Vector Calculus - Corral PDF

1
0.8
0.6
0.4
0.2
z
0
-0.2 -10
-0.4 -5
-10 0 x
-5
0 5
y 5 10
10
Vector
Calculus
Michael Corral
Vector Calculus
Michael Corral
Schoolcraft College
About the author:
Michael Corral is an Adjunct Faculty member of the Department of Mathematics at
Schoolcraft College. He received a B.A. in Mathematics from the University of California
at Berkeley, and received an M.A. in Mathematics and an M.S. in Industrial & Operations
Engineering from the University of Michigan.
This text was typeset in LATEX 2 with the KOMA-Script bundle, using the GNU Emacs text
editor on a Fedora Linux system. The graphics were created using MetaPost, PGF, and
Gnuplot.
Copyright 2008 Michael Corral.

Permission is granted to copy, distribute and/or modify this document under the terms of the
GNU Free Documentation License, Version 1.2 or any later version published by the Free
Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
Texts. A copy of the license is included in the section entitled GNU Free Documentation
License.
Preface
This book covers calculus in two and three variables. It is suitable for a one-semester course,
normally known as Vector Calculus, Multivariable Calculus, or simply Calculus III.
The prerequisites are the standard courses in single-variable calculus (a.k.a. Calculus I and
II).
I have tried to be somewhat rigorous about proving results. But while it is important for
students to see full-blown proofs - since that is how mathematics works - too much rigor and
emphasis on proofs can impede the flow of learning for the vast majority of the audience at
this level. If I were to rate the level of rigor in the book on a scale of 1 to 10, with 1 being
completely informal and 10 being completely rigorous, I would rate it as a 5.
There are 420 exercises throughout the text, which in my experience are more than
enough for a semester course in this subject. There are exercises at the end of each sec-
tion, divided into three categories: A, B and C. The A exercises are mostly of a routine
computational nature, the B exercises are slightly more involved, and the C exercises usu-
ally require some effort or insight to solve. A crude way of describing A, B and C would be
Easy, Moderate and Challenging, respectively. However, many of the B exercises are
easy and not all the C exercises are difficult.
There are a few exercises that require the student to write his or her own computer pro-
gram to solve some numerical approximation problems (e.g. the Monte Carlo method for
approximating multiple integrals, in Section 3.4). The code samples in the text are in the
Java programming language, hopefully with enough comments so that the reader can figure
out what is being done even without knowing Java. Those exercises do not mandate the use
of Java, so students are free to implement the solutions using the language of their choice.
While it would have been simple to use a scripting language like Python, and perhaps even
easier with a functional programming language (such as Haskell or Scheme), Java was cho-
sen due to its ubiquity, relatively clear syntax, and easy availability for multiple platforms.
Answers and hints to most odd-numbered and some even-numbered exercises are pro-
vided in Appendix A. Appendix B contains a proof of the right-hand rule for the cross prod-
uct, which seems to have virtually disappeared from calculus texts over the last few decades.
Appendix C contains a brief tutorial on Gnuplot for graphing functions of two variables.
This book is released under the GNU Free Documentation License (GFDL), which allows
others to not only copy and distribute the book but also to modify it. For more details, see
the included copy of the GFDL. So that there is no ambiguity on this matter, anyone can
make as many copies of this book as desired and distribute it as desired, without needing
my permission. The PDF version will always be freely available to the public at no cost
(go to http://www.mecmath.net). Feel free to contact me at [email protected] for
iii
iv Preface
any questions on this or any other matter involving the book (e.g. comments, suggestions,
corrections, etc). I welcome your input.
Finally, I would like to thank my students in Math 240 for being the guinea pigs for the
initial draft of this book, and for finding the numerous errors and typos it contained.
January 2008 M ICHAEL C ORRAL

Contents
Preface iii
1 Vectors in Euclidean Space 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Vector Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Cross Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Lines and Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.6 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.7 Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.8 Vector-Valued Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.9 Arc Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2 Functions of Several Variables 65

2.1 Functions of Two or Three Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.2 Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.3 Tangent Plane to a Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.4 Directional Derivatives and the Gradient . . . . . . . . . . . . . . . . . . . . . . . 78
2.5 Maxima and Minima . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.6 Unconstrained Optimization: Numerical Methods . . . . . . . . . . . . . . . . . . 89
2.7 Constrained Optimization: Lagrange Multipliers . . . . . . . . . . . . . . . . . . 96
3 Multiple Integrals 101

3.1 Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.2 Double Integrals Over a General Region . . . . . . . . . . . . . . . . . . . . . . . . 105
3.3 Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.4 Numerical Approximation of Multiple Integrals . . . . . . . . . . . . . . . . . . . 113
3.5 Change of Variables in Multiple Integrals . . . . . . . . . . . . . . . . . . . . . . . 117
3.6 Application: Center of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.7 Application: Probability and Expected Value . . . . . . . . . . . . . . . . . . . . . 128
4 Line and Surface Integrals 135

4.1 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.2 Properties of Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.3 Greens Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
v
vi Contents
4.4 Surface Integrals and the Divergence Theorem . . . . . . . . . . . . . . . . . . . . 156

4.5 Stokes Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.6 Gradient, Divergence, Curl and Laplacian . . . . . . . . . . . . . . . . . . . . . . . 177
Bibliography 187
Appendix A: Answers and Hints to Selected Exercises 189
Appendix B: Proof of the Right-Hand Rule for the Cross Product 192
Appendix C: 3D Graphing with Gnuplot 196
GNU Free Documentation License 201
History 209
Index 210
1 Vectors in Euclidean Space
1.1 Introduction
In single-variable calculus, the functions that one encounters are functions of a variable
(usually x or t) that varies over some subset of the real number line (which we denote by R).
For such a function, say, y = f ( x), the graph of the function f consists of the points ( x, y) =
( x, f ( x)). These points lie in the Euclidean plane, which, in the Cartesian or rectangular
coordinate system, consists of all ordered pairs of real numbers (a, b). We use the word
Euclidean to denote a system in which all the usual rules of Euclidean geometry hold. We
denote the Euclidean plane by R2 ; the 2 represents the number of dimensions of the plane.
The Euclidean plane has two perpendicular coordinate axes: the x-axis and the y-axis.
In vector (or multivariable) calculus, we will deal with functions of two or three variables
(usually x, y or x, y, z, respectively). The graph of a function of two variables, say, z = f ( x, y),
lies in Euclidean space, which in the Cartesian coordinate system consists of all ordered
triples of real numbers (a, b, c). Since Euclidean space is 3-dimensional, we denote it by R3 .
The graph of f consists of the points ( x, y, z) = ( x, y, f ( x, y)). The 3-dimensional coordinate
system of Euclidean space can be represented on a flat surface, such as this page or a black-
board, only by giving the illusion of three dimensions, in the manner shown in Figure 1.1.1.
Euclidean space has three mutually perpendicular coordinate axes ( x, y and z), and three
mutually perpendicular coordinate planes: the x y-plane, yz-plane and xz-plane (see Figure
1.1.2).
z z
c
P(a, b, c)
yz-plane
b y xz-plane y
0 0
a x y-plane
x x
Figure 1.1.1 Figure 1.1.2
1
2 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE
The coordinate system shown in Figure 1.1.1 is known as a right-handed coordinate

system, because it is possible, using the right hand, to point the index finger in the positive
direction of the x-axis, the middle finger in the positive direction of the y-axis, and the thumb
in the positive direction of the z-axis, as in Figure 1.1.3.
z
Figure 1.1.3 Right-handed coordinate system
An equivalent way of defining a right-handed system is if you can point your thumb up-
wards in the positive z-axis direction while using the remaining four fingers to rotate the
x-axis towards the y-axis. Doing the same thing with the left hand is what defines a left-
handed coordinate system. Notice that switching the x- and y-axes in a right-handed
system results in a left-handed system, and that rotating either type of system does not
change its handedness. Throughout the book we will use a right-handed system.
For functions of three variables, the graphs exist in 4-dimensional space (i.e. R4 ), which
we can not see in our 3-dimensional space, let alone simulate in 2-dimensional space. So
we can only think of 4-dimensional space abstractly. For an entertaining discussion of this
subject, see the book by A BBOTT.1
So far, we have discussed the position of an object in 2-dimensional or 3-dimensional space.
But what about something such as the velocity of the object, or its acceleration? Or the
gravitational force acting on the object? These phenomena all seem to involve motion and
direction in some way. This is where the idea of a vector comes in.
1 One thing you will learn is why a 4-dimensional creature would be able to reach inside an egg and remove the
yolk without cracking the shell!

1.1 Introduction 3
You have already dealt with velocity and acceleration in single-variable calculus. For
example, for motion along a straight line, if y = f ( t) gives the displacement of an object after
time t, then d y/ dt = f ( t) is the velocity of the object at time t. The derivative f ( t) is just a
number, which is positive if the object is moving in an agreed-upon positive direction, and
negative if it moves in the opposite of that direction. So you can think of that number, which
was called the velocity of the object, as having two components: a magnitude, indicated
by a nonnegative number, preceded by a direction, indicated by a plus or minus symbol
(representing motion in the positive direction or the negative direction, respectively), i.e.
f ( t) = a for some number a 0. Then a is the magnitude of the velocity (normally called
the speed of the object), and the represents the direction of the velocity (though the + is
usually omitted for the positive direction).
For motion along a straight line, i.e. in a 1-dimensional space, the velocities are also con-
tained in that 1-dimensional space, since they are just numbers. For general motion along a
curve in 2- or 3-dimensional space, however, velocity will need to be represented by a multi-
dimensional object which should have both a magnitude and a direction. A geometric object
which has those features is an arrow, which in elementary geometry is called a directed line
segment. This is the motivation for how we will define a vector.
Definition 1.1. A (nonzero) vector is a directed line segment drawn from a point P (called
its initial point) to a point Q (called its terminal point), with P and Q being distinct

points. The vector is denoted by PQ . Its magnitude is the length of the line segment,

denoted by PQ , and its direction is the same as that of the directed line segment. The
zero vector is just a point, and it is denoted by 0.
To indicate the direction of a vector, we draw an arrow from its initial point to its terminal
point. We will often denote a vector by a single bold-faced letter (e.g. v) and use the terms
magnitude and length interchangeably. Note that our definition could apply to systems
with any number of dimensions (see Figure 1.1.4 (a)-(c)).
z
y R
S Q
Q Q R
P
S

PQ
v P R y
x
0
0
RS PQ RS v
P
S 0 R P Q x S x
(a) One dimension (b) Two dimensions (c) Three dimensions
Figure 1.1.4 Vectors in different dimensions

A few things need to be noted about the zero vector. Our motivation for what a vector is
included the notions of magnitude and direction. What is the magnitude of the zero vector?
We define it to be zero, i.e. k 0 k = 0. This agrees with the definition of the zero vector as just
a point, which has zero length. What about the direction of the zero vector? A single point
really has no well-defined direction. Notice that we were careful to only define the direction
of a nonzero vector, which is well-defined since the initial and terminal points are distinct.
Not everyone agrees on the direction of the zero vector. Some contend that the zero vector
has arbitrary direction (i.e. can take any direction), some say that it has indeterminate
direction (i.e. the direction can not be determined), while others say that it has no direction.
Our definition of the zero vector, however, does not require it to have a direction, and we will
leave it at that.2
Now that we know what a vector is, we need a way of determining when two vectors are
equal. This leads us to the following definition.
Definition 1.2. Two nonzero vectors are equal if they have the same magnitude and the
same direction. Any vector with zero magnitude is equal to the zero vector.
By this definition, vectors with the same magnitude and direction but with different initial
points wouldpbe equal. For example, in Figure 1.1.5 the vectors u, v and w all have the same
magnitude 5 (by the Pythagorean Theorem). And we see that u and w are parallel, since
they lie on lines having the same slope 21 , and they point in the same direction. So u = w,
even though they have different initial points. We also see that v is parallel to u but points
in the opposite direction. So u 6= v.
y
u
4
3
2 v
w
1
x
0 1 2 3 4
Figure 1.1.5
So we can see that there are an infinite number of vectors for a given magnitude and
direction, those vectors all being equal and differing only by their initial and terminal points.
Is there a single vector which we can choose to represent all those equal vectors? The answer
is yes, and is suggested by the vector w in Figure 1.1.5.
2 In the subject of linear algebra there is a more abstract way of defining a vector where the concept of direction
is not really used. See A NTON and R ORRES.

1.1 Introduction 5
Unless otherwise indicated, when speaking of the vector with a given magnitude and
direction, we will mean the one whose initial point is at the origin of the coordinate
system.
Thinking of vectors as starting from the origin provides a way of dealing with vectors in
a standard way, since every coordinate system has an origin. But there will be times when
it is convenient to consider a different initial point for a vector (for example, when adding
vectors, which we will do in the next section).
Another advantage of using the origin as the initial point is that it provides an easy cor-
respondence between a vector and its terminal point.
Example 1.1. Let v be the vector in R3 whose initial point is at the origin and whose ter-
minal point is (3, 4, 5). Though the point (3, 4, 5) and the vector v are different objects, it is
convenient to write v = (3, 4, 5). When doing this, it is understood that the initial point of v
is at the origin (0, 0, 0) and the terminal point is (3, 4, 5).
z z
P(3, 4, 5) v = (3, 4, 5)
y y
0 0
x x
(a) The point (3,4,5) (b) The vector (3,4,5)
Figure 1.1.6 Correspondence between points and vectors
Unless otherwise stated, when we refer to vectors as v = (a, b) in R2 or v = (a, b, c) in R3 ,

we mean vectors in Cartesian coordinates starting at the origin. Also, we will write the zero
vector 0 in R2 and R3 as (0, 0) and (0, 0, 0), respectively.
The point-vector correspondence provides an easy way to check if two vectors are equal,
without having to determine their magnitude and direction. Similar to seeing if two points
are the same, you are now seeing if the terminal points of vectors starting at the origin
are the same. For each vector, find the (unique!) vector it equals whose initial point is
the origin. Then compare the coordinates of the terminal points of these new vectors: if
those coordinates are the same, then the original vectors are equal. To get the new vectors
starting at the origin, you translate each vector to start at the origin by subtracting the
coordinates of the original initial point from the original terminal point. The resulting point
will be the terminal point of the new vector whose initial point is the origin. Do this for
each original vector then compare.

Example 1.2. Consider the vectors PQ and RS in R3 , where P = (2, 1, 5),Q = (3, 5, 7), R =

(1, 3, 2) and S = (2, 1, 0). Does PQ = RS ?

Solution: The vector PQ is equal to the vector v with initial point (0, 0, 0) and terminal point
Q P = (3, 5, 7) (2, 1, 5) = (3 2, 5 1, 7 5) = (1, 4, 2).

Similarly, RS is equal to the vector w with initial point (0, 0, 0) and terminal point S R =
(2, 1, 0) (1, 3, 2) = (2 1, 1 (3), 0 (2)) = (1, 4, 2).

So PQ = v = (1, 4, 2) and RS = w = (1, 4, 2).

PQ = RS
z Q

PQ (3, 5, 7)
P
(2, 1, 5)

Translate PQ to v
v=w
(1, 4, 2)
0 y

Translate RS to w S
(2, 1, 0)
S
R R
(1, 3, 2) x
Figure 1.1.7
Recall the distance formula for points in the Euclidean plane:
For points P = ( x1 , y1 ), Q = ( x2 , y2 ) in R2 , the distance d between P and Q is:

q
d= ( x2 x1 )2 + ( y2 y1 )2 (1.1)
By this formula, we have the following result:

For a vector PQ in R2 with initial point P = ( x1 , y1 ) and terminal point

Q = ( x2 , y2 ), the magnitude of PQ is:
q
PQ = ( x2 x1 )2 + ( y2 y1 )2 (1.2)
1.1 Introduction 7
Finding the magnitude of a vector v = (a, b) in R2 is a special case of formula (1.2) with
P = (0, 0) and Q = (a, b) :
For a vector v = (a, b) in R2 , the magnitude of v is:

p
kvk = a2 + b 2 (1.3)
To calculate the magnitude of vectors in R3 , we need a distance formula for points in

Euclidean space (we will postpone the proof until the next section):
Theorem 1.1. The distance d between points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) in R3 is:

q
d= ( x2 x1 )2 + ( y2 y1 )2 + ( z2 z1 )2 (1.4)
The proof will use the following result:
Theorem 1.2. For a vector v = (a, b, c) in R3 , the magnitude of v is:

p
kvk = a2 + b 2 + c 2 (1.5)
Proof: There are four cases to consider: p p

Case 1: a = b = c = 0. Then v = 0, so k v k = 0 = 02 + 02 + 02 = a2 + b2 + c2 .
Case 2: exactly two of a, b, c are 0. Without loss of generality, we assume that a = b = 0 and
c 6= 0 (the other two possibilities are handled in a similar p manner).
p Then v =p(0, 0, c), which
is a vector of length | c | along the z-axis. So k v k = | c| = c = 0 + 02 + c2 = a2 + b2 + c2 .
2 2
Case 3: exactly one of a, b, c is 0. Without loss of generality, we assume that a = 0, b 6= 0

and c 6= 0 (the other two possibilities are handled in a similar manner). Then vp= (0, b, c),
which is a vector 2 2
p p in the yz-plane, so by the Pythagorean Theorem we have k v k = b + c =
02 + b 2 + c 2 = a 2 + b 2 + c 2 .
Case 4: none of a, b, c are 0. Without loss of generality, we can as- z
Q(a, b, c)
sume that a, b, c are all positive (the other seven possibilities are
handled in a similar manner). Consider the points P = (0, 0, 0), v
c
Q = (a, b, c), R = (a, b, 0), and S = (a, 0, 0), as shown in Figure
1.1.8. Applying the Pythagorean Theorem to the right trian- y
gle PSR gives |PR |2 = a2 + b2 . A second application of the P
a 0
Pythagorean Theorem, this time to the right triangle PQR , S
q p b R
gives k v k = |PQ | = |PR |2 + |QR |2 = a2 + b2 + c2 . x
This proves the theorem. QED Figure 1.1.8
Example 1.3. Calculate the following:

2
(a) The magnitude of the vector PQ in R with P = (1, 2) and Q = (5, 5).
p p p p
Solution: By formula (1.2), PQ = (5 (1))2 + (5 2)2 = 36 + 9 = 45 = 3 5.

