x = 2/3 tan θ , then dx = 2/3 sec2 θ dθ , and
! ! 8√ 9
1 1 2
dx = " # sec2 θ dθ
x 2 (9x 2 + 2) 2 3
tan θ (2 tan2 θ + 2)
2
9
!
3 sec2 θ
= √ dθ
2 2 tan2 θ sec2 θ
FIGURE 8.3 Triangle to fit ! !
√ 3 3
x = ( 2/3) tan θ = √ cot 2 θ dθ = √ (csc2 θ − 1) dθ
2 2 2 2
3
= √ (− cot θ − θ ) + C
2 2
9x 2 + 2
3x :√ " #;
−3 2 3x
= √ + Tan−1 √ +C (Figure 8.3)
2 2 3x 2
" #
2 −1 3 −1 3x
= − √ Tan √ + C.
2x 2 2 2
√ √
(b) If we set x = 3 sin θ , then dx = 3 cos θ dθ , and
! & ! & √
2
x 3− x2 dx = (3 sin2 θ ) 3 − 3 sin2 θ( 3 cos θ dθ)
! &
=9 sin2 θ 1 − sin2 θ cos θ dθ
!
=9 sin2 θ cos2 θ dθ
!
9
= sin2 2θ dθ (by 8.5b)
4
! " #
9 1 − cos 4θ
= dθ (by 8.5d)
FIGURE 8.4 Triangle to fit 4 2
√ " #
x= 3 sin θ 9 1
= θ − sin 4θ + C (by 8.4d)
8 4
" #
9 x 9
3 = Sin−1 √ − (2 sin 2θ cos 2θ ) + C (by 8.5b)
x 8 3 32
" #
9 x 9
= Sin−1 √ − (2 sin θ cos θ )(1 − 2 sin2 θ ) + C (by 8.5b,d)
8 3 16
3− x2
" # " #√ " #
9 −1 x 9 x 3 − x2 2x 2
= Sin √ − √ √ 1− +C (Figure 8.4)
8 3 8 3 3 3
" #
9 x x&
= Sin −1
√ − 3 − x 2 (3 − 2x 2 ) + C.
8 3 8
EXAMPLE 8.12
Use a trigonometric substitution to evaluate the integral in Example 8.4.
SOLUTION If we set x = 3 tan θ , then dx = 3 sec2 θ dθ , and
! !
x3 27 tan3 θ
√ dx = √ 3 sec2 θ dθ
9 + x2 9 + 9 tan2 θ
!
tan3 θ sec2 θ
= 27 dθ
sec θ
!
= 27 tan θ (sec2 θ − 1) sec θ dθ (by 8.7)

!
x = 3 tan θ = 27 (sec3 θ tan θ − sec θ tan θ ) dθ
" #
1 3
= 27 sec θ − sec θ +C (by 8.6b)
9 + x2 3
x 8√ 93 √
9 + x2 9 + x2
=9 − 27 +C (Figure 8.5)
3 3
3
1 &
= (9 + x 2 )3/2 − 9 9 + x 2 + C.
3
EXAMPLE 8.13
The Application Preview posed the problem of finding the path y = f (x) followed by the
Application Preview point midway between the rear wheels of a trailer given that the pivot point of the tractor and
Revisited trailer follows the curve Y = g(X) (Figure 8.6).
FIGURE 8.6 Path followed by point midway between rear wheels of tractor-trailer
y
Y = g(X)
Tractor
(X, Y)
Trailer
y = f(x)
Pivot point
x
(x, y)
SOLUTION The key is to notice that the line joining (x, y) to (X, Y ) is always tangent to
y = f (x) . Consequently,
dy Y −y g(X) − y
= = .
dx X−x X−x
This is a differential equation that must be solved for y = f (x) subject to the initial position of
the midpoint between the rear wheels on the x -axis. To solve this equation we must eliminate
X . This can be done by letting L be the fixed distance between (x, y) and (X, Y ) , in which
case
L2 = (X − x)2 + (Y − y)2 = (X − x)2 + [g(X) − y ]2 .
Given g(X) , we can theoretically solve this equation for X in terms of x , y , and L , and
substitute into the differential equation to express dy/dx in terms of x and y . To do so for
given g(X) could be a formidable task if g(X) is at all complicated. We shall consider only
the simplest possible situation in which the pivot point begins at the origin and moves up the
y -axis. In√this case, X is always zero, and L2 = x 2 + (Y − y)2 . From this equation,
Y − y = L2 − x 2 , and the differential equation for y = f (x) is
√
dy Y −y L2 − x 2
= = .
dx X−x −x
Integration with respect to x gives
! √ 2
L − x2
y =− dx.
x
We now set x = L sin θ and dx = L cos θ dθ .
! ! !
L cos θ 1 − sin2 θ
y =− L cos θ dθ = −L dθ = L (sin θ − csc θ ) dθ
L sin θ sin θ
= L[− cos θ − ln | csc θ − cot θ |] + C
8√ 7 √ 79
L2 − x 2 7L L2 − x 2 77
7
= −L + ln 7 − 7 +C
L 7x x 7
7 7
& 7 L − √L2 − x 2 7
7 7
= − L2 − x 2 − L ln 7 7+C
7 x 7
7 7
& 7 L − √L2 − x 2 L + √L2 − x 2 7
7 7
= − L2 − x 2 − L ln 7 √ 7+C
7 x L + L2 − x 2 7
& 7 7
7 x 7
= − L2 − x 2 − L ln 77 √ 7+C
L + L2 − x 2 7
7 7
7 L + √L2 − x 2 7 &
7 7
= L ln 7 7 − L2 − x 2 + C.
7 x 7
Since y = 0 when x = L , it follows that C = 0, and
8 √ 9
L+ L2 − x 2 &
y = L ln − L2 − x 2 .
x
EXERCISES 8.4 ! !
1 x+5
In Exercises 1–20 evaluate the indefinite integral. 9. √ dx 10. dx
! ! x2 − 5 10x 2 + 2
1 1
1. √ dx 2. √ dx ! ! √
x 2x 2 − 4 9 − 5x 2 1 4 − x2
11. √ dx 12. dx
! ! x x2 + 3 x
1 1
3. dx 4. √ dx ! ! √
10 + x 2 x2 4 − x2 x2 x 2 − 16
! & ! & 13. dx 14. dx
(2 − 9x 2 )3/2 x2
5. 7 − x 2 dx 6. x 5x 2 + 3 dx ! !
1 1
! ! 15. √ dx 16. √ dx
& 1 x2 2x 2 +7 x3 x2 − 4
7. x 3 4 + x 2 dx 8. dx
1 − x2
! √ ! ! !
9 − z2 y3 1 1
17. dz 18. & dy ∗ 33. dx ∗ 34. dx
z4 y2 + 4 ! &x − x3 ! & x 3 (4x 2 − 1)3/2
! ! 2
1 x +2 ∗ 35. x 2 − 4 dx ∗ 36. 1 + 3x 2 dx
19. dx 20. dx
(4x 2 − 9)3/2 x3 + x !
x2
∗ 37. √ dx
x2 − 5
∗ 21. Show that if a > 0, then
∗ 38. Find the length of the curve 8y 2 = x 2 (1 − x 2 ) .
! 7 7 ∗ 39. Find the length of the parabola y = x 2 from (0, 0) to (1, 1) .
1 1 7a + x 7
dx = ln 77 7 + C.
∗ 40. At what distance from the centre of a circle should a line be drawn
a2 − x 2 2a a − x7
in order that the second moment of area of the circle about that line will
2 2 2 2 2 2 be equal to twice the second moment of the circle about a line through
∗ 22. Verify that the area inside the ellipse b x +a y = a b is π ab .
its centre?
∗ 23. Find the area of the region common to the circles x 2 + y 2 = 4 ∗ 41. When water in the soil moves to the surface of the soil, depth z
and x 2 + y 2 = 4x . below the surface is related to suction head x by the equation
∗ 24. Given that the equation 2x 2 + y 2 − 2xy + 4x − 4y + 3 = 0 !
1
describes an ellipse, find its area. z(x) = − dx,
1 + V /K
∗ 25. A boy initially at O (figure below) walks along the edge of a pier
(the y -axis) towing his sailboat by a string of length L . where V > 0 is a constant, and K = k/(cx 2 + 1) with k > 0
and c > 0 being constants. Show that if z(0) = H , then
(a) If the boat starts at Q and the string always remains straight,
6
show that the equation of the curved path y = f (x) fol- k Vc
lowed by the boat must satisfy the differential equation z(x) = H − √ Tan −1 x,
V c(k + V ) k+V
√
dy L2 − x 2 whereas if z(Hw − L) = L , then
=− .
dx x : 6
k −1 Vc
(b) Find y = f (x) . z(x) = L + √ Tan (Hw − L)
V c(k + V ) k+V
6 ;
y Vc
−1
− Tan x .
k+V
∗ 42. When a flexible cable of constant mass per unit length ρ hangs
between two fixed points A and B (figure below), the shape y = f (x)
L
of the cable must satisfy the differential equation
Pier (x, y) < " #2
d 2y ρg dy
2
= 1 + ,
dx H dx
Q
O
Shoreline x where g > 0 and H > 0 are constants. If we set k = ρg/H and
L
2 2 2
p = dy/dx , then
∗ 26. The parabola x = y divides the circle x + y = 4 into two
parts. Find the second moment of area of the smaller part about the dp & 1 dp
x -axis. = k 1 + p2 or & = k.
dx 1+ p2 dx
∗ 27. Find the centroid of the region in Exercise 26.
∗ 28. Find the horizontal line that divides the ellipse b2 x 2 + a 2 y 2 = It follows that !
1
a 2 b2 into two parts so that the area of the lower part is twice the area & dp = kx + C.
of the upper part. 1 + p2
∗ 29. If a thin circular plate has radius 2 units and constant mass per unit (a) Evaluate the integral shown to find p = dy/dx .
area ρ , find its moment of inertia about any tangent line to its edge. (b) Integrate once more to find the shape of the cable (see also
∗ 30. A cylindrical oil can with horizontal axis has radius r and length Example 3.39 in Section 3.13.)
h . If the density of the oil is ρ , find the force on each end of the can y B
when it is full.
A
In Exercises 31–37 evaluate the indefinite integral.
! ! Lowest point
2x 4 − x 2 on cable
∗ 31. dx ∗ 32. (7 − x 2 )3/2 dx
2x 2 + 1 x
∗∗ 43. Repeat Exercise 21 of Section 7.6 if the log is circular with radius (b) Use part (a) to prove that
10 cm. 8
! 1
√ √ 9
∗∗ 44. Find the area inside the loop of the strophoid y 2 (a + x) = 1 2 1 1+ 2
x 2 (a − x) . √ dx = 1 − + ln .
0 x+1+ x2 + 1 2 2 2
∗∗ 45. (a) By rationalizing the denominator, show that
√ ∗∗ 46. Show that ! 6
1
1 x+1− x2 + 1 1+x
√ = . dx = π.
x+1+ x2 + 1 2x −1 1−x
8.5 Completing the Square and Trigonometric Substitutions

√
Trigonometric substitutions
√ √ reduce integrals containing square roots of the form a 2 − b2 x 2 ,
a 2 + b2 x √
2 , and b2 x 2 − a 2 to trigonometric integrals. For integrals containing square roots
of the form ax 2 + bx + c we can again reduce the integral to a trigonometric integral by a
trigonometric substitution, if we first complete the square.
Consider the integral
!
1
dx.
(x 2 + 2x + 5)3/2
If we complete the square in the denominator, we obtain
! !
1 1
dx = dx.
(x 2 + 2x + 5)3/2 [(x + 1)2 + 4]3/2
Had the denominator been (x 2 + 4)3/2 , we would have set x = 2 tan θ . It is natural, then, for
the denominator [(x + 1)2 + 4]3/2 to set
x + 1 = 2 tan θ,
in which case dx = 2 sec2 θ dθ , and

x + 1 = 2 tan θ ! ! !
1 1 2 sec2 θ
dx = 2 sec2 θ dθ = dθ
(x 2 + 2x + 5)3/2 (4 tan2 θ + 4) 3/2 8 sec3 θ
!
x 2 + 2x +5 1 1
= cos θ dθ = sin θ + C
x+1 4 4
x+1
= √ + C. (Figure 8.7)
2
4 x + 2x + 5
2
Consider another example,

! &
4 + x − x 2 dx.
Completion of the square leads to

! & ! &
4 + x − x 2 dx = −(x − 1/2)2 + 17/4 dx.
If we now set √
1 17
x− = sin θ,
2 2
8.5 Completing the Square and Trigonometric Substitutions 515
then √
17
dx = cos θ dθ,
2
and
! & ! 6 √ ! √ √
17 2 17 17 17 17
4+x− x2 dx = − sin θ + cos θ dθ = cos θ cos θ dθ
4 4 2 2 2
! ! " #
17 17 1 + cos 2θ
= cos2 θ dθ = dθ
FIGURE 8.8
√ Triangle to fit 4 4 2
2x − 1 = 17 sin θ " #
17 1 17
= θ + sin 2θ + C = (θ + sin θ cos θ ) + C
8 2 8
17 " # ) * √
17 −1 2x − 1 17 2x − 1 2 4 + x − x2
2x − 1 = Sin √ + √ √ + C (Figure 8.8)
8 17 8 17 17
" # &
17 −1 2x − 1 1
= Sin √ + (2x − 1) 4 + x − x 2 + C.
8 17 4
17 − (2x − 1)2
As a final illustrative example, we consider
= 2 4 + x − x2 ! ! !
x 1 x 1 x
√ dx = √ 6 dx = √ <" #2 dx.
2x 2 + 3x − 6 2 2 3x 2 3 57
x + −3 x+ −
2 4 16
If we set √
3 57
x+ = sec θ,
4 4
then √
57
dx = sec θ tan θ dθ
4
and
! √ ! √
x 1
( 57/4) sec θ − 3/4 57
√ dx = √ & sec θ tan θ dθ
2
2x + 3x − 6 2 2
(57/16) sec θ − 57/16 4
! √ √
1 ( 57/4) sec θ − 3/4 57
= √ √ sec θ tan θ dθ
2 ( 57/4) tan θ 4
! 8√ 9
1 57 2 3
= √ sec θ − sec θ dθ
2 4 4
√
57 3
FIGURE 8.9
√ Triangle to fit = √ tan θ − √ ln | sec θ + tan θ | + C
4x + 3 = 57 sec θ 4 2 4 2
√ : √ √ ;
57 2 2 2x 2 + 3x − 6
4x + 3
= √ √
(4x + 3)2 − 57 4 2 57
7 √ √ 7
= 2 2 2x 2 + 3x − 6 7 4x + 3 2 + 3x − 6 7
3 7 2 2 2 x 7
− √ ln 7 √ + √ 7 + C (Figure 8.9)
57 4 2 7 57 57 7
√ 7 7
2x 2 + 3x − 6 3 7 √ & 7
= − √ ln 74x + 3 + 2 2 2x 2 + 3x − 67 + D,
2 4 2
where
3
D = C + √ ln 57.
8 2
We have shown that completing the square is the natural generalization of the technique
of
√ trigonometric substitutions. It reduces integrals containing square roots of the form
ax 2 + bx + c to trigonometric integrals.
EXAMPLE 8.14
Find the volume of a donut.
SOLUTION A donut is generated (as the volume of the solid of revolution) when the circle
(x − a)2 + y 2 = b2 (a > b)
is rotated around the y -axis (Figure 8.10). Cylindrical shells for vertical rectangles yield the
FIGURE 8.10 Donut
volume by shells
volume generated by the upper semicircle, which can be doubled to give
y
! a+b
(x − a)2 + y2 = b2
V =2 2π xy dx.
a−b
x y &
If we solve the equation of the circle for y = ± b2 − (x − a)2 , then
a−b a a+b x
! a+b &
V =2 2π x b2 − (x − a)2 dx.
a−b
We now set x − a = b sin θ , in which case dx = b cos θ dθ , and

! π/2
V = 4π (a + b sin θ )b cos θ b cos θ dθ
−π/2
! π/2
= 4π b 2 (a cos2 θ + b cos2 θ sin θ ) dθ
−π/2
! π/2 ) " # *
2 1 + cos 2θ 2
= 4π b a + b cos θ sin θ dθ
−π/2 2
/ " # 0π/2
a 1 b
= 4π b 2 θ + sin 2θ − cos3 θ
2 2 3 −π/2
= ' π (>
2 a π
= 4π b + = 2π 2 ab2 .
2 2 2
With horizontal rectangles (Figure 8.11), the washer method yields
FIGURE 8.11 Volume of
donut by washers ! b
y V =2 (π x22 − π x12 ) dy.
0
x2
&
x1 We can solve the equation of the circle for x = a ± b2 − y 2 ; hence,
y & &
x
x1 = a − b2 − y 2 , x2 = a + b 2 − y 2 .
a
Thus )'
(x − a)2 + y 2 = b2 ! b & (2 ' & (2 *
V = 2π 2
a+ b −y 2 2
− a− b −y 2 dy,
0
and this simplifies to

! b &
V = 2π 4a b2 − y 2 dy.
0
If we set y = b sin θ , then dy = b cos θ dθ , and

! π/2
V = 8π a b cos θ b cos θ dθ
0
! π/2 " #
1 + cos 2θ
= 8π ab2 dθ
0 2
/ 0π/2
1
= 4π ab2 θ + sin 2θ
2 0
= 4π ab (π/2) = 2π ab2 .
2 2
EXERCISES 8.5
In Exercises 1–10 evaluate the indefinite integral. ∗ 12. One of the gates in a dam is circular with radius 1 m. If the gate
! ! is closed and the surface of the water is 3 m above the top of the gate,
x 1 find the force due to water pressure on the gate.
1. √ dx 2. √ dx
27 + 6x − x 2 x 2 + 2x + 2 ∗ 13. Evaluate
! ! !
1 1 1
3. dy 4. dx dx
(y 2 + 4y)3/2 3x − x2 −4 3x − x 2
! √ ! (a) by completing the square, and setting x − 3/2 = (3/2) sin θ and
x 2 + 2x − 3 x
5. dx 6. dx (b) by multiplying numerator and denominator by −1, completing the
x+1 (4x − x 2 )3/2 square, and setting x − 3/2 = (3/2) sec θ . (c) Explain the difference
! & !
2x − 3 in the two answers.
7. −y 2 + 6y dy 8. dx
x2 + 6x + 13
! !
5 − 4x 1 In Exercises 14–16 evaluate the integral.
9. √ dx ∗ 10. & dx
12x − 4x 2 −8 x 6 + 4 ln x + (ln x)2 ! & !
1
∗ 11. Use the substitution z = 1/x to evaluate ∗ 14. x 2 − 2x − 3 dx ∗∗ 15. √ dx
x 2x − x 2
! !
1 1
√ dx. ∗∗ 16. √ dx
2
x x + 6x + 3 (2x + 5) 2x − 3 + 8x − 12
8.6 Partial Fractions

Partial fractions is a method that we apply to rational functions, integrals of the form
!
N (x)
dx, (8.8)
D(x)
where N (x) and D(x) are polynomials in x , and the degree of N (x) is less than the degree
of D(x) . When the degree of N (x) is greater than or equal to that of D(x) , we divide D(x)
into N (x) . For example, in the integral
!
x 4 + 4x 3 + 2 x + 4
dx,
x3 + 1
the numerator has degree 4 and the denominator degree 3. By long division, we obtain
! ! " #
x 4 + 4x 3 + 2 x + 4 x
dx = x+4+ dx.
x3 + 1 x3 + 1
The first two terms on the right, namely, x + 4, can be integrated immediately, and partial
fractions can be applied to the remaining term, x/(x 3 + 1) .
As a general rule then: If the degree of N (x) is greater than or equal to the degree of D(x) ,
divide D(x) into N (x) to produce a quotient polynomial Q(x) and a remainder polynomial
R(x) of degree less than that of D(x) ; that is,
N (x) R(x)
= Q(x) + ,
D(x) D(x)
where deg R < deg D . Integral 8.8 can then be written in the form
! ! !
N (x) R(x)
dx = Q(x) dx + dx.
D(x) D(x)
The first integral on the right is trivial. This leaves integration of the rational function
!
R(x)
dx, (8.9)
D(x)
where deg R < deg D .
To use partial fractions on integral 8.9 we must factor the denominator D(x) . In Section
1.2, we stated that every real polynomial, and specifically D(x) , can be factored into real linear
factors ax + b and irreducible real quadratic factors ax 2 + bx + c . A quadratic factor is
irreducible if b2 − 4ac < 0. In the complete factorization of D(x) , these factors may or may
not be repeated; that is, the factorization of D(x) contains terms of the form
(ax + b)n and (ax 2 + bx + c)n ,
where n ≥ 1 is an integer. When n = 1 the factor is nonrepeated, and when n > 1, it is
repeated (it has multiplicity n ). For example, in the factorization
D(x) = (x − 1)(2x + 1)3 (3x 2 + 4x + 5)(x 2 + 1)2 ,
x − 1 and 3x 2 + 4x + 5 are nonrepeated, and 2x + 1 and x 2 + 1 are repeated.
It is worthwhile noting here that D(x) = (x − 1)(2x − 2) does not have two distinct
linear factors since we can write D(x) = 2(x − 1)2 . Likewise, D(x) = (x 2 + 1)(3x 2 + 3)
does not have distinct quadratic factors; it has a repeated quadratic factor D(x) = 3(x 2 + 1)2 .
Distinct factors must have different zeros.
Having factored D(x) , we can separate the rational function in 8.9 into fractional compo-
nents. We call this the partial fraction decomposition of the integrand. We illustrate with the
rational function (x 2 + 2)/[(x − 1)(2x + 1)3 (3x 2 + 4x + 5)(x 2 + 1)2 ], and then state gen-
eral rules for all decompositions. The partial fraction decomposition of this particular rational
function is
x2 + 2 A B C D
= + + +
(x − 1)(2x + 1)3 (3x 2+ 4x + 5)(x 2 + 1)2 x−1 2x + 1 (2x + 1)2 (2x + 1)3
Ex + F Gx + H Ix + J
+ + + 2 ,
3x 2 + 4x + 5 x2 + 1 (x + 1)2
where A , B, . . . , J are constants. The first term corresponds to the nonrepeated linear factor
x − 1, the next three correspond to the repeated linear factor 2x + 1, the fifth term corresponds
to the nonrepeated quadratic factor 3x 2 + 4x + 5, and the last two terms result from the repeated
quadratic x 2 + 1.
Let us now state general rules for the partial fraction decomposition of rational functions
R(x)/D(x) . There are three rules:
1. For each repeated or nonrepeated linear factor (ax + b)n in D(x) , include the following
terms in the decomposition:
A1 A2 An
+ 2
+ ··· + . (8.10a)
ax + b (ax + b) (ax + b)n
The number of terms corresponds to the power n on (ax + b)n .
2. For each repeated or nonrepeated irreducible quadratic factor (ax 2 + bx + c)n in D(x) ,
include the following terms in the decomposition:
B 1 x + C1 B2 x + C2 Bn x + Cn
2
+ 2 2
+ ··· + . (8.10b)
ax + bx + c (ax + bx + c) (ax 2 + bx + c)n
Again the number of terms corresponds to the power n on (ax 2 + bx + c)n .

3. The complete decomposition is the sum of all terms in 8.10a and all terms in 8.10b.
What is important here is to realize that all terms in expression 8.10a are immediately integrable.
Terms in expression 8.10b are unlikely to be mental integrations, but if b = 0, they can be
integrated with a trigonometric substitution. If b != 0, they can be integrated by completing the
square and using a trigonometric substitution.
The following examples will clarify the above rules.
EXAMPLE 8.15
What form do partial fraction decompositions for the following rational functions take?
x 2 + 2x + 3 x 2 + 3x − 1
(a) (b)
3x 3 − x 2 − 3x + 1 x4 + x3 + x2 + x
x 3 + 3x 2 − x 3x 5 − 1
(c) (d)
x 5 + x 4 + 2x 3 + 2x 2 + x + 1 (3x 2 + 5)(x 2 + 2x + 3)(2x − 1)2
SOLUTION
(a) Since 3x 3 − x 2 − 3x + 1 = (3x − 1)(x − 1)(x + 1) , we have nonrepeated linear
factors. Rules 1 and 3 give the partial fraction decomposition
x 2 + 2x + 3 A B C
= + + .
3x 3 − x 2 − 3x + 1 3x − 1 x−1 x+1
(b) The factorization x 4 + x 3 + x 2 + x = x(x + 1)(x 2 + 1) has two nonrepeated linear
factors and a nonrepeated quadratic factor. The three rules lead to
x 2 + 3x − 1 A B Cx + D
= + + 2 .
x4 3
+x +x +x 2 x x+1 x +1
(c) Since x 5 + x 4 + 2x 3 + 2x 2 + x + 1 = (x + 1)(x 2 + 1)2 , the partial fraction
decomposition is
x 3 + 3x 2 − x A Cx + D Ex + F
5 4 3 2
= + 2 + 2 .
x + x + 2x + 2x + x + 1 x+1 x +1 (x + 1)2
(d) Since x 2 + 2x + 3 can be factored no further, the decomposition is
3x 5 − 1 Ax + B Cx + D E F
= + 2 + + .
(3x 2 + 5)(x 2 + 2x + 3)(2x − 1)2 3x 2 + 5 x + 2x + 3 2x − 1 (2x − 1)2
To illustrate how to calculate coefficients in partial fraction decompositions, we use part (b) of
Example 8.15. We bring the right side of the decomposition to a common denominator,
x 2 + 3x − 1 A B Cx + D
4 3 2
= + + 2
x +x +x +x x x+1 x +1
A(x + 1)(x 2 + 1) + Bx(x 2 + 1) + x(x + 1)(Cx + D)
= ,
x4 + x3 + x2 + x
and equate numerators,
x 2 + 3x − 1 = A(x + 1)(x 2 + 1) + Bx(x 2 + 1) + x(x + 1)(Cx + D).
There are two methods for finding the constants A , B , C , and D .

Method 1. First, we gather together terms in the various powers of x on the right side of the
equation:
x 2 + 3x − 1 = (A + B + C)x 3 + (A + C + D)x 2 + (A + B + D)x + A.
Now, Exercise 35 in Section 3.8 states that two polynomials of the same degree can be equal for
all values of x if and only if coefficients of corresponding powers of x are identical. Since we
have equal cubic polynomials in the equation above, we equate coefficients of x 3 , x 2 , x , and
x 0 (meaning terms with no x ’s):
x3 : 0 = A + B + C,
x2 : 1 = A + C + D,
x : 3 = A + B + D,
x0 : −1 = A.
The solution of these four linear equations in four unknowns is A = −1, B = 3/2, C = −1/2,
and D = 5/2. Hence, the partial fraction decomposition of (x 2 + 3x − 1)/(x 4 + x 3 + x 2 + x)
is
x 2 + 3x − 1 1 3/2 −x/2 + 5/2
=− + + .
x4 + x3 + x2 + x x x+1 x2 + 1
Method 2. In this method we substitute convenient values of x into the equation
x 2 + 3x − 1 = A(x + 1)(x 2 + 1) + Bx(x 2 + 1) + x(x + 1)(Cx + D).
Clearly, x = 0 is most convenient, since it yields the value of A :
−1 = A(1)(1);
that is, A = −1. Convenient also is x = −1:
(−1)2 + 3(−1) − 1 = B(−1)(2);
it gives B = 3/2. The values x = 1 and x = 2 yield the equations
1 + 3(1) − 1 = A(2)(2) + B(1)(2) + 1(2)(C + D),
4 + 3(2) − 1 = A(3)(5) + B(2)(5) + 2(3)(2C + D).
When A = −1 and B = 3/2 are substituted into these, the resulting equations are
C + D = 2,
4C + 2D = 3.
The solution of these is C = −1/2 and D = 5/2.

It is also possible to use a combination of methods 1 and 2 for finding coefficients in a partial
fraction decomposition; that is, substitute some values of x and equate some coefficients. The
total number must be equal to the number of unknown coefficients in the decomposition.
Once we have completed the partial fraction decomposition of a rational function, we can
integrate the function by finding antiderivatives of the component fractions. For example, with
the partial fraction decomposition of (x 2 + 3x − 1)/(x 4 + x 3 + x 2 + x) , it is very simple to
find its indefinite integral,
! ! " #
x 2 + 3x − 1 1 3/2 −x/2 + 5/2
dx = − + + dx
x4 + x3 + x2 + x x x+1 x2 + 1
! ! ! !
1 3 1 1 x 5 1
=− dx + dx − dx + dx
x 2 x+1 2 x2 + 1 2 x2 + 1
3 1 5
= − ln |x| + ln |x + 1| − ln (x 2 + 1) + Tan−1 x + C.
2 4 2
Other integrations by partial fraction decompositions are illustrated in the following
example.
EXAMPLE 8.16
! !
x x2 + 1
(a) 4 2
dx (b) 3 2
dx
! x + 6x + 5 ! 2x 2− 5x + 4x − 1
1 x
(c) dx (d) dx
x3 + 8 x3 + 8
SOLUTION
(a) Since x 4 + 6x 2 + 5 = (x 2 + 1)(x 2 + 5) , the partial fraction decomposition of
x/(x 4 + 6x 2 + 5) has the form
x Ax + B Cx + D
= 2 + 2 .
x4 2
+ 6x + 5 x +1 x +5
When we bring the right side to a common denominator and equate numerators, the
result is
x = (Ax + B)(x 2 + 5) + (Cx + D)(x 2 + 1).

We now multiply out the right side and equate coefficients of like powers of x ,
x3 : 0 = A + C,
x2 : 0 = B + D,
x : 1 = 5A + C,
x0 : 0 = 5B + D.
The solution of these equations is A = 1/4, B = 0, C = −1/4, and D = 0, and

therefore
! ! " #
x x/4 −x/4
dx = + dx
x + 6x 2 + 5
4 x2 + 1 x2 + 5
1 1
= ln (x 2 + 1) − ln (x 2 + 5) + C.
8 8
(b) Since 2x 3 − 5x 2 + 4x − 1 = (2x − 1)(x − 1)2 , the partial fraction decomposition

takes the form
x2 + 1 A B C
= + +
2x 3 2
− 5x + 4x − 1 2x − 1 x−1 (x − 1)2
A(x − 1)2 + B(x − 1)(2x − 1) + C(2x − 1)
= .
(2x − 1)(x − 1)2
When we equate numerators,
x 2 + 1 = A(x − 1)2 + B(x − 1)(2x − 1) + C(2x − 1).
We now set x = 0, x = 1, and x = 1/2:
x =0: 1 = A + B − C,
x =1: 2 = C(1),
x = 1/2 : 5/4 = A(1/4).
These give A = 5, B = −2, and C = 2, and therefore

! ! ) *
x2 + 1 5 2 2
dx = − + dx
2x 3 − 5x 2 + 4x − 1 2x − 1 x−1 (x − 1)2
5 2
= ln |2x − 1| − 2 ln |x − 1| − + C.
2 x−1
(c) Since x 3 + 8 = (x + 2)(x 2 − 2x + 4) , we set
1 1 A Bx + C
= = + 2
x3 +8 (x + 2)(x 2 − 2 x + 4) x+2 x − 2x + 4
A(x 2 − 2x + 4) + (Bx + C)(x + 2)
= ,
(x + 2)(x 2 − 2x + 4)
and now equate numerators:
1 = A(x 2 − 2x + 4) + (Bx + C)(x + 2).
We set x = −2 and equate coefficients of x 2 and 1:
x = −2 : 1 = 12A,
x2 : 0 = A + B,
1 : 1 = 4A + 2C.
These give A = 1/12, B = −1/12, and C = 1/3, and therefore

! ! " #
1 1/12
−x/12 + 1/3
dx = + 2 dx
x3 + 8 x+2 x − 2x + 4
!
1 1 4−x
= ln |x + 2| + dx.
12 12 (x − 1)2 + 3
√
For the√ integral on the right, we set x − 1 = 3 tan θ , in which case
dx = 3 sec2 θ dθ , and
! ! √
1 1 1 4 − (1 + 3 tan θ ) √
dx = ln |x + 2| + 3 sec2 θ dθ
x3 + 8 12 12 3 sec2 θ
! ' √ (
1 1
= ln |x + 2| + √ 3− 3 tan θ dθ
12 12 3
FIGURE 8.12
√ Triangle to
1 1 ' √ (
fit x − 1 = 3 tan θ = ln |x + 2| + √ 3θ + 3 ln | cos θ | + C
12 12 3
√ " # 7 √ 7
x 2 − 2x + 4 1 3 x−1 17 3 7
7 7
x−1 = ln |x + 2| + Tan−1 √ + ln 7 √ 7+C
12 12 3 12 7 x − 2x + 4 7
2
3 (Figure 8.12)
√ " #
1 3 x−1 1
= ln |x + 2| + Tan−1 √ − ln (x 2 − 2x + 4) + D,
12 12 3 24
where D = C + (1/24) ln 3.
(d) Do not be misled into partial fractions in this example; the rational function is imme-
diately integrable:
!
x2 1
dx = ln |x 3 + 8| + C.
x3 +8 3
There are a number of useful devices that can sometimes simplify calculating coefficients in
partial fraction decompositions. We indicate two of them here. First, if ax + b is a nonrepeated,
linear factor of the denominator of a partial fraction decomposition, then its coefficient in the
decomposition can be obtained by “covering up” this term and substituting x = −b/a into
what remains. For example, to find the coefficient A in
x A B
= + ,
(x − 1)(2x + 3) x−1 2x + 3
we cover up the x − 1 and substitute x = 1 into x/(2x + 3) . We obtain 1/(2 + 3) = 1/5.
This is A . To find B , we cover up 2x + 3 and substitute x = −3/2 into x/(x − 1) . The result
is B = −(3/2)/(−3/2 − 1) = 3/5. Thus,
x 1/5 3/5
= + .
(x − 1)(2x + 3) x−1 2x + 3
Remember, however, that cover up can be used only on nonrepeated, linear factors. It could be
used to find A , but not B or C , in
x2 + 3 A Bx + C
2
= + 2 .
(3x − 1)(x + 2) 3x − 1 x +2
Second, when the denominator consists of only one repeated linear factor, say (x + 3)4 ,
rewriting the numerator in powers of x + 3 can sometimes give the partial fraction decomposition
very quickly, especially when the numerator is a linear or quadratic polynomial. For example,
to find the partial fraction decomposition of (x − 4)/(x + 3)4 , we write
x−4 (x + 3) − 7 1 7
4
= 4
= 3
− .
(x + 3) (x + 3) (x + 3) (x + 3)4
Likewise,
x2 + 2 (x + 4)2 − 8x − 14 (x + 4)2 − 8(x + 4) + 18

= =
(x + 4)6 (x + 4)6 (x + 4)6
1 8 18
= − + .
(x + 4 )4 (x + 4 )5 (x + 4)6
We have a chemical problem to solve. Two substances A and B react to form a third
substance C in such a way that 2 grams of A react with 3 grams of B to produce 5 grams
of C. When 20 grams of A and 40 grams of B are originally brought together, 3 grams
of C are formed in the first hour. Our problem is to find the amount of C present in the
mixture at any time.
SOLUTION We need to assume something about the rate at which A and B combine to
give C. Consultation with chemical engineers suggests that for many chemical reactions,
the rate at which chemicals react is proportional to the amounts that are present in the
mixture; the more chemicals, the faster the reaction. To express this algebraically, we
let x be the number of grams of C at time t in the mixture. Some of this came from A
and some from B. Specifically, 2x/5 grams came from A and 3x/5 grams came from B.
This means that there are 20 − 2x/5 grams of A and 40 − 3x/5 grams of B remaining.
The rate at which C is formed is represented by its derivative dx/dt and because it is
proportional to the amounts of A and B present in the mixture, we write
" #" #
dx 2x 3x
= K 20 − 40 −
dt 5 5
2K
= (50 − x)(200 − 3x) = k(50 − x)(200 − 3x),
25
where we have set k = 2K/25. Notice that we multiplied the amounts of A and B
in the mixture rather than add them. By adding them, we would have the unacceptable
situation of dx/dt being greater than zero even when one of the reactants vanishes. Since
t = 0 when A and B are brought together, x(t) must also satisfy x(0) = 0. In addition,
x(1) = 3. We use partial fractions to write the differential equation in the form
" #
1 1/50 3/50
k dt = dx = − dx.
(50 − x)(200 − 3x) 50 − x 200 − 3x
It is separated, and solutions are defined implicitly by
1+ ,
kt + C = − ln (50 − x) + ln (200 − 3x) .
50
Absolute values are unnecessary because x cannot exceed 50. We now solve this equation
for x by writing
" #
200 − 3x
50(kt + C) = ln ,
50 − x
and exponentiating,
200 − 3x
= De50kt ,
50 − x
where D = e50C . Cross multiplying gives (50 − x)De50kt = 200 − 3x , and this can
200 − 50De50kt
be solved for x = . The initial condition x(0) = 0 requires D = 4, in
3 − De50kt
which case $ %
200 − 200e50kt 200 1 − e50kt
x = = grams .
3 − 4e50kt 3 − 4e50kt
Since x(1) = 3,
" #
200(1 − e50k ) 1 191
3 = *⇒ k = ln .
3 − 4e50k 50 188
Thus, the number of grams of C in the mixture after t hours is

FIGURE 8.13 Amount of
substance produced in chemical + ,
200 1 − et ln (191/188)
reaction x(t) = .
x
3 − 4et ln (191/188)
50 We could plot this function or draw it using techniques from Chapter 4. What is
interesting to note is that with only the differential equation and some physical reasoning,
we can get a very good idea of the shape of the graph. The graph begins at (0, 0) , and has
horizontal asymptote x = 50. The differential equation shows us that dx/dt , the slope
of the graph is a maximum at t = 0 when x is smallest, and decreases as x increases.
t The graph can therefore have no critical points or points of inflection. It must appear as
in Figure 8.13.
Nuclear scientists are approaching us this time. The volume (in cubic metres) of a plug-
flow-reactor to operate for 90% fractional conversion is given by the integral
! 0.022
1 (0.096 − x/2)3
V = dx.
393 0 (0.024 − x)(0.024 − x/2)2
Evaluation is often done numerically because it is claimed that the integration is too
complex. We are asked to evaluate the integral exactly so as to reduce errors in numerical
integration.
SOLUTION First, we consider the indefinite integral
! !
(0.096 − x/2)3 1 (0.192 − x)3
I = dx = dx.
(0.024 − x)(0.024 − x/2)2 2 (0.024 − x)(0.048 − x)2
Let us set a = 0.024 in which case partial fractions lead to
! ! ) *
1 (8a − x)3 1 343a 216a 2 324a
I = dx = 1+ − − dx
2 (a − x)(2a − x)2 2 a−x (2a − x)2 2a − x
) *
1 216a 2
= x − 343a ln |a − x| − + 324a ln |2a − x| + C.
2 2a − x
Consequently, the volume of the reactor should be

)
1
V = 0.022 − 343(0.024) ln |0.024 − 0.022|
2(393)
*
216(0.024)2
− + 324(0.024) ln |0.048 − 0.022|
0.048 − 0.022
= 0.023 m3 .
EXERCISES 8.6
In Exercises 1–16 evaluate the indefinite integral. (a) By separating the differential equation show that v(t) =
50(1 − e−t/15 )/(1 + e−t/15 ) .
! ! (b) Find the position of the car as a function of time.
x+2 1
1. dx 2. dy
x 2 − 2x + 1 y 3 + 3y 2 + 3y + 1
∗ 19. When a raindrop with mass m falls in air, it is acted on by gravity
! ! 2 and also by a force due to air resistance that is proportional to the
1 x + 2x − 4
3. dz 4. dx square of its instantaneous speed. According to Newton’s second law,
z3 + z x 2 − 2x − 8 the differential equation describing the velocity of the raindrop is
! !
x y+1
5. dx 6. dy dv
(x − 4)2 y 3 + y 2 − 6y m = mg − kv 2 ,
dt
! !
3x + 5 x3
7. dx 8. dx where g = 9.81, and k > 0 is a constant. Separate the differential
x3 − x2 −x+1 (x 2 + 2)2 equation to show that if the raindrop exits vertically downward from
! ! the cloud with velocity v0 , then
1 y2
9. dx 10. dy ) " # *
x2 −3 y2 + 3y + 2 V − v0 −2kV t/m
! ! V 1− e
z2 + 3z − 2 y 2 + 6y + 4 V + v0
v(t) = " # ,
11. dz 12. dy V − v0 −2kV t/m
z3 + 5z y 4 + 5y 2 + 4 1+ e
! ! V + v0
x x2 + 3 √
13. dx 14. dx where V = mg/k . Can you interpret V physically?
x 4 + 7x 2 + 6 x4 + x2 − 2
! ∗ 20. If we wish to know the velocity of the raindrop in Exercise 19 as it
3t + 4
15. dt strikes the earth, it is preferable to find velocity as a function of distance
t4 − 3t 3 + 3t 2 −t
fallen (instead of time). By expressing the differential equation in the
! form
x3 + 6
16. dx dv
x 4 + 2x 3 − 3x 2 − 4x + 4 mv = mg − kv 2 ,
dy
∗ 17. Find the length of the curve y = ln (1 − x 2 ) from x = 0 to where y is distance fallen by the raindrop, find the velocity of the
x = 1/2. raindrop when it strikes the earth if it falls from height h .
∗ 18. A car of mass 1500 kg starts from rest at an intersection and moves ∗ 21. The exponential growth of bacteria in the two exercises of Sec-
in the positive x -direction. The engine exerts a constant force of mag- tion 5.5 is unrealistic in the long term when there is a limited food
nitude 2500 N, and air friction causes a resistive force whose magnitude supply. The logistic model introduces a quantity C called the carry-
in newtons is equal to the square of the speed of the car in metres per ing capacity for the environment in which the bacteria are living. As
second. Newton’s second law gives the following differential equation the number N(t) of bacteria approaches C , its growth rate must slow
for the velocity of the car: down. The logistic model to describe this is the differential equation
" #
dv dN N
1500 = 2500 − v 2 . = kN 1 − .
dt dt C
!
Notice that when N is small, dN/dt is approximately equal to kN , x 4 + 8x 3 − x 2 + 2x + 1
∗ 32. dx
thus preserving early exponential growth. The factor 1 − N/C causes x5 + x4 + x2 + x
dN/dt → 0 as N → C . Solve this differential equation for N(t)
when k = 1, C = 106 , and N(0) = 100. ∗ 33. During the initial stages of flow in a pipeline of length L , the
∗ 22. Show that the solution of the logistic model in Exercise 21 for an velocity of the flow v must satisfy the differential equation
initial population N (0) = N0 is
dv gHe
C = ,
N = " # . dt L
C − N0 −kt
1+ e where g = 9.81 and He is the effective head of the line, given by
N0
8 9
∗ 23. Find the centroid of the region bounded by the curves (x + 2)2 y = v2
4 − x , x = 0, y = 0, and (x, y ≥ 0) . He = H 1 − 2 ,
vf
∗ 24. Suppose that N represents the number of people in a population,
and x(t) the number that are infected by some disease at any given where H is the constant head and vf is the final velocity in the pipeline.
time t . It is often assumed that the rate of infection is proportional to (a) Using the initial condition v(0) = 0, show that
the product of the number of infected and not infected,
" #
dx Lvf vf + v
= kx(N − x), t = ln .
dt 2gH vf − v
where k > 0 is a constant. Thus, x(t) must satisfy this differential (b) Solve the equation in part (a) to obtain v as a function of t .
equation subject to an initial condition such as, perhaps, x(0) = 1.
Therefore, one infected person is introduced into the population at time ∗ 34. In the study of frictional fluid flow in a duct, the following indefinite
t = 0. Find x(t) . integral is encountered:
∗ 25. Chemical reactions such as that in Project 12 are called second- !

M(1 − M 2 )
order reactions (because of the x 2 term on the right). In general, they " # dM,
take the form k−1 2
dx M4 1 + M
= k(a − x)(b − x). 2
dt
Solve this differential equation in the cases that (a) a = b , (b) a != b . where k is a constant. Evaluate this integral.
∗ 26. The velocity v of water, flowing from a tap that is suddenly turned ∗ 35. If an integrand is a rational function of sin x and cos x , it can be
on, varies initially according to reduced to a rational function of t by the substitution
'x (
dv
a = v02 − v 2 , t = tan .
dt 2
where a > 0 is a constant and v0 is the steady-state velocity. Find This is often called the Weierstrass substitution. Show that with this
v(t) using the initial condition v(0) = 0. substitution
2 2t 1 − t2
In Exercises 27–32 evaluate the indefinite integral. dx = dt, sin x = , cos x = .
1 + t2 1+t 2 1 + t2
!
x3 + x + 2
∗ 27. dx
x 5 + 2x 3 + x
!
1 In Exercises 36–39 use the substitution of Exercise 35 to evaluate the
∗ 28. dx
x5 + x4 + 2x 3 + 2x 2 + x + 1 integral.
! ! !
1 1
∗ 29. dx ∗ 36. sec x dx ∗ 37. dx
(x 2 + 5)(x 2 + 2x + 3) 3 + 5 sin x
! ! !
1 1 1
∗ 30. dx ∗ 38. dx ∗∗ 39. dx
1 + x3 1 − 2 cos x sin x + cos x
!
sin x ∗ 40. Show that the answer to Exercise 36 can be expressed in the usual
∗ 31. dx
cos x(1 + cos2 x) form, ln | sec x + tan x| + C .
∗∗ 41. (a) Use the change of variable in Exercise 35 to show that (c) Show that
" #
x 2 −1 sin x
! ' + Tan
1 2 x( 3 3 2 − cos x
dx = Tan −1 3 tan + C.
5 − 4 cos x 3 2
is also an antiderivative of (5 − 4 cos x)−1 . Use it to eval-
uate the definite integral in part (b).
(b) What happens when the antiderivative in part (a) is used to !
evaluate x2 + x + 3
∗∗ 42. Evaluate dx .
x4 + + 2x 2 + 11x − 5
x3
! 2π !
1 2x 3 + 8x 2 − 3x + 5
dx ? ∗∗ 43. Evaluate dx .
0 5 − 4 cos x x4 + 3x 3 + x 2 + 2x − 12
8.7 Numerical Integration

To evaluate the definite integral of a continuous function f (x) with respect to x from x = a to
x = b , we have used the first fundamental theorem of integral calculus: Find an antiderivative for
f (x) , substitute the limits x = b and x = a , and subtract. The evaluation procedure depends
on our ability to produce an antiderivative for f (x) . When a function f (x) is complicated,
FIGURE 8.14 Definite
it may be difficult or even impossible to find its antiderivative. In such a case, it may be
integral as area when f (x) ≥ 0
necessary to approximate the definite integral of f (x) on some interval a ≤ x ≤ b , rather than
y evaluate it analytically. We discuss three methods for doing this: the rectangular, trapezoidal,
and Simpson’s rules. Each method divides the interval a ≤ x ≤ b into subintervals and
approximates f (x) with an easily integrated function on each subinterval.
y = f (x)
If f (x) ≥ 0 on a ≤ x ≤ b (Figure 8.14), the definite integral of f (x) with respect to x
can be interpreted as the area bounded by y = f (x) , y = 0, x = a , and x = b . We have
carefully pointed out that it is not always wise to think of a definite integral as area, but for our
discussion here it is convenient to do so. Our problem is to approximate the area in Figure 8.14
when it is difficult or impossible to find an antiderivative for f (x) .
a b x
Rectangular Rule
The first method is to return to definition 7.1 for area in terms of approximating rectangles. We
subdivide the interval a ≤ x ≤ b into n subintervals by points a = x0 < x1 < · · · < xn−1 <
xn = b . For simplicity, we choose n equal subdivisions, in which case
" #
b−a
xi = a + i ,
n
and denote the width of each subinterval by h = (b − a)/n . Area under y = f (x) , above
y = 0, and between x = a and x = b , which is given by the definite integral of f (x) from a
to b , is approximated by the rectangles in Figure 8.15,
! b n
5 n
5
f (x) dx ≈ f (xi ) h = h f (xi ). (8.11)
a i=1 i=1
This is the rectangular rule for approximating the definite integral; we have replaced the area
under y = f (x) with n rectangles. Another way of looking at it is to say that we have replaced
the original function f (x) by a function that is constant on each subinterval (Figure 8.16),
but the constant value varies from subinterval to subinterval. Such a function is called a step
function. The definite integral of this function from x = a to x = b is the right side of 8.11.
It approximates the definite integral of f (x) .
Graphically, it is reasonable to expect that an increase in the number of subdivisions results
in an increase in the accuracy of the approximation.
FIGURE 8.15 Approximation of area FIGURE 8.16 Interpretation of rectan-

by rectangles of equal widths gular rule as integration of step function
y y
f (b)
y = f (x)
y = f ( x)
f ( xn) f (xi)
f (xi)
f ( x4) f (x1)
a x2 x4 xi−1 xn−2 b x a x2 xi b x
x1 x3 xi xn−1 x1 xi−1 xn−1
EXAMPLE 8.17
Use a subdivision of the interval 1 ≤ x ≤ 3 into 5, 10, and 20 equal parts to approximate the
definite integral
! 3
sin x dx
1
with the rectangular rule.
SOLUTION With 5 equal parts, h = 2/5 and xi = 1 + 2i/5. Consequently,
! 5 " #
3
25 2i
sin x dx ≈ sin 1 +
1 5 5
i=1
" #
2 7 9 11 13
= sin + sin + sin + sin + sin 3
5 5 5 5 5
= 1.370.
With 10 equal parts, h = 2/10 = 1/5 and xi = 1 + i/5, so that
! 10 " #
3
15 i
1 5 5
i=1
"
1 6 7 8 9 11
= sin + sin + sin + sin + sin 2 + sin
5 5 5 5 5 5
#
12 13 14
+ sin + sin + sin + sin 3
5 5 5
= 1.455.
With 20 equal parts we have
! 20 " #
3
1 5 i
1 10 10
i=1
" #
1 11 6 13 29
= sin + sin + sin + · · · + sin + sin 3
10 10 5 10 10
= 1.494.
When we compare these results with the correct answer,

! 3 - .3
sin x dx = − cos x = − cos 3 + cos 1 = 1.530 294 8,
1 1
we see that as n increases from 5 to 10 to 20, the approximation improves, but it must be
increased even further to give a reasonable approximation to this integral.
Trapezoidal Rule
Regarding area, it is clear in Figure 8.15 that were we to join successive points (xi , f (xi )) on
the curve y = f (x) with straight-line segments as in Figure 8.17a, the area under this broken
straight line would be a much better approximation to the area under y = f (x) than that
provided by the rectangular rule. Effectively, we now approximate the area by n trapezoids.
FIGURE 8.17a Approximation of area FIGURE 8.17b Enlargement of i th

by trapezoids of equal widths subinterval
y y
y = f ( x)
( xi , f ( xi )) ( xi , f ( xi ))
( xi − 1, f ( xi −1)) ( xi − 1, f ( xi − 1))
( x3, f ( x3))
( x2 , f ( x2))
( x1, f ( x1))
a x1 x2 x3 xi − 1 xi xn − 2 b x
x
xi − 1 xi
h xn − 1
Since the area of a trapezoid is its width multiplied by the average of its parallel lengths, it
follows that the area of the i th trapezoid in Figure 8.17b is given by
) *
f (xi ) + f (xi−1 )
h ,
2
where again h = xi − xi−1 . As a result, the area under y = f (x) can be approximated by the
sum
) *
f (x1 ) + f (x0 ) f (x2 ) + f (x1 ) f (xn ) + f (xn−1 )
h + + ··· +
2 2 2
h
= [f (a) + 2f (x1 ) + 2f (x2 ) + · · · + 2f (xi ) + · · · + 2f (xn−1 ) + f (b)]
2
: n−1
;
h 5
= f (a) + 2 f (xi ) + f (b) .
2
i=1
We write therefore that

! : n−1
;
b
h 5
f (x) dx ≈ f (a) + 2 f (xi ) + f (b) , (8.12a)
a 2
i=1
where xi = a + ih = a + i(b − a)/n , and call this the trapezoidal rule for approximating
a definite integral. Note that if 8.12a is written in the form
! b ) * n
5
f (a) − f (b)
f (x) dx ≈ h +h f (xi ), (8.12b)
a 2
i=1
the summation on the right, except for the first two terms, is the rectangular rule. In other words,
the extra numerical calculation involved in using the trapezoidal rule rather than the rectangular
rule is minimal, but it would appear that the accuracy is increased significantly. For this reason,
the trapezoidal rule supplants the rectangular rule in most applications.
EXAMPLE 8.18
Use the trapezoidal rule to approximate the definite integral in Example 8.17.
SOLUTION With 5 equal partitions,
! : 4 " # ;
3
2 /5 5 2i
sin x dx ≈ sin 1 + 2 sin 1 + + sin 3 = 1.5098.
1 2 5
i=1
With 10 equal parts,

! : 9 " # ;
3
1/5 5 i
1 2 5
i=1
Finally, with n = 20, we have

! : 19 " # ;
3
1/10 5 i
1 2 10
i=1
As expected, these approximations are significantly better than corresponding results using the
rectangular rule.
Simpson’s Rule
The rectangular rule replaces a function f (x) with a step function; the trapezoidal rule replaces
FIGURE 8.18 Approxi- f (x) with a succession of linear functions — geometrically, a broken straight line. So far as
mation of curve by parabolas for
ease of integration is concerned, the next simplest function is a quadratic function. Consider,
Simpson’s rule
then, replacing the curve y = f (x) by a succession of parabolas on the subintervals xi−1 ≤
y x ≤ xi . Now the equation of a parabola with a vertical axis of symmetry is of the form
y = f ( x) y = ax 2 + bx + c with three constants a , b , and c to be determined. If this parabola is to
2
y = ax + bx + c approximate y = f (x) on xi−1 ≤ x ≤ xi , we should have the parabola pass through the end
( xi , f ( xi )) points (xi−1 , f (xi−1 )) and (xi , f (xi )) . But this imposes only two conditions on a , b , and c ,
( xi + 1, f ( xi + 1)) not three. To take advantage of this flexibility, we demand that the parabola also pass through
( xi − 1, f ( xi − 1)) the point (xi+1 , f (xi+1 )) (Figure 8.18).

In other words, instead of replacing y = f (x) with n parabolas, one on each subinterval
xi−1 ≤ x ≤ xi , we replace it with n/2 parabolas, one on each pair of subintervals xi−1 ≤
xi − 1 xi xi + 1 x x ≤ xi+1 . Note that this requires n to be an even integer. These three conditions imply that a ,
b , and c must satisfy the equations
2
axi− 1 + bxi−1 + c = f (xi−1 ), (8.13a)
axi2 + bxi + c = f (xi ), (8.13b)

2
axi+ 1 + bxi+1 + c = f (xi+1 ). (8.13c)
These equations determine the values for a , b , and c , but it will not be necessary to actually
solve them. Suppose for the moment that we have solved equations 8.17 for a , b , and c , and
we continue our main discussion. With the parabola y = ax 2 + bx + c replacing the curve
y = f (x) on the interval xi−1 ≤ x ≤ xi+1 , we approximate the area under y = f (x) with
that under the parabola, namely,
! xi+1
a b
(ax 2 + bx + c) dx = 3
(xi+ 3
1 − xi−1 ) +
2
(xi+ 2
1 − xi−1 ) + c(xi+1 − xi−1 )
xi−1 3 2
xi+1 − xi−1 2 2
= [2a(xi+ 1 + xi+1 xi−1 + xi−1 )
6
+ 3b(xi+1 + xi−1 ) + 6c].
Now, if we use xi+1 − xi−1 = 2h , xi+1 = xi + h , and xi−1 = xi − h to write everything
in terms of h and xi , we obtain
! xi+1
2h ? + ,
(ax 2 + bx + c) dx = 2a (xi + h)2 + (xi + h)(xi − h) + (xi − h)2
xi−1 6
+ , @
+ 3b (xi + h) + (xi − h) + 6c
h? @
= a(6xi2 + 2h2 ) + 6bxi + 6c .
3
But if we add equation 8.13a, equation 8.13c, and four times equation 8.13b, we find that
2 2 2
f (xi−1 ) + 4f (xi ) + f (xi+1 ) = a(xi− 1 + 4xi + xi+1 ) + b(xi−1 + 4xi + xi+1 ) + 6c
+ ,
= a (xi − h)2 + 4xi2 + (xi + h)2
+ ,
+ b (xi − h) + 4xi + (xi + h) + 6c
= a(6xi2 + 2h2 ) + 6bxi + 6c.

Thus, the area under the parabola may be written in the form
! xi+1
h+ ,
(ax 2 + bx + c) dx = f (xi−1 ) + 4f (xi ) + f (xi+1 ) ,
xi−1 3
and the right side is free of a , b , and c . The same expression would have resulted had we solved
equations 8.13 for a , b , and c and then evaluated the integral of ax 2 + bx + c from xi−1 to
xi+1 . The derivation above, however, is much simpler.
When we add all such integrals over the n/2 subintervals xi−1 ≤ x ≤ xi+1 between
x = a and x = b , we obtain
h+ , h+ ,
f (x0 ) + 4f (x1 ) + f (x2 ) + f (x2 ) + 4f (x3 ) + f (x4 ) + · · ·
3 3
h+ ,
+ f (xn−2 ) + 4f (xn−1 ) + f (xn )
3
h+
= f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + · · ·
3
,
+ 2f (xn−2 ) + 4f (xn−1 ) + f (xn ) .
In other words, the definite integral of f (x) from x = a to x = b can be approximated by
! b
h+
f (x) dx ≈ f (a) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + · · ·
a 3
,
+ 2f (xn−2 ) + 4f (xn−1 ) + f (b) , (8.14)
where xi = a + ih = a + i(b − a)/n . This result is called Simpson’s rule for approximating
a definite integral. Although the formula does not display it explicitly (except by counting
terms), do not forget that n must be an even integer.
EXAMPLE 8.19
Approximate the definite integral in Example 8.17 using Simpson’s rule with n = 10 and
n = 20.
SOLUTION With n = 10,
! 3
1/5 = 6 7 8 13 14 >
sin x dx ≈ sin 1 + 4 sin + 2 sin + 4 sin + · · · + 2 sin + 4 sin + sin 3
1 3 5 5 5 5 5
= 1.530 308 5.
With n = 20,
! 3
1/10 = 11 6 13 14 29 >
sin x dx ≈ sin 1 + 4 sin + 2 sin + 4 sin + · · · + 2 sin + 4 sin + sin 3
1 3 10 5 10 5 10
= 1.530 295 7.
Table 8.1 lists the approximations in Examples 8.17–8.19. The correct answer for the integral is
1.530 294 8 (to seven decimal places). It is clear that each method gives a better approximation
as the value of n increases, and that Simpson’s rule is by far the most accurate.
TABLE 8.1
Rectangular rule Trapezoidal rule Simpson’s rule

n=5 1.370 1.5098
n = 10 1.455 1.5252 1.530 308 5
n = 20 1.494 1.5290 1.530 295 7
In practice, we use the rectangular, trapezoidal, and Simpson’s rules to approximate definite
integrals that cannot be handled analytically, and we will not therefore have the correct answer
with which to compare the approximation. We would still like to make some statement about
the accuracy of the approximation, however, since what good is the approximation otherwise?
The following two theorems give error estimates for the trapezoidal rule and for Simpson’s rule.
THEOREM 8.1
If f .. (x) exists on a ≤ x ≤ b , and Tn is the error in approximating the definite integral
of f (x) from x = a to x = b using the trapezoidal rule with n equal subdivisions,
! : n−1
;
b
h 5
Tn = f (x) dx − f (a) + 2 f (xi ) + f (b) ,
a 2
i=1
then
M(b − a)3
|Tn | ≤ , (8.15)
12n2
where M is the maximum value of |f .. (x)| on a ≤ x ≤ b .
THEOREM 8.2
If f .... (x) exists on a ≤ x ≤ b , and Sn is the error in approximating the definite integral
of f (x) from x = a to x = b using Simpson’s rule with n equal subdivisions,
! : n/2 n/2−1
;
b
h 5 5
Sn = f (x) dx − f (a) + 4 f (x2i−1 ) + 2 f (x2i ) + f (b) ,
a 3
i=1 i=1
then
M(b − a)5
|Sn | ≤ , (8.16)
180n4
where M is the maximum value of |f .... (x)| on a ≤ x ≤ b .
Proofs of these theorems can be found in books on numerical analysis. Note that because
of the n4 factor in the denominator of 8.16, the accuracy of Simpson’s rule increases much more
rapidly than does that of the trapezoidal rule.
For the function f (x) = sin x in Examples 8.17–8.19,
f .. (x) = − sin x and f .... (x) = sin x.
Consequently, in both cases we can state that M = 1, and therefore
(3 − 1)3 2 (3 − 1)5 8
|Tn | ≤ 2
= and |Sn | ≤ = .
12n 3n2 180n 4 45n4
For n = 10 and n = 20, we find that
2 8
|T10 | ≤ < 0.0067 and |S10 | ≤ < 0.000 018;
3(10)2 45(10)4
2 8
|T20 | ≤ < 0.0017 and |S20 | ≤ < 0.000 001 2.
3(20)2 45(20)4
Differences between the correct value for the integral and the approximations listed in Table 8.1
corroborate these predictions.
EXAMPLE 8.20
What is the maximum possible error in using the trapezoidal rule with 100 equal subdivisions
to approximate
! 3
1
√ dx ?
1 1 + x3
SOLUTION According to formula 8.15, if T100 is the maximum possible error, then
M(3 − 1)3 2M
|T100 | ≤ = ,
12(100)2 3 × 104
where
√ M is the maximum of (the absolute value of) the second derivative of the integrand
1/ 1 + x 3 on 1 ≤ x ≤ 3. Now
) *
d2 1 d −3x 2 3x(5x 3 − 4)
2
√ = 3 3 2
= .
dx 1 + x3 dx 2(1 + x ) / 4(1 + x 3 )5/2
Instead of maximizing the absolute value of this function on the interval 1 ≤ x ≤ 3, which
would require another derivative, we note that the maximum value of the numerator is obtained
for x = 3, and the minimum value of the denominator occurs at x = 1. It follows therefore
that the second derivative cannot possibly be larger than
3(3)[5(3)3 − 4] 1179
= √ .
4 (1 + 1)5/2 16 2
√
Thus, M must be less than or equal to 1179/(16 2) , and we can state that
" #
2 1179
|T100 | ≤ √ ≤ 0.0035;
3 × 104 16 2
that is, the error in using the trapezoidal rule with 100 equal subdivisions to approximate the
definite integral cannot be any larger than 0.0035.
EXAMPLE 8.21
How many equal subdivisions of the interval 0 ≤ x ≤ 2 guarantee an error of less than 10−5
in the approximation of the definite integral
! 2
2
e−x dx
0
using Simpson’s rule?

SOLUTION According to formula 8.16, the error in using Simpson’s rule with n equal sub-
divisions to approximate this definite integral is
M(2 − 0)5
|Sn | ≤ ,
180n4
2
where M is the maximum of the (absolute value of the) fourth derivative of e−x on the interval
0 ≤ x ≤ 2. It is a short calculation to find
d 4 ' −x 2 ( 2
e = 4(3 − 12x 2 + 4x 4 )e−x .
dx 4
2
Instead of maximizing the absolute value of this function, we note that e−x has a maximum
value of 1 (when x = 0). Furthermore, because |3 − 12x 2 + 4x 4 | ≤ 3 + 12x 2 + 4x 4 , which
has a maximum at x = 2, it follows that
M ≤ 4[3 + 12(2)2 + 4(2)4 ](1) = 460.

Consequently,
460(2)5 3680
|Sn | ≤ = .
180n4 45n4
The error is less than 10−5 if n is chosen sufficiently large that
3680
< 10−5 ,
45n4
that is, if
" #1/4
3680
n> = 53.5.
45 × 10−5
Since n must be an even integer, the required accuracy is guaranteed if n is chosen greater than
or equal to 54.
Numerical techniques are indispensable in situations where the function to be integrated is not
known, but what is available is a set of tabulated values for the function, perhaps experimental
FIGURE 8.19 Approxi-
data. For example, when the truss in Figure 8.19 is subjected to a force F at its centre, point A
mating work to deflect truss from
tabulated values of deflection
deflects an amount y from its equilibrium position. Forces (in kilonewtons) required to produce
deflections from 0 to 5 cm at intervals of 0.5 cm are listed in Table 8.2.
F
TABLE 8.2
A y=0
y 0 0 .5 1.0 1.5 2 .0 2 .5 3.0 3.5 4 .0 4 .5 5.0
F 0 1.45 2.90 4.40 5.90 7.43 9.05 10.7 13.2 15.3 18.0
y
The work done by F in deflecting A by 5 cm is the definite integral of F with respect to y

from y = 0 to y = 5. The trapezoidal rule and Simpson’s rule can be used to approximate
this definite integral even though we do not have a formula for F as a function of y . Indeed,
formulas 8.12 and 8.14 do not use the form of the function to be integrated, only its values at the
subdivision points. We have these points in Table 8.2. Since there are 11 points and, therefore,
10 subdivisions on the interval 0 ≤ y ≤ 5, the trapezoidal rule gives
! 5
1/2
F (y) dy ≈ [F (0) + 2F (0.5) + 2F (1.0) + · · · + 2F (4.0) + 2F (4.5) + F (5.0)]
1 2
1
= [0 + 2(1.45) + 2(2.90) + · · · + 2(13.2) + 2(15.3) + 18.0]
4
= 39.67 kN · cm = 396.7 J.
Simpson’s rule gives

! 5
1/2
F (y) dy ≈ [F (0) + 4F (0.5) + 2F (1.0) + · · · + 2F (4.0) + 4F (4.5) + F (5.0)]
1 3
1
= [0 + 4(1.45) + 2(2.90) + · · · + 2(13.2) + 4(15.3) + 18.0]
6
= 39.54 kN · cm = 395.4 J.
What we lose in this type of application is the ability to predict a maximum possible error in the
approximation since we cannot find M for formulas 8.15 and 8.16.
The error bounds in formulas 8.15 and 8.16 are somewhat idealistic in the sense that they
are error predictions based on the use of exact numbers. For instance, they predict that by
increasing n indefinitely, any degree of accuracy is attainable. Theoretically this is true, but
practically it is not. No matter how we choose to evaluate the summations in 8.12 and 8.14, be
it by hand, by an electronic hand calculator, or by a high-speed computer, each calculation is
rounded off to a certain number of decimals. The final sum takes into account many, many of
these “approximate numbers,” and must therefore be inherently inaccurate. We call this round-
off error and it is very difficult to predict how extensive it is. It depends on both the number and
nature of the operations involved in 8.12 and 8.14. In the approximation of a definite integral
by the trapezoidal rule or Simpson’s rule, there are two sources of error. Formulas 8.15 and
8.16 predict errors due to the methods themselves; round-off errors may also be appreciable for
large n .
We should emphasize once again that although we have used area as a convenient vehicle
by which to explain the approximation of definite integrals by the rectangular, trapezoidal, and
Simpson’s rules, it is not necessary for f (x) to be nonnegative. All three methods can be used
to approximate the definite integral of a function f (x) , be it positive, negative, or sometimes
positive and sometimes negative on the interval of integration. The only condition that we have
imposed is that f (x) be continuous. In view of our discussion of improper integrals in Section
7.10, even this is not always necessary.
EXERCISES 8.7
In Exercises 1–10 use the trapezoidal rule and Simpson’s rule with 10
equal subdivisions to approximate the definite integral. In each case,
evaluate the integral analytically to get an idea of the accuracy of the 3.8
4.6 3.6 3.6
approximation. 5.8 3.8
! ! 7.0 6.8
2
1 3
1 3.4 6.0
1. dx 2. √ dx
1 x 2 x+2
! 1 ! 1/ 2
3. tan x dx 4. ex dx
0 0 ∗ 21. An aerial photograph of an oil spill shows the pattern in the figure
! ! below. Assuming that the oil slick has a uniform depth of 1 cm, estimate
1 √ −2
1 the number of cubic metres of oil in the spill.
5. x + 1 dx 6. dx
−1 −3 x3
350 m
! 1 ! 1
1
7. cos x dx 8. dx
1/ 2 0 3 + x2
180 m 110 m 200 m 440 m 210 m 180 m
! 3 ! 1/ 2
1 x2
9. dx 10. xe dx
1 x2 + x 0
50 m
In Exercises 11–14 use the trapezoidal rule and Simpson’s rule with
10 equal subdivisions to approximate the definite integral.
! 2 ! 1
1 2
11. dx 12. ex dx
0 1+ x3 0
! 2 & ! 0
13. 1 + x 4 dx 14. sin (x 2 ) dx
1 −1 ∗ 22. The numerical techniques of this section can also be used to ap-
15. Show graphically that if y = f (x) is concave downward on the proximate many improper integrals. Consider
interval a ≤ x ≤ b , then the trapezoidal rule underestimates the ! 4
definite integral of f (x) on this interval. ex
√ dx.
0 x
16. What happens to the errors in 8.15 and 8.16 when the number of
partitions is doubled?
(a) Why can the trapezoidal rule and Simpson’s rule not be
17. The definite integral used directly to approximate this integral?
√
! (b) Show that the change of variable u = x replaces this im-
b
2 proper integral with an integral that is not improper and use
e−x dx
a Simpson’s rule with 20 equal subdivisions to approximate
its value.
is very important in mathematical statistics. Use Simpson’s rule with
(c) Could you use the rectangular rule on the improper integral?
n = 16 to evaluate the integral for a = 0 and b = 1.
∗ 18. Use Simpson’s rule with 10 equal intervals to approximate the ∗ 23. (a) Show that the definite integral
definite integral for the length of the parabola y = x 2 between x = 0 <
and x = 1. Compare the answer to that of Exercise 39 in Section 8.4.
! 3
4 81 − 5x 2
L= dx
∗ 19. Use the trapezoidal rule and Simpson’s rule with 10 equal subdi- 3 0 9 − x2
visions to approximate the definite integral for the length of the curve
y = sin x from x = 0 to x = π/2. represents the length L of the ellipse 4x 2 + 9y 2 = 36.
∗ 20. The swimming pool that follows has an average depth of 1.8 m. (b) Show that when we set x = 3 sin θ , the θ -integral is no
It is to be drained and filled with dirt. To estimate the volume of dirt longer improper.
required, measurements across the pool are taken at 1-m intervals. Use (c) Use the trapezoidal rule and Simpson’s rule with eight equal
Simpson’s rule to find the estimate. subdivisions to approximate the θ -integral in part (b).
∗ 24. To approximate the improper integral ∗ 27. Repeat Exercise 26 for the integral from x = 1 to x = 3 for the
function tabulated below.
! ∞
1
dx x 1.0 1.2 1.4 1.6 1.8 2 .0 2 .2 2 .4 2 .6 2 .8 3.0
1 1 + x4
y 0.84 1.12 1.40 1.60 1.75 1.82 1.79 1.62 1.34 0.94 0.42
set x = 1/t , and use the trapezoidal rule and Simpson’s rule with 10
equal subdivisions on the resulting integral. ∗ 28. Show that when f (x) is a cubic polynomial, evaluation of the
definite integral of f (x) from x = a to x = b by Simpson’s rule
∗ 25. Use the technique of Exercise 24 to approximate always gives the exact answer. Illustrate with an example.
! ∞
x2
dx. In Exercises 29–32 how many equal subdivisions of the interval of
1 x4 + x2 + 1
integration guarantee an error of less than 10−4 in the approximation
of the definite integral using (a) the trapezoidal rule and (b) Simpson’s
∗ 26. Use the trapezoidal rule and Simpson’s rule to approximate the
rule?
definite integral from x = −1 to x = 4 for the function tabulated in
the following table.
! 4 ! π/4
1
∗ 29. dx ∗ 30. cos x dx
1 x 0
x −1.0 −0.5 0 0 .5 1.0 1.5 2 .0 2 .5 3.0 3.5 4 .0 ! 1/ 3 ! 5
1
∗ 31. e2x dx ∗ 32. √ dx
y 2.287 0.395 0 0.145 0.310 0.334 1.819 0.123 0.021 −0.037 −0.055 0 4 x+2
SUMMARY
Antidifferentiation is a far more complicated process than differentiation; it is not possible to state
a set of rules and formulas that will suffice for most functions. Certainly, there are integration
formulas that we must learn, but most of those given in this chapter have been differentiation
formulas listed in Chapter 3 written in terms of integrals rather than derivatives. We have stressed
the importance of knowing these obvious integration formulas, but beyond this, it becomes an
organizational problem — organizing a difficult integral into a form that utilizes these simple
formulas. The three most important techniques for doing this are substitutions, integration by
parts, and partial fractions.
There is a variety of substitutions for the evaluation of indefinite integrals, many suggested
by the form of the integrand. For example, a term in x 1/n ( n an integer) suggests the substitution
u = x√1/n . Trigonometric√ substitutions are most important.√They eliminate square roots of the
form a 2 ± b2 x 2 and b2 x 2 − a 2 , and the general root ax 2 + bx + c .
Integration by parts is a powerful integration technique. It is used to evaluate antideriva-

tives of transcendental functions that are multiplied by powers of x , it leads to the method of
integration by reproduction, and it is used to develop reduction formulas.
The method of partial fractions decomposes complicated rational functions into simple frac-
tions that are either immediately integrable or amenable to other methods, such as trigonometric
substitutions.
Even with the techniques that we have studied and tables, there are many functions that
either cannot be antidifferentiated at all or can be antidifferentiated only with extreme difficulty.
To approximate definite integrals of such functions, numerical techniques such as the rectangular
rule, the trapezoidal rule, and Simpson’s rule are essential.
KEY TERMS
Integration by parts Integration-by-parts formula

Integration by reproduction Trigonometric substitutions
Partial fraction decomposition Rectangular rule
Trapezoidal rule Simpson’s rule
REVIEW
EXERCISES
! ! &
1
In Exercises 1–50 evaluate the indefinite integral. 27. dx 28. x 3 4 − x 2 dx
x2 + 4x − 5
! !
! ! cos 2x 6x
√ 1 29. dx 30. dx
1. 2 − x dx 2. dx 1 − sin 2x 4 − x2
(x + 3)2 ! !
1 1
! 2 ! 2 31. √ dx 32. dx
x +3 x +3 x ln x x2 + 4x − 4
3. dx 4. dx ! !
x x+1 sin x 1
33. dx 34. dx
! 2 ! 1 + cos2 x x4 + x3
x +3 x ! !
5. dx 6. √ dx 1
x2 + 1 x+3 35. x sec2 (3x) dx 36. √ dx
! ! 16 − 3x + x 2
! √ !
7. sin2 x cos3 x dx 8. x sin x dx x2 − 4
37. dx 38. x 2 Tan −1 x dx
x2
! ! ! !
9. tan2 (2x) dx 10.
x
dx x2 ln x
39. dx 40. dx
x2 + 2x − 3 x 3 + 3x 2 + 3x + 1 x
! ! √ ! !
1 2− x x2 1
11. √ dx 12. √ dx 41. dx 42. dx
x+5 1 + 4x 6 x(9 + x 2 )2
4 − 3x 2
! !
! ! x2 + 2 x2 + 2
x e x 43. dx 44. dx
13. dx 14. √ dx x + 5x 2 + 4x
3 x3 + 4x 2 + 4x
3x 2 + 4 1 − e 2x ! !
x2 + 2 3x 2 + 2x + 4
! ! 45. dx 46. dx
x x 3 + x 2 + 4x x 3 + x 2 + 4x
15. x 2 ln x dx 16. dx ! !
(x + 1)2
2 √
47. x Sin −1 x dx 48. cot x csc4 x dx
! 2 ! 3
x x ! !
17. dx 18. dx √ 1
(x 2 + 1)2 (x 2 + 1)2 49. ln ( x + 1) dx 50. dx
(4x − x 2 )3/2
! ! " #2
x+1 x+1
19. dx 20. dx
x 3 − 4x x−1 In Exercises 51–55 use the trapezoidal rule and Simpson’s rule with
10 equal partitions to approximate the definite integral.
! !
x2 ! ! 1√
21. dx 22. Cos −1 x dx 2
sin x
(1 + 3x 3 )4 51. dx 52. sin x dx
1 x 0
! ! ! !
4 3
23. sin x cos 2x dx 24. sin x cos 5x dx 1 1
53. dx 54. dx
2 ln x −1 1 + ex
! ! ! 1
1 1
25. e3x cos 2x dx 26. √ dx 55. dx
x 2 + 4x − 5 0 (1 + x 4 )2
! !
In Exercises 56–71 evaluate the indefinite integral. 1 x 4 + 3x 2 − 2x + 5
∗ 66. dx ∗ 67. dx
1 + cos 2x x 2 − 3x + 7
! ! ! !
1 1
∗ 56. √ dx ∗ 57. ln (1 + x 2 ) dx ∗ 68. sin2 x cos 3x dx ∗ 69. dx
x 1/ 3 − x x 3 (4 − x 2 )3/2
! ! ! & !
x 1
∗ 58. dx ∗ 59. 3
csc x dx ∗ 70. 1 − x 2 Sin −1 x dx ∗ 71. √ dx
4
x + 16 x+ x2 + 4
! !
1 1 ∗ 72. (a) Use the substitution z2 = (1 + x)/(1 − x) to show that
∗ 60. dx ∗ 61. √ dx
(3x − x 2 )3/2 x3 x2 −9 ! 6 6
&
! ! 1+x −1 1+x
√ dx = 2 Tan − 1 − x 2 + C.
∗ 62. sin x dx ∗ 63. sin (ln x) dx 1−x 1−x
! ! (b) Evaluate the integral in part (a) by multiplying

√ numerator
x 4 + 3x 2 + 1 and denominator of the integrand by 1 + x . Verify that
∗ 64. x cos x sin 3x dx ∗ 65. dx
x(x 2 + 1)2 this answer is the same as that in part (a).
CHAPTER
9 Parametric Equations and Polar
Coordinates
Application Preview The figure below shows a mechanism in which rod AB , pinned at A , rotates counterclockwise
about A . Slider B moves along rod CD , causing it to rotate also. End E of rod CE is confined
to slide along a horizontal line. Lengths of the members AB , CD , and CE are as shown.
y
C l3 E
l2
B
l1
A θ D
x
d > l1
THE PROBLEM Determine the path followed by joint C . (See Example 9.6 on page 548
for the solution.)
In this chapter we introduce parametric representations for curves and polar coordinates.
Parametric representations of curves offer an alternative to the explicit and implicit forms used
in Chapters 1–8. Polar coordinates provide an alternative way to identify points in a plane. They
are more efficient than Cartesian coordinates in many applications. In particular, we shall see
how they provide a unified approach to conic sections.
9.1 Parametric Equations

A curve is defined explicitly by equations of the form y = f (x) or x = g(y) , and implicitly
by an equation F (x, y) = 0. A third method is described in the following definition.
DEFINITION 9.1
A curve is said to be defined parametrically if it is given in the form
x = x(t), y = y(t), α ≤ t ≤ β. (9.1)
541
542 Chapter 9 Parametric Equations and Polar Coordinates
Variable t is called a parameter ; it is a connecting link between x and y . Each value of t in

the interval α ≤ t ≤ β is substituted
! into"the parametric equations x = x(t) and y = y(t)
in 9.1, and the pair (x, y) = x(t), y(t) represents a point on the curve. When the interval
for t is unspecified, we assume that it consists of all values for which both x(t) and y(t) are
defined.
Graphing calculators and computers are adept at plotting curves defined parametrically.
Given !the functions" x(t) and y(t) , and the interval α ≤ t ≤ β , they plot sufficiently many
points x(t), y(t) and join them with straight lines to give an excellent rendering of the curve.
EXAMPLE 9.1
Plot the curve
x = 2 − t, y = t + 5, −6 ≤ t ≤ 5.
What does it look like? Verify this analytically.
SOLUTION The plot in Figure 9.1 appears to be a straight line joining the points (8, −1) and
(−3, 10) . This is easily verified by solving x = 2 − t for t = 2 − x and substituting into
y = t + 5:
y = (2 − x) + 5 = −x + 7.
FIGURE 9.1 Parametric plot of a straight line
y
10
8
6
4
2
−2 2 4 6 8 x
When x(t) and y(t) in equations 9.1 are linear functions, the curve is always a straight line or
line segment. If α and β are finite, they define ends of the line segment; if they are not finite, the
whole line results. Line segments can also be defined by nonlinear functions (see Exercise 31).
EXAMPLE 9.2
Plot the curve
x = t 2 − 2 t + 4, y = 3 − 2t, −2 ≤ t ≤ 2 .
What does it look like? Verify this analytically.
SOLUTION The plot in Figure 9.2 appears to be a parabola joining the points (12, 7) and
(4, −1) . To verify this, we solve the second equation for t = (3 − y)/2, and substitute into
the first:
# $2 # $
3−y 3−y y2 y 13
x = −2 +4 = − + .
2 2 4 2 4
FIGURE 9.2 Parametric plot of part of a parabola
y
7
6
5
4
3
2
1
0
2 4 6 8 10 12 x
−1
EXAMPLE 9.3
Discuss the curve defined parametrically by
x = 2 cos θ, y = 2 sin θ, 0 ≤ θ < 2π.
SOLUTION If the given equations are squared and added, the result is
x 2 + y 2 = 4 cos2 θ + 4 sin2 θ = 4,
an implicit definition of the curve. The given equations therefore define the circle parametri-
cally. Seldom is it possible to give a geometric interpretation for parameter t in equations 9.1.
Parameter θ for a circle is an exception; it can be interpreted as the angle in Figure 9.3. Values
0 ≤ θ < 2π describe the complete circle in a counterclockwise direction beginning at (2, 0) .
Additional values of θ duplicate existing points. For example, using 0 ≤ θ < 4π traces the
circle twice. Specifying an interval of length less than √ 2π describes part of the circle. For
instance, 0 ≤ θ ≤ π gives the upper semicircle y = 4 − x 2 .
FIGURE 9.3 Parametric plot of a circle
y
2
(x, y)
1
−2 −1 1 2 x
−1
−2
EXAMPLE 9.4
Find an implicit definition for the curve
x = t − t 3, y = t + t 3,
and plot the curve.
SOLUTION By adding and subtracting the given equations, we have
x + y = 2t and y − x = 2t 3 .
It follows from the first of these that t = (x + y)/2, and when this is substituted into the
second, we obtain an implicit definition of the curve
(x + y)3
y−x = .
4
Unfortunately, as is always the case when an electronic device is used to plot parametric equa-
tions, Figure 9.4a does not indicate which values of t give which points on the curve. We can
remedy this by relating graphs of x(t) and y(t) in Figures 9.4b and c to the curve in Figure 9.4a.
For example, beginning with t = 0, Figures 9.4b and c give x = 0 and y = 0, and therefore
the point (0, 0) in Figure 9.4a.
√ As t increases
√ from t = 0 to t = 1, values of x increase from
0 to a maximum value 2/(3 3) at t = 1/ 3, and then decrease to 0. Simultaneously, values
of y increase steadily from 0 to 2. This gives the first quadrant part of the curve in Figure 9.4a.
As t increases beyond 1, values of x decrease through negative numbers while y continues
to increase. This is reflected in that part of the graph in Figure 9.4a in the second quadrant.
Because x = t − t 3 and y = t + t 3 are odd functions, replacing t by −t reverses the signs
of x and y . This means that corresponding to each point (x, y) on the graph for which t > 0,
we must have the point (−x, −y) corresponding to −t . This gives third- and fourth-quadrant
parts of the curve in Figure 9.4a.
FIGURE 9.4a Parametric FIGURE 9.4b Plot of FIGURE 9.4c Plot of

plot of x = t − t 3 , y = t + t 3 x = t − t3 y = t + t3
y x 1.5 y
3 1.5
1.0
2 1.0
0.5
1 0.5
−3 −2 −1 1 2 3 t −3 −2 −1 1 2 3 t
−1.5 −1 −0.5 0.5 1 1.5 x −0.5
−1 −0.5
−1.0
−2 −1.0
−1.5
−3 −1.5
It is worthwhile noting that were calculators and/or computers not programmed to do parametric
plots, we would draw parametric curves precisely as outlined in this example. Piece the curve
together from separate graphs of x(t) and y(t) . This would be the case if x(t) or y(t) , or both,
contained unspecified parameters. The following example is an illustration.
EXAMPLE 9.5
Draw the curve defined parametrically by
x = a cos t, y = b sin 2t,

where a > 0 and b > 0 are constants.
FIGURE 9.5a Parametric plot of x = 2 cos t , y = sin 2t FIGURE 9.5b Graph of x = a cos t , y = b sin 2t
y y
1 b
0.5
−2 −1 1 2 x a x
−0.5
−1
SOLUTION Because a cos t and b sin 2t are periodic with periods 2π and π , respectively, it
follows that all values of x and y are obtained for 0 ≤ t ≤ 2π . We begin by plotting the curve
using specific values for a and b , say a = 2 and b = 1. The result is shown in Figure 9.5a.
The required curve can be obtained by changing scales on the x - and y -axes (Figure 9.5b).
To confirm this, we draw the curve using the technique of Example 9.4. We draw graphs
of x = a cos t and y = b sin 2t (Figures 9.6a and b). Value t = 0 gives the point (a, 0) in
Figure 9.6c. As t increases from t = 0 to t = π/2, values of x decrease from a to 0 and y
increases from 0 to b , and then decreases from b to 0. This gives the first-quadrant part of the
curve in Figure 9.6c. As t increases from π/2 to π , values of x decrease from 0 to −a and y
decreases from 0 to −b , and then increases from −b to 0. This adds the third-quadrant part of
the curve in Figure 9.6d. Continuation leads to the full curve in Figure 9.5b.
FIGURE 9.6a Graph of x = a cos t FIGURE 9.6b Graph of y = b sin 2t

x y
a b
t t
FIGURE 9.6c First-quadrant graph of x = a cos t , FIGURE 9.6d First- and third-quadrant graph of
y = b sin 2t x = a cos t , y = b sin 2t
y y
b b
−a
a x a x
−b
Parametric equations are frequently used in problems concerning the motion of objects in a
FIGURE 9.7 Path of a
projectile is a parabola
plane or in space. For example, suppose a stone is thrown horizontally over the edge of a cliff
100 m above a river (Figure 9.7). If the initial speed of the stone is 30 m/s, the path followed by
the stone is described parametrically by
%
2 100
x = 30t, y = 100 − 4.905t , 0 ≤ t ≤ ,
4.905
100 where t is time in seconds and x and y are in metres.
√ For these equations, t has been chosen
equal to zero at the instant of projection, and t = 100/4.905 is the time at which the stone
strikes the river.
y
When parameter t is eliminated from this parametric representation of the trajectory, the
resulting equation is
x & x '2
y = 100 − 4.905 = 100 − 0.005 45x 2 .
30
It clearly indicates that the stone follows a parabolic path. However, if we were to discuss the
motion of the stone as regards, say, velocity and acceleration, we would find it necessary to
return to the parametric representation that describes the motion in the horizontal and vertical
directions.
We say that equations 9.1 define y parametrically as a function of x if the curve so defined
represents a function; that is, if every vertical line that intersects the curve does so exactly once.
The parametric equations in Example 9.1 therefore define a function parametrically, but in
Examples 9.2–9.5 they do not. However, each of the curves in these latter examples can be
divided into subcurves that do represent functions. For instance, in Example 9.4, each of the
subcurves √
x = t − t 3, y = t + t 3, t ≤ −1/ 3; √
√
x = t − t 3 , y = t + t 3 , −1/ 3 ≤ t √ ≤ 1/ 3;
x = t − t 3, y = t + t 3, t ≥ 1/ 3;
defines a function, and does so parametrically (Figures 9.8); together all three curves make up
the original curve.

Division of curve in Figure 9.4a into pieces each of which represents a function
y y y
1.0 4
0.5 1 x
−1 0.5 3
2
−2 −0.4 −0.2 0.2 0.4 x
−0.5 1
−3
−4 −1.0 −1 −0.5 x
When equations 9.1 do define y parametrically as a function of x , it is straightforward to

find the derivative of the function. If we denote this function by y = f (x) , then the chain rule
applied to y = f (x) , x = x(t) gives
dy dy dx
= .
dt dx dt
Provided then that dx/dt %= 0, we may solve for
dy
dy
= dt . (9.2)
dx dx
dt
This is called the parametric rule for differentiation of a function defined parametrically by
equations 9.1; it defines the derivative of y with respect to x in terms of derivatives of the given
functions x(t) and y(t) with respect to t .
For example, the curve in Figure 9.9 is defined parametrically by
1 2
x(t) = , y(t) = .
t2 +1 t (t 2 + 1)
It can be subdivided into two parts, one corresponding to values of t > 0 and the other to
values of t < 0, and each defines y as a function of x . The derivative of either function is
dy −2(3t 2 + 1)
dy (t 3 + t)2 3t 2 + 1
= dt = = .
dx dx −2 t t3
dt (t 2 + 1)2
Neither function has a relative maximum or minimum, and this is consistent with the fact that
dy/dx never vanishes. The derivative is undefined for t = 0, as is y . We can also calculate
d 2 y/dx 2 in spite of the fact that dy/dx is in terms of t rather than x . Once again it is the
chain rule that comes to the rescue:
# $
d dy
# $ # $
d 2y d dy d dy dt dt dx
= = =
dx 2 dx dx dt dx dx dx
dt
t 3 (6t) − (3t 2 + 1)(3t 2 )
t6 −3t 4 − 3t 2 (t 2 + 1)2 3(t 2 + 1)3
= = · = .
−2 t t6 −2 t 2t 5
(t 2 + 1)2
This derivative is positive for t > 0 and negative for t < 0, agreeing with the fact that the curve
in Figure 9.9 is concave upward when t > 0 and concave downward when t < 0.
FIGURE 9.9 Plot of x = (t 2 + 1)−1 , y = 2[t (t 2 + 1)]−1
y
7.5
5.0
t>0
2.5
0.2 0.4 0.6 0.8 x

−2.5
t<0
−5.0
−7.5
EXAMPLE 9.6
Application Preview Find parametric equations for the path followed by joint C of the mechanism in the Application
Revisited Preview. Plot the path for l1 = 1/2 m, l2 = 2 m, and d = 1 m.
FIGURE 9.10 Determination of path followed by point C in mechanism
y
C l3 E
l2
B
l1
A θ D
x
d > l1
SOLUTION Coordinates of B in terms of the angle of rotation θ from the positive x -axis
(Figure 9.10) are
xB = l1 cos θ, yB = l1 sin θ.
The equation of line BD is y = m(x − d) , and since B is on the line,
l1 sin θ
l1 sin θ = m(l1 cos θ − d) &⇒ m= .
l1 cos θ − d
Because point C(xc , yc ) lies on this line, it follows that
l1 sin θ
yc = (xc − d).
l1 cos θ − d
Furthermore, the length of CD is l2 so that
(xc − d)2 + yc2 = l22 .
Consequently,
l12 sin2 θ
(xc − d)2 + (xc − d)2 = l22 .
(l1 cos θ − d)2
This equation can be solved for xc :
(
)
) l22 l2 (d − l1 cos θ )
xc = d ± ) =d±+ .
* l12 sin2 θ d + l12 − 2dl1 cos θ
2
1+
(l1 cos θ − d)2
Since xc is always less than d , we choose
l2 (d − l1 cos θ )
xc = d − + .
d 2 + l12 − 2dl1 cos θ
The y -coordinate of C is now

-
,
l22 (d − l1 cos θ )2 ±l1 l2 | sin θ |
yc = ± l22 − (xc − d)2 = ± l22 − 2 2
= + .
d + l1 − 2dl1 cos θ d 2 + l12 − 2dl1 cos θ
We can see that yc is positive when θ is an angle in the first or second quadrant, and yc is
negative when θ is in the third or fourth quadrant. Hence
l1 l2 sin θ
FIGURE 9.11 Motion of C yc = + .
d2 + l12 − 2dl1 cos θ
y
1 When l1 = 1/2, l2 = 2, and d = 1, parametric equations for the path followed by C are
2[1 − (1/2) cos θ ] 2(2 − cos θ )

xc = 1 − √ = 1− √ ,
1 + 1/4 − 2(1/2) cos θ 5 − 4 cos θ
−1 1 x
−
2 (1/2)(2) sin θ 2 sin θ
yc = √ = √ .
1 + 1/4 − 2(1/2) cos θ 5 − 4 cos θ
−1
A plot of the curve for 0 ≤ θ ≤ 2π is shown in Figure 9.11.
EXAMPLE 9.7
The tire of a car rolling along the x -axis without slipping picks up a stone at the origin. The
path followed by the stone is called a cycloid. Show that parametric equations for the cycloid
in terms of the angle θ through which the tire has rotated since picking up the stone are
x = R(θ − sin θ ), y = R(1 − cos θ ),
where R is the radius of the tire. Plot the cycloid for 0 ≤ θ ≤ 4π . In what direction is the
stone travelling when it meets the road?
SOLUTION The x -coordinate of P in Figure 9.12a is equal to (OB( minus (P C( . Since
length (P B( along the circle is equal to (OB( and (P C( = (P A( sin θ = R sin θ , it
follows that
x = Rθ − R sin θ = R(θ − sin θ ).
Furthermore,
y = (AB( − (AC( = R − R cos θ = R(1 − cos θ ).
FIGURE 9.12a Development of path followed by stone caught in tread of a tire
Stone Tire
A
R
P (x, y) C
0 B x
FIGURE 9.12b Path followed by stone for two revolutions of tire
y
2R
x
The plot in Figure 9.12b for 0 ≤ θ ≤ 4π shows the path of the stone for two revolutions
of the tire. (It was plotted for R = 1 and scales were then changed from 1 to R .)
The slope of the cycloid is given by
dy dy/dθ R(sin θ ) sin θ

= = = .
dx dx/dθ R(1 − cos θ ) 1 − cos θ
It is undefined when θ = 2nπ , values of θ at which the stone meets the road. Using L’Hôpital’s
rule, we calculate
sin θ cos θ sin θ
lim = lim = −∞ and lim = ∞.
θ →2nπ − 1 − cos θ θ →2nπ − sin θ θ →2nπ + 1 − cos θ
These show that the stone is moving vertically downward as it meets the road, and then vertically
upward as it leaves the road.
Many of the applications of integration in Chapter 7 can be adapted to curves defined paramet-
rically. We illustrate in the next example and the following consultation project.
EXAMPLE 9.8
Find the area bounded by the curve in Example 9.5.
SOLUTION Because of the symmetry of the curve, the area is four times that in the first
quadrant (Figure 9.13). If we use vertical rectangles of area y dx , then the required area is
FIGURE 9.13 Area bounded by

parametrically defined curve
x
dx
. a . 0 . π/2
A=4 y dx = 4 b sin 2t(−a sin t) dt = 4ab sin t sin 2t dt.
0 π/2 0
With identity 1.48a,

. π/2 / 0π/2
1 1 8
A = 4ab (− cos 3t + cos t) dt = 2ab − sin 3t + sin t = ab.
0 2 3 0 3
Integrals 7.17 in Section 7.3 define the length of a curve. We express them in a slightly different
form,
. B +
L= (dx)2 + (dy)2 ,
A
FIGURE 9.14 Length of a

curve defined parametrically
where A and B are the initial and final points on the curve (Figure 9.14). When the curve is
defined parametrically by 9.1, this integral becomes
y -#
B (b, d )
. β $2 # $2
dx dy
L= + dt. (9.3)
α dt dt
A (a, c) Verification of this formula is left to Exercise 49. We use it in the following project.
x
In a rolling mill, the right circle in Figure 9.15a represents a cylinder of radius R that rolls
around a second cylinder of the same radius. Attached to point P on the end of the right
cylinder is part of a linkage P Z . Many questions are being asked about the motion of
point P , one of which is the distance that it travels as the right cylinder rolls once around
the left cylinder. These questions can be answered if the equation for the curve followed
by P can be found. Our task is to find it.

Distance travelled by a point on one circle as it rolls around another circle
y y
S
R R
Qρθ
ρ
R
θW P(x,y)
P x O U T V x
φ = angle WQP
Z
SOLUTION Suppose we let θ be the angle though which the centre of the right cylinder
rotates as shown in Figure 9.15b. At this position, we can say that the x -coordinate of P is
x = (OU ( + (U V ( = R cos θ + (P Q( sin φ.
Angles θ , φ , and ρ are related by the equations
(π/2 − θ ) + φ + ρ = π, and θ + 2ρ = π.
When these are solved for ρ and results are equated, we obtain φ = 3θ/2. Further-
more, if angle θ is bisected at S to divide triangle P QS into two congruent right-angled
triangles, we see that (P Q(/2 = R sin (θ/2) . Hence,
x = R cos θ + 2R sin (θ/2) sin (3θ/2)
= R(cos θ − cos 2θ + cos θ ) = R(2 cos θ − cos 2θ ).
The y -coordinate of P is
y = (U Q( − (QW ( = R sin θ − (P Q( cos φ
= R sin θ − 2R sin (θ/2) cos (3θ/2)
= R(sin θ + sin θ − sin 2θ ) = R(2 sin θ − sin 2θ ).
We have therefore found parametric equations for the path of P ,
x = R(2 cos θ − cos 2θ ), y = R(2 sin θ − sin 2θ ).
According to formula 9.3, the distance that P travels for one complete rotation of the
cylinder is
. π +
L=2 R 2 (−2 sin θ + 2 sin 2θ )2 + R 2 (2 cos θ − 2 cos 2θ )2 dθ
0
√ . + √ .
π π √
= 4 2R 1 − (cos θ cos 2θ + sin θ sin 2θ ) dθ = 4 2R 1 − cos θ dθ
0 0
√ . π , . π
2
= 4 2R 1 − [1 − 2 sin (θ/2)] dθ = 8R sin (θ/2) dθ = 16R.
0 0
EXERCISES 9.1
# $ 1/ 3 # $4
In Exercises 1–12 first draw the curve and then plot it. 1+u 1−u
19. x = , y=
1−u 1+u
1. x = 2 + t, y = 3t − 1
√ 1
2. x = t 2 + 3t + 4, y = 1 − t 20. x = −t 2 + 3t + 5, y =
t 2 + 2t − 5
3. x = 1 + 2 cos t, y = 2 + 2 sin t, 0 ≤ t ≤ 2π
4. x = −2 + 4 cos t, y = 3 + 4 sin t, 0≤t ≤π 21. Find equations for the tangent and normal lines to the curve
5. x = 1 + cos t, y = −1 − sin t, 0 ≤ t ≤ π/4 1 1

x =t+ , y =t−
6. x = t + 1/t, y = t − 1/t t t
7. x = 2 cos t, y = 4 sin t, 0 ≤ t < 2π at the point corresponding to t = 4.
8. x = 1 + 3 cos t, y = −2 + 2 sin t, 0≤t ≤π 22. Find point(s) on the curve
9. x = −1 + sin t, y = −1 − 3 cos t, 0 ≤ t ≤ π/2 t3 3t 2

x = − 3t, y = +t
10. x = t − t 2 , y = t + t 2 3 2
where the slope of the tangent line to the curve is equal to 1.
11. x = t 2 + 1, y = t 3 + 3
12. x = 2 cot θ, y = 2 sin2 θ, −π/2 ≤ θ ≤ π/2, θ %= 0

In Exercises 23–26 assume that y is defined parametrically as a func-
tion of x , and find dy/dx and d 2 y/dx 2 .
In Exercises 13–20 assume that y is defined parametrically as a func-
1 1
tion of x , and find dy/dx . ∗ 23. x = t 2 + , y = t 2 −
t t
13. x = t 3 + 3t − 2, y = t 2 − 1 √ √
∗ 24. x = t − 1, y = t + 1
u u2
14. x = , y= 2 ∗ 25. x = 2u + 5, y = 7 − 14u
u−1 u −1
√ ∗ 26. x = v 2 + 2v + 3, y = 2v − 4
15. x = (v + 2) v − 1, y = 2v 3 + 3
3
% %
2+t 2−t ∗ 27. Is there a difference between the two curves
16. x = , y=
2−t 2+t
y = 2x 2 − 1 and x = cos t, y = 2 cos2 t − 1?
17. x = s 3/2 − s 2/3 , y = s 2 + 2s
∗ 28. What curve is described by the parametric equations
4
t
18. x = (2t + 3) , y =
t +6 x = h + a cos θ, y = k + b sin θ, 0 ≤ θ < 2π ?
∗ 29. Find parametric equations for a circle with centre (h, k) and ra- ∗ 44. The cycloid x = R(θ − sin θ) , y = R(1 − cos θ) , 0 ≤ θ ≤ 2π
dius r . of Example 9.7 and the x -axis
∗ 30. Show that the straight line through two points P1 (x1 , y1 ) and
P2 (x2 , y2 ) has parametric equations
In Exercises 45–47 find the length of the curve.
x = x1 + (x2 − x1 )t, y = y1 + (y2 − y1 )t.
45. x = 3 + 4 cos t, y = −2 + 4 sin t, 0 ≤ t < 2π
∗ 31. Show that the equations x = 2 sin2 t , y = 4 cos2 t , which are not ∗ 46. x = e −t
sin t, y = e −t
cos t, 0≤t ≤1
linear, define a straight-line segment.
∗ 47. x = t + ln t, y = t − ln t, 1≤t ≤2
∗ 32. Draw the following curves and determine whether they are related:
(a) x = sec θ, y = tan θ, −π/2 < θ < π/2
48. Set up, but do not evaluate, a definite integral representing the
(b) x = cosh φ, y = sinh φ length of the ellipse
# $ # $
1 1 1 1
(c) x = t+ , y= t− , t ≥0
2 t 2 t x = a cos θ, y = b sin θ, 0 ≤ θ < 2π.
∗ 49. Verify formula 9.3 for the length of a curve.

In Exercises 33–36 find parametric equations for the curve.
∗ 50. The equations x = t 2 + 2t − 1, y = t + 5, 1 ≤ t ≤ 4 define
x+1 a curve parametrically. Find parametric equations that describe this
33. y = 34. x + y 3 + xy = 5y 2 curve but have values of the parameter in the intervals (a) 0 ≤ t ≤ 3
x−2
and (b) 0 ≤ t ≤ 1.
∗ 35. x 2 + y 2 + 2x − 4y = 0 ∗ 36. 4 − x 2 + 2y 2 = 0
∗ 51. Suppose x(t) and y(t) in equations 9.1 are continuous on α ≤
t ≤ β and have derivatives on α < t < β , and that x , (t) %= 0 on
∗ 37. Two particles move along straight lines '1 and '2 defined para- α < t < β . Show that Cauchy’s generalized mean value theorem
metrically by (Theorem 3.18) implies that there exists a number c between α and β
such that
'1 : x = 1 − t, y = t, t ≥ 0; y(β) − y(α) y , (c)
= , .
x(β) − x(α) x (c)
'2 : x = 4t − 5, y = 2t − 1, t ≥ 0;
Interpret this result geometrically.
where t is time. When are they closest together?
∗ 52. A particle travels around the circle x 2 + y 2 = 4 counterclockwise
∗ 38. If x = x(t) and y = y(t) define y as a function of x , show that at constant speed, making 2 revolutions each second. If the particle
starts at point (2, 0) at time t = 0, find parametric equations for its
dx d 2 y dy d 2 x position in terms of t .
2
d y 2
−
= dt dt# $dt dt 2 .
2 3 ∗ 53. (a) Find the area under one arch of the cycloid in Example 9.7.
dx dx
(b) Find the length of one arch of the cycloid. What does it
dt represent physically?
∗ 54. Draw or plot the strophoid
In Exercises 39–42 find the area bounded by the curve.
1 − t2 t (1 − t 2 )
39. The ellipse x = a cos t , y = b sin t , 0 ≤ t ≤ 2π x = , y = ,
1+ t2 1 + t2
∗ 40. The astroid x = cos3 t , y = sin3 t , 0 ≤ t ≤ 2π
and find points at which its tangent line is horizontal.
∗ 41. The deltoid x = 2 cos t + cos 2t , y = 2 sin t − sin 2t , 0 ≤ t ≤
2π ∗ 55. Plot the path followed by joint C in the mechanism of Example
∗ 42. The droplet x = 2 cos t − sin 2t , y = sin t , 0 ≤ t ≤ 2π 9.6 when l1 = 1, l2 = 3, and d = 1/2.
∗ 56. (a) Find the x -coordinate of slider E in Example 9.6.
In Exercises 43–44 find the volume of the solid of revolution when the (b) Plot xE as a function of θ when l1 = 1/2 m, l2 = 2 m,
area bounded by the curve(s) is rotated about the x -axis. l3 = 4 m, d = 1 m, and yE = 2 m. From the graph
estimate the length of the stroke of E (the length of the line
∗ 43. The curve x = a cos t , y = b sin 2t of Example 9.5 segment along which E moves).
∗ 57. End A of shaft AB with length 4 cm (figure below), moves around

the circle of radius 1 cm counterclockwise at 60 rpm. As it does so, y
a slider at end B moves back and forth along the line y = −3 cm.
Assume that there is no binding when A is at positions (0, ±1) so that Stone
b Tire
the slider moves between quadrants three and four at these times. R
y (x, y)
1
A x
0
1
x ∗ 59. A string is wound around the circle x 2 + y 2 = r 2 in the figure
4 below with one end at (r, 0) . If the string is unwound while being held
taut, the curve that the end traces is called an involute of the circle.
Show that parametric equations for the involute in terms of the angle θ
B shown are
y = −3
x = r cos θ + rθ sin θ, y = r sin θ − rθ cos θ.
(a) Given that A starts at position (1, 0) at time t = 0, find
and plot a formula for the x -coordinate of B in terms of t y
for two revolutions of A . Hint: It is necessary to piece a
function together.
(b) Estimate maximum left and right positions of the slider from
the graph in part (a), and then find these positions exactly r
by using the fact that velocity will be zero there. Q String
P(x, y)
(c) Is the velocity function continuous as the slider passes be-
tween quadrants three and four? r x
(d) Use a graph of the velocity function to estimate maximum

speed of the slider.
(e) Use a graph of the acceleration function to estimate maxi- In Exercises 60–62 try to draw the curve and then plot it.
mum |a(t)| .
∗ 60. x = cos θ, y = sin 3θ (curve of Lissajous)
∗ 58. If the stone in Example 9.7 is embedded in the side of the tire rather 2
than the tread, its path is called a trochoid (figure following). Show that 3t 3t
∗ 61. x = , y= (folium of Descartes)
if the distance from the centre of the tire to the stone is b , parametric 1 + t3 1 + t3
equations for the trochoid are ∗ 62. x = cos3 θ, y = sin3 θ (astroid or hypocycloid of four cusps)
x = Rθ − b sin θ, y = R − b cos θ. ∗∗ 63. A cow is attached to the side of a silo of radius 5 m with a rope of
10 m. Determine the grazing area of the cow.
9.2 Polar Coordinates

In this section we introduce polar coordinates, an alternative coordinate system for the plane.
Many problems that have complex solutions using Cartesian coordinates become much simpler
in polar coordinates.
FIGURE 9.16 Polar coor-
Polar coordinates are defined by choosing a point O in the plane called the pole and a
dinates of a point half-line originating at O called the polar axis (Figure 9.16). If P is a point in the plane, we
join O and P . The first polar coordinate of P , denoted by r , is the length of line segment
P OP . The other polar coordinate is the angle θ through which the polar axis must be rotated
r to coincide with line segment OP . Counterclockwise rotations are regarded as positive, and
clockwise rotations as negative. In Figure 9.17, position OQ is reached through a positive
O rotation of π/6 radians; therefore, for point Q , θ = π/6. But clearly we could arrive at this
Pole Polar axis position in many other ways. We could, for instance, rotate the polar axis counterclockwise
9.2 Polar Coordinates 555
through any number of complete revolutions, bringing it back to its original position, and then
FIGURE 9.17 Polar coor- rotate a further π/6 radians. Alternatively, we could rotate in a clockwise direction any number
dinates of two specific points of complete revolutions, and then a further −11π/6 radians. In other words, polar coordinate
R Q
θ for Q could be any of the values π/6 + 2nπ , where n is an integer. Possible values of θ
2 for point R in Figure 9.17 are 3π/4 + 2nπ . For point Q , r = 2, and for R , r = 1. Polar
4 coordinates r and θ for a point are written in the form (r, θ ) so that possible polar coordinates
1 6 for Q and R are (2, π/6 + 2nπ ) and (1, 3π/4 + 2nπ ) .
O This situation is not like that for Cartesian coordinates, where each point has only one set
of coordinates (x, y) , and every ordered pair of real numbers specifies one point. With polar
coordinates, every ordered pair of real numbers (r, θ ) , where r must be nonnegative, represents
one and only one point, but every point has an infinity of possible representations. We should
point out that in some applications this is not a desirable situation. For instance, in the branch
of mathematics called tensor analysis, it is necessary that polar coordinates assign exactly one
FIGURE 9.18 Relation-
pair of coordinates to each point. This can be accomplished in any region that does not contain
ships between polar and Cartesian the pole by demanding, for instance, that −π < θ ≤ π . When we use polar coordinates to
coordinates find areas in Section 9.4, we must also be particular about our choice of θ . For now, however,
no advantage is gained by imposing restrictions on θ , and we therefore accept the fact that if
y
(r, θ ) are polar coordinates of a point, so are (r, θ + 2nπ ) for any integer n . Note also that
P
polar coordinates for the pole are (0, θ) for any θ whatsoever.
If we introduce into a plane both a system of Cartesian coordinates (x, y) and a system
r of polar coordinates (r, θ ) , then relations exist between the two. Suppose the pole of polar
y
coordinates and the origin of Cartesian coordinates are chosen as the same point, and that the
polar axis is chosen as the positive x -axis (Figure 9.18). In this case, Cartesian and polar
x Polar axis x coordinates of any point P are related by the equations
Pole
x = r cos θ, y = r sin θ. (9.4)

dinates of point (1, 1)
These equations define Cartesian coordinates of a point in terms of its polar coordinates; that
y is, given its polar coordinates (r, θ ) , we can calculate its Cartesian coordinates (x, y) by means
(1, 1)
of 9.4. For example, if polar coordinates of a point are (3, 2) , then its Cartesian coordinates are
1
r x = 3 cos 2 = −1.25, y = 3 sin 2 = 2.73.
Equations 9.4 implicitly define polar coordinates of a point in terms of its Cartesian coordinates.
x
For instance, if Cartesian coordinates of a point are (1, 1) (Figure 9.19), its polar coordinates
1
must satisfy
1 = r cos θ, 1 = r sin θ.
If we square and add these equations, we have
1 + 1 = r 2 cos2 θ + r 2 sin2 θ = r 2 ,
√ √
and therefore r = 2. It follows that cos θ = sin √ θ = 1/ 2, from which we get θ =
π/4 + 2nπ . Thus, polar coordinates of the point are ( 2, π/4 + 2nπ ) .
Equations 9.4 define r and θ implicitly in terms of x and y , but obviously it would be
preferable to have explicit definitions. There is no problem expressing r explicitly in terms of
x and y ,
+
r = x2 + y2, (9.5)
but the case for θ is not so simple. If we substitute expression 9.5 for r into equations 9.4, we
obtain
x y
cos θ = + , sin θ = + . (9.6)
x2 + y2 x2 + y2
Except for the pole, these two equations determine all possible values of θ for given x and y .
What we would like to do is obtain one equation, if possible, that defines θ . If we divide the
second of these equations by the first, we have
y
tan θ = , (9.7)
x
and this equation suggests that we set
&y '
θ = Tan −1 . (9.8)
x
dinates of (−1, 1)
Unfortunately, neither equation 9.7 nor 9.8 is satisfactory. For instance, given the point P with
P (−1, 1) y Cartesian coordinates (−1, 1) (Figure 9.20), equation 9.7 yields tan θ = −1, the solutions of
which are θ = −π/4 + nπ . These are angles in the second and fourth quadrants, so only
half of them are acceptable polar angles for P . Equation 9.8 gives θ = −π/4, which is not a
possible polar angle for P .
x
We suggest that all angles satisfying 9.7 be found, and then a diagram be used to determine
those angles that are acceptable values for θ . We illustrate in the following example.
EXAMPLE 9.9
Find all polar coordinates for points with the following Cartesian coordinates:
(a) (1, 2) (b) (−2, 3) (c) (3, −1) (d) (−2, −4)

dinates of four points SOLUTION
√ √
y (a) r = 12 + 22 = 5. Angles that satisfy tan θ = 2/1 are 1.11 + nπ . Since the
point is in the first quadrant (Figure 9.21), acceptable
√ values for θ are 1.11 + 2nπ .
(−2, 3) (1, 2) Polar coordinates of the point are therefore ( 5, 1.11 + 2nπ ) .
+ √
x (b) r = (−2)2 + 32 = 13. Angles satisfying tan θ = −3/2 are −0.98 + nπ .
(3, −1) Since the point is in the second quadrant, possible√values for θ are (π −0.98)+2nπ =
(−2, −4) 2.16 + 2nπ . Polar coordinates are therefore ( 13, 2.16 + 2nπ ) .
+ √
(c) r = 32 + (−1)2 = 10. Since values of θ satisfying tan θ = −1/3 are
−0.32√+ nπ , and the point is in the fourth quadrant, it follows that polar coordinates
are ( 10, −0.32 + 2nπ ) .
+ √
(d) r = (−2)2 + (−4)2 = 2 5. Since angles satisfying tan θ√= 2 are 1.11 + nπ ,
and the point is in the third quadrant, polar coordinates are (2 5, 4.25 + 2nπ ) .
The results in equations 9.4–9.8 are valid only when the pole and origin coincide and the polar
axis and positive x -axis are identical. For a different arrangement, these relations must be
changed accordingly. For example, if the pole is at the point with Cartesian coordinates (h, k)
and the polar axis is as shown in Figure 9.22, equations 9.4 are replaced by

dinates with pole at (h, k)
y P
r Polar
axis
(h, k)
Pole Horizontal
line
x = h + r cos (θ + α), y = k + r sin (θ + α). (9.9)

You will verify these equations in Exercise 13.
The usual choice of pole and polar axis is that in Figure 9.18, and unless otherwise stipulated,
we assume this to be the case.
EXERCISES 9.2
In Exercises 1–8 plot the point having the given set of Cartesian coor- In Exercises 9–12 plot the point having the given set of polar coordi-
dinates, and find all possible polar coordinates. nates, and find its Cartesian coordinates.
√
1. (1, −1) 2. (−1, 3) 9. (2, π/4) 10. (6, −π/6)
√
3. (4, 3) 4. (−2 3, 2) 11. (7, 1) 12. (3, −2.4)
5. (2, 6) 6. (−1, −4) 13. Verify that polar and Cartesian coordinates as shown in Figure 9.22
7. (7, −5) 8. (−5, 2) are related by equations 9.9.
9.3 Curves in Polar Coordinates

A curve is defined explicitly in Cartesian coordinates by equations y = f (x) or x = g(y) ,
and implicitly by F (x, y) = 0. A point is on a curve if and only if its Cartesian coordinates
(x, y) satisfy the equation of the curve.
Analogously, a curve is defined explicitly in polar coordinates if its equation is expressed
in either of the forms
r = f (θ ) or θ = g(r), (9.10)
and is defined implicitly when its equation is given in the form
F (r, θ ) = 0. (9.11)
A point is on a curve if at least one of its sets of polar coordinates (r, θ ) satisfies the equation of
the curve. All sets of polar coordinates for a point need not satisfy the equation. For example,
the origin or!pole has polar"coordinates (0, θ) for any θ whatsoever. But only those coordinates
of the form 0, (2n + 1)π satisfy the equation r = 1 + cos θ . Likewise, the polar coordinates
(1, π) satisfy the equation r = sin (θ/2) , but the coordinates (1, 3π ) of the same point do not
satisfy this equation.
Often we are required to transform the equation of a curve from Cartesian coordinates to
polar coordinates, and vice versa. To transform from Cartesian to polar is straightforward:
Replace each x with r cos θ and each y with r sin θ . For example, the equation x 2 + y 2 = 9
describes a circle centred at the origin with radius 3. In polar coordinates, its equation is
9 = (r cos θ )2 + (r sin θ )2 = r 2 cos2 θ + r 2 sin2 θ = r 2 .
Consequently, r = 3 is the equation of this circle in polar coordinates, a much simpler equation
than x 2 + y 2 = 9.
EXAMPLE 9.10
Find equations in polar coordinates for the following curves:
+
(a) 2x + 3y = 3 (b) x 2 − 2x + y 2 = 0 (c) x2 + y2 = x 2 + y 2 − 4x
SOLUTION
(a) For 2x + 3y = 3, we obtain
3
3 = 2r cos θ + 3r sin θ or r = .
2 cos θ + 3 sin θ
(b) For the equation x 2 − 2x + y 2 = 0, we have
0 = −2x + (x 2 + y 2 ) = −2r cos θ + r 2 = r(r − 2 cos θ ).
Thus,
r =0 or r = 2 cos θ.
Since r = 0 defines the pole, and this point also satisfies r = 2 cos θ (for θ = π/2),
it follows that we need only write r = 2 cos θ .
+
(c) For the curve with equation x 2 + y 2 = x 2 + y 2 − 4x , we obtain
r 2 = r − 4r cos θ = r(1 − 4 cos θ ).

Thus,
0 = r 2 − r(1 − 4 cos θ ) = r(r − 1 + 4 cos θ ),
from which we have
r =0 or r = 1 − 4 cos θ.
Again the pole satisfies the second of these equations, and therefore the equation of
the curve in polar coordinates is r = 1 − 4 cos θ .
To transform equations of curves from polar to Cartesian coordinates can sometimes be more
difficult, principally because we have no substitution for θ . If, however, the equation involves
cos θ and/or sin θ , we use equations 9.6.
EXAMPLE 9.11
Find equations in Cartesian coordinates for the following curves:
(a) r = 1 + cos θ (b) r 2 cos 2θ = 1 (c) r 2 = 9 sin 2θ
SOLUTION
(a) We use equations 9.5 and 9.6 to write
+ x
x2 + y2 = 1 + + ,
x2 + y2
+
and multiplication by x 2 + y 2 gives
+
x2 + y2 = x + x2 + y2.
(b) For r 2 cos 2θ = 1, we use double-angle formula 1.46b to write the equation in terms
of cos θ rather than cos 2θ , and then use equations 9.5 and 9.6:
# $
x2 2
2 x − y
2
1 = r 2 (2 cos2 θ − 1) = (x 2 + y 2 ) 2 − 1 = (x 2
+ y ) = x2 − y2.
x2 + y2 x2 + y2
(c) This time we use double-angle formula 1.45 on sin 2θ :

y x
x 2 + y 2 = 18 sin θ cos θ = 18 + + or
x2 + y2 x2 + y2
(x 2 + y 2 )2 = 18xy.
Examples 9.10 and 9.11 illustrate that equations for some curves are simpler when expressed
in polar coordinates. These polar representations can prove very efficient in producing graphs,
whether we are plotting with an electronic device or drawing by hand. We illustrate with the
curve r = 1 + cos θ of Example 9.11. To plot this curve by graphing calculator or computer,
we supply the function r = 1 + cos θ and the range of values of θ . Because 1 + cos θ is
2π -periodic, the range −π ≤ θ ≤ π suffices. Other values of θ create duplications. The
result is shown in Figure 9.23; it is called a cardioid.
To draw the cardioid by hand we first create a Cartesian coordinate system consisting of a
horizontal θ -axis and a vertical r -axis. On this set of axes we graph the function r = f (θ ) =
1 + cos θ (Figure 9.24). This is not the required curve; it is a graph of the function f (θ ) ,
illustrating values of r for various values of θ . It represents an “infinite table of values” for r
as a function of θ .
FIGURE 9.23 Plot of cardioid in polar coordinates FIGURE 9.24 Cartesian graph of r = 1 + cos θ
r
2
2.0
1.0
1.5
0.5
1.0
0.5 1.0 1.5 2.0 0 0.5
−0.5
−
2 2
−1.0
−
2
To draw the cardioid r = 1 + cos θ , we now read pairs of polar coordinates (r, θ ) from
Figure 9.24, interpreting r as radial distance and θ as rotation. Suppose we begin with the two
points (2, 0) and (1, π/2) . Figure 9.24 indicates that as θ increases from 0 to π/2, values of
r decrease from 2 to 1. This means that as we rotate from the θ = 0 line to the θ = π/2
line in the first quadrant, radial distances from the origin become smaller. This is shown in
Figure 9.25a. As rotation is increased from π/2 to π , Figure 9.24 shows that radial distances
continue to decrease, eventually reaching 0 at an angle of π radians as in Figure 9.25b. Notice
that the line θ = π is tangent to the curve at the pole, reflecting the fact that the pole is attained
for an angle of π radians. Consideration of the graph in Figure 9.24 to the left of θ = 0 leads
to that part of the cardioid below the θ = 0 and θ = π lines in Figure 9.25c. The symmetry
of Figure 9.24 about the r -axis is reflected in the symmetry of Figure 9.25c about the θ = 0
and θ = π lines. Since the function r = f (θ ) = 1 + cos θ is 2π -periodic, only values of θ
in the interval −π < θ ≤ π need be considered. Values outside this interval retrace previous
points.
FIGURE 9.25a Plot of FIGURE 9.25b Plot of FIGURE 9.25c Plot of

π
r = 1 + cos θ from 0 to 2
r = 1 + cos θ from 0 to π r = 1 + cos θ from −π to π
2 2 2
1.0 1.0 1.0
0.5 0.5 0.5
0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 0
−0.5
−1.0
−
2
EXAMPLE 9.12
Plot and draw the curve r 2 = 9 sin 2θ .
√
SOLUTION The explicit definition of the curve is r = 3 sin 2θ . For sin 2θ to be nonnega-
tive, θ must be restricted to the intervals −π ≤ θ ≤ −π/2 and 0 ≤ θ ≤ π/2. When these
are submitted to whatever electronic device you use to make polar plots, the result is as shown
in Figure 9.26.
Lacking a device that does polar plots we can draw the curve √
by first drawing a graph of the
function sin 2θ in Figure 9.27a. A graph of the function r = 3 sin 2θ then follows (Figure
9.27b). Reading pairs of polar coordinates from this graph gives the curve in Figure 9.26. It is
called a lemniscate.
FIGURE 9.26 Polar plot of lemniscate
2
2
−2 −1 1 2 0
−1
−2
−
2
√
FIGURE 9.27a Cartesian graph of sin 2θ FIGURE 9.27b Cartesian graph of r = 3 sin 2θ
r r
1 3
−
2 2 −
2 2
−1
EXAMPLE 9.13
Draw the curve r = |a − b cos θ | , where b > a > 0 are constants.
SOLUTION We begin by drawing the function −b cos θ in Figure 9.28a. A shift of a units
vertically gives the graph in Figure 9.28b. Absolute values lead to Figure 9.28c. Interpreting r
and θ as distance and rotation leads to the curve in Figure 9.28d. Angles at which r = 0 are
θ = ±Cos −1 (a/b) .
FIGURE 9.28a Cartesian graph of −b cos θ FIGURE 9.28b Cartesian graph of a − b cos θ
r r
b a+b
−
−b a−b
FIGURE 9.28c Cartesian graph of r = |a − b cos θ| FIGURE 9.28d Polar graph of r = |a − b cos θ|
r
a+b
−a − b b−a
b−a
0
−a
−
To find points of intersection of two curves whose equations are given in Cartesian coordinates,
we solve the equations simultaneously for all (real) solutions. Each solution represents a distinct
point of intersection. For curves whose equations are given in polar coordinates, the situation is
somewhat more complex because we have multiple names for points. To find points of intersec-
tion, we again solve the equations simultaneously for all solutions. Each solution represents a
point of intersection; but as points have many sets of polar coordinates, some of these solutions
may represent the same point. In addition, it may also happen that one set of polar coordinates for
a point of intersection satisfies one equation, whereas a different set satisfies the other equation.
Particularly troublesome in this respect is the pole, which has so many sets of polar coordinates.
The best way to handle these difficulties is to graph the curves.
EXAMPLE 9.14
Find points of intersection for the curves r = sin θ and r = 1 − sin θ .
SOLUTION If we set sin θ = 1 − sin θ , then sin θ = 1/2. All solutions of this equation are
defined by  π

+ 2nπ

6
θ = 5π

 + 2nπ
6
where n is an integer. Graphs of the curves in Figure 9.29 indicate that these values of θ give
the points of intersection (1/2, π/6) and (1/2, 5π/6) . The figure also indicates that the origin
is a point of intersection of the curves. We did not obtain this point by solving r = sin θ and
r = 1 − sin θ because different values of θ yield r = 0 in r = sin θ and r = 1 − sin θ . To
obtain r = 0 from r = sin θ , θ must be one of the values nπ , whereas to obtain r = 0 from
r = 1 − sin θ , θ must be one of the values π/2 + 2nπ . Thus, both curves pass through the
pole, but the pole cannot be obtained by solving the equations of the curves simultaneously.
FIGURE 9.29 Intersection points of polar curves r = sin θ and r = 1 − sin θ
0.5
1 0.5 0.5 1 0
−0.5
−1
−1.5
−2
−
Slopes of Curves in Polar Coordinates

When a curve has polar equation r = f (θ ) , α ≤ θ ≤ β , substitution into equations 9.4 gives
x = f (θ ) cos θ, y = f (θ ) sin θ, α ≤ θ ≤ β. (9.12)
These are parametric equations for the curve, where the parameter is the polar angle θ . Equation
9.2 gives
dy
dy dθ f , (θ ) sin θ + f (θ ) cos θ
= = , . (9.13)
dx dx f (θ ) cos θ − f (θ ) sin θ
dθ
This formula defines the slope of the tangent line to a curve, which has polar equation r = f (θ ) .
EXAMPLE 9.15
Find points on the cardioid r = 1 + sin θ at which the tangent line is horizontal.
SOLUTION The graph of the cardioid in Figure 9.30 indicates three points at which the tangent
line is horizontal. To find them we use equation 9.13 to write
dy (cos θ ) sin θ + (1 + sin θ ) cos θ
0 = = .
dx (cos θ ) cos θ − (1 + sin θ ) sin θ
Since the numerator must vanish, we set
0 = cos θ sin θ + cos θ + sin θ cos θ = cos θ (1 + 2 sin θ ).
From cos θ = 0, we choose θ = π/2 and from 1 + 2 sin θ = 0, we take θ = −π/6

and −5π/6. Thus, points√at which the cardioid has a horizontal tangent line have Cartesian
coordinates (0, 2) and (± 3/4, −1/4) .
FIGURE 9.30 Points at which tangent line to r = 1 + sin θ is horizontal
1.5
0.5
1 0.5 0.5 1 0
−
Lengths of Curves in Polar Coordinates

The length of a curve defined parametrically by x = x(t) and y = y(t) is given by formula
9.3: -
. β # $2 # $2
dx dy
L= + dt.
α dt dt
If we substitute from equations 9.12 into this formula with t replaced by θ , we obtain
. β +
L= [f , (θ ) cos θ − f (θ ) sin θ ]2 + [f , (θ ) sin θ + f (θ ) cos θ ]2 dθ
α
. β +
= [f , (θ )]2 + [f (θ )]2 dθ .
α
Thus, we may write for the length of a curve r = f (θ ) , α ≤ θ ≤ β ,

- # $2
. β
dr
L= r2 + dθ . (9.14)
α dθ
EXAMPLE 9.16
Find the length of the cardioid r = 1 − cos θ .
SOLUTION According to equation 9.14 (see Figure 9.31),
. 2π +
L= (1 − cos θ )2 + (sin θ )2 dθ
0
. 2π √ √
= 2 1 − cos θ dθ
0
- 5 # $6
√ . 2π
θ
= 2 1 − 1 − 2 sin2 dθ
0 2
. 2π # $ / # $02π
θ θ
=2 sin dθ = 2 −2 cos = 8.
0 2 2 0
FIGURE 9.31 Length of polar curve r = 1 − cos θ
0.5
2 1.5 1 0.5 0
0.5
EXAMPLE 9.17
The plate cam in Figure 9.32 rotates about an axis through the origin and perpendicular to the
plate. The follower moves back and forth along the x -axis as point A on its end remains in
contact with the cam. Suppose r = a + b cos θ , where a > b > 0 are constants, is the polar
equation of the edge of the cam, and that the cam rotates at ω revolutions per second. Show
that the follower exhibits simple harmonic motion (called a harmonic cam) and find a formula
for its velocity.
FIGURE 9.32 Velocity of follower in a plate cam
r = a + bcosθ
Follower
x
A
Cam
SOLUTION As the cam rotates, the value of r represents the x -coordinate of A ; that is,
x = r = a + b cos θ . If we choose time t = 0 when the cam is in the position shown, then
θ = 2π ωt , and x(t) = a + b cos (2π ωt) . This represents simple harmonic motion for the
motion of A . The velocity is
dx
v(t) = = −2π bω sin (2π ωt).
dt
EXERCISES 9.3
In Exercises 1–10 find an equation for the curve in polar coordinates. In Exercises 22–25 find all points of intersection for the curves.
Draw, and then plot, the curve.
22. r = 2, r 2 = 8 cos 2θ
1. x + 2y = 5 23. r = cos θ, r = 1 + cos θ
24. r = 1 + cos θ, r = 2 − 2 cos θ
2. y = −x
25. r = 1, r = 2 cos 2θ
3. x 2 + y 2 = 3
In Exercises 26–29 find the slope of the curve at the given value of θ .
4. x 2 − 2x + y 2 − 2y + 1 = 0
26. r = 9 cos 2θ at θ = π/6
5. y = 4x 2 6. x 2 + 2y 2 = 3
27. r 2 = 9 sin 2θ at θ = −5π/6
+
2 2 2 2
7. x + y = x 8. x + y = x 2 + y 2 − x 28. r = 3 − 5 cos θ at θ = 3π/4
2
9. (x + y ) = x 2 2
10. y = 1/x 2 29. r = 2 cos (θ/2) at θ = π/2
30. Find the slope of the tangent line to the curve r = 3/(1 − sin θ)
at the point with polar coordinates (6, π/6) in two ways: (a) by using
In Exercises 11–20 find the equation of the curve in Cartesian coordi- 9.13; (b) by finding the equation of the curve in Cartesian coordinates,
nates. Draw, and then plot, the curve. and calculating dy/dx .
∗ 31. Show that if f (θ) is an even function, then the curve r = f (θ)
11. r = 5 12. θ = 1 is symmetric about the lines θ = 0 and θ = π (or x -axis). Illustrate
13. r = 3 sin θ 14. r 2 = 4 sin 2θ with two examples.
15. r = 3 + 3 sin θ 16. r = 2 sin 2θ

In Exercises 32–39 draw, and then plot, the curve.
17. r 2 = −4 cos 2θ 18. r = 3 − 4 cos θ
∗ 32. r = sin 3θ ∗ 33. r = cos 2θ
19. r = 5 csc θ 20. r = cot 2 θ csc θ
∗ 34. r = sin 4θ ∗ 35. r 2 = θ
∗ 36. r = eθ ∗ 37. r = 2 sin (θ/2)
21. Draw and plot the curves (a) r = 2 + 2 sin θ , (b) r = 2 + 4 sin θ ,
and (c) r = 4 + 2 sin θ . ∗ 38. r = −2 cos (θ/2) ∗ 39. r = 1 + cos (θ + π/6)
∗ 40. At what times and positions is the speed of the follower in Example ∗ 49. (a) Show that the polar equivalent for the equation of the circle
9.17 maximum and minimum? (x − a)2 + y 2 = R 2 is
∗ 41. Find the length of the cardioid r = a(1 + sin θ) . ( a is a constant.) +
r = a cos θ ± R 2 − a 2 sin2 θ.
∗ 42. (a) The electrostatic charge distribution consisting of a charge
q > 0 at the point with polar coordinates (s, 0) and a (b) Show that when a = R the equation reduces to r =
charge −q at (s, π ) is called a dipole. When s is very 2a cos θ . Does this represent the entire circle?
small, the lines of force for the dipole are defined by the (c) Do you need both equations in part (a) to describe the en-
equation r = A sin2 θ , where each value of the constant tire circle when a > R ? If so, which part of the circle
A > 0 defines a particular line of force. Plot or draw lines corresponds to which equation?
of force for A = 1, 2, and 3.
(d) Repeat part (c) when a < R .
(b) The equipotential lines for the dipole are defined by r 2 =
B cos θ , where B is a constant. Plot or draw equipotential ∗∗ 50. (a) A patrol boat at point A in the figure below spots a subma-
lines for B = ±1, ±2, and ±3. rine submerging at point B at a time that we call t = 0. The
submarine, unaware of the patrol boat, follows a straight-
∗ 43. Draw, and then plot, the bifolium (x 2 + y 2 )2 = x 2 y . line path at constant speed v along some angle φ relative
∗ 44. Curves with equations of the form r = a(1 ± cos θ) or r = to BA (unknown to the patrol boat). The patrol boat heads
a(1 ± sin θ ) ( a > 0 a constant) are called cardioids. directly toward point B at speed V > v for k/(v + V )
(a) Draw all such curves. units of time arriving at point C . Show that the submarine
and patrol boat are equidistant from B at t = k/(v + V ) .
(b) Find equations for the cardioids in Cartesian coordinates.
(b) We set up a system of polar coordinates with B as pole
∗ 45. Curves with equations of the form r 2 = a 2 cos 2θ or r 2 = and BA as polar axis. Let the distance (BC( be de-
a 2 sin 2θ ( a > 0 a constant) are called lemniscates. noted by r0 . Suppose that the patrol boat now follows
the logarithmic
+ spiral r = r0 eθ/α still at speed V , where
(a) Draw all such curves. 2 2
α = V /v − 1. Show that the patrol boat must inter-
(b) Find equations for the lemniscates in Cartesian coordinates.
cept the submarine.
∗ 46. (a) Draw the curves r = b ± a cos θ and r = b ± a sin θ ,
where a and b are positive constants, in the three cases
a b .
(b) Find equations for the curves in Cartesian coordinates.
(c) Compare these curves with the cardioids of Exercise 44
when a = b . Path of submarine
∗ 47. A curve with equation of the form r = a sin nθ or r = a cos nθ ,
where a > 0 is a constant and n > 0 is an integer, is called a rose.
Show that the rose has n petals.
∗ 48. Show that the roses r = |a sin nθ| and r = |a cos nθ| , where B C A
a > 0 is a constant and n > 0 is an integer, have 2n petals. k
9.4 Areas in Polar Coordinates

In Section 7.1 we used definite integrals to find areas bounded by curves whose equations are
conveniently expressed in Cartesian coordinates. In this section we indicate how to find areas
bounded by curves whose equations are expressed in polar coordinates. We require the formula
FIGURE 9.33 Area of
sector of a circle
1
r 2 (θ2 − θ1 ) (9.15)
2
r for the area of the shaded sector of the circle in Figure 9.33. This formula results from the fact
that the area of the sector is the fractional part (θ2 − θ1 )/(2π ) of the area π r 2 of the circle.
Consider finding the area of the region in Figure 9.34a bounded by the radial lines θ = α
and θ = β and the curve r = f (θ ) . We divide the region into subregions by means of n + 1
radial lines θ = θi , where
α = θ0 < θ1 < θ2 < · · · < θn−1 < θn = β.