2
(b) The magnitude of the vector v = (8 p, 3) in R . p
2 2
Solution: By formula (1.3), k v k = 8 + 3 = 73.
2
(c) The distance between the points P = (2, 1, 4)
pand Q = (4, 2, 3) in R .
Solution: By formula
p p (1.4), the distance d = (4 2) + (2 (1)) + (3 4)2 =
2 2
4 + 9 + 49 = 62.
3
(d) The magnitude of the vector v = (5 p, 8, 2) in R . p p
Solution: By formula (1.5), k v k = 52 + 82 + (2)2 = 25 + 64 + 4 = 93.
Exercises
A
1. Calculate the magnitudes of the following vectors:
(a) v = (2, 1) (b) v = (2, 1, 0) (c) v = (3, 2, 2) (d) v = (0, 0, 1) (e) v = (6, 4, 4)

2. For the points P = (1, 1, 1), Q = (2, 2, 2), R = (2, 0, 1), S = (3, 1, 2), does PQ = RS ?

3. For the points P = (0, 0, 0), Q = (1, 3, 2), R = (1, 0, 1), S = (2, 3, 4), does PQ = RS ?
B
4. Let v = (1, 0, 0) and w = (a, 0, 0) be vectors in R3 . Show that k w k = | a | k v k.
5. Let v = (a, b, c) and w = (3a, 3 b, 3 c) be vectors in R3 . Show that k w k = 3 k v k.
C
z
6. Though we will see a simple proof of Theorem 1.1
in the next section, it is possible to prove it using Q(x2 , y2 , z2 )
methods similar to those in the proof of Theorem P(x1 , y1 , z1 )
1.2. Prove the special case of Theorem 1.1 where the R(x2 , y2 , z1 )
points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) satisfy the fol-
lowing conditions: y
0
x2 > x1 > 0, y2 > y1 > 0, and z2 > z1 > 0. S(x1 , y1 , 0)
(Hint: Think of Case 4 in the proof of Theorem 1.2,
T(x2 , y2 , 0)
and consider Figure 1.1.9.) x U(x2 , y1 , 0)
Figure 1.1.9
1.2 Vector Algebra 9
1.2 Vector Algebra

Now that we know what vectors are, we can start to perform some of the usual algebraic
operations on them (e.g. addition, subtraction). Before doing that, we will introduce the
notion of a scalar.
Definition 1.3. A scalar is a quantity that can be represented by a single number.
For our purposes, scalars will always be real numbers.3 Examples of scalar quantities are
mass, electric charge, and speed (not velocity).4 We can now define scalar multiplication of
a vector.
Definition 1.4. For a scalar k and a nonzero vector v, the scalar multiple of v by k,
denoted by kv, is the vector whose magnitude is | k | k v k, points in the same direction as v if
k > 0, points in the opposite direction as v if k < 0, and is the zero vector 0 if k = 0. For the
zero vector 0, we define k0 = 0 for any scalar k.
Two vectors v and w are parallel (denoted by v w) if one is a scalar multiple of the other.
You can think of scalar multiplication of a vector as stretching or shrinking the vector, and
as flipping the vector in the opposite direction if the scalar is a negative number (see Figure
1.2.1).
v 2v 3v 0.5v v 2v
Figure 1.2.1
Recall that translating a nonzero vector means that the initial point of the vector is
changed but the magnitude and direction are preserved. We are now ready to define the
sum of two vectors.
Definition 1.5. The sum of vectors v and w, denoted by v + w, is obtained by translating
w so that its initial point is at the terminal point of v; the initial point of v + w is the initial
point of v, and its terminal point is the new terminal point of w.
3 The term scalar was invented by 19th century Irish mathematician, physicist and astronomer William Rowan
Hamilton, to convey the sense of something that could be represented by a point on a scale or graduated ruler.
The word vector comes from Latin, where it means carrier.
4 An alternate definition of scalars and vectors, used in physics, is that under certain types of coordinate trans-
formations (e.g. rotations), a quantity that is not affected is a scalar, while a quantity that is affected (in a
certain way) is a vector. See M ARION for details.
Intuitively, adding w to v means tacking on w to the end of v (see Figure 1.2.2).
v+w
w
w w
v v v
(a) Vectors v and w (b) Translate w to the end of v (c) The sum v + w
Figure 1.2.2 Adding vectors v and w
Notice that our definition is valid for the zero vector (which is just a point, and hence can
be translated), and so we see that v + 0 = v = 0 + v for any vector v. In particular, 0 + 0 = 0.
Also, it is easy to see that v + (v) = 0, as we would expect. In general, since the scalar
multiple v = 1 v is a well-defined vector, we can define vector subtraction as follows:
v w = v + (w). See Figure 1.2.3.
v v
w vw
w w
v
(a) Vectors v and w (b) Translate w to the end of v (c) The difference v w
Figure 1.2.3 Subtracting vectors v and w
Figure 1.2.4 shows the use of geometric proofs of various laws of vector algebra, that is,
it uses laws from elementary geometry to prove statements about vectors. For example, (a)
shows that v + w = w + v for any vectors v, w. And (c) shows how you can think of v w as
the vector that is tacked on to the end of w to add up to v.
v w
v+w
w+v vw
w w v vw w
v+w vw
v w v
(a) Add vectors (b) Subtract vectors (c) Combined add/subtract
Figure 1.2.4 Geometric vector algebra
Notice that we have temporarily abandoned the practice of starting vectors at the origin.
In fact, we have not even mentioned coordinates in this section so far. Since we will deal
mostly with Cartesian coordinates in this book, the following two theorems are useful for
performing vector algebra on vectors in R2 and R3 starting at the origin.
Theorem 1.3. Let v = (v1 , v2 ), w = (w1 , w2 ) be vectors in R2 , and let k be a scalar. Then
(a) kv = ( kv1 , kv2 )
(b) v + w = (v1 + w1 , v2 + w2 )
Proof: (a) Without loss of generality, we assume that v1 , v2 > 0 (the other possibilities are
handled in a similar manner). If k = 0 then kv = 0v = 0 = (0, 0) = (0v1 , 0v2 ) = ( kv1 , kv2 ), which
is what we needed to show. If k 6= 0, then ( kv1 , kv2 ) lies on a line with slope kv 2 v2
kv1 = v1 , which
is the same as the slope of the line on which v (and hence kv) lies, and ( kv1 , kv2 ) points in
the same directionqon that line asq kv. Also, by formula q (1.3) the magnitude of ( kv1 , kv2 ) is
p
( kv1 )2 + ( kv2 )2 = k2 v12 + k2 v22 = k2 (v12 + v22 ) = | k | v12 + v22 = | k | k v k. So kv and ( kv1 , kv2 )
have the same magnitude and direction. This proves (a).
(b) Without loss of generality, we assume that y

v2 + w 2
v1 , v2 , w1 , w2 > 0 (the other possibilities are han- v
dled in a similar manner). From Figure 1.2.5, we w2 w w2
w v+w
see that when translating w to start at the end of v2
w1
v, the new terminal point of w is (v1 + w1 , v2 + w2 ), v x
so by the definition of v + w this must be the ter- 0 w1 v1 v1 + w 1
minal point of v + w. This proves (b). QED
Figure 1.2.5
Theorem 1.4. Let v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 , let k be a scalar. Then

(a) kv = ( kv1 , kv2 , kv3 )
(b) v + w = (v1 + w1 , v2 + w2 , v3 + w3 )
The following theorem summarizes the basic laws of vector algebra.
Theorem 1.5. For any vectors u, v, w, and scalars k, l , we have

(a) v + w = w + v Commutative Law
(b) u + (v + w) = (u + v) + w Associative Law
(c) v + 0 = v = 0 + v Additive Identity
(d) v + (v) = 0 Additive Inverse
(e) k( l v) = ( kl )v Associative Law
(f) k(v + w) = kv + kw Distributive Law
(g) ( k + l )v = kv + l v Distributive Law
Proof: (a) We already presented a geometric proof of this in Figure 1.2.4(a).

(b) To illustrate the difference between analytic proofs and geometric proofs in vector alge-
bra, we will present both types here. For the analytic proof, we will use vectors in R3 (the
proof for R2 is similar).
Let u = ( u 1 , u 2 , u 3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 . Then
u + (v + w) = ( u 1 , u 2 , u 3 ) + ((v1 , v2 , v3 ) + (w1 , w2 , w3 ))
= ( u 1 , u 2 , u 3 ) + (v1 + w1 , v2 + w2 , v3 + w3 ) by Theorem 1.4(b)
= ( u 1 + (v1 + w1 ), u 2 + (v2 + w2 ), u 3 + (v3 + w3 )) by Theorem 1.4(b)
= (( u 1 + v1 ) + w1 , ( u 2 + v2 ) + w2 , ( u 3 + v3 ) + w3 ) by properties of real numbers
= ( u 1 + v1 , u 2 + v2 , u 3 + v3 ) + (w1 , w2 , w3 ) by Theorem 1.4(b)
= (u + v) + w
This completes the analytic proof of (b). Figure 1.2.6 provides the geometric proof.
u + (v + w) = (u + v) + w
v+w w
u u+v
v
Figure 1.2.6 Associative Law for vector addition
(c) We already discussed this on p.10.

(d) We already discussed this on p.10.
(e) We will prove this for a vector v = (v1 , v2 , v3 ) in R3 (the proof for R2 is similar):
k( l v) = k( lv1 , lv2 , lv3 ) by Theorem 1.4(a)

= ( klv1 , klv2 , klv3 ) by Theorem 1.4(a)
= ( kl )(v1 , v2 , v3 ) by Theorem 1.4(a)
= ( kl )v
(f) and (g): Left as exercises for the reader. QED
A unit vector is a vector with magnitude 1. Notice that for any nonzero vector v, the
vector k vv k is a unit vector which points in the same direction as v, since k v1 k > 0 and k vv k =
kvk
kvk = 1. Dividing a nonzero vector v by k v k is often called normalizing v.
There are specific unit vectors which we will often use, called the basis vectors:
i = (1, 0, 0), j = (0, 1, 0), and k = (0, 0, 1) in R3 ; i = (1, 0) and j = (0, 1) in R2 .
These are useful for several reasons: they are mutually perpendicular, since they lie on
distinct coordinate axes; they are all unit vectors: k i k = k j k = k k k = 1; every vector can
be written as a unique scalar combination of the basis vectors: v = (a, b) = a i + b j in R2 ,
v = (a, b, c) = a i + b j + c k in R3 . See Figure 1.2.7.
z z
2 v = (a, b, c)
y y 1
2 v = (a, b)
ck
k
y y
1 bj i 0 j 0
1 2 ai
j x 1
x
0 2 bj
i 1 2 0 ai x x
(a) R2 (b) v = a i + b j (c) R3 (d) v = a i + b j + c k
Figure 1.2.7 Basis vectors in different dimensions
When a vector v = (a, b, c) is written as v = a i + b j + c k, we say that v is in component

form, and that a, b, and c are the i, j, and k components, respectively, of v. We have:
v = v1 i + v2 j + v3 k, k a scalar = kv = kv1 i + kv2 j + kv3 k

v = v1 i + v2 j + v3 k, w = w1 i + w2 j + w3 k = v + w = (v1 + w1 )i + (v2 + w2 )j + (v3 + w3 )k
q
v = v1 i + v2 j + v3 k = k v k = v12 + v22 + v32
Example 1.4. Let v = (2, 1, 1) and w = (3, 4, 2) in R3 .
(a) Find v w.
Solution: v w = (2 3, 1 (4), 1 2) = (1, 5, 3)
(b) Find 3v + 2w.

Solution: 3v + 2w = (6, 3, 3) + (6, 8, 4) = (12, 5, 1)
(c) Write v and w in component form.

Solution: v = 2 i + j k, w = 3 i 4 j + 2 k
(d) Find the vector u such that u + v = w.

Solution: By Theorem 1.5, u = w v = (v w) = (1, 5, 3) = (1, 5, 3), by part(a).
(e) Find the vector u such that u + v + w = 0.

Solution: By Theorem 1.5, u = w v = (3, 4, 2) (2, 1, 1) = (5, 3, 1).
(f) Find the vector u such that 2u + i 2 j = k.

Solution: 2u = i + 2 j + k = u = 12 i + j + 12 k
v
(g) Find the unit vector kvk .
v p 1 p2 , p1 , p
1
Solution: kvk = (2, 1, 1) =
22 +12 +(1)2 6 6 6
We can now easily prove Theorem 1.1 from the previous section. The distance d between
two points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) in R3 is the same as the length of the vector w v,
where the vectors v and w are defined as v = ( x1 , y1 , z1 ) and p w = ( x2 , y2 , z2 ) (see Figure 1.2.8).
So since w v = ( x2 x1 , y2 y1 , z2 z1 ), then d = k w v k = ( x2 x1 )2 + ( y2 y1 )2 + ( z2 z1 )2 by
Theorem 1.2.
z P(x1 , y1 , z1 )
wv
v Q(x2 , y2 , z2 )
w y
0
x
Figure 1.2.8 Proof of Theorem 1.2: d = k w v k
Exercises
A
1. Let v = (1, 5, 2) and w = (3, 1, 1).

(a) Find v w. (b) Find v + w. (c) Find v
kvk . (d) Find 21 (v w) .

(e) Find 1 (v + w) .
2 (f) Find 2 v + 4 w. (g) Find v 2 w.
(h) Find the vector u such that u + v + w = i.
(i) Find the vector u such that u + v + w = 2 j + k.
(j) Is there a scalar m such that m(v + 2 w) = k? If so, find it.
2. For the vectors v and w from Exercise 1, is k v w k = k v k k w k? If not, which quantity

is larger?
3. For the vectors v and w from Exercise 1, is k v + w k = k v k + k w k? If not, which quantity

is larger?
B
4. Prove Theorem 1.5(f) for R3 . 5. Prove Theorem 1.5(g) for R3 .
C
6. We know that every vector in R3 can be written as a scalar combination of the vectors i,
j, and k. Can every vector in R3 be written as a scalar combination of just i and j, i.e. for
any vector v in R3 , are there scalars m, n such that v = m i + n j? Justify your answer.
1.3 Dot Product 15
1.3 Dot Product

You may have noticed that while we did define multiplication of a vector by a scalar in the
previous section on vector algebra, we did not define multiplication of a vector by a vector.
We will now see one type of multiplication of vectors, called the dot product.
Definition 1.6. Let v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) be vectors in R3 .

The dot product of v and w, denoted by v w, is given by:
v w = v1 w1 + v2 w2 + v3 w3 (1.6)
Similarly, for vectors v = (v1 , v2 ) and w = (w1 , w2 ) in R2 , the dot product is:
v w = v1 w1 + v2 w2 (1.7)
Notice that the dot product of two vectors is a scalar, not a vector. So the associative law
that holds for multiplication of numbers and for addition of vectors (see Theorem 1.5(b),(e)),
does not hold for the dot product of vectors. Why? Because for vectors u, v, w, the dot
product u v is a scalar, and so (u v) w is not defined since the left side of that dot product
(the part in parentheses) is a scalar and not a vector.
For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form, the dot product
is still v w = v1 w1 + v2 w2 + v3 w3 .
Also notice that we defined the dot product in an analytic way, i.e. by referencing vector
coordinates. There is a geometric way of defining the dot product, which we will now develop
as a consequence of the analytic definition.
Definition 1.7. The angle between two nonzero vectors with the same initial point is the
smallest angle between them.
We do not define the angle between the zero vector and any other vector. Any two nonzero
vectors with the same initial point have two angles between them: and 360 . We will
always choose the smallest nonnegative angle between them, so that 0 180 . See
Figure 1.3.1.
360

360 360
(a) 0 < < 180 (b) = 180 (c) = 0
Figure 1.3.1 Angle between vectors
We can now take a more geometric view of the dot product by establishing a relationship
between the dot product of two vectors and the angle between them.
Theorem 1.6. Let v, w be nonzero vectors, and let be the angle between them. Then
v w
cos = (1.8)
kvk kwk
Proof: We will prove the theorem for vectors in R3 (the proof for R2 is similar). Let v =
(v1 , v2 , v3 ) and w = (w1 , w2 , w3 ). By the Law of Cosines (see Figure 1.3.2), we have
k v w k2 = k v k2 + k w k2 2 k v k k w k cos (1.9)
(note that equation (1.9) holds even for the degenerate cases = 0 and 180 ).
z
vw
v

w y
0
x
Figure 1.3.2
Since v w = (v1 w1 , v2 w2 , v3 w3 ), expanding k v w k2 in equation (1.9) gives
k v k2 + k w k2 2 k v k k w k cos = (v1 w1 )2 + (v2 w2 )2 + (v3 w3 )2

= (v12 2v1 w1 + w12 ) + (v22 2v2 w2 + w22 ) + (v32 2v3 w3 + w32 )
= (v12 + v22 + v32 ) + (w12 + w22 + w32 ) 2(v1 w1 + v2 w2 + v3 w3 )
= k v k2 + k w k2 2(v w) , so
2 k v k k w k cos = 2(v w) , so since v 6= 0 and w 6= 0 then
v w
cos = , since k v k > 0 and k w k > 0. QED
kvk kwk
Example 1.5. Find the angle between the vectors v = (2, 1, 1) and w = (3, 4, 1).
p p
Solution: Since v w = (2)(3) + (1)(4) + (1)(1) = 1, k v k = 6, and k w k = 26, then
v w 1 1
cos = = p p = p 0.08 = = 85.41
kvk kwk 6 26 2 39
Two nonzero vectors are perpendicular if the angle between them is 90 . Since cos 90 =
0, we have the following important corollary to Theorem 1.6:
Corollary 1.7. Two nonzero vectors v and w are perpendicular if and only if v w = 0.
We will write v w to indicate that v and w are perpendicular.

1.3 Dot Product 17
Since cos > 0 for 0 < 90 and cos < 0 for 90 < 180 , we also have:
Corollary 1.8. If is the angle between nonzero vectors v and w, then

> 0 for 0 < 90

v w is 0 for = 90

< 0 for 90 < 180
By Corollary 1.8, the dot product can be thought of as a way of telling if the angle be-
tween two vectors is acute, obtuse, or a right angle, depending on whether the dot product
is positive, negative, or zero, respectively. See Figure 1.3.3.
w w w
90 < 180
= 90

0 < 90
v v v
(a) v w > 0 (b) v w < 0 (c) v w = 0
Figure 1.3.3 Sign of the dot product & angle between vectors
Example 1.6. Are the vectors v = (1, 5, 2) and w = (3, 1, 1) perpendicular?

Solution: Yes, v w since v w = (1)(3) + (5)(1) + (2)(1) = 0.
The following theorem summarizes the basic properties of the dot product.
Theorem 1.9. For any vectors u, v, w, and scalar k, we have

(a) v w = w v Commutative Law
(b) ( kv) w = v ( kw) = k(v w) Associative Law
(c) v 0 = 0 = 0 v
(d) u (v + w) = u v + u w Distributive Law
(e) (u + v) w = u w + v w Distributive Law
(f) | v w | k v k k w k Cauchy-Schwarz Inequality5
Proof: The proofs of parts (a)-(e) are straightforward applications of the definition of the
dot product, and are left to the reader as exercises. We will prove part (f).
(f) If either v = 0 or w = 0, then v w = 0 by part (c), and so the inequality holds trivially. So
assume that v and w are nonzero vectors. Then by Theorem 1.6,
v w = cos k v k k w k , so
| v w | = | cos | k v k k w k , so
| v w | k v k k w k since | cos | 1. QED
5 Also known as the Cauchy-Schwarz-Buniakovski Inequality.

Using Theorem 1.9, we see that if u v = 0 and u w = 0, then u ( kv + l w) = k(u v)+ l (u w) =

k(0) + l (0) = 0 for all scalars k, l . Thus, we have the following fact:
If u v and u w, then u ( kv + l w) for all scalars k, l .
For vectors v and w, the collection of all scalar combinations kv + l w is called the span
of v and w. If nonzero vectors v and w are parallel, then their span is a line; if they are
not parallel, then their span is a plane. So what we showed above is that a vector which is
perpendicular to two other vectors is also perpendicular to their span.
The dot product can be used to derive properties of the magnitudes of vectors, the most
important of which is the Triangle Inequality, as given in the following theorem:
Theorem 1.10. For any vectors v, w, we have

(a) k v k2 = v v
(b) k v + w k k v k + k w k Triangle Inequality
(c) k v w k k v k k w k
Proof: (a) Left as an exercise for the reader.

(b) By part (a) and Theorem 1.9, we have
k v + w k2 = (v + w) (v + w) = v v + v w + w v + w w
= k v k2 + 2(v w) + k w k2 , so since a | a | for any real number a, we have
k v k2 + 2 | v w | + k w k2 , so by Theorem 1.9(f) we have
k v k2 + 2 k v k k w k + k w k2 = (k v k + k w k)2 and so
k v + w k k v k + k w k after taking square roots of both sides, which proves (b).
(c) Since v = w + (v w), then k v k = k w + (v w) k k w k+k v w k by the Triangle Inequality,

so subtracting k w k from both sides gives k v k k w k k v w k. QED
The Triangle Inequality gets its name from the fact that in any triangle, v+w
no one side is longer than the sum of the lengths of the other two sides (see
w
Figure 1.3.4). Another way of saying this is with the familiar statement the v
shortest distance between two points is a straight line.
Figure 1.3.4
Exercises
A
1. Let v = (5, 1, 2) and w = (4, 4, 3). Calculate v w.
2. Let v = 3 i 2 j k and w = 6 i + 4 j + 2 k. Calculate v w.
For Exercises 3-8, find the angle between the vectors v and w.
1.3 Dot Product 19
3. v = (5, 1, 2), w = (4, 4, 3) 4. v = (7, 2, 10), w = (2, 6, 4)
5. v = (2, 1, 4), w = (1, 2, 0) 6. v = (4, 2, 1), w = (8, 4, 2)
7. v = i + 2 j + k, w = 3 i + 6 j + 3 k 8. v = i, w = 3 i + 2 j + 4k
9. Let v = (8, 4, 3) and w = (2, 1, 4). Is v w? Justify your answer.
10. Let v = (6, 0, 4) and w = (0, 2, 1). Is v w? Justify your answer.
11. For v, w from Exercise 5, verify the Cauchy-Schwarz Inequality | v w | k v k k w k.
12. For v, w from Exercise 6, verify the Cauchy-Schwarz Inequality | v w | k v k k w k.
13. For v, w from Exercise 5, verify the Triangle Inequality k v + w k k v k + k w k.
14. For v, w from Exercise 6, verify the Triangle Inequality k v + w k k v k + k w k.

B
Note: Consider only vectors in R3 for Exercises 15-25.
15. Prove Theorem 1.9(a). 16. Prove Theorem 1.9(b).
17. Prove Theorem 1.9(c). 18. Prove Theorem 1.9(d).
19. Prove Theorem 1.9(e). 20. Prove Theorem 1.10(a).
21. Prove or give a counterexample: If u v = u w, then v = w.

C
22. Prove or give a counterexample: If v w = 0 for all v, then w = 0.
23. Prove or give a counterexample: If u v = u w for all u, then v = w.

24. Prove that k v k k w k k v w k for all v, w.
v
25. For nonzero vectors v and w, the projection of v onto w (some-
times written as pro j w v) is the vector u along the same line L
as w whose terminal point is obtained by dropping a perpendic- L
ular line from the terminal point of v to L (see Figure 1.3.5). w
u
Show that
|v w| Figure 1.3.5
kuk = .
kwk
(Hint: Consider the angle between v and w.)
26. Let , , and be the angles between a nonzero vector v in R3 and the vectors i, j, and
k, respectively. Show that cos2 + cos2 + cos2 = 1.
(Note: , , are often called the direction angles of v, and cos , cos , cos are called
the direction cosines.)
1.4 Cross Product

In Section 1.3 we defined the dot product, which gave a way of multiplying two vectors. The
resulting product, however, was a scalar, not a vector. In this section we will define a product
of two vectors that does result in another vector. This product, called the cross product, is
only defined for vectors in R3 . The definition may appear strange and lacking motivation,
but we will see the geometric basis for it shortly.
Definition 1.8. Let v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) be vectors in R3 . The cross product
of v and w, denoted by v w, is the vector in R3 given by:
v w = (v2 w3 v3 w2 , v3 w1 v1 w3 , v1 w2 v2 w1 ) (1.10)
z
Example 1.7. Find i j.
1 k = i j
Solution: Since i = (1, 0, 0) and j = (0, 1, 0), then
y
i j = ((0)(0) (0)(1), (0)(0) (1)(0), (1)(1) (0)(0)) i 0 j 1
= (0, 0, 1) 1
x
=k
Figure 1.4.1
Similarly it can be shown that j k = i and k i = j.
In the above example, the cross product of the given vectors was perpendicular to both
those vectors. It turns out that this will always be the case.
Theorem 1.11. If the cross product v w of two nonzero vectors v and w is also a nonzero
vector, then it is perpendicular to both v and w.
Proof: We will show that (v w) v = 0:
(v w) v = (v2 w3 v3 w2 , v3 w1 v1 w3 , v1 w2 v2 w1 ) (v1 , v2 , v3 )
= v2 w3 v1 v3 w2 v1 + v3 w1 v2 v1 w3 v2 + v1 w2 v3 v2 w1 v3
= v1 v2 w3 v1 v2 w3 + w1 v2 v3 w1 v2 v3 + v1 w2 v3 v1 w2 v3
= 0 , after rearranging the terms.
v w v by Corollary 1.7.
The proof that v w w is similar. QED
As a consequence of the above theorem and Theorem 1.9, we have the following:
Corollary 1.12. If the cross product v w of two nonzero vectors v and w is also a nonzero
vector, then it is perpendicular to the span of v and w.
1.4 Cross Product 21
The span of any two nonzero, nonparallel vectors v, w in R3 is a plane P , so the above
corollary shows that v w is perpendicular to that plane. As shown in Figure 1.4.2, there
are two possible directions for v w, one the opposite of the other. It turns out (see Appendix
B) that the direction of v w is given by the right-hand rule, that is, the vectors v, w, v w
form a right-handed system. Recall from Section 1.1 that this means that you can point your
thumb upwards in the direction of v w while rotating v towards w with the remaining four
fingers.
z
v w
v
y
0
w
P
x v w
Figure 1.4.2 Direction of v w
We will now derive a formula for the magnitude of v w, for nonzero vectors v, w:
k v w k2 = (v2 w3 v3 w2 )2 + (v3 w1 v1 w3 )2 + (v1 w2 v2 w1 )2

= v22 w32 2v2 w2 v3 w3 + v32 w22 + v32 w12 2v1 w1 v3 w3 + v12 w32 + v12 w22 2v1 w1 v2 w2 + v22 w12
= v12 (w22 + w32 ) + v22 (w12 + w32 ) + v32 (w12 + w22 ) 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 )
and now adding and subtracting v12 w12 , v22 w22 , and v32 w32 on the right side gives
= v12 (w12 + w22 + w32 ) + v22 (w12 + w22 + w32 ) + v32 (w12 + w22 + w32 )
(v12 w12 + v22 w22 + v32 w32 + 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 ))
= (v12 + v22 + v32 )(w12 + w22 + w32 )
((v1 w1 )2 + (v2 w2 )2 + (v3 w3 )2 + 2(v1 w1 )(v2 w2 ) + 2(v1 w1 )(v3 w3 ) + 2(v2 w2 )(v3 w3 ))
so using (a + b + c)2 = a2 + b2 + c2 + 2ab + 2ac + 2 bc for the subtracted term gives
= (v12 + v22 + v32 )(w12 + w22 + w32 ) (v1 w1 + v2 w2 + v3 w3 )2

= k v k2 k w k2 (v w)2
(v w)2
= k v k2 k w k2 1 , since k v k > 0 and k w k > 0, so by Theorem 1.6
k v k2 k w k2
= k v k2 k w k2 (1 cos2 ) , where is the angle between v and w, so
k v w k2 = k v k2 k w k2 sin2 , and since 0 180 , then sin 0, so we have:
If is the angle between nonzero vectors v and w in R3 , then
k v w k = k v k k w k sin (1.11)
It may seem strange to bother with the above formula, when the magnitude of the cross
product can be calculated directly, like for any other vector. The formula is more useful for
its applications in geometry, as in the following example.
Example 1.8. Let PQR and PQRS be a triangle and parallelogram, respectively, as shown
in Figure 1.4.3.
P S P S
h w h

Q b R Q v R
Figure 1.4.3
Think of the triangle as existing in R3 , and identify the sides QR and QP with vectors v
and w, respectively, in R3 . Let be the angle between v and w. The area A PQR of PQR is
1
2 bh, where b is the base of the triangle and h is the height. So we see that
b = kvk and h = k w k sin
1
A PQR = k v k k w k sin
2
1
= kv wk
2
So since the area A PQRS of the parallelogram PQRS is twice the area of the triangle PQR ,
then
A PQRS = k v k k w k sin
By the discussion in Example 1.8, we have proved the following theorem:

Theorem 1.13. Area of triangles and parallelograms
(a) The area A of a triangle with adjacent sides v, w (as vectors in R3 ) is:
1
A= kv wk
2
(b) The area A of a parallelogram with adjacent sides v, w (as vectors in R3 ) is:
A = kv wk
It may seem at first glance that since the formulas derived in Example 1.8 were for the
adjacent sides QP and QR only, then the more general statements in Theorem 1.13 that the
formulas hold for any adjacent sides are not justified. We would get a different formula for
the area if we had picked PQ and PR as the adjacent sides, but it can be shown (see Exercise
26) that the different formulas would yield the same value, so the choice of adjacent sides
indeed does not matter, and Theorem 1.13 is valid.
Theorem 1.13 makes it simpler to calculate the area of a triangle in 3-dimensional space
than by using traditional geometric methods.
Example 1.9. Calculate the area of the triangle PQR , where P = (2, 4, 7), Q = (3, 7, 18),
and R = (5, 12, 8).
z
Solution: Let v = PQ and w = PR , as in Figure 1.4.4. Then
Q(3, 7, 18)
v = (3, 7, 18)(2, 4, 7) = (1, 3, 25) and w = (5, 12, 8)(2, 4, 7) =
(7, 8, 15), so the area A of the triangle PQR is R(5, 12, 8)
1 1 v
A= k v w k = k (1, 3, 25) (7, 8, 15) k w y
2 2 0
1
= ((3)(15) (25)(8), (25)(7) (1)(15), (1)(8) (3)(7)) x
P(2, 4, 7)
2
1
= (155, 190, 29) Figure 1.4.4
2
1p 1p
= (155)2 + (190)2 + 292 = 60966
2 2
A 123.46
Example 1.10. Calculate the area of the parallelogram PQRS , where P = (1, 1), Q = (2, 3),
R = (5, 4), and S = (4, 2).
y
Solution: Let v = SP and w = SR , as in Figure 1.4.5. Then
v = (1, 1) (4, 2) = (3, 1) and w = (5, 4) (4, 2) = (1, 2). But 4 R
2
these are vectors in R , and the cross product is only defined Q
3 2 3 w
for vectors in R . However, R can be thought of as the subset
of R3 such that the z-coordinate is always 0. So we can write 2
S
v = (3, 1, 0) and w = (1, 2, 0). Then the area A of PQRS is v
1
P
x
A = k v w k = (3, 1, 0) (1, 2, 0)
0 1 2 3 4 5
= ((1)(0) (0)(2), (0)(1) (3)(0), (3)(2) (1)(1))
Figure 1.4.5
= (0, 0, 5)
A=5
The following theorem summarizes the basic properties of the cross product.
Theorem 1.14. For any vectors u, v, w in R3 , and scalar k, we have

(a) v w = w v Anticommutative Law
(b) u (v + w) = u v + u w Distributive Law
(c) (u + v) w = u w + v w Distributive Law
(d) ( kv) w = v ( kw) = k(v w) Associative Law
(e) v 0 = 0 = 0 v
(f) v v = 0
(g) v w = 0 if and only if v w
Proof: The proofs of properties (b)-(f) are straightforward. We will prove parts (a) and (g)
and leave the rest to the reader as exercises.
(a) By the definition of the cross product and scalar multipli- z
v w
cation, we have: v
v w = (v2 w3 v3 w2 , v3 w1 v1 w3 , v1 w2 v2 w1 ) y
= (v3 w2 v2 w3 , v1 w3 v3 w1 , v2 w1 v1 w2 ) 0
w
= (w2 v3 w3 v2 , w3 v1 w1 v3 , w1 v2 w2 v1 )
= w v x w v
Figure 1.4.6
Note that this says that v w and w v have the same mag-
nitude but opposite direction (see Figure 1.4.6).
(g) If either v or w is 0 then v w = 0 by part (e), and either v = 0 = 0w or w = 0 = 0v, so v

and w are scalar multiples, i.e. they are parallel.
If both v and w are nonzero, and is the angle between them, then by formula (1.11),
v w = 0 if and only if k v k k w k sin = 0, which is true if and only if sin = 0 (since k v k > 0
and k w k > 0). So since 0 180 , then sin = 0 if and only if = 0 or 180 . But the
angle between v and w is 0 or 180 if and only if v w. QED
Example 1.11. Adding to Example 1.7, we have
i j = k j k = i k i = j
j i = k k j = i i k = j
i i = j j = k k = 0
Recall from geometry that a parallelepiped is a 3-dimensional solid with 6 faces, all of
which are parallelograms.6
6 An equivalent definition of a parallelepiped is: the collection of all scalar combinations k v + k v + k v of
1 1 2 2 3 3
some vectors v1 , v2 , v3 in R3 , where 0 k 1 , k 2 , k 3 1.
Example 1.12. Volume of a parallelepiped: Let the vectors u, v, w in R3 represent adjacent

sides of a parallelepiped P , with u, v, w forming a right-handed system, as in Figure 1.4.7.
Show that the volume of P is the scalar triple product u (v w).
Solution: Recall that the volume vol(P ) of a par- v w
allelepiped P is the area A of the base parallel-
ogram times the height h. By Theorem 1.13(b),
the area A of the base parallelogram is k v w k. u
And we can see that since v w is perpendicular h
to the base parallelogram determined by v and
w
w, then the height h is k u k cos , where is the
angle between u and v w. By Theorem 1.6 we
v
know that
u (v w)
cos = . Hence, Figure 1.4.7 Parallelepiped P
kuk kv wk
vol(P ) = A h
k u k u (v w)
= kv wk
kuk kv wk
= u (v w)
In Example 1.12 the height h of the parallelepiped is k u k cos , and not k u k cos , be-
cause the vector u is on the same side of the base parallelograms plane as the vector v w
(so that cos > 0). Since the volume is the same no matter which base and height we use,
then repeating the same steps using the base determined by u and v (since w is on the same
side of that bases plane as u v), the volume is w (u v). Repeating this with the base
determined by w and u, we have the following result:
For any vectors u, v, w in R3 ,
u (v w) = w (u v) = v (w u) (1.12)
(Note that the equalities hold trivially if any of the vectors are 0.)
Since v w = w v for any vectors v, w in R3 , then picking the wrong order for the three
adjacent sides in the scalar triple product in formula (1.12) will give you the negative of the
volume of the parallelepiped. So taking the absolute value of the scalar triple product for
any order of the three adjacent sides will always give the volume:
Theorem 1.15. If vectors u, v, w in R3 represent any three adjacent sides of a paral-

lelepiped, then the volume of the parallelepiped is | u (v w) |.
Another type of triple product is the vector triple product u (v w). The proof of the
following theorem is left as an exercise for the reader:
Theorem 1.16. For any vectors u, v, w in R3 ,
u (v w) = (u w)v (u v)w (1.13)
An examination of the formula in Theorem 1.16 gives some idea of the geometry of the
vector triple product. By the right side of formula (1.13), we see that u (v w) is a scalar
combination of v and w, and hence lies in the plane containing v and w (i.e. u (v w), v
and w are coplanar). This makes sense since, by Theorem 1.11, u (v w) is perpendicular
to both u and v w. In particular, being perpendicular to v w means that u (v w) lies
in the plane containing v and w, since that plane is itself perpendicular to v w. But then
how is u (v w) also perpendicular to u, which could be any vector? The following example
may help to see how this works.
Example 1.13. Find u (v w) for u = (1, 2, 4), v = (2, 2, 0), w = (1, 3, 0).
Solution: Since u v = 6 and u w = 7, then
u (v w) = (u w)v (u v)w
= 7 (2, 2, 0) 6 (1, 3, 0) = (14, 14, 0) (6, 18, 0)
= (8, 4, 0)
Note that v and w lie in the x y-plane, and that u (v w) also lies in that plane. Also,
u (v w) is perpendicular to both u and v w = (0, 0, 4) (see Figure 1.4.8).
z
vw u
y
0 w
v
u (v w)
x
Figure 1.4.8
For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form, the cross product

is written as: v w = (v2 w3 v3 w2 )i + (v3 w1 v1 w3 )j + (v1 w2 v2 w1 )k. It is often easier to use the
component form for the cross product, because it can be represented as a determinant. We
will not go too deeply into the theory of determinants7 ; we will just cover what is essential
for our purposes.
7 See A NTON and R ORRES for a fuller development.
A 2 2 matrix is an array of two rows and two columns of scalars, written as

a b a b
or
c d c d
where a, b, c, d are scalars. The determinant of such a matrix, written as

a b
or det a b ,
c d c d
is the scalar defined by the following formula:

a b

c = ad bc
d
It may help to remember this formula as being the product of the scalars on the downward
diagonal minus the product of the scalars on the upward diagonal.
Example 1.14.
1 2
3 4 = (1)(4) (2)(3) = 4 6 = 2

A 3 3 matrix is an array of three rows and three columns of scalars, written as

a1 a2 a3 a1 a2 a3
b1 b2 b3 or b1 b2 b3 ,
c1 c2 c3 c1 c2 c3
and its determinant is given by the formula:

a 1 a2 a 3
b2 b 3 b1 b 3 b1 b 2
b1 b2 b 3 = a 1
a 2 + a 3 (1.14)

c c2 c3 c1 c3 c1 c2
1 c2 c3
One way to remember the above formula is the following: multiply each scalar in the first
row by the determinant of the 2 2 matrix that remains after removing the row and column
that contain that scalar, then sum those products up, putting alternating plus and minus
signs in front of each (starting with a plus).
Example 1.15.

1 0 2
4 1 3 = 1 1 3 0 4 3 + 2 4 1

= 1(2 0) 0(8 3) + 2(0 + 1) = 0
0 2 1 2 1 0
1 0 2
We defined the determinant as a scalar, derived from algebraic operations on scalar en-
tries in a matrix. However, if we put three vectors in the first row of a 3 3 matrix, then
the definition still makes sense, since we would be performing scalar multiplication on those
three vectors (they would be multiplied by the 22 scalar determinants as before). This gives
us a determinant that is now a vector, and lets us write the cross product of v = v1 i + v2 j + v3 k
and w = w1 i + w2 j + w3 k as a determinant:

i j k
v2 v3 v1 v3 v1 v2
v w = v1 v2 v3 =

i w w j + w w k

w w w w2 w3 1 3 1 2
1 2 3
= (v2 w3 v3 w2 )i + (v3 w1 v1 w3 )j + (v1 w2 v2 w1 )k
Example 1.16. Let v = 4 i j + 3 k and w = i + 2 k. Then

i j k
1 3 4 3 4 1
v w = 4 1 3 =
i j + k = 2 i 5 j + k
1 0 2 1 2 1 0
0 2
The scalar triple product can also be written as a determinant. In fact, by Example 1.12,
the following theorem provides an alternate definition of the determinant of a 3 3 matrix
as the volume of a parallelepiped whose adjacent sides are the rows of the matrix and form
a right-handed system (a left-handed system would give the negative volume).
Theorem 1.17. For any vectors u = ( u 1 , u 2 , u 3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) in R3 :

u1 u2 u 3

u (v w) = v1 v2 v3 (1.15)
w w2 w3
1
Example 1.17. Find the volume of the parallelepiped with adjacent sides u = (2, 1, 3), v =
(1, 3, 2), w = (1, 1, 2) (see Figure 1.4.9).
Solution: By Theorem 1.15, the volume vol(P ) of the parallelepiped z
P is the absolute value of the scalar triple product of the three
adjacent sides (in any order). By Theorem 1.17, v
u
2 1 3
y
u (v w) = 1 3 2
1 1 2 0
x w
3 2 1 2 1 3
= 2 1 + 3

1 2 1 2 1 1 Figure 1.4.9 P
= 2(8) 1(0) + 3(4) = 28, so
vol(P ) = |28 | = 28.
Interchanging the dot and cross products can be useful in proving vector identities:

u w u z
Example 1.18. Prove: (u v) (w z) =
for all vectors u, v, w, z in R3 .
v w v z
Solution: Let x = u v. Then
(u v) (w z) = x (w z)
= w (z x) (by formula (1.12))
= w (z (u v))
= w ((z v)u (z u)v) (by Theorem 1.16)
= (z v)(w u) (z u)(w v)
= (u w)(v z) (u z)(v w) (by commutativity of the dot product).

u w u z
=
v w v z
Exercises
A
For Exercises 1-6, calculate v w.
1. v = (5, 1, 2), w = (4, 4, 3) 2. v = (7, 2, 10), w = (2, 6, 4)
3. v = (2, 1, 4), w = (1, 2, 0) 4. v = (1, 3, 2), w = (7, 2, 10)
5. v = i + 2 j + k, w = 3 i + 6 j + 3 k 6. v = i, w = 3 i + 2 j + 4k
For Exercises 7-8, calculate the area of the triangle PQR .
7. P = (5, 1, 2), Q = (4, 4, 3), R = (2, 4, 0) 8. P = (4, 0, 2), Q = (2, 1, 5), R = (1, 0, 1)
For Exercises 9-10, calculate the area of the parallelogram PQRS .
9. P = (2, 1, 3), Q = (1, 4, 5), R = (2, 5, 3), S = (3, 2, 1)
10. P = (2, 2), Q = (1, 4), R = (6, 6), S = (3, 0)
For Exercises 11-12, find the volume of the parallelepiped with adjacent sides u, v, w.
11. u = (1, 1, 3), v = (2, 1, 4), w = (5, 1, 2) 12. u = (1, 3, 2), v = (7, 2, 10), w = (1, 0, 1)
For Exercises 13-14, calculate u (v w) and u (v w).
13. u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2) 14. u = (1, 0, 2), v = (1, 0, 3), w = (2, 0, 2)
15. Calculate (u v) (w z) for u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2), z = (2, 1, 4).
B
16. If v and w are unit vectors in R3 , under what condition(s) would v w also be a unit
vector in R3 ? Justify your answer.
17. Show that if v w = 0 for all w in R3 , then v = 0.
18. Prove Theorem 1.14(b). 19. Prove Theorem 1.14(c).
20. Prove Theorem 1.14(d). 21. Prove Theorem 1.14(e).
22. Prove Theorem 1.14(f). 23. Prove Theorem 1.16.
24. Prove Theorem 1.17. (Hint: Expand both sides of the equation.)
25. Prove the following for all vectors v, w in R3 :

(a) k v w k2 + | v w |2 = k v k2 k w k2
(b) If v w = 0 and v w = 0, then v = 0 or w = 0.
C
26. Prove that in Example 1.8 the formula for the area of the triangle PQR yields the
same value no matter which two adjacent sides are chosen. To do this, show that 21 k u
(w) k = 12 k v w k, where u = PR , w = PQ , and v = QR , w = QP as before. Similarly,
show that 12 k (u) (v) k = 21 k v w k, where u = RP and v = RQ .
27. Consider the vector equation a x = b in R3 , where a 6= 0. Show that:

(a) a b = 0
b a
(b) x = + ka is a solution to the equation, for any scalar k
k a k2
28. Prove the Jacobi identity: u (v w) + v (w u) + w (u v) = 0
29. Show that u, v, w lie in the same plane in R3 if and only if u (v w) = 0.
30. For all vectors u, v, w, z in R3 , show that
(u v) (w z) = (z (u v))w (w (u v))z
and that
(u v) (w z) = (u (w z))v (v (w z))u
Why do both equations make sense geometrically?

1.5 Lines and Planes 31
1.5 Lines and Planes

Now that we know how to perform some operations on vectors, we can start to deal with some
familiar geometric objects, like lines and planes, in the language of vectors. The reason
for doing this is simple: using vectors makes it easier to study objects in 3-dimensional
Euclidean space. We will first consider lines.
Line through a point, parallel to a vector

Let P = ( x0 , y0 , z0 ) be a point in R3 , let v = (a, b, c) be a nonzero vector, and let L be the line
through P which is parallel to v (see Figure 1.5.1).
z
tv L
P(x0 , y0 , z0 )
r + tv
r t>0
r + tv y
t<0 v
0
x
Figure 1.5.1
Let r = ( x0 , y0 , z0 ) be the vector pointing from the origin to P . Since multiplying the vector
v by a scalar t lengthens or shrinks v while preserving its direction if t > 0, and reversing
its direction if t < 0, then we see from Figure 1.5.1 that every point on the line L can be
obtained by adding the vector tv to the vector r for some scalar t. That is, as t varies over all
real numbers, the vector r + tv will point to every point on L. We can summarize the vector
representation of L as follows:
For a point P = ( x0 , y0 , z0 ) and nonzero vector v in R3 , the line L through P parallel to v

is given by
r + tv, for < t < (1.16)
where r = ( x0 , y0 , z0 ) is the vector pointing to P .
Note that we used the correspondence between a vector and its terminal point. Since
v = (a, b, c), then the terminal point of the vector r + tv is ( x0 + at, y0 + bt, z0 + ct). We then get
the parametric representation of L with the parameter t:
For a point P = ( x0 , y0 , z0 ) and nonzero vector v = (a, b, c) in R3 , the line L through P

parallel to v consists of all points ( x, y, z) given by
x = x0 + at, y = y0 + bt, z = z0 + ct, for < t < (1.17)
Note that in both representations we get the point P on L by letting t = 0.