9.4 Areas in Polar Coordinates 567
FIGURE 9.34a Area bounded by curves using polar coordinates FIGURE 9.34b Approximating area with sectors of circles
2 2
Polar axis 0 Polar axis 0
On that part of the! curve r =" f (θ ) between θ = θi−1 and θ = θi , we pick any point with
polar coordinates f (θi∗ ), θi∗ as in Figure 9.34b. If between the lines θ = θi−1 and θ = θi
we draw the arc of a circle with centre at the pole and radius f (θi∗ ) , a sector is formed with area
1
)Ai = [f (θi∗ )]2 )θi ,
2
where )θi = θi − θi−1 . Since this sector approximates that part of the required area between
the radial lines θ = θi−1 and θ = θi , we can say that an approximation to the required area is
n
7 n
7 1
)Ai = [f (θi∗ )]2 )θi .
2
i=1 i=1
By increasing the number of sectors indefinitely, and at the same time requiring each of the
)θi to approach zero, we obtain a better and better approximation, and in the limit
n
7 1
area = lim [f (θi∗ )]2 )θi .
()θi (→0 2
i=1
But this limit is the definition of the definite integral of the function (1/2)[f (θ )]2 with respect
to θ from θ = α to θ = β , and we therefore write
FIGURE 9.35 Sector area
for polar coordinates . β
1
area = [f (θ )]2 dθ . (9.16)
α 2
2
In order to arrive at this integral in any given problem, without memorizing it, we use the
procedure adopted for definite integrals in Cartesian coordinates discussed in Chapter 7. We
draw at angle θ a representative sector of angular width dθ and radius r (Figure 9.35). The
area of this sector is
r 1 2 1
r dθ = [f (θ )]2 dθ.
2 2
If areas of all such sectors from angle α to angle β are added together, and the limit is taken
0
as their widths approach zero, the required area is obtained. But this is the process defined by
the definite integral, and we therefore write equation 9.16 for the area. Definite integral 9.16
exists when f (θ ) is continuous for α ≤ θ ≤ β .
EXAMPLE 9.18
Find the area inside the cardioid r = 1 + sin θ .
FIGURE 9.36 Area inside SOLUTION The area of the representative sector in Figure 9.36 is
cardioid r = 1 + sin θ 1 2 1
r dθ = (1 + sin θ )2 dθ,
2 2
2
and we must add over all sectors interior to the cardioid. Since areas on either side of the
θ = π/2 line are identical, we calculate the area to the right and double the result. To find the
area to the right of the line θ = π/2, we must identify angular positions of the first and last
2 sectors. The first sector is at the pole, and the equation of the cardioid indicates that r = 0
r when sin θ = −1, that is, when θ = −π/2 + 2nπ . But which of these values of θ shall we
choose? Similarly, the last sector occurs when r = 2, in which case sin θ = 1, and θ could be
1 any of the values π/2 + 2nπ . Again, which shall we choose? If we choose α = −π/2 and
1 0 β = π/2, then all values of θ in the interval −π/2 ≤ θ ≤ π/2 yield points on the right half
of the cardioid with no duplications. Consequently,
. π/2 . π/2
1 2
area = 2 (1 + sin θ ) dθ = (1 + 2 sin θ + sin2 θ ) dθ
−π/2 2 −π/2
. π/2 # $
1 − cos 2θ
= 1 + 2 sin θ + dθ
−π/2 2
/ 0π/2
3θ sin 2θ 3π
= − 2 cos θ − = .
2 4 −π/2 2
EXAMPLE 9.19
Find the area common to the circles x 2 + y 2 = 4 and x 2 + y 2 = 4x .
FIGURE 9.37 Area com- SOLUTION Equations for the circles in polar coordinates are r = 2 and r = 4 cos θ , and
mon to circles x 2 + y 2 = 4 and they intersect in the points with polar coordinates (2, ±π/3) (Figure 9.37). If As is the area
x 2 + y 2 = 4x above the x -axis, outside x 2 + y 2 = 4 and inside x 2 + y 2 = 4x , then the area common to the
circles is the area of either circle less twice As :
2 area = π(2)2 − 2As .
The area of the representative element is the difference in the areas of two sectors:
2,
x 2 + y2 = 4, 3
As 1 1
r=2 (4 cos θ )2 dθ − (2)2 dθ = 2(4 cos2 θ − 1) dθ.
2 2
Since all sectors in As can be identified by values of θ between 0 and π/3, the required area is
0 . .
π/3 π/3
area = 4π − 2 2(4 cos2 θ − 1) dθ = 4π − 4 [2(1 + cos 2θ ) − 1] dθ
0 0
2, − 8
3 √ 9
π/3 π 3 8π √
= 4π − 4 {θ + sin 2θ }0 = 4π − 4 + = − 2 3.
3 2 3
−
2
EXERCISES 9.4
In Exercises 1–10 find the area of the region enclosed by the curve. 5. r 2 = − cos θ 6. r = 2 − 2 cos θ
1. r = 3 sin θ 2. r = −6 cos θ 7. r = 4 − 4 cos θ 8. r = 4 − 2 cos θ

3. r = 2 sin 2θ 2
4. r = 2 sin 2θ 9. r = sin 3θ 10. r = 2(cos θ + sin θ)
In Exercises 11–21 find the area of the indicated region. 18. Inside r = 4 + 3 sin θ but outside r = 2
19. Inside r = |1 − 4 cos θ|
11. Outside r = 3 but inside r = 6 sin θ
20. Inside the bifolium r = sin θ cos2 θ
12. Inside both r = 1 and r = 1 − sin θ 21. Bounded by θ = π and r = θ , 0 ≤ θ ≤ π
13. Inside r = 2 sin 2θ but outside r = 1 ∗∗ 22. (a) Show that in polar coordinates the strophoid
14. Inside both r = 2 + 2 cos θ and r = 2 − 2 cos θ a−x

y2 = x2 ,
a+x
15. Inside both r = sin θ and r = cos θ
where a > 0 is a constant, takes the form r =
16. Inside both r = cos θ and r = 1 − cos θ
a cos 2θ sec θ .
17. Inside r = 1 − 4 cos θ (b) Draw or plot the curve and find the area inside its loop.
9.5 Definitions of Conic Sections

In Section 1.4 we used conic sections to illustrate the algebraic-geometric interplay of plane
analytic geometry. They can be visualized as curves of intersection of a plane with a pair of
right circular cones (Figure 9.38). Certainly, this suggests why the conic sections are so named,
FIGURE 9.38
(a) Circle: Plane perpendicular to (b) Ellipse: Plane cuts completely across (c) Parabola: Plane cuts only one
axis of cones one cone but not perpendicular to axis cone but not completely across
(d) Hyperbola: Plane cuts (e) Pair of straight lines: Plane passes (f) One straight line: Plane passes
both cones through vertex and cuts through both cones through vertex and touches both cones
but because of the three-dimensional nature of the cone, an analysis of conic sections from this
point of view is not yet possible. In this section we use plane analytic geometry to develop
definitions for parabolas, ellipses, and hyperbolas.
The Parabola
DEFINITION 9.2
A parabola is the curve traced out by a point that moves in a plane so that its distances
from a fixed point called the focus and a fixed line called the directrix are always the
same.
Suppose the focus of a parabola is the point (p, q) and the directrix is a line y = r parallel
to the x -axis as in Figure 9.39a. If P (x, y) is any point on the parabola, then the fact that its
distance from (p, q) must be equal to its distance from y = r is expressed as
+
(x − p)2 + (y − q)2 = |y − r|. (9.17)
With the absolute values, this equation includes the case of a directrix above the focus, as in
Figure 9.39b. If we square both sides of the equation and rearrange terms, we obtain
(x − p)2 = (y − r)2 − (y − q)2 = 2y(q − r) + r 2 − q 2 .
We can solve this equation for y in terms of x ; the result is
1
y = [(x − p)2 + (q 2 − r 2 )]. (9.18)
2(q − r)
We could rewrite this equation in our accustomed form y = ax 2 + bx + c for a parabola, but
the present form is more informative. First, the line x = p through the focus and perpendicular
to the directrix is the line of symmetry for the parabola. Second, the parabola opens upward if
q > r , in which case the focus is above the directrix, and opens downward if q < r . Finally,
the vertex of the parabola is found by setting x = p , in which case y = (q + r)/2, halfway
between the focus and directrix.
FIGURE 9.39a Parabola with direc- FIGURE 9.39b Parabola

trix below focus with directrix above focus
y y y=r Directrix
Focus ( p, q) Vertex
P (x, y)
Focus ( p, q)
P (x, y)

Vertex
with vertical directrix
x
x = r; y
y=r Directrix x
Directrix
P (x, y)
Focus
( p, q) A similar analysis shows that when the directrix is parallel to the y -axis (Figure 9.40), the
Vertex equation of the parabola is of the form
1
x = [(y − q)2 + (p 2 − r 2 )]. (9.19)
x 2(p − r)
Given the focus and directrix (parallel to a coordinate axis) of a parabola, we can easily
find its equation: use formulas 9.18 or 9.19, or follow the algebraic steps leading from 9.17
to 9.18. Conversely, given the equation of a parabola in the form y = ax 2 + bx + c or
x = ay 2 + by + c , we can identify its focus and directrix (see Exercises 52 and 53).
EXAMPLE 9.20
Find the equation of the parabola that has focus (2, 4) and directrix x = 6.
with vertical directrix SOLUTION If (x, y) is any point on the parabola (Figure 9.41), the fact that its distance from
(2, 4) is equal to its distance from x = 6 is expressed as
y x=6
+
(x − 2)2 + (y − 4)2 = 6 − x.
(2, 4) If we square both sides and simplify, the result is x = (16 + 8y − y 2 )/8.
(x, y) The Ellipse

2 6 x
DEFINITION 9.3
An ellipse is the curve traced out by a point that moves in a plane so that the sum of its
distances from two fixed points called foci remains constant.
The equation of an ellipse is simplest when the foci lie on either the x - or y -axis and are
FIGURE 9.42 Ellipse in
terms of two foci
equidistant from the origin. Suppose the foci are (c, 0) and (−c, 0) (Figure 9.42), and the sum
of the distances from these foci to a point on the ellipse is 2a , where a > c ≥ 0.
y
If P (x, y) is any point on the ellipse, Definition 9.3 implies that
P (x, y) + +
(x + c)2 + y 2 + (x − c)2 + y 2 = 2a. (9.20)
−c c x If we transpose the second term to the right-hand side and square both sides, we obtain
+
(x + c)2 + y 2 = 4a 2 − 4a (x − c)2 + y 2 + (x − c)2 + y 2 ,
+
a 2 − cx = a (x − c)2 + y 2 .
Squaring once again leads to
a 4 − 2a 2 cx + c2 x 2 = a 2 (x 2 − 2cx + c2 + y 2 ) or
x 2 (a 2 − c2 ) + a 2 y 2 = a 4 − a 2 c2 .
Division by a 2 (a 2 − c2 ) gives
x2 y2
+ = 1. (9.21)
a2 a 2 − c2
It is customary to denote y -intercepts of an ellipse by ±b ( b > 0) (Figure 9.43a), in which
case b2 = a 2 − c2 , and the equation of the ellipse becomes
x2 y2
+ =1 or b2 x 2 + a 2 y 2 = a 2 b2 . (9.22)
a2 b2
FIGURE 9.43a Ellipse FIGURE 9.43b Ellipse

with horizontal major axis and with vertical major axis and hori-
vertical minor axis zontal minor axis
y y
Minor axis b b
a c
b
−a −c c a x
−a a x
Major axis
−b
−c
−b
The line segment across the ellipse and through the foci is called the major axis of the
ellipse; it has length 2a (see Figure 9.43a). The midpoint of the major axis is called the centre
of the ellipse. The line segment across the ellipse, through its centre, and perpendicular to the
major axis is called the minor axis; it has length 2b . Note that the line segment joining either
end of the minor axis to a focus (Figure 9.43a) has length a , and the triangle formed specifies
the relationship among a , b , and c , namely a 2 = b2 + c2 .
A similar analysis shows that when the foci of the ellipse are on the y -axis, equidistant
from the origin, the equation of the ellipse is again in form 9.22. In this case, 2b is the length
of the major axis, 2a is the length of the minor axis, and b2 = a 2 + c2 (Figure 9.43b).
What we should remember is that an equation of form 9.22 always specifies an ellipse. Foci
are on the longer axis and can be located using c2 = |a 2 − b2 | . The length of the major axis
represents the sum of the distances from any point on the ellipse to the foci.
EXAMPLE 9.21
Draw the ellipse 16x 2 + 9y 2 = 144, indicating its foci.
FIGURE 9.44 Ellipse 16x 2 + 9y 2 = 144 showing foci
y
4
−3 3 x
− 7
−4
SOLUTION If we write the ellipse in the form x 2 /9 + y 2 /16 = 1, its x - and y -intercepts
are ±3 and ±4. A sketch of the ellipse
√ is therefore√
as shown in Figure 9.44. The foci must lie
on the y -axis at distances ±c = ± 42 − 32 = ± 7 from the origin.
If a = b in equation 9.22, then x 2 + y 2 = a 2 , and this is the equation for a circle with radius
a and centre at the origin. But if a = b , the distance from the origin to each focus of the ellipse
must be c = 0. In other words, a circle may be regarded as a degenerate ellipse whose foci are
at one and the same point. It is also true that when c is very small compared to half the length of
FIGURE 9.45a Ellipse the major axis (Figure 9.45a), the ellipse is shaped very much like a circle. On the other hand,
when c is much less than a when these lengths are almost equal (Figure 9.45b), the ellipse is long and narrow.
y When the centre of an ellipse is at point (h, k) and the foci lie on either the line x = h or
b y = k (Figures 9.46), the equation for the ellipse is somewhat more complex than 9.22. If 2a
and 2b are again the lengths of the axes of the ellipse, a calculation similar to that leading from
9.20 to 9.22 gives (see Exercise 55)
a x (x − h)2 (y − k)2
−c c + = 1. (9.23)
a2 b2
Alternatively, the curves in Figures 9.46 are those in Figures 9.42 and 9.43b translated h
units in the x -direction and k units in the y -direction. According to Section 1.5, equations for
FIGURE 9.45b Ellipse
the translated ellipses can be obtained by replacing x and y in x 2 /a 2 + y 2 /b2 = 1 by x − h
when c is approximately equal to
a and y − k , respectively.
y
b
Ellipse with centre at (h, k)
y y
−c c a x x=h
(h − c, k) (h + c, k) (h, k + b)
Focus
(h, k) (h, k + c)
y=k
Focus Focus (h + a, k)
(h, k)
(h − a, k)
x Focus x
(h, k − b) (h, k − c)
EXAMPLE 9.22
Draw the ellipse 16x 2 + 25y 2 − 160x + 50y = 1175.
SOLUTION If we complete the squares on the x - and y -terms, we obtain
16(x − 5)2 + 25(y + 1)2 = 1600 or
(x − 5)2 (y + 1)2
= 1.+
100 64
The centre of the ellipse is (5, −1) , and lengths of its major and minor axes are 20 and 16,
respectively (Figure 9.47).
FIGURE 9.47 Ellipse 16x 2 + 25y 2 − 160x + 50y = 1175 drawn by completing the squares
y
(5, 7)
(−5, −1) x
(5, −1) (15, −1)
(5, −9)
The Hyperbola
DEFINITION 9.4
A hyperbola is the path traced out by a point that moves in a plane so that the difference
between its distances from two fixed points called foci remains constant.
Like the ellipse, the simplest hyperbolas have foci on either the x - or y -axis, equidistant
from the origin. Suppose the foci are (±c, 0) (Figure 9.48) and the difference in the distances
from P (x, y) to these foci is 2a . Then Definition 9.4 implies that
+ +
| (x + c)2 + y 2 − (x − c)2 + y 2 | = 2a. (9.24)
This equation can be simplified by a calculation similar to that leading to 9.22; the result is
x2 y2
− = 1, (9.25)
a2 b2
where b2 = c2 − a 2 . The hyperbola has x -intercepts equal to ±a .
FIGURE 9.48 Hyperbola FIGURE 9.49 Hyperbola

with foci on x -axis with foci on y -axis
y y
c
x 2 y2 y2 x 2
− =1 P (x, y) − =1
a 2 b2 b2 a 2
b
P (x, y)
x
−c −a a c x −b
−c
When the foci are on the y -axis (Figure 9.49), the equation of the hyperbola becomes
y2 x2
− = 1, (9.26)
b2 a2
where 2b is the constant difference in the distances from a point (x, y) to the foci, and a 2 =
c2 − b2 . This hyperbola intersects the y -axis at ±b .
That part of the line segment joining the foci of a hyperbola that is between the two branches
of the curve is called the transverse axis of the hyperbola; it has length 2a in Figure 9.50a and
2b in Figure 9.50b. The midpoint of the transverse axis is called the centre of the hyperbola.
The line segment perpendicular to the transverse axis, through its centre, and of length 2b in
Figure 9.50a and 2a in Figure 9.50b is called the conjugate axis. Asymptotes of both hyperbolas
are the lines y = ±bx/a .
What we should remember is that an equation of form 9.25 or 9.26 specifies a hyperbola.
The foci lie on the extension of the transverse axis and can be located using c2 = a 2 + b2 .
The length of the transverse axis represents the difference of the distances from any point on
the hyperbola to the foci.

Transverse and conjugate axes of hyperbolas
y y
b b c
y=− x y= x
a a b
Transverse
axis
b
−c −a a y2 x 2 x
a c x Conjugate − =1
Transverse axis −b b2 a 2
axis −c
Conjugate b
x 2 y2 y=− x
axis − =1 b a
a2 b2 y= x
a
EXAMPLE 9.23
FIGURE 9.51 Hyperbola

16x 2 − 9y 2 = 144 showing foci Draw the hyperbola 16x 2 − 9y 2 = 144, indicating its foci.
y
4 SOLUTION If we express the hyperbola in the form x 2 /9 − y 2 /16 = 1, its x -intercepts

are ±3. With asymptotes
√ y = ±4x/3, we obtain Figure 9.51. The foci lie on the x -axis at
distances ±c = ± 42 + 32 = ±5 from the origin.
−5 −3 3 5 x
When the centre of a hyperbola is at point (h, k) and its foci are on the lines x = h or y = k ,
equations 9.25 and 9.26 are modified in exactly the same way as equation 9.22 was modified
for an ellipse. We replace each x by x − h and each y by y − k (see also Exercise 56).
Consequently, equations for the hyperbolas in Figures 9.52 are
(x − h)2 (y − k)2 (y − k)2 (x − h)2

− =1 and − = 1. (9.27)
a2 b2 b2 a2

Hyperbolas with centres at (h, k)
b b
y y=k+ (x − h) y y=k+ (x − h)
a b a
b y=k− (x − h)
y=k− (x − h) a
a
(h − a, k)
(h + a, k) (h, k + c)
y=k (h, k + b)
(h − c, k) (h, k) (h + c, k) (h, k)
(h, k − b)
x
x
(h, k − c)
x=h
EXAMPLE 9.24
Draw the hyperbola x 2 − y 2 + 4x + 10y = 5.
SOLUTION If we complete squares on x - and y -terms, we find
FIGURE 9.53 Hyperbola
x 2 − y 2 + 4x + 10y = 5 drawn by (x + 2)2 − (y − 5)2 = −16 or
completing the squares
(y − 5)2 (x + 2)2
y − = 1.
16 16
The centre of the hyperbola is (−2, 5) , and the length of its transverse axis (along x = −2) is
(−2, 9) 8. If we solve the equation for y , we obtain
(−2, 5) +
y = 5± (x + 2)2 + 16;
(−2, 1)
−2 x asymptotes of the hyperbola are then y = 5 ± (x +√ 2) . The hyperbola can
√ now be sketched as
in Figure 9.53. Its foci are at the points (−2, 5 + 4 2) and (−2, 5 − 4 2) .
y = 5 − (x + 2)
y = 5 + (x + 2)
If P is a point on a conic section, the focal radii at P are the lines joining P to the foci
(Figure 9.54). As a result, a parabola has one focal radius at each point, and an ellipse and
hyperbola each have two. One of the properties of conics that makes them so useful is the fact
that the normal line to the conic at any point bisects the angle between the focal radii. For the
parabola, the normal bisects the angle between the focal radius and the line through P parallel
to the axis of symmetry of the parabola. We will verify these facts in Exercises 59 and 60. To
obtain one physical significance of these results, suppose each conic in Figure 9.54 is rotated
about the x -axis to form a surface of revolution, which we regard as a mirror. It is a law of
optics that when a ray of light strikes a reflecting surface, the angle between incident light and
the normal to the surface is always equal to the angle between reflected light and the normal.
Consequently, if a beam of light travels in the negative x -direction and strikes the parabolic
mirror in Figure 9.54a, all light is reflected toward the focus. Conversely, if F is a source of
light, all light striking the mirror is reflected parallel to the x -axis. If either focus of the ellipse
in Figure 9.54b is a light source, all light striking the elliptic mirror is reflected toward the other
focus. Similarly, if either focus of the hyperbola in Figure 9.54c is a source, all light striking
the mirror is reflected in a direction that would make it seem to originate at the other focus.
Conversely, if light that is directed at one focus first strikes the mirror, it is reflected toward the
other focus. This is precisely why we have parabolic reflectors in automobile headlights and
searchlights, parabolic and hyperbolic reflectors in telescopes, and elliptic ceilings in whispering
rooms.
Thus far, we have defined parabolas, ellipses, and hyperbolas in terms of distances; for the
parabola we use a focus and a directrix, and for the ellipse and hyperbola two foci. In Section
9.6 we show that ellipses and hyperbolas can also be defined in terms of a focus and directrix.
FIGURE 9.54a Parabola FIGURE 9.54b Ellipse as FIGURE 9.54c Hyper-

as a reflector a reflector bola as a reflector
y y y
P
P
P
F x F2 F1 x F2 F1 x
EXERCISES 9.5
In Exercises 1–14 identify the equation as representing a straight line, ∗ 40. Find the height of the parabolic arch in the figure below.
a circle, a parabola, an ellipse, a hyperbola, or none of these.
1. 2x + 3y = y 2 2. x 2 + y 2 − 3x + 2y = 25
3. 2x − y = 3 4. x 2 + y 3 = 3x + 2
5. 5x 2 = 11 − 2y 2 6. 2x 2 − 3y 2 + 5 = 0 4
7. y 2 − x + 3y = 14 − x 2 8. x 2 + 2x = 3y + 4 3
9. y 2 + x 2 − 2x + 6y + 15 = 0
5
10. x 2 + 2y 2 + 24 = 0 ∗ 41. Explain how an ellipse can be drawn with a piece of string, two
tacks, and a pencil.
11. 5 + y 2 = 3x 2 12. x 2 + 2y 2 = 24
13. y 3 = 3x + 4 14. 3 − x = 4y ∗ 42. Find the equation of the ellipse traced out by a point that moves
so that the sum of its distances from (±4, 0) is always equal to 10
(a) by using equation 9.22 with suitable values for a and b and (b) by
In Exercises 15–36 draw the curve. Identify foci for each ellipse and establishing and simplifying an equation similar to 9.20.
hyperbola.
∗ 43. Find the equation of the hyperbola traced out by a point that moves
x2 y2
15. y = 2x 2 − 1 16. + =1 so that the difference between its distances from (0, ±3) is always
25 36 equal to 1 (a) by using equation 9.26 with suitable values for a and b
y2 and (b) by establishing and simplifying an equation similar to 9.24.
17. x 2 − =1 18. 3x = 4y 2 − 1
16
∗ 44. Show that the equation of every straight line, every circle, and every
2
y x2 2 2 conic section discussed in this section can be obtained by appropriate
19. − =1 20. 7x + 3y = 16
4 25 choices of constants A , C , D , E , and F in the equation
21. x + y 2 = 1 22. 2y 2 + x = 3y + 5
Ax 2 + Cy 2 + Dx + Ey + F = 0.
23. 9x 2 + 289y 2 = 2601 24. y 2 = 10(2 − x 2 )
25. 3x 2 − 4y 2 = 25 26. y 2 − x 2 = 5
∗ 45. Show that the equation of the tangent line to the ellipse b2 x 2 +
27. y = −x + 6x − 9 2 2
28. 2x − 3y = 5 2 a 2 y 2 = a 2 b2 at a point (x0 , y0 ) is b2 xx0 + a 2 yy0 = a 2 b2 .
2 2
29. 3x + 6y = 21 30. x 2 + 16y 2 = 2 ∗ 46. Show that the equation of the tangent line to the hyperbola b2 x 2 −
31. y 2 − 3x 2 = 1 32. x = −(4 + y)2 a 2 y 2 = a 2 b2 at a point (x0 , y0 ) is b2 xx0 − a 2 yy0 = a 2 b2 .
33. x 2 + 2x + 4y 2 − 16y + 13 = 0 ∗ 47. Find the point P on that part of the ellipse 2x 2 + 3y 2 = 14 in the
34. x 2 − 6x − 4y 2 − 24y = 11 first quadrant where the tangent line at P is perpendicular to the line
joining P and (2, 5) .
35. 9x 2 + y 2 − 18x − 6y = 0
∗ 48. Find the area inside the ellipse b2 x 2 + a 2 y 2 = a 2 b2 .
36. 9x 2 − 16y 2 − 18x − 64y = 91
∗ 49. Among all rectangles that can be inscribed inside the ellipse b2 x 2 +
37. Find the equation of a hyperbola that passes through the point a 2 y 2 = a 2 b2 and have sides parallel to the axes, find the one with
(1, 2) and has asymptotes y = ±4x . largest possible area.
38. Find the equation of an ellipse through the points (−2, 4) and
(3, 1) . ∗ 50. A prolate spheroid is the solid of revolution obtained by rotating
an ellipse about its major axis. An oblate spheroid is obtained by
∗ 39. Find the width of the elliptic arch in the figure below. rotating the ellipse about its minor axis. Find volumes for the prolate
and oblate spheroids generated by the ellipse b2 x 2 + a 2 y 2 = a 2 b2 if
1 3 a > b.
2
5 2 ∗ 51. A sharp noise originating at one focus F1 of an ellipse is reflected
by the ellipse toward the other focus F2 . Explain why all reflected
noise arrives at F2 at exactly the same time.
∗ 52. Use equation 9.18 to show that when a parabola is written in the parallel to the x -axis. Hint: Draw the tangent line P R at P and show
form y = ax 2 + bx + c , the following formulas identify its focus that (P F ( = (RF ( .
(p, q) and directrix y = r :
∗∗ 58. When a beam of light travelling in the negative x -direction (figure
b 1 1 below) strikes a parabolic mirror with cross section represented by
p=− , q= (1 + 4ac − b2 ), r= (−1 + 4ac − b2 ). x = ay 2 + c , all light rays are reflected to the focus F of the mirror.
2a 4a 4a
Show that all photons that pass simultaneously through x = d arrive
∗ 53. What are formulas for the focus (p, q) and directrix x = r for a at F at the same time.
parabola of the type x = ay 2 + by + c ?
∗ 54. Use the formulas in Exercises 52 and 53 to identify the focus and y
x=d
directrix for any parabolas in Exercises 15–36.
∗ 55. Show that when the centre of an ellipse is at point (h, k) and its
foci are on the line x = h or y = k , Definition 9.3 leads to equation
9.23.
∗ 56. Show that when the centre of a hyperbola is at point (h, k) and its x
F
foci are on the line x = h or y = k , Definition 9.4 leads to equation
9.27.
∗ 57. Prove that the normal line to the parabola x = ay 2 in the figure
below bisects the angle between the focal radius F P and the line P Q
∗∗ 59. Prove that the normal line to an ellipse or hyperbola bisects the
y angle between the focal radii.
P Q ∗∗ 60. Prove that the normal line to the parabola y = ax 2 + bx + c

at any point P bisects the angle between the focal radius and the line
through P parallel to the y -axis.
R F x
∗∗ 61. A line segment through the focus of a parabola with ends on the
parabola is called a focal chord . It is an established result that tangents
x = ay 2 to a parabola at the ends of a focal chord are perpendicular to each other
and intersect on the directrix. Prove this for the parabola y = ax 2 .
9.6 Conic Sections in Polar Coordinates

In Section 9.5 we defined parabolas using a focus and directrix and ellipses and hyperbolas using
two foci. In this section we show that all three conics can be defined using a focus and directrix
and that, in polar coordinates, one equation represents all three conics.
Let F be a fixed point (the focus), and l be a fixed line (the directrix) that does not pass
through F , as in Figure 9.55a. We propose to find the equation of the curve traced out by a
point that moves so that its undirected distances from F and l always remain in a constant ratio
* called the eccentricity. To do this we set up polar coordinates with F as pole and polar axis
directed away from l and perpendicular to l as in Figure 9.55b.

Development of conic sections from focus and directrix
l l P
r
F
Focus F Polar axis
d Pole
d
Directrix
If (r, θ ) are polar coordinates for any point P on the required curve, the fact that the ratio
of the distances from F and l to P is equal to * is expressed as
r
= *. (9.28)
d + r cos θ
When this equation is solved for r , we have
*d
r = , (9.29)
1 − * cos θ
and this is the polar equation of the curve traced out by the point. Certainly, this curve should be
a parabola when * = 1. We now verify this, and show that the curve is an ellipse when * < 1
and a hyperbola when * > 1. To do this we transform the equation into the usual Cartesian
coordinates x = r cos θ and y = r sin θ . From equations 9.5 and 9.6, we obtain
+ *d
x2 + y2 = *x ,
1− +
x2 + y2

+
x 2 + y 2 = *(d + x), (9.30a)
or when squared,
x 2 + y 2 = * 2 (d + x)2 . (9.30b)
When * = 1, the x 2 -terms in 9.30b cancel, and the equation reduces to that for a parabola:
1
x = (y 2 − d 2 ), (9.31)
2d
as in Figure 9.56a. When * %= 1, we write
x 2 + y 2 = * 2 (d 2 + 2dx + x 2 ) or
(1 − * 2 )x 2 − 2d* 2 x + y 2 = * 2 d 2 .
When * < 1, we divide by 1 − * 2 :
2d* 2 y2 *2d 2
x2 − 2
x+ = ,
1−* 1 − *2 1 − *2
and complete the square on the x -terms:
# $2 # $2
d* 2 y2 *2d 2 d 2*4 *d
x− + = + = . (9.32)
1 − *2 1 − *2 1 − *2 (1 − * 2 )2 1 − *2
Comparing equation 9.32 with 9.23, we conclude that 9.32 is the equation for an ellipse with
centre at position x = d* 2 /(1 − * 2 ) on the x -axis. Since one focus is at the origin, it follows
that the other must also be on the x -axis at x = 2d* 2 /(1 − * 2 ) as in Figure 9.56b.
FIGURE 9.56a Eccentric- FIGURE 9.56b Eccentric- FIGURE 9.56c Eccentric-

ity * = 1 leads to parabola ity * < 1 leads to ellipse ity * > 1 leads to hyperbola
y y y
l l Other
focus Other l
focus
Centre
F x F x Centre
F x
A similar calculation shows that when * > 1, points on the right half of the hyperbola
# $2 # $2
d* 2 y2 *d
x+ 2 − 2 = (9.33)
* −1 * −1 *2 − 1
are obtained as in Figure 9.56c. Equation 9.30a is not satisfied by points with x -coordinates
less than −d , and therefore points on the left half of the hyperbola, shown dotted, do not satisfy
9.29.
We have shown that equation 9.29 defines an ellipse when 0 < * < 1, a parabola when
* = 1, and a hyperbola when * > 1, and this provides a unifying approach to conic sections.
All three conics can be studied using a focus and a directrix.
It is clear that equation 9.29 can yield only a parabola that opens to the right and has its
focus on the x -axis, and an ellipse and hyperbola with foci on the x -axis, one at the origin. To
obtain parabolas that open to the left, or up, or down, and ellipses and hyperbolas with foci on
the y -axis, we must change the position of the directrix. The conic sections in Figures 9.57
have directrix to the right of the focus. They have equations
*d
r = . (9.34a)
1 + * cos θ
The conic sections in Figures 9.58 and 9.59 have equations of the form, respectively,
*d *d
r = and r = . (9.34b)
1 − * sin θ 1 + * sin θ

Conic sections in the form of r = *d/(1 + * cos θ)
y y y
l l l
Polar Polar Polar

axis axis axis
F x F x F x

Conic sections in the form of r = *d/(1 − * sin θ)
y y y
Polar Polar Polar

axis axis axis
F F F
x x x
l l l

Conic sections in the form of r = *d/(1 + * sin θ)
y y y
l Polar l l
axis
F x F x F x
Polar Polar
axis axis
In Figures 9.56–9.59, one focus of the conic is chosen as the pole. In other words, the
simplicity of equations 9.29 and 9.34 to describe conic sections is a direct consequence of the
fact that the pole is at a focus, and the directrix is either parallel or perpendicular to the polar
axis.
EXAMPLE 9.25
Draw the curve r = 15/(3 + 2 cos θ ) .
SOLUTION If we write the equation in the form 9.34a,
5
FIGURE 9.60 Ellipse r =
r = !2" ,
1+ 3
cos θ
15/(3 + 2 cos θ)
y the eccentricity * = 2/3 indicates that the curve is an ellipse. Both foci lie on the x -axis,
and one is at the origin. The ends of the major axis occur when θ = 0 and θ = π , and for
3 5 these values r = 3 and r = 15 (Figure 9.60). It now follows that the centre of the ellipse
is at x = −6, and its other focus is at x = −12. If b > 0 denotes half the length of the
minor axis (it is also the maximum y -value on the ellipse,
√occurring when x = −6), then
−15 −12 −6 3 x b2 = a 2 − c2 = 92 − 62 = 45. Consequently, b = 3 5, and the ellipse is as shown in
−3 5 Figure 9.60. This information now permits us to write the equation of the ellipse in Cartesian
coordinates:
(x + 6)2 y2
+ = 1.
81 45
+
An+ alternative approach in Example 9.25 would be to substitute r = x 2 + y 2 and cos θ =
x/ x 2 + y 2 into the polar equation for the conic, simplify it to (x + 6)2 /81 + y 2 /45 = 1, and
then draw the ellipse from this equation. We illustrate this method in the following example.
EXAMPLE 9.26
Draw the curve r = 2/(3 − 4 sin θ ) .

+ +
SOLUTION If we set r = x 2 + y 2 and sin θ = y/ x 2 + y 2 , then
+
+ 2 2 x2 + y2
x2 + y2 = = + .
4y 3 x 2 + y 2 − 4y
3− +
x2 + y2
+
Division by x 2 + y 2 leads to
+
3 x 2 + y 2 − 4y = 2 or
+
3 x 2 + y 2 = 4y + 2 .
If we now square both sides, we obtain
9(x 2 + y 2 ) = 4 + 16y + 16y 2 or
9x 2 − 7y 2 − 16y = 4.
If we complete the square on the y -terms, we obtain
# $2
2 8 64 36
9x − 7 y + = 4− =− .
7 7 7
Division by −36/7 yields the equation
# $2
8
y+
7 x2
− = 1.
FIGURE 9.61 Hyperbola 36 4
r = 2/(3 − 4 sin θ) 49 7
8 3x
y y=− + This equation describes a hyperbola with centre (0, −8/7) and y -intercepts equal to −8/7 ±
7 7
2 6/7 = −2/7, −2. Its asymptotes are
−
7
%
x 8 6 7x 2 8 3x
8 y =− ± =− ±√ .
− 7 7 4 7 7
7
8 3x
y=− −
−2 7 7 The hyperbola is
+shown in Figure 9.61, but only the top half is described by r = 2/(3 − 4 sin θ ) .
The equation 3 x 2 + y 2 = 4y + 2 does not permit y ≤ −2.
EXAMPLE 9.27
Find a polar representation for the ellipse
(x − 1)2 y2
+ = 1.
4 9
FIGURE 9.62 Ellipse SOLUTION To find a polar representation for the ellipse (Figure 9.62), we could use the usual
(x − 1)2 /4 + y 2 /9 = 1 polar coordinates defined by x = r cos θ and y = r sin θ , but the resulting equation would not
y be simple. Try it. We know that a simple polar representation must result if the pole is chosen
at a focus of the ellipse and polar axis either parallel or √
perpendicular√
to the directrix. For this
3 Polar ellipse, foci are on the line x = 1 at distances ±c = ± 9 − 4 = ± 5 from the x -axis. Let
√
(1, 5 ) axis us choose the pole at position (1, 5) and the polar axis parallel to the x -axis, and therefore
parallel to the directrix. According
√ to equation 9.9, polar and Cartesian coordinates are related
−1 1 3 x by x = 1 + r cos θ and y = 5 + r sin θ . If we substitute these into the equation for the
ellipse, we obtain
√
−3 r 2 cos2 θ ( 5 + r sin θ )2
+ =1
4 9
or √
9r 2 cos2 θ + 4(5 + 2 5r sin θ + r 2 sin2 θ ) = 36.
This equation can be expressed as a quadratic equation in r :
√
r 2 (9 cos2 θ + 4 sin2 θ ) + r(8 5 sin θ ) − 16 = 0.
Solutions for r are

√ +
−8 5 sin θ ± 320 sin2 θ + 64(9 cos2 θ + 4 sin2 θ )
r =
2(9 cos2 θ + 4 sin2 θ )
√ +
−8 5 sin θ ± 576(cos2 θ + sin2 θ )
=
2(9 cos2 θ + 4 sin2 θ )
√
4(±3 − 5 sin θ )
= .
9 cos2 θ + 4 sin2 θ
Since r must be nonnegative, we must choose +3, and not −3, and therefore
√ √
4 (3 − 5 sin θ ) 4 (3 − 5 sin θ )
r = 2 2
=
9(1 − sin θ ) + 4 sin θ 9 − 5 sin2 θ
√
4 (3 − 5 sin θ ) 4
= √ √ = √ .
(3 − 5 sin θ )(3 + 5 sin θ ) 3+ 5 sin θ
EXERCISES 9.6
4 4
In Exercises 1–10 draw the conic section. 5. r = 6. r =
3 − 4 sin θ 4 − 3 sin θ
3 16 1 1
1. r = 2. r = 7. r = 8. r =
1 + cos θ 3 + 5 cos θ 2 − 2 cos θ 2 + sin θ
4 16 sec θ 4 csc θ
3. r = 4. r = 9. r = 10. r =
3 − 3 sin θ 5 + 3 cos θ 3 + 6 sec θ 7 csc θ − 2
In Exercises 11–16 find the Cartesian equation for the curve. ∗ 21. (a) Show that the eccentricity * of an ellipse or hyperbola is
always equal to the distance from its centre to either focus,
3 1 divided by half the length of the major or transverse axis.
11. r = 12. r =
1 − sin θ 3 + cos θ (b) Discuss the eccentricities of the ellipses in Figures 9.45a
1 2 and 9.45b.
13. r = 14. r =
1 + 2 cos θ 1 − 3 cos θ ∗ 22. A circle has been described as a degenerate ellipse in the sense that
4 4 its foci coincide. What happens to the eccentricity of an ellipse if the
15. r = 16. r = distance between its foci approaches zero? Hint: See Exercise 21.
6 − 3 sin θ 5 + 5 cos θ
∗∗ 23. Paths of celestial objects can be described by equations of the form
r = a/(1 + * cos θ) , where a > 0 is a constant. To find the time to
In Exercises 17–20 find a polar equation in one of the four forms of travel from one point on the path to another, it is necessary to evaluate
9.29 or 9.34 for the conic. the integral .
1
∗ 17. x 2 − 16y 2 = 16 ∗ 18. 4x 2 + 9y 2 = 36 I = dθ.
(1 + * cos θ)2
(y + 1)2 Do so in the cases that (a) * = 0, (b) 0 < * < 1, (c) * = 1, and (d)
∗ 19. (x − 1)2 + =1 ∗ 20. x 2 − 9(y − 2)2 = 9
4 * > 1.
Website
Information on Translation and Rotation of Axes has been placed on the Text Enrichment
Site: www.pearsoned.ca/text/trim.
SUMMARY
In Chapter 1 we defined curves explicitly and implicitly. In this chapter we added a third
description — the parametric definition
x = x(t), y = y(t), α ≤ t ≤ β.
When such a curve defines a function, the derivative of the function can be calculated using the
parametric rule:
dy
dy
= dt ;
dx dx
dt
second- and higher-order derivatives can be calculated using the chain rule. The length of such
a curve is defined by the definite integral
-# $2 # $2
. β
dx dy
+ dt.
α dt dt
Polar coordinates provide an alternative way to identify the positions of points in a plane.
They use distance from a point, called the pole, and rotation of a half line, called the polar
axis. Many curves that have complex equations in Cartesian coordinates can be represented very
simply using polar coordinates. Particularly simple are multileaved roses, cardioids, lemniscates,
and some circles. Polar coordinates also provide a unified approach to parabolas, ellipses, and
hyperbolas. Each curve can be described as the path traced out by a point that moves so that
the ratio * of its distances from a fixed point and a fixed line remain constant. The curve is
an ellipse, a parabola, or a hyperbola depending on whether 0 < * < 1, * = 1, or * > 1,
respectively. Ellipses and hyperbolas can also be defined as curves traced out by a point that
moves so that the sum and difference of its distances from two fixed points remain constant.
When a curve r = f (θ ) in polar coordinates encloses a region R , the definite integral
. β
1
[f (θ )]2 dθ
α 2
with appropriate choices of α and β can be used to find the area of R .
Parametric equations for a polar curve r = f (θ ) , α ≤ θ ≤ β are
x(θ ) = f (θ ) cos θ, y(θ ) = f (θ ) sin θ, α ≤ θ ≤ β.
Its slope is given by

dy f , (θ ) sin θ + f (θ ) cos θ
= , ,
dx f (θ ) cos θ − f (θ ) sin θ
and its length can be calculated with the definite integral
- # $2
. β
dr
r2 + dθ .
α dθ
KEY TERMS
Parametric equations Parametrically defined function
Parametric rule Polar coordinates
Pole Polar axis
Slopes of curves in polar coordinates Lengths of curves in polar coordinates
Areas in polar coordinates Parabola
Focus Directrix
Ellipse Major axis
Centre of an ellipse Minor axis
Hyperbola Transverse axis
Conjugate axis Eccentricity
REVIEW
EXERCISES
In Exercises 1–30 draw, and then plot, the curve. 15. r 2 = 4 cos2 θ 16. r = 3 cos 2θ
+
1. r = cos θ 2. r = − sin θ 17. x + y = x 2 + y 2 18. x 2 + y 2 = x + y
3 3 19. r = 4 cos 3θ 20. r = sin2 θ − cos2 θ

3. r = 4. r =
1 + 2 sin θ 2 + sin θ 21. x 2 + y 2 − 3x + 2y = 1 22. x 2 − 3y 2 + 4 = 0
1 3 23. 2x 2 + 3y 2 − 6y = 0 24. y 2 + x + 2y = 3
5. r = 6. r =
1 + sin θ 2 − 2 cos θ 25. r(2 cos θ − sin θ) = 3 26. r = cos (θ/2)
2
7. r + 1 = 2 sin θ 8. r = 4 cos θ 27. r = sin θ 2
28. x 2 = y 2 + y
9. x = 2 cos t, y = 3 sin t, 0 ≤ t < 2π 29. x = e−t , y = ln t, t > 0 30. x = sin θ, y = sin 2θ
10. x = 4 + t, y = 5 − 3t 2
In Exercises 31–35 find the area of the region indicated.
11. x = sin2 t, y = cos2 t, 0 ≤ t ≤ 2π
31. Inside r = 2 + 2 cos θ
12. x = 1 + 4 sin t, y = −2 + 4 cos t, 0 ≤ t ≤ π 32. Inside r = 4 but outside r = 4 sin 2θ
33. Common to r = 2 and r 2 = 9 cos 2θ
13. (x 2 + y 2 )3 = x
+ 34. Inside r = sin2 θ
2 2
14. x + y = 2 x 2 + y 2 35. Common to r = 1 + sin θ and r = 2 − 2 sin θ
In Exercises 36–37 assume that y is defined as a function of x , and ∗ 39. x = t 2 , y = t3, 0≤t ≤1
find dy/dx and d 2 y/dx 2 .
∗ 40. x = e cos t ,
t
y = et sin t , 0 ≤ t ≤ π/2
3 3
∗ 36. x = t + 2t, y = 3t − t ∗ 41. Find the equation of the tangent line to the curve r = 2 − 2 sin θ
∗ 37. x = 2 sin u, y = 3 cos u at the point with polar coordinates (1, π/6) .
In Exercises 38–40 find the length of the curve.
∗ 38. r = 2 + 2 cos θ
CHAPTER
10 Infinite Sequences and Series
Application Preview The amount A of a certain chemical in a reactor decreases exponentially in time
A(t) = A0 ekt ,
where A0 is the amount at time t = 0, and k < 0 is a constant. After time T , the amount in
the reactor is A0 ekT , at which time an additional amount A0 is added, resulting in an amount
A0 + A0 ekT = A0 (1 + ekT ) at that time. This amount then decreases exponentially until an
additional amount A0 is added at time 2T , and so on, and so on. Once the amount of chemical
reaches a critical level L in the reactor, the process must be terminated.
THE PROBLEM Find an equation that determines the time at which the process must be
terminated. (For a solution to the problem, see Example 10.38 on page 662.)
Sequences and series play an important role in many areas of applied mathematics. Se-
quences of numbers were first encountered in Section 4.1, although we did not use the term
sequences at the time. Newton’s iterative procedure produces a set of numbers x1 , x2 , x3 , . . .
each of which approximates a root of an equation f (x) = 0. The first number is chosen as
some initial approximation to the solution of the equation, and subsequent numbers defined by
the formula
f (xn )
xn+1 = xn −
f " (xn )
are better and better approximations. This ordered set of numbers is called a sequence. Each
number in the set corresponds to a positive integer, and each is calculated according to a stated
formula.
A series of numbers is the sum of the numbers in a sequence. If the numbers are x1 , x2 ,
x3 , . . . , the corresponding series is denoted symbolically by
∞
!
x n = x1 + x2 + x3 + · · · ,
n=1
where the three dots indicate that the addition is never-ending. It is all very well to write an
expression like this, but it does not have meaning. No matter how fast we add, or how fast a
calculator adds, or even how fast a supercomputer adds, an infinity of numbers can never be
added together in a finite amount of time. We shall give meaning to such expressions, and show
that they are really the only sensible way to define many of the more common transcendental
functions, such as trigonometric, exponential, and hyperbolic.
The first two sections of this chapter and Section 10.8 are devoted to sequences and the
remaining ten to series. This is not to say that series are more important than sequences; they
are not. Discussions on series invariably become discussions on sequences associated with
series. We have found that difficulties with this chapter can usually be traced back to a failure to
distinguish between the two concepts. Special attention to the material in Sections 10.1, 10.2,
and 10.8 will be rewarded; a cursory treatment leads to confusion in other sections.
587
588 Chapter 10 Infinite Sequences and Series
10.1 Infinite Sequences of Numbers

Sequences of numbers are defined as follows.
DEFINITION 10.1
An infinite sequence of numbers is a function f whose domain is the set of positive
integers.
For example, when f (n) = 1/n , the following numbers are associated with the positive
integers:
1 1 1
1, , , , ....
2 3 4
The word infinite simply indicates that an infinity of numbers is defined by the sequence, as
there is an infinity of positive integers, but it indicates nothing about the nature of the numbers.
Often, we write the numbers f (n) in a line separated by commas,
f (1), f (2), . . . , f (n), . . . (10.1a)
and refer to this array as the sequence rather than the rule by which it is formed. Since this
notation is somewhat cumbersome, we adopt a notation similar to that used for the sequence
defined by Newton’s iterative procedure in Section 4.1 We set c1 = f (1) , c2 = f (2) , . . . ,
cn = f (n) , . . . , and write for 10.1a
c1 , c2 , c3 , . . . , c n , . . . . (10.1b)
The first number c1 is called the first term of the sequence, c2 the second term, and for general
n , cn is called the nth term (or general term) of the sequence. For the example above, we have
1 1
c1 = 1, c2 = , , etc.
c3 =
2 3
In some applications, it is more convenient to define a sequence as a function whose domain
is the set of integers larger than or equal to some fixed integer N , and N can be positive, negative,
or zero. Later in this chapter, we find it convenient to initiate the assignment with N = 0. For
now we prefer to use Definition 10.1 where N = 1, in which case we have the natural situation
where the first term of the sequence corresponds to n = 1, the second term to n = 2, and so
on.
EXAMPLE 10.1
The general terms of four sequences are
1 n
(a) (b) (c) (−1)n |n − 3| (d) (−1)n+1 .
2n−1 n+1
Write out the first six terms of each sequence.
SOLUTION The first six terms of these sequences are
1 1 1 1 1 1 2 3 4 5 6
(a) 1, , , , , ; (b) , , , , , ;
2 4 8 16 32 2 3 4 5 6 7
(c) − 2, 1, 0, 1, −2, 3; (d) 1, −1, 1, −1, 1, −1.
The sequences in Example 10.1 are said to be defined explicitly; we have an explicit formula
for the nth term of the sequence in terms of n . This allows easy determination of any term in
the sequence. For instance, to find the one-hundredth term, we simply replace n by 100 and
perform the resulting arithmetic. Contrast this with the sequence in the following example.
EXAMPLE 10.2
The first term of a sequence is c1 = 1 and every other term is to be obtained from the formula
√
cn+1 = 5 + 2 + cn , n ≥ 1.
Calculate c2 , c3 , c4 , and c5 .
SOLUTION To obtain c2 we set n = 1 in the formula:
√ √ √
c1+1 = c2 = 5 + 2 + c1 = 5 + 2+1 = 5+ 3 ≈ 6.732.
To find c3 , we set n = 2:
" # %
√ √ $ √
c3 = 5 + 2 + c2 = 5 + 2+ 5+ 3 = 5+ 7+ 3 ≈ 7.955.
Similarly,
" %
√ √
c4 = 5 + 2 + c3 = 5 + 7+ 7+ 3 ≈ 8.155 and
& " %
√ √
c5 = 5 + 2 + c4 = 5 + 7+ 7+ 7+ 3 ≈ 8.187.
When the terms of a sequence are defined by a formula such as the one in Example 10.2, the
sequence is said to be defined recursively. The terms for a sequence obtained from Newton’s
iterative procedure are so defined. To find the 100th term of a recursively defined sequence, we
must know the 99th ; to find the 99th , we must know the 98th ; to find the 98th , we need the 97th ;
and so on down the line. In other words, to find a term in the sequence, we must first find every
term that precedes it. Obviously, it is much more convenient to have an explicit definition for cn
in terms of n , but this is not always possible. It can be very difficult to find an explicit formula
for the nth term of a sequence that is defined recursively.
Sometimes it is impossible to give an algebraic formula for the terms of a sequence. This
is illustrated in Exercise 31.
When the general term of a sequence is known explicitly, any term in the sequence is
obtained by substituting the appropriate value of n . In other words, the general term specifies
every term in the sequence. We therefore use the general term to abbreviate the notation for a
sequence by writing the general term in braces. Specifically, for the sequence in Example 10.1a,
we write ' (∞
1 1 1 1 1
1
= 1, , , , . . . , n−1 , . . . ,
2 n−
1 2 4 8 2
where 1 and ∞ indicate that the first term corresponds to the integer n = 1, and that there is
an infinite number of terms in the sequence. In general, we write
{cn }∞
1 = c1 , c 2 , c 3 , . . . , c n , . . . . (10.2)
If, as is the case in this section, the first term of a sequence corresponds to the integer n = 1,
we abbreviate the notation further and simply write {cn } in place of {cn }∞ 1 .
Since a sequence {cn } is a function whose domain is the set of positive integers, we can
represent {cn } graphically. The sequences of Example 10.1 are shown in Figures 10.1.

Plots of terms of sequences of Example 10.1
cn cn
1
1
1 3
1 cn = 4
2n−1 1 n
2 2 cn =
1 n+1
1 4
8
2 4 6 8 10 12 14 n 2 4 6 8 10 12 14 n
cn cn
cn = (−1) n |n − 3| cn = (−1) n+1
10
2
5 1
2 4 6 8 10 12 14 n 2 4 6 8 10 12 14 n
−1
−5
−2
−10
In most applications of sequences we are interested in a number called the limit of the
sequence. Intuitively, a number L is called the limit of a sequence {cn } if as we go farther and
farther out in the sequence, the terms get arbitrarily close to L and stay close to L . If such a
number L exists, we write
L = lim cn , (10.3)
n→∞
and say that the sequence {cn } converges to L . If no such number exists, we say that the
sequence does not have a limit, or that the sequence diverges.
For the sequences of Example 10.1, it is evident that:
1 n
(a) lim = 0. (b) lim = 1.
n→∞ 2n−1 n→∞ n+1
(c) lim (−1)n |n − 3| does not exist. (d) lim (−1)n+1 does not exist.
n→∞ n→∞
Note how the points on the graphs in Figures 10.1a and b cluster around the limits 0 and 1 as n
gets larger and larger. No such clustering occurs in the remaining two figures.
It is usually, but not always, easy to determine whether an explicitly defined sequence has
a limit, and what that limit is. It is like finding the limit of a function f (x) as x → ∞ . For
example, if we divide numerator and denominator of the sequence {(n2 +n− 3)/(2n2 +n+ 4)}
by n2 ,
1 3
n2 + n − 3 1+ − 2
lim = lim n n = 1.
2
n→∞ 2n + n + 4 n→∞ 1 4 2
2+ + 2
n n
Seldom is it obvious whether a recursively defined sequence has a limit. For instance, it is not
at all clear whether the recursive sequence of Example 10.2 has a limit. In spite of the fact
that differences between successive terms are approaching zero, and terms of the sequence are
therefore
√ getting closer together, the sequence might not have a limit. For instance, the sequence
{ n} does not have a limit, yet it is easy to show that differences between successive terms get
smaller and smaller as n increases. (See Exercise 30 for another example.) There are ways to
verify that a sequence has a limit, and some of these will be discussed in Section 10.8.
When a sequence arises in applications, it may be perfectly clear that the sequence is
convergent. Such is often the case in the field of numerical analysis, where recursive sequences
commonly arise in the form of iterative procedures. We have already encountered Newton’s
iterative procedure as one example. When this method is applied to the equation
f (x) = x 3 − 3x + 1 = 0
with initial approximation x1 = 0.7 to find the root between 0 and 1, the sequence obtained is
f (xn ) x 3 − 3xn + 1
x1 = 0.7, xn+1 = xn − "
= xn − n 2 , n ≥ 1.
f (xn ) 3xn − 3
The first three terms of this sequence are illustrated in Figure 10.2, and the tangent line construc-
tion by which they are obtained makes it clear that the sequence must converge to the solution
of the equation between 0 and 1. In other words, it is not necessary for us to verify convergence
of the sequence algebraically; it is obvious geometrically.
FIGURE 10.2 Graphical illustration of sequence from Newton’s iterative procedure
y
1
y = x 3 − 3x + 1
0.75
0.5 (x2 , f (x2))

0.25
x2
0.2 x3 0.4 0.6 x1 0.8 x

−0.25
−0.5
(x1, f (x1))
−0.75
To find the solution of the equation we evaluate terms of the sequence algebraically until
they repeat:
x1 = 0.7, x2 = 0.205, x3 = 0.342, x4 = 0.347 285,

x5 = 0.347 296 355, x6 = 0.347 296 355.
Obviously, x6 = 0.347 296 355 must be close to the root of x 3 − 3x + 1 = 0 between 0 and
1. How close can be verified with the zero intermediate value theorem from Section 1.11. For
instance, to verify that 0.347 296 is an approximation to the solution correctly rounded to six
decimal places, we calculate
f (0.347 295 5) = 2.3 × 10−6 and f (0.347 296 5) = −3.8 × 10−7 .
The fact that function values are opposite in sign verifies that 0.347 296 is correct to six decimal
places.
Because Figure 10.2 illustrated that the sequence defined by Newton’s iterative procedure
must converge to a root of f (x) = x 3 − 3x + 1 = 0, it was unnecessary to verify existence of a
limit of the sequence algebraically. Are we proposing that a graph of the function f (x) should
always be drawn when using Newton’s method to solve equations? Not really, although we are
of the philosophy that pictures should be drawn whenever they are helpful. Numerical analysts
have proved that under very mild restrictions, the sequence defined by Newton’s method always
converges to a root of the equation provided that the initial approximation is sufficiently close
to that root. See Exercise 68 for further discussion of this point. In practice, we approximate
the solution of an equation f (x) = 0 using Newton’s method exactly as we did in Section 4.1,
and as illustrated above. We set up the appropriate sequence, choose an initial approximation,
and iterate to find further approximations. We do not verify, algebraically or geometrically, that
the sequence has a limit. If terms get closer and closer together, we suspect that the sequence is
convergent, that is, that terms are getting closer and closer to the limit. We do not know this for
sure, nor do we try to verify it. The real problem is to solve the equation f (x) = 0. We have
what we believe is an approximation to the solution, and we verify this directly with the zero
intermediate value theorem.
Another way to solve for the same root of x 3 − 3x + 1 = 0 is to rewrite the equation in
the form
1
x = (x 3 + 1),
3
and use this to define the following recursive sequence:
1 1
x1 = , xn+1 = (xn3 + 1), n ≥ 1.
3 3
The initial term was chosen somewhat arbitrarily. In practice, it should be as close to the root
as possible. The first eight terms of the sequence are
x1 = 1/3, x5 = 0.347 293 53,

x2 = 0.345 679, x6 = 0.347 296 01,
x3 = 0.347 102 19, x7 = 0.347 296 31,
x4 = 0.347 272 95, x8 = 0.347 296 35.
The fact that x7 and x8 agree to six decimal places leads us to believe that the sequence
converges and that its limit is approximately equal to 0.347 296. We cannot be certain of this,
nor do we attempt to verify it. What we want is an approximation to the root of x 3 − 3x + 1 = 0
between 0 and 1. We can verify directly that x = 0.347 296 is an approximation accurate to
six decimal places, as above.
This method of finding the root of an equation is often called the method of successive
approximations or fixed-point iteration. What we do is rearrange the equation f (x) = 0
into the form x = g(x) , and define a recursive sequence,
x1 = A, xn+1 = g(xn ), n ≥ 1,
where A is some initial approximation to the root (the closer the better). We iterate hoping that
terms of the sequence get closer together. If they do, we terminate iterations when we suspect
that we have an approximation to the solution of the equation with the required accuracy, and
verify this directly with the zero intermediate value theorem. If terms do not seem to converge,
we may try a different rearrangement of the equation into the form x = g(x) , or another
method.
We can visualize the method of successive approximations geometrically. In terms of the
example above, we are finding where the graph of the function f (x) = x 3 − 3x + 1 crosses the
x -axis (Figure 10.3). There are three points, and we have been concentrating on the intercept
between x = 0 and x = 1. When we rewrite the equation in the form x = (x 3 + 1)/3, an
alternative interpretation is possible. We are searching for x -coordinates of points of intersection
of the curves y = x and y = (x 3 + 1)/3 (Figure 10.4).
FIGURE 10.3 Solutions of equation x 3 − 3x + 1 = 0 FIGURE 10.4 Solutions of x 3 − 3x + 1 = 0 as x -coordinates of

points of intersection of curves y = x , y = (x 3 + 1)/3
y y
7.5
2
5
y = x 3 − 3x + 1 1 3
1 y= (x + 1)
2.5 3
−2 −1 1 2 x
−3 −2 −1 1 2 3x
−2.5 −1
y=x
−5
−2
To show how the method of successive approximations works, we have expanded that part of
Figure 10.4 between x = 0.33 and x = 0.35 (Figure 10.5). To find x2 we substitute x1 = 1/3
into (x 3 + 1)/3. It is the y -coordinate of the point A on the curve y = (x 3 + 1)/3. If we
proceed horizontally from A to line y = x , then point B has coordinates (x2 , x2 ) . To find x3
algebraically, we evaluate (x 3 + 1)/3 at x2 . Geometrically, x3 is the y -coordinate of point C .
If we move horizontally to the line y = x , coordinates of D are (x3 , x3 ) . Continuation leads
to the sequence converging to the required root.
We illustrate the method again in the following example.
FIGURE 10.5 Sequence from successive approximations applied to x = (x 3 + 1)/3
y
0.35
1 3
y= (x + 1) C
3
A D (x3 , x3)
0.345 B (x2 , x2)
0.34
y=x
0.335
x2 x3
x
x1 0.335 0.34 0.345 0.35
EXAMPLE 10.3
Use successive approximations to approximate the solution of f (x) = x 3 + 25x − 50 = 0
between x = 1 and x = 2 with error less than 10−6 . Illustrate the sequence graphically.
SOLUTION The equation can be written in the form x = (50 − x 3 )/25, leading to the
sequence
50 − xn3
x1 = 1, xn+1 = , n ≥ 1.
25
The next 15 terms are
x2 = 1.96, x3 = 1.70, x4 = 1.80,
x5 = 1.77, x6 = 1.778 2, x7 = 1.775 1,
x8 = 1.776 3, x9 = 1.775 81, x10 = 1.776 00,
x11 = 1.775 93, x12 = 1.775 954, x13 = 1.775 945,
x14 = 1.775 948 2, x15 = 1.775 946 9, x16 = 1.775 947 4.
The terms certainly appear to be getting closer together (agreeing to six decimal places at this
stage), therefore suggesting that the sequence has a limit. We do not verify this. We have what
we feel is an approximation to the required solution of x 3 + 25x − 50 = 0 with sufficient
accuracy. To verify that 1.775 947 is a solution with error less than 10−6 , we use the zero
intermediate value theorem with the calculations
f (1.775 946) = −4.4 × 10−5 , f (1.775 948) = 2.5 × 10−5 .
The terms of the sequence are shown in Figure 10.6.
FIGURE 10.6 Sequence from successive approximations applied to x = (50 − x 3 )/25
y
3
2.5 y=x
2
50 − x 3
y=
1.5 25
0.5
x1 x3 x2
0 0.5 1 1.5 2 2.5 3 x
We are being presented with a chemical problem. A precipitate rests at the bottom of a
beaker that holds volume V of a mother liquid. The mother liquid is to be removed in
the following way. Volume Ṽ of pure water is added to the beaker and the liquids are
mixed. Then Ṽ of the mixture is removed, leaving volume V of a mixture that contains
less mother liquid than originally held by the beaker. This is called a decantation. Our
problem is to determine how many decantations are necessary before the amount of mother
liquid is less than 1% of its original amount.
SOLUTION What we need is a formula Cn for the amount of mother liquid in the
beaker after the nth decantation. After the first addition of pure water, the concentration
of mother liquid is V /(V + Ṽ ) . After mixture is removed, the amount of mother liquid
remaining is
V Ṽ V2
C1 = V − = .
V + Ṽ V + Ṽ
After the addition of pure water for the (n + 1)st time, the concentration of mother liquid
is Cn /(V + Ṽ ) . After mixture is removed, the amount of mother liquid remaining is
) *
Cn Ṽ V
Cn+1 = Cn − = Cn .
V + Ṽ V + Ṽ
Iteration of this recursive formula gives

) * ) *
V V3 V V4
C2 = C1 = , C3 = C2 = ,
V + Ṽ (V + Ṽ )2 V + Ṽ (V + Ṽ )3
and so on. The pattern emerging is
V n+1
Cn = .
(V + Ṽ )n
This will be less than 1% of V when
V n+1 V Vn 1
< )⇒ < .
(V + Ṽ )n 100 (V + Ṽ )n 100
When we take logarithms,

) * ) *
V 1 − ln 100
n ln < ln )⇒ n> .
V + Ṽ 100 ln [V /(V + Ṽ )]
The smallest such integer n satisfying this inequality is the minimum number of decanta-
tions required. For instance, if V = 100 cm 3 and Ṽ = 200 cm 3 , then
ln 100
n>− = 4 .2 .
ln [100/300]
Five decantations would be required.
EXERCISES 10.1
2
In Exercises 1–20 determine whether the sequence is convergent or 13. c1 = 2, cn+1 = , n≥1
divergent. Find limits for convergent sequences. cn − 1
' ( cn
1 + , 14. c1 = 4, cn+1 = − 2, n ≥ 1
1. 2. 3n + 1 n
n ' ( ' (
-) * . n+1 2n + 3
n+1 15. 16.
3 2n + 3 n2 − 5
3. {3} 4. ' (
4 n2 + 5n − 4 + ,
-) * . -) 17. 18. ne−n
n+1 *n+5 . n2 + 2n − 2
4 15 ' ( ' (
5. 6. − ln n n
3 16 19. 20. Tan −1 n
' ( n n+1
/ # nπ $0 n
7. sin 8.
2 n2 + n + 2
' ( In Exercises 21–25 find an explicit formula for the general term of the
(−1)n + −1
, sequence. In each case assume that the remaining terms follow the
9. 10. Tan n
n pattern suggested by the given terms.
' (
ln n 1 3 7 15 31
11. 21. , , , , ,...
n2 + 1 2 4 8 16 32
/ 1 0 7 10 13 16 19
12. (−1)n n2 + 1 22. 4, , , , , ,...
4 9 16 25 36
ln 2 ln 3 ln 4 − ln 5
23. √ , − √ , √ , √ , . . . 46. f (x) = x 4 − 120x + 20; x1 = 0 ; use x = (x 4 + 20)/120
2 3 4 5
47. f (x) = x 3 − 2x 2 − 3x + 1; x1 = 3; use x = (2x 2 + 3x −
∗ 24. 1, 0, 1, 0, 1, 0, . . . 1)/x 2
∗ 25. 1, 1, −1, −1, 1, 1, −1, −1, . . .
48. f (x) = 8x 3 − x 2 − 1; x1 = 0 ; use x = (1/2)(1 + x 2 )1/3
In Exercises 26–29 show how L’Hôpital’s rule can be used to evaluate

In Exercises 49–54 find a rearrangement of the equation that leads,
the limit for the sequence.
through the method of successive approximations, to a four-decimal-
' ( ' ( place approximation to the root of the equation.
ln n n3 + 1
26. √ 27.
n en ∗ 49. x 3 − 6x 2 + 11x − 7 = 0 between x = 3 and x = 4
' ) *( ') * (
4 n+5 n 4 2
∗ 50. x − 3x − 3x + 1 = 0 between x = 0 and x = 1
∗ 28. n sin ∗ 29.
n n+3 ∗ 51. x 4 + 4x 3 − 50x 2 + 100x − 50 = 0 between x = 0 and x = 1
∗ 52. sin2 x = 1 − x 2 between x = 0 and x = 1
∗ 30. If a sequence converges, then differences between successive terms
in the sequence must approach zero. The converse is not always true. ∗ 53. sec x = 2/(1 + x 4 ) between x = 0 and x = 1
The following example is an illustration. Show that differences between
∗ 54. ex + e−x − 10x = 0 between x = 0 and x = 1
successive terms of the sequence {ln n} approach zero, but the sequence
itself diverges.
∗ 55. (a) Use Newton’s iterative procedure with x1 = 1 to approx-
∗ 31. (a) The nth term of a sequence is the nth prime integer (greater imate the root, between 0 and 1, of x 4 − 15x + 2 = 0,
than 1) when all such primes are listed in ascending size. accurate to six decimal places.
List its first 10 terms. (b) Use the method of successive approximations with x1 = 1
(b) Can you give a formula for the nth term? and xn+1 = (xn4 + 2)/15 to approximate the root in part
(a), accurate to six decimal places.
(c) Use Newton’s method to approximate the root between 2
In Exercises 32–43 use Newton’s iterative procedure with the given
and 3.
initial approximation x1 to define a sequence of approximations to a
solution of the equation. Determine graphically whether the sequence (d) What happens if the sequence in part (b) is used to approx-
has a limit. Approximate any limit that exists to seven decimal places. imate the root between 2 and 3 with x1 = 2 and x1 = 3?
∗ 56. A superball is dropped from the top of a building 20 m high. Each
32. x1 = 1, x 2 + 3x + 1 = 0 time it strikes the ground, it rebounds to 99% of the height from which
33. x1 = −1, x 2 + 3x + 1 = 0 it fell.
(a) If dn denotes the distance travelled by the ball between the
34. x1 = −1.5, x 2 + 3x + 1 = 0 nth and (n + 1)th bounces, find a formula for dn .
35. x1 = −3, x 2 + 3x + 1 = 0 (b) If tn denotes the time between the nth and (n + 1)th
bounces, find a formula for tn .
36. x1 = 4, x 3 − x 2 + x − 22 = 0
∗ 57. A dog sits at a farmhouse patiently watching for his master to return
37. x1 = 2, x 3 − x 2 + x − 22 = 0
from the fields. When the farmer is 1 km from the farmhouse, the dog
38. x1 = 2, x 5 − 3x + 1 = 0 immediately takes off for the farmer. When he reaches the farmer, he
turns and runs back to the farmhouse, whereupon he again turns and
39. x1 = 1, x 5 − 3x + 1 = 0 runs to the farmer. The dog continues this frantic action until the farmer
40. x1 = 0, x 5 − 3x + 1 = 0 reaches the farmhouse. If the dog runs twice as fast as the farmer, find
the distance dn run by the dog from the point when he reaches the
41. x1 = 4/5, x 5 − 3x + 1 = 0 farmer for the nth time to the point when he reaches the farmer for the
(n + 1)th time. Ignore any accelerations of the dog in the turns.
42. x1 = 0.85, x 5 − 3x + 1 = 0
∗ 58. The equilateral triangle in the left figure below has perimeter P . If
43. x1 = −2, x 5 − 3x + 1 = 0 each side of the triangle is divided into three equal parts, an equilateral
triangle is drawn on the middle segment of each side, and the figure
transformed into the middle figure, what is the perimeter P1 of this
In Exercises 44–48 illustrate that the method of successive approxi-
figure? If each side of this figure is now subdivided into three equal
mations with the suggested rearrangement of the equation f (x) = 0,
along with the initial approximation x1 , leads to a sequence that con-
verges to a root of the equation. Find the root accurate to four decimal
places.
44. f (x) = x 2 − 2x − 1; x1 = 2; use x = 2 + 1/x
45. f (x) = x 3 + 6x + 3; x1 = −1; use x = −(1/6)(x 3 + 3)

portions and equilateral triangles are similarly constructed to result that is, Fn is a linear combination of cn and the M terms immediately
in the right figure, what is the perimeter P2 of this figure? If this before cn . The b0 , b1 , . . . , bM are fixed constants. (The running
subdivision process is continued indefinitely, what is the perimeter Pn averagers in Exercises 63 and 64 are causal FIR filters.) If cn =
after the nth subdivision? What is limn→∞ Pn ? {n/(n + 1)} , M = 2 with b0 = 1, b1 = 2, and b2 = −1, calculate
the first 10 terms of {Fn }∞
3 .
∗ 59. A stone of mass 100 g is thrown vertically upward with speed
20 m/s. Air exerts a resistive force on the stone proportional to its ∗ 66. Repeat Exercise 65 with cn = {(1/n2 ) sin (n/3)} , M = 3, b0 =
speed, and has magnitude 1/10 N when the speed of the stone is 10 1, b1 = −2, b2 = 3, b3 = −4. List terms rounded to four decimal
m/s. It can be shown that the height y (in metres) above the projection places.
point attained by the stone is given by
2 3 ∗ 67. (a) Show that if α is the only root of the equation x = g(x) and
y = −98.1t + 1181 1 − e−t/10 , the left figure below is a graph of g(x) , then the right figure
exhibits geometrically the sequence of approximations of
where t is time (measured in seconds with t = 0 at the instant of α determined by the method of successive approximations.
projection).
y y y=x
(a) The time taken for the stone to return to its projection point
can be obtained by setting y = 0 and solving the equation y = g (x)
for t . Do so (correct to two decimal places). y = g (x)
(b) Find the time for the stone to return if air resistance is ig-
nored.
∗ 60. When the beam in the figure below vibrates vertically, there are cer-
tain frequencies of vibration, called natural frequencies of the system. x x4 x3 x2 x1 x
They are solutions of the equation (b) Illustrate graphically how the sequence defined by the
x −x method of successive approximations converges to the root
e −e x = α for the equation x = g(x) if g(x) is as shown in
tan x =
ex + e−x the left figure below.
divided by 20π . Find the two smallest natural frequencies.
y y
10 m y = g (x)
y = g (x)
3
2
1 3
2
123 x1 x 1
123 x1 x
∗ 61. If An is the area of the figure with perimeter Pn in Exercise 58,
find a formula for An . (c) Illustrate graphically that the method fails for the function
∗ 62. What are the next two terms in the sequence 1, 11, 21, 1211, g(x) in the right figure above.
111221, 312211, 13112221? (d) Based on the results of parts (a)–(c), what determines suc-
cess or failure of the method of successive approximations?
∗ 63. (a) Plot the first 20 terms of the sequence {cn } , where cn =
Part (e) provides a proof of the correct answer.
(1.02)n + 0.5 cos (π n/4 + π/4) .
(e) Prove that when x = g(x) has a root x = α , the method
(b) In digital signal processing, the sequence {An }∞3 defined of successive approximations with an initial approximation
in terms of cn by An = (cn + cn−1 + cn−2 )/3 is called
of x1 always converges to α if |g " (x)| ≤ a < 1 on the
a causal three-point running averager. Plot its first 18
interval |x −α| ≤ |x1 −α| . In other words, when an equa-
terms.
tion f (x) = 0 is rearranged into the form x = g(x) for
∗ 64. Repeat part (b) of Exercise 63 with a seven-point running averager. the method of successive approximations, success or failure
depends on whether the derivative of g(x) is between −1
∗ 65. Suppose that terms of a sequence {cn } represent a discrete-time and 1 near the required root.
signal. A FIR (finite impulse response) filter is a sequence {Fn } whose
terms are linear combinations of the terms of {cn } . It is causal if Fn is
∗∗ 68. Suppose that f "" (x) exists on an open interval containing a root
of the form
x = α of the equation f (x) = 0. Use the result of Exercises 67 to
prove that if f " (α) -= 0, Newton’s iterative sequence always converges
Fn = b0 cn + b1 cn−1 + · · · + bM−1 cn−M+1 + bM cn−M to α provided the initial approximation x1 is chosen sufficiently close
to α . Hint: First use Exercise 67(e) to show that Newton’s sequence
!M converges to α if on the interval |x − α| ≤ |x1 − α| , |ff "" /(f " )2 | ≤
= bk cn−k ; a < 1. [In actual fact, Newton’s method often works even when
k=0 f " (α) = 0.]
10.2 Sequences of Functions

In many applications, we encounter sequences of functions as opposed to sequences of numbers.
A sequence of functions is the assignment of functions to positive integers. For instance, the
first five functions in the sequence {x 2 + 10xe−nx } on the interval 0 ≤ x ≤ 1 are
x 2 + 10xe−x , x 2 + 10xe−2x , x 2 + 10xe−3x , x 2 + 10xe−4x , x 2 + 10xe−5x .
FIGURE 10.7a Sequence {x 2 + 10xe−nx } for 0 ≤ x ≤ 1 FIGURE 10.7b Sequence {x 2 + 10xe−nx } for −1 ≤ x ≤ 0
y y
5
n=1 n=1 x
4 −50
n=2
3 −100
n=2
2 −150
n=3
1 n=4 n=3 −200
y = x2 n=5
1 x −250
n=4 n=5
They are plotted in Figure 10.7a. As n gets larger and larger, values of x 2 + 10xe−nx get closer
and closer to x 2 for all x in the interval 0 ≤ x ≤ 1, and we say that the limit function for this
sequence of functions is f (x) = x 2 . We write that
lim (x 2 + 10xe−nx ) = x 2 , 0 ≤ x ≤ 1.
n→∞
This is also true for x > 1, but not for x < 0. Figure 10.7b shows that for x < 0, the functions
take on increasingly large negative values as n increases.
It is important to realize that a sequence of functions contains many sequences of numbers;
simply substitute a value of x into the functions. For example, if we set x = 1 in the sequence
{x 2 + 10xe−nx } , we obtain the sequence of numbers {1 + 10e−n } . They are the y -coordinates
of the points at the ends of the curves in Figure 10.7a, and the sequence clearly converges to 1. In
fact, for any nonnegative value x0 , the sequence of numbers {x02 + 10x0 e−nx0 } is visualized as
heights of points of intersection of the curves in Figure 10.7a with the line x = x0 (Figure 10.8).
Each such sequence has limit x02 .
FIGURE 10.8 Sequence of points obtained from a sequence of functions
y
5 n=1
4
3
n=2
2
n=3
1 n=4
y = x2 n=5
x0 1 x
EXAMPLE 10.4
Plot the first five functions of the sequence {(x − 1)/[n + n2 (x − 1)2 ]} on the interval −5 ≤
x ≤ 5. What is the limit function for the sequence?
SOLUTION The first five functions are plotted in Figure 10.9. Geometrically, and alge-
braically, it is clear that
x−1
lim =0 for all x.
n→∞ n + n2 (x − 1)2
/ 0
x−1
FIGURE 10.9 Sequence of functions n+n2 (x−1)2
y
0.4
n=1
0.2
n=2
−4 −2 2 4 x
−0.2
−0.4
The sequence in the next example is indispensable to future discussions.
EXAMPLE 10.5
Show that the limit of the sequence of functions {|x|n /n!} is zero for all x ; that is, verify that
|x|n
lim = 0, −∞ < x < ∞.
n→∞ n!
SOLUTION We can see this intuitively. As n increases by 1, a fixed, extra |x| -factor appears
in the numerator, but an ever increasing factor n occurs in the denominator. When n surpasses
x , the fraction will get smaller and smaller. Let us prove this analytically.
Suppose that x is any fixed value, and let m = .|x|/ , where .x/ is the floor function
of Exercise 68 in Section 1.5. It is the largest integer that does not exceed x . (If x = 3.4,
then .|3.4|/ = 3; if x = −22.6, then .| − 22.6|/ = .22.6/ = 22; and if x = −4, then
.|− 4|/ = .4/ = 4.) Since |x|n is |x| multiplied by itself n times, and n! has n multiplications,
we may write that
) *) *) * ) *
|x|n |x| |x| |x| |x|
= ··· .
n! 1 2 3 n
If n is chosen greater than m = .|x|/ , then
) *) *) * ) *) * ) *
|x|n |x| |x| |x| |x| |x| |x|
= ··· ··· .
n! 1 2 3 m m+1 n
Suppose we let
) *) *) * ) *
|x| |x| |x| |x|
M = ··· ,
1 2 3 m
a fixed constant. Then
) * ) * *)
|x|n |x|
|x| |x|
=M ···
n! m+1
m+2 n
) *) * ) *
|x| |x| |x|
<M ··· ,
m+1 m+1 m+1
where we have replaced m+ 2, m+ 3, . . . , n in the denominators by the smaller integer m+ 1.

Since there are n − m bracketed terms on the right, all equal to each other, we have
) *n−m ) *−m ) *n
|x|n |x| |x| |x|
<M =M .
n! m+1 m+1 m+1
Now, M [|x|/(m + 1)]−m is a fixed number as far as the limit on n is concerned. Furthermore,
|x|/(m + 1) is also constant and is less than 1. Consequently,
) *−m ) *n
|x|n |x| |x|
lim ≤ lim M
n→∞ n! n→∞ m+1 m+1
) *−m ) *n
|x| |x|
=M lim
m+1 n→∞ m + 1
) *−m
|x|
=M (0 )
m+1
= 0.
Since |x|n /n! ≥ 0, it follows that limn→∞ |x|n /n! = 0.
EXERCISES 10.2
In Exercises 1–14 fn (x) is the nth term in a sequence of functions 7. fn (x) = n2 x n (2 − x) , 0 ≤ x ≤ 2
{fn (x)} . Plot graphs of the first five functions in the sequence. Deter- 8. fn (x) = n2 x n (1 − x 2 ) , −1 ≤ x ≤ 1
mine whether the sequence has a limit.
2 + nx 2
9. fn (x) = ,0≤x≤2
nx 1 + nx
1. fn (x) = ,0≤x≤1
1 + n2 x 2 ∗ 10. fn (x) = (sin x)1/n , 0 < x < π
n2 x ∗ 11. fn (x) = (sin x)1/n , 0 ≤ x ≤ π
2. fn (x) = ,0≤x≤1 ) *1/n
1 + n3 x 2 sin x
∗ 12. fn (x) = ,0<x<π
nx 2 x
3. fn (x) = ,0≤x≤1 ) *1/n
1 + nx sin x
) * ∗ 13. fn (x) = ,0<x≤π
1 1 1 x
4. fn (x) = + sin ,1≤x≤2
x n nx
∗ 14. fn (x) = n2 xe−nx , 0 ≤ x < ∞
5. fn (x) = nx n (1 − x) , 0 ≤ x ≤ 1 ∗ 15. For what values of x does the sequence of functions {(1−x n )/(1−
6. fn (x) = nx n (1 − x) , 0 ≤ x ≤ 2 x)} have a limit?
10.3 Taylor Polynomials, Remainders, and Series

FIGURE 10.10 Oscilla- Figure 10.10 shows a pendulum consisting of a mass on the end of a string of length L . If the
tions of a pendulum mass is pulled slightly to the side and released, it swings back and forth for some time to come.
The position of the mass can be described by the angle θ that the string makes with the vertical.
By analyzing the forces acting on the mass, it can be shown that when air resistance is neglected,
θ , as a function of time t , must satisfy the differential equation
d 2θ g
+ sin θ = 0, (10.4)
L
dt 2 L
where g = 9.81 is the acceleration due to gravity. Thus, θ (t) is a function whose second
derivative is −g/L times the sine of itself. This is a very difficult differential equation to solve;
in fact, no combination of the simple functions with which we are familiar satisfies equation
10.4. What is sometimes done in examples like this is replace sin θ with a simpler function
f (θ ) that simultaneously approximates sin θ and for which the differential equation
d 2θ g
2
+ f (θ ) = 0 (10.5)
dt L
is solvable. Although the solution of equation 10.5 only approximates the solution of 10.4,
if f (θ ) approximates sin θ very closely, the solution of 10.5 may be sufficiently close to the
solution of 10.4 to give real insight into the motion of the pendulum. In particular, as we shall
show shortly, when θ is very small, sin θ can be approximated by θ itself. In other words, for
small oscillations of the pendulum, equation 10.4 can be replaced by
d 2θ g
2
+ θ = 0. (10.6)
dt L
It is straightforward to verify that
" "
g g
θ (t) = A cos t + B sin t
L L
is a solution of equation 10.6 for any constants A and B whatsoever. If motion is initiated at
time t = 0 by pulling the mass an amount θ0 to the right and then releasing it, then θ (t) must
satisfy the additional conditions
θ ( 0 ) = θ0 , θ " (0 ) = 0 .
These imply that A = θ0 and B = 0, and hence

"
g
θ (t) = θ0 cos t.
L
This function yields simple harmonic motion for the pendulum, as we expect.
In this section we show how to approximate complicated functions by polynomials. The
interval on which the approximation is required and the accuracy of the approximation dictate the
degree of the polynomial. To begin with, we recall the Mean Value Theorem from Section 3.14.
It states that when f (x) is continuous on the closed interval between c and some given value
of x , and f " (x) exists on the open interval between c and x , there exists a number between c
and x , call it z0 , such that
f (x) − f (c)
= f " (z0 ),
x−c
or,
f (x) = f (c) + f " (z0 )(x − c). (10.7)
This theorem was a corollary to Cauchy’s generalized mean value theorem in Section 3.14.
We give a direct proof of it here because a similar but more complicated argument leads to an
extension called Taylor’s Remainder Formula, a result that is fundamental to our studies in this
chapter.
Consider the following function of y for fixed c and x ,
) *
x−y
F (y) = f (x) − f (y) − [f (x) − f (c)].
x−c
Because f is continuous on the closed interval between c and x , F (y) is continuous for y in
this interval. Furthermore,
1
F " (y) = −f " (y) + [f (x) − f (c)].
x−c
Since f " (x) exists on the open interval between c and x , so also does F " (y) . Finally, F (c) =
0 = F (x) . Rolle’s Theorem (Theorem 3.17 in Section 3.14) implies that there exists a number
z0 between c and x at which
1
0 = F " (z0 ) = −f " (z0 ) + [f (x) − f (c)],
x−c
and this gives equation 10.7.
Suppose now that f (x) and f " (x) are continuous on the closed interval between c and x ,
and f "" (x) exists on the open interval between c and x . Consider the following function of y ,
) *2
" x−y
F (y) = f (x) − f (y) − f (y)(x − y) − [f (x) − f (c) − f " (c)(x − c)].
x−c
Because f (x) and f " (x) are continuous on the closed interval between c and x , F (y) is
continuous for y in this interval. Furthermore,
2(x − y)
F " (y) = −f " (y) − f "" (y)(x − y) + f " (y) + [f (x) − f (c) − f " (c)(x − c)].
(x − c)2
Since f " (x) is continuous on the closed interval between c and x , and f "" (x) exists on the
open interval between c and x , it follows that F " (y) exists on the open interval. Finally,
F (c) = 0 = F (x) , and therefore Rolle’s Theorem implies the existence of a number z1
between c and x at which
2(x − z1 )
0 = F " (z1 ) = −f " (z1 )−f "" (z1 )(x −z1 )+f " (z1 )+ [f (x)−f (c)−f " (c)(x −c)].
(x − c)2
This can be rearranged into the form
f "" (z1 )
f (x) = f (c) + f " (c)(x − c) + (x − c)2 . (10.8)
2
By assuming that f (x) , f " (x) , and f "" (x) are continuous on the closed interval between c and
x , and f """ (x) exists on the open interval, it can be shown in a similar way that there exists a
number z2 between c and x such that
f "" (c) f """ (z2 )

f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 . (10.9)
2 3!
These results can be extended indefinitely if we assume that f (x) has a sufficient number of
derivatives on the closed interval between c and x . The complete result is contained in the
following theorem.
THEOREM 10.1 (Taylor’s Remainder Formula)

If f (x) and its first n derivatives are continuous on the closed interval between c and x ,
and if f (x) has an (n + 1)th derivative on the open interval between c and x , then there
exists a point zn between c and x such that
f "" (c) f """ (c)

f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 + · · ·
2! 3!
f (n) (c) f (n+1) (zn )
+ (x − c)n + (x − c)n+1 . (10.10)
n! (n + 1)!
The notation f (n) (c) represents the nth derivative of f (x) evaluated at x = c .
In the remainder of this section, we assume that f (x) has derivatives of all orders on the
closed interval between c and some value of x . Taylor’s remainder formula can then be written
down for all values of n . For n = 0, 1, 2, and 3, we obtain
f (x) = f (c) + f " (z0 )(x − c), (10.11a)

f "" (z1 )
f (x) = f (c) + f " (c)(x − c) + (x − c)2 , (10.11b)
2!
f "" (c) f """ (z2 )
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 , (10.11c)
2! 3!
f "" (c) f """ (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3
2! 3!
f """" (z3 )
+ (x − c)4 . (10.11d)
4!
The last terms in these equations are called Taylor remainders, and when they are denoted by
R0 , R1 , R2 , and R3 , respectively,
f (x) = f (c) + R0 , (10.12a)
"
f (x) = f (c) + f (c)(x − c) + R1 , (10.12b)
f "" (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + R2 , (10.12c)
2!
f "" (c) f """ (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 + R3 , (10.12d)
2! 3!
and in general,
f "" (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + · · ·
2!
f (n) (c)
+ (x − c)n + Rn . (10.13)
n!
We call Rn the remainder for the simple reason that, if we drop the remainders from equations
10.12, then what remains is a sequence of polynomial approximations to f (x) , called the Taylor
polynomials of f (x) about c :
f (x) ≈ P0 (x) = f (c), (10.14a)

"
f (x) ≈ P1 (x) = f (c) + f (c)(x − c), (10.14b)
""
f (c)
f (x) ≈ P2 (x) = f (c) + f " (c)(x − c) + (x − c)2 , (10.14c)
2!
f "" (c) f """ (c)
f (x) ≈ P3 (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 . (10.14d)
2! 3!
The nth Taylor polynomial Pn (x) is a polynomial of degree n . What happens in practice is
that as n increases, the polynomials Pn (x) approximate f (x) more and more closely on some
finite or infinite interval containing c . The following examples illustrate this.
EXAMPLE 10.6
Find the first six Taylor polynomials for sin x about x = 0. Plot the polynomials and sin x to
show how the approximations improve as more terms are included.
SOLUTION The value of f (x) = sin x and its first five derivatives at x = 0 are
f (0) = 0, f " (0) = cos 0 = 1, f "" (0) = − sin 0 = 0, f """ (0) = − cos 0 = −1,
f (4) (0) = sin 0 = 0, f (5) (0) = cos 0 = 1.

The first six Taylor polynomials are
P0 (x) = f (0) = 0,
P1 (x) = f (0) + f " (0)(x − 0) = x,
f "" (0)
P2 (x) = f (0) + f " (0)(x − 0) + (x − 0)2 = x,
2!
f "" (0) f """ (0) x3
P3 (x) = f (0) + f " (0)(x − 0) + (x − 0)2 + (x − 0)3 = x − ,
2! 3! 3!
x3
P4 (x) = x − ,
3!
x3 x5
P5 (x) = x − + .
3! 5!
Because even-ordered derivatives of sin x vanish, even-numbered polynomials are the same as
their odd predecessors. We have shown polynomials P1 (x) , P3 (x) , and P5 (x) along with
sin x for positive values of x in Figure 10.11. We have included P7 (x) and P9 (x) to further
emphasize the following facts. The higher the degree of the polynomial, the more closely it
approximates sin x for small values of x , and the larger the interval on which the approximation
is reasonable.
FIGURE 10.11 Taylor polynomials of sin x
y
3
P1 P5 P9
2
1
sin x
2 4 6 8 10 x
−1 P0
−2 P3
P7
−3
EXAMPLE 10.7
Find the first five Taylor polynomials of ex about x = 0. Plot the polynomials and ex .
SOLUTION Since all derivatives of ex at x = 0 are equal to 1, the first five polynomials are
P0 (x) = 1,
P1 (x) = 1 + x,
x2
P2 (x) = 1 + x + ,
2!
x2 x3
P3 (x) = 1 + x + + ,
2! 3!
x2 x3 x4
P4 (x) = 1 + x + + + .
2! 3! 4!
They are illustrated in Figure 10.12. When x > 0, the polynomials are always less than ex and
approach ex as more terms are included. For x < 0, the polynomials are alternately higher and
lower than ex , gradually getting closer and closer to ex .
FIGURE 10.12 Taylor polynomials of ex
P4 P
y 3
P2
6
ex
4 P1
P2 2
P4
P0
P0
ex
−3 P3 −2 −1 1 2 3 x
P1 −2
EXAMPLE 10.8
Find the first five Taylor polynomials for ln x about x = 1. Plot the polynomials and ln x .
SOLUTION Since
d 14 d2 1 4
ln x|x=1 = 0, ln x|x=1 = 44 = 1, 2
ln x|x=1 = − 2 44 = −1,
dx x x=1 dx x x=1
d3 2 44 d4 −3! 44
ln x = = 2, ln x = = −3!,
x 3 4 x=1 x 3 4 x=1
|x= 1 |x= 1
dx 3 dx 4
the first five polynomials are
P0 (x) = 0,
P1 (x) = x − 1,
1
P2 (x) = (x − 1) − (x − 1)2 ,
2
1 1
P3 (x) = (x − 1) − (x − 1)2 + (x − 1)3 ,
2 3
1 1 1
P4 (x) = (x − 1) − (x − 1)2 + (x − 1)3 − (x − 1)4 .
2 3 4
They are plotted in Figure 10.13. For 0 < x < 1, the polynomials are greater than ln x and
approach ln x as more terms are included. For 1 < x ≤ 2, the polynomials are alternately
higher and lower than ln x , gradually getting closer and closer to ln x . The plot of the Taylor
polynomials on the interval 1 ≤ x ≤ 3 in Figure 10.14 indicates that they do not approach ln x
for x > 2.
FIGURE 10.13 Taylor polynomials of ln x on 0 < x ≤ 2 FIGURE 10.14 Taylor polynomials of ln x on 1 < x ≤ 3
y P1 y
1 P3 P3 P1
ln x
P4 1.5
P0 P2
0.5 1 1.5 2 x In x
1
P1
−1 P 0.5
2 P2
P3
−2 P4 1.5 2 2.5 3x
ln x −0.5
−3 P4
To make this clearer, below is a list of the first 30 Taylor polynomials evaluated at x = 2.1.
At first these numbers get closer together, but then they separate. We have shown them in Figure
10.15.
0 .0 1.1 0.495 0.938 67 0.572 64 0.894 74
0.599 48 0.877 87 0.609 92 0.871 92 0.612 54 0.871 92
0.610 38 0.875 94 0.605 69 0.883 17 0.595 99 0.893 31
0.584 43 0.906 32 0.569 94 0.922 33 0.552 32 0.941 64
0.531 23 0.964 62 0.506 23 0.991 79 0.476 75 1.023 75
FIGURE 10.15 Taylor polynomials of ln x evaluated at x = 2.1
Pn (2.1)
0.8
0.6
0.4
0.2
5 10 15 20 25 n
The plots in Examples 10.6–10.8 support our earlier contention that as n increases, the Taylor
polynomials of a function approximate the function more and more closely on some finite or
infinite interval. We would like to be able to verify analytically what we see geometrically. To
do this we return to the Taylor remainders. The nth remainder is
f (n+1) (zn )
Rn = Rn (c, x) = (x − c)n+1 , (10.15)
(n + 1)!
where zn (which, as the notation suggests, varies with n ) is always between c and x . We have
written Rn (c, x) to emphasize the fact that remainders depend on the point of expansion c , the
value of x being considered, and the choice of n . The sequence of remainders {Rn (c, x)} is a
sequence of functions. It is these remainders that we drop from Taylor’s remainder formula in
equations 10.12 to form Taylor polynomials in equations 10.14. The polynomials in 10.14 will
approximate f (x) more and more closely as n increases if the neglected remainders Rn (c, x)
get smaller and smaller. In other words, to verify algebraically that the sequence of Taylor
polynomials approximates a function more and more closely with increasing n on an interval
I , it is sufficient to show that for each x in I ,
lim Rn (c, x) = 0. (10.16)
n→∞
Interval I is called the interval of convergence. It is tedious to write groups of equations

like 10.14 and say that, the more and more terms that are added, the better and better is the
approximation. We invent a notation to stand for this. We write
f "" (c) f """ (c)

f (x) = f (c) + f " (c)(x − c) + (x − c)2 + (x − c)3 + · · · . (10.17)
2! 3!
The · · · indicate that additional terms of the same form are to be added indefinitely. This is
called the Taylor series of f (x) about the point c . Be clear on what we mean by 10.17, because
it cannot be taken literally. There is an infinity of terms on the right side of the equation, and
substitution of a value for x (and c ) gives an infinity of numbers to add. No matter how fast you
add, your calculator adds, or a super computer adds, an infinity of numbers cannot be added in
a finite amount of time. Equation 10.17 does not therefore mean that adding the terms on the
right gives f (x) . It means any one of the following three equivalent statements:
1. The more and more terms that are added on the right, the closer and closer the sum gets to
f (x) .
2. By adding sufficiently many terms, we can make the sum on the right as close to f (x) as
desired.
3. If the Taylor series is truncated at any number of terms, the resulting Taylor polynomial
approximates f (x) ; the larger n , the better the approximation.
To use the terminology from Section 10.2, we say that the Taylor polynomials Pn (x) converge
to f (x) , meaning again that they get closer and closer to f (x) , the larger the value of n . We
also say that the Taylor series converges to f (x) .
When c = 0, the Taylor series becomes
f "" (0) f """ (0)

f (x) = f (0) + f " (0)x + x2 + x3 + · · · . (10.18)
2! 3!
We call this the Maclaurin series for f (x) . It consists of powers of x rather than x − c .
To obtain the Taylor (or Maclaurin) series for a function f (x) , we evaluate all derivatives
of f (x) at x = c (or x = 0), and use formula 10.17 (or 10.18). To show that the series
converges to f (x) , we show that Taylor remainders approach 0, and we must find an interval I
on which this happens. This can be quite difficult because Rn (c, x) is defined in terms of zn , an
unknown value except that it must be between c and x . This problem is usually circumvented
by finding a maximum value for |Rn (c, x)| , and showing that this maximum value approaches
zero. We illustrate in the next three examples.
EXAMPLE 10.9
Find the Maclaurin series for sin x and show that it converges to sin x for all x .
SOLUTION Using the derivatives of sin x in Example 10.6, we can write Taylor’s remainder
formula for sin x and c = 0,
x3 x5 x7 dn 4 xn
sin x = x − + − + ··· + ( sin x) 4 + Rn (0, x),
3! 5! 7! dx n 4 x=0 n!
where
d n+1 4 x n+1
Rn (0, x) = ( sin x) 4 .
dx n+1 4 x=zn (n + 1)!
But the (n + 1)th derivative of sin x is ± sin x or ± cos x , so that

4 n+1 4 4
4d 4 4
4 4
4 dx n+1 (sin x) 4 x=zn 4 ≤ 1.
Hence,
|x|n+1
|Rn (0, x)| ≤ .
(n + 1)!
But according to Example 10.5, limn→∞ |x|n /n! = 0 for any x whatsoever, and therefore
lim |Rn (0, x)| = 0 )⇒ lim Rn (0, x) = 0.

n→∞ n→∞
The Maclaurin series for sin x therefore converges to sin x for all x , and we may write
x3 x5 x7
sin x = x − + − + ···
3! 5! 7!
!∞
(−1)n 2n+1
= x ,
n=0
(2n + 1)!
and this is true for all x ; that is, the interval of convergence is −∞ < x < ∞ . Let us not
forget what we mean by writing the Maclaurin series for sin x . If the series is truncated at any
value of n , a Taylor polynomial is obtained. These polynomials approximate sin x more and
more accurately for small x , and the interval on which the polynomials approximate sin x with
any degree of accuracy gets larger and larger. We illustrated this graphically in Figure 10.11.
In a similar way it can be shown that the Maclaurin series for cos x is
∞
! (−1)n x2 x4 x6
cos x = x 2n = 1 − + − + ···, −∞ < x < ∞.
n=0
(2n)! 2! 4! 6!
Sometimes it is necessary to consider points on either side of the point of expansion separately.
This is illustrated in the next example.
EXAMPLE 10.10
Find the Maclaurin series for ex and show that it converges to ex for all x .
SOLUTION Since
d n x 44
(e ) 4 = ex |x=0 = 1,
dx n x=0
Taylor’s remainder formula for ex and c = 0 gives
x2 x3 xn
ex = 1 + x + + + ··· + + Rn (0, x),
2! 3! n!
where
d n+1 x 44 x n+1 zn x
n+1
Rn (0, x) = (e ) 4 x=z (n + 1)! = e .
dx n+1 n (n + 1)!
Now, if x < 0, then x < zn < 0, and
|x|n+1
|Rn (0, x)| < e0 ,
(n + 1)!
which approaches zero as n becomes infinite (see Example 10.5). If x > 0, then 0 < zn < x ,
and
|x|n+1
|Rn (0, x)| < ex ,
(n + 1)!
which again has limit zero as n approaches infinity. Thus, for any x whatsoever, we can say
that limn→∞ Rn (0, x) = 0, and the Maclaurin series for ex converges to ex :
!∞
x2 x3 1 n
ex = 1 + x + + + ··· = x , −∞ < x < ∞.
2! 3! n!
n=0
Separate discussions for x < 0 and x > 0 are reflected in how the Taylor polynomials converge
to ex in Figure 10.12. For x > 0, polynomials are always less than ex and increase toward ex .
For x < 0, polynomials are alternately larger and smaller than ex , gradually approaching ex .
The following example is more difficult. Remainders do not approach zero for all x so that the
Taylor series does not converge to the function for all x . The problem would have been even
more difficult had we not suggested values of x to consider.
EXAMPLE 10.11
Find the Taylor series for ln x about the point 1 and show that, for 1/2 ≤ x ≤ 2, it converges
to ln x .
SOLUTION Using the derivatives of ln x at c = 1 in Example 10.8, we obtain Taylor’s
remainder formula for ln x and c = 1,
1 2! (−1)n+1 (n − 1)!
ln x = (x − 1) − (x − 1)2 + (x − 1)3 + · · · + (x − 1)n + Rn (1, x),
2! 3! n!
where
d n+1 4 (x − 1)n+1 (−1)n n! (x − 1)n+1

Rn (1, x) = ( ln x) 4 =
dx n+1 4 x=zn (n + 1)! (zn )n+1 (n + 1)!
) *
(−1)n x − 1 n+1
= ,
n+1 zn
and zn is between 1 and x .
If 1 < x ≤ 2, then the largest value of x − 1 is 1. Furthermore, zn must be larger than 1.
It follows that
) *n+1
1 1 1
|Rn (1, x)| < =
n+1 1 n+1
and therefore
lim Rn (1, x) = 0.
n→∞
If 1/2 ≤ x < 1, then −1/2 ≤ x − 1 < 0. Combine this with x < zn < 1, and we can
state that −1 < (x − 1)/zn < 0. Then,
1
|Rn (1, x)| < and lim Rn (1, x) = 0.
n+1 n→∞
Thus, for 1/2 ≤ x ≤ 2, the sequence of remainders {Rn (1, x)} approaches zero, and the
Taylor series converges to ln x for those values of x :
1 1
ln x = (x − 1) − (x − 1)2 + (x − 1)3 + · · ·
2 3
∞
! n+1
(−1) 1
= (x − 1)n , ≤ x ≤ 2.
n 2
n=1
This series actually converges to ln x on the larger interval 0 < x ≤ 2 as we saw graphically
in Figure 10.13. We will show this algebraically in Example 10.22.
The above examples were chosen for their simplicity. It is simple to find nth derivatives of
sin x , ex , and ln x ; that remainders approach zero follows immediately. You may disagree that
the examples were simple, but this only serves to emphasize the following point. If these are
the simplest possible examples, and they are not simple, imagine the difficulty in finding and
showing that Taylor series of other functions converge to the function. The problem is so great
that we do our utmost to avoid calculating derivatives and considering remainders. Based on
discussions in Section 10.4, we will find alternatives in Section 10.5.
EXERCISES 10.3
In Exercises 1–5 use Taylor’s remainder formula to show that the Tay- ∗ 16. In Section 4.4 we stated the second-derivative test for determining
lor series for the function f (x) about the point indicated converges to whether a critical point x0 at which f " (x0 ) = 0 yields a relative
f (x) for all x . In each case, plot the first six Taylor polynomials to maximum or a relative minimum. Use Taylor’s remainder formula to
show how they approximate f (x) more closely as the degree of the verify this result when f " (x) and f "" (x) are continuous on an open
polynomial increases. interval containing x0 .
∗ 1. f (x) = cos x about x = 0 ∗ 17. Extend the result of Exercise 16 to verify the extrema test of Ex-
ercise 35 in Section 4.4.
∗ 2. f (x) = e5x about x = 0
∗ 18. There is an integral form for the remainder Rn (c, x) in Taylor’s
∗ 3. f (x) = sin (10x) about x = 0 remainder formula that is sometimes more useful than the derivative
form in Theorem 10.1: If f (x) and its first n derivatives are continuous
∗ 4. f (x) = sin x about x = π/4
on the closed interval between c and x , then
∗ 5. f (x) = e2x about x = 1
f "" (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2
2!
In Exercises 6–15 find the Taylor series for the function f (x) about
the point c . Plot enough polynomials and the function to determine the
f n (c)
+ ··· + (x − c)n + Rn (c, x),
interval on which the series converges to the function. n!
∗ 6. f (x) = e2x , c = 0 where 5 x

1
∗ 7. f (x) = cos 3x , c = 0 Rn (c, x) = (x − t)n f (n+1) (t) dt.
n! c
∗ 8. f (x) = sin x , c = π/2 Use the following outline to prove this result.
∗ 9. f (x) = 1/(1 − x) , c = 0 (a) Show that
∗ 10. f (x) = 1/(2 − x) , c = 1 5 x

f (x) = f (c) + f " (t) dt.
∗ 11. f (x) = x/(1 + 2x) , c = 0 c
∗ 12. f (x) = 1/(1 + 3x)2 , c = 0 (b) Use integration by parts with u = f " (t) , du = f "" (t) dt ,
dv = dt , and v = t − x on the integral in part (a) to
∗ 13. f (x) = ln x , c = 2
obtain
√
∗ 14. f (x) = 1 + 3x , c = 0 5 x
"
1/ 3
f (x) = f (c) + f (c)(x − c) + (x − t)f "" (t) dt.
∗ 15. f (x) = 1/(4 + x) ,c=2 c
(c) Use integration by parts with u = f "" (t) , du = f """ (t) dt , (b) Use L’Hôpital’s rule to show that for every positive integer
dv = (x − t) dt , and v = −(1/2)(x − t)2 to obtain n,
f "" (c) 2
f (x) = f (c) + f " (c)(x − c) + (x − c)2 e−1/x
2! lim = 0.
x→0 x n
5 x
1
+ (x − t)2 f """ (t) dt.
2! c
(c) Prove, by mathematical induction, that f (n) (0) = 0 for
(d) Continue this process to obtain the integral form for Taylor’s n ≥ 1.
remainder formula.
∗∗ 19. (a) Draw a graph of the function (d) What is the Maclaurin series for f (x) ?
' 2
e−1/x , x =
- 0, (e) For what values of x does the Maclaurin series of f (x)
f (x) =
0, x = 0. converge to f (x) ?
10.4 Power Series

Our work on Taylor series in Section 10.3 began with a function and developed the Taylor (or
Maclaurin) series for the function, a sequence of polynomials that approximate the function
more and more closely as more and more terms are included. In this section we begin with a
series called a power series and ask two questions: For what values of x does the power series
converge, and to what function does it converge?
A power series in x is an (infinite) series of the form
∞
!
a n x n = a 0 + a 1 x + a2 x 2 + · · · + a n x n + · · · , (10.19)
n=0
where coefficients an are constants. For example,

∞
!
xn = 1 + x + x2 + x3 + · · · + xn + · · ·
n=0
and
∞
! (−1)n x x2 x3
xn = 1 − + 2 − 3 + ···
2 n n! 2 (1!) 2 (2!) 2 (3!)
n=0
are power series in x .
To say that the Taylor series
f "" (c)
f (x) = f (c) + f " (c)(x − c) + (x − c)2 + · · ·
2!
of a function f (x) converges to f (x) is to say that the sequence of Taylor polynomials {Pn (x)} ,
where
f (n) (c)
Pn (x) = f (c) + f " (c)(x − c) + · · · + (x − c)n ,
n!
converges to f (x) ; that is, as more and more terms of the series are included, the closer and
closer the sum gets to f (x) . We use the same idea to define what we mean by convergence
of power series 10.19. Power series 10.19 is said to converge to a function f (x) , or have sum
f (x) , on an interval I if the limit of the sequence {Pn (x)} of polynomials where
Pn (x) = a0 + a1 x + · · · + an x n
is f (x) for each x in I . In such a case we write

∞
!
f (x) = a n x n = a0 + a1 x + · · · + a n x n + · · · , x in I , (10.20)
n=0
and say that the power series converges to f (x) , or has sum f (x) . Once again what we mean
is that as more and more terms on the right are included, the closer and closer the sum gets to
f (x) on the interval I . We call I the interval of convergence for the power series.
There are two aspects to convergence for power series, interval I and sum f (x) . It is
relatively simple to find the interval of convergence for a power series, but much more difficult
to find its sum.
One of the easiest, and perhaps the most important, power series is
∞
!
ax n = a + ax + ax 2 + ax 3 + · · · , (10.21)
n=0
where a is some nonzero constant. It is so important in applications that it is given a special

name; it is called a geometric series. A geometric series is a series in which every term after
the first is obtained by multiplying the previous term by the same amount, called the common
ratio. For geometric series 10.21, the common ratio is x ; each term is x times the previous
term. To show that the series converges we must show that the sequence of polynomials
Pn (x) = a + ax + ax 2 + · · · + ax n
has a limit. If we multiply Pn (x) by x , then
x Pn (x) = ax + ax 2 + ax 3 + · · · + ax n+1 .
When these are subtracted,
Pn (x) − x Pn (x) = a − ax n+1 ,

from which
a(1 − x n+1 )
Pn (x) = . (10.22)
1−x
This is a simplified form for Pn (x) . Since
-
0, −1 < x < 1
lim x n+1 = 1, x =1
n→∞
does not exist, otherwise
it follows that when the sequence {Pn (x)} is restricted to the interval |x| < 1, it has a limit,
a
lim Pn (x) = , |x| < 1.
n→∞ 1−x
We can therefore write that
∞
! a
x n = a + ax + ax 2 + · · · = , |x| < 1. (10.23)
1−x
n=0
The geometric series has sum a/(1 − x) on the interval (of convergence) |x| < 1. It does not
have a sum for |x| ≥ 1.
Figures 10.16 show some of the polynomials Pn (x) and a/(1 − x) when a = 1. They
illustrate how polynomials approximate 1/(1 − x) more closely as n increases. Polynomials
are defined for all x and 1/(1 − x) is defined for all x -= 1, but curves y = Pn (x) approach
y = 1/(1 − x) only for |x| < 1. For x > 0, polynomials approach 1/(1 − x) from below;
for x < 0, they oscillate about 1/(1 − x) , but gradually approach 1/(1 − x) .

Partial sums of geometric series 1 + x + x 2 + · · ·
y y
3 3
2 y = P4 (x)
2 y = P1(x)
1 1
y= 1 y= 1
1−x 1−x
−2 −1 1 2x −2 −1 1 2x
−1 −1
−2 −2
−3 −3
y y
3 3
y = P7 (x) y = P10 (x)
2 2
1 1
y= 1 y= 1
1−x 1−x
−2 −1 1 2x −2 −1 1 2x
−1 −1
−2 −2
−3 −3
6∞
Obviously, the interval of convergence for a power series n=0 an x n always includes the
value x = 0, but what other possibilities are there? For geometric series 10.23, the interval
of convergence is the interval −1 < x < 1. The terminology itself, interval of convergence,
suggests that the values of x for which a power series converges form some kind of interval.
This is indeed true, as we shall soon see.
Every Maclaurin series is a power series in x . Examples 10.9 and 10.10 developed Maclau-
rin series for sin x and ex ,
x3 x5 x2 x3
sin x = x − + + ···, ex = 1 + x + + + ···.
3! 5! 2! 3!
They are power series that add to the functions with intervals of convergence −∞ < x < ∞ .
The power series
∞
!
n! x n = 1 + x + 2! x 2 + 3! x 3 + · · ·
n=0
converges only for x = 0. This is a direct result of the fact that
lim n! |x|n = ∞
n→∞
for any given nonzero value of x . How, then, could the addition of more and more terms ever
converge? These examples have illustrated
6∞ that there are at least three possible types of intervals
of convergence for power series n=0 an x n :
1. The power series converges only for x = 0;
2. The power series converges for all x ;
3. There exists a number R > 0 such that the power series converges for |x| < R , diverges
for |x| > R , and may or may not converge for x = ±R .
These are in fact the only possibilities for an interval of convergence. In 3 we call R the
radius of convergence of the power series. It is half the length of the interval of convergence,
or the distance we may proceed in either direction along the x -axis from x = 0 and expect
convergence of the power series, with the possible exceptions of x = ±R . In order to have a
radius of convergence associated with every power series, we say in 1 and 2 above that R = 0
and R = ∞ , respectively.
6∞
Every power series n=0 an x n now has a radius of convergence R . If R = 0, the power
series converges only for x = 0; if R = ∞ , the power series converges for all x ; and if
0 < R < ∞ , the power series converges for |x| < R , diverges for |x| > R , and may or may
not converge for x = ±R . For many power series the radius of convergence can be calculated
according to the following theorem (see Section 10.12 for a proof).
THEOREM 10.2
6∞ n
The radius of convergence of a power series n=0 an x is given by
4 4
4 an 4
R = lim 44 4 or (10.24a)
n→∞ an+1 4
1
R = lim √ , (10.24b)
n→∞ n |an |
provided that either limit exists or is equal to infinity.
6∞
n
When we know that the radius of convergence of a power series n=0 an x is R , and
0 < R < ∞ , we know all values of x for which the series converges with the exception
of two values of x . The series converges for |x| < R , diverges for |x| > R , and may or
may not converge for x = ±R . Some series converge at both endpoints, some at neither,
and some at one end but not the other. There is no simple test that can distinguish among
these situations for all power series. Every power series must be checked individually as to
whether it converges at the endpoints of its interval of convergence. At x = ±R , a power series
reduces to a series of numbers. Sections 10.9–10.12 contain tests that determine whether series
of numbers converge or diverge, and therefore testing whether endpoints of power series with
finite radii of convergence can be included in the intervals of convergence will have to wait until
we have covered this material. In the meantime, we will call the interval of convergence the
open interval of convergence whenever we have not tested for inclusion of endpoints.
When Theorem 10.2 determines the radius of convergence of a power series, it also deter-
mines the open interval of convergence. We illustrate in the following examples.
EXAMPLE 10.12
∞
! (−1)n
Find the open interval of convergence for the power series xn .
n=1
n 52n
SOLUTION Since
4 4
4 (−1)n 4
4 4 4 4 ) *
4 an 4 4 4
R = lim 44 4 = lim 4 n52 n
4 = lim 25 n + 1 = 25,
n→∞ a 4 n→∞ 44 (−1)n+1 44 n→∞ n
n+1
4 2 2 4
(n + 1)5 n+
the open interval of convergence is −25 < x < 25. The series converges for these values of
x , diverges for x < −25 and x > 25, but we do not know whether it converges for x = −25
or x = 25.
EXAMPLE 10.13
!∞
(n!)2 n
Find the open interval of convergence for the power series x .
n=0
(2n)!
SOLUTION Since
(n!)2
4 4 7 8
4 an 4 (2n)! (n!)2 (2n + 2)(2n + 1)(2n)!
R = lim 44 4 = lim
4 = lim
n→∞ a n+1 n→∞ [(n + 1)!]2 n→∞ (2n)! (n + 1)2 (n!)2
(2 n + 2 )!
(2n + 2)(2n + 1)
= lim = 4,
n→∞ (n + 1)2
the open interval of convergence is −4 < x < 4.
EXAMPLE 10.14
!∞
1 n
Find the open interval of convergence for the power series x .
n=1
nn
SOLUTION Since
1 1
R = lim √ = lim √ = lim n = ∞,
n→∞ n |an | n→∞ n 1/nn n→∞
the series converges for all x . Its interval of convergence is −∞ < x < ∞ , which is not just
its open interval.
EXAMPLE 10.15
∞
! 1
Find the open interval of convergence for the power series x 2n+1 .
n=1
n2 2 n
SOLUTION Since coefficients of even powers of x are 0, the sequence {an /an+1 } is not
defined. We cannot therefore find its radius of convergence directly using Theorem 10.2. Instead,
we write
! ∞ !∞
1 2n+1 1 2 n
x =x (x )
n=1
n2 2 n n=1
n2 2 n
and set y = x 2 in the series:
!∞ !∞
1 2 n 1 n
2 2n
(x ) = 2 2n
y .
n=1
n n=1
n
According to equation 10.24a, the radius of convergence of this series in y is
1 ) *2
n2 2 n n+1
Ry = lim = lim 2 = 2.
n→∞ 1 n→∞ n
(n + 1)2 2n+1
√ √
Since x = ± y , it follows that the radius of convergence of the power series in x is Rx = 2.
√ √
The open interval of convergence is therefore − 2 < x < 2.
Sums of Power Series

Finding the interval of convergence of a power series is only half the convergence problem.
The other half is finding the sum of the power series. What function does the power series
approximate as more and more terms are included? Sometimes it is possible to relate a given
series to a series with known sum. Four very important series with known sums are
a
= a + ax + ax 2 + ax 3 + · · · , −1 < x < 1, (10.25a)
1−x
x2 x3
ex = 1 + x + + + ··· , −∞ < x < ∞, (10.25b)
2! 3!
x3 x5 x7
sin x = x − + − + ··· , −∞ < x < ∞, (10.25c)
3! 5! 7!
x2 x4 x6
cos x = 1 − + − + ··· , −∞ < x < ∞. (10.25d)
2! 4! 6!
EXAMPLE 10.16
∞
! (−1)n+1
Find the sum of the power series xn .
2n
n=0
SOLUTION We write
∞
! ∞
! ) *
(−1)n+1 n −x n
x = (−1)
2n 2
n=0 n=0
1 1 1
= −1 + x − x2 + x3 − · · · ,
2 22 23
and note that this is a geometric series with first term a = −1 and common ratio −x/2. If we
therefore replace x by −x/2 in equation 10.25a, we have
∞
! (−1)n+1 −1 −2
xn = = .
2n 1 − (−x/2) x+2
n=0
This is valid for −1 < −x/2 < 1, or, −2 < x < 2.
EXAMPLE 10.17
∞
! (−1)n 2n
Find the sum of the power series x 2n+2 .
n=0
(2n)!
SOLUTION The series can be expressed in the form
(−1)n #√ $2n
∞
! ∞
!
(−1)n 2n 2n+2 2
x =x 2x
n=0
(2n)! n=0
(2n)!
9 √ √ √ :
( 2 x)2 ( 2 x)4 ( 2 x)6
= x2 1 − + − + ··· .
2! 4! 6!
√
It is x 2 times the Maclaurin series for cos x with x replaced by 2 x . In other words,
∞
! (−1)n 2n √
x 2n+2 = x 2 cos 2 x,
n=0
(2n)!
√
valid for −∞ < 2 x < ∞ , or −∞ < x < ∞ .
Other methods for finding sums of power series are discussed in Section 10.6.
For many power series there is no known function to which the power series converges. In
such cases we write
∞
!
f (x) = an x n (10.26)
n=0
and say that the power series defines the value of f (x) at each x in the interval of convergence.
For instance, a very important series in engineering and physics is
!∞
(−1)n 2n
x ,
22n (n!)2
n=0
which converges for all x . It arises so often in applications that it is given a special name, the
Bessel function of the first kind of order zero, and is denoted by J0 (x) . In other words, we write
!∞
(−1)n 2n
J0 (x) = x ,
22n (n!)2
n=0
and say that the power series defines the value of J0 (x) for each x . We get an idea of what
J0 (x) looks like by plotting the polynomial approximations obtained by adding more and more
terms of the series. Figure 10.17 shows the first five polynomials for x ≥ 0:
P1 (x) = 1,
x2
P2 (x) = 1 − ,
22
x2 x4
P3 (x) = 1 − + ,
22 24 (2!)2
x2 x4 x6
P4 (x) = 1 − + − ,
22 24 (2!)2 26 (3!)2
x2 x4 x6 x8
P5 (x) = 1 − + − + .
22 24 (2!)2 26 (3!)2 28 (4!)2
They are even functions, so their graphs would be symmetric about the y -axis. As the
number of terms is increased, the polynomials approximate J0 (x) in Figure 10.18 more and
more closely.
We have called 10.19 a power series in x . It is also said to be a power series about 0 (meaning
the point 0 on the x -axis), where we note that for any interval of convergence whatsoever, 0 is
always at its centre. This suggests that power series about other points might be considered, and
this is indeed the case. The general power series about a point c on the x -axis is
∞
!
an (x − c)n = a0 + a1 (x − c) + a2 (x − c)2 + · · · (10.27)
n=0
and is said to be a power series in x − c .