In formula (1.17), if a 6= 0, then we can solve for the parameter t: t = ( x x0 )/a. We can also
solve for t in terms of y and in terms of z if neither b nor c, respectively, is zero: t = ( y y0 )/ b
and t = ( z z0 )/ c. These three values all equal the same value t, so we can write the following
system of equalities, called the symmetric representation of L:
For a point P = ( x0 , y0 , z0 ) and vector v = (a, b, c) in R3 with a, b and c all nonzero, the line
L through P parallel to v consists of all points ( x, y, z) given by the equations
x x0 y y0 z z0
= = (1.18)
a b c
What if, say, a = 0 in the above scenario? We can not divide by z

zero, but we do know that x = x0 + at, and so x = x0 + 0 t = x0 . Then the L
symmetric representation of L would be:
y
y y0 z z0 0
x = x0 , = (1.19) x = x0
b c x0
Note that this says that the line L lies in the plane x = x0 , which is
x
parallel to the yz-plane (see Figure 1.5.2). Similar equations can be
derived for the cases when b = 0 or c = 0. Figure 1.5.2
You may have noticed that the vector representation of L in formula (1.16) is more compact
than the parametric and symmetric formulas. That is an advantage of using vector notation.
Technically, though, the vector representation gives us the vectors whose terminal points
make up the line L, not just L itself. So you have to remember to identify the vectors r + tv
with their terminal points. On the other hand, the parametric representation always gives
just the points on L and nothing else.
Example 1.19. Write the line L through the point P = (2, 3, 5) and parallel to the vector
v = (4, 1, 6), in the following forms: (a) vector, (b) parametric, (c) symmetric. Lastly: (d) find
two points on L distinct from P .
Solution: (a) Let r = (2, 3, 5). Then by formula (1.16), L is given by:
r + tv = (2, 3, 5) + t(4, 1, 6), for < t <
(b) L consists of the points ( x, y, z) such that
x = 2 + 4 t, y = 3 t, z = 5 + 6 t, for < t <
(c) L consists of the points ( x, y, z) such that
x2 y3 z5
= =
4 1 6
(d) Letting t = 1 and t = 2 in part(b) yields the points (6, 2, 11) and (10, 1, 17) on L.
Line through two points

z
Let P1 = ( x1 , y1 , z1 ) and P2 = ( x2 , y2 , z2 ) be distinct points
P2 (x2 , y2 , z2 )
in R3 , and let L be the line through P1 and P2 . Let r1 = P1 (x1 , y1 , z1 ) r2 r1
( x1 , y1 , z1 ) and r2 = ( x2 , y2 , z2 ) be the vectors pointing to P1 r2
and P2 , respectively. Then as we can see from Figure L r1 r1 + t(r2 r1 )
1.5.3, r2 r1 is the vector from P1 to P2 . So if we multiply y
the vector r2 r1 by a scalar t and add it to the vector 0
r1 , we will get the entire line L as t varies over all real
x
numbers. The following is a summary of the vector, para-
metric, and symmetric forms for the line L: Figure 1.5.3
Let P1 = ( x1 , y1 , z1 ), P2 = ( x2 , y2 , z2 ) be distinct points in R3 , and let r1 = ( x1 , y1 , z1 ), r2 =

( x2 , y2 , z2 ). Then the line L through P1 and P2 has the following representations:
Vector:
r1 + t(r2 r1 ) , for < t < (1.20)
Parametric:
x = x1 + ( x2 x1 ) t, y = y1 + ( y2 y1 ) t, z = z1 + ( z2 z1 ) t, for < t < (1.21)
Symmetric:
x x1 y y1 z z1
= = (if x1 6= x2 , y1 6= y2 , and z1 6= z2 ) (1.22)
x2 x1 y2 y1 z2 z1
Example 1.20. Write the line L through the points P1 = (3, 1, 4) and P2 = (4, 4, 6) in
parametric form.
Solution: By formula (1.21), L consists of the points ( x, y, z) such that
x = 3 + 7 t, y = 1 + 3 t, z = 4 2 t, for < t <
Distance between a point and a line
Let L be a line in R3 in vector form as r + tv (for < t < ), P

and let P be a point not on L. The distance d from P to L is the w d
length of the line segment from P to L which is perpendicular to L L

(see Figure 1.5.4). Pick a point Q on L, and let w be the vector from Q v
Q to P . If is the angle between w and v, then d = k w k sin . So
Figure 1.5.4
since k v w k = k v k k w k sin and v 6= 0, then:
kv wk
d= (1.23)
kvk
Example 1.21. Find the distance d from the point P = (1, 1, 1) to the line L in Example 1.20.
Solution: From Example 1.20, we see that we can represent L in vector form as: r + tv, for

r = (3, 1, 4) and v = (7, 3, 2). Since the point Q = (3, 1, 4) is on L, then for w = QP =
(1, 1, 1) (3, 1, 4) = (4, 0, 5), we have:

i j k
3 2 7 2 7 3
v w = 7 3 2 =
i j+
k = 15 i 43 j 12 k , so
4 0 0 5 4 5 4 0
5
p p
k v w k 15 i 43 j 12 k 152 + (43)2 + (12)2 2218
d= = = p = p = 5.98
kvk (7, 3, 2) 72 + 32 + (2)2 62
It is clear that two lines L 1 and L 2 , represented in vector form as r1 + sv1 and r2 + tv2 ,
respectively, are parallel (denoted as L 1 L 2 ) if v1 and v2 are parallel. Also, L 1 and L 2 are
perpendicular (denoted as L 1 L 2 ) if v1 and v2 are perpendicular.
In 2-dimensional space, two lines are either identical, parallel, or they z
intersect. In 3-dimensional space, there is an additional possibility: two L1
lines can be skew, that is, they do not intersect but they are not parallel.
L2 y
However, even though they are not parallel, skew lines are on parallel
planes (see Figure 1.5.5). 0
3 x
To determine whether two lines in R intersect, it is often easier to use
the parametric representation of the lines. In this case, you should use dif- Figure 1.5.5
ferent parameter variables (usually s and t) for the lines, since the values of the parameters
may not be the same at the point of intersection. Setting the two ( x, y, z) triples equal will
result in a system of 3 equations in 2 unknowns ( s and t).
Example 1.22. Find the point of intersection (if any) of the following lines:
x+1 y2 z1 y8 z+3
= = and x+3 = =
3 2 1 3 2
Solution: First we write the lines in parametric form, with parameters s and t:
x = 1 + 3 s, y = 2 + 2 s, z = 1s and x = 3 + t, y = 8 3 t, z = 3 + 2 t
The lines intersect when (1 + 3 s, 2 + 2 s, 1 s) = (3 + t, 8 3 t, 3 + 2 t) for some s, t:
1 + 3 s = 3 + t : t = 2 + 3 s
2 + 2 s = 8 3 t : 2 + 2 s = 8 3(2 + 3 s) = 2 9 s 2 s = 9 s s = 0 t = 2 + 3(0) = 2
1 s = 3 + 2 t : 1 0 = 3 + 2(2) 1 = 1 X (Note that we had to check this.)
Letting s = 0 in the equations for the first line, or letting t = 2 in the equations for the second
line, gives the point of intersection (1, 2, 1).
We will now consider planes in 3-dimensional Euclidean space.

Plane through a point, perpendicular to a vector
Let P be a plane in R3 , and suppose it contains a point P0 = ( x0 , y0 , z0 ). Let n = (a, b, c) be
a nonzero vector which is perpendicular to the plane P . Such a vector is called a normal
vector (or just a normal) to the plane. Now let ( x, y, z) be any point in the plane P . Then
the vector r = ( x x0 , y y0 , z z0 ) lies in the plane P (see Figure 1.5.6). So if r 6= 0, then r n
and hence n r = 0. And if r = 0 then we still have n r = 0.
n
r
(x, y, z) (x0 , y0 , z0 )
Figure 1.5.6 The plane P
Conversely, if ( x, y, z) is any point in R3 such that r = ( x x0 , y y0 , z z0 ) 6= 0 and n r = 0,

then r n and so ( x, y, z) lies in P . This proves the following theorem:
Theorem 1.18. Let P be a plane in R3 , let ( x0 , y0 , z0 ) be a point in P , and let n = (a, b, c) be a

nonzero vector which is perpendicular to P . Then P consists of the points ( x, y, z) satisfying
the vector equation:
n r = 0 (1.24)
where r = ( x x0 , y y0 , z z0 ), or equivalently:
a( x x0 ) + b( y y0 ) + c( z z0 ) = 0 (1.25)
The above equation is called the point-normal form of the plane P .
Example 1.23. Find the equation of the plane P containing the point (3, 1, 3) and perpen-
dicular to the vector n = (2, 4, 8).
Solution: By formula (1.25), the plane P consists of all points ( x, y, z) such that:
2( x + 3) + 4( y 1) + 8( z 3) = 0
If we multiply out the terms in formula (1.25) and combine the constant terms, we get an
equation of the plane in normal form:
ax + b y + cz + d = 0 (1.26)
For example, the normal form of the plane in Example 1.23 is 2 x + 4 y + 8 z 22 = 0.

Plane containing three noncollinear points

In 2-dimensional and 3-dimensional space, two points determine a line. Two points do
not determine a plane in R3 . In fact, three collinear points (i.e. all on the same line) do not
determine a plane; an infinite number of planes would contain the line on which those three
points lie. However, three noncollinear points do determine a plane. For if Q , R and S are

noncollinear points in R3 , then QR and QS are nonzero vectors which are not parallel (by

noncollinearity), and so their cross product QR QS is perpendicular to both QR and QS .

So QR and QS (and hence Q , R and S ) lie in the plane through the point Q with normal

vector n = QR QS (see Figure 1.5.7).

n = QR QS

QR
Q
R
S QS
Figure 1.5.7 Noncollinear points Q, R, S
Example 1.24. Find the equation of the plane P containing the points (2, 1, 3), (1, 1, 2) and
(3, 2, 1).

Solution: Let Q = (2, 1, 3), R = (1, 1, 2) and S = (3, 2, 1). Then for the vectors QR = (1, 2, 1)

and QS = (1, 1, 2), the plane P has a normal vector

n = QR QS = (1, 2, 1) (1, 1, 2) = (5, 3, 1)
So using formula (1.25) with the point Q (we could also use R or S ), the plane P consists of
all points ( x, y, z) such that:
5( x 2) 3( y 1) + ( z 3) = 0
or in normal form,
5 x 3 y + z 10 = 0
We mentioned earlier that skew lines in R3 lie on separate, parallel planes. So two skew
lines do not determine a plane. But two (nonidentical) lines which either intersect or are
parallel do determine a plane. In both cases, to find the equation of the plane that contains
those two lines, simply pick from the two lines a total of three noncollinear points (i.e. one
point from one line and two points from the other), then use the technique above, as in
Example 1.24, to write the equation. We will leave examples of this as exercises for the
reader.
Distance between a point and a plane

The distance between a point in R3 and a plane is the length of the line segment from
that point to the plane which is perpendicular to the plane. The following theorem gives a
formula for that distance.
Theorem 1.19. Let Q = ( x0 , y0 , z0 ) be a point in R3 , and let P be a plane with normal form
ax + b y + cz + d = 0 that does not contain Q . Then the distance D from Q to P is:
| ax0 + b y0 + cz0 + d |
D= p (1.27)
a2 + b 2 + c 2
Proof: Let R = ( x, y, z) be any point in the plane P (so that ax + b y + cz + d = 0) and let

r = RQ = ( x0 x, y0 y, z0 z). Then r 6= 0 since Q does not lie in P . From the normal form
equation for P , we know that n = (a, b, c) is a normal vector for P . Now, any plane divides
R3 into two disjoint parts. Assume that n points toward the side of P where the point Q
is located. Place n so that its initial point is at R , and let be the angle between r and
n. Then 0 < < 90 , so cos > 0. Thus, the distance D is cos k r k = | cos | k r k (see Figure
1.5.8).
n Q
r D
D

P
R
Figure 1.5.8
nr
By Theorem 1.6 in Section 1.3, we know that cos = , so
knk krk

n r n r
| a( x0 x) + b( y0 y) + c( z0 z) |
D = | cos | k r k = krk = = p
knk krk knk a2 + b 2 + c 2
| ax0 + b y0 + cz0 (ax + b y + cz) | | ax0 + b y0 + cz0 ( d ) | | ax0 + b y0 + cz0 + d |
= p = p = p
a2 + b 2 + c 2 a2 + b 2 + c 2 a2 + b 2 + c 2
If n points away from the side of P where the point Q is located, then 90 < < 180 and
so cos < 0. The distance D is then | cos | k r k, and thus repeating the same argument as
above still gives the same result. QED
Example 1.25. Find the distance D from (2, 4, 5) to the plane from Example 1.24.
Solution: Recall that the plane is given by 5 x 3 y + z 10 = 0. So
| 5(2) 3(4) + 1(5) 10 | |17 | 17

D= p = p = p 2.87
2 2
5 + (3) + 1 2 35 35
Line of intersection of two planes
Note that two planes are parallel if they have normal vectors that
are parallel, and the planes are perpendicular if their normal vectors
are perpendicular. If two planes do intersect, they do so in a line (see
L
Figure 1.5.9). Suppose that two planes P1 and P2 with normal vectors
n1 and n2 , respectively, intersect in a line L. Since n1 n2 n1 , then
n1 n2 is parallel to the plane P1 . Likewise, n1 n2 n2 means that Figure 1.5.9
n1 n2 is also parallel to P2 . Thus, n1 n2 is parallel to the intersection
of P1 and P2 , i.e. n1 n2 is parallel to L. Thus, we can write L in the following vector form:
L : r + t(n1 n2 ) , for < t < (1.28)
where r is any vector pointing to a point belonging to both planes. To find a point in both
planes, find a common solution ( x, y, z) to the two normal form equations of the planes. This
can often be made easier by setting one of the coordinate variables to zero, which leaves you
to solve two equations in just two unknowns.
Example 1.26. Find the line of intersection L of the planes 5 x 3 y + z 10 = 0 and 2 x + 4 y

z + 3 = 0.
Solution: The plane 5 x 3 y + z 10 = 0 has normal vector n1 = (5, 3, 1) and the plane 2 x +
4 y z + 3 = 0 has normal vector n2 = (2, 4, 1). Since n1 and n2 are not scalar multiples, then
the two planes are not parallel and hence will intersect. A point ( x, y, z) on both planes will
satisfy the following system of two equations in three unknowns:
5 x 3 y + z 10 = 0
2x + 4 y z + 3 = 0
Set x = 0 (why is that a good choice?). Then the above equations are reduced to:
3 y + z 10 = 0
4y z + 3 = 0
The second equation gives z = 4 y + 3, substituting that into the first equation gives y = 7.
Then z = 31, and so the point (0, 7, 31) is on L. Since n1 n2 = (1, 7, 26), then L is given by:
r + t(n1 n2 ) = (0, 7, 31) + t(1, 7, 26), for < t <
or in parametric form:
x = t, y = 7 + 7 t, z = 31 + 26 t, for < t <

Exercises
A
For Exercises 1-4, write the line L through the point P and parallel to the vector v in the
following forms: (a) vector, (b) parametric, and (c) symmetric.
1. P = (2, 3, 2), v = (5, 4, 3) 2. P = (3, 1, 2), v = (2, 8, 1)
3. P = (2, 1, 3), v = (1, 0, 1) 4. P = (0, 0, 0), v = (7, 2, 10)
For Exercises 5-6, write the line L through the points P1 and P2 in parametric form.
5. P1 = (1, 2, 3), P2 = (3, 5, 5) 6. P1 = (4, 1, 5), P2 = (2, 1, 3)
For Exercises 7-8, find the distance d from the point P to the line L.
7. P = (1, 1, 1), L : x = 2 2 t, y = 4 t, z = 7 + t
8. P = (0, 0, 0), L : x = 3 + 2 t, y = 4 + 3 t, z = 5 + 4 t
For Exercises 9-10, find the point of intersection (if any) of the given lines.
9. x = 7 + 3 s, y = 4 3 s, z = 7 5 s and x = 1 + 6 t, y = 2 + t, z = 3 2 t
x6 x 11 y 14 z + 9
10. = y + 3 = z and = =
4 3 6 2
For Exercises 11-12, write the normal form of the plane P containing the point Q and per-
pendicular to the vector n.
11. Q = (5, 1, 2), n = (4, 4, 3) 12. Q = (6, 2, 0), n = (2, 6, 4)
For Exercises 13-14, write the normal form of the plane containing the given points.
13. (1, 0, 3), (1, 2, 1), (6, 1, 6) 14. (3, 1, 3), (4, 4, 3), (0, 0, 1)
15. Write the normal form of the plane containing the lines from Exercise 9.
16. Write the normal form of the plane containing the lines from Exercise 10.
For Exercises 17-18, find the distance D from the point Q to the plane P .
17. Q = (4, 1, 2), P : 3 x y 5 z + 8 = 0 18. Q = (0, 2, 0), P : 5 x + 2 y 7 z + 1 = 0
For Exercises 19-20, find the line of intersection (if any) of the given planes.
19. x + 3 y + 2 z 6 = 0, 2 x y + z + 2 = 0 20. 3 x + y 5 z = 0, x + 2 y + z + 4 = 0
B
x6
21. Find the point(s) of intersection (if any) of the line = y + 3 = z with the plane
4
x + 3 y + 2 z 6 = 0. (Hint: Put the equations of the line into the equation of the plane.)
1.6 Surfaces
In the previous section we discussed planes in Euclidean space. A plane is an example of
a surface, which we will define informally8 as the solution set of the equation F ( x, y, z) = 0
in R3 , for some real-valued function F . For example, a plane given by ax + b y + cz + d = 0
is the solution set of F ( x, y, z) = 0 for the function F ( x, y, z) = ax + b y + cz + d . Surfaces are
2-dimensional. The plane is the simplest surface, since it is flat. In this section we will
look at some surfaces that are more complex, the most important of which are the sphere
and the cylinder.
Definition 1.9. A sphere S is the set of all points ( x, y, z) in R3 which are a fixed distance r
(called the radius) from a fixed point P0 = ( x0 , y0 , z0 ) (called the center of the sphere):
S = { ( x, y, z) : ( x x0 )2 + ( y y0 )2 + ( z z0 )2 = r 2 } (1.29)
Using vector notation, this can be written in the equivalent form:
S = { x : k x x0 k = r } (1.30)
where x = ( x, y, z) and x0 = ( x0 , y0 , z0 ) are vectors.
Figure 1.6.1 illustrates the vectorial approach to spheres.

z
z k x x0 k = r
kxk = r
x x
x x0
y ( x0 , y0 , z0 )
0
x0
y
0
x
x
(a) radius r , center (0, 0, 0) (b) radius r , center ( x0 , y0 , z0 )
Figure 1.6.1 Spheres in R3
Note in Figure 1.6.1(a) that the intersection of the sphere with the x y-plane is a circle
of radius r (i.e. a great circle, given by x2 + y2 = r 2 as a subset of R2 ). Similarly for the
intersections with the xz-plane and the yz-plane. In general, a plane intersects a sphere
either at a single point or in a circle.
8 See ON EILL for a deeper and more rigorous discussion of surfaces.
1.6 Surfaces 41
Example 1.27. Find the intersection of the sphere x2 + y2 + z2 = 169 with the plane z = 12.
z
p The sphere is centered at the origin and has radius
Solution:
13 = 169, so it does intersect the plane z = 12. Putting z = 12
z = 12 into the equation of the sphere gives
y
x2 + y2 + 122 = 169
0
x2 + y2 = 169 144 = 25 = 52
which is a circle of radius 5 centered at (0, 0, 12), parallel to x

the x y-plane (see Figure 1.6.2). Figure 1.6.2
If the equation in formula (1.29) is multiplied out, we get an equation of the form:
x2 + y2 + z2 + ax + b y + cz + d = 0 (1.31)
for some constants a, b, c and d . Conversely, an equation of this form may describe a sphere,
which can be determined by completing the square for the x, y and z variables.
Example 1.28. Is 2 x2 + 2 y2 + 2 z2 8 x + 4 y 16 z + 10 = 0 the equation of a sphere?

Solution: Dividing both sides of the equation by 2 gives
x 2 + y2 + z 2 4 x + 2 y 8 z + 5 = 0
( x2 4 x + 4) + ( y2 + 2 y + 1) + ( z2 8 z + 16) + 5 4 1 16 = 0
( x 2)2 + ( y + 1)2 + ( z 4)2 = 16
which is a sphere of radius 4 centered at (2, 1, 4).
Example 1.29. Find the points(s) of intersection (if any) of the sphere from Example 1.28
and the line x = 3 + t, y = 1 + 2 t, z = 3 t.
Solution: Put the equations of the line into the equation of the sphere, which was ( x 2)2 +
( y + 1)2 + ( z 4)2 = 16, and solve for t:
(3 + t 2)2 + (1 + 2 t + 1)2 + (3 t 4)2 = 16

( t + 1)2 + (2 t + 2)2 + ( t 1)2 = 16
6 t2 + 12 t 10 = 0
4
The quadratic formula gives the solutions t = 1 p . Putting those two values into the
6
equations of the line gives the following two points of intersection:

4 8 4 4 8 4
2 + p , 1 + p , 4 p and 2 p , 1 p , 4 + p
6 6 6 6 6 6
If two spheres intersect, they do so either at a single point or in a circle.
Example 1.30. Find the intersection (if any) of the spheres x2 + y2 + z2 = 25 and x2 + y2 + ( z
2)2 = 16.
Solution: For any point ( x, y, z) on both spheres, we see that
x2 + y2 + z2 = 25 x2 + y2 = 25 z2 , and
x2 + y2 + ( z 2)2 = 16 x2 + y2 = 16 ( z 2)2 , so
16 ( z 2)2 = 25 z2 4z 4 = 9 z = 13/4
x + y = 25 (13/4)2 = 231/16
2 2
p
231
The intersection is the circle x2 + y2 = 231
16 of radius 4 3.8 centered at (0, 0, 13
4 ).
The cylinders that we will consider are right circular cylinders. These are cylinders ob-
tained by moving a line L along a circle C in R3 in a way so that L is always perpendicular
to the plane containing C . We will only consider the cases where the plane containing C is
parallel to one of the three coordinate planes (see Figure 1.6.3).
z z
r
z
r y
y r
y 0
0
0
x x x
(a) x2 + y2 = r 2 , any z (b) x2 + z2 = r 2 , any y (c) y2 + z2 = r 2 , any x
Figure 1.6.3 Cylinders in R3
For example, the equation of a cylinder whose base circle C lies in the x y-plane and is
centered at (a, b, 0) and has radius r is
( x a)2 + ( y b)2 = r 2 , (1.32)
where the value of the z coordinate is unrestricted. Similar equations can be written when
the base circle lies in one of the other coordinate planes. A plane intersects a right circular
cylinder in a circle, ellipse, or one or two lines, depending on whether that plane is parallel,
oblique9 , or perpendicular, respectively, to the plane containing C . The intersection of a
surface with a plane is called the trace of the surface.
9 i.e. at an angle strictly between 0 and 90 .
1.6 Surfaces 43
The equations of spheres and cylinders are examples of second-degree equations in R3 , i.e.
equations of the form
Ax2 + B y2 + Cz2 + Dx y + Exz + F yz + Gx + H y + I z + J = 0 (1.33)
for some constants A , B, . . . , J . If the above equation is not that of a sphere, cylinder, plane,
line or point, then the resulting surface is called a quadric surface.
One type of quadric surface is the ellipsoid, given z
c
by an equation of the form:
x 2 y2 z 2
+ + =1 (1.34) y
a2 b 2 c 2 0 b
a
In the case where a = b = c, this is just a sphere.
In general, an ellipsoid is egg-shaped (think of an
ellipse rotated around its major axis). Its traces in x
the coordinate planes are ellipses. Figure 1.6.4 Ellipsoid
Two other types of quadric surfaces are the hyperboloid of one sheet, given by an
equation of the form:
x 2 y2 z 2
+ =1 (1.35)
a2 b 2 c 2
and the hyperboloid of two sheets, whose equation has the form:
x 2 y2 z 2
=1 (1.36)
a2 b 2 c 2
z z
y y
0 0
x x
Figure 1.6.5 Hyperboloid of one sheet Figure 1.6.6 Hyperboloid of two sheets
For the hyperboloid of one sheet, the trace in any plane parallel to the x y-plane is an
ellipse. The traces in the planes parallel to the xz- or yz-planes are hyperbolas (see Figure
1.6.5), except for the special cases x = a and y = b; in those planes the traces are pairs of
intersecting lines (see Exercise 8).
For the hyperboloid of two sheets, the trace in any plane parallel to the x y- or xz-plane is
a hyperbola (see Figure 1.6.6). There is no trace in the yz-plane. In any plane parallel to the
yz-plane for which | x | > | a |, the trace is an ellipse.
The elliptic paraboloid is another type of quadric surface, z
whose equation has the form:
x 2 y2 z
+ = (1.37)
a2 b 2 c
The traces in planes parallel to the x y-plane are ellipses, though
in the x y-plane itself the trace is a single point. The traces in
planes parallel to the xz- or yz-planes are parabolas. Figure y
1.6.7 shows the case where c > 0. When c < 0 the surface is 0
turned downward. In the case where a = b, the surface is called x
a paraboloid of revolution, which is often used as a reflecting sur- Figure 1.6.7 Paraboloid
face, e.g. in vehicle headlights.10
A more complicated quadric surface is the hyperbolic paraboloid, given by:
x 2 y2 z
= (1.38)
a2 b 2 c
100
50
z 0
-10
-50
-5
-100
-10
0
-5 x
0 5
y 5
10 10
Figure 1.6.8 Hyperbolic paraboloid
10 For a discussion of this see pp. 157-158 in H ECHT.

1.6 Surfaces 45
The hyperbolic paraboloid can be tricky to draw; using graphing software on a computer
can make it easier. For example, Figure 1.6.8 was created using the free Gnuplot package
(see Appendix C). It shows the graph of the hyperbolic paraboloid z = y2 x2 , which is the
special case where a = b = 1 and c = 1 in equation (1.38). The mesh lines on the surface are
the traces in planes parallel to the coordinate planes. So we see that the traces in planes
parallel to the xz-plane are parabolas pointing upward, while the traces in planes parallel
to the yz-plane are parabolas pointing downward. Also, notice that the traces in planes
parallel to the x y-plane are hyperbolas, though in the x y-plane itself the trace is a pair of
intersecting lines through the origin. This is true in general when c < 0 in equation (1.38).
When c > 0, the surface would be similar to that in Figure 1.6.8, only rotated 90 around
the z-axis and the nature of the traces in planes parallel to the xz- or yz-planes would be
reversed.
The last type of quadric surface that we will consider is the z
elliptic cone, which has an equation of the form:
x 2 y2 z 2
+ =0 (1.39)
a2 b 2 c 2
The traces in planes parallel to the x y-plane are ellipses, ex- y
cept in the x y-plane itself where the trace is a single point. 0
The traces in planes parallel to the xz- or yz-planes are hyper-
bolas, except in the xz- and yz-planes themselves where the
traces are pairs of intersecting lines. x
Notice that every point on the elliptic cone is on a line which
lies entirely on the surface; in Figure 1.6.9 these lines all go
Figure 1.6.9 Elliptic cone
through the origin. This makes the elliptic cone an example of
a ruled surface. The cylinder is also a ruled surface.
What may not be as obvious is that both the hyperboloid of one sheet and the hyperbolic
paraboloid are ruled surfaces. In fact, on both surfaces there are two lines through each
point on the surface (see Exercises 11-12). Such surfaces are called doubly ruled surfaces,
and the pairs of lines are called a regulus.
It is clear that for each of the six types of quadric surfaces that we discussed, the surface
can be translated away from the origin (e.g. by replacing x2 by ( x x0 )2 in its equation). It can
be proved11 that every quadric surface can be translated and/or rotated so that its equation
matches one of the six types that we described. For example, z = 2 x y is a case of equation
(1.33) with mixed variables, e.g. with D 6= 0 so that we get an x y term. This equation does

not match any of the types we considered. However, by rotating the p x- and y -axes pby 45 in

the x y-plane by means of the coordinate transformation x = ( x y )/ 2, y = ( x + y )/ 2, z = z ,

then z = 2 x y becomes the hyperbolic paraboloid z = ( x )2 ( y )2 in the ( x , y , z ) coordinate

system. That is, z = 2 x y is a hyperbolic paraboloid as in equation (1.38), but rotated 45 in
the x y-plane.
11 See Ch. 7 in P OGORELOV.
Exercises
A
For Exercises 1-4, determine if the given equation describes a sphere. If so, find its radius
and center.
1. x2 + y2 + z2 4 x 6 y 10 z + 37 = 0 2. x2 + y2 + z2 + 2 x 2 y 8 z + 19 = 0
3. 2 x2 + 2 y2 + 2 z2 + 4 x + 4 y + 4 z 44 = 0 4. x2 + y2 z2 + 12 x + 2 y 4 z + 32 = 0
5. Find the point(s) of intersection of the sphere ( x 3)2 + ( y + 1)2 + ( z 3)2 = 9 and the line
x = 1 + 2 t , y = 2 3 t , z = 3 + t .
B
6. Find the intersection of the spheres x2 + y2 + z2 = 9 and ( x 4)2 + ( y + 2)2 + ( z 4)2 = 9.
7. Find the intersection of the sphere x2 + y2 + z2 = 9 and the cylinder x2 + y2 = 4.

x2 y2 z2
8. Find the trace of the hyperboloid of one sheet a2
+ b2 c2
= 1 in the plane x = a, and the
trace in the plane y = b.
x2 y2 z
9. Find the trace of the hyperbolic paraboloid a2
b2 = c in the x y-plane.
C
10. It can be shown that any four noncoplanar points (i.e. points that do not lie in the same
plane) determine a sphere.12 Find the equation of the sphere that passes through the
points (0, 0, 0), (0, 0, 2), (1, 4, 3) and (0, 1, 3). (Hint: Equation (1.31))
11. Show that the hyperboloid of one sheet is a doubly ruled surface, i.e. each point on
the surface is on two lines lying entirely on the surface. (Hint: Write equation (1.35) as
x2 2 y2
a2
zc2 = 1 b2 , factor each side. Recall that two planes intersect in a line.)
12. Show that the hyperbolic paraboloid is a doubly ruled surface. (Hint: Exercise 11)
z
13. Let S be the sphere with radius 1 centered at (0, 0, 1), (0, 0, 2)
and let S be S without the north pole point (0, 0, 2). Let
(a, b, c) be an arbitrary point on S . Then the line passing (a, b, c)
S
through (0, 0, 2) and (a, b, c) intersects the x y-plane at some 1
point ( x, y, 0), as in Figure 1.6.10. Find this point ( x, y, 0) in y
terms of a, b and c. 0
(x, y, 0)
(Note: Every point in the x y-plane can be matched with a x
point on S , and vice versa, in this manner. This method is Figure 1.6.10
called stereographic projection, which essentially identifies
all of R2 with a punctured sphere.)
12 See W ELCHONS and K RICKENBERGER, p. 160, for a proof.

1.7 Curvilinear Coordinates 47
1.7 Curvilinear Coordinates

The Cartesian coordinates of a point ( x, y, z) are determined by z (x, y, z)
following straight paths starting from the origin: first along the
z
x-axis, then parallel to the y-axis, then parallel to the z-axis, as y
in Figure 1.7.1. In curvilinear coordinate systems, these paths can
x 0
be curved. The two types of curvilinear coordinates which we will
consider are cylindrical and spherical coordinates. Instead of ref- y
x
erencing a point in terms of sides of a rectangular parallelepiped,
Figure 1.7.1
as with Cartesian coordinates, we will think of the point as ly-
ing on a cylinder or sphere. Cylindrical coordinates are often used when there is symmetry
around the z-axis; spherical coordinates are useful when there is symmetry about the origin.
Let P = ( x, y, z) be a point in Cartesian coordinates in R3 , and let P0 = ( x, y, 0) be the
projection of P upon the x y-plane. Treating ( x, y) as a point in R2 , let ( r, ) be its polar
coordinates (see Figure 1.7.2). Let be the length of the line segment from the origin to P ,
and let be the angle between that line segment and the positive z-axis (see Figure 1.7.3).
is called the zenith angle. Then the cylindrical coordinates ( r, , z) and the spherical
coordinates ( , , ) of P ( x, y, z) are defined as follows:13
Cylindrical coordinates ( r, , z): z P(x, y, z)

q
x = r cos r= x 2 + y2 z
y y
y = r sin = tan1 x 0
x r
z=z z=z
y P0 (x, y, 0)
x
where 0 if y 0 and < < 2 if y < 0
Figure 1.7.2
Cylindrical coordinates
Spherical coordinates ( , , ): z P(x, y, z)

q
x = sin cos = x 2 + y2 + z 2 z

y
y = sin sin = tan 1 y 0
x x

1 p 2 z2 2
z = cos = cos
x + y +z y P0 (x, y, 0)
x
where 0 if y 0 and < < 2 if y < 0 Figure 1.7.3
Spherical coordinates
Both and are measured in radians. Note that r 0, 0 < 2, 0 and 0 .
Also, is undefined when ( x, y) = (0, 0), and is undefined when ( x, y, z) = (0, 0, 0).
13 This standard definition of spherical coordinates used by mathematicians results in a left-handed system.
For this reason, physicists usually switch the definitions of and to make ( , , ) a right-handed system.
Example 1.31. Convert the point (2, 2, 1) from Cartesian coordinates to (a) cylindrical
and (b) spherical coordinates.
p p 2 5
Solution: (a) r = (2)2 + (2)2 = 2 2, = tan1 1
2 = tan (1) = 4 , since y = 2 < 0.
p 5
(r, , z) = 2 2, 4 , 1
p p
(b) = (2)2 + (2)2 + 12 = 9 = 3, = cos1 13 1.23 radians.
5
( , , ) = 3, 4 , 1.23
For cylindrical coordinates ( r, , z), and constants r 0 , 0 and z0 , we see from Figure 1.7.4
that the surface r = r 0 is a cylinder of radius r 0 centered along the z-axis, the surface = 0
is a half-plane emanating from the z-axis, and the surface z = z0 is a plane parallel to the
x y-plane.
z z z
r0 z0
y y y
0
0 0
0
x x x
(a) r = r 0 (b) = 0 (c) z = z0
Figure 1.7.4 Cylindrical coordinate surfaces
For spherical coordinates ( , , ), and constants 0 , 0 and 0 , we see from Figure 1.7.5
that the surface = 0 is a sphere of radius 0 centered at the origin, the surface = 0 is a
half-plane emanating from the z-axis, and the surface = 0 is a circular cone whose vertex
is at the origin.
z
z z
0
0
y y
0
0
y
0 0
x x x
(a) = 0 (b) = 0 (c) = 0
Figure 1.7.5 Spherical coordinate surfaces
Figures 1.7.4(a) and 1.7.5(a) show how these coordinate systems got their names.
1.7 Curvilinear Coordinates 49
Sometimes the equation of a surface in Cartesian coordinates can be transformed into a

simpler equation in some other coordinate system, as in the following example.
Example 1.32. Write the equation of the cylinder x2 + y2 = 4 in cylindrical coordinates.

p
Solution: Since r = x2 + y2 , then the equation in cylindrical coordinates is r = 2.
Using spherical coordinates to write the equation of a sphere does not necessarily make
the equation simpler, if the sphere is not centered at the origin.
Example 1.33. Write the equation ( x 2)2 + ( y 1)2 + z2 = 9 in spherical coordinates.

Solution: Multiplying the equation out gives
x2 + y2 + z2 4 x 2 y + 5 = 9 , so we get
2 4 sin cos 2 sin sin 4 = 0 , or
2 2 sin (2 cos sin ) 4 = 0
after combining terms. Note that this actually makes it more difficult to figure out what the
surface is, as opposed to the Cartesian equation where you could immediately identify the
surface as a sphere of radius 3 centered at (2, 1, 0).
Example 1.34. Describe the surface given by = z in cylindrical coordinates.

Solution: This surface is called a helicoid. As the (vertical) z coordinate increases, so does
the angle , while the radius r is unrestricted. So this sweeps out a (ruled!) surface shaped
like a spiral staircase, where the spiral has an infinite radius. Figure 1.7.6 shows a section
of this surface restricted to 0 z 4 and 0 r 2.
14
12
10
8
z 6
4
-2
2 -1.5
-1
0-2 -0.5
-1.5 -1 0
-0.5 0.5 x
0 0.5 1
y 1 1.5 1.5
2 2
Figure 1.7.6 Helicoid = z

Exercises
A
For Exercises 1-4, find the (a) cylindrical and (b) spherical coordinates of the point whose
Cartesian coordinates are given.
p p p p
1. (2, 2 3, 1) 2. (5, 5, 6) 3. ( 21, 7, 0) 4. (0, 2, 2)
For Exercises 5-7, write the given equation in (a) cylindrical and (b) spherical coordinates.
5. x2 + y2 + z2 = 25 6. x2 + y2 = 2 y 7. x2 + y2 + 9 z2 = 36
B
8. Describe the intersection of the surfaces whose equations in spherical coordinates are
= 2 and = 4 .
9. Show that for a 6= 0, the equation = 2a sin cos in spherical coordinates describes a
sphere centered at (a, 0, 0) with radius | a |.
C
10. Let P = (a, , ) be a point in spherical coordinates, with a > 0 and 0 < < . Then P
lies on the sphere = a. Since 0 < < , the line segment from the origin to P can be
extended to intersect the cylinder given by r = a (in cylindrical coordinates). Find the
cylindrical coordinates of that point of intersection.
11. Let P1 and P2 be points whose spherical coordinates are ( 1 , 1 , 1 ) and ( 2 , 2 , 2 ), respec-
tively. Let v1 be the vector from the origin to P1 , and let v2 be the vector from the origin
to P2 . For the angle between v1 and v2 , show that
cos = cos 1 cos 2 + sin 1 sin 2 cos( 2 1 ).
This formula is used in electrodynamics to prove the addition theorem for spherical har-
monics, which provides a general expression for the electrostatic potential at a point due
to a unit charge. See pp. 100-102 in J ACKSON.
12. Show that the distance d between the points P1 and P2 with cylindrical coordinates
( r 1 , 1 , z1 ) and ( r 2 , 2 , z2 ), respectively, is
q
d = r 21 + r 22 2 r 1 r 2 cos( 2 1 ) + ( z2 z1 )2 .
13. Show that the distance d between the points P1 and P2 with spherical coordinates
( 1 , 1 , 1 ) and ( 2 , 2 , 2 ), respectively, is
q
d = 21 + 22 2 1 2 [sin 1 sin 2 cos( 2 1 ) + cos 1 cos 2 ] .
1.8 Vector-Valued Functions 51
1.8 Vector-Valued Functions