FIGURE 10.17 Polynomial approximations for J0 (x) FIGURE 10.18 Bessel’s function J0 (x) of the first
kind of order zero
y P3 (x) P5 (x) y
1
3
0.8
2 0.6
0.4 y = J0 (x)
1 P1 (x)
0.2
1 2 3 4 5 6x
2 4 6 8 10 x
−1 −0.2
−2 −0.4
P2 (x) P4 (x)
A power series in x−c has an interval of convergence and a radius of convergence analogous
to a power series in x . In particular, every power series in x − c has a radius of convergence
R such that if R = 0, the power series converges only for x = c ; if R = ∞ , the power series
converges for all x ; and if 0 < R < ∞ , the series converges for |x − c| < R , diverges for
|x − c| > R , and may or may not converge for x = c ± R . The radius of convergence is
again given by equations 10.24, provided that the limits exist or are equal to infinity. For power
series in x − c , then, the point c is the centre of the interval of convergence.
EXAMPLE 10.18
Find the open interval of convergence of the power series
22 (x + 2) + 23 (x + 2)2 + 24 (x + 2)3 + · · · + 2n+1 (x + 2)n + · · · .
SOLUTION By writing the power series in sigma notation,
∞
!
2n+1 (x + 2)n ,
n=1
we can use equation 10.24a (or 10.24b) to calculate its radius of convergence:
4 4 n+1
4 an 4
R = lim 44 4 = lim 2 1
= .
n→∞ an+1 4 n→∞ 2 2
n+ 2
Since this is a power series about −2, the open interval of convergence is −5/2 < x < −3/2.
We can do better. This is a geometric series with first term 4(x + 2) and common ratio 2(x + 2) .
According to equation 10.25a, its sum is
∞
! 4(x + 2) −4(x + 2)
2n+1 (x + 2)n = = ,
1 − 2(x + 2) 2x + 3
n=1
valid for −1 < 2(x + 2) < 1 )⇒ −5/2 < x < −3/2. This is now known to be the
interval of convergence, not just the open interval of convergence.
EXERCISES 10.4
In Exercises 1–25 find the open interval of convergence for the power In Exercises 26–37 find the sum of the power series.
series.
∞ ∞ ∞
! ∞
!
! 1 n ! 1
1. x 2. nx2 n 26. x 3n 27. (−e)n x n
n n=0
4 n
n=1
n=1 n=1
∞
! ∞
!
1 ∞
! ∞
!
3. xn 4. n2 3n x n 1
(n + 1)3 28. 2n
(x − 1)n 29. (x + 5)2n
n=0 n=0
n=1
3 n=2
∞
! ∞
!
1
5. (x − 1)n 6. (−1)n n3 (x + 3)n ∞
! ∞
!
2n (−1)n 5n
n=0 n=4 ∗ 30. x 4n ∗ 31. xn
∞ ∞ ) *2 n=0
(2n)! n=0
n!
! 1 ! n−1
7. √ (x + 2)n 8. 2n (x − 4)n
n n+2 ∞ ∞
n=1 n=2 ! (−1)n ! (−3)n
2n+2
∞ ∞ ∗ 32. x ∗ 33. (x + 1)n
! 1 !
n=0
32n+1 (2n + 1)! n=0
n!
∗ 9. x 2n ∗ 10. (−1)n x 3n
n=1
n2 n=0
∞
! !∞
∞ ∞ (−1)n (−1)n+1
! n−1 ! 1 ∗ 34. xn ∗ 35. (x + 1)2n+3
∗ 11. (2x)n ∗ 12. √ x 3n+1 n=1
n! n=0
(2 n + 1)!
n=0
n+1 n=0
n+1
∞
! ∞
! ∞
! !∞
(−1)n (−e)n 2n (−1)n 4n+4
∗ 13. x 2n+1 ∗ 14. xn ∗ 36. (x − 1/2)n ∗ 37. x
3n n2 n=0
n! n=0
22n (2n)!
n=0 n=1
x 4 9 n2
∗ 15. + x2 + x3 + · · · + xn + · · ·
9 3 4 6
3 32n ∗ 38. If m is a nonnegative integer, the Bessel function of order m of
2 2 3 3 n n the first kind is defined by the power series
∗ 16. x + 2 x + 3 x + · · · + n x + · · ·
1 1 1
∗ 17. (x + 10)6 + (x + 10)7 + (x + 10)8 ∞
!
36 49 64 (−1)n
1 Jm (x) = x 2n+m .
+··· + n
(x + 10) + · · · n=0
22n+m n! (n
+ m)!
n2
∗ 18. 3x + 8(3x)2 + 27(3x)3 + · · · + n3 (3x)n + · · ·
3 27 3n (a) Write out the first five terms of J0 (x) , J1 (x) , and Jm (x) .
∗ 19. x2 + x4 + x6 + · · · + x 2n + · · ·
4 16 (n + 1)2
(b) Find the interval of convergence for each Jm (x) .
1 3 1 6 1 3n
∗ 20. 1 + x + x + ··· + x + ···
5 25 5n
!∞ ∗ 39. The hypergeometric series is
1 n
∗ 21. x
n=2
ln n
∞ αβ α(α + 1)β(β + 1) 2
! 1 1+ x+ x
∗ 22. xn γ 2! γ (γ + 1)
n=2
n2 ln n
α(α + 1)(α + 2)β(β + 1)(β + 2) 3
!∞ + x + ···,
(n!)3 n 3! γ (γ + 1)(γ + 2)
∗ 23. x
n=1
(3n)!
∞
! where α , β , and γ are all constants.
2 · 4 · 6 · . . . · (2n)
∗ 24. xn
3 · 5 · 7 · . . . · (2n + 1)
n=1 (a) Write this series in sigma notation.
∞
! 2
[1 · 3 · 5 · . . . · (2n + 1)]
∗ 25. xn (b) What is the radius of convergence of the hypergeometric
n=1
22n (2n)! series if γ is not zero or a negative integer?
10.5 Taylor Series Expansions of Functions

Let us summarize what we have seen in Sections 10.3 and 10.4. Given a function f (x) and a
point c , the Taylor series of f (x) about c is
∞
! f (n) (c)
f (x) = (x − c)n ,
n=0
n!
and the series converges

6∞ to f (x) at all values of x for which Taylor remainders approach zero.
Given a power series n=0 an (x − c)n , there is an interval of convergence inside of which the
series has a sum. If this sum is f (x) , we write
∞
!
f (x) = an (x − c)n , x in the interval of convergence.
n=0
Clearly, Taylor series are power series. A power series would be a Taylor series if an were
equal to f (n) (c)/n!. If this were the case, then power series and Taylor series would be one
and the same. The following theorem allows us to prove this.
THEOREM 10.3
6∞
If f (x) = n=0 an (x − c)n , and the radius of convergence R is greater than zero, then
each of the following series has radius of convergence R :
∞
!
f " (x) = nan (x − c)n−1 , (10.28a)
n=0
5 !∞
an
f (x) dx = (x − c)n+1 + C. (10.28b)
n + 1
n=0
Due to the difficulty in proving this theorem, and in order to preserve the continuity of our
discussion, we omit a proof. Note that the theorem is stated in terms of radii of convergence
rather than intervals of convergence. This is due to the fact that in differentiating a power series
we may lose the endpoints of the original interval of convergence, and in integrating we may pick
them up. It could be stated in terms of open intervals of convergence, however: term-by-term
differentiation and integration of power series preserve open intervals of convergence.
The next theorem implies that power series and Taylor series are one and the same, that
every power series is a Taylor series, the Taylor series of its sum.
THEOREM 10.4
6∞
If f (x) is the sum of the power series n=0 an (x − c)n with R > 0, then the series is
the Taylor series of f (x) .
PROOF When we set x = c in

∞
!
f (x) = an (x − c)n = a0 + a1 (x − c) + a2 (x − c)2 + · · · ,
n=0
we obtain f (c) = a0 . If we differentiate the power series according to Theorem 10.3, we

obtain
f " (x) = a1 + 2a2 (x − c) + 3a3 (x − c)2 + · · · .
When we substitute x = c , the result is
f " (c) = a1 .
If we differentiate the power series for f " (x) , we obtain
f "" (x) = 2a2 + 3 · 2a3 (x − c) + 4 · 3a4 (x − c)2 + · · · ,
and substitute x = c ,
f "" (c)
f "" (c) = 2a2 or a2 = .
2!
Continued differentiation and substitution leads to the result that for all n ,
f (n) (c)
an = .
n!
6∞
The power series f (x) = n=0 an (x − c)n is therefore the Taylor series of f (x) .
The following corollary is an immediate consequence of this theorem.
COROLLARY 10.4.1
6∞ 6∞
If two power series n=0 an (x − c)n and n=0 bn (x − c)n with positive radii of con-
vergence have identical sums,
∞
! ∞
!
an (x − c)n = bn (x − c)n ,
n=0 n=0
then an = bn for all n .
Theorem 10.4 shows that Sections 10.3 and 10.4 were dealing with the same problem but
coming at it from different directions. In Section 10.3, f (x) and x = c were given, and we
developed the Taylor series for f (x) about x = c . Theorem 10.4 shows 6∞that this is the only
power series for f (x) about x = c . In Section 10.4, a power series, n=0 an (x − c)n , was
given, and we determined its interval of convergence and sum f (x) . Theorem 10.4 indicates
that the power series is actually the Taylor series of f (x) about x = c .
This equivalence of power series and Taylor series simplifies the problem of finding the
Taylor series for a function f (x) about a point x = c immeasurably. Instead of finding f (n) (c)
and showing that Taylor remainders approach zero (as we did is Section6 10.3), we can proceed
∞
as follows. If, by any method whatsoever, we can find a power series n=0 an (x − c)n that
has sum f (x) , then it must be the Taylor series of f (x) about x = c . In the remainder of this
section we show how easy it is to do this. In essence, we take series with known sums, such as
10.25, and construct other series from them.
EXAMPLE 10.19
Find (a) the Maclaurin series for 1/(4 + 5x) and (b) the Taylor series about 5 for 1/(13 − 2x) .
SOLUTION
(a) We write
1 1 1/4
= ) * =
4 + 5x 5x 5x
4 1+ 1+
4 4
and interpret the right side as the sum of a geometric series with first term 1/4 and
common ratio −5x/4. Equation 10.25a then gives
∞ ) *) *n 4 4
1 ! 1 5x 4 5x 4
= − , 4− 4 < 1
4 + 5x 4 4 4 44
n=0
∞
! (−1)n 5n 4
= x n, |x| < .
4n+1 5
n=0
(b) By a similar procedure, we have
1 1 1 1/3
= = 7 8 =
13 − 2x 3 − 2(x − 5) 2 2
3 1− (x − 5) 1− (x − 5)
3 3
∞ 7 8n 4 4
! 2 42 4
= (1/3) (x − 5) , 4 (x − 5)4 < 1,
3 43 4
n=0
∞
! 2n 3
= (x − 5)n , |x − 5| < .
3n+1 2
n=0
In both examples, properties of geometric series gave not only the required series, but also
their intervals of convergence. To appreciate the simplicity of these solutions, we suggest using
Taylor remainders in an attempt to obtain the series with the same intervals of convergence. You
will quickly abort.
Addition and Subtraction of Power Series

According to the following theorem, convergent power series can be added and subtracted in
their common interval of convergence.
THEOREM 10.5
∞
! ∞
!
n
If f (x) = an (x − c) and g(x) = bn (x − c)n have positive radii of conver-
n=0 n=0
gence, then
∞
!
f (x) ± g(x) = (an ± bn )(x − c)n , (10.29)
n=0
valid for every x that is common to the intervals of convergence of the two series.
We use this result in the following example.
EXAMPLE 10.20
Find the Maclaurin series for f (x) = 5x/(x 2 − 3x − 4) .
SOLUTION We decompose f (x) into its partial fractions,

5x 4 1
f (x) = = + ,
x 2 − 3x − 4 x−4 x+1
and expand each of these terms in a Maclaurin series,
) *
4 −1 x x2
x−4
= x = − 1 + 4 + 42 + · · · , |x| < 4, and
1−
4
1
= 1 − x + x2 − x3 + · · · , |x| < 1.
1+x
Addition of these series within their common interval of convergence gives the Maclaurin series
for f (x) :
) *
5x x x2 x3 2 3
= −1 − − 2 − 3 − · · · + 1 − x + x 2 − x 3 + · · ·
x 2 − 3x − 4 4 4 4
) * ) * ) *
1 1 2 1
= −1 − x+ 1− x + −1 − x3 + · · ·
4 42 43
∞ 7
! 8
1
= (−1) − n x n ,
n
|x| < 1.
4
n=1
Differentiation and Integration of Power Series

Perhaps the most powerful technique for generating Taylor series is to differentiate or integrate
known expansions according to Theorem 10.3.
EXAMPLE 10.21
Find Maclaurin series for the following functions:
x
(a) cos x (b)
(2 − x)3
SOLUTION
(a) We derived the Maclaurin series for sin x in Example 10.9,
x3 x5 x7
sin x = x − + − + ···.
3! 5! 7!
We then stated that the Maclaurin series for cos x could be derived in a similar way.
Term-by-term differentiation of the sine series is faster,
) *
d d x3 x5 x7
cos x = (sin x) = x− + − + ···
dx dx 3! 5! 7!
x2 x4 x6
= 1− + − + ···
2! 4! 6!
∞
! (−1)n
= x 2n , −∞ < x < ∞.
n=0
(2n)!
(b) We begin with the Maclaurin series for 1/(2 − x) ,

7 #x $ # x $2 8
1 1 1
2−x
= # x$ = 2 1 + 2 + 2 + ···
2 1−
2
1 x x2 x3
= + + + + ···, |x| < 2.
2 22 23 24
Term-by-term differentiation of this series gives
1 1 2x 3x 2 4x 3
= + + + + ···,
(2 − x)2 22 23 24 25
with open interval of convergence |x| < 2. Another differentiation yields
2 2 3 · 2x 4 · 3x 2
= + + + ···, |x| < 2.
(2 − x)3 23 24 25
Multiplication by x/2 now gives
x 2x 3 · 2x 2 4 · 3x 3
= + + + ···
(2 − x)3 24 25 26
∞
! n(n + 1)
= x n, |x| < 2.
2n+3
n=1
EXAMPLE 10.22
Find the Taylor series about 1 for ln x .
SOLUTION Noting that ln x is an antiderivative of 1/x , we first expand 1/x in a Taylor
series about 1:
1 1
= = 1 − (x − 1) + (x − 1)2 − (x − 1)3 + · · · , |x − 1| < 1.
x (x − 1) + 1
If we integrate this series term by term, we have
7 8
1 2 1 3 1 4
ln |x| = x − (x − 1) + (x − 1) − (x − 1) + · · · + C.
2 3 4
Substitution of x = 1 implies that 0 = 1 + C ; that is, C = −1, and hence,

1 1
ln |x| = (x − 1) − (x − 1)2 + (x − 1)3 − · · ·
2 3
∞
! (−1)n+1
= (x − 1)n .
n=1
n
According to Theorem 10.3, the radius of convergence of this series is also R = 1; that is,
the open interval of convergence is 0 < x < 2. We can therefore delete the absolute values
around x .
Comparison of the solutions in Examples 10.11 and 10.22 indicates once again the advantage
of avoiding the use of Taylor’s remainder formula.
Multiplication and Division of Power Series

In Example 10.20 we added the Maclaurin series for 4/(x − 4) and 1/(1 + x) to obtain the
Maclaurin series for 5x/(x 2 − 3x − 4) . An alternative procedure might be to multiply the two
series since
) *) *
5x 1 1
= 5x
x 2 − 3x − 4 x−4 x+1
) *
5x x x2 x3 2 3
= 1 + + 2 + 3 + · · · 1 − x + x2 − x3 + · · · .
−4 4 4 4
The rules of algebra demand that we multiply every term of the first series by every term of the
second. If we do this and group all products with like powers of x , we obtain
7 ) * ) *
5x 5x 1 1 1
= 1+ −1 + x + 1 − + 2 x2
x 2 − 3x − 4 −4 4 4 4
) * 8
1 1 1 3
+ −1 + − + x + ··· .
4 42 43
It is clear that the coefficient of x n is a finite geometric series to which we can apply for-
mula 10.22:
7 8 9 2 3n+1 :
n 1 1 (−1)n n 1 − − 41
(−1) 1− + − ··· + = (−1) 1
4 42 4n 1+ 4
9 *n+1 : )
n4 1
= (−1) 1− − .
5 4
Consequently,
∞
9 ) *n+1 :
5x 5x ! n4 1
= (−1) 1− − xn
x 2 − 3x − 4 −4 n=0
5 4
∞
! 7 8
n+1 (−1)n+1
= (−1) 1− x n+1
4n+1
n=0
∞ 7
! 8
1
= (−1)n+1 − x n+1
4n+1
n=0
∞ 7
! 8
1
= (−1) − n x n .
n
4
n=1
For this example, then, multiplication as well as addition of power series leads to the Maclaurin
series. Clearly, addition of power series is much simpler for this example, but we have at least
demonstrated that power series can be multiplied together. That this is generally possible is
stated in the following theorem.
THEOREM 10.6
6∞ n
6∞
If f (x) = n=0 an (x − c) and g(x) = n=0 bn (x − c)n have positive radii of
convergence R1 and R2 , respectively, then
∞
!
f (x)g(x) = dn (x − c)n , (10.30a)
n=0
where
n
!
dn = ai bn−i = a0 bn + a1 bn−1 + · · · + an−1 b1 + an b0 , (10.30b)
i=0
and the radius of convergence is the smaller of R1 and R2 .
EXAMPLE 10.23
Find the Maclaurin series for f (x) = [1/(x − 1)] ln (1 − x) .
SOLUTION If we integrate the Maclaurin series
1
= 1 + x + x2 + x3 + · · · , |x| < 1,
1−x
we find that
) *
x2 x3 x4
− ln |1 − x| = x+ + + + · · · + C.
2 3 4
By setting x = 0, we obtain C = 0, and
x2 x3 x4
ln (1 − x) = −x − − − − ···.
2 3 4
We have dropped absolute value signs since the radius of convergence of the series is 1. We
now form the Maclaurin series for f (x) :
1 −1
ln (1 − x) = ln (1 − x)
x−1 1−x
) *
2 3 x2 x3
= 1 + x + x2 + x3 + · · · x + + + ···
2 3
) * ) *
1 1 1
= x+ 1+ x2 + 1 + + x3 + · · ·
2 2 3
∞ )
! *
1 1 1
= 1+ + + ··· + x n.
2 3 n
n=1
Since both of the multiplied series have radius of convergence 1, so also does the Maclaurin series
for (x − 1)−1 ln (1 − x) . In other words, its open interval of convergence is −1 < x < 1.
EXAMPLE 10.24
Find the first three nonzero terms in the Maclaurin series for tan x .
6∞ n
SOLUTION If tan x = n=0 an x , then by setting tan x = sin x/ cos x , we have
∞
!
sin x = cos x an x n .
n=0
We now substitute Maclaurin series for sin x and cos x :

) *
x3 x5 x7 x2 x4 x6 2 3
x− + − + ··· = 1− + − + · · · a 0 + a 1 x + a2 x 2 + · · · .
3! 5! 7! 2! 4! 6!
According to the corollary to Theorem 10.4, two power series can be identical only if corre-
sponding coefficients are equal. We therefore multiply the right side and equate coefficients of
like powers of x :
x0 : 0 = a0 ;
x : 1 = a1 ;
a0
x2 : 0 = a2 − , which implies a2 = 0;
2!
1 a1 1 1 1
x3 : − = a3 − , which implies a3 = − = ;
3! 2! 2! 3! 3
a2 a0
x4 : 0 = a4 − + , from which a4 = 0;
2! 4!
1 a3 a1 1 1 1 2
x5 : = a5 − + , from which a5 = + − = .
5! 2! 4! 5! 6 4! 15
The first three nonzero terms in the Maclaurin series for tan x are therefore
1 2
tan x = x + x3 + x5 + · · · .
3 15
We could obtain the same result by long division of the Maclaurin series for sin x by that of
cos x shown below. Long division can produce a few terms of a Maclaurin series, but seldom
does it suggest a pattern for all terms in the series.
x3 2x5
x + x3 + 2x5 + …
3 15
x2 x4 x3 x5
1 − x3 + 2x5 − … x − x3 + 2x5 − …
2 24 6 120
x3 x5
x − x3 + 2x5 − …
2 24
x3 x5
x + x3 − 2x5 + …
3 30
x3 x5
x + x3 − 2x5 + …
3 6
2x5
x + x3 + 2x5 + …
15
2x5
x + x3 + 2x5 + …
15
Binomial Expansion
One of the most widely used power series is the binomial expansion. We are well acquainted
with the binomial theorem, which predicts the product (a + b)m for any positive integer m :
m ) *
! m
(a + b)m = a n bm−n . (10.31)
n=0
n
With the usual definition of the binomial coefficients,

) *
m m! m(m − 1)(m − 2) · · · (m − n + 1)
= = ,
n (m − n)! n! n!
the binomial theorem becomes
m(m − 1)
(a + b)m = a m + ma m−1 b + a m−2 b2 + · · · + mabm−1 + bm . (10.32)
2!
Even when m is not a positive integer, this form for the binomial theorem remains almost intact.
To show this, we consider the power series
∞
! m(m − 1)(m − 2) · · · (m − n + 1)
1+ xn
n=1
n!
for any real number m except a nonnegative integer. The radius of convergence of this power
series is
4 4
4 m(m − 1)(m − 2) · · · (m − n + 1) (n + 1)! 4
4
R = lim 4 4
n→∞ n! m(m − 1)(m − 2) · · · (m − n) 4
4 4
4n+14
4
= lim 4 4 = 1.
n→∞ m − n 4
The open interval of convergence is therefore |x| < 1. Whether the series converges at the
end-points x = ±1 depends on the value of m . For the time being, we will work on the interval
|x| < 1, and at the end of the discussion, we will state the complete result. Let us denote the
sum of the series by
∞
! m(m − 1)(m − 2) · · · (m − n + 1)
f (x) = 1 + x n, |x| < 1.
n=1
n!
If we differentiate this series term by term according to Theorem 10.3,

∞
! m(m − 1) · · · (m − n + 1)
f " (x) = x n−1 , |x| < 1,
n=1
(n − 1)!
and then multiply both sides by x , we have

∞
! m(m − 1) · · · (m − n + 1)
xf " (x) = x n, |x| < 1.
n=1
(n − 1)!
If we add these results, we obtain

∞
! ∞
!
" " m(m − 1) · · · (m − n + 1) n−1 m(m − 1) · · · (m − n + 1)
f (x) + xf (x) = x + x n.
n=1
(n − 1)! n=1
(n − 1)!
We now change the variable of summation in the first sum:

∞
! ∞
!
" m(m − 1) · · · (m − n) n m(m − 1) · · · (m − n + 1)
(1 + x)f (x) = x + x n.
n=0
n! n=1
(n − 1)!
When these summations are added over their common range, beginning at n = 1, and the n = 0
term in the first summation is written out separately, the result is
∞
! ) *
" m(m − 1) · · · (m − n + 1) m − n
(1 + x)f (x) = m + + 1 xn
n=1
(n − 1)! n
9 ∞
:
! m(m − 1) · · · (m − n + 1)
= m 1+ xn
n=1
n!
= mf (x).
Consequently, the function f (x) must satisfy the differential equation
f " (x) m
= .
f (x) 1+x
Integration immediately gives
ln |f (x)| = m ln |1 + x| + C or f (x) = D(1 + x)m .
To evaluate D , we note that from the original definition of f (x) as the sum of the power series,
f (0) = 1, and this implies that D = 1. Thus,
f (x) = (1 + x)m ,
and we may write finally that

∞
!
m m(m − 1)(m − 2) · · · (m − n + 1)
(1 + x) = 1+ xn (10.33a)
n=1
n!
m(m − 1) m(m − 1)(m − 2)

= 1 + mx + x2 + x3 + · · · , (10.33b)
2! 3!
valid for |x| < 1. This is called the binomial expansion of (1 + x)m ; it is the Maclaurin
series for (1 + x)m . We have verified the binomial expansion for m any real number except a
nonnegative integer, but in the case of a nonnegative integer, the series terminates after m + 1
terms and is therefore valid for these values of m also. We mentioned earlier that the binomial
expansion may also converge at the endpoints x = ±1, depending on the value of m . The
complete result states that 10.33 is valid for
−∞ < x < ∞ if m is a nonnegative integer,

−1 < x < 1 if m ≤ −1,
−1 < x ≤ 1 if −1 < m < 0,
−1 ≤ x ≤ 1 if m > 0 but not an integer.
It is not difficult to generalize this result to expand (a + b)m for real m . If |b| < |a| , we
write ) * m
b
(a + b)m = a m 1 +
a
and now expand the bracketed term by means of 10.33:

9 ) * ) * :
b m(m − 1) b 2
(a + b)m = a m 1+m + + ··· |b| < |a|,
a 2! a
m(m − 1)
= a m + ma m−1 b + a m−2 b2 + · · · |b| < |a|, (10.34)
2!
which, as we predicted, is equation 10.32 except that the series does not terminate. We recom-
mend use of either of equations 10.33 over 10.34; this necessitates creation of the 1. We illustrate
in the following example.
EXAMPLE 10.25
Use the binomial expansion to find the Maclaurin series for x/(2 − x)3 .
SOLUTION We demonstrate how to use both of equations 10.33a and b. By 10.33a,
x x x # x $−3
= # x $ = 1 −
(2 − x)3 23 1 − 23 2
2
9 :
(−3)(−4)(−5) · · · (−3 − n + 1) # x $n
!∞
x
= 1+ −
23 n! 2
n=1
9 ∞
:
x ! (−1)n (3)(4)(5) · · · (n + 2) (−1)n x n
= 1+
23 n! 2n
n=1
9 ∞
:
x ! (2)(3)(4) · · · (n + 2) n
= 1+ x
23 2n+1 n!
n=1
9 ∞
:
x ! (n + 1)(n + 2) n
= 1+ x
23 n+1 2
n=1
∞
!
x (n + 1)(n + 2)
= + x n+1
23 2n+4
n=1
∞
!
x n(n + 1)
= + xn
23 2n+3
n=2
∞
! n(n + 1)
= x n, |x| < 2.
2n+3
n=1
With Equation 10.33b, we write
x x x # x $−3
= # x $ = 1 −
(2 − x)3 23 1 − 23 2
2
7 # x$ 8
x (−3)(−4) # x $2 (−3)(−4)(−5) # x $3
= 1 + (−3) − + − + − +···
23 2 2! 2 3! 2
7 8
x 3x 3·4 2 3·4·5 3 3·4·5·6 4
= 1+ + 2 x + 3 x + x + ···
23 2 2 2! 4 2 3! 2 4!
7 8
x 3x 2·3·4 2·3·4·5 2·3·4·5·6
= 1+ + x2 + x3 + x4 + · · ·
23 2 23 2! 24 3! 25 4!
7 8
x 3x 3·4 2 4·5 3 5·6 4
= 1+ + x + x + x + ···
23 2 23 24 25
∞
x ! (n + 1)(n + 2)
= xn
23 2n+1
n=0
∞
! (n + 1)(n + 2)
= x n+1
2n+4
n=0
∞
! n(n + 1)
= x n, |x| < 2.
2n+3
n=1
This result was also obtained in Example 10.21 by differentiation of the Maclaurin series for
1/(2 − x) .
EXAMPLE 10.26
Find the Maclaurin series for Sin −1 x .
SOLUTION By the binomial expansion, we have
) * 2 32 3 2 32 32 3
1 1 2
− 21 − 23 2 2
− 21 − 23 − 25
√ = 1+ − (−x ) + (−x ) + (−x 2 )3 + · · ·
1 − x2 2 2! 3!
1 3 3·5 3·5·7
= 1 + x2 + x4 + x6 + x8 + · · · , |x| < 1.
2 22 2! 23 3! 24 4!
Integration of this series gives
) *
1 3 3·5 3·5·7
Sin −1 x = x+ x3 + x5 + x7 + x 9 + · · · + C.
2·3 22 2! 5 23 3! 7 24 4! 9
Evaluation of both sides of this equation at x = 0 gives C = 0. According to Theorem 10.3,

the radius of convergence of this series must be 1, and we can write
!∞
1 · 3 · 5 · . . . · (2n − 1) 2n+1
Sin−1 x = x + x
2n n! (2n + 1)
n=1
!∞
1 · 2 · 3 · 4 · 5 · . . . · (2n − 2)(2n − 1)(2n) 2n+1
=x+ x
2 · 4 · . . . · (2n)2n n! (2n + 1)
n=1
∞
! (2n)!
=x+ x 2n+1
n=1
(2n + 1)22n (n!)2
∞
! (2n)!
= x 2n+1 , |x| < 1.
n=0
(2n + 1)22n (n!)2
EXAMPLE 10.27
When measuring the velocity v of gas flow using a Pitot tube, the following equation for pressure
P of the gas is encountered:
7 ) * 8
k − 1 # v $2 k/(k−1)
P (v) = P0 1 + ,
2 c
where P0 , c , and k are constants. The first two terms in the binomial expansion of P represent
the situation for an incompressible gas. The third term in the expansion is sometimes regarded
as the error in using the tube to determine velocity for compressible flow. Find the first three
terms in the binomial expansion.
SOLUTION Using equation 10.33,
9 ) * *) ) *) *) * :
k − 1 # v $2 1
k k k k − 1 2 # v $4
P (v) = P0 1 + + −1 + ···
2 k−1c 2 k−1 k−1 2 c
) *
k 2 k 4
= P0 1 + 2 v + 4 v + · · · .
2c 8c
EXERCISES 10.5
1
In Exercises 1–30 find the Maclaurin or Taylor series of the function ∗ 15. f (x) = √ about x = 0
about the indicated point. 1+x
∗ 16. f (x) = ln (1 + 2x) about x = 0
1
1. f (x) = about x = 0
3x + 2 ∗ 17. f (x) = (1 + 3x)3/2 about x = 0
1 ∗ 18. f (x) = ln x about x = 2
2. f (x) = about x = 0
4+ x2 ∗ 19. f (x) = ln (x + 3) about x = −1
2 ∗ 20. f (x) = 1/x about x = 4
3. f (x) = cos (x ) about x = 0
5x 1
4. f (x) = e about x = 0 ∗ 21. f (x) = about x = 0
(x + 2)3
5. f (x) = ex about x = 3
1
∗ 22. f (x) = about x = 3
6. f (x) = e1−2x about x = 0 (2 − x)2
7. f (x) = e1−2x about x = −1 1
∗ 23. f (x) = about x = 1
(x + 3)2
8. f (x) = cosh x about x = 0
1
9. f (x) = sinh x about x = 0 ∗ 24. f (x) = about x = 0
x 2 + 8x + 15
10. f (x) = x 4 + 3x 2 − 2x + 1 about x = 0 ∗ 25. f (x) = Tan −1 x about x = 0
4
11. f (x) = x + 3x − 2x + 1 2
about x = −2
√
∗ 26. f (x) = x + 3 about x = 0
1 √
∗ 12. f (x) = about x = 2 ∗ 27. f (x) = x + 3 about x = 2
x+3
x ∗ 28. f (x) = (1 − 2x)1/3 about x = 1
∗ 13. f (x) = about x = 1 2
2x + 5 x
∗ 29. f (x) = about x = 0
x 2 (1 + x 2 )2
∗ 14. f (x) = about x = 2
3 − 4x ∗ 30. f (x) = x(1 − x)1/3 about x = 0
x
In Exercises 31–33 find the first four nonzero terms in the Maclaurin ∗ 45. f (x) = ∗ 46. f (x) = xe−2x
series for the function. (4 + 3x)2
∗ 31. f (x) = tan 2x ∗ 32. f (x) = sec x

In Exercises 47–48 use Taylor series to find a formula for the nth deriva-
x
∗ 33. f (x) = e sin x tive of the function at x = 2.
∗ 34. Find the Maclaurin series for cos2 x . 1

∗ 47. f (x) = ∗ 48. f (x) = xe−x
3+x
In Exercises 35–38 find the Maclaurin series for the function.
∗ 49. Show that even-order derivatives of x 2 sin 2x at x = 0 are equal
1 to zero.
∗ 35. f (x) = ∗ 36. f (x) = Sin −1 (x 2 )
x 6 − 3x 3 − 4 2
9 ∗ 50. Show that odd-order derivatives of e−x at x = 0 vanish.
√ :
2x 2 + 4 1 + x/ 2
∗ 37. f (x) = ∗ 38. f (x) = ln √
x 2 + 4x + 3 1 − x/ 2 Show that Bessel functions of the first kind (defined in Exercise 38 of
Section 10.4) satisfy the properties in Exercises 51–52 .
∗ 39. Prove the corollary to Theorem 10.4.
∗ 51. 2mJm (x) − xJm−1 (x) = xJm+1 (x)
∗ 40. Prove 6
that if a power series with positive radius of convergence has
∞
sum zero, n=0 an (x − c)n = 0, then an = 0 for all n .
∗ 52. Jm−1 (x) − Jm+1 (x) = 2Jm" (x)
∗ 41. If, during a working day, one person drinks from a fountain every
30 s (on the average), then the probability that exactly n people drink in ∗ 53. If the function (1 − 2µx + x 2 )−1/2 is expanded in a Maclaurin
a time interval of length t seconds is given by the Poisson distribution: series in x ,
) *n
1 t ∞
!
Pn (t) = e−t/30 . 1
n! 30 1 = Pn (µ)x n ,
1 − 2µx + x 2 n=0
6∞
Calculate n=0 Pn (t) and interpret the result.
the coefficients Pn (µ) are called the Legendre polynomials. Find
∗ 42. A certain experiment is to be performed until it is successful. The P0 (µ) , P1 (µ) , P2 (µ) , and P3 (µ) .
probability that it will be successful in any given attempt is p (0 <
p < 1), and therefore the probability that it will fail is q = 1 − p . ∗ 54. (a) If we define f (x) = x/(ex − 1) at x = 0 as f (0) = 1,
The expected number of times that the experiment must be performed it turns out that f (x) has a Maclaurin series expansion
in order to be successful can be shown to be represented by the infinite with positive radius of convergence. When this expansion
series ∞ ∞ is expressed in the form
! !
npq n−1 = np(1 − p)n−1 .
x B2 2 B3 3
n=1 n=1 = 1 + B1 x + x + x + ···,
ex − 1 2! 3!
(a) What is the sum of this series?
(b) If p is the probability that a single die will come up 6, is the coefficients B1 , B2 , B3 , . . . are called the Bernoulli
the answer in part (a) what you would expect? numbers. Write this equation in the form
∗ 43. Find the Maclaurin series for the error function erf( x ) defined by ) *
x B2 2
5 x x = (e − 1) 1 + B1 x + x + ··· ,
2 2 2!
erf( x ) = √ e−t dt.
π 0
and substitute the Maclaurin series for ex to find the first
∗ 44. Find Maclaurin series for the Fresnel integrals C(x) and S(x) five Bernoulli numbers.
defined by (b) Show that the odd Bernoulli numbers all vanish for n ≥ 3.
5 x 5 x
C(x) = cos (π t 2 /2) dt, S(x) = sin (π t 2 /2) dt. ∗ 55. Show that
0 0 ∞
!
ex(t−1/t)/2 = Jn (x)t n .
n=0
In Exercises 45–46 use Maclaurin series to find a formula for the nth
derivative of the function at x = 0. For a definition of Jm (x) , see Exercise 38 in Section 10.4.
10.6 Sums of Power Series

Theorem 10.3 provides an important technique for finding sums of power series. If a series with
unknown sum can be reduced to a series with known sum by differentiations or integrations,
then the unknown sum can be obtained when these operations are reversed.
EXAMPLE 10.28
∞
!
Find the sum of the series (n + 1)x n .
n=0
SOLUTION Without the factor n+ 1, the series would be geometric. Integration of (n+ 1)x n
removes the factor. This is the idea; now let us formulate it mathematically. The radius of
convergence of the series is
4 4
4n + 14
R = lim 44 4 = 1.
n→∞ n + 2 4
If we denote the sum of the series by S(x) ,
∞
!
S(x) = (n + 1)x n ,
n=0
and use Theorem 10.3 to integrate the series term by term, we obtain
5 ∞
!
S(x) dx = x n+1 + C.
n=0
But the series on the right is a geometric series with sum x/(1 − x) , provided that |x| < 1,
and we may therefore write
5
x
S(x) dx = + C, |x| < 1.
1−x
If we now differentiate this equation, we obtain
(1 − x)(1) − x(−1) 1
S(x) = 2
= ,
(1 − x) (1 − x)2
and therefore
∞
! 1
(n + 1)x n = .
n=0
(1 − x)2
The
6∞open interval of convergence is −1 < x < 1. If we set x = 1, the series reduces to
n=0 (n + 1) = 1 + 2 + 3 + · · · , which clearly does not have a sum. At x = −1, we
have 1 − 2 + 3 − 4 + · · · , which diverges also. The interval of convergence is therefore
−1 < x < 1.
EXAMPLE 10.29
!∞
1 n
Find the sum of the series x .
n=1
n
SOLUTION We can remove the factor 1/n by differentiation, and thereby produce a geometric
series. We proceed as follows. The radius of convergence of the series is
4 4
4 1/n 4
R = lim 44 4 = 1.
n→∞ 1/(n + 1) 4
If we denote the sum of the series by
!∞
1 n
S(x) = x ,
n=1
n
and differentiate the series term by term, we have

∞
!
S " (x) = x n−1 .
n=1
This is a geometric series with sum 1/(1 − x) , so that
1
S " (x) = .
1−x
Integration now gives
S(x) = − ln (1 − x) + C.
Since S(0) = 0, it follows that
0 = − ln (1) + C.
Hence, C = 0, and S(x) = − ln (1 − x) . We have shown therefore that
!∞
1 n
x = − ln (1 − x).
n=1
n
The open interval of convergence is −1 < x < 1.
EXAMPLE 10.30
∞
! (−1)n (2n + 1)
Find the sum of the series x 2n .
n=0
(2n)!
SOLUTION Without the factor 2n + 1, we see a cosine series (note the x 2n and the (2n)!).
The factor can be eliminated by integration. We now know how to proceed. The radius of
convergence of the series is
4 4
4 (−1)n (2n + 1) 4
4 4 7 8
4 (2n)! 4 (2n + 1) (2n + 2)(2n + 1)(2n)!
4
R = lim 4 4 = lim ·
n→∞ 4 (−1)n+1 (2n + 3) 44 n→∞ (2n)! 2n + 3
4 4
(2 n + 2 )!
(2n + 1)2 (2n + 2)
= lim = ∞.
n→∞ 2n + 3
If we denote the sum of the series by S(x) ,

∞
! (−1)n (2n + 1)
S(x) = x 2n ,
n=0
(2n)!
and use Theorem 10.3 to integrate the series term by term, we obtain
5 ∞
! ∞
!
(−1)n 2n+1 (−1)n
S(x) dx = x +C = x x 2n + C = x cos x + C.
n=0
(2n)! n=0
(2n)!
Differentiation now gives

S(x) = −x sin x + cos x.
EXERCISES 10.6
∞
! ∞ )
! *
In Exercises 1–15 find the sum of the power series. n+1
∗ 9. n3n x 2n ∗ 10. xn
∞ ∞ n=2 n=0
n+2
! !
∗ 1. nx n−1 ∗ 2. n(n − 1)x n−2
∞ )
! * ∞
!
n=1 n=2 n+1 (−1)n (2n + 1)
∗ 11. xn ∗ 12. x 2n+1
∞
! ∞
! n=1
n! n=0
(2n + 1)!
∗ 3. (n + 1)x n−1 ∗ 4. n2 x n−1
n=1 n=1 ∞ ∞
! (−1)n (n + 2) ! (2n + 3)2n
∞
! ∞
! 1
∗ 13. x 2n ∗ 14. x 2n
∗ 5. (n2 + 2n)x n ∗ 6. xn n=0
(2n)! n=1
n!
n=1 n=0
n+1
!∞ ∞
! ∞
!
(−1)n 2n+1 (−1)n (−1)n+1 (2n − 1)
∗ 7. x ∗ 8. x 2n ∗ 15. x 2n+1
n=0
2n + 1 n=1
n n=0
(2n)!
10.7 Applications of Taylor Series and Taylor’s Remainder

Formula
When a function is approximated by its nth -degree Taylor polynomial,
f "" (c) f (n) (c)

f (x) ≈ f (c) + f " (c)(x − c) + (x − c)2 + · · · + (x − c)n ,
2! n!
the error in doing so is the Taylor remainder,
f (n+1) (zn )
Rn (c, x) = (x − c)n+1 ,
(n + 1)!
where zn is between c and x . The smaller Rn is, the better the approximation. Because zn is
unknown, we cannot find Rn (c, x) . What we do is replace f (n+1) (zn ) by some larger value,
thereby obtaining a maximum value for the error. For instance, consider using the first three
terms of the Maclaurin series for ex to approximate ex on the interval 0 ≤ x ≤ 1/2. Taylor’s
remainder formula, with c = 0, states that
x2
ex = 1 + x + + R2 (0, x),
2
where
d 3 x 44 x 3 ez x 3
R2 (0, x) = (e ) 4 = ,
dx 3 x=z 3! 6
and z is between 0 and x . Although z is unknown — except that it is between 0 and x , we can
say that because only the values 0 ≤ x ≤ 21 are under consideration, z must be less than 21 . It
follows that √
e1/2 x 3 e(1/2)3
R2 (0, x) < ≤ < 0.035.
6 6
Thus the quadratic function 1 + x + x 2 /2 approximates ex on the interval 0 ≤ x ≤ 1/2 with
error no greater than 0.035.
In the following example, we determine the number of terms of a Maclaurin series required
to guarantee a certain accuracy.
EXAMPLE 10.31
How many terms in the Maclaurin series for ln (1 + x) guarantee a truncation error of less than
10−6 for any x in the interval 0 ≤ x ≤ 1/2?
SOLUTION The nth derivative of ln (1 + x) is
dn (−1)n+1 (n − 1)!
ln ( 1 + x) = , n ≥ 1,
dx n (x + 1)n
and therefore Taylor’s remainder formula with c = 0 states that
x2 x3 x4 (−1)n+1 n
ln (1 + x) = x − + − + ··· + x + Rn (0, x),
2 3 4 n
where
d n+1 4 x n+1 (−1)n n! x n+1

4
Rn (0, x) = ln ( 1 + x)4 =
dx n+1 x=zn (n + 1)! (zn + 1)n+1 (n + 1)!
(−1)n
= x n+1 .
(n + 1)(zn + 1)n+1
Since zn is between 0 and x and 0 ≤ x ≤ 1/2, we can state that x must be less than or equal
to 1/2, and zn must be greater than 0. Hence,
1 1
|Rn (0, x)| < (1/2)n+1 = .
(n + 1)(1)n+1 (n + 1)2n+1
This is less than 10−6 if

1
< 10−6 or (n + 1)2n+1 > 106 .
(n + 1) 2n+1
A calculator quickly indicates that the smallest value of n for which this is true is n = 15.
Consequently, if ln (1 + x) is approximated by the 15th -degree polynomial
x2 x3 x4 x 15
ln (1 + x) ≈ x − + − + ··· +
2 3 4 15
on the interval 0 ≤ x ≤ 1/2, the truncation error is less than 10−6 .

Two young electrical engineers are having a disagreement and we must settle the argument.
Figure 10.19 shows two charges of equal size q , but of opposite signs, distance d apart.
When d is small, the configuration is called a dipole. One engineer argues that when
point P is very far away from the dipole, so that r is very much larger than d , the charges
effectively cancel one another, and the potential due to them is zero. The other engineer
disagrees. We must decide which of them is correct.
FIGURE 10.19 Potential SOLUTION The potential at P due to charge q is given by the formula V = q/(4π &0 r1 ) ,
due to a dipole where &0 is a constant. Similarly, the potential at P due to charge −q is V = −q/(4π &0 r2 ) .
P The potential at P due to both charges is
r2 r1
) *
r 1 q q
V = − .
−q θ q 4 π &0 r1 r2
Q R S
d/2 d/2 The cosine law applied to triangles P RS and P QR gives
d2
r12 = r 2 + − rd cos θ,
4
d2 d2
r22 = r 2 + − rd cos (π − θ ) = r 2 + + rd cos θ.
4 4
Hence,
; <
q 1 1
V = 1 −1 .
4 π &0 r 2 + d 2 /4 − rd cos θ r 2 + d 2 /4 + rd cos θ
The binomial expansion can be used to write
1 1
1 = "
r 2 + d 2 /4 + rd cos θ d2 d
r 1 + 2 + cos θ
4r r
7 ) 2 *8−1/2
1 d d
= 1+ + cos θ
r 4r 2 r
7 ) 2 *
1 1 d d
= 1− + cos θ
r 2 4r 2 r
) *2 :
(−1/2)(−3/2) d 2 d
+ + cos θ + · · · .
2! 4r 2 r
When d is very much less than r , so that terms in d 2 /r 2 , d 3 /r 3 , etc., are negligible
compared to d/r ,
) *
1 1 d
1 ≈ 1− cos θ .
r 2 + d 2 /4 + rd cos θ r 2r
) *
1 1 d
Similarly, 1 ≈ 1+ cos θ . Hence,
2 2
r + d /4 − rd cos θ r 2r
7 ) * ) *8
q 1 d 1 d qd cos θ
V ≈ 1+ cos θ − 1− cos θ = .
4 π &0 r 2r r 2r 4 π &0 r 2
This shows that the potential due to the dipole does not vanish at large distances from it.
It is small because d is small and r is large, but it is not zero. Only when P is on the
perpendicular bisector of the line joining the charges (so that θ = π/2) is the potential
zero.
Limits
We have customarily used L’Hôpital’s rule to evaluate limits of the indeterminate form 0/0.
Maclaurin and Taylor series can sometimes be used to advantage. Consider
x − sin x
lim .
x→0 x3
Three applications of L’Hôpital’s rule give a limit of 1/6. Alternatively, if we substitute the
Maclaurin series for sin x ,
7 ) *8
x − sin x 1 x3 x5
lim = lim 3 x − x − + − ···
x→0 x3 x→0 x 3! 5!
7 8
1 x2 1
= lim − + ··· = .
x→0 6 5! 6
Here is another example.
EXAMPLE 10.32
λ− 5
Evaluate lim , where c > 0 is a constant (see also Exercise 58 in Section 4.11).
λ→0+ ec/λ − 1
SOLUTION We begin by making the change of variable v = 1/λ in the limit:
λ− 5 v5
lim = lim .
λ→0+ ec/λ − 1 v→∞ ecv − 1
We now expand ecv into its Maclaurin series,
λ− 5 v5
lim = lim ) *
λ→0+ ec/λ − 1 v→∞ c2 v 2
1 + cv + + ··· − 1
2!
v5
= lim 2 2
.
v→∞ cv
cv + + ···
2!
If we now divide numerator and denominator by v 5 , we obtain
λ− 5 1
lim = lim 2 3
= 0.
λ→0+ e −1
c/λ v→∞ c c c c4 c5 c6 v
+ 3 + + + + + ···
v4 2v 3! v 2 4! v 5! 6!
Evaluation of Definite Integrals

In Section 8.8 we developed three numerical techniques for approximating definite integrals
of functions f (x) that have no obvious antiderivatives: the rectangular rule, the trapezoidal
rule, and Simpson’s rule. Each method divides the interval of integration into a number of
subintervals and approximates f (x) by a more elementary function on each subinterval. The
rectangular rule replaces f (x) by a step function, the trapezoidal rule uses a succession of linear
functions, and Simpson’s rule uses quadratic functions.
Another possibility is to replace f (x) by a truncated power series (a polynomial) over the
entire interval of integration. For instance, consider the definite integral
5 1/2
sin x
dx,
0 x
where (sin x)/x is defined as 1 at x = 0. The integral defines the area in Figure 10.20.
FIGURE 10.20 Area under the curve

y = x −1 sin x
y
1
sin x
0.8 y=
x
0.6
0.4
0.2
0 0.5 1 1.5 2 x
Taylor’s remainder formula for sin x gives
x3 x5 dn 4 xn
4
sin x = x − + − ··· + ( sin x) 4 + Rn (0, x)
3! 5! dx n x=0 n!
where
d n+1 (sin x) 44 x n+1
Rn (0, x) = 4
dx n+1 x=zn (n + 1)!
and zn is between 0 and x . Therefore,
sin x x2 x4 1
= 1− + − ··· + Rn (0, x),
x 3! 5! x
dn 4 x n−1
4
where the term before x −1 Rn (0, x) is ( sin x) 4 . When we take definite integrals,
dx n x=0 n!
5 1/2 5 1/2 7 8
sin x x2 x4 1
dx = 1− + − · · · + Rn (0, x) dx
0 x 0 3! 5! x
' ( 1/2 5 1/2
x3 x5 1
= x− + − ··· + Rn (0, x) dx,
3 · 3! 5 · 5! 0 0 x
5 1/2
1 (1/2)3 (1/2)5 1
= − + − ··· + Rn (0, x) dx.
2 3 · 3! 5 · 5! 0 x
Now 45 4 5 1/2 4 n+1 4

4 1/2
1 4 4 1 d (sin x) 44 x n+1 44
4 4
Rn (0, x) dx 4 ≤ 4
4x 4 x dx n+1 4x=zn (n + 1)! 4 dx.
0 0
4 n+1 4
4 d (sin x) 44 4
4
Since 4 4 4 ≤ 1, it follows that
dx n+ 1 x=zn 4
45 1/2
4 5 1/2 ' (1/2
4
4 1 4 xn x n+1 (1/2)n+1
4 Rn (0, x) dx 44 ≤ dx = = .
0 x 0 (n + 1)! (n + 1)(n + 1)! 0 (n + 1)(n + 1)!
Thus, if we write
5 1/2
sin x 1 (1/2)3 (1/2)5
dx ≈ − + = 0.493 107 639,
0 x 2 3 · 3! 5 · 5!
then the maximum error is

(1/2)6 (1/2)7
or ,
6 · 6! 7 · 7!
depending on whether we regard x − x 3 /3! + x 5 /5! as a fifth- or sixth-degree approximation
for sin x . Since (1/2)7 /(7 · 7!) < 0.000 000 222 gives a smaller error, we can say that
5 1/2
sin x
0.493 107 639 − 0.000 000 222 < dx < 0.493 107 639 + 0.000 000 222;
0 x
that is,
5 1/2
sin x
0.493 107 417 < dx < 0.493 107 861.
0 x
Consequently, using only three terms of the Maclaurin series for (sin x)/x , we can say that to
five decimal places
5 1/2
sin x
dx = 0.493 11.
0 x
An easier analysis is given in Example 10.54 of Section 10.13. It uses alternating series
instead of Taylor’s remainder formula.
Differential Equations
Many differential equations arising in physics and engineering have solutions that can be ex-
pressed only in terms of infinite series. One such equation is Bessel’s differential equation of
order zero for a function y = f (x) :
xy "" + y " + xy = 0.
Before considering this somewhat difficult differential equation, we introduce the ideas through
an easier example.
EXAMPLE 10.33
Determine whether the differential equation
dy
− 2y = x
dx
6∞ n
has a solution that can be expressed as a power series y = n=0 an x with positive radius of
convergence.
SOLUTION If
∞
!
y = f (x) = a n x n = a 0 + a1 x + a2 x 2 + · · ·
n=0
is to be a solution of the differential equation, we may substitute the power series into the
differential equation:
2 3 2 3
a1 + 2a2 x + 3a3 x 2 + 4a4 x 3 + · · · − 2 a0 + a1 x + a2 x 2 + · · · = x.
We now gather together like terms in the various powers of x :
0 = (a1 − 2a0 ) + (2a2 − 2a1 − 1)x + (3a3 − 2a2 )x 2 + (4a4 − 2a3 )x 3 + · · · .
Since the power series on the right has sum zero, its coefficients must all vanish (see Exercise
40 in Section 10.5), and therefore we must set
a1 − 2 a0 = 0 ,
2 a2 − 2 a1 − 1 = 0 ,
3a3 − 2a2 = 0,
4 a4 − 2 a3 = 0 ,
and so on. These equations imply that
a1 = 2 a0 ;
1 1
a2 = (1 + 2 a1 ) = (1 + 4a0 );
2 2
2 2
a3 = a2 = (1 + 4a0 );
3 3!
2 22
a4 = a3 = (1 + 4a0 ).
4 4!
The pattern emerging is
2n−2
an = (1 + 4a0 ), n ≥ 2.
n!
Thus,
1 2n−2
f (x) = a0 + 2a0 x + (1 + 4a0 )x 2 + · · · + (1 + 4a0 )x n + · · ·
2 n!
) *
1 22 23 2n
= a0 + 2 a0 x + (1 + 4 a0 ) x2 + x3 + · · · + xn + · · · .
4 2! 3! n!
We can find the sum of the series in parentheses by noting that the Maclaurin series for e2x is
(2x)2 (2x)3
e2x = 1 + (2x) + + + ···.
2! 3!
Therefore, the solution of the differential equation is
1 = >
y = f (x) = a0 + 2a0 x + (1 + 4a0 ) e2x − 1 − 2x
4
1 x 1 1 x
=− − + (1 + 4a0 )e2x = Ce2x − − .
4 2 4 4 2
Using power series to solve the differential equation in Example 10.33 is certainly not the most
expedient method. A far simpler method will be discussed in Section 15.3. But the example
clearly illustrated the procedure by which power series are used to solve differential equations.
We now apply the procedure to Bessel’s differential equation of order zero.
EXAMPLE 10.34
6∞ n
Find a power series solution y = n=0 an x , with positive radius of convergence, for Bessel’s
differential equation of order zero,
xy "" + y " + xy = 0.
SOLUTION In this example we abandon the · · · notation

6∞ of Example 10.33, and maintain
sigma notation throughout. When we substitute y = n=0 an x n into the differential equation,
we obtain
∞
! ∞
! ∞
!
0 = x n(n − 1)an x n−2 + nan x n−1 + x an x n
n=2 n=1 n=0
∞
! ∞
! ∞
!
= n(n − 1)an x n−1 + nan x n−1 + an x n+1 .
n=2 n=1 n=0
In order to bring these three summations together as one, and combine terms in like powers of
x , we lower the index of summation in the last term by 2:
∞
! ∞
! ∞
!
0 = n(n − 1)an x n−1 + nan x n−1 + an−2 x n−1 .
n=2 n=1 n=2
We now combine the three summations over their common interval, beginning at n = 2, and
write separately the n = 1 term in the second summation,
∞
!
0 = a1 + [n(n − 1)an + nan + an−2 ]x n−1 .
n=2
But the only way a power series can be equal to zero is for all of its coefficients to be equal to
zero; that is,
a1 = 0 ; n(n − 1)an + nan + an−2 = 0, n ≥ 2.
Thus,
an−2
an = − , n ≥ 2,
n2
a recursive relation defining the unknown coefficient an of x n in terms of the coefficient an−2
of x n−2 . Since a1 = 0, it follows that
0 = a1 = a3 = a5 = · · · .
a0
For n = 2, a2 = − .
22
a2 a0 a0
For n = 4, a4 = − = = .
42 2 2 42 2 (2!)2
4
a4 −a0 a0
For n = 6, a6 = − = =− .
62 2 (2!)2 62
4 2 (3!)2
6
The solution is therefore

a0 a0 a0
y = a0 − x2 + x4 − x6 + ···
22 2 (2!)2
4 2 (3!)2
6
!∞
(−1)n 2n
= a0 x .
22n (n!)2
n=0
The function defined by the infinite series
!∞
(−1)n 2n
J0 (x) = x , −∞ < x < ∞,
22n (n!)2
n=0
is called the zero-order Bessel function of the first kind.
EXERCISES 10.7
5 1 5 0 .3
In Exercises 1–10 find a maximum possible error in using the given ∗ 13. x 11 sin x dx ∗ 14.
2
e−x dx
terms of the Taylor series to approximate the function on the interval −1 0
specified.
x2 x3 In Exercises 15–20 use series to evaluate the limit.

∗ 1. ex ≈ 1 + x + + for 0 ≤ x ≤ 0.01
2 6
tan x 1 − cos x
x2 x3 ∗ 15. lim ∗ 16. lim
∗ 2. e ≈ 1 + x +
x
+ for 0 ≤ x < 0.01
x→0 x x→0 x2
2 6 √
(1 − cos x)2 1+x−1
x 2
x 3 ∗ 17. lim ∗ 18. lim
∗ 3. e ≈ 1 + x +
x
+ for −0.01 ≤ x ≤ 0 x→0 3x 4 x→0 x
2 6 ) *
1
x2 x3 ∗ 19. lim x sin
∗ 4. ex ≈ 1 + x + + for |x| ≤ 0.01 x→∞ x
2 6
) x *
e + e−x 1
x3 ∗ 20. lim −
∗ 5. sin x ≈ x − for 0 ≤ x ≤ 1 x→0 ex − e−x x
3!
x2 x4
∗ 6. cos x ≈ 1 − + for |x| ≤ 0.1 In Exercises 21–24 determine where the Maclaurin series for the func-
2! 4!
tion may be truncated in order to guarantee the accuracy indicated.
x2 x3
∗ 7. ln (1 − x) ≈ −x − − 0 ≤ x ≤ 0.01 ∗ 21. sin (x/3) on |x| ≤ 4 with error less than 10−3
2 3
√
1 ∗ 22. 1/ 1 + x 3 on 0 < x < 1/2 with error less than 10−4
∗ 8. ≈ 1 + 3x + 6x 2 + 10x 3 for |x| < 0.2
(1 − x)3
∗ 23. ln (1 − x) on |x| < 1/3 with error less than 10−2
9x 3 81x 5
∗ 9. sin 3x ≈ 3x − + for |x| < π/100 ∗ 24. cos2 x on |x| < 0.1 with error less than 10−3
2 40
1 1 1
∗ 10. ln x ≈ (x − 1) − (x − 1)2 + (x − 1)3 − (x − 1)4
2 3 4 In Exercises 25–30 find a series solution in powers of x for the differ-
for |x − 1| ≤ 1/2
ential equation.
In Exercises 11–14 evaluate the integral correct to three decimal places.

∗ 25. y " + 3y = 4 ∗ 26. y "" + y " = 0
5 1 5 1/ 2 ∗ 27. xy " − 4y = 3x ∗ 28. 4xy "" + 2y " + y = 0
sin x 2
∗ 11. dx ∗ 12. cos (x ) dx
0 x 0 ∗ 29. y "" + y = 0 ∗ 30. xy "" + y = 0
∗ 31. Find the natural logarithm of 0.999 999 999 9 accurate to 15 deci- Show that when (k − 1)M02 /2 < 1, Ps can be expressed in the form
mal places. 7 ) * 8
∗ 32. In special relativity theory, the kinetic energy K of an object mov- 1 M2 2−k
Ps = P0 + ρ0 V02 1 + 0 + M04 + · · · ,
ing with speed v is defined by 2 4 24
K = c2 (m − m0 ), where ρ0 = kP0 /c02 .
where c is a constant (the speed of light), m0 is the rest mass of the ∗ 35. The ellipse b2 x 2 +a 2 y 2 = a 2 b2 can be represented parametrically
object, and m is its mass when moving with speed v . The masses m by
and m0 are related by x = a cos t, y = b sin t, 0 ≤ t < 2π.
m0 (a) Show that the length of the circumference of the ellipse is

m= 1 . defined by the definite integral
1 − v 2 /c2
5 π/2 1 a2
Use the binomial expansion to show that L = 4b 1 − k 2 sin2 t dt, k2 = 1 − .
0 b2
) *
1 3 v4 5 v6
K = m0 v 2 + m0 c2 + + ··· , (b) Use the binomial expansion to show that
2 8 c4 16 c6
) *
and hence, to a first approximation, kinetic energy is defined by the k2 3k 4
L = 2π b 1 − − − ···
classical expression m0 v 2 /2. 4 64
∗ 33. Stagnation pressure P0 and pressure P are related by the Mach so that to a first approximation, L is the circumference of
number M in incompressible flow of an ideal gas by the equation a circle of radius b .
P0 kM 2 ∗ 36. The well function for leaky aquifers is defined by the convergent
= 1+ , improper integral
P 2
5 ∞
where k is a constant. For compressible flow, the relation is 1 2 /(4αx)]
W (α, β) = e−[αx+β dx.
7 ) * 8k/(k−1) 1 x
P0 k−1
= 1+ M2 . 2
(a) Show that if e−β /(4αx) is replaced by a Maclaurin series
P 2
in β 2 /(4αx) , and integration is done term-by-term,
Assuming that ) *
k−1 ∞
! (−1)n β 2n
M 2 < 1, W (α, β) = En+1 (α),
2 4n α n n!
n=0
expand the latter relation to show that for small M , P0 /P in compress-
ible flow can be approximated by P0 /P in incompressible flow. where En (α) is the exponential integral
∗ 34. The figure below shows uniform, two-dimensional, compressible, 5 ∞
e−αx
adiabatic flow of a frictionless fluid around a circular object. The pres- En (α) = dx.
sure is P0 and the velocity is V0 in the undisturbed flow to the left of 1 xn
the object.
(b) Show that the exponential integrals satisfy the recursion
relation
1
V0 En+1 (α) = [e−α − αEn (α)].
n
S
P0 ∗ 37. Planck’s law for the energy density ) of blackbody radiation of
wavelength λ states that
8π chλ−5
At the stagnation point S , the velocity of the fluid is zero, and we )(λ) = ,
ech/(λkT ) − 1
let Ps be the stagnation pressure. It is known that
7 ) * 8k/(k−1) where h > 0 is Planck’s constant, c is the (constant) speed of light,
Ps k−1 and T is temperature, also assumed constant. Show that for long wave-
= 1+ M02 ,
P0 2 lengths, Planck’s law reduces to the Rayleigh–Jeans law:
where M0 = V0 /c0 is the Mach number of the flow ( c0 is the velocity of 8π kT

)(λ) = .
pressure propagation in the undisturbed flow), and k > 1 is a constant. λ4
∗ 38. In the figure below, charges of q > 0 and −q coulombs are a (b) Use the binomial expansion to show that when d is very
distance d apart. The configuration is called an electric dipole. much less than x , E can be approximated by
(a) The electric field at point P due to these charges is
qd
E = .
q q 2π &0 x 3
E = − ,
4π &0 (x − d/2)2 4π &0 (x + d/2)2
∗ 39. Liquid flows in the semicircular flume in the figure below.
where &0 is a constant. Verify that E can be expressed in (a) Find an expression for the ratio of the cross-sectional area
the form A of the flow to the product of h and d .
9) *− 2 ) *− 2 : (b) Use the result in part (a) to find an approximation for the
q d d ratio that includes terms of order θ 2 .
E = 1− − 1+ .
4π &0 x 2 2x 2x
R Semicircular
flume
d d θ
2 2
h
A
−q q P
x d
10.8 Convergence of Sequences of Numbers
In Section 10.1 we introduced the basic ideas of convergence for sequences of numbers. Ex-
amples were so simple that we had no difficulty determining whether sequences converged or
diverged. As this is not always the case, we discuss two important ways to show that a se-
quence converges. Our concern is particularly with recursively defined sequences, since it is
usually straightforward to determine whether an explicit sequence is convergent or divergent.
For instance, consider the sequence
√
c1 = 1, cn+1 = 5 + 2 + cn , n ≥ 1,
of Example 10.2 in Section 10.1. Its next four terms are
c2 = 6.732, c3 = 7.955, c4 = 8.155, c5 = 8.187.
We suspect that the sequence has a limit. Why? Because terms are rapidly getting closer and
closer to each other, or to put it another way, differences between terms are rapidly approaching
zero.
√ But this does not guarantee convergence. Differences between terms of the sequences
{ n} and {ln n} get smaller and smaller as n increases, but neither sequence has a limit.
The following two definitions lead to Theorem 10.7, which can be very useful in verifying
convergence of sequences like this.
DEFINITION 10.2
A sequence {cn } is said to be
(i) increasing if cn+1 > cn for all n ≥ 1; (10.35a)

(ii) nondecreasing if cn+1 ≥ cn for all n ≥ 1; (10.35b)
(iii) decreasing if cn+1 < cn for all n ≥ 1; (10.35c)
(iv) nonincreasing if cn+1 ≤ cn for all n ≥ 1. (10.35d)
If a sequence satisfies any one of these four properties, it is said to be monotonic.
DEFINITION 10.3
A sequence {cn } is said to have an upper bound U (be bounded above by U ) if
cn ≤ U (10.36a)
for all n ≥ 1. It has a lower bound V (is bounded below by V ) if
cn ≥ V (10.36b)
for all n ≥ 1. If a sequence has both an upper bound and a lower bound, it is said to be a
bounded sequence.
Note that if U is an upper bound for a sequence, then any number greater than U is also
an upper bound. If V is a lower bound, so too is any number smaller than V .
Let us illustrate these definitions with some simple explicit sequences before stating our
first convergence theorem and applying it to the above recursive sequence. The sequence
' (
1 1 1 1
= 1, , , , ...
2n−1 2 4 8
is decreasing, has an upper bound U = 1, and a lower bound V = 0. The sequence

' (
n 1 2 3 4
= , , , , ...
n+1 2 3 4 5
is increasing, has an upper bound U = 5, and a lower bound V = −2. The sequence
{(−1)n+1 } = 1, −1, 1, −1, . . .
is not monotonic, has an upper bound U = 1, and a lower bound V = −3.

√
The above recursive sequence c1 = 1, cn+1 = 5 + 2 + cn appears to be increasing,
V = 0 is obviously a lower bound, and U = 10 appears to be an upper bound. The reason
why we would verify these conjectures is contained in the following theorem.
THEOREM 10.7
A bounded, monotonic sequence has a limit.
FIGURE 10.21a Graph of increasing and bounded sequence FIGURE 10.21b Graph of decreasing and bounded sequence
cn cn
U 0.5
0.9
0.4
0.8
0.3
0.7 0.2
V
0.6
0.1
5 10 15 20 n
0 5 10 15 20 n
To expand on this statement somewhat, consider a sequence {cn } whose terms are illustrated
graphically in Figure 10.21a. Suppose that the sequence is increasing and therefore monotonic,
and that U is an upper bound for the sequence. We have shown the upper bound as a horizontal
line in the figure; c1 is a lower bound. Our intuition suggests that because the terms in the
sequence always increase, and they never exceed U, the sequence must have a limit. Theorem
10.7 confirms this. The theorem does not suggest the value of the limit, but obviously it must
be less than or equal to U .
Similarly, when a sequence is decreasing or nonincreasing and has a lower bound V (Fig-
ure 10.21b), it must approach a limit that is greater than or equal to V .
Another way of stating Theorem 10.7 is as follows.
COROLLARY 10.7.1
A monotonic sequence has a limit if and only if it is bounded.
We are now prepared

√ to give a complete and typical discussion for the recursive sequence
c1 = 1, cn+1 = 5 + 2 + cn .
EXAMPLE 10.35
Verify that the sequence
√
c1 = 1, cn+1 = 5 + 2 + cn , n ≥ 1,
has a limit, and find it.

SOLUTION The first five terms of the sequence are
c1 = 1, c2 = 6.732, c3 = 7.955, c4 = 8.155, c5 = 8.187.
They suggest that the sequence is increasing; that is, cn+1 > cn . To prove this we use mathe-
matical induction (see Appendix A). Certainly, the inequality is valid for n = 1 since c2 > c1 .
Suppose that k is an integer for which ck+1 > ck . Then
2 + ck+1 > 2 + ck ,
from which
1 √
2 + ck+1 > 2 + ck .
It follows that 1 √
5+ 2 + ck+1 > 5 + 2 + ck .
The left side is ck+2 and the right side is ck+1 . Therefore, we have proved that ck+2 > ck+1 .
Hence, by mathematical induction cn+1 > cn for all n ≥ 1. Since the sequence is increasing,
its first term c1 = 1 must be a lower bound. Certainly, any upper bound, if one exists, must be
at least 8.187 ( c5 ). We can take any number greater than 8.187 and use mathematical induction
to test whether it is indeed an upper bound. It appears that U = 10 might be a reasonable
guess for an upper bound for this sequence, and we verify this by induction as follows. Clearly,
c1 < 10. We suppose that k is some integer for which ck < 10. Then
√ √ √
ck+1 = 5 + 2 + ck < 5 + 2 + 10 = 5 + 12 < 10.
By mathematical induction, cn < 10 for n ≥ 1.

Since the sequence is monotonic and bounded, Theorem 10.7 guarantees that it has a limit,
call it L . To evaluate L , we take limits on each side of the equation defining the sequence
recursively:
2 √ 3
lim cn+1 = lim 5 + 2 + cn .
n→∞ n→∞
It is important to note that this cannot be done until the conditions of Theorem 10.7 have√ been
checked. Since terms
√ c n of the sequence approach L as n → ∞ , it follows that 5 + 2 + cn
approaches 5 + 2 + L . Furthermore, as n → ∞ , cn+1 must also approach L . Do not make
the mistake of saying that cn+1 approaches L + 1 as n → ∞ . Think about what lim n→∞ cn+1
means. We conclude therefore that
√
L = 5+ 2 + L.
If we transpose the 5 and square both sides of the equation, we obtain the quadratic equation
L2 − 11L + 23 = 0,
with solutions √
11 ± 29
L= .
2
√
Only the positive√square root satisfies the original√equation L = 5 + 2 + L defining L , so
that L = (11 + 29)/2. The other root, (11 − 29)/2 ≈ 2.8, can also be eliminated on the
grounds that all terms beyond the first are greater than 6.
Sequences do not have to be monotonic to be convergent. Convergence can occur for other
reasons. The following example illustrates a second common way for sequences to converge.
Again we have chosen a recursive sequence as illustration because for an explicit sequence, the
limit is usually obvious. Consider the recursive sequence
1
c1 = 2 , cn+1 = 2 + , n ≥ 1.
cn
The first six terms of the sequence are
c1 = 2 , c2 = 2.5, c3 = 2 .4 , c4 = 2.417, c5 = 2.4138, c6 = 2.414 29.
The terms seem to be clustering around a number close to 2.414, one larger, one smaller, one
larger, and so on (Figure 10.22). Unfortunately, lack of monotony precludes the possibility of
using Theorem 10.7 to discuss convergence. In Theorem 10.8, we discuss properties that imply
convergence for sequences of this type, but it is helpful first to illustrate these properties with a
specific example such as the one above. What are the properties that lead us to believe that this
FIGURE 10.22 Graph of an oscillating sequence
cn
3
2.5
2
1.5
1
0.5
0 n
1 2 3 4 5 6 7
sequence converges? First, the terms are not monotonic; they are “up-down-up-down-up-down.”
How do we say this mathematically? Differences between successive terms in the sequence are
c2 − c1 = 2.5 − 2 = 0.5,
c3 − c2 = 2.4 − 2.5 = −0.1,
c4 − c3 = 2.417 − 2.4 = 0.017,
c5 − c4 = 2.4138 − 2.417 = −0.0032,
c6 − c5 = 2.414 29 − 2.4138 = 0.000 49.
The fact that these differences are alternately positive and negative implies that the terms in
the sequence {cn } are up-down-up-down-up-down. This is not enough to guarantee conver-
gence, however. For example, the sequence 1, −1, 1, −1, 1, −1, . . . is up-down-up-down-
up-down, but it does not converge. The added feature of the sequence above is that absolute
values of the differences
|c2 − c1 | = 0.5, |c3 − c2 | = 0.1, |c4 − c3 | = 0.017,
|c5 − c4 | = 0.0032, |c6 − c5 | = 0.000 49
seem to form a decreasing sequence with limit zero. We have said “seem” because we have
not proved that these properties hold for all differences, only the first five. We shall provide
a general verification after stating the next theorem. However, it is the up-down-up-down-
up-down nature of the sequence together with the fact that absolute values of the differences
decrease and approach zero that lead us to believe that the sequence {cn } has a limit. This is
formalized in the following theorem.
THEOREM 10.8
Suppose a sequence {cn } has the following properties:
1. Differences cn+1 − cn alternate in sign.
2. Absolute values of these differences |cn+1 − cn | are decreasing.
3. Absolute values of the differences |cn+1 − cn | approach 0.
Then the sequence {cn } converges, and its limit lies between any two successive terms in
the sequence.
A sequence that satisfies condition 1 of this theorem is said to be an oscillating sequence.

Its terms are up-down-up-down-up-down. When terms also satisfy conditions 2 and 3, they
oscillate about a limit — one below, one above, one below, one above, and so on, but gradually
they get closer and closer to the limit.
We now verify that the sequence c1 = 2, cn+1 = 2 + 1/cn on the previous page is
oscillating and convergent. The difference between the (n + 1)th and nth terms is
) * ) *
1 1 −(cn − cn−1 )
cn+1 − cn = 2+ − 2+ = .
cn cn−1 cn cn−1
Since all terms in the sequence are clearly positive, the denominator of this expression is positive.
It follows that the difference cn+1 − cn must have the sign opposite to the previous difference
cn − cn−1 ; that is, the differences cn+1 − cn alternate in sign. Furthermore, since 2 is a lower
bound for all terms in the sequence, it follows that for n ≥ 2,
|cn − cn−1 | |cn − cn−1 | |cn − cn−1 |

|cn+1 − cn | = < = .
cn cn−1 (2)(2) 4
But if each difference is less than one-quarter the previous difference, the differences must be
decreasing and have limit zero. By Theorem 10.8 this sequence has a limit L that we obtain by
taking limits in the recursive definition:
) *
1 1
lim cn+1 = lim 2+ )⇒ L = 2+ .
n→∞ n→∞ cn L
√ √
Of the two solutions 1 ± 2 of this equation, only L = 1 + 2 lies between c1 and c2 .
Notice that we did not use mathematical induction to verify that the sequence above is oscil-
lating and convergent. This is characteristic of oscillating sequences; mathematical induction is
not required to verify the three properties of Theorem 10.8. It may be necessary to use induction
to prove ancilliary results (such as a lower bound for the sequence), but induction is not needed
to verify the properties of Theorem 10.8.
An oscillating sequence results when the method of successive approximations is applied
to the equation f (x) = x 3 + 25x − 50 = 0 for the root between x = 1 and x = 2. With
50 − xn3
x1 = 1, xn+1 = , n ≥ 1,
25
the next 15 terms are
x2 = 1.96, x3 = 1.70, x4 = 1.80,
x5 = 1.77, x6 = 1.7782, x7 = 1.7751,
x8 = 1.7763, x9 = 1.775 81, x10 = 1.776 00,
x11 = 1.775 93, x12 = 1.775 954, x13 = 1.775 945,
x14 = 1.775 948 2, x15 = 1.775 946 9, x16 = 1.775 947 4.
The terms are indeed oscillating but appear to be approaching a limit. Were we given this
recursive sequence without reference to the equation x 3 + 25x − 50 = 0, and asked to discuss
its convergence, we would verify the conditions of Theorem 10.8. In the context of solving the
equation, however, we would not do this. We have what we feel is a potential candidate for
the solution of the equation x 3 + 25x − 50 = 0 between x = 1 and x = 2. To verify that
1.775 947 is the solution correctly rounded to six decimal places, we use the zero intermediate
value theorem with the calculations
f (1.775 946 5) = −2.7 × 10−5 , f (1.775 947 5) = 7.3 × 10−6 .
Now that we know what it means for a sequence to have a limit, and how to find limits, we
can be more precise. To give a mathematical definition for the limit of a sequence, we start with
our intuitive description and make a succession of paraphrases, each of which is one step closer
to a precise definition:
A sequence {cn } has limit L if terms get arbitrarily close to L , and stay close to L , as n
gets larger and larger.
A sequence {cn } has limit L if terms can be made arbitrarily close to L by choosing n
sufficiently large.
A sequence {cn } has limit L if differences |cn − L| can be made arbitrarily close to 0 by
choosing n sufficiently large.
A sequence {cn } has limit L if given any real number & > 0, no matter how small, we can
make differences |cn − L| less than & by choosing n sufficiently large.
Finally, we arrive at the following definition.
DEFINITION 10.4
A sequence {cn } has limit L if for any given & > 0, there exists an integer N such that
for all n > N ,
|cn − L| < &.
This definition puts our intuitive idea of a limit in precise terms. For those who have
studied Section 2.6, note the similarity between Definition 10.4 and Definition 2.2. For a better
understanding of Definition 10.4, it is helpful to consider its geometric interpretation. The
inequality |cn − L| < & , when written in the form L − & < cn < L + & , is interpreted as a
horizontal band of width 2& around L (Figure 10.23a). Definition 10.4 requires that no matter
how small & , we can find a stage, denoted by N , beyond which all terms in the sequence are
contained in the horizontal band. For the sequence and & in Figure 10.23a, N must be chosen
as shown. For the same sequence, but a smaller & , N must be chosen correspondingly larger
(Figure 10.23b). Proofs of some results in the rest of this chapter require a working knowledge
of this definition. As an example, we use it to verify that a sequence cannot have two limits, a
fact that we have implicitly assumed throughout our discussions.

Illustration of a limit of a sequence
cn cn
L L
all terms in band all terms in band
4 N 8 12 16 20 n 4 8N 12 16 20 n
THEOREM 10.9
A sequence can have at most one limit.
PROOF We prove this by showing that a sequence cannot have two distinct limits. Suppose
to the contrary that a sequence {cn } has two distinct limits L1 and L2 , where L2 > L1 , and
let L2 − L1 = δ . If we set & = δ/3, then according to Definition 10.4, there exists an integer
N1 such that for all n > N1 ,
|cn − L1 | < & = δ/3;
that is, for n > N1 , all terms in the sequence are within a distance δ/3 of L1 .
But since L2 is also supposed to be a limit, there exists an N2 such that for n > N2 , all
terms in the sequence are within a distance & = δ/3 of L2 :
|cn − L2 | < & = δ/3.
FIGURE 10.24 Illustration of proof that a sequence cannot have two limits
cn
All terms
L2 in here
All terms
L1 in here
N1 N2 n
But this is impossible (Figure 10.24) if L1 and L2 are a distance δ apart. This contradiction
therefore implies that L1 and L2 are the same; that is, {cn } cannot have two limits.
The following theorem, which states some of the properties of convergent sequences, can
also be proved using Definition 10.4. Only the first two parts, however, are straightforward (see
Exercises 58, 66, and 67).
THEOREM 10.10
If sequences {cn } and {dn } have limits C and D , then:
(i) {kcn } has limit kC if k is a constant.

(ii) {cn ± dn } has limit C ± D.
(iii) {cn dn } has limit CD.
(iv) {cn /dn } has limit C/D provided that D -= 0, and none of the dn = 0.
EXERCISES 10.8
In Exercises 1–24 determine whether the statement is true or false. 9. A sequence can be both nonincreasing and nondecreasing.
Verify any statement that is true, and give a counterexample to any
statement that is false. 10. If a sequence is monotonic and has a limit, it must be bounded.
1. A sequence can be increasing and nondecreasing. 11. If a sequence is bounded and has a limit, it must be monotonic.
2. A monotonic sequence must be bounded.
12. If a sequence is bounded, but not monotonic, it cannot have a limit.
3. An increasing sequence must have a lower bound.
4. A decreasing sequence must have a lower bound. 13. If a sequence has a limit, it must be bounded.
5. An increasing sequence with a lower bound must have a limit. 14. If a sequence is not monotonic, it cannot have a limit.
6. An increasing sequence with an upper bound must have a limit.
15. If all terms of a sequence {cn } are less than U , and L =
7. If a sequence diverges, then either limn→∞ cn = ∞ or limn→∞ cn limn→∞ cn exists, then L must be less than U .
= −∞ .
8. A sequence cannot be both increasing and decreasing. 16. A sequence {cn } has a limit if and only if {cn2 } has a limit.
17. An oscillating sequence must converge. ∗ 43. (a) Prove that if a sequence converges, then differences be-
18. If terms of a sequence are alternately positive and negative, the tween successive terms in the sequence must approach zero.
sequence is oscillating. (b) The following example illustrates that the converse is not
∗ 19. A sequence {cn } of positive numbers converges if cn+1 < cn /2. true. Show that differences between successive terms of
the sequence {ln n} approach zero, but the sequence itself
∗ 20. A sequence {cn } of numbers converges if cn+1 < cn /2.
diverges.
∗ 21. If an increasing sequence has a limit L , then L must be equal to
the smallest upper bound for the sequence.
∗ 22. If an infinite number of terms of a sequence all have the same value In Exercises 44–47 show that the sequence is convergent and find its
a , then either a is the limit of the sequence or the sequence has no limit. limit.
∗ 23. If absolute values of differences between terms of an oscillating 8 − cn
sequence decrease, the sequence converges.
∗ 44. c1 = 2, cn+1 = , n≥1
5
∗ 24. If terms of an oscillating sequence approach zero, then terms must 3
be alternately positive and negative. ∗ 45. c1 = 20, cn+1 = 12 + , n≥1
cn
1
∗ 46. c1 = 1, cn+1 = , n≥1
In Exercises 25–41 discuss, with proofs, whether the sequence is 2 + cn
monotonic, and whether it has an upper bound, a lower bound, and 1
a limit. ∗ 47. c1 = 10, cn+1 = , n≥1
3 + 2cn
1 2 3 ∗ 48. Show that the sequence
∗ 25. c1 = 1, cn+1 = cn3 + 12 , n≥1
10 3
1 2 3 c1 = 1, cn+1 = , n ≥ 1,
∗ 26. c1 = 0, cn+1 = 4
cn + 5 , n≥1 1 + cn
12
√ is convergent and find its limit. Hint: You may find it helpful to prove
∗ 27. c1 = 3, cn+1 = 5 + cn , n≥1 that 1 ≤ cn ≤ 2 for n ≥ 1. Do this by mathematical induction.
√
∗ 28. c1 = 1, cn+1 = 5 + cn , n≥1 ∗ 49. Repeat Exercise 48 if c1 = 0. In this case you might want to show
√ that 1 ≤ cn ≤ 2 for n ≥ 4.
∗ 29. c1 = 5, cn+1 = 1 + 6 + cn , n≥1
√ ∗ 50. Show that the sequence
∗ 30. c1 = 3, cn+1 = 1 + 6 + cn , n≥1
√ 4
∗ 31. c1 = 1, cn+1 = 4 − 5 − cn , n≥1 c1 = 1, cn+1 = , n ≥ 1,
√ 2 + 5cn
∗ 32. c1 = 4, cn+1 = 4 − 5 − cn , n≥1 is convergent and find its limit. Hint: You may find it helpful to prove
1 that 1/2 ≤ cn ≤ 1 for n ≥ 1. Do this by mathematical induction.
∗ 33. c1 = 2, cn+1 = , n≥1
3 − cn ∗ 51. Show that the sequence
1 √
∗ 34. c1 = 1, cn+1 = , n≥1 c1 = 1, cn+1 = 26 − cn , n ≥ 1,
4 − 2cn
7 is convergent and find its limit. Hint: Show that 4 ≤ cn ≤ 5 for
∗ 35. c1 = 1, cn+1 = , n≥1 n ≥ 2, and then consider (cn+1 )2 − (cn )2 .
16 − 8c2 n
∗ 52. Show that the sequence
7 √
∗ 36. c1 = 0, cn+1 = , n≥1 c1 = 4, cn+1 = 20 − 3cn , n ≥ 1,
16 − 8cn2
4cn is convergent and find its limit.
∗ 37. c1 = 1, cn+1 = , n≥1
4 + cn ∗ 53. Show that the sequence
3cn cn
∗ 38. c1 = 4, cn+1 = , n≥1 c1 = 2, cn+1 = , n ≥ 1,
2 + cn 4cn − 1
2cn2 does not converge.
∗ 39. c1 = 2, cn+1 = , n≥1
3 + cn ∗ 54. Show that the sequence in Exercise 53 diverges for any first term
3 cn + 2 except c1 = 0 and c1 = 1/2.
∗ 40. c1 = , cn+1 = , n≥1 ∗ 55. Show that the sequence
2 4 − cn
3 − cn 5cn
∗ 41. c1 = 0, cn+1 = , n≥1 c1 = 1, cn+1 = , n ≥ 1,
5 − 2cn 2cn − 1
∗ 42. Show that the sequence is convergent and find its limit. Hint: Express cn+1 in the form
1 cn+1 = 5/(2 − 1/cn ) and show that 2 ≤ cn ≤ 4 for n ≥ 3.
c1 = 1, cn+1 = , n≥1 ∗ 56. Show that the sequence
4 − cn − cn2
√
3
is monotonic and bounded. Find an approximation to its limit accurate c1 = 3, cn+1 = 10 − cn , n ≥ 1,
to five decimal places. is convergent and find its limit.
∗ 57. Show that the sequence ∗ 64. A particular breed of rabbits grows and reproduces according to the
following schedule. Each female rabbit becomes an adult and produces
cn a pair of babies (one male and one female) at the age of 2 months and
c1 = a, cn+1 = , n ≥ 1,
bcn − 1 produces exactly one pair every month thereafter. Suppose we begin
with one pair of adult rabbits who produce a pair of babies at the end
where a and b are nonzero constants diverges unless ab = 2.
of the first month. If no rabbits die, find a sequence {Rn } representing
∗ 58. Prove parts (i) and (ii) of Theorem 10.10. the number of adult female rabbits after n months. Do you recognize
∗ 59. Find bounds for the sequence it?
1 + cn ∗ 65. Prove that the sequence
c1 = 1, cn+1 = , n ≥ 1.
1 + 2cn 1
c1 = d, cn+1 = a + 2bcn , n≥1
∗ 60. Determine whether the following sequence is monotonic, has
bounds, and has a limit: ( a , b√
, and d all positive constants) is increasing
√ if and only if d <
cn cn−1 b + a + b2 . What happens when d = b + a + b2 ?
c1 = −30, c2 = −20, cn+1 = 5 + + , n ≥ 2.
2 3 ∗∗ 66. In this exercise we outline a proof of part (iii) of Theorem 10.10.
∗ 61. A sequence {cn } has only positive terms. Prove that: (a) Verify that
(a) if limn→∞ cn = L < 1, there exists an integer N such
that for all n ≥ N , |cn dn − CD| ≤ |cn ||dn − D| + |D||cn − C|.
L+1 (b) Show that given any & > 0, there exist positive integers
cn < ;
2 N1 , N2 , and N3 such that
(b) if limn→∞ cn = L > 1, there exists an integer N such |cn | < |C| + 1 whenever n > N1 ,
that for all n ≥ N ,
&
L+1 |dn − D| < whenever n > N2 ,
cn > . 2 (|C| + 1)
2
&
|cn − C| < whenever n > N3 .
These results are used in Theorem 10.16, Section 10.11. 2|D| + 1
∗ 62. Prove that if limn→∞ cn = L , there exists an integer N such that
for all n ≥ N , (c) Use these results to prove part (iii) of Theorem 10.10.
cn < L + 1.
∗∗ 67. In this exercise we outline a proof for part (iv) of Theorem 10.10.
This result is used in Theorem 10.15, Section 10.10.
∗ 63. The Fibonacci sequence found in many areas of applied mathe- (a) Verify that when dn -= 0 and D -= 0,
matics is defined by 4 4
4 cn 4
4 − C 4 ≤ |cn − C| + |C||dn − D| .
c1 = 1, c2 = 1, cn+1 = cn + cn−1 , n ≥ 2. 4d D4 |dn | |D||dn |
n
(a) Evaluate the first 10 terms of this sequence. (b) Show that given any & > 0, there exist positive integers
(b) Is the sequence monotonic, is it bounded, and does it have N1 , N2 , and N3 such that
a limit?
(c) Prove that |D|
|dn | > whenever n > N1 ,
2
cn2 − cn−1 cn+1 = (−1)n+1 , n ≥ 2.
&|D|
|cn − C| < whenever n > N2 ,
(d) Verify that an explicit formula for cn is 4
9; √ <n ; √ <n : &|D|2
1 1+ 5 1− 5 |dn − D| < whenever n > N3 .
cn = √ − . 4|C| + 1
5 2 2
(c) Now prove part (iv) of Theorem 10.10.
(e) Define a sequence {bn } as the ratio of terms in the Fibonacci
sequence ∗∗ 68. Find an explicit formula for the recursive sequence
c n+1
bn = .
cn cn + cn−1
c1 = 1, c2 = 2 , cn+1 = , n ≥ 2.
Is this sequence monotonic? Does it have a limit? 2
10.9 Infinite Series of Numbers

Infinite series of numbers arise in two ways. When {cn } is an infinite sequence of numbers, the
expression
∞
!
c n = c 1 + c2 + c3 + · · · + c n + · · · (10.37)
n=1
is called an infinite series of numbers, or simply a series. From the sequences of Example
10.1, we may form the following series of numbers.
EXAMPLE 10.36
Write out the first six terms of the following series:
∞
! ∞
! ∞
! ∞
!
1 n
(a) (b) (c) (−1)n |n − 3| (d) (−1)n+1
2n−1 n+1
n=1 n=1 n=1 n=1
SOLUTION The first six terms of these series are shown below.
∞
! 1 1 1 1 1 1
(a) = 1+ + + + + + ···
2n−1 2 4 8 16 32
n=1
∞
! n 1 2 3 4 5 6
(b) = + + + + + + ···
n=1
n+1 2 3 4 5 6 7
∞
!
(c) (−1)n |n − 3| = −2 + 1 − 0 + 1 − 2 + 3 − · · ·
n=1
∞
!
(d) (−1)n+1 = 1 − 1 + 1 − 1 + 1 − 1 + · · ·
n=1
Infinite series of numbers also arise when specific values of x are substituted
6∞ into power or
Taylor series. For example, if we substitute x = 1 into the power series n=1 x n /(n2n ) , we
obtain
!∞
1 1 1 1 1
= + + + ··· + + ···;
n 2n 2 2 · 22 3 · 23 n 2n
n=1
if we set x = −2, we obtain
∞
! (−2)n 1 1 1 (−1)n
= −1 + − + − ··· + + ···.
n=1
n 2n 2 3 4 n
As mentioned earlier in this chapter, it is not possible to add an infinity of numbers in a finite
amount of time, and therefore expression 10.37 is as yet meaningless. To attach a meaning we
take the same approach as we did for Taylor series and power series. We illustrate with a simple
example, and then give a formal definition. Consider the infinite series of Example 10.36(a),
∞
! 1 1 1 1
= 1+ + + + ···.
2n−1 2 4 8
n=1
If we start adding terms, a pattern soon emerges. Indeed, if we denote by Sn the sum of the first
n terms of this series, we find that
S1 = 1,
1 3
S2 = 1 + = , (sum of first two terms),
2 2
1 1 7
S3 = 1 + + = , (sum of first three terms),
2 4 4
1 1 1 15
S4 = 1 + + + = , (sum of first four terms),
2 4 8 8
1 1 1 1 31
S5 = 1 + + + + = , (sum of first five terms),
2 4 8 16 16
and so on. We see that the sum of the first n terms of the series is given by the formula
2n − 1 1
Sn = = 2− .
2n−1 2n−1
As we add more and more terms of this series together, the sum Sn gets closer and closer to
2. It is always less than 2, but Sn can be made arbitrarily close to 2 by choosing n sufficiently
large. If this series is to have a sum, the only6
reasonable sum is 2. In practice, this is precisely
∞
what we do; we define the sum of the series n=1 1/2n−1 to be 2.
Let us now take this idea and define sums for general infinite series. We begin by defining
a sequence {Sn } as follows:
S 1 = c1 ,
S 2 = c1 + c2 ,
S 3 = c1 + c2 + c3 ,
.. ..
. .
S n = c1 + c2 + c3 + · · · + c n ,
.. ..
. .
6∞
called the sequence of partial sums of the series n=1 cn . The nth term of the sequence {Sn }
represents the sum of the first n terms of the series. If this sequence has a limit, say S , then the
more terms of the series that we add together, the closer the sum gets to S . It seems reasonable,
then, to call S the sum of the series. We therefore make the following definition.
DEFINITION 10.5
6n 6∞
Let Sn = k=1 ck be the nth partial sum of a series n=1 cn . If the sequence of partial
sums {Sn } has limit S ,
S = lim Sn ,
n→∞
we call S the sum of the series and write
∞
!
cn = S;
n=1
if {Sn } does not have a limit, we say that the series does not have a sum.
If a series has sum S , we say that the series converges to S , which means that its sequence
of partial sums converges to S . If a series does not have a sum, we say that the series diverges,
which means that its sequence of partial sums diverges. Partial sums are to infinite series of
numbers what Taylor polynomials are to Taylor series. Before proceeding with examples we
need to make one comment about terminology and one about notation. In the next few sections
we will be concentrating on series of numbers as opposed to power series or series of functions.
We shall usually write series instead of series of numbers; the context will always make it clear
when we are discussing series of numbers. Whenever the first term of a series corresponds to
n = 1, we shall drop the limits n =6 1 and ∞ on the sigma
6 notation when the notation appears
in text. For instance, we shall write cn in place of ∞ n=1 cn . Limits will always be retained
when sigma notation appears in a displayed equation,
6 or when the lower limit is not n = 1.
n−1
According to Definition
6 10.5, the series 1/ 2 of Example 10.36(a) has sum 2. Since
every term of the series n/(n + 1) in Example 10.36(b) is greater than or equal to 1/2, it
follows that the sum of the first n terms is Sn ≥ n(1/2) = n/2. As the sequence of partial
sums is therefore unbounded, it cannot possibly have a limit (see the corollary to Theorem 10.7).
The series does6 not therefore have a sum; it diverges. Examination of the first few partial sums
of the series (−1)n |n − 3| in Example 10.36(c) leads to the result that for n > 1,

 n−4
 , if n is even,
Sn = 2
 1 − n , if n is odd.

2
Since the sequence {Sn } does not have a limit, the series diverges. The partial sums of the series
6
(−1)n+1 are
S1 = 1, S2 = 0, S3 = 1, S4 = 0, . . . .
Since this sequence does not have a limit, the series does not have a sum.
Perhaps the most important series of numbers, and certainly the series that occurs most
frequently in applications, is the geometric series. We discussed it in the context of power
series in Section 10.4, but it is worthwhile repeating the ideas here, unencumbered by other
considerations. Geometric series are of the form
∞
!
ar n−1 = a + ar + ar 2 + ar 3 + · · · . (10.38)
n=1
Each term is obtained by multiplying the preceding term by the same constant r , the common
ratio. If {Sn } is the sequence of partial sums for this series, then
Sn = a + ar + ar 2 + · · · + ar n−1 .
If we multiply this equation by r ,
rSn = ar + ar 2 + ar 3 + · · · + ar n .
When we subtract these equations, the result is
Sn − rSn = a − ar n .
Hence, for r -= 1, we obtain

a(1 − r n )
Sn = .
1−r
Furthermore, when r = 1,
Sn = na,
and therefore the sum of the first n terms of a geometric series is

- a(1 − r n )
, r -= 1
Sn = 1−r (10.39a)
na, r = 1.
To determine whether a geometric series has a sum, we consider the limit of this sequence.
Certainly, limn→∞ na does not exist, unless trivially a = 0. In addition, limn→∞ r n = 0
when |r| < 1, and does not exist when |r| > 1. Nor does it exist when r = −1. Thus, we
may state that
- a
, |r| < 1
lim Sn = 1−r
n→∞
does not exist, |r| ≥ 1.
The geometric series therefore has the sum
∞
! a
ar n−1 = (10.39b)
1−r
n=1
if |r| < 1, but otherwise diverges.

The series in parts (a) and (d) of Example 10.36 are geometric. The first has common
ratio 1/2 and therefore converges to 1/(1 − 1/2) = 2; the second has common ratio −1 and
therefore diverges.
We encounter the harmonic series
!∞
1 1 1 1
= 1 + + + + ··· (10.40)
n 2 3 4
n=1
quite often in our work. It does not have a sum, and we can show this by considering the
following partial sums of the series:
S1 = 1,
1 3
S2 = 1 + = ,
2 2
1 1 3 1 1 4
S 4 = S2 + + > + + = ,
3 4 2 4 4 2
1 1 1 1 4 1 1 1 1 5
S 8 = S4 + + + + > + + + + = ,
5 6 7 8 2 8 8 8 8 2
1 1 1 5 1 1 1 6
S16 = S8 + + + ··· + > + + + ··· + = .
9 10 16 2 16 16 16 2
This procedure can be continued indefinitely and shows that the sequence of partial sums is
unbounded. The harmonic series therefore diverges (see once again the corollary to Theorem
10.7).
In each of the examples above, we used Definition 10.5 for the sum of a series to determine
whether the series converges or diverges; that is, we formed the sequence of partial sums {Sn }
in order to consider its limit. For most examples, it is either too difficult or impossible to
evaluate Sn in a simple form, and in such cases consideration of the limit of the sequence {Sn }
is impractical. Consequently, we must develop alternative ways to decide on the convergence
of a series. We do this in Sections 10.10–10.12.
We6now discuss some fairly simple but important results on convergence of series. If a
series cn has a finite number of its terms altered in any fashion whatsoever, the new series
converges if and only if the original series converges. The new series may converge to a different
sum, but it converges if the original series converges. For example, if we double the first three
terms of the geometric series in Example 10.36(a), but do not change the remaining terms, the
new series is
1 1 1
2+1+ + + + ···.
2 8 16
It is not geometric. But its nth partial sum, call it Sn , is very closely related to that of the geometric
series, call it Tn . In fact, for n ≥ 4, we can say that Sn = Tn + 7/4 (7/4 is the total change in
the first three terms). Since limn→∞ Tn = 2, it follows that limn→∞ Sn = 2 + 7/4 = 15/4;
that is, the new series converges, but its sum is 7/4 greater than that of the geometric series.
On the other hand, suppose we change the first 100 terms of the harmonic series, which
diverges, to 0, but leave the remaining terms unaltered. The new series is
100 terms
C DE F 1 1
0 + 0 + ··· + 0+ + + ···.
101 102
We know that the sequence of partial sums {Tn } of the harmonic series is increasing and un-
bounded. The nth partial sum Sn of the new series, for n > 100, is equal to Tn − k , where k is
the sum of the first 100 terms of the harmonic series. But because limn→∞ Tn = ∞ , so must
limn→∞ Sn = ∞ ; that is, the new series diverges.
The following theorem indicates that convergent series can be added and subtracted, and
multiplied by constants. These properties can be verified using Definition 10.5.
THEOREM 10.11
6∞ 6∞
If series n=1 cn and n=1 dn have sums C and D , then:
∞
!
(i) kcn = kC (when k is a constant). (10.41a)
n=1
∞
!
(ii) (cn ± dn ) = C ± D. (10.41b)
n=1
Our first convergence test is a corollary to the following theorem.
THEOREM 10.12
6∞
If a series n=1 cn converges, then limn→∞ cn = 0.
6
PROOF If series cn has a sum S , its sequence of partial sums
S1 , S2 , . . . , S n , . . .
has limit S . The sequence

0, S1 , S2 , . . . , Sn−1 , . . .
must also have limit S ; it is the sequence of partial sums with an additional term equal to 0 at
the beginning. According to Theorem 10.10, if we subtract these two sequences, the resulting
sequence must have limit S − S = 0; that is,
S1 − 0, S2 − S1 , · · · , Sn − Sn−1 , · · · → 0.
But S1 − 0 = S1 = c1 , S2 − S1 = c2 , and so on; that is, we have shown that lim n→∞ cn = 0.
6
Theorem 10.12 states that a necessary condition for a series cn to converge is that the
sequence {cn } of its terms must approach zero. What we really want are sufficient conditions to
guarantee convergence or divergence of a series. We can take the contrapositive of the theorem
and obtain the following.
COROLLARY 10.12.1 ( nth -Term Test)

6∞
If limn→∞ cn -= 0, or does not exist, then the series n=1 cn diverges.
This is our first convergence test, the nth -term test. It is, in fact, a test for divergence rather
than convergence, stating that6 if limn→∞ cn exists and is equal to anything but zero, or the limit
does not exist, then the series cn diverges. Note well that the nth -term test never indicates that
a series converges6 . Even if limn→∞ cn = 0, we can conclude nothing about the convergence
or
6 divergence of cn ; it may converge
6 or it may diverge. For example, the harmonic series
1/n and the geometric series 1/2n−1 both satisfy the condition limn→∞ cn = 0, yet one
series diverges and the other converges. The nth -term test therefore may indicate that a series
diverges, but it never indicates6 that a series converges.6
In particular, both series (−1)n |n − 3| and (−1)n+1 of Example 10.36 diverge by
the nth -term test.
To understand series, it is crucial to distinguish clearly among three 6 entities: the series
itself, its sequence of partial sums, and its sequence of terms. For any series cn , we can form
its sequence of terms {cn } and its sequence of6partial sums {Sn } . Each of these sequences may
give information about the sum of the series cn , but in very different ways.
6
The sum of the series is defined by its sequence of partial sums in that cn has a sum only
if {Sn } has a limit. The sequence of partial sums therefore tells us definitely whether the series
converges or diverges, provided that we can evaluate Sn in a simple form.
The sequence of terms {cn } , on the other hand, may or may not tell us whether the series
diverges. If {cn } has no limit or has a limit other than zero, we know that the series does not
have a sum. If {cn } has limit zero, we obtain no information about convergence of the series,
and must continue our investigation.
The interval of convergence of a power series (or Taylor series) is all values of x for
which the series converges. It is determined, except possibly for endpoints, by the radius of
convergence. As we uncover convergence tests for series of numbers in this and the next three
sections, we will be able to discuss endpoints of intervals of convergence, and consequently, be
able to convert open intervals of convergence to intervals of6 convergence. As a first example,
∞
we show that the open interval of convergence for the series n=1 x 2n /(n 4n ) is its interval of
convergence; that is, endpoints are not included.
EXAMPLE 10.37
!∞
1 2n
Find the interval of convergence for the power series x .
n 4n
n=1
SOLUTION If we set y = x 2 , then
!∞ !∞
1 2n 1 n
x = y .
n 4n n 4n
n=1 n=1
According to equation 10.24a, the radius of convergence of this series is

4 4
4 1 4
4 4 ) *
4 4
Ry = lim 44 n 4n 4 = lim 4 n + 1 = 4.
n→∞ 4 1 4 n→∞ n
4
4 (n + 1)4n+1 4
Consequently, the radius of convergence of the power series in x is Rx = 2, and the open
interval of convergence is −2 < x < 2. At x = ±2, the series becomes
!∞ !∞
1 2n 1
(±2) = ,
n 4n n
n=1 n=1
the harmonic series, which diverges. The interval of convergence is therefore −2 < x < 2.
EXAMPLE 10.38
Solve the problem in the Application Preview.
Application Preview SOLUTION Were we to plot the amount of chemical in the reactor, the result would be
Revisited somewhat like that in Figure 10.25.
FIGURE 10.25 Amount of chemical in a reactor
A0(1 + ekT)
A0
T 2T 3T 4T t
We need to find a formula for the amount of chemical in the reactor at times t = nT ,
taking for granted that the amount A0 has been injected at this time. We already know that
A(0) = A0 , and A(T ) = A0 (1 + ekT ) . During the time interval T < t < 2T , the amount
A0 (1 + ekT ) decreases exponentially so that during this time interval
A(t) = A0 (1 + ekT )ek(t−T ) .
As t → 2T − , this approaches
lim A(t) = A0 (1 + ekT )ek(2T −T ) = A0 (1 + ekT )ekT .
t→2T −
Once the injection of A0 at t = 2T is made,
A(2T ) = A0 + A0 (1 + ekT )ekT = A0 (1 + ekT + e2kT ).
There is a pattern emerging here that we could verify by mathematical induction, but suppose
we accept that
A(nT ) = A0 (1 + ekT + e2kT + · · · + enkT ).
This is a geometric sum, and if we use formula 10.39a,
7 8 7 8
1 − (ekT )n+1 1 − e(n+1)kT
A(nT ) = A0 = A0 .
1 − ekT 1 − ekT
We can now say that the process should be terminated for the smallest integer n such that
7 8
1 − e(n+1)kT
A0 > L.
1 − ekT
This can be solved for
7 8
1 L kT
n> ln 1 − (1 − e ) − 1.
kT A0
EXERCISES 10.9
In Exercises 1–10 determine whether the series converges or diverges. ∗ 23. Find the total distance travelled by the superball in Exercise 56 of
Find the sum of each convergent series. To get a feeling for a series, it Section 10.1 before it comes to rest.
is helpful to write out its first few terms. Try it.
∗ 24. Find the time taken for the superball in Exercise 56 of Section 10.1
∞
! ∞
! to come to rest.
n+1 2n
1. 2.
2n 5n+1 ∗ 25. What distance does the dog run from the time when it sees the
n=1 n=1
farmer until the farmer reaches the farmhouse in Exercise 57 of Section
∞
! # nπ $ !∞ ) *n 10.1?
n
3. cos 4. ∗ 26. Find a simplified formula for the area An in Exercise 61 of Section
n=1
2 n=1
n+1
10.1. What is limn→∞ An ?
∞
! ∞
!
72n+3 7n+3 ∗ 27. Find all values of x satisfying the inequality
5. 6.
32n−2 32n−2
n=1 n=1
1 + x + x 2 + · · · + x n ≥ 0,
&
∞
! ∞
!
n2 − 1 cos (nπ ) where n ≥ 1 is an integer.
7. 8.
n2 + 1 2n
n=1 n=1 ∗ 28. (a) According to equation 10.39a, the nth partial sum of the
geometric series 1 + r + r 2 + r 3 + · · · is Sn = (1 −
∞
! ∞
!
4n + 3n r n )/(1 − r) . If Tn denotes the nth partial sum of the series
9. 10. Tan −1 n
3n
n=1 n=1 1 + 2r + 3r 2 + 4r 3 + · · · ,
show that
In Exercises 11–14 we have given a repeating decimal. Express the
decimal as a geometric series and use formula 10.39b to express it as a Tn − Sn = r(Tn − nr n−1 ).
rational number.
11. 0.666 666 . . . 12. 0.131 313 131 3 . . . Solve this equation for Tn and take the limit as n → ∞ to
show that
13. 1.347 346 346 346 . . . 14. 43.020 502 050 205 . . .
∞
! 1
nr n−1 = , |r| < 1.
In Exercises 15–17 complete the statement and give a short proof to n=1
(1 − r)2
substantiate your claim. 6∞
6 6 6 (b) Verify this result by setting S(r) = n=1 nr n−1 and in-
15. If cn and
dn converge, then (cn + dn ) . . . tegrating.
6 6 6
16. If cn converges and dn diverges, then (cn + dn ) . . .
6 6 6 In Exercises 29–32 use the result of Exercise 28 to find the sum of the
17. If cn and dn diverge, then (cn + dn ) . . . series.
1 2 3 4
29. + + + + ···
In Exercises 18–21 determine whether the series converges or diverges. 2 22 23 24
Find the sum of each convergent series.
2 4 6 8
∞ ∞ 30. + + + + ···
! 2n + 3n ! 3n − 1 5 25 125 625
18. 19.
n=1
4n n=1
2n 2 3 4 5
∗ 31. + + + + ···
3 27 243 2187
∞
! ∞
!
n2 + 22n 2n + 4n − 8n
20. 21. 12 48 192 768
4n 2 3n ∗ 32. + + + + ···
n=1 n=1 5 25 125 625
∗ 22. Find the sum of the series ∗ 33. Two people flip a single coin to see who can first flip a head.
∞
Show that the probability that the first person to flip wins the game is
! 1 represented by the series
.
n(n + 1)
n=1 1 1 1 1
+ + + ··· + + ···.
th
Hint: Use partial fractions on the n term and find the sequence of 2 8 32 22n−1
partial sums. What is the sum of this series?
∗ 34. Two people throw a die to see who can first throw a six. Find the ∗∗ 44. The Laplace transform of a function f (t) was defined by an im-
probability that the person who throws first wins the game. proper integral in the instructions to Exercises 27–30 in Section 8.2.
Show that when f (t) is a continuous function with period p , its
Laplace transform is given by the ordinary integral
In Exercises 35–38 show that the interval of convergence of the power 5 p
series is the same as the open interval of convergence. 1
F (s) = f (t)e−st dt.
1− e−ps 0
∞
! ∞
!
1
∗ 35. x n
∗ 36. 2 n n
n3 x ∗∗ 45. Suppose the voltage Vin applied to the circuit in the left figure
n=0
2n n=1 below is as shown in the right figure. Currents and voltages in the
) *2 circuit are initially zero. The rectifier passes current freely (with no
∞
! ∞
!
n−1 resistance) in the forward direction, but prevents current in the reverse
∗ 37. 2n (x − 4)n ∗ 38. (−1)n x 3n direction.
n=2
n+1 n=0
(a) Show that the differential equation describing the voltage
across the capacitor in the time interval 2(n − 1)T < t <
∗ 39. One of Zeno’s paradoxes describes a race between Achilles and (2n − 1)T , n = 1, 2, . . . , is
a tortoise. Zeno claims that if the tortoise is given a head start, then
no matter how fast Achilles runs, he can never catch the tortoise. He dV R1 + R2 1
reasons as follows: In order to catch the tortoise, Achilles must first + τ V = αV , where τ = and α = .
dt R1 R2 C R1 C
make up the length of the head start. But while he is running this
distance, the tortoise “runs” a further distance. While Achilles makes up Hint: Use the fact that the current through R1 must be the
this distance, the tortoise covers a further distance, and so on. It follows sum of the currents through R2 and C in order to find the
that Achilles is always making up distance covered by the tortoise, and voltage across R1 when the voltage across C is V .
therefore can never catch the tortoise. If the tortoise is given a head (b) Solve the differential equation in part (a) by multiplying
start of length L and Achilles runs c > 1 times as fast as the tortoise, it by eτ t . By denoting the voltage across the capacitor at
use infinite series to show that Achilles does in fact catch the tortoise time t = 2nT by Vn , show that for 2(n − 1)T < t <
and that the distance he covers in doing so is cL(c − 1)−1 . (2n − 1)T ,
∗ 40. Find the time between 1:05 and 1:10 when the minute and hour ; <
hands of a clock point in the same direction (a) by reasoning with infinite αV αV
V = + Vn−1 − − e−τ [t−2(n−1)T ] .
series as in Exercise 39 and (b) by finding expressions for the angular τ τ
displacements of the hands as functions of time.
What is V at t = (2n − 1)T ?
∗ 41. Repeat Exercise 40 for the instant between 10:50 and 10:55 when
the hands coincide.
Rectifier
∗ 42. A child has a large number of identical cubical blocks, which she Vin
1 V
stacks as shown in the following figure. The top block protrudes its 2 R1
length over the second block, which protrudes 41 its length over the third Vin R2 C
block, which protrudes 16 its length over the fourth block, and so on.
Assuming the centre of mass of each block is at its geometric centre, T 2T 3T 4T t
show that the centre of mass of the top n blocks lies directly over the (c) Show that the differential equation describing the voltage
edge of the (n + 1)th block. Now deduce that if a sufficient number of across the capacitor in the time interval (2n − 1)T < t <
blocks is piled, the top block can be made to protrude as far over the 2nT is
bottom block as desired without the stack falling.
dV 1
+ σ V = 0, where σ = .
dt R2 C
(d) Solve the differential equation in
2 part (c) and
3 use the fact
1
_ that lim V should be V (2n − 1)T as calculated
t→(2n−1)T +
2 in part (b) to show that for (2n − 1)T < t < 2nT ,
1
_ 9 ; < :
4 αV αV
V = + Vn−1 − e −τ T
e−σ [t−(2n−1)T ] .
τ τ
1
_
6
(e) Set t = 2nT and V = Vn in the function in part (d) to
1
_ show that Vn = pVn−1 + q , where
8
6∞ αV
∗ 43. Prove the following result: If converges, then its terms p = e−T (τ +σ ) and q = (1 − e−τ T )e−σ T .
n=1 cn τ
can be grouped in any manner, and the resulting series is convergent
with the same sum as the original series. Now find an explicit formula for the sequence {Vn } .
∗∗ 46. It is customary to assume that when a drug is administered to the the nth and (n + 1)th injection is given by
human body, it will be eliminated exponentially; that is, if A represents
7 8
the amount of drug in the body, then 1 − eknT
An (t) = A0 e−kt , (n − 1)T < t < nT .
1 − ekT
A = A0 e−kt ,
(b) Sketch graphs of these functions on one set of axes.
where k > 0 is a constant and A0 is the amount injected at time (c) What is the amount of drug in the body immediately after
t = 0. Suppose n successive injections of amount A0 are administered the nth injection for very large n ; that is, what is
at equally spaced time intervals T , the first injection at time t = 0.
lim An ((n − 1)T )?
(a) Show that the amount of drug in the body at time t between n→∞
10.10 Integral, Comparison, and Limit Comparison Tests

In Section 10.9 we derived the nth -term test, a test for divergence of a series of numbers. In
order to develop further convergence and divergence tests, we consider two classes of series of
numbers:
1. series with terms that are all nonnegative;

2. series with both positive and negative terms.
A series with terms that are all nonpositive is the negative of a series of type 1, and therefore
any test applicable to series of type 1 is easily adapted to a series with nonpositive terms.
DEFINITION 10.6
6∞
A series n=1 cn is said to be nonnegative if each term is nonnegative: cn ≥ 0.
6 6
For example, the harmonic series 1/n and the geometric series 1/2n−1 are both
nonnegative series. We have already seen that both series have the property that limn→∞ cn = 0,
yet the harmonic series diverges and the geometric series converges. Examination of terms of
these series reveals that 1/2n−1 approaches 0 much more quickly than does 1/n . In general,
whether a nonnegative series does or does not have a sum depends on how quickly its sequence of
terms {cn } approaches zero. The study of convergence of nonnegative series is an investigation
into the question: How fast must terms of a nonnegative series approach zero in order that the
series have a sum? This is not a simple problem; only a partial answer is provided through the
tests in this section and the next. We begin with the integral test.
THEOREM 10.13 (Integral test)

6∞
Suppose that the terms in a series n=1 cn are denoted by cn = f (n) , and f (x) is a
continuous, positive, decreasing
G ∞ function for x ≥ 1. Then the series converges if and
only if the improper integral 1 f (x) dx converges.
PROOF To prove this result, we return to the definition of the sum of a series as the limit of
its sequence of partial sums. Suppose first that the improper integral converges to value K . We
know that this value can be interpreted as the area under the curve y = f (x) , above the x -axis,
and to the right of the line x = 1 in Figure 10.26a. In Figure 10.26b the area of the nth rectangle is
FIGURE 10.26a Comparison area for integral test
y = f (x)
1 x
FIGURE 10.26b Rectangles to represent terms of a series to prove convergence
y = f (x)
A = f (1) A = f (n)
A = f (2)
A = f (3) A = f (4)
1 2 3 4 n−1 n x
f (n) = cn , and clearly it is less6

than the area under the curve from x = n − 1 to x = n . Now
the nth partial sum of the series cn is
S n = c1 + c2 + · · · + c n
= f (1) + f (2) + · · · + f (n)
5 2 5 3 5 n
< f (1) + f (x) dx + f (x) dx + · · · + f (x) dx
1 2 n−1
5 n
= f (1) + f (x) dx
1
5 ∞
< f (1) + f (x) dx
1
= f (1) + K.
What this shows is that the sequence
6 {Sn } is bounded [since f (1) + K is independent of n ].
Because all terms of the series cn are positive, the sequence {Sn } of partial sums must be
increasing.
6 It follows by Theorem 10.7 that the sequence {Sn } must have a limit; that is, the
series cn converges.
Conversely, suppose now that the improper integral diverges; the area under the curve
y = f (x) , above the x -axis and to the right of the line x = 1, is “infinite.” This time we draw
rectangles to the right of the vertical lines at n = 1, 2, . . . (Figure 10.27). Then
S n = c1 + c2 + · · · + c n
= f (1) + f (2) + · · · + f (n)
5 2 5 3 5 n+1
> f (x) dx + f (x) dx + · · · + f (x) dx
1 2 n
5 n+1
= f (x) dx.
1
G n+1
Since limn→∞ 1
f (x) dx = ∞ , it follows that limn→∞ Sn = ∞ , and the series therefore
diverges.
FIGURE 10.27 Rectangles to represent terms of a series to prove divergence
A = f (1) y = f (x)
A = f (2) A = f (n)
A = f (3)
1 2 3 4 n n+1 x
To use the integral test it is necessary to antidifferentiate the function f (x) , obtained by
replacing n ’s in cn with x ’s. When the antiderivative appears obvious, and other conditions of
Theorem 10.13 are met, the integral test may be the easiest way to decide on convergence of the
series; when the antiderivative is not obvious, it may be better to try another test.
EXAMPLE 10.39
Determine whether the following series converge or diverge:
∞
! ∞
! ∞
!
1 1
(a) (b) (c) ne−n
n=1
n2 + 1 n=2
n ln n n=1
SOLUTION
(a) Since f (x) = 1/(x 2 + 1) is continuous, positive, and decreasing for x ≥ 1, and
5 ∞
1 + ,∞ π π π
dx = Tan −1 x 1 = − = ,
1 x2 + 1 2 4 4
6
it follows that the series 1/(n2 + 1) converges.
(b) Since this series begins with n = 2, we modify the integral test by considering the
improper integral of f (x) = 1/(x ln x) over the interval x ≥ 2. Since f (x) is
positive, continuous, and decreasing for x ≥ 2, and
5 ∞
1
dx = {ln (ln x)}2∞ = ∞,
2 x ln x
6∞
it follows that n=2 1/(n ln n) diverges.
(c) Since xe−x is continuous, positive, and decreasing for x ≥ 1, and
5 ∞ + ,∞ 2
xe−x dx = −xe−x − e−x 1 = ,
1 e
the series converges.
We mentioned before that two very important series are geometric series and harmonic series.
Convergence of geometric series was discussed in Section 10.9. The harmonic series belongs
to a type of series called p -series, defined as follows:
!∞
1 1 1
= 1 + p + p + ···. (10.42)
n p 2 3
n=1
When p = 1, the p -series becomes the harmonic series, which we know diverges. The series
diverges for p < 0 by the nth -term test. Consider the case when p > 0, but p -= 1. The
function 1/x p is continuous, positive, and decreasing for x ≥ 1, and

5 ∞ ' (∞  ∞, p <1
1 1
dx = = 1
1 xp −(p − 1)x p−1 1
 , p > 1.
p−1
According to the integral test, the p -series converges when p > 1 and diverges when p < 1.
Let us summarize results for geometric and p -series:
Geometric Series p -series
∞
- a ∞ '
! , |r| < 1 ! 1 converges, p > 1
n−1
ar = r−1 = (10.43)
n p diverges, p ≤ 1.
n=1 diverges, |r| ≥ 1. n=1
It is unfortunate that no general formula can be given for the sum of the p -series when
p > 1. Some interesting cases that arise frequently are cited below:
!∞ !∞ !∞
1 π2 1 1 π4
2
= , ≈ 1.202 056 903 1, = . (10.44)
n=1
n 6
n=1
n3 n=1
n 4 90
Many series can be shown to converge or diverge by comparing them to known convergent and
divergent series. This is the essence of the following two tests.
THEOREM 10.14 (Comparison Test)

6∞ 6∞
If 0 ≤ cn ≤ an for all
6 n and n=1 an converges,
6∞ then n=1 cn converges. If cn ≥
dn ≥ 0 for all n and ∞ d
n=1 n diverges, then n=1 cn diverges.
6
PROOF Suppose first that an converges
6 and 0 ≤ cn ≤ an for all n . Since an ≥ 0 the
sequence of partial sums for the series an must be nondecreasing. Since the series converges,
the sequence of partial sums must also be bounded (corollary to Theorem 6 10.7). Because
0 ≤ cn ≤ an for all n , it follows that the sequence of partial sums6of cn is nondecreasing
and bounded also. This sequence therefore has a limit and series cn converges. A similar
argument can be made for the divergent case.
6
Theorem 10.14 states that if6
the terms of a nonnegative series cn are smaller than those of
a known convergent series,
6 then c n must converge; if they are larger than a known nonnegative
divergent series, then cn must diverge.
In order for the comparison test to be useful, we require a catalogue of known convergent
and divergent series with which we may compare other series. The geometric and p -series in
equation 10.43 are extremely useful in this respect.
EXAMPLE 10.40
∞
! !∞ !∞
ln n 2 n2 − 1 2 n2 + 1
(a) (b) (c)
n 15n4 + 14 15n4 − 14
n=2 n=1 n=1
SOLUTION
(a) For this series, we note that when n ≥ 3,
ln n 1
> .
n n
6∞
Since6 n=3 1/n diverges (harmonic series with first6
two terms changed to zero), so
∞ ∞
does n=3 (ln n)/n by the comparison test. Thus n=2 (ln n)/n diverges. This
can also be verified with the integral test.
6
(b) For the series (2n2 − 1)/(15n4 + 14) , we note that
2 n2 − 1 2 n2 2
< . =
15n4
+ 14 15n2 15n4
6 6
Since 2/(15n2 ) = (2/15) 1/n2 converges ( p -series with p = 2), so does
the given series by the comparison test.
(c) For this series we note that the inequality
2 n2 + 1 3
≤
15n4 − 14 n2
is valid if and only if
45n4 − 42 ≥ 2n4 + n2 .
But this inequality is valid if and only if
n2 (43n2 − 1) ≥ 42,
which is obviously true for n ≥ 1. Consequently,
2 n2 + 1 3
≤ ,
15n4 − 14 n2
6 6
and since 3/n2 = 3 1/n2 converges, so does the given series.
Two observations about Example 10.40 are worthwhile:
1. To use the comparison test we must first have a suspicion as to whether the given series
converges or diverges in order to discuss the correct inequality; that is, if we suspect that the
given series converges, we search for a convergent series for the right side of the inequality
≤ , and if we suspect that the given series diverges, we search for a divergent series for the
right side of the opposite inequality ≥ .
2. Recall that when the terms of a nonnegative series approach zero, whether the series has
a sum depends on how fast these terms approach zero. The only difference in nth terms
of Examples 10.40(b) and (c) is the position of the negative sign. For very large n this
difference becomes negligible since each nth term can, for large n , be closely approximated
by 2/(15n2 ) . We might then expect similar analyses for these examples, yet they are quite
different. In addition, it is natural to ask where we obtained the factor 3 in part (c). The
answer is: “by trial and error.”
Each of these observations points out weaknesses in the comparison test, but these problems
can be eliminated in many examples with the following test.
THEOREM 10.15 (Limit Comparison Test)

If 0 ≤ cn and 0 < bn , and
cn
lim
= -, 0 < - < ∞, (10.45)
bn
n→∞
6∞ 6∞ 6∞
then series n=1 cn converges if n=1 bn converges, and diverges if n=1 bn diverges.
6
PROOF Suppose that series bn converges. Since sequence {cn /bn } converges to - , we
can use the result of Exercise 62 in Section 10.8 and say that for all n greater than or equal to
some integer N ,
cn
< -+1 or cn < (- + 1)bn .
bn
6∞ 6∞ 6∞
Since the series n=N (- + 1)b6n = (- + 1) n=N bn converges, so also must n=N cn
∞
(by the comparison test). Hence, n=1 cn converges. A similar argument can be made for the
divergent case.
If limn→∞ cn /bn = - , then for very large n we can say that cn ≈ -bn . It follows that if
{bn } approaches zero, {cn } approaches zero 1/- times as fast as {bn } . Theorem 10.15 implies
then that if the sequences of nth terms of two nonnegative series approach zero at proportional
rates, the series converge or diverge together.
6
The limit comparison test avoids inequalities. To use it, we must find a series bn so that
the limit of the ratio cn /bn is finite and greater than zero. To obtain this series, it is sufficient in
many examples simply to answer the question: What does the given series really look like for
very large n ? In both Examples 10.40(b) and (c), we see that for large n ,
2 n2 2
cn ≈ = .
15n4 15n2
Consequently, we calculate in Example 10.40(c) that

) *
2 n2 + 1 2 1
n 2+
4 n2 15n2
- = lim 15n − 14 = lim ) *· = 1.
n→∞ 2 n→∞ 14 2
n4 15 −
15n2 n4
6 6
Since 2/(15n2 ) converges, so does (2n2 + 1)/(15n4 − 14) .
EXAMPLE 10.41
∞ √ ∞ n
! n2 + 2 n − 1 ! 2 +1
(a) (b)
n=1
n5/2 + 15n − 3 n=1
3n + 5
SOLUTION
(a) For very large n , the nth term of the series can be approximated by
√
n2 + 2 n − 1 n 1
≈ 5/2 = 3/2 .
n5/2 + 15n − 3 n n
We calculate therefore that
√ "
n2 + 2 n − 1 2 1
5/2
n 1+ − 2
- = lim n + 15n − 3 = lim ) n n * · n3/2 = 1.
n→∞ 1 n→∞ 5/2 15 3
n 1 + 3/2 − 5/2
n3/2 n n
6
Since 1/n3/2 converges ( p -series with p = 3/2), so does the given series by the
limit comparison test.
(b) For large n ,
) *n
2n + 1 2
≈ .
3n + 5 3
We calculate therefore that
) *
2n + 1 n 1
2 1+
2n 3n
- = lim 3) +*n5 = lim
n
) * · n = 1.
n→∞ 2 n→∞ 5 2
3n 1+ n
3 3
6
Since (2/3)n converges (a geometric series with r = 2/3), the given series
converges by the limit comparison test.
EXAMPLE 10.42
∞
! n
Find the interval of convergence for the power series (x − 1)2n .
n=1
(n + 1)3 2n
2
SOLUTION If we set y = (x − 1) , then
∞
! !∞
n 2n n
3
(x − 1) = y n.
(n + 1) 2 n (n + 1)3 2n
n=1 n=1
The radius of convergence of the series in y is

4 4
4 n 4
4 4 3
4 (n + 1)3 2n 4
4
Ry = lim 4 4 = lim 2n(n + 2) = 2.
n→∞ 4 n+1 4 n→∞ (n + 1)4
4
4 (n + 2)3 2n+1 4
√
The radius of convergence of the
√power series in √
x − 1 is therefore Rx = 2,√and the open
interval of convergence is 1 − 2 < x < 1 + 2. At endpoints x = 1 ± 2, the power
series becomes
∞
! n
.
n=1
(n + 1)3
Since
n
(n + 1)3 n3
- = lim = lim = 1,
n→∞ 1 n→∞ (n + 1)3
n2
6 2
6 3
and 1/n converges, so does
√ n/(n + 1)√ (by the limit comparison test). The interval of
convergence is therefore 1 − 2 ≤ x ≤ 1 + 2.
EXERCISES 10.10
∞
! ∞
!
In Exercises 1–22 determine whether the series converges or diverges. 2n + n 1 + ln2 n
17. 18.
n=1
3n + 1 n=2
n ln2 n
∞
! ∞
!
1 1
1. 2. ∞
! 1 + 1/n
∞
! ln (n + 1)
n=1
2n + 1 n=1
4n − 3 19. 20.
n=1
en n=1
n+1
∞
! ∞
!
1 1 ∞ ∞
3. 4. ! 2
! 1
n=1
2n2 + 4 n=1
5n2 − 3n − 1 21. ne−n 22. √
3
n=1 n=2 n ln n
∞
! ∞
!
1 n2
5. 6.
n3 − 1 n − 6n2 + 5
4
n=2 n=4 In Exercises 23–26 find the interval of convergence of the power series.
∞
! ∞
!
1 n−5 !∞ !∞
7. 8. 1 2n 1 2n
(2n − 1)(2n + 1) n2 + 3n − 2 ∗ 23. x ∗ 24. 2
x
n=1 n=1 n n
n=1 n=1
∞
! ∞
!
1 !∞ ∞
!
9. 10. n2 e−2n 1 n3n
ln n ∗ 25. (x − 1)4n ∗ 26. (x + 2)2n
n=2 n=1
n=1
n2n n=0
(n + 1)3
∞ √ ∞ √
! n2 + 2n − 3 ! n+5
11. 12.
n=2
n2 + 5 n=1
3
n +3 In Exercises 27–29 find values of p for which the series converges.
&
!∞ ∞
! ∞ ∞
n2 + 2n + 3 1 ! 1 ! 1
13. 14. ∗ 27. ∗ 28.
n=1
2n4 − n n=2
n2 ln n np ln n n(ln n)p
n=2 n=2
∞
! #π $ !∞ √ 2 ∞
!
1 n +1 1
15. sin 16. Tan −1 n ∗∗ 29.
n=1
2n n n=1
n3 n=2
(ln n)p
10.11 Limit Ratio and Limit Root Tests

In this section we consider two additional tests to determine whether nonnegative
6 series converge
or diverge. The first test, the limit ratio test, indicates whether a series cn resembles a
geometric series for large n .
THEOREM 10.16 (Limit Ratio Test)

Suppose that cn > 0 and
cn+1
lim = L. (10.46)
n→∞ cn
Then:
6∞
(i) 1 cn converges if L < 1.
6n=
∞
(ii) 1 cn diverges if L > 1 (or if lim n→∞ cn+1 /cn = ∞ ).
6n=
∞
(iii) n=1 cn may converge or diverge if L = 1.

Schematic for proof of limit ratio test
L L+1 1 1 L+1 L
2 2
PROOF
(i) By the result of part (a) in Exercise 61 of Section 10.8 (see also Figure 10.28a), we
can say that if limn→∞ cn+1 /cn = L < 1, there exists an integer N such that for
all n ≥ N ,
cn+1 L+1
< .
cn 2
Consequently,
) *
L+1
cN + 1 < cN ;
2
) * ) *2
L+1 L+1
cN +2 < cN + 1 < cN ;
2 2
) * ) *3
L+1 L+1
cN + 3 < cN + 2 < cN ;
2 2
and so on. Hence,

) * ) *2
L+1 L+1
c N + c N + 1 + c N + 2 + · · · < cN + cN + cN + · · · .
2 2
Since the right side of this inequality is a geometric6series with common ratio
∞
(L +6 1)/2 < 1, it follows by the comparison test that n=N cn converges. There-
∞
fore, n=1 cn converges also.
(ii) By the result of part (b) in Exercise 61 of Section 10.8 (see also Figure 10.28b), if
limn→∞ cn+1 /cn = L > 1, there exists an integer N such that for all n ≥ N ,
cn+1 L+1
> > 1.
cn 2
When limn→∞ cn+1 /cn = ∞ , it is also true that for n greater than or equal to some
N , cn+1 /cn must be greater than 1. This implies that for all n > N , cn > cN , and
therefore
lim cn -= 0.
n→∞
6
Hence cn diverges by the nth -term test.
(iii) To show6 that the limit
6 ratio test is inconclusive when L = 1, consider the two p -
series 1/n and 1/n2 . For each series, L = 1, yet the first series diverges and
the second converges.
In a nonrigorous way we can justify the limit ratio test from 6the following standpoint. If
limn→∞ cn+1 /cn = L , then for large n , each term of series cn is essentially L times the
term before it; that is, the series resembles a geometric series with common ratio L . We would
expect convergence of the series if L < 1 and divergence if L > 1. We might also anticipate
some indecision about the L = 1 case, depending on how this limit is reached, since this case
corresponds to the common ratio that separates convergent and divergent geometric series.
EXAMPLE 10.43
∞ n
! ∞
! ∞
!
2 n100 nn
(a) (b) (c)
n=1
n4 n=1
1 · 3 · 5 · . . . · (2n − 1)
n=1
n!
SOLUTION
6
(a) For the series 2n /n4 ,
2n+1
) *4
(n + 1)4 n
L = lim = lim 2 = 2,
n→∞ 2n n→∞ n+1
n4
and the series therefore diverges by the limit ratio test.

(b) For this series,
(n + 1)100
) *
1 · 3 · 5 · . . . · (2n + 1) n + 1 100 1
L = lim = lim = 0,
n→∞ n100 n→∞ n 2n + 1
1 · 3 · 5 · . . . · (2n − 1)
and the series therefore converges by the limit ratio test.

(c) Since
(n + 1)n+1
) *
(n + 1)! n+1 n
L = lim = lim =e >1
n→∞ nn n→∞ n
n!
6 n
(see equation 1.68, the series n /n! diverges.
We present one last test for nonnegative series; there are many others.
THEOREM 10.17 (Limit Root Test)

Suppose that cn ≥ 0 and
√
lim n
cn = R. (10.47)
n→∞
Then:
6∞
(i) 1 cn converges if R < 1.
6n=
∞ √
(ii) 1 cn diverges if R > 1 (or if lim n→∞
n c = ∞ ).
n
6n=
∞
(iii) n=1 nc may converge or diverge if R = 1.
PROOF
√
(i) By the result of part (a) in Exercise 61 of Section 10.8, if limn→∞ n cn = R < 1,
there exists an integer N such that for all n ≥ N ,
√
n
R+1
cn < .
2
Consequently,
) *N
√ R+1 R+1
N
cN < or cN < ;
2 2
) *N +1
√ R+1 R+1
N+1
cN + 1 < or cN + 1 < ;
2 2
) *N +2
√ R+1 R+1
N+2
cN + 2 < or cN + 2 < ;
2 2
and so on. Hence,

) *N ) *N +1 ) *N + 2
R+1 R+1 R+1
c N + cN + 1 + cN + 2 + · · · < + + + ···.
2 2 2
Since the right side of this inequality is a convergent geometric6series, the left side
∞
must be a convergent
6∞ series also by the comparison test. Thus, n=N n converges,
c
and so must n=1 cn .
√
(ii) When limn→∞ n cn = R > 1, there exists an integer N such that for all n ≥ N ,
√ R+1
n
cn > >1 or cn > 1,
2
√
as in part (b) of Exercise 61 in Section 10.8. When lim n→∞ n cn = ∞ , it is also
√
true that for n greater than or equal to some N , n cn must be greater than 1. But it
now follows that limn→∞ cn -= 0, and the series diverges by the nth -term test.
(iii) To show that 6 the test is inconclusive
6 when R = 1, we show that R = 1 for the
two p -series 1/n and 1/n2 , diverging and converging, respectively. For the
harmonic series, let R = limn→∞ (1/n)1/n . If we take logarithms, then
9 ) * n1 : ) *1/n ) *
1 1 1 ln n
ln R = ln lim = lim ln = lim 1/n ln = − lim
n→∞ n n→∞ n n→∞ n n→∞ n
1/n
= − lim (by L’Hôpital’s rule)
n→∞ 1
= 0.
Hence, R = 1 for 6this divergent series. A similar analysis gives R = 1 for the
convergent series 1/n2 .
EXAMPLE 10.44
∞ )
! *2 ∞
!
n+1 n n
(a) (b)
n=1
n n=1
(ln n)n
SOLUTION
(a) Since
9) *n2 :1/n ) *n
n+1 n+1
R = lim = lim = e > 1,
n→∞ n n→∞ n
the series diverges by the limit root test.

(b) For this series,
7 81/n
n n1/n
R = lim = lim .
n→∞ (ln n)n n→∞ ln n
If we set L = limn→∞ n1/n , then

# $ 1
ln L = ln lim n1/n = lim ln n
n→∞ n→∞ n
1
= lim n (by L’Hôpital’s rule)

n→∞ 1
= 0.
Thus, L = 1, and it follows that
n1/n
R = lim = 0.
n→∞ ln n
The series therefore converges.
We have developed six tests to determine whether series of numbers converge or diverge: the
nth -term, integral, comparison, limit comparison, limit ratio, and limit root tests.
The form of the nth term of a series often suggests which test should be used. Keep the
following ideas in mind when choosing a test:
1. The limit ratio test can be effective on factorials, products of the form 1 · 3 · 5 · . . . ·(2n− 1) ,
and constants raised to powers involving n (2n , 3−n , etc.).
2. The limit root test thrives on functions of n raised to powers involving n (see Example
10.44).
3. The limit
√ comparison
√ test is successful on rational functions of n , and fractional powers as
well ( n , 3 n/(n + 1) , etc.).
4. The integral test can be effective when the nth term is easily integrated. Logarithms often
require the integral test.
6
By definition, a series cn converges if and only if its sequence of partial sums {Sn } converges.
The difficulty with using this definition to discuss convergence of a series is that Sn can seldom
be evaluated in a simple form, and therefore consideration of limn→∞ Sn is impossible. The
tests above have the advantage of avoiding partial sums. On the other hand, they have one
disadvantage. Although they may indicate that a series does indeed have a sum, the tests in
no way suggest the value of the sum. The problem of calculating the sum often proves more
difficult than showing that it exists in the first place. In Section 10.13 we discuss various ways
to calculate and approximate sums for known convergent series.
EXERCISES 10.11
∞
!
In Exercises 1–20 determine whether the series converges or diverges. 1 + 1/n
13.
n=1
en
∞
! !∞
en 1 ∞
! ) *
1. 2. 2 · 4 · . . . · (2n) 1
n4 n! 14.
n=1 n=1
n=1
3 · 5 · . . . · (2n + 1) n2
∞
! !∞
n3 1 ∞
! nn
∞
! (n + 1)n
3. 4. ∗ 15. ∗ 16.
n=1
2n n=1
nn (n + 1)n+1 nn+1
n=1 n=1
∞ ∞ ∞
! ∞
!
! (n − 1)(n − 2) ! (2n)! n4 + 3 2n + n2 3n
5. 6. ∗ 17. ∗ 18.
n2 2n (n!)2 n=1
5n/2 n=1
4n
n=1 n=1
√ ∞
! ∞
!
∞
! ∞
! n2 2n − n (2n)!
n+1 3−n + 2−n ∗ 19. ∗ 20. 52n
7. 8. n3 + 1 (3n)!
n=1
nn+1/2 n=1
4−n + 5−n n=1 n=1
∞ ∞
! e−n ! 2 · 4 · . . . · (2n) ∗ 21. For what integer values of a is the series
9. √ 10.
n+π 4 · 7 · . . . · (3n + 1) ∞
n=1 n=1 ! (n!)2
∞
! ∞
! ) *n (an)!
nn−1 3 n=1
11. 1
12. n
n=1
3 (n − 1)!
n−
n=1
4 convergent?
10.12 Absolute and Conditional Convergence, Alternating Series

The convergence tests of Sections 10.10 and 10.11 are applicable to nonnegative series, series
whose terms are all nonnegative. Series with infinitely many positive and negative terms are more
complicated. It is fortunate, however, that all our tests are still useful in discussing convergence
of series with positive and negative terms. What makes this possible is the following definition
and Theorem 10.18.
DEFINITION 10.7
6∞
A
6series n=1 cn is said to be absolutely convergent if the series of absolute values
∞
n=1 |cn | converges.
At first glance it might seem that absolute convergence is a strange concept indeed. What
possible good could it do to consider the series of absolute values, which is quite different from
the original series? The fact is that when the series of absolute values converges, it automatically
follows from the next theorem that the original series converges also. And since the series of
absolute values has all nonnegative terms, we can use the comparison, limit comparison, limit
ratio, limit root, or integral test to consider its convergence.
THEOREM 10.18
If a series is absolutely convergent, then it is convergent.
6
PROOF Let {Sn } be the sequence of partial sums of the absolutely convergent series cn .
Define sequences {Pn } and {Nn } , where Pn is the sum of all positive terms in Sn , and Nn is
the sum of the absolute values of all negative terms in Sn . Then
Sn = Pn − Nn .
6
The sequence of partial sums for the series of absolute values |cn | is
{Pn + Nn },
and this sequence must be nondecreasing and bounded. Since each of the sequences {Pn } and
{Nn } is nondecreasing and a part of {Pn + Nn } , it follows that each is bounded, and therefore
has a limit, say, P 6
and N , respectively. As a result, sequence {Pn − Nn } = {Sn } has limit
P − N , and series cn converges to P − N .
EXAMPLE 10.45
Show that the following series are absolutely convergent:

!∞
(−1)n n
(a)
n=1
(n + 1)2n
1 1 1 1 1 1 1 1 1 1
(b) 1− − + + + − − − − + ··· + − ···
22 33 44 55 66 77 88 99 1010 1515
SOLUTION
6
(a) The series of absolute values is n/[(n + 1)2n ]. We use the limit comparison test
to show that it converges. Since
n
(n + 1)2n
- = lim = 1,
n→∞ 1
2n
6
and (1/2)n is convergent (a geometric series with r = 1/2), it follows that the
series of absolute values converges. The given series therefore converges absolutely.
6
(b) The series of absolute values is 1/nn . Since
) *1/n
1 1
R = lim = lim = 0,
n→∞ nn n→∞ n
6
the series 1/nn converges by the limit root test. The given series therefore con-
verges absolutely.
In Example 10.45, absolute convergence of the given series implies convergence of the series,
but it is customary to omit such a statement. It is important to realize, however, that it is the
given series that is being analyzed, and its convergence is guaranteed by Theorem 10.18.
We now ask whether series can converge without converging absolutely. If there are such
series, and indeed there are, we must devise new convergence tests. We describe these series as
follows.
DEFINITION 10.8
A series that converges but does not converge absolutely is said to converge conditionally.
The most important type of series with both positive and negative terms is an alternating
series. As the name suggests, an alternating series has terms that are alternately positive and
negative. For example,
∞
! (−1)n+1 1 1 1 1
= 1− + − + − ···
n 2 3 4 5
n=1
is an alternating series called the alternating harmonic series.

Given an alternating series to examine for convergence, we first test for absolute conver-
gence as in Example 10.45(a). Should this fail, we check for conditional convergence with the
following test.
THEOREM 10.19 (Alternating Series Test)

6∞
An alternating series n=1 cn converges if the sequence of absolute values of the terms
{|cn |} is decreasing and has limit zero.
PROOF If {Sn } is the sequence of partial sums of the series, then differences of partial sums
are Sn − Sn−1 = cn . Because the cn alternate in sign, so also do the differences Sn − Sn−1 .
This means that the partial sums {Sn } form an oscillating sequence. Since absolute values
|Sn − Sn−1 | = |cn | are decreasing and approach zero, it follows by Theorem 10.8 of Section
10.8 that the sequence {Sn } of partial sums has a limit.
EXAMPLE 10.46
Determine whether the following series converge absolutely, converge conditionally, or diverge:
∞ ∞ √ ∞
! (−1)n+1 ! n2 + 5n ! 4n
(a) (b) (−1) n
(c) (−1)n+1
n=1
n n=1
n3/2 n=1
n5 3n
SOLUTION
6 n+1
(a) The alternating harmonic series
6 (−1) /n is not absolutely convergent because
the series of absolute values 1/n diverges. Since the sequence
6 of absolute values
of the terms {1/n} is decreasing with limit zero, the series (−1)n+1 /n converges
conditionally.
(b) For this alternating series, we first consider the series of absolute values
∞ √
! n2 + 5n
.
n=1
n3/2
We use the limit comparison test on this series. Since

√ "
n2 + 5n 5
3/ 2
n 1+
- = lim n = lim n · n1/2 = 1,
n→∞ 1 n→∞ n3/2
n1/2
6 6√ 2
and 1/n1/2 diverges, so does n + 5n/n3/2 . The original series does not
therefore converge
√ absolutely. We now resort to the alternating series test. The
sequence { n2 + 5n/n3/2 } of absolute values of the terms of the series is decreasing
if 1 √
(n + 1)2 + 5(n + 1) n2 + 5n
< .
(n + 1)3/2 n3/2
When we square and cross multiply, the inequality becomes
n3 (n2 + 7n + 6) < (n2 + 5n)(n + 1)3

= n5 + 8n4 + 18n3 + 16n2 + 5n;
that is,
n4 + 12n3 + 16n2 + 5n > 0,
√
which is obviously valid because n ≥61. Since√ limn→∞ n2 + 5n/n3/2 = 0,
we conclude that the alternating series (−1) ( n2 + 5n/n3/2 ) converges con-
n
ditionally.
6 n 5 n
(c) If we apply the limit ratio test to the series of absolute values 4 /(n 3 ) , we have
4n+1
) *5
(n + 1)5 3n+1 4 n 4
L = lim n = lim = .
n→∞ 4 n→∞ 3 n+1 3
n5 3n
6 n
Since L > 1, the series 4 /(n5 3n ) diverges. The original alternating series does
not therefore converge absolutely. But L = 4/3 implies that for large n , each term
in the series of absolute values is approximately 4/3 times the term that precedes it,
and therefore
4n
lim 5 n = ∞.
n→∞ n 3
Consequently,
4n
lim (−1)n+1 5 n
n→∞ n3
cannot possibly exist, and the given series diverges by the nth -term test.
We have noted several times that the essential question for convergence of a nonnegative series
is: Do the terms approach zero quickly enough to guarantee convergence of the series? With a
series that has infinitely many positive and negative terms, this question is inappropriate. Such
a series may converge because of a partial cancelling effect; for example, a negative term may
offset the effect of a large positive term. This kind of process may produce a convergent series
even though the series would be divergent if all terms were replaced by their absolute values. A
specific example is the alternating harmonic series which converges (conditionally) because of
this cancelling effect, whereas the harmonic series itself, which has no cancellations, diverges.
Absolute and conditional convergence are particularly important when discussing endpoints
of intervals of convergence for power series since one of the endpoints often leads to an alternating
series. We illustrate in the following examples.
EXAMPLE 10.47
∞
! 1
Find the interval of convergence for the power series xn .
n=0
(n + 1)2n
SOLUTION Since the radius of convergence is
4 4
4 1 4
4 4 ) *
4 (n + 1)2n 4
R = lim 44 4 = lim 2 n + 2 = 2,
n→∞ 4 1 4 n→∞ n+1
4
4 (n + 2)2n+1 4
6∞
the open interval of convergence is −2 < x < 2. At x = 2, the series is n=0 1/(n + 1) ,
the harmonic series, which diverges. At x = −2, we obtain the alternating harmonic series
6 ∞ n
n=0 (−1) /(n + 1) , which converges conditionally. The interval of convergence is therefore
−2 ≤ x < 2.
EXAMPLE 10.48
!∞
n(−1)n
Find the interval of convergence for the power series (x − 3)n .
( 2n + 5)3
n=1
SOLUTION Since the radius of convergence is

4 4
4 n(−1)n 4
4 4
4 (2n + 5)3 4 n(2n + 7)3
R = lim 44 4 = lim = 1,
n→∞ 4 (n + 1)(−1)n+1 44 n→∞ (n + 1)(2n + 5)3
4 4
(2n + 7)3
the open interval of convergence is 2 < x < 4. At x = 2, the series becomes
!∞ !∞
n(−1)n n n
(−1) = .
( 2n + 5)3 ( 2n + 5)3
n=1 n=1
Since
4 n 4
4 4
4 3 4 8n3
4 (2n + 5) 4
- = lim 4 4 = lim = 1,
n→∞ 4 1 4 n→∞ (2n + 5)3
4 4
8n2
6 6 6
and 1/(8n2 ) = (1/8) 1/n2 converges ( p = 2 series), so also does n/(2n + 5)3 (by
the limit comparison test). At x = 4, the power series becomes
!∞
n(−1)n
.
n=1
(2n + 5)3
This is an alternating series that converges absolutely, as indicated in the discussion of x = 2.

The interval of convergence is therefore 2 ≤ x ≤ 4.
In Sections 10.10–10.12, we have obtained a number of tests for determining whether series
of numbers converge or diverge. To test a series for convergence, we suggest the following
procedure:
1. Try the nth -term test for divergence.

2. If {cn } has limit zero and the series is nonnegative, try the comparison, limit comparison,
limit ratio, limit root, or integral test.
3. If {cn } has limit zero and the series contains both positive and negative terms, test for
absolute convergence using the tests in 2. If this fails and the series is alternating, test for
conditional convergence with the alternating series test.
Each of the comparison, limit comparison, limit ratio, limit root, integral, and alternating series
tests requires conditions to be satisfied for all terms of the series. Specifically, the comparison,
limit comparison, and limit root tests require that cn ≥ 0 for all n ; the limit ratio test requires
cn > 0; the integral test requires f (n) to be positive, continuous, and decreasing; and the
alternating series test requires {|cn |} to be decreasing and {cn } to be alternately positive and
negative. None of these requirements is essential for all n ; in fact, so long as they are satisfied
for all terms in the series beyond some point, say 6for n greater than or equal to6some integer
∞
N , the particular test may
6∞ be used on the series n=N cn . The original series ∞ n=1 cn then
converges if and only if n=N cn converges.
Before leaving this section, we prove Theorem 10.2. We verify 10.24a using the limit ratio
test; verification of 10.24b is 6
similar, using the limit root test. If the limit ratio test is applied to
the series of absolute values |an x n | ,
4 4
|an+1 x n+1 | 4 an+1 4
4 4.
L = lim = |x| lim
n→∞ |an x n | n→∞ 4 an 4
Assuming that limit 10.24a exists or is equal to infinity, there are three possibilities:
(i) If limn→∞ |an /an+1 | = 0, then limn→∞ |an+1 /an | = ∞ . Therefore, L = ∞ ,
and the power series diverges for all x -= 0. In other words,
4 4
4 an 4
R = 0 = lim 44 4.
n→∞ an+1 4
(ii) If limn→∞ |an /an+1 | = ∞ , then limn→∞ |an+1 /an | = 0. Therefore, L = 0, and
the power series converges absolutely for all x . Consequently,
4 4
4 an 4
R = ∞ = lim 44 4.
4
n→∞ a n+1
(iii) If limn→∞ |an /an+1 | = R , then limn→∞ |an+1 /an | = 1/R . In this case, L =
|x|/R . Since the power series converges absolutely for L < 1 and diverges for
L > 1, it follows that absolute convergence occurs for |x| < R and divergence for
|x| > R . This implies that R is the radius of convergence of the power series.
EXERCISES 10.12
∞ ) *n ∞ √
In Exercises 1–14 determine whether the series converges absolutely, ! n ! n2 + 3
∗ 11. (−1)n ∗ 12. (−1)n
converges conditionally, or diverges. n+1 n2 + 5
n=1 n=1
∞
! ∞
!
ln n cos (nπ/10) Cot −1 n
∗ 13. (−1)n−1 ∗ 14.
∞
! n
∞
! n n=2
n n=1
n3 + 5n
1. (−1)n 2. (−1)n
n=1
n3 + 1 n=1
n2 + 1
∞ ∞ In Exercises 15–22 find the interval of convergence of the power series.

! cos (nπ/2) ! n3
3. 4. (−1)n !∞ ∞
!
2n2 3n 1 n 1
n=1 n=1 ∗ 15. x ∗ 16. xn
n=1
n n=0
(n + 1)2
!∞ ∞
!
(−1)n+1 n3
n
∞ ∞
5. √ 6. (−1) ! 1 ! 1
n=1
n n=1
n3 ∗ 17. (x − 1)n ∗ 18. √ (x + 2)n
n=1
n2 n
n=1
n
∞
! ∞
!
n n sin (nπ/4) ∞
! n−1
∞
! 1
7. (−1)n 8. ∗ 19. (2x)n ∗ 20. √ x 3n+1
n2 +n+1 2n n2 + 1
n=1 n=1
n=0 n=0
n+1
∞ ) * ∞ √ ∞ ∞
! n ! 3n − 2 ! 1 ! 1
9. (−1)n+1 10. (−1)n+1 ∗ 21. xn ∗ 22. (x − 2)n
n=1
n+1 n=1
n n=2
ln n n=2
n2 ln n
∗ 23. Discuss convergence of the series ∗ 25. Discuss convergence of the series
∞
! sin (nx)
.
n=1
n2
∞
!
6 6 p nn
∗ 24. Prove that if cn converges absolutely, then cn converges (−1)n .
absolutely for all integers p > 1. n=1
(n + 1)n+1
10.13 Exact and Approximate Values for Sums of Series of Numbers

In Sections 10.9–10.12 we have concentrated on whether series of numbers converge or diverge.
But this is only half the problem. The comparison, limit comparison, limit ratio, limit root,
integral, and alternating series tests may determine whether a series converges or diverges, but
they do not determine the sum of the series in the case of a convergent series. This part of the
problem, as suggested before, can sometimes be more complicated.
If the convergent series is a geometric series, no problem exists; we can use formula 10.39b
to find its sum. It may also happen that the nth partial sum Sn of the series can be calculated in
a simple form, in which case the sum of the series is limn→∞ Sn . Cases of the latter type are
very rare. By substituting values of x into power series with known sums, we obtain formulas
for sums of series of numbers. For instance, in Example 10.10, we verified that the Maclaurin
series for ex is
!∞
x 1 n
e = x , −∞ < x < ∞.
n=0
n!
By substituting x = 1, we obtain a series that converges to e ,
!∞
1 1 1 1
e = = 1+ + + + ···.
n ! 1! 2! 3!
n=0
Another illustration is contained in the following example.
EXAMPLE 10.49
Use Example 10.22 to show that the sum of the alternating harmonic series is ln 2.
SOLUTION According to Example 10.22, the Taylor series about x = 1 for ln x is
∞
! (−1)n+1 1 1
ln x = (x − 1)n = (x − 1) − (x − 1)2 + (x − 1)3 − · · · ,
n 2 3
n=1
with open interval of convergence 0 < x < 2. At x = 0, the series becomes the negative
of the harmonic series that diverges, and at x = 2, it becomes the conditionally convergent,
alternating harmonic series; that is,
∞
! (−1)n+1 1 1 1
ln 2 = = 1− + − + ···.
n 2 3 4
n=1
Convergence of the Taylor series at x = 2 does not, by itself, imply convergence to ln 2, as we

are suggesting. It is, however, true, and this is a direct application of the following theorem.
THEOREM 10.20
6∞ n
If the Taylor series n=0 an (x − c) of a function f (x) converges at the endpoint
x = c + R of its interval of convergence, and if f (x) is continuous at x = c + R , then
the Taylor series evaluated at c + R converges to f (c + R) . The same result is valid at
the other endpoint, x = c − R .
This example suggests another possibility for summing convergent series of numbers. Find
a power series with known sum that reduces to the given series of numbers upon substitution of
a value of x . We illustrate this in the next two examples.
EXAMPLE 10.50
∞
! n
Find .
3n
n=1
SOLUTION This series results if we set x = 1/3 in the power series
∞
!
S(x) = nx n .
n=1
The radius of convergence of this series is

4 4
4 n 4
4
R = lim 4 4 = 1.
n→∞ n + 1 4
If we divide both sides by x ,

∞
!
1
S(x) = nx n−1 ,
x n=1
and integrate according to Theorem 10.3,
5 ∞
!
1
S(x) dx = x n.
x n=1
This is a geometric series with sum

5
1 x
S(x) dx = , |x| < 1.
x 1−x
Differentiation now gives
1 (1 − x)(1) − x(−1) 1
S(x) = = .
x (1 − x)2 (1 − x)2
Thus,
∞
! x
S(x) = n xn = .
n=1
(1 − x)2
When we set x = 1/3,
∞
! n 1/3 3
= = .
3n (1 − 1/3)2 4
n=1
EXAMPLE 10.51
∞
! (−1)n
Find the sum of the series .
n=0
(2n + 1)2n
SOLUTION There are many power series that reduce to this series √ upon substitution of a
specific value of x . For instance, substitution of −1/2, 1, and 1/ 2 into the following power
series, respectively, lead to the given series:
∞ ∞ ∞ √
! 1 ! (−1)n ! 2(−1)n
n
x , x n, x 2n+1 .
2n + 1 (2n + 1)2n 2n + 1
n=0 n=0 n=0
Which should we consider? Although it is not the simplest, the third series looks most promising;
the fact that the power on x corresponds to the coefficient in the denominator suggests that we
can find the sum of this series. We therefore set
∞ √
! 2(−1)n
S(x) = x 2n+1 .
2n + 1
n=0
To find the radius of convergence of this series, we set y = x 2 , in which case

∞ √ ∞ √
! 2(−1)n ! 2(−1)n
2n+1
x =x y n.
2n + 1 2n + 1
n=0 n=0
The radius of convergence of the y -series is

4 √ 4
4 2(−1)n 44
4 ) *
4 2n + 1 4 2n + 3
4
Ry = lim 4 √ 4 = lim = 1.
n→∞ 4 2(−1)n+1 44 n→∞ 2n + 1
4 4
2n + 3
The radius of convergence for the power series in x is therefore Rx = 1 also. If we differentiate
the series with respect to x ,
∞ ∞ √ √
! √ ! √ 2 2
" n 2n 2 n
S (x) = 2(−1) x = 2(−x ) = = .
1 − (−x 2 ) 1 + x2
n=0 n=0
Antidifferentiation now gives

5 √
2 √
S(x) = dx = 2 Tan −1 x + C.
1+ x2
Since S(0) = 0, it follows that C = 0, and
∞ √
! 2(−1)n √
x 2n+1 = 2 Tan −1 x.
2n + 1
n=0
√
If we now set x = 1/ 2,
∞ √ ) *2n+1 ) *
! 2(−1)n 1 √ 1
−1
√ = 2 Tan √ .
2n + 1 2 2
n=0
Consequently,
∞
! ) *
(−1)n √ 1
−1
= 2 Tan √ .
n=0
(2n + 1)2n 2
Approximating the Sum of a Series of Numbers

The techniques described above do not find sums for all convergent series of numbers. In fact,
there are many series for which we would find it impossible to find a sum. But in applications we
might be satisfied with a reasonable approximation to the sum of a series, and we therefore turn
our attention to the problem of estimating the sum6of a convergent series. The easiest method
for estimating the sum S of a convergent series cn is simply to choose the partial sum SN
for some N as an approximation; that is, truncate the series after N terms and choose
S ≈ S N = c1 + c2 + · · · + c N .
But an approximation is of value only if we can make some definitive6statement about its
∞
accuracy. In truncating the series, we have neglected the infinity of terms n=N +1 cn , and the
6∞
accuracy of the approximation is therefore determined by the size of n=N +1 cn ; the smaller
6is,
it
∞
the better the approximation. The problem is that we do not know the exact value of
n=N +1 cn ; if we did, there would be no need to approximate
6∞ the sum of the original series in
the first place. What we must do is estimate the sum n=N +1 cn .
When the integral test or the alternating series test is used to prove that a series converges,
simple formulas give accuracy estimates on the truncated series. Let us illustrate these first.
Truncating an Alternating Series

FIGURE 10.29 Approx-
imating the sum of a convergent 6obtain the truncation error, an estimate of the accuracy of a truncated
It is very simple to
alternating series alternating series cn provided that the sequence {|cn |} is decreasing with limit zero. For
example, suppose that c1 >6 0 (a similar discussion can be made when c1 < 0). If {Sn } is the
Sn sequence of partial sums of cn , then even partial sums can be expressed in the form
S1
S3
S5 S2n = (c1 + c2 ) + (c3 + c4 ) + · · · + (c2n−1 + c2n ).
S7
S Since {|cn |} is decreasing ( |cn | > |cn+1 | ), each term in the parentheses is positive. Conse-
S6 S8 quently, the subsequence {S2n } of even partial sums of {Sn } is increasing and approaches the
S4 6
S2 sum of the series cn from below (Figure 10.29). In a similar way, we can show that the
subsequence {S2n−1 } of odd partial6 sums is decreasing and approaches the sum of the series
from above. It follows that the sum cn must be between any two terms of the subsequences
1 2 3 4 5 6 7 8 n {S2n } and {S2n−1 } . In particular, the sum of the alternating series must be between any two
successive partial sums. Furthermore, when the alternating series is truncated, the maximum
possible error is the next term.
EXAMPLE 10.52
6∞ n+1
Use the first 20 terms of the series n=1 (−1) /n3 to estimate its sum. Obtain an error
estimate.
SOLUTION The sum of the first 20 terms of the series is 0.901 485. Since the series is
alternating, and absolute values of terms are decreasing with limit zero, the maximum possible
error in this estimate is the 21 st term, 1/213 < 0.000 108. Thus,
∞
! (−1)n+1
0.901 485 < < 0.901 593.
n=1
n3
In practical situations, we often have to decide how many terms of a series to take in order to
guarantee a certain degree of accuracy. Once again this is easy for alternating series whose
terms satisfy the conditions of the alternating series test.
EXAMPLE 10.53
6∞ n+1
How many terms in the series n=2 (−1) /(n3 ln n) ensure a truncation error of less than
10−5 ?
SOLUTION Because absolute values of terms are decreasing and have limit zero, the maxi-
mum error in truncating this alternating series when n = N is
(−1)N +2
.
(N + 1)3 ln (N + 1)
The absolute value of this error is less than 10−5 when
1
< 10−5 or
(N + 1)3 ln (N + 1)
(N + 1)3 ln (N + 1) > 105 .
A calculator quickly reveals that the smallest integer for which this is valid is N = 30. Thus,
the truncated series has the required accuracy after the 29th term (the first term corresponds to
n = 2, not n = 1).
In Section 10.7 we illustrated the use of Taylor’s remainder formula to estimate the error when
definite integrals are approximated using Taylor series. Alternating series sometimes provide
an easier alternative. We redo the example of Section 10.7 to demonstrate.
EXAMPLE 10.54
5 1/2
sin x
Approximate dx to five decimal places using the Maclaurin series for sin x .
0 x
SOLUTION Using the Maclaurin series for sin x , we obtain
5 1/2 5 1/2 ) *
sin x 1 x3 x5
dx = x− + − · · · dx
0 x 0 x 3! 5!
5 1/2 ) *
x2 x4
= 1− + − · · · dx
0 3! 5!
' (1/2
x3 x5 1 (1/2)3 (1/2)5
= x− + − ··· = − + − ···.
3 · 3! 5 · 5! 0 2 3 · 3! 5 · 5!
This is a convergent alternating series. To find a five decimal approximation, we calculate partial
sums until two successive sums agree to five decimals:
1
S1 = ,
2
(1/2)3
S 2 = S1 − = 0.493 056,
3 · 3!
(1/2)5
S 3 = S2 + = 0.493 108,
5 · 5!
(1/2)7
= 0.493 108.
S 4 = S3 −
7 · 7!
Consequently, to five decimals the value of the integral is 0.493 11.
Truncating a Series Whose Convergence Was

Established with the Integral Test
6∞
Suppose now that a series n=1 cn has been shown to converge with the integral test; that is,
the integral
5 ∞
f (x) dx
1
converges where f (n) = cn . If the series is truncated after the N th term, the error cN +1 +
cN +2 · · · is shown as the sum of the areas of the rectangles in Figure 10.30. Clearly, the sum
of these areas is less than the area under y = f (x) to the right of x = N . In other words, the
error in truncating the series with the N th term must be less than
5 ∞
f (x) dx. (10.48)
N
FIGURE 10.30 Approximating the sum of a series using the integral test
y
y = f (x)
cN+1 = f (N + 1)
cN+2 = f (N + 2)
f (2) cN+3 = f (N + 3)
f (3)
1 2 3 N N+1 N+2 N+3 x
EXAMPLE 10.55
∞
! 1
Obtain an error estimate if the series is truncated when n = 100.
n=2
n(ln n)4
SOLUTION The error cannot be larger than

5 ∞ ' (∞
1 −1 1
dx = = < 0.0035.
100 x(ln x)4 3(ln x)3 100 3(ln 100)3
When convergence of a series is established by the comparison,

6∞ limit comparison, limit ratio, or
limit root tests, we often estimate the truncation error n=N +1 cn by comparing it to something
that is summable. We illustrate this in the following two examples.
EXAMPLE 10.56
In the first paragraph of this section we indicated that e is the sum of the series
!∞
1
.
n=0
n!
Use the first 10 terms to find an approximation for e .