Now that we are familiar with vectors and their operations, we can begin discussing func-
tions whose values are vectors.
Definition 1.10. A vector-valued function of a real variable is a rule that associates a

vector f( t) with a real number t, where t is in some subset D of R1 (called the domain of f).
We write f : D R3 to denote that f is a mapping of D into R3 .
For example, f( t) = ti + t2 j + t3 k is a vector-valued function in R3 , defined for all real num-

bers t. We would write f : R R3 . At t = 1 the value of the function is the vector i + j + k,
which in Cartesian coordinates has the terminal point (1, 1, 1).
A vector-valued function of a real variable can be written in component form as
f( t) = f 1 ( t)i + f 2 ( t)j + f 3 ( t)k
or in the form
f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t))
for some real-valued functions f 1 ( t), f 2 ( t), f 3 ( t), called the component functions of f. The first
form is often used when emphasizing that f( t) is a vector, and the second form is useful
when considering just the terminal points of the vectors. By identifying vectors with their
terminal points, a curve in space can be written as a vector-valued function.
z
3
Example 1.35. Define f : R R by f( t) = (cos t, sin t, t).
This is the equation of a helix (see Figure 1.8.1). As the value of
t increases, the terminal points of f( t) trace out a curve spiraling
y
upward. For each t, the x- and y-coordinates of f( t) are x = cos t f(2) 0
and y = sin t, so
f(0)
x2 + y2 = cos2 t + sin2 t = 1.
x
Thus, the curve lies on the surface of the right circular cylinder
Figure 1.8.1
x2 + y2 = 1.
It may help to think of vector-valued functions of a real variable in R3 as a generalization

of the parametric functions in R2 which you learned about in single-variable calculus. Much
of the theory of real-valued functions of a single real variable can be applied to vector-valued
functions of a real variable. Since each of the three component functions are real-valued, it
will sometimes be the case that results from single-variable calculus can simply be applied
to each of the component functions to yield a similar result for the vector-valued function.
However, there are times when such generalizations do not hold (see Exercise 13). The
concept of a limit, though, can be extended naturally to vector-valued functions, as in the
following definition.
Definition 1.11. Let f( t) be a vector-valued function, let a be a real number and let c be a
vector. Then we say that the limit of f( t) as t approaches a equals c, written as lim f( t) = c,
t a
if lim k f( t) c k = 0. If f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)), then
t a

lim f( t) = lim f 1 ( t), lim f 2 ( t), lim f 3 ( t)
t a t a t a t a
provided that all three limits on the right side exist.
The above definition shows that continuity and the derivative of vector-valued functions
can also be defined in terms of its component functions.
Definition 1.12. Let f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)) be a vector-valued function, and let a be a real
number in its domain. Then f( t) is continuous at a if lim f( t) = f(a). Equivalently, f( t) is
t a
continuous at a if and only if f 1 ( t), f 2 ( t), and f 3 ( t) are continuous at a.
df
The derivative of f( t) at a, denoted by f (a) or (a), is the limit
dt
f(a + h) f(a)
f (a) = lim
h0 h
if that limit exists. Equivalently, f (a) = ( f 1 (a), f 2 (a), f 3 (a)), if the component derivatives
exist. We say that f( t) is differentiable at a if f (a) exists.
Recall that the derivative of a real-valued function of a single variable is a real number,
representing the slope of the tangent line to the graph of the function at a point. Similarly,
the derivative of a vector-valued function is a tangent vector to the curve in space which
the function represents, and it lies on the tangent line to the curve (see Figure 1.8.2).
z
f ( a)
f(a L
+
f(a) h)

f( a f( t)
)
f(a + h) y
0
x
Figure 1.8.2 Tangent vector f (a) and tangent line L = f(a) + sf (a)
Example 1.36. Let f( t) = (cos t, sin t, t). Then f ( t) = ( sin t, cos t, 1) for all t. The tangent line
L to the curve at f(2) = (1, 0, 2) is L = f(2) + s f (2) = (1, 0, 2) + s(0, 1, 1), or in parametric
form: x = 1, y = s, z = 2 + s for < s < .
A scalar function is a real-valued function. Note that if u( t) is a scalar function and

f( t) is a vector-valued function, then their product, defined by ( u f)( t) = u( t) f( t) for all t, is a
vector-valued function (since the product of a scalar with a vector is a vector).
The basic properties of derivatives of vector-valued functions are summarized in the fol-
lowing theorem.
Theorem 1.20. Let f( t) and g( t) be differentiable vector-valued functions, let u( t) be a

differentiable scalar function, let k be a scalar, and let c be a constant vector. Then
d
(a) (c) = 0
dt
d df
(b) ( kf) = k
dt dt
d df dg
(c) (f + g) = +
dt dt dt
d df dg
(d) (f g) =
dt dt dt
d du df
(e) ( u f) = f+u
dt dt dt
d df dg
(f) (f g) = g + f
dt dt dt
d df dg
(g) (f g) = g+f
dt dt dt
Proof: The proofs of parts (a)-(e) follow easily by differentiating the component functions
and using the rules for derivatives from single-variable calculus. We will prove part (f),
and leave the proof of part (g) as an exercise for the reader.
(f) Write f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)) and g( t) = ( g 1 ( t), g 2 ( t), g 3 ( t)), where the component functions
f 1 ( t), f 2 ( t), f 3 ( t), g 1 ( t), g 2 ( t), g 3 ( t) are all differentiable real-valued functions. Then
d d
(f( t) g( t)) = ( f 1 ( t) g 1 ( t) + f 2 ( t) g 2 ( t) + f 3 ( t) g 3 ( t))
dt dt
d d d
= ( f 1 ( t) g 1 ( t)) + ( f 2 ( t) g 2 ( t)) + ( f 3 ( t) g 3 ( t))
dt dt dt
d f1 d g1 d f2 d g2 d f3 d g3
= ( t) g 1 ( t) + f 1 ( t) ( t) + ( t) g 2 ( t) + f 2 ( t) ( t) + ( t) g 3 ( t) + f 3 ( t) ( t)
dt dt dt dt dt dt
df d f2 d f3
1
= ( t), ( t), ( t) ( g 1 ( t), g 2 ( t), g 3 ( t))
dt dt dt
dg d g2 d g3
1
+ ( f 1 ( t), f 2 ( t), f 3 ( t)) ( t), ( t), ( t)
dt dt dt
df dg
= ( t) g( t) + f( t) ( t) for all t. QED
dt dt
Example 1.37. Suppose f( t) is differentiable. Find the derivative of k f( t) k.

Solution: Since k f( t) k is a real-valued function of t, then by the Chain Rule for real-valued
d d
functions, we know that k f( t) k2 = 2k f( t) k k f( t) k.
dt dt
2 d 2 d
But k f( t) k = f( t) f( t), so k f( t) k = (f( t) f( t)). Hence, we have
dt dt
d d
2k f( t) k k f( t) k = (f( t) f( t)) = f ( t) f( t) + f( t) f ( t) by Theorem 1.20(f), so
dt dt
= 2f ( t) f( t) , so if k f( t) k 6= 0 then
d f ( t) f( t)
k f( t) k = .
dt k f( t) k
d
We know that k f( t) k is constant if and only if k f( t) k = 0 for all t. Also, f( t) f ( t) if and
dt
only if f ( t) f( t) = 0. Thus, the above example shows this important fact:
If k f( t) k 6= 0, then k f( t) k is constant if and only if f( t) f ( t) for all t.
This means that if a curve lies completely on a sphere (or circle) centered at the origin, then
the tangent vector f ( t) is always perpendicular to the position vector f( t).
cos t sin t at
Example 1.38. The spherical spiral f( t) = p ,p ,p , for a 6= 0.
1 + a2 t2 1 + a2 t2 1 + a2 t2
Figure 1.8.3 shows the graph of the curve when a = 0.2. In the exercises, the reader will be
asked to show that this curve lies on the sphere x2 + y2 + z2 = 1 and to verify directly that
f ( t) f( t) = 0 for all t.
1
0.8
0.6
0.4
z 0.2
0
-0.2
-0.4
-0.6 -1
-0.8 -0.8
-0.6
-1-1 -0.4
-0.2
-0.8 -0.6 0
-0.4 -0.2 0.2 x
0 0.4
0.2 0.4 0.6
y 0.6 0.8 0.8
1 1
Figure 1.8.3 Spherical spiral with a = 0.2

Just as in single-variable calculus, higher-order derivatives of vector-valued functions are

obtained by repeatedly differentiating the (first) derivative of the function:
d d dnf d d n 1 f
f ( t) = f ( t) , f ( t) = f ( t) , ... , = (for n = 2, 3, 4, . . .)
dt dt dt n dt dt n1
We can use vector-valued functions to represent physical quantities, such as velocity, ac-
celeration, force, momentum, etc. For example, let the real variable t represent time elapsed
from some initial time ( t = 0), and suppose that an object of constant mass m is subjected
to some force so that it moves in space, with its position ( x, y, z) at time t a function of
t. That is, x = x( t), y = y( t), z = z( t) for some real-valued functions x( t), y( t), z( t). Call
r( t) = ( x( t), y( t), z( t)) the position vector of the object. We can define various physical quan-
tities associated with the object as follows:14
position: r( t) = ( x( t), y( t), z( t))

dr
velocity: v( t) = r ( t) = r ( t) =
dt

= ( x ( t), y ( t), z ( t))
dv
t) = v ( t) =
acceleration: a( t) = v(
dt
d2r
= r ( t) = r ( t) = 2
dt

= ( x ( t), y ( t), z ( t))
momentum: p( t) = mv( t)
dp
t) = p ( t) =
force: F( t) = p( (Newtons Second Law of Motion)
dt
The magnitude k v( t) k of the velocity vector is called the speed of the object. Note that since
the mass m is a constant, the force equation becomes the familiar F( t) = ma( t).
Example 1.39. Let r( t) = (5 cos t, 3 sin t, 4 sin t) be the position vector of an object at time t 0.
Find its (a) velocity and (b) acceleration vectors.
Solution: (a) v( t) = r ( t) = (5 sin t, 3 cos t, 4 cos t)
(b) a( t) = v( t) = (5 cos t, 3 sin t, 4 sin t)
p
Note that k r( t) k = 25 cos2 t + 25 sin2 t = 5 for all t, so by Example 1.37 we know that r( t)
r ( t) = 0 for all t (which we can verify from part (a)). In fact, k v( t) k = 5 for all t also. And not
only does r( t) lie on the sphere of radius 5 centered at the origin, but perhaps not so obvious
is that it lies completely within a circle of radius 5 centered at the origin. Also, note that
a( t) = r( t). It turns out (see Exercise 16) that whenever an object moves in a circle with
constant speed, the acceleration vector will point in the opposite direction of the position
vector (i.e. towards the center of the circle).
14 We will often use the older dot notation for derivatives when physics is involved.
Recall from Section 1.5 that if r1 , r2 are position vectors to distinct points then r1 + t(r2 r1 )
represents a line through those two points as t varies over all real numbers. That vector
sum can be written as (1 t)r1 + tr2 . So the function l( t) = (1 t)r1 + tr2 is a line through
the terminal points of r1 and r2 , and when t is restricted to the interval [0, 1] it is the line
segment between the points, with l(0) = r1 and l(1) = r2 .
In general, a function of the form f( t) = (a 1 t + b 1 , a 2 t + b 2 , a 3 t + b 3 ) represents a line in R3 . A
function of the form f( t) = (a 1 t2 + b 1 t + c 1 , a 2 t2 + b 2 t + c 2 , a 3 t2 + b 3 t + c 3 ) represents a (possibly
degenerate) parabola in R3 .
Example 1.40. Bzier curves are used in Computer Aided Design (CAD) to approximate
the shape of a polygonal path in space (called the Bzier polygon or control polygon). For
instance, given three points (or position vectors) b0 , b1 , b2 in R3 , define
b10 ( t) = (1 t)b0 + tb1

b11 ( t) = (1 t)b1 + tb2
b20 ( t) = (1 t)b10 ( t) + tb11 ( t)
= (1 t)2 b0 + 2 t(1 t)b1 + t2 b2
for all real t. For t in the interval [0, 1], we see that b10 ( t) is the line segment between b0 and
b1 , and b11 ( t) is the line segment between b1 and b2 . The function b20 ( t) is the Bzier curve
for the points b0 , b1 , b2 . Note from the last formula that the curve is a parabola that goes
through b0 (when t = 0) and b2 (when t = 1).
As an example, let b0 = (0, 0, 0), b1 = (1, 2, 3), and b2 = (4, 5, 2). Then the explicit formula for
the Bzier curve is b20 ( t) = (2 t + 2 t2 , 4 t + t2 , 6 t 4 t2 ), as shown in Figure 1.8.4, where the line
segments are b10 ( t) and b11 ( t), and the curve is b20 ( t).
(1, 2, 3)
2.5
2
(4, 5, 2)
1.5
z
1 (0, 0, 0)
0
0.5 0.5
1
1.5
2 x
0 0 2.5
1 2 3
3 4 3.5
5 4
y
Figure 1.8.4 Bzier curve approximation for three points

In general, the polygonal path determined by n 3 noncollinear points in R3 can be used

to define the Bzier curve recursively by a process called repeated linear interpolation. This
curve will be a vector-valued function whose components are polynomials of degree n 1,
and its formula is given by de Casteljaus algorithm.15 In the exercises, the reader will be
given the algorithm for the case of n = 4 points and asked to write the explicit formula for
the Bzier curve for the four points shown in Figure 1.8.5.
2
(4, 5, 2)
(0, 1, 1)
1.5
1
z
(0, 0, 0)
0.5
0
0.5
(2, 3, 0) 1
1.5
2 x
0 0 2.5
1 2 3
3 4 3.5
5 4
y
Figure 1.8.5 Bzier curve approximation for four points
Exercises
A
For Exercises 1-4, calculate f ( t) and find the tangent line at f(0).
2
1. f( t) = ( t + 1, t2 + 1, t3 + 1) 2. f( t) = ( e t + 1, e2 t + 1, e t + 1)
3. f( t) = (cos 2 t, sin 2 t, t) 4. f( t) = (sin 2 t, 2 sin2 t, 2 cos t)
For Exercises 5-6, find the velocity v( t) and acceleration a( t) of an object with the given
position vector r( t).
5. r( t) = ( t, t sin t, 1 cos t) 6. r( t) = (3 cos t, 2 sin t, 1)
B
cos t sin t at
7. Let f( t) = p ,p ,p , with a 6= 0.
1 + a2 t2 1 + a2 t2 1 + a2 t2
(a) Show that k f( t) k = 1 for all t.
(b) Show directly that f ( t) f( t) = 0 for all t.
8. If f ( t) = 0 for all t in some interval (a, b), show that f( t) is a constant vector in (a, b).
15 See pp. 27-30 in FARIN.
9. For a constant vector c 6= 0, the function f( t) = tc represents a line parallel to c.

(a) What kind of curve does g( t) = t3 c represent? Explain.
(b) What kind of curve does h( t) = e t c represent? Explain.
(c) Compare f (0) and g (0). Given your answer to part (a), how do you explain the
difference in the two derivatives?
d df d2f
10. Show that f =f 2.
dt dt dt
11. Let a particle of (constant) mass m have position vector r( t), velocity v( t), acceleration
a( t) and momentum p( t) at time t. The angular momentum L( t) of the particle with
respect to the origin at time t is defined as L( t) = r( t) p( t). If F( t) is the force acting on
the particle at time t, then define the torque N( t) acting on the particle with respect to
the origin as N( t) = r( t) F( t). Show that L ( t) = N( t).
d df dg dh
12. Show that (f (g h)) = (g h) + f h + f g .
dt dt dt dt
13. The Mean Value Theorem does not hold for vector-valued functions: Show that for f( t) =
(cos t, sin t, t), there is no t in the interval (0, 2) such that
f(2) f(0)
f ( t) = .
2 0
C
14. The Bzier curve b30 ( t) for four noncollinear points b0 , b1 , b2 , b3 in R3 is defined by the
following algorithm (going from the left column to the right):
b10 ( t) = (1 t)b0 + tb1 b20 ( t) = (1 t)b10 ( t) + tb11 ( t) b30 ( t) = (1 t)b20 ( t) + tb21 ( t)
b11 ( t) = (1 t)b1 + tb2 b21 ( t) = (1 t)b11 ( t) + tb12 ( t)
b12 ( t) = (1 t)b2 + tb3
(a) Show that b30 ( t) = (1 t)3 b0 + 3 t(1 t)2 b1 + 3 t2 (1 t)b2 + t3 b3 .
(b) Write the explicit formula (as in Example 1.40) for the Bzier curve for the points
b0 = (0, 0, 0), b1 = (0, 1, 1), b2 = (2, 3, 0), b3 = (4, 5, 2).
15. Let r( t) be the position vector for a particle moving in R3 . Show that
d
(r (v r)) = k r k2 a + (r v)v (k v k2 + r a)r.
dt
16. Let r( t) be the position vector in R3 for a particle that moves with constant speed c > 0
in a circle of radius a > 0 in the x y-plane. Show that a( t) points in the opposite direction
as r( t) for all t. (Hint: Use Example 1.37 to show that r( t) v( t) and a( t) v( t), and hence
a( t) r( t).)
17. Prove Theorem 1.20(g).

1.9 Arc Length 59
1.9 Arc Length

Let r( t) = ( x( t), y( t), z( t)) be the position vector of an object moving in R3 . Since k v( t) k is the
speed of the object at time t, it seems natural to define the distance s traveled by the object
from time t = a to t = b as the definite integral
Zb Zb q
s= k v( t) k dt = x ( t)2 + y ( t)2 + z ( t)2 dt , (1.40)
a a
which is analogous to the case from single-variable calculus for parametric functions in R2 .
This is indeed how we will define the distance traveled and, in general, the arc length of a
curve in R3 .
Definition 1.13. Let f( t) = ( x( t), y( t), z( t)) be a curve in R3 whose domain includes the inter-
val [a, b]. Suppose that in the interval (a, b) the first derivative of each component function
x( t), y( t) and z( t) exists and is continuous, and that no section of the curve is repeated. Then
the arc length L of the curve from t = a to t = b is
Zb Zb q

L= k f ( t) k dt = x ( t)2 + y ( t)2 + z ( t)2 dt (1.41)
a a
A real-valued function whose first derivative is continuous is called continuously differ-

entiable (or a C 1 function), and a function whose derivatives of all orders are continuous
is called smooth (or a C function). All the functions we will consider will be smooth. A
smooth curve f( t) is one whose derivative f ( t) is never the zero vector and whose component
functions are all smooth.
Note that we did not prove that the formula in the above definition actually gives the
length of a section of a curve. A rigorous proof requires dealing with some subtleties, nor-
mally glossed over in calculus texts, which are beyond the scope of this book.16
Example 1.41. Find the length L of the helix f( t) = (cos t, sin t, t) from t = 0 to t = 2.
Solution: By formula (1.41), we have
Z2 p Z2 p Z2 p
2
L= ( sin t)2 + (cos t)2 + 12 dt = 2
sin t + cos t + 1 dt = 2 dt
0 0 0
p p
= 2(2 0) = 2 2
Similar to the case in R2 , if there are values of t in the interval [a, b] where the derivative
of a component function is not continuous then it is often possible to partition [a, b] into
subintervals where all the component functions are continuously differentiable (except at
the endpoints, which can be ignored). The sum of the arc lengths over the subintervals will
be the arc length over [a, b].
16 In particular, Duhamels principle is needed. See the proof in T AYLOR and M ANN, 14.2 and 18.2.
Notice that the curve traced out by the function f( t) = (cos t, sin t, t) from Example 1.41 is
also traced out by the function g( t) = (cos 2 t, sin 2 t, 2 t). For example, over the interval [0, ],
g( t) traces out the same section of the curve as f( t) does over the interval [0, 2]. Intuitively,
this says that g( t) traces the curve twice as fast as f( t). This makes sense since, viewing the
functions as position p vectors and their p derivatives as velocity vectors, the speeds of f( t) and
g( t) are k f ( t) k = 2 and k g ( t) k = 2 2, respectively. We say that g( t) and f( t) are different
parametrizations of the same curve.
Definition 1.14. Let C be a smooth curve in R3 represented by a function f( t) defined on an

interval [a, b], and let : [ c, d ] [a, b] be a smooth one-to-one mapping of an interval [ c, d ]
onto [a, b]. Then the function g : [ c, d ] R3 defined by g( s) = f(( s)) is a parametrization of
C with parameter s. If is strictly increasing on [ c, d ] then we say that g( s) is equivalent
to f( t).
s t f( t)
f
[ c, d ] [a, b] R3
g( s) = f(( s)) = f( t)
Note that the differentiability of g( s) follows from a version of the Chain Rule for vector-
valued functions (the proof is left as an exercise):
Theorem 1.21. Chain Rule: If f( t) is a differentiable vector-valued function of t, and t =

( s) is a differentiable scalar function of s, then f( s) = f(( s)) is a differentiable vector-valued
function of s, and
d f d f dt
= (1.42)
ds dt ds
for any s where the composite function f(( s)) is defined.
Example 1.42. The following are all equivalent parametrizations of the same curve:
f( t) = (cos t, sin t, t) for t in [0, 2]

g( s) = (cos 2 s, sin 2 s, 2 s) for s in [0, ]
h( s) = (cos 2 s, sin 2 s, 2 s) for s in [0, 1]
To see that g( s) is equivalent to f( t), define : [0, ] [0, 2] by ( s) = 2 s. Then is smooth,

one-to-one, maps [0, ] onto [0, 2], and is strictly increasing (since ( s) = 2 > 0 for all s).
Likewise, defining : [0, 1] [0, 2] by ( s) = 2 s shows that h( s) is equivalent to f( t).
1.9 Arc Length 61
A curve can have many parametrizations, with different speeds, so which one is the best
to use? In some situations the arc length parametrization can be useful. The idea behind
this is to replace the parameter t, for any given smooth parametrization f( t) defined on [a, b],
by the parameter s given by Z t
s = s( t) = k f ( u) k du. (1.43)
a
In terms of motion along a curve, s is the distance traveled along the curve after time t
has elapsed. So the new parameter will be distance instead of time. There is a natural
correspondence between s and t: from a starting point on the curve, the distance traveled
along the curve (in one direction) is uniquely determined by the amount of time elapsed, and
vice versa.
Since s is the arc length of the curve over the interval [a, t] for each t in [a, b], then it is a
function of t. By the Fundamental Theorem of Calculus, its derivative is
Z
ds d t
s ( t) = = k f ( u) k du = k f ( t) k for all t in [a, b].
dt dt a
Since f( t) is smooth, then k f ( t) k > 0 for all t in [a, b]. Thus s ( t) > 0 and hence s( t) is strictly
increasing on the interval [a, b]. Recall that this means that s is a one-to-one mapping of the
interval [a, b] onto the interval [ s(a), s( b)]. But we see that
Za Zb
s( a) = k f ( u) k du = 0 and s( b) = k f ( u) k du = L = arc length from t = a to t = b
a a
So the function s : [a, b] [0, L] is a one-to-one, differentiable ( s )

mapping onto the interval [0, L]. From single-variable calculus, s t
we know that this means that there exists an inverse function [0, L] [a, b]
: [0, L] [a, b] that is differentiable and the inverse of s : [a, b]
s( t)
[0, L]. That is, for each t in [a, b] there is a unique s in [0, L] such
that s = s( t) and t = ( s). And we know that the derivative of is Figure 1.9.1 t = (s)
1 1
( s) = =
s (( s)) k f (( s)) k
So define the arc length parametrization f : [0, L] R3 by
f( s) = f(( s)) for all s in [0, L].
Then f( s) is smooth, by the Chain Rule. In fact, f( s) has unit speed:
f ( s) = f (( s)) ( s) by the Chain Rule, so

1
= f (( s)) , so
k f (( s)) k
k f ( s) k = 1 for all s in [0, L].
So the arc length parametrization traverses the curve at a normal rate.