SOLUTION The sum of the first 10 terms is

!9
1
= 2.718 281 526.
n !
n=0
The truncation error in using this as an approximation for e is
!∞
1 1 1 1 1
= + + + + ···
n=10
n! 10! 11! 12! 13!
) *
1 1 1 1
= 1+ + + + ···
10! 11 11 · 12 11 · 12 · 13
) *
1 1 1 1
< 1+ + 2 + 3 + ··· (a geometric series)
10! 11 11 11
1 1
= (using equation 10.39b)
10! 1 − 1/11
11
= < 0.000 000 304.
10 · 10!
We may write, therefore, that
!∞
1
2.718 281 526 < < 2.718 281 830,
n=0
n!
and to six decimal places, e = 2.718 282.
EXAMPLE 10.57
6∞
How many terms in the convergent series n=1 n/[(n + 1)3n ] ensure a truncation error of less
than 10−5 ?
SOLUTION If this series is truncated after the N th term, the error is
∞
! n N +1 N +2
= 1
+ + ···
(n + 1)3n (N + 2)3N + (N + 3)3N +2
n=N +1
1 1
< + + ··· (a geometric series)
3N +1 3N +2
1
= 3N +1 (using equation 10.39b)
1
1− 3
1
= .
2 · 3N
Consequently, the error is guaranteed to be less than 10−5 if N satisfies the inequality
1
< 10−5 or
2 · 3N
ln (105 /2)
2 · 3N > 105 )⇒ N > = 9.85.
ln 3
Thus, 10 or more terms yield the required accuracy.
EXERCISES 10.13
!∞ !∞
In Exercises 1–10 verify that the sum of the series is as indicated. 1 1
∗ 19. (5 terms) ∗ 20. (10 terms)
∞ ∞ n=1
nn n=1
n2n
! 2n ! (−1)n
1. = e2 2. = sin 1 ∞
! 1 #π $
n=0
n! n=0
(2n + 1)! ∗ 21. sin (15 terms)
n=1
2n n
∞
! ∞
!
(−1)n 32n (−1)n 1
!∞ !∞
3. = cos 3 4. = −1 2n − 1 2n + 1
n=0
(2n)! n=1
n! e ∗ 22. (20 terms) ∗ 23. (20 terms)
n=2
3n + n n=2
3n + n
∞
! (−1)n 1 !∞
(−1)n
5. =− ∗ 24. (100 terms)
n=1
2 2n 5 n
n=1
∞
! (−1)n+1 22n+3
∗ 6. = −8(1 + cos 2)
n=2
(2n)! In Exercises 25–27 how many terms in the series guarantee an approx-
imation to the sum with a truncation error of less than 10−4 ?
!∞ !∞
2n 1 ∞ ∞
∗ 7. = ln 3 ∗ 8. = ln 2 ! (−1)n ! 1
n=1
n3n n=1
n2n ∗ 25. ∗ 26.
n=1
n2 n=1
n2 4n
∞
! ) *
(−1)n 1 !∞
∗ 9. = 3 sin −1 2n
32n (2n + 1)! 3 ∗ 27.
n=1
n=1
n!
∞
! n
∗ 10. =2 ∞
n=1
2n !
∗ 28. Suppose the series e−n sin2 n is truncated after the 10th term.
n=1
∗ 11. Find the Maclaurin series for Tan −1 x and use it to evaluate Obtain an error estimate by (a) using 10.48 and (b) using the fact that
e−n sin2 n < e−n . Which gives the better estimate?
!∞
(−1)n
.
n=1
2n + 1 In Exercises 29–36 evaluate the integral correct to three decimal places.
Compare the work in Exercises 29, 30, 32, and 34 to that in Exercises
∗ 12. Find the Maclaurin series for x/(1 + x 2 )2 and use it to evaluate 11–14 of Section 10.7.
5 1 5 1/ 2
∞ sin x
! n(−1)n ∗ 29. dx ∗ 30. cos (x 2 ) dx
. 0 x 0
n=1
32n 5 2 /3 5 1
1
∗ 31. dx ∗ 32. x 11 sin x dx
0 x4 + 1 −1
5 1/ 2 5 0 .3
In Exercises 13–14 approximate the sum of the series if it is truncated 1 2
after the N th term. Use 10.48 to find an error estimate. ∗ 33. √ dx ∗ 34. e−x dx
0 1 + x3 0
5 0
∞
! n2 !∞
n 1
13. , N = 10 14. , N =5 ∗ 35. ln (1 − x) dx
(n + 1)4
3 e 3n − 0 .1 x−1
n=2 n=1 5 1/ 2
1
∗ 36. dx
0 x 6 − 3x 3 − 4
In Exercises 15–16 use 10.48 to estimate the error when the series is
truncated after the N th term.
∗ 37. In determining the radiated power from a half-wave antenna, it is
∞
! !∞ ) * necessary to evaluate
1 1 1
15. , N = 100 16. sin , N = 20 5 2π
n2 +1 n2 n 1 − cos θ
n=1 n=1 dθ.
0 θ
Find a two-decimal-place approximation for this integral.
In Exercises 17–24 use the number of terms indicated to find an ap-
∗ 38. A very important function in engineering and physics is the error
proximation to the sum of the series. In each case, obtain an estimate
function erf (x) defined by
of the truncation error. 5 x
2 2
∞
! (−1)n+1
∞
! (−1)n erf (x) = √ e−t dt
17. (3 terms) 18. (20 terms) π 0
n3 3n n4
n=2 n=1 Calculate erf (1) correct to three decimal spaces.
Summary 691
∗ 39. This exercise shows that we must be very careful in predicting the (c) If En = Sn − S are the differences between the sum of the
accuracy of a result. Consider the series series and its first four partial sums, show that
) *
1 1 1 E1 = 0.000 100 2,
S = 3.125 100 1 − 0.000 090 18 1 + + 2 + 3 + ··· .
10 10 10 E2 = 0.000 010 02,
(a) Show that the sum of this series is exactly S = 3.124 999 9. E3 = 0.000 001 002,
To two decimal places, then, the value of S is 3.12.
E4 = 0.000 000 100 2.
(b) Verify that the first four partial sums of the series are
What can you say about the accuracy of S1 , S2 , S3 ,
S1 = 3.125 100 1, and S4 as approximations to S ?
(d) If S is approximated by any of S1 , S2 , S3 , or S4 to two
S2 = 3.125 009 92, decimals, the result is 3.13, not 3.12 as in part (a). Thus, in
S3 = 3.125 000 902, spite of the accuracy predicted in part (c), S1 , S2 , S3 , and
S4 do not predict S correctly to two decimal places. Do
S4 = 3.125 000 000 2. they predict S correctly to three or four decimal places?
SUMMARY
An infinite sequence of numbers is the assignment of numbers to positive integers. In most
applications of sequences, the prime consideration is whether the sequence has a limit. If the
sequence has its terms defined explicitly, then our ability to take limits of continuous functions
(limits at infinity in Chapter 2 and L’Hôpital’s rule in Chapter 4) can be very helpful. If the
sequence is defined recursively, existence of the limit can sometimes be established by showing
that the sequence is monotonic and bounded, or that it is oscillatory and convergent.
An expression of the form
∞
!
c n = c1 + c2 + · · · + c n + · · ·
n=1
is called an infinite series. We define the sum of this series as the limit of its sequence of partial
sums {Sn } , provided that the sequence has a limit. Unfortunately, for most series we cannot find
a simple formula for Sn , and therefore analysis of the limit of the sequence {Sn } is impossible.
To remedy this, we developed various convergence tests that avoided the sequence {Sn } : nth
term, comparison, limit comparison, limit ratio, limit6 root, integral, and alternating series tests.
Note the sequences that are associated with a series cn :
{Sn } sequence of partial sums for the definition of a sum;
{cn } sequence of terms for the nth -term test;
{cn /bn } sequence for the limit comparison test;
{cn+1 /cn } sequence for the limit ratio test;

√
{ n cn } sequence for the limit root test;
{|cn |} sequence for the alternating series test.
Depending on the limits of these sequences — if they exist — we may be able to infer
something about convergence of the series.
Infinite sequences and series of functions are important in applications — in particular,
power series. (As scientists, you will see other types of series: Fourier series, for example.) We
considered situations where a6power series was given and the sum was to be determined. We
saw that every power series an (x − c)n has a radius of convergence R and an associated
interval of convergence. If R = 0, the interval of convergence consists of only one point
x = c ; if R = ∞ , the power series converges for all x ; and if 0 < R < ∞ , the interval of
convergence must be one of four possibilities: c − R < x < c + R , c − R ≤ x < c + R ,

c − R < x ≤ c + R , or c − R ≤ x ≤ c + R . The radius of convergence is given by
limn→∞ |an /an+1 | or limn→∞ |an |−1/n provided that the limits exist or are equal to infinity.
If at each point in the interval of convergence of the power6 series the value of a function f (x)
is the same as the sum of the series, we write f (x) = an (x − c)n and call f (x) the sum
of the series.
We also considered situations where a function f (x) and a point c are given, and ask
whether f (x) has a power series expansion about c . We saw that there can be at most one
power series expansion of f (x) about c with a positive radius of convergence, and this series
must be its Taylor series. One way to verify that f (x) does indeed have a Taylor series about
c and that this series converges to f (x) is to show that the sequence of Taylor’s remainders
{Rn (c, x)} exists and has limit zero. Often, however, it is much easier to find Taylor series by
adding, multiplying, differentiating, and integrating known series.
When a Taylor series is truncated, Taylor’s remainder Rn (c, x) represents the truncation
error and, in spite of the fact that Rn is expressed in terms of some unknown point zn , it is
often possible to calculate a maximum value for the error. Sometimes Rn (c, x) can be avoided
altogether. For instance, if the Taylor series is an alternating series, then the maximum possible
truncation error is the value of the next term.
Power series are often used in situations that require approximations. Taylor series provide
polynomial approximations to complicated functions, and they offer an alternative to the nu-
merical techniques of Section 8.8 in the evaluation of definite integrals. Power series are also
useful in situations that do not require approximations. They are sometimes helpful in evaluating
limits, and they are the only way to solve many differential equations.
KEY TERMS
Sequence Series
Infinite sequence of numbers Term
Explicit sequences Recursive sequence
Limit of a sequence Convergent sequence
Divergent sequence Method of successive approximations
or fixed-point iteration
Taylor Remainder Formula Taylor remainders
Taylor polynomials Interval of convergence
Taylor series Maclaurin series
Power series Geometric series
Common ratio Radius of convergence
Open interval of convergence Binomial expansion
Sums of power series Increasing sequence
Nondecreasing sequence Decreasing sequence
Nonincreasing sequence Monotonic sequence
Upper bound Lower bound
Bounded sequence Oscillating sequence
Successive approximations Infinite series of numbers
Sequence of partial sums Convergent series
Divergent series Harmonic series
nth -term test Nonnegative series
Integral test p -series
Comparison test Limit comparison test
Limit ratio test Limit root test
Absolutely convergent series Conditionally convergent series
Alternating series Alternating harmonic series
Alternating series test Truncation error
REVIEW
EXERCISES
!∞ ) * !∞ ) *
In Exercises 1–6 discuss, with all necessary proofs, whether the se- 1 1 1 −1 1
19. Cos −1 20. Cos
quence is monotonic and has an upper bound, a lower bound, and a n n n2 n
n=1 n=1
limit.
' ( !∞
2 · 4 · 6 · . . . · (2n) !∞
3 · 6 · 9 · . . . · (3n)
n2 − 5n + 3 21. 22.
∗ 1. n! (2n)!
n2 + 5n + 4 n=1 n=1
1 &
∗ 2. c1 = 1, cn+1 = (1/2) cn2 + 1, n≥1 ∞
! ∞
! ) *3
n2 + 1 1
' ( 23. 24. (−1)n+1 1 +
Tan −1 (1/n) n2 + 5 n
∗ 3. n=1 n=1
n2 + 1
!∞ ∞
!
√ 1 10n
∗ 4. c1 = 7, cn+1 = 15 + cn − 2, n≥1 25. 2
sin n 26.
n=1
n n=1
53n+2
2
∗ 5. c1 = 6, cn+1 = 6 + , n≥1 ∞ ∞
cn ! ln n ! 1
27. (−1)n 28. nπ
1 n=1
n n=1
e
∗ 6. c1 = 6, cn+1 = , n≥1
5 + 4cn ∞ ∞
! 2n + 2−n ! 1
7. Use Newton’s iterative procedure and the method of successive ap- 29. 30. √ cos (nπ )
proximations to approximate the root of the equation n=1
3n n=1
n
) *2
x+5
x =
x+4 In Exercises 31–38 find the interval of convergence for the power se-
ries.
between x = 1 and x = 2, accurate to 5 decimal places.
!∞ !∞
8. For what values of k does the sequence n+1 n 1 n
∗ 31. x ∗ 32. x
n2 + 1 2
n 2n
c1 = k, cn+1 = cn2 , n≥1 n=0 n=1
∞
! !∞
1 n
converge? ∗ 33. (n + 1)3 x n ∗ 34. x
n=0 n=1
nn
9. Find an explicit definition for the sequence
% ∞ ∞ "
! 1 ! n+1
c1 = 1, cn+1 = 1+ c2 , n ≥ 1. ∗ 35. (x − 2)n ∗ 36. (x + 3)n
n
n=0
4 n
n=2
n−1
10. Use the derivative of the function f (x) = (ln x)/x to prove that ∞ ∞
! ! 2n
the sequence {ln n /n} is decreasing for n ≥ 3. ∗ 37. n3n x 2n ∗ 38. x 3n
n=1 n=1
n
In Exercises 11–30 determine whether the series converges or diverges.
In the case of a convergent series that has both positive and negative
terms, indicate whether it converges absolutely or conditionally. In Exercises 39–47 find the power series expansion of the function
about the indicated point.
∞
! ∞
!
n2 − 3n + 2 n2 + 5n + 3
11. 12. √
n3 + 4n n4 − 2n + 5 39. f (x) = 1 + x2 , about x = 0
n=1 n=1
∞
! 52n
∞
! n2 + 3 40. f (x) = ex+5 , about x = 0
13. 14.
n=1
n! n=1
n3n ∗ 41. f (x) = cos (x + π/4) , about x = 0
!∞ ∞
! ) * ∗ 42. f (x) = x ln (2x + 1) , about x = 0
(ln n)2 n+1
15. √ 16. (−1)n
n=1
n n=1
n2 ∗ 43. f (x) = sin x , about x = π/4
∞ ) * ∞ ) *
! n+1 ! 1 ∗ 44. f (x) = x/(x 2 + 4x + 3) , about x = 0
17. (−1)n 18. Cos −1
n=1
n3 n=1
n ∗ 45. f (x) = ex , about x = 3
√
∗ 46. f (x) = (x + 1) ln (x + 1) , about x = 0 ∗ 50. Find the Maclaurin series for f (x) = 1 + sin x valid for
2
−π/2 ≤ x ≤ π/2 by first showing that f (x) can be written in
∗ 47. f (x) = x 3 ex , about x = 0 the form
2 f (x) = sin (x/2) + cos (x/2).

∗ 48. How many terms in the Maclaurin series for f (x) = e−x guar-
antee a truncation error of less than 10−5 for all x in the interval
0 ≤ x ≤ 2? Why is the restriction −π/2 ≤ x ≤ π/2 necessary?
∗ 49. Find a power series solution in powers of x for the differential ∗ 51. On a calculator take the cosine of 1 (radian). Take the cosine of
equation this result, and take it again, and again, and again, . . . . What happens?
y "" − 4y = 0. Interpret what is going on.
CHAPTER
11 Vectors and Three-Dimensional
Analytic Geometry
Application Preview The figure below shows a boom OA carrying a mass M . The boom is supported by cables AB
and AC .
z C
m5 m
. 2
7
m
8
4.
B O
5.8 m 9.6 m
A
y
x
M
THE PROBLEM If tensions in the cables must not exceed 20 000 N, what is the maximum
mass that can be supported by the boom? (See Example 11.10 on page 719 for the solution.)
Chapters 1–10 dealt with single-variable calculus — differentiation and integration of func-
tions f (x) of one variable. In Chapters 11–14 we study multivariable calculus. Discussions
of three-dimensional analytic geometry and vectors in Sections 11.1–11.5 prepare the way. In
Sections 11.9–11.13 we differentiate and integrate vector functions, and apply the results to the
geometry of curves in space and the motion of objects.
11.1 Rectangular Coordinates in Space

FIGURE 11.1 Coordinate The coordinate of a point on the real line is its directed distance from the origin. Cartesian
axes and coordinate planes coordinates of a point in a plane are its directed distances from the coordinate axes. In space,
Cartesian coordinates are directed distances from three fixed planes called the coordinate planes.
z
In particular, we draw through a point O, called the origin, three mutually perpendicular lines
called the x -, y -, and z -axes (Figure 11.1). Each of the axes is coordinatized with some unit
4
distance (which need not be the same for all three axes). These three coordinate axes determine
3
2 the three coordinate planes: The xy -coordinate plane is that plane containing the x - and y -axes,
1 the yz -coordinate plane contains the y - and z -axes, and the xz -coordinate plane contains the
1O
x - and z -axes.
2 1 2 If P is any point in space, we draw lines from P perpendicular to the three coordinate
3 3 4
4 5 planes (Figure 11.2). The directed distance from the yz -coordinate plane to P is parallel to
y the x -axis, and is called the x -coordinate of P . Similarly, y - and z -coordinates are defined as
x directed distances from the xz - and xy -coordinate planes to P . These three coordinates of P ,
695
696 Chapter 11 Vectors and Three-Dimensional Analytic Geometry
FIGURE 11.2 A point as perpen- FIGURE 11.3 A point in space FIGURE 11.4 Coordinates of four spe-
dicular distances from coordinate planes as distances along coordinates axes cific points
z z z
5
(2, −3, 4) 4
x 3 (−2, 5, 3)
z
y y x
P P 2 −3
O −3 1 (1, 1, 1)
O
x y −2 1
1
z z 2 −1 2 3 4
y y 3
4 −2 5
5 y
x x −3
x (3, 4, −1)
written (x, y, z) , are called the Cartesian or rectangular coordinates of P . Note that if we
draw lines through P that are perpendicular to the axes, then the directed distances from O to
points of intersection of these perpendiculars with the axes are also the Cartesian coordinates of
P (Figure 11.3).
By either definition, each point in space has a unique ordered set of Cartesian coordinates
(x, y, z) ; conversely, every ordered triple of real numbers (x, y, z) is the set of coordinates for
one and only one point in space. For example, points with coordinates (1, 1, 1) , (2, −3, 4) ,
(3, 4, −1) , and (−2, 5, 3) are shown in Figure 11.4.
The coordinate systems in Figures 11.1–11.4 are called right-handed coordinate systems,
because if we curl the fingers on our right hand from the positive x -direction toward the positive
y -direction, then the thumb points in the positive z -direction (Figure 11.5). The coordinate
system in Figure 11.6, on the other hand, is a left-handed coordinate system, since the thumb
of the left hand points in the positive z -direction when the fingers of this hand are curled from
the positive x -direction to the positive y -direction. We always use right-handed systems in this
book, as should everyone.
Suppose we construct for any two points P1 and P2 with coordinates (x1 , y1 , z1 ) and
(x2 , y2 , z2 ) , respectively, a box with sides parallel to the coordinate planes and with line segment
P1 P2 as diagonal (Figure 11.7). Because triangles P1 AB and P1 BP2 are right-angled, we can
write
"P1 P2 "2 = "P1 B"2 + "BP2 "2
= "P1 A"2 + "AB"2 + "BP2 "2
= (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 .
FIGURE 11.5 Right-handed FIGURE 11.6 Left-handed coor- FIGURE 11.7 Distance between
coordinate system dinate system two points in space
z z z P2 (x2 , y2, z2)

P1 (x1, y1, z1)
y x A
x y x B y
11.1 Rectangular Coordinates in Space 697
In other words, the length of the line segment joining two points P1 (x1 , y1 , z1 ) and
P2 (x2 , y2 , z2 ) is
!
"P1 P2 " = (x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 . (11.1)
This is the analogue of formula 1.10 for the length of a line segment joining two points in the
xy -plane.
Just as the x - and y -axes divide the xy -plane into four regions called quadrants, the xy -,
yz -, and xz -coordinate planes divide xyz -space into eight regions called octants. The region
where x -, y -, and z -coordinates are all positive is called the first octant. There is no commonly
accepted way to number the remaining seven octants.
EXERCISES 11.1
1. Draw a Cartesian coordinate system and show the points (1, 2, 1) , ∗ 15. If P and Q in the figure below have coordinates (x1 , y1 , z1 ) and
(−1, 3, 2) , (1, −2, 4) , (3, 4, −5) , (−1, −2, −3) , (−2, −5, 4) , (x2 , y2 , z2 ) , show that coordinates of the point R midway between P
(8, −3, −6) , and (−4, 3, −5) . and Q are
" #
2. Find the length of the line segment joining the points (1, −2, 5) x1 + x2 y1 + y2 z1 + z2
, , .
and (−3, 2, 4) . 2 2 2
√ √
3. Prove that √
the triangle with vertices (2, 0, 4 2) , (3, −1, 5 2) ,
and (4, −2, 4 2) is right-angled and isosceles.
z
4. A cube has sides of length 2 units. What are coordinates of its corners
if one corner is at the origin, three of its faces lie in the coordinate planes,
and one corner has all three coordinates positive? Q (x 2, y2, z2)
R
5. Show that the (undirected, perpendicular) distances from
! a point
(x, y, z) to the!x -, y -, and z -axes are, respectively, y 2 + z2 , P (x 1, y1, z1)
√
x 2 + z2 , and x 2 + y 2 .
y
In Exercises 6–9 find the (undirected) distances from the point to (a) x
the origin, (b) the x -axis, (c) the y -axis, and (d) the z -axis.
∗ 16. (a) Find the midpoint of the line segment joining the points
6. (2, 3, −4) 7. (1, −5, −6) P (1, −1, −3) and Q(3, 2, −4) .
8. (4, 3, 0) 9. (−2, 1, −3)
(b) If the line segment joining P and Q is extended its own
10. Prove that the three points (1, 3, 5) , (−2, 0, 3) , and (7, 9, 9) are length beyond Q to a point R , find the coordinates of R .
collinear.
11. Find that point in the xy -plane that is equidistant from the points ∗ 17. The four-sided object in the figure below is a tetrahedron. If the
(1, 3, 2) and (2, 4, 5) and has a y -coordinate equal to three times its four vertices of the tetrahedron are as shown, prove that the three lines
x -coordinate. joining the midpoints of opposite edges (one of which is P Q ) meet at
12. Find an equation describing all points that are equidistant from a point that bisects each of them.
the points (−3, 0, 4) and (2, 1, 5) . What does this equation describe
z
geometrically?
√ √ √ √ √ (d, e, f )
13. (a) If ( 3 − 3, 2 + 2 3, 2 3 − 1) and (2 3, 4, 3 − 2)
are two vertices of an equilateral triangle, and if the third
vertex lies on the z -axis, find the third vertex.
(0, 0, 0)
(b) Can you find a third vertex on the x -axis?
P O
∗ 14. A birdhouse is built from a box z
1/2 m on each side with a roof y
as shown in the figure to the Q
2 (a, 0, 0)
right. If the distance from each
corner of the roof to the peak is (b, c, 0)
5 x
3/4 m, find coordinates of the 1
2
nine corners of the house. (The ∗ 18. Let A , B , C , and D be the vertices of a quadrilateral in space (not
sides of the box are parallel to 1 1 necessarily planar). Show that the line segments joining midpoints of
the coordinate planes.) 2 2 opposite sides of the quadrilateral intersect in a point that bisects each.
4 y
x
∗∗ 19. Generalize the result of Exercise 15 to prove that if a point R

"P R" r1
divides the length P Q so that = , where r1 and r2 are
"RQ" r2
positive integers, then the coordinates of R are
r1 x2 + r2 x1 r1 y2 + r2 y1 r1 z2 + r2 z1 1
z 2
x = , y = , z= .
r1 + r2 r1 + r2 r1 + r2
x
y 10
∗∗ 20. A man 2 m tall walks along the edge of a straight road 10 m wide
(figure right). On the other edge of the road stands a streetlight 8 m
high. A building runs parallel to the road and 1 m from it. If Cartesian 10 8
coordinates are set up as shown (with x - and y -axes in the plane of the
road), find coordinates of the tip of the man’s shadow when he is at the
position shown.
11.2 Curves and Surfaces

An equation involving the x - and y -coordinates of points in the xy -plane usually specifies
FIGURE 11.8 In the xy -
plane, x 2 + y 2 = 4 describes a
a curve. For example, the equation x 2 + y 2 = 4 describes a circle of radius 2 centred at
circle the origin (Figure 11.8). We now ask what is defined by an equation involving the Cartesian
coordinates (x, y, z) of points in space. For example, the equation z = 0 describes all points
y
in the xy -plane since all such points have a z -coordinate equal to zero. Similarly, y = 2
2 x 2 + y2 = 4 describes all points in the plane parallel to and 2 units to the right of the xz -plane. What does
the equation x 2 + y 2 = 4 describe? In other words, regarded as a restriction on the x -, y -, and
z -coordinates of points in space, rather than a restriction on the x - and y -coordinates of points
in the xy -plane, what does it represent? Because the equation says nothing about z , there is no
2 x restriction whatsoever on z . In other words, the z -coordinate can take on all possible values,
but x - and y -coordinates must be restricted by x 2 + y 2 = 4. If we consider those points in the
xy -plane ( z = 0) that satisfy x 2 + y 2 = 4, we obtain the circle in Figure 11.8. In space, each
of these points has coordinates (x, y, 0) , where x and y still satisfy x 2 + y 2 = 4 (Figure 11.9).
If we now take any point Q that is either directly above or directly below a point P (x, y, 0)
on this circle, it has exactly the same x - and y -coordinates as P ; only its z -coordinate differs.
Thus the x - and y -coordinates of Q also satisfy x 2 + y 2 = 4. Since we can do this for any
FIGURE 11.9 In space,
x 2 + y 2 = 4 describes a cylinder
point P on the circle, it follows that x 2 + y 2 = 4 describes the right-circular cylinder of radius
2 and infinite extent in Figure 11.9.
z By reasoning similar to that used above, we can show that the equation 2x+y = 2 describes
the plane in Figure 11.10 parallel to the z -axis and standing on the straight line 2x + y = 2,
z = 0 in the xy -plane. !
x 2 + y2 = 4
Finally, consider the equation x 2 + y 2 + z2 = 9. Since x 2 + y 2 + z2 is the distance
from the origin to a point with coordinates (x, y, z) , this equation describes all points that are
3 units away from the origin. In other words, x 2 + y 2 + z2 = 9 describes points on a sphere
Q (x, y, z) of radius 3 centred at the origin.
It appears that one equation in the coordinates (x, y, z) of points in space specifies a
surface. The shape of the surface is determined by the form of the equation. If one equation in
the coordinates (x, y, z) specifies a surface, it is easy to see what two simultaneous equations
2 specify. For instance, suppose we ask for all points in space whose coordinates satisfy both of
the equations
2 P (x, y, 0)
y x 2 + y 2 = 4, z = 1.
x By itself, x 2 + y 2 = 4 describes the cylinder in Figure 11.9. The equation z = 1 describes
all points in a plane parallel to the xy -plane and 1 unit above it. To ask for all points that
satisfy x 2 + y 2 = 4 and z = 1 simultaneously is to ask for all points that lie on both surfaces.
FIGURE 11.10 Plane with FIGURE 11.11 Curve of intersection of

equation 2x + y = 2 cylinder x 2 + y 2 = 4 and plane z = 1
z z
x 2 + y 2 = 4,
z=1
2x + y = 2
z=1
x 2 + y2 = 4
1 2
y
y
x
x
Consequently, the equations x 2 + y 2 = 4, z = 1 describe the curve of intersection of the two

surfaces — the circle in Figure 11.11.
The equation x = 0 describes the yz -plane; the equation y = 0 describes the xz -plane.
If we put the two equations together, x = 0 and y = 0, we obtain all points that lie on both
the yz -plane and the xz -plane (i.e., the z -axis). In other words, equations for the z -axis are
x = 0, y = 0.
Finally, x 2 + y 2 + z2 = 9 is the equation of a sphere of radius 3 centred at the origin, and
y = 2 is the equation of a plane parallel to the xz -plane and 2 units to the right. Together, the
equations x 2 + y 2 + z2 = 9, y = 2 describe the curve of intersection of the two surfaces —
the circle in Figure 11.12. Note that by substituting y = 2 into the equation of the sphere, we
can write alternatively that x 2 + z2 = 5, y = 2. This pair of equations is equivalent to the
original pair because all points that satisfy x 2 + y 2 + z2 = 9, y = 2 also satisfy x 2 + z2 = 5,
y = 2, and vice versa. This new pair of equations provides an alternative way of visualizing
the curve. Again y√= 2 is the plane of Figure 11.12, but x 2 + z2 = 5 describes a right-circular
cylinder of radius 5 and infinite extent around the y -axis (Figure 11.13). Our discussion has
shown that the cylinder and plane intersect in the same curve as the sphere and plane.
In summary, we have illustrated that one equation in the coordinates (x, y, z) of a point
specifies a surface; two simultaneous equations specify a curve, the curve of intersection of the
two surfaces (provided, of course, that the surfaces do intersect).
FIGURE 11.12 Curve of inter- FIGURE 11.13 Curve of

section of sphere x 2 + y 2 + z2 = 9 and intersection of cylinder x 2 + z2 = 5
plane y = 2 and plane y = 2
z z
y=2 x2 2
+z =5 y=2
x 2 + y 2 + z2 = 9 x 2 + y 2 + z2 = 9,
y=2 x 2 + z 2 = 5,
3 y=2
3 3
y y
x
x
In Chapters 2–9 we learned to appreciate the value of plotting and drawing curves in the
xy -plane. Sometimes a plot or drawing serves as a device by which we can interpret algebraic
statements geometrically (such as the mean value theorem or the interpretation of a critical point
of a function as a point where the tangent line to the graph of the function is horizontal, vertical,
or does not exist). Sometimes they play an integral part in the solution of a problem (such as
when the definite integral is used to find areas, volumes, etc.). Sometimes a plot or drawing is a
complete solution to a problem (such as to determine whether a given function has an inverse).
We will find that plotting and drawing surfaces can be just as useful for multivariable calculus
in Chapters 12–14.
One of the most helpful techniques for drawing a surface is to imagine the intersection of the
surface with various planes — in particular, the coordinate planes. From these cross-sections of
the surface, it is sometimes possible to visualize the entire surface. For example, if we intersect
the surface z = x 2 + y 2 with the yz -plane, we obtain the parabola z = y 2 , x = 0. Similarly,
the parabola z = x 2 , y = 0 is the intersection curve with the xz -plane. These curves, shown in
Figure 11.14a, would lead us to suspect that the surface z = x 2 + y 2 might be shaped as shown
in Figure 11.14b. To verify this, we intersect the surface with a plane z = k ( k a constant),
giving the curve
z = x2 + y2 or x2 + y2 = k
z=k z = k.
cross-sections of z = x 2 + y 2 with planes z = k are circles
The latter equations indicate that √
centred on the z -axis with radii k that increase as k increases. This certainly confirms the
sketch in Figure 11.14b.
FIGURE 11.14a Cross-sections of surface FIGURE 11.14b Illustration of

z = x 2 + y 2 with xz - and yz -coordinate planes surface z = x 2 + y 2
z z
z= ,x2
y=0
z = y 2,
x=0
z = x2 + y2
y y
x x
Intersections of the surface y = z + x 2 with the xy -, xz -, and yz -coordinate planes give

two parabolas and a straight line, shown in Figure 11.15a. These really do not help us visualize
the surface. If, however, we intersect the surface with planes z = k , we obtain the parabolas
y = z + x2 or y = x2 + k
z=k z = k.
These parabolas, shown in Figure 11.15b, indicate that the surface y = z + x 2 should be drawn
as in Figure 11.15c.
FIGURE 11.15a Cross-sections of FIGURE 11.15b Cross-sections of FIGURE 11.15c Illustra-

surface y = z + x 2 with coordinate planes surface y = z + x 2 with planes z = k tion of surface y = z + x 2
z z y = x 2+ k, z
y = z, z=k
x=0 y = z + x2
y = x 2,
z=0
y y y
x
x z = −x 2, x
y=0
We can sometimes “build” surfaces in much the same way that we “built” curves in single-
variable calculus. For the surface z = 1 − x 2 − y 2 , we first draw the surface z = x 2 + y 2 in
Figure 11.14b. To draw z = −(x 2 + y 2 ) , we turn z = x 2 + y 2 upside down (Figure 11.16a),
and finally we see that z = 1 − x 2 − y 2 is z = −(x 2 + y 2 ) shifted upward 1 unit (Figure
11.16b).
FIGURE 11.16a Illustration FIGURE 11.16b Illustration

of surface z = −(x 2 + y 2 ) of surface z = 1 − x 2 − y 2
z z
1
z = 1 − x 2 − y2
1
1
y y
z = − (x 2 + y2)
x x
EXAMPLE 11.1
Draw the surface defined!by each of the following equations: √
(a) z = 4x + 2y − x 2 − y 2 − 4 (b) y = 1 + x 2 + z2
SOLUTION
(a) If we square the equation, and at the same time complete squares on −x 2 + 4x and
−y 2 + 2y , we have
z2 = −(x − 2)2 − (y − 1)2 + 1,
or,
(x − 2)2 + (y − 1)2 + z2 = 1.
!
Because (x − 2)2 + (y − 1)2 + z2 is the distance from a point (x, y, z) to
(2, 1, 0) , this equation states that (x, y, z) must always be a unit distance from
(2, 1, 0) [i.e., the equation (x − 2)2 + (y − 1)2 + z2 = 1 defines a sphere of radius
1 centred at (2, 1, 0) ] (Figure 11.17a). Because the original equation requires z to be
nonnegative, the required surface is the upper half of this sphere — the hemisphere
in Figure 11.17b.
FIGURE 11.17a Sphere described by FIGURE 11.17b Hemisphere de-

!
(x − 2)2 + (y − 1)2 + z2 = 1 scribed by z = 4x + 2 y − x 2 − y 2 − 4
z z
(2, 1, 0) y (2, 1, 0) y
x x
√
(b) If we intersect the surface y = x 2 + z2 with the xy -plane, we obtain the broken
(i.e., bent) straight line y = |x| , z = 0 in Figure 11.18a. Intersections of the surface
with planes y = k ( k a constant) give
!
y = x 2 + z2 or x 2 + z2 = k 2
y =k y = k.
These√define circles of radii k in the planes y = k (Figure 11.18b). Consequently,

y = √ x 2 + z2 defines the right-circular cone in Figure 11.18c. The surface y =
1 + x 2 + z2 can now be obtained by shifting the cone 1 unit in the y -direction
(Figure 11.18d).
FIGURE 11.18a Cross- FIGURE 11.18b Cross-

√ √
section of surface y = x 2 + z2 sections of surface y = x 2 + z2
with xy -plane with planes y = k
z z
x 2 + z 2 = k 2,
y=k
y = |x|,
z=0
y y
x x
FIGURE 11.18c Cone FIGURE 11.18d Cone

√ √
described by y = x 2 + z2 described by y = 1 + x 2 + z2
z z
y = x 2 + z2
y = 1 + x 2 + z2
y y
x x
Cylinders
Suppose that l is a straight line and C is a curve that lies in some plane (the xy -plane in Figure
11.19a). A cylinder is the surface traced out by a line that moves along C always remaining
parallel to l (Figure 11.19b). The right-circular cylinder in Figure 11.9 is generated by moving a
vertical line around the circle x 2 +y 2 = 4, z = 0 in the xy -plane. Although we might not like to
think of it as such, the plane in Figure 11.10 is a cylinder. The surface in Figure 11.15c is a cylin-
der; move lines parallel to y = z , x = 0 in Figure 11.15a along the parabola y = x 2 , z = 0.
FIGURE 11.19a Definition of a cylinder using line and curve FIGURE 11.19b Visualization of a cylinder
z z
l l
y y
C Cylinder
x x
When one of the coordinates x , y , z is missing from the equation of a surface, a cylinder
results. The right-circular cylinder in Figure 11.9 is an example ( z is missing). For such
cylinders, a line parallel to the axis of the missing variable (the line x = 2, y = 0, say, for
x 2 + y 2 = 4) plays the role of l , and the cross-section of the cylinder with the plane of the
remaining two variables (circle x 2 + y 2 = 4, z = 0 in the xy -plane) plays the role of C . All
cross-sections of the cylinder with planes perpendicular to the axis of the missing variable are
identical to C .
The equation z = x 2 is free of y . Each cross-section of this surface with a plane y = k
is the parabola z = x 2 in the plane y = k . Consequently, z = x 2 is the equation for the
parabolic cylinder in Figure 11.20. The surface yz = 1, x > 0 is the hyperbolic cylinder in
Figure 11.21. All cross-sections in planes parallel to the yz -plane are hyperbolas.
FIGURE 11.20 Parabolic cylinder z = x 2 FIGURE 11.21 Hyperbolic cylinder yz = 1, x > 0
z z
2,
z=x
y=0
yz = 1
y
x
y
Quadric Surfaces
A quadric surface is a surface whose equation is quadratic in x , y , and z , the most general
such equation being
Ax 2 + By 2 + Cz2 + Dxy + Eyz + F xz + Gx + Hy + I z + J = 0. (11.2)

For the most part, we encounter quadric surfaces whose equations are of the form
Ax 2 + By 2 + Cz2 = J or Ax 2 + By 2 = I z,
or these equations with x , y , and z interchanged. Surfaces with these equations fall into nine
major classes, depending on whether the constants are positive, negative, or zero. They are
illustrated in Figures 11.22–11.30.
The names of these surfaces are derived from the fact that their cross-sections are ellipses,
hyperbolas, or parabolas. For example, cross-sections of the hyperbolic paraboloid with planes
z = k are hyperbolas x 2 /a 2 − y 2 /b2 = k . Cross-sections with planes x = k are parabolas
z = k 2 /a 2 − y 2 /b2 , as are cross-sections with planes y = k .
FIGURE 11.22 Elliptic cylinder FIGURE 11.23 Hyperbolic cylinder FIGURE 11.24 Parabolic cylinder
z z z
z= Ax 2
x 2 y2 x 2 y2
+ =1 − =1
a 2 b2 a 2 b2
−a
b
a y a y
y
x
x
x
FIGURE 11.25 Ellipsoid FIGURE 11.26 Elliptic paraboloid FIGURE 11.27 Elliptic cone
z z z
x 2 y2 z 2
+ + =1
a2 b2 c 2
c
x 2 y2 x 2 y2
z= + z2 = +
a 2 b2 a2 b2
b
a y y y
x x
x
FIGURE 11.28 Hyperbolic paraboloid FIGURE 11.29 Elliptic hyperboloid FIGURE 11.30 Elliptic
of one sheet hyperboloid of two sheets
z z z
x 2 y2
z= −
a 2 b2 x 2 y2 z 2
+ − =1
a 2 b2 c 2 z 2 x 2 y2
c − − =1
c 2 a 2 b2
b
y a y −c
y
x
x x
In applications of multiple integrals in Chapter 13, it is often necessary to project a space

curve into one of the coordinate planes and find equations for the projection. To illustrate,
consider the curve of intersection of the cylinder x 2 + z2 = 4 and the plane 2y + z = 4
(the first octant part of which is shown in Figure 11.31). Since the curve of intersection lies on
the cylinder x 2 + z2 = 4, its projection in the xz -plane is the circle x 2 + z2 = 4, y = 0.
To find its projection in the xy -plane, we eliminate z between the equations 2y + z = 4 and
x 2 + z2 = 4. The result is x 2 + (4 − 2y)2 = 4, or x 2 + 4(y − 2)2 = 4. This shows that the
curve of intersection lies on the elliptic cylinder x 2 + 4(y − 2)2 = 4, and therefore it projects
onto the ellipse x 2 + 4(y − 2)2 = 4, z = 0 in the xy -plane. The projection of the curve
in the yz -plane is that part of the line 2y + z = 4, x = 0 between the points (0, 1, 2) and
(0, 3, −2).
FIGURE 11.31 Projections of a curve in the coordinate planes
z
x 2 + z 2 = 4,
2y + z = 4
4 Projection
in yz -plane
2 2y + z = 4
x2 + z2 = 4
2
2
Projection
in xz-plane y
x
Projection
in xy -plane
Graphing calculators and computers can plot surfaces provided that the equation of the
surface is solved for z in terms of x and y . (This may not always be convenient.) We have
shown some computer-generated plots in Figures 11.32 together with values for x and y specified
in generating the plots. With the complexity of the expressions for z , drawing these surfaces by
hand would be a formidable task.
To appreciate the shape of a plotted surface, it is often necessary to vary the point in space
from which the surface is viewed. Computers usually have this ability; graphing calculators
may not.

Computer plots of surfaces
z z
y
y
x
x
z = 3 x 2 + y2 − x 2 − y2 z = x ( y − 1)2e−(x2 + y2)/4
−2 ≤ x ≤ 2, −2 ≤ y ≤ 2 −5 ≤ x ≤ 5, −5 ≤ y ≤ 5
z z
y
y
x
z = xy 2 cos x x
z = e x sin y
−5 ≤ x ≤ 5, −10 ≤ y ≤ 10 −1 ≤ x ≤ 1, −10 ≤ y ≤ 10
FIGURE 11.32e FIGURE 11.32f
z z
x y
z = sin (2x + y2 ) z = | x − y| + | x + y|
−2 ≤ x ≤ 2, −2 ≤ y ≤ 2 x
−1 ≤ x ≤ 1, −1 ≤ y ≤ 1
Drawings and plots of curves were invaluable in Chapter 7 when we applied definite integrals
to numerous geometric and physical problems. We encounter many of these applications in
Chapter 13, but applied to volumes in space rather than areas in the plane. Our ability to
visualize and draw surfaces in space proves more indispensable in these problems than plots
from computers and graphing calculators. This is especially! so when a picture contains a number
of intersecting surfaces. For example, the surfaces z = 4 − x 2 − y 2 and z = x 2 bound
volume in the first octant. The drawing in Figure 11.33a gives an excellent visualization of the
volume; the plot in Figure 11.33b is not as satisfactory. It is often advantageous first to produce
a plot of a surface and use it to render a drawing, adding whatever information is important for
the application at hand.
FIGURE 11.33a Hand drawing of two intersecting surfaces FIGURE 11.33b Computer plot of the two intersecting surfaces
z z
2
z = 4 − x2 − y2
z = x2
y
2
y x
2
x
11.3 Vectors 707
EXERCISES 11.2
In Exercises 1–35 draw the surface defined by the equation. Whenever In Exercises 46–55 find equations for projections of the curve in the
possible, confirm your drawing with a plot generated by a computer or xy -, yz -, and xz -coordinate planes. In each case draw the curve.
graphing calculator.
∗ 46. x + y = 3, 2y + 3z = 4
1. 2y + 3z = 6 2. 2x − 3y = 0
∗ 47. x + y + z = 4, 2x − y + z = 6
3. y = x 2 + 2 4. z = x 3
∗ 48. x 2 + y 2 = 4, z = 4
5. y 2 + z2 = 1 6. x 2 + y 2 + z2 = 4
∗ 49. x 2 + y 2 = 4, y = x
7. x 2 + 4y 2 = 1 8. y 2 − z2 = 4
! ∗ 50. x 2 + y 2 = 4, x = z
9. z = 2(x 2 + y 2 ) 10. x = y 2 + z2
! ∗ 51. x 2 + y 2 = 4, x + y + z = 2
11. x = 1 − y2 12. z = 2 − x
∗ 52. y 2 + z2 = 3, x 2 + z2 = 3
13. x 2 = y 2 14. x = z2 + 2
! ∗ 53. z = x 2 + y 2 , x + z = 1
15. z = y + 3 16. 4z = 3 x 2 + y 2 !
∗ 54. z = x 2 + y 2 , z = 6 − x 2 − y 2
17. x 2 − 2x + z2 = 0 18. yz = 1
19. x 2 + y 2 + (z − 1)2 = 3 20. z + 5 = 4(x 2 + y 2 ) ∗ 55. x 2 + y 2 + z2 = 1, y = x
21. x 2 + z2 = y 2 22. x 2 + z2 = y 2 + 1
23. y 2 + z2 = x 24. x 2 + y 2 + 4z2 = 1 In Exercises 56–61 find equations for the projection of the curve in the
specified plane. Draw each curve.
25. 9z2 = x 2 + y 2 + 1 26. (y 2 + z2 )2 = x + 1
∗ 56. z = x 2 − y 2 , z = 2x + 4y in the xy -plane
27. z2 + 4y 2 = 1 28. y − z2 = 0
29. x 2 − z2 = 4 30. x 2 + y 2 /4 + z2 /9 = 1 ∗ 57. x 2 + y 2 − 4z2 = 1, x + y = 2 in the xz -plane
31. z = x 2 /4 + y 2 /25 32. x 2 = z2 + 9y 2 ∗ 58. y = z + x 2 , y + z = 1 in the xy -plane

!
33. z = y 2 /16 − x 2 /4 34. x 2 + y 2 /4 − z2 /25 = 1 ∗ 59. x = 1 + 2y 2 + 4z2 , x 2 + 9y 2 + 4z2 = 36 in the yz -plane
35. z2 − 9x 2 − 16y 2 = 1 ∗ 60. z = x 2 + y 2 , z = 4(x − 1)2 + 4(y − 1)2 in the xy -plane
∗ 61. x 2 + y 2 − 2y = 0, z2 = x 2 + y 2 in the xz -plane
In Exercises 36–45 draw the curve defined by the equations.
36. x 2 + y 2 = 2, z = 4 37. x + 2y = 6, y − 2z = 3 In Exercises 62–71 draw whatever is defined by the equation or equa-
2 2 2 2 2 2
38. z = x + y , x + y = 5 39. x + y = 1, x + z = 1 tions.
!
40. z = x2 + y2, y = x 41. z + 2x 2 = 1, y = z ∗ 62. (x − 2)2 + y 2 + z2 = 0 ∗ 63. x = 0, y = 5
! √ √
42. z = 4 − x 2 − y 2 , x 2 + y 2 − 2y = 0 ∗ 64. x + y = 1, z = x ∗ 65. x + y = 15, y − x = 4
43. z = y, y = x 2 ∗ 66. z = 1 − (x 2 + y 2 )1/3 ∗ 67. z = |x|

44. x 2 + z2 = 1, y 2 + z2 = 1 ∗ 68. z = x 2 , y = z2 ∗ 69. x = ln (y 2 + z2 )
45. z = x 2 , z = y 2 ∗ 70. z = |x − y| ∗ 71. x = 2, y = 4, z2 − 1 = 0
11.3 Vectors
Physical quantities that have associated with them only a magnitude can be represented by real
numbers. Some examples are temperature, density, area, moment of inertia, speed, and pressure.
They are called scalars. There are many quantities, however, that have associated with them
both magnitude and direction, and these quantities cannot be described by a single real number.
Velocity, acceleration, and force are perhaps the most notable concepts in this category. To
represent such quantities mathematically, we introduce vectors.
DEFINITION 11.1
A vector is defined as a directed line segment.
To denote a vector we use a letter in boldface type, such as v. In Figures 11.34a–c we show
two vectors u and v along a line, three vectors u, v, and w in a plane, and three vectors u, v,
and w in space, respectively. It is customary to place an arrowhead on a vector and call this
end the tip of the vector. The other end is called the tail of the vector, and the direction of the
vector is from tail to tip. A vector then has both direction and length.
FIGURE 11.34a Vectors FIGURE 11.34b Vectors FIGURE 11.34c Vectors

along the x -axis in the xy -plane in space
v O u y z
x v v
w
x u
u
w y
x
Definition 11.1 for a vector says nothing about its point of application (i.e., where its tail
should be placed). This means that we may place the tail anywhere we wish. This suggests the
following definition for equality of vectors.
DEFINITION 11.2
Two vectors are equal if and only if they have the same length and direction. Their points
of application are irrelevant.
FIGURE 11.35 Equality of
vectors u and v but not of w and u
z
For example, vectors u and v in Figure 11.35 have exactly the same length and direction,
and are therefore one and the same. Although the vector w in the same figure is parallel to u
and v and has the same length, it points in the opposite direction and is not, therefore, the same
u v as u and v.
w
y Components of Vectors
x
We realized in Chapter 1 that to solve geometric problems, it is often helpful to represent them
algebraically. In fact, our entire development of single-variable calculus has hinged on our
ability to represent a curve by an algebraic equation and also to draw the curve described by
an equation. We now show that vectors can be represented algebraically. Suppose we denote
by PQ the vector from point P to point Q in Figure 11.36. If P and Q have coordinates
(x1 , y1 , z1 ) and (x2 , y2 , z2 ) in the coordinate system shown, then the length of PQ is
!
(x2 − x1 )2 + (y2 − y1 )2 + (z2 − z1 )2 . (11.3)
Note also that if we start at point P , proceed x2 − x1 units in the x -direction, then y2 − y1
units in the y -direction, and finally z2 − z1 units in the z -direction, we arrive at Q . In other
words, the three numbers x2 − x1 , y2 − y1 , and z2 − z1 characterize both the direction
and the length of the vector joining P to Q . Because of this we make the following agree-
ment.
11.3 Vectors 709
FIGURE 11.36 Vector from point P to point Q
P (x1, y1, z1) Q (x2, y2, z2)
x2 − x 1
z2 − z 1
x y2 − y 1
y
DEFINITION 11.3
If the tail of a vector v is at P (x1 , y1 , z1 ) and its tip is at Q(x2 , y2 , z2 ) , then v shall be
represented by the triple of numbers x2 − x1 , y2 − y1 , z2 − z1 . In such a case we enclose
the numbers in parentheses and write
v = (x2 − x1 , y2 − y1 , z2 − z1 ). (11.4)
The equal sign in 11.4 means “is represented by.” The number x2 − x1 is called the x -
component of v, y2 − y1 the y -component, and z2 − z1 the z -component. Vectors in the
xy -plane have only an x - and a y -component:
v = (x2 − x1 , y2 − y1 ),
where (x1 , y1 ) and (x2 , y2 ) are the coordinates of the tail and tip of v. Vectors along the x -axis
have only an x -component x2 − x1 , where x1 and x2 are the coordinates of the tail and tip of v.
We now have an algebraic representation for vectors. Each vector has associated with it a set
of components that can be found by subtracting the coordinates of its tail from the coordinates
of its tip. Conversely, given a set of real numbers (a, b, c) , there is one and only one vector
with these numbers as components. We can visualize this vector by placing its tail at the origin
and its tip at the point with coordinates (a, b, c) (Figure 11.37). Alternatively, we can place
the tail of the vector at any point (x1 , y1 , z1 ) and its tip at the point (x1 + a, y1 + b, z1 + c) .
It is worth emphasizing once again that the same components of a vector are obtained for
any point of application whatsoever. For example, the two vectors in Figure 11.38 are identical,
and in both cases the components (2, 2) are obtained by subtracting the coordinates of the tail
from those of the tip. What we are saying is that Definition 11.2 for equality of vectors can be
stated algebraically as follows.
THEOREM 11.1
Two vectors are equal if and only if they have the same components.
FIGURE 11.37 Tail of a vector can be placed at any point FIGURE 11.38 Two equal vectors
z (a, b, c) y
(x1 + a, y1 + b, z1 + c) 4 (1, 4)
3 (4, 2)
2
(−1, 2)
1 (2, 0)
1 2 3 4 x
x (x1, y1, z1) y
EXAMPLE 11.2
Find the components of a vector in the xy -plane that has length 5, its tail at the origin, and
FIGURE 11.39 Components
of vector with length 5 and angle π
makes an angle of π/6 radians with the positive x -axis.
6
radians with x -axis
SOLUTION Figure 11.39 illustrates that there are two such vectors, u and v. From the triangles
y shown, it is clear that
Q √
u "OP " = 5 cos(π/6) = 5 3/2 and "P Q" = "P R" = 5 sin(π/6) = 5/2.
5 √ √
Consequently, Q and R have coordinates Q = (5 3/2, 5/2) and R = (5 3/2, −5/2) , and
O 6 P
x " √ # " √ #
5 3 5 5 3 5
6 u = , , v = ,− .
5 2 2 2 2
v
EXAMPLE 11.3
Find the components of the vector in the xy -plane that has its tail at the point (4, 5) , has length
FIGURE 11.40 Components
of a vector given its length and its
3, and points directly toward the point (2, −3) .
direction
√ √
SOLUTION In Figure 11.40 "P Q" = 22 + 82 = 2 17. Because of similar triangles,
y P (4, 5) we can write that
"ST " "QR" 3(2) 3

3 = or "ST " = √ = √ .
"P S" "P Q" 2 17 17
S T
Similarly,
x "P R" 3(8) 12
"P T " = "P S" = √ = √ .
"P Q" 2 17 17
R Since "ST " and "P T " represent differences
√ in the x -√
and y -coordinates of P and S (except
Q (2, −3)
for signs), the components of PS are (−3/ 17, −12/ 17) .
Unit Vectors and Scalar Multiplication

If the x -, y -, and z -components of a vector v are (vx , vy , vz ) , often called the Cartesian
components of v, then these components represent the differences in the coordinates of its tip
and tail. But then, according to equation 11.3, the length of the vector, which we denote by |v| ,
is $
|v | = vx2 + vy2 + vz2 . (11.5)
In words, the length of a vector is the square root of the sum of the squares of its components.
DEFINITION 11.4
A vector v is said to be a unit vector if it has length equal to 1 unit; that is, v is a unit
vector if
vx2 + vy2 + vz2 = 1. (11.6)
To indicate that a vector is a unit vector, we place a circumflex ˆ above it: v̂.
11.3 Vectors 711
EXAMPLE 11.4
What is the length of the vector from (1, −1, 0) to (2, −3, −5) ?
SOLUTION Since the components of the vector are (1, −2, −5) , its length is
! √
(1)2 + (−2)2 + (−5)2 = 30.
We now have vectors, which are directed line segments, and real numbers, which are scalars.
We know that scalars can be added, subtracted, multiplied, and divided, but can we do the same
with vectors, and can we combine vectors and scalars? In the remainder of this section we show
how to add and subtract vectors and multiply vectors by scalars; in Section 11.4 we define two
ways to multiply vectors. Each of these operations can be approached either algebraically or
geometrically. The geometric approach uses the geometric properties of vectors, namely, length
and direction; the algebraic approach uses components of vectors. Neither method is suitable
FIGURE 11.41 Geometric for all situations. Sometimes an idea is more easily introduced with a geometric approach;
illustration of scalar multiplication sometimes an algebraic approach is more suitable. We choose whichever we feel expresses the
of vectors idea more clearly. But, whenever we take a geometric approach, we are careful to follow it
z up with the algebraic equivalent; conversely, when an algebraic approach is taken, we always
illustrate the geometric significance of the results.
To introduce multiplication of a vector by a scalar, consider the vectors u and v in Figure
w 11.41, both of which have their tails at the origin; v is in the same direction as u but is twice
v
as long as u. In such a situation we would like to say that v is equal to 2u and write v = 2u.
u Vector w is in the opposite direction to r and is three times as long as r, and we would like to
denote this vector by w = −3r. Both of these situations are realized if we adopt the following
r y definition for multiplication of a vector by a scalar.
x
DEFINITION 11.5
If λ > 0 is a scalar and v is a vector, then λv is the vector that is in the same direction as
v and λ times as long as v; if λ < 0, then λv is the vector that is in the opposite direction
to v and |λ| times as long as v.
This is a geometric definition of scalar multiplication; it describes the length and direction
of λv. We now show that the components of λv are λ times the components of v. In Figure
11.42 we show a box with faces parallel to the coordinate planes and λv as diagonal, and have
given vector v components (vx , vy , vz ) . From the pairs of similar triangles OAB and OCD ,
and OBE and ODF , we can write that
"OC" "CD" "OD" "DF " |λv|

= = = = = λ.
vx vy "OB" vz |v |
Hence
"OC" = λvx , "CD" = λvy , "DF " = λvz ,
where "OC" , "CD" , and "DF " are the components of λv. In other words, the components
of λv are λ times the components of v:
λv = λ(vx , vy , vz ) = (λvx , λvy , λvz ). (11.7)

FIGURE 11.42 Components of

λv are λ times the components of v
z
F
E
v
O vz
vx
A vy
C B y
x D
To multiply a vector by a scalar, then, we multiply each component by the scalar.
EXAMPLE 11.5
Find components for the unit vector in the same direction as v = (2, −2, 1) .
!
SOLUTION The length of v is |v| = (2)2 + (−2)2 + 12 = 3. According to our definition
of multiplication of a vector by a scalar, the vector 13 v must have length 1 ( 13 that of v) and the
same direction as v. Consequently, a unit vector in the same direction as v is
" #
1 1 2 2 1
v̂ = v = (2, −2, 1) = ,− , .
3 3 3 3 3
This example illustrates that a unit vector in the same direction as a given vector v is
v
v̂ = . (11.8)
|v |
EXAMPLE 11.6
Find components for the vector of length 4 in the direction opposite that of v = (1, 2, −3) .
√ √
SOLUTION Since |v| = 1+4+9= 14, a unit vector in the same direction as v is
1
v̂ = √ v.
14
The vector of length 4 in the opposite direction to v must therefore be

" # " # " #
−4 −4 4 8 12
(−4)v̂ = √ v = √ (1, 2, −3) = − √ , − √ , √ .
14 14 14 14 14
11.3 Vectors 713
With the operation of scalar multiplication, we can simplify the solution of Example 11.3. The
FIGURE 11.43 Addition
vector that points from (4, 5) to (2, −3) is v = (−2, −8) , and therefore the unit vector in this
of parallel vectors direction is
z 1 1 1
v̂ = √ (−2, −8) = √ (−2, −8) = √ (−1, −4).
4 + 64 2 17 17
u+v
The required vector of length 3 is
v
" #
3 3 12
u 3v̂ = √ (−1, −4) = − √ , − √ .
17 17 17
y
x
Addition and Subtraction of Vectors
FIGURE 11.44 Triangular In Figure 11.43 we show two parallel vectors u and v and have placed the tail of v on the tip
addition of vectors
of u. It would seem natural to denote the vector that has its tail at the tail of u and its tip at
z the tip of v by u + v. For instance, if u and v were equal, then we would simply be saying
that u + u = 2u. We use this idea to define addition of vectors even when the vectors are not
u parallel.
v
u+v DEFINITION 11.6
The sum of two vectors u and v, denoted by u + v, is the vector from the tail of u to the
tip of v when the tail of v is placed on the tip of u.
x y
Because the three vectors u, v, and u + v then form a triangle (Figure 11.44), we call this
FIGURE 11.45 Parallelo-
triangular addition of vectors.
gram addition of vectors Note that were we to place tails of u and v both at the same point (Figure 11.45), and
complete the parallelogram with u and v as sides, the diagonal of this parallelogram would also
z
represent the vector u + v. This is an equivalent method for geometrically finding u + v, and
it is called parallelogram addition of vectors.
u
Algebraically, vectors are added component by component; that is, if u = (ux , uy , uz )
and v = (vx , vy , vz ) , then
u +v u + v = (ux + vx , uy + vy , uz + vz ). (11.9)
v
To verify this we simply note that differences in the coordinates of P and Q in Figure 11.46 are
x y (ux , uy , uz ) , and differences in those of Q and R are (vx , vy , vz ) . Consequently, differences
in the coordinates of P and R must be (ux + vx , uy + vy , uz + vz ) .
It is not difficult to show (see Exercise 26) that vector addition and scalar multiplication
obey the following rules:
FIGURE 11.46 To add
vectors, add their components u + v = v + u; (11.10a)
z (u + v) + w = u + (v + w); (11.10b)
Q λ(u + v) = λu + λv; (11.10c)
u
(λ + µ)v = λv + µv. (11.10d)
P
v
u+ v If we denote the vector (−1)v by −v, then the components of −v are the negatives of those of
R
v:
x y
−v = (−1)v = (−1)(vx , vy , vz ) = (−vx , −vy , −vz ). (11.11)
This vector has the same length as v, but is opposite in direction to v (Figure 11.47). When v
is added to −v, the resultant vector has components that are all zero:
FIGURE 11.47 Vector v + (−v) = (vx , vy , vz ) + (−vx , −vy , −vz ) = (0, 0, 0).
−v has the same length as v , but
is opposite in direction to v
This vector, called the zero vector, is denoted by 0, and has the property that
z
v+0 = 0+v = v (11.12)
v for any vector v whatsoever.

To subtract a vector v from u, we add −v to u.
−v
y DEFINITION 11.7
The difference u − v between two vectors u and v is the vector
x
u − v = u + (−v). (11.13)
In Figure 11.48, u − v is determined by a triangle, and in Figure 11.49 by a parallelogram.

Alternatively, if we denote by r the vector joining the tip of v to the tip of u in Figure 11.50,
then, by triangle addition, we have v + r = u. Addition of −v to each side of this equation
gives
−v + v + r = −v + u or 0 + r = −v + u.
Thus r = u − v, and u − v is the vector joining the tip of v to the tip of u. Definition 11.7
implies that vectors are subtracted component by component:
u − v = (ux , uy , uz ) − (vx , vy , vz ) = (ux − vx , uy − vy , uz − vz ). (11.14)
FIGURE 11.48 u−v FIGURE 11.49 u−v FIGURE 11.50 u − v can

can be obtained by adding u and can be obtained by adding u and be obtained directly with triangular
−v with triangular addition −v with parallelogram addition subtraction
z z z
v
v r=u−v
v
u u
u
u−v −v y y y
x x −v u−v x
EXAMPLE 11.7
If u = (1, 1, 1) , v = (−2, 3, 0) , and w = (−10, 10, −2) , find:
4
(a) 3u + 2v − w (b) 2u − 4v + w (c) |u |v + w
|v |
11.3 Vectors 715
SOLUTION
(a)
3u + 2v − w = 3(1, 1, 1) + 2(−2, 3, 0) − (−10, 10, −2)
= (3, 3, 3) + (−4, 6, 0) + (10, −10, 2)

= (9, −1, 5)
(b)
2u − 4v + w = 2(1, 1, 1) − 4(−2, 3, 0) + (−10, 10, −2)
= (0 , 0 , 0 ) = 0
√ √ ! √
(c) Since |u| = 12 + 12 + 12 = 3 and |v| = (−2)2 + 32 = 13,
4 √ 4
|u |v + w = 3(−2, 3, 0) + √ (−10, 10, −2)
|v | 13
" #
√ 40 √ 40 8
= −2 3 − √ , 3 3 + √ , − √ .
13 13 13
Forces
We have already mentioned that quantities such as temperature, area, and density have associated
with them only a magnitude and are therefore represented by scalars. There are many quantities,
however, that have associated with them both magnitude and direction, and these are described
by vectors. The most notable of this group are forces. When we speak of a force, we mean a
push or pull of some size in some specific direction. For example, when the boy in Figure 11.51a
pulls his wagon, he exerts a force in the direction indicated by the handle. Suppose that he pulls
with a force of 10 N and that the angle between the handle and the horizontal is π/4 radians. To
represent this force as a vector F1 , we choose the coordinate system in Figure 11.51b, and make
the agreement that the length of F1 be equal to the magnitude of the force. Since F1 represents
a force of 10 N, it follows that the length of F1 is 10 units. Furthermore, because F1 makes an
angle of π/4 radians with the positive x - and y -axes, the difference
√ in the x -coordinates (and
the y -coordinates)√ tip and tail must be 10 cos(π/4) = 5 2. The components of F1 are
of its √
therefore F1 = (5 2, 5 2) . If the boy’s young sister drags her feet on the ground, then she
effectively exerts a force F2 in the negative x -direction. If the magnitude of this force is 3 N,
then its vector representation is F2 = (−3, 0) . Finally, if the combined weight of the wagon
and the girl is 200 N, then the force F3 of gravity on the wagon and its load is F3 = (0, −200) .
FIGURE 11.51a Boy pulling FIGURE 11.51b Direction of forces FIGURE 11.51c Resultant
wagon exerts a force in direction of handle exerted by boy and girl dragging their feet of forces of boy, girl, and gravity
F1 y F1
10
y 4 F2
F2 x
F3 F
In mechanics we replace the individual forces F1 , F2 , and F3 by a single force that has the same
effect on the wagon as all three forces combined. This force, called the resultant force of F1 ,
F2 , and F3 , is represented by the vector F, which is the sum of the vectors F1 , F2 , and F3 :
F = F1 + F2 + F3
√ √
= (5 2, 5 2) + (−3, 0) + (0, −200)
√ √
= (5 2 − 3, 5 2 − 200).
The magnitude of this force corresponds to the length of F,

$ √ √
|F| = (5 2 − 3)2 + (5 2 − 200)2 = 193.0,
and must therefore be 193.0 N. Its direction is shown in Figure 11.51c, where
" √ #
−1 200 − 5 2
θ = Tan √ = 1.55 radians.
5 2−3
By the x -, y -, and z -components (vx , vy , vz ) of a vector v, we mean that if we start at a

point P (Figure 11.52) and proceed vx units in the x -direction, vy units in the y -direction, and
vz units in the z -direction to a point Q , then v is the directed line segment joining P and Q . To
FIGURE 11.52 Geometric illustration FIGURE 11.53 Unit vectors

of components of a vector along the coordinate axes
z
z
v Q
P
k
vx vz j
i
vy
y
x x
y
phrase this another way, we introduce three special vectors parallel to the coordinate axes. We
define î as a unit vector in the positive x -direction, ĵ as a unit vector in the positive y -direction,
and k̂ as a unit vector in the positive z -direction. We have shown these vectors with their tails
at the origin in Figure 11.53, and it is clear that their components are
î = (1, 0, 0), ĵ = (0, 1, 0), k̂ = (0, 0, 1). (11.15)
But note, then, that we can write the vector v = (vx , vy , vz ) in the form
v = (vx , 0, 0) + (0, vy , 0) + (0, 0, vz )
= vx (1, 0, 0) + vy (0, 1, 0) + vz (0, 0, 1)
= vx î + vy ĵ + vz k̂.
In other words, every vector in space can be written as a linear combination of the three vectors î,
ĵ, and k̂ (i.e., as a constant times î plus a constant times ĵ plus a constant times k̂). Furthermore,
the constants multiplying î, ĵ, and k̂ are the Cartesian components of the vector. This result is
11.3 Vectors 717
equally clear geometrically. In Figure 11.52, we have shown the vector v from P to Q . If we
define points A and B as shown in Figure 11.54, then
v = PQ = PB + BQ = PA + AB + BQ.
But because PA is a vector in the positive x -direction and has length vx , it follows that PA = vx î.
Similarly, AB = vy ĵ and BQ = vz k̂, and therefore
v = vx î + vy ĵ + vz k̂. (11.16)
To say then that vx , vy , and vz are the x -, y -, and z -components of a vector v is to say that v
can be written in form 11.16.
Some authors refer to vx , vy , and vz as the scalar components of the vector v, and the
vectors vx î, vy ĵ, and vz k̂ as the vector components of v. By component, we always mean
scalar component.
Vectors in the xy -plane have only an x - and a y -component, and can therefore be written
in terms of î and ĵ. If v = (vx , vy ) , then we write equivalently that v = vx î + vy ĵ (Figure
11.55). Vectors along the x -axis have only an x -component and can therefore be written in the
form v = vx î.
We use this new notation in the following example.
FIGURE 11.54 Vectors in FIGURE 11.55 Vectors in

space can be expressed as a linear the xy -plane can be expressed as
combination of î , ĵ , and k̂ a linear combination of î and ĵ
y
z
v
vy j
P v Q
j
i PB BQ
PA k i vx i
A
j x
AB B
x
y
EXAMPLE 11.8
The force F exerted on a point charge q1 coulombs by a charge q2 coulombs is defined by
Coulomb’s law as
q1 q2
F = r̂ N,
FIGURE 11.56 Electrostatic 4 π $0 r 2
force of charge q2 on charge q1
where $0 is a positive constant, r is the distance in metres between the charges, and r̂ is a unit
q1 vector in the direction from q2 to q1 (Figure 11.56). When q1 and q2 are both positive charges
or both negative charges, then F is repulsive, and when one is positive and the other is negative,
q2 r̂ F is attractive. In particular, suppose that charges of 2 C and −2 C are placed at (0, 0, 0) and
r (3, 0, 0) , respectively, and a third charge of 1 C is placed at (1, 1, 1) . According to Coulomb’s
law, the 2 C charge will exert a repulsive force on the 1 C charge, and the −2 C charge will exert
an attractive force on the 1 C charge. Find the resultant of these two forces on the 1 C charge.
SOLUTION If F1 is the force exerted on the 1 C charge by the −2 C charge (Figure 11.57),
then
(1)(−2)
F1 = r̂,
4 π $0 r 2
! √
where the distance between the charges is r = (−2)2 + 12 + 12 = 6. The vector from
FIGURE 11.57 Force
on 1-C charge at (1, 1, 1) due to
(3, 0, 0) to (1, 1, 1) is (−2, 1, 1) , and therefore
2-C charge at (0, 0, 0) and
−2-C charge at (3, 0, 0) 1
r̂ = √ (−2î + ĵ + k̂).
z 6
(1, 1, 1) Consequently,
2
F2
F1 −2 1 1
y F1 = · √ (−2î + ĵ + k̂) = √ (2î − ĵ − k̂).
4π $0 (6) 6 12 6 π $0
−2
(3, 0, 0)
Similarly, the force F2 exerted on the 1 C charge by the charge at the origin is
x
(1)(2) 1 1
F2 = · √ (î + ĵ + k̂) = √ (î + ĵ + k̂).
4π $0 (3) 3 6 3 π $0
The resultant of these forces is
1 √ √ √
F = F1 + F2 = √ [(2 + 2 2)î + (2 2 − 1)ĵ + (2 2 − 1)k̂] N.
12 6 π $0
FIGURE 11.58 Resultant of n forces on a mass
z
F2
F1
F3
Fn
y
x
Suppose a number of forces F1 , F2 , . . . , Fn act on the mass M in Figure 11.58. The resultant
of these forces is F = F1 + F2 + · · · + Fn . It is a principle of statics that “the mass will remain
motionless if the sum of all forces on it is zero.” In such circumstances, the mass is said to be
in equilibrium under the action of the forces.
F = F1 + F2 + · · · + Fn = 0. (11.17a)
This is a vector equation. It is equivalent to three scalar equations obtained by invoking the
principle that vectors are equal if and only if their components are equal; that is, if F =
Fx î + Fy ĵ + Fz k̂, the equivalent to 11.17a is
Fx = 0 , Fy = 0 , Fz = 0 . (11.17b)
The following examples use this principle to advantage.

11.3 Vectors 719
EXAMPLE 11.9
Two cables AB and AC are tied together at A and attached to a vertical wall at B and C as
shown in Figure 11.59a. Determine the range of values of the magnitude of the force P for
which both cables remain taut when a mass of 100 kg hangs at A . Force P acts only in the
direction shown.
SOLUTION Suppose we establish the coordinate system in Figure 11.59b. For equilibrium
when both cables are taut, the x -components of all forces acting at A must sum to zero, as must
the y -components. If magnitudes of the tensions in AC and BC are denoted by TAC and TBC ,
then
0 = −P cos θ + TAC + TAB cos φ, 0 = P sin θ + TAB sin φ − 100g.
FIGURE 11.59a Force exerted on two cables FIGURE 11.59b Tensions in cable due to
attached to a wall by P and hanging mass force P and gravity acting on mass
B B
1m TAB 1
P P
1
3
A 1m C A TAC C x
3
100 kg 100 kg
√
Since tan θ = 3√
/4, it follows that cos θ = 4/5 and sin θ = 3/5. Furthermore, cos φ = 1/ 10
and sin φ = 3/ 10, and therefore
4P TAB 3P 3TAB
0 = − + TAC + √ , 0 = + √ − 100g.
5 10 5 10
The minimum value of P can be determined by decreasing it to the point when TAC becomes
zero;
4P TAB 3P 3TAB
0 = − +√ , 0 = + √ − 100g.
5 10 5 10
By eliminating TAB and solving for P we obtain P = 100g/3 N. By setting TAB = 0, we
obtain the maximum value of P ,
4P 3P
0 = − + TAC , 0 = − 100g.
5 5
The second of these gives P = 500g/3 N. Consequently, both cables remain taut for 100g/3 <
P < 500g/3.
EXAMPLE 11.10
Find the maximum mass that can be supported by the boom in the Application Preview (shown
Application Preview again in Figure 11.60a). Assume that the mass of the boom is negligible in comparison to the
Revisited mass.
SOLUTION We use that fact that the sum F of all forces acting on the end A of the boom
must be zero. There is the weight W (Figure 11.60b), tensions TAB and TAC , and a reaction
R by the boom itself. They are R = (0, R, 0) , W = (0, 0, −Mg) , and
" #
AB TAB (4.8, −9.6, 5.8) TAB (2.4, −4.8, 2.9)
TAB = TAB = √ = ,
|AB| 2 2
4.8 + 9.6 + 5.8 2 6.1
" #
AC TAC (−7.2, −9.6, 5) TAC (−7.2, −9.6, 5)
TAC = TAC = √ = .
|AC| 2 2
7.2 + 9.6 + 5 2 13
Hence,
TAB (2.4, −4.8, 2.9) TAC (−7.2, −9.6, 5)

F = (0, R, 0) + (0, 0, −Mg) + + .
6.1 13
FIGURE 11.60a Mass supported by a boom FIGURE 11.60b Forces acting on end of boom
z C C(−7.2, 0, 5)
m5 m z
2
7.
m
8
4. TAC
B O B(4.8, 0, 5.8)
R
5.8 m
A A(0, 9.6, 0)
9.6 m TAB
y y
x x
M W = −Mg
When we equate components of this vector to zero,
2.4TAB 7.2TAC 4.8TAB 9.6TAC 2.9TAB 5TAC

0= − , 0 = R− − , 0 = −Mg + + .
6.1 13 6.1 13 6.1 13
The first and third of these can be solved for TAB = (3/2)Mg and TAC = (130/92)Mg .
Since the tension in cable AB is larger than that in AC , we require
3Mg 40 000
≤ 20 000 &⇒ M ≤ ≈ 1359.
2 3g
Maximum mass that can be supported by the boom is 1359 kg.
EXERCISES 11.3
If u = (1, 3, 6) , v = (−2, 0, 4) , and w = (4, 3, −2) , find compo- If u = 2î + ĵ and v = −î + 3ĵ , find components of the vector in
nents for the vector in Exercises 1–10 . Exercises 11–14 and illustrate the vector geometrically.
1. 3u − 2v 2. 2w + 3v 11. u + v 12. u − v
3. w − 3u − 3v 4. v̂ 13. 2û 14. v̂ + û
5. 2ŵ − 3v 6. |v|v − 2|v̂|w
7. (15 − 2|w|)(u + v) 8. |3u|v − |−2v|u In Exercises 15–24 find the Cartesian components for the spatial vector
v−w described. In each case, draw the vector.
9. |2u + 3v − w|ŵ 10.
|v + w| 15. From (1, 3, 2) to (−1, 4, 5)
11.3 Vectors 721
16. With length 5 in the positive x -direction and (2, −1, −2) .
17. With length 2 in the negative z -direction ∗ 32. Newton’s universal law of gravitation states that the force of at-
traction F, in newtons, exerted on a mass of m kilograms by a mass of
18. With tail at (1, 1, 1) , length 3, and pointing toward the point M kilograms is
(1, 3, 5) GmM
F= r̂,
r2
19. With positive y -component, length 1, and parallel to the line
where G = 6.67 × 10 −11 is a constant, r is the distance in metres
through (1, 3, 6) and (−2, 1, 4)
between the masses, and r̂ is a unit vector in the direction from m to M .
20. In the same direction as the vector from (1, 0, −1) to (3, 2, −4) If point masses, each of 5 kg, are situated at (5, 1, 3) and (−1, 2, 1) ,
but only half as long what is the resultant force on a mass of 10 kg at (2, 2, 2) ?
21. With positive and equal x - and y -components, length 10, and z -
∗ 33. Illustrate geometrically the triangle inequality for vectors |u + v|
component equal to 4
≤ |u| + |v| . Prove the result algebraically.
∗ 34. Determine to the nearest newton the tensions in the cables AC and
22. Has its tail at the origin, makes angles of π/3 and π/4 radians
BC in the figure below.
with the positive x - and y -axes, respectively, and has length 5/2
23. From (1, 3, −2) to the midpoint of the line segment joining
B
(2, 4, −3) and (1, 5, 6)
24. Has its tail at the origin, makes equal angles with the positive
coordinate axes, has all positive components, and has length 2 C
25. If P , Q , and R are the points with coordinates (3, 2, −1) , A 200 kg
(0, 1, 4) , and (6, 5, −2) , respectively, find coordinates of a point S
in order that PQ = RS . ∗ 35. Determine to the nearest newton the tensions in the cables AC and
BC in the figure below.
26. Prove that vector addition and scalar multiplication have the prop-
erties in equations 11.10. A B
π/4 5π/36
27. Draw all spatial vectors of length 1 that have equal x - and y -
components and tails at the origin.
C
28. Draw all spatial vectors of length 2 that have their tails at the origin θ = π/3
and make an angle of π/4 with the positive z -axis. F = 500 N
∗ 36. For the cables in Exercise 35, it is known that the maximum allow-
∗ 29. If u = 3î + 2ĵ − 4k̂ and v = î + 6ĵ + 5k̂ , find scalars λ and
able tension in cable AC is 600 N and that in BC is 750 N. Determine
ρ so that the vector w = 5î − 18ĵ − 32k̂ can be written in the form the maximum force |F| that may be applied at C , and the corresponding
w = λu + ρ v . angle θ .
∗ 30. Find a vector T of length 3 along the tangent line to the curve ∗ 37. A 200-kg crate is supported by the rope-and-pulley arrangement
y = x 2 at the point (2, 4) (figure below). in the figure below. Determine the magnitude and direction of the force
F that should be exerted on the free end of the rope.
y
T 0.75 m
(2, 4)
y = x2 2.4 m
x F
∗ 31. Use Coulomb’s law (see Example 11.8) to find the force on a charge
of 2 C at the origin due to equal charges of 3 C at the points (1, 1, 2) 200 kg
∗ 38. A 16 kg, horizontal, triangular plate ABC is supported by three (Hint: See equation 7.43.)
wires in the figure below. Determine the tension in each wire.
F
D
B
L L
600 mm
h
A C
mm C
200
mm
200
20
0m
B m Vectors u, v, and w are said to be linearly dependent if there exist three
40
0m scalars a , b , and c , not all zero, such that a u + bv + cw = 0 . If this
m equation can only be satisfied with a = b = c = 0, the vectors are
said to be linearly independent. In Exercises 42–45 determine whether
A the vectors are linearly dependent or linearly independent.
∗ 39. A container with weight 360 N is supported by cables, AB and
∗ 42. u = (1, 1, 1), v = (2, 1, 3), w = (4, 2, 6)
AC in the figure below. Knowing that F = 60î N, determine the force
P = P ĵ that must be applied at A to maintain the configuration. What ∗ 43. u = (1, 1, 1), v = (2, 1, 3), w = (1, 6, 4)
are the tensions in the cables?
∗ 44. u = (−1, 3, −5), v = (2, 4, −1), w = (3, 11, −7)
z
220 mm
C
∗ 45. u = (4, 2, 6), v = (1, 3, −2), w = (7, 1, 4)
160 mm D
B 240 mm ∗ 46. Use vectors to show that the line segment joining the midpoints
of two sides of a triangle is parallel to the third side and its length is
120 mm y one-half the length of the third side (figure below).
x
A 480 mm z
P
F A
D E
∗ 40. A 200-kg mass is hung by means of two cables, AB and AC , B
which are attached to the top of a vertical wall (figure below). Deter- C
mine the magnitude of a horizontal force F perpendicular to the wall
that will hold the weight in place. What are the tensions in the cables? x
y
10 m C
∗ 47. Use vectors to show that the medians of a triangle (figure below)
8m all meet in a point with coordinates
1.2 m
B " #
A F x1 + x2 + x3 y1 + y2 + y3 z1 + z2 + z3
, , .
200 kg 2m 3 3 3
12 m
z
A (x1, y1, z1 )
∗ 41. Two bars, AB and BC , are pinned at B as well as at each of the
ends A and C (figure follows). Initially each bar is of length L and
point B is at a distance h above the line AC . The bars are identical,
each having cross-sectional area A and Young’s modulus E . A vertical
force with magnitude F is applied at B . Show that the displacement B (x2 , y2 , z2 )
y of B is related to F by the equation
C (x3, y3, z3)
% &
2AE L
F = (h − y) ! −1 . x
L y 2 − 2hy + L2 y
11.3 Vectors 723
∗ 48. If n point masses mi are located at points (xi , yi ) in the xy -plane,

equations 7.31 and 7.32 define the centre of mass (x, y) of the system. z
Show that these two scalar equations are represented by the one vector 80 mm
240 mm P
equation
y
n
' B
M(x, y) = mi ri ,
i=1 x A
where ri is the vector joining the origin to the point (xi , yi ) . F
∗∗ 53. Two wires AC and BC are attached to the top of pole CO in

∗ 49. If u = ux î + uy ĵ and v = vx î + vy ĵ , show that every vector w
the figure below. The force exerted by the pole is vertical, and the
in the xy -plane can be written in the form w = λu + ρ v provided
that ux vy − uy vx *= 0. z
C 2500 N
∗ 50. Two identical springs with constant k are joined at point C in the Wire Pole
figure below and have their ends fixed at A and B . In this position the
springs are unstretched and uncompressed. A 3 Wire
O
(a) If a weight W is attached to their join, and slowly lowered
until the system is in equilibrium, find an equation relating
6
W , L , k , and the distance y from C to D . x y
(b) Find an approximation for W in terms of y when y is very 9 B
small compared to L .
2500 N force applied at C is in the negative x -direction. Determine
the tensions in the wires and the vertical force exerted by the pole.
C
A L L B ∗∗ 54. In the left figure below, the xy -plane is the interface between two
materials that both transmit light. If a ray of light strikes the surface in
y a direction defined by the unit vector û = (ux , uy , uz ) , then some of
the light is reflected along a vector v̂ , and some is refracted along ŵ .
D The three vectors û , v̂ , and ŵ all lie in a plane that is perpendicular to
the xy -plane.
W (a) If the angle of incidence i in the right figure below is equal
to the angle of reflection φ , find components for v̂ in terms
of those of û .
∗∗ 51. In the figure below, two springs (with constants k1 and k2 and
unstretched lengths l ) are fixed at points A and B . They are joined (b) If the angle of refraction θ is related to the angle of in-
to a sleeve that slides along the x -axis. Find the resultant force of the cidence by n1 sin i = n2 sin θ , where n1 and n2 are the
springs on the sleeve at any point between O and C . indices of refraction of the two materials, find components
of ŵ .
y
L z z
A B Reflected
ray
k1 k2
l l u u xy -plane
Incident i
y v n1
ray v
O C x x n2
w
w
Refracted
∗∗ 52. In the following figure, collars A and B are connected by a 440- ray
mm wire and slide freely on frictionless rods, one along the y -axis, and
the other parallel to the z -axis and passing through the x -axis 240 mm
from the origin. If a force F of magnitude 450 N is applied to collar A , ∗∗ 55. Vectors v1 , v2 , . . . , vn are drawn from the centre of a regular
determine the force P required to keep collar B at the position shown. n -sided polygon in the plane to each of its vertices. Show that the sum
What is the tension in the wire? of these vectors is the zero vector.
11.4 Scalar and Vector Products

The Scalar Product of Vectors
In Section 11.3 we learned how to multiply a vector by a scalar and how to add and subtract
vectors. The next question naturally is: Can vectors be multiplied? The answer is yes, and in
fact we define two products for vectors, one of which yields a scalar and the other a vector. The
first is defined algebraically as follows.
DEFINITION 11.8
The scalar product (dot product or inner product) of two vectors u and v with Cartesian
components (ux , uy , uz ) and (vx , vy , vz ) is defined as
u · v = ux v x + uy v y + uz v z . (11.18)
If u and v have only x - and y -components, Definition 11.8 reduces to
u · v = ux v x + uy v y , (11.19)
and if they have only x -components, it becomes
u · v = ux vx . (11.20)
It is straightforward to check that
î · î = ĵ · ĵ = k̂ · k̂ = 1, (11.21a)
î · ĵ = ĵ · k̂ = î · k̂ = 0, (11.21b)
and that the scalar product is commutative and distributive:
u · v = v · u, (11.22a)
u · (λv + ρ w) = λ(u · v) + ρ(u · w). (11.22b)
EXAMPLE 11.11
If u = (−2, 1, 3) and v = (3, −2, −1) , evaluate each of the following:
(a) u · v; (b) 3u · (2u − 4v).
SOLUTION
(a) u · v = (−2)(3) + (1)(−2) + (3)(−1) = −11
(b) 3u · (2u − 4v) = (−6, 3, 9) · (−16, 10, 10)
= (−6)(−16) + (3)(10) + (9)(10)

= 216
By taking the scalar product of a vector v = (vx , vy , vz ) with itself, we obtain

√
v · v = vx2 + vy2 + vz2 = |v|2 ⇐⇒ |v| = v · v. (11.23)
Because Definition 11.8 for the scalar product of two vectors u and v is phrased in terms of
the components of u and v, and these components depend on the coordinate system used, it
follows that this definition also depends on the fact that we have used Cartesian coordinates.
Were we to use a different set of coordinates (such as polar coordinates), then the definition
of u · v in terms of components in that coordinate system might be different. For this reason
we now find a geometric definition for the scalar product (which is therefore independent of
coordinate systems).
THEOREM 11.2
If two nonzero vectors u and v are placed tail to tail, and θ is the angle between them
(0 ≤ θ ≤ π ) , then
u · v = |u||v| cos θ. (11.24)
FIGURE 11.61 Proof that PROOF The cosine law applied to the triangle in Figure 11.61 gives
u · v = |u||v| cos θ
|QR|2 = |PQ|2 + |PR|2 − 2|PQ||PR| cos θ
Q
v u−v or
|u − v|2 = |v|2 + |u|2 − 2|v||u| cos θ.
P u R Consequently,
1( )
|u||v| cos θ = | u | 2 + |v | 2 − |u − v | 2 ,
2
and if (ux , uy , uz ) and (vx , vy , vz ) are the Cartesian components of u and v, then
1
|u||v| cos θ = {(u2x + u2y + u2z ) + (vx2 + vy2 + vz2 )
2
− [(ux − vx )2 + (uy − vy )2 + (uz − vz )2 ]}
= u x v x + uy v y + uz v z
= u · v.
An immediate consequence of this result is the following.
COROLLARY 11.2.1
Two nonzero vectors u and v are perpendicular if and only if
u · v = 0. (11.25)
For example, the vectors u = (1, 2, −1) and v = (4, 2, 8) are perpendicular since
u · v = (1)(4) + (2)(2) + (−1)(8) = 0.
Expression 11.24 doesn’t just tell us whether or not the angle between two vectors is π/2
radians; it can be used to determine the angle between any two nonzero vectors u and v, when
they are placed tail to tail. We first solve 11.24 for cos θ :
u·v
cos θ = . (11.26)
|u||v|
Since principal values of the inverse cosine function lie between 0 and π , precisely the range
for θ , we can write that " #
−1 u · v
θ = Cos . (11.27)
|u||v|
EXAMPLE 11.12
Find the angle between the vectors u = (2, −3, 1) and v = (5, 2, 4) .
SOLUTION According to formula 11.27,
" # " #
u·v 10 − 6 + 4
θ = Cos−1 = Cos−1 √ √
|u||v| 4 + 9 + 1 25 + 4 + 16
" #
8
= Cos−1 √ = 1.25 radians.
3 70
EXAMPLE 11.13
Find the angle between the lines x + 2y = 3 and 4x − 3y = 5 in the xy -plane.
FIGURE 11.62 The angle SOLUTION Since the slope of x + 2y = 3 is −1/2, a vector along this line is u = (−2, 1) .
between lines in the xy -plane Similarly, a vector along 4x − 3y = 5 is v = (3, 4) . If θ is the angle between these vectors,
using vectors along the lines and therefore between the lines (Figure 11.62), then
y " # " # " #
4x − 3y = 5 −1 u·v −1 −6 + 4 −1 −2
θ = Cos = Cos √ √ = Cos √ = 1.75 radians.
|u||v| 5 25 5 5
x + 2y = 3
The acute angle between the lines is π − 1.75 = 1.39 radians. Formula 1.60 gives the same
3 v result.
2 u
5 3 x
4
−5
3
The Vector Product of Vectors
In many applications we need to find a vector perpendicular to two given vectors. For instance,
consider finding a vector r = (a, b, c) perpendicular to two given vectors u = (ux , uy , uz )
and v = (vx , vy , vz ) . The corollary to Theorem 11.2 requires a , b , and c to satisfy
0 = r · u = aux + buy + cuz ,
0 = r · v = avx + bvy + cvz .
When we solve these equations, we find that there is an infinite number of solutions all repre-
sented by
a = s(uy vz − uz vy ), b = s(uz vx − ux vz ), c = s(ux vy − uy vx ),

where s is any real number. In other words, any vector of the form
r = s(uy vz − uz vy , uz vx − ux vz , ux vy − uy vx )
is perpendicular to u = (ux , uy , uz ) and v = (vx , vy , vz ) . When we choose s = 1, the

resulting vector is called the vector product of u and v.
DEFINITION 11.9
The vector product (cross product or outer product) of two vectors u and v with
Cartesian components (ux , uy , uz ) and (vx , vy , vz ) is defined as
u × v = (uy vz − uz vy )î + (uz vx − ux vz )ĵ + (ux vy − uy vx )k̂. (11.28)

To eliminate the need for memorizing the exact placing of the six components of u and v in
this definition, we borrow the notation for determinants from linear algebra. A brief discussion
of determinants and their properties is given in Appendix B. We set up a 3 × 3 determinant with
î, ĵ, and k̂ across the top row and the components of u and v across the second and third rows:
* *
* î ĵ k̂ *
* *
* ux uy uz *.
*v vy v *
x z
In actual fact, this is not a determinant, since three entries are vectors and six are scalars. If
we ignore this fact, and apply the rules for expansion of a 3 × 3 determinant along its first row
(namely, î times the 2 × 2 determinant obtained by deleting the row and column containing î,
minus ĵ times the 2 × 2 determinant obtained by deleting the row and column containing ĵ,
plus k̂ times the 2 × 2 determinant obtained by deleting the row and column containing k̂), we
obtain * *
* î ĵ k̂ * * * * * * *
* * * u uz * * u uz * * u uy *
* ux uy uz * = * y * î − * x * ĵ + * x * k̂.
*v vy vz *
vy vz vx vz vx vy
x
But the value of a 2 × 2 determinant is
* *
*a b*
* * = ad − bc.
c d
Consequently,
* *
* î ĵ k̂ *
* *
* ux uy uz * = (uy vz − uz vy )î + (uz vx − ux vz )ĵ + (ux vy − uy vx )k̂,
*v vy v *
x z
and this is the same as the right side of equation 11.28. We may therefore write, as a memory-
saving device, that
* *
* î ĵ k̂ *
* *
u × v = * ux uy uz *, (11.29)
*v vy vz *
x
so long as we evaluate the right side using the general rules for expansion of a determinant along
its first row.
For example, if u = (1, −1, 2) and v = (2, 3, −5) , then
* *
* î ĵ k̂ *
* *
u × v = * 1 −1 2 * = (5 − 6)î − (−5 − 4)ĵ + (3 + 2)k̂ = −î + 9ĵ + 5k̂.
*2 3 −5
*
It is straightforward to verify that
î × î = ĵ × ĵ = k̂ × k̂ = 0, (11.30a)
î × ĵ = k̂, ĵ × k̂ = î, k̂ × î = ĵ, (11.30b)
and that the cross product is anticommutative and distributive:
u × v = −v × u , (11.31a)
u × (λv + ρ w) = λ(u × v) + ρ(u × w). (11.31b)
Our preliminary analysis indicated that u × v is perpendicular to both u and v. The

following theorem relates the length of u × v to lengths of u and v.
THEOREM 11.3
If θ is the angle between two vectors u and v, then
|u × v| = |u||v| sin θ. (11.32)
PROOF Since θ is an angle between 0 and π , sin θ must be positive, and we can write from
equation 11.26 that
+ " #2
! u·v 1 !
sin θ = 1− cos2 θ = 1− = | u | 2 | v | 2 − (u · v ) 2 .
|u||v| |u||v|
Consequently,
!
|u||v| sin θ = | u | 2 | v | 2 − (u · v ) 2 .
If u = (ux , uy , uz ) and v = (vx , vy , vz ) , then
|u|2 |v|2 sin2 θ = (u2x + uy2 + u2z )(vx2 + vy2 + vz2 ) − (ux vx + uy vy + uz vz )2
= u2x (vx2 + vy2 + vz2 ) + uy2 (vx2 + vy2 + vz2 )
+ u2z (vx2 + vy2 + vz2 ) − (u2x vx2 + u2y vy2 + u2z vz2
+ 2 ux vx uy vy + 2 ux vx uz vz + 2 uy vy uz vz )
= (u2y vz2 − 2uy vy uz vz + u2z vy2 ) + (u2x vz2 − 2ux vx uz vz + u2z vx2 )
+ (u2x vy2 − 2ux vx uy vy + u2y vx2 )
= (uy vz − uz vy )2 + (ux vz − uz vx )2 + (ux vy − uy vx )2

= |u × v | 2
or
|u × v| = |u||v| sin θ.
We now know that u × v is perpendicular to u and v, and has length |u||v| sin θ , where θ
FIGURE 11.63 There are
two directions perpendicular to two
is the angle between u and v. Figure 11.63 illustrates that there are only two directions that are
given vectors, one in the opposite perpendicular to u and v, and one is the negative of the other. Let us denote by ŵ the unit vector
direction to the other along that direction which is perpendicular to u and v and is determined by the right-hand rule
(curl the fingers of the right hand from u toward v and the thumb points in direction ŵ).
v
We now show that u × v always points in the direction determined by the right-hand rule
w
(i.e., in direction ŵ rather than −ŵ). To see this, we place u and v tail to tail and establish
a coordinate system with this common point as origin and the positive x -axis along u (Figure
u 11.64a). Let the plane determined by u and v be the xy -plane. In this coordinate system, u has
only an x -component, u = ux î(ux > 0) , and v has only x - and y -components, v = vx î +vy ĵ.
–w
The cross product of u and v is therefore
* *
* î ĵ k̂ *
* *
u × v = * ux 0 0 * = ux vy k̂.
*v vy 0
*
x
For v in Figure 11.64a, vy is clearly positive and therefore ux vy , the component of u × v, is

also positive. But then u × v is indeed determined by the right-hand rule since ŵ = k̂.

The direction of u × v is always determined by the right-hand rule
z z z
vy
u × v = ux vy k u × v = ux vy k vx
v
vy
k=w k
k=w
v vx
v j
j vy vx
i i w = −k
i j vy y
u u
y y
u
v u × v = ux vy k
vx x
x x
When the tip of v lies in the second quadrant of the xy -plane (Figure 11.64b), u × v is
once again in the positive z -direction, the direction of ŵ. When the tip of v is in the third or
fourth quadrant (Figure 11.64c), vy < 0, and u × v therefore has a negative z -component. But
in this case ŵ = −k̂, and the direction of u × v is once again determined by the right-hand
rule.
What we have now established is the following coordinate-free definition for the vector
product of two vectors u and v:
u × v = (|u||v| sin θ )ŵ. (11.33)
The unit vector ŵ defines the direction of u × v and the factor |u||v| sin θ is its length.
The fact that the vector product u × v is perpendicular to both u and v makes it a powerful
tool in many applications. We shall see some of them in Sections 11.5 and 11.6.
EXAMPLE 11.14
Find the cross product of the vectors u = î + 2ĵ and v = 3î − 2ĵ + k̂, and show that the result
is indeed perpendicular to u and v.
FIGURE 11.65 Cross-product
of two vectors is perpendicular to both SOLUTION Using formula 11.29,
z * *
* î ĵ k̂ *
* *
v u × v = * 1 2 0 * = 2î − ĵ − 8k̂.
* 3 −2 1 *
u
y
x We have shown the vectors in Figure 11.65. We can check that u × v is perpendicular to u and
v using 11.25:
u×v
(u × v) · u = (2î − ĵ − 8k̂) · (î + 2ĵ) = 2(1) + (−1)(2) = 0,
(u × v) · v = (2î − ĵ − 8k̂) · (3î − 2ĵ + k̂) = 2(3) + (−1)(−2) + (−8)(1) = 0.
EXERCISES 11.4
If u = 2î − 3ĵ + k̂ , v = ĵ − k̂ , and w = 6î − 2ĵ + 3k̂ , evaluate the Find direction angles for the vectors in Exercises 37–40 .
scalar or find the components of the vector in Exercises 1–10 .
37. (1, 2, −3) 38. (0, 1, −3)
1. u · v 2. (v · w)u
39. (−1, −2, 6) 40. (−2, 3, 4)
3. (2u − 3v) · w 4. 2î · û ∗ 41. Verify the results in equations 11.31.
5. |2u|v · w 6. (3u − 4w) · (2î + 3u − 2v) ∗ 42. Show that the cross product is not associative; that is, in general
(105u + 240v) · (105u + 240v) u × (v × w) *= (u × v) × w .
7. w · ŵ 8.
|105u + 240v|2 ∗ 43. (a) If u *= 0, show that if the conditions u · v = u · w and
9. |u − v + k̂|(ĵ + w) · k̂ 10. u · v + v · w − (u + w) · v u × v = u × w are both satisfied, then v = w .
(b) Show that if one of the conditions in part (a) is satisfied but
If u = (3, 1, 4) , v = (−1, 2, 0) , and w = (−2, −3, 5) , evaluate the other is not, then v cannot be equal to w .
the scalar or find the components of the vector in Exercises 11–20 . ∗ 44. The scalar u · v × w is called the scalar triple product of u , v ,
and w .
11. v × w 12. (−3u) × (2v)
13. u · (v × w) 14. û × ŵ (a) Find u · v × w if u = (6, −1, 0) , v = (1, 3, 4) , and
15. ((3u) × w) + (u × v) 16. u × (3v − w) w = (−2, −1, 4) .
w×u (b) Prove that u · v × w = u × v · w .
17.
|u × v| (c) Show that |u · v × w| can be interpreted as the volume of
18. (u × w) − (u × v) + (u × (2u + v)) the parallelepiped with u , v , and w as coterminal sides in
19. (u × v) × w 20. u × (v × w) the figure below.
(d) Verify that three nonzero vectors u , v , and w all lie in the
In Exercises 21–24 determine whether the vectors are perpendicular. same plane if and only if u · v × w = 0.
21. (1, 2), (3, 5) 22. (2, 4), (−8, 4)
23. (1, 3, 6), (−2, 1, −4) 24. (2, 3, −6), (−6, 6, 1) z
w
In Exercises 25–30 find the angle between the vectors.
25. (3, 4), (2, −5) 26. (1, 6), (−4, 7)
27. (4, 2, 3), (1, 5, 6) 28. (3, 1, −1), (−2, 1, 4) v
29. (2, 0, 5), (0, 3, 0) 30. (1, 3, −2), (−2, −6, 4)
u
In Exercises 31–33 find components for the vector. x y
31. Perpendicular to the vectors (1, 3, 5) and (−2, 1, 4)

32. Perpendicular to the y -axis and the vector joining the points
In Exercises 45–46 prove the identity.
(2, 4, −3) and (1, 5, 6)
33. Perpendicular to the triangle with vertices (−1, 0, 3) , (5, 1, 2) , ∗ 45. (u × v) · (w × r) = (u · w)(v · r) − (u · r)(v · w)
and (−6, 2, 4)
34. Verify the results in equations 11.22. ∗ 46. u × (v × w) = (u · w)v − (u · v)w
∗ 35. Use equations 11.23 to prove sin A sin B sin C
∗ 47. Use vectors to prove the sine law = = for the
|u + v|2 + |u − v|2 = 2|u|2 + 2|v|2 . a b c
triangle in the figure below. Hint: Note that PQ + QR + RP = 0 .
This is often called the parallelogram law. Why? Cross this equation with PQ and take lengths.
∗ 36. The angles between a vector v = (vx , vy , vz ) and the vectors î , z
ĵ , and k̂ are called direction angles α , β , and γ of v . Show that P
% & " # c
A
−1 v · î vx Q
α = Cos = Cos−1 ,
|v| |v| B
b a
% & " #
v · ĵ vy C
β = Cos−1 = Cos−1 ,
|v| |v| R
% & " # x y
v · k̂ vz
γ = Cos−1 = Cos−1 . ∗ 48. Show that the vector (|v|u + |u|v) / ||u|v + |v|u| is a unit vector
|v| |v|
that bisects the angle between u and v .
11.5 Planes and Lines

Planes
A plane in space can be characterized in various ways: by two intersecting lines, by a line and a
point not on the line, or by three noncollinear points. For our present purposes, we use the fact
that given a point P (x0 , y0 , z0 ) and a vector (A, B, C) (Figure 11.66), there is one and only
one plane through P that is perpendicular to (A, B, C) .
FIGURE 11.66 A plane is characterized by a point in it and a vector perpendicular to it
z
(A, B, C )
Q (x, y, z)
P (x0, y0, z0)
x
y
To find the equation of this plane we note that if Q(x, y, z) is any other point in the plane,
then vector PQ = (x −x0 , y −y0 , z−z0 ) lies in the plane. But PQ must then be perpendicular
to (A, B, C) ; hence, by the corollary to Theorem 11.2,
(A, B, C) · (x − x0 , y − y0 , z − z0 ) = 0.
Because this equation must be satisfied by every point (x, y, z) in the plane (and at the
same time is not satisfied by any point not in the plane), it must be the equation of the plane
through P and perpendicular to (A, B, C) . If we expand the scalar product, we obtain the
equation of the plane in the form
A(x − x0 ) + B(y − y0 ) + C(z − z0 ) = 0.
This result is worth stating as a theorem.
THEOREM 11.4
The equation for the plane through the point (x0 , y0 , z0 ) perpendicular to the vector
(A, B, C) is
A(x − x0 ) + B(y − y0 ) + C(z − z0 ) = 0. (11.34)
Equation 11.34 can also be written in the form
Ax + By + Cz + D = 0, (11.35)
where D = −(Ax0 + By0 + Cz0 ) , and this equation is said to be linear in x , y , and z . We
have shown then that every plane has a linear equation, and the coefficients A , B , and C of x ,
y , and z in the equation are the components of a vector (A, B, C) that is perpendicular to the
plane. Instead of saying that (A, B, C) is perpendicular to the plane Ax + By + Cz + D = 0,
we often say that (A, B, C) is normal to the plane or that (A, B, C) is a normal vector to
the plane.
EXAMPLE 11.15
Find an equation for the plane through the point (4, −3, 5) and normal to the vector (4, −8, 3) .
SOLUTION According to 11.34, the equation of the plane is
4(x − 4) − 8(y + 3) + 3(z − 5) = 0 or 4x − 8y + 3z = 55.
EXAMPLE 11.16
Determine whether the planes x + 2y − 4z = 10 and 2x + 4y − 8z = 11 are parallel.
SOLUTION Normal vectors to these planes are N1 = (1, 2, −4) and N2 = (2, 4, −8) .
Since N2 = 2N1 , the normal vectors are in the same direction, and therefore the planes are
parallel.
Lines
In Section 11.2 we indicated that space curves can be described by two simultaneous equations
in x , y , and z , and that such a representation describes the curve as the intersection of two
surfaces. A straight line results when the surfaces are planes. We shall discuss this further, but
we prefer to begin our discussion of straight lines with vectors much as we did for planes. A
straight line in space is characterized by a point on it and a vector parallel to it. There is one
and only one line * through the point P (x0 , y0 , z0 ) and in the direction v = (vx , vy , vz ) in
Figure 11.67.
If Q(x, y, z) is any point on this line, then the vector r joining the origin to Q has
FIGURE 11.67 A line is
characterized by a point on it and
components r = (x, y, z) . Now this vector can be expressed as the sum of r0 , the vector from
a vector along it 0 to P , and PQ:
r = r0 + PQ.
z
*
PQ But PQ is in the same direction as v; therefore, it must be some scalar multiple of v (i.e.,
P(x0, y0, z0) Q(x, y, z) PQ = t v). Consequently, the vector joining O to any point on * can be written in the form
v r
r0 r = r0 + t v, (11.36)
0 for an appropriate value of t . Because the components of r = (x, y, z) are also the coordinates
of the point (x, y, z) on * , this equation is called the vector equation of the line * through
y (x0 , y0 , z0 ) in direction v. If we substitute components for r, r0 , and v, then
x
(x, y, z) = (x0 , y0 , z0 ) + t (vx , vy , vz ) = (x0 + tvx , y0 + tvy , z0 + tvz ).
Since two vectors are equal if and only if corresponding components are identical, we can write
that
x = x0 + vx t, (11.37a)
y = y0 + vy t, (11.37b)
z = z0 + vz t. (11.37c)
These three scalar equations are equivalent to vector equation 11.36; they are called parametric
equations for line * . They illustrate once again that a line in space is characterized by a point
(x0 , y0 , z0 ) on it and a vector (vx , vy , vz ) along it. Each value of t substituted into 11.37 yields
a point (x, y, z) on * , and conversely, every point on * is represented by some value of t . For
instance, t = 0 yields P , and t = 1 gives the point at the tip of v in Figure 11.67.
If none of vx , vy , and vz is equal to zero, we can solve equations 11.37 for t and equate
the three expressions to obtain
x − x0 y − y0 z − z0
= = . (11.38)
vx vy vz
These are called symmetric equations for the line * through (x0 , y0 , z0 ) parallel to v =
(vx , vy , vz ) . There are only two independent equations in 11.38, which therefore substantiates
our previous result that a curve (in this case, a line) can be described by two equations in x , y ,
and z . We could, for instance, write
vx (y − y0 ) = vy (x − x0 ) and vz (y − y0 ) = vy (z − z0 )
or
v y x − vx y = v y x 0 − vx y 0 , v y z − vz y = v y z 0 − v z y 0 .
Since the first of these is linear in x and y and the second is linear in y and z , each describes a
plane. The line has been described as the curve of intersection of two planes.
EXAMPLE 11.17
Find, if possible, vector, parametric, and symmetric equations for the line through the points
(−1, 2, 1) and (3, −2, 1) .
SOLUTION A vector along the line is (3, −2, 1) − (−1, 2, 1) = (4, −4, 0) , and so too is
(1, −1, 0) . A vector equation for the line is
r = (−1, 2, 1) + t (1, −1, 0) = (t − 1, −t + 2, 1).
Parametric equations are therefore
x = t − 1, y = −t + 2, z = 1.
Because the z -component of every vector along the line is zero, we cannot write full symmetric
equations for the line. By eliminating t between the x - and y -equations, however, we can write
y−2
x+1 = , z = 1.
−1
If we set x + 1 = 2 − y , z = 1, or x + y = 1, z = 1, we represent the line as the intersection
of the planes x + y = 1 and z = 1 (Figure 11.68).
FIGURE 11.68 Line represented as intersection of two planes
z
x+y=1
(−1, 2, 1)
(3, −2, 1)
1
1 z=1
y
x
EXAMPLE 11.18
Find the equation of the plane containing the origin and the line 2x + y − z = 4, x + z = 5.
SOLUTION We can easily find two more points on the plane. For instance, if we set x = 0,
then the equations of the line require z = 5 and y = 9; and if we set z = 0, then x = 5 and
y = −6. Thus P (0, 9, 5) and Q(5, −6, 0) , as well as O(0, 0, 0) , are points on the plane. It
follows then that OP = (0, 9, 5) and OQ = (5, −6, 0) are vectors in the plane, and a vector
normal to the plane is
* *
* î ĵ k̂ *
* *
OP × OQ = * 0 9 5 * = (30, 25, −45).
* 5 −6 0 *
The vector (6, 5, −9) is also normal to the plane, and the equation of the plane is
0 = (6, 5, −9) · (x − 0, y − 0, z − 0) = 6x + 5y − 9z.
EXAMPLE 11.19
Find symmetric equations for the line x + y − 2z = 6, 2x − 3y + 4z = 10.
FIGURE 11.69 A vector SOLUTION To find symmetric equations, we require a vector parallel to the line and a point on
along the line of intersection of it. By setting x = 0 and solving y − 2z = 6, −3y + 4z = 10, we obtain y = −22, z = −14.
two planes can be obtained by Consequently, (0, −22, −14) is a point on the line. To find a vector along the line, we could
crossing normal vectors to the find another point on the line, say, by setting z = 0 and solving x + y = 6, 2x − 3y = 10 for
plane
x = 28/5, y = 2/5. A vector along the line is therefore (28/5, 2/5, 0) − (0, −22, −14) =
x + y − 2z = 6 (28/5, 112/5, 14) , and so is (5/14)(28/5, 112/5, 14) = (2, 8, 5) .
Alternatively, we know that (1, 1, −2) and (2, −3, 4) are vectors that are normal to the
planes x + y − 2z = 6 and 2x − 3y + 4z = 10, and a vector along the line of intersection
of the planes must be perpendicular to both of these vectors (Figure 11.69). Consequently, a
vector along the line of intersection is
(2, −3, 4)
* *
* î ĵ k̂ *
(1, 1, −2) * *
*1 1 −2 * = (−2, −8, −5).
* 2 −3 4
*
Symmetric equations for the line are therefore

2x − 3y + 4z = 10
x y + 22 z + 14
= = .
2 8 5
EXERCISES 11.5
In Exercises 1–10 find the equation for the plane. 6. Containing the lines (x − 1)/6 = y/8 = (z + 2)/2 and (x +
1)/3 = (y − 2)/4 = z + 5
1. Through the point (1, −1, 3) and normal to the vector (4, 3, −2) 7. Containing the line x − y + 2z = 4, 2x + y + 3z = 6 and the
2. Through the point (2, 1, 5) and normal to the vector joining (2, 1, 5) point (1, −2, 4)
and (4, 2, 3) 8. Containing the lines x + 2y + 4z = 21, x − y + 6z = 13 and
x = 2 + 3t , y = 4, z = −3 + 5t
3. Containing the points (1, 3, 2) , (−2, 0, −2) , (1, 4, 3)
9. Containing the lines 3x + 4y = −6, x + 2y + z = 2 and 2y +
4. Containing the point (2, −4, 3) and the line (x − 1)/3 = (y + 3z = 19, 3x − 2y − 9z = −58
5)/4 = z + 2
10. Containing the line x + y − 4z = 6, 2x + 3y + 5z = 10 and (a)
5. Containing the lines x = 2y = (z + 1)/4 and x = t , y = 2t , perpendicular to the xy -plane, (b) perpendicular to the xz -plane, and
z = 6t − 1 (c) perpendicular to the yz -plane.
11. The acute angle between two intersecting planes is defined as the ∗ 22. Does the line (x − 3)/2 = y − 2 = (z + 1)/4 lie in the plane
acute angle between their normals. Find the acute angle between the x − y + 2z = −1?
planes x − 2y + 4z = 6 and 2x + y = z + 4. ∗ 23. Show that a vector perpendicular to the line Ax + By + C = 0
in the xy -plane is (A, B) .
In Exercises 12–21 find vector, parametric, and symmetric (if possible) ∗ 24. Show that if a plane has nonzero intercepts a , b , and c on the x -,
equations for the straight line. y -, and z -axes, then its equation is x/a + y/b + z/c = 1.
∗ 25. Find equations for the four faces of the tetrahedron in Exercise 17
12. Through the point (1, −1, 3) and parallel to the vector (2, 4, −3)
of Section 11.1.
13. Through the point (−1, 3, 6) and parallel to the vector (2, −3, 0) ∗ 26. Find equations for the nine planes forming the sides, bottom, and
14. Through the points (2, −3, 4) and (5, 2, −1) roof of the birdhouse in Exercise 14 of Section 11.1.
15. Through the points (−2, 3, 3) and (−2, −3, −3) ∗ 27. Verify that the equation of the plane passing through the three
16. Through the points (1, 3, 4) and (1, 3, 5) points P1 (x1 , y1 , z1 ) , P2 (x2 , y2 , z2 ) , and P3 (x3 , y3 , z3 ) can be writ-
17. Through the point (1, −3, 5) and parallel to the line ten in the form P1 P · P1 P2 × P1 P3 = 0, where P (x, y, z) is any
point in the plane.
x y−2 z+4 ∗ 28. The region Sxy in the xy -plane bounded by the straight lines x =
= =
5 3 −2 0, x = 1, y = 0, and y = 1 is a rectangle with unit area.
18. Through the point (2, 0, 3) and parallel to the line x = 4 + t , (a) Show that the region in the plane y =√z that projects onto
y = 2, z = 6 − 2t Sxy is also a rectangle, but with area 2.
∗ 19. Through the point of intersection of the lines
(b) Show that the region in the plane x + y − 2√ z = 0 that
x−1 y+4 z−2 x−1 y+4 z−2 projects onto Sxy is a parallelogram with area 6/2.
= = and = = ,
2 −3 5 6 3 4 (c) Generalize the results of parts (a) and (b) to show that if S
is the area in a plane Ax + By + Cz + D = 0 ( C *= 0)
and parallel to the line joining the points (1, 3, −2) and (2, −2, 1)
that projects onto Sxy , then the area of S is sec γ , where
∗ 20. 2x − y = 5, 3x + 4y + z = 10
γ is the acute angle between k̂ and the normal to the given
∗ 21. Through the point (−2, 3, 1) and parallel to the line x + y =
plane.
3, 2x − y + z = −2
11.6 Geometric Applications of Scalar and Vector Products
FIGURE 11.70 Components Components of Vectors in Arbitrary Directions

of a vector can be found by drawing
perpendiculars from the tip of the vector We have defined what is meant by Cartesian components (vx , vy , vz ) of a vector v. They are
to the coordinate axes
scalars that multiply î, ĵ, and k̂ so that v = vx î + vy ĵ + vz k̂. They can be represented in terms
z of scalar products as
vx = v · î, vy = v · ĵ, vz = v · k̂. (11.39)
vz
Geometrically, they can be found by drawing v, î, ĵ, and k̂ all at the origin and dropping
perpendiculars from the tip of v to the x -, y -, and z -axes (Figure 11.70). We now generalize
k v our definition of a component along an axis to define the component of a vector in any direction
i whatsoever.
vx j vy
y
DEFINITION 11.10
x
To define the component of a vector v in a direction u, we place v and u tail to tail at
a point P and draw the perpendicular from the tip of v to the line containing u (Figure
11.71a). The directed distance P R is called the component of v in the direction u. If R
is on the same side of P as the tip of u, P R is taken as positive; and if R is on that side
of P opposite to the tip of u, P R is negative.
FIGURE 11.71a Component In Figure 11.71a the component of v in direction u is positive, and in Figure 11.71b the
of vector v in direction u component is negative. Note that the length of u is irrelevant; it is only the direction of u that
v determines the component of v in the direction u. If θ is the angle between u and v, then
||P R|| = |v| cos θ.

u
P R The right side of this equation looks very much like the scalar product of v and a vector that
makes an angle θ with v. It lacks only the length of this second vector. Clearly, u is a vector
FIGURE 11.71b Situation that makes an angle θ with v, but we cannot write ||P R|| = |v||u| cos θ , since the length of u
where component of v in direction of need not be 1. If, however, û is the unit vector in the same direction as u, then we can write
u is negative
||P R|| = |v||û| cos θ. (11.40)
R
In other words, we have the following theorem.
v
THEOREM 11.5
The component of a vector v in a direction u is v · û, where û is the unit vector in
P
direction u.
u This result agrees with equations 11.39 for the x -, y -, and z -components of v.
EXAMPLE 11.20
Find components of v = (1, 2, −3) in directions specified by the following vectors:
(a) u = (−1, 3, 4) (b) u = (1, 3, −2) (c) u = (4 , 4 , 4 )
SOLUTION Components of v in these directions are

(−1, 3, 4) 7
(a) v · û = (1, 2, −3) · √ = −√
26 26
(1, 3, −2) 13
(b) v · û = (1, 2, −3) · √ =√
14 14
(4 , 4 , 4 )
(c) v · û = (1, 2, −3) · √ =0
4 3
This last result means that v is perpendicular to (4, 4, 4) .
Distances Between Points, Lines, and Planes

There are three distances between geometric objects in a plane — between points, from a point
to a line, and between parallel lines. They were all covered in Section 1.3. Distance formulas
1.10 and 1.16 handle the first two situations. Distance between parallel lines can be calculated
by finding a point on one line and using 1.16 for the distance from this point to the other line.
In space, there are six distances of interest — point to point, point to line, point to plane, line
to line, line to plane, and plane to plane. Formula 11.1 gives distance between points. We can
develop a formula for the distance from a point (x1 , y1 , z1 ) to a plane Ax + By + Cz + D = 0
very similar to formula 1.16 for distance from a point to a line in the xy -plane. Let R be the foot
of the perpendicular from a point P (x1 , y1 , z1 ) to the plane in Figure 11.72, and Q(x, y, z)
be any other point in the plane. The length of vector PR is the distance from P to the plane,
and this length is the component of PQ in direction PR. Hence, |PR| = PQ · P ,R. A normal
vector to the plane is (A, B, C) , and therefore
(A, B, C)
,
PR = ±√ .
A2 + B 2 + C 2
FIGURE 11.72 Distance form a point to a plane
P (x1, y1, z1)
Q (x, y, z)
R
Ax + By + Cz + D = 0
Thus,
- .
±(A, B, C)
|PR| = PQ · √ .
A2 + B 2 + C 2
We choose the + or − to guarantee a positive result. This can also be accomplished by inserting
absolute values, * *
* (A, B, C) *
|PR| = **PQ · √ *.
A +B +C *
2 2 2
Since PQ = (x − x1 , y − y1 , z − z1 ) ,
* *
* (A, B, C) *
*
|PR| = *(x − x1 , y − y1 , z − z1 ) · √ *
A2 + B 2 + C 2 *
|Ax + By + Cz − (Ax1 + By1 + Cz1 )|
= √ .
A2 + B 2 + C 2
Because Q is in the plane, we can replace Ax + By + Cz with −D , and the formula then
becomes
|Ax1 + By1 + Cz1 + D|
√ . (11.41)
A2 + B 2 + C 2
Notice, as we suggested, the similarity of this formula to 1.16.
EXAMPLE 11.21
Find the distance from the point (1, 2, 5) to the plane x + y + 2z = 4.
SOLUTION According to 11.41, the distance is
|1 + 2 + 2(5) − 4| 9
√ = √ .
1+1+4 6
With distance formula 11.41, it is straightforward to find the distance between two parallel
planes, or between a plane and a line parallel to the plane.
EXAMPLE 11.22
Show that the line x = 3 + t , y = 1 − 2t , z = 4 + 3t and the plane x + 2y + z = 6 are
parallel, and find the distance between them.
SOLUTION A vector along the line is v = (1, −2, 3) and a vector normal to the plane is
N = (1, 2, 1) . Since v · N = 0, the vectors are perpendicular, and this confirms that the line is
parallel to the plane. [We could also confirm this by substituting from the equations of the line
into the plane, 6 = (3 + t) + 2(1 − 2t) + (4 + 3t) = 9, a contradiction. The line and plane
do not therefore intersect.] To find the distance between the line and plane, we find the distance
from (3, 1, 4) , a point on the line, to the plane:
|3 + 2(1) + 4 − 6| 3
√ = √ .
1+4+1 6
For the distance between parallel planes, choose a point on one plane and find the distance to
the other plane. There is also a formula (see Exercise 31). This leaves two distances — point
to line and line to line. We illustrate each by example.
EXAMPLE 11.23
Find the distance from the point (1, 3, 6) to the line
y−2 z−4
*: x−1 = = .
2 3
FIGURE 11.73 Distance SOLUTION Clearly, Q(1, 2, 4) is a point on the line, and therefore the required distance d
from a point to a line is the component of PQ in the direction PR (Figure 11.73). By equation 11.40, then, d =
z |PQ · P, ,
R| . To find PR we need a vector in the direction PR. Since v = (1, 2, 3) is a vector
P (1, 3, 6) along * , the vector
* *
d
* î ĵ k̂ *
* *
v × PQ = * 1 2 3 * = −î + 2ĵ − k̂
Q (1, 2, 4) R * 0 −1 −2 *
is perpendicular to both PQ and v. It now follows that a vector along PR is
v v × PQ * *
* î ĵ k̂ *
* *
(v × PQ) × v = * −1 2 −1 * = 8î + 2ĵ − 4k̂.
* 1 2 3
*
y
x
Finally, then,
(8, 2, −4) (4, 1, −2)
,
PR = √ = √
64 + 4 + 16 21
and * * √
* (4, 1, −2) ** 21
*
d = *(0, −1, −2) · √ *=
21
.
7
Cross products can also be used to find the distance from a point to a line (see Exercises 29 and
30). The above method parallels that for other distances in this section; it uses scalar products.
EXAMPLE 11.24
Find the (shortest) distance between the lines
x−1 y+3
*1 : = = z − 4 and *2 : x = −1 + t, y = 2t, z = 3 − 2t.
2 3
FIGURE 11.74 Shortest distance between nonintersecting lines
P R (1, −3, 4)
Q S (−1, 0, 3)
SOLUTION There will be a point P on *1 and a point Q on *2 such that line segment P Q
is perpendicular to *1 and *2 (Figure 11.74). The length of this line segment is the shortest
distance between the lines. There is no problem finding a vector in the same direction as PQ:
cross vectors (2, 3, 1) and (1, 2, −2) along *1 and *2 , respectively,
* *
* î ĵ k̂ *
* *
*2 3 1 * = (−8, 5, 1).
* 1 2 −2 *
Now comes the hard part to visualize. If we take any point on *1 , say R(1, −3, 4) , and
any point on *2 , say S(−1, 0, 3) , then the component of vector RS along PQ is the length
of PQ. (If you are not convinced, consider the following argument. It is easier to see that the
,
component of PS along PQ is the length of PQ so that |PQ| = |PS · PQ| . But PS = PR + RS,
and therefore,
|PQ| = |(PR + RS) · P, ,
Q| = |PR · P ,
Q + RS · PQ|.
,
Since PR is perpendicular to P ,
Q, it follows that PR · PQ = 0, and
|PQ| = |RS · P,
Q|.
,
The right side is the component of RS along PQ.) Consequently,
* *
* (−8, 5, 1) ** √
*
|PQ| = *(−2, 3, −1) · √ = 10.
90
*
This procedure fails when the lines are parallel, but in this case, we can pick a point on one line
and find the distance to the other line as in Example 11.23.
Areas of Triangles and Parallelograms

Vector products provide a simple way to find areas of triangles and parallelograms.
THEOREM 11.6
If A , B , and C are vertices of a triangle, then
1
area of +ABC = |AB × AC|. (11.42)
2
PROOF Area of triangle ABC in Figure 11.75 is

1 1 1
|AC||BD| = |AC||AB| sin θ = |AB × AC|.
2 2 2
FIGURE 11.75 Area of a triangle in FIGURE 11.76 Area of a

terms of vectors representing two sides triangle with given vertices
z z
B
B (2, −3, 2) C (4, 1, 5)
A
A (1, 1, 1)
D
C
y
y
x x
For the triangle with vertices A(1, 1, 1) , B(2, −3, 2) , and C(4, 1, 5) in Figure 11.76,
* *
* î ĵ k̂ *
* *
AB × AC = * 1 −4 1 * = (−16, −1, 12).
*3 0 4
*
The area of the triangle is, therefore,
1 1√ 1√
FIGURE 11.77 Area of a
|(−16, −1, 12)| = 256 + 1 + 144 = 401.
2 2 2
parallelogram
The following corollary is a direct consequence of Theorem 11.6.
z
B D
COROLLARY 11.6.1
The area of a parallelogram with coterminal sides AB and AC (Figure 11.77) is
A
C area = |AB × AC|. (11.43)
y
x
EXERCISES 11.6
In Exercises 1–4 find the area of the triangle with given vertices. 10. Along the line x = 1 − t , y = 3 + 2t , z = 4 − 3t in the direction
of increasing x
1. (1, 0), (4, 2), (2, 6)
2. (−1, 0, 3), (5, 1, 2), (−6, 2, 4)
3. (1, 1, 1), (−3, 4, −2), (−1, −2, 3) In Exercises 11–28 find the distance indicated.
4. (1, 2, 3), (3, 5, 10), (−3, −4, −11) 11. From the point (3, 2) to the line 2x + 3y = 18
12. Between the lines x + 4 = y and 3x + 7 = 3y
In Exercises 5–6 find the area of the parallelogram with given vertices.
13. From the point (1, 3, 4) to the plane x + y − 2z = 0
5. (1, 2, 3), (4, 3, 7), (−1, 3, 6), (2, 4, 10) 14. From the point (−2, 3, −5) to the plane 2x + y + 4z = 6
6. (1, −2, 4), (3, 5, 7), (4, 6, 8), (2, −1, 5)
15. Between the line x = 1 − t , y = 2 + 3t , z = 4 − 2t and the
plane 2x + 4y + 5z = 10
In Exercises 7–10 find the component of v = (1, −4, 3) in the direc- 16. Between the line x + y + z = 6, 2x − y + z = 3 and the plane
tion specified. 2x − 7y − z = 5
7. (1, 2, −3) 17. From the line x − 1 = 3(y + 4) = −z − 1 to the plane 2x −
8. In the direction from (−1, 2, 3) to (4, −3, 2) 3y + z = 4
9. Perpendicular to the plane x + y + 2z = 4 in the direction of 18. From the line x = 1 − 6t , y = 2 + 4t , z = −t to the plane
increasing z x + y − 2z = 1
19. Between the planes 2x + 3y − z = 15 and 4x + 6y − 2z = 7 z = −6 − 3t form a triangle and find its area.
20. Between the planes x − y + 2z = 4 and 3x − 3y + 6z = 10
∗ 21. From the point (1, 2, −3) to the line x = 2(y + 1) = (z − 4)/2 In Exercises 34–35 verify that v̂ and ŵ are perpendicular, and then
∗ 22. From the point (3, −2, 0) to the line x = t , y = 3 − 2t , z = 4 +t find scalars λ and ρ so that u = λv̂ + ρ ŵ .
√ √ √ √
∗ 23. From the point (1, −1, 2) to the line x+y−z = 2, 2x+3y+z = 4 ∗ 34. u = (2, 1) ; v̂ = (1/ 2, 1/ 2) , ŵ = (1/ 2, −1/ 2)
∗ 24. From the point (1, 3, 3) to the line x = 2 − t , y = 1 + 2t , z = 3 √ √
∗ 35. u = 3î − 2ĵ ; v̂ = (î − 2ĵ)/ 5, ŵ = (2î + ĵ)/ 5
∗ 25. Between the lines (x−1)/2 = (y+3)/3 = 4−z and x = −1+t ,
y = 2t , z = 3 − 2t
∗ 26. Between the lines x = t , y = 3t − 1, z = 1 + 2t and x = 2t + 1, In Exercises 36–37 verify that û , v̂ , and ŵ are mutually perpendicular,
y = 1 − t , z = 4 + 2t and then find scalars λ , ρ , and µ so that r = λû + ρ v̂ + µŵ .
√ √
∗ 27. Between the lines x + y − z = 4, 2x − z = 4 and x = ∗ 36. r = (1,√3, −4) ; û = (2, 1, 0)/ 5, v̂ = (−1, 2, 3)/ 14, ŵ =
(y + 1)/2 = (z − 1)/3 (3, −6, 5)/ 70
∗ 28. Between the lines x +y +z = 2, x −y +z = 3 and 2x −y +z = √ √
∗ 37. r = 2î −√k̂ ; û = (î + ĵ + k̂)/ 3, v̂ = (î + ĵ − 2k̂)/ 6,
3, y − 4z = 5
ŵ = (î − ĵ)/ 2
∗ 29. Show that the distance from point P (1, 3, 6) to the line x − 1 =
(y − 2)/2 = (z − 4)/3 in Figure 11.73 can also be found as follows:
|PR| = |PQ| sin θ = |PQ||v̂| sin θ = |PQ × v̂|, ∗ 38. If u = (3, 2) , v = (1, −3) , w = (6, 2) , verify that v and w are
perpendicular, and find scalars λ and ρ so that u = λv + ρ w . Be
where θ is the angle between PQ and line * . sure to recognize that v and w are not unit vectors.
∗ 30. Use the technique of Exercise 29 to find the distance in Exercise ∗ 39. If u = (1, 0, 1) , v = (1, 1, −1) , w = (−1, 2, 1) , and r =
24. (−2, −3, 4) ,verify that u , v , and w are mutually perpendicular, and
∗ 31. Prove that the (undirected) distance between two parallel planes find scalars λ , ρ , and µ so that r = λu + ρ v + µw . As in Exercise
Ax + By + Cz + D1 = 0 and Ax + By + Cz + D2 = 0 is 38, take into account that u , v , and w are not unit vectors.
|D1 − D2 | ∗ 40. Find the equation of a plane normal to î − 2ĵ + 3k̂ and 2 units
√ . from the point (1, 2, 3) .
A2 + B 2 + C 2
∗ 41. If a , b , and c (all positive constants) are the intercepts of a plane
with the x -, y -, and z -axes, and p is the length of the perpendicular
∗ 32. Show that the lines joining the midpoints of the sides of any quadri- from the origin to the plane, show that
lateral form a parallelogram.
∗ 33. Show that the three lines (x − 4)/3 = (y − 8)/4 = (z+ 7)/(−4) ; 1 1 1 1
= + + .
(x + 5)/3 = (y + 2)/2 = (z + 1)/2; and x = 1, y = 5 + t , p2 a2 b2 c2
11.7 Physical Applications of Scalar and Vector Products
Work
In Section 7.4, we described work as the product of force and distance. Now that we have
FIGURE 11.78 Work represented forces by vectors, we can be more precise. In particular, if a particle moves along
done by a force along a line the line in Figure 11.78 from P to Q , then vector PQ represents its displacement. If a constant
force F acts on the particle during this motion, then the work done by F is defined as
F F
F F
F W = F · PQ. (11.44)
Q It is important to keep in mind exactly when this definition of work can be used: For
constant forces acting along straight lines, and by constant F, we mean that F is constant in
P both magnitude and direction. Note that when F and PQ are both in the same direction, then
the angle between the vectors is zero. In this case,
W = |F||PQ| cos(0) = |F||PQ|,

and this is essentially the equation dealt with in Section 7.4. When F and PQ are not in the
same direction,
W = |F||PQ| cos θ.
Since |F| cos θ is the component of F along PQ, this equation simply states that when F and
PQ are not in the same direction, |PQ| should be multiplied by the component of F in direction
PQ. The component of F perpendicular to the displacement does no work.
EXAMPLE 11.25
If the boy in Figure 11.79 pulls the wagon handle with a force of 10 N at an angle of π/4 radians
with the horizontal, how much work does he do in walking 20 m in a straight line?
FIGURE 11.79 Work
done by boy pulling wagon SOLUTION The force F exerted by the boy has magnitude |F| = 10, and points in a direction
that makes an angle of π/4 radians with the displacement vector. If d is the displacement vector,
then " #
π 200 √
W = F · d = |F||d| cos θ = 10(20) cos = √ = 100 2 J.
4 2
4
When motion is along a straight line, but F is not constant in direction or magnitude or both,
we must use integration. The following example illustrates such a situation.
EXAMPLE 11.26
The spring in Figure 11.80 is fixed at A and moves the sleeve frictionlessly along the rod from
B to C . If the spring is unstretched when the sleeve is at C , find the work done by the spring.
FIGURE 11.80 Work done by a spring pulling a sleeve along a line
y
1 m
2
D
x
B x dx C
1m
SOLUTION If we set up a coordinate system as shown, then at position D the force F exerted
by the spring on the sleeve has magnitude
/0 1
1 1
|F | = k (1 − x)2 + − ,
4 2
where k is the spring constant. Since the spring is always stretched during motion, the direction
of F is along the vector DA. Clearly, then, F changes in both magnitude and direction as the
sleeve moves from B to C . For a small displacement dx at position D , the amount of work
done by F is (approximately)
/0 1
1 1 1−x
F · (dx î) = |F| dx(cos θ ) = k (1 − x)2 + − dx !
4 2 (1 − x)2 + 1/4
/ 1
1/2
= k(1 − x) 1 − ! dx.
(1 − x)2 + 1/4
Total work done by the spring as the sleeve moves from B to C must therefore be
2 / 1
1
1
W = k(1 − x) 1 − ! dx
0 2 (1 − x)2 + 1/4
2 1/ 1
1−x
=k 1−x− ! dx
0 2 (1 − x)2 + 1/4
3 41
x2 1!
=k x− + (1 − x)2 + 1/4
2 2 0
k √
= (3 − 5) J.
4
Moments
Moments play a fundamental role in structural engineering. We encountered them in Section
7.7 in the context of centres of mass, but lacking vectors we could not treat them fully. (There
was a hint of vectors when we defined the first moment of an object as its mass multiplied by
some directed distance, directed distances now being associated with vectors.) We discuss first
moments of forces here, rather than first moments of masses. Forces were present in Section
7.7, the force of gravity on the masses, but by pushing vectors and forces into the background,
we concentrated on the balance concept. Other applications of moments require a vectorial
approach.
When a force is to produce a desired effect, where to apply the force may be just as important
as the size of the force. For example, it is easier to insert screws using a screwdriver with a large
handle than a small one; a long prybar can be more effective in removing spikes than a short
one; it is easier to spin a roulette wheel by pulling on its edge rather than somewhere close to
its centre. The direction to apply a force is also important. When spinning a roulette wheel,
we don’t pull away from the centre of the wheel (or toward the centre); we pull in a direction
perpendicular to the line joining the centre of the wheel and the point of application of the force.
In turning a revolving door, the best direction is perpendicular to the plane of the door. These
ideas are captured in the following definition.
DEFINITION 11.11
The moment of a force F, applied at a point Q , about a point P (Figure 11.81) is defined
as
M = r × F, where r = PQ. (11.45)
The magnitude of M is |M| = |r||F| sin θ , making it clear that M depends on both the
point of application (Q) and the direction of F (angle θ relative to r). It is a maximum when
F is perpendicular to r. The direction of M is perpendicular to both r and F. If r and F are in
the plane of the page in Figure 11.81, then M points out of the page towards the reader.
FIGURE 11.81 Moment of a force about a point
r
P
Q
Units of moment are newtons multiplied by metres (N · m), the same as for work. For work,
1 N · m is called a joule. We do not associate joules with moment; moment is a vector, work is
a scalar.
EXAMPLE 11.27
What is the moment of the force F = 3î − 2ĵ + 8k̂ N, applied at the point (1, 2, −1) about the
point (3, 1, 4) , all distances in metres?
SOLUTION The moment is
* *
* î ĵ k̂ *
* *
M = * −2 1 −5 * = −2î + ĵ + k̂ N · m.
* 3 −2 8
*
EXAMPLE 11.28
What is the moment of the force F = ĵ − k̂ N, applied at the point (3, −2, 3) about the point
(3, 1, 0) , all distances in metres?
SOLUTION The moment is
* *
* î ĵ k̂ *
* *
M = * 0 −3 3 * = 0.
*0 1 −1
*
It is zero because F is along the same line as r. (This is like trying to close a door by pushing
toward the hinges.)
EXAMPLE 11.29
The horizontal rectangular plate in Figure 11.82a is supported by brackets at A and B and by
a wire CD , where D is directly above B . If the tension in the wire is 200 N, determine the
moment about A of the force exerted by the wire on point C .
FIGURE 11.82a Plate supported by hinges at A and B and wire CD FIGURE 11.82b Moment of tension in wire about A
D z
D(0, 0, 240)
240 mm
80 mm
300 m B
m
B
m
m
y
0
24
80 mm A(240, 0, 0)
T
A
x C(320, 300, 0)
C
SOLUTION With the coordinate system in Figure 11.82b, the tension in the wire acting on C
is
200(−320, −300, 240)
T = √ = 8(−16, −15, 12).
3202 + 3002 + 2402
Since AC = (0.08, 0.3, 0) , the required moment is

* *
* î ĵ k̂ *
* *
M = AC × T = * 0.08 0.3 0 * = (28.8, −7.68, 28.8) N · m.
* −128 −120 96 *
FIGURE 11.83 Moment

In discussing rigid bodies it is useful to define the moment of a force about a line. When a force
of a force about a line F acts at a point Q , its moment about a point P (Figure 11.83) is M = r × F = PQ × F. If
* is any line through P , we define the moment of F about * as the component of M along * .
If v̂ is a unit vector along * , then
F
v̂ moment of F about * = (PQ × F) · v̂ = PQ × F · v̂. (11.46)
P Q The parentheses are unnecessary since the expression makes sense only if the vector product is
performed first. This is a scalar triple product.
EXAMPLE 11.30
Find the moment of the force in Example 11.27 about the line through the points P (3, 1, 4) and
Q(4, −1, 3) .
SOLUTION According to the calculations in Example 11.27, the moment about P is M =

−2î + ĵ + k̂. The moment about the line through P and Q is the component of M along P Q .
Since a unit vector along P Q is PQ/|PQ| , the moment about the line is
(1, −2, −1) 5

(−2, 1, 1) · √ = − √ N · m.
6 6
FIGURE 11.84 Moment

The moment of the force F about the line * in Figure 11.83 is independent of the choice of point
of a force about a line is indepen- P on * used in its determination. To show this, suppose that R is any other point on * (Figure
dent of point chosen on line 11.84). Then
R moment of F about * = PQ × F · v̂ = (PR + RQ) × F · v̂ = PR × F · v̂ + RQ × F · v̂.

F
v̂ Since PR × F is perpendicular to PR, and v̂ is parallel to PR, it follows that PR × F is
perpendicular to v̂, and their scalar product is zero. Consequently,
P Q
moment of F about * = RQ × F · v̂,
and this is formula 11.46 with P replaced by R .

The moment of a force about a line is not unique because the unit vector v̂ along the line
is determined only as to sign; that is, if v̂ is a unit vector along the line, so is −v̂. In other
words, moments of forces about lines are determined only as to sign. This could be remedied
by specifying a direction along the line, but this is not normally done. It is agreed that when a
number of moments are required about a line, the same direction will be taken along the line in
all calculations.
EXAMPLE 11.31
What are the moments about the x -, y -, and z -axes due to the force in Example 11.29?
SOLUTION Since A is on the x -axis in Figure 11.82, and the moment about A is M =
28.8î − 7.68ĵ + 28.8k̂ N · m, the moment about the x -axis is 28.8 N · m, the x -component of
M. To find moments about the y - and z -axes, we find the moment due to the tension T about
B , the origin, since it lies on both axes. It is
* *
* î ĵ k̂ *
* *
MB = BC × T = * 0.32 0 .3 0 * = (28.8, −30.72, 0) N · m.
* −128 −120 96 *
Consequently, moments about the y - and z -axes are −30.72 N · m and 0 N · m, respectively.
Also confirmed is the moment 28.8 N · m about the x -axis.
EXERCISES 11.7
In Exercises 1–5 calculate the moment of the force about the given ∗ 13. Suppose that a force F = Fx î + Fy ĵ + Fz k̂ acts at a point
point. P (x1 , y1 , z1 ) , and v̂ = vx î + vy ĵ + vz k̂ are the components of a
vector along a line * . Show that if Q(x0 , y0 , z0 ) is any point on * ,
1. F = 2î + 3ĵ − 4k̂ at (1, 3, 2) about (−1, 4, 2) then the moment of F about * is the scalar triple product
* *
2. F = î + 2ĵ at (1, 1, 0) about (2, 1, −5) * vx vy vz *
* *
v̂ · PQ × F = * x0 − x1 y0 − y1 z0 − z1 * .
* F *
3. F = −î + 3k̂ at (0, 0, 0) about (−1, 3, 0) x Fy Fz
4. F = 3î − ĵ + 4k̂ at (1, 1, 1) about (2, 2, 2) ∗ 14. Repeat Example 11.26 if the spring has an unstretched length l
that is less than the length of AC .
5. F = 6î at (0, 1, 3) about (2, 0, 0)
∗ 15. Two positive charges q1 and q2 are placed in the xy -plane at
positions (5, 5) and (−2, 3) , respectively. A third positive charge q3
In Exercises 6–9 calculate the moment of the force about the indicated is moved along the x -axis from x = 1 to x = −1. Find the total work
line(s). done by the electrostatic forces of q1 and q2 on q3 .
6. F = 2î + 3ĵ − 4k̂ at (1, 3, 2) about (a) the line through the points ∗ 16. When the rocket in the figure below passes close to the spherical
(1, −3, 2) and (−2, 4, 3) , (b) the coordinate axes, and (c) the line asteroid, it is attracted to the asteroid by a gravitational force with
x = 2 + 3t , y = 4 − 2t , z = 1 + 5t magnitude GmM/r 2 , where m and M are the masses of the rocket and
asteroid, r is the distance from the rocket to the centre of the asteroid,
7. F = 6î − 5ĵ + k̂ at (−2, 3, 1) about the line (x − 3)/2 = y + 1 = and G is Newton’s gravitational constant. Determine the work the
z/4 rocket must do against this force in order to follow the straight-line
path from A to B .
8. F = 4î − 2k̂ at (6, −2, 1) about the line x = y = z − 1
Rocket
A B
9. F = î + ĵ − k̂ at (−1, −1, −2) about the line x − y + z = 2,
x + 2y + 3z = 4
10. If M = (Mx , My , Mz ) are the components of the moment about

R r
the origin due to a force F = (Fx , Fy , Fz ) acting at a point P (x, y, z) ,
what are the moments due to F about the x -, y -, and z -axes?
4
11. Suppose that a force F acts at a point Q , and P is a point on a line
* . Prove that when PQ × F is perpendicular to * , then the moment of Asteroid
F about * is zero.
∗ 12. Prove that if a force F acts at a point P , and * is a line through ∗ 17. If F is a constant force, show that the work done by F on an object
P , then the moment due to F about * is zero. moving around any closed polygon is zero.
∗ 18. Two springs with constants k and 2k are joined together at C and ∗ 21. Rod AB is held in place by cord AC in the figure below. If the
have their other ends fixed at points A and B in the figure below. When tension in the cord is 1500 N, determine the magnitude of the moment
their join is at C , neither spring is compressed or stretched. If the join about B due to the tension at A .
is pulled along the straight line CD perpendicular to AB , what work
is done?
C
l k 3.6 m
B
C D
l 2k 2.4 m
A
L 4.5 m
B
∗ 22. A precast concrete wall section is temporarily held by two cables

∗ 19. A small boat hangs from davits, one of which is shown in the as shown in the figure below. If the tension in cable BD is 900 N,
figure below. The tension in the rope ABAD is 410 N. Determine the determine (a) the moment about O of the force exerted by cable BD
moment about C of the resultant force of the three tensions in the rope at B , (b) the moment due to this force about the z -axis, and (c) the line
exerted on the davit at A . BC .
z z
1m a=2m
A A 2.5 m
B
B E
O
C 2m
1.58 m
D
C y
y D 2m
x
x 1m
∗ 23. The 6-m boom OB in the figure below has a fixed end O . A cable
∗ 20. A 200-N force is applied to the bracket ABC in the figure below. is stretched from the free end B to a point A in the vertical wall of the
Determine the moment of the force about A . xz -plane. If the tension in the cable is 1900 N, find the moment about
O of the tension at B .
z 200 N
60 mm z A
30˚ 4m
60˚
2.4 m
C
B
25 mm O
A x
50 mm
y 6m B
y
x
∗ 24. Forces of 700 N, 1000 N, and 1200 N are applied to the bracket
in the figure below. The 700-N force is directed toward point E . The z F
1000-N force is parallel to the xy -plane, and the 1200-N force is in the
yz -plane. What is the total moment of all three forces about A ? 100 mm
1.3
B C
z
O
75 mm 1000 N
C
50 mm 200 mm
50 mm
A B A
1
1200 N x 150 mm
y
100 mm O 700 N D y ∗ 30. The 575-mm vertical rod CD in the figure below is welded to the
midpoint C of the 1250-mm rod AB . Determine the moments about
AB of (a) the 1175-N force F1 and (b) the 870-N force F2 .
100 mm E(100, 150, −50)
x z
600 mm 800 mm
D
∗ 25. A force F acts at a point A with coordinates (a, a, a) . Show that F2
the sum of the moments of F about the coordinate axes is zero. A H
F1
∗ 26. The rectangular plate in the figure below is hinged at A and B and C
supported by a cable that passes over a frictionless hook at E . If the 750 mm 425 mm
O
tension in the cable is 1349 N, determine moments about the coordinate
B
axes of the force exerted by the cable (a) at C and (b) at D .
x 400 mm 300 mm y
z 525 mm 450 mm
0.90 m G
2.30 m ∗ 31. The frame ACD in the figure below is hinged at A and D and
supported by a cable that passes through a ring at B and is attached to
hooks at G and H . Knowing that the tension in the cable is 1125 N,
E
determine the moment about diagonal AD of the force exerted on the
frame by (a) portion BH of the cable and (b) portion BG of the cable.
O A
B 1.50 m z
280 mm
C
x y G 700 mm
2.25 m
D H
∗ 27. The moment about the x -axis of the resultant force of the three
740 mm O 600 mm
tensions in the rope in Exercise 19 exerted on the davit at A must not
exceed 375 N · m in absolute value. What is the largest allowable tension
in the rope when a = 2 m? D
A y
x 600 mm
∗ 28. What is the largest allowable distance a in Exercise 27 when the 400 mm B
tension in the rope is 300 N? 400 mm C
∗ 29. The force F in the following figure has magnitude 125 N and acts in ∗ 32. Two forces F1 and F2 in space have the same magnitude. Prove
a direction perpendicular to the handle BC of the crank. Find moments that the moment of F1 about the line of action of F2 is equal to the
of the force about the coordinate axes. moment of F2 about the line of action of F1 .
11.8 Hanging Cables

Hanging cables are used in many engineering applications — suspension bridges, transmission
lines, aerial gondolas, and so on. They are flexible members capable of withstanding tension
forces but not shear forces. The shape of a hanging cable is determined by the loads that it
supports, be they concentrated loads, distributed loads, or a combination of these.
Suppose the load per unit length in the x -direction (including the weight of the cable itself)
of the cable in Figure 11.85 is w(x) N/m. It is important to realize that this is not the load per
unit length as measured along the cable. For simplicity, we have chosen a coordinate system
where the lowest point of the cable is on the y -axis. Three forces act on that part of the cable
between x = 0 and an arbitrary point x , a horizontal tension T0 at x = 0, a tangential tension
T at x , and the vertical load. They are in equilibrium so that we may equate horizontal and
vertical components of their resultant to zero,
2 x
0 = −T0 + T cos θ, 0 = T sin θ − w(t) dt. (11.47)
0
If we eliminate T , we obtain 2 x
T0 tan θ = w(t) dt.
0
Since tan θ is the slope of the cable at x , we can write
2 x
dy
T0 = w(t) dt.
dx 0
Since this equation is valid at every point on the cable we may differentiate with respect to x ,
d 2y w(x)
2
= . (11.48)
dx T0
Two integrations of this differential equation give the shape of the cable once the load w(x) is
specified.
FIGURE 11.85 Differential equation describing the shape of a hanging cable
y B
y = y(x)
T
A
(x, y)
T0
x
x
The simplest load is w(x) = w , a constant. Uniformly distributed loads along the horizon-
tal would be realized in a suspension bridge where the weight of the cable would be negligible
compared to that of the roadway. Two integrations of 11.48 give
wx 2
y(x) = + C1 x + C2 .
2 T0
Thus, uniformly loaded cables are parabolic. The fact that y , (0) = 0 implies that C1 = 0, and
therefore y(x) = wx 2 /(2T0 ) + C2 . The position of the x -axis determines C2 . If the origin is
chosen as the lowest point of the cable, then C2 = 0 and y(x) = wx 2 /(2T0 ) .
When the supports A and B of a cable have the same elevation, the distance L between
them is called the span of the cable, and the vertical distance h from the supports to the lowest
point of the cable is called the sag (Figure 11.86).
FIGURE 11.86 Span and sag of a hanging cable
y
A L B
h
We now consider a cable carrying a load uniformly distributed along the cable itself (rather
than in the horizontal direction). Such would be the case for a cable hanging under its own
weight. Let w newtons per metre be the constant weight per unit length along the curve,
and choose a coordinate system with the y -axis through the lowest point on the cable (Figure
11.87). Since length along the cable corresponding to an
! ! increment dx along the x -axis is
1 + (dy/dx)2 dx , the weight of this much cable is w 1 + 2
!(dy/dx) dx . The weight per
unit length in the x -direction along the cable is therefore w 1 + (dy/dx)2 . We substitute
this for the function w(x) in equation 11.48:
+ " #2
d 2y w dy
= 1+ . (11.49)
dx 2 T0 dx
FIGURE 11.87 Cable with a load uniformly distributed along the cable
y B
A
(x, y)
T0
dx
x
x
This differential equation defines the shape of the curve. To solve it, we first set v = dy/dx
and dv/dx = d 2 y/dx 2 ,
dv w! 1 w
= 1 + v2 &⇒ √ dv = dx,
dx T0 1+ v2 T0
a separated differential equation. Solutions are defined implicitly by
2 2 2
1 w 1 wx
√ dv = dx &⇒ √ dv = + D.
1 + v2 T0 1 + v2 T0
In the integral we substitute v = tan θ and dv = sec2 θ dθ ,

2 2 !
wx 1 2
+D = sec θ dθ = sec θ dθ = ln | sec θ + tan θ | = ln | 1 + v 2 + v|.
T0 sec θ
Since v is the slope of the cable and the coordinate system has been chosen so that y , (0) =
v(0) = 0, it follows that D = 0. Consequently,
! wx !
ln ( 1 + v 2 + v) = &⇒ 1 + v 2 + v = ewx/T0 .
T0
√
When we square 1 + v 2 = −v + ewx/T0 , the result is
dy e2wx/T0 − 1 1
1 +v 2 = v 2 − 2vewx/T0 +e2wx/T0 &⇒ v = = = (ewx/T0 −e−wx/T0 ).
dx 2ewx/T0 2
Integration now gives
" #
T0 wx/T0 −wx/T0 T0 wx
y = (e +e )+C = cosh + C. (11.50)
2w w T0
This curve is called a catenary. If we choose the x -axis to pass through the minimum point of
the cable (Figure 11.88), then C = −T0 /w and
- " # .
T0 wx
y = cosh −1 . (11.51)
w T0
FIGURE 11.88 Cable with origin chosen at lowest point
y B
Two important properties of the catenary are derived in the next two examples.
EXAMPLE 11.32
Verify that the length of catenary 11.50 from x = 0 to any point with x -coordinate x is
s = (T0 /w) sinh (wx/T0 ) .
SOLUTION The length of the catenary from x = 0 to an arbitrary x is
+
2 # "2 x+ " #
x
dy 2 2
wt
s = 1+ dt = 1 + sinh dt
0 dt 0 T0
2 x " # 3 " #4x " #
wt T0 wt T0 wx
= cosh dt = sinh = sinh .
0 T0 w T0 0 w T0
EXAMPLE 11.33
Verify that the tension T at any point P (x, y) on the catenary 11.50 is related to T0 by the
equation T = T0 + wy .
SOLUTION The tension at any point in the cable is given by the first of equations 11.47,
+ " #2
! dy
T = T0 sec θ = T0 1 + tan2 θ = T0 1 +
dx
+ " # " # " #
wx wx wy
= T0 1 + sinh 2 = T0 cosh = T0 + 1 = T0 + wy.
T0 T0 T0
EXAMPLE 11.34
A uniform cable with mass per unit length 5 kg/m is suspended between the two points A and
B in Figure 11.89a. The span is 150 m and the sag is 30 m. Determine the length of the cable
and the maximum and minimum tensions in it.
FIGURE 11.89a Uniform cable FIGURE 11.89b Coordinate

with given span and sag system applied to cable
150 A(−75, 30) y B(75, 30)

A B
30
x
SOLUTION If we adopt the coordinate system in Figure 11.89b, the equation of the catenary
takes form 11.51. With point (75, 30) on the curve,
- " # . " #
T0 75w 30w 75w
30 = cosh −1 &⇒ = cosh − 1.
w T0 T0 T0
If we set z = w/T0 , the equation
f (z) = cosh (75z) − 30z − 1 = 0
can be solved numerically for z . For instance, Newton’s iterative procedure with z1 = 0.01
and
cosh (75zn ) − 30zn − 1
zn+1 = zn −
75sinh (75zn ) − 30
gives the approximations z2 = 0.010 17, z3 = 0.010 16, and z4 = 0.010 16. Consequently,
the minimum tension in the cable is T0 = w/z = 5(9.81)/0.010 16 = 4828 N. According to
Example 11.32, the length of the cable is
" #
2 T0 75w 2
L= sinh = sinh [75(0.010 16)] = 164.9 m.
w T0 0.010 16
According to Example 11.33, the maximum tension, at x = 75, is
T = T0 + 30w = 4828 + 30(5)(9.81) = 6300 N.
If a uniform cable is pulled very, very tight, it becomes almost horizontal, and it should be
possible to consider its weight per unit length w along the curve as a uniform weight per unit
length in the x -direction. To show that this is the case, we express equation 11.51 for the
catenary in the form
- " # . -" # .
T0 wx T0 w2 x 2 w4 x 4
y = cosh −1 = 1+ + + · · · − 1 .
w T0 w 2T02 24T04
For a very tight cable, T0 will be very large, so that if we retain only the first two terms of the
Maclaurin series,
" #
T0 w2 x 2 wx 2
y ≈ = .
w 2T02 2 T0
What we are saying is that as the tension in a catenary is increased, the more closely it can be
approximated by a parabola.
A cable, with uniform weight per unit length w , is to be suspended over a chasm with both
ends of the cable at the same elevation. Design specification for the cable requires that
tension must never exceed Tm . We are to determine the maximum allowable horizontal
span for the cable.
SOLUTION We can take the equation for the cable in form 11.51,
- " # .
T0 wx
y = cosh −1 ,
w T0
where T0 is the tension at the lowest point in the cable, provided we use the coordinate
system in Figure 11.90. If h is the sag and L is the span, then
- " # .
T0 wL
h= cosh −1 .
w 2 T0
FIGURE 11.90 Span of hanging cable
y
A(L/2, h)
Maximum tension occurs at A and according to Example 11.33 is given by

- " # . # "
wL wL
Tm = T0 + wh = T0 + T0 cosh − 1 = T0 cosh .
2 T0 2 T0
Since Tm is specified and w is known, this equation implicitly defines L as a function of

T0 , and to find the maximum value of L , we set dL/dT0 = 0. Implicit differentiation
gives
" # " #" #
wL wL −wL w dL
0 = cosh + T0 sinh + ,
2 T0 2 T0 2T02 2T0 dT0
and this implies that
- " #.
dL 2 wL wL
= − coth = 0.
dT0 w 2 T0 2 T0
If we set z = wL/(2T0 ) , this equation becomes
z − coth z = 0 &⇒ z tanh z = 1.
This equation cannot be solved exactly, but when it is solved numerically the result is
z = 1.200. It now follows that
wL 2.400T0
= 1.200 &⇒ L= .
2 T0 w
5 6
To replace T0 , we note that Tm = T0 cosh wL 2 T0
= T0 cosh 1.200. Thus, maximum
span is
2.400Tm sech (1.200) 1.325Tm
L= = .
w w
EXERCISES 11.8
1. A cable with ends at the same elevation has a span of 100 m and a 6. Use the formula in Exercise 5(a) and the two-term approximation
sag of 5 m. It is subjected to a uniform horizontal load of 1000 N/m. in part (b) to calculate the length of the cable in Exercise 1.
Determine the minimum tension in the cable. 7. Complete Exercise 6 for the cable in Exercise 4.
2. Cable AB in the figure below supports a uniform, horizontally dis-
tributed load of 1100 N/m. The lowest point of the cable is 3 m below 8. The centre span of the Verrazano-Narrows bridge consists of two
support A , and the support at B is 6 m higher than the support at A . uniform roadways suspended from four cables. The design of the bridge
Determine the minimum and maximum tensions in the cable. includes the effect of extreme temperature changes, which cause the
sag of the centre span to vary from hw = 115.8 m in winter to
B hs = 118.2 m in summer. If the span is 1278 m, use the two-term
6m A approximation in Exercise 5(b) to determine the change in the length
3m of the cables at these temperature extremes.
∗ 9. Before being fed into a printing press located to the right of D
1100 N/m
40 m in the figure below, a continuous sheet of paper having mass per unit
length 300 g/m passes over rollers at A and B . Assuming that the
3. Cable AB in the figure below supports a uniform, horizontally dis- curve formed by the sheet is parabolic, determine (a) the location of the
tributed load of w newtons per metre. The lowest point of the cable is lowest point C and (b) the maximum tension in the sheet.
1 m below the support at A , and the support at B is 2 m higher than
the support at A . Determine a formula for the minimum tension in the 0.25 m
1.125 m
cable in terms of w .
A
B 100 mm 75 mm D
B
2m C
A ∗ 10. A 40-m rope is strung between the roofs of two buildings, each
1m 14 m high. The maximum tension is 350 N and the lowest point of
the cable is 6 m above the ground. Determine the horizontal distance
w newtons per metre between the buildings and the mass of the rope.
20 m
∗ 11. A 50-m steel measuring tape has mass 1.6 kg. If the tape is stretched
4. The centre span of the George Washington Bridge consists of a uni- between two points at the same elevation until the tension at each end
form roadway suspended from four cables. The uniform load supported is 60 N, determine the span of the tape.
by each cable is 142 kN/m along the horizontal. If the span is 1050 m
and the sag is 94.8 m, determine the minimum and maximum tensions ∗ 12. An aerial tramway cable of length 150 m and with mass per unit
in each cable. length 4 kg/m is suspended between two points at the same elevation.
∗ 5. (a) Show that the length of a cable that supports a uniform, If the sag is 37.5 m, find the span of the cable and the maximum tension
horizontally distributed load of w , from its minimum point in the cable.
to a point P (xP , yP ) , is given by ∗ 13. A counterweight D of mass 40 kg is attached to a cable that passes
+ over a small pulley at A and is attached to a support at B (figure below).
2 xP The system is motionless. Knowing that L = 15 m and h = 5 m,
w2 x 2
L(P ) = 1+ dx. determine the length of the cable from A to B and the mass per unit
0 T02 length of the cable. Neglect the mass of the cable from A to D .
Show that the value of this integral is
+ + 
" #2 " # 20 m
xP wxP T0  wxP 2 wxP 
L(P ) = 1+ + ln 1+ + .
2 T0 2w T0 T0 A B
D h
(b) The expression in part (a) for L(P ) is unwieldy in per- 40 kg
forming calculations involving the length of a cable subject C
to uniform loads.
$ In practice, it is often approximated by ∗∗ 14. A uniform cord 900 mm long passes over a frictionless pulley at
first expanding 1 + (w 2 xP2 /T02 ) as an infinite series and B and is attached to a rigid support at A (figure below). If L = 300
integrating term by term. Show that this leads to mm, determine the smaller of the two sags for which the cord is in
/ 1 equilibrium.
" #2 " #4
2 yP 2 yP
L(P ) = xP 1 + − + ··· , L
3 xP 5 xP
A B
h
convergent for yP /xP < 1/2. In situations where yP is
much less than xP , the series is truncated after the first two
terms.
∗ 15. To the left of point B in the figure below, the long cable rests on a ∗∗ 16. A cable of uniform weight per unit length w is suspended between
rough horizontal surface. If the cable has mass per unit length 2.7 kg/m, two points at the same elevation a distance L apart. Determine the sag-
determine the force F to maintain equilibrium. to-span ratio for which the maximum tension is as small as possible.
D E F
h=3m
A B
2.7 m
11.9 Differentiation and Integration of Vectors

In Section 11.7, vectors were used to represent forces and moments; they can also be used to
describe many other physical quantities, such as position, velocity, acceleration, electric and
magnetic fields, and fluid flow. In applications such as these, vectors seldom have constant
components; instead, they have components that are either functions of position, or functions of
some parameter, such as time, or both. For instance, the spring force in Example 11.26 varies
in both magnitude and direction as the sleeve moves from B to C . Consequently, components
of the vector F representing this force are functions of position x between B and C :
F = Fx î + Fy ĵ = Fx (x)î + Fy (x)ĵ.
When a particle moves along a curve C in the xy -plane defined parametrically by
C : x = x(t), y = y(t), α ≤ t ≤ β,
FIGURE 11.91 Position its position (x, y) relative to the origin is represented by the vector r = x î + y ĵ (Figure 11.91).
vector of particle moving along a
This vector is called the position vector or displacement vector of the particle relative to the
curve in the xy -plane
origin. We will have more to say about it in Section 11.13. Note, however, that if we substitute
y from the parametric equations for C , we have
(x, y) C
r = x(t)î + y(t)ĵ,
r = x i + yj
which indicates that the displacement vector has components that are functions of the parame-
x ter t .
In this section we consider the general situation in which the components vx , vy , and vz
of a vector v are functions of some parameter t ,
v = vx (t)î + vy (t)ĵ + vz (t)k̂, (11.52)
and show how the operations of differentiation and integration can be applied to such a vector.
In Sections 11.11 and 11.12 we will use these results to discuss the geometry of curves, and this
will pave the way for an analysis of the motion of particles in Section 11.13.
Because v in 11.52 has components that are functions of t , we say that v itself is a vector-
valued function of t , and write v = v(t) . Each of the component functions vx (t) , vy (t) ,
and vz (t) has a domain, and their common domain is called the domain of the vector-valued
function v(t) . Given that this domain is some interval α ≤ t ≤ β , we express 11.52 more
fully in the form
v = v(t) = vx (t)î + vy (t)ĵ + vz (t)k̂, α ≤ t ≤ β. (11.53)
To differentiate and integrate vector-valued functions, we first require the concept of a limit.
DEFINITION 11.12
If v(t) = vx (t)î + vy (t)ĵ + vz (t)k̂, then
- . - . - .
lim v(t) = lim vx (t) î + lim vy (t) ĵ + lim vz (t) k̂, (11.54)
t→t0 t→t0 t→t0 t→t0
provided that each of the limits on the right exists.
This definition states that to take the limit of a vector-valued function, we take the limit of
each component separately. As an illustration, consider the following example.
EXAMPLE 11.35
If v = v(t) = (t 2 + 1)î + 3t ĵ − (sin t)k̂, calculate limt→5 v(t) .
SOLUTION According to Definition 11.12,
- . - . - .
2
lim v(t) = lim (t + 1) î + lim (3t) ĵ − lim sin t k̂
t→5 t→5 t→5 t→5
= 26î + 15ĵ − (sin 5)k̂.
In the remainder of this section we define continuity, derivatives, and antiderivatives for vector-
valued functions. Each definition is an exact duplicate of the corresponding definition for a scalar
function y(t) , except that y(t) is replaced by v(t) . We then show that the vector definition can
be rephrased in terms of components of the vector. We begin with continuity in the following
definition.
DEFINITION 11.13
A vector-valued function v(t) is said to be continuous at a point t0 if
v(t0 ) = lim v(t). (11.55)

t→t0
It is a simple matter to prove the next theorem.
THEOREM 11.7
A vector-valued function is continuous at a point if and only if its components are contin-
uous at that point.
PROOF If v(t) = vx (t)î + vy (t)ĵ + vz (t)k̂, then according to 11.55, v(t) is continuous at
t0 if and only if
vx (t0 )î + vy (t0 )ĵ + vz (t0 )k̂ = lim [vx (t)î + vy (t)ĵ + vz (t)k̂].
t→t0
Definition 11.12 implies that we can write this condition in the form
- . - . - .
vx (t0 )î + vy (t0 )ĵ + vz (t0 )k̂ = lim vx (t) î + lim vy (t) ĵ + lim vz (t) k̂.
t→t0 t→t0 t→t0
But because two vectors are equal if and only if their components are equal, we can say that
v(t) is continuous at t0 if and only if
vx (t0 ) = lim vx (t); vy (t0 ) = lim vy (t); vz (t0 ) = lim vz (t)

t→t0 t→t0 t→t0
[i.e., v(t) is continuous at t0 if and only if its components are continuous at t0 ].

For example, the vector-valued function
v(t) = (t − 1)î + (1/t)ĵ + (t 2 − 1)−1 k̂
is discontinuous for t = 0 [since vy (0) is not defined] and for t = ±1 [since vz (±1) is not
defined].
The derivative of a scalar function y(t) is its instantaneous rate of change:
dy y(t + h) − y(t)
= lim .
dt h→0 h
The derivative of a vector-valued function is also a rate of change defined by a similar limit.
DEFINITION 11.14
The derivative of a vector-valued function v(t) is defined as
dv v(t + h) − v(t)
= lim , (11.56)
dt h→0 h
In practice, we seldom use the definition of a derivative to calculate dy/dt for a scalar func-
tion y(t) ; formulas such as the power, product, quotient, and chain rules are more convenient. It
would be helpful to have corresponding formulas for derivatives of vector-valued functions. The
following theorem shows that to differentiate a vector-valued function, we simply differentiate
its Cartesian components.
THEOREM 11.8
If v(t) = vx (t)î + vy (t)ĵ + vz (t)k̂, then
dv dvx dvy dvz

= î + ĵ + k̂, (11.57)
dt dt dt dt
provided that the derivatives on the right exist.
PROOF If we substitute the components of v(t + h) and v(t) into Definition 11.14, then we
have
3 4
dv [vx (t + h)î + vy (t + h)ĵ + vz (t + h)k̂] − [vx (t)î + vy (t)ĵ + vz (t)k̂]
= lim
dt h→0 h
3- . - . - . 4
vx (t + h) − vx (t) vy (t + h) − vy (t) vz (t + h) − vz (t)
= lim î + ĵ + k̂
h→0 h h h
(according to equation 11.14), and

- . - .
dv vx (t + h) − vx (t) vy (t + h) − vy (t)
= lim î + lim ĵ
dt h→0 h h→0 h
- .
vz (t + h) − vz (t)
+ lim k̂
h→0 h
(according to equation 11.54). Since each of the limits on the right exists, we can write that
dv dvx dvy dvz

= î + ĵ + k̂.
dt dt dt dt
Theorem 11.8 gives us a working rule for differentiating vector-valued functions: To dif-
ferentiate a vector-valued function, we differentiate its Cartesian components.
EXAMPLE 11.36
If v(t) = t 2 î + (3t 3 − 2t)ĵ + 5k̂, find v, (3) .
dv
= 2t î + (9t 2 − 2)ĵ.
dt
Consequently, v, (3) = 6î + 79ĵ.
The sum rule 3.9 for differentiation of scalar functions has its counterpart in the sum rule for
vector-valued functions,
d du dv
(u + v ) = + (11.58)
dt dt dt
(see Exercise 22). There are three types of products associated with vectors: the product of
a scalar and a vector, the dot product of two vectors, and the cross product of two vectors.
Corresponding to each, we have a product rule for differentiation, but all resemble the product
rule for scalar functions.
THEOREM 11.9
If f (t) is a differentiable function and u(t) and v(t) are differentiable vector-valued
functions, then
d df dv
(f v) = v+f , (11.59a)
dt dt dt
d dv du
(u · v ) = u · + · v, (11.59b)
dt dt dt
d dv du
(u × v ) = u × + × v. (11.59c)
dt dt dt
For a proof of these results, see Exercise 23.

EXAMPLE 11.37
If f (t) = t 2 + 2t + 3, u(t) = t î + t 2 ĵ − 3k̂, and v(t) = t (î + ĵ + k̂) , use 11.59 to evaluate:
d d d
(a) (f u) (b) (u · v) (c) (u × v)
dt dt dt
SOLUTION
(a) With 11.59a,
d df du
(f u) = u+f = (2t + 2)(t î + t 2 ĵ − 3k̂) + (t 2 + 2t + 3)(î + 2t ĵ)
dt dt dt
= (3t 2 + 4t + 3)î + (4t 3 + 6t 2 + 6t)ĵ − 6(t + 1)k̂.
(b) With 11.59b,
d dv du
(u · v ) = u · + ·v
dt dt dt
= (t î + t 2 ĵ − 3k̂) · (î + ĵ + k̂) + (î + 2t ĵ) · (t î + t ĵ + t k̂)
= (t + t 2 − 3) + (t + 2t 2 ) = 3t 2 + 2t − 3.
(c) With 11.59c,
d dv du
(u × v ) = u × + × v = (t î + t 2 ĵ − 3k̂) × (î + ĵ + k̂) + (î + 2t ĵ) × (t î + t ĵ + t k̂)
dt dt dt
* * * *
* î ĵ k̂ * * î ĵ k̂ *
* * * *
= * t t 2 −3 * + * 1 2 t 0 *
*1 1 1 * *t t t *
= [(t 2 + 3)î − (3 + t)ĵ + (t − t 2 )k̂] + [2t 2 î − t ĵ + (t − 2t 2 )k̂]

= (3t 2 + 3)î − (3 + 2t)ĵ + (2t − 3t 2 )k̂.
If vector-valued functions can be differentiated, then they can be antidifferentiated. Formally,

we make the following statement.
DEFINITION 11.15
A vector-valued function V(t) is said to be an antiderivative of v(t) on the interval
α < t < β if
dV
= v(t) for α < t < β. (11.60)
dt
For example, an antiderivative of v(t) = 2t î − ĵ + 3t 2 k̂ is
V(t) = t 2 î − t ĵ + t 3 k̂.
If we add to V(t) in 11.60 any vector with constant components, denoted by C, then V(t) + C
is also an antiderivative of v(t) . We call this vector the indefinite integral of v(t) , and write
2
v(t) dt = V(t) + C. (11.61)
For our example, then,

2
(2t î − ĵ + 3t 2 k̂) dt = t 2 î − t ĵ + t 3 k̂ + C.
Because vectors can be differentiated component by component, it follows that they may also
be integrated component by component; that is, if v(t) = vx (t)î + vy (t)ĵ + vz (t)k̂, then
2 -2 . -2 . -2 .
v(t) dt = vx (t) dt î + vy (t) dt ĵ + vz (t) dt k̂. (11.62)
EXAMPLE 11.38
√
Find the indefinite integral of v(t) = t − 1 î + et ĵ + 6t 2 k̂.
SOLUTION According to equation 11.62,
2 - .
2 3/2
v(t) dt = (t − 1) + C1 î + (et + C2 )ĵ + (2t 3 + C3 )k̂
3
2
= (t − 1)3/2 î + et ĵ + 2t 3 k̂ + C,
3
where C = C1 î + C2 ĵ + C3 k̂ is a constant vector.
EXERCISES 11.9
2
In Exercises 1–5 find the largest possible domain for the vector-valued d
16. [t (u × v)] 17. [f (t)u(t)] dt
function. dt
2 2
√ 18. [3g(t)v(t) + u(t)] dt 19. [f (t)u · v] dt
1. v(t) = t 2 î + t − 1ĵ + k̂
2
dv dv dv
2. v(t) = (sin t)î + (cos t)ĵ − t 3 k̂ 20. u × − f (t)u · v 21. u · −v· u(t) dt
dt dt dt
3. v(t) = (Sin −1
t)î − t ĵ + (t + 1)k̂
2
4. v(t) = ln (t + 4)(î + ĵ) 22. Prove equation 11.58.

23. Verify the results in equations 11.59.
5. v(t) = et î + (cos2 t)ĵ − (et cos t)k̂ ∗ 24. Prove that for differentiable functions u(t) , v(t) , and w(t) ,
d du dv dw
(u · v × w ) = ·v×w+u· ×w+u·v× .
If f (t) = t 2 + 3, g(t) = 2t 3 − 3t , u(t) = t î − t 2 ĵ + 2t k̂ , and dt dt dt dt
v(t) = î − 2t ĵ + 3t 2 k̂ , find the scalar or the components of the vector ∗ 25. Prove that if a differentiable function v(t) has constant length, then
in Exercises 6–21 . at any point at which d v/dt *= 0 , the vector d v/dt is perpendicular
to v .
du d ∗ 26. If v = v(s) is a differentiable vector-valued function and s = s(t)
6. 7. [f (t)v(t)]
dt dt is a differentiable scalar function, prove that
d d
8. [g(t)u(t)] 9. (u × v ) dv d v ds
dt dt = .
d d dt ds dt
10. (u × t v ) 11. (2 u · v )
dt dt This result is called the chain rule for differentiation of vector-valued
2 functions.
d
12. (3u + 4v) 13. u(t) dt ∗ 27. Show that the following definition for the limit of a vector-valued
dt
2 function is equivalent to Definition 11.12: A vector-valued function
14.
d
[f (t)u + g(t)v] 15. 4v(t) dt v(t) is said to have limit V as t approaches t0 if given any $ > 0, there
dt exists a δ > 0 such that |v(t) − V| < $ whenever 0 < |t − t0 | < δ .
11.10 Parametric and Vector Representations of Curves

In Section 11.2, we presented curves in space as the intersection of two surfaces. For example,
each of the equations
x 2 + y 2 + z2 = 9, y =2
describes a surface (the first is a sphere and the second a plane), and together they describe the
curve of intersection of the surfaces — the circle in Figure 11.92.
FIGURE 11.92 Curve of intersection of a sphere and a plane is a circle
z
y=2
x 2 + y 2 + z 2 = 9,
x2 + y2 + z2 = 9
y=2
3
3 3
y
Parametric Representation of Curves

In many applications it is more convenient to have a curve defined parametrically.
DEFINITION 11.16
A curve in space is defined parametrically by three functions:
C : x = x(t), y = y(t), z = z(t), α ≤ t ≤ β. (11.63)
FIGURE 11.93 Curves

Each value of t in the interval α ≤ t ≤ β is substituted into the three functions, and the
are directed from initial point to triple (x, y, z) = (x(t), y(t), z(t)) specifies a point on the curve. Definition 11.16 clearly
final point corresponds to parametric Definition 9.1 for a plane curve.
It is customary to assign a direction to a curve by calling that point on C corresponding to
z
Final t = α the initial point and that point corresponding to t = β the final point, and the direction of
point C is that direction along C from initial point to final point (Figure 11.93). Note in particular that
C the direction of a curve always corresponds to the direction in which the parameter increases
along the curve. Because of this, whenever we describe a curve in nonparametric form but with
Direction of C
a specified direction, we must be careful in setting up parametric equations to ensure that the
Initial parameter increases in the appropriate direction.
point
When a curve is described as the curve of intersection of two surfaces, we often obtain
y parametric equations for the curve by specifying one of x , y , or z as a function of t , and then
x solving the equations for the other two as functions of t . Considerable ingenuity is sometimes
required in arriving at a suitable initial function of t . We illustrate this in the following example.
EXAMPLE 11.39
Find parametric equations for each of the following curves:
(a) z − 1 = x 2 + y 2 , x − y = 0 directed so that z increases when x and y are positive.
(b) x + 2y + z = 4, 2x + y + 3z = 6 directed so that y increases along the curve.
(c) x 2 + (y − 1)2 = 4, z = x directed so that y increases when x is positive.
SOLUTION
(a) The curve of intersection of the circular paraboloid z = 1 + x 2 + y 2 and the plane
y = x is shown in Figure 11.94. If we choose x as the parameter along the curve by
setting x = t , then y = x and z = 1 + x 2 + y 2 imply that
x = t, y = t, z = 1 + 2t 2 .
When t > 0, so are x and y , and for these values of t , z increases as t increases.
This means that these are acceptable parametric equations for the curve.
(b) The straight-line intersection of the two planes is shown in Figure 11.95. If we choose
y as the parameter by setting y = t (thus forcing y to increase as t increases), then
x + z = 4 − 2t, 2x + 3z = 6 − t.
The solution of these equations for x and z in terms of t gives the parametric equations
x = 6 − 5t, y = t, z = −2 + 3t.
(c) The curve of intersection of the right-circular cylinder x 2 +(y− 1)2 = 4 and the plane
z = x is shown in Figure 11.96. We saw in Section 9.1 that trigonometric functions
are particularly useful for circles. If we set x = 2 cos t , then y = 1 ± 2 sin t . A set
of parametric equations for the curve is therefore
x = 2 cos t, y = 1 + 2 sin t, z = 2 cos t, 0 ≤ t ≤ 2π.
Any range of values of t of length 2π traces the curve exactly once. To check that
these equations specify the correct direction along the curve, we note that t = 0
gives the point (2, 1, 2) and t = π/2 gives (0, 3, 0) . Since values of t between 0
and π/2 give one-quarter of the curve (rather than three-quarters), points are indeed
generated in the required direction indicated by the arrowhead in Figure 11.96. Had
we chosen the equations
x = 2 cos t, y = 1 − 2 sin t, z = 2 cos t, 0 ≤ t ≤ 2π,
we would have generated the same set of points traced in the opposite direction.
FIGURE 11.94 Curve of FIGURE 11.95 Line of inter- FIGURE 11.96 Curve of intersection of
intersection of paraboloid and plane section of two planes circular cylinder and plane
z z x 2 + (y − 1)2 = 4
z = 1 + x 2 + y 2,
y=x z
x + 2y + z = 4, x 2 + (y − 1)2 = 4,
4 2x + y + 3z = 6 z=x
z=1+ x2 + y2 x + 2y + z = 4 z=x
2
y=x
2x + y + 3z = 6
3 2 −1
4 3
6 y
y y
x
x
x
Because computers and graphing calculators produce excellent plots of parametrically defined
three-dimensional curves, there is incentive to represent curves parametrically rather than as the
intersection of two surfaces. Plots of the curves in Example 11.39 are shown in Figures 11.97.

Computer plots of the curves in Example 11.39
z z
z
y y
x x
y
x
DEFINITION 11.17
A curve C : x = x(t) , y = y(t) , z = z(t) , α ≤ t ≤ β , is said to be a continuous
curve if each of the functions x(t) , y(t) , and z(t) is continuous for α ≤ t ≤ β .
Geometrically, this implies that the curve is at no point separated. Each of the curves in
Example 11.39 is therefore continuous.
A curve is said to be a closed curve if its initial and final points are the same. Circles and
ellipses are closed curves. Straight-line segments, parabolas, and hyperbolas are not closed.
Vector Representation of Curves

The position vector or displacement vector of a point P (x, y, z) in space is
r = (x, y, z) = x î + y ĵ + zk̂.
We visualize it as the vector drawn from the origin to P (Figure 11.98). If we consider only
points that lie on a curve defined parametrically by 11.63, then for these points we can write that
r = r(t) = x(t)î + y(t)ĵ + z(t)k̂, α ≤ t ≤ β. (11.64)
As t varies from t = α to t = β , the tip of this vector traces the curve C from initial point to
final point. We call 11.64 the vector representation of a curve.
FIGURE 11.98 Vector representation of a curve
z
B
P (x, y, z) C
r = x i + yj + zk
y
x
EXAMPLE 11.40
Draw and plot the curve with position vector
r = r(t) = (2 cos t)î + (3 sin t)ĵ + (t/10)k̂, 0 ≤ t ≤ 6π.
SOLUTION When we set x = 2 cos t , y = 3 sin t , and z = (t/10) , it is clear that x 2 /4 +
y 2 /9 = 1. This means that the curve lies on the elliptic cylinder x 2 /4 + y 2 /9 = 1. As t
increases from 0 to 6π , values of z increase linearly from 0 to 3π/5. What we have therefore
is three loops that rise around the elliptic cylinder. It is part of what is called an elliptic helix.
It is drawn in Figure 11.99a and plotted in Figure 11.99b. If the curve is given width, it could
represent three coils of a spring or three windings of an inductor.
FIGURE 11.99a Drawing an elliptic helix FIGURE 11.99b Computer plot of an elliptic helix
z z
y
x
3
2
y
x
EXAMPLE 11.41
Draw and plot the curve with position vector
r = r(t) = t î + t 2 ĵ + t k̂, t ≥ 0.
SOLUTION When we set x = t , y = t 2 , and z = t , then z = x and y = x 2 . These imply
that r = r(t) describes points on the curve of intersection of the surfaces y = x 2 and z = x
(Figure 11.100a). Because t ≥ 0, only that half of the curve of intersection in the first octant is
defined by r = r(t) . The curve is plotted in Figure 11.100b.
FIGURE 11.100a Curve r = t î + t 2 ĵ + t k̂ FIGURE 11.100b Computer
drawn as intersection of surfaces y = x 2 , z = x plot of curve r = t î + t 2 ĵ + t k̂
z z
z=x
r = r (t)
y
y
x
x
y = x2
EXERCISES 11.10
!
In Exercises 1–10 find parametric and vector representations for the ∗ 8. z = 4 − x 2 − y 2 , x 2 +y 2 − 2y = 0 directed so that z decreases
curve. Draw and plot each curve. when x is positive
√
∗ 9. x = z , z = y 2 directed away from the origin in the first octant
1. x + 2y + 3z = 6, y − 2z = 3 directed so that z increases along
!
the curve ∗ 10. z = x 2 + y 2 , y = x 2 directed so that y decreases in the first
2. x 2 + y 2 = 2, z = 4 directed so that y increases in the first octant octant.
3. x 2 + y 2 = 2, x + y + z = 1 directed so that y decreases when x

In Exercises 11–15 draw and plot the curve with the given position
is positive
vector.
4. z = x 2 + y 2 , x 2 + y 2 = 5 directed clockwise as viewed from the
11. r(t) = t î + t ĵ + t 2 k̂ , t ≥ 0
origin
12. r(t) = (2 cos t)î + (2 sin t)ĵ + 3t k̂ , 0 ≤ t ≤ 4π
5. z + 2x 2 = 1, y = z directed so that x decreases along the curve
! 13. r(t) = (t − 2)î + (2 − 3t)ĵ + 5t k̂
6. z = x 2 + y 2 , y = x directed so that y increases when x is
positive 14. r(t) = (t 2 − t)î + t ĵ + 5k̂
7. z = x + y , y = x 2 directed so that x increases along the curve 15. r(t) = (cos t)î + (sin t)ĵ + (cos t)k̂ , 0 ≤ t ≤ π
11.11 Tangent Vectors and Lengths of Curves

If C is a curve in the xy -plane (Figure 11.101), then the tangent line to C at P is defined as
FIGURE 11.101 Line the limiting position of the line P Q as Q moves along C toward P (see Section 3.1). We take
joining two points on a curve the same approach in defining tangent vectors to curves in an arbitrary plane or in space. On
curve C defined by 11.63, we let P and Q be the points corresponding to the parameter values
y
P t and t + h . Position vectors of P and Q are then
Q
B r(t) = x(t)î + y(t)ĵ + z(t)k̂
C
A
and
x r(t + h) = x(t + h)î + y(t + h)ĵ + z(t + h)k̂
(Figure 11.102), and the vector joining P to Q is
FIGURE 11.102 Limit of
vector joining two points on a curve
PQ = r(t + h) − r(t).
leads to tangent vector to curve
If we let h approach zero, then Q moves along C toward P , and the direction of PQ
z becomes closer to what seems to be a reasonable definition of the tangent direction to C at
r (t + h) − r (t)
P P . Perhaps then we should define limh→0 [r(t + h) − r(t)] as a tangent vector to C at P .
A
Unfortunately, the limit vector has length zero, and therefore
r (t) Q
B
C lim [r(t + h) − r(t)] = 0.
r (t + h) h→0
If, however, we divide r(t + h) − r(t) by h , then the resulting vector
x y r(t + h) − r(t)
h
is not equal to PQ, but it does have the same direction as PQ. Consider, then, taking the limit
of this vector as h approaches zero:
r(t + h) − r(t)
lim .
h→0 h
FIGURE 11.103 Tangent
vector to a curve If the limit vector exists, then it will be tangent to C at P . But according to equation 11.56,
this limit defines the derivative d r/dt :
z
dr r(t + h) − r(t) dx dy dz
C = lim = î + ĵ + k̂, (11.65)
dt h→ 0 h dt dt dt
A
B provided that each of the derivatives dx/dt , dy/dt , and dz/dt exists. We have just established
r
dr the following result.
dt
x y
THEOREM 11.10
If r = r(t) = x(t)î + y(t)ĵ + z(t)k̂, α ≤ t ≤ β , is the vector representation of a curve
FIGURE 11.104 Two C , then at any point on C at which x , (t) , y , (t) , and z, (t) all exist and do not vanish
tangent vectors exist at each point simultaneously,
on a curve, one in the opposite
dr dx dy dz
direction to the other T = = î + ĵ + k̂ (11.66)
dt dt dt dt
z
is a tangent vector to C (Figure 11.103).
B
C
There are two tangent directions at any point on a curve. One of these has been shown to
dr be d r/dt ; the other must be −d r/dt (Figure 11.104). How can we tell which one is d r/dt ?
A −
dt dr A closer analysis of the limit in 11.65 indicates the following (see Exercise 17).
dt
x y COROLLARY 11.10.1
The tangent vector d r/dt to a curve C : x = x(t) , y = y(t) , z = z(t) , α ≤ t ≤ β ,
FIGURE 11.105 Smooth always points in the direction in which the parameter t increases along C .
curve
z
C
DEFINITION 11.18
A curve C : x = x(t) , y = y(t) , z = z(t) , α ≤ t ≤ β , is said to be a smooth curve
A
if the derivatives x , (t) , y , (t) , and z, (t) are all continuous for α < t < β and do not
vanish simultaneously for α < t < β .
B
x y
Since x , (t) , y , (t) , and z, (t) are the components of a tangent vector to C , this definition
implies that along a smooth curve, small changes in t produce small changes in the direction of
FIGURE 11.106 Piecewise-
the tangent vector. In other words, the tangent vector turns gradually, or “smoothly.” The curve
smooth curve
in Figure 11.105 is smooth; that in Figure 11.106 is not because abrupt changes in the direction
z of the curve occur at P and Q . According to the following definition, this curve is piecewise
P
smooth.
Q
C DEFINITION 11.19
A B
A curve is said to be a piecewise-smooth curve if it is continuous and can be divided into
a finite number of smooth subcurves.
x y
EXAMPLE 11.42
For the curve in Example 11.41, find a tangent vector at the point (3, 9, 3) .
SOLUTION A tangent vector to this curve at any point on the curve is
dr dx dy dz
= î + ĵ + k̂ = î + 2t ĵ + k̂.
dt dt dt dt
Since t = 3 yields the point (3, 9, 3) , a tangent vector at this point is r, (3) = î + 6ĵ + k̂.
EXAMPLE 11.43
Find a tangent vector at the point (2, 0, 3) to the helix
FIGURE 11.107 Tangent 3t

x = 2 cos t, y = 2 sin t, z= , t ≥ 0.
vector to a helix 2π
z Is the helix smooth?
SOLUTION A tangent vector to the helix at any point is
" #
dr dx dy dz 3
dr = î + ĵ + k̂ = (−2 sin t)î + (2 cos t)ĵ + k̂.
(2, 0, 3) dt dt dt dt 2π
dt
Since t = 2π yields the point (2, 0, 3) , a tangent vector at this point is r, (2π ) = 2ĵ +
y
(3/(2π ))k̂ (Figure 11.107). Since x , (t) , y , (t) , and z, (t) are all continuous functions, and they
are never simultaneously zero, the helix is indeed smooth.
x
Unit Tangent Vectors

When a curve in the xy -plane is defined parametrically by
C : x = x(t), y = y(t), α ≤ t ≤ β, (11.67)
a tangent vector to C is
dr dx dy
= T = î + ĵ, (11.68)
dt dt dt
and this tangent vector points in the direction in which t increases along C . To produce a unit
tangent vector to C at any point, we divide T by its length:
T d r/dt
T̂ = = . (11.69)
|T| |d r/dt|
We now show that if length along C is used as the parameter by which to specify its points, then
division by |T| is unnecessary.
In Section 7.3 we showed that small lengths along a plane curve C can be approximated by
FIGURE 11.108 Length
along a curve
straight-line lengths along tangent lines to the curve, and that the total length of a smooth curve
from A to B (Figure 11.108) is
y
2 B !
C L= (dx)2 + (dy)2 .
s A
P (x(t), y(t)) With parametric equations 11.67 we can write this formula as a definite integral with respect
to t (see also equation 9.3):
x +-" #2 " #2 . +
2 β 2 β " #2 " #2
dx dy dx dy
L= + (dt)2 = + dt. (11.70)
α dt dt α dt dt
Furthermore, if we denote by s = s(t) the length of that part of C from its initial point A
(where t = α ) to any point P (x(t), y(t)) on C (Figure 11.108), then s(t) is defined by the
integral
+" #2 " #2
2 t
dx dy
s(t) = + dt. (11.71)
α dt dt
It follows, then, that the derivative of s(t) is
+" #2 " #2
ds dx dy
= + . (11.72)
dt dt dt
But according to 11.68, d r/dt is a tangent vector to C with the same length:
* * +" # " #2
* dr *
* *= dx 2 dy ds
* dt * + = . (11.73)
dt dt dt
When this equation is multiplied by dt , it gives

!
|d r| = (dx)2 + (dy)2 = ds. (11.74)
This equation states that ds is the length of the tangent vector d r = dx î + dy ĵ, and therefore
FIGURE 11.109 Small lengths
along the tangent line to a curve in the ds is a measure of length along the tangent line to C . In spite of this we often think of ds as
xy -plane are regarded as small lengths a measure of small lengths along C itself (Figure 11.109), and that ds is approximated by the
along the curve itself tangential straight-line length
y !
dr |d r| = (dx)2 + (dy)2 .
dt B
dr
( ) Note too that if we use length s along C as a parameter, then the chain rule applied to
ds C
r = r(s) , s = s(t) gives
A dr d r ds
= . (11.75)
x dt ds dt
(The chain rule is proved in Exercise 26 of Section 11.9.) Consequently, equation 11.73 implies
that
dr d r/dt d r/dt
= = . (11.76)
ds ds/dt |d r/dt|
What we have shown, then, is that if we choose length along a curve C as the parameter by
which to specify points on the curve ( C : x = x(s) , y = y(s) ), then the vector
dr dx dy
T̂ = = î + ĵ (11.77)
ds ds ds
is a unit tangent vector to C . In addition, the corollary to Theorem 11.10 implies that d r/ds
points in the direction in which s increases along C . This suggests perhaps that we should always
set up parametric equations for a curve with length along the curve as parameter. Theoretically,
this is quite acceptable, but practically it is impossible. For most curves we have enough difficulty
just finding a set of parametric equations, let alone finding that set with length along the curve as
parameter. If we then use a parameter t other than length along the curve, a unit tangent vector
is calculated according to 11.69.
These results can be extended to space curves as well. When a smooth curve C has
parametric equations 11.63, equation 11.69 still defines a unit tangent vector to C , but because
C is a space curve, d r/dt is calculated according to 11.66.
Corresponding to formula 11.71 for length along a curve in the xy -plane, length along a
smooth curve in space is defined by the definite integral:
+" #2 " #2 " #2
2 t
dx dy dz
s(t) = + + dt. (11.78)
α dt dt dt
These two results imply that
* * +" # " #2 " #2
FIGURE 11.110 Small * dr *
* *= dx 2 dy dz ds
lengths along the tangent line to + + = , (11.79)
a curve in space are regarded as
* dt * dt dt dt dt
small lengths along the curve itself
the three-space analogue of 11.73. Once again we are led to the fact that when s is used as
z parameter along C , then
C
B dr dx dy dz
T̂ = = î + ĵ + k̂ (11.80)
ds ds ds ds
ds is a unit tangent vector to C . In addition, multiplication of 11.79 by dt yields
(
A dr !
)
dt dr |d r| = ds = (dx)2 + (dy)2 + (dz)2 , (11.81)
x indicating that small lengths ds along C (Figure 11.110) are defined in terms of small lengths
y
|d r| along the tangent line to C .
EXAMPLE 11.44
Find a unit tangent vector to the curve
C : x = sin t, y = 2 cos t, z = 2t/π, t ≥0
at the point (0, −2, 2) .
SOLUTION A tangent vector to C at any point is
" #
dr dx dy dz 2
= î + ĵ + k̂ = cos t î − 2 sin t ĵ + k̂.
dt dt dt dt π
Since t = π yields the point (0, −2, 2) , a tangent vector at this point is
r, (π ) = −î + (2/π )k̂.
A unit tangent vector is then
−î + (2/π )k̂ −π î + 2k̂
T̂ = ! = √ .
1 + 4/π 2 4 + π2
EXAMPLE 11.45
Find the length of that part of the curve x = y 2/3 , x = z2/3 between the points (0, 0, 0) and
(4, 8, 8) .
SOLUTION If we use x = t , y = t 3/2 , z = t 3/2 , 0 ≤ t ≤ 4, as parametric equations for
the curve, then
+" # #2
" #2 " + " #2 " #2
2 2 4
4
dy 2 dx dz 3√ 3√
L= + + dt = 1+ t + t dt
0 dt dt dt 0 2 2
2 40 3 " #3/2 4 4
9t 4 9t 4 √
= 1 + dt = 1+ = (19 19 − 1).
0 2 27 2 0 27
EXERCISES 11.11
In Exercises 1–5 express the curve in vector form and find the unit In Exercises 11–14 find the length of the curve. Draw each curve.
tangent vector T̂ at each point on the curve.
11. x = 2 cos t, y = 2 sin t, z = 3t, 0 ≤ t ≤ 2π
1. x = sin t, y = cos t, z = t, −∞ < t < ∞
2. x = t, y = t 2 , z = t 3 , t ≥ 1 12. x = 2 − 5t, y = 1 + t, z = 6 + 4t, −1 ≤ t ≤ 0
3. x = (t − 1)2 , y = (t + 1)2 , z = −t, −3 ≤ t ≤ 4 13. x = t 3 , y = t 2 , z = t 3 , 0 ≤ t ≤ 1

4. x + y = 5, x 2 − y = z from (5, 0, 25) to (0, 5, −5) 14. x = t, y = t 3/2 , z = 4t 3/2 , 1 ≤ t ≤ 4
2 2
5. x + y + z = 4, x + y = 4, y ≥ 0 from (2, 0, 2) to (−2, 0, 6)
∗ 15. In Definition 11.18 why are the derivatives assumed not to vanish
simultaneously? Hint: Consider the curve x = t 3 , y = t 2 , z = 0.
In Exercises 6–10 find T̂ at the point.
6. √x =√4 cos ∗ 16. Find a unit tangent vector to the curve r = (cos t + t sin t)î +
√ t, y = 6 sin t, z = 2 sin t, −∞ < t < ∞ ;
(2 2, 3 2, 2) (sin t − t cos t)ĵ , 0 ≤ t ≤ 2π , called an involute of a circle.
7. x = 2 − 5t, y = 1 + t, z = 6 + 4t, −∞ < t < ∞ ; (7, 0, 2) ∗ 17. Show that the tangent vector d r/dt to a curve described by equa-
!
2 2
8. x + y + z = 4, z = 2
x2 + y2 , directed so that x increases tion 11.64 always points in the direction in which t increases along the
√ curve.
when y is positive; (1, 1, 2)
9. x = y 2 + 1, z = x + 5, directed so that y increases along the
∗ 18. (a) What happens when equation 11.66 is used to determine
curve; (5, 2, 10) a tangent vector to the curve x = t 2 , y = t 3 , z = t 2 ,
−∞ < t < ∞ , at the origin?
10. x 2 + (y − 1)2 = 4, z = x , directed so that z decreases when y
is negative; (2, 1, 2) (b) Can you devise a way in which to obtain a tangent vector?
11.12 Normal Vectors, Curvature, and Radius of Curvature

In discussing curves we distinguish between two types of properties: intrinsic and not intrinsic.
An intrinsic property is one that is independent of the parameter used to specify the curve; a
property that is not intrinsic is parameter dependent. To illustrate, the tangent vector T = d r/dt
in 11.66 is not intrinsic; a change of parameter results in a change in the length of T. The unit
tangent vector T̂, on the other hand, is intrinsic; there is only one unit tangent vector in the
direction of the curve. The length of a curve from its initial point to an arbitrary point is an
intrinsic property; a change of parameter along the curve does not affect length between points.
Because length along a curve is an intrinsic property, it is customary in theoretical discus-
sions to use it as the parameter by which to specify points on the curve. When C is a smooth
curve in the xy -plane, parametric equations for C in terms of length s along C take the form
C : x = x(s), y = y(s), 0 ≤ s ≤ L. (11.82)
FIGURE 11.111 There is

essentially one normal direction to Normal Vectors to Curves
a curve in the xy -plane
The normal line at a point P on a smooth curve C in the xy -plane is that line which is
y Normal
vector perpendicular to the tangent line to C at P (Figure 11.111). Any vector along this normal line
C is said to be a normal vector to the curve at P . Since the unit tangent vector to C at P is
P T
dr dx dy
Normal T̂ = = î + ĵ,
line ds ds ds
x it follows that
dy dx
N̂ = − î + ĵ (11.83)
ds ds
is a unit normal vector to C at P (note that T̂ · N̂ = 0). Because there is only one direction
normal to C at P , every normal vector to C at P must be some multiple λN̂ of N̂.
The situation is quite different for space curves (Figure 11.112). If T̂ is the unit tangent
FIGURE 11.112 There
is a plane of normal vectors to a vector to a smooth curve C , then there is an entire plane of normal vectors to C at P . In the
curve in space following discussion, we single out two normal vectors called the principal normal and the
binormal. Suppose that C is defined parametrically by
z
C : x = x(s), y = y(s), z = z(s), 0 ≤ s ≤ L, (11.84)
P
C and that T̂ is the unit tangent vector to C defined by 11.80. Because T̂ has unit length,
T 1 = T̂ · T̂.
x
y
If we use equation 11.59b to differentiate this equation with respect to s , we have
" #
d T̂ d T̂ d T̂
0 = · T̂ + T̂ · = 2 T̂ · .
ds ds ds
But if neither of the vectors T̂ nor d T̂/ds is equal to zero, then the fact that their scalar product
is equal to zero implies that they are perpendicular. In other words,
FIGURE 11.113 Principal
normal and binormal to a curve d T̂
N = (11.85)
z ds
C
N is a normal vector to C at any point. The unit normal vector in this direction,
B N d T̂/ds
N̂ = = , (11.86)
T |N | |d T̂/ds|
x y is called the principal normal (vector) to C (Figure 11.113).
Because N̂ is defined in terms of intrinsic properties T̂ and s for a curve, it must also be an
intrinsic property. It follows, then, that no matter what parameter is used to specify points on a
curve, N̂ is always the same. But how do we find N̂ when a curve C is specified in terms of a
parameter other than length along C , say, in the form
C : x = x(t), y = y(t), z = z(t), α ≤ t ≤ β? (11.87)
If s = s(t) is length along C (measured from t = α ), then by the chain rule
d T̂ d T̂ dt
= ,
ds dt ds
where dt/ds must be positive since both s and t increase along C . Consequently,
d T̂/ds (d T̂/dt)(dt/ds) (d T̂/dt)(dt/ds) d T̂/dt

N̂ = = = = . (11.88)
|d T̂/ds| |(d T̂/dt)(dt/ds)| |d T̂/dt|dt/ds |d T̂/dt|
In other words, for any parametrization of C whatsoever, the vector d T̂/dt always points in the
direction of the principal normal, and to find N̂, we simply find the unit vector in the direction
of d T̂/dt .
In the study of space curves, a second normal vector to C , called the binormal (vector), is
defined by
B̂ = T̂ × N̂. (11.89)
Since the cross product of two vectors is always perpendicular to each of the vectors, it follows
that the binormal is perpendicular to both T̂ and N̂, and must therefore indeed be a normal vector
to C (Figure 11.113).
We have singled out three vectors at each point P on a curve C : a unit tangent vector T̂
and two unit normal vectors N̂ and B̂. As P moves along C , these vectors constantly change
direction but always have unit length.
EXAMPLE 11.46
Find T̂, N̂, and B̂ for the curve x = t , y = t 2 , z = t 2 , t ≥ 0.
SOLUTION The unit tangent vector T̂ is defined by
d r/dt (1, 2t, 2t) (1, 2t, 2t)

T̂ = = √ = √ .
|d r/dt| 2
1 + 4t + 4t 2 1 + 8t 2
The principal normal N̂ lies along the vector N = d T̂/dt , and according to equation 11.59a,
we can write that
" #
d T̂ d 1 1 d
N = = √ (1, 2t, 2t) + √ (1, 2t, 2t)
dt dt 1 + 8t 2 1 + 8t 2 dt
−8t 1
= 2 3 2
(1, 2t, 2t) + √ (0 , 2 , 2 )
(1 + 8t ) /
1 + 8t 2
1
= [−8t (1, 2t, 2t) + (1 + 8t 2 )(0, 2, 2)]
(1 + 8t 2 )3/2
(−8t, 2, 2)
= .
(1 + 8t 2 )3/2
The principal normal is therefore
N (−8t, 2, 2) (−4t, 1, 1)
N̂ = = √ = √ .
|N | 64t 2 +4+4 2 + 16t 2
The binormal is
(1, 2t, 2t) (−4t, 1, 1)

B̂ = T̂ × N̂ = √ ×√
1 + 8t 2 2 + 16t 2
* *
1* î ĵ k̂ *
* *
= √ √ * 1 2t 2t *
1 + 8t 2 + 16t * −4t
2 2
1 1
*
1
= √ √ √ (0, −1 − 8t 2 , 1 + 8t 2 )
2 1 + 8t 2 1 + 8t 2
1 + 8t 2
= √ (0, −1, 1)
2(1 + 8t 2 )
(0, −1, 1)
= √ .
2
The significance of the fact that the binormal has constant direction can be seen from a
drawing of the curve. Because the parametric equations imply that y = x 2 and z = y , the
curve is the curve of intersection of these two surfaces (Figure 11.114). Since the curve lies in
the plane −y + z = 0, and a normal vector to this plane is (0, −1, 1) , it follows that (0, −1, 1)
is always normal to the curve. But this is precisely the direction of B̂. In other words, constant
B̂ implies that the curve lies in a plane that has B̂ as normal (see Exercise 30).
FIGURE 11.114 Tangent, principal normal, and

binormal to the curve x = t , y = t 2 , z = t 2 , t ≥ 0
z z=y
y = x2
x
EXAMPLE 11.47
Show that for a smooth curve C : x = x(t) , y = y(t) , α ≤ t ≤ β in the xy -plane, the
principal normal is
" #
dy d 2 x dx d 2 y (dy/dt, −dx/dt)
N̂ = sgn 2
− 2
! ,
dt dt dt dt (dx/dt)2 + (dy/dt)2
where the signum function sgn( u ) is defined by


 1, if u > 0,
sgn(u) = 0, if u = 0,

−1, if u < 0.
SOLUTION The unit tangent vector to C is
(dx/dt, dy/dt)
T̂ = ! .
(dx/dt)2 + (dy/dt)2
For simplicity in notation, we use a dot “ . ” above a variable to indicate that the variable is
differentiated with respect to t . For example, ẋ = dx/dt and ẍ = d 2 x/dt 2 . With this
notation,
(ẋ, ẏ)
T̂ = ! .
ẋ 2 + ẏ 2
By equation 11.88, N̂ = (d T̂/dt)/|d T̂/dt| , where

" #
d T̂ d (ẋ, ẏ) d 1 1
= ! = ! (ẋ î + ẏ ĵ) + ! (ẍ î + ÿ ĵ)
dt 2
dt ẋ + ẏ 2 dt 2
ẋ + ẏ 2 ẋ + ẏ 2
2
- .
−ẋ ẍ − ẏ ÿ 1
= 2 2 3 2
(ẋ î + ẏ ĵ) + ! (ẍ î + ÿ ĵ)
(ẋ + ẏ ) /
ẋ + ẏ 2
2
1
= [−(ẋ ẍ + ẏ ÿ)(ẋ î + ẏ ĵ) + (ẋ 2 + ẏ 2 )(ẍ î + ÿ ĵ)]
(ẋ 2 + ẏ 2 )3/2
1
= [(−ẋ 2 ẍ − ẋ ẏ ÿ + ẋ 2 ẍ + ẏ 2 ẍ)î
(ẋ 2 + ẏ 2 )3/2
+ (−ẋ ẏ ẍ − ẏ 2 ÿ + ẋ 2 ÿ + ẏ 2 ÿ)ĵ]
1
= [ẏ(ẏ ẍ − ẋ ÿ)î + ẋ(ẋ ÿ − ẏ ẍ)ĵ]
(ẋ 2 + ẏ 2 )3/2
ẏ ẍ − ẋ ÿ
= (ẏ î − ẋ ĵ).
(ẋ 2 + ẏ 2 )3/2
If ẏ ẍ − ẋ ÿ is positive, then
ẏ î − ẋ ĵ
N̂ = ! ;
ẋ 2 + ẏ 2
whereas if ẏ ẍ − ẋ ÿ is negative, then
−ẏ î + ẋ ĵ
N̂ = ! .
ẋ 2 + ẏ 2
In other words,
" #
ẏ î − ẋ ĵ
N̂ = sgn(ẏ ẍ − ẋ ÿ) ! .
ẋ 2 + ẏ 2
Curvature and Radius of Curvature

When length s along a smooth curve C is used as the parameter by which to identify points
on the curve, the vector T̂ = d r/ds is a unit tangent vector to C . Suppose we differentiate
T̂ with respect to s to form d T̂/ds = d 2 r/ds 2 . Since T̂ has constant unit length, only its
direction can change; therefore, the derivative d T̂/ds must be a measure of the rate of change
of the direction of T̂. Since T̂ is really our way of specifying the direction of the curve itself,
we can also say that d T̂/ds is a measure of how fast the direction of C changes. But exactly
how does a vector d T̂/ds that has both magnitude and direction measure the rate of change of
the direction of C ? We illustrate by example that it cannot be the direction of d T̂/ds ; it must
be its magnitude that measures the rate of change of the direction of C . In Figure 11.115 we
show a number of circles in the xy -plane, all of which are tangent to the y -axis at the origin.
Parametric equations for the circle with centre (R, 0) and radius R in terms of length s along
the circle [as measured from (2R, 0) ] are
x = R + R cos(s/R), y = R sin(s/R), 0 ≤ s < 2π R.
Consequently,
" # " #
dr s s
T̂ = = − sin î + cos ĵ
ds R R
and " # " #

d T̂ 1 s 1 s
= − cos î − sin ĵ.
ds R R R R
At the origin, s = π R , and
d T̂ ** 1
* = î.
ds s=π R R
Thus, for each of the circles in Figure 11.115, the vector d T̂/ds has exactly the same direction.
FIGURE 11.115 Circles
tangent to the y -axis at (0, 0)
Yet the rate of change of the direction of T̂ is not the same for each circle; the direction changes
more rapidly as the radius of the circle decreases. We must conclude, therefore, that it cannot be
y the direction of d T̂/ds that measures the rate of change of T̂. Since a vector has only length and
direction, it must be the length of d T̂/ds that measures this rate of change. The circles in Figure
11.115 certainly support this claim; the length of d T̂/ds is 1/R , and this quantity increases
R as the radii of the circles decrease. This agrees with the fact that the rate at which T̂ turns
R (2R, 0) x increases as R decreases. According to the following definition, we call |d T̂/ds| curvature
and 1/|d T̂/ds| radius of curvature.
DEFINITION 11.20
If x = x(s) , y = y(s) , z = z(s) , 0 ≤ s ≤ L , are parametric equations for a smooth
curve in terms of length s along the curve, we define the curvature of the curve at a point
as * *
* d T̂ *
κ(s) = ** **, (11.90)
ds
its radius of curvature as
1
ρ(s) = , (11.91)
κ(s)
and its circle of curvature as that circle in the plane of T̂ and N̂ with centre at r(s)+ρ(s)N̂
and radius ρ(s) .
The circle of curvature is illustrated in Figure 11.116.

For the circles in Figure 11.115 we have already shown that d T̂/ds = R −1 î. Consequently,
FIGURE 11.116 Circle
of curvature of a curve
for these circles, the curvature is always R −1 , and the radius of curvature is R , the radius of
the circle. In other words, for a circle, the circle of curvature is the circle itself, its radius of
z Circle of curvature is its radius, and its curvature is the inverse of its radius. For the case when a curve is
curvature
not a circle, we show in Exercise 27 that at any point on the curve the circle of curvature is in
some sense the best-fitting circle to the curve at that point.
N
Because curvature and radius of curvature have been defined in terms of intrinsic properties
N
N T̂ and s for a curve, they must also be intrinsic properties. It follows, then, that no matter what
C parameter is used to specify points on a curve, curvature and radius of curvature are always the
T
same. The following theorem shows how to calculate κ and ρ when the curve is specified in
r terms of a parameter other than length along the curve.
x y THEOREM 11.11
When a smooth curve is defined parametrically by
C : x = x(t), y = y(t), z = z(t), α ≤ t ≤ β,
its curvature κ(t) is given by

|ṙ × r̈|
κ(t) = , (11.92)
|ṙ|3
where ṙ = d r/dt and r̈ = d 2 r/dt 2 .
PROOF If s(t) is length along C (measured from t = α ), then by the chain rule
* * * *
* d T̂ * * d T̂ dt *
*
κ =* *=* * * * = |d T̂/dt| = |d T̂/dt| = |d T̂/dt|
ds dt ds * |ds/dt| ds/dt |ṙ|
(see equation 11.79). Now, we can write ṙ in the form ṙ = |ṙ|T̂; therefore, using 11.59a, we
have
" #
d d T̂
r̈ = |ṙ| T̂ + |ṙ|
dt dt
" #
d d T̂/dt
= |ṙ| T̂ + (|ṙ||d T̂/dt|)
dt |d T̂/dt|
" # " * *#
d * d T̂ *
= |ṙ| T̂ + |ṙ|** * N̂.
dt dt *
If we take the cross product of this vector with ṙ, we get
" # " * *#
d * d T̂ *
ṙ × r̈ = |ṙ| ṙ × T̂ + |ṙ|** * ṙ × N̂
dt dt *
" * *#
* d T̂ *
= |ṙ|** ** ṙ × N̂. (since ṙ is parallel to T̂)
dt
Because ṙ is perpendicular to N̂, it follows that |ṙ × N̂| = |ṙ||N̂| sin(π/2) = |ṙ| , and therefore
" * *#
* d T̂ *
|ṙ × r̈| = |ṙ|** * |ṙ|.
dt *
Consequently,
* *
* d T̂ *
* * = |ṙ × r̈|
* dt * |ṙ|2
and
|ṙ × r̈|
κ = κ(t) = .
|ṙ|3
For the radius of curvature, we have the following.
COROLLARY 11.11.1
When a smooth curve is defined in terms of an arbitrary parameter t ,
|ṙ|3
ρ(t) = . (11.93)
|ṙ × r̈|
EXAMPLE 11.48
Find curvature and radius of curvature for the curve in Example 11.46.
|ṙ × r̈| |(1, 2t, 2t) × (0, 2, 2)|

κ(t) = =
|ṙ|3 |(1, 2t, 2t)|3
** **
1 ** î ĵ k̂ ** 1
** **
= 2 2 3 2
** 1 2 t 2 t ** = |(0, −2, 2)|
( 1 + 4 t + 4 t ) ** 0 2 2 **
/ (1 + 8t 2 )3/2
√
2 2
= ;
(1 + 8t 2 )3/2
1 (1 + 8t 2 )3/2
ρ(t) = = √ .
κ(t) 2 2
Note in particular that as t increases, so does ρ , a fact that is certainly supported by Figure
11.114.
EXAMPLE 11.49
Show that for a smooth curve y = y(x) in the xy -plane,
|y ,, |
κ(x) = .
[1 + (y , )2 ]3/2
SOLUTION When we use x as parameter along the curve y = y(x) , parametric equations
are x = x , y = y(x) . Then
ṙ = (1, y , (x)), r̈ = (0, y ,, (x)),
and * *
* î ĵ k̂ *
* *
ṙ × r̈ = * 1 y , 0 * = y ,, k̂.
* 0 y ,, 0 *
Thus,
|ṙ × r̈| |y ,, | |y ,, |
κ(x) = = = .
|ṙ|3 |(1, y , )|3 [1 + (y , )2 ]3/2
According to equations 11.86 and 11.90, we can write that

* *
d T̂ * d T̂ *
* *
= * * N̂ = κ N̂. (11.94)
ds * ds *
This is called the first Frenet–Serret formula of differential geometry. We now derive the
second of these formulas. Differentiation of B̂ = T̂ × N̂ with respect to s gives
d B̂ d T̂ d N̂ d N̂ d N̂
= × N̂ + T̂ × = κ N̂ × N̂ + T̂ × = T̂ × .
ds ds ds ds ds
This equation implies that d B̂/ds is perpendicular to T̂ (and d N̂/ds ). Differentiation of

B̂ · B̂ = 1 with respect to s immediately implies that d B̂/ds is perpendicular to B̂ also. It
follows then that d B̂/ds must be some multiple of N̂ ( N̂ being perpendicular to T̂ and B̂), and
we may therefore write that
d B̂
= −τ N̂. (11.95)
ds
This is the second Frenet–Serret formula. Quantity τ is called the torsion of the curve. We
can find formulas for τ in terms of s and in terms of an arbitrary parameter along the curve.
The scalar product of equation 11.95 with N̂ gives
% &
d B̂ d d N̂ d T̂
τ =− · N̂ = −N̂ · (T̂ × N̂) = −N̂ · T̂ × + × N̂
ds ds ds ds
% &
d N̂ d N̂
= −N̂ · T̂ × + κ N̂ × N̂ = −N̂ · T̂ × .
ds ds
Since T̂ = d r/ds and N̂ = κ −1 d T̂/ds = κ −1 d 2 r/ds 2 , it follows that

" #
1 d 2r
dr 1 d 3r 1 dκ d 2 r
τ =− 2
· × −
κ ds ds κ ds 3 κ 2 ds ds 2
1 d 2r dr d 3r 1 dκ d 2 r d r d 2r
=− · × 3 + 3 · ×
κ 2 ds 2 ds ds κ ds ds 2 ds ds 2
1 dr d 2r d 3r
= · × , (11.96)
κ 2 ds ds 2 ds 3
which expresses τ in terms of s . To express it in terms of an arbitrary parameter t along the
curve rather than length s , we use
dr dt d r dt
= = ṙ,
ds ds dt ds
# " " #2
d 2r d 2t d r dt d d r dt d 2t dt
2
= 2
+ = 2
ṙ + r̈,
ds ds dt ds dt dt ds ds ds
" # " #" 2 # 2 " #2 " #
d 3r d 3t d r d 2 t d d r dt dt d t d r dt d d 2 r dt
= + 2 +2 +
ds 3 ds 3 dt ds dt dt ds ds ds 2 dt 2 ds dt dt 2 ds
" #3
d 3t d 2 t dt dt ···
= ṙ + 3 2 r̈ + r.
ds 3 ds ds ds
With these,
" # / 2 " #2 1 / 3 " #3 1
1 dt d t dt d t d 2 t dt dt ···
τ = ṙ · ṙ + r̈ × ṙ + 3 2 r̈ + r
κ2 ds ds 2 ds ds 3 ds ds ds
" #6
1 dt |ṙ|6 1
= (ṙ · r̈ × ···
r) = (ṙ · r̈ × ···
r),
κ2 ds |ṙ × r̈|2 |ṙ|6
and hence,
ṙ · r̈ × ···
r
τ (t) = . (11.97)
|ṙ × r̈|2
The third Frenet–Serret formula expresses d N̂/ds in terms of T̂ and N̂; it is developed in
Exercise 29.
EXERCISES 11.12
In Exercises 1–5 find N̂ and B̂ at each point on the curve. (b) F = t 2 î + t 4 ĵ is a vector that is defined at each point P on
C . Denote by FT and FN the components of F in the di-
1. x = sin t, y = cos t, z = t, −∞ < t < ∞
rections T̂ and N̂ at P . Find FT and FN as functions of t .
2. x = t, y = t 2 , z = t 3 , t ≥ 1
3. x = (t − 1)2 , y = (t + 1)2 , z = −t, −3 ≤ t ≤ 4 (c) Express F in terms of T̂ and N̂ .
4. x + y = 5, x 2 − y = z , from (5, 0, 25) to (0, 5, −5)

∗ 24. Repeat Exercise 23 for the curve C : x = 2 cos t, y = 2 sin t ,
5. z = x, x 2 + y 2 = 4, y ≥ 0, from (2, 0, 2) to (−2, 0, −2)
and the vector F = x 2 î + y 2 ĵ .
In Exercises 6–10 find N̂ and B̂ at the point. ∗ 25. The vectors T̂ , N̂ , and B̂ were calculated at each point on the curve
x = t, y = t 2 , z = t 2 in Example 11.46. If F = t 2 î + 2t ĵ − 3k̂
6. √x =√4 cos
√ t, y = 6 sin t, z = 2 sin t, −∞ < t < ∞ ; is a vector defined along C , find the components of F in the directions
(2 2, 3 2, 2) T̂ , N̂ , and B̂ . Express F in terms of T̂ , N̂ , and B̂ .
7. x = 2 − 5t, y = 1 + t, z = 6 + 4t 3 , −∞ < t < ∞ ; (7, 0, 2)
! ∗ 26. Calculate T̂ , N̂ , and B̂ for the curve x = cos t, y = sin t, z = t .
8. x 2 + y 2 + z2 = 4, z = x 2 + y 2 , directed so that x increases
√ Express the vector F = x î + xy 2 ĵ + k̂ in terms of T̂ , N̂ , and B̂ .
when y is positive; (1, 1, 2)
9. x = y 2 + 1, z = x + 5, directed so that y increases along the ∗ 27. In this exercise we discuss our claim that the circle of curvature is
curve; (5, 2, 10) the best-fitting circle to the curve at a point.
2 2
10. x + (y − 1) = 4, x = z , directed so that z decreases when y
(a) Is it true that the circle of curvature at a point on a curve
is negative; (2, 1, 2)
passes through that point?
In Exercises 11–18 find the curvature and the radius of curvature of (b) Show that the circle of curvature and curve share the same
the curve (if they exist). Draw each curve. tangent line at their common point.
11. (x − h)2 + (y − k)2 = R 2 , z = 0, directed counterclockwise (c) Verify that the circle of curvature and curve have the same
curvature at their common point.
12. x = x0 + at, y = y0 + bt, z = z0 + ct, −∞ < t < ∞
( x0 , y0 , z0 , a, b, c all constants)
∗ 13. x = t, y = t 2 , z = 0, t ≥ 0 ∗ 28. If φ is the angle between î and T̂ for a curve in the xy -plane
(figure below), show that
∗ 14. x = e cos t, y = e sin t, z = t, −∞ < t < ∞
t t
∗ 15. x = t, y = t 3 , z = t 2 , t ≥ 0 * *
* dφ *
∗ 16. x = 2 cos t, y = 2 sin t, z = 2 sin t, 0 ≤ t < 2π κ(s) = ** **.
ds
∗ 17. x = t + 1, y = t 2 − 1, z = t + 1, −∞ < t < ∞
∗ 18. x = t 2 , y = t 4 , z = 2t, −1 ≤ t ≤ 5 y
∗ 19. At which points on the ellipse b2 x 2 +a 2 y 2 = a 2 b2 ( a > b ) is the
curvature a maximum, and at which points is the curvature a minimum?
C T
∗ 20. Show that curvature for a smooth curve x = x(t), y = y(t), α ≤
i
t ≤ β , in the xy -plane can be expressed in the form
* *
* dy d 2 x dx d 2 y **
* −
* dt dt 2 dt dt 2 * x
κ(t) = /" # " #2 1 3 / 2
.
dx 2 dy
+ ∗∗ 29. The third Frenet–Serret formula is
dt dt
∗ 21. Show that the only curves for which curvature is identically equal
to zero are straight lines. d N̂
= τ B̂ − κ T̂.
∗ 22. What happens to curvature at a point of inflection on the graph of ds
a function y = f (x) ?
∗ 23. Let C be the curve x = t, y = t 2 in the xy -plane. Verify this result by showing that N̂ = B̂ × T̂ and then calculating
(a) At each point on C calculate the unit tangent vector T̂ and
d N̂/ds .
the principal normal N̂ . What is B̂ ? (See Example 11.47 ∗∗ 30. Show that a curve lies in a plane if and only if its torsion vanishes.
for N̂ .)
11.13 Displacement, Velocity, and Acceleration

In Sections 4.8 and 5.2 we introduced the concepts of displacement, velocity, and acceleration
for moving objects, but indicated that our terminology at that time was somewhat loose. In
particular, we stated that if x = x(t) represents the position of a particle moving along the
x -axis, then the instantaneous velocity of the particle is
dx
v = , (11.98)
dt
provided, of course, that t is time, and the acceleration of the particle is
dv d 2x
a = = . (11.99)
dt dt 2
We illustrated by examples that given any one of x(t) , v(t) , or a(t) and sufficient initial
conditions, it is always possible to find the other two. There was nothing wrong with the
calculations in the examples — they were correct — but our terminology was not quite correct.
We now rectify this situation and give precise definitions of velocity and acceleration.
Suppose a particle moves along some curve C in space (under perhaps the influence of
various forces), and that C is defined as a function of time t by the parametric equations
C : x = x(t), y = y(t), z = z(t), t ≥ 0. (11.100)
The position of the particle can then be described as a function of time by its position or
displacement vector:
r = r(t) = x(t)î + y(t)ĵ + z(t)k̂, t ≥ 0. (11.101)
The velocity v of the particle at any time t is defined as the time rate of change of its displacement
vector:
dr
v = . (11.102)
dt
Velocity, then, is a vector, and because of Theorem 11.8, the components of velocity are the
derivatives of the components of displacement:
dr dx dy dz
v = = î + ĵ + k̂. (11.103)
dt dt dt dt
But according to Theorem 11.10, the vector d r/dt is tangent to the curve C (Figure 11.117).
FIGURE 11.117 Velocity
is always tangent to curve along
In other words, if a particle is at position P , and we draw its velocity vector with tail at P , then
which the particle travels v is tangent to the trajectory.
In some applications it is the length or magnitude of velocity that is important, not its
z direction. This quantity, called speed, is therefore defined by
C +" #2 " #2 " #2
dx dy dz
v |v | = + + . (11.104)
P dt dt dt
r
Equation 11.79 implies that if s(t) is length along the trajectory C [where s(0) = 0], then
x |v| = ds/dt . In other words, speed is the time rate of change of distance travelled along C .
y
It is important to understand this difference between velocity and speed. Velocity is the
time derivative of displacement; speed is the time derivative of distance travelled. Velocity is a
vector; speed is a scalar — the magnitude of velocity.
The acceleration of the particle as it moves along the curve C in equations 11.100 is defined
as the rate of change of velocity with respect to time:
dv d 2r d 2x d 2y d 2z
a = = 2 = î + ĵ + k̂. (11.105)
dt dt dt 2 dt 2 dt 2
Acceleration, then, is also a vector; it is the derivative of velocity, and therefore its components are
the derivatives of the components of the velocity vector. Alternatively, it is the second derivative
of displacement and has components that are the second derivatives of the components of the
displacement vector.
In the special case in which C is a curve in the xy -plane, definitions of displacement,
velocity, speed, and acceleration become, respectively,
r = x(t)î + y(t)ĵ, (11.106a)

dr dx dy
v = = î + ĵ, (11.106b)
dt dt dt
+" # " #2
dx 2 dy
|v | = + , (11.106c)
dt dt
dv d 2r d 2x d 2y
a = = 2 = î + ĵ. (11.106d)
dt dt dt 2 dt 2
For motion along the x -axis,
r = x(t)î, (11.107a)
dr dx
v = = î, (11.107b)
dt dt
* *
* dx *
|v| = ** **, (11.107c)
dt
dv d 2r d 2x
a = = 2 = î. (11.107d)
dt dt dt 2
If we compare equations 11.107b and d with equations 11.98 and 11.99, we see that for motion
along the x -axis, x(t) , v(t) , and a(t) are the components of the displacement, velocity, and
acceleration vectors, respectively. Because these are the only components of r(t) , v(t) , and
a(t) , it follows that consideration of the components of the vectors is equivalent to consideration
of the vectors themselves. For one-dimensional motion, then, we can drop the vector notation
and work with components (and this is precisely the procedure that we followed in Sections 4.8
and 5.2).
Newton’s second law describes the effects of forces on the motion of objects. It states that
if an object of mass m is subjected to a force F, then the time rate of change of its momentum
(mv) is equal to F:
d
F = (mv). (11.108)
dt
In most cases, the mass of the object is constant, and this equation then yields its acceleration:
dv
F = m = ma . (11.109)
dt
If F is known as a function of time t , F = F(t) , then 11.109 defines the acceleration of the
object as a function of time,
1
a(t) = F(t),
m
and integration of this equation leads to expressions for the velocity v(t) and position r(t) as
functions of time.
EXAMPLE 11.50
A projectile is fired at angle θ to the horizontal with speed v0 (Figure 11.118). Find the distance
R from the firing place that the projectile strikes the ground (called the range of the projectile).
What is the maximum height attained by the projectile?
FIGURE 11.118 Path of a projectile
V0
x
R
SOLUTION Since the acceleration of the projectile is d v/dt = a = −g ĵ, its velocity is
v(t) = −gt ĵ + C. If we choose t = 0 at the instant the projectile is fired, then v(0) = v0 =
dr
v0 cos θ î + v0 sin θ ĵ, and therefore v0 = v(0) = C. Integration of = −gt ĵ + v0 gives
dt
1
r(t) = − gt 2 ĵ + v0 t + D.
2
Since r(0) = 0, it follows that D = 0, and

" #
1 2 1 2
r(t) = − gt ĵ + v0 t = (v0 cos θ t)î + − gt + v0 sin θ t ĵ.
2 2
The projectile strikes the ground when r(t) = R î, in which case
" #
1 2
R î = (v0 cos θ t)î + − gt + v0 sin θ t ĵ.
2
When we equate components,
1
R = v0 cos θ t, 0 = − gt 2 + v0 sin θ t.
2
The second of these implies that t = (2v0 /g) sin θ , and when this is substituted into the first,
" #
2v0 sin θ v02 sin 2θ
R = v0 cos θ = .
g g
The projectile attains maximum height when the y -component of its velocity is zero, 0 =
−gt + v0 sin θ &⇒ t = (v0 /g) sin θ . The height of the shell at this time is the y -component
of its displacement,
" #2 " #
1 v0 sin θ v0 sin θ v02 sin2 θ
− g + v0 sin θ = .
2 g g 2g
EXAMPLE 11.51
A particle starts at time t = 0 from position (1, 1) with speed 2 m/s in the negative y -direction.
It is subjected to an acceleration that is given as a function of time by
1
a(t) = √ î + 6t ĵ m/s2 .
t +1
Find its velocity and position as functions of time.
√
SOLUTION If a = d v/dt = (1/ t + 1)î + 6t ĵ, then
√
v = 2 t + 1 î + 3t 2 ĵ + C,
where C is some constant vector. Because the initial velocity of the particle is 2 m/s in the
negative y -direction, v(0) = −2ĵ. Consequently, −2ĵ = 2î + C &⇒ C = −2î − 2ĵ. The
velocity, then, of the particle at any time t ≥ 0 is
√
v(t) = (2 t + 1 − 2)î + (3t 2 − 2)ĵ m/s.
Because v = d r/dt , integration gives

- .
4 3/2
r = (t + 1) − 2t î + (t 3 − 2t)ĵ + D.
3
Since the particle starts from position (1, 1) , r(0) = î + ĵ, and
4 1
î + ĵ = î + D, or D = − î + ĵ.
3 3
The displacement of the particle is therefore
- .
4 3/2 1
r(t) = (t + 1) − 2t − î + (t 3 − 2t + 1)ĵ m.
3 3
EXAMPLE 11.52
The mass M in Figure 11.119a is dropped from point B . Show that if a mass m is fired from
any position A directly at M at the instant M is released, m will always collide with M .
FIGURE 11.119a Mass FIGURE 11.119b Coor-

m is fired from A at mass M at dinate system to analyze motions
B as M is dropped of masses m and M
B y
H B
M
m M
v m
A
A x
h
SOLUTION We choose the coordinate system in Figure 11.119b, and take time t = 0 at the
instant both masses begin motion. To show that the masses collide, we show that they have
the same displacement vector for some time t . The acceleration of each mass is a = −9.81ĵ.
Integration of this gives velocities of the masses,
vm = −9.81t ĵ + C, vM = −9.81t ĵ + D.
Since M is dropped, its initial velocity is zero, vM (0) = O. This implies that D = O. If m
is fired at angle θ with speed v > 0, then vm (0) = v cos θ î + v sin θ ĵ. This implies that
C = v cos θ î + v sin θ ĵ. Integrations of
d rm d rM
= −9.81t ĵ + v cos θ î + v sin θ ĵ and = −9.81t ĵ
dt dt
give
rm = v cos θ t î + (−4.905t 2 + v sin θ t)ĵ + E and rM = −4.905t 2 ĵ + F.
Since rm (0) = O and rM (0) = hî + H ĵ, we obtain E = O and F = hî + H ĵ. Thus,
rm = v cos θ t î + (−4.905t 2 + v sin θ t)ĵ and rM = hî + (H − 4.905t 2 )ĵ. The masses
collide if and when
rm = rM ⇐⇒ v cos θ t î + (−4.905t 2 + v sin θ t)ĵ = hî + (H − 4.905t 2 )ĵ.
v cos θ t = h, −4.905t 2 + v sin θ t = H − 4.905t 2 ,
from which
h H
t = and t = .
v cos θ v sin θ
These are compatible since tan θ = H / h . The time for collision to occur increases as h
increases, v decreases, and/or θ increases.
Tangential and Normal Components of Velocity and

Acceleration
For some types of motion it is inconvenient to express velocity and acceleration of a particle
in terms of Cartesian components; sometimes it is an advantage to resolve these vectors into
components that are tangent and normal to the path of the particle. When the trajectory C of a
particle is specified as a function of time t by 11.100, its velocity v = d r/dt is tangent to C ,
and we can therefore write
v = |v|T̂. (11.110)
In other words, the tangential component of velocity is speed, and v has no component
normal to the trajectory. Differentiation of this equation gives the particle’s acceleration:
" #
dv d d T̂
a = = |v| T̂ + |v|
dt dt dt
" # " * *#
d * d T̂ * d T̂/dt
= |v| T̂ + |v|** **
dt dt |d T̂/dt|
" # " * *#
d * d T̂ *
= |v| T̂ + |v|** ** N̂. (11.111)
dt dt
FIGURE 11.120 Tangential and

We have therefore expressed a in terms of the unit tangent vector T̂ to C and the princi-
normal components of acceleration pal normal N̂ (Figure 11.120). We call d(|v|)/dt and |v||d T̂/dt| the tangential and normal
components of acceleration, respectively. If aT and aN denote these components, we can write
z that
C
N a = aT T̂ + aN N̂, (11.112a)
a
aN
T aT where
* *
x y d * d T̂ *
aT = a · T̂ = |v|, aN = a · N̂ = |v|** **. (11.112b)
dt dt
Note that the tangential component of acceleration is the time rate of change of speed. Since
acceleration is the rate of change of velocity, the normal component of acceleration must deter-
mine the rate of change of the direction of v. What is significant here is that a is expressed in
terms of T̂ and N̂; it is not necessary to use the binormal B̂. The acceleration vector of a particle
is always in the plane of T̂ and N̂.
To calculate aN using 11.112b is often quite complicated. A far easier formula results if
we take the scalar product of a as defined by 11.112a with itself:
a · a = (aT T̂ + aN N̂) · (aT T̂ + aN N̂)
= aT2 T̂ · T̂ + 2aT aN T̂ · N̂ + aN2 N̂ · N̂

= aT2 + aN2 ,
since T̂ · N̂ = 0 and T̂ · T̂ = N̂ · N̂ = 1. Consequently,
aN2 = a · a − aT2 = |a|2 − aT2 ,
and because aN is always positive (see equation 11.112b),

$
aN = |a|2 − aT2 . (11.113)
Kepler’s Laws for Planetary Motion

Based on Tycho Brahe’s (1546–1601) astronomical measurements, Johannes Kepler (1571–
1630) postulated three laws of planetary motion. The first states that planets move in elliptic
orbits with the sun at one focus of the ellipse. This can be proved with Newton’s second law
and Newton’s universal law of gravitation. If m is the mass of a planet, Newton’s second law
requires the acceleration of the planet to satisfy F = ma, where F is the resultant of all forces
acting on the planet. If we assume that the only force acting on the planet is the force of attraction
of the sun, with mass M , then F = −(GmM/r 2 )r̂, where r̂ is the unit vector in the direction
from the sun to the planet. It follows then that
GM
a = − r̂, (11.114)
r2
and the acceleration of the planet always points toward the sun. Let us choose a coordinate
system in space with origin at the sun (Figure 11.121).
FIGURE 11.121 Motion of planet around sun
M Sun
r y
a
b
m
Planet
x
Pick any two points on the orbit of the planet and let these points and M define the xy -plane.
We first show that the orbit of the planet always lies in the xy -plane. According to property
11.59c,
d dv dr
(r × v ) = r × + × v = r × a + v × v.
dt dt dt
But v × v is always the zero vector, and so is r × a (r and a are parallel, see equation 11.114).
Thus,
d
(r × v ) = 0 &⇒ r × v = C, (11.115)
dt
where C is a constant vector. But this means that the displacement vector r and the velocity
vector v are always perpendicular to C. In other words, C is in the z -direction (or negative
z -direction), and r and v are in the xy -plane; the planet moves in the xy -plane. To show that
the planet follows an elliptic path we find a formula for r which represents the polar coordinate
in the xy -plane. First we note that
" #
d dr d r̂
C = r × v = (r r̂) × (r r̂) = r r̂ × r̂ + r
dt dt dt
" #
dr d r̂
= r (r̂ × r̂) + r 2 r̂ ×
dt dt
" #
d r̂
= r 2 r̂ × .
dt
If we cross this with a = −(GM/r 2 )r̂,

- " #. - " #.
GM 2 d r̂ d r̂
a×C = − r̂ × r r̂ × = −GM r̂ × r̂ × .
r2 dt dt
We now use Exercise 46 in Section 11.4:

-" # .
d r̂ d r̂
a × C = −GM r̂ · r̂ − (r̂ · r̂) .
dt dt
Differentiation of r̂ · r̂ = 1 with respect to t gives
d r̂ d r̂ d r̂
r̂ · + · r̂ = 0 &⇒ r̂ · = 0.
dt dt dt
Thus,
d r̂ d r̂ d
a × C = GM &⇒ GM = (v × C).
dt dt dt
Integration with respect to t yields
v × C = GM r̂ + b,
where b is a constant vector. Since v × C and r̂ are both in the xy -plane, so is b. Suppose the
x -axis is chosen along b (Figure 11.121). The dot product of r with the equation above gives
r · (v × C) = GM r̂ · r + b · r = GMr + |b|r cos θ.
Thus,
r · (v × C)
r = .
GM + |b| cos θ
Since r · (v × C) = (r × v) · C = C · C = |C|2 , it follows that
|C|2 |C|2 /(GM)

r = = .
GM + |b| cos θ 1 + [|b|/(GM)] cos θ
If we set $ = |b|/(GM) , and d = |C|2 /|b| , then
$d
r = . (11.116)
1 + $ cos θ
According to equation 9.34a, this is a conic section with the origin as a focus. Since planets
are known to follow closed paths, the conic section must be an ellipse. The second and third of
Kepler’s laws are discussed in Exercises 45 and 46.
EXAMPLE 11.53
A particle is confined to move in a circular path of radius R and centre (h, k) in the xy -plane
if and only if its position vector is r = x î + y ĵ, where
x = h + R cos ω(t), y = k + R sin ω(t),
and ω(t) is some function of time t . Determine the form of ω(t) if the acceleration of the
particle is directed radially toward the centre of the circle.
SOLUTION The principal normal N̂ at any point on the circle is directed toward the centre of
the circle. Hence, the acceleration must be along N̂ and the tangential component must vanish:
d
0 = aT = |v | &⇒ |v| = C = constant.
dt
Since v = −Rω, (t) sin ω(t)î + Rω, (t) cos ω(t)ĵ, it follows that C = |v| = R|ω, (t)| . Thus,
C Ct
ω, (t) = ± &⇒ ω(t) = ± + D.
R R
In other words, ω(t) must be a linear function of t for acceleration to be directed radially toward
the centre of the circle.
The metal roof of a structure is a hemisphere with a large radius a = 10 metres (Figure
11.122a); the bottom of the roof is H = 20 metres above the ground. During the winter,
large chunks of ice that form on the roof break off, slide down the roof, and fall to the
ground. An annular empty zone is to be created on the ground in order to prevent human
injury or property damage. Our problem is to determine minimum width for the zone.
SOLUTION Once a chunk of ice breaks off, it picks up speed as it slides down the roof,
and may leave the roof before it reaches the bottom edge. In order to determine a minimum
radius for the empty zone, we shall find the farthest point at which ice can be expected to
hit the ground. Chunks that strike the ground farthest from the roof are those that attain
the greatest speed on the roof, and therefore leave the roof earliest. We shall assume that
there is no friction between roof and ice in order to maximize speed. In addition, chunks
that attain greatest speed are ones that slide from the very top of the roof.
FIGURE 11.122a Chunk of ice sliding down a frictionless sphere
Roof Chunk of ice
Empty zone
FIGURE 11.122b
(x, y)
N T
W = −9.81mj
a x
We consider, then, a mass m of ice starting from rest at the top of the roof as it slides
down the circle in Figure 11.122b. It is acted on by gravity and the reaction of the sphere.
As long as the mass is on the roof, the reaction of the roof on the mass is perpendicular
to the sphere. It seems reasonable, then, to work with tangential and normal components
of motion to the sphere (or circle). When the mass is at position (x, y) , tangential and
normal components of the weight W = −9.81mĵ are
W = 9.81m cos θ T̂ + 9.81m sin θ N̂.

If the reaction of the sphere on m is denoted by N = −N N̂, where N is therefore its
magnitude, then the total force on m is
F = 9.81m cos θ T̂ + (−N + 9.81m sin θ ) N̂.
The mass leaves the sphere when N = 0. To find where this happens, we use
Newton’s second law F = ma with F as above, and a given by equation 11.111,
# % * *& "
d * d T̂ *
* *
9.81m cos θ T̂ + (−N + 9.81m sin θ )N̂ = m |v| T̂ + m |v| * * N̂.
dt * dt *

* *
d * d T̂ *
* *
9.81 cos θ = |v|, −N + 9.81m sin θ = m|v| * * .
dt * dt *
Now, x = a cos θ and y = a sin θ , so that

+" #2 " #2 + " #2 " #2
dx dy dθ dθ dθ
|v | = + = a 2 sin2 θ + a 2 cos2 θ = −a .
dt dt dt dt dt
Hence,
" #
d dθ d 2θ
9.81 cos θ = −a = −a 2 .
dt dt dt
Multiplication by dθ/dt gives
" #2
dθ d 2 θ dθ a d dθ
−9.81 cos θ =a 2 = ,
dt dt dt 2 dt dt
and we may integrate with respect to t ,
" #2
a dθ
−9.81 sin θ = + C.
2 dt
Since dθ/dt = 0 when θ = π/2, it follows that C = −9.81, and

" #2 " #2
a dθ dθ 19.62
−9.81 sin θ = − 9.81 &⇒ = (1 − sin θ ).
2 dt dt a
To tackle the normal components, we first calculate that
d dθ
T = (a cos θ, a sin θ ) = a(− sin θ, cos θ ) &⇒ T̂ = (sin θ, − cos θ ).
dt dt
Thus, * *
d T̂ dθ * d T̂ * dθ
* *
= (cos θ, sin θ ) &⇒ * *=− .
dt dt * dt * dt
Substitution into −N + 9.81m sin θ = m|v||d T̂/dt| gives

" #" # " #2
dθ dθ dθ
−N + 9.81m sin θ = m −a − = ma .
dt dt dt
Replacing (dθ/dt)2 by (19.62/a)(1 − sin θ ) , we obtain
−N + 9.81m sin θ = 19.62m(1 − sin θ ) &⇒ N = 9.81m(3 sin θ − 2).
Clearly, N = 0 when θ = Sin−1 (2/3) , and this is the angle at which the mass leaves
the sphere. Its speed at this point is
+ " # 0
dθ 19.62 2 19.62a
|v| = −a =a 1− = .
dt a 3 3
Once the mass leaves the roof, the only force acting on it is gravity, and therefore its
acceleration is
d 2r dr
2
= −g ĵ &⇒ = −gt ĵ + C.
dt dt
If
0 we choose time t = 0 when the mass leaves the roof, then its velocity at this time is
19.62a
(sin θ î − cos θ ĵ) , where angle θ is defined above. If we denote this by v0 , then
3
C = v0 , and integration of
dr 1
= −gt ĵ + v0 gives r = − gt 2 ĵ + v0 t + D.
dt 2
Since the initial position of the mass when it leaves the roof is r0 = a cos θ î + a sin θ ĵ,
it follows that D = r0 , and the position of the mass after it leaves the roof is
1
r(t) = − gt 2 ĵ + v0 t + r0 .
2
The mass hits the ground when the y -component of r(t) is equal to −H ,
0
1 2 19.62a
−H = − gt − cos θ t + a sin θ.
2 3
√
When we substitute a = 10, H = 20, sin θ = 2/3, and cos θ = 5/3, we obtain the
quadratic equation
0
2 109 80
4.905t + = 0,
t−
3 3
the positive solution of which is t = 1.797. The x -coordinate of the mass at this time is
0 " # %√ &
196.2 2 5
(1.797) + 10 = 17.14 m.
3 3 3
This is the minimum radius of the outer edge of the empty zone. In other words, it must
be 7.14 metres wide.
EXERCISES 11.13
In Exercises 1–5 find the velocity, speed, and acceleration of a particle ∗ 16. A particle travels counterclockwise around the circle (x − h)2 +
if the given equations represent its position as a function of time. (y − k)2 = R 2 in the figure below. Show that the speed of the particle
√ √ at any time is |v| = ωR , where ω = dθ/dt is called the angular
1. x(t) = t 2 + 1, y(t) = t t 2 + 1, t ≥ 0 speed of the particle.
2. x(t) = t + 1/t, y(t) = t − 1/t, t ≥ 1 y

(x, y)
3. x(t) = sin t, y(t) = 3 cos t, z(t) = sin t, 0 ≤ t ≤ 10π
(h, k)
4. x(t) = t 2 + 1, y(t) = 2tet , z(t) = 1/t 2 , 1 ≤ t ≤ 5 R
2
5. x(t) = e−t , y(t) = t ln t, z(t) = 5, t ≥ 1
x
In Exercises 6–7 a particle at (1, 2, −1) starts from rest at time t = 0. ∗ 17. A particle travels around the circle x 2 + y 2 = 4 counterclockwise
Find its position as a function of time if the given function defines its at constant speed, making 2 revolutions each second. If x and y are
acceleration. measured in√ metres, what is the velocity of the particle when it is at the
point (1, − 3) ?
6. a(t) = 3t 2 î + (t + 1)ĵ − 4t 3 k̂, t ≥ 0
∗ 18. (a) Show that if an object moves with constant speed in a cir-
cular path of radius R , the magnitude of its acceleration is
7. a(t) = 3î + ĵ/(t + 1) , t ≥ 0
3
|a| = |v|2 /R .
(b) If a satellite moves with constant speed in a circular orbit
In Exercises 8–9 find the tangential and normal components of acceler- 200 km above the earth’s surface, what is its speed? Hint:
ation for a particle moving with position defined by the given functions Use Newton’s universal law of gravitation (see Exercise
(where t is time). 32 in Section 11.3) to determine the acceleration a of the
satellite. Assume that the earth is a sphere with radius
8. x(t) = t, y(t) = t 2 + 1, t ≥ 0 6370 km and density 5.52 × 10 3 kg/m 3 .
9. x(t) = cos t, y(t) = sin t, z = t, t ≥ 0 ∗ 19. Two particles move along curves C1 and C2 in the figure below.
If at some instant of time the particles are at positions P1 and P2 , then
the vector P1 P2 is the displacement of P2 with respect to P1 . Clearly,
10. Show that the normal component of acceleration of a particle can OP1 + P1 P2 = OP2 . Show that when this equation is differentiated
be expressed in the form aN = |v|2 /ρ = κ|v|2 . with respect to time, we have
11. Find the kinetic energy for each particle in Exercises 1–5 if its
vP1 /O + vP2 /P1 = vP2 /O ,
mass is 2 g. Assume that x , y , and z are measured in metres and t in
seconds. where vP1 /O and vP2 /O are velocities of P1 and P2 with respect to the
12. A particle starts at the origin and moves along the curve 4y = x 2 origin, and vP2 /P1 is the velocity of P2 with respect to P1 . Can this
to the point (4, 4) . equation be rewritten in the form
(a) If the y -component of its acceleration is always equal to 2 vP1 /O + vO/P2 = vP1 /P2 ?
and the y -component of its velocity is initially zero, find
the x -component of its acceleration.
z
(b) If the x -component of its acceleration is equal to 24t 2 ( t
being time) and the x -component of its velocity is initially P1
zero, find the y -component of its acceleration.
OP1 C1
13. A particle moves along the curve x(t) = t, y(t) = t 3 − 3t 2 + P1P2
2t, 0 ≤ t ≤ 5 in the xy -plane (where t is time). Is there any point at O C2
which its velocity is parallel to its displacement?
OP2 P2
14. A particle moves along the curve y = x 3 − 2x + 3 so that its x y
x -component of velocity is always equal to 5. Find its acceleration.
∗ 20. A plane flies on a course N30 ◦ E with airspeed 650 km/h (i.e.,
15. If a particle starts at time t = 0 from rest at position (3, 4) and the speed of the plane relative to the air is 650). If the air is moving
experiences an acceleration a = −5t 4 î − (2t 3 + 1)ĵ , find its speed at at 40 km/h due east, find the ground velocity and speed of the plane.
t = 2. Hint: Use Exercise 19.
∗ 21. A plane flies with speed 600 km/h in still air. The plane is to fly in ∗ 30. The English longbow in medieval times was regarded to be accurate
a straight line from city A to city B , where B is 1000 km northwest at 100 m or more. For an arrow to travel a horizontal distance of 100 m
of A . What should be its bearing if the wind is blowing from the west with maximum height 10 m, find the initial speed and angle of projection
at 50 km/h? How long will the trip take? of the arrow. Ignore air friction.
∗ 22. A straight river is 200 m wide and the water flows at 3 km/h. If
∗ 31. The block of mass M in the figure below slides on a thin film of
you can paddle your canoe at 4 km/h in still water, in what direction
oil. The film thickness is h and the area of the block in contact with the
should you paddle if you wish the canoe to go straight across the river?
film is A . When released, mass m exerts tension in the cord, causing
How long will it take to cross?
block M to accelerate. When the speed of M is v , the viscous force
∗ 23. (a) In the figure below a cannon is fired up an inclined plane. acting on it due to the film is F = µAv/ h where µ is the viscosity of
If the speed at which the ball is ejected from the cannon is the oil. Find the speed of M as a function of time t . Neglect friction
S , show that the range R of the ball is given by in the pulley and air resistance.
2S 2 cos θ sin (θ − α)
R = , M
g cos2 α
Oil film
where g is the acceleration due to gravity. h
m
(b) What angle θ maximizes R ?
∗ 32. Water issues from the nozzle of a fire hose at speed S in the figure
R below. Show that the maximum height attainable by the water on the
building is given by (S 4 − g 2 d 2 )/(2gS 2 ) , where g is the acceleration
due to gravity.
Building
∗ 24. What constant acceleration must a particle experience if it is to

travel from (1, 2, 3) to (4, 5, 7) along the straight line joining the
points, starting from rest and covering the distance in 2 units of time?
∗ 25. Calculate the normal component aN of the acceleration of a particle
using equations 11.112 and 11.113 if its position is given by x = t 2 + 1,
y = 2t 2 − 1, z = t 2 + 5t , t ≥ 0 ( t being time). d
√
∗ 26. A particle moves along the curve x(t) = 2 + 1 − t 2 , y(t) = t ,
0 ≤ t ≤ 1/2, where t is time. Is there a time at which its acceleration ∗ 33. A boy stands on a cliff 50 m high that overlooks a river 85 m wide
is perpendicular to its velocity? (figure below). If he can throw a stone at 25 m/s, can he throw it across
∗ 27. (a) Show that motion along a straight line is the result in both the river?
of the following situations:
(i) The initial velocity is zero, and the acceleration
is constant.
(ii) The initial velocity is nonzero, and the acceler-
ation is constant and parallel to the initial ve-
locity. 50
(b) Can we generalize the results of part (a) and state that con-
stant acceleration produces straight-line motion? Illustrate.
∗ 28. A particle starts from position r0 = (x0 , y0 , z0 ) at time t = t0
with velocity v0 . If it experiences constant acceleration a , show that
85
1 2
r = r0 + v0 (t − t0 ) + a(t − t0 ) . ∗ 34. A golfer can drive a maximum of 300 m in the air on a level fairway.
2
From the tee in the figure below, can he expect to clear the stream?
∗ 29. A ladder 8 m long has its upper end against a vertical wall and its
lower end on a horizontal floor. Suppose that the lower end slips away Tee
from the wall at constant speed 1 m/s. 310 m
Green
(a) Find the velocity and acceleration of the middle point of
the ladder when the foot of the ladder is 3 m from the wall. 20 m
(b) How fast does the middle point of the ladder strike the floor? Stream
∗ 35. (a) A projectile is fired at angle θ to the horizontal from a height ∗ 39. If the stone in Exercise 38 is embedded in the side of the tire, its
h above the ground with speed v (figure below). Show that path is called a trochoid (see Exercise 58 in Section 9.1).
the range R of the projectile is given by the formula
% 0 & (a) Find velocity, speed, and acceleration of the stone if the tire
v 2 cos θ 2
2gh rolls so that its centre has constant speed S . (Assume that
R = sin θ + sin θ + ,
g v2 x = 0 at time t = 0.)
where g = 9.81.
(b) What are normal and tangential components of the stone’s
acceleration?
y v
∗ 40. Circles C1 and C2 in the figure below represent cross-sections of
two cylinders. The left cylinder remains stationary while the right one
h Ground level rolls (without slipping) around the left one, and the cylinders always
x
R remain in contact. If the right cylinder picks up a speck of dirt at
(b) What angle maximizes R for given v and h ? point (R, 0) , the path that the dirt traces out during one revolution is a
cardioid.
(c) Suppose the projectile is a shot, thrown by an Olympic
athlete. What is the angle in part (b) if v = 13.7 m/s and
(a) Show that parametric equations for the cardioid are
h = 2.25 m?
(d) Prove that the maximum height attained by the projectile is
x = R(2 cos θ − cos 2θ), y = R(2 sin θ − sin 2θ).
R 2 tan2 θ
h+ .
4(h + R tan θ)
(b) Verify that if the point of contact moves at constant speed
∗ 36. A cannon is located on a plane inclined at angle α to the horizontal S , with t = 0 when the speck of dirt is picked up, then
(figure below). If a projectile is fired from the cannon at angle β to the θ = St/R .
plane, prove that for the projectile to hit the plane horizontally,
" # (c) Find velocity, speed, and acceleration of the speck of dirt.
sin 2α
β = Tan −1 .
3 − cos 2α (d) What are normal and tangential components of the dirt’s
y acceleration?
x
(x, y)
∗ 37. If r is the position vector of a particle with mass m moving under
C1 C2
the action of a force F , the torque of F about the origin is τ = r × F .
The angular momentum of m about O is defined as H = r ×mv . Use
−R 3R x
Newton’s second law in the form 11.108 to show that τ = d H/dt .
∗ 38. When a stone is embedded in the tread of a tire and the tire rolls
(without slipping) along the x -axis (figure below), the path that it traces
is called a cycloid (see Example 9.7 in Section 9.1).
(a) Verify that if the centre of the tire moves at constant speed S , ∗ 41. Show that the path of a particle lies on a sphere if its displacement
with t = 0 when the stone is at the origin, then θ = St/R . and velocity are always perpendicular during its motion.
(b) Find velocity, speed, and acceleration of the stone at any
time. ∗ 42. Suppose in Exercise 22 that because of an injured elbow, you can
(c) What are the normal and tangential components of the paddle only at 2 km/h. What should be your heading to travel straight
stone’s acceleration? to a point L kilometres downstream on the opposite shore? Are there
any restrictions on L ?
y
∗ 43. Suppose that position vectors of a system of n masses mi are
2R denoted>by ri and forces acting on these masses are Fi . Show that
(x, y) n
if F = i=1 Fi , then the acceleration a of the centre of mass
>n of the
system (see Section 7.7) is given by F = M a , where M = i=1 mi .
x ∗ 44. If the force acting on a particle is always tangent to the particle’s

trajectory, what can you conclude about the trajectory?
∗ 45. Kepler’s second law states that the line joining the sun to a planet (c) Use parts (a) and (b) to obtain
sweeps out equal areas in equal time intervals. To show this, let A(t) be
the area swept out by the line beginning at some time t0 in the following dA |C|
figure and ending at time t . = = constant.
dt 2
(a) Verify that
2 Does this verify Kepler’s second law?
θ (t)
1 2
A(t) = r dθ,
θ0 2 ∗ 46. Kepler’s third law states that the square of the period of revolution
of a planet is proportional to the cube of the length of the major axis of
and use equation 6.19 to prove that
its orbit.
dA 1 dθ (a) Use part (c) of Exercise 45 to show that if P is the time
= r2 .
dt 2 dt taken for one complete revolution, then
y 2π ab
Position of planet P =
Position ,
at time t0 |C|
of planet
A (t)
at time t r(t0)
where 2a and 2b are lengths of the major and minor axes
of the elliptic orbit.
r(t)
Sun (b) Use equation 11.116 to show that b2 /a = |C|2 /(GM) ,
x
and therefore
4π 2 a 3
P2 = .
GM
(b) Show that r̂ = cos θ î + sin θ ĵ and d r̂/dt are perpendic-

∗∗ 47. Suppose that in Project 18 the mass is√given an initial speed v0 at
ular, and deduce from equation 11.115 that
the top of the sphere. Prove that if v0 ≤ ag , the angle θ at which
" #
dθ |C| −1 2 v02
= 2 . the mass leaves the sphere is Sin + .
dt r 3 3ag
SUMMARY
We have now established the groundwork for multivariable calculus. We discussed curves
and surfaces in space and introduced vectors. We described points by Cartesian coordinates
(x, y, z) and then illustrated that an equation F (x, y, z) = 0 in these coordinates usually
defines a surface. When a second equation G(x, y, z) = 0 also defines a surface, the pair of
simultaneous equations F (x, y, z) = 0, G(x, y, z) = 0 describes the curve of intersection of
the two surfaces (provided the surfaces do intersect). It is often more useful to have parametric
equations for a curve, and these can be obtained by specifying one of x , y , or z as a function
of a parameter t and solving the given equations for the other two in terms of t : x = x(t) ,
y = y(t) , z = z(t) .
The most common surfaces that we encountered were planes and quadric surfaces. Every
plane has an equation of the form Ax + By + Cz + D = 0; conversely, every such equation
describes a plane (provided that A , B , and C are not all zero). A plane is uniquely defined by
a vector that is perpendicular to it [and ( A , B , C ) is one such vector] and a point on it. Quadric
surfaces are surfaces whose equations are quadratic in x , y , and z , the most important of which
were sketched in Figures 11.22–11.30.
Every straight line in space is characterized by a vector along it and a point on it. (Contrast
this with the characterization of a plane described above.) If ( a , b , c ) are the components of a
vector along a line and ( x0 , y0 , z0 ) are the coordinates of a point on it, then vector, symmetric,
and parametric equations for the line are, respectively,
(x, y, z) = (x0 , y0 , z0 ) + t (a, b, c);

Summary 795
x − x0 y − y0 z − z0
= = ;
a b c
x = x0 + at,
y = y0 + bt,
z = z0 + ct.
Geometrically, vectors are defined as directed line segments; algebraically, they are represented
by ordered sets of real numbers ( vx , vy , vz ), called their Cartesian components. Vectors can
be added or subtracted geometrically using triangles or parallelograms; algebraically, they are
added and subtracted component by component. Vectors can also be multiplied by scalars to
give parallel vectors of different lengths.
We defined two products of vectors: the scalar product and the vector product. The scalar
product of two vectors u = (ux , uy , uz ) and v = (vx , vy , vz ) is defined as
u · v = ux vx + uy vy + uz vz = |u||v| cos θ,
$
where |u| = u2x + uy2 + u2z is the length of u and θ is the angle between u and v. If the
components of u and v are known, this equation can be used to find the angle θ between the
vectors. The scalar product has many uses: finding components of vectors in arbitrary directions,
calculating distances between geometric objects, and finding mechanical work, among others.
The vector product of two vectors u and v is
* *
* î ĵ k̂ *
* *
u × v = * ux uy uz * = |u||v| sin θ ŵ,
*v vy v *
x z
where ŵ is the unit vector perpendicular to u and v determined by the right-hand rule. Because of
the perpendicularity property, the vector product is indispensable in finding vectors perpendicular
to other vectors. We used this fact when finding a vector along the line of intersection of two
planes, a vector perpendicular to the plane containing three given points, and distances between
geometric objects. It can also be used to find areas of triangles and parallelograms.
If a curve is represented vectorially in the form r(t) = x(t)î + y(t)ĵ + z(t)k̂, then a unit
vector tangent to the curve at any point is
d r/dt
T̂ = .
|d r/dt|
Two unit vectors normal to the curve are the principal normal N̂ and the binormal B̂:
d T̂/dt
N̂ = ; B̂ = T̂ × N̂.
|d T̂/dt|
These three vectors form a moving triad of mutually perpendicular unit vectors along the curve.
The curvature of a curve, defined by κ(t) = |ṙ × r̈|/|ṙ|3 , measures the rate at which the
curve changes direction: The larger κ is, the faster the curve turns. The reciprocal of curvature
ρ = κ −1 is called radius of curvature. It is the radius of that circle which best approximates
the curve at any point.
If parametric equations for a curve represent the position of a particle and t is time, then
the velocity and acceleration of the particle are, respectively,
dr dv d 2r
v = ; a = = 2;
dt dt dt
and its speed is the magnitude of velocity, |v| .
Tangential and normal components of velocity and acceleration of the particle are defined
by
v = |v|T̂; a = aT T̂ + aN N̂;
where
* *
* d T̂ * $
d * *
aT = |v | and aN = |v|* * = |a|2 − aT2 .
dt * dt *
What these results say is that velocity is always tangent to the trajectory of the particle, and
its acceleration always lies in the plane of the velocity vector and the principal normal.
KEY TERMS
Coordinate planes Origin
Cartesian or rectangular coordinates Right-handed coordinate system
Left-handed coordinate system Octants
Cylinder Quadric surface
Scalars Vectors
Tip of a vector Tail of a vector
x -, y -, and z -components of v Cartesian components of v
Unit vector Triangular addition of vectors
Parallelogram addition of vectors Zero vector
Resultant force Scalar components
Vector components Scalar, dot, or inner product
Vector, cross, or outer product Equation for a plane
Normal or normal vector to the plane Vector equation of a line
Parametric equations for a line Symmetric equations for a line
Component Work
First moments of forces Span of a cable
Sag of a cable Catenary
Position or displacement vector Continuous vector-valued function
Derivative of a vector-valued function Antiderivative of a vector-
Indefinite integral of a vector-valued function valued function
Parametric representation of curves Continuous curve
Closed curve Vector representation of curve
Tangent vectors to curves Smooth curve
Piecewise-smooth curve Unit tangent vector to a curve
Principal normal vector Binormal vector
Curvature of a curve Radius of curvature of a curve
Circle of curvature of a curve Frenet–Serret formulas
Torsion of a curve Velocity
Speed Acceleration
Tangential components of velocity Normal components of velocity
and acceleration and acceleration
Kepler’s laws
REVIEW
EXERCISES
In Exercises 1–10 find the value of the scalar or the components of 33. Containing the line x − y + z = 3, 3x + 4y = 6 and the point
the vector if u = (1, 3, −2) , v = (2, 4, −1) , w = (0, 2, 1) , and (2, 2, 2)
r = (2, 0, −1) .
34. Containing the lines x = 3t, y = 1 + 2t, z = 4 − t , and
1. 2u − 3w + r 2. u · (v × w) x=y=z
3. (3u × 4v) − w 4. 3u × (4v − w)
5. |u|v − |v|r 6. (u + v) · (r − w) In Exercises 35–39 find the distance.
7. (u + v) × (r − w) 8. (u × v) × (r × w)
35. Between the points (1, 3, −2) and (6, 4, 1)
2r
9. (u · v)r − 3(v · w)u 10. + 3(v + u) 36. From the point (6, 2, 1) to the plane 6x + 2y − z = 4
v·w
37. From the line x − y + z = 2, 2x + y + z = 4 to the plane
x−y =5
In Exercises 11–26 draw whatever the equation, or equations, describe
in space. 38. From the line x − y + z = 2, 2x + y + z = 4 to the plane
3x + 6y = 4
11. x − y + 2z = 6 12. x 2 + z2 = 1
! ∗ 39. From the point (6, 2, 3) to the line x−y+z = 6, 2x+y+4z = 1
13. x = y 2 + z2 14. x − y = 5, 2x + y = 6
40. Find the area of the triangle with vertices (1, 1, 1) , (−2, 1, 0) ,
15. x 2 + y 2 + z2 = 6z + 10 and (6, 3, −2) .
16. x 2 + y 2 + z2 = 6z − 10 41. If the points in Exercise 40 are three vertices of a parallelogram,
what are possibilities for the fourth vertex? What are areas of these
17. x + y = 5, 2x − 3y + 6z = 1, y = z
parallelograms?
18. x = t 2 , y = t, z = t 3
19. x = t, y = t 3 + 1
In Exercises 42–43 find the unit tangent vector T̂ , the principal normal
x−1 y+5
20. = =z vector N̂ , and the binormal vector B̂ for the curve.
3 2
21. z = 4 − x − 2y 2
2 42. x = 2 sin t, y = 2 cos t, z = t
22. y 2 + z2 = 1, y = z 43. x = t 3 , y = 2t 2 , z = t + 4
23. y 2 + z2 = 1, x = z 44. If a particle has a trajectory defined by x = t, y = t 2 , z =
t 2 , where t is time, find its velocity, speed, and acceleration at any
24. x 2 + y 2 = z2 + 1
time. What are normal and tangential components of its velocity and
25. x = y 2 , x = z2 acceleration?
26. z2 = x 2 − y 2 ∗ 45. A force F = x −2 (2î + 3ĵ) acts on a particle moving from x = 1
to x = 4 along the x -axis. How much work does it do?
In Exercises 27–30 find equations for the line. ∗ 46. A ball rolls off a table 1 m high with speed 0.5 m/s (figure below).
27. Through the points (−2, 3, 0) and (1, −2, 4) y
28. Through (6, 6, 2) and perpendicular to the plane 5x − 2y +z = 4 x
29. Parallel to the line x − y = 5, 2x + 3y + 6z = 4 and through
the origin
1m
30. Perpendicular to the line x = t + 2, y = 3 − 2t, z = 4 + t ,
intersecting this line, and through the point (1, 3, 2)
(a) With what speed does it strike the floor?