In practice, parametrizing a curve f( t) by arc length requires you to evaluate the integral
Rt
s = a k f ( u) k du in some closed form (as a function of t) so that you could then solve for t in
terms of s. If that can be done, you would then substitute the expression for t in terms of s
(which we called ( s)) into the formula for f( t) to get f( s).
Example 1.43. Parametrize the helix f( t) = (cos t, sin t, t), for t in [0, 2], by arc length.
Solution: By Example 1.41 and formula (1.43), we have
Zt Zt p p

s= k f ( u) k du = 2 du = 2 t for all t in [0, 2].
0 0
s
So we can solve for t in terms of s: t = ( s) = p .
2
s s s p
f(s) = cos p , sin p , p for all s in [0, 2 2]. Note that k f (s) k = 1.
2 2 2
Arc length plays an important role when discussing curvature and moving frame fields,
in the field of mathematics known as differential geometry.17 The methods involve using
an arc length parametrization, which often leads to an integral that is either difficult or
impossible to evaluate in a simple closed form. The simple integral in Example 1.43 is
the exception, not the norm. In general, arc length parametrizations are more useful for
theoretical purposes than for practical computations.18 Curvature and moving frame fields
can be defined without using arc length, which makes their computation much easier, and
these definitions can be shown to be equivalent to those using arc length. We will leave this
to the exercises.
The arc length for curves given in other coordinate systems can also be calculated:
Theorem 1.22. Suppose that r = r ( t), = ( t) and z = z( t) are the cylindrical coordinates of
a curve f( t), for t in [a, b]. Then the arc length L of the curve over [a, b] is
Zb p
L= r ( t)2 + r ( t)2 ( t)2 + z ( t)2 dt (1.44)
a
Proof: The Cartesian coordinates ( x( t), y( t), z( t)) of a point on the curve are given by
x( t) = r ( t) cos ( t), y( t) = r ( t) sin ( t), z ( t) = z ( t)
so differentiating the above expressions for x( t) and y( t) with respect to t gives
x ( t) = r ( t) cos ( t) r ( t) ( t) sin ( t), y ( t) = r ( t) sin ( t) + r ( t) ( t) cos ( t)

17 See ON EILL for an introduction to elementary differential geometry.
18 For example, the usual parametrizations of Bzier curves, which we discussed in Section 1.8, are polynomial
functions in R3 . This makes their computation relatively simple, which, in CAD, is desirable. But their arc
length parametrizations are not only not polynomials, they are in fact usually impossible to calculate at all.
1.9 Arc Length 63
and so
x ( t)2 + y ( t)2 = ( r ( t) cos ( t) r ( t) ( t) sin ( t))2 + ( r ( t) sin ( t) + r ( t) ( t) cos ( t))2

= r ( t)2 (cos2 + sin2 ) + r ( t)2 ( t)2 (cos2 + sin2 )
2 r ( t) r ( t) ( t) cos sin + 2 r ( t) r ( t) ( t) cos sin
= r ( t)2 + r ( t)2 ( t)2 , and so
Zb q
L= x ( t)2 + y ( t)2 + z ( t)2 dt
a
Zb p
= r ( t)2 + r ( t)2 ( t)2 + z ( t)2 dt QED
a
Example 1.44. Find the arc length L of the curve whose cylindrical coordinates are r = e t ,
= t and z = e t , for t over the interval [0, 1].
Solution: Since r ( t) = e t , ( t) = 1 and z ( t) = e t , then
Z1 p
L= r ( t)2 + r ( t)2 ( t)2 + z ( t)2 dt
0
Z1 p
= e2 t + e2 t (1) + e2 t dt
0
Z1 p p
= e t 3 dt = 3( e 1)
0
Exercises
A
For Exercises 1-3, calculate the arc length of f( t) over the given interval.
1. f( t) = (3 cos 2 t, 3 sin 2 t, 3 t) on [0, /2]

p
2. f( t) = (( t2 + 1) cos t, ( t2 + 1) sin t, 2 2 t) on [0, 1]
3. f( t) = (2 cos 3 t, 2 sin 3 t, 2 t3/2 ) on [0, 1]
4. Parametrize the curve from Exercise 1 by arc length.
5. Parametrize the curve from Exercise 3 by arc length.
B
6. Let f( t) be a differentiable curve such that f( t) 6= 0 for all t. Show that

d f( t) f( t) (f ( t) f( t))
= .
dt f( t) k f( t) k3
Exercises 7-9 develop the moving frame field T, N, B at a point on a curve.
7. Let f( t) be a smooth curve such that f ( t) 6= 0 for all t. Then we can define the unit tangent
vector T by
f ( t)
T( t) = .
k f ( t) k
Show that
f ( t) (f ( t) f ( t))
T ( t) = .
k f ( t ) k3
8. Continuing Exercise 7, assume that f ( t) and f ( t) are not parallel. Then T ( t) 6= 0 so we

can define the unit principal normal vector N by
T ( t)
N( t) = .
k T ( t) k
Show that
f ( t) (f ( t) f ( t))
N( t) = .
k f ( t) k k f ( t) f ( t) k
9. Continuing Exercise 8, the unit binormal vector B is defined by
B( t) = T( t) N( t).
Show that
f ( t) f ( t)
B( t) = .
k f ( t) f ( t) k
Note: The vectors T( t), N( t) and B( t) form a right-handed system of mutually perpendic-
ular unit vectors (called orthonormal vectors) at each point on the curve f( t).
10. Continuing Exercise 9, the curvature is defined by
k T ( t) k k f ( t) (f ( t) f ( t)) k
( t) = = .
k f ( t) k k f ( t ) k4
Show that
k f ( t) f ( t) k
( t) = 3
and that T ( t) = k f ( t) k ( t) N( t).
k f ( t) k
Note: ( t) gives a sense of how curved the curve f( t) is at each point.
11. Find T, N, B and at each point of the helix f( t) = (cos t, sin t, t).
12. Show that the arc length L of a curve whose spherical coordinates are = ( t), = ( t)
and = ( t) for t in an interval [a, b] is
Zb q
L= ( t)2 + ( ( t)2 sin2 ( t)) ( t)2 + ( t)2 ( t)2 dt.
a
2 Functions of Several Variables
2.1 Functions of Two or Three Variables
In Section 1.8 we discussed vector-valued functions of a single real variable. We will now
examine real-valued functions of a point (or vector) in R2 or R3 . For the most part these
functions will be defined on sets of points in R2 , but there will be times when we will use
points in R3 , and there will also be times when it will be convenient to think of the points as
vectors (or terminal points of vectors).
A real-valued function f defined on a subset D of R2 is a rule that assigns to each point
( x, y) in D a real number f ( x, y). The largest possible set D in R2 on which f is defined is
called the domain of f , and the range of f is the set of all real numbers f ( x, y) as ( x, y)
varies over the domain D . A similar definition holds for functions f ( x, y, z) defined on points
( x, y, z) in R3 .
Example 2.1. The domain of the function
f ( x, y) = x y
is all of R2 , and the range of f is all of R.
1
f ( x, y) =
x y
is all of R2 except the points ( x, y) for which x = y. That is, the domain is the set D = {( x, y) :
x 6= y}. The range of f is all real numbers except 0.

q
f ( x, y) = 1 x2 y2
is the set D = {( x, y) : x2 + y2 1}, since the quantity inside the square root is nonnegative if
and only if 1 ( x2 + y2 ) 0. We see that D consists of all points on and inside the unit circle
in R2 (D is sometimes called the closed unit disk). The range of f is the interval [0, 1] in R.
65
66 CHAPTER 2. FUNCTIONS OF SEVERAL VARIABLES
f ( x, y, z) = e x+ y z
is all of R3 , and the range of f is all positive real numbers.
A function f ( x, y) defined in R2 is often written as z = f ( x, y), as was mentioned in Section

1.1, so that the graph of f ( x, y) is the set {( x, y, z) : z = f ( x, y)} in R3 . So we see that this
graph is a surface in R3 , since it satisfies an equation of the form F ( x, y, z) = 0 (namely,
F ( x, y, z) = f ( x, y) z). The traces of this surface in the planes z = c, where c varies over R,
are called the level curves of the function. Equivalently, the level curves are the solution
sets of the equations f ( x, y) = c, for c in R. Level curves are often projected onto the x y-plane
to give an idea of the various elevation levels of the surface (as is done in topography).
Example 2.5. The graph of the function

p
sin x2 + y2
f ( x, y) = p
x 2 + y2
is shown below. Note that the level curves (shown both on the surface and projected onto the
x y-plane) are groups of concentric circles.
1
0.8
0.6
0.4
0.2
z
0
-0.2 -10
-5
-0.4
-10 0
-5 x
0 5
y 5
10 10
p
sin x2 + y2
Figure 2.1.1 The function f (x, y) = p 2 2
x +y
You may be wondering what happens to the function in Example 2.5 at the point ( x, y) =
(0, 0), since both the numerator and denominator are 0 at that point. The function is not
defined at (0, 0), but the limit of the function exists (and equals 1) as ( x, y) approaches (0, 0).
We will now state explicitly what is meant by the limit of a function of two variables.
2.1 Functions of Two or Three Variables 67
Definition 2.1. Let (a, b) be a point in R2 , and let f ( x, y) be a real-valued function defined
on some set containing (a, b) (but not necessarily defined at (a, b) itself). Then we say that
the limit of f ( x, y) equals L as ( x, y) approaches (a, b), written as
lim f ( x, y) = L , (2.1)
( x,y)(a,b)
if given any > 0, there exists a > 0 such that

q
| f ( x, y) L | < whenever 0 < ( x a)2 + ( y b)2 < .
A similar definition can be made for functions of three variables. The idea behind the
above definition is that the values of f ( x, y) can get arbitrarily close to L (i.e. within of
L) if we pick ( x, y) sufficiently close to (a, b) (i.e. inside a circle centered at (a, b) with some
sufficiently small radius ).
If you recall the epsilon-delta proofs of limits of real-valued functions of a single variable,
you may remember how awkward they can be, and how they can usually only be done easily
for simple functions. In general, the multivariable cases are at least equally awkward to go
through, so we will not bother with such proofs. Instead, we will simply state that when the
function f ( x, y) is given by a single formula and is defined at the point (a, b) (e.g. is not some
indeterminate form like 0/0) then you can just substitute ( x, y) = (a, b) into the formula for
f ( x, y) to find the limit.
Example 2.6.
xy (1)(2) 2
lim = =
( x,y)(1,2) x2 + y2 2
1 +2 2 5
xy
since f ( x, y) = x2 + y2
is properly defined at the point (1, 2).
The major difference between limits in one variable and limits in two or more variables
has to do with how a point is approached. In the single-variable case, the statement x a
means that x gets closer to the value a from two possible directions along the real number
line (see Figure 2.1.2(a)). In two dimensions, however, ( x, y) can approach a point (a, b) along
an infinite number of paths (see Figure 2.1.2(b)).
y
x x (a, b)
x
0 a x 0
(a) x a in R (b) ( x, y) (a, b) in R2
Figure 2.1.2 Approaching a point in different dimensions

Example 2.7.
xy
lim does not exist
( x,y)(0,0) x 2 + y2
Note that we can not simply substitute ( x, y) = (0, 0) into the function, since doing so gives an
indeterminate form 0/0. To show that the limit does not exist, we will show that the function
approaches different values as ( x, y) approaches (0, 0) along different paths in R2 . To see this,
suppose that ( x, y) (0, 0) along the positive x-axis, so that y = 0 along that path. Then
xy x0
f ( x, y) = = =0
x 2 + y2 x 2 + 02
along that path (since x > 0 in the denominator). But if ( x, y) (0, 0) along the straight line
y = x through the origin, for x > 0, then we see that
xy x2 1
f ( x, y) = 2 2
= 2 2
=
x +y x +x 2
which means that f ( x, y) approaches different values as ( x, y) (0, 0) along different paths.
Hence the limit does not exist.
Limits of real-valued multivariable functions obey the same algebraic rules as in the
single-variable case, as shown in the following theorem, which we state without proof.
Theorem 2.1. Suppose that lim f ( x, y) and lim g( x, y) both exist, and that k is
( x,y)(a,b) ( x,y)(a,b)
some scalar. Then:
h i h i
(a) lim [ f ( x, y) g( x, y)] = lim f ( x, y) lim g( x, y)
( x,y)(a,b) ( x,y)(a,b) ( x,y)(a,b)
h i
(b) lim k f ( x, y) = k lim f ( x, y)
( x,y)(a,b) ( x,y)(a,b)
h ih i
(c) lim [ f ( x, y) g( x, y)] = lim f ( x, y) lim g( x, y)
( x,y)(a,b) ( x,y)(a,b) ( x,y)(a,b)
lim f ( x, y)
f ( x, y) ( x,y)(a,b)
(d) lim = if lim g( x, y) 6= 0
( x,y)(a,b) g( x, y) lim g( x, y) ( x,y)(a,b)
( x,y)(a,b)
(e) If | f ( x, y) L | g( x, y) for all ( x, y) and if lim g( x, y) = 0, then lim f ( x, y) = L.

( x,y)(a,b) ( x,y)(a,b)
Note that in part (e), it suffices to have | f ( x, y) L | g( x, y) for all ( x, y) sufficiently close
to (a, b) (but excluding (a, b) itself).
2.1 Functions of Two or Three Variables 69
Example 2.8. Show that

y4
lim = 0.
( x,y)(0,0) x 2 + y2
Since substituting ( x, y) = (0, 0) into the function gives the indeterminate form 0/0, we need
an alternate method for evaluating this limit. We will use Theorem 2.1(e). First, notice that
p 4 p 4 p 4
y4 = y2 and so 0 y4 x2 + y2 for all ( x, y). But x2 + y2 = ( x2 + y2 )2 . Thus, for
all ( x, y) 6= (0, 0) we have

y4 ( x2 + y2 )2 2 2
x2 + y2 x2 + y2 = x + y 0 as ( x, y) (0, 0).

y4
Therefore lim = 0.
( x,y)(0,0) x2 + y2
Continuity can be defined similarly as in the single-variable case.
Definition 2.2. A real-valued function f ( x, y) with domain D in R2 is continuous at the

point (a, b) in D if lim f ( x, y) = f (a, b). We say that f ( x, y) is a continuous function if
( x,y)(a,b)
it is continuous at every point in its domain D .
Unless indicated otherwise, you can assume that all the functions we deal with are con-
tinuous. In fact, we can modify the function from Example 2.8 so that it is continuous on all
of R2 .
Example 2.9. Define a function f ( x, y) on all of R2 as follows:

0
if ( x, y) = (0, 0)
f ( x, y) = 4
y

2 if ( x, y) 6= (0, 0)
x + y2
Then f ( x, y) is well-defined for all ( x, y) in R2 (i.e. there are no indeterminate forms for any
( x, y)), and we see that
b4
lim f ( x, y) = = f (a, b) for (a, b) 6= (0, 0).
( x,y)(a,b) a2 + b 2
So since
lim f ( x, y) = 0 = f (0, 0) by Example 2.8,
( x,y)(0,0)
then f ( x, y) is continuous on all of R2 .

Exercises
A
For Exercises 1-6, state the domain and range of the given function.
1
1. f ( x, y) = x2 + y2 1 2. f ( x, y) =
x 2 + y2
p x2 + 1
3. f ( x, y) = x 2 + y2 4 4. f ( x, y) =
y
p
5. f ( x, y, z) = sin( x yz) 6. f ( x, y, z) = ( x 1)( yz 1)
For Exercises 7-18, evaluate the given limit.
7. lim cos( x y) 8. lim ex y

( x,y)(0,0) ( x,y)(0,0)
x 2 y2 x y2
9. lim 10. lim
( x,y)(0,0) x 2 + y2 ( x,y)(0,0) x 2 + y4
2 2
x 2x y + y x y2
11. lim 12. lim
( x,y)(1,1) x y ( x,y)(0,0) x 2 + y2
2 2
x y x 2 2 x y + y2
13. lim 14. lim
( x,y)(1,1) x y ( x,y)(0,0) x y

y4 sin( x y) 2 2 1
15. lim 16. lim ( x + y ) cos
( x,y)(0,0) x 2 + y2 ( x,y)(0,0) xy
x
17. lim 1
( x,y)(0,0) y 18. lim cos
( x,y)(0,0) xy
B
1 2 2 2
( x + y )/2
19. Show that f ( x, y) = 2 2 e , for > 0, is constant on the circle of radius r > 0
centered at the origin. This function is called a Gaussian blur, and is used as a filter in
image processing software to produce a blurred effect.
20. Suppose that f ( x, y) f ( y, x) for all ( x, y) in R2 . Show that f ( x, y) = f ( y, x) for all ( x, y) in

R2 .
p
21. Use the substitution r = x2 + y2 to show that
p
sin x2 + y2
lim p =1 .
( x,y)(0,0) x 2 + y2
(Hint: You will need to use LHpitals Rule for single-variable limits.)
C
22. Prove Theorem 2.1(a) in the case of addition. (Hint: Use Definition 2.1.)
23. Prove Theorem 2.1(b).

2.2 Partial Derivatives 71
2.2 Partial Derivatives

Now that we have an idea of what functions of several variables are, and what a limit of
such a function is, we can start to develop an idea of a derivative of a function of two or more
variables. We will start with the notion of a partial derivative.
Definition 2.3. Let f ( x, y) be a real-valued function with domain D in R2 , and let (a, b) be
a point in D . Then the partial derivative of f at (a, b) with respect to x, denoted by
f
(a, b), is defined as
x
f f (a + h, b) f (a, b)
(a, b) = lim (2.2)
x h0 h
f
and the partial derivative of f at (a, b) with respect to y, denoted by (a, b), is defined
y
as
f f (a, b + h) f (a, b)
(a, b) = lim . (2.3)
y h0 h
Note: The symbol is pronounced del.1
Recall that the derivative of a function f ( x) can be interpreted as the rate of change of
that function in the (positive) x direction. From the definitions above, we can see that the
partial derivative of a function f ( x, y) with respect to x or y is the rate of change of f ( x, y) in
the (positive) x or y direction, respectively. What this means is that the partial derivative of
a function f ( x, y) with respect to x can be calculated by treating the y variable as a constant,
and then simply differentiating f ( x, y) as if it were a function of x alone, using the usual
rules from single-variable calculus. Likewise, the partial derivative of f ( x, y) with respect to
y is obtained by treating the x variable as a constant and then differentiating f ( x, y) as if it
were a function of y alone.
f f
Example 2.10. Find ( x, y) for the function f ( x, y) = x2 y + y3 .
( x, y) and
x y
Solution: Treating y as a constant and differentiating f ( x, y) with respect to x gives
f
( x, y) = 2 x y
x
and treating x as a constant and differentiating f ( x, y) with respect to y gives
f
( x, y) = x2 + 3 y2 .
y
1 It is not a Greek letter. The symbol was first used by the mathematicians A. Clairaut and L. Euler around
1740, to distinguish it from the letter d used for the usual derivative.
f f f f
We will often simply write and instead of ( x, y) and ( x, y).
x y x y
f f sin( x y2 )
Example 2.11. Find and .
for the function f ( x, y) =
x y x2 + 1
Solution: Treating y as a constant and differentiating f ( x, y) with respect to x gives
f ( x2 + 1)( y2 cos( x y2 )) (2 x) sin( x y2 )

=
x ( x2 + 1)2
and treating x as a constant and differentiating f ( x, y) with respect to y gives
f 2 x y cos( x y2 )
= .
y x2 + 1
f f
Since both and are themselves functions of x and y, we can take their partial
x y
derivatives with respect to x and y. This yields the higher-order partial derivatives:
2 f f 2 f f
= =
x2 x x y2 y y
2 2
f f f f
= =
y x y x x y x y
3
f 2 f 3 f 2 f
= =
x3 x x2 y3 y y2
3
f 2 f 3 f 2 f
= =
y x2 y x2 x y2 x y2
3 f 2 f 3 f 2 f
= =
y2 x y y x x2 y x x y
3
f 2 f 3 f 2 f
= =
x y x x y x y x y y x y
..
.
f f 2 f 2 f 2 f 2 f
Example 2.12. Find the partial derivatives , , 2
, 2
, and for the
x y x y y x x y
2
function f ( x, y) = e x y + x y3 .
2.2 Partial Derivatives 73
Solution: Proceeding as before, we have
f 2 f 2
= 2 x ye x y
+ y3 = x2 e x y
+ 3 x y2
x y
2 f 2 2 f 2
= (2 x ye x y + y3 ) = ( x 2 e x y + 3 x y2 )
x2 x y2 y
2 2 2
= 2 ye x y
+ 4 x 2 y2 e x y
= x4 e x y
+ 6x y
2 f 2 2 f 2
= (2 x ye x y + y3 ) = ( x 2 e x y + 3 x y2 )
y x y x y x
2 2 2 2
= 2 xe x y
+ 2 x3 ye x y
+ 3 y2 = 2 xe x y
+ 2 x3 ye x y
+ 3 y2
Higher-order partial derivatives that are taken with respect to different variables, such
2 f 2 f
as y x and x y , are called mixed partial derivatives. Notice in the above example that
2 f 2 f 2 f
y x
= x y
. It turns that this will usually be the case. Specifically, whenever both y x
and
2 f
are continuous at a point (a, b), then they are equal at that point.2 All the functions
x y
we will deal with will have continuous partial derivatives of all orders, so you can assume in
the remainder of the text that
2 f 2 f
= for all ( x, y) in the domain of f .
y x x y
In other words, it doesnt matter in which order you take partial derivatives. This applies
even to mixed partial derivatives of order 3 or higher.
The notation for partial derivatives varies. All of the following are equivalent:
f
: f x ( x, y) , f 1 ( x, y) , D x ( x, y) , D 1 ( x, y)
x
f
: f y ( x, y) , f 2 ( x, y) , D y ( x, y) , D 2 ( x, y)
y
2 f
: f xx ( x, y) , f 11 ( x, y) , D xx ( x, y) , D 11 ( x, y)
x2
2 f
: f yy ( x, y) , f 22 ( x, y) , D yy ( x, y) , D 22 ( x, y)
y2
2 f
: f x y ( x, y) , f 12 ( x, y) , D x y ( x, y) , D 12 ( x, y)
y x
2 f
: f yx ( x, y) , f 21 ( x, y) , D yx ( x, y) , D 21 ( x, y)
x y
2 See pp. 214-216 in T AYLOR and M ANN for a proof.
Exercises
A
f f
For Exercises 1-16, find x
and y
.
1. f ( x, y) = x2 + y2 2. f ( x, y) = cos( x + y)
p
3. f ( x, y) = x2 + y + 4 x+1
4. f ( x, y) =
y+1
5. f ( x, y) = e x y + x y 6. f ( x, y) = x2 y2 + 6 x y + 4 x 8 y + 2
7. f ( x, y) = x4 8. f ( x, y) = x + 2 y
p
9. f ( x, y) = x2 + y2 10. f ( x, y) = sin( x + y)
p
11. f ( x, y) = 3 x2 + y + 4 xy+1
12. f ( x, y) =
x+ y
2
+ y2 )
13. f ( x, y) = e( x 14. f ( x, y) = ln( x y)
15. f ( x, y) = sin( x y) 16. f ( x, y) = tan( x + y)