In Exercises 31–34 find the equation for the plane.
(b) What is its displacement vector relative to the point where
31. Through the points (1, 3, 2) , (2, −1, 0) , and (6, 1, 3) it left the table when it strikes the floor?
32. Through the point (1, 2, −1) and perpendicular to the line y = (c) If it rebounds in the direction shown but loses 20% of its
z, x + y = 4 speed in the bounce, find the position of its second bounce.
∗ 47. In the figure below, a spring (with constant k ) is fixed at A and ∗ 48. Find Cartesian components of the spring force F on the sleeve in
attached to a sleeve at C . The sleeve is free to slide without friction on Example 11.26.
a vertical rod, and when the spring is horizontal (at B ), it is unstretched.
If the sleeve is slowly lowered, there is a position at which the vertical ∗ 49. If a toy train travels around the oval track in the figure below with
component of the spring force on C is balanced by the force of gravity constant speed, show that its acceleration at A (the point at which the
on the sleeve (ignoring the weight of the spring itself). If the mass of circular end meets the straight section) is discontinuous.
the sleeve is m , find an equation determining s in terms of d , m , k ,
and g = 9.81, the acceleration due to gravity.
A
d
R
A B
Semicircle
s
C
CHAPTER
12 Differential Calculus of
Multivariable Functions
Application Preview The figure on the left below shows gas confined in a cylinder, closed on one end, with a piston
on the other. If the piston moves to the left, it compresses the gas in the cylinder. The volume
V that the gas occupies decreases and the pressure P that it exerts on the piston increases.
Conversely, when the piston moves to the right, the volume of the gas increases and the pressure
on the piston decreases. The table below gives the pressure for various volumes of gas. They
have been plotted in the figure on the right.
V 54.3 61.82 72.4 88.7 118.6 194.0

P 61.2 49.5 37.6 28.4 19.2 10.1
P
60
Cylinder
40
Piston
Gas
20
50 100 150 200 V
THE PROBLEM What function represents the data in the table; or, equivalently, were the
data points to be joined by a smooth curve, what would be its equation? (See the discussion on
pages 882 and 883 for the solution.)
Few quantities in real life depend on only one variable; most depend on a multitude of inter-
related variables. In order to understand such complicated relationships, we initiate discussions
in this chapter with derivatives of functions of more than one variable. Much of the theory and
many of our examples involve functions of two or three variables, because in these cases we
can give geometric as well as analytic explanations. If the situation is completely analogous for
functions of more variables, then it is likely that no mention of this fact will be made; on the
other hand, if the situation is different for a higher number of variables, we will be careful to
point out these differences.
12.1 Multivariable Functions

If a variable T depends on other variables x , y , z , and t , we write T = f (x, y, z, t) and
speak of T as a function of x , y , z , and t . For example, T might be temperature, x , y , and z
might be the coordinates of points in some region of space, and t might be time. The stopping
distance D of a car depends on many factors: the initial speed s , the reaction time t of the driver
to move from the accelerator to the brake, the texture T of the road, the moisture level M on
the road, and so on. We write D = f (s, t, T , M, . . .) to represent this functional dependence.
The function P = f (I, R) = I 2 R represents the power necessary to maintain a current I
through a wire with resistance R .
799
800 Chapter 12 Differential Calculus of Multivariable Functions
More precisely, a variable z is said to be a function of two independent variables x and y

FIGURE 12.1 A function if x and y are not related and each pair of values of x and y determines a unique value of z .
f (x, y) of two independent vari- We write z = f (x, y) to indicate that z is a function of x and y . Each possible pair of values
ables can be represented geomet- x and y of the independent variables can be represented geometrically as a point (x, y) in the
rically as a surface with equation xy -plane. The totality of all points for which f (x, y) is defined forms a region in the xy -plane
z = f (x, y) called the domain of the function. Figure 12.1, for example, illustrates a rectangular domain.
z If for each point (x, y) in the domain we plot a point f (x, y) units above the xy -plane, we
z = f (x, y) obtain a surface, such as the one in Figure 12.1. Each point on this surface has coordinates
(x, y, z) that satisfy the equation
z = f (x, y), (12.1)
(x, y, z)
and therefore 12.1 is the equation of the surface. This surface is a pictorial representation of the
function.
It is clear that functions of more than two independent variables cannot be represented
pictorially as surfaces. For example, if u = f (x, y, z) is a function of three independent
(x, y) y
x variables, values (x, y, z) of these independent variables can be represented geometrically as
points in space. To graph u = f (x, y, z) as above would require a u -axis perpendicular to
Domain of f (x, y)
the x -, y -, and z -axes, a somewhat difficult task geometrically. We can certainly think of u =
f (x, y, z) as defining a surface in four-dimensional xyzu -space, but visually we are stymied.
Although every function f (x, y) of two independent variables can be represented geo-
metrically as a surface, not every surface represents a function f (x, y) . A given surface does
represent a function f (x, y) if and only if every vertical line (in the z -direction) that intersects
the surface does so in exactly one point. For example, a sphere such as x 2 + y 2 + z2 = 1
does not determine z as a function of x and y . Most, but not all, vertical lines that intersect the
2 2 2
sphere do so in two points. It is also clear algebraically that
! x + y + z = 1 does not define
2 2
z as a function of x and y . Solving for z gives z = ± 1 − x − y , two solutions for each
x and y satisfying x 2 + y 2 < 1.
EXAMPLE 12.1
Draw the surface defined by the function f (x, y) = x 2 + 4y 2 .
SOLUTION To draw the surface z = x 2 + 4y 2 , we note that if the surface is intersected with
a plane z = k > 0, then the ellipse x 2 + 4y 2 = k , z = k is obtained. As k increases, the
ellipse becomes larger. In other words, cross-sections of this surface are ellipses that expand
with increasing z . If we intersect the surface with the yz -plane ( x = 0), we obtain the
parabola z = 4y 2 , x = 0. Similarly, intersection of the surface with the xz -plane gives the
parabola z = x 2 , y = 0. These facts lead to Figure 12.2a. A computer plot of the surface for
−2 ≤ x ≤ 2 and −2 ≤ y ≤ 2 is shown in Figure 12.2b.
FIGURE 12.2a Cross-sections of surface with xz -plane, FIGURE 12.2b Computer plot of
yz -plane, and planes z = k lead to surface z = x 2 + 4y 2 surface z = x 2 + 4y 2
z x 2 + 4y 2 = k, z
z=k
z = 4y 2,
x=0
z = x 2, z = x 2 + 4y 2
y=0
y
y
x x
12.1 Multivariable Functions 801
EXAMPLE 12.2
The ends of a taut string are fixed at x = 0 and x = 2 on the x -axis. At time t = 0, the string
is given a displacement in the y -direction of y = sin(π x/2) (Figure 12.3a). If the string is
then released, its displacement thereafter is given by
y = f (x, t) = sin(π x/2) cos(8π t).
Physically, this function need only be considered for t ≥ 0 and 0 ≤ x ≤ 2, and is plotted
for 0 ≤ t ≤ 1/2 in Figure 12.3b. Interpret physically the intersections of this surface with
planes t = t0 ( t0 = a constant) and x = x0 ( x0 = a constant).
FIGURE 12.3a Initial displacement of a string FIGURE 12.3b Computer plot representing
that is displaced then released displacement of points in string
y y
1
0.8
0.6
0.4 t
0.2
x
0.5 1 1.5 2 x
SOLUTION Grid lines on the surface are curves of intersection of the surface with vertical
planes x = x0 and t = t0 . The curve of intersection of the surface with a plane t = t0
represents the position of the string at time t0 . For t = t0 , the equation of the curve is
y(x, t0 ) = cos(8π t0 ) sin(π x/2) . Thus, the string vibrates up and down always in the shape
of a sine curve, its amplitude at time t0 being | cos(8π t0 )| .
The curve of intersection of the surface with a plane x = x0 , 0 < x0 < 2, is a graphical
history of the vertical displacement of the particle in the string at position x0 . The equation of
the curve is y(x0 , t) = sin(π x0 /2) cos(8π t) . The particle at x0 undergoes simple harmonic
motion, the amplitude being |sin(π x0 /2)| .
Another way to visualize a function f (x, y) of two independent variables is through level
curves. Curves f (x, y) = C are drawn in the xy -plane for various values of C . Effectively,
the surface z = f (x, y) is sliced with a plane z = C , and the curve of intersection is projected
into the xy -plane. Each curve joins all points for which f (x, y) has the same value; or it joins
all points that have the same height on the surface z = f (x, y) . A few level curves for the
surface in Figure 12.2 are shown in Figure 12.4; they are ellipses. Level curves for the function
in Figure 12.5a are shown in Figure 12.5b.
FIGURE 12.4 Level curves of the function f (x, y) = x 2 + 4y 2
y
3 C = 36 x 2 + 4y2 = C
C = 16
C=1 C=4
−6 1 2 4 6 x
−3
FIGURE 12.5a Computer plot of sur- FIGURE 12.5b Level curves of

face z = x 2 − y 2 defined by f (x, y) = x 2 − y 2 f (x, y) = x 2 − y 2
z y
x 2 − y2 = C
z = x 2 − y2
−1
−1
1 4 1
1 0 0
4 4 1
x
−1
4
y −1
This technique is used on topographical maps to indicate land elevation, on marine charts to
indicate water depth, and on climatic maps to indicate curves of constant temperature (isotherms)
and curves of constant barometric pressure (isobars).
EXERCISES 12.1
1. If f (x, y) = x 3 y + x sin y , evaluate (a) f (1, 2) , (b) f (−2, −2) , ∗ 24. f (x, y) = ln (x 2 + y 2 )
(c) f (x 2 + y, x − y 2 ) , and (d) f (x + h, y) − f (x, y) .
∗ 25. f (x, y) = x 2 − y 2
2. If f (x, y, z) = x 2 y 2 − x 4 + 4zx 2 , show that f (a + b, a −
b, ab) = 0.
∗ 26. A closed box is to have total surface area 30 m 2 . Find a formula
for the volume of the box in terms of its length l and width w .
In Exercises 3–6 find and illustrate geometrically the largest possible ∗ 27. (a) A company wishes to construct a storage tank in the form
domain for the function. of a rectangular box. If material for sides and top costs
! $1.25/m 2 and material for the bottom costs $4.75/m 2 , find
2 2
3. f (x, y) = 4− x2 − y2 4. f (x, y) = ln (1 − x + y )
the cost of building the tank as a function of its length l ,
5. f (x, y) = Sin −1 (x 2 y + 1) 6. f (x, y, z) = 1/(x 2 + y 2 + z2 ) width w , and height h .
12xy − x 2 y 2 (b) If the tank must hold 1000 m 3 , find the construction cost in
7. For what values of x and y is the function f (x, y) =
2(x + y) terms of l and w .
equal to zero? Illustrate these values as points in the xy -plane. What
(c) Repeat parts (a) and (b) if the 12 edges of the tank must be
is the largest domain of the function?
welded at a cost of $7.50/m of weld.
∗ 28. A rectangular box is inscribed inside the ellipsoid x 2 /a 2 +y 2 /b2 +
In Exercises 8–21 draw the surface defined by the function. Plot the z2 /c2 = 1 with sides parallel to the coordinate planes and corners on
surface as a check. the ellipsoid. Find a formula for the volume of the box in terms of x
and y .
8. f (x, y) = y 2 9. f (x, y) = 4 − x − 2y
! ∗ 29. (a) A silo is to be built in the shape of a right-circular cylinder
2 2
10. f (x, y) = x + y 11. f (x, y) = x2 + y2
surmounted by a right-circular cone. If the radius of each
12. f (x, y) = y + x 13. f (x, y) = 1 − x 3 is 6 m, find a formula for the volume V of the silo as a
14. f (x, y) = 2(x 2 + y 2 ) 15. f (x, y) = 1 − x 2 − 4y 2 function of the heights H and h of the cylinder and cone.
16. f (x, y) = xy ∗ 17. f (x, y) = y − x 2 (b) If the total surface area of the silo must be 200 m 2 (not
−x 2 −y 2 including the base), find V as a function
√ of h . (The area
∗ 18. f (x, y) = e ∗ 19. f (x, y) = |x − y|
! of the curved surface of a cone is π r r 2 + h2 .)
2 2
∗ 20. f (x, y) = x − y ∗ 21. f (x, y) = 1 + x 2 − y 2 ∗ 30. The figure below shows parameters taken into account in the anal-
ysis of the distance travelled by a long jumper:
In Exercises 22–25 draw level curves f (x, y) = C corresponding to θ = angle to the horizontal at which jumper leaves the ground;
the values C = −2, −1, 0, 1, 2. T = horizontal distance from toe to centre of mass G at takeoff;
! L = horizontal distance from heel to centre of mass on landing;
∗ 22. f (x, y) = 4 − 4x 2 + y 2 h = vertical distance between centre of mass on takeoff and landing;
∗ 23. f (x, y) = y − x 2 R = horizontal distance between centre of mass on takeoff and landing.
Using the formula developed in Exercise 35 of Section 11.13, the the faces at x = 0 and x = π are both kept at temperature
total length of the jump is 0◦ C for time t > 0. If the initial temperature (at time
t = 0) of the rod is given by 100 sin x , 0 ≤ x ≤ π , then
D(θ, v) = T + R + L = T + L the temperature thereafter is
" # $
v 2 cos θ 2
2gh T = f (x, t) = 100e−kt sin x (k > 0 constant).
+ sin θ + sin θ + 2 ,
g v
(b) Draw the surface T = f (x, t) .
where g = 9.81. Typical values for T , L , and h are 0.35 m, 0.9 m,
and 0.5 m, respectively. (c) Interpret physically the curves of intersection of this surface
(a) Calculate D(0.35, 9.0) . with planes x = x0 and t = t0 .
(b) What is the percentage change in D , from that in part (a),
∗ 33. A cow’s daily diet consists of three foods: hay, grain, and sup-
if takeoff speed is increased by 10%?
plements. The animal is always given 11 kg of hay per day, 50% of
(c) What is the percentage change in D , from that in part (a), which is digestive material and 12% of which is protein. Grain is 74%
if takeoff angle is increased by 10%? digestive and 8.8% protein. Supplements are 62% digestive material
and 34% protein. Hay costs $27.50 for 1000 kg, whereas grain and
v supplements cost $110 and $175, respectively, for 1000 kg. A healthy
cow’s daily diet must contain between 9.5 and 11.5 kg of digestive ma-
G
h terial and between 1.9 and 2.0 kg of protein. Find a formula for the
G cost per day, C , of feeding a cow in terms of the number of kilograms
of grain, G , and supplements, S , fed to the cow daily. What is the
T R L domain of this function?
∗ 31. A long piece of metal 1 m wide is bent in two places A and B (figure ∗ 34. The Easy University is buying computers. It has three models to
below) to form a channel with three straight sides. Find a formula for choose from. Each model A computer, with 64 MB of memory and a 3
the cross-sectional area of the channel in terms of x , θ , and φ . GB hard drive, costs $1300; model B, with 32 MB of memory and a 4
GB drive, costs $1200; and economy model C, with 16 MB of memory
x and a 1 GB drive, costs $1000. For reasons related to accreditation,
A B the university needs at least 2000 MB of memory and 150 GB of disk
space. If the computer lab must have 100 computers, set up a formula
∗ 32. (a) A uniform circular rod has flat ends at x = 0 and x = π for the cost of outfitting the lab in terms of the numbers x and y of
on the x -axis. The round side of the rod is insulated and computers of models A and B. What is the domain of the function?
12.2 Limits and Continuity

The concepts of limit and continuity for multivariable functions are exactly the same as for
functions of one variable; on the other hand, the work involved with the application of these
concepts is more complicated for multivariable functions.
Intuitively, a function f (x, y) is said to have limit L as x and y approach x0 and y0 if
f (x, y) gets arbitrarily close to L , and stays close to L , as x and y get arbitrarily close to
x0 and y0 . To say this in a precise mathematical way, it is convenient to represent pairs of
independent variables as points (x, y) in the xy -plane. We then have the following definition.
DEFINITION 12.1
A function f (x, y) has limit L as (x, y) approaches (x0 , y0 ) , written
lim f (x, y) = L, (12.2)

(x,y)→(x0 ,y0 )
if, given any $ > 0, we can find a δ > 0 such that
|f (x, y) − L| < $
!
whenever 0 < (x − x0 )2 + (y − y0 )2 < δ and (x, y) is in the domain of f (x, y) .
In other words, f (x, y) has limit L as (x, y) approaches (x0 , y0 ) if f (x, y) can be made
arbitrarily close to L (within $ ) by choosing points (x, y) sufficiently close to (x0 , y0 ) (within
a circle of radius δ ). Note the similarity of this definition to that for the limit of a function f (x)
of one variable in Section 2.6.
It is clear that
lim (x 2 + 2xy − 5) = 3,
(x,y)→(2,1)
but the limit

y2 − x2
lim
(x,y)→(0,0) y 2 + x 2
presents a problem, since both numerator and denominator approach zero as x and y approach
zero.
To conclude that limx→a f (x) = L , the limit must be L no matter how x approaches
a — be it through numbers larger than a , through numbers smaller than a , or through any
other approach. For limit 12.2, the limit of f (x, y) must also be L for all possible ways
of approaching (x0 , y0 ) . But in this case there might be a multitude of ways of approaching
(x0 , y0 ) . We might be able to approach (x0 , y0 ) along straight lines with various slopes, along
parabolas, along cubics, and so on. Definition 12.1 implies, then, that the limit exists only if it
is independent of the manner of approach. It is assumed, however, that we approach (x0 , y0 )
only through points (x, y) that lie in the domain of definition of the function.
For the second example above, suppose we approach the origin along the straight line
y = mx . Along this line,
y2 − x2 m2 x 2 − x 2 m2 − 1
lim = lim = .
(x,y)→(0,0) y 2 + x 2 x→0 m2 x 2 + x 2 m2 + 1
Because this result depends on m , we have shown that as (x, y) approaches (0, 0) along various
straight lines, the function (y 2 − x 2 )/(y 2 + x 2 ) approaches different numbers. We conclude,
therefore, that the function does not have a limit as (x, y) approaches (0, 0) . In Figure 12.6a
we show a portion of the surface z = (y 2 − x 2 )/(y 2 + x 2 ) to illustrate our conclusion. The
curve of intersection of the surface with the plane y = 2x is shown in Figure 12.6b. Heights
of points on the curve represent values of (y 2 − x 2 )/(y 2 + x 2 ) at points on the line y = 2x
in the xy -plane; they are always equal to 3/5. For other vertical planes containing the z -axis,
curves of intersection are at different heights.
FIGURE 12.6a Computer FIGURE 12.6b Curve of intersection of

y2 − x2 y2 − x2
plot of surface z = surface z = and plane y = 2x
y2 + x2 y2 + x2
z z
y = 2x
y y
x x
EXAMPLE 12.3
x2 − y2
Evaluate lim if it exists.
(x,y)→(0,0) x + y
SOLUTION Because points on the line y = −x are not within the domain of definition of
the function, we can write that
x2 − y2 (x − y)(x + y)
lim = lim = lim (x − y) = 0.
(x,y)→(0,0) x + y (x,y)→(0,0) x+y (x,y)→(0,0)
The concept of continuity for multivariable functions is contained in the following definition.
DEFINITION 12.2
A function f (x, y) is said to be a continuous function at a point (x0 , y0 ) if
1. f (x, y) is defined at (x0 , y0 )
2. lim f (x, y) exists
(x,y)→(x0 ,y0 )
3. The value of the function in 1 and its limit in 2 are the same.
All three conditions can be combined by writing the single equation
lim f (x, y) = f (x0 , y0 ). (12.3)

(x,y)→(x0 ,y0 )
The function f (x, y) = (y 2 − x 2 )/(y 2 + x 2 ) in Figure 12.6 is discontinuous at (0, 0) .

The function is undefined at (0, 0) and the limit as (x, y) → (0, 0) does not exist. Geo-
metrically, a function f (x, y) is continuous at a point (x0 , y0 ) if the surface z = f (x, y) is
2 2
not separated at the point (x0 , y0 , f (x0 , y0 )) . The function f (x, y) = 1 − e−1/(x +y ) is
discontinuous at (0, 0) since it is undefined for x = % 0 and y = 0. The surface
& has a hole at
(0, 0) (Figure 12.7). The function f (x, y) = sgn (x − 1)2 + (y − 2)2 (see Exercise 47 in
Section 2.4) is discontinuous at (1, 2) ; it has value 0 at x = 1 and y = 2, and value 1 for all
other values of x and y (Figure 12.8). The function h(x 2 + y 2 − 1) where h(x) is the heaviside
function of Section 2.5 has value zero everywhere inside the circle x 2 + y 2 < 1, does not have
a value on the circle, and has value 1 outside the circle (Figure 12.9). It is discontinuous at each
point on the circle. The surface z = h(x 2 + y 2 − 1) is that part of the xy -plane inside the
circle, and that part of the plane z = 1 above the outside of the circle.
FIGURE 12.7 Function FIGURE 12.8 Function FIGURE 12.9 Function

2 +y 2 )
f (x, y) = 1 − e−1/(x is f (x, y) = sgn[(x − 1)2 + (y − 2)2 ] is f (x, y) = h(x 2 + y 2 − 1)
discontinuous at (0, 0) discontinuous at (1, 2)
z
z z
1 Plane z = 1
z=1
1
1
x y
y y
(1, 2)
x x
EXERCISES 12.2
In Exercises 1–20 evaluate the limit, if it exists. x 3 + 4(y + 2)3
∗ 31. lim Hint: Approach (0, −2) along
(x,y)→(0,−2) 3x 3 − (y + 2)2
x2 − 1 x 3 + 2y 3
1. lim 2. lim straight lines.
(x,y)→(2,−3) x + y (x,y)→(1,1) x 3 + 4y 3
2x − 3y xyz
3. lim 4. lim x 2 − 2x − y 2 + 2y
(x,y)→(3,2) x + y (x,y,z)→(2,3,−1) x 2 + y 2 + z2
∗ 32. lim Hint: Approach (1, 1)
x (x,y)→(1,1) x 2 − 2x + y 2 − 2y + 2
−1
5. lim 6. lim Tan [x/(yz)] along straight lines.
(x,y)→(1,0) y (x,y,z)→(0,π/2,1)
7. lim Tan −1 (yz/x) 8. lim Tan −1 |yz/x|

(x,y,z)→(0,π/2,1) (x,y,z)→(0,π/2,1)
x 2 − 2x + y 2 + 2y − 2
2
|x − y | 2 2
|x + y | 2 ∗ 33. lim Hint: Approach (1, 1)
9. lim 10. lim
(x,y)→(1,1) x 2 − y 2 − 2x + 2y
(x,y)→(3,4) x 2 − y 2 (x,y)→(3,4) x 2 + y 2 along straight lines.
x2 − y2 x2 − y2
11. lim ∗ 12. lim √ √
(x,y)→(2,1) x − y (x,y)→(2,2) x − y x+y− x−y
∗ 34. lim
x2 − y2 x−y (x,y)→(1,0) y
∗ 13. lim ∗ 14. lim
(x,y)→(0,0) x − y (x,y)→(0,0) x + y
x2 − y2 sin (x 2 + y 2 )
15. lim ∗ 35. lim
(x,y,z)→(0,0,0) y 2 + z2 + 1 (x,y)→(0,0) x2 + y2
(x − 2)2 (y + 1)
16. lim
(x,y)→(2,1) x−2 ∗ 36. sin (x − y)
17. lim |2x − y − z| (a) Does the limit lim exist? Explain. Is
(x,y,z)→(1,1,1) (x,y)→(1,1)x−y
the function continuous at (1, 1) ?
3x 3 − y 3
∗ 18. lim
(x,y)→(0,0) 2 x 3 + 4y 3 (b) If we define the function everywhere by giving it the value
' (
−1 1 along the line y = x , does the limit of the function exist
∗ 19. lim Sec −1 at (1, 1) ? Is the function continuous at (1, 1) ?
(x,y)→(0,0) x2 + y2
∗ 20. lim Sec −1(x 2 + y 2 )
(x,y)→(0,0)
∗ 37. Give a mathematical definition for lim f (x, y, z) = L .
(x,y,z)→(x0 ,y0 ,z0 )
In Exercises 21–26 find all points of discontinuity for the function.

x2 − 1 xy ∗ 38. Is the following statement true or false? If a function f (x, y) is
21. f (x, y) = 22. f (x, y) = undefined at every point on a curve C , then for any point (x0 , y0 ) on
x+y x2 + y2
C, lim f (x, y) does not exist. Explain. Give an example.
1 1 (x,y)→(x0 ,y0 )
23. f (x, y) = 24. f (x, y, z) =
1 − x2 − y2 xyz
x+y  2 2
25. f (x, y) = |x − y| ∗ 26. f (x, y) = 2  x y
x y + xy 2 , if (x, y) (= (0, 0)
) + ∗ 39. Let f (x, y) = x 4 + y 4
* 
∗ 27. Evaluate lim 2
cos (x + y) − 1 − sin (x + y) where 0, if (x, y) = (0, 0).
(x,y)→(a,a)
0 ≤ a ≤ π/2. (a) Show that f (x, y) is continuous in each variable separately
at (0, 0) . In other words, show that f (x, 0) and f (0, y)
In Exercises 28–35 evaluate the limit, if it exists. are continuous at x = 0 and y = 0.
x4 + y2 (b) Show that f (x, y) is not continuous at (0, 0) .

∗ 28. lim Hint: Approach (0, 0) along parabolas.
(x,y)→(0,0) x 4 − y 2
x 6 − 2y 2
∗ 29. lim Hint: Approach (0, 0) along cubic curves. ∗∗ 40. Prove that:
(x,y)→(0,0) 3x 6 + y 2
(x − 1)2 + y 2 (a) lim(x,y)→(0,0) (xy + 5) = 5

∗ 30. lim Hint: Approach (1, 0) along
(x,y)→(1,0) 3(x − 1)2 − 2y 2
straight lines. (b) lim(x,y)→(1,1) (x 2 + 2xy + 5) = 8
12.3 Partial Derivatives

We now define partial derivatives of multivariable functions and interpret these derivatives
algebraically and geometrically.
DEFINITION 12.3
The partial derivative of a function f (x, y) with respect to x is
∂f f (x + 'x, y) − f (x, y)
= lim , (12.4)
∂x 'x→0 'x
and the partial derivative with respect to y is
∂f f (x, y + 'y) − f (x, y)

= lim . (12.5)
∂y 'y→0 'y
It is evident from equation 12.4 that the partial derivative of f (x, y) with respect to x is
simply the ordinary derivative of f (x, y) with respect to x , where y is considered a constant.
Similarly, ∂f/∂y is the ordinary derivative of f (x, y) with respect to y , holding x constant.
For the partial derivative of a function of more than two independent variables, we again
permit one variable to vary, but hold all others constant. For example, the partial derivative of
f (x, y, z, t, . . .) with respect to z is
∂f f (x, y, z + 'z, t, . . .) − f (x, y, z, t, . . .)
= lim . (12.6)
∂z 'z→0 'z
Hence, we differentiate with respect to z while treating x, y, t, . . . as constants.
Other notations for the partial derivative are common. In particular, for ∂f/∂x when
z = f (x, y) , there are also
( (
∂z ∂f ∂z
, fx , zx , , and ,
∂x ∂x y ∂x y
the last two indicating that the variable y is held constant when differentiation with respect to
x is performed.
EXAMPLE 12.4
Find ∂z/∂x and ∂z/∂y if z = sin (x 2 + y 3 ) + exy .
SOLUTION For this function,
∂z ∂z
= 2x cos (x 2 + y 3 ) + yexy and = 3y 2 cos (x 2 + y 3 ) + xexy .
∂x ∂y
EXAMPLE 12.5
Find ∂f/∂x at the point (1, 2, 3) if f (x, y, z) = x 2 /y 4 + 3xz + 4.
SOLUTION Since ∂f/∂x = 2x/y 4 + 3z ,
∂f // 73
/ = 2(1)/24 + 3(3) = .
∂x (1, 2, 3) 8
For a function y = f (x) of one variable, we defined differentials dx and dy in such a way
that the derivative dy/dx could be thought of as a quotient. This is not done for functions of
∂f
more than one variable. Although we write the partial derivative in the form ∂f/∂x (for
∂x
typographical reasons), we never consider it as a quotient.
Algebraically, the partial derivative ∂f/∂x represents the rate of change of f (x, y, . . .)
with respect to x when all other variables in f (x, y, . . .) are held constant. For instance,
V = π r 2 h/3 represents the volume of a right-circular cone with height h and radius r , and
therefore ∂V /∂r = 2π rh/3 represents the rate of change of the volume of the cone as the base
radius changes and the height remains fixed. Similarly, ∂V /∂h = π r 2 /3 is the rate of change
FIGURE 12.10 Geometric
of the volume of the right-circular cone as the height changes and the radius is kept fixed. We
interpretation of partial derivatives
of a function f (x, y) of two
shall learn later how to calculate the rate of change of the volume when the radius and height
variables are both changing.
We can interpret the partial derivative of a function geometrically when the function can
z
be interpreted geometrically, namely, when there are only two independent variables. Consider,
(x0 , y0 , z0) then, a function f (x, y) that is represented geometrically as a surface z = f (x, y) in Figure
z = f (x, y) 12.10. If we intersect this surface with a plane y = y0 = a constant, we obtain a curve with
equations
y = y0 , z = f (x, y0 ). (12.7)
y = y0 Because this curve lies in the plane y = y0 , we can talk about its tangent line at the point
(x0 , y0 , z0 ) , where z0 = f (x0 , y0 ) . The slope of this tangent is the derivative of z with respect
to x , but because y is being held constant at y0 , it must be the partial derivative of z with
y respect to x . In other words, the slope of the tangent line to the curve in Figure 12.10 at the
x (x0 , y0)
point (x0 , y0 , z0 ) is ∂f/∂x|(x0 ,y0 ) . Similarly, the partial derivative ∂f/∂y evaluated at (x0 , y0 )
represents the slope of the tangent line to the curve of intersection of z = f (x, y) and the plane
x = x0 at the point (x0 , y0 , z0 ) .
EXERCISES 12.3
In Exercises 1–20 evaluate ∂f/∂x and ∂f/∂y . 24. ∂f/∂x at (1, −1, 1, −1) if f (x, y, z, t) = zt/(x 2 + y 2 − t 2 )
1. f (x, y) = x 3 y 2 + 2xy 2. f (x, y) = 3xy − 4x 4 y 4 !

25. ∂f/∂t if f (x, y, t) = x t 2 − y 2 /t 2 + (x/y) Sec −1 (t/3)
3. f (x, y) = x 4 /y 3 4. f (x, y) = x/(x + y) − x/y
26. ∂f/∂x if f (x, y, z) = Cot −1 (1 + x + y + z)
5. f (x, y) = x/(2x 2 + y) 6. f (x, y) = sin (xy)
!
7. f (x, y) = x cos (x + y) 8. f (x, y) = x2 + y2 27. ∂f/∂y at (1, 2, 3) if f (x, y, t) = Sin −1 (xyt)/Cos −1 (xyt)
!
9. f (x, y) = x x 2 − y 2 10. f (x, y) = tan (2x 2 + y 2 )
28. ∂f/∂x if f (x, y, z) = x 3 /y + x sin (yz/x)
11. f (x, y) = e x+y
12. f (x, y) = e xy
29. ∂f/∂t if f (x, y, z, t) = xyz ln (x 2 + y 2 + z2 )

13. f (x, y) = xyexy 14. f (x, y) = ln (x 2 + y 2 )
√
15. f (x, y) = (x + 1) ln (xy) 16. f (x, y) = sin (yex ) 30. ∂f/∂z if f (x, z) = (z2 /2) Sin −1 (x/z)+ (x/2) z2 − x 2
!
17. f (x, y) = Tan −1 (x/y) 18. f (x, y) = 3
1 − cos3 (x 2 y)
∂f ∂f
sin x 0 √ 1 ∗ 31. If f (x, y) = x 3 y/(x−y) , show that x +y = 3f (x, y) .
19. f (x, y) = 20. f (x, y) = ln sec x+y ∂x ∂y
cos y
∗ 32. If f (x, y, z) = (x 4 + y 4 + z4 )/(xyz) , show that
∂f ∂f ∂f
In Exercises 21–30 evaluate the derivative indicated. x +y +z = f (x, y, z) .
∂x ∂y ∂z
2 +y 2
21. ∂f/∂x if f (x, y, z) = xyzex
∗ 33. If f (x, y, z) = (x 2 + y 2 ) cos [(y + z)/x ], show that
22. ∂f/∂z if f (x, z) = Tan −1 [1/(x 2 + z2 )] ∂f ∂f ∂f
x +y +z = 2f (x, y, z) .
23. ∂f/∂y at (1, 1, 0) if f (x, y, z) = xy(x 2 + y 2 + z2 )1/3 ∂x ∂y ∂z
∗ 34. To evaluate ∂f/∂x for f (x, y) at the point (1, 2) , state which of (a) ρ = constant, u = (2x 2 − xy + z2 )t , v = (x 2 − 4xy +
the following are acceptable: y 2 )t , w = (−2xy − yz + y 2 )t
(a) Differentiate f (x, y) with respect to x holding y constant, (b) ρ = xy+zt, u = x 2 y+t, v = y 2 z−2t 2 , w = 5x+2z
and then set x = 1 and y = 2.
(b) Set x = 1 and y = 2, and then differentiate with respect
to x . ∗ 40. A gas-filled pneumatic strut behaves like the piston-cylinder ap-
paratus shown below. At one instant when the piston is L = 0.15 m
(c) Set y = 2, differentiate with respect to x , and set x = 1.
away from the closed end of the cylinder, the gas density is uniform at
(d) Set x = 1, differentiate with respect to x , and set y = 2. ρ = 18 kg/m 3 , and the piston begins to move away from the closed
end at a constant rate of 12 m/s. The gas motion is one-dimensional
∗ 35. Temperature at points (x, y) in a semicircular plate defined by and proportional to distance from the closed end. It varies linearly
x 2 + y 2 ≤ 4, y ≥ 0 is given by T (x, y) = 16x 2 − 24xy + 40y 2 . from zero at the closed end of the cylinder to 12 m/s at the piston. Gas
Find, if possible, (a) Tx (1, 1) , (b) Ty (1, 1) , (c) Tx (1, 0) , (d) Ty (1, 0) , density is always uniform throughout the cylinder, but varies in time.
(e) Tx (0, 2) , and (f) Ty (0, 2) . Use the equation of continuity in Exercise 39 to find the density of the
gas as a function of time.
∗ 36. In the figure below, two identical bars AB and BC are pinned at
B as well as at A and C . Each bar is initially of length L , and initially
point B lies a distance h above line AC . When a vertical force with
magnitude F is applied at B , the vertical displacement x > 0 of B is
related to F by the equation
' (' (
h−x L
F = 2AE √ −1 ,
L L2 − 2hx + x 2
where A is the cross-sectional area of the bar, and E is Young’s mod-
ulus for !the bars. Show that ∂F /∂x vanishes for x = h − (L3 − ∗ 41. In complex variable theory, two functions u(x, y) and v(x, y) are
2 1/ 3 said to be harmonic conjugates in a region R if in R they satisfy the
Lh ) 1 − (1 − h2 /L2 )1/3 .
Cauchy–Riemann equations
F
∂u ∂v ∂v ∂u
= , =− .
B ∂x ∂y ∂x ∂y
L x L
h Show that the following pairs of functions are harmonic conjugates:
(a) u(x, y) = −3xy 2 +y+x 3 , v(x, y) = 3x 2 y−y 3 −x+5
A C
(b) u(x, y) = (x 2 + x + y 2 )/(x 2 + y 2 ), v(x, y) =
−y/(x 2 + y 2 )
(c) u(x, y) = ex (x cos y−y sin y), v(x, y) = ex (x sin y+
∗ 37. Can you find a function f (x, y) so that fx (x, y) = 2x − 3y and y cos y)
fy (x, y) = 3x + 4y ?
∗ 38. Suppose a , b , and c are the lengths of the sides of a triangle ∗ 42. If r and θ are polar coordinates, then the Cauchy–Riemann equa-
and A , B , and C are the opposite angles. Find (a) aA (b, c, A) , (b) tions in Exercise 41 for functions u(r, θ) and v(r, θ) take the form
Aa (a, b, c) , (c) ab (b, c, A) , and (d) Ab (a, b, c) .
∂u 1 ∂v 1 ∂u ∂v
∗ 39. The equation of continuity for three-dimensional unsteady flow of = , =− , r (= 0.
a compressible fluid is ∂r r ∂θ r ∂θ ∂r
∂ρ ∂ ∂ ∂ Show that the following pairs of functions satisfy these equations:

+ (ρu) + (ρv) + (ρw) = 0,
∂t ∂x ∂y ∂z (a) u(r, θ) = (r 2 + r cos θ)/(1 + r 2 + 2r cos θ), v(r, θ) =
r sin θ/(1 + r 2 + 2r cos θ)
where ρ(x, y, z, t) is the density of the fluid, and uî + v ĵ + w k̂ is √ √
(b) u(r, θ) = r cos (θ/2), v(r, θ) = r sin (θ/2)
the velocity of the fluid at position (x, y, z) and time t . Determine
whether the continuity equation is satisfied if: (c) u(r, θ) = ln r, v(r, θ) = θ
12.4 Gradients
Suppose a function f (x, y, z) is defined at each point in some region of space, and that at each
point of the region all three partial derivatives
∂f ∂f ∂f
, ,
∂x ∂y ∂z
exist. For example, if f (x, y, z) represents the present temperature at each point in the room
in which you are working, then these derivatives represent rates of change of temperature in
directions parallel to the x -, y -, and z -axes, respectively. There is a particular combination of
these derivatives that proves very useful in later work. This combination is contained in the
following definition.
DEFINITION 12.4
If a function f (x, y, z) has partial derivatives ∂f/∂x , ∂f/∂y , and ∂f/∂z at each point
in some region D of space, then at each point in D we define a vector called the gradient
of f (x, y, z) , written grad f or ∇f , by
∂f ∂f ∂f
grad f = ∇f = î + ĵ + k̂. (12.8)
∂x ∂y ∂z
For a function f (x, y) of only two independent variables, we have
∂f ∂f
∇f = î + ĵ. (12.9)
∂x ∂y
EXAMPLE 12.6
If f (x, y, z) = x 2 yz − 2x/y , find ∇f at (1, −1, 3) .
SOLUTION Since ∇f = (2xyz − 2/y)î + (x 2 z + 2x/y 2 )ĵ + (x 2 y)k̂, we have
∇f|(1,−1,3) = −4î + 5ĵ − k̂.
EXAMPLE 12.7
If f (x, y, z) = Tan−1 (xy/z) , what is ∇f ?
SOLUTION
) ' (+ ) ' (+ ) ' (+
1 y 1 x 1 −xy
∇f = î + ĵ + k̂
1 + (xy/z)2 z 1 + (xy/z)2 z 1 + (xy/z)2 z2
yz xz xy
= î + 2 ĵ − 2 k̂
z2 2
+x y 2 z +x y2 2 z + x2y2
= (yzî + xzĵ − xy k̂)/(z2 + x 2 y 2 ).
Gradients arise in a multitude of applications in applied mathematics — heat conduction, elec-

tromagnetic theory, and fluid flow, to name a few — and two of the properties that make them so
indispensable are discussed in detail in Sections 12.8 and 12.9. Examples 12.8 and 12.9 suggest
these properties, but we make no attempt at a complete discussion here. For the moment we
simply want you to be familiar with the definition of gradients and be able to calculate them.
12.4 Gradients 811
EXAMPLE 12.8
FIGURE 12.11 Gradient The equation F (x, y, z) = 0, where F (x, y, z) = x 2 + y 2 + z2 − 4, defines a sphere of
of function defining a sphere is radius 2 centred at the origin. Show that the gradient vector ∇F at any point on the sphere is
perpendicular to the sphere perpendicular to the sphere.
z SOLUTION If P (x, y, z) is any point on the sphere (Figure 12.11), then the position vector
∇F r = x î + y ĵ + zk̂ from the origin to P is clearly perpendicular to the sphere. On the other
hand,
∇F = 2x î + 2y ĵ + 2zk̂ = 2r.
P (x, y, z)
r Consequently, at any point P on the sphere, ∇F is also perpendicular to the sphere.
y
This example suggests that gradients may be useful in finding perpendiculars to surfaces (and,
x as we will see, perpendiculars to curves).
EXAMPLE 12.9
The function f (x, y) = 2x 2 − 4x + 3y 2 + 2y + 6 is defined at every point in the xy -plane. If
we start at the origin (0, 0) and move along the positive x -axis, the rate of change of the function
is fx (0, 0) = −4; if we move along the y -axis, the rate of change is fy (0, 0) = 2. Calculate
the rate of change of f (x, y) at (0, 0) if we move toward the point (1, 1) along the line y = x ,
and show that it is equal to the component of ∇f|(0,0) in the direction v = (1, 1) .
SOLUTION The difference in values of f (x, y) at any point (x, y) and (0, 0) is f (x, y) −
f (0, 0) = (2x 2 − 4x + 3y 2 + 2y + 6) − (6) = 2x 2 − 4x + 3y 2 + 2y . If we divide this by
the length of the line joining (0, 0) and (x, y) , we obtain
f (x, y) − f (0, 0) 2x 2 − 4x + 3y 2 + 2y
! = ! .
x2 + y2 x2 + y2
The limit of this quotient as (x, y) approaches (0, 0) along the line y = x should yield the
required rate of change. We therefore set y = x and take the limit as x approaches zero through
positive numbers:
2x 2 − 4x + 3x 2 + 2x x(5x − 2) √
lim √ = lim+ √ = − 2.
x→0+ x2 + x2 x→0 2x
This quantity, then, is the rate of change of f (x, y) at (0, 0) along the line y = x toward the
point (1, 1) .
On the other hand,
∇f|(0,0) = (4x − 4, 6y + 2)|(0,0) = (−4, 2),
and the component of this vector in the direction v = (1, 1) is
(1, 1) −4 + 2 √
∇f|(0,0) · v̂ = (−4, 2) · √ = √ = − 2.
2 2
This example indicates that gradients may be useful in calculating rates of change of functions
in directions other than those parallel to the coordinate axes.
EXAMPLE 12.10
The electrostatic potential at a point (x, y, z) in space due to a charge q fixed at the origin is
given by
q
V = ,
4 π $0 r
!
where r = x 2 + y 2 + z2 . If a second charge Q is placed at (x, y, z) , it experiences a force
F, where
qQ
F = r,
4 π $0 r 3
where r = x î + y ĵ + zk̂. Show that F = −Q∇V .
SOLUTION We show that ∇V = −F/Q ,
' ( ' (
q q 1
∇V = ∇ = ∇ !
4 π $0 r 4 π $0 x 2 + y 2 + z2
) +
q −x −y −z
= î + ĵ + k̂
4π $0 (x 2 + y 2 + z2 )3/2 (x 2 + y 2 + z2 )3/2 (x 2 + y 2 + z2 )3/2
−q −q F
= 3
(x î + y ĵ + zk̂) = 3
r = − .
4 π $0 r 4 π $0 r Q
EXERCISES 12.4
In Exercises 1–10 find the gradient of the function. 16. The equation F (x, y, z) = Ax + By + Cz + D = 0 defines a
plane in space. Show that at any point on the plane the vector ∇F is
1. f (x, y, z) = x 2 y + xz + yz2 perpendicular to the plane.
2. f (x, y, z) = x 2 yz 17. Use the result of Exercise 16 to illustrate that a vector along the
2
3. f (x, y, z) = x y/z − 2xz 6 line
F (x, y, z) = 2x + 3y − 2z + 4 = 0,
4. f (x, y) = x 2 y + xy 2
5. f (x, y) = sin (x + y) G(x, y, z) = x − y + 3z + 6 = 0
6. f (x, y, z) = Tan −1 (xyz) is ∇F × ∇G . Find parametric equations for the line.

7. f (x, y) = Tan −1 (y/x) 18. Prove that if f (x, y, z) and g(x, y, z) both have gradients, then
8. f (x, y, z) = ex+y+z ∇(fg) = f ∇g + g∇f . What does this remind you of?
9. f (x, y) = 1/(x 2 + y 2 ) ∗ 19. Repeat Example 12.9 for the functions (a) f (x, y) = x 2 + y 2
! and (b) f (x, y) = 2x 3 − 3y .
10. f (x, y, z) = 1/ x 2 + y 2 + z2
∗ 20. The equation F (x, y) = x 3 +xy +y 4 − 5 = 0 implicitly defines
a curve in the xy -plane. Show that at any point on the curve, ∇F is a
In Exercises 11–15 find the gradient of the function at the point. normal vector to the curve.
11. f (x, y) = xy + x + y at (1, 3)

12. f (x, y, z) = cos (x + y + z) at (−1, 1, 1) Draw the surface defined by the equations in Exercises 21–22. At what
points on the surface is ∇F not defined?
13. f (x, y, z) = (x 2 + y 2 + z2 )2 at (0, 3, 6)
!
14. f (x, y) = e−x
2 −y 2
at (2, 2) 21. F (x, y, z) = z − x2 + y2 = 0
15. f (x, y, z) = xy ln (x + y) at (4, −2) 22. F (x, y, z) = z − |x − y| = 0
!
∗ 23. If f (x, y) = 1 − x 2 − y 2 , find ∇f . Find the point (x, y) at ∗ 27. Repeat Exercise 26 if ∇f = (x î + y ĵ + zk̂)/ x 2 + y 2 + z2 .
which ∇f = 0 , and illustrate graphically the nature of the surface
z = f (x, y) at this point. ∗ 28. If f (x, y) and g(x, y) have first partial derivatives in a region
R of the xy -plane, and if in R , ∇f = ∇g , how are f (x, y) and
∗ 24. If the gradient of a function f (x, y) is ∇f = (2xy − y)î + g(x, y) related?
(x 2 − x)ĵ , what is f (x, y) ? ∗ 29. If ∇f = 0 for all points in some region R of space, what can we
say about f (x, y, z) in R ?
∗ 25. Repeat Exercise 24 if ∇f = (2x/y + 1)î + (−x 2 /y 2 + 2)ĵ .
∗∗ 30. Show that if the equation F (x, y) = 0 implicitly defines a curve C
∗ 26. If the gradient of a function f (x, y, z) is ∇f = yzî + (xz + in the xy -plane, then at any point on C the vector ∇F is perpendicular
2yz)ĵ + (xy + y 2 )k̂ , what is f (x, y, z) ? to C .
12.5 Higher-Order Partial Derivatives

If f (x, y) = x 3 y 2 + yex , then
∂f ∂f
= 3x 2 y 2 + yex and = 2x 3 y + e x .
∂x ∂y
Since each of these partial derivatives is a function of x and y , we can take further partial
derivatives. The partial derivative of ∂f/∂x with respect to x is called the second partial
derivative of f (x, y) with respect to x , and is written
' (
∂ ∂f ∂ 2f
= = 6xy 2 + yex .
∂x ∂x ∂x 2
Similarly, we have three more second partial derivatives:
' (
∂ ∂f ∂ 2f
= = 6x 2 y + ex ,
∂y ∂x ∂y ∂x
' (
∂ ∂f ∂ 2f
= = 6x 2 y + ex ,
∂x ∂y ∂x ∂y
' (
∂ ∂f ∂ 2f
= = 2x 3 .
∂y ∂y ∂y 2
Note that the second partial derivatives ∂ 2 f/∂x ∂y and ∂ 2 f/∂y ∂x are identical. This is not a
peculiarity of this function; according to the following theorem, it is to be expected.
THEOREM 12.1
If f (x, y) , ∂f/∂x , ∂f/∂y , ∂ 2 f/∂x ∂y , and ∂ 2 f/∂y ∂x are all defined inside a circle
centred at a point P , and are continuous at P , then at P
∂ 2f ∂ 2f
= . (12.10)
∂x ∂y ∂y ∂x
Corresponding to the subscript notation fx (x, y) for ∂f/∂x , we have the following nota-
tions for second partial derivatives:
∂ 2f ∂ 2f ∂ 2f ∂ 2f
= fxx , = fyx , = fxy , = fyy .
∂x 2 ∂x ∂y ∂y ∂x ∂y 2
Notice the reversal in order of x and y in the middle terms. In subscript notation, derivatives are
taken in the order in which they appear (left to right, y first, x second, in fyx ). In ∂ 2 f/∂x ∂y ,
derivatives are done right to left, y first, x second. Because of Theorem 12.1, the order is usually
irrelevant anyway.
Partial derivatives of orders higher than two are also possible. For the function f (x, y) =
x 3 y 2 + yex above, we have
' (
∂ 3f ∂ ∂ 2f ∂ 0 1
fxxy = 2
= = 6xy 2 + yex = 12xy + ex
∂y ∂x ∂y ∂x 2 ∂y
and
) ' 2 (+ ) +
∂ ∂ ∂ f ∂ ∂ 0 2 x
1 ∂ % 2&
fyxyx = = 6x y + e = 6x = 12x.
∂x ∂y ∂x ∂y ∂x ∂y ∂x
For most functions with which we will be concerned, Theorem 12.1 can be extended to say
that a mixed partial derivative may be calculated in any order whatsoever. For example, if we
require ∂ 10 f/∂x 3 ∂y 7 , where f (x, y) = ln(y y ) + x 2 y 10 , it is advantageous to reverse the
order of differentiation:
' (
∂ 10 f ∂ 10 f ∂ 7 ∂ 3f
= = = 0.
∂x 3 ∂y 7 ∂y 7 ∂x 3 ∂y 7 ∂x 3
EXAMPLE 12.11
2y 3
Show that the function f (x, y) = Tan −1 satisfies the equation
x
∂ 2f ∂ 2f
+ = 0.
∂x 2 ∂y 2
SOLUTION Since
∂f 1 2 y3 −y
= 2
− 2 = 2 ,
∂x y x x + y2
1+ 2
x
∂ 2f 2xy
the second derivative with respect to x is 2
= 2 . Since
∂x (x + y 2 )2
' (
∂f 1 1 x
= = 2 ,
∂y y2 x x + y2
1+ 2
x
∂ 2f −2xy
the second derivative with respect to y is 2
= 2 . When added, these second
∂y (x + y 2 )2
derivatives cancel one another, and this completes the proof.
The equation
∂ 2f ∂ 2f
+ =0 (12.11)
∂x 2 ∂y 2
for a function f (x, y) is one of the most important equations in applied mathematics. It is called
Laplace’s equation in two variables ( x and y ). Laplace’s equation for a function f (x, y, z)
of three variables is
∂ 2f ∂ 2f ∂ 2f
+ + = 0. (12.12)
∂x 2 ∂y 2 ∂z2
A function is said to be a harmonic function in a region R if it satisfies Laplace’s equation in

R and has continuous second partial derivatives in R . In particular, the function f (x, y) in
Example 12.11 is harmonic in any region that does not contain points on the y -axis. The next
two examples illustrate areas of applied mathematics in which Laplace’s equation is prominent.
A third is contained in Exercise 30.
EXAMPLE 12.12
Show that the electrostatic potential function of Example 12.10 is harmonic in any region not
containing the origin.
% &
SOLUTION With V = q/(4π $0 ) (x 2 + y 2 + z2 )−1/2 ,
) +
∂V q −x
=
∂x 4π $0 (x 2 + y 2 + z2 )3/2
and
) +
∂ 2V −q 1 −3x 2
= + 2
∂x 2 4π $0 (x 2 + y 2 + z2 )3/2 (x + y 2 + z2 )5/2
) +
q 2x 2 − y 2 − z 2
= .
4π $0 (x 2 + y 2 + z2 )5/2
Similarly,
) + ) +
∂ 2V q 2y 2 − x 2 − z 2 ∂ 2V q 2z 2 − x 2 − y 2
= , = .
∂y 2 4π $0 (x 2 + y 2 + z2 )5/2 ∂z2 4π $0 (x 2 + y 2 + z2 )5/2
Addition of these shows that Vxx + Vyy + Vzz = 0. Since second partial derivatives are
continuous in any region not containing (0, 0, 0) , V (x, y, z) is harmonic therein.
EXAMPLE 12.13
Figure 12.12 shows a 1-m by 1-m metal plate that is insulated top and bottom. Temperature along
the edges x = 0, x = 1, and y = 1 is held at 0◦ C, while that along y = 0 is f (x) = 4 sin π x .
Steady-state temperature at points inside the plate is then
4 % &
T (x, y) = eπ(1−y) − e−π(1−y) sin π x.
eπ − e−π
Show that T (x, y) is harmonic inside the plate.

FIGURE 12.12 Temperature in a plate
0•C
1
0•C 0•C
T = 4 sin x 1 x
SOLUTION Partial derivatives of T (x, y) are
∂T 4π % π(1−y) −π(1−y)
&
= π e − e cos π x,
∂x e − e−π
∂ 2T −4π 2 % π(1−y) −π(1−y)
&
= e − e sin π x,
∂x 2 eπ − e−π
∂T 4 % &
= π −π
−π eπ(1−y) − π e−π(1−y) sin π x,
∂y e −e
∂ 2T 4 % 2 π(1−y) &
2
= π −π
π e − π 2 e−π(1−y) sin π x.
∂y e −e
Clearly, ∂ 2 T /∂x 2 + ∂ 2 T /∂y 2 = 0. Since second partial derivatives are continuous for 0 <
x < 1 and 0 < y < 1, the temperature function T (x, y) is harmonic in this region. In
practice, we are not given T (x, y) ; we must find it; that is, we must solve Laplace’s equation
∂ 2T ∂ 2T
2
+ = 0, 0 < x < 1, 0 < y < 1,
∂x ∂y 2
subject to the conditions
T (0, y) = 0, 0 < y < 1,
T (1, y) = 0, 0 < y < 1,
T (x, 1) = 0, 0 < x < 1,
T (x, 0) = 4 sin π x, 0 < x < 1.
This is called a boundary-value problem; it is treated in books dealing with partial differential
equations. Laplace’s equation is a partial differential equation, that is, a differential equation
with partial derivatives.
Longitudinal Vibrations of Bars

Figure 12.13 shows a bar of length L and uniform cross-section. If a longitudinal force (in
the x -direction) is applied to each cross-section, the bar stretches and/or compresses. (The bar
acts like a very stiff spring.) Suppose we denote displacement of the cross-section normally at
position x by y(x, t) , where t is time. This allows for the situation when the applied force
F (x, t) varies along the bar and is also a function of time. It can be shown that y(x, t) must
satisfy the following partial differential equation, called the one-dimensional wave equation,
∂ 2y E ∂ 2y F (x, t)
2
= 2
+ , 0 < x < L, t > 0, (12.13a)
∂t ρ ∂x ρ
where ρ is the density of the material in the bar and E is Young’s modulus of elasticity. It is a
constant that depends on the material of the bar; the larger the value of E , the more resistant the
bar is to stretch or compression. Given F (x, t) , ρ , and E , the objective is to solve the wave
equation for y(x, t) , therefore giving positions of cross-sections of the bar for all time.
FIGURE 12.13 Longitudinal vibrations of a bar
End of bar Cross-section at position x when bar

fixed at x = 0 unstretched and uncompressed
x=0 L x
x
x=0 Position of cross-section L x

x normally at x
By itself, the wave equation has many solutions; other conditions must be stipulated. New-
ton’s second law governs motion (and it was used in developing the wave equation). Our
experience with particle motion suggests that we require two initial conditions, one specifying
the initial positions of cross-sections of the bar, and a second specifying their velocities. In other
words, accompanying 12.13a will be initial conditions of the form
y(x, 0) = f (x), 0 < x < L, (12.13b)
yt (x, 0) = g(x), 0 < x < L. (12.13c)

There will also be boundary conditions specifying what is happening at the ends of the bar.
For instance, if end x = 0 is clamped in position and, therefore, not allowed to move, y(x, t)
satisfies
y(0, t) = 0, t > 0. (12.13d)
If end x = L of the bar is allowed to move freely, then the boundary condition there is
yx (L, t) = 0, t > 0. (12.13e)
There are many types of boundary conditions that can occur at x = 0 and x = L . These are
two examples.
Let us take a very simple illustration. The problem is much easier if displacement is not
a function of time, in which case we solve for what are called static displacements of the bar.
For this to occur, the applied force must be independent of time. Displacement becomes only
a function of position, y(x) , initial conditions are dropped, time disappears from the boundary
conditions, and the partial differential equation becomes an ordinary differential equation,
d 2y
0 = E + F (x), 0 < x < L, (12.14a)
dx 2
y(0) = 0, (12.14b)
y + (L) = 0. (12.14c)
If, for instance, all cross-sections are subjected to the same force F (a constant), then
d 2y F F x2
=− ,⇒ y(x) = − + Cx + D.
dx 2 E 2E
The boundary conditions require that
FL FL
0 = y(0) = D, 0 = y + (L) = − +C ,⇒ .C =
E E
Thus, displacement of the cross-section of the bar from its equilibrium position x is
F x2 F Lx Fx
y(x) = − + = (2L − x).
2E E 2E
Its graph is shown in Figure 12.14. To get an idea of the magnitude of these displacements, we
find that for a 1-m steel bar with E = 2.0 × 1011 N/m 2 , and an applied force F = 105 N, the
right end has displacement
105 % &
y(1) = 2(1) − 1 = 2.5 × 10−5 m;
2 (2 × 1011 )
that is, the bar stretches by only 0.025 mm.
FIGURE 12.14 Static displacements of a bar subjected to a longitudinal force
FL2
2E
L x
The one-dimensional wave equation 12.13a is satisfied by longitudinal vibrations of bars.

It is also satisfied by transverse vibrations of strings and rotational vibrations of bars. These
will be discussed in the exercises.
EXERCISES 12.5
In Exercises 1–20 find the derivative. 11. ∂ 3 f/∂x 2 ∂y if f (x, y) = x 2 ey + y 2 ex
1. ∂ 2 f/∂x 2 if f (x, y) = x 2 y 2 − 2x 3 y 12. ∂ 2 f/∂x 2 if f (x, y) = Tan −1 (y/x)
2. ∂ 3 f/∂y 3 if f (x, y) = 2x/y + 3x 3 y 4 13. ∂ 3 f/∂x ∂y 2 if f (x, y, z) = cot (x 2 + y 2 + z2 )
2 2
3. ∂ f/∂z if f (x, y, z) = sin (xyz)
14. ∂ 2 f/∂x ∂y at (−2, −2) if f (x, y) = Sin −1 (x 2 + y 2 )−1
4. ∂ 2 f/∂y ∂z if f (x, y, z) = xyzex+y+z
! 15. ∂ 10 f/∂x 7 ∂y 3 if f (x, y) = x 7 ex y 2 + 1/y 6
5. ∂ 2 f/∂y ∂x if f (x, y) = x2 + y2
16. ∂ 8 f/∂x 8 if f (x, y, z) = x 8 y 9 z10
6. ∂ 3 f/∂x 2 ∂y if f (x, y) = ex+y − x 2 /y 2
7. ∂ 3 f/∂y 3 at (1, 3) if f (x, y) = 3x 3 y 3 − 3x/y 17. ∂ 6 f/∂x 2 ∂y 2 ∂z2 if f (x, y, z) = 1/x 2 + 1/y 2 + 1/z2
8. ∂ 3 f/∂x ∂y ∂z at (1, 0, −1) if f (x, y, z) = x 2 y 2 + x 2 z2 + y 2 z2 18. ∂ 4 f/∂x 3 ∂y if f (x, y) = cos (x + y 3 )

! !
9. ∂ 2 f/∂x 2 if f (x, y) = 1 − x2 − y2 19. ∂ 4 f/∂x ∂y ∂z ∂t if f (x, y, z, t) = x 2 + y 2 + z2 − t 2
!
10. ∂ 2 f/∂z2 if f (x, y, z) = ln x 2 + y 2 + z2 20. ∂ 2 f/∂x ∂y if f (x, y) = Sec −1 (xy)
∗ 21. If z = x 2 + xy + y 2 sin (x/y) , show that ∗ 32. When the bar in Figure 12.13 is turned vertically and clamped at
x = 0 (figure below), static deflections of the bar must satisfy the
∂z ∂z ∂ 2z ∂ 2z ∂ 2z following problem:
x +y = 2z = x 2 2 + 2xy + y2 2 .
∂x ∂y ∂x ∂x ∂y ∂y
d 2y
∗ 22. If u = x + y + zey/x , show that 0 = E + F (x), 0 < x < L,
dx 2
∂ 2u 2
2∂ u
2
2∂ u
x2 + y + z
∂x 2 ∂y 2 ∂z2 y(0) = 0, y + (L) = 0,
∂ 2u ∂ 2u ∂ 2u where F (x) is the weight of that part of the bar below the cross-section
+ 2xy + 2yz + 2xz = 0.
∂x ∂y ∂y ∂z ∂x ∂z that would be at position x if the bar were unstretched. Find the length
of the bar as it stretches under its own weight.
In Exercises 23–28 find a region (if possible) in which the function is x=0 x=0
harmonic.
x x
∗ 23. f (x, y) = x 2 − y 2 + 2xy + y
∗ 24. f (x, y) = ln (x 2 + y 2 ) Unstretched Stretched
3 2 position position
∗ 25. f (x, y) = x y − 3xy y
2 3
∗ 26. f (x, y, z) = 3x yz − y z + xy
!
∗ 27. f (x, y, z) = 1/ x 2 + y 2 + z2
∗ 28. f (x, y, z) = x 3 y 3 z3 x=L
∗ 29. If V (x, y, z) represents the electrostatic potential at a point

(x, y, z) due to a system of n point charges at points (xi , yi , zi ) , ∗ 33. When the bar in Figure 12.13 is subjected to a force per unit area
does V (x, y, z) satisfy Laplace’s equation? of magnitude F on its right end, and no other forces act on the bar,
equations 12.14 for static deflections become
∗ 30. The gravitational potential at a point (x, y, z) in space due to a
uniform spherical mass distribution (mass M ) at the !origin is defined
d 2y
as V = GM/r , where G is a constant and r = x 2 + y 2 + z2 . 0 = E , 0 < x < L,
Show that V (x, y, z) satisfies Laplace’s equation 12.12. dx 2
∗ 31. The figure below shows a plate bounded by the lines x = 0,
y = 0, x = 1, and y = 1. Temperature along the first three sides y(0) = 0, y + (L) = F /E.
is kept at 0◦ C, while that along y = 1 varies according to f (x) =
sin(3πx) − 2 sin(4π x) , 0 ≤ x ≤ 1. The temperature at any point Find displacements for cross-sections, and the length of the bar.
interior to the plate is then
∗ 34. (a) Show that when F (x, t) ≡ 0, wave equation 12.13a is
3πy −3πy
T (x, y) = C(e −e ) sin(3π x) satisfied by functions of the form
+ D(e4πy − e−4πy ) sin(4π x), y(x, t) = (A sin λx + B cos λx)(C sin cλt + D cos cλt),
where C = (e3π − e−3π )−1 and D = −2(e4π − e−4π )−1 . Show that c2 = E/ρ,
T (x, y) is harmonic in the region 0 < x < 1, 0 < y < 1, and that
it also satisfies the boundary conditions T (0, y) = 0, T (1, y) = 0,
where λ , A , B , C , and D are arbitrary constants.
T (x, 0) = 0, and T (x, 1) = f (x) .
(b) Show that boundary conditions 12.13d and e require that
y B = 0 and λ = (2n − 1)π/(2L) , where n is an integer.
T = f (x)
1 (c) If g(x) is equal to zero [but not f (x) ], what is C ?
(d) If f (x) is equal to zero [but not g(x) ], what is D ?
0•C 0•C
∗ 35. If the ends of the taut string in the figure below are fastened at x = 0
and x = L on the x -axis, and if the string vibrates in the y -direction
only, small displacements y(x, t) must satisfy the one-dimensional
0•C 1 x wave equation
∗ 38. A taut string has its ends fixed at x = 0 and x = L on the x -axis.
2 2 At time t = 0, the string is moved so as to take the shape of the sine
∂ y ∂ y F (x, t)
= c2 2 + , 0 < x < L, t > 0, curve f (x) = 3 sin(π x/L) , and then released. If the only force taken
∂t 2 ∂x ρ into account as acting on the string is its tension, then displacements of
the string must satisfy
where c2 = T /ρ , T is the (constant) tension of the string, and ρ is
the (constant) mass per unit length of the string. Function F (x, t) is ∂ 2y ∂ 2y
the result of all forces per unit length acting at position x and time 2
= c2 2 , 0 < x < L, t > 0,
∂t ∂x
t in the string (except for tension in the string). Accompanying this
equation will be initial conditions 12.13b and c specifying the position y(0, t) = y(L, t) = 0, t > 0,
and velocity of the string at time t = 0, and the boundary conditions
y(x, 0) = f (x) = 3 sin(π x/L), 0 < x < L,
y(0, t) = 0, y(L, t) = 0, t > 0,
yt (x, 0) = 0, 0 < x < L.
since both ends of the string are attached to the x -axis. (a) Show that y(x, t) = 3 sin(π x/L) cos(π ct/L) satisfies
(a) Show that when F (x, t) ≡ 0, the wave equation is satisfied this problem.
by a function of the form (b) Plot the surface y(x, t), interpreting cross-sections physi-
cally. (Use L = 1 and c = 2.)
y(x, t) = (A sin λx + B cos λx)(C sin cλt + D cos cλt), (c) Plot y(x, t) as a function of x for t = 0, L/(8c) , L/(4c) ,
3L/(8c) , and L/(2c) .
where λ , A , B , C , and D are arbitrary constants.
∗ 39. Show that when f (x) = 3 sin(π x/L) − 2 sin(2π x/L) , in Exer-
(b) Show that the boundary conditions require B = 0 and cise 38, the problem is satisfied by y(x, t) = 3 sin(π x/L) cos(π ct/L)
λ = nπ/L , where n is an integer. − 2 sin(2π x/L) cos(2π ct/L) . Plot y(x, t) as a function of x for
t = 0, L/(8c) , L/(4c) , 3L/(8c) , and L/(2c) .
(c) If g(x) is equal to zero [but not f (x) ], what is C ?
(d) If f (x) is equal to zero [but not g(x) ], what is D ?
∗ 40. (a) If the length of the string in Exercise 38 is L = 1 m, and
4
x/100, 0 ≤ x ≤ 1/2,
y f (x) =
(1 − x)/100, 1/2 ≤ x ≤ 1,
displacements can be expressed in the form of an infinite

y (x, t) series
x L x 5∞
1 (−1)n+1
y(x, t) = sin(2n − 1)π x cos(2n − 1)π ct.
∗ 36. Static deflections of the string in Exercise 35 occur when F (x, t) 25π 2 n=1 (2n − 1)2
is only a function of x , and initial conditions are ignored. The problem
then is Show that this function satisfies the partial differential equa-
d 2y tion, boundary conditions, and the second initial condition
0 = T + F (x), 0 < x < L, in Exercise 38. (Assume that differentiation and summation
dx 2 operations can be interchanged.)
y(0) = 0, y(L) = 0. (b) Suppose the first 20 terms of the series are used to approx-
imate y(x, t) . Plot the sum of these terms as a function of
When the only force acting on the string (other than internal tension) x for the times t = 0, 1/(16c) , 1/(8c) , 3/(16c) , 1/(4c) ,
is gravity, then F (x) = −9.81ρ . Show that the solution for y(x) is 5/(16c) , 3/(8c) , 7/(16c) , and 1/(2c) . Does the string
a parabola. What underlying assumption(s) make this problem and its appear to retain its broken-line shape?
solution different from Example 3.39 in Section 3.13? ∗ 41. The uniform circular rod in the figure below has flat ends at x = 0
and x = L . If the round side of the rod is perfectly insulated, heat
∗ 37. A string of length L is stretched tightly along the x -axis. Its right flows in only the x -direction. When no heat sources exist within the
end is fixed on the x -axis, and its left end at x = 0 is looped around the rod, temperature T (x, t) at points in the rod must satisfy the one-
y -axis and is free to move thereon without friction. The string is slowly dimensional heat conduction equation
lowered under the influence of gravity to take up a static position. The
shape of the string is defined by the differential equation ∂T ∂ 2T
= k 2, 0 < x < L, t > 0,
2
∂t ∂x
d y
0 = T − 9.81ρ, 0 < x < L,
dx 2
y + (0) = 0, y(L) = 0.
x=0 L x
Solve this problem. Perfect insulation
where k > 0 is a constant called the thermal diffusivity of the rod. Ac- ∗ 43. If the ends x = 0 and x = L of the rod in Exercise 41 are held
companying this equation will be two boundary conditions describing at temperatures T0◦ C and TL◦ C for t > 0, temperature T (x, t) must
what is happening at the ends of the rod. For instance, if the left end satisfy
is kept at temperature 0◦ C and the right end is insulated, T (x, t) must
satisfy ∂T ∂ 2T
T (0, t) = 0, Tx (L, t) = 0, t > 0. = k 2, 0 < x < L, t > 0,
∂t ∂x
In addition, there will be an initial condition describing the temperature
T (0, t) = T0 , T (L, t) = TL , t > 0,
distribution in the rod at time t = 0,
T (x, 0) = f (x), 0 < x < L.
T (x, 0) = f (x), 0 < x < L.
(a) Show that the heat conduction equation is satisfied by func- After a very long time, the temperature at each point in the rod will
tions of the form remain constant, but temperature will vary from point to point. Tem-
2t
perature is said to have reached a steady-state situation. It can be found
T (x, t) = (A sin λx + B cos λx)e−kλ by removing the term ∂T /∂t from the heat conduction equation, and
the initial condition. Temperature becomes a function of only x , which
where λ , A , and B are arbitrary constants. satisfies
(b) Show that the boundary conditions require B = 0 and d 2T
λ = (2n − 1)π/(2L) , where n is an integer. 0 = , 0 < x < L,
dx 2
(c) When the initial temperature of the rod is f (x) = x , so
that it increases linearly from 0◦ C at its left end to L◦ C at
T (0) = T0 , T (L) = TL .
its right end, temperature thereafter is
∞ Find T (x) .
8L 5 (−1)n+1 2 π 2 kt/(4L2 ) (2n − 1)π x
T (x, t) = e−(2n−1) sin .
π2 n=1
(2n − 1)2 2L ∗ 44. If heat is added to the rod in Exercise 41, the heat conduction
equation takes the form
Show that this function satisfies the partial differential equa-
tion and boundary conditions. (Assume that differentiation ∂T ∂ 2T
and summation operations can be interchanged.) = k 2 + F (x, t), 0 < x < L, t > 0,
∂t ∂x
(d) Plot the sum of the first 20 terms of the series as a function of
x for the times t = 0, 10, 100, 1000, and 10 000. Use a 1-m where F (x, t) is the heat source term. If temperatures of the ends of
rod and k = 1.14 × 10−4 m 2 /s (typical for copper). Does the rod are held at T0◦ C and TL◦ C, then T (x, t) must also satisfy the
the t = 0 plot approximate f (x) , and do the remaining boundary conditions
plots reflect the boundary conditions?
∗ 42. If both ends of the rod in Exercise 41 are held at temperature 0◦ C, T (0, t) = T0 , T (L, t) = TL , t > 0.
the problem for T (x, t) becomes
If F (x, t) is independent of t , then the steady-state temperature of the
∂T ∂ 2T rod (see Exercise 43) must satisfy
= k 2, 0 < x < L, t > 0,
∂t ∂x
T (0, t) = T (L, t) = 0, t > 0, d 2T
0 = k + F (x), 0 < x < L,
dx 2
T (x, 0) = f (x), 0 < x < L.
(a) Show that for the functions in part (a) of Exercise 41, the
T (0) = T0 , T (L) = TL .
boundary conditions require that B = 0 and λ = nπ/L ,
Find T (x) in the case that F (x) is a constant value F .
where n is an integer.
(b) When the initial temperature of the rod is f (x) = x(L − ∗ 45. The figure below shows a bar of length L and uniform cross-
x) , temperature thereafter is section. When torque is applied to cross-sections, these cross-sections
∞
8L4 5 1 2 π 2 kt/L2 (2n − 1)π x Torque-free position
T (x, t) = e−(2n−1) sin .
π3 n=1
(2n − 1)3 L
Plot the sum of the first 20 terms of this series as a function

of x for the times t = 0, 10, 100, 1000, and 10 000. Use a y
1-m rod and k = 1.14 × 10−4 m 2 /s. Does the t = 0 plot x=0
approximate f (x) ? x L x
are forced to rotate. Let x represent distance from the left end of the bar, ∗ 46. Show that when τ (x) is unspecified, static rotations in Exercise
and y(x, t) the angular displacement of the cross-section at position 45 can be expressed in the form
x and time t from its torque-free position. It can be shown that when
τ (x, t) represents the torque per unit length at position x and time 6 x 6 v
1
t , and ρ is the (constant) density of the bar, y(x, t) must satisfy the y(x) = Cx − τ (u) du dv,
E 0 0
one-dimensional wave equation
∂ 2y 2 where
2∂ y τ (x, t) 6
2
= c 2
+ , 0 < x < L, t > 0, 1 L
∂t ∂x ρ C = τ (x) dx.
E 0
where c2 = E/ρ and E is Young’s modulus of elasticity of the ma-
terial in the bar under shear. Accompanying the partial differential ∗ 47. Two functions u(x, y) and v(x, y) are said to be harmonic con-
equation will be initial conditions 12.13b and c specifying the initial jugates if they satisfy the Cauchy–Riemann equations of Exercise 41
displacements and velocities of cross-sections at time t = 0. In addi- in Section 12.3. Show that if u(x, y) and v(x, y) are harmonic con-
tion, there will be two boundary conditions describing end conditions. jugates and have continuous second partial derivatives in a region R ,
For example, if the left end is clamped so that no rotation is possi- then each is harmonic in R .
ble, and the right end is free to move, the boundary conditions will be
12.13d and e. The problem is therefore identical to that for longitudinal ∗ 48. (a) Show that the function u(x, y) = x 2 − y 2 is harmonic in
vibrations of bars. Static rotations occur when τ (x, t) is independent the entire xy -plane.
of t , the time derivative is removed from the wave equation, and initial
conditions are deleted. Displacements y(x) must then satisfy (b) Use the Cauchy–Riemann equations in Exercise 41 of Sec-
tion 12.3 to find a function v(x, y) so that u and v are
d 2y harmonic conjugates.
0 = E + τ (x), 0 < x < L,
dx 2
∗ 49. Repeat Exercise 48 if u(x, y) = ex cos y + x .
y(0) = 0, y + (L) = 0,
at least for the boundary conditions described above. Solve this problem ∗ 50. For what values of n does the function (x 2 + y 2 + z2 )n satisfy
when τ (x) = x so that torque increases linearly along the bar. equation 12.12? In what regions are the functions harmonic?
12.6 Chain Rules for Partial Derivatives

If y = f (u) and u = g(x) , the chain rule for the derivative dy/dx of the composite function
f [g(x)] is
dy dy du
= . (12.15)
dx du dx
Equation 12.15 can be extended in terms of more intermediate variables, say y = f (u) ,
u = g(s) , s = h(x) , in which case
dy dy du ds
= . (12.16)
dx du ds dx
For multivariable functions, variations in chain rules are countless. We discuss two examples
in considerable detail, and then show schematic diagrams that easily lead to chain rules for even
the most complicated functional situations.
Suppose z is a function of u and v and each of u and v is a function of x and y ,
z = f (u, v), u = g(x, y), v = h(x, y). (12.17)
By the substitutions
z = f [g(x, y), h(x, y)], (12.18)
we express z as a function of x and y , and can then calculate the partial derivative ∂z/∂x .
However, if the functions in 12.17 are at all complicated, you can imagine how difficult the
composite function in 12.18 might be to differentiate. As a result, we search for an alternative
procedure for calculating ∂z/∂x , namely, the appropriate chain rule. It is contained in the
following theorem.
THEOREM 12.2
Let u = g(x, y) and v = h(x, y) be continuous and have first partial derivatives with
respect to x at a point (x, y) , and let z = f (u, v) have continuous first partial derivatives
inside a circle centred at the point (u, v) = (g(x, y) , h(x, y)) . Then
∂z ∂z ∂u ∂z ∂v
= + . (12.19)
∂x ∂u ∂x ∂v ∂x
PROOF This result can be proved in much the same way as chain rule 3.20a was proved in
Section 3.7. By Definition 12.3,
∂z f [g(x + 'x, y), h(x + 'x, y)] − f [g(x, y), h(x, y)]
= lim .
∂x 'x→0 'x
Now the increment 'x in x produces changes in u and v , which we denote by
'u = g(x + 'x, y) − g(x, y), 'v = h(x + 'x, y) − h(x, y).
If we write u and v whenever g(x, y) and h(x, y) are evaluated at (x, y) , and substitute for
g(x + 'x, y) and h(x + 'x, y) in the definition for ∂z/∂x , then
∂z f (u + 'u, v + 'v) − f (u, v)
= lim
∂x 'x→0 'x
[f (u + 'u, v + 'v) − f (u, v + 'v)] + [f (u, v + 'v) − f (u, v)]
= lim
'x→0 'x
) +
f (u + 'u, v + 'v) − f (u, v + 'v) f (u, v + 'v) − f (u, v)
= lim + .
'x→0 'x 'x
We assumed that the derivative
∂z f (u, v + 'v) − f (u, v)
= lim
∂v 'v→0 'v
exists at (u, v) . An equivalent way to express the fact that this limit exists is to say that
f (u, v + 'v) − f (u, v) ∂z

= + $1 ,
'v ∂v
where $1 must satisfy the condition that lim'v→0 $1 = 0. We can write, therefore, that
f (u, v + 'v) − f (u, v) = [zv (u, v) + $1 ] 'v.
Similarly, we can write that
f (u + 'u, v + 'v) − f (u, v + 'v) = [zu (u, v + 'v) + $2 ] 'u,
where lim'u→0 $2 = 0 (provided that 'v is sufficiently small). When these expressions are
substituted into the limit for ∂z/∂x , we have
4 7
∂z 'u 'v
= lim [zu (u, v + 'v) + $2 ] + [zv (u, v) + $1 ] .
∂x 'x→0 'x 'x
We now examine each part of this limit. Clearly,
'u ∂u 'v ∂v
lim = and lim = .
'x→0 'x ∂x 'x→0 'x ∂x
In addition, because g(x, y) and h(x, y) are continuous, 'u → 0 and 'v → 0 as 'x → 0.
Consequently,
lim $1 = lim $1 = 0 and lim $2 = lim $2 = 0.

'x→0 'v→0 'x→0 'u→0
Finally, because ∂z/∂u is continuous,
lim zu (u, v + 'v) = lim zu (u, v + 'v) = zu (u, v).

'x→0 'v→0
When all these results are taken into account, we have
∂z ∂u ∂v ∂z ∂u ∂z ∂v
= zu (u, v) + zv (u, v) = + ,
∂x ∂x ∂x ∂u ∂x ∂v ∂x
which completes the proof.
Chain rule 12.19 defines ∂z/∂x in terms of derivatives of the given functions in 12.17. We
could be more explicit by indicating which variable is being held constant in each of the five
derivatives: ( ( ( ( (
∂z ∂z ∂u ∂z ∂v
= + . (12.20)
∂x y ∂u v ∂x y ∂v u ∂x y
From the point of view of rates of change, this result seems quite reasonable. The left side is
the rate of change of z with respect to x holding y constant. The first term ( ∂z/∂u )( ∂u/∂x )
accounts for the rate of change of z with respect to those x ’s that affect z through u . The second
term, ( ∂z/∂v )( ∂v/∂x ), accounts for the rate of change of z with respect to those x ’s that affect
z through v . The total rate of change is then the sum of the two parts.
Consider now the functional situation
z = f (u, v), u = g(x, y, s), v = h(x, y, s), x = p(t), y = q(t), s = r(t).

(12.21)
By the substitutions
z = f [g(p(t), q(t), r(t)), h(p(t), q(t), r(t))], (12.22)
we express z as a function of t alone, and can therefore pose the problem of calculating dz/dt .
If we reason as in the preceding paragraph, the appropriate chain rule for dz/dt must account
for all t ’s affecting z through u and v . We obtain, then,
dz ∂z du ∂z dv
= + ,
dt ∂u dt ∂v dt
where we have written du/dt and dv/dt because u and v can be expressed entirely in terms
of t :
u = g [p(t), q(t), r(t)], v = h[p(t), q(t), r(t)].
Chain rules for each of du/dt and dv/dt (similar to 12.19) yield
du ∂u dx ∂u dy ∂u ds dv ∂v dx ∂v dy ∂v ds
= + + , = + + .
dt ∂x dt ∂y dt ∂s dt dt ∂x dt ∂y dt ∂s dt
Finally, then,
' ( ' (
dz ∂z ∂u dx ∂u dy ∂u ds ∂z ∂v dx ∂v dy ∂v ds
= + + + + + , (12.23)
dt ∂u ∂x dt ∂y dt ∂s dt ∂v ∂x dt ∂y dt ∂s dt
which expresses dz/dt in terms of derivatives of the given functions in 12.21.

These two examples suggest the complexities that may be involved in finding chain rules for
complicated composite functions. Fortunately, there is an amazingly simple method that gives
the correct chain rule in every situation. The method is not designed to help you understand the
chain rule, but to find it quickly. We suggest that you test your understanding by developing a
few chain rules in the exercises with a discussion such as in the second example above, and then
check your result by the quicker method.
In the first example we represent the functional situation described in 12.17 by the schematic
z diagram to the left. At the top of the diagram is the dependent variable z , which we wish to
differentiate. In the line below z are the variables u and v in terms of which z is initially
defined. In the line below u and v are x ’s and y ’s illustrating that each of u and v is defined
u v in terms of x and y .
Here are the rules to obtain the partial derivative from schematic diagrams, in general,
x y x y
followed by ∂z/∂x for the specific example:
1. Take all possible paths in the schematic from the differentiated variable to the differentiating
variable.
2. For each straight-line segment in a given path, differentiate the upper variable with respect
to the lower variable and multiply together all such derivatives in that path.
3. Add the products together to form the complete chain rule.
To calculate ∂z/∂x from the schematic to the left, we note that there are two paths from z to
x , one through u and one through v . For the path through u we form the product
∂z ∂u
,
∂u ∂x
and for the path through v ,
∂z ∂v
.
∂v ∂x
The complete chain rule is then the sum of these products,
∂z ∂z ∂u ∂z ∂v
= + ,
∂x ∂u ∂x ∂v ∂x
and this result agrees with 12.19. The schematic diagram also indicates which variables are to
be held constant in the derivatives on the right (as in 12.20). All other variables on the same
z level are held constant.
For the second example in equations 12.21 the schematic diagram is to the left. There are
six possible paths from z to t , so that the chain rule for dz/dt must have six terms. We find
u v
dz ∂z ∂u dx ∂z ∂u dy ∂z ∂u ds ∂z ∂v dx ∂z ∂v dy ∂z ∂v ds
x y s x y s
= + + + + + ,
dt ∂u ∂x dt ∂u ∂y dt ∂u ∂s dt ∂v ∂x dt ∂v ∂y dt ∂v ∂s dt
and this agrees with 12.23. Note too that if when forming a derivative from the schematic dia-
t t t t t t gram, there are two or more lines emanating from a variable, then we obtain a partial derivative;
if there is only one line, then we have an ordinary derivative.
EXAMPLE 12.14
Find chain rules for ( (
∂z ∂z
and
∂x y ∂y x
if
z = f (r, s, x), r = g(x, y), s = h(x, y).
z SOLUTION From the schematic diagram to the left,

( ( ( ( (
r s x
∂z ∂z ∂r ∂z ∂s
= + ,
∂y x ∂r
s,x ∂y x ∂s r,x ∂y x
x y x y ( ( ( ( ( (
∂z ∂z ∂r ∂z ∂s ∂z
= + + .
∂x y ∂r s,x ∂x y ∂s r,x ∂x y ∂x r,s
In Example 12.14 it is essential that we indicate which variables to hold constant in the partial
derivatives. If we were to omit these designations, then in the second result we would have a
term ∂z/∂x on both sides of the equation but they would have different meanings. The term
∂z/∂x)y indicates the derivative of z with respect to x holding y constant if z were expressed
entirely in terms of x and y ; the term ∂z/∂x)r,s indicates the derivative of the given function
f (r, s, x) with respect to x holding r and s constant.
EXAMPLE 12.15
Find dz/dt if
1
z = x 3 y 2 + x sin y + tx, x = 2t + , y = t 2 et .
t
x y t dz ∂z dx ∂z dy ∂z
= + +
dt ∂x dt ∂y dt ∂t
t t = (3x 2 y 2 + sin y + t)(2 − 1/t 2 ) + (2x 3 y + x cos y)(2tet + t 2 et ) + x.
When a chain rule is used to calculate a derivative, the result usually involves all intermediate
variables. For instance, the derivative dz/dt in Example 12.15 involves not only t , but the
intermediate variables x and y as well. Were dz/dt required at t = 1, values of x and y for
t = 1 would be calculated — x(1) = 3 and y(1) = e — and all three values substituted to
obtain
dz //
= [3(3)2 (e)2 + sin(e) + 1](2 − 1)
dt /t=1
+ [2(3)3 e + (3)cos(e)](2e + e) + 3
= 1378.6.
EXAMPLE 12.16
Find ∂ 2 z/∂x 2 if
y
z = s 2 t + 2 sin t, s = xy − y, t = x2 + .
x
∂z ∂z ∂s ∂z ∂t
= + = (2st)(y) + (s 2 + 2 cos t)(2x − y/x 2 ).
s t ∂x ∂s ∂x ∂t ∂x
x y x y Now ∂z/∂x is a function of s , t , x , and y , and therefore in order to find

' (
∂ 2z ∂ ∂z
= ,
∂x 2 ∂x ∂x
we form a schematic diagram for ∂z/∂x . From this schematic diagram, we obtain
∂z
∂x ' ( ' ( ' (
∂ 2z ∂ ∂z ∂s ∂ ∂z ∂t ∂ ∂z
= + +
∂x 2 ∂s ∂x ∂x ∂t ∂x ∂x ∂x ∂x s,t,y
s t x y
= [2ty + 2s(2x − y/x 2 )](y) + [2sy − 2 sin t (2x − y/x 2 )](2x − y/x 2 )
x y x y + (s 2 + 2 cos t)(2 + 2y/x 3 ).
EXAMPLE 12.17
Temperature T at points in the atmosphere depends on both position (x, y, z) and time t :
T = T (x, y, z, t) . When a weather balloon is released to take temperature readings, it is not
free to take readings at just any point, only at those points along the path that the winds force
the balloon to follow. This path is a curve in space represented parametrically by
C : x = x(t), y = y(t), z = z(t), t ≥ 0,
t again being time. If we substitute from the equations for C into the temperature function,
then T becomes a function of t alone,
T = T [x(t), y(t), z(t), t ],

T
and this function of time describes the temperature at points along the path of the balloon. For
the derivative of this function with respect to t , the schematic diagram yields
x y z t
dT ∂T dx ∂T dy ∂T dz ∂T
= + + + .
dt ∂x dt ∂y dt ∂z dt ∂t
t t t
The question we pose is: What is the physical difference between dT /dt and ∂T /∂t ?
SOLUTION Temperature at a point in space is independent of the observer measuring it;
hence T [x(t), y(t), z(t), t ] is the temperature at points on C as measured by both the balloon
and any observer fixed in the xyz -reference system. If, however, these two observers calculate
the rate of change of temperature with respect to time at some point (x, y, z) on C , they
calculate different results. The observer fixed in the xyz -reference system (not restricted to
move along C ) calculates the rate of change of T with respect to t as the derivative of the
function T (x, y, z, t) partially with respect to t holding x , y , and z constant (i.e., the fixed
observer calculates ∂T /∂t as the rate of change of temperature in time). The balloon, on the
other hand, has no alternative but to take temperature readings as it moves along C ; thus its
measurement of T as a function of t is
T [x(t), y(t), z(t), t ].

Therefore, when the balloon calculates the time variation of temperature, it is calculating dT /dt .
It follows, then, that the terms
∂T dx ∂T dy ∂T dz
+ +
∂x dt ∂y dt ∂z dt
describe that part of dT /dt caused by the motion of the balloon through space.
Discussions like those in Example 12.17 are prominent in the study of fluid motion (gas or
liquid). Sometimes rates of change from the point of view of a fixed observer are important;
other times, rates of change as measured by an observer moving with the fluid are appropriate.
Many important applications of the chain rule occur in the field of partial differential equa-
tions. The following example is an illustration.
EXAMPLE 12.18
The one-dimensional wave equation
∂ 2y 2
2∂ y
= c , c = constant
∂t 2 ∂x 2
for functions y(x, t) describes transverse vibrations of taut strings, and longitudinal and rota-
tional vibrations of metal bars. Show that if f (u) and g(v) are twice-differentiable functions
of u and v , then y(x, t) = f (x + ct) + g(x − ct) satisfies the wave equation.
y SOLUTION The schematic diagram to the left describes the functional situation
y = f (u) + g(v)
u v
where u = x + ct and v = x − ct . The chain rule for ∂y/∂t is
x t x t
∂y ∂y ∂u ∂y ∂v
= + = cf + (u) − cg + (v).
∂t ∂u ∂t ∂v ∂t
The schematic diagram for ∂y/∂t leads to

∂y
∂t ' ( ' (
∂ 2y ∂ ∂y ∂u ∂ ∂y ∂v
= +
∂t 2 ∂u ∂t ∂t ∂v ∂t ∂t
u v
= [cf ++ (u)]c + [−cg ++ (v)](−c)
x t x t
= c2 [f ++ (u) + g ++ (v)].
∂ 2y
A similar calculation gives = f ++ (u) + g ++ (v) . Hence y(x, t) does indeed satisfy the
∂x 2
wave equation.
We have suggested how important Laplace’s equation is to engineering, particularly in elec-

trostatics, heat conduction, fluid flow, and deflection of plates. The two-dimensional Laplace
equation in Cartesian coordinates is 12.11. In the next example, we find its form in polar
coordinates.
EXAMPLE 12.19
Find Laplace’s equation in polar coordinates.
SOLUTION Cartesian coordinates are related to polar coordinates by the equations x =
r cos θ , y = r sin θ . The inverse transformation is
! 2y 3
r = x2 + y2, θ = Tan−1 .
x
The second of these is not always correct; it may need ±π added to it. Since the derivation
V here requires only derivatives of θ , and not θ itself, the ±π is inconsequential. Suppose
V = f (x, y) is a function that satisfies Laplace’s equation ∂ 2 V /∂x 2 + ∂ 2 V /∂y 2 = 0. We
can express V in terms of r and θ by writing V = F (r, θ ) = f (r cos θ, r sin θ ) . The
r
schematic to the left represents the functional situation where V = F (r, θ ) and r and θ are
x y x y expressed in terms of x and y . From it,
∂V ∂V ∂r ∂V ∂θ
= + ,
∂x ∂r ∂x ∂θ ∂x
where
∂r x r cos θ
= ! = = cos θ,
∂x x2 + y2 r
' (
∂θ 1 −y −y − sin θ
= 2 2
= 2 2
= .
∂x 1 + (y/x) x x +y r
Thus,
∂V ∂V sin θ ∂V
= cos θ − .
∂x ∂r r ∂θ
The combination of r and θ derivatives on the right of this equation must be applied to V when
∂V it is expressed in terms of r and θ to give the partial derivative with respect to x when V is
∂x
expressed in terms of x and y . To find ∂ 2 V /∂x 2 , we use the schematic to the left.
' ( ' (
r ∂ 2V ∂ ∂V ∂r ∂ ∂V ∂θ
= +
∂x 2 ∂r ∂x ∂x ∂θ ∂x ∂x
x y x y ' (
2
∂ V sin θ ∂V sin θ ∂ 2 V
= cos θ 2 + 2 − cos θ
∂r r ∂θ r ∂r ∂θ
' (' (
∂V ∂ 2V cos θ ∂V sin θ ∂ 2 V sin θ
+ − sin θ + cos θ − − −
∂r ∂θ ∂r r ∂θ r ∂θ 2 r
∂ 2V 2 sin θ cos θ ∂ 2 V sin2 θ ∂ 2 V
= cos2 θ 2
− +
∂r r ∂r ∂θ r 2 ∂θ 2
2 sin θ cos θ ∂V sin2 θ ∂V
+ + ,
r2 ∂θ r ∂r
where we have assumed that mixed partial derivatives are equal. A similar calculation gives
∂ 2V 2
2 ∂ V 2 sin θ cos θ ∂ 2 V cos2 θ ∂ 2 V 2 sin θ cos θ ∂V cos2 θ ∂V
= sin θ + + − + .
∂y 2 ∂r 2 r ∂r∂θ r 2 ∂θ 2 r2 ∂θ r ∂r
When these are added together, the result is Laplace’s equation in polar coordinates,
∂ 2V 1 ∂V 1 ∂ 2V
+ + = 0. (12.24)
∂r 2 r ∂r r 2 ∂θ 2
Calculations in this example allow us to emphasize a point that we made in Section 12.3. We
said that although the derivative dy/dx can be considered as a quotient of differentials, we
never consider a partial derivative as a quotient. To do so in Example 12.19 would lead to errors.
We calculated ∂r/∂x = cos θ . Notice that ∂x/∂r = cos θ , and therefore to regard ∂r/∂x as
the reciprocal of ∂x/∂r is incorrect.
Homogeneous Functions
Homogeneous functions arise in numerous areas of applied mathematics. A function f (x, y, z)
is said to be a positively homogeneous function of degree n if for every t > 0,
f (tx, ty, tz) = t n f (x, y, z). (12.25)
For example, the function f (x, y, z) = x 2 + y 2 + z2 is homogeneous of degree 2; the func-

tion f (x, y) = x 3 cos(y/x) + x 2 y + xy 2 is homogeneous of degree 3; and f (x, y, z, t) =
√
x 2 + z2 (x 2 y + yt 2 ) is homogeneous of degree 4. Partial derivatives of homogeneous func-
tions satisfy many identities. In particular, their first derivatives satisfy Euler’s theorem.
THEOREM 12.3 (Euler’s Theorem)

If f (x, y, z) is positively homogeneous of degree n , and has continuous first partial
derivatives, then
∂f ∂f ∂f
x +y +z = nf (x, y, z). (12.26)
∂x ∂y ∂z
f PROOF To verify 12.26, we differentiate 12.25 with respect to t , holding x, y, and z constant.
For the derivative of the left side we introduce variables u = tx , v = ty , and w = tz , and use
the schematic to the left. The result is
u v w
∂f ∂u ∂f ∂v ∂f ∂w
+ + = nt n−1 f (x, y, z)
t x t y t z ∂u ∂t ∂v ∂t ∂w ∂t
or
∂f ∂f ∂f
+y
x +z = nt n−1 f (x, y, z).
∂u ∂v ∂w
When we set t = 1, we obtain u = x , v = y , w = z , and the equation above becomes 12.26.
The results of Exercises 31–33 in Section 12.3 are special cases of 12.26.
EXERCISES 12.6
!
In Exercises 1–10 we have defined a general functional situation and u= x 2 + y 2 + z2 , x = 2st , y = s 2 + t 2 , z = st
a specific example. Find the chain rule for the indicated derivative
in the general situation, and then use that result to calculate the same 4. dz/du if z = f (x, y, v) , x = g(u) , y = h(u) , v = k(u) ;
derivative in the specific example. z = x 2 yv 3 , x = u3 + 2u , y = ln (u2 + 1) , v = ueu
(
1. dz/dt if z = f (x, t), x = g(t) ; z = xt 2 /(x + t), x = e3t ∂u
( 5. if u = f (x, y, s) , x = g(t) , y = h(r) , s = k(r, t) ;
∂z ∂r! t
2. if z = f (x, y) , x = g(s, t) , y = h(s, t) ; z = x 2 ey + u = x 2 + y 2 s , x = t/(t + 5) , y = Sin −1 (r 2 + 5) , s = tan (rt)
∂t s
y ln x , x = s 2 cos t , y = 4 Sec −1 (t 2 + 2s) (
( ∂z
∂u 6. if z = f (x) , x = g(y) , y = h(r, t) ; z = 3x+2 , x =
3. if u = f (x, y, z) , x = g(s, t) , y = h(s, t) , z = k(s, t) ; ∂t r
∂s t y 2 + 5, y = csc (r 2 + t)
(
∂u !
7. if u = f (x, y, z), z = g(x, y) ; u = y/ x 2 + y 2 + z2 , ∗ 18. The radius and height of a right-circular cone are 10 and 20 cm,
∂x y respectively. If the radius is increasing at 1 cm/min and the height is
z = x/y decreasing at 2 cm/min, how fast is the volume changing? Do you need
( multivariable calculus to solve this problem?
∂x
8. if x = f (r, s, t) , r = g(y) , s = h(y, z) , t = k(y, z) ;
∂y z ∗ 19. If two sides of a triangle have lengths x and y and the angle
x = s 2 r 2 t 2 , r = y −5 , s = 1/(y 2 + z2 ) , t = 1/y 2 + 1/z2 between them is θ , then the area of the triangle is A = (1/2)xy sin θ .
( How fast is the area changing when x is 1 m, y is 2 m, and θ is 1/3
∂z radian, if x and y are each increasing at 1/2 m/s and θ is decreasing
9. if z = f (x, y) , x = g(r) , y = h(r) , r = k(s, t) ;
∂t s at 1/10 radian per second?
z = ex+y , x = 2r + 5, y = 2r − 5, r = t ln (s 2 + t 2 )
∗ 20. When a rocket rises from the earth’s surface, its mass decreases
10. dz/dt if z = f (x, y, u) , x = g(v) , u = h(x, y) , v = k(t) , because fuel is being consumed at the rate of 50 kg/s. Use Newton’s
y = p(t) ; z = x 2 + y 2 + u2 , x = v 3 − 3v 2 , u = 1/(x 2 − y 2 ) , universal law of gravitation (see Example 7.34 in Section 7.10) to deter-
v = e t , y = e 4t mine how fast the force of gravity of the earth on the rocket is changing
when the rocket is 100 km above the earth’s surface and climbing at
2 km/s. Assume that the mass of the rocket at this height is 12 × 106
In Exercises 11–15 find the derivative. kg.
(
∂ 2z
∗ 11. if z = x 2 y 2 + xey , x = s + t 2 , y = s − t 2 ∗ 21. If z = f (u, v), u = g(x, y), v = h(x, y) , find the chain rule
∂t 2 s
for the second derivative ∂ 2 z/∂x 2 .
∗ 12. d 2 x/dt 2 if x = y 2 + yt − t 2 , y = t 2 et
( ∗ 22. Determine which of the following functions are positively homo-
∂ 2u geneous:
∗ 13. if u = x 2 + y 2 + z2 + xyz, x = s 2 + t 2 , y = s 2 − t 2 ,
∂s 2 t (a) f (x, y) = x 2 + xy + 3y 2
z = st
(b) f (x, y) = x 2 y + xy − 2xy 2
∗ 14. d 2 z/dv 2 if z = sin (xy), x = 3 cos v, y = 4 sin v
(c) f (x, y, z) = x 2 sin (y/z) + y 2 + y 3 /z
!
∗ 15. ∂ 2 u/∂x ∂y if u = y/ x 2 + y 2 + z2 , z = x/y (d) f (x, y, z) = xey/z − xyz
16. Suppose that u is a differentiable function of r and r =

∗ ! (e) f (x, y, z, t) = x 4 + y 4 + z4 + t 4 − xyzt
x 2 + y 2 + z2 . Show that (f) f (x, y, z, t) = ex
2 +y 2
(z2 + t 2 )
' (2 ' (2 ' (2 ' (2
∂u ∂u ∂u du (g) f (x, y, z) = cos (xy) sin (yz)
+ + = . !
∂x ∂y ∂z dr (h) f (x, y) = x 2 + xy + y 2 ey/x (2x 2 − 3y 2 )
∗ 17. Consider a gas that is moving through some region D of space. If
we follow a particular particle of the gas, it traces out some curved path ∗ 23. (a) Suppose that the circular plate with radius R in the figure
(figure below) below has its lower edge held at 0 V and its upper edge held
at 1 V. Show that the electrostatic potential
C : x = x(t), y = y(t), z = z(t), t ≥ 0. ' (
1 1 2Rr sin θ
Suppose the density of the gas at any point in the region D at V (r, θ) = + Tan−1
2 π R2 − r 2
time t is denoted by ρ(x, y, z, t) . We can write that along C ,
ρ = ρ [x(t), y(t), z(t), t ].
satisfies Laplace’s equation 12.24 for r < R .
z
y
(x, y, z)
C 1 1
R x
y
x
(a) Obtain the chain rule defining dρ/dt in terms of ∂ρ/∂t 0 0
and derivatives of x , y , and z with respect to t .
(b) Explain the physical difference between dρ/dt and ∂ρ/∂t .
' (2 ' (2
(b) The function in part (a) is not defined for r = R , but it ∂f ∂f
∗ 33. + = 0; x = u cos v , y = u sin v ;
can be shown that as r → R , values of V approach 1 for ∂x ∂y
0 < θ < π , and approach 0 for −π < θ < 0. Solve the
' (2 ' (2
∂F 1 ∂F
expression in part (a) for r in terms of V and θ , and use + 2 =0
∂u u ∂v
the result to plot equipotential curves for V = 1/8, 1/4,
3/8, 1/2, 5/8, 3/4, and 7/8. Set R = 1 to do this. ∗ 34. In many problems in elasticity theory, the Airy’s stress function
+(x, y) must satisfy the biharmonic equation
∗ 24. Verify that
' ( ∂ 4+ ∂ 4+ ∂ 4+
1 1 −1 2Rr sin θ + 2 + = 0.
V (r, θ ) = (V1 + V2 ) + (V1 − V2 ) Tan ∂x 4 ∂x 2 ∂y 2 ∂y 4
2 π R2 − r 2
Use Example 12.19 to show that in polar coordinates, the equation can
satisfies 12.24 for r < R . It represents potential in the circle in the
be expressed in the form
figure in Exercise 23 when potential on the upper edge is V1 and that
on the lower edge is V2 . ' (' (
∂2 1 ∂ 1 ∂2 ∂ 2+ 1 ∂+ 1 ∂ 2+
2
+ + 2 2 2
+ + 2 = 0.
∗ 25. If f (s) and g(t) are differentiable functions, show that ∇f (x 2 − ∂r r ∂r r ∂θ ∂r r ∂r r ∂θ 2
y 2 ) · ∇g(xy) = 0.
∗ 26. If f (s) is a differentiable function, show that f (x − y) satisfies ∗ 35. An observer travels along the curve x = t 2 , y = 3t 3 + 1, z =
the equation 2t + 5, where x , y , and z are in metres and t ≥ 0 is in seconds. If the
∂f ∂f density ρ of a gas (in kg/m 3 ) is given by ρ = (3x 2 + y 2 )/(z2 + 5) ,
=− .
∂y ∂x find the time rate of change of the density of the gas as measured by
the observer when t = 2 s.
∗ 27. If f (s) is a differentiable function, show that u(x, y) = f (4x −
!
3y) + 5(y − x) satisfies the equation ∗ 36. If f (r) is a differentiable function and r = x 2 + y 2 + z2 ,
show that
∂u ∂u f + (r)
3 +4 = 5. ∇f = (x î + y ĵ + zk̂).
∂x ∂y r
∗ 28. If f (s) and g(t) are twice differentiable, show that the function ∗ 37. If f (x, y) = 0 defines y as a function of x , show that
u(x, y) = xf (x + y) + yg(x + y) satisfies
d 2y fxx fy2 − 2fxy fx fy + fyy fx2
∂ 2u ∂ 2u ∂ 2u 2
=− .
2
−2 + 2 = 0. dx fy3
∂x ∂x ∂y ∂y
∗ 29. If f (s) and g(t) are twice differentiable, show that f (x − y) + ∗ 38. If f (x, y) is a harmonic function, show that the function
g(x + y) satisfies F (x, y) = f (x 2 − y 2 , 2xy) is also harmonic.
∂ 2u ∂ 2u
− 2 = 0. ∗ 39. (a) Show that f (x, y) = ln(x 2 + y 2 ) satisfies Laplace’s
∂x 2 ∂y equation 12.11.
∗ 30. Show that if f (v) is differentiable, then u(x, y) = x 2 f (y/x) (b) Transform f (x, y) into polar coordinates and show that
satisfies the function satisfies 12.24.
∂u ∂u
x +y = 2u. ∗ 40. Find an identity satisfied by the second partial derivatives of a
∂x ∂y
function f (x, y, z) that is positively homogeneous of degree n .
∗∗ 41. It is postulated in one of the theories of traffic flow that the average
In Exercises 31–33 suppose that f (x, y) satisfies the first partial dif-
speed u at a point x on a straight highway (along the x -axis) is related
ferential equation. Show that with the change of independent variables,
to the concentration k of traffic by the differential equation
function F (u, v) = f [x(u, v) , y(u, v)] must satisfy the second par-
tial differential equation. ∂u ∂u ∂k
' (2 ' (2 u + = −c2 k n ,
∂f ∂f ∂x ∂t ∂x
∗ 31. + = 0; u = (x + y)/2, v = (x − y)/2;
∂x ∂y where t is time, and c > 0 and n are constants.
' (2 ' (2
∂F ∂F (a) Use chain rules for ∂u/∂x and ∂u/∂t in the functional
+ =0
∂u ∂v situation u = f (k) and k = g(x, t) to show that
' (
∂ 2f ∂ 2f ∂ 2F du ∂k ∂k ∂k
∗ 32. − = 0; u = (x +y)/2 , v = (x −y)/2; =0 u + + c2 k n = 0.
∂x 2 ∂y 2 ∂u ∂v dk ∂x ∂t ∂x
(b) The equation of continuity for traffic flow states that ∗∗ 43. Two equal masses m are connected by springs having equal spring
constant k so that the masses are free to slide on a frictionless table
∂k ∂(ku) (see figure below). The walls A and B are fixed.
+ = 0.
∂t ∂x
Use these last two equations to obtain the differential equa- m m
k k k
tion relating speed and concentration:
A B
du
= −ck (n−1)/2 . x1 x2
dk
(a) Use Newton’s second law to show that the differential equa-
(c) Solve the differential equation in part (b) for u = f (k) . tions for the motions of the masses are
∗∗ 42. A bead slides from rest at the origin on a frictionless wire in a mẍ1 = k(x2 − 2x1 ), mẍ2 = k(x1 − 2x2 ),
vertical plane to the point (x0 , y0 ) under the influence of gravity (figure
below). As it does so, gravitational potential energy is converted into where x1 and x2 are the displacements of the masses from
kinetic energy. At (x, y) , the bead has lost potential energy mgy . their equilibrium positions, ẍ1 = d 2 x1 /dt 2 and ẍ2 =
If its kinetic energy is mv 2 /2 at this point, then 2
! mgy = mv /2, so d 2 x2 /dt 2 .
√ 2 2
that v = 2gy . To travel a small distance (dx) + (dy) along
! (b) The Euler–Lagrange equations from theoretical mechanics
the curve at (x, y) with velocity v takes time (dx)2 + (dy)2 /v . for this system are
Hence, the total time to traverse the entire curve is
' ( ' (
6 ! d ∂L ∂L d ∂L ∂L
x0
(dx)2 + (dy)2 − = 0, − = 0,
t = dt ∂ ẋ1 ∂x1 dt ∂ ẋ2 ∂x2
0 v
8
6 x0 ' (2 6 x0 8 where L is defined as the kinetic energy of the two masses
1 dy 1 1 + (y + )2 less the energy stored in the springs. Show that
= √ 1+ dx = √ dx.
0 2gy dx 2g 0 y
m
L(x1 , x2 , ẋ1 , ẋ2 ) = (ẋ12 + ẋ22 ) − k(x12 + x22 − x1 x2 ).
The problem of finding the shape of wire that makes t as small as 2
possible is called the brachistochrone problem. It is shown in the
(c) Obtain the equations in part (a) from the Euler–Lagrange
calculus of variations that y = f (x) must satisfy the equation
equations in part (b).
' (
d ∂F ∂F
− = 0, ∗∗ 44. Suppose that the second-order partial differential equation
dx ∂y + ∂y
' (
∂ 2z ∂ 2z ∂ 2z ∂z ∂z
where 8 p 2 +q + r 2 = F x, y, z, ,
1 + (y + )2 ∂x ∂x∂y ∂y ∂x ∂y
+
F (y, y ) = .
y ( p , q , and r are constants) is subjected to the change of variables
(a) Show that f (x) must satisfy the differential equation 1 +
s = ax + by, t = cx + dy,
(y + )2 + 2yy ++ = 0.
(b) Show that the curve that satisfies the equation in part (a) is where a , b , c , and d are constants. Show that the partial differential
the cycloid defined parametrically by equation in s and t is
x = a(θ − sin θ), y = a(1 − cos θ), ' (
∂ 2z ∂ 2z ∂ 2z ∂z ∂z
P + Q + R = G s, t, z, , ,
where a is a constant. ∂s 2 ∂s∂t ∂t 2 ∂s ∂t
(c) Show that it does not matter what point on the cycloid the
where Q2 − 4P R = (q 2 − 4pr)(ad − bc)2 .
bead starts from, the time to get to (x0 , y0 ) is always the
same.
∗∗ 45. Show that if a solution u = f (x, y, z) of the three-dimensional
Laplace
! equation 12.12 can be expressed in the form u = g(r) , where
r = x 2 + y 2 + z2 , then f (x, y, z) must be of the form
x
C
y = f (x) f (x, y, z) = ! + D,
x 2 + y 2 + z2
y (x, y) (x0, y0) where C and D are constants.
12.7 Implicit Differentiation

In Section 3.8 we introduced the technique of implicit differentiation in order to obtain the
derivative of a function y = f (x) defined implicitly by an equation
F (x, y) = 0. (12.27)
Essentially, the technique is to differentiate all terms in the equation with respect to x , considering
all the while that y is a function of x . For example, if y is defined implicitly by
x 2 y 3 + 3xy = 3x + 2,
implicit differentiation gives
dy dy dy 3 − 2xy 3 − 3y
2xy 3 + 3x 2 y 2 + 3y + 3x =3 ,⇒ = .
dx dx dx 3x 2 y 2 + 3x
With the chain rule we can actually present a formula for dy/dx . Since equation 12.27,
when written in the form
F [x, f (x)] = 0,
must be valid for all x in the domain of the function f (x) , we can differentiate it with respect
F to x . From the schematic diagram to the left, the derivative of the left side of the equation is
x y dF ∂F ∂F dy
= + .
dx ∂x ∂y dx
x
If we equate this to the derivative of the right side of the equation, we find
dy
F x + Fy =0
dx
or
dy Fx
=− . (12.28)
dx Fy
For the function defined implicitly above by x 2 y 3 + 3xy − 3x − 2 = 0, equation 12.28

gives
dy 2xy 3 + 3y − 3
=− ,
dx 3x 2 y 2 + 3x
and this result is identical to that obtained by implicit differentiation.
Similarly, if the equation
F (x, y, z) = 0 (12.29)
defines z implicitly as a function of x and y , the schematic diagram to the left immediately
yields
F ∂F ∂F ∂z ∂F ∂F ∂z
+ = 0, + = 0.
x y z ∂x ∂z ∂x ∂y ∂z ∂y
x y
From these we obtain the results
∂z Fx ∂z Fy
=− , =− . (12.30)
∂x Fz ∂y Fz
We do not suggest that formulas 12.28 and 12.30 be memorized. On the contrary, we obtain
results in this section that include 12.28 and 12.30 as special cases. To develop these results we
work with three equations in five variables:
F (x, y, u, v, w) = 0, G(x, y, u, v, w) = 0, H (x, y, u, v, w) = 0. (12.31)

We assume that these equations define u , v , and w as functions of x and y for some domain
of values of x and y (and do so implicitly). It might even be possible to solve the system and
obtain explicit definitions of the functions
u = f (x, y), v = g(x, y), w = h(x, y). (12.32)
We pose the problem of finding the six first-order partial derivatives of u , v , and w with respect
to x and y , supposing that it is undesirable or even impossible to obtain the explicit form of the
functions. To do this, we note that were results 12.32 known and substituted into 12.31, then
F [x, y, f (x, y), g(x, y), h(x, y)] = 0,

G[x, y, f (x, y), g(x, y), h(x, y)] = 0,
H [x, y, f (x, y), g(x, y), h(x, y)] = 0
would be identities in x and y . As a result we could differentiate each equation with respect to
x , obtaining from the schematic diagram
∂F ∂F ∂u ∂F ∂v ∂F ∂w
+ + + = 0,
∂x ∂u ∂x ∂v ∂x ∂w ∂x
∂G ∂G ∂u ∂G ∂v ∂G ∂w
+ + + = 0, (12.33)
∂x ∂u ∂x ∂v ∂x ∂w ∂x
∂H ∂H ∂u ∂H ∂v ∂H ∂w
+ + + = 0,
∂x ∂u ∂x ∂v ∂x ∂w ∂x
F, G, H
x y u v w
x y x y x y
or
∂u ∂v ∂w
Fu + Fv + Fw = −Fx ,
∂x ∂x ∂x
∂u ∂v ∂w
Gu + Gv + Gw = −Gx , (12.34)
∂x ∂x ∂x
∂u ∂v ∂w
Hu + Hv + Hw = −Hx .
∂x ∂x ∂x
We have in 12.34 three equations in the three unknowns ∂u/∂x , ∂v/∂x , and ∂w/∂x , and
because the equations are linear in the unknowns, solutions can be obtained using Cramer’s
rule. † In particular,
/ / / /
/ −Fx Fv Fw / / Fx Fv Fw /
/ / / /
/ −Gx Gv Gw / / Gx Gv Gw /
∂u / −H H Hw / /H Hv Hw /
= / x v
/ = −/ x /. (12.35)
∂x / Fu F v Fw / / Fu Fv Fw /
/ / / /
/ Gu Gv Gw / / Gu Gv Gw /
/H H Hw / /H Hv Hw /
u v u
†
Cramer’s rule is discussed in Appendix B.
The two determinants on the right of 12.35 involve only derivatives of the given functions F ,
G , and H , and we have therefore obtained a method for finding ∂u/∂x that avoids solving
12.31 for u , v , and w . We could list similar formulas for the remaining five derivatives, but
first we introduce some simplifying notation.
DEFINITION 12.5
The Jacobian determinant of functions F , G , and H with respect to variables u , v ,
∂(F, G, H )
and w is denoted by and is defined as the determinant
∂(u, v, w)
/ ∂F ∂F ∂F /
/ /
/ /
/ / / ∂u ∂v ∂w /
/ Fu F v F w / / /
∂(F, G, H ) / / / ∂G ∂G ∂G /
= / Gu Gv Gw / = / / /. (12.36)
∂(u, v, w) /H H H / / ∂u ∂v ∂w //
u v w / /
/ ∂H ∂H ∂H /
/ /
∂u ∂v ∂w
With this notation we can write 12.35 in the form
∂(F, G, H )
∂u ∂(x, v, w)
=− . (12.37)
∂x ∂(F, G, H )
∂(u, v, w)
The remaining derivatives of v = g(x, y) and w = h(x, y) with respect to x can also be
obtained from equations 12.34 by Cramer’s rule:
∂(F, G, H ) ∂(F, G, H )
∂v ∂(u, x, w) ∂w ∂(u, v, x)
=− , =− . (12.38)
∂x ∂(F, G, H ) ∂x ∂(F, G, H )
∂(u, v, w) ∂(u, v, w)
A similar procedure yields
∂(F, G, H ) ∂(F, G, H ) ∂(F, G, H )

∂u ∂(y, v, w) ∂v ∂(u, y, w) ∂w ∂(u, v, y)
=− , =− , =− . (12.39)
∂y ∂(F, G, H ) ∂y ∂(F, G, H ) ∂y ∂(F, G, H )
∂(u, v, w) ∂(u, v, w) ∂(u, v, w)
Formulas 12.37–12.39 apply only to the situation in which equations 12.31 define u , v ,
and w as functions of x and y . It is, however, fairly evident how to construct formulas in other
situations. Here are the rules:
1. The partial derivative has a Jacobian divided by a Jacobian (and do not forget the negative
sign).
2. In the denominator, it is the Jacobian of the functions defining the original equations with
respect to the dependent variables.
3. The only difference in the Jacobian in the numerator is that the dependent variable that is
being differentiated is replaced by the independent variable with respect to which differen-
tiation is being performed.
The results in equations 12.37–12.39 are valid provided, of course, that the Jacobian
∂(F, G, H )
(= 0.
∂(u, v, w)
In actual fact, it is this condition that guarantees that equations 12.31 do define u , v , and w as
functions of x and y in the first place.
As a second example, the equations
F (x, y, s, t) = x +y 2 − 2xs +t + 1 = 0, G(x, y, s, t) = x 2 −y 4 − 2y 2 +y + 3s + 2t 3 + 2 = 0
define x and y as functions of s and t . To find ∂x/∂s when s = 1 and t = 0, we first calculate
∂(F, G) / / / /
/ Fs F y / / −2 x 2y /
/ / / 3 /
∂x ∂(s, y) Gs Gy 3 −4 y − 4 y + 1
=− = −/ / = −/ /.
∂s ∂(F, G) / Fx F y / / 1 − 2s 2y /
/ / / 3 /
∂(x, y) G x G y 2 x − 4 y − 4 y + 1
When s = 1 and t = 0, the equations defining x and y reduce to
x + y 2 − 2x + 1 = 0, x 2 − y 4 − 2y 2 + y + 5 = 0.
The first gives x = 1 + y 2 , which we substitute into the second:
0 = (1 + y 2 )2 − y 4 − 2y 2 + y + 5 = y + 6.
Thus, y = −6, and x = 1 + 36 = 37. With these values, the partial derivative is
/ /
/ −74 −12 /
/ /
∂x 3 889
=−/ / = 65 750.
∂s / −1 −12 /
/ /
74 889
EXAMPLE 12.20
If x 2 y 2 z3 + zx sin y = 5 defines z as a function of x and y , find ∂z/∂x .
SOLUTION If we set F (x, y, z) = x 2 y 2 z3 + zx sin y − 5 = 0, then
∂(F )
∂z ∂(x) Fx 2xy 2 z3 + z sin y
=− =− =− 2 2 2 .
∂x ∂(F ) Fz 3x y z + x sin y
∂(z)
EXAMPLE 12.21
The equations
x 2 y 3 z3 + uvw + 1 = 0, x 2 + y 2 + z2 + u3 + v 3 + w 2 = 6, u + v + w = x + 2y,
define u , v , and w as functions of x , y , and z . Find ∂v/∂z when x = 1, y = 0, z = 2,

u = 1, v = −1, and w = 1.
SOLUTION If we set
F (x, y, z, u, v, w) = x 2 y 3 z3+uvw+1, G(x, y, z, u, v, w) = x 2+y 2+z2+u3+v 3+w 2−6,
H (x, y, z, u, v, w) = u + v + w − x − 2y,
then
/ / / /
/ Fu Fz Fw / / vw 3x 2 y 3 z2 uv /
∂(F, G, H ) / / / 2 /
/ Gu Gz Gw / / 3u 2z 2w /
∂v ∂(u, z, w) /H Hz Hw / / 1 0 1
/
=− = −/ u / =− / / .
∂z ∂(F, G, H ) / Fu Fv Fw / / vw uw uv /
/ / / 2 /
∂(u, v, w) / Gu Gv Gw / / 3u 3v 2 2w /
/H Hv Hw / / 1 1 1
/
u
Instead of expanding these determinants, and then substituting values for the variables, we
substitute first, and then expand:
/ /
/ −1 0 −1 /
/ /
/ 3 4 2/
∂v / 1 0 1
/
= −/ / = 0.
∂z / −1 1 −1 /
/ /
/ 3 3 2/
/ 1 1 1
/
EXERCISES 12.7
In Exercises 1–4 y is defined implicitly as a function of x . Find define x , y , and z as functions of u and v , find ∂x/∂u)v at the values
dy/dx . x = 1, y = 1, u = π/2, v = 0, and z = 0.
1. x 3 y 2 − 2xy + 5 = 0 2. (x + y)2 = 2x
∗ 15. If the equation F (x, y, z) = 0 defines each of x , y , and z as a
3. x(x − y) − 4y 3 = 2exy + 6 4. sin (x + y) + y 2 = 12x 2 + y function of the other two, show that
In Exercises 5–8 z is defined implicitly as a function of x and y . Find ' (' (' (
∂z ∂x ∂y
∂z/∂x and ∂z/∂y . = −1.
∂x y ∂y z ∂z x
5. x 2 sin z − yez = 2x 6. x 2 z2 + yz + 3x = 4
2 2 3
7. z sin y + y sin x = z 8. Tan −1 (yz) = xz ∗ 16. If z = ex cos y , where x and y are functions of t defined by
In Exercises 9–13 find the required derivative. Assume that the system x 3 + ex − t 2 − t = 1, yt 2 + y 2 t − t + y = 0,
of equations does define the function(s) indicated.
9. ∂u/∂x and ∂v/∂y if x 2 − y 2 + u2 + 2v 2 = 1, x 2 + y 2 = find dz/dt .

2 + u2 + v 2
10. ∂x/∂t if sin (x + t) − sin (x − t) = z
∗ 17. Find ∂s/∂u)v if s = x 2 + y 2 , and x and y are functions of u
11. ∂φ/∂x)y,z if x = r sin φ cos θ, y = r sin φ sin θ, z = and v defined by
r cos φ
12. dz/dx if x 2 + y 2 − z2 + 2xy = 1, x 3 + y 3 − 5y = 4
u = x2 − y2, v = x 2 − y.
13. ∂u/∂y)x if xyu + vw = 4, y 2 + u2 − u2 v = y, yw +
xu + v + 4 = 0
∗ 14. Given that the equations ∗ 18. Find ∂z/∂y)x if z = u3 v + sin (uv) , and u and v are functions
of x and y defined by
x 2 − y cos (uv) + z2 = 0, x 2 + y 2 − sin (uv) + 2z2 = 2,
xy − sin u cos v + z = 0 x = eu cos v, y = eu sin v.

∗ 19. Given that z3 − xz − y = 0 defines z as a function of x and y ,

show that
∂ 2z 3z2 + x
∂(u, v) ∂(s, t) ∂(u, v)
=− 2 . = .
∂x∂y (3z − x)3 ∂(s, t) ∂(x, y) ∂(x, y)
(b) If the equations F (u, v, x, y) = 0, G(u, v, x, y) = 0
∗ 20. If the equations x = u2 − v 2 , y = 2uv , define u and v as define u and v as functions of x and y , and also define x
functions of x and y , find ∂ 2 u/∂x 2 .
and y as functions of u and v , show that
∗ 21. (a) Given that the equation z4 x + y 3 z + 9x 3 = 2 defines z as ∂(u, v) 1
a function of x and y , and x as a function of y and z , are = .
∂(x, y) ∂(x, y)
∂z/∂x and ∂x/∂z reciprocals?
∂(u, v)
(b) Given that the equations z4 x+y 3 z+ 9x 3 = 2, x 2 y+xz =
1 define z as a function of x , and x as a function of z , are
dz/dx and dx/dz reciprocals? ∗∗ 24. Suppose the system of m linear equations in n unknowns ( n > m )
(c) Given that the equations u2 − v = 3x + y , u − 2v 2 = n
5
x − 2y define u and v as functions of x and y , and also aij xj = ci , i = 1, . . . , m
define x and y as functions of u and v , are ∂u/∂x and j =1
∂x/∂u reciprocals?
defines x1 , x2 , . . . , xm as functions of xm+1 , xm+2 , . . . , xn . Show
∗ 22. Given that the equations x 2 − 2y 2 s 2 t − 2st 2 = 1, x 2 + 2y 2 s 2 t + that if 1 ≤ i ≤ m and m + 1 ≤ j ≤ n , then
5st 2 = 1 define s and t as functions of x and y , find ∂ 2 t/∂y 2 .
∂xi Dij
=− ,
∗∗ 23. (a) Suppose the equations F (u, v, s, t) = 0, G(u, v, s, t) = ∂xj D
0 define u and v as functions of s and t , and the equations
H (s, t, x, y) = 0, I (s, t, x, y) = 0 define s and t as where D = |aij |m×m , and Dij is the same as determinant D except
functions of x and y . Show that that its i th column is replaced by the j th column of [aij ]m×n .
12.8 Directional Derivatives

FIGURE 12.15 Rate of If a function f (x, y, z) is defined throughout some region of space, then at any point (x0 , y0 , z0 )
change of a function f (x, y, z) in we can calculate its partial derivatives ∂f/∂x , ∂f/∂y , and ∂f/∂z . These derivatives define
an arbitrary direction v rates of change of f (x, y, z) at (x0 , y0 , z0 ) in directions parallel to the x -, y -, and z -axes.
But what if we want the rate of change of f (x, y, z) at (x0 , y0 , z0 ) in some arbitrary direction
z
defined by a vector v (Figure 12.15)? By the rate of change of f (x, y, z) in the direction v, we
v mean the rate of change with respect to distance as measured along a line through (x0 , y0 , z0 )
in direction v. Let us define s as a measure of directed distance along this line, taking s = 0
at (x0 , y0 , z0 ) and positive s in the direction of v. What we want, then, is the derivative of
(x0, y0, z0)
f (x, y, z) with respect to s at s = 0. To express f (x, y, z) in terms of s , we use parametric
equations of the line through (x0 , y0 , z0 ) along v. If v̂ = (vx , vy , vz ) is a unit vector in the
y direction of v, then parametric equations for this line (see equations 11.37) are
x
x = x0 + vx s, y = y0 + vy s, z = z0 + vz s. (12.40)
From the schematic diagram to the left, we obtain

f
df ∂f dx ∂f dy ∂f dz ∂f ∂f ∂f
x y z = + + = vx + vy + vz ,
ds ∂x ds ∂y ds ∂z ds ∂x ∂y ∂z
s s s
where all partial derivatives of f (x, y, z) are to be evaluated at (x0 , y0 , z0 ) . We call this a
directional derivative. It is given an alternative notation in the following definition.
DEFINITION 12.6
The directional derivative of a function f (x, y, z) in the direction v̂ = (vx , vy , vz ) at
the point (x0 , y0 , z0 ) is
∂f ∂f ∂f
Dv f = vx + vy + vz . (12.41)
∂x ∂y ∂z
Now vx , vy , and vz are the components of the unit vector v̂ in the direction of v, and
∂f/∂x , ∂f/∂y , and ∂f/∂z are the components of the gradient of f (x, y, z) . We can write,
therefore, that
Dv f = ∇f · v̂. (12.42)
Consequently, the derivative (rate of change) of a function in any given direction is the scalar
product of the gradient of the function and a unit vector in the required direction. We state this
in the following theorem.
THEOREM 12.4
The directional derivative of a function in any direction is the component of the gradient
of the function in that direction.
EXAMPLE 12.22
Find Dv f at (4, 0, 16) if f (x, y, z) = x 3 ey + xz and v is the vector from (4, 0, 16) to
(−2, 1, 4) .
SOLUTION Since
∇f|(4,0,16) = [(3x 2 ey + z)î + x 3 ey ĵ + x k̂]|(4,0,16)
= 64î + 64ĵ + 4k̂
and
v (−6, 1, −12) −1
v̂ = = √ = √ (6, −1, 12),
|v | 36 + 1 + 144 181
we have
(6, −1, 12) 368
Dv f = −(64, 64, 4) · √ = −√ .
181 181
The fact that the derivative is negative means that f (x, y, z) is decreasing in direction v.
The directional derivative gives us insight into some of the properties of the gradient vector. In
particular, we have the next theorem.
THEOREM 12.5
The gradient ∇f of a function f (x, y, z) defines the direction in which the function
increases most rapidly, and the maximum rate of change is |∇f | .
PROOF Theorem 12.4 states that the directional derivative of f (x, y, z) in a direction v
is the component of ∇f in that direction. Figure 12.16, which shows components of ∇f in
FIGURE 12.16 Gradient of a various directions, makes it clear that Dv f is greatest when v is parallel to ∇f . Alternatively,
function points in the direction in which if θ is the angle between v and ∇f , then
the function increases most rapidly
v
Dv f = ∇f · v̂ = |∇f ||v̂| cos θ = |∇f | cos θ.
v
∇f Obviously Dv f is a maximum when cos θ is a maximum (i.e., when cos θ = 1 or θ = 0) and
Component
of ∇f this occurs when v is parallel to ∇f . Finally, when v is parallel to ∇f , we have Dv f = |∇f | ,
along v v and this completes the proof.
Note that for any function f (x, y, z) ,
v
∂f ∂f ∂f
v Dî f = , Dĵ f = , Dk̂ f = .
∂x ∂y ∂z
In other words, the partial derivatives of a function are its directional derivatives along the
coordinate directions.
EXAMPLE 12.23
Find the direction at the point (1, 2, −3) in which the function f (x, y, z) = x 2 y + xyz
increases most rapidly.
SOLUTION According to Theorem 12.5, f (x, y, z) increases most rapidly in the direction
∇f|(1,2,−3) = (2xy + yz, x 2 + xz, xy)|(1,2,−3) = (−2, −2, 2).
You might feel that because the definition of the directional derivative Dv f does not involve a
limit process, it is some strange new type of differentiation. To show that this is not the case, let
us return to the calculation of the derivative of f (x, y, z) at (x0 , y0 , z0 ) in the direction v shown
in Figure 12.15. With parametric equations 12.40 for the line through (x0 , y0 , z0 ) along v, the
value of f (x, y, z) at any point (x, y, z) along this line is f (x0 + vx s, y0 + vy s, z0 + vz s) .
If we take the difference between this value and f (x0 , y0 , z0 ) and divide by the distance s
between (x0 , y0 , z0 ) and (x, y, z) , then the limit of this expression as s → 0+ should define
the derivative of f (x, y, z) at (x0 , y0 , z0 ) in the direction v; that is,
f (x0 + vx s, y0 + vy s, z0 + vz s) − f (x0 , y0 , z0 )
Dv f = lim+ . (12.43)
s→0 s
It can be shown that this limit (and this is perhaps the form we might have expected the derivative
to take) also leads to the result contained in 12.42 (see Exercise 35).
Consider a curve C in space that is defined parametrically by
FIGURE 12.17 Rate of
change of a function along a curve C : x = x(t), y = y(t), z = z(t), α ≤t ≤β
z
B (Figure 12.17). Imagine that C is the path traced out by some particle as it moves through space
under the action of some system of forces, and suppose that f (x, y, z) is a function defined
C along C . Perhaps the particle is a weather balloon and f (x, y, z) is temperature at points
along its trajectory C . In such applications we are frequently asked for the rate of change of
A f (x, y, z) with respect to distance travelled along C . If we use s as a measure of distance
along C (taking s = 0 at A ), then the required rate of change is df/ds . Since the coordinates
of points (x, y, z) on C can be regarded as functions of s (although it might be difficult to find
x y these functions explicitly), the chain rule gives
df ∂f dx ∂f dy ∂f dz
= + +
ds ∂x ds ∂y ds ∂z ds
' ( ' (
f ∂f ∂f ∂f dx dy dz
= , , · , ,
x y z ∂x ∂y ∂z ds ds ds
s s s
dr
= ∇f · .
ds
In Section 11.11 we saw that d r/ds is a unit tangent vector T̂ to C . Consequently,
df
= ∇f · T̂.
ds
But this equation states that df/ds is the directional derivative of f (x, y, z) along the tangent
direction to the curve C . In other words, to calculate the rate of change of a function f (x, y, z)
with respect to distance as measured along a curve C , we calculate the directional derivative of
f (x, y, z) in the direction of the tangent vector to C .
EXAMPLE 12.24
Find the rate of change of the function f (x, y, z) = x 2 y − xz along the curve y = x 2 , z = x
in the direction of decreasing x at the point (2, 4, 2) .
SOLUTION Since parametric equations for the curve are C : x = −t , y = t 2 , z = −t ,
a tangent vector to C at any point is T = (−1, 2t, −1) . At (2, 4, 2) , t = −2, and the
tangent vector is T = (−1, −4, −1) . A unit tangent vector to C at (2, 4, 2) in the direction
of decreasing x is therefore
(−1, −4, −1) −1

T̂ = √ = √ (1, 4, 1).
18 3 2
The rate of change of f (x, y, z) in this direction is
(1, 4, 1)
∇f · T̂ = (2xy − z, x 2 , −x)|(2,4,2) · √
−3 2
−1 28
= √ (14, 4, −2) · (1, 4, 1) = − √ .
3 2 3 2
FIGURE 12.18 First and
second directional derivatives of a
function f (x, y) of two variables
z In preparation for maxima and minima of multivariable functions in Section 12.10, we now
discuss directional derivatives for a function f (x, y) of two independent variables. Such a
z = f (x, y)
function can be represented graphically as a surface z = f (x, y) (Figure 12.18).
(x0, y0, z0) For a direction v at (x0 , y0 ) in the xy -plane,
Dv f = ∇f · v̂,
where ∇f is evaluated at (x0 , y0 ) . Algebraically, this is the rate of change of f (x, y) in
x direction v. Geometrically, it is the rate of change of the height z of the surface as we move
(x0, y0) y along the curve of intersection of the surface and a vertical plane containing the vector v, or the
slope of this curve. Each direction v at (x0 , y0 ) defines an angle α with a line through (x0 , y0 )
v parallel to the positive x -axis, and for this direction
v
v̂ = cos α î + sin α ĵ.
We can write, then,

' (
∂f ∂f
Dv f = ∇f · v̂ = î + ĵ · (cos α î + sin α ĵ)
∂x ∂y
∂f ∂f
= cos α + sin α. (12.44)
∂x ∂y
If Dv f represents the slope of the curve of intersection of the surface and the vertical plane
through v, then Dv (Dv f ) represents the rate of change of this slope. Now
Dv (Dv f ) = ∇(Dv f ) · v̂
and
' ( ' 2 (
∂ 2f ∂ 2f ∂ f ∂ 2f
∇(Dv f ) = cos α + sin α î + cos α + sin α ĵ.
∂x 2 ∂x ∂y ∂y ∂x ∂y 2
Thus,
' ( ' 2 (
∂ 2f ∂ 2f ∂ f ∂ 2f
Dv (Dv f ) = cos α + sin α cos α + cos α + sin α sin α
∂x 2 ∂x ∂y ∂y ∂x ∂y 2
∂ 2f 2 ∂ 2f ∂ 2f
= cos α + 2 cos α sin α + sin2 α. (12.45)
∂x 2 ∂x ∂y ∂y 2
We call Dv (Dv f ) the second directional derivative of f (x, y) at (x0 , y0 ) in direction v. If it is
positive, then the curve of intersection is concave upward, whereas if it is negative, the curve is
concave downward. We will find these results useful in Section 12.10 when we discuss relative
extrema of functions of two independent variables.
EXERCISES 12.8
In Exercises 1–8 calculate the directional derivative of the function at 10. f (x, y) = x 2 + y at (−1, 3) along the curve y = −3x 3 in the
the point and in the direction indicated. direction of decreasing x
1. f (x, y, z) = 2x 2 − y 2 + z2 at (1, 2, 3) in the direction of the 11. f (x, y, z) = xy + z2 at (1, 0, −2) along the curve y = x 2 −
vector from (1, 2, 3) to (3, 5, 0) 1, z = −2x in the direction of increasing x
2. f (x, y, z) = x 2 y + xz at (−1, 1, −1) in the direction of the 12. f (x, y, z) = x 2 y+xy 3 z at (2, −1, 2) along the curve x 2 −y 2 =
vector that joins (3, 2, 1) to (3, 1, −1) 3, z = x in the direction of increasing x
3. f (x, y) = xey + y at (3, 0) in the direction of the vector from
(3, 0) to (−2, −4)
In Exercises 13–18 find the direction in which the function increases
4. f (x, y, z) = ln (xy + yz + xz) at (1, 1, 1) in the direction from most rapidly at the point. What is the rate of change in that direction?
(1, 1, 1) toward the point (−1, −2, 3)
5. f (x, y) = Tan −1 (xy) at (1, 2) along the line y = 2x in the 13. f (x, y, z) = x 4 yz − xy 3 + z at (1, 1, −3)
direction of increasing x 14. f (x, y) = 2xy + ln (xy) at (2, 1/2)
6. f (x, y) = sin (x + y) at (2, −2) along the line 3x + 4y = −2 !
15. f (x, y, z) = 1/ x 2 + y 2 + z2 at (1, −3, 2)
in the direction of decreasing y
!
7. f (x, y, z) = x 3 y sin z at (3, −1, −2) along the line x = 3 + 16. f (x, y, z) = −1/ x 2 + y 2 + z2 at (1, −3, 2)
t, y = −1 + 4t, z = −2 + 2t in the direction of decreasing x
17. f (x, y, z) = Tan −1 (xyz) at (3, 2, −4)
8. f (x, y, z) = x 2 y + y 2 z + z2 x at (1, −1, 0) along the line x +
2y + 1 = 0, x − y + 2z = 2 in the direction of decreasing z 18. f (x, y) = xyexy at (1, 1)
19. In what direction is the rate of change of f (x, y, z) = xyz small-

In Exercises 9–12 find the rate of change of the function with respect
est at the point (2, −1, 3) ?
to distance travelled along the curve.
∗ 20. In what directions (if any) is the rate of change of the function
9. f (x, y) = 2x − 3y at (1, 1) along the curve y = x 2 in the f (x, y) = x 2 y + y 3 at the point (1, −1) equal to (a) 0, (b) 1, and (c)
direction of increasing x 20?
∗ 21. In what directions (if any) is the rate of change of the function ∗ 28. If we know the rate of change of a function f (x, y, z) at a point
f (x, y, z) = xy + z at the point (0, 1, −2) equal to (a) 0, (b) 1, and P on a curve C , proceeding in one direction along C , what is the rate
(c) −20? of change in the opposite direction along C ?
∗ 22. Must there always be a direction in which the rate of change of a ∗ 29. What is the rate of change of a function f (x, y, z) in a direction
function at a point is equal to (a) 0 and (b) 3? perpendicular to ∇f ?
∗ 30. The rate of change of a function f (x, y) at a point (x0 , y0 ) in
∗ 23. In the derivation of 12.41, why was it necessary to use a unit vector
direction î + 2ĵ is 3 and the rate of change in direction −2î − ĵ is −1.
v̂ to determine parametric equations for the line through (x0 , y0 , z0 )
along v? In other words, why could we not use the components of v Find its rate of change in direction 2î + 3ĵ .
itself to write parametric equations for the line? ∗ 31. Rates of change of a function f (x, y, z) at a point (x0 , y0 , z0 ) in
directions î + ĵ , 2î − k̂ , and î − ĵ + k̂ are 1, 2, and −3, respectively.
∗ 24. How fast is the distance to the origin changing with respect to What is its partial derivative with respect to z at the point?
distance travelled along the curve x = 2 cos t, y = 2 sin t, z = 3t
at any point on the curve? What is the rate of change when t = 0? ∗ 32. Find the second directional derivative of the function f (x, y) =
Would you expect this? x 3 y 2 at the point (1, 1) in the direction of the vector (1, −2) .
∗ 33. Find the second directional derivative of the function f (x, y, z) =
∗ 25. Find points on the curve C : x = t, y = 1 − 2t, z = t at which
x 2 + 2y 2 + 3z2 at the point (−2, −1, 3) in the direction (1, 1, −1) .
the rate of change of f (x, y, z) = x 2 + xyz with respect to distance
travelled along the curve vanishes. ∗ 34. The path followed by a stone embedded in the tread of a tire is
a cycloid given parametrically by x = R(θ − sin θ), y = R(1 −
∗ 26. Repeat Exercise 25 for the curve C : z = x, x = y 2 and the cos θ), θ ≥ 0 (see Example 9.7 in Section 9.1).
function f (x, y, z) = x 2 − y 2 + z2 . (a) How fast is the distance from the origin changing with re-
spect to distance travelled along the curve at the points cor-
∗ 27. The path of a particle is defined parametrically by x = (cos t + responding to θ = π/2 and θ = π ?
t sin t)î + (sin t − t cos t)ĵ , where t is time. Plot the path called an (b) How fast is the y -coordinate changing at these points?
involute of a circle. Show that the rate of change of the distance of the
(c) How fast is the x -coordinate changing at these points?
particle from the origin, with respect to distance travelled, is always
positive. ∗∗ 35. Verify that expression 12.43 for Dv f leads to formula 12.42.
12.9 Tangent Lines and Tangent Planes

Tangent Lines to Curves
One equation in the coordinates x , y , and z of points in space,
F (x, y, z) = 0, (12.46)
usually defines a surface. (There are exceptions. The equation x 2 + y 2 + z2 = 0 defines a

point, and x 2 + y 2 + z2 = −1 defines nothing.) When each of the equations
F (x, y, z) = 0, G(x, y, z) = 0 (12.47)
defines a surface, then together they define the curve of intersection of the two surfaces (provided,
of course, that the surfaces do intersect). Theoretically, we can find parametric equations for
the curve by setting x equal to some function of a parameter t , say x = x(t) , and then solving
equations 12.47 for y and z in terms of t : y = y(t) and z = z(t) . The parametric definition,
therefore, takes the form
x = x(t), y = y(t), z = z(t), α ≤ t ≤ β, (12.48)
where α and β specify the endpoints of the curve. Practical difficulties arise in choosing x(t)
and solving for y(t) and z(t) . For some examples, it might be more convenient to specify y(t)
and solve for x(t) and z(t) or, alternatively, to specify z(t) and solve for x(t) and y(t) . We
considered examples of such conversions in Section 11.10.
In Section 11.11 we indicated that when a curve C is defined parametrically by 12.48, a

tangent vector to C at any point P is
dr dx dy dz
= î + ĵ + k̂ (12.49)
dt dt dt dt
(Figure 12.19). The tangent line to C at P is defined as the line through P having direction
FIGURE 12.19 Tangent
d r/dt . If (x0 , y0 , z0 ) are the coordinates of P and t0 is the value of t yielding P , then the
line to a curve in space
vector equation for the tangent line at P is
z
d r //
Tangent line (x, y, z) = (x0 , y0 , z0 ) + u / (12.50a)
dt t=t0
to C at P C
(see equation 11.36). Parametric equations for the tangent line are
r P x = x0 + x + (t0 )u, y = y0 + y + (t0 )u, z = z0 + z+ (t0 )u, (12.50b)

dr + + +
dt and in the case where none of x (t0 ) , y (t0 ) , and z (t0 ) vanishes, we can also write symmetric
equations for the tangent line:
y
x x − x0 y − y0 z − z0
+
= + = + . (12.50c)
x (t0 ) y (t0 ) z (t0 )
EXAMPLE 12.25
Find equations for the tangent line to the elliptic helix
C : x = 2 cos t, y = 4 sin t, z = 2t/π
√ √
at P ( 2, 2 2, 1/2) .
SOLUTION Since t = π/4 at P , a tangent vector to C at P is
d r // √ √
/ = (−2 sin t, 4 cos t, 2/π )|t=π/4 = (− 2, 2 2, 2/π ).
dt t=π/4
Symmetric equations for the tangent line are therefore
√ √
x− 2 y−2 2 z − 1/2
√ = √ = .
− 2 2 2 2/π
We have shown the tangent line to the helix in Figure 12.20.
FIGURE 12.20 Tangent line to an elliptic helix
y
P(√2, 2√2, 1/2)
x
EXAMPLE 12.26
Find equations for the tangent line to the curve z = 1 − x 2 , x + y + z = 2 at the point
P (1/2, 3/4, 3/4) .
SOLUTION Parametric equations for the curve are
x = t, y = 1 − t + t 2, z = 1 − t 2.
Since t = 1/2 at P , a tangent vector to the curve at P is

/
d r //
= (1, −1 + 2t, −2t)|t=1/2 = (1, 0, −1).
dt /t=1/2
Because the y -component vanishes, we cannot write full symmetric equations for the tangent
line, although we could write partial symmetric equations involving x and z . Alternatively,
parametric equations for the tangent line are
1 3 3
x = + u, y = , z= − u.
2 4 4
The line is shown in Figure 12.21.
FIGURE 12.21 Tangent line to curve of intersection of parabolic cylinder and plane
z
2
x+y+z=2
1
Tangent z = 1 − x2
line
P
2
y
Tangent Planes to Surfaces

We now consider the problem of finding the equation for the tangent plane at a point P on a
surface S (Figure 12.22). We define the tangent plane as that plane which contains all tangent
lines at P to curves in S through P (provided, of course, that such a plane exists). Suppose
that the surface is defined by the equation
F (x, y, z) = 0, (12.51)
and that
C : x = x(t), y = y(t), z = z(t), α ≤t ≤β
is any curve in S through P . Since C is in S , the equation
F [x(t), y(t), z(t)] = 0

is valid for all t in α ≤ t ≤ β . If F (x, y, z) has continuous first partial derivatives, and x(t) ,
y(t) , and z(t) are all differentiable, we may differentiate this equation using the chain rule:
∂F dx ∂F dy ∂F dz
+ + = 0.
∂x dt ∂y dt ∂z dt
This equation, which holds at all points on C , and in particular at P , can be expressed vectorially
as ' ( ' (
∂F ∂F ∂F dx dy dz dr
0 = , , · , , = ∇F · .
∂x ∂y ∂z dt dt dt dt
But if the scalar product of two vectors vanishes, the vectors are perpendicular (see equation
11.25). Consequently, ∇F is perpendicular to the tangent vector d r/dt to C at P . Since C
is an arbitrary curve in S , it follows that ∇F at P is perpendicular to the tangent line to every
curve C in S at P . In other words, ∇F at P must be perpendicular to the tangent plane to S
FIGURE 12.22 Tangent plane at a FIGURE 12.23 The gradient of the

point P on a surface S contains all tangent function defining a surface is perpendicular to
vectors at P to curves in S the tangent plane to the surface
z z Tangent
plane
S
S
∇F
x y y
x
at P (Figure 12.23). If the coordinates of P are (x0 , y0 , z0 ) , then the equation of the tangent
plane to S at P is
0 = ∇F|P · (x − x0 , y − y0 , z − z0 )
= Fx (x0 , y0 , z0 )(x − x0 ) + Fy (x0 , y0 , z0 )(y − y0 )

+ Fz (x0 , y0 , z0 )(z − z0 ) (12.52)
(see equation 11.34).
EXAMPLE 12.27
Find the equation of the tangent plane to the surface xyz3 + yz2 = 4 at the point (1, 2, 1) .
SOLUTION A vector perpendicular to the tangent plane is
∇(xyz3 + yz2 − 4)|(1,2,1) = (yz3 , xz3 + z2 , 3xyz2 + 2yz)|(1,2,1) = (2, 2, 10).
But then the vector (1, 1, 5) must also be perpendicular to the tangent plane, and the equation
of the plane is therefore
0 = (1, 1, 5) · (x − 1, y − 2, z − 1) = x + y + 5z − 8.
We have shown in this section that if the equation F (x, y, z) = 0 defines a surface S , and
if there is a tangent plane to S at a point P , then the vector ∇F|P is normal to the tangent
plane (Figure 12.23). It is customary to state in this situation that ∇F|P is normal to the surface
itself at P , rather than to the tangent plane to the surface. This fact proves to be another of the
important properties of the gradient vector, and is worth stating as a theorem.
THEOREM 12.6
If the equation F (x, y, z) = 0 defines a surface S , and F (x, y, z) has continuous first
partial derivatives, then at any point on S the vector ∇F is perpendicular to S .
A geometric application of this fact is contained in the following example.
EXAMPLE 12.28
Find equations for the tangent line at the point (1, 2, 2) to the curve C : x 2 + y 2 + z2 = 9,
4(x 2 + y 2 ) = 5z2 .
SOLUTION Equation 12.49 indicates that to find a tangent vector to C , we should first have
parametric equations for C . These can be obtained by first solving each equation for x 2 + y 2
and equating the results:
9 − z2 = 5z2 /4.
This equation implies that z = ±2, the positive result being required here. On C , then,
x 2 + y 2 = 5, and parametric equations for C are
√ √
x = 5 cos t, y = 5 sin t, z = 2, 0 ≤ t < 2π.
According to 12.49, a tangent vector to C at (1, 2, 2) is

' (
dx dy dz // √ √
, , / = (− 5 sin t, 5 cos t, 0)|t=Sin−1 (2/√5) = (−2, 1, 0).
dt dt dt (1,2,2)
The tangent line therefore has equations
x−1 y−2
= , z = 2, or x + 2y = 5, z = 2.
−2 1
The fact that gradients can be used to find normals to surfaces suggests an alternative
solution. It is clear from Figure 12.24 that if we define F (x, y, z) = x 2 + y 2 + z2 − 9, then
FIGURE 12.24 Cross product of gradients of sphere and

cone yields a vector tangent to the curve of intersection
z
∇F
(1, 2, 2)
∇F × ∇G
G = 4(x 2 + y 2 ) − 5z 2 = 0 F = x 2 + y2 + z 2 − 9 = 0
∇G
x
∇F evaluated at (1, 2, 2) is perpendicular not only to the surface F (x, y, z) = 0, but also
to the curve C . Similarly, if G(x, y, z) = 4(x 2 + y 2 ) − 5z2 , then ∇G at (1, 2, 2) is also
perpendicular to C . Since a vector along the tangent line to C at (1, 2, 2) is perpendicular to
both of these vectors, it follows that a vector along the tangent line is
(∇F × ∇G)|(1,2,2) = [(2x, 2y, 2z) × (8x, 8y, −10z)]|(1,2,2)

= (2, 4, 4) × (8, 16, −20)
/ /
/ î ĵ k̂ /
/ /
= 8/1 2 2/
/ 2 4 −5 /
= 8(−18, 9, 0)
= 72(−2, 1, 0).
Once again, we have obtained (−2, 1, 0) as a tangent vector to the curve, and equations for the
tangent line can be written down as before.
Example 12.28 illustrates that when a curve is defined as the intersection of two surfaces
F (x, y, z) = 0, G(x, y, z) = 0 (Figure 12.25), then a vector tangent to the curve is
T = ∇F × ∇G. (12.53)
Thus to find a tangent vector to a curve we use 12.49 when the curve is defined parametrically.
When the curve is defined as the intersection of two surfaces, we can either find parametric
equations and use 12.49, or use 12.53. Note, too, that in order to find tangent lines to curves, it
is not necessary to have a direction assigned to the curves.
FIGURE 12.25 Cross product of gradients of two surfaces FIGURE 12.26 Unit
yields a vector tangent to their curve of intersection normal vector to a surface
z z Tangent
plane to S
∇F × ∇G n
∇F
n
∇G
S
G (x, y, z) = 0
F (x, y, z) = 0
x y
x y
At each point on a surface S at which S has a tangent plane (Figure 12.26), we have defined
a normal vector to S as a vector normal to the tangent plane to S . If we denote by n̂ a unit
normal vector to S , then the direction of n̂ clearly varies as we move from point to point on
S . We say that n̂ is a function of position (x, y, z) on S . Furthermore, at each point at which
S has a unit normal vector, it has two such vectors, one in the opposite direction to the other.
We say that a surface S is a smooth surface if it can be assigned a unit normal n̂ that varies
continuously on S . What this means geometrically is that for small changes in position, the unit
normal n̂ will undergo small changes in direction. The sphere in Figure 12.27 is smooth, as is
the paraboloid in Figure 12.28.
FIGURE 12.27 A sphere FIGURE 12.28 A paraboloid FIGURE 12.29 The surface of the
is a smooth surface is a smooth surface cylinder is a piecewise-smooth surface
z z z
x2 + y2 + z2 = R2
z=1
z = x 2 + y2 x2 + y2 = 1
y
y y
x
x x
z = −1
The surface bounding the cylindrical volume in Figure 12.29 is not smooth; a unit normal
that varies continuously over the surface cannot be assigned at points on the circles x 2 +y 2 = 1,
z = ±1. This surface can, however, be divided into a finite number of subsurfaces, each of
which is smooth. In particular, we choose the three subsurfaces S1 : z = 1, x 2 + y 2 ≤ 1;
S2 : z = −1, x 2 + y 2 ≤ 1; S3 : x 2 + y 2 = 1, −1 < z < 1. Such a surface is said to be a
piecewise-smooth surface.
EXERCISES 12.9
In Exercises 1–20 find equations for the tangent line to the curve at the In Exercises 21–26 find an equation for the tangent plane to the surface
point. at the point.
! √
1. y = x 2 , z = 0 at (−2, 4, 0) 21. z = x 2 + y 2 at (1, 1, 2)
2. x = t, y = t 2 , z = t 3 at (1, 1, 1) 22. x = x 2 − y 3 z at (2, −1, −2)
3. x = cos t, y = sin t, z = cos t at (1, 0, 1) 23. x 2 y + y 2 z + z2 x + 3 = 0 at (2, −1, −1)
2 24. x + y + z = 4 at (1, 1, 2)
4. y = x , z = x at (−2, 4, −2)
25. x = y sin (π z/2) at (−1, −1, 1)
5. x 2 = y, z + x = y at (1, 1, 0)
26. x 2 + y 2 + 2y = 1 at (1, 0, 3)
6. x = 2 − t 2 , y = 3 + 2t, z = t at (1, 5, 1)
√ √ ∗ 27. Show that the curve x = 2(t 3 + 2)/3, y = 2t 2 , z = 3t − 2
7. x = 2 cos t, y = 3 sin t, z = 5 at ( 2, −3/ 2, 5)
intersects the surface x 2 + 2y 2 + 3z2 = 15 at right angles at the point
8. x 2 y 3 + xy = 68 at (1, 4) (2, 2, 1) .
9. x + y + z = 4, x − y = 2 at (0, −2, 6) ∗ 28. Verify that the curve x 2 − y 2 + z2 = 1, xy + xz = 2 is tangent
to the surface xyz − x 2 − 6y + 6 = 0 at the point (1, 1, 1) .
10. x = e−t cos t, y = e−t sin t, z = t at (1, 0, 0)
∗ 29. Show that the equation of the tangent plane to a surface S : z =
11. x = t 2 + 1, y = 2t − 4, z = t 3 + 3 at (2, −6, 2) f (x, y) at a point (x0 , y0 , z0 ) on S can be written in the form
√
12. y 2 + z2 = 6, x + z = 1 at (2, − 5, −1) z − z0 = (x − x0 )fx (x0 , y0 ) + (y − y0 )fy (x0 , y0 ).
2 2 2 2 2 2
√
13. x + y + z = 4, z = x + y at (1, 1, − 2)
√ √
14. x = t, y = 1, z = 1 + t 2 at (4, 1, 17) In Exercises 30–32 find the indicated derivative for the function.
√
15. x = 1 + cos t, y = 2 − sin t, z = 4 + t at (2, 2, 2) ∗ 30. f (x, y, z) = 2x 2 + y 2 z2 at (3, 1, 0) with respect to distance
along the curve x + y + z = 4, x − y + z = 2 in the direction of
16. x = z2 + z3 , y = z − z4 at (12, −14, 2)
increasing x
17. x = y 2 + 3y 3 − 2y + 5, z = 0 at (7, 1, 0) ∗ 31. f (x, y, z) = xyz + xy + xz + yz at (1, −2, 5) perpendicular
18. 2x 2 + y 2 + 2y = 3, z = x + 1 at (0, 1, 1) to the surface z = x 2 + y 2
√ √ ∗ 32. f (x, y, z) = x 2 + y 2 − z2 at (3, 4, 5) with respect to distance
19. x = t 2 , y = t, z = t + t 4 at (1, 1, 2) along the curve x 2 + y 2 − z2 = 0, 2x 2 + 2y 2 − z2 = 25 in the
20. x = t sin t, y = t cos t, z = 2t at (0, 2π, 4π ) direction of decreasing x
∗ 33. If F (x, y) = 0 defines a curve implicitly in the xy -plane, prove equations for the tangent line to C at P can be written in the form
that at any point on the curve ∇F is perpendicular to the curve.
x − x0 y − y0 z − z0
∗ 34. Find the equation of the tangent plane to the ellipsoid x 2 /a 2 + = = ,
∂(F, G, H ) // ∂(F, G, H ) // ∂(F, G, H ) //
y 2 /b2 + z2 /c2 = 1 at any point (x0 , y0 , z0 ) on the surface. ∂(t, y, z) / P ∂(x, t, z) / P ∂(x, y, t) / P
∗ 35. Find all points on the surface z = x 2 /4 − y 2 /9 at which the
provided that none of the Jacobians vanishes.
tangent plane is parallel to the plane x + y + z = 4.
∗ 38. Find all points on the paraboloid z = x 2 + y 2 − 1 at which the
∗ 36. Find all points on the surface z2 = 4(x 2 +y 2 ) at which the tangent normal to the surface coincides with the line joining the origin to the
plane is parallel to the plane x − y + 2z = 3. point.
∗ 37. Suppose that the equations F (x, y, z, t) = 0, G(x, y, z, t) = 0, ∗∗ 39. Show that the sum of the intercepts on the x -, y -, and z -axes of
√ √ √ √
H (x, y, z, t) = 0 implicitly define parametric equations for a curve the tangent plane to the surface x + y + z = a at any point
C ( t being the parameter). If P (x0 , y0 , z0 ) is a point on C , show that is a .
12.10 Relative Maxima and Minima

We now study relative extrema of functions of more than one independent variable. Most of the
discussion will be confined to functions f (x, y) of two independent variables because we can
discuss the concepts geometrically as well as algebraically. Unfortunately, not all results are
easily extended to functions of more than two independent variables, and we will therefore be
careful to point out these limitations.
Before beginning the discussion, we briefly review maxima–minima results for functions
f (x) of one variable. We do this because maxima–minima theory for multivariable functions is
essentially the same as that for single-variable functions. In fact, every definition that we make
and every result that we discuss in this section has its counterpart in single-variable theory.
Hence, a synopsis of single-variable results is in order. Unfortunately, proving results in the
multivariable case is considerably more complicated than in the single-variable case, but if we
can keep central ideas foremost in our minds and constantly make comparisons with single-
variable calculus, we will find that discussions are not nearly as difficult as they might otherwise
be.
Critical points of a function f (x) are points at which f + (x) is either equal to zero or does
not exist. Geometrically, this means points at which the graph of f (x) has a horizontal tangent
line, a vertical tangent line, or no tangent line at all. Critical points for continuous functions
can yield relative maxima, relative minima, horizontal points of inflection, vertical points of
inflection, or just corners. There are two tests to determine whether a critical point x0 yields a
relative maximum or a relative minimum. The first-derivative test states that if f + (x) changes
from a positive quantity to a negative quantity as x increases through x0 , then x0 gives a relative
maximum; if f + (x) changes from negative to positive, then a relative minimum is obtained.
The second-derivative test indicates the nature of a critical point at which f + (x0 ) = 0 whenever
f ++ (x0 ) (= 0. If f ++ (x0 ) > 0, then a relative minimum is obtained, and if f ++ (x0 ) < 0, a
relative maximum is found.
We begin our study of extrema theory for multivariable functions by defining critical points
for functions of two independent variables.
DEFINITION 12.7
A point (x0 , y0 ) in the domain of a function f (x, y) is said to be a critical point of
f (x, y) if
∂f // ∂f //
/ = 0, / =0 (12.54)
∂x (x0 ,y0 ) ∂y (x0 ,y0 )
or if one (or both) of these partial derivatives does not exist at (x0 , y0 ) .
There are two ways to interpret critical points of f (x, y) geometrically. In Section 12.3,
we interpreted ∂f/∂x at (x0 , y0 ) as the slope of the tangent line to the curve of intersection of
the surface z = f (x, y) and the plane y = y0 , and ∂f/∂y as the slope of the tangent line to
the curve of intersection with x = x0 . It follows, then, that (x0 , y0 ) is critical if both curves
have horizontal tangent lines or if either curve has a vertical tangent line or no tangent line at
all. Alternatively, recall that the equation of the tangent plane to the surface z = f (x, y) at
(x0 , y0 ) is
z − z0 = fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 )
(see Exercise 29 in Section 12.9). If both partial derivatives vanish, then the tangent plane is
horizontal with equation z = z0 . For example, at each of the critical points in Figures 12.30–
12.33, ∂f/∂x = ∂f/∂y = 0 and the tangent plane is horizontal. The remaining functions in
Figures 12.34–12.38 have critical points at which either ∂f/∂x or ∂f/∂y or both do not exist.
In Figures 12.34–12.37, the surfaces do not have tangent planes at critical points, and in Figure
12.38, the tangent plane is vertical at each critical point. Consequently, (x0 , y0 ) is a critical
point of a function f (x, y) if at (x0 , y0 ) the surface z = f (x, y) has a horizontal tangent
plane, a vertical tangent plane, or no tangent plane at all.
FIGURE 12.30 Tangent plane FIGURE 12.31 Tangent plane FIGURE 12.32 Tangent plane hori-
horizontal at critical point (0, 0) horizontal at critical point (0, 1) zontal at critical point (0, 0)
z z z
z = f (x, y)
= 1 − x 2 − (y − 1) 2
z = f (x, y)
= x2 + y2 z = f (x, y)
(0, 1) 1 = 1 + x2 − y 2
2 y
Critical point y x y Critical point
x at which Critical point
at which at which
fx = fy = 0 fx = f y = 0
fx = f y = 0 x
FIGURE 12.33 Tangent plane FIGURE 12.34 No tangent plane FIGURE 12.35 No tangent plane at
horizontal at critical points (x, 0) at critical point (1, 0) critical point (x, x)
z z z
z = f (x, y)
z = f (x, y)
= y3 = (x − 1) 2 + y 2 z = f (x, y)
= |x − y|
Critical
points
y at which
fx and fy y
Critical do not
points exist
at which
x fx = fy = 0 y x
(1, 0)
Critical point at which

fx and fy do not exist
x
FIGURE 12.36 No tangent FIGURE 12.37 No tangent plane FIGURE 12.38 Vertical tangent
plane at critical point (0, 0) at critical point (0, 0) planes at critical points (x, 1)
z = f (x, y) z z z
= |x| + |y| z = f (x, y) z = f (x, y) = (y − 1)1/3
= 1 − (x 2 + y 2 )1/3
x
y Critical y
y
Critical points points
x Critical point
x at which at which fx = 0
at which
fx and fy do but fy does
fx and fy do not exist
not exist not exist
EXAMPLE 12.29
Find all critical points for the function
f (x, y) = x 2 y − 2xy 2 + 3xy + 4.
SOLUTION For critical points, we first solve
∂f
0 = = 2xy − 2y 2 + 3y = y(2x − 2y + 3),
∂x
∂f
0 = = x 2 − 4xy + 3x = x(x − 4y + 3).
∂y
To satisfy these two equations simultaneously, there are four possibilities:
1. x = 0, y = 0, which gives the critical point (0, 0) ;
2. y = 0, x − 4y + 3 = 0, which gives the critical point (−3, 0) ;
3. x = 0, 2x − 2y + 3 = 0, which gives the critical point (0, 3/2) ;
4. 2x − 2y + 3 = 0, x − 4y + 3 = 0, which gives the critical point (−1, 1/2) .
Since ∂f/∂x and ∂f/∂y are defined for all x and y , these are the only critical points. The plot
of the surface z = f (x, y) for −10 ≤ x ≤ 10, −10 ≤ y ≤ 10 in Figure 12.39 does not
really illustrate the critical points. In other words, computer plots of functions of two variables
are not as helpful in determining critical points of functions as were plots of functions of one
variable in Chapter 4.
FIGURE 12.39 Computer plot of z = x 2 y − 2xy 2 + 3xy + 4
x
Critical points for functions of more than two independent variables can be defined algebraically,
but because we have no geometric representation for such functions, there is no geometric
interpretation for their critical points. For example, if f (x, y, z, t) is a function of independent
variables x , y , z , and t , then (x0 , y0 , z0 , t0 ) is a critical point of f (x, y, z, t) if all four of its
first-order partial derivatives vanish at (x0 , y0 , z0 , t0 ) ,
∂f // ∂f // ∂f // ∂f //
/ = / = / = / = 0, (12.55)
∂x (x0 ,y0 ,z0 ,t0 ) ∂y (x0 ,y0 ,z0 ,t0 ) ∂z (x0 ,y0 ,z0 ,t0 ) ∂t (x0 ,y0 ,z0 ,t0 )
or if at least one of the partial derivatives does not exist at the point. Note that because the partial
derivatives of a function are the components of the gradient of the function, we can say that a
critical point of a function is a point at which its gradient is either equal to zero or undefined.
EXAMPLE 12.30
Find all critical points for the function
!
f (x, y, z) = xyz x 2 + y 2 + z2 .
SOLUTION For critical points, we consider the equations
∂f ! x 2 yz
0 = = yz x 2 + y 2 + z2 + !
∂x x 2 + y 2 + z2
yz
= ! (2x 2 + y 2 + z2 ),
x 2 + y 2 + z2
∂f ! xy 2 z
0 = = xz x 2 + y 2 + z2 + !
∂y x 2 + y 2 + z2
xz
= ! (x 2 + 2y 2 + z2 ),
2 2
x +y +z 2
∂f ! xyz2
0 = = xy x 2 + y 2 + z2 + !
∂z x 2 + y 2 + z2
xy
= ! (x 2 + y 2 + 2z2 ).
x + y 2 + z2
2
The partial derivatives are clearly undefined for x = y = z = 0, and therefore the origin
(0, 0, 0) is a critical point. If x , y , and z are not all zero, then the terms in parentheses cannot
vanish, and we must set
yz = 0, xz = 0, xy = 0.
If any two of x , y , and z vanish, but the third does not, then these equations are satisfied. In
other words, every point on the x -axis, every point on the y -axis, and every point on the z -axis
is critical.
We now turn our attention to the classification of critical points of a function f (x, y) of two
independent variables. Critical points (0, 1) in Figure 12.31 and (0, 0) in Figure 12.37 yield
“high” points on the surfaces. We describe this property in the following definition.
DEFINITION 12.8
A function f (x, y) is said to have a relative maximum f (x0 , y0 ) at a point (x0 , y0 )
if there exists a circle in the xy -plane centred at (x0 , y0 ) such that for all points (x, y)
inside this circle
f (x, y) ≤ f (x0 , y0 ). (12.56)
The “low” points on the surfaces at (0, 0) in Figure 12.30, (1, 0) in Figure 12.34, and (0, 0)
in Figure 12.36 are relative minima according to the following.
DEFINITION 12.9
A function f (x, y) is said to have a relative minimum f (x0 , y0 ) at a point (x0 , y0 )
if there exists a circle in the xy -plane centred at (x0 , y0 ) such that for all points (x, y)
inside this circle
f (x, y) ≥ f (x0 , y0 ). (12.57)
Every critical point in Figure 12.35 yields a relative minimum of f (x, x) = 0.
DEFINITION 12.10
If a critical point of a function f (x, y) at which ∂f/∂x = ∂f/∂y = 0 yields neither a
relative maximum nor a relative minimum, it is said to yield a saddle point.
The critical point (0, 0) in Figure 12.32 therefore gives a saddle point, as does each of the
critical points in Figure 12.33. Saddle points for surfaces z = f (x, y) are clearly the analogues
of horizontal points of inflection for curves y = f (x) . In both cases the derivative(s) of the
function vanishes but there is neither a relative maximum nor a relative minimum.
The critical points in Figure 12.36 [except (0, 0) ] are the counterparts of corners for the
graph of a function f (x) . They are points at which one or both of the partial derivatives of
f (x, y) do not exist, but like corners for f (x) , they do not necessarily yield relative extrema.
Critical points in Figure 12.38 are the analogues of vertical points of inflection for a function
f (x) .
Our discussion has made it clear that:
(a) At a relative maximum or minimum of f (x, y) , either ∂f/∂x and ∂f/∂y both
vanish, or one or both of the partial derivatives do not exist.
(b) Saddle points may also occur where ∂f/∂x = ∂f/∂y = 0, and points where the
derivatives do not exist may fail to yield relative extrema.
In other words, every relative extremum of f (x, y) occurs at a critical point, but critical points
do not always give relative extrema.
Given the problem of determining all relative maxima and minima of a function f (x, y) ,
we should first find its critical points. But how do we decide whether these critical points yield
relative maxima, relative minima, saddle points, or none of these? We do not have a practical
test that is equivalent to the first-derivative test for functions of one variable, but we do have a test
that corresponds to the second-derivative test. For functions of two independent variables the
situation is more complicated, however, since there are three second-order partial derivatives,
but the idea of the test is essentially the same. It determines whether certain curves are concave
upward or concave downward at the critical point. The complete result is contained in the
following theorem.
THEOREM 12.7
Suppose (x0 , y0 ) is a critical point of f (x, y) at which ∂f/∂x and ∂f/∂y both vanish.
Suppose further that fx , fy , fxx , fxy , and fyy are all continuous at (x0 , y0 ) . Define
A = fxx (x0 , y0 ), B = fxy (x0 , y0 ), C = fyy (x0 , y0 ).
If:
(i) B 2 − AC < 0 and A < 0, then f (x, y) has a relative maximum at (x0 , y0 ) ;
(ii) B 2 − AC < 0 and A > 0, then f (x, y) has a relative minimum at (x0 , y0 ) ;
(iii) B 2 − AC > 0, then f (x, y) has a saddle point at (x0 , y0 ) ;
(iv) B 2 − AC = 0, no conclusion can be made.
PROOF
(i) Suppose we intersect the surface z = f (x, y) with a plane parallel to the z -axis,
through the point (x0 , y0 , 0) , and making an angle α with the line through (x0 , y0 , 0)
parallel to the positive x -axis (Figure 12.40). The slope of the curve of intersection of
these surfaces at the point (x0 , y0 , f (x0 , y0 )) is given by the directional derivative
∂f // ∂f //
Dv f|(x0 ,y0 ) = / cos α + sin α
∂x (x0 ,y0 ) ∂y / (x0 ,y0 )
(see equation 12.44). Since (x0 , y0 ) is a critical point at which ∇f = 0, it follows

that
Dv f|(x0 ,y0 ) = 0 for all α.
In Figure 12.40 we have illustrated the critical point as a relative maximum.
But how do we verify that this is indeed the case? If we can show that each and
every curve of intersection of the surface with a vertical plane through (x0 , y0 , 0) is
concave downward at (x0 , y0 , 0) , then (x0 , y0 ) must give a relative maximum. But
to discuss concavity of a curve, we require the second derivative — in this case, the
second directional derivative of f (x, y) . According to equation 12.45, the second
directional derivative of f (x, y) at (x0 , y0 ) in the direction v̂ = (cos α, sin α) is
∂ 2 f // 2 ∂ 2 f // ∂ 2 f //
Dv (Dv f ) = / cos α + 2 / cos α sin α + / sin2 α
∂x 2 (x0 ,y0 ) ∂x∂y (x0 ,y0 ) ∂y 2 (x0 ,y0 )
= A cos2 α + 2B cos α sin α + C sin2 α,
FIGURE 12.40 Critical point yielding a relative maximum for the function
z z = f (x, y)
(x0, y0, f (x0, y0))
x y
(x0, y0, 0)
v
where we understand that here Dv (Dv f ) is implicitly suffixed by (x0 , y0 ) . In order,

therefore, to verify that (x0 , y0 ) gives a relative maximum, it is sufficient to show that
Dv (Dv f ) is negative for each value of α in the interval 0 ≤ α < 2π . However,
because Dv (Dv f ) is unchanged if α is replaced by α + π , it is sufficient to verify
that Dv (Dv f ) is negative for 0 ≤ α < π .
For any of these values of α except π/2, we can write
Dv (Dv f ) = cos2 α(A + 2B tan α + C tan2 α),
and if we set u = tan α ,
Dv (Dv f ) = cos2 α(A + 2Bu + Cu2 ).
It is evident that Dv (Dv f ) < 0 for all α (= π/2 if and only if
Q(u) = A + 2Bu + Cu2 < 0 for −∞ < u < ∞.
Were we to draw a graph of the quadratic Q(u) , we would see that it crosses the
u -axis where
√ √
−2 B ± 4B 2 − 4AC −B ± B 2 − AC
u= = .
2C C
But because B 2 − AC < 0, there are no real solutions of this equation, and therefore
Q(u) never crosses the u -axis. Since Q(0) = A < 0, it follows that Q(u) < 0
for all u . We have shown, then, that
Dv (Dv f ) < 0 for all α (= π/2.
When α = π/2, Dv (Dv f ) = C . Since B 2 − AC < 0 and A < 0, it follows that

C < 0 also. Consequently, if B 2 − AC < 0 and A < 0, then Dv (Dv f ) < 0 for
all α , and (x0 , y0 ) yields a relative maximum.
(ii) If B 2 −AC < 0 and A > 0, a similar argument leads to the conclusion that (x0 , y0 )
yields a relative minimum; the only difference is that inequalities are reversed.
(iii) If B 2 − AC > 0, then Q(u) has real distinct zeros, in which case Q(u) is some-
times negative and sometimes positive. This means that the curve of intersection is
sometimes concave upward and sometimes concave downward, and the point (x0 , y0 )
therefore gives a saddle point.
(iv) If B 2 − AC = 0, the classification of the point determined by (x0 , y0 ) depends on
which of A , B , and C vanish, if any.
To illustrate that we can obtain a relative maximum, a relative minimum, or a saddle point
for a critical point at which B 2 − AC = 0, consider the three functions f (x, y) = −y 2 ,
f (x, y) = y 2 , and f (x, y) = y 3 in Figures 12.41–12.43. The point (0, 0) is a critical point
for each function, and at this point B 2 − AC = 0. Yet (0, 0) yields a relative maximum for
f (x, y) = −y 2 , a relative minimum for f (x, y) = y 2 , and a saddle point for f (x, y) = y 3 .
In fact, every point on the x -axis is a relative maximum for f (x, y) = −y 2 , a relative minimum
for f (x, y) = y 2 , and a saddle point for f (x, y) = y 3 .
FIGURE 12.41 Critical FIGURE 12.42 Critical FIGURE 12.43 Critical

points yield relative maxima points yield relative minima points yield saddle points
z z z
y y y
z = y2
z = y3
z= −y 2
x x x
EXAMPLE 12.31
Find and classify critical points for each of the following functions as yielding relative
maxima, relative minima, saddle points, or none of these:
(a) f (x, y) = 4xy − x 4 − y 4
(b) f (x, y) = x 4 y 3
SOLUTION
(a) Critical points of f (x, y) are given by
∂f ∂f
0 = = 4y − 4x 3 , 0 = = 4x − 4y 3 .
∂x ∂y
Solutions of these equations are (0, 0) , (1, 1) , and (−1, −1) . We now calculate
∂ 2f ∂ 2f ∂ 2f
= −12x 2 , = 4, = −12y 2 .
∂x 2 ∂x∂y ∂y 2
We could tabulate results to determine the nature of the critical points.
TABLE 12.1
Critical point A B C B 2 − AC Nature
(0 , 0 ) 0 4 0 16 Saddle point
FIGURE 12.44 Diagram to show
sign of function f (x, y) = x 4 y 3
(1, 1) −12 4 −12 −128 Relative maximum
y (−1, −1) −12 4 −12 −128 Relative maximum

+ 0 +
(b) For critical points we solve
0 0 ∂f ∂f
x 0 = = 4x 3 y 3 , 0 = = 3x 4 y 2 .
∂x ∂y
− 0 −
Every point on the x - and y -axes is critical, and at each of these points f (x, y) = 0.
The second-derivative test fails to classify these critical points. Figure 12.44 shows a
value of zero for the function on the axes and the sign of f (x, y) in the four quadrants.
It implies that the points (0, y) for y > 0 yield relative minima; (0, y) for y < 0
yield relative maxima; and (x, 0) yield saddle points.
EXAMPLE 12.32
It is straightforward to verify that (0, 0) is the only critical point of the function f (x, y) =
x 2 − 6xy 2 + y 4 and that the quantity B 2 − AC of Theorem 12.7 is equal to zero at this critical
point. Show graphically and algebraically that the critical point gives a saddle point.
SOLUTION The value of the function at (0, 0) is f (0, 0) = 0. There is a relative minimum at
(0, 0) if f (x, y) ≥ 0 in some circle around (0, 0) ; there is a relative maximum if f (x, y) ≤ 0
in some such circle; and there is a saddle point if f (x, y) takes on negative and positive values
in every circle centred at (0, 0) . The plot in Figure 12.45 indicates that the last situation prevails.
To show this algebraically, we first note that values of f (x, y) are positive along the x -axis
and the y -axis away from the origin. On the parabola x = y 2 , values of the function are
(y 2 )2 − 6(y 2 )y 2 + y 4 = −4y 4 ≤ 0.
Thus, (0, 0) yields a saddle point.
FIGURE 12.45 For x 2 − 6xy 2 + y 4 , (0, 0) yields a saddle point
This completes our discussion of relative extrema of functions of two independent variables.
Our next step should be to extend the theory to functions of more than two variables. It is a
simple matter to give definitions of relative maxima and minima for such functions; they are
almost identical to Definitions 12.8 and 12.9 (see Exercise 19). On the other hand, to develop
a theorem for functions of more than two independent variables that is analogous to Theorem
12.7 is beyond the scope of this book. We refer the interested reader to more advanced books.
EXERCISES 12.10
2 +y 2 )
In Exercises 1–14 find all critical points for the function and classify 7. f (x, y) = xye−(x
each as yielding a relative maximum, a relative minimum, a saddle
point, or none of these. 8. f (x, y) = x 2 − 2xy + y 2
9. f (x, y) = (x 2 + y 2 )2/3
1. f (x, y) = x 2 + 2xy + 2y 2 − 6y
10. f (x, y) = x 4 y 3
2. f (x, y) = 3xy − x 3 − y 3
3. f (x, y) = x 3 − 3x + y 2 + 2y 11. f (x, y) = 2xy 2 + 3xy + x 2 y 3
4. f (x, y) = x 2 y 2 + 3x ∗ 12. f (x, y) = |x| + y 2

5. f (x, y) = xy − x 2 + y 2 ∗ 13. f (x, y) = (1 − x)(1 − y)(x + y − 1)
6. f (x, y) = x sin y ∗ 14. f (x, y) = x 4 + y 4 − x 2 − y 2 + 1
In Exercises 15–18 find all critical points for the function. point in D at which either fxx or fxy does not vanish.
∗ 21. Find and classify the critical points for the function f (x, y) =
15. f (x, y, z) = x 2 + y 2 − z2 + 3x − 2y + 5 y 2 − 4x 2 y + 3x 4 .
∗ 22. Find and classify the critical points of f (x, y) = x 4 + 3xy 2 + y 2
16. f (x, y, z, t) = x 2 y 2 z2 + t 2 x 2 + 3x as yielding relative maxima, relative minima, or saddle points.
17. f (x, y, z) = xyz + x 2 yz − y ∗ 23. The equation 2x 2 + 3y 2 + z2 − 12xy + 4xz = 35 defines

function z = f (x, y) . Show that the point x = 1 and y = 2 is a
18. f (x, y, z) = xyzex
2 +y 2 +z2 critical point for the function with value 5 at (1, 2) . Does it yield a
relative extremum for the function?
∗ 19. Give definitions for a relative maximum and a relative minimum ∗∗ 24. (a) Show that the function f (x, y, z) = x 2 + y 2 + z2 − xyz
for a function f (x, y, z) at a point (x0 , y0 , z0 ) . has a critical point (0, 0, 0) . What are the other critical
points?
∗ 20. Suppose that f (x, y) is harmonic in the region D : x 2 +y 2 < 1. (b) Use the definition in Exercise 19 to show that f (x, y, z)
Show that f (x, y) cannot have a relative maximum or minimum at any has a relative minimum at (0, 0, 0) .
12.11 Absolute Maxima and Minima

Absolute maxima and minima are more important than relative maxima and minima when it
comes to applications. In this section and in Section 12.12 we discuss the theory of absolute
extrema and consider a number of applications. Once again we begin with functions f (x, y)
of two independent variables and base our discussion on the theory of absolute extrema for
functions of one variable.
We learned in Section 4.7 that a function f (x) that is continuous on a finite interval
a ≤ x ≤ b must have an absolute maximum and an absolute minimum on that interval.
Furthermore, these absolute extrema must occur at either critical points or at the ends x = a
and x = b of the interval. Consequently, to find the absolute extrema of a function f (x) , we
evaluate f (x) at all critical points, at x = a , and at x = b ; the largest of these numbers is the
absolute maximum of f (x) on a ≤ x ≤ b , and the smallest is the absolute minimum.
The procedure is much the same for a function f (x, y) that is continuous on a region R
that is finite and includes all the points on its boundary. First, however, we define exactly what
we mean by absolute extrema of f (x, y) and consider a number of simple examples. We will
then be able to make general statements about the nature of all absolute extrema, and proceed
to the important area of applications.
DEFINITION 12.11
The absolute maximum of a function f (x, y) on a region R is f (x0 , y0 ) if (x0 , y0 ) is
in R and
f (x, y) ≤ f (x0 , y0 ) (12.58)
for all (x, y) in R . The absolute minimum of f (x, y) on R is f (x0 , y0 ) if (x0 , y0 ) is
in R and
f (x, y) ≥ f (x0 , y0 ) (12.59)
for all (x, y) in R .
In Figures 12.46–12.51, we have shown six functions defined on the circle R : x 2 +y 2 ≤ 1.

The absolute maxima and minima of these functions for this region are shown in Table 12.2.
FIGURE 12.46 Absolute maximum FIGURE 12.47 Absolute maximum FIGURE 12.48 Absolute maxi-
= 1, Absolute minimum = 0 = 2, Absolute minimum = 0 mum = 2, Absolute minimum = 0
z z z
z = x 2 + y2
(−1, 0, 2)
(1, 0, 2)
(−1, 0, 2)
z= (x − 1) 2 + y 2
−1
1 −1
1 y
1 −1
y
x z = 1 + x 2 − y2 1
x 1 y
x
FIGURE 12.49 Absolute maximum FIGURE 12.50 Absolute maximum FIGURE 12.51 Absolute maxi-
√
= 4, Absolute minimum = 2 = 1, Absolute minimum = 0 mum = 2, Absolute minimum = 0
z z z
4 1
z = 4 − 2x 2 − y 2
z = 1 − (x 2 + y 2 )1/ 3
z= x−y
1
1 y
−1
x 1
1 y
1 x
1 y
x
TABLE 12.2
Value of Value of
Function Position of absolute Position of absolute
f (x, y) absolute maximum maximum absolute minimum minimum
Every point on
x2 + y2 x2 + y2 = 1 1 (0 , 0 ) 0
2 2
1+x −y (±1, 0) 2 (0, ±1) 0
!
(x − 1)2 + y2 (−1, 0) 2 (1, 0) 0
2 2
4 − 2x − y (0 , 0 ) 4 (±1, 0) 2
Every point on
1 − (x 2 + y 2 )1/3 (0 , 0 ) 1 x2 + y2 = 1 0
Every point on
√ √ √
|x − y| (±1/ 2, ∓1/ 2) , 2
√ y = x, √ 0
−1/ 2 ≤ x ≤ 1/ 2
For each of the functions in these figures, absolute extrema occur at either a critical point or
a point on the boundary of R . This result is true for any continuous function defined on a
finite region that includes all the points on its boundary. Although this result may seem fairly
obvious geometrically, to prove it analytically is very difficult; we will be content to assume its
validity and carry on from there.
Sometimes a drawing or plot of the surface defined by a function makes absolute maxima and
FIGURE 12.52 Absolute
extrema of a function continuous
minima clear. This may not always be the case, however, and we therefore turn our attention to
on a region R and its boundary determining absolute extrema algebraically. Suppose, then, that a continuous function f (x, y)
is given and we are required to find its absolute extrema on a finite region R (which includes
y its boundary points). The previous discussion indicated that the extrema must occur either at
C critical points or on the boundary of R . Consequently, we should first determine all critical
points of f (x, y) in R , and evaluate f (x, y) at each of these points. These values should now
R be compared to the maximum and minimum values of f (x, y) on the boundary of R . But how
do we find the maximum and minimum values of f (x, y) on the boundary? If the boundary
of R is denoted by C (Figure 12.52), and if C has parametric equations x = x(t) , y = y(t) ,
x
α ≤ t ≤ β , then on C we can express f (x, y) in terms of t , and t alone:
f [x(t), y(t)], α ≤ t ≤ β.
To find the maximum and minimum values of f (x, y) on C is now an absolute extrema problem
for a function of one variable. The function f [x(t), y(t)] should therefore be evaluated at each
of its critical points and at t = α and t = β . A plot of f [x(t), y(t)] could be valuable here.
If the boundary of R consists of a number of curves (Figure 12.53), then this boundary
FIGURE 12.53 Boundary
of a region may consist of more
procedure must be performed for each part. In other words, on each part of the boundary we
than one curve express f (x, y) as a function of one variable, and then evaluate this function at its critical points
and at the ends of that part of the boundary to which it applies.
y
The absolute maximum of f (x, y) on R is then the largest of all values of f (x, y) evaluated
at the critical points inside R , the critical points on the boundary of R , and the endpoints of
C3 each part of the boundary. The absolute minimum of f (x, y) on R is the smallest of all these
values.
R C2 Recall that to find the absolute extrema of a function f (x) , continuous on a ≤ x ≤ b ,
we evaluate f (x) at all critical points and at the boundary points x = a and x = b . The
x procedure that we have established here for f (x, y) is much the same — the difference is
C1 that, for f (x, y) , the boundary consists not of two points, but of entire curves. Evaluation of
f (x, y) on the boundary therefore reduces to one or more extrema problems for functions of
one variable. Note too that for f (x, y) [or f (x) ], it is not necessary to determine the nature
of the critical points; it is necessary only to evaluate f (x, y) at these points.
EXAMPLE 12.33
Find the maximum value of the function z = f (x, y) = 4xy − x 4 − 2y 2 on the region R :
−2 ≤ x ≤ 2, −2 ≤ y ≤ 2.
SOLUTION A plot of the function in Figure 12.54 suggests that the maximum value occurs
at a critical point in the first and third quadrants, or on the edges of the square. We confirm this
with the procedure outlined above. Critical points of f (x, y) are given by
∂f ∂f
0 = = 4y − 4x 3 , 0 = = 4x − 4y.
∂x ∂y
Solutions of these equations are (0, 0) , (1, 1) , and (−1, −1) , and the values of f (x, y) at
these critical points are
f (0 , 0 ) = 0 , f (1, 1) = 1 , f (−1, −1) = 1 .
We denote the four parts of the boundary of R by C1 , C2 , C3 , and C4 (Figure 12.55).

FIGURE 12.54 Computer plot of FIGURE 12.55 Region

4xy − x 4 − 2y 2 on −2 ≤ x ≤ 2, −2 ≤ y ≤ 2 and its four bounding curves
z y
C3 : y = 2
2
y
2
−2 x
C4 : x = −2 C2 : x = 2
x −2
C1 : y = −2
On C1 , y = −2, in which case
z = −8x − x 4 − 8, −2 ≤ x ≤ 2 .
For critical points of this function, we solve
dz
0 = = −8 − 4 x 3 .
dx
The only solution is x = −21/3 , at which the value of z is
z = 8 · 21/3 − 24/3 − 8 = −0.44 .
On C2 , x = 2, in which case
z = 8y − 16 − 2y 2 , −2 ≤ y ≤ 2 .
Critical points are defined by

dz
0 = = 8 − 4y.
dy
The only solution y = 2 defines one of the corners of the square, and at this point
z = −8 .
On C3 , y = 2 and
z = 8x − x 4 − 8, −2 ≤ x ≤ 2 .
dz
0 = = 8 − 4x 3 .
dx
At the single point x = 21/3 ,
z = 8 · 21/3 − 24/3 − 8 = −0.44 .
On the final curve C4 , x = −2 and
z = −8y − 16 − 2y 2 , −2 ≤ y ≤ 2 .
Critical points are given by

dz
0 = = −8 − 4y.
dy
The solution y = −2 defines another corner of the square at which
z = −8 .
We have now evaluated f (x, y) at all critical points inside R and at all critical points on the
four parts of the boundary of R . It remains only to evaluate f (x, y) at the corners of the square.
Two corners have already been accounted for; the other two give
f (2, −2) = −40 , f (−2, 2) = −40 .
The largest value of f (x, y) is the largest of the numbers in the boxes, namely, 1, and this is
therefore the maximum value of f (x, y) on R .
EXAMPLE 12.34
Temperature in degrees Celsius at each point (x, y) in a semicircular plate defined by x 2 +y 2 ≤
1, y ≥ 0 is given by
T (x, y) = 16x 2 − 24xy + 40y 2 .
Find the hottest and coldest points in the plate.
SOLUTION For critical points of T (x, y) , we solve
∂T ∂T
0 = = 32x − 24y, 0 = = −24x + 80y.
∂x ∂y
The only solution of these equations, (0, 0) , is on the boundary. On the upper edge of the plate
(Figure 12.56), we set x = cos t , y = sin t , 0 ≤ t ≤ π , in which case
T = 16 cos2 t − 24 cos t sin t + 40 sin2 t, 0 ≤ t ≤ π.
A plot of this function in Figure 12.57 shows the absolute maximum at the relative maximum
and the absolute minimum at the relative minimum. To locate them we solve for critical points
of this function:
dT
0 = = −32 cos t sin t − 24(− sin2 t + cos2 t) + 80 sin t cos t
dt
= 24(sin 2t − cos 2t).
FIGURE 12.56 Maximum and minimum FIGURE 12.57 Plot of temperature on semicircu-
temperatures on a semicircular domain lar part of boundary
y T
45
1 40
x2 + y2 = 1
35
30
25
−1 1 x 20
15
0.5 1 1.5 2 2.5 3 t

If we divide by cos 2t (since cos 2t = 0 does not lead to a solution of this equation), we
have
tan 2t = 1.
The only solutions of this equation in the interval 0 ≤ t ≤ π are t = π/8 and t = 5π/8.
When t = π/8, T = 11.0 ; and when t = 5π/8, T = 45.0 . At the ends of this part of the
boundary, t = 0 and t = π , and T (1, 0) = 16 and T (−1, 0) = 16 . On the lower edge
of the plate, y = 0, in which case
T = 16x 2 , −1 ≤ x ≤ 1.
The only critical point of this function is x = 0, at which T = 0 . The hottest point in the
plate is therefore (cos(5π/8), sin(5π/8)) = (−0.38, 0.92) , where the temperature is 45.0◦ C,
and the coldest point is (0, 0) with temperature 0◦ C.
EXAMPLE 12.35
Find the point on the first octant part of the plane 6x + 3y + 4z = 6 closest to the point (4, 6, 7) .
Assume that lines of intersection of the plane with the coordinate planes are part of the surface.
FIGURE 12.58 Minimum distance from (4, 6, 7) to plane 6x + 3y + 4z = 6 in first octant
z
(4, 6, 7)
3
2
D
(x, y, z)
1 2 y
SOLUTION The distance D from (4, 6, 7) to any point (x, y, z) in space is defined by
D 2 = (x − 4)2 + (y − 6)2 + (z − 7)2 .
We must minimize D , but consider only points (x, y, z) that satisfy the equation of the plane
and lie in the first octant (Figure 12.58). At the moment, D 2 is a function of three variables
x , y , and z , but they are not all independent because of the planar restriction. If we solve the
FIGURE 12.59 Triangular
domain for distance function
equation of the plane for z in terms of x and y , and substitute,
y
' (2
2 2 2 6 − 6x − 3y
D = f (x, y) = (x − 4) + (y − 6) + −7
2 4
' (2
6x + 3y + 22
C3 : 2x + y = 2 = (x − 4)2 + (y − 6)2 + ,
C1 4
where x and y are independent variables. Now D is minimized when D 2 is minimized, and
we therefore find the point (x, y) that minimizes D 2 . The values of x and y that yield points
C2 1 x on the first octant part of the plane are those in the triangle of Figure 12.59.
For critical points of D 2 , we solve
∂f 3
0 = = 2(x − 4) + (6x + 3y + 22),
∂x 4
∂f 3
0 = = 2(y − 6) + (6x + 3y + 22).
∂y 8
The solution of these equations is (−140/61, 174/61) , an unacceptable point since it does not
lie in the triangle of Figure 12.59. The point on the triangle in Figure 12.58 closest to (4, 6, 7)
therefore lies along one of the edges of the triangle. We can find it by minimizing f (x, y)
along the edges of the triangle in Figure 12.59. On C1 , x = 0 in which case
' (2
2 2 3y + 22
D = F (y) = f (0, y) = 16 + (y − 6) + , 0 ≤ y ≤ 2.
4
For critical points we solve
3 6
0 = F + (y) = 2(y − 6) + (3y + 22) ,⇒ y = .
8 5
The value of F (y) here is F (6/5) = 80 . We will evaluate D 2 at the corners of the triangle
later. On C2 , y = 0 in which case
' (2
2 2 3x + 11
D = G(x) = f (x, 0) = (x − 4) + 36 + , 0 ≤ x ≤ 1.
2
For critical points we solve
3 17
0 = G+ (x) = 2(x − 4) + (3x + 11) ,⇒ x =− ,
2 13
an unacceptable value. On C3 , 2x + y = 2, in which case
' (2
2 2 2 6x + 6 − 6x + 22
D = H (x) = (x − 4) + (2 − 2x − 6) +
4
= (x − 4)2 + 4(x + 2)2 + 49, 0 ≤ x ≤ 1.
4
0 = H + (x) = 2(x − 4) + 8(x + 2) ,⇒ x =− ,
5
again an unacceptable value. We now evaluate D 2 at the corners of the triangle in Figure 12.59,
f (0, 0) = 329/4 , f (1, 0) = 94 , f (0, 2) = 81 .
The point on the first quadrant part of the plane

√ 6x +√3y + 4z = 6 closest to (4, 6, 7) is
therefore (0, 6/5, 3/5) , where the distance is 80 = 4 5.
EXERCISES 12.11
In Exercises 1–8 find the maximum and minimum values of the func- ∗ 21. Find the maximum and minimum values of the function f (x, y, z) =
tion on the region. xyz on the sphere x 2 + y 2 + z2 = 1.
∗ 22. Find the maximum and minimum values of f (x, y) = x 2 − y 2
∗ 1. f (x, y) = x 2 + y 3 on R : x 2 + y 2 ≤ 1 on the circle x 2 + y 2 = 1.
∗ 2. f (x, y) = x 2 +x+3y 2 +y on the region R bounded by y = x+1, ∗ 23. Find the maximum and minimum values of f (x, y) = |x − y|
y = 1 − x , y = x − 1, y = −x − 1 on the circle x 2 + y 2 = 1.
∗ 3. f (x, y) = 3x + 4y on region R bounded by the lines x + y = 1, ∗ 24. Find the maximum and minimum values of f (x, y) = x 2 − y 2
x + y = 4, y + 1 = x , y − 1 = x on the curve |x| + |y| = 1.
∗ 25. Find the maximum and minimum values of f (x, y) = |x − 2y|
∗ 4. f (x, y) = x 2 y + xy 2 + y on R : −1 ≤ x ≤ 1, −1 ≤ y ≤ 1 on the curve |x| + |y| = 1.
∗ 5. f (x, y) = 3x 2 + 2xy − y 2 + 5 on R : 4x 2 + 9y 2 ≤ 36 ∗ 26. If P is the perimeter of a triangle with sides of lengths x , y , and
z , the area of the triangle is
∗ 6. f (x, y) = x 3 − 3x + y 2 + 2y on the triangle bounded by x = 8 '
0, y = 0, x + y = 1 (' (' (
P P P P
A= −x −y −z ,
∗ 7. f (x, y) = x 3 + y 3 − 3x − 12y + 2 on the square −3 ≤ x ≤ 3, 2 2 2 2
−3 ≤ y ≤ 3
where P = x + y + z . Show that A is maximized for fixed P when
∗ 8. f (x, y) = x 3 + y 3 − 3x − 3y + 2 on the circle x 2 + y 2 ≤ 1 the triangle is equilateral.
∗ 27. Show that for any triangle with interior angles A , B , and C ,
∗ 9. Find maximum and minimum values of the function f (x, y, z) =
xy 2 z3 on that part of the plane x + y + z = 6 for which (a) x > sin (A/2) sin (B/2) sin (C/2) ≤ 1/8.
0, y > 0, z > 0 and (b) x ≥ 0, y ≥ 0, z ≥ 0.
Hint: Find the maximum value of the function f (A, B, C) =
∗ 10. Find the point on the plane x + y − 2z = 6 closest to the origin. sin (A/2) sin (B/2) sin (C/2) .
∗ 11. Find the shortest distance from (−1, 1, 2) to the plane 2x − 3y + ∗ 28. Show that | cos x + cos y + sin x sin y| ≤ 2 for all x and y .
6z = 14. ∗ 29. A silo is in the shape of a right-circular cylinder surmounted by a
∗ 12. Find the point on the surface z = x 2 + y 2 closest to the point right-circular cone. If the radius of each is 6 m and the total surface
(1, 1, 0) . area must be 200 m 2 (not including the base), what heights for the cone
and cylinder yield maximum enclosed volume?
∗ 13. Find the point on that part of the plane x + y + 2z = 4 in the first ∗ 30. What values of x and y maximize the production function
octant that is closest to the point (3, 3, 1) . For this question assume P (x, y) = kx α y β , where k , α , and β are positive constants
that the curves of intersection of the plane with the coordinate planes ( α + β = 1), when x and y must satisfy Ax + By = C , where
are part of the surface. A , B , and C are positive constants?
∗ 14. The electrostatic potential at each point in the region 0 ≤ x ≤ ∗ 31. A long piece of metal 1 m wide is bent at A and B , as shown in the
1, 0 ≤ y ≤ 1 is given by V (x, y) = 48xy − 32x 3 − 24y 2 . Find the figure below, to form a channel with three straight sides. If the bends
maximum and minimum potentials in the region. are equidistant from the ends, where should they be made in order to
obtain maximum possible flow of fluid along the channel?
∗ 15. When a rectangular box is sent through the mail, the post office
demands that the length of the box plus twice the sum of its height
and width be no more than 250 cm. Find the dimensions of the box
satisfying this requirement that encloses the largest possible volume.
∗ 16. An open tank in the form of a rectangular parallelepiped is to be A B
built to hold 1000 L of acid. If the cost per unit area of lining the base
of the tank is three times that of the sides, what dimensions minimize
the cost of lining the tank? ∗ 32. Find maximum and minimum values of the function f (x, y, z) =
xy + xz on the region x 2 + y 2 + z2 ≤ 1.
∗ 17. Prove that minimum distance from a point (x1 ,√y1 , z1 ) to a plane
∗∗ 33. Find maximum and minimum values of the function f (x, y, z) =
Ax+By+Cz+D = 0 is |Ax1 +By1 +Cz1 +D|/ A2 + B 2 + C 2 .
x 2 yz on the region x 2 + y 2 ≤ 1, 0 ≤ z ≤ 1.
∗ 18. Prove that for triangles the point that minimizes the sum of the ∗∗ 34. A company produces two products, X and Y, each of which must
squares of the distances to the vertices is the centroid. pass through three stages of manufacture. In phase I, up to eight units
2 2 2 2 2 per hour of X can be processed and up to four units per hour of Y .
∗ 19. Find the point on the curve x − xy + y − z = 1, x + y = 1
For phase II, the numbers of units per hour of X and Y are 3 and 6,
closest to the origin.
respectively, whereas the total number of units that can be handled in
∗ 20. Find the dimensions of the box with largest possible volume that phase III is nine per two-hour shift for each of X and Y . The profit per
can fit inside the ellipsoid x 2 /a 2 + y 2 /b2 + z2 /c2 = 1, assuming that unit of X is $200 and per unit of Y is $300. How many units of each
its edges are parallel to the coordinate axes. product should be manufactured for maximum profit?
∗∗ 35. A cow’s daily diet consists of three foods: hay, grain, and supple- space. If the computer lab must have 100 computers, how many of each
ments. The cow is always given 11 kg of hay per day, 50% of which is should the university purchase in order to minimize cost?
digestive material and 12% of which is protein. Grain is 74% digestive ∗∗ 39. Exercises 34, 35, and 38 are examples of linear programming
material and 8.8% protein, whereas supplements are 62% digestive ma- problems that abound in applications of mathematics. The general two-
terial and 34% protein. The cost of hay is $27.50 for 1000 kg, and grain dimensional linear programming problem is to maximize the function
and supplements cost $110 and $175 for 1000 kg. A healthy cow’s diet f (x, y) = cx + dy where c and d are positive constants. Points to be
must contain between 9.5 and 11.5 kg of digestive material and between considered must satisfy m inequalities of the form Ai x + Bi y ≤ Ci ,
1.9 and 2.0 kg of protein. Determine the daily amounts of grain and i = 1, . . . , m , where Ai , Bi , and Ci are positive constants, and x ≥ 0
supplements that the cow should be fed in order that total food costs be and y ≥ 0. These inequalities describe a polygon in the first quadrant
kept to a minimum. of the xy -plane. Give an argument to show that f (x, y) is maximized
∗∗ 36. Find the area of the largest triangle that has vertices on the circle at one or more of the vertices of the polygon.
x2 + y2 = r 2 . ∗∗ 40. In Consulting Project 5 in Section 4.7, we considered seismic
∗∗ 37. A rectangle is surmounted by an isosceles triangle as shown in prospecting with two media. The figure below shows the situation
the figure below. Find x , y , and θ in order that the area of the figure for three media, the problem being to predict depths d1 and d2 . Show
will be as large as possible under the restriction that its perimeter must that the time for a signal to be emitted by the source, pass through
be P . medium 1 with speed v1 , pass through medium 2 with speed v2 , travel
along the interface between media 2 and 3 with speed v3 , then through
medium 2 and medium 1 to the receiver is given by
2d1 sec θ1 2d2 sec θ2 s − 2d1 tan θ1 − 2d2 tan θ2

t = + + .
v1 v2 v3
Verify that t is a minimum when θ1 and θ2 are given by
θ1 = Sin−1 (v1 /v3 ), θ2 = Sin−1 (v2 /v3 ).

y
s
Source Receiver
Surface
x
Medium 1 v1 d1
∗∗ 38. The Easy University is buying computers. It has three models to

choose from. Each model A computer, with 64 MB of memory and a
3 GB hard drive, costs $1300; model B, with 32 MB of memory and a
Medium 2 v2 d2
4 GB drive, costs $1200; and economy model C, with 16 MB of memory
and a 1 GB drive, costs $1000. For reasons related to accreditation,
the university needs at least 2000 MB of memory and 150 GB of disk Medium 3 v3
12.12 Lagrange Multipliers

Many applied maxima and minima problems result in constraint problems. In particular,
Examples 12.34 and 12.35 contain such problems. In Example 12.34, to find extreme values of
T on the edge of the plate we maximized and minimized T (x, y) = 16x 2 − 24xy + 40y 2
subject first to the constraint x 2 + y 2 = 1, and then to the constraint y = 0. Our method
there was to substitute from the constraint equation into T (x, y) in order to obtain a function
of one variable. In Example 12.35, to find the minimum distance from (4, 6, 7) to the plane
6x + 3y + 4z = 6, we minimized D 2 = (x − 4)2 + (y − 6)2 + (z − 7)2 subject to the
constraint 6x + 3y + 4z = 6. Again we substituted from the constraint to obtain D 2 as a
function of two independent variables.
A natural question to ask is whether problems of this type can be solved without substituting
from the constraint equation, for if the constraint equation is complicated, substitution may
be very difficult or even impossible. To show that there is indeed an alternative, consider
the situation in which a function f (x, y, z) is to be maximized or minimized subject to two
constraints:
F (x, y, z) = 0, (12.60a)
G(x, y, z) = 0. (12.60b)
Algebraically, we are to find extreme values of f (x, y, z) , considering only those values of x ,
y , and z that satisfy equations 12.60. Geometrically, we can interpret each of these conditions
as specifying a surface, so that we are seeking extreme values of f (x, y, z) , considering only
those points on the curve of intersection C of the surfaces F (x, y, z) = 0 and G(x, y, z) = 0
(Figure 12.60).
Extreme values of f (x, y, z) along C will occur either at critical points of the function or
FIGURE 12.60 Extreme
values of f (x, y, z) along the curve
at the ends of the curve. But what derivative or derivatives of f (x, y, z) are we talking about
of intersection of two surfaces when we say critical points? Since we are concerned only with values of f (x, y, z) on C , we
must mean the derivative of f (x, y, z) along C (i.e., the directional derivative in the tangent
z
F (x, y, z) = 0 direction to C ). If T, then, is a tangent vector to C , critical points of f (x, y, z) along C are
given by
T 0 = DT f = ∇f · T̂,
C
or at points where the directional derivative is undefined. According to equation 12.53, a tangent
vector to C is T = ∇F × ∇G , and hence a unit tangent vector is
∇F × ∇G
T̂ = .
|∇F × ∇G|
x The directional derivative of f (x, y, z) at points along C is therefore given by
G (x, y, z) = 0 y
∇F × ∇G
DT f = ∇f · .
|∇F × ∇G|
It follows, then, that critical points of f (x, y, z) are points (x, y, z) that satisfy the equation
∇f · ∇F × ∇G = 0,
or points at which the left side is not defined. Now vector ∇F × ∇G is perpendicular to both
∇F and ∇G . Since ∇f is perpendicular to ∇F × ∇G (their dot product is zero), it follows
that ∇f must lie in the plane of ∇F and ∇G . Consequently, there exist scalars λ and µ such
that
∇f = (−λ) ∇F + (−µ) ∇G,
or
∇f + λ ∇F + µ ∇G = 0. (12.61)
This vector equation is equivalent to the three scalar equations
∂f ∂F ∂G
+λ +µ = 0, (12.62a)
∂x ∂x ∂x
∂f ∂F ∂G
+λ +µ = 0, (12.62b)
∂y ∂y ∂y
∂f ∂F ∂G
+λ +µ = 0, (12.62c)
∂z ∂z ∂z
and these equations must be satisfied at a critical point at which the directional derivative of
f (x, y, z) vanishes. Note, too, that at a point at which the directional derivative of f (x, y, z)
does not exist, one of the partial derivatives in these equations does not exist. In other words,
we have shown that critical points of f (x, y, z) are points that satisfy equations 12.62 or
points at which the equations are undefined. These equations, however, contain five unknowns:
x , y , z , λ , and µ . To complete the system we add equations 12.60 since they must also
be satisfied by a critical point. Equations 12.60 and 12.62 therefore yield a system of five
equations in the five unknowns x , y , z , λ , and µ ; the first three unknowns (x, y, z) define
a critical point of f (x, y, z) along C . The advantage of this system of equations lies in the
fact that differentiations in 12.62 involve only the given functions (and no substitutions from the
constraint equations are necessary). What we have sacrificed is a system of three equations in the
three unknowns (x, y, z) for a system of five equations in the five unknowns (x, y, z, λ, µ) .
Let us not forget that the original problem was to find extreme values for the function
f (x, y, z) subject to constraints 12.60. What we have shown so far is that critical points at
which the directional derivative of f (x, y, z) vanishes can be found by solving equations 12.60
and 12.62. In addition, critical points at which the directional derivative of f (x, y, z) does not
exist are points at which equations 12.62 are not defined. What remains is to evaluate f (x, y, z)
at all critical points and at the ends of C . If C is a closed curve (i.e., if C rejoins itself), then
f (x, y, z) needs to be evaluated only at the critical points. This turns out to be very important
in practice.
Through the directional derivative and tangent vectors to curves, we have shown that equa-
tions 12.60 and 12.62 define critical points of a function f (x, y, z) that is subject to two
constraints: F (x, y, z) = 0 and G(x, y, z) = 0. But what about other situations? Let us
say, for example, that we require extreme values of a function f (x, y, z, t) subject to a single
constraint F (x, y, z, t) = 0. How shall we find critical points of this function? Fortunately,
as we now show, there is a very simple method that yields equations 12.60 and 12.62, and this
method generalizes to other situations also.
To find critical points of f (x, y, z) subject to constraints 12.60, we define a function
L(x, y, z, λ, µ) = f (x, y, z) + λF (x, y, z) + µG(x, y, z),
and regard it as a function of five independent variables x , y , z , λ , and µ . To find critical

points of this function, we would first solve the equations obtained by setting each of the partial
derivatives of L equal to zero:
∂L ∂f ∂F ∂G
0 = = +λ +µ ,
∂x ∂x ∂x ∂x
∂L ∂f ∂F ∂G
0 = = +λ +µ ,
∂y ∂y ∂y ∂y
∂L ∂f ∂F ∂G
0 = = +λ +µ ,
∂z ∂z ∂z ∂z
∂L
0 = = F (x, y, z),
∂λ
∂L
0 = = G(x, y, z).
∂µ
In addition, we would consider points at which the partial derivatives of L do not ex-
ist. Clearly, this means points (x, y, z) at which any of the partial derivatives of f (x, y, z) ,
F (x, y, z) , and G(x, y, z) do not exist. But these are equations 12.60 and 12.62. We have
shown, then, that finding critical points (x, y, z) of f (x, y, z) subject to F (x, y, z) = 0 and
G(x, y, z) = 0 is equivalent to finding critical points (x, y, z, λ, µ) of L(x, y, z, λ, µ) .
The two unknowns λ and µ that accompany a critical point (x, y, z) of f (x, y, z) are called
Lagrange multipliers. They are not a part of the solution (x, y, z) to the original problem,
but have been introduced as a convenience by which to arrive at that solution. The function
L(x, y, z, λ, µ) is often called the Lagrangian of the problem.
The method for other constraint problems should now be evident. Given a function
f (x, y, z, t, . . .) of n variables to maximize or minimize subject to m constraints
F1 (x, y, z, t, . . .) = 0, F2 (x, y, z, t, . . .) = 0, . . . , Fm (x, y, z, t, . . .) = 0, (12.63)

we introduce m Lagrange multipliers λ1 , λ2 , . . . , λm into a Lagrangian of n + m independent

variables x, y, z, t, . . . , λ1 , λ2 , . . . , λm :
L(x, y, z, t, . . . λ1 , λ2 , . . . , λm ) = f (x, y, z, t, . . .) + λ1 F1 (x, y, z, t, . . .) + · · ·

+ λm Fm (x, y, z, t, . . .). (12.64)
Critical points (x, y, z, t, . . .) of f (x, y, z, t, . . .) are then determined by the equations defin-
ing critical points of L(x, y, z, t, . . . , λ1 , λ2 , . . . , λm ) , namely,
∂L ∂f ∂F1 ∂Fm
0 = = + λ1 + · · · + λm , (12.65a)
∂x ∂x ∂x ∂x
∂L ∂f ∂F1 ∂Fm
0 = = + λ1 + · · · + λm , (12.65b)
∂y ∂y ∂y ∂y
.. .. ..
. . .
∂L
0 = = F1 (x, y, z, t, . . .), (12.65c)
∂λ1
∂L
0 = = F2 (x, y, z, t, . . .), (12.65d)
∂λ2
.. .. ..
. . .
∂L
0 = = Fm (x, y, z, t, . . .). (12.65e)
∂λm
To use Lagrange multipliers in Example 12.35, where we were minimizing D 2 = (x −

4) + (y − 6)2 + (z − 7)2 subject to 6x + 3y + 4z = 6, we define the Lagrangian
2
L(x, y, z, λ) = D 2 + λ(6x + 3y + 4z − 6)
= (x − 4)2 + (y − 6)2 + (z − 7)2 + λ(6x + 3y + 4z − 6).
Critical points of L(x, y, z, λ) are defined by
∂L
0 = = 2(x − 4) + 6λ,
∂x
∂L
0 = = 2(y − 6) + 3λ,
∂y
∂L
0 = = 2(z − 7) + 4λ,
∂z
∂L
0 = = 6x + 3y + 4z − 6.
∂λ
The solution of this linear system is (x, y, z, λ) = (−140/61, 174/61, 171/61, 128/61) ,
yielding as before the critical point (−140/61, 174/61, 171/61) of D 2 .
EXAMPLE 12.36
Find the maximum and minimum values of the function f (x, y, z) = xyz on the sphere
x 2 + y 2 + z2 = 1.
SOLUTION If we define the Lagrangian
L(x, y, z, λ) = xyz + λ(x 2 + y 2 + z2 − 1),

then critical points of L , and therefore of f (x, y, z) , are defined by the equations
∂L
0 = = yz + 2λx,
∂x
∂L
0 = = xz + 2λy,
∂y
∂L
0 = = xy + 2λz,
∂z
∂L
0 = = x 2 + y 2 + z2 − 1.
∂λ
If we multiply the first equation by y and the second by x , and equate the resulting expressions
for 2λxy , we have
y 2 z = x 2 z.
Consequently, either z = 0 or y = ±x .
Case I: z = 0. In this case the equations reduce to
λx = 0, λy = 0, xy = 0, x 2 + y 2 = 1.
The first implies that either x = 0 or λ = 0. If x = 0, then y = ±1, and we have two critical
points (0, ±1, 0) . If λ = 0, then the third equation requires x = 0 or y = 0. We therefore
obtain two additional critical points (±1, 0, 0) .
Case II: y = x . In this case the equations reduce to
xz + 2λx = 0, x 2 + 2λz = 0, 2x 2 + z2 = 1.
The first implies that either x = 0 or z = −2λ . If x = 0, then z = ±1, and we have the√two
critical points (0, 0, ±1) . If z = −2λ , then the last two equations imply that x = ±1/ 3,
and we obtain the four critical points
√ √ √ √ √ √
(±1/ 3, ±1/ 3, 1/ 3) and (±1/ 3, ±1/ 3, −1/ 3).
Case III: y = −x . This case is similar to that for y = x , and leads to the additional four
critical points
√ √ √ √ √ √
(±1/ 3, ∓1/ 3, 1/ 3) and (±1/ 3, ∓1/ 3, −1/ 3).
Because x 2 + y 2 + z2 = 1 is a surface without a boundary, we complete the problem by

evaluating f (x, y, z) at each of the critical points:
f (±1, 0, 0) = f (0, ±1, 0) = f (0, 0, ±1) = 0,

√ √ √ √ √ √ √
f (±1/ 3, ±1/ 3, 1/ 3) = f (±1/ 3, ∓1/ 3, −1/ 3) = 3/9,
√ √ √ √ √ √ √
f (±1/ 3, ±1/ 3, −1/ 3) = f (±1/ 3, ∓1/ 3, 1/ 3) = − 3/9.
√
The maximum
√ and minimum values of f (x, y, z) on x 2 + y 2 + z2 = 1 are therefore 3/ 9
and − 3/9.
To compare the Lagrangian solution in this example to that without a Lagrange multiplier, see
Exercise 21 in Section 12.11.
EXAMPLE 12.37
A company manufactures wheelbarrows at n of its plants. The cost of manufacturing xi wheel-
barrows at the i th plant is xi2 /ci , where ci > 0 are known constants. The total cost of
manufacturing x1 wheelbarrows at plant 1, x2 at plant 2, . . . , xn at plant n , is therefore
x12 x2 x2
f (x1 , . . . , xn ) = + 2 + ··· + n.
c1 c2 cn
The production engineer wishes to schedule a total of D wheelbarrows among the plants. How
many should each plant produce if costs are to be minimized, and what is minimum cost?
SOLUTION We must minimize f (x1 , . . . , xn ) subject to the constraint x1 + x2 + · · · +

xn = D . If we define the Lagrangian
L(x1 , . . . , xn , λ) = f (x1 , . . . , xn ) + λ(x1 + x2 + · · · + xn − D),
then critical points of L (and therefore of f ) are given by the n + 1 equations
∂L ∂f 2 xi
0 = = +λ = + λ, i = 1, . . . , n,
∂xi ∂xi ci
∂L
0 = = x1 + x2 + · · · + xn − D.
∂λ
These give xi = −ci λ/2 for each i , and if we substitute these into the last equation,
c1 λ c2 λ cn λ −2 D
− − − ··· − =D ,⇒ λ= .
2 2 2 c1 + · · · + cn
Thus, production levels at the plants should be
ci D
xi = , i = 1, . . . , n.
c1 + · · · + cn
Minimum cost is
' (2 ' (2
1 c1 D 1 cn D D2
+ ··· + = .
c1 c1 + · · · + cn cn c1 + · · · + cn c1 + · · · + cn
We are being approached by a hydraulic engineer who is fabricating an open channel from
long pieces of metal joined end to end. We take the width of the metal as 1 metre, although
the solution is easily adapted to any width. Each piece of metal is bent to form the channel.
The engineer was once a student of this text and has solved Exercise 31 in Section 12.11,
but he is not convinced that this is the optimum shape of the channel. For instance, why
is it necessary that the two sides of the channel be of the same length? We are to show
that the solution to Exercise 31 in Section 12.11 is indeed the best of all possible channels
with a maximum of two bends.
SOLUTION Figure 12.61a shows a channel with two bends but allows for different
lengths of the sides of the channel and therefore different angels θ and φ . Volume of flow
along the channel is maximized when its cross-sectional area is maximized, and the area
for this channel is
1 1
A= y 2 sin θ cos θ + xy sin θ + (1 − x − y)2 sin φ cos φ,
2 2
where, in order that both sides of the channel have the same height, the condition y sin θ =
(1 − x − y) sin φ must be satisfied. If we regard x , θ , and φ as independent variables,
and this equation as a restriction on y , then A must be maximized for values of x , θ , and
φ in the box of Figure 12.61b.
FIGURE 12.61a Bending a FIGURE 12.61b Domain

piece of metal to form a channel for optimization problem
y cos θ (1 − x − y) cos φ π/2

φ
(1 − x − y) sin φ
y sin θ
y 1−x−y
θ φ
π/2
x 1
x θ
We shall find critical values of A inside the box, and then turn to the six faces of the
box for maxima thereon. We shall have two difficulties. One is the complexity of the
equations that must be solved, and the other is to not lose sight of exactly where we are in
the solution process. For critical points of A interior to the box, we form the Lagrangian
1
L(x, y, θ, φ, λ) = y 2 sin θ cos θ + xy sin θ
2
1
+ (1 − x − y)2 sin φ cos φ + λ[y sin θ − (1 − x − y) sin φ ].
2
Critical points of this function are defined by the equations
∂L
0 = = y sin θ − (1 − x − y) sin φ cos φ + λ sin φ, (12.66a)
∂x
∂L
0 = = y sin θ cos θ + x sin θ − (1 − x − y) sin φ cos φ
∂y
+ λ(sin θ + sin φ), (12.66b)
∂L 1
0 = = y 2 (cos2 θ − sin2 θ ) + xy cos θ + λy cos θ, (12.66c)
∂θ 2
∂L 1
0 = = (1 − x − y)2 (cos2 φ − sin2 φ) − λ(1 − x − y) cos φ, (12.66d)
∂φ 2
∂L
0 = = y sin θ − (1 − x − y) sin φ. (12.66e)
∂λ
If we subtract the first of these from the second, we obtain
y sin θ cos θ +x sin θ −y sin θ +λ sin θ = 0 ,⇒ sin θ (y cos θ +x −y +λ) = 0.
One possibility is sin θ = 0 ,⇒ θ = 0, but this yields minimum area. The other
possibility is that
y cos θ + x − y + λ = 0. (12.66f)
When we solve this for λ and substitute into equation 12.66c,
1
y 2 (cos2 θ − sin2 θ ) + xy cos θ + y cos θ (y − x − y cos θ ) = 0
2
' (
1
,⇒ y 2 cos θ − = 0.
2
Thus, y = 0, which gives minimum area, or cos θ = 1/2 ,⇒ θ = π/3. If we subtract

equation 12.66e from 12.66a, we find
−(1 − x − y) sin φ cos φ + (1 − x − y) sin φ + λ sin φ = 0

,⇒ sin φ [(1 − x − y)(1 − cos φ) + λ] = 0.
Either sin φ = 0 ,⇒ φ = 0, giving minimum A , or (1 − x − y)(1 − cos φ) + λ = 0.

When we substitute this into equation 12.66d,
1
0 = (1 − x − y)2 (cos2 φ − sin2 φ) + (1 − x − y)2 (1 − cos φ) cos φ.
2
This equation implies that x + y = 1, giving minimum A , or,
1 1 π
(cos2 φ − sin2 φ) − cos2 φ + cos φ = 0 ,⇒ cos φ = ,⇒ φ = .
2 2 3
Equation 12.66e now gives x + 2y = 1. When this is substituted into equations 12.66a,
c, two equations in y and λ result, and the solution for y is y = 1/3. This in turn gives
x = 1/3, and the area of the channel with these values of x , y , θ and φ is
' (2 " √ $ ' ( ' ( ' ( "√ $
1 1 3 1 1 1 3
A= +
2 3 2 2 3 3 2
' (2 " √ $ ' ( √

1 1 1 3 1 3
+ 1− − = .
2 3 3 2 2 12
We now turn to the six faces of the box defining permissible values of x , θ , and φ . The
three faces x = 1, φ = 0, and θ = 0 give minimum values 0 for area of the channel.
This leaves faces x = 0, φ = π/2, and θ = π/2. We need discuss only one of the last
two faces, since discussions would be identical for the other. Consider first then the face
x = 0. In this case the metal has only one bend as shown in Figure 12.62a. Area of the
cross-section of the channel is
1 1
A= y 2 sin θ cos θ + (1 − y)2 sin φ cos φ,
2 2
subject to the restriction y sin θ = (1 − y) sin φ . If θ and φ are taken as independent

variables, this function must be maximized on the square in Figure 12.62b.
FIGURE 12.62a Triangular FIGURE 12.62b Domain for

channel optimization of triangular channel
y cos θ (1 − y) cos φ φ
π/2
y sin θ = (1 − y) sin φ
y 1−y
θ φ
π/2 θ
First we find critical points interior to the square, and then consider the four lines
forming the boundary of the square. For critical points of A interior to the square, we
form the Lagrangian
1 1
L(y, θ, φ, λ) = y 2 sin θ cos θ + (1 −y)2 sin φ cos φ +λ[y sin θ −(1 −y) sin φ ].
2 2
Critical points of this function are defined by the equations
∂L
0 = = y sin θ cos θ − (1 − y) sin φ cos φ + λ(sin θ + sin φ), (12.67a)
∂y
∂L 1
0 = = y 2 (cos2 θ − sin2 θ ) + λy cos θ, (12.67b)
∂θ 2
∂L 1
0 = = (1 − y)2 (cos2 φ − sin2 φ) − λ(1 − y) cos φ, (12.67c)
∂φ 2
∂L
0 = = y sin θ − (1 − y) sin φ. (12.67d)
∂λ
The second of these implies that y = 0, which minimizes A , or that
y(cos2 θ − sin2 θ ) + 2λ cos θ = 0. (12.67e)
The third yields y = 1, which minimizes A , or
(1 − y)(cos2 φ − sin2 φ) − 2λ cos φ = 0. (12.67f)
When we substitute from equation 12.67d into 12.67a, we obtain

' (
1−y
0 = (1 − y) sin φ cos θ − (1 − y) sin φ cos φ + λ sin φ + sin φ . (12.67g)
y
Either sin φ = 0, which leads to minimum A , or
y(1 − y) cos θ − y(1 − y) cos φ + λ = 0. (12.67h)
When we solve this for λ = −y(1 − y)(cos θ − cos φ) , and substitute into equation
12.67e,
y(cos2 θ − sin2 θ ) − 2y(1 − y) cos θ (cos θ − cos φ) = 0.

This requires y = 0, a minimizing value, or
cos2 θ − sin2 θ − 2(1 − y) cos θ (cos θ − cos φ) = 0. (12.67i)
When we substitute for λ in equation 12.67f,
(1 − y)(cos2 φ − sin2 φ) + 2y(1 − y) cos φ(cos θ − cos φ) = 0,
from which y = 1, a minimum, or
cos2 φ − sin2 φ + 2y cos φ(cos θ − cos φ) = 0. (12.67j)
We now solve equation 12.67d for y = sin φ/(sin θ + sin φ) , and substitute into 12.67i, j,
2 sin θ cos θ (cos θ − cos φ)

cos2 θ − sin2 θ − = 0,
sin θ + sin φ
2 sin φ cos φ(cos θ − cos φ)
cos2 φ − sin2 φ + = 0.
sin θ + sin φ
These imply that

cos2 θ − sin2 θ cos2 φ − sin2 φ
=− .
2 sin θ cos θ 2 sin φ cos φ
The only way for this equation to hold is for θ = φ = π/4, and this implies that y = 1/2.
The area of the channel for these values is
' (2 ' ( ' (2 ' (
1 1 1 1 1 1 1
A= + = .
2 2 2 2 2 2 8
We should now consider area on

FIGURE 12.63 Triangular chan-
the four sides of the square. Along
nel with vertical side
θ = 0 and φ = 0, area is zero. Con-
sideration of φ = π/2 is the mirror
image of θ = π/2. For θ = π/2, we
are considering channels in the shape y 1−y
in Figure 12.63. Area is
1 ! 1 !
A = y (1 −y)2 −y 2 = y 1 − 2y,
2 2
where 0 ≤ y ≤ 1/2. For critical
points, we solve ' (
1 ! y 1
0 = 1 − 2y − √ ,⇒ y = .
2 1 − 2y 3
√
The area of the channel when y = 1/3 is A = 3/18 . The area is zero when y = 0
and y = 1/2. This completes the discussion of the face x = 0 of the box in Figure
12.61b.
Our final consideration is face θ = π/2 of the box. In this case, the channel has the
shape in Figure 12.64a. Its area is
1
A = xy + (1 − x − y)2 sin φ cos φ,
2
where y = (1 − x − y) sin φ . This function must be maximized over the rectangle in

Figure 12.64b.
FIGURE 12.64a Channel FIGURE 12.64b Domain for opti-

with one vertical side mization of channel with one vertical side
(1 − x − y) cos φ φ
π/2
(1 − x − y) sin φ
y 1−x−y
φ
1 x
x
The associated Lagrangian,
1
L(x, y, φ, λ) = xy + (1 − x − y)2 sin φ cos φ + λ[y − (1 − x − y) sin φ ],
2
has critical points defined by
∂L
0 = = y − (1 − x − y) sin φ cos φ + λ sin φ, (12.68a)
∂x
∂L
0 = = x − (1 − x − y) sin φ cos φ + λ(1 + sin φ), (12.68b)
∂y
∂L 1
0 = = (1 − x − y)2 (cos2 φ − sin2 φ) − λ(1 − x − y) cos φ, (12.68c)
∂φ 2
∂L
0 = = y − (1 − x − y) sin φ. (12.68d)
∂λ
The first two equations imply that λ = y − x . We substitute this into equation 12.68a,
and into 12.68c, after removing the extra factor 1 − x − y ,
0 = y − (1 − x − y) sin φ cos φ + (y − x) sin φ, (12.68e)
0 = (1 − x − y)(cos2 φ − sin2 φ) − 2(y − x) cos φ. (12.68f)
When we multiply the first of these by 2 cos φ , the second by sin φ , and add,
2y cos φ − 2(1 − x − y) sin φ cos2 φ + (1 − x − y)(cos2 φ − sin2 φ) sin φ = 0.
We can use equation 12.68d to eliminate x from this equation,
2y cos φ − 2y cos2 φ + (cos2 φ − sin2 φ)y = 0.
This implies that y = 0, which gives minimum A , or,
1 π
2 cos φ − 2 cos2 φ + (cos2 φ − sin2 φ) = 0 ,⇒ cos φ = ,⇒ φ = .
2 3
Equation 12.68f now gives
' ( ' (
1 3 1
(1 − x − y) − − 2(y − x) =0 ,⇒ y = 3x − 1.
4 4 2
When we substitute this in equation 12.68d,
√ √
3 3− 3
3x − 1 − (1 − x − 3x + 1) =0 ,⇒ x = .
2 3
√
This now gives y = 2 − 3, and area of the channel is
" √ $ 9 √ :2 " √ $ ' ( √
3− 3 √ 1 3− 3 √ 3 1 2− 3
A= (2 − 3)+ 1− −(2 − 3) = .
3 2 3 2 2 2
We now consider A along the four

edges of the rectangle in Figure 12.64b.
Along φ = 0 and x = 1, area is a min- FIGURE 12.65 Channel with
imum. We have already dealt with the two vertical sides
case x = 0 (see Figure 12.63). This
leaves the case φ = π/2, a rectangu-
lar channel as shown in Figure 12.65.
Its area is (1 − x)/2
1
A = x(1 − x), 0 ≤ x ≤ 1.
2
x
Critical points of this function are given
by 0 = 1 − 2x , from which x = 1/2.
Area of the channel is 1/8 . Area is
zero when x = 0 and x = 1.
This completes our calculations; we have considered area of the channel for values of
x , θ , and φ interior to the box of Figure 12.61b,
√ over each face
√ of the box, along√each edge,
and at each corner.√The boxed numbers are 3/12, 1/8, 3/18 and (2 − 3)/2, the
largest of which is 3/12. This is the solution to Exercise 31 in Section 12.11, confirming
that it provides the channel with maximum flow from all channels with a maximum of two
bends. We might notice that if more bends or curved sides were allowed, the area could
be increased beyond this. For instance, if the channel is semicircular, its area is 1/(2π ) .
EXERCISES 12.12
In Exercises 1–8 use Lagrange multipliers to find maximum and min- ∗ 6. f (x, y, z) = x 2 y + z subject to x 2 + y 2 = 1, z = y
imum values of the function subject to the constraints. In each case,
∗ 7. f (x, y, z) = x 2 + y 2 + z2 subject to x 2 + y 2 + z2 = 2z, x +
interpret the constraints geometrically.
y+z=1
!
∗ 1. f (x, y) = x 2 + y subject to x 2 + y 2 = 4 ∗ 8. f (x, y, z) = xyz − x 2 z subject to x 2 + y 2 = 1, z = x 2 + y 2
∗ 2. f (x, y, z) = 5x − 2y + 3z + 4 subject to x 2 + 2y 2 + 4z2 = 9

∗ 9. Use Lagrange multipliers to solve Exercise 11 in Section 12.11.
2
∗ 3. f (x, y) = x + y subject to (x − 1) + y = 1 2 ∗ 10. Use Lagrange multipliers to solve Exercise 12 in Section 12.11.
∗ 4. f (x, y, z) = x 3 + y 3 + z3 subject to x 2 + y 2 + z2 = 9
2 2 2
∗ 5. f (x, y, z) = xyz subject to x + 2y + 3z = 12 ∗ 13. Use Lagrange multipliers to solve Exercise 27 in Section 12.11.
∗ 14. Use Lagrange multipliers to solve Exercise 29 in Section 12.11. ∗ 27. Find the maximum value of f!(x, y, z) = x 2 yz − xzy 2 subject
to constraints x 2 + y 2 = 1, z = x 2 + y 2 .
∗ 28. The equation 3x 2 + 4xy + 6y 2 = 140 describes an ellipse that
∗ 16. Use Lagrange multipliers to solve Exercise 23 in Section 12.11. has its centre at the origin, but major and minor axes are not along the
x - and y -axes. Find coordinates of the ends of the major and minor
∗ 17. Use Lagrange multipliers to solve Exercise 24 in Section 12.11. axes.
∗ 18. Use Lagrange multipliers to solve Exercise 25 in Section 12.11. ∗ 29. Use Lagrange multipliers to find the point on the first octant part of
the plane Ax + By + Cz = D ( A, B, C, D all positive constants)
∗ 19. Use Lagrange multipliers to solve Exercise 17 in Section 12.11. that maximizes the function f (x, y, z) = x p y q zr , where p , q , and
r are positive constants.
∗ 20. Suppose that F (x, y) = 0 and G(x, y) = 0 define two curves
C1 and C2 in the xy -plane. Let P (x0 , y0 ) and Q(X0 , Y0 ) be the ∗ 30. The folium of Descartes has parametric equations
points on C1 and C2 that minimize the distance between C1 and C2 .
3at 3at 2
If C1 and C2 have tangent lines at P and Q , show that the line P Q x = , y = (a > 0)
is perpendicular to these tangent lines. 1+ t3 1 + t3
∗ 21. What are production levels and minimum cost in Example 12.37 (see Exercise 52 in Section 3.8 and Exercise 61 in Section 9.1). Find
for the production of 500 wheelbarrows if there are four plants with the point in the first quadrant farthest from the origin in two ways:
c1 = 26, c2 = 24, c3 = 23, and c4 = 27? (a) Express D 2 = x 2 + y 2 in terms of t and maximize this
function of one variable.
∗ 22. When a thermonuclear reactor is built in the form of a right-circular (b) Show that an implicit equation for the curve is x 3 + y 3 =
cylinder, neutron diffusion theory requires its radius and height to sat- 3axy and maximize D 2 = x 2 + y 2 subject to this con-
isfy the equation straint.
' (2 ' (2 ∗∗ 31. To find the point on the curve x 2 −xy +y 2 +z2 = 1, x 2 +y 2 = 1
2.4048 π closest to the origin, we must minimize the function f (x, y, z) =
+ = k,
r h x 2 + y 2 + z2 subject to the constraints defined by the equations of the
curve. Show that this can be done by (a) using two Lagrange multipliers;
(b) expressing f (x, y, z) in terms of x and y alone, u = 1 − xy , and
where k is a constant. Find r and h in terms of k if the reactor is to
minimizing this function subject to x 2 + y 2 = 1 (with one Lagrange
occupy as small a volume as possible.
multiplier);
√ (c) expressing f (x, y, z) in terms of x alone, u = 1 ±
∗ 23. Find the points on the curve x 2 + xy + y 2 = 1 closest to and x 1 − x 2 , and minimizing these functions on appropriate intervals;
farthest from the origin. and (d) writing x = cos t, y = sin t along the curve, expressing
f (x, y, z) in terms of t , u = 1 − sin t cos t , and minimizing this
function on appropriate intervals.
∗∗ 32. Find the smallest and largest distances from the origin to the curve
In Exercises 24–26 use Lagrange multipliers to find maximum and x 2 + y 2 /4 + z2 /9 = 1, x + y + z = 0.
minimum values of the function.
∗∗ 33. Find the maximum value of f (x, y, z) = (xy + x 2 )/(z2 + 1)
subject to the constraint x 2 (4 − x 2 ) = y 2 .
∗ 24. f (x, y) = 3x 2 + 2xy − y 2 + 5 for 4x 2 + 9y 2 ≤ 36
∗∗ 34. Find the maximum and minimum values of the function f (x, y,
∗ 25. f (x, y) = x 2 y + xy 2 + y for −1 ≤ x ≤ 1, −1 ≤ y ≤ 1 z) = 2x 2 y 2 + 2y 2 z2 + 3z considering only values of x , y , and z
satisfying the equations z = x 2 + y 2 , x 2 + 3y 2 = 1. Do this with
∗ 26. f (x, y, z) = xy + xz for x 2 + y 2 + z2 ≤ 1 and without Lagrange multipliers.
12.13 Least Squares

The basic tool of applied mathematics is the function. Given a function such as f (x) =
x 2 − 2x − 3, and any value of x , say x = 4, we calculate the value of the function at this x
as f (4) = 5. In many problems, the function is not given; it must be found. What we might
have is a set of experimental data suggesting that various quantities are related to one another,
but the data do not specify exactly how they are related. For example, suppose a variable y is
known to depend on a variable x , and an experiment is performed to measure 10 values of y
corresponding to 10 values of x . The results are listed in Table 12.3.
TABLE 12.3
x 1 2 3 4 5 6 7 8 9 10
y 6.05 8.32 10.74 13.43 15.90 18.38 20.93 23.32 24.91 28.36
Consider the problem of finding that function y = y(x) that best describes these data. Two
considerations are important — simplicity and accuracy. We would like as simple a function
as possible to describe how y depends on x . At the same time, we want the function to be as
accurate as possible. For example, we could find a polynomial of degree nine to fit the data
exactly; it would give the exact y -value for each value of x . But this would hardly be a simple
representation. It would also be unreasonable from the following standpoint. Because y -values
are determined experimentally (we assume that x -values are exact), they will be subject to errors
both random and systematic. It is pointless to use a ninth-degree polynomial to reproduce the
data exactly, since it therefore also reproduces the inherent errors. What we want is a simple
function that best fits the trend of the data. To discover this trend, we plot the data of Table 12.3
as points (x, y) in Figure 12.66.
FIGURE 12.66 Finding the best-fitting line to data points in Table 12.3
y
25
20
15
10
2 4 6 8 10 x
We are immediately impressed by the fact that although the points do not all lie on a straight
line, to describe them by a straight line would be a simple representation, and reasonably
accurate. We therefore look for a linear function
y = y(x) = ax + b (12.69)
to describe the data in Table 12.3. Many lines could be drawn to fit the points in Figure 12.66
reasonably accurately, and each line would be characterized by different values of a and b .
Our problem then is to find that line (or those values of a and b ) that best fits the points.
Mathematicians have developed a method called least squares to arrive at a best fit.
We denote the x -values in Table 12.3 by xi = i , i = 1, . . . , 10, and corresponding
observed values of y by y i . The linear function y(x) = ax + b predicts values yi = y(xi ) =
axi + b at the xi . Differences between observed and predicted values of y are
yi − y i = (axi + b) − y i .
We define a quantity S as the sum of the squares of these differences for all xi :
10
5 10
5
S = (yi − y i )2 = (axi + b − y i )2 . (12.70)
i=1 i=1
Given any a and b , S is a measure of the degree to which the points vary from that line — the
better the fit, the smaller the value of S . The method of least squares states that one way to
approximate the points in Figure 12.66 by a straight line is to choose that line which minimizes
the function S = S(a, b) . In other words, find values of a and b that minimize S(a, b) . For
critical points of this function we solve
10
5
∂S
0 = = 2xi (axi + b − y i ),
∂a
i=1
10
5
∂S
0 = = 2(axi + b − y i ).
∂b
i=1
If we rewrite these equations in the form

" 10
$ " 10 $ 10
5 5 5
xi2 a+ xi b = xi y i , (12.71a)
i=1 i=1 i=1
" 10 $ 10
5 5
xi a + 10b = yi , (12.71b)
i=1 i=1
we have a pair of linear equations in a and b . With the data in Table 12.3, we obtain
385a + 55b = 1139.27,
55a + 10b = 170.34,

which have solutions a = 2.45 and b = 3.54. Since there is only one critical point, and we
know that S(a, b) must have an absolute minimum for some a and b , it follows that these
values must minimize S(a, b) . The straight line
y = y(x) = 2.45x + 3.54
is therefore the best straight-line fit (in the least-squares sense) to the data in Table 12.3. This is,
in fact, the line in Figure 12.66. It is important to remember that least squares assumes that we
have a prior knowledge of the type of function to be determined (in this case a linear function),
and the method then proceeds to find the best such function.
Application Preview Polynomials of higher order (quadratics, cubics, etcetera.) can be fitted to data points by
Revisited least-squares (see Exercises 4–7). Other types of functions can also be used, often by reducing
the problem to a straight-line situation. In the Application Preview we posed the problem of
finding the function that represents the tabulated values below, or the equation of a curve that
approximates the points plotted in Figure 12.67a.
TABLE 12.4
V 54.3 61.82 72.4 88.7 118.6 194.0

P 61.2 49.5 37.6 28.4 19.2 10.1
FIGURE 12.67a Data FIGURE 12.67b Plot of

points for pressure and volume logarithms of pressure and volume
P p
60 4
3
40
2
20
1
50 100 150 200 V 4 4.5 5 v

Thermodynamics suggests that when the compression and expansion of the gas is adiabatic,
pressure P and volume V are related by an equation of the form P = b/V a for some constants
a and b . A direct application of the above procedure leads to complicated nonlinear equations
for a and b . Instead, we take logarithms of the equation P = b/V a , and write
ln P = ln b − a ln V .
If we define new variables, p = ln P , B = ln b , v = ln V , and A = −a , then
p = Av + B.
This is the equation of a straight line in the vp -plane. We have tabulated p = ln P and
v = ln V below, and plotted p against v in Figure 12.67b. The fact that the points seem
reasonably collinear is confirmation of the fact that the data points in Figure 12.67a are adequately
described by a function of the form P = b/V a .
TABLE 12.5
v = ln V 3.998 4.124 4.282 4.485 4.776 5.268

p = ln P 4.114 3.902 3.627 3.346 2.955 2.313
Fitting the straight line p = B + Av to the data in Table 12.5 leads to the following equations
for A and B corresponding to equations 12.71:
" 6
$ " 6 $ 6
5 5 5
vi2 A+ vi B = vi p i ,
i=1 i=1 i=1
" 6 $ 6
5 5
vi A + 6B = pi .
i=1 i=1
These give
121.9758A + 26.9295B = 89.3605,
26.9295A + 6B = 20.2570,
the solution of which is A = −1.4043 and B = 9.6788. Consequently, least-squares estimates

for A and B give
p = −1.4043v + 9.6788 ,⇒ ln P = 9.6788 − 1.4043 ln V .
When we exponentiate,
P = e9.6788−1.4043 ln V = 15975V −1.4043 ,
and this function approximates the data in Table 12.4, or the points in Figure 12.67a. It is
important to notice that we did not apply the method of least squares directly to the function
P = b/V a and the data in Table 12.4. Were we to do so, it would be very difficult to solve
the resulting equations for a and b . Try it! Were we successful in doing so, values of a and b
would differ from those above, but not significantly.
Representation of data by other types of functions is discussed in the exercises.
EXERCISES 12.13
1. The following table shows the ages M (in months) and the average test scores S (obtained on a standard intelligence test) for children between
the ages of 9 and 12 years.
Age M (months) 108 112 116 120 124 128 132 136 140 144
Average score S 62 67 72 77 83 87 94 100 105 109
Find least-squares estimates for a linear function to describe S = S(M) .

2. The table below shows the average systolic blood pressure P of 13 children.
Age A (years) 4 5 6 7 8 9 10 11 12 13 14 15 16
Pressure P 85 87 90 92 95 98 100 105 108 110 112 115 118
Use least squares to find the equation of a straight line fitting these data.
3. The production of steel in the United States for the years 1946–1956 is shown in the table below.
Year 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
Steel (tonnes ×106 ) 55.2 76.9 80.5 70.9 88.0 95.6 84.7 101.5 80.3 106.4 104.7
Find the least-squares estimate for a linear function S = at + b to describe the data by (a) taking t = 0 in year zero and (b) taking t = 0 in year
1946. Plot the data and line.
∗ 4. (a) Plot the 16 points in the following table.
x 3.00 3.25 3.50 3.75 4.00 4.25 4.50 4.75 5.00 5.25 5.50 5.75 6.00 6.25 6.50 6.75
y 31.5 30.4 29.2 28.1 26.9 26.4 25.3 25.2 25.1 25.2 25.4 26.3 27.0 28.2 29.3 29.9
Do they seem to follow a parabolic path?

(b) If y = ax 2 + bx + c is the equation of a parabola that is to approximate the function y = f (x) described by these points, then the
following sum of differences between observed and predicted values is a measure of the accuracy of the fit:
16
5 0 12
S = S(a, b, c) = axi2 + bxi + c − y i ,
i=1
where (xi , y i ) are the points in the table. To find the best possible fit in the least-squares sense, we choose a , b , and c to minimize S .
Show that S has only one critical point (a, b, c) that is defined by the linear equations
" 16 $ " 16 $ " 16 $ 16
5 5 5 5
4 3 2
xi a+ xi b+ xi c = xi2 y i ,
i=1 i=1 i=1 i=1
" 16
$ " 16 $ " 16 $ 16
5 5 5 5
xi3 a + xi2 b+ xi c = xi y i ,
i=1 i=1 i=1 i=1
" 16
$ " 16 $ 16
5 5 5
xi2 a+ xi b + 16c = yi .
i=1 i=1 i=1
(c) Solve these equations for a , b , and c .
∗ 5. (a) Fit a least-squares quadratic (as in Exercise 4) to the following data.
x 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2
y 7.06 11.34 15.62 19.50 25.62 31.94 37.02 44.32 51.56 58.72 67.08 75.91
(b) Calculate the value of S at the critical values of a , b , and c .

∗ 6. The following table relates head H and discharge Q from a pump. Find the best-fitting parabola H = a + bQ2 for the data.
Q (L/s) 0 31.5 50.4 63.0 69.3 75.6 88.2 94.5

H (m) 37.8 34.4 30.8 27.3 24.7 21.3 13.1 7.8
∗ 7. (a) Fit a least-squares cubic polynomial y = ax 3 + bx 2 + cx + d to data in Exercise 5.

(b) Calculate the value of S at the critical values of a , b , c , and d . How does it compare with that in Exercise 5(b)?
∗ 8. Least squares can also be used with exponential functions. Suppose, for example, that we wish to fit a curve to the data in the following table:
x 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
y 140 180 230 290 365 455 565 670 785 1000 1230
(a) To determine whether the data can be described by a function of the form y(x) = beax , we take logarithms on both sides of the equation,
ln y = ln b + ax . If we define Y = ln y and B = ln b , then Y = ax + B , and this is the equation of a straight line in the xY -plane.
In other words, to test whether a set of points can be described by an exponential beax , we plot Y = ln y against x , and if these points
can be approximated by a straight line, then the original data should be describable by an exponential function. Do this for the data in
the table.
(b) Find least-squares estimates for a and B , and hence find the best-fitting exponential for the original data.
∗ 9. In the study of longshore sand transport on beaches, the following data were recorded at the El Moreno Beach on the Baja in California. Longshore
energy flux F is in units of joules per metre per second, and immersed weight transport W is in units of newtons per second.
F 6.0 15 18 20 30 38 43 104
W 6.1 9.9 20.8 14.6 25.7 42.8 45.1 84.4
Show that W can be represented in the form W = aF b and find least-squares estimates for a and b . Plot data points and the least-squares function
to show the adequacy of the fit.
10. The following table shows the number of kilometres per litre recorded by six trucks at speeds of 80, 90, and 100 km/h.
Vehicle 80 90 100
Truck 1 2.23 2.20 2.05
Truck 2 2.35 2.18 2.00
Truck 3 2.37 2.16 1.97
Truck 4 2.27 2.13 2.10
Truck 5 1.95 1.91 1.80
Truck 6 2.16 1.96 1.92
Use this information to solve Exercise 59 in Section 4.7. Use w = $20 and p = $0.60.
∗ 11. The following table represents census figures for the population (in millions) for the United States from 1790 to 1910.
Year 1790 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910
Population 3.9 5.3 7.2 9.6 12.9 17.1 23.2 31.4 39.8 50.2 62.9 76.0 92.0
Show that these data can be represented by an exponential function and find its least-squares estimates.
∗ 12. (a) To fit a power function y = bx a to data, we take logarithms, obtaining ln y = ln b + a ln x . When we set Y = ln y , B = ln b , and
X = ln x , this equation becomes Y = B + aX , that for a straight line. This suggests that we plot Y against X to see if data points
(X, Y ) are reasonably collinear. Do so for the data in the following table.
x 10 16 25 40 60
y 94 118 147 180 223
(b) Find least-squares estimates for a and b by fitting a straight line to Y as a function of X .
∗ 13. The following table represents the length of time t (in seconds) that an athlete could hold a load F (in newtons) in the position shown in the
figure below. Show that t can be represented in the form t = aF b , and find least-squares estimates for a and b .
F 70 100 200 300 400

F
t 470 288 84 52 32
∗ 14. The number N of bacteria per unit volume in a culture after t hours is tabulated below.
t 0 1 2 3 4 5 6
N 32 47 65 92 132 190 275
(a) Plot the data points (t, ln N) to show that it is reasonable to fit an equation of the form N = beat to the data.
(b) Find least-squares estimates for a and b .
∗ 15. The following table gives experimental values of the pressure P of a given mass of gas corresponding to various volumes V . Use least squares
to find estimates for constants a and b if thermodynamics suggests an equation of the form P V a = b to describe the data.
V 54.3 61.8 72.4 88.7 118.6 194.0

P 61.2 49.5 37.6 28.4 19.2 10.1
∗ 16. (a) You are given n pairs of observations (xi , y i ) and are required to fit a curve of the form y = b/x 2 to them. By direct application of
least squares, find a formula for b .
(b) Apply the formula in part (a) to the data below.
x 5 10 15 20 25 30
y 0.022 43 0.005 06 0.002 97 0.001 47 0.000 98 0.000 67
(c) Plot the data points and least-squares curve.

∗ 17. (a) Fit the curve y = ln(b + ax) to the following data by converting the given relation to linear form
ey = ax + b
and using as data ey i and xi .
x 2 3 4 5 6 7
y 1.952 2.156 2.413 2.549 2.670 2.821
(b) Can you use least squares directly on the logarithm function? Explain.
∗ 18. Devise a method for using least squares to obtain a curve of the form
1
y =
ax + b
to represent the following data.
x 5 6 7 8 9
y 1.335 1.431 1.247 1.197 1.118
∗∗ 19. A Cobb–Douglas production function has the form P (x, y) = kx q y 1−q , where P is the number of items produced per unit time, x is the
number of employees, and y is the operating budget for that time. The numbers k > 0 and 0 < q < 1 are constants. Find least-squares estimates
for k and q given the following production data.
Workers, x 100 110 90 100 95 105 110

Budget, y (dollars) 10 000 9000 9000 12 000 11 000 9500 10 000
Production, P 800 810 720 860 810 800 850
Hint: Write P /y = k(x/y)q and take logarithms.
12.14 Differentials
If y = f (x) is a function of one variable, the differential of y , defined by dy = f + (x) dx ,
is an approximation to the increment 'y = f (x + dx) − f (x) for small dx . In particular,
dy is the change in y corresponding to the change dx in x if we follow the tangent line to the
curve at (x, y) instead of the curve itself.
We take the same approach in defining differentials for multivariable functions. First con-
sider a function f (x, y) of two independent variables that can be represented geometrically as
a surface with equation z = f (x, y) (Figure 12.68). If we change the values of x and y by
amounts 'x = dx and 'y = dy , then the corresponding change in z is
'z = f (x + dx, y + dy) − f (x, y).
Geometrically, this is the difference in the heights of the surface at the points (x + dx, y + dy)
and (x, y) . If we draw the tangent plane to the surface at (x, y) , then very near (x, y) the
height of the tangent plane approximates the height of the surface (Figure 12.69). In particular,
the height of the tangent plane at (x + dx, y + dy) for small dx and dy approximates the
height of the surface. We define the differential dz as the change in z corresponding to the
changes dx and dy in x and y if we follow the tangent plane at (x, y) instead of the surface
itself. To find dz in terms of dx and dy , we note that the vector joining the points (x, y, z) and
(x + dx, y + dy, z + dz) has components (dx, dy, dz) , and this vector lies in the tangent
plane. Since a normal vector to the tangent plane is
∇(z − f (x, y)) = (−fx , −fy , 1),
it follows that the vectors (−fx , −fy , 1) and (dx, dy, dz) must be perpendicular. Conse-
quently,
∂f ∂f
0 = (−fx , −fy , 1) · (dx, dy, dz) = − dx − dy + dz,
∂x ∂y
and hence
∂f ∂f
dz = dx + dy. (12.72)
∂x ∂y
Note that if y is held constant in the function f (x, y) , then dy = 0 and 12.72 for dz reduces
to the definition of the differential of a function of one variable.
FIGURE 12.68 'z is exact differ- FIGURE 12.69 dz is difference
ence in heights of surface at (x, y) and at in heights of surface at (x, y) and tangent
(x + dx, y + dy) plane at (x + dx, y + dy)
z z
z = f (x, y)
z = f (x, y)
(x, y, z) (x, y, z)
∆z (dx, dy, dz)

(x + dx, y + dy, z + ∆z)
(x + dx, y + dy, z + dz)
(x, y) (x, y)
x dy x dy
dx dx
y y
(x + dx, y + dy) (x + dx, y + dy)
In Section 4.12 we indicated that we must be careful in using the differential dy as an

approximation for the change 'y in a function f (x) . For the same reasons, we must be
judicious in our use of dz as an approximation for 'z . Indeed, we have stated that dz is an
approximation for 'z for small dx and dy , but the difficulty is deciding how small is small and
how good is the approximation. In addition, note that if (x, y) is a critical point of the function
f (x, y) , then either dz = 0 for all dx and dy , or dz is undefined. In other words, dz cannot
be used to approximate 'z at a critical point.
EXAMPLE 12.38
If the radius of a right-circular cone is changed from 10 cm to 10.1 cm and the height is changed
from 1 m to 0.99 m, use differentials to approximate the change in its volume.
SOLUTION The volume of a cone of radius r and height h is given by the formula V =
π r 2 h/3. The differential of this function is
∂V ∂V 2 1
dV = dr + dh = π rh dr + π r 2 dh.
∂r ∂h 3 3
If r = 10, dr = 0.1, h = 100, and dh = −1, then
2 1 100π
dV = π(10)(100)(0.1) + π(10)2 (−1) = cm3 .
3 3 3
Equation 12.72 suggests the following definition for the differential of a function of more than
two independent variables.
DEFINITION 12.12
If u = f (x, y, z, t, . . . , w) , then the differential of f (x, y, z, t, . . . , w) is defined as
∂f ∂f ∂f ∂f ∂f
du = dx + dy + dz + dt + · · · + dw. (12.73)
∂x ∂y ∂z ∂t ∂w
EXAMPLE 12.39
1
FIGURE 12.70 Using The area of the triangle in Figure 12.70 is given by the formula A = ab sin θ . If when
2
differentials to approximate change 1 1
in area of a triangle θ = π/3, a and b are changed by % and θ by %, use differentials to find the approximate
3 2
percentage change in A .
a
SOLUTION Since
b ∂A ∂A ∂A
dA = da + db + dθ
∂a ∂b ∂θ
1 1 1
= b sin θ da + a sin θ db + ab cos θ dθ,
2 2 2
the approximate percentage change in A is

' ( ' (
dA 100 1 1 1
100 = b sin θ da + a sin θ db + ab cos θ dθ
A A 2 2 2
' (
da db
= 100 + + cot θ dθ .
a b
1 1
Since a and b are changed by % and θ by %,
3 2
' ( ' ( ' (
da db 1 dθ 1
100 = 100 = and 100 = .
a b 3 θ 2
Thus, ' (
dA 1 1 θ 2 θ
100 = + + cot θ = + cot θ,
A 3 3 2 3 2
and when θ = π/3,
' ( ' (' (
dA 2 1 π 1
100 = + √ = 0.97.
A 3 2 3 3
The approximate percentage change in A is therefore 1%.
When n resistances Ri ( i = 1, . . . , n ) are connected in series, then the resultant resis-

tance is
R = R 1 + R2 + · · · + R n .
When they are connected in parallel, the effective resistance, R, is given by
1 1 1 1
= + + ··· + .
R R1 R2 Rn
We are being asked the following question. If each resistance Ri is changed by the same
small percentage c , what is the percentage change in the series and parallel combinations?
SOLUTION Let us consider the series case first. Because percentage changes in the Ri
are small, we shall use differentials to calculate the percentage change in R ,
∂R ∂R
dR = dR1 + · · · + dRn = dR1 + · · · + dRn .
∂R1 ∂Rn
The percentage change in R is therefore
dR 100(dR1 + · · · + dRn )
100 =
R R1 + · · · + Rn
100dR1 100dR2
= ' (+ ' (
R2 Rn R1 Rn
R1 1+ + ··· + R2 + 1 + ··· +
R1 R1 R2 R2
100dRn
+ ··· + ' (
R1 Rn−1
Rn + ··· + +1
Rn Rn
100dRi
Since the percentage change in each resistance Ri is c, it follows that Ri = c . Hence,
100dR c c
= +
R R2 Rn R1 Rn
1+ + ··· + + 1 + ··· +
R1 R1 R2 R2
c
+ ··· +
R1 Rn−1
+ ··· + +1
Rn Rn
cR1 cR2 cRn

= + + ··· + = c.
R1 + · · · + Rn R1 + · · · + Rn R1 + · · · + Rn
' (−1
1 1
For the parallel case, we write that R = + ··· + , and take differentials,
R1 Rn
' (−2 ' (
1 11 1
dR = − + ··· + − 2 dR1 − · · · − 2 dRn
R1 Rn R1 Rn
' (
1 1
= R2 2
dR1 + · · · + 2 dRn .
R1 Rn
The percentage change in R is therefore

' (
dR 1 1
100 = 100R dR1 + · · · + 2 dRn
R R12 Rn
' ( ' (
R dR1 R dRn
= 100 + ··· + 100
R1 R1 Rn Rn
cR cR
= + ··· +
R1 Rn
' (
1 1
= cR + ··· + = c.
R1 Rn
Thus, whether resistances are connected in series or parallel, small equal percentage
changes in the individual resistances results in the same percentage change in the resultant
resistance.
EXERCISES 12.14
In Exercises 1–10 find the differential of the function. 8. f (x, y, z, t) = xy + yz + zt + xt
1. f (x, y) = x 2 y − sin y 9. f (x, y, z, w) = xy tan (zw)

2 +y 2 +z2 −t 2
10. f (x, y, z, t) = ex
2. f (x, y) = Tan −1 (xy)
3. f (x, y, z) = xyz − x 3 ez 11. A right-circular cone has radius 10 cm and height 20 cm. If its
radius increases by 0.1 cm and its height decreases by 0.3 cm, use
4. f (x, y, z) = sin (xyz) − x 2 y 2 z2 differentials to find the approximate change in its volume. Compare
this with the actual change in volume.
5. f (x, y, z) = ln (x 2 + y 2 + z2 )
12. When the ellipse b2 x 2 + a 2 y 2 = a 2 b2 is rotated about the x -
6. f (x, y) = Sin −1
(xy) axis, the volume V of the spheroid is 4π ab2 /3. If a and b are each
increased by 1%, use differentials to find the approximate percentage
7. f (x, y) = Sin −1 (x + y) + Cos −1 (x + y) change in V .
12.15 Taylor Series for Multivariable Functions

Taylor series for functions of one variable can be used to generate Taylor series for multivariable
functions. For simplicity, we once again work with functions of two independent variables.
Extensions to functions of more than two independent variables will be clear. Suppose that a
function f (x, y) has continuous partial derivatives of all orders in some open circle centred at
the point (c, d) (Figure 12.71).
12.15 Taylor Series for Multivariable Functions 891
FIGURE 12.71 Finding the Taylor series of a function f (x, y) about a point (c, d)
(c, d )
Parametric equations for the line through (c, d) in direction v = (vx , vy ) are
x = c + vx t, y = d + vy t.
If we substitute these values into f (x, y) , we obtain a function F (t) of one variable,
F (t) = f (c + vx t, d + vy t),
which represents the value of f (x, y) at points along the line through (c, d) in direction v. If
we expand this function into its Maclaurin series, we obtain
F ++ (0)
F (t) = F (0) + F + (0)t + t2 + · · · . (12.74)
2!
The schematic diagram to the left gives

F
∂F dx ∂F dy
x y F + (t) = + = fx (x, y)vx + fy (x, y)vy ,
∂x dt ∂y dt
t t and therefore
F + (0) = fx (c, d)vx + fy (c, d)vy .
The schematic for F + (t) gives the second derivative of F (t) ,
F +(t)
∂ dx ∂ dy
x y
F ++ (t) = [F + (t)] + [F + (t)]
∂x dt ∂y dt
t t ∂ ∂
= [fx (x, y)vx + fy (x, y)vy ]vx + [fx (x, y)vx + fy (x, y)vy ]vy
∂x ∂y
= fxx (x, y)vx2 + 2fxy (x, y)vx vy + fyy (x, y)vy2 ,
and therefore
F ++ (0) = fxx (c, d)vx2 + 2fxy (c, d)vx vy + fyy (c, d)vy2 .
A similar calculation gives
F +++ (0) = fxxx (c, d)vx3 + 3fxxy (c, d)vx2 vy + 3fxyy (c, d)vx vy2 + fyyy (c, d)vy3 ,
and the pattern is emerging. When these results are substituted into 12.74,
F (t) = f (c + vx t, d + vy t)
= f (c, d) + [fx (c, d)vx + fy (c, d)vy ]t
t2
+ [fxx (c, d)vx2 + 2fxy (c, d)vx vy + fyy (c, d)vy2 ] + ···. (12.75)
2!
We now let v be the vector from (c, d) to point (x, y) , so that vx = x − c and vy = y − d ,
and at the same time set t = 1. Then F (1) = f (c + x − c, d + y − d) = f (x, y) , and
12.75 becomes
1
f (x, y) = f (c, d) + [fx (c, d)(x − c) + fy (c, d)(y − d)] + [fxx (c, d)(x − c)2
2!
+ 2fxy (c, d)(x − c)(y − d) + fyy (c, d)(y − d)2 ] + · · · . (12.76)
This is the Taylor series for f (x, y) about the point (c, d) . It gives the value of the
function at the point (x, y) in terms of values of the function and its derivatives at the point
(c, d) .
EXAMPLE 12.40
Find the first six nonzero terms in the Taylor series for f (x, y) = sin(2x + 3y) about (0, 0) .
SOLUTION We calculate that
f (0 , 0 ) = 0 ,
fx (0, 0) = 2 cos(2x + 3y)|(0,0) = 2,
fy (0, 0) = 3 cos(2x + 3y)|(0,0) = 3,
fxx (0, 0) = −4 sin(2x + 3y)|(0,0) = 0,
fxy (0, 0) = −6 sin(2x + 3y)|(0,0) = 0,
fyy (0, 0) = −9 sin(2x + 3y)|(0,0) = 0,
fxxx (0, 0) = −8 cos(2x + 3y)|(0,0) = −8,
fxxy (0, 0) = −12 cos(2x + 3y)|(0,0) = −12,
fxyy (0, 0) = −18 cos(2x + 3y)|(0,0) = −18,
fyyy (0, 0) = −27 cos(2x + 3y)|(0,0) = −27.
Formula 12.76 then gives
1 1
sin(2x + 3y) = 0 + [2x + 3y ] + [0] + [−8x 3 − 36x 2 y − 54xy 2 − 27y 3 ] + · · ·
2! 3!
1
= (2x + 3y) − (2x + 3y)3 + · · · .
3!
This series could also have been obtained by substituting 2x + 3y for x in the Maclaurin series
for sin x . This is not always an alternative.
Summary 893
EXERCISES 12.15
√
1. If f (x, y) = F (x)G(y) , is the Taylor series of f (x, y) about ∗ 10. 1 + xy about (2, 1)
(0, 0) the product of the Maclaurin series for F (x) and G(y) ?
∗ 11. ex sin (3x − y) about (−1, 0)
2. What are the cubic terms in 12.76?
∗ 12. (x + y)2 ln (x + y) about (0, 1)
In Exercises 3–8 find the Taylor series of the function about the point
∗ 13. Tan −1 (3x + 2y) about (1, −1)
by using Taylor series for functions of one variable.
3. cos (xy) about (0, 0) 4. e2x−3y about (1, −1)

∗ 14. x 8 y 10 about (0, 0)
√
5. x 2 y 1 + x about (0, 0) 6. ln (1 + x 2 + y 2 ) about (0, 0)
∗ 15. What are the terms in the Taylor series for a function f (x, y, z)
1 xy 2 about the point (c, d, e) corresponding to those in equation 12.76?
7. about (3, −4) 8. about (−1, 0)
1+x+y 1 + y2
∗ 16. Express 12.76 in sigma notation. Hint: Think about an operator,
In Exercises 9–14 find the Taylor series of the function up to and in-
) +n
∂ ∂
cluding quadratic terms. (x − a) + (y − b) ,
∂x ∂y
xy
∗ 9. about (−1, 1)
x2 + y2 which is expanded as a binomial to operate on functions f (x, y) .
SUMMARY
We began the study of multivariable functions in this chapter, concentrating our attention on
differentiation and its applications. We introduced two types of derivatives for a multivari-
able function: partial derivatives and directional derivatives. Partial derivatives are directional
derivatives in directions parallel to the coordinate axes. The directional derivative of a function
f (x, y, z) in the direction v is given by the formula Dv f = ∇f · v̂, where v̂ is the unit vector
in the direction of v, and the gradient ∇f is evaluated at the point at which Dv f is required.
This formula leads to the fact that the gradient ∇f (x, y, z) points in the direction in which
f (x, y, z) increases most rapidly, and |∇f | is the (maximum) rate of change of f (x, y, z) .
A second property of gradient vectors (that is related to the first) is that if F (x, y, z) = C , C a
constant, is the equation of a surface, then at any point on the surface ∇F is perpendicular to the
surface. This property, along with the fact that perpendicularity to a surface is synonymous with
perpendicularity to its tangent plane, enables us to find equations for tangent planes to surfaces
and tangent lines to curves.
We illustrated various ways to calculate partial derivatives of a multivariable function,
depending on whether the function is defined explicitly, implicitly, or as a composite function.
Since partial derivatives are ordinary derivatives with other variables held constant, there is no
difficulty calculating partial derivatives when the function is defined explicitly; we simply use
the rules from single-variable calculus. When the partial derivative of a composite function
is required, we use a schematic diagram illustrating functional dependences to develop the
appropriate chain rule. Partial derivatives for functions defined implicitly are calculated using
Jacobians.
Critical points of a multivariable function are points at which all of its first partial derivatives
vanish or at which one or more of these partial derivatives does not exist. Critical points can
yield relative maxima, relative minima, saddle points, or none of these. For functions of two
independent variables, a second-derivative test exists that may determine whether a critical point
at which the partial derivatives vanish yields a relative maximum, a relative minimum, or a saddle
point. This test is analogous to that for functions of one variable.
A continuous function of two independent variables defined on a region that includes its
boundary always takes on a maximum value and a minimum value. To find these values we
evaluate the function at each of its critical points and compare these numbers to the maximum and
minimum values of the function on its boundary. Finding the extreme values on the boundary
involves one or more extrema problems for a function of one variable, the number of such
problems depending on the complexity of the boundary.
There are two methods for finding extreme values of a function when the variables of
the function are subject to constraints: solve the constraint equations for dependent variables
and express the given function in terms of independent variables, or use Lagrange multipliers.
Lagrange multipliers eliminate the necessity for solving the constraint equations, but they do,
on the other hand, give a larger system of equations to solve for critical points.
Differentials of multivariable functions can be used to approximate changes in functions
when small changes are made to its independent variables. Taylor series can also be used to
approximate multivariable functions.
The method of least squares fits a function y = f (x) of known form to a set of data. It
minimizes the sum of the squares of the differences between measured and predicted values
of y .
KEY TERMS
Domain Limit
Continuous function Partial derivative
Gradient Second partial derivative
Laplace’s equation Harmonic function
Chain rule Positively homogeneous function
Euler’s Theorem Implicit differentiation
Jacobian determinant Directional derivative
Tangent line to a curve Tangent plane to a surface
Normals to surfaces Smooth surface
Piecewise-smooth surface Critical point
Relative maximum Relative minimum
Saddle point Absolute maximum
Absolute minimum Constraint problems
Lagrange multipliers Lagrangian
Least squares Differential
Taylor series
REVIEW
EXERCISES
In Exercises 1–20 find the derivative. 12. ∂r/∂x)y if x = r cos θ, y = r sin θ

2 3 −1
1. ∂f/∂x if f (x, y) = x /y − Sin (xy) 13. ∂θ/∂x)y,z if x = r sin φ cos θ, y = r sin φ sin θ , z = r cos φ
2 2 2 2 2
2. ∂ f/∂y if f (x, y, z) = ln (x + y + z ) 14. ∂u/∂r)θ if u = x 2 − y 2 x 3 , x = r cos θ, y = r sin θ
3. ∂ 3 f/∂x 2 ∂y if f (x, y, z, t) = x 3 ey −xzt 2 −sin (x + y + z + t)
15. ∂ 2 u/∂r 2 )θ if u = x 2 − y 2 x 3 , x = r cos θ, y = r sin θ
2 −1
4. ∂z/∂x if z x + Tan z + y = 3x
16. d 2 u/dt 2 if u = x/z2 − z/x 2 , x = t 3 − 3, z = 1/t 3
2
5. ∂u/∂y if u cos y + y cos (xu) + z = 5x
17. dz/dt if z = y − xy 2 + x, and x 2 − y 2 + xt = 2t, xy = 4t 2
6. df/dt if f (x, y) = x 2 + y 2 − exy , x = t 3 + 3t, y = t ln t
18. ∂ 2 z/∂x 2 if xz − x 2 z3 + y 2 = 3
7. dy/dx if x = y 3 + 3y 2 − 2y + 4
19. dy/dx if yx − x 2 z2 + 5x = 3, 2xz − 3x 2 y 2 = 4z4
8. ∂u/∂x)y if u2 + v 2 − xy = 5, 3u − 2v + x 2 u = 2v 3
√ √ 20. ∂u/∂t)v if u = xyt 2 − 3 Sin −1 (xy), x = v 2 t 2 − 2t, y =
9. ∂ 2 f/∂u∂v if f (u, v) = u2 / v − v/ u
v tan t
10. df/dt if f (x, y) = xy − x 2 − y 2 , x = tet , y = te−t
∗ 21. If u = (x 2 + y 2 )[1 + sin (x/z)], show that
2 2
11. ∂z/∂t)u if z = x − y , x = 2u − 3v + 3uvt , y = 2 ∂u ∂u ∂u
x +y +z = 2u .
u cos (vt), v = t 2 − 2t ∂x ∂y ∂z
∗ 22. If u = 2x 2 − 3y 2 + xy , show that In Exercises 37–40 find all critical points for the function and classify
∂ 2u ∂ 2u ∂ 2u each as yielding a relative maximum, a relative minimum, or a saddle
x 2 2 + 2xy + y 2 2 = 2u . point.
∂x ∂x ∂y ∂y
∗ 23. If f (s) is a differentiable function, show that f (3x − 2y) satisfies 37. f (x, y) = x 3 + 3y 2 − 6x + 4
∂f ∂f
2 +3 = 0. 38. f (x, y) = yex
∂x ∂y
39. f (x, y) = x 2 − xy + y 2 + x − 4y
∗ 24. If f (s, t) has continuous first partial derivatives, show that the
∂f ∂f 40. f (x, y) = (x 2 + y 2 − 1)2
function f (x 2 − y 2 , y 2 − x 2 ) satisfies y +x = 0.
∂x ∂y
∗ 41. If f (x, y) = (x 2 +y 2 )F (x, y) where F (x, y) = x 3 /y −y 3 /x ,
verify that
In Exercises 25–30 find the directional derivative.
' 2 (
25. f (x, y) = x 2 sin y at (3, −1) in the direction v = (2, 4) ∂ 2f ∂ 2f 2 2 ∂ F ∂ 2F
+ = (x + y ) + + 12F (x, y).
∂x 2 ∂y 2 ∂x 2 ∂y 2
26. f (x, y, z) = x 2 + y 2 + z2 at (1, 0, 1) in the direction from
(1, 0, 1) to (2, −1, 3) ∗ 42. Find maximum and minimum values of the function f (x, y) =
−1 xy on the circle x 2 + y 2 ≤ 1.
∗ 27. f (x, y, z) = z Tan (x + y) at (−1, 2, 5) in the direction per-
pendicular to the surface z = x 2 + y 2 with positive z -component ∗ 43. Find maximum and minimum values of the function f (x, y, z) =
2x + 3y − 4z on the sphere x 2 + y 2 + z2 ≤ 2.
∗ 28. f (x, y, z) = x 2 +y − 2z at (1, −1, 2) along the line x −y +z =
4, 2x + 4y + 2 = 0 in the direction of increasing x ∗ 44. Find the points on the curve x 2 + x 4 + y 2 = 1 closest to and
farthest from the origin.
∗ 29. f (x, y) = ln (x + y) at (3, 10) along the curve y = x 2 + 1 in
the direction of decreasing y ∗ 45. Find the point(s) on the surface z2 = 1 + xy closest to the origin.
∗ 46. Generalize Review Exercise 32 in Chapter 4 to incorporate a third
∗ 30. f (x, y, z) = 2xyz − x 2 − z2 at (0, 1, 1) along the curve x 2 +
crop, say sunflowers, with a yield of r dollars per hectare and a pro-
y 2 + z2 = 2, y = z in the direction of increasing x
portional loss cz .
∗ 47. If the equation u = f (x − ut) defines u implicitly as a function
In Exercises 31–33 find the equation of the tangent plane to the surface. of x and t , show that
31. z = x 2 + y 2 at (1, 3, 10) 32. x 2 + z3 = y 2 at (−1, 3, 2) ∂u ∂u

+u = 0.
33. x 2 + y 2 = z2 + 1 at (1, 0, 0) ∂t ∂x
∗ 48. Find the first six nonzero terms in the Taylor series for x 3 sin (x 2 y)
about the point (1, π/4) .
In Exercises 34–36 find equations for the tangent line to the curve.
∗ 49. Find the best possible line, in the least-squares sense, to fit the data
34. x = t 2 + 1, y = t 2 − 1, z = t 3 + 5t at (2, 0, 6) in the following table.
35. x + y + z = 0, 2x − 3y − 6z = 11 at (1, 1, −2) x 1 2 3 4 5 6 7 8
2 2
36. z = xy, x + y = 2 at (1, 1, 1) y 1.2 4.6 8.4 12.2 15.6 19.7 23.0 26.9
CHAPTER
13 Multiple Integrals
Application Preview In the figure below, a freshwater marsh is drained to the ocean through an automatic tide gate
that is L metres wide and 0.9 m high. The gate is held by an L -metre-long hinge at A and
bears on a sill at B . The water level in the marsh is 1.8 m, and the density of ocean water
is 1030 kg/m 3 compared to 1000 kg/m 3 for the fresh water in the marsh. The marsh and the
ocean both create forces on the gate, and when water levels are the same, the ocean creates a
greater force, thus keeping the gate closed. As the ocean level falls, the force it exerts on the
gate decreases, and eventually the gate opens.
Marsh
1.8 A Ocean
h
0.9
THE PROBLEM At what depth of ocean water will the tide gate open? (See Example 13.8
on page 914 for the solution.)
The definite integral of a function f (x) of one variable is defined as the limit of a sum of
the form
f (x1∗ ) !x1 + f (x2∗ ) !x2 + · · · + f (xn∗ ) !xn , (13.1)
where the norm of the partition approaches zero. We have seen that definite integrals can be
used to calculate area, volume, work, fluid force, and moments. In spite of the fact that some of
these are two- and three-dimensional concepts, we have been careful to emphasize that a definite
integral with respect to x is an integration along the x -axis, and a definite integral with respect
to y is an integration along the y -axis. In other words, independent of how we interpret the
result of the integration, a definite integral is a limit summation along a line. Generalizations
of these limiting sums to functions of two and three independent variables lead to definitions of
double and triple integrals.
13.1 Double Integrals and Double Iterated Integrals

Suppose a function f (x, y) is defined in some region R of the xy -plane that has finite area
(Figure 13.1). To define the double integral of f (x, y) over R , we first divide R into n
subregions of areas !A1 , !A2 , . . . , !An , in any manner whatsoever. In each subregion
!Ai (i = 1, . . . , n) we choose an arbitrary point (xi∗ , yi∗ ) and form the sum
n
!
f (x1∗ , y1∗ ) !A1 + f (x2∗ , y2∗ ) !A2 + · · · + f (xn∗ , yn∗ ) !An = f (xi∗ , yi∗ ) !Ai .
i=1
(13.2)
The norm of the partition of R into subareas !Ai is the area of the largest of the subareas,
denoted by "!Ai " = maxi=1,...,n !Ai .
896
Suppose we increase the number of terms in 13.2 by increasing the number of subareas
FIGURE 13.1 Subdivide
R into smaller regions to define !Ai and decreasing the norm "!Ai " . If the sum approaches a limit as the number of subareas
the double integral of f (x, y) becomes increasingly large and each subarea shrinks to a point, we call the limit the double
over R integral of f (x, y) over the region R , and denote it by
y
"" n
!
f (x, y) dA = lim f (xi∗ , yi∗ ) !Ai . (13.3)
R "!Ai "→0
i=1
∆ A1
∆ Ai ∆ A2 The notation "!Ai " → 0 does not necessarily require that every !Ai shrink to a point. We
(x*i, y*i ) implicitly assume, however, that this is always the case.
∆ A3
In some texts the R is placed below the integral signs as in
R ""
x f (x, y) dA.
R
We will use the notation in equation 13.3, but be aware of the alternative.
If the limit in 13.3 were dependent on the choice of subdivision !Ai or choice of star points
(xi∗ , yi∗ ) , double integrals would be of little use. We therefore demand that the limit of the sum
be independent of the manner of subdivision of R and choice of star points in the subregions.
The following theorem indicates that for continuous functions this is always the case.
THEOREM 13.1
Let C be a closed, piecewise-smooth curve that encloses a region R with finite area. If
f (x, y) is a continuous function inside and on C , then the double integral of f (x, y)
over R exists.
For a continuous function, then, the double integral exists, and any choice of subdivision
and star points leads to the same value through limiting process 13.3. Note that continuity was
also the condition that guaranteed existence of the definite integral in Theorem 6.2.
We cannot overemphasize the fact that a double integral is simply the limit of a sum.
Moreover, any limit of form 13.3 may be interpreted as the double integral of a function f (x, y)
over the region defined by the !Ai .
The following properties of double integrals are easily proved using definition 13.3:
1. If the double integral of f (x, y) over R exists and c is a constant, then
"" ""
cf (x, y) dA = c f (x, y) dA. (13.4)
R R
2. If double integrals of f (x, y) and g(x, y) over R exist, then

"" "" ""
[f (x, y) + g(x, y)] dA = f (x, y) dA + g(x, y) dA. (13.5)
R R R
3. If a region R is subdivided by a piecewise-smooth curve into two parts R1 and R2 that
have at most boundary points in common (Figures 13.2), and the double integral of f (x, y)
over R exists, then
"" "" ""
f (x, y) dA = f (x, y) dA + f (x, y) dA. (13.6)
R R1 R2
4. The area of a region R can be obtained by integrating the function f (x, y) = 1 over R :
""
area of R = dA. (13.7)
R
In spite of the fact that double integrals are defined as limits of sums, we do not evaluate them as
such. Just as definite integrals are evaluated with antiderivatives, we evaluate double integrals
with double iterated integrals.
898 Chapter 13 Multiple Integrals
FIGURE 13.2a A region R FIGURE 13.2b Division of R into two parts
y y
R R1 R2
x x
Double Iterated Integrals

We have already seen that a function f (x, y) of two independent variables has two first-order
partial derivatives, one with respect to x holding y constant, and one with respect to y holding
x constant. We now reverse this process and define “partial” integration of f (x, y) with respect
to x and y . Quite naturally, we define a partial antiderivative of f (x, y) with respect to x as
an antiderivative of f (x, y) with respect to x , holding y constant. For example, since
∂ 3
(x + x 2 y) = 3x 2 + 2xy,
∂x
x 3 + x 2 y is an antiderivative with respect to x of 3x 2 + 2xy . But so is x 3 + x 2 y + y . In fact,
for any differentiable function C(y) of y , x 3 + x 2 y + C(y) is an antiderivative of 3x 2 + 2xy
with respect to x . Since this expression represents all antiderivatives of 3x 2 + 2xy , we call it
the partial indefinite integral of 3x 2 + 2xy with respect to x , and write
"
(3x 2 + 2xy) dx = x 3 + x 2 y + C(y).
Similarly, the partial indefinite integral of 3x 2 + 2xy with respect to y is

"
(3x 2 + 2xy) dy = 3x 2 y + xy 2 + D(x),
where D(x) is an arbitrary differentiable function of x .