2 f 2 f 2 f
For Exercises 17-26, find ,
x2 y2
and y x
(use Exercises 1-8, 14, 15).
17. f ( x, y) = x2 + y2 18. f ( x, y) = cos( x + y)

p
19. f ( x, y) = x2 + y + 4 x+1
20. f ( x, y) =
y+1
21. f ( x, y) = e x y + x y 22. f ( x, y) = x2 y2 + 6 x y + 4 x 8 y + 2
23. f ( x, y) = x4 24. f ( x, y) = x + 2 y
25. f ( x, y) = ln( x y) 26. f ( x, y) = sin( x y)
B
27. Show that the function f ( x, y) = sin( x + y) + cos( x y) satisfies the wave equation
2 f 2 f
=0 .
x2 y2
The wave equation is an example of a partial differential equation.
28. Let u and v be twice-differentiable functions of a single variable, and let c 6= 0 be a con-
stant. Show that f ( x, y) = u( x + c y) + v( x c y) is a solution of the general one-dimensional
wave equation3
2 f 1 2 f
=0 .
x 2 c 2 y2
3 Conversely, it turns out that any solution must be of this form. See Ch. 1 in W EINBERGER.
2.3 Tangent Plane to a Surface 75
2.3 Tangent Plane to a Surface

f f
In the previous section we mentioned that the partial derivatives x and y can be thought
of as the rate of change of a function z = f ( x, y) in the positive x and y directions, respectively.
dy
Recall that the derivative dx of a function y = f ( x) has a geometric meaning, namely as the
slope of the tangent line to the graph of f at the point ( x, f ( x)) in R2 . There is a similar
f f
geometric meaning to the partial derivatives x and y of a function z = f ( x, y): given a
point (a, b) in the domain D of f ( x, y), the trace of the surface described by z = f ( x, y) in the
plane y = b is a curve in R3 through the point (a, b, f (a, b)), and the slope of the tangent line
f f
L x to that curve at that point is x (a, b). Similarly, y (a, b) is the slope of the tangent line
L y to the trace of the surface z = f ( x, y) in the plane x = a (see Figure 2.3.1).
z z f
z = f (x, y) slope = y
(a, b)
f
(a, b, f (a, b)) slope = x
(a, b) z = f (x, y)
(a, b, f (a, b))
Ly
Lx
b y y
0 0
a
(a, b) (a, b)
x x
D D
(a) Tangent line L x in the plane y = b (b) Tangent line L y in the plane x = a
Figure 2.3.1 Partial derivatives as slopes
dy
Since the derivative dx of a function y = f ( x) is used to find the tangent line to the graph
of f (which is a curve in R2 ), you might expect that partial derivatives can be used to define
a tangent plane to the graph of a surface z = f ( x, y). This indeed turns out to be the case.
First, we need a definition of a tangent plane. The intuitive idea is that a tangent plane just
touches a surface at a point. The formal definition mimics the intuitive notion of a tangent
line to a curve.
Definition 2.4. Let z = f ( x, y) be the equation of a surface S in R3 , and let P = (a, b, c) be

a point on S . Let T be a plane which contains the point P , and let Q = ( x, y, z) represent a

generic point on the surface S . If the (acute) angle between the vector PQ and the plane
T approaches zero as the point Q approaches P along the surface S , then we call T the
tangent plane to S at P .
Note that since two lines in R3 determine a plane, then the two tangent lines to the surface
z = f ( x, y) in the x and y directions described in Figure 2.3.1 are contained in the tangent
plane at that point, if the tangent plane exists at that point. The existence of those two
tangent lines does not by itself guarantee the existence of the tangent plane. It is possible
that if we take the trace of the surface in the plane x y = 0 (which makes a 45 angle with
the positive x-axis), the resulting curve in that plane may have a tangent line which is not
in the plane determined by the other two tangent lines, or it may not have a tangent line
f f
at all at that point. Luckily, it turns out4 that if x and y exist in a region around a point
(a, b) and are continuous at (a, b) then the tangent plane to the surface z = f ( x, y) will exist
at the point (a, b, f (a, b)). In this text, those conditions will always hold.
Suppose that we want an equation of the tangent plane T z
to the surface z = f ( x, y) at a point (a, b, f (a, b)). Let L x and z = f (x, y)
L y be the tangent lines to the traces of the surface in the
(a, b, f (a, b)) L
planes y = b and x = a, respectively (as in Figure 2.3.2), and y
suppose that the conditions for T to exist do hold. Then the Lx
equation for T is T y
0
A ( x a) + B( y b) + C ( z f (a, b)) = 0 (2.4)
x
Figure 2.3.2 Tangent plane
where n = ( A, B, C ) is a normal vector to the plane T . Since
T contains the lines L x and L y , then all we need are vectors vx and v y that are parallel to L x
and L y , respectively, and then let n = vx v y .
f f z
Since the slope of L x is x (a, b), then the vector vx = (1, 0, x (a, b)) is f
vx = (1, 0, x (a, b))
parallel to L x (since vx lies in the xz-plane and lies in a line with slope
f
x
(a,b) f
1 =x
(a, b). See Figure 2.3.3). Similarly, the vector
f
f (a, b)
vy = (0, 1, y (a, b)) is parallel to L y . Hence, the vector x
x
0
i j k 1

f f f
n = vx v y = 1 0 x
(a, b) = x (a, b) i y (a, b) j + k Figure 2.3.3
f
0 1 (a, b)
y
is normal to the plane T . Thus the equation of T is

f f
x (a, b) ( x a) y (a, b) ( y b) + z f (a, b) = 0 . (2.5)
Multiplying both sides by 1, we have the following result:
The equation of the tangent plane to the surface z = f ( x, y) at the point (a, b, f (a, b)) is
f f
x
(a, b) ( x a) + y (a, b) ( y b) z + f (a, b) = 0 (2.6)
4 See T AYLOR and M ANN, 6.4.

2.3 Tangent Plane to a Surface 77
Example 2.13. Find the equation of the tangent plane to the surface z = x2 + y2 at the point
(1, 2, 5).
f f
Solution: For the function f ( x, y) = x2 + y2 , we have x
= 2 x and y
= 2 y, so the equation of
the tangent plane at the point (1, 2, 5) is
2(1)( x 1) + 2(2)( y 2) z + 5 = 0 , or
2x + 4 y z 5 = 0 .
In a similar fashion, it can be shown that if a surface is defined implicitly by an equation

of the form F ( x, y, z) = 0, then the tangent plane to the surface at a point (a, b, c) is given by
the equation
F
x
(a, b, c) ( x a) + Fy (a, b, c) ( y b) + Fz (a, b, c) ( z c) = 0 . (2.7)
Note that formula (2.6) is the special case of formula (2.7) where F ( x, y, z) = f ( x, y) z.
Example 2.14. Find the equation of the tangent plane to the surface x2 + y2 + z2 = 9 at the
point (2, 2, 1).
F F F
Solution: For the function F ( x, y, z) = x2 + y2 + z2 9, we have x
= 2 x, y
= 2 y, and z
= 2 z,
so the equation of the tangent plane at (2, 2, 1) is
2(2)( x 2) + 2(2)( y 2) + 2(1)( z + 1) = 0 , or

2x + 2 y z 9 = 0 .
Exercises
A
For Exercises 1-6, find the equation of the tangent plane to the surface z = f ( x, y) at the
point P .
1. f ( x, y) = x2 + y3 , P = (1, 1, 2) 2. f ( x, y) = x y, P = (1, 1, 1)
3. f ( x, y) = x2 y, P = (1, 1, 1) 4. f ( x, y) = xe y , P = (1, 0, 1)
p
5. f ( x, y) = x + 2 y, P = (2, 1, 4) 6. f ( x, y) = x2 + y2 , P = (3, 4, 5)
For Exercises 7-10, find the equation of the tangent plane to the given surface at the point
P.
p
2 y2 z2
7. x4 + 9 + 16 = 1, P = 1, 2, 2 311 8. x2 + y2 + z2 = 9, P = (0, 0, 3)
p
9. x2 + y2 z2 = 0, P = (3, 4, 5) 10. x2 + y2 = 4, P = ( 3, 1, 0)
2.4 Directional Derivatives and the Gradient

f f
For a function z = f ( x, y), we learned that the partial derivatives x and y represent the
(instantaneous) rate of change of f in the positive x and y directions, respectively. What
about other directions? It turns out that we can find the rate of change in any direction
using a more general type of derivative called a directional derivative.
Definition 2.5. Let f ( x, y) be a real-valued function with domain D in R2 , and let (a, b) be a
point in D . Let v be a unit vector in R2 . Then the directional derivative of f at (a, b) in
the direction of v, denoted by D v f (a, b), is defined as
f ((a, b) + hv) f (a, b)

D v f (a, b) = lim (2.8)
h0 h
Notice in the definition that we seem to be treating the point (a, b) as a vector, since we
are adding the vector hv to it. But this is just the usual idea of identifying vectors with their
terminal points, which the reader should be used to by now. If we were to write the vector v
as v = (v1 , v2 ), then
f (a + hv1 , b + hv2 ) f (a, b)
D v f (a, b) = lim . (2.9)
h0 h
f f
From this we can immediately recognize that the partial derivatives x and y are special
cases of the directional derivative with v = i = (1, 0) and v = j = (0, 1), respectively. That is,
f f
x
= D i f and y = D j f . Since there are many vectors with the same direction, we use a unit
vector in the definition, as that represents a standard vector for a given direction.
f f
If f ( x, y) has continuous partial derivatives x and y (which will always be the case in
this text), then there is a simple formula for the directional derivative:
Theorem 2.2. Let f ( x, y) be a real-valued function with domain D in R2 such that the
f f
partial derivatives x and y exist and are continuous in D . Let (a, b) be a point in D , and
let v = (v1 , v2 ) be a unit vector in R2 . Then
f f
D v f (a, b) = v1 (a, b) + v2 (a, b) . (2.10)
x y
f
Proof: Note that if v = i = (1, 0) then the above formula reduces to D v f (a, b) = x
(a, b),
f
which we know is true since D i f = x
, as we noted earlier. Similarly, for v = j = (0, 1) the
f f
formula reduces to D v f (a, b) = y
(a, b), which is true since D j f = y
. So since i = (1, 0) and
2
j = (0, 1) are the only unit vectors in R with a zero component, then we need only show the
formula holds for unit vectors v = (v1 , v2 ) with v1 6= 0 and v2 6= 0. So fix such a vector v and
fix a number h 6= 0.
2.4 Directional Derivatives and the Gradient 79
Then
f (a + hv1 , b + hv2 ) f (a, b) = f (a + hv1 , b + hv2 ) f (a + hv1 , b) + f (a + hv1 , b) f (a, b) . (2.11)
Since h 6= 0 and v2 6= 0, then hv2 6= 0 and thus any number c between b and b + hv2 can be
written as c = b + hv2 for some number 0 < < 1. So since the function f (a+ hv1 , y) is a real-
valued function of y (since a + hv1 is a fixed number), then the Mean Value Theorem from
single-variable calculus can be applied to the function g( y) = f (a + hv1 , y) on the interval
[ b, b + hv2 ] (or [ b + hv2 , b] if one of h or v2 is negative) to find a number 0 < < 1 such that
f g( b + hv2 ) g( b) f (a + hv1 , b + hv2 ) f (a + hv1 , b)

(a + hv1 , b + hv2 ) = g ( b + hv2 ) = =
y b + hv2 b hv2
and so
f
f (a + hv1 , b + hv2 ) f (a + hv1 , b) = hv2 (a + hv1 , b + hv2 ) .
y
By a similar argument, there exists a number 0 < < 1 such that
f
f (a + hv1 , b) f (a, b) = hv1 (a + hv1 , b) .
x
Thus, by equation (2.11), we have
f f
f (a + hv1 , b + hv2 ) f (a, b) hv2 y (a + hv1 , b + hv2 ) + hv1 x (a + hv1 , b)
=
h h
f f
= v2 (a + hv1 , b + hv2 ) + v1 (a + hv1 , b)
y x
so by formula (2.9) we have
f (a + hv1 , b + hv2 ) f (a, b)

D v f (a, b) = lim
h0 h

f f
= lim v2 (a + hv1 , b + hv2 ) + v1 (a + hv1 , b)
h0 y x
f f f f
= v2 (a, b) + v1 (a, b) by the continuity of and , so
y x x y
f f
D v f (a, b) = v1 (a, b) + v2 (a, b)
x y
after reversing the order of summation. QED

f f
Note that D v f (a, b) = v x
(a, b), y (a, b) . The second vector has a special name:
Definition 2.6. For a real-valued function f ( x, y), the gradient of f , denoted by f , is the
vector f f
f = , (2.12)
x y
in R2 . For a real-valued function f ( x, y, z), the gradient is the vector
f f f
f = , , (2.13)
x y z
in R3 . The symbol is pronounced del.5
Corollary 2.3. D v f = v f
2 3
directional derivative of f ( x, y) = x y + x y at the point (1, 2) in the
Example 2.15. Find the
direction of v = p1 , p1 .
2 2
Solution: We see that f = ( y2 + 3 x2 y, 2 x y + x3 ), so

D v f (1, 2) = v f (1, 2) = p1 , p1 (22 + 3(1)2 (2), 2(1)(2) + 13 ) = 15
p
2 2 2
f f
A real-valued function z = f ( x, y) whose partial derivatives x and y exist and are con-
tinuous is called continuously differentiable. Assume that f ( x, y) is such a function and that
f 6= 0. Let c be a real number in the range of f and let v be a unit vector in R2 which is
tangent to the level curve f ( x, y) = c (see Figure 2.4.1).
y
v f
f ( x, y) = c
x
0
Figure 2.4.1
5 Sometimes the notation grad( f ) is used instead of f .

2.4 Directional Derivatives and the Gradient 81
The value of f ( x, y) is constant along a level curve, so since v is a tangent vector to this
curve, then the rate of change of f in the direction of v is 0, i.e. D v f = 0. But we know that
D v f = v f = k v k k f k cos , where is the angle between v and f . So since k v k = 1 then
D v f = k f k cos . So since f 6= 0 then D v f = 0 cos = 0 = 90 . In other words, f v,
which means that f is normal to the level curve.
In general, for any unit vector v in R2 , we still have D v f = k f k cos , where is the angle
between v and f . At a fixed point ( x, y) the length k f k is fixed, and the value of D v f then
varies as varies. The largest value that D v f can take is when cos = 1 ( = 0 ), while the
smallest value occurs when cos = 1 ( = 180 ). In other words, the value of the function
f increases the fastest in the direction of f (since = 0 in that case), and the value of
f decreases the fastest in the direction of f (since = 180 in that case). We have thus
proved the following theorem:
Theorem 2.4. Let f ( x, y) be a continuously differentiable real-valued function, with f 6= 0.

Then:
(a) The gradient f is normal to any level curve f ( x, y) = c.
(b) The value of f ( x, y) increases the fastest in the direction of f .
(c) The value of f ( x, y) decreases the fastest in the direction of f .
Example 2.16. In which direction does the function f ( x, y) = x y2 + x3 y increase the fastest
from the point (1, 2)? In which direction does it decrease the fastest?
Solution: Since f = (y2 + 3 x2 y, 2 x y + x3 ), then f (1, 2) = (10, 5) 6= 0. A unit vector
in that
f 2 p1 2 p1
direction is v = k f k = p , . Thus, f increases the fastest in the direction of p , and
5 5 5 5

2 p
decreases the fastest in the direction of p , 1 .
5 5
Though we proved Theorem 2.4 for functions of two variables, a similar argument can
be used to show that it also applies to functions of three or more variables. Likewise, the
directional derivative in the three-dimensional case can also be defined by the formula D v f =
v f .
Example 2.17. The temperature T of a solid is given by the function T ( x, y, z) = e x + e2 y +

e4 z , where x, y, z are space coordinates relative to the center of the solid. In which direction
from the point (1, 1, 1) will the temperature decrease the fastest?
Solution: Since f = ( e x , 2 e2 y , 4 e4 z ), then the temperature will decrease the fastest in
the direction of f (1, 1, 1) = ( e1 , 2 e2 , 4 e4 ).
Exercises
A
For Exercises 1-10, compute the gradient f .
1
1. f ( x, y) = x2 + y2 1 2. f ( x, y) =
x 2 + y2
p
3. f ( x, y) = x 2 + y2 + 4 4. f ( x, y) = x2 e y
5. f ( x, y) = ln( x y) 6. f ( x, y) = 2 x + 5 y
7. f ( x, y, z) = sin( x yz) 8. f ( x, y, z) = x2 e yz
p
9. f ( x, y, z) = x2 + y2 + z2 10. f ( x, y, z) = x2 + y2 + z2
For Exercises
11-14, find the directional derivative of f at the point P in the direction of
v= p1 , p1 .
2 2
1
11. f ( x, y) = x2 + y2 1, P = (1, 1) 12. f ( x, y) = , P = (1, 1)
x 2 + y2
p
13. f ( x, y) = x2 + y2 + 4, P = (1, 1) 14. f ( x, y) = x2 e y , P = (1, 1)
For Exercises 15-16,

find the directional derivative of f at the point P in the direction of
1 1 1
v= p ,p ,p .
3 3 3
15. f ( x, y, z) = sin( x yz), P = (1, 1, 1) 16. f ( x, y, z) = x2 e yz , P = (1, 1, 1)
17. Repeat Example 2.16 at the point (2, 3).
18. Repeat Example 2.17 at the point (3, 1, 2).
B
For Exercises 19-26, let f ( x, y) and g( x, y) be continuously differentiable real-valued func-
tions, let c be a constant, and let v be a unit vector in R2 . Show that:
19. ( c f ) = c f 20. ( f + g) = f + g
g f f g
21. ( f g) = f g + g f 22. ( f / g) = if g( x, y) 6= 0
g2
23. D v f = D v f 24. D v ( c f ) = c D v f
25. D v ( f + g) = D v f + D v g 26. D v ( f g) = f D v g + g D v f
p
27. The function r ( x, y) = x2 + y2 is the length of the position vector r = x i + y j for each
1
point ( x, y) in R2 . Show that r = r when ( x, y) 6= (0, 0), and that ( r 2 ) = 2 r.
r
2.5 Maxima and Minima 83
2.5 Maxima and Minima

The gradient can be used to find extreme points of real-valued functions of several variables,
that is, points where the function has a local maximum or local minimum. We will consider
only functions of two variables; functions of three or more variables require methods using
linear algebra.
Definition 2.7. Let f ( x, y) be a real-valued function, and let (a, b) be a point in the domain
of f . We say that f has a local maximum at (a, b) if f ( x, y) f (a, b) for all ( x, y) inside some
disk of positive radius centered at (a, b), i.e. there is some sufficiently small r > 0 such that
f ( x, y) f (a, b) for all ( x, y) for which ( x a)2 + ( y b)2 < r 2 .
Likewise, we say that f has a local minimum at (a, b) if f ( x, y) f (a, b) for all ( x, y)
inside some disk of positive radius centered at (a, b).
If f ( x, y) f (a, b) for all ( x, y) in the domain of f , then f has a global maximum at
(a, b). If f ( x, y) f (a, b) for all ( x, y) in the domain of f , then f has a global minimum at
(a, b).
Suppose that (a, b) is a local maximum point for f ( x, y), and that the first-order partial
derivatives of f exist at (a, b). We know that f (a, b) is the largest value of f ( x, y) as ( x, y)
goes in all directions from the point (a, b), in some sufficiently small disk centered at (a, b).
In particular, f (a, b) is