In this chapter we are concerned only with partial definite integrals. Limits on a partial
definite integral with respect to x must not depend on x , but may depend on y . In general, then,
a partial definite integral with respect to x is of the form
" h(y)
f (x, y) dx; (13.8)
g(y)
and a partial definite integral with respect to y is of the form

" h(x)
f (x, y) dy. (13.9)
g(x)
Each of these partial definite integrals is evaluated by substituting the limits into a corre-
sponding antiderivative. For example,
" x+2
2
(2y + xey ) dy = {y 2 + xey }x+
x2
x2
2
= {(x + 2)2 + xex+2 } − {(x 2 )2 + xex }
2
= (x + 2)2 − x 4 + x(ex+2 − ex ).
Once antidifferentiation in 13.8 is completed and the limits substituted, the result is a function
of y alone. It is then possible to integrate this function with respect to y between any two limits,
say from y = c to y = d :
" d #" h(y) $
f (x, y) dx dy.
c g(y)
In practice we omit the braces and simply write

" d" h(y)
f (x, y) dx dy, (13.10)
c g(y)
understanding that in the evaluation we proceed from the inner integral to the outer. This is
called a double iterated integral first with respect to x and then with respect to y (or, more
concisely, with respect to x and y ). Double iterated integrals with respect to y and x take the
form " " b h(x)
f (x, y) dy dx. (13.11)
a g(x)
EXAMPLE 13.1
Evaluate each of the following double iterated integrals:
" 1" x−1 " 0" 0
2 y x
(a) (x + e ) dy dx (b) % dx dy
0 x −1 y x + y2
2
SOLUTION
" 1" x−1 " 1
2
(a) y
(x + e ) dy dx = {x 2 y + ey }x−
x
1
dx
0 x 0
" 1
= [x 2 (x − 1) + ex−1 − x 3 − ex ] dx
0
" 1
= (−x 2 + ex−1 − ex ) dx
0
# $1
x3 x−1 x
= − +e −e
3 0
& '
1
= − + 1 − e − (e−1 − 1)
3
5
= − e − e −1
3
" 0" 0 " 0 (% )0 " 0 % %
x
(b) % dx dy = x 2 + y 2 dx = ( y 2 − 2y 2 ) dy
−1 y x2 + y2 −1 y −1
" 0 √ √ " 0
= (−y + 2y) dy = ( 2 − 1) y dy
−1 −1
# $0 √
√ y2 1− 2
= ( 2 − 1) =
2 −1 2
EXERCISES 13.1
" 0" 4x 2 % " 0" 0
In Exercises 1–30 evaluate the double iterated integral. x
∗ 27. y − x 4 dy dx ∗ 28. % dx dy
" 2" y+2 " 3 " √18−2y 2 −2 x 4 −2 y x2 + y2
1. (x 2 − xy) dx dy 2. √ x dx dy " 2" y3 % " 1" x %
−1 y −3 − 18−2y 2 ∗ 29. 1 + y dx dy ∗∗ 30. x 2 + y 2 dy dx
" 1" x " 0" 2 −1 −1 0 0
3. (2xy + 3y 2 ) dy dx 4. (1 + y)2 dx dy ∗ 31. In two-dimensional, steady-state, incompressible flow, the velocity

0 x2 −1 y
" 4" " 2" v of the flow has two components, v = u(x, y)î + v(x, y)ĵ , which
π/2 y must satisfy the continuity equation
5. x sin y dy dx 6. ex+y dx dy
3 0 1 1
" 1" " 1" ∂u ∂v
5 2x + = 0.
7. 2
(x + y ) dy dx 2
8. 3 3
(xy + x y ) dy dx ∂x ∂y
−1 −x −1 x
" 1" 1 " 2" 2x If u(x, y) = kx , where k is a constant, find all possible functions
1
9. (x + y) dy dx 4
10. dy dx v(x, y) .
0 x 1 x (x + y)3
" 1" 3x " 1" e
√ y
11. x + y dy dx 12. dx dy In Exercises 32–35 you are given one of the velocity components for
0 0 −1 1 x
two-dimensional, steady-state, incompressible flow v = u(x, y)î +
" 4" x2 v(x, y)ĵ . Use the continuity equation of Exercise 31 to find all possible
13. √
(x 2 + 2xy − 3y 2 ) dy dx values for the other component.
1 x
" 2" 2x 2 ∗ 32. u(x, y) = x 2 + y 2
14. x cos y dy dx
0 x2 ∗ 33. u(x, y) = Tan −1 (y/x)
" 1" tan x " 1" y3 %
1 1 ∗ 34. v(x, y) = x x 2 + y 2
15. dy dx 16. dx dy
0 1 1+ y2 0 0 1 + y2
" 3" " 2" ∗ 35. v(x, y) = sin x cos y
1 x
x 2 3/ 2
17. % dy dx 18. (8 − 2x ) dy dx
2 0 1 − y2 0 −x
√
Stream functions ψ(x, y) for two-dimensional, steady-state, incom-
" 1" x " 0" x 2 9+x pressible flow satisfy
1
∗ 19. % dy dx ∗ 20. dy dx
0 0 1 − y2 −9 0
∂ψ ∂ψ
" 2" 2 " 0" 0 % = −v(x, y), = u(x, y),
∂x ∂y
∗ 21. √ y 2 dy dx ∗ 22. x x 2 + y 2 dx dy
0 4−x 2 −1 y
" 3" " 1" where v = u(x, y)î + v(x, y)ĵ is the velocity of the flow. In Exer-
2x Cos−1 x
1 cises 36–39 find all stream functions for the flow with given velocity.
∗ 23. dy dx ∗ 24. x cos y dy dx
2 1 (xy + x 2 )2 0 0
" 1" √ ∗ 36. v = x î − y ĵ
2y %
3
∗ 25. √ x x 2 − y 2 dx dy ∗ 37. v = (x 2 + y 2 )î − 2xy ĵ
0 y 2 +y
% %
" 1" √y 2 +y % ∗ 38. v = −y x 2 + y 2 î + x x 2 + y 2 ĵ
∗ 26. √ x 3 x 2 − y 2 dx dy
0 2y ∗ 39. v = − cos x sin y î + (sin x cos y + x)ĵ
13.2 Evaluation of Double Integrals by Double Iterated

Integrals
According to Theorem 13.1, if a function f (x, y) is continuous on a finite region R with a
piecewise-smooth boundary, then double integral 13.3 exists, and its evaluation by means of
that limit is independent of both the manner of subdivision of R into areas !Ai and choice of
starpoints (xi∗ , yi∗ ) . We now show that if we make particular choices of !Ai , double integrals
can be evaluated by means of double iterated integrals in x and y .
FIGURE 13.3 Proof that double integrals can be evaluated with double iterated integrals
y
ym = d
ym−1
R
(xi , yj )
yj
yj−1
y2
y1
y0 = c
x1 x2 xi x
a = x0 xi−1 xn−1 b = xn
Consider first a rectangle R with edges parallel to the x - and y -axes as shown in Figure 13.3.
We divide R into smaller rectangles by a network of n + 1 vertical lines and m + 1 horizontal
lines identified by abscissae,
a = x0 < x1 < x2 < · · · < xn−1 < xn = b,
and ordinates,
c = y0 < y1 < y2 < · · · < ym−1 < ym = d.
If the (i, j )th rectangle is that rectangle bounded by the lines x = xi−1 , x = xi , y = yj −1 ,
and y = yj , then its area is !xi !yj , where !xi = xi − xi−1 and !yj = yj − yj −1 . We
choose as star point in the (i, j )th rectangle the upper right corner: (xi∗ , yj∗ ) = (xi , yj ) . With
this rectangular subdivision of R and choice of star points, Definition 13.3 for the double integral
of f (x, y) over R takes the form
"" n !
! m
f (x, y) dA = lim f (xi , yj ) !xi !yj . (13.12a)
R "!xi !yj "→0
i=1 j =1
Since "!xi !yj " → 0 if the norms "!xi " and "!yj " individually approach zero, we
can write that
"" n !
! m
f (x, y) dA = lim f (xi , yj ) !xi !yj . (13.12b)
R "!xi "→0
"!yj "→0 i=1 j =1
Suppose we choose to first perform the limit on y and then the limit on x , and therefore write
 
"" n 
! m
! 
f (x, y) dA = lim lim f (xi , yj )!yj !xi .
R "!xi "→0 "!yj "→0 
i=1 j =1
Since xi is constant in the limit with respect to y , the y -limit is the definition of the definite
integral of f (xi , y) with respect to y from y = c to y = d ; that is,
m
! " d
lim f (xi , yj ) !yj = f (xi , y) dy.
"!yj "→0 c
j =1
Consequently,
"" n #"
! d $
f (x, y) dA = lim f (xi , y)dy !xi .
R "!xi "→0 c
i=1
Because the term in braces is a function of xi alone, we can interpret this limit as a definite
integral with respect to x :
"" " b #" d $ " b" d
f (x, y) dA = f (x, y) dy dx = f (x, y) dy dx, (13.13)
R a c a c
a double iterated integral. By reversing the order of taking limits, we can show similarly that
the double integral can be evaluated with a double iterated integral with respect to x and y :
"" " d" b
f (x, y) dA = f (x, y) dx dy. (13.14)
R c a
We have shown, then, that for the special case of a rectangle R with sides parallel to the
axes, a double integral over R can be evaluated by using double iterated integrals. Conversely,
every double iterated integral with constant limits represents a double integral over a rectangle.
The double iterated integral simply indicates that a rectangular subdivision has been chosen to
evaluate the double integral.
We have just stated that the choice of a double iterated integral to evaluate a double integral
implies that the region of integration has been subdivided into small rectangles. We now show
that the x - and y -integrations themselves can be interpreted geometrically. These interpretations
will simplify the transition to more difficult regions of integration.
In the subdivision of R into rectangles, suppose we denote the dimensions of a representative
rectangle at position (x, y) by dx and dy (Figure 13.4). In the inner integral
" d
f (x, y) dy dx
c
of equation 13.13, x is held constant and integration is performed in the y -direction. This
(partial) definite integral is therefore interpreted as summing over the rectangles in the vertical
strip of width dx at position x . The limits y = c and y = d identify the initial and terminal
positions of this vertical strip. It is important to note that we are not adding the areas of the
rectangles of dimensions dx and dy in the strip. On the contrary, each rectangle of area dy dx
is multiplied by the value of f (x, y) for that rectangle,
f (x, y) dy dx,
and it is these quantities that are added.
FIGURE 13.4 Addition process in a double iterated integral with respect to y and x over a rectangle
y
R
d
dy
(x, y)
a dx b x
The x -integration in equation 13.13 is interpreted as adding over all strips starting at x = a
FIGURE 13.5 Addition
and ending at x = b . The limits on x therefore identify positions of the first and last strips.
process in a double iterated inte- Although our diagram illustrates finite rectangles of dimensions dx and dy and finite strips of
gral with respect to x and y over width dx , we must keep in mind that the integrations take limits as these dimensions approach
a rectangle zero.
Analogously, the double iterated integral in equation 13.14 is interpreted as adding over
y R
horizontal strips, as shown in Figure 13.5. Inner limits indicate where each strip starts and stops,
d and outer limits indicate the positions of first and last strips.
The transition now to more general regions is quite straightforward. For the double integral
dy of f (x, y) over the region in Figure 13.6, we use a double iterated integral with respect to y
(x, y)
and x . The y -integration adds the quantities f (x, y) dy dx over rectangles in a vertical strip.
c We write
" h(x)
a dx b x f (x, y) dy dx,
g(x)
where g(x) and h(x) indicate that each vertical strip starts on the curve y = g(x) and ends
on the curve y = h(x) . The x -integration now adds over all strips, beginning at x = a and
ending at x = b :
"" " b" h(x)
f (x, y) dA = f (x, y) dy dx. (13.15)
FIGURE 13.6 Integration R a g(x)
with respect to y and x adds first A double iterated integral in the reverse order is not convenient for this region because horizontal
inside a vertical strip, and then strips neither all start on the same curve nor all end on the same curve.
over all strips
For the region in Figure 13.7, we obtain
y y = h (x) "" " d" h(y)
f (x, y) dA = f (x, y) dx dy. (13.16)
R c g(y)
dy
R
The limits on double iterated integrals have been interpreted schematically as follows:
dx y = g (x) " position of last " where each and every

horizontal strip horizontal strip stops
f (x, y) dx dy;
position of first where each and every
horizontal strip horizontal strip starts
a b x " position of last " where each and every

vertical strip vertical strip stops
f (x, y) dy dx.
position of first where each and every
vertical strip vertical strip starts
With these interpretations on the limits, you can see how important it is to have a well-labelled
diagram.
FIGURE 13.7 Integration with respect to x and y adds first inside a horizontal strip, and
then over all strips
d
R x = h( y)
dx
dy
x = g( y)
x
EXAMPLE 13.2
Evaluate the double integral of f (x, y) = xy 2 + x 2 over the region bounded by the curves
y = x 2 and x = y 2 .
FIGURE 13.8 Integration
of xy 2 + x 2 over R using vertical SOLUTION If we use vertical strips (Figure 13.8), we have
strips
"" " 1" √ " 1 # $√x
y
x
xy 3
x = y2 (xy 2 + x 2 ) dA = (xy 2 + x 2 ) dy dx = + x2y dx
(1, 1) R 0 x2 0 3 x2
" 1 & ' # 7/2 $1
4 x7 8x x8 x5
y= x2 = x 5/2 − − x 4 dx = − −
0 3 3 21 24 5 0
dy
R
8 1 1 39
= − − = .
21 24 5 280
dx
x
There are two distinct parts to every double integral: first, the function f (x, y) being integrated,
which is the integrand; second, the region R over which integration is being performed, and
this region determines the limits on the corresponding double iterated integral. Note that we do
not use f (x, y) to determine limits on the double iterated integral; the region determines the
limits. Conversely, if we are given a double iterated integral, then we know that it represents
the double integral of its integrand over some region, and the region is completely defined by
the limits on the iterated integral. This point is emphasized in the following example.
EXAMPLE 13.3
" 2" 2
2
Evaluate the double iterated integral ex dx dy .
0 y
2
SOLUTION The function ex does not have an elementary antiderivative with respect to x ,
and it is therefore impossible to evaluate the double iterated integral as it now stands. But the
2
double iterated integral represents the double integral of ex over some region R in the xy -
plane. To find R we note that the inner integral indicates horizontal strips that all start on the
line x = y and stop on the line x = 2 (Figure 13.9a). The outer limits state that the first and
last strips are at y = 0 and y = 2, respectively. This defines R as the triangle bounded by
FIGURE 13.9a Limits can be used to FIGURE 13.9b Reversing order of

determine the region of integration integration leads to simpler integral
y y
(2, 2)
y=x x=2
y=x
x=2
dy
x dx x
the straight lines y = x , x = 2, and y = 0 (Figure 13.9b). If we now reverse the order of
integration and use vertical strips, we have
" 2" 2 "" " 2" x " 2
2 2 2 2
ex dx dy = ex dA = ex dy dx = {yex }x0 dx
0 y R 0 0 0
" 2 # $2
2 1 2 e4 − 1
= xex dx = ex = .
0 2 0 2
This example points out that an iterated integral in one order may be much easier to evaluate
than the corresponding iterated integral in the opposite order.
EXERCISES 13.2
In Exercises 1–12 evaluate the double integral over the region. In Exercises 13–18 evaluate the double iterated integral by reversing
the order of integration.
""
1. (x 2 + y 2 ) dA , where R is bounded by y = x 2 , x = y 2 " 2" √4−x 2
R
∗ 13. (4 − y 2 )3/2 dy dx
"" 0 0
√ " 1"
2. (4 − x 2 − y) dA , where R is bounded by x = 4 − y, 1
R ∗ 14. sin (x 2 ) dx dy
x = 0, y = 0 0 y
"" " 0" 2
3. (x + y) dA , where R is bounded by x = y 3 + 2, ∗ 15. y(x 2 + y 2 )8 dx dy

R −2 −y
x = 1, y = 1 " 0" x
x
"" ∗ 16. % dy dx
−2 −2 x + y2
2
4. xy 2 dA , where R is bounded by x + y + 1 = 0, x + y 2 = 1
R " 2" x 2 /2
x
"" ∗ 17. % dy dx
0 0 1 + x2 + y2
5. xey dA , where R is bounded by y = x , y = 0, x = 1
R " 2" 0
x
"" ∗ 18. % dy dx
0 −x 2 /2 1 + x2 + y2
6. (x + y) dA , where R is bounded by x 2 + y 2 = 9
R
"" ∗ 19. Verify that if m ≤ f (x, y) ≤ M for all (x, y) in R , then
√
7. x 2 y dA , where R is bounded by y = x + 4, y = 0, ""
R
x+y =2 m(area of R) ≤ f (x, y) dA ≤ M(area of R).
R
""
√
8. (xy + y 2 − 3x 2 ) dA , where R is bounded by y = |x|, ∗ 20. Evaluate the double integral of f (x, y) = 1/ 2x − x 2 over the
R region in the first quadrant bounded by y 2 = 4 − 2x .
y = 1, y = 2
""
9. (1 − x)2 dA , where R is bounded by x + y = 1, x + y = In Exercises 21–28 either the integral has value 0 or it can be evaluated
R by doubling the double integral over half the region. By drawing the
−1, x − y = 1, y − x = 1 region and examining the integrand, determine which situation prevails.
"" Do not evaluate the integral.
10. (x + y) dA , where R is bounded by x = y 2 , x 2 −y 2 = 12 ""
R %
21. x 2 y 3 dA , where R is bounded by x = 4 − y2 , x = 0
"" R
11. x dA , where R is bounded by y = 3x , y = x , x + y = 4 "" %
R 22. x 2 y 2 dA , where R is bounded by x = 4 − y2 , x = 0
"" ""
R
12. y 2 dA , where R is bounded by x = 0, y = 1, y = 1/2, 23. (x + y) dA , where R is the square with vertices (±3, 0)
%
R
R
x = 1/ y 4 + 12y 2 and (0, ±3)
""
24. x 7 cos (x 2 ) dA , where R is bounded by y = 4 − |x|, In Exercises 35–41 evaluate the double integral over the region.
R
""
y = x2
"" ∗ 35. x 2 dA , where R is bounded by x 2 + y 2 = 4
x 2 +y 2 2 2 R
25. e dA , where R is bounded by y = 4−4x , y = x −1 ""
R
"" ∗ 36. (6 − x − 2y) dA , where R is bounded by x 2 + y 2 = 4
2 3
26. cos (x y) dA , where R is bounded by y = 0, y = x − x R
R
""
"" ∗ 37. 6x 5 dA , where R is the region under x + 5y = 16, above
27. sin (x 2 y) dA , where R is bounded by y = 0, y = x 3 − x R
R y = x − 4, and bounded by x = (y − 2)2
"" ""
√ √
28. (x 2 y 3 + xy 2 ) dA , where R is bounded by |x| + |y| = 1 ∗ 38. yex dA , where R is bounded by y = x , x + y = 2, y = 0
R R
"" %
The average value of a function f (x, y) over a region R with area A
∗ 39. 1 + y dA , where R is bounded by x = −1, y = 2, x =
R
is defined as "" y 3
1
f = f (x, y) dA. "" %
A R
∗ 40. y x 2 + y 2 dA , where R is bounded by y = x , x =
In Exercises 29–32 find the average value of the function over the R
−1, y = 0
region.
""
29. f (x, y) √
= xy over the region in the first quadrant bounded by ∗ 41. (x 2 + y 2 ) dA , where R is bounded by x 2 + y 2 = 9
x = 0, y = 1 − x 2 , y = 0 R
" 1" 1
30. f√(x, y) = x + y over the region bounded by y = x , y = 0,
∗ 42. Evaluate the double iterated integral |x − y| dy dx .
y = 2−x 0 0
∗ 31. f (x, y) = x over the region between y = sin x and y = 0 for ""
0 ≤ x ≤ 2π ∗ 43. Evaluate the double integral |y − 2x 2 + 1| dA , where R is
R
∗ 32. f (x, y) = ex+y over the region bounded by y = x + 1, y = the square bounded by x = ±1, y = ±1.
x − 1, y = 1 − x , y = −1 − x
∗ 44. For the accelerating slit system of a mass spectrometer, the number
∗ 33. The Cobb–Douglas production function for a widget is P (x, y) = of ions within unit solid angle of the electron beam, at the plane of the
10 000x 0.3 y 0.7 , where P is the number of widgets produced each first slit, is
month, x is the number of employees, and y is the monthly operating
budget in thousands of dollars. If the company uses anywhere between
" 2d " ∞
2nc L x 2 (1 − a 2 y 2 /c2 )
45 and 55 workers each month, and its operating budget varies from n= dx dy,
π 0 0 (1 + x 2 )(x 2 + a 2 y 2 /c2 )
$8000 to $12 000 per month, what is the average number of widgets
produced each month?
where nc is the number of ions with initial velocity c , L is the length
∗ 34. Repeat Exercise 33 using the production function P (x, y) = of the slit, d is the width of the slit, and a > 0 is a constant. Show that
10 000x 0.7 y 0.3 . n = 2nc dL(1 − ad/c) .
13.3 Areas and Volumes of Solids of Revolution

FIGURE 13.10 Double Because equation 13.7 represents the area of a region R as a double integral, and double integrals
iterated integrals can be used to are evaluated by means of double iterated integrals, it follows that areas can be calculated using
find areas of regions
double iterated integrals. In particular, to find the area of the region in Figure 13.10, we subdivide
y R into rectangles of dimensions dx and dy and therefore of area dA = dy dx . Areas of these
rectangles are then added in the y -direction to give the area of a vertical strip
R
y = h(x) " h(x)
dy dy dx,
g(x)
y = g(x)
dx
a b x
where limits indicate that every vertical strip starts on the curve y = g(x) and ends on the
curve y = h(x) . Finally, areas of the vertical strips are added together to give the total area:
" b" h(x)
area = dy dx,
a g(x)
where a and b indicate the x -positions of first and last strips.
EXAMPLE 13.4
√
Find the area of the region bounded by the curves xy = 2, x = 2 y , y = 4.
SOLUTION If we choose horizontal strips for this region (Figure 13.11), we have
" 4" √
2 y " 4 & '
√ 2
area = dx dy = 2 y− dy
1 2/y 1 y
# $4
4 3/2 28
= y − 2 ln |y| = − 2 ln 4.
3 1 3
For vertical strips (Figure 13.12), we require two iterated integrals because, to the left of the line
x = 2, strips begin
√ on the hyperbola xy = 2, whereas to the right of x = 2, they begin on the
parabola x = 2 y . We obtain
" 2 " 4 " 4" 4 " 2 & ' " 4 & '
2 x2
area = dy dx + dy dx = 4− dx + 4− dx
1/2 2/x 2 x 2 /4 1/2 x 2 4
# $4
x3 28
= {4x − 2 ln |x|}21/2 + 4x − = − 2 ln 4.
12 2 3
FIGURE 13.11 When horizon- FIGURE 13.12 When

tal strips are chosen, only one double vertical strips are chosen, two dou-
iterated integral is needed ble iterated integrals are needed
y y
1 y=4 1 y=4
,4 ,4
2 2
(4, 4) (4, 4)
dx dy dy
dy
xy = 2 x = 2√y
dx dx
xy = 2 x = 2√y
(2, 1) (2, 1)
x x
If we compare finding areas by definite integrals (Section 7.1) and finding the same areas by
double integrals, it is clear that no great advantage is derived by using double integrals. In fact,
it is probably more work because we must perform two, rather than one, integrations, although
the first integration is trivial. The advantage of double integrals is therefore not in finding area;
it is in finding volumes of solids of revolution, centres of mass, moments of inertia, and fluid
forces, among other applications.
Volumes of Solids of Revolution

If the region in Figure 13.13 is rotated around the x -axis, the volume of the resulting solid of
FIGURE 13.13 With def-
inite integrals, volumes of washers
revolution can be evaluated by using the washer method introduced in Section 7.2:
or cylindrical shells are calculated " b
y volume = {π [h(x)]2 − π [g(x)]2 } dx. (13.17)
a
y = h(x)
If this region is rotated around the y -axis, the volume generated is calculated by using the
cylindrical shell method:
" b
volume = 2π x [h(x) − g(x)] dx. (13.18)
a
y = g(x)
Thus, once we have chosen to use vertical rectangles, the axis of revolution determines whether
a b x we use washers or cylindrical shells. We now show that with double integrals one method works
for all problems.
To rotate this region around the x -axis we subdivide R into small areas dA (Figure 13.14).
FIGURE 13.14 With
double integrals, volumes of rings
If the area dA at a point (x, y) is rotated about the x -axis, it generates a “ring” with cross-
are calculated sectional area dA . Since (x, y) travels a distance 2πy in traversing the ring, it follows that
the volume in the ring is approximately 2πy dA . To find the total volume obtained by rotating
y
R about the x -axis, we add the volumes of all such rings and take the limit as the areas shrink
to points. But this is what we mean by the double integral of 2πy over the region R , and we
R
therefore write ""
dA volume = 2πy dA. (13.19)
(x, y)
R
a On the other hand, if dA is rotated about the y -axis, it again forms a ring, but with
approximate volume 2π x dA . The total volume, then, when R is rotated about the y -axis is
b x ""
volume = 2π x dA. (13.20)
R
Since double iterated integrals are used to evaluate double integrals, it follows that we can
FIGURE 13.15 With
double iterated integrals, volumes
set up double iterated integrals to find volumes represented by equations 13.19 and 13.20. The
of rectangular rings are calculated decision to use a double iterated integral with respect to y and x implies a subdivision of R
into rectangles of dimensions dx and dy (Figure 13.15). The volume of the ring formed when
y this rectangle is rotated around the x -axis is 2πy dy dx . If we choose to integrate first with
y = h(x) respect to y , we are adding over all rectangles in a vertical strip
dy " h(x)
(x, y) 2πy dy dx,
g(x)
where limits indicate that all vertical strips start on the curve y = g(x) and end on the curve
dx
a y = h(x) . This integral is the volume generated by rotating the vertical strip around the x -axis.
y = g(x) Integration now with respect to x adds over all strips to give the required volume:
b x
" b" h(x)
volume = 2πy dy dx. (13.21)
a g(x)
Note that when we actually do perform the inner integration, we get

" b " b
volume = {πy 2 }h(x)
g(x) dx = {π [h(x)]2 − π [g(x)]2 } dx,
a a
and this is the result contained in equation 13.17.

When R is rotated around the y -axis, the rectangular area dy dx generates a ring of volume
2π x dy dx . Addition over the rectangles in a vertical strip
" h(x)
2π x dy dx
g(x)
gives the volume generated by rotating the strip about the y -axis. Finally, integration with
respect to x adds over all strips to give
" b" h(x)
volume = 2π x dy dx. (13.22)
a g(x)
This time the inner integration leads to

" b " b
volume = {2π xy}h(x)
g(x) dx = 2π x [h(x) − g(x)] dx,
a a
the same result as in equation 13.18.

The advantage, then, in using double integrals to find volumes of solids of revolution is
that it requires only one idea, that of rings. The first integration leads to washers or cylindrical
shells, but we need never think about this.
EXAMPLE 13.5
Find volumes of the solids of revolution if the region bounded by the curves y = 2x − x 2 ,
y = x 2 − 2x is rotated around:
(a) the y -axis (b) x = −3 (c) y = 2 (d) y = x + 2.
SOLUTION
(a) If we use vertical strips (Figure 13.16), then
" 2" 2x−x 2 " 2

2
volume = 2π x dy dx = 2π {xy}2xx−x
2 −2x dx
0 x 2 −2 x 0
" 2
= 2π x{(2x − x 2 ) − (x 2 − 2x)} dx
0
" 2 # $2
2 3 2x 3 x4 16π
= 4π (2x − x ) dx = 4π − = .
0 3 4 0 3
FIGURE 13.16 Area is subdivided into rectangles, vol-

umes of rings calculated, and added over vertical strips
y y=x+2
2 y=2
x = −3 y = 2x − x 2
dx
dy
−3 2 x
y = x 2 − 2x
(b) In this case the radius of the ring formed by rotating the rectangle about x = −3 is
x + 3, and therefore
" 2" 2x−x 2 " 2
2
volume = 2π(x + 3) dy dx = 2π {y(x + 3)}2xx−x
2 −2x dx
0 x 2 −2 x 0
" 2 " 2
= 2π (x + 3)(4x − 2x 2 ) dx = 4π (6x − x 2 − x 3 ) dx
0 0
# $2
2 x3 x4 64π
= 4π 3x − − = .
3 4 0 3
(c) When the rectangle is rotated around y = 2, the radius of the ring is 2 − y , and
hence
" 2" 2x−x 2 " 2 # $2x−x 2
1
volume = 2π(2 − y) dy dx = 2π − (2 − y)2 dx
0 x 2 −2 x 0 2 x 2 −2 x
" 2 " 2
=π [(2 − x 2 + 2x)2 − (2 − 2x + x 2 )2 ] dx = 8π (2x − x 2 ) dx
0 0
# 3
$2
x 32π
= 8π x 2 − = .
3 0 3
(d) Using formula 1.16, the distance

√from the area element
√ dy dx at position (x, y) to the
line y = x + 2 is |x − y + 2|/ 2 = (x − y + 2)/ 2. 0 Hence the volume√of1 the ring
obtained by rotating the rectangle around the line is 2π(x − y + 2)/ 2 dydx ,
and the volume of the solid of revolution is
" 2" 2x−x 2 " 2 # $2x−x 2
√ √ 1 2
volume = 2 π(x − y + 2) dy dx = 2π − (x − y + 2) dx
0 x 2 −2 x 0 2 x 2 −2 x
" 2
π
= −√ [(x − 2x + x 2 + 2)2 − (x − x 2 + 2x + 2)2 ] dx
2 0
" 2 √
π π 2 32
= √ (16x − 4x 3 ) dx = √ 8x 2 − x 4 0 = 8 2 π.
2 0 2
Notice that double integrals allow us to calculate volumes of solids of revolution about any line
(part (d) of Example 13.5). With definite integrals we were restricted to vertical and horizontal
lines.
EXERCISES 13.3
In Exercises 1–10 use a double integral to find the area of the region 5. y = 4/x 2 , y = 5 − x 2 6. y = xe−x , y = x, x = 2
bounded by the curves.
7. x = 4y − 4y 2 , y = x − 3, y = 1, y = 0
8. x = y(y − 2), x + y = 12
1. y = 4x 2 , x = 4y 2 2. y = x 2 , y = 5x + 6
9. y = x 3 − x 2 − 2x + 2, y = 2
3. x = y 2 , x = 3y − 2 4. y = x 3 + 8, y = 4x + 8 10. x + y = 1, x + y = 5, y = 2x + 1, y = 2x + 6
In Exercises 11–20 use a double integral to find the volume of the solid ∗ 28. y 2 = x 4 (9 + x)
of revolution obtained by rotating the region bounded by the curves
∗ 29. y = (x 2 + 1)/(x + 1), x + 3y = 7
around the line.
√ ∗ 30. (x + 2)2 y = 4 − x, x = 0, y = 0 ( x ≥ 0, y ≥ 0)
11. y = − 4 − x, x = 0, y = 0 about y = 0
12. 4x 2 + 9y 2 = 36 about y = 0
In Exercises 31–35 use a double integral to find the volume of the solid
13. y = (x − 1)2 , y = 1 about x = 0 of revolution obtained by rotating the region bounded by the curves
around the line.
14. y = x 2 + 4, y = 2x 2 about y = 0
∗ 31. y = 4/(x 2 + 1)2 , y = 1 about x = 0
15. x − 1 = y 2 , x = 5 about x = 1
√ ∗ 32. y = x 2 − 2, y = 0 about y = −1
16. x = y 3 , y = 2 − x, y = 0 about y = 1
∗ 33. y = |x 2 − 1|, x = −2, x = 2, y = −1 about y = −2
17. y = 4x 2 − 4x, y = x 3 about y = −2 %
∗ 34. x = 4 + 12y 2 , x − 20y = 24, y = 0 about y = 0
18. x = 3y − y 2 , x = y 2 − 3y about y = 4
∗ 35. y = (x + 1)1/4 , y = −(x + 1)2 , x = 0 about x = 0
19. x = 2y − y 2 − 2, x = −5 about x = 1
√
20. x + y = 4, y = 2 x − 1, y = 0 about y = −1 ∗ 36. Find the area of the region common to the two circles x 2 + y 2 = 4
and x 2 + y 2 = 6x .
In Exercises 21–30 use a double integral to find the area of the region
bounded by the curves. In Exercises 37–40 find the volume of the solid of revolution obtained
by rotating the region bounded by the curves around the line.
∗ 21. y = 2x 3 , y = 4x + 8, y = 0
√ ∗ 37. x = 1, y = 1, x = 0, y = 0 about x + y = 2
∗ 22. y = x/ x + 3, x = 1, x = 6, y = −x 2
√ √ ∗ 38. y = x 2 , y = 2x + 3 about y = 2x + 3
∗ 23. y = x − 2, y = 4 − −x, y = 4 − x, y = 0 (−16 ≤ x ≤ √
3) ∗ 39. y = x, y = 0, x = 1 about y = 3x + 2
√ ∗ 40. x = 2y, y = x − 1, y = 0 about x + y + 1 = 0
∗ 24. y = x 3 − x, x + y + 1 = 0, x = y + 1
∗ 25. y 2 = x 2 (4 − x 2 ) ∗∗ 41. Prove that the area above the line y = h and under the circle
2 2 2 2
∗ 26. x + y = 4, x + y = 4x (interior to both) x 2 + y 2 = r 2 ( r > h ) is given by
% %
∗ 27. x = 1/ 4 − y 2 , 4x + y 2 = 0, y + 1 = 0, y − 1 = 0 A = π r 2 /2 − h r 2 − h2 − r 2 Sin −1 (h/r).
13.4 Fluid Pressure

In Section 7.6 we defined pressure at a point in a fluid as the magnitude of the force per unit
area that would act on any surface placed at that point. We discovered that at a depth d > 0
below the surface of a fluid, pressure is given by
P = 9.81ρd, (13.23)
where ρ is the density of the fluid. With these ideas and the definite integral, we were able to
calculate fluid forces on flat surfaces in the fluid. In particular, the magnitude of the total force
on each side of the vertical surface in Figure 13.17 is given by the definite integral
" b
force = −9.81ρy [h(y) − g(y)] dy. (13.24)
a
Although horizontal rectangles are convenient for this problem, it is clear that they are not
reasonable for the surface in Figure 13.18.
FIGURE 13.17 With FIGURE 13.18 Use of FIGURE 13.19 Double

definite integrals, use horizontal horizontal rectangles to calculate integrals to calculate forces on
rectangles to calculate fluid forces force on this plate is inconvenient submerged plates are most efficient
y y y
Fluid surface a Fluid surface b Fluid surface
x x x
b
x = g(y) x = h(y) y = h(x)
dA
dy
R
a
y = g(x)
Double integrals, on the other hand, can be applied with equal ease to both surfaces. To see
this we consider, first, force on each side of the surface in Figure 13.19. If the surface is divided
into small areas dA , then the force on dA is P dA , where P is pressure at that depth. Total
force on R is the sum of the forces on all such areas in R as areas dA shrink to a point. But
once again this is the double integral, and we therefore write
""
force = P dA. (13.25)
R
To set up a double iterated integral in order to evaluate this double integral, we use our interpre-
tation of the double iterated integral as a limit of a sum in which the areas dA have been chosen
as rectangles. In particular, for the surface in Figure 13.17, we draw rectangles of dimensions
FIGURE 13.20 To find dx and dy , as shown in Figure 13.20. The force on this rectangle is its area dx dy multiplied
the force on this plate, use hori- by pressure −9.81ρy at that depth, −9.81ρy dx dy . Addition of these quantities over all
zontal strips rectangles in a horizontal strip gives the force on the strip,
y " h(y)
Fluid surface −9.81ρy dx dy,
x = g( y) x g(y)
b
where the limits indicate that all horizontal strips start on the curve x = g(y) and end on the
dx x = h( y) curve x = h(y) . Integration with respect to y now adds over all horizontal strips to give the
dy
total force on the surface:
(x, y) " b" h(y)
a
force = −9.81ρy dx dy. (13.26)
a g(y)
When we perform the inner integration, we obtain

" b " b
To find h(y)
FIGURE 13.21 force = {−9.81ρyx}g(y) dy = −9.81ρy [h(y) − g(y)] dy,
the force on this plate, use vertical a a
strips
and this is the result contained in equation 13.24.
y For the surface in Figure 13.18 we again draw rectangles of dimensions dx and dy and
a Fluid surface b
calculate the force on such a rectangle, −9.81ρy dy dx . In this case it is more convenient to
dx x
add over rectangles in a vertical strip to give the force on the strip (Figure 13.21):
" h(x)
y = h(x) −9.81ρy dy dx.
dy g(x)
(x, y)
Force on the entire surface can now be found by adding over all vertical strips:
" b" h(x)
y = g(x)
force = −9.81ρy dy dx. (13.27)
a g(x)
EXAMPLE 13.6
The face of a dam is parabolic with breadth 100 m and height 50 m. Find the magnitude of the
total force due to fluid pressure on the face when the water is 1 m from the top.
FIGURE 13.22 Force of SOLUTION If we use the coordinate system in Figure 13.22, then the edge of the dam has an
1
water on vertical face of a dam equation of the form y = kx 2 . Since (50, 50) is a point on this curve, it follows that k = 50 .
y Because force on the left half of the dam is the same as that on the right half, we can integrate
for the right half and double the result; that is,
dx (50, 50) " √
35 2 " 49
force = 2 9.81(1000)(49 − y) dy dx
x2
(x, y) dy y= 0 x 2 /50
50 √ # √
" 35 2 $49 " 35 2 & '
y2 2401 49x 2 x4
x = 19 620 49y − dx = 19 620 − + dx
0 2 x 2 /50 0 2 50 5000
# $35√2
2401x 49x 3 x5
= 19 620 − + = 6.22 × 108 N.
2 150 25 000 0
EXAMPLE 13.7
1
A tank in the form of a right-circular cylinder of radius m and length 10 m has its axis
2
horizontal. If it is full of water, find the force due to water pressure on each end of the tank.
FIGURE 13.23 Force on SOLUTION Since the force on that part of the end to the left of the y -axis (Figure 13.23) is
the end of a full cylindrical tank identical to the force on that part to the right, we double the force on the right half; that is,
y " 1/2 " √1/4−y 2
1
2
force = 2 9.81(1000)( 21 − y) dx dy
−1/2 0
dx " 1/2 " √1/4−y 2 " 1/2 " √1/4−y 2

dy = 9810 dx dy − 19 620 y dx dy.
(x, y) −1/2 0 −1/2 0
1 x
2 The first double iterated integral represents the area of one-half the end of the tank. Consequently,
4 & '2 5 " 6
1/2
1 1 1 1
x2 + y2 = force = 9810 π − 19 620 y − y 2 dy
4 2 2 4
−1/2
7 & '3/2 81/2
4905π 1 1 4905π
= − 19 620 − − y2 = N.
4 3 4 4
−1/2
FIGURE 13.24 Centre of
pressure on a submerged plate is a
point for equivalency of moments
of forces
y The centroid of a planar region is a point at which the area of the region can be concentrated as
far as first moments of the region are concerned. It is advantageous to define a point called the
x centre of pressure for a surface submerged in a fluid; it is a point where a single force equal to
that of the fluid on the surface has the same first moment about any line as does the fluid force
R dA on the surface. For example, the fluid force on the region R in Figure 13.24 is given by the
(x, y) double integral ""
F = P dA, (13.28)
R
where P = −ρgy in the coordinate system shown. It is the sum of fluid forces P dA on
elemental areas dA at points (x, y) .
Each elemental force P dA creates a moment yP dA about the x -axis, and the total first
moment of all such elemental moments is
""
yP dA. (13.29a)
R
Similarly, the first moment of the fluid force on the surface about the y -axis is
""
xP dA. (13.29b)
R
The centre of pressure of R is the point (xc , yc ) defined by

"" ""
F xc = xP dA, F yc = yP dA, (13.30)
R R
where F is given by 13.28.

For example, due to the symmetry in Example 13.6, the centre of pressure of the dam is on
the y -axis. Its y -coordinate is given by
""
1
yc = yP dA,
F R
where R represents the dam, and F = 6.22 × 108 N is the total force on the dam. We integrate
over the right half and double the result:
" √ √ # $49
35 2 " 49 " 35 2
2 19 620 49y 2 y3
yc = y(9810)(49 − y) dy dx = − dx
F 0 x 2 /50 F 0 2 3 x 2 /50
" √ & '
35 2
19 620 117 649 49x 4 x6
= − + dx
F 0 6 5000 375 000
# $35√2
19 620 117 649x 49x 5 x7
= − + = 21.0 m.
F 6 25 000 7(375 000) 0
EXAMPLE 13.8
Find the depth h of ocean water in the Application Preview at which the tide gate opens (Fig-
Application Preview ure 13.25a).
Revisited
FIGURE 13.25a Tide gate is on the verge FIGURE 13.25b Calcula-
of opening when the sum of the moments of the tion of moment of marsh water on
marsh and the ocean on the gate is zero gate about hinge
y
Water level of marsh
Marsh 1.8
1.8 A Ocean
h
Hinge
0.9 0.9
dx dy
B (x, y) Gate
x
SOLUTION The forces of the marsh and ocean on the gate create moments about the hinge. If
we take counterclockwise moments about the hinge (marsh) as positive, and clockwise moments
(ocean) as negative, the gate will be on the verge of opening when the sum of these moments
is zero. Since the force of the marsh on a rectangle of dimensions dx and dy on the gate
(Figure 13.25b) is 1000g(1.8 − y) dx dy , the moment of this force about the hinge is (0.9 −
y)1000g(1.8 − y) dx dy . The total moment about the hinge of the force of the marsh on the
gate is
" 0.9" L " 0 .9 & '
81 27y 2
Mm = 1000g(0.9 − y)(1.8 − y) dx dy = 1000gL − +y dy
0 0 0 50 10
# $0.9
81y 27y 2 y3 1215gL
= 1000gL − + = .
50 20 3 0 2
Similarly, the moment about the hinge of the force of the ocean water on the gate, when depth
of the ocean water is h , is
" 0.9" L
Mo = − 1030g(0.9 − y)(h − y) dx dy
0 0
" 0 .9 9 & ' :
9h 9 2
= −1030gL − h+ y+y dy
0 10 10
# & ' $0.9
9hy 9 y2 y3
= −1030gL − h+ +
10 10 2 3 0
& 9 ' :
81h 81 9 243
= −1030gL − h+ + .
100 200 10 1000
The sum of these moments is zero when
9 & ' :
1215gL 81h 81 9 243
0 = − 1030gL − h+ + *⇒ h = 1.756.
2 100 200 10 1000
The gate is on the verge of opening when the ocean water is 1.756 m deep.
EXERCISES 13.4
In Exercises 1–8 the surface is submerged vertically in a fluid with 7. A triangle of side lengths 3, 3, and 4, with the longest side vertical,
density ρ . Find the force due to fluid pressure on one side of the and the uppermost vertex 1 unit below the surface
surface.
8. A semicircle of radius 5 with the (diameter) base in the surface
1. An equilateral triangle of side length 2 with one edge in the surface
∗ 9. The vertical end of a water trough is an isosceles triangle with width
2. A parabolic segment of base 12 and height 4 with the base in the 2 m and depth 1 m. Find the force of the water on each end when the
surface trough is one-half filled (by volume) with water.
3. A square of side length 3 with one diagonal vertical and the upper- ∗ 10. A dam across a river has the shape of a parabola 36 m across the
most vertex in the surface top and 9 m deep at the centre. Find the maximum force due to water
pressure on the dam.
4. A triangle of side lengths 5, 5, and 8, with the longest side uppermost,
horizontal, and 3 units below the surface
In Exercises 11–15 the surface is submerged vertically in a fluid with
5. A triangle of side lengths 5, 5, and 8, with the longest side below density ρ . Find the force due to fluid pressure on one side of the surface.
the opposite vertex, horizontal, and 6 units below the surface
∗ 11. A circle of radius 2 with centre 3 units below the surface
6. A trapezoid with vertical parallel sides of lengths 6 and 8, and a third ∗ 12. A rectangle of side lengths 2 and 5, with one diagonal vertical and
side perpendicular to the parallel sides, of length 5, and in the surface the uppermost vertex in the surface
∗ 13. An ellipse with major and minor axes of lengths 8 and 6, and with ∗ 19. A semicircle with radius r when the diameter is horizontal, above
the major axis horizontal and 5 units below the surface the rest of the semicircle, and h units below the surface
∗ 14. A parallelogram of side lengths 4 and 5, with one of the longer
sides horizontal and in the surface, and two sides making an angle of ∗ 20. An equilateral triangle of side length L when one edge is horizontal
π/6 radians with the surface and in the surface
∗ 15. A triangle of side lengths 2, 3, and 4 with the longest side vertical, ∗ 21. A square with sides of length L when one diagonal is vertical and
the side of length 2 above the side of length 3, and the uppermost vertex the uppermost vertex is in the surface
1 unit below the surface
∗ 16. An oil can is in the form of a right-circular cylinder of radius r and ∗ 22. The nonhypotenuse sides of a right-angled triangle have lengths
height h . If the axis of the can is horizontal, and the can is full of oil L and & , where & < L . The vertex containing the right angle is at
with density ρ , find the force due to fluid pressure on each end. the origin in the surface of the fluid and the shortest side is along the
positive x -axis. Find the centre of pressure of the triangle.
∗ 17. Find the force due to water pressure on each side of the flat vertical
plate in the figure below. ∗ 23. The centroid of a plane figure is a fixed point. The centre of
pressure, on the other hand, changes depending on depth below the
Fluid surface
surface. Use the result of Exercise 18 to verify this.
Semicircle 1m
∗ 24. A square plate of side length 2 m has one side on the bottom of
a swimming pool 3 m deep. The plate is inclined at an angle of π/4
radians
√ with the bottom of the pool so that its horizontal upper edge is
4m 3 − 2 m below the surface. Find the force due to water pressure on
3m each side of the plate.
∗∗ 25. A thin triangular piece of wood with sides of lengths 2 m, 2 m,

and 3 m floats in a pond. A piece of rope is tied to the vertex op-
posite the longest side. A rock is then attached to the other end of
Semicircle
the rope and lowered into the water. When the rock sits on the bot-
tom (and the rope is taut), the longest side of the wood still floats in
the surface of the water, but the opposite vertex is 1 m below the sur-
In Exercises 18–21 find the centre of pressure of the surface submerged face. Find the force due to water pressure on each side of the piece of
vertically in a fluid. wood.
∗ 18. A circle with radius r when its centre is h > r units below the ∗∗ 26. Show that the centre of pressure of a plane surface is always below
surface its centroid.
13.5 Centres of Mass and Moments of Inertia

We now show how double integrals can be used to replace definite integrals in calculating first
moments, centres of mass, and moments of inertia of thin plates. Consider a thin plate with
mass per unit area ρ such as that in Figure 13.26. Note that, unlike our discussion in Section
FIGURE 13.26 Double
7.7 where we assumed ρ constant, we have made no such assumption here. In other words,
integrals are very efficient in cal-
culating moments and centres of density could be a function of position, ρ = ρ(x, y) .
mass The centre of mass (x, y) of the plate is a point at which a particle of mass M (equal to
the total mass of the plate) has the same first moments about the x - and y -axes as the plate
y
itself. If we divide the plate into small areas dA , then the mass in dA is ρ dA . Addition over
all such areas in R as each dA shrinks to a point gives the mass of the plate
dA
""
R M = ρ dA. (13.31)
(x, y) R
Since the first moment of the mass in dA about the y -axis is xρ dA , it follows that the first
moment of the entire plate about the y -axis is
x
""
xρ dA.
R
But this must be equal to the first moment of the particle of mass M at (x, y) about the y -axis,
and hence ""
Mx = xρ dA. (13.32)
R
This equation can be solved for x once the double integral on the right and M have been
calculated.
Similarly, y is determined by the equation
""
My = yρ dA, (13.33)
R
where the double integral on the right is the first moment of the plate about the x -axis.
FIGURE 13.27 Vertical In any given problem, the double integrals in 13.31–13.33 are evaluated by means of dou-
strips are chosen for this plate ble iterated integrals. For example, if we divide the plate in Figure 13.27 into rectangles of
dimensions dx and dy and use vertical strips, then we have
y
y = h(x) " b" h(x)

M = ρ dy dx, (13.34a)
a g(x)
dy " b"
(x, y) h(x)
Mx = xρ dy dx, (13.34b)
a g(x)
dx " b" h(x)
a y = g(x) b x My = yρ dy dx. (13.34c)
a g(x)
Once again we point out that equations 13.34 should not be memorized as formulas. Indeed,
each can be derived as needed. For instance, to obtain equation 13.34c, we reason that the first
moment of the mass in a rectangle of dimensions dx and dy at position (x, y) about the x -axis
is yρ dy dx . Addition over the rectangles in a vertical strip gives the first moment of the strip
about the x -axis,
" h(x)
yρ dy dx,
g(x)
and integration with respect to x now adds over all strips to give the first moment of the entire
plate about the x -axis. Note that if ρ is constant and inner integrations are performed in each
of equations 13.34, then
" b " b
M = {ρy}h(x)
g(x) dx = ρ [h(x) − g(x)] dx,
a a
" b " b
Mx = {ρxy}h(x)
g(x) dx = ρx [h(x) − g(x)] dx,
a a
" b # 2 $h(x) " b
y ρ
My = ρ dx = {[h(x)]2 − [g(x)]2 } dx.
a 2 g(x) a 2
These are equations 7.35–7.37 (with different names for the curves), but the simplicity of the
discussion leading to the double iterated integrals certainly demonstrates its advantage over use
of the definite integral described in Section 7.7.
EXAMPLE 13.9
Find the centre of mass of a thin plate with constant mass per unit area ρ if its edges are defined
by the curves y = 2x − x 2 and y = x 2 − 4.
FIGURE 13.28 Vertical SOLUTION For vertical strips as shown in Figure 13.28,
strips are chosen for this plate
" 2 " 2x−x 2 " 2
y M = ρ dy dx = ρ [(2x − x 2 ) − (x 2 − 4)] dx
y = 2x − x 2 −1 x 2 −4 −1
dx
# $2
2x 3
2 = ρ x2 − + 4x = 9ρ.
x 3 −1
(x, y) dy If the centre of mass of the plate is (x, y) , then

y = x2 − 4
" 2 " 2x−x 2 " 2
Mx = xρ dy dx = ρ x [(2x − x 2 ) − (x 2 − 4)] dx
−1 x 2 −4 −1
# $2
−4 2x 3 x4 2 9ρ
(−1, −3) =ρ − + 2x = .
3 2 −1 2
9ρ 1 1
Thus, x = · = . Since
2 9ρ 2
" 2 " 2x−x 2 " 2 # $2x−x 2
y2
My = yρ dy dx = ρ dx
−1 x 2 −4 −1 2 x 2 −4
" 2
ρ ρ 27ρ
= (−4x 3 + 12x 2 − 16)dx = {−x 4 + 4x 3 − 16x}2−1 = − ,
2 −1 2 2
27ρ 1 3
we find y = − · =− .
2 9ρ 2
EXAMPLE 13.10
Find the first moment of area about the line y = −2 for the region bounded by the curves
x = |y|3 and x = 2 − y 2 .
FIGURE 13.29 Horizontal SOLUTION Method 1 The first moment about y = −2 of a rectangle of dimensions
strips are chosen for this area’s dx and dy at position (x, y) is (y + 2) dx dy (Figure 13.29). For the entire plate, then, the
first moments required first moment is
y " " " 1"
0 2−y 2 2−y 2
(1, 1) (y + 2) dx dy + (y + 2) dx dy
−1 −y 3 0 y3
x= y3 " "
0 1
dx 2−y 2 2−y 2
dy = {x(y + 2)}−y 3 dy + {x(y + 2)}y 3 dy
−1 0
" 0 " 1
2 x = (y 4 + y 3 − 2y 2 + 2y + 4) dy + (−y 4 − 3y 3 − 2y 2 + 2y + 4) dy
dx
dy −1 0
x = −y 3
# 5 4 3
$0 # $1
x= 2− y2 y y 2y 2 y5 3y 4 2y 3 2
= + − + y + 4y + − − − + y + 4y
(1, −1) 5 4 3 −1 5 4 3 0
y = −2
17
= .
3
Method 2 By symmetry, the centroid of the region is somewhere along the x -axis. Hence,
the required first moment is 2 A , where A is the area of the region and 2 is the distance from
y = −2 to the centroid. Since the area of the region is equally distributed about the x -axis, we
obtain the required first moment as
" 1" 2−y 2 " 1
2(2)(area of plate above x -axis) = 4 dx dy = 4 (2 − y 2 − y 3 ) dy
0 y3 0
# $1
y3 y4 17
= 4 2y − − = .
3 4 0 3
EXAMPLE 13.11
Find the first moment about the line 2x + y = 1 of a thin plate with constant mass per unit area
ρ if its edges are defined by the curves y = x + 2 and y = x 2 .
SOLUTION We could concentrate the mass M of the plate at its centre of mass (x, y) , and
multiply M by the distance from 2x + y = 1 to (x, y) . Because calculation of (x, y) requires
three integrations, we prefer a more direct approach. The distance from 2x + y = 1 to a
rectangle of dimensions dx and dy at position (x, y) (Figure 13.30) is given by formula 1.16,
|2x + y − 1| |2x + y − 1|
√ = √ .
4+1 5
If we take distances to the right of√the line as positive and those to the left as negative, the
directed distance is (2x + y − 1)/ 5. It now follows that the first moment of the plate around
the line 2x + y = 1 is
" 2" x+2 & ' " 2# $x+2
2x + y − 1 ρ 1 2
√ ρ dy dx = √ (2x + y − 1) dx
−1 x 2 5 5 −1 2 x2
" 2
ρ
= √ [(3x + 1)2 − x 4 − 4x 3 − 2x 2 + 4x − 1] dx
2 5 −1
# $2
ρ 1 3 x5 4 2x 3 2
= √ (3x + 1) − −x − + 2x − x
2 5 9 5 3 −1
√
36 5 ρ
= .
25
FIGURE 13.30 Vertical strips are chosen to find the first moment of this plate
y y = x2
4 y=x+2
y = −2x + 1
(2, 4)
(x, y)
2
(−1, 1)
x
−1 1 2
This example illustrates that double integrals allow us to calculate first moments about arbitrary
lines, unlike definite integrals, which restrict first moments to horizontal and vertical lines.
Calculating moments of inertia (second moments) of thin plates is as easy as calculating

first moments if we use double integrals. In particular, the mass in area dA in Figure 13.26
is ρ dA ; thus its moments of inertia about the x - and y -axes are, respectively, y 2 ρ dA and
x 2 ρ dA . Moments of inertia of the entire plate about the x - and y -axes are therefore given by
the double integrals:
"" ""
2
Ix = y ρ dA and Iy = x 2 ρ dA. (13.35)
R R
The product moment of inertia of this plate with respect to the x - and y -axes is defined as
""
Ixy = xyρ dA. (13.36)
R
Unlike Ix and Iy , which are always positive, Ixy can be positive, negative, or zero.
For a plate such as that shown in Figure 13.31, we evaluate these double integrals by means
of double iterated integrals with respect to x and y :
" b" h(y)

Ix = y 2 ρ dx dy, (13.37a)
a g(y)
" b" h(y)
Iy = x 2 ρ dx dy, (13.37b)
a g(y)
" b" h(y)
Ixy = xyρ dx dy. (13.37c)
a g(y)
FIGURE 13.31 Double integrals are advantageous when calculating moments of inertia of plates
x = h(y)
x = g (y) dx
dy
(x, y)
EXAMPLE 13.12
Find the moments of inertia about the x - and y -axes and the product moment of inertia of a
3
thin plate
√ with constant mass per unit area ρ if its edges are defined by the curves y = x ,
y = 2 − x , and x = 0.
FIGURE 13.32 Vertical SOLUTION With vertical strips as shown in Figure 13.32, the moments of inertia are
strips are chosen to calculate mo-
ments of inertia of this plate " 1" √
2−x " 1 # $√2−x " 1
2 y3 ρ
y y = √2 − x Ix = y ρ dy dx = ρ dx = [(2 − x)3/2 − x 9 ] dx
0 x3 0 3 x3 3 0
# $1 √
dx ρ 2 x 10 16 2 − 5
5/2
(1, 1) = − (2 − x) − = ρ,
3 5 10 0 30
" 1" √ " "
(x, y) dy 2−x 1 2 2 3√2−x 1 √
Iy = x 2 ρ dy dx = ρ x y x 3 dx = ρ (x 2 2 − x − x 5 ) dx.
y = x3 0 x3 0 0
2 x If we set u = 2 − x in the first term,

" 1 # $1 " 2
2
√ x6 √ ρ
Iy = ρ (2 − u) u(−du) − ρ =ρ (4 u − 4u3/2 + u5/2 ) du −
2 6 0 1 6
# $2 √
8u3/2 8u5/2 2u7/2 ρ 256 2 − 319
=ρ − + − = ρ.
3 5 7 1 6 210
" 1" √
2−x " 1 # $√2−x " 1
xy 2 ρ
Ixy = xyρ dy dx = ρ dx = (2x − x 2 − x 7 ) dx
0 x3 0 2 x3 2 0
# $1
ρ 2 x3 x8 13ρ
= x − − = .
2 3 8 0 48
EXAMPLE 13.13
Find the second moments of area about the lines y = −1 and x + y = 1 of the region bounded
by the curves x = y 2 and x = 2y .
SOLUTION The second moment of area about the line y = −1 (Figure 13.33) is
" 2" 2y " 2 " 2
2 2 2y
(y + 1) dx dy = {x(y + 1) }y 2 dy = (y + 1)2 (2y − y 2 ) dy
0 y2 0 0
" 2 # $2
4 2 y5 3 2 28
= (−y + 3y + 2y) dy = − +y +y = .
0 5 0 5
FIGURE 13.33 Horizontal strips are chosen for moments of inertia of this plate
x = y2
(4, 2)
x+y=1 dx
x = 2y
1 dy
(x, y)
1 x
y = −1
√
Since the undirected distance from x + y = 1 to (x, y) is |x + y − 1|/ 2, the second moment
about this line is
" 2" 2y & '2 " 2 # $2y
x+y−1 1 1 3
√ dx dy = (x + y − 1) dy
0 y2 2 2 0 3 y2
" 2
1
= [(3y − 1)3 − y 6 − 3y 5 + 5y 3 − 3y + 1] dy
6 0
# $2
1 1 4 y7 y6 5y 4 3y 2 62
= (3y − 1) − − + − +y = .
6 12 7 2 4 2 0 21
Principal Axes and Principal Moments of Inertia

Consider finding the lines through the origin about which the moments of inertia of the plate in
Figure 13.34 are largest and smallest relative to all other lines through the origin. Using distance
formula 1.16, we can say that the moment of inertia about any line y = mx is
"" & '2 ""
|y − mx| 1
I (m) = √ ρ dA = (y 2 − 2mxy + m2 x 2 )ρ dA
R m2 + 1 m2 + 1 R
1
= (Ix − 2mIxy + m2 Iy ). (13.38)
m2 + 1
FIGURE 13.34 Moments of inertia have maximum and minimum values about principal axes
y
y = mx
dA
x
Critical points of I (m) are given by
dI −2 m 1
0 = = (Ix − 2mIxy + m2 Iy ) + 2 (−2Ixy + 2mIy )
dm (m2 + 1)2 m +1
−2m(Ix − 2mIxy + m2 Iy ) + (m2 + 1)(−2Ixy + 2mIy )
=
(m2 + 1)2
2[(m2 − 1)Ixy − mIx + mIy ]
= .
(m2 + 1)2
Solutions of the quadratic equation Ixy m2 + (Iy − Ix )m − Ixy = 0 are

% ; & '
Ix − I y ± (Iy − Ix )2 + 4(Ixy )2 Ix − Iy Ix − I y 2
m= = ± 1+ . (13.39)
2Ixy 2Ixy 2Ixy
Lines with these slopes are perpendicular, as is easily seen by setting a = (Ix − Iy )/(2Ixy ) ,
√ √
and noting that (a + 1 + a 2 )(a − 1 + a 2 ) = a 2 − (1 + a 2 ) = −1. They are called
the principal axes of the plate at the origin. Substitution of 13.39 into 13.38 leads to messy
algebra. Instead, we note that Ixy m2 + (Iy − Ix )m − Ixy = 0 can be expressed in the form
(m2 + 1)Ixy + m(Iy − Ix ) = 2Ixy . Substituting this into 13.38 gives the moments of inertia
around the principal axes:
1
I = [Ix − m(m2 + 1)Ixy − m2 (Iy − Ix ) + m2 Iy ]
m2 + 1
1
= [(m2 + 1)Ix − m(m2 + 1)Ixy ]
m2 +1
4 % 5
Ix − I y ± (Ix − Iy )2 + 4(Ixy )2
= Ix − mIxy = Ix −
2
;& '2
I x + Iy I x − Iy
= ∓ + (Ixy )2 . (13.40)
2 2
That these yield (absolute) maximum and minimum values for I (m) is shown in Exercise 42.
They are called the principal moments of inertia of the plate about the origin; they are moments
of inertia about the principal axes.
EXAMPLE 13.14
Show that principal moments of inertia about the origin for the uniform rectangular plate in
Figure 13.35 are Ix and Iy .
SOLUTION Because of the symmetry of the plate about the axes, Ixy = 0, in which case
equation 13.39 does not define principal directions. If we return to function 13.38, we find that
1 I x − Iy
I (m) = (Ix + m2 Iy ) = Iy + .
m2 +1 m2 + 1
If Ix > Iy (as is the case in Figure 13.35), then this is an even function with respect to m ,
decreasing from I (0) = Ix to limm→∞ I = Iy ; that is, principal moments of inertia are Ix
and Iy . If Ix < Iy , then this even function increases from I (0) = Ix to limm→∞ I = Iy , and
once again Ix and Iy are principal moments of inertia.
FIGURE 13.35 Principal moments of inertia of a rectangular plate about its centre
y
b
x
−a a
−b
Notice that if the rectangle is a square ( a = b ), then Ix = Iy , and the moment of inertia of the
plate about every line through the origin has the same value. In this case we say that every pair
of perpendicular lines through the origin constitutes a pair of principal axes.
EXERCISES 13.5
In Exercises 1–10 find the centroid of the region bounded by the curves. ∗ 28. Find the moment of inertia of a uniform rectangular plate a units
long and b units wide about a line through the centre of the plate and
1. x = y + 2, x = y 2 perpendicular to the plate.
2 2
2. y = 8 − 2x , y + x = 4
3. y = x 2 − 1, y + (x + 1)2 = 0 In Exercises 29–31 find the second moment of area of the region
bounded by the curves about the line.
4. x + y = 5, xy = 4
5. y = ex , y = 0, x = 0, x = 1 ∗ 29. 4x 2 + 9y 2 = 36 about y = −2
√ ∗ 30. x = y 2 , x + y = 2 about x = −1
6. y = 4 − x 2 , y = x, x = 0
√
7. y = 1/(x − 1), y = 1, y = 2, x = 0 ∗ 31. y = a 2 − x 2 , y = a, x = a (a > 0) about the x -axis
8. x = 4y − 4y 2 , x = y + 3, y = 1, y = 0
In Exercises 32–35 find the first and second moments of area of the
9. y = |x 2 − 1|, y = 2
region bounded by the curves about the line.
10. y = x, y = 2x, 2y = x + 3
∗ 32. x = y 2 − 2, y = x about x + y = 1
In Exercises 11–15 find the second moment of area of the region ∗ 33. x = y 2 , x + y = 2 about y − x = 2
bounded by the curves about the line. ∗ 34. y = x 3 , x = y 2 about 2x + y = 3
2 3
11. y = x , y = x about the y -axis ∗ 35. y = 2 − x 2 , y = |x| about y = x
12. y = x, y = 2x + 4, y = 0 about the x -axis ∗ 36. A triangular plate with constant mass per unit area ρ is bounded
13. y = x 2 , 2y = x 2 + 4 about y = 0 by the coordinate axes and the line hx + by = hb , where h and b are
positive constants. Find its product moments of inertia about (a) the x -
14. y = x 2 − 4, y = 2x − x 2 about x = −2 and y -axes and (b) the axes through the centre of mass parallel to the
% x - and y -axes.
15. x = 1/ y 4 + 12y 2 , x = 0, y = 1/2, y = 1 about y = 0
∗ 37. Show that the absolute value of the product moment of inertia with
respect to the x - and y -axes of a plate with constant mass per unit area
16. Find the first moment about the line y = −2 of a thin plate of is always less than one-half the sum of the moments of inertia about
constant mass per unit area ρ if its edges are defined by the curves the x - and y -axes.
y = 2 − 2x 2 and y = x 2 − 1.
∗ 38. Show that for a plate with constant mass per unit area, the product
moment of inertia with respect to the principal axes through a point is
In Exercises 17–20 find the product moment of inertia with respect to always zero.
the x - and y -axes for the plate defined by the curves if it has constant
mass per unit area ρ . ∗ 39. Suppose that Ix and Iy are moments of inertia of a thin plate with
constant mass per unit area about the x - and y -axes. Let Ix- and Iy- be
17. x 2 + y 2 = r 2 18. y = x 2 , y = x 3 moments of inertia of the plate about any other pair of perpendicular
lines through the origin. Show that Ix- + Iy- = Ix + Iy .
∗ 19. x = −y 2 , x + y + 2 = 0
∗ 20. x = −2y, y = −x, x + 3y + 2 = 0 ∗ 40. Find the principal axes and principal moments of inertia about the
origin for the uniform square plate bounded by the lines x = 0, y = 0,
x = a, y = a.
In Exercises 21–27 find the centroid of the region bounded by the ∗ 41. Find the principal axes and principal moments of inertia about the
curves. origin for the uniform rectangular plate bounded by the lines x = 0,
√ y = 0, x = a , y = b , where a > b > 0.
∗ 21. x = y + 2, y = x, y = 0
∗ 22. y + x 2 = 0, x = y + 2, x + y + 2 = 0, y = 2 (above ∗ 42. Show that the values of I in equation 13.40 are indeed maximum
y + x 2 = 0) and minimum values of I (m) as defined by 13.38 for −∞ < m < ∞ .
∗ 23. y 2 = x 4 (1 − x 2 ) (right loop) ∗ 43. Suppose that a thin plate with constant mass per unit area is sym-
∗ 24. 3x 2 + 4y 2 = 48, (x − 2)2 + y 2 = 1 metric about a line & and P is a point on & . Show that the product
√ moment of inertia of the plate about & and a line through P perpendic-
∗ 25. y = ln x, y + x − 1 = 0, x = 2 ular to & vanishes.
√ ∗ 44. Show that if a thin plate with constant mass per unit area has an
∗ 26. y = 2 − x, 15y = x 2 − 4
√ axis of symmetry, then the axis of symmetry must be a principal axis
∗ 27. y = x 1 − x 2 , x ≥ 0 and the x -axis about any point on the line.
∗ 45. Show that if θ is the angle of inclination of a principal axis (about parallel line through the centre of mass plus the mass multiplied by the
2Ixy square of the distance between the lines.
the origin), then tan 2θ = .
Iy − Ix ∗ 49. Suppose that a thin plate with constant mass per unit area occupies
a region R of the xy -plane. Let x = x1 and x = x2 be any two
∗ 46. Prove the theorem of Pappus: If a plane area is revolved about a vertical lines and Ix1 and Ix2 be moments of inertia of the plate about
coplanar axis not crossing the area, the volume generated is equal to these lines. Show that
the product of the area and the circumference of the circle described by
the centroid of the area. Ix2 = Ix1 + M [x22 − x12 + 2x(x1 − x2 )],
∗ 47. A thin flat plate of area A is immersed vertically in a fluid with where M is the mass of the plate and x is the x -coordinate of its centre
density ρ . Show that the total force (due to fluid pressure) on each side of mass. Does this result reduce to the parallel axis theorem of Exercise
of the plate is equal to the product of 9.81, A , ρ , and the depth of the 48 when one of the lines passes through the centre of mass?
centroid of the plate below the surface of the fluid. Use this result to
find the forces in some of the problems in Exercises 13.4, say 1, 3, 4, 5, ∗ 50. Suppose that a thin plate with constant mass per unit area occupies
7, 11, 12, 13, 16, and 17. For those problems involving triangles recall a region R of the xy -plane, and (x, y) is its centre of mass. Let Ixy be
the result of Exercise 43 in Section 7.7 or Exercise 47 in Section 11.3. the product moment of inertia of the plate about the x - and y -axes, and
Ixy be the product moment of inertia of the plate about the lines through
∗ 48. Prove the parallel axis theorem for thin plates: The moment of (x, y) parallel to the x - and y -axes. Verify that Ixy = Ixy − Mx y ,
inertia of a thin plate (with constant mass per unit area) with respect to where M is the mass of the plate. (This is called the parallel axis
any coplanar line is equal to the moment of inertia with respect to the theorem for product moments of inertia.)
13.6 Surface Area

To find the length of a curve in Section 7.3 we approximated the curve by tangent line segments.
To find the area of a surface we follow a similar procedure by approximating the surface with
tangential planes. In particular, consider finding the area of a smooth surface S given that every
vertical line that intersects the surface does so in exactly one point (Figure 13.36). If Sxy is the
region in the xy -plane onto which S projects, we divide Sxy into n subregions with areas !Ai
in any fashion whatsoever, and choose a point (xi , yi ) in each !Ai . At the point (xi , yi , zi )
on the surface S that projects onto (xi , yi ) , we draw the tangent plane to S . Suppose we now
project !Ai upward onto S and onto the tangent plane at (xi , yi , zi ) and denote these projected
areas by !Si and !ST i , respectively.
FIGURE 13.36 Curved surface area is defined in terms of flat areas on tangent planes to the surface
z
γi Tangent plane
k
ni
∆STi
(xi, yi , zi )
∆Si
S
∆A1
x
Sxy ∆A2
y
∆Ai
(xi , yi )
Now !ST i is an approximation to !Si and as long as !Ai is small, a reasonably good
FIGURE 13.37 Area on
a slanted plane is related to its approximation. In fact, the smaller !Ai , the better the approximation. We therefore define the
projection in the xy -plane through area of S as follows:
n
!
the angle between the planes
area of S = lim !ST i , (13.41)
n→∞
k i=1
γ
ni i where in taking the limit we demand that each !Ai shrink to a point. We have therefore defined
area on a curved surface in terms of flat areas on tangent planes to the surface. The advantage
∆STi of this definition is that we can calculate !ST i in terms of !Ai . To see how, we denote by n̂i
the unit normal vector to S at (xi , yi , zi ) with positive z -component, and by γi the acute angle
between n̂i and k̂. Now !ST i projects onto !Ai , and γi is the acute angle between the planes
containing !Ai and !ST i (Figure 13.37). It follows that !Ai and !ST i are related by
∆ Ai
!Ai = cos γi !ST i (13.42)
(see Exercise 28 in Section 11.5). Note that if !ST i is horizontal, then γi = 0 and !ST i =
!Ai ; and if !ST i tends toward the vertical (γi → π/2) , then !ST i becomes very large for
fixed !Ai .
Because the surface projects in a one-to-one fashion onto the area Sxy in the xy -plane, we
can take the equation for S in the form z = f (x, y) . A vector normal to S at any point is
therefore & '
∂f ∂f
∇(z − f (x, y)) = − , − , 1 ;
∂x ∂y
hence & ' & '
∂f ∂f ∂z ∂z
− ,− ,1 − ,− ,1
∂x ∂y ∂x ∂y
n̂ = ; = ;
& '2 & '2 & '2 & '2 .
∂f ∂f ∂z ∂z
1+ + 1+ +
∂x ∂y ∂x ∂y
Since n̂i · k̂ = |n̂i ||k̂| cos γi = cos γi , it follows that
1
cos γi = n̂i · k̂ = < .
1 + zx2 (xi , yi ) + zy2 (xi , yi )
When we substitute this expression into 13.42, we obtain the result that area !ST i on the tangent
plane to z = f (x, y) at the point (xi , yi , zi ) is related to its projection !Ai in the xy -plane
according to
<
!ST i = 1 + zx2 (xi , yi ) + zy2 (xi , yi ) !Ai . (13.43)
Formula 13.41 for the area of S can now be written in the form
n <
!
area of S = lim 1 + zx2 (xi , yi ) + zy2 (xi , yi ) !Ai , (13.44)
"!Ai "→0
i=1
where the summation is carried out over all areas !Ai in Sxy as each !Ai shrinks to a point.
%
But this is the definition of the double integral of the function 1 + (∂z/∂x)2 + (∂z/∂y)2
over the region Sxy . In other words, on the basis of formula 13.41, areas on surfaces can be
calculated according to
; & '2 & '2
""
∂z ∂z
area of S = 1+ + dA. (13.45)
Sxy ∂x ∂y
Note
% the analogy between equations 13.45 and 7.15. In equation 7.15 we think of
1 + (dy/dx)2 dx as a small length % along a curve C : y = f (x) that projects onto the
length dx along the x -axis. In fact, 1 + (dy/dx)2 dx is along the tangent line to C , but we
think of it as along C itself (see Section 11.11). The total length of C is then found by adding
% all projections of C from x = a to x = b . Similarly, in equation 13.45 we think of
over
FIGURE 13.38 Divide the 1 + (∂z/∂x)2 + (∂z/∂y)2 dA as a small area on a surface S : z = f (x, y) that projects
area of this surface into two parts
onto the area dA in the xy -plane. It is, in fact, a small area on the tangent plane to S , but it is
S1 and S2 with curve C and find
usually easier to think of it as being on S itself. The total area of S is then the addition over the
both areas
projection Sxy of S .
z
Note, too, that when S is smooth, ∂z/∂x and ∂z/∂y are continuous, thus guaranteeing
existence of double integral 13.45. If S is piecewise smooth, we divide it into smooth subsurfaces
and integrate over each piece separately.
This discussion has been based on the assumption that S projects one-to-one onto some
S1 region Sxy in the xy -plane. If this condition is not met, one possibility is to subdivide S into
C parts, each of which projects one-to-one onto the xy -plane. The total area of S is then the sum
of the areas of its parts. For instance, to find the area of the surface S in Figure 13.38, we could
subdivide S into S1 and S2 along the curve C . The area of S is then the sum of the areas of S1
and S2 , each of which projects one-to-one onto the xy -plane.
S2
Alternatively, we could note that there is nothing sacred about projecting S onto the xy -
plane. We could develop similar results if S projects one-to-one onto either the yz - or the
y
xz -plane. If Syz and Sxz represent these projections, then
x
; & '2 & '2
""
∂x ∂x
area of S = 1+ + dA, (13.46a)
Syz ∂y ∂z
; & '2 & '2
""
∂y ∂y
area of S = 1+ + dA. (13.46b)
Sxz ∂x ∂z
EXAMPLE 13.15
Find the area of that part of the plane x + 2y + 3z = 6 in the first octant.
FIGURE 13.39 To find
the area of x + 2y + 3z = 6 in
the first octant, project it onto the
SOLUTION This area projects one-to-one onto the triangular area Sxy in the xy -plane bounded
xy -plane by the lines x = 0, y = 0, x + 2y = 6 (Figure 13.39). Since z = 2 − x/3 − 2y/3,
z ; & '2 & '2 ; & '2 & '2

"" ""
∂z ∂z 1 2
area = 1+ + dA = 1+ − + − dA
2 Sxy ∂x ∂y Sxy 3 3
S √ √ √
"" 9 :
14 14 14 1 √
= dA = (area of Sxy ) = (3)(6) = 3 14.
3 3 Sxy 3 3 2
y
x + 2y = 6, We could also use formula 11.42.
6 z=0
x
EXAMPLE 13.16
Find the area of the surface z = x 3/2 that projects onto the rectangle in the xy -plane bounded
by the straight lines x = 0, x = 2, y = 1, and y = 3.
SOLUTION Since the surface projects one-to-one onto the rectangle (Figure 13.40), we find
that
; & '2 & '2 ; & '2

"" ""
∂z ∂z 3
area = 1+ + dA = 1+ x 1/2 dA
Sxy ∂x ∂y Sxy 2
" 2" 3 " 2 " 2
1 √ 1 √ √
= 4 + 9x dy dx = {y 4 + 9x}31 dx = 4 + 9x dx
2 0 1 2 0 0
# $2
2 2 √
= (4 + 9x)3/2 = (22 22 − 8).
27 0 27
FIGURE 13.40 To find the area of z = x 3/2 above the rectangle, project it onto the rectangle
z = x 3/ 2
dx 3
y=1
2
dy y
Sxy
x
x=2 y=3
EXAMPLE 13.17
√
Find the area of the cone y = x 2 + z2 to the left of the plane y = 1.
SOLUTION Method 1 The surface projects one-to-one onto the interior of the circle
x 2 + z2 = 1 in the xz -plane (Figure 13.41a). Since the area of the surface is four times that in
the first octant, if we let Sxz be the quarter-circle x 2 + z2 ≤ 1, x ≥ 0, z ≥ 0, then
; & '2 & '2

""
∂y ∂y
area = 4 1+ + dA
Sxz ∂x ∂z
; & '2 & '2
""
x z
=4 1+ √ + √ dA
Sxz x 2 + z2 x 2 + z2
;
"" "" √
x2 z2
=4 1+ 2 + dA = 4 2 dA
Sxz x + z2 x 2 + z2 Sxz
√ √ √
= 4 2(area of Sxz ) = 4 2[ 41 π(1)2 ] = 2 π.
FIGURE 13.41a To FIGURE 13.41b To

find the surface area of the cone, find the surface area of the cone,
project it onto the xz -plane project it onto the xy -plane
z z
x 2 + z 2 = 1,
y=0
y = x2 + z2 y = x2 + z2
Sxz
dx
1 x
y 1 y
x
y = x, Sxy
z=0
y=1
Method 2 Suppose instead that we project that part of the surface in the first octant onto
the triangle Sxy in Figure 13.41b. We write the equation of this part of the surface in the form
%
z = y 2 − x 2 and calculate
; & '2 & '2

""
∂z ∂z
area = 4 1+ + dA
Sxy ∂x ∂y
= @ A2 @ A2
"">
> −x y
=4 ?1+ % + % dA
Sxy y2 − x2 y2 − x2
"" ; √ ""
x2 y2 y
=4 1+ 2 2
+ 2 2
dA = 4 2 % dA.
Sxy y −x y −x Sxy y2 − x2
To evaluate this double integral, it is advantageous to integrate first with respect to y :
√ " 1" 1
y √ " 1 %
area = 4 2 % dy dx = 4 2 { y 2 − x 2 }1x dx
0 x 2
y −x 2 0
√ " 1 %
=4 2 1 − x 2 dx.
0
If we now set x = sin θ , then dx = cos θ dθ , and
√ " π/2 √ " π/2 & '

1 + cos 2θ
area = 4 2 cos θ cos θ dθ = 4 2 dθ
0 0 2
# $π/2
√
1 √
= 2 2 θ + sin 2θ = 2 π.
2 0
EXERCISES 13.6
In Exercises 1–6 find the area required. ∗ 12. The area of z = ln (1 + x + y) in the first octant cut off by
y = 1 − x2
1. The area of 2x + 3y + 6z = 1 in the first octant
2. The area of x + 2y − 3z + 4 = 0 for which x ≤ 0, y ≤ 0 and ∗ 13. Find the area of the surface z = ln x that projects onto the rectangle
z≥0 in the xy -plane bounded by the lines x = 1, x = 2, y = 0, y = 2.
%
3. The area of z = 1 − 4 x 2 + y 2 above the xy -plane ∗ 14. Verify that the area of the √
curved portion of a right-circular cone
√ of radius r and height h is π r r 2 + h2 .
4. The area of z = 2xy cut out by the planes x = 1, x = 2, y = 1,
y=3
∗ 5. The area in the first octant cut out from the surface z = x + y by In Exercises 15–19 set up, but do not evaluate, double iterated integrals
the plane x + 2y = 4 to find the required area.
∗ 6. The area of z = x 3/2 + y 3/2 in the first octant cut off by the plane
x+y =1 ∗ 15. The area of y = x 2 + z2 cut off by y + z = 1
∗ 16. The area of y = z2 + x inside x 2 + y 2 = 1

In Exercises 7–12 set up, but do not evaluate, double iterated integrals
to find the required area. ∗ 17. The area of y 2 = z + x 2 inside x 2 + y 2 = 4
%
∗ 7. The area of x 2 + y 2 + z2 = 2 inside the cone z = x 2 + y 2
∗ 18. The area of z = x 3 + y 3 that is in the first octant and between the
2 2
∗ 8. The area of 4x = y + z cut off by x = 4 planes x + y = 1 and x + y = 2
∗ 9. The area in the first octant cut from y = xz by the cylinder x 2 + ∗ 19. The area of x 2 + y 2 = z2 + 1 between the planes z = 1 and
z2 = 1 z=4
∗ 10. The area of z = (x 2 + y 2 )2 below z = 4
∗ 20. Find the area of that part of the surface z = 2x 2 + 3y bounded
2 2
∗ 11. The area of y = 1 − x − 3z to the right of the xz -plane by the planes x = 2, y = 0, and y = x .
13.7 Double Iterated Integrals in Polar Coordinates

So far we have used only double iterated integrals in x and y to evaluate double integrals. But
for some problems this is not convenient. For instance, the double integral of a continuous
FIGURE 13.42 Double function f (x, y) over the region R in Figure 13.42 requires three double iterated integrals in
integrals in polar coordinates over x and y . In other words, a subdivision of R into rectangles by coordinate lines x = constant
this area are convenient
and y = constant is simply not convenient for this region. For such an area, polar coordinates
y are more suitable.
x2 + y2 = b2 Polar coordinates with the origin as pole and the positive x -axis as polar axis are defined
y = m2x by
R x = r cos θ, y = r sin θ
(see Section 9.2). We wish to obtain double iterated integrals in polar coordinates that
y = m1 x represent the double integral of f (x, y) over the region R in Figure 13.42. To do this we return
x 2 + y 2 = a2 to definition 13.3 for a double integral, and choose a subdivision of R into subareas convenient
x to polar coordinates. When using Cartesian coordinates we drew coordinate lines x = constant
and y = constant. When using polar coordinates we draw coordinate curves r = constant and
θ = constant. In particular, we subdivide R by a network of n + 1 circles r = ri , where
a = r0 < r1 < r2 < · · · < rn−1 < rn = b,
and m + 1 radial lines θ = θj , where
α = θ0 < θ1 < θ2 < · · · < θm−1 < θm = β

(Figure 13.43a). If !Aij represents the area bounded by the circles r = ri−1 and r = ri and
the radial lines θ = θj −1 and θ = θj (Figure 13.43b), then it is straightforward to show that
1
!Aij = (ri2 − ri−
2
1 )(θj − θj −1 ).
2
If we set !ri = ri − ri−1 and !θj = θj − θj −1 , then
& '
1 ri + ri−1
!Aij = (ri + ri−1 )(ri − ri−1 )(θj − θj −1 ) = !ri !θj . (13.47)
2 2
Our next task in using equation 13.3 for the double integral of f (x, y) over R is to choose
a star point in each !Aij . If we select
& '
ri + ri−1
(ri∗ , θj∗ ) = , θj ,
2
FIGURE 13.43a Division of region FIGURE 13.43b Area of small

into subregions using polar coordinates element using polar coordinates
y * *
r = ri
m−1
R
r = ri−1 ∆Aij
θj − 1
r2
r1 ∆Aij
r0 = a ri rn = b
ri−1 rn−1
then
!Aij = ri∗ !ri !θj ,
and by equation 13.3,
"" m !
! n
f (x, y) dA = lim f (ri∗ cos θj∗ , ri∗ sin θj∗ ) !Aij
R "!Aij "→0
j =1 i=1
m !
! n
= lim f (ri∗ cos θj∗ , ri∗ sin θj∗ )ri∗ !ri !θj .
"!ri "→0
"!θj "→0 j =1 i=1
If we take the limit first as "!ri " → 0 and then as "!θj " → 0, we obtain the double
iterated integral
"" " β" b
f (x, y) dA = f (r cos θ, r sin θ )r dr dθ. (13.48)
R α a
Reversing the order of taking limits reverses the order of the iterated integrals:
"" " b" β
f (x, y) dA = f (r cos θ, r sin θ )r dθ dr. (13.49)
R a α
For the region R of Figure 13.43a, then, there are two double iterated integrals in polar coordi-
nates representing the double integral of f (x, y) over R .
We have interpreted double iterated integrals in Cartesian coordinates as integrations over
horizontal or vertical strips. Double iterated integrals in polar coordinates can also be interpreted
geometrically. Take, for instance, equation 13.48. A double iterated integral in polar coordinates
implies a subdivision of the region R into areas as shown in Figure 13.44. Let us denote small
variations in r and θ for a representative piece of area at position (r, θ ) by dr and dθ . If dr
and dθ are very small (as is implied in the definition of the double integral), then this piece of
area is almost rectangular with an approximate area of (r dθ ) dr . In polar coordinates, then,
we think of dA in equation 13.48 as being replaced by
dA = r dr dθ. (13.50)
Each such area at (r, θ ) is multiplied by the value of f (x, y) at (r, θ ) to give the product
f (r cos θ, r sin θ )r dr dθ.
FIGURE 13.44 Area element in polar coordinates ex- FIGURE 13.45 Double iterated integral with respect to r
pressed in terms of differentials is dA = r drdθ and θ adds first inside a wedge, and then over all wedges
y Length = dr y
B
dr
r=b
dr
r=a
x x
The inner integral

" b
f (r cos θ, r sin θ )r dr dθ
a
with respect to r holds θ constant and is therefore interpreted as a summation over the small
areas in a wedge dθ from r = a to r = b . The θ -integration then adds over all wedges starting
at θ = α and ending at θ = β . Limits on θ therefore identify positions of first and last wedges.
If the order of integration is reversed (equation 13.49), then the inner integral
" d
f (r cos θ, r sin θ )r dθ dr
c
holds r constant. We interpret this as an addition over the small areas in a ring dr , where the
limits indicate that each and every ring starts on the curve θ = α and ends on the curve θ = β .
The outer r -integration is an addition over all rings with the first ring at r = a and the last at
r = b.
Double iterated integrals in polar coordinates for more general regions are now quite simple.
For the region R of Figure 13.45,
"" " β" h(θ )
f (x, y) dA = f (r cos θ, r sin θ )r dr dθ.
R α g(θ )
EXAMPLE 13.18
Evaluate the double iterated integral
" 1" √
−x 2 +x
y 2 dy dx.
0 0
SOLUTION The limits identify the region of integration as the interior of the semicircle in
Figure 13.46a. The integrand suggests an interpretation of the integral as the second moment
of area of this semicircle about the x -axis. Since the semicircle R in Figure 13.46b has exactly
the same second moment about the x -axis, we can state that
" 1" √ "" " π"

−x 2 +x 1/2
2 2
y dy dx = y dA = (r 2 sin2 θ )r dr dθ
0 0 R 0 0
" # $1/2 "
π
r4 2 1 π
= sin θ dθ = sin2 θ dθ
0 4 0 64 0
" π # $π
1 1 sin 2θ π
= (1 − cos 2θ ) dθ = θ− = .
128 0 128 2 0 128
FIGURE 13.46a Limits FIGURE 13.46b Inte-

indicate that the region of integra- gration of x 2 over this semicircle
tion is the semicircle gives the same result
y y
y = −x 2 + x x2 + y2 = 1
or 4
or
2
1 1
x−1 + y2 = R r=
2
2 4
1 1 x 1 x
2 2
EXAMPLE 13.19
FIGURE 13.47 Centroid Find the centroid of the region R in Figure 13.47.
of the semicircular annulus
y
SOLUTION Evidently, x = 0, and the area of the region is A = (π b2 − π a 2 )/2. Since
x2 + y2 = b2
or "" " π" " # $b
r=b b π
r3
Ay = y dA = (r sin θ )r dr dθ = sin θ dθ
R R 0 a 0 3 a
" π
1 3 3 1 2
= (b − a ) sin θ dθ = (b3 − a 3 ){− cos θ }π0 = (b3 − a 3 ),
3 0 3 3
a b x
x 2 + y 2 = a2 2 2 4 b2 + ab + a 2
or it follows that y = (b3 − a 3 ) = .
3 π(b2 − a 2 ) 3π a+b
r=a
EXAMPLE 13.20
%
Find the area of that portion of the sphere x 2 + y 2 + z2 = 2 inside the cone z = x2 + y2 .
SOLUTION If S is that portion of the sphere that is inside the cone and also in the first octant
(Figure 13.48), then the required area is four times that of S ; that is,
; & '2 & '2
""
∂z ∂z
area = 4 1+ + dA,
Sxy ∂x ∂y
where Sxy is the projection of S on the xy -plane. The curve of intersection of the cone and the
sphere has equations
x 2 + y 2 + z 2 = 2, or equivalently, x 2 + y 2 = 1,
%
z = x2 + y2, z = 1.
FIGURE 13.48 Polar coordinates are advantageous in

calculating the area of that part of a sphere inside a cone
z
2 S
x 2 + y2 + z2 = 2
z = x 2 + y2
1
1
2
2 y
Sxy x 2 + y 2 = 1,
x z=0
Consequently, Sxy is the interior of the quarter-circle x 2 + y 2 ≤ 1, x ≥ 0, y ≥ 0. On S ,
∂z x ∂z y
=− and =− ,
∂x z ∂y z
so that
; ;
"" ""
x2 y2 x 2 + y 2 + z2
area = 4 1 + 2 + 2 dA = 4 dA
Sxy z z Sxy z2
"" 6 √ ""
2 1
=4 2
dA = 4 2 % dA.
Sxy z Sxy 2 − x2 − y2
If we now use polar coordinates to evaluate this double integral, we have

√ " π/2" 1
1 √ " π/2 %
area = 4 2 √ r dr dθ = 4 2 {− 2 − r 2 }10 dθ
0 0 2 − r2 0
√ √ " π/2 √ √ √ √
π
= 4 2( 2 − 1) dθ = 4 2( 2 − 1) = 2 2π( 2 − 1).
0 2
Here is perhaps our trickiest consultation project. Liquid flow through a pipe of radius R
is controlled by a circular valve of the same radius that moves right-to-left across the pipe
(Figure 13.49). In order to calibrate the amount of flow through the pipe, we are being
asked to find a formula for the area of flow through the pipe as a function of the position
a of the centre of the valve; that is, when the centre of the valve is at position x = a ,
what is the area inside the pipe not covered by the valve?
FIGURE 13.49 Valve to restrict flow in a circular pipe
y
R
Pipe Valve
a x
SOLUTION Why do we claim that this project is so tricky? It would seem to be a simple
matter to use a double integral in polar coordinates to find the area common to the pipe
and the valve and subtract this from π R 2 , the area of the pipe. In principle this is correct,
but setting up the double iterated integral for the common area has hidden difficulties. We
shall see this as we proceed. In polar coordinates, the equation for the pipe is r = R ;
in Cartesian coordinates, the equation for the valve is (x − a)2 + y 2 = R 2 . When we
change this to polar coordinates, the result is
(r cos θ − a)2 + (r sin θ )2 = R 2 *⇒ r 2 − 2ar cos θ + (a 2 − R 2 ) = 0.
When we use the quadratic formula to find an explicit definition of the curve, we obtain
%
r = a cos θ ± R 2 − a 2 sin2 θ.
To find the point of intersection of the pipe and valve in the first quadrant, we set
% a
R = a cos θ − R 2 − a 2 sin2 θ, which simplifies to cos θ = .
2R
B a C
Let us denote the solution by θ = Cos −1 . Figure 13.50 shows the situation when
2R
the valve is only slightly closed.
FIGURE 13.50 Flow when valve almost fully open
y
R
Pipe Valve
a x
To find the area of the flow through the pipe, we subtract the area common to pipe and
valve from π R 2 . The common area is
" θ" R
A=2 √ r dr dθ
0 a cos θ − R 2 −a 2 sin2 θ
" θ %
= [R 2 − (a cos θ − R 2 − a 2 sin2 θ)2 ] dθ
0
" θ %
= [R 2 − a 2 cos2 θ + 2a cos θ R 2 − a 2 sin2 θ − R 2 + a 2 sin2 θ ] dθ
0
" θ %
=a [a(− cos 2θ ) + 2 cos θ R 2 − a 2 sin2 θ ] dθ.
0
On the second term, we make the change of variable R sin φ = a sin θ , in which
case R cos φ dφ = a cos θ dθ . If R sin φ = a sin θ , then
# $θ " φ
2 − sin 2θ R
A=a + 2a cos φ R cos φ dφ
2 0 0 a
" & ' # $φ
a2 φ
1 + cos 2φ a2 sin 2φ
=− sin 2θ + 2R 2 dφ = − sin 2θ +R 2 φ +
2 0 2 2 2 0
2 2 2
a R D E a D E
=− sin 2θ + 2φ + sin 2φ = − sin 2θ + R 2 φ + sin φ cos φ
2 2 2
4 6 5
a Baa 2 C
= −a 2 sin θ cos θ + R 2 Sin−1 sin θ + sin θ 1 − 2 sin2 θ
R R R
6 @ 6 A
B a C a 2 a a 2
= −a 2 1− + R 2 Sin−1 1−
2R 4R 2 R 4R 2
6 ; & '
a2 a2 a2
+ aR 1 − 1− 2 1−
4R 2 R 4R 2
@ √ A √
2 −1 a 4 R 2 − a2 a(a 2 − R 2 ) 4R 2 − a 2
= R Sin − .
2R 2 2R 2
Area of the flow is therefore

@ √ A √
2 2 −1 a 4 R 2 − a2 a(a 2 − R 2 ) 4R 2 − a 2
A = π R − R Sin + .
2R 2 2R 2
It is worth noting that when a = 2R , the largest value for a , this reduces to π R 2 , as it
should.
As the valve continues to close, it reaches the position in Figure 13.51a where the line
joining the origin to the point of intersection of the curves is tangent to the valve. When
the valve is to the left of this position, the previous calculation of A is no longer valid.
To find the value of a at this stage, we calculate the slope of the valve by differentiating
(x − a)2 + y 2 =%R 2 , to get 2(x − a) + 2yy - = 0, and evaluate y - at the point of
intersection (a/2, R 2 − a 2 /4) ,
a/2 − a a
y- = − % = √ .
2 2
R − a /4 4R − a 2
2
%
When we equate this to the slope of the line joining the origin to (a/2, R 2 − a 2 /4 ) ,
we obtain %
a R 2 − a 2 /4
√ , =
a/2
4R 2 − a 2
√
and when this equation is solved for a , the result
√ is a = 2R . Thus, the above calculation
for A is valid only in the interval R ≤ a ≤ 2R .

Flow when valve is at intermediate position
y y
R R
Pipe Valve Pipe Valve
a x a x
When the centre of the valve is just to the left of this position (Figure 13.51b), the
area common to the pipe and valve is the sum of A as calculated above and twice the
small area A between the dashed line and the valve,
" θ" √
a cos θ + R 2 −a 2 sin2 θ
A=2 √ r dr dθ,
θ a cos θ − R 2 −a 2 sin2 θ
where θ = Cos −1 [a/(2R)], as above. Angle θ is defined by the equation

% % & '
−1 R
a cos θ − R 2 − a 2 sin2 θ = a cos θ + R 2 − a 2 sin2 θ *⇒ θ = Sin .
a
Thus,
" θ % %
A= [(a cos θ + R 2 − a 2 sin2 θ)2 − (a cos θ − R 2 − a 2 sin2 θ)2 ] dθ
θ
" θ %
= 4a cos θ R 2 − a 2 sin2 θ dθ .
θ
When we use the above substitution R sin φ = a sin θ to evaluate this integral, the result is
4 @ √ A √ 5
2 −1 a 4R 2 − a 2 a(2R 2 − a 2 ) 4R 2 − a 2
A=R π − 2Sin − .
2R 2 2R 4
Area of the flow is therefore

@ √ A
a 4 R 2 − a2
A = π R 2 − A − A = π R 2 − R 2 Sin−1
2R 2
√
a(a 2 − R 2 ) 4R 2 − a 2
+
2R 2
4 @ √ A √ 5
2
2
−1 a 4 R − a
2 a(2R 2 − a 2 ) 4R 2 − a 2
− R π − 2Sin −
2R 2 2R 4
@ √ A √
2 2 a 4R 2 − a 2
2 −1 a 4 R − a
= R Sin + .
2R 2 2
√
This is valid when the position of the centre of the valve
√ satisfies R ≤ a ≤ 2R . Note
that this calculation agrees with that for A when a = 2R ; both give A = (π + 2)R 2 /2.
Finally, we calculate that for 0 ≤ a ≤ R (Figure 13.51c),
" π" R
A=2 √ r dr dθ
θ a cos θ + R 2 −a 2 sin2 θ
" π %
= [R 2 − (a cos θ + R 2 − a 2 sin2 θ)2 ] dθ
θ
" π %
= [−a 2 cos 2θ − 2a cos θ R 2 − a 2 sin2 θ ] dθ.
θ
Once again we use the substitution R sin φ = a sin θ to evaluate the second term, the
final result being
@ √ A √
a 4 R 2 − a2 a 4R 2 − a 2
2 −1
A = R Sin + .
2R 2 2
Putting all these results together, the area A of flow as a function of position a of the
centre of the valve is
 @ √ A √

 2 a 4R 2 − a 2 a 4R 2 − a 2 √
 − 1
 R Sin

2R 2
+
2
, 0 ≤ a ≤ 2R
A= @ √ A √

 a 4 R 2 −a 2 a(a 2
−R 2
) 4R 2 − a 2 √
 π R −R Sin
 2 2 − 1
+ , 2R < a ≤ 2R .
 2R 2 2R 2
FIGURE 13.51c Flow when valve almost closed
y
R
Pipe Valve
a x
EXERCISES 13.7
In Exercises 1–5 evaluate the double integral of the function over the ∗ 24. The roof of an exhibition hall consists of cylindrical concrete shells
region R . as shown below. The length of each shell (not shown) is 50 m. The
underside of each shell is half of a circular cylinder with radius 7 m.
2 +y 2
1. f (x, y) = ex , where R is bounded by x 2 + y 2 = a 2 The thickness of the shell (measured radially) is 10 cm at the base and
% decreases linearly with angle θ to 5 cm at the top. Find the volume of
2. f (x, y) = x , where R is bounded by x = 2y − y 2 , x = 0
% √ each shell.
3. f (x, y) = x 2 + y 2 , where R is bounded by y = 9 − x 2 , y =
x, x = 0
%
4. f (x, y) = 1/ x 2 + y 2 , where R is the region outside x 2 +y 2 =
4 and inside x 2 + y 2 = 4x
%
5. f (x, y) = 1 + 2x 2 + 2y 2 , where R is bounded by x 2 + y 2 = 5 cm
1, x 2 + y 2 = 4
7m
In Exercises 6–7 evaluate the double iterated integral.
10 cm
" 1" √1−x 2% " 0 " √4−y 2
6. x 2 + y 2 dy dx 7. √ x 2 dx dy
0 0 − 2 −y ∗ 25. Find the area inside the circle x 2 + y 2 = 4x and outside the circle
x 2 + y 2 = 1.
In Exercises 8–12 find the area of the region bounded by the curves. ∗ 26. Find the volume of the solid of revolution when a circle of radius
R is rotated about a tangent line.
√
8. Outside x 2 + y 2 = 9 and inside x 2 + y 2 = 2 3 y ∗ 27. A circular plate of radius R (figure below) has a uniform charge
9. r = 9(1 + cos θ) distribution of ρ coulombs per square metre. If P is a point directly
10. r = cos 3θ above the centre of the plate and dA is a small area on the plate, then
the potential at P due to dA is given by
∗ 11. Common to r = 2 and r 2 = 9 cos 2θ
∗ 12. Common to r = 1 + sin θ and r = 2 − 2 sin θ 1 ρ dA
,
4π,0 s
∗ 13. Find the
% centroid of the region bounded by the curves y = x, y =
−x, x = 2 − y 2 . where s is the distance from P to dA .
∗ 14. Find the second moment of area for a circular plate of radius R (a) Show that in terms of polar coordinates, the potential V at
about any diameter. P due to the entire plate is
∗ 15. A water tank in the form of a right-circular cylinder with radius R
and length h has its axis horizontal. If it is full, what is the force due " π " R
ρ r
to water pressure on each end? V = √ dr dθ ,
4π ,0 −π 0 r2 + d2
In Exercises 16–18 find the area of the surface. where d is the distance from P to the centre of the plate.
(b) Evaluate the double iterated integral to find V .
∗ 16. The area of z = x 2 + y 2 below z = 4
∗ 17. The area of x 2 + y 2 + z2 = 4 inside x 2 + y 2 = 1 z
∗ 18. The area of z = xy inside x 2 + y 2 = 9
P (0, 0, d)
∗ 19. Prove that the area of a sphere of radius R is 4π R 2 .

∗ 20. Find the area of the hyperbolic paraboloid z = x 2 − y 2 between
the cylinders x 2 + y 2 = 1 and x 2 + y 2 = 4. s
In Exercises 21–22 find the volume of the solid of revolution obtained R

dA
by rotating the region bounded by the curve about the line. y
R
∗ 21. r = cos2 θ about the x -axis
x
∗ 22. r = 1 + sin θ about the y -axis
∗ 28. Use Coulomb’s law (see Example 11.8 in Section 11.3) to find the
∗ 23. Find the area of the region bounded by the curve (x 2 + y 2 )3 = force on a charge q at point P due to the charge on the plate in Exercise
4a 2 x 2 y 2 . 27. What happens to this force as the radius of the plate gets very large?
In Exercises 29–31 find the area of the region bounded by the curves. ∗ 36. Suppose we denote P R 2 /(4nL) in Exercise 35 by Vmax , the
maximum velocity at the centre of the blood vessel, v = Vmax [1 −
∗ 29. (x 2 + y 2 )2 = 2xy (r/R)2 ]. Suppose the vessel converges to a smaller vessel of radius
∗ 30. Inside both r = 6 cos θ and r = 4 − 2 cos θ R1 = αR , 0 < α < 1. If blood flow in the smaller vessel also has
∗ 31. r = cos2 θ sin θ a parabolic profile v = Umax [1 − (r/R1 )2 ], find Umax in terms of
Vmax and α . Hint: Assume that the volume flow rates in the two
∗ 32. Find the centroid of the region bounded by the cardioid r = 1 + vessels must be the same.
cos θ . ∗ 37. Repeat Exercise 36 for flow in a square pipe that reduces from a
∗ 33. Find the second moment of area about the x -axis for the region width of L to one of width αL (0 < α < 1). Assume a velocity profile
bounded by r 2 = 9 cos 2θ . in the larger pipe of the form v = Vmax (1 − 4x 2 /L2 )(1 − 4y 2 /L2 ) ,
∗ 34. Evaluate the double integral of −L/2 ≤ x ≤ L/2, −L/2 ≤ y ≤ L/2, and a similar profile in the
smaller pipe.
;
1 − x2 − y2 ∗∗ 38. Find the area of that part of the sphere x 2 + y 2 + z2 = a 2 inside
f (x, y) = (x 2 + y 2 )2 = a 2 (x 2 − y 2 ) .
1 + x2 + y2
∗∗ 39. Find the area of that portion of the surface x 2 + z2 = a 2 cut out
by x 2 + y 2 = a 2 .
over the region inside the circle x 2 + y 2 = 1.
∗ 35. The figure below illustrates a piece of an artery or vein with circular ∗∗ 40. A very important integral in statistics is
cross-section (radius R ). The speed of blood flowing through the blood " ∞
2
vessel is not uniform because of the viscosity of the blood and friction I = e−x dx.
at the walls. Poiseuille’s law states that for laminar blood flow, the 0
speed v of blood at a distance r from the centre of the vessel is given
by To evaluate the integral we set
P
v = (R 2 − r 2 ), " ∞
4nL 2
I = e−y dy,
where P is the pressure difference between the ends of the vessel, L 0
is the length of the vessel, and n is the viscosity of the blood. Find the
amount of blood flowing over a cross-section of the blood vessel per and then multiply these two equations. Do this to prove that I =
√
unit time. π /2.
∗∗ 41. Use the result of Exercise 40 to evaluate the gamma function
R " ∞
-(n) = x n−1 e−x dx
0
L at n = 1/2.
13.8 Triple Integrals and Triple Iterated Integrals

Triple integrals are defined in much the same way as double integrals. Suppose f (x, y, z) is
a function defined in some region V of space that has finite volume (Figure 13.52). We divide
FIGURE 13.52 Subdivide
V into n subregions of volumes !V1 , !V2 , . . . , !Vn in any manner whatsoever, and in each
a region V into smaller volumes subregion !Vi (i = 1, . . . , n) we choose an arbitrary point (xi∗ , yi∗ , zi∗ ) . We then form the
to define the triple integral of sum
f (x, y, z) f (x1∗ , y1∗ , z1∗ ) !V1 + f (x2∗ , y2∗ , z2∗ ) !V2 + · · · + f (xn∗ , yn∗ , zn∗ ) !Vn
! n
z
= f (xi∗ , yi∗ , zi∗ ) !Vi . (13.51)
V i=1
∆Vi If this sum approaches a limit as the number of subregions becomes increasingly large and every
subregion shrinks to a point, we call the limit the triple integral of f (x, y, z) over the region
V and denote it by
(x*i, y*i, z*i,) """ n
!
f (x, y, z) dV = lim f (xi∗ , yi∗ , zi∗ ) !Vi . (13.52)
"!Vi "→0
x y
V i=1
As in the case of double integrals, we require that this limit be independent of the manner
of subdivision of V and the choice of star points in the subregions. This is guaranteed for
continuous functions by the following theorem.
THEOREM 13.2
Let S be a piecewise-smooth surface that encloses a region V with finite volume. If
f (x, y, z) is continuous inside and on S , then the triple integral of f (x, y, z) over V
exists.
Properties analogous to those in equations 13.4–13.7 hold for triple integrals, although we
will not list the first three here. Corresponding to equation 13.7, the volume of a region V is
given by the triple integral
"""
volume of V = dV . (13.53)
V
We evaluate triple integrals with triple iterated integrals. If we use Cartesian coordinates
there are six possible triple iterated integrals of a function f (x, y, z) , corresponding to the six
permutations of the product of the differentials dx , dy , and dz :
dz dy dx, dz dx dy, dx dz dy, dx dy dz, dy dx dz, dy dz dx.
The general triple iterated integral of f (x, y, z) with respect to z , y , and x is of the form
" b" g2 (x)" h2 (x,y)
f (x, y, z) dz dy dx. (13.54)
a g1 (x) h1 (x,y)
Because the first integration with respect to z holds x and y constant, the limits on z may
therefore depend on x and y . Similarly, the second integration with respect to y holds x
constant, and the limits may be functions of x .
EXAMPLE 13.21
" 1" x 2" x+y
Evaluate the triple iterated integral xyz dz dy dx .
0 0 xy
SOLUTION
" 1" x 2" " 1" x2 # $x+y
x+y
xyz2
xyz dz dy dx = dy dx
0 0 xy 0 0 2 xy
" 1" x2
1
= [xy(x + y)2 − xy(xy)2 ] dy dx
2 0 0
" 1" x2
1
= (x 3 y + 2x 2 y 2 + xy 3 − x 3 y 3 ) dy dx
2 0 0
" 1 # $x 2
1 x 3y 2 2x 2 y 3 xy 4 x 3y 4
= + + − dx
2 0 2 3 4 4 0
" 1
1
= (6x 7 + 8x 8 + 3x 9 − 3x 11 ) dx
24 0
# $1
1 3x 8 8x 9 3x 10 x 12 19
= + + − = .
24 4 9 10 4 0 270
Because of the analogy between double and triple integrals, we accept without proof that triple
integrals can be evaluated with triple iterated integrals. We must, however, examine how triple
iterated integrals bring about the summations represented by triple integrals, for it is only by
understanding this process thoroughly that we can obtain limits for triple iterated integrals.
In Section 13.2 we discussed in considerable detail evaluation of double integrals by means
of double iterated integrals. In particular, we showed that double iterated integrals in Cartesian
coordinates represent the subdivision of an area into small rectangles by coordinate lines x =
constant and y = constant. The first integration creates a summation over rectangles in a strip,
and the second integration adds over all strips. It is fairly straightforward to generalize these
ideas to triple integrals. Consider evaluating the triple integral
"""
f (x, y, z) dV
V
over the region V in Figure 13.53a bounded above by the surface z = h(x, y) , below by the
region R in the xy -plane, and on the sides by a cylindrical wall standing on the curve bounding R .
The choice of a triple iterated integral in Cartesian coordinates to evaluate this triple integral
implies a subdivision of V into small rectangular parallelepipeds (boxes, for short) by means
of coordinate planes x = constant, y = constant, and z = constant. The dimensions of a
representative box at position (x, y, z) in V are denoted by dx, dy , and dz , with resulting
volume dx dy dz . If we decide on a triple iterated integral with respect to z, y , and x , then
the first integration on z holds x and y constant. This integration therefore adds the quantities
f (x, y, z) dz dy dx over boxes in a vertical column of cross-sectional dimensions dx and dy .
Lower and upper limits on z identify where each and every column starts and stops, and must
consequently be 0 and h(x, y) :
" h(x,y)
f (x, y, z) dz dy dx.
0
Since this integration produces a function of x and y alone, the remaining integration with
respect to y and x is essentially a double iterated integral in the xy -plane. These integrations
must account for all columns in V and therefore the region in the xy -plane over which this
double iterated integral is performed is the region R upon which all columns in V stand. Since
the y -integration adds inside a strip in the y -direction and limits identify where all strips start
and stop, they must therefore be g1 (x) and g2 (x) . We now have
" g2 (x)" h(x,y)

g1 (x) 0
Finally, the x -integration adds over all strips and the limits are x = a and x = b :
""" " b" g2 (x)" h(x,y)

f (x, y, z) dV = f (x, y, z) dz dy dx.
V a g1 (x) 0
Suppose now that V is the region bounded above by the surface z = h2 (x, y) and below by
z = h1 (x, y) (Figure 13.53b). In this case, the limits on the first integration with respect to z
are h1 (x, y) and h2 (x, y) since every column starts on the surface z = h1 (x, y) and ends on
the surface z = h2 (x, y) :
" h2 (x,y)
h1 (x,y)
For the volume of Figure 13.53a we interpreted the final two integrations as a double iterated
integral in the xy -plane over the region R from which all columns emanated. For the present
volume we interpret R as the region in the xy -plane onto which all columns project. We then
obtain
""" " b" g2 (x)" h2 (x,y)
f (x, y, z) dV = f (x, y, z) dz dy dx.
V a g1 (x) h1 (x,y)
FIGURE 13.53a Columns start on the xy -plane z = 0 FIGURE 13.53b Columns start and stop on surfaces
and end on surface z = h(x, y) . R is the area in the xy -plane z = h1 (x, y) and z = h2 (x, y) . R is the area in the xy -plane
on which all columns stand onto which all columns project
z z
z = h (x, y) z = h 2(x, y)
dz
V z = h1(x, y)
dz
a a
b b
y R y
x R x
dx dx
y = g1(x) dy y = g1(x) dy
y = g2(x) y = g2 (x)
Schematically, we have obtained the following interpretation for the limits of triple iterated
integrals in Cartesian coordinates:
 

 dz dy dx 
 dz dx dy 
 

" position of " where every " where every 
 

last strip strip stops column stops dx dz dy
f (x, y, z) .
position of where every where every
 
 dx dy dz 
first strip strip starts column starts 
 

 dy dx dz 
 
dy dz dx
EXAMPLE 13.22
Set up the six triple iterated integrals in Cartesian coordinates for the triple integral of a function
f (x, y, z) over the region V in the first octant bounded by the surfaces
y 2 + z2 = 1, y = x, z = 0, x = 0.
SOLUTION The triple integral of f (x, y, z) over V is given by each of the following triple
iterated integrals (see Figure 13.54):
" 1" 1" √1−y 2 " 1" y" √1−y 2
f (x, y, z) dz dy dx, f (x, y, z) dz dx dy,
0 x 0 0 0 0
" 1" √
1−z2" y " 1" √1−y 2" y
f (x, y, z) dx dy dz, f (x, y, z) dx dz dy,
0 0 0 0 0 0
" 1" √ √ √ √
1−x 2" 1−z2 " 1" 1−z2" 1−z2
f (x, y, z) dy dz dx, f (x, y, z) dy dx dz.
0 0 x 0 0 x
FIGURE 13.54 The six triple iterated integrals for the

volume bounded by y = x , y 2 + z2 = 1, z = 0, x = 0
z y=x
x 2 + z2 = 1, 1 y 2 + z 2 = 1,
y=0 x=0
y2 + z2 = 1
1
1 y
x
EXAMPLE 13.23
Evaluate the triple integral of f (x, y, z) = xyz over the region V bounded by the surfaces
√
z= y, y + z = 2, x = 0, z = 0, x = 2.
SOLUTION If we choose a triple iterated integral with respect to z , y , and x , some columns
√
end on the parabolic cylinder z = y (Figure 13.55) and others end on the plane y + z = 2.
We therefore require two such iterated integrals,
""" " 2" 1" √
y " 2" 2" 2−y
xyz dz dy dx = xyz dz dy dx + xyz dz dy dx
V 0 0 0 0 1 0
Columns " 2" 1 # $√y " 2" 2 # $2−y

FIGURE 13.55 xyz2 xyz2
in the z -direction necessitate two = dy dx + dy dx
triple iterated integrals 0 0 2 0 0 1 2 0
" 2" 1 " 2" 2
z 1 1
z= y (0, 1, 1)
= xy 2 dy dx + x(4y − 4y 2 + y 3 ) dy dx
2 0 0 2 0 1
" 2 # $
3 1 " 2 # & '$2
1 xy 1 4y 3 y4
y+z=2
= dx + x 2y 2 − + dx
2 0 3 0 2 0 3 4 1
" 2 " 2
1 5
2 = x dx + x dx
y
6 0 24 0
# $2
2 3 x2
=
x 8 2 0
x=2
3
= .
4
Only one iterated integral is required if integration is first performed with respect to y ,
namely
" 1" 2" 2−z " 2" 1" 2−z
xyz dy dx dz or xyz dy dz dx.
0 0 z2 0 0 z2
13.9 Volumes 945
EXERCISES 13.8
In Exercises 1–12 evaluate the triple integral over the region. In Exercises 14–17 set up, but do not evaluate, a triple iterated integral
""" for the triple integral.
1. (x 2 z + yex ) dV , where V is bounded by x = 0, x = 1, """
V
y = 1, y = 2, z = 0, z = 1 ∗ 14. (x 2 + y 2 + z2 ) dV , where V is bounded by
""" % V
z= 1 − x2 − y2, z = x2
2. x dV , where V is bounded by x = 0, y = 0, z = 0, x + """
V
y+z=4 ∗ 15. xz sin (x + y) dV , where V is bounded by y 2 = 1 +
""" V √
3. sin (y + z) dV , where V is bounded by z = 0, y = 4x 2 + 4z 2 , y = 4 + x 2
V """
2x, y = 0, x = 1, z = x + 2y ∗ 16. xyz dV , where V is bounded by z = x 2 + 4y 2 , 2x +
""" % V
4. xy dV , where V is enclosed by z = 1 − x2 − y2 , z = 0 8y + z = 4
V """
"""
∗ 17. x 2 y 2 z2 dV , where V is bounded by x = y 2 +z2 , x + 1 =
5. dV , where V is bounded by x 2 + y 2 = 1, x 2 + z2 = 1 V
"""
V (y 2 + z2 )2
6. (x 2 + 2z) dV , where V is bounded by z = 0, y + z =
V
In Exercises 18–23 evaluate the triple integral over the region.
4, y = x 2
""" """
7. x 2 y 2 z2 dV , where V is bounded by z = 1 + y, y + z = ∗ 18. (y + x 2 ) dV , where V is bounded by x + z2 = 1, z =
V V
1, x = 1, x = 0, z = 0 x + 1, y = 1, y = −1
""" """
8. xyz dV , where V is the first octant region cut out by z = ∗ 19. (xy + z) dV , where V is bounded by y + z = 1, z =
V% V
2 2
x + y , z = x2 + y2 2y, z = y, x = 0, x = 3
""" """
9. dV , where V is bounded by z = x 2 , y + z = 4, y = 0 ∗ 20. dV , where V is bounded by z = 0, x 2 + y 2 = 1, x +
V V
""" y+z=2
"""
10. (x + y + z) dV , where V is bounded by x = 0, x = 1,
V ∗ 21. dV , where V is bounded by z = x 2 +y 2 , z = 4 −x 2 −y 2
z = 0, y + z = 2, y = z V
""" """
11. xyz dV , where V is bounded by z = 1, z = x 2 /4 +y 2 /9 ∗ 22. (x + y + z) dV , where V is bounded by 2z = y 2 −
V V
""" x2, z = 1 − x2
12. x 2 y dV , where V is the first octant region bounded by """
V ∗ 23. |yz| dV , where V is bounded by z2 = 1 + x 2 + y 2 ,
z = 1, z = x 2 /4 + y 2 /9 % V
z = 4 − x2 − y2
∗ 13. Set up the six triple iterated integrals in Cartesian coordinates for ∗ 24. Set up, but do not evaluate, triple iterated integrals to evaluate the
the triple integral of a function f (x, y, z) over the region enclosed by triple integral of the function f (x, y, z) = x 2 +y 2 +z 2
% over the region
2 2 2
the surfaces y = 1 − x 2 , z = 0, and y = z . bounded by the surfaces x + y = z + 1, 2z = x 2 + y 2 , z = 0.
13.9 Volumes
Because the volume of a region V is represented by triple integral 13.53 and triple integrals
are evaluated by means of triple iterated integrals, it follows that volumes can be evaluated with
triple iterated integrals. For example, to evaluate the volume of the region in Figure 13.53b
using a triple iterated integral in x, y , and z , we subdivide V into boxes of dimensions dx, dy ,
and dz and therefore of volume dz dy dx . Integration with respect to z adds these volumes in
the z -direction to give the volume of a vertical column (Figure 13.56):
" h2 (x,y)
dz dy dx.
h1 (x,y)
Limits indicate that all columns start on the surface z = h1 (x, y) and end on the surface
z = h2 (x, y) . Integration with respect to y now adds volumes of columns that project onto a
strip in the y -direction:
" g2 (x)" h2 (x,y)
dz dy dx,
g1 (x) h1 (x,y)
where the limits indicate that all strips start on the curve y = g1 (x) and end on the curve
y = g2 (x) . Evidently, this integration yields the volume of a slab as shown in Figure 13.56.
Finally, integration with respect to x adds the volumes of all such slabs in V :
" b" g2 (x)" h2 (x,y)

dz dy dx,
a g1 (x) h1 (x,y)
where a and b designate positions of first and last strips.
FIGURE 13.56 The integrations when a triple iterated integral is used to find volume
z
z = h 2(x, y)
z = h 1(x, y)
b
y
x
y = g1(x)
y = g2(x)
EXAMPLE 13.24
Find the volume bounded by the planes z = x + y , y = 2x , z = 0, x = 0, y = 2.
SOLUTION If we use vertical columns (Figure 13.57), we have
" 1" 2 " " 1" 2 " 1 # $2

x+y
y2
volume = dz dy dx = (x + y) dy dx = xy + dx
0 2x 0 0 2x 0 2 2x
" 1 # $1
2 x2 2x 3 5
=2 (1 + x − 2x ) dx = 2 x + − = .
0 2 3 0 3
13.9 Volumes 947
FIGURE 13.57 The volume bounded by the planes z = x + y , y = 2x , z = 0, x = 0, y = 2
z
z=x+y
2
x y
y=2
y = 2x
EXAMPLE 13.25
Find the volume in the first octant cut from the cyclinder x 2 + z2 = 4 by the plane y + z = 6.
SOLUTION With columns in the y -direction (Figure 13.58), we have
" 2" √ √
4−x 2" 6−z " 2" 4−x 2
volume = dy dz dx = (6 − z) dz dx
0 0 0 0 0
" 2 # $√4−x 2
z2
= 6z − dx
0 2 0
" 2 # % $
1 2
= 6 4− x2 − (4 − x ) dx.
0 2
In the first term we set x = 2 sin θ , from which we get dx = 2 cos θ dθ , and
" π/2 # $2
x3
volume = 6 (2 cos θ ) · 2 cos θ dθ + −2x +
0 6 0
" π/2
8
= 12 (1 + cos 2θ ) dθ −
0 3
# $π/2
1 8 8
= 12 θ + sin 2θ − = 6π − .
2 0 3 3
FIGURE 13.58 Volume in first octant cut from cylinder x 2 + z2 = 4 by plane y + z = 6
y+z=6
6
2 y
x
x2 + z2 = 4
Had we used iterated integrals with respect to z, y , and x in this example, we would have had
two integrals,
" 2" √ √
6− 4−x 2" 4−x 2 " 2" 6 " 6−y
volume = dz dy dx + √ dz dy dx.
0 0 0 0 6− 4−x 2 0
We are being consulted by an architect who wants to know the volume of air contained in
a certain type of structure. It has a horizontal, polygonal base (Figure 13.59a) on which
stand vertical walls all of the same height H . Above this is a roof formed in the following
way. There is a peak, height h above the top of the walls, that is joined to the tops of the
walls by planes.
FIGURE 13.59a Volume FIGURE 13.59b Roof

in structure portion of structure
P
H
SOLUTION There is no problem with the lower part of the structure; the volume of air
contained in it is the area of the base multiplied by H . To calculate the volume of the
upper part, we take it aside as in Figure 13.59b and drop a perpendicular from the top
to the polygonal base. Now join the foot of the perpendicular P to each vertex of the
polygon. Finally draw vertical planes each containing one of these lines, the perpendicular
from top to base, and a slanted edge of the roof. What this does is divide the volume into
tetrahedrons (four-sided figures) each with one horizontal and two vertical planes. The
remaining face of the tetrahedron is a slanted face of the roof. If we find the volume in one
such tetrahedron, we can add to find the total volume in the upper portion of the structure.
To find the volume of any one
FIGURE 13.60 Volume in one part
of the tetrahedrons, let us place its
of roof structure
triangular base in the xy -plane with
one vertex at the origin, a second z
D(0, 0, h)
vertex on the x -axis, and the third
vertex in the first quadrant (Figure
13.60). The top of the tetrahedron
will be on the z -axis. Denote coordi-
nates of the tetrahedron by O(0, 0, 0) , 0
A(a, 0, 0) , B(b, c, 0) , and D(0, 0, h) .
Equations of lines AB and OB are x B(b, c, 0)
A(a, 0, 0) y
y = c(x−a)/(b−a) and y = cx/b ,
respectively. A normal vector to plane
ABD is
13.9 Volumes 949
G G
G î ĵ k̂ G
G G
AB × AD = G b − a c 0 G = (ch, ah − bh, ac).
G −a 0 h G
The equation of plane ABD is therefore ch(x − a) + (ah − bh)y + acz = 0. The
volume of the tetrahedron can now be calculated with a triple integral,
" c" a+(b−a)y/c " [(bh−ah)y−ch(x−a)]/(ac)
V = dz dx dy
0 by/c 0
" c" a+(b−a)y/c 9 :
(bh − ah)y − ch(x − a)
= dx dy
0 by/c ac
" c# $a+(b−a)y/c
1 ch
= h(b − a)xy − (x − a)2 dy
ac 0 2 by/c
" c7 9 : 9 :2
1 (b − a)y ch (b − a)y
= h(b − a)y a + − a+ −a
ac 0 c 2 c
& ' & '2 8
by ch by
− h(b − a)y + −a dy
c 2 c
7
1 h y3 y2 h y3
= (b − a)2 + ah(b − a) − (b − a)2
ac c 3 2 2c 3
& '3 8c
bh y3 c2 h by
− (b − a) + −a
c 3 6b c
0
1
= ach.
6
Since the area of the triangular base is ac/2, we have shown that the volume of the
tetrahedron is the area of the triangular base multiplied by one-third of the height h . It
follows that the volume of the upper structure is the area of the polygonal base multiplied
by h/3. Finally, then, if A represents the area of the polygonal base, the volume of air in
the structure is & '
h
V =A H + .
3
EXERCISES 13.9
In Exercises 1–17 find the volume of the region bounded by the sur- 6. x + y + z = 4, y = 3z, x = 0, y = 0
faces.
7. x 2 + y 2 = 4, y 2 + z2 = 4
1. y = x 2 , y = 1, z = 0, z = 4
8. y = x 2 − 1, y = 1 − x 2 , x + z = 1, z = 0
2. x = z2 , z = x 2 , y = 0, y = 2
3. x = 3z, z = 3x, y = 1, y = 0, x = 2 9. z = 16 − x 2 − 4y 2 , x + y = 1, z = 16, x ≥ 0, y ≥ 0
4. x + y + z = 6, y = 4 − x 2 , z = 0, y = 0 10. z = x 2 + y 2 , x = 1, z = 0, x = y, x = 2y
5. z = x 2 + y 2 , y = x 2 , y = 4, z = 0 11. z = 1 − x 2 − y 2 , z = 0
12. x − z = 0, x + z = 3, y + z = 1, z = y + 1, z = 0 22. f (x, y, z) = x 2 +y 2 +z2 over the region bounded by the surfaces
x = 0, x = 1, y + z = 2, y = 2, z = 2
13. x + y + z = 2, x 2 + y 2 = 1, z = 0
14. y + z = 1, z = 2y, z = y, x = 0, x + y + z = 4
∗ 23. Find the volume bounded by the surfaces z = x 2 − y 2 and z =
15. x 2 + 4y 2 = z, x 2 + 4y 2 = 12 − 2z 4 − 2(x 2 + y 2 ) .
∗ 16. y = 1 − z2 , y = z2 − 1, x = 1 − z2 , x = z2 − 1
∗ 24. Verify that the surfaces z = x 2 − y 2 and z = 4 − x 2 − y 2 do not
∗ 17. x + 3y + 2z = 6, z = 0, y = x, y = 2x bound a finite volume.
∗ 18. Find the volume in the first octant bounded by the plane 2x + y + ∗ 25. Find the volume bounded by the surfaces x + z = 2, z =
z = 2 and inside the cylinder y 2 + z2 = 1. 0, 4y 2 = x(2 − z) .
∗ 19. A pyramid has a square base with side length b and has height h ∗ 26. Find the volume bounded by the surfaces z = (x − 1)2 +y 2 , 2x +
at its centre. z = 2.
(a) Find its volume by taking cross-sections parallel to the base
(see Section 7.9). ∗ 27. Find the volume inside the ellipsoid x 2 /a 2 + y 2 /b2 + z2 /c2 = 1.
(b) Find its volume using triple integrals.
∗∗ 28. The bottom and sides of a boat are defined by the surface equation
x = 10(1 − y 2 − z2 ) , 0 ≤ x ≤ 10, where all dimensions are in
The average value of a function f (x, y, z) over a region with volume metres.
V is defined as
""" (a) Find the volume of water displaced by the boat when the
1 water level on the side of the boat is d metres below the top
f = f (x, y, z) dV .
V V of the boat.
In Exercises 20–22 find the average value of the function over the (b) Archimedes’ principle states that the buoyant force on an
region. object when immersed or partially immersed in a fluid is
equal to the weight of the fluid displaced by the object. Find
20. f (x, y, z) = xy over the region bounded by the surfaces x = 0, the maximum weight of the boat and contents just before
y = 0, z = 0, x + y + z = 1 sinking.
21. f (x, y, z) = x +y +z over the region in the first octant bounded
by the surfaces z = 9 − x 2 − y 2 , z = 0 and for which 0 ≤ x ≤ ∗∗ 29. Find the volume inside all three surfaces x 2 +y 2 = a 2 , x 2 +z2 =
1, 0 ≤ y ≤ 1 a 2 , y 2 + z2 = a 2 .
13.10 Centres of Mass and Moments of Inertia

In this section we discuss centres of mass and moments of inertia for three-dimensional objects
of density ρ(x, y, z) (mass per unit volume). If we divide the object occupying region V in
Figure 13.61 into small volumes dV , then the mass in dV is ρ dV . The triple integral
"""
M = ρ dV (13.55)
V
FIGURE 13.61 Centre of adds the masses of all such volumes (of ever-decreasing size) to produce the total mass M of
mass of a 3-dimensional object the object.
z Corresponding to equations 13.32 and 13.33 for first moments of planar masses about the
coordinate axes are the following formulas for first moments of the object about the coordinate
V planes:
"""
dV
first moment of object about yz-plane = xρ dV , (13.56a)
V
"""
first moment of object about xz-plane = yρ dV , (13.56b)
x y V
"""
first moment of object about xy -plane = zρ dV . (13.56c)
V
The centre of mass of the object is defined as that point (x, y, z) at which a particle of mass
M would have the same first moments about the coordinate planes as the object itself. Since the
first moments of M at (x, y, z) about the coordinate planes are Mx, My , and Mz , it follows
that we can use the equations
""" """ """
Mx = xρ dV , My = yρ dV , Mz = zρ dV , (13.57)
V V V
to solve for (x, y, z) once M and the integrals on the right have been evaluated.
If we use a triple iterated integral with respect to z , y , and x to evaluate the third of these,
say, for the object in Figure 13.56, then
""" " b" g2 (x)" h2 (x,y)
zρ dV = zρ dz dy dx.
V a g1 (x) h1 (x,y)
Quantity zρ dz dy dx is the first moment about the xy -plane of the mass in an elemental box
of dimensions dx , dy , and dz . The z -integration then adds these moments over boxes in the
z -direction to give the first moment about the xy -plane of the mass in a vertical column:
" h2 (x,y)
zρ dz dy dx.
h1 (x,y)
The y -integration then adds the first moments of columns that project onto a strip in the y -
direction: " " g2 (x) h2 (x,y)
zρ dz dy dx.
g1 (x) h1 (x,y)
This quantity therefore represents the first moment about the xy -plane of the slab in Figure 13.56.
Finally, the x -integration adds first moments of all such slabs to give the total first moment of
V about the xy -plane:
" b" g2 (x)" h2 (x,y)
zρ dz dy dx.
a g1 (x) h1 (x,y)
EXAMPLE 13.26
Find the centre of mass of an object of constant density ρ if it is bounded by the surfaces
z = 1 − y 2 , x = 0, z = 0, x = 2.
FIGURE 13.62 Centre of mass of uniform solid

bounded by z = 1 − y 2 , x = 0, z = 0, x = 2
z = 1 − y2
2
x x=2
SOLUTION From the symmetry of the object (Figure 13.62), we see x = 1 and y = 0. Now
" 2" 1" 1−y 2 " 2" 1
M =2 ρ dz dy dx = 2ρ (1 − y 2 ) dy dx
0 0 0 0 0
" 2 # $1 " 2
y3 4ρ 8ρ
= 2ρ y− dx = dx =
0 3 0 3 0 3
and
" 2" 1" 1−y 2 " 2" 1 # $1−y 2
z2
Mz = 2 zρ dz dy dx = 2ρ dy dx
0 0 0 0 0 2 0
" 2" 1
=ρ (1 − 2y 2 + y 4 ) dy dx
0 0
" 2 # $1 " 2
2y 3 y5 8ρ 16ρ
=ρ y− + dx = dx = .

Calculus For Engineers, 4th Edition

Uploaded by

Copyright:

Available Formats

Calculus For Engineers, 4th Edition

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Calculus For Engineers, 4th Edition

Uploaded by

Copyright:

Available Formats

Calculus for Engineers

Editor-in-Chief: Gary Bennett

Printed and bound in United States.

Chapter 1 Calculus Preparation 1

Chapter 2 Limits and Continuity 104

Chapter 3 Differentiation 149

Chapter 4 Applications of Differentiation 237

Chapter 5 The Indefinite Integral and the Antiderivative 335

Chapter 6 The Definite Integral 374

Chapter 7 Applications of the Definite Integral 410

Chapter 8 Techniques of Integration 490

Chapter 9 Parametric Equations and Polar Coordinates 541

Chapter 10 Infinite Sequences and Series 587

10.11 Limit Ratio and Limit Root Tests 672

Chapter 11 Vectors and Three-Dimensional Analytic Geometry 695

Chapter 12 Differential Calculus of Multivariable Functions 799

12.15 Taylor Series for Multivariable Functions 890

Chapter 13 Multiple Integrals 896

Chapter 14 Vector Calculus 982

Chapter 15 Differential Equations 1047

Appendix A Mathematical Induction A–1

Appendix B Determinants B–1

Appendix C Complex Numbers C–1

Answers to Even-Numbered Exercises E–1

About the Author

• Electronic calculators or software packages such as Maple, Mathematica, MathCad, and

• A concise Summary is given near the end of each chapter.

Treatment of Particular Topics

New in This Edition

Right circular cylinder: Right circular cone:

Cosine law- c2 = a 2 + b2 − 2ab cos C

tan A + tan B tan A − tan B

1.2 Solving Polynomial Equations

DIAGNOSTIC TEST FOR SECTION 1.2

5. x 2 + 6x − 4 = 0 6. 36x 2 − 108x + 81 = 0 13. Zeros of x 3 + 6x 2 − x − 30 are x = −5, −3, 2.

A real polynomial in x of degree n , where n ≥ 0 is an integer, is an expression of the

Pn (x) = an x n + an−1 x n−1 + · · · + a1 x + a0 = 0 (1.2)

is called a polynomial equation of degree n . Examples are 3x 3 − 2x + 5 = 0 and 2x 2 + 5x −

Application of this formula to the equation 2x 2 + 4x − 5 = 0 gives

and after that the quartic,

THEOREM 1.1 (Factor Theorem)

P (x) = x 4 + 2x 3 + x 2 − 2x − 2 = (x − 1)(x 3 + 3x 2 + 4x + 2).

What this means is that equation 1.8a can be replaced by

P (x) = (x − 1)(x 3 + 3x 2 + 4x + 2) = 0. (1.8b)

P (x) = (x − 1)(x + 1)(x 2 + 2x + 2) = 0. (1.8c)

x 4 + 2x 3 + x 2 − 2x − 2 = (x − 1)(x + 1)(x 2 + 2x + 2), (1.9a)

0 = (x − 2)(x 2 + 3x − 10) = (x − 2)(x + 5)(x − 2) = (x − 2)2 (x + 5).

Thus, x = 2 is a solution with multiplicity 2, and x = −5 is a solution with multiplicity 1.

THEOREM 1.3 (Rational Root Theorem)

±1, ±2, ±4, ±1/5, ±2/5, ±4/5.

0 = (x − 1)(x − 1)(x 2 + x − 2) = (x − 1)2 (x + 2)(x − 1) = (x − 1)3 (x + 2).

7. x 2 + 8x + 16 = 0 8. 4x 2 − 36x + 81 = 0 ∗ 28. 25x 4 − 120x 3 + 109x 2 − 36x + 4 = 0

9. 2x 2 + 5x − 10 = 0 10. 4x 2 − 8x + 9 = 0 ∗ 29. x 5 + 9x 4 + 47x 3 + 125x 2 + 18x − 200 = 0

11. x 3 − 3x 2 + 3x − 1 = 0 12. 8x 3 + 12x 2 + 6x + 1 = 0 ∗ 30. x 6 + 16x 4 − 81x 2 − 1296 = 0

∗ 16. x 4 + 7x 3 + 9x 2 − 21x − 36 = 0 33. x 3 − 11x 2 + 4x + 60, x = −2, 3, 10

∗ 17. x 4 − 16 = 0 34. 24x 3 + 22x 2 − 27x + 5, x = −5/3, 1/4, 1/2

35. x 4 − 2x 3 + 2x − 1, x = −1, x = 1 (muliplicity 3)

ANSWERS TO DIAGNOSTIC TEST FOR SECTION 1.2

1. −5/9 (1 mark) 8. 1 with multiplicity 2, 2 (3 marks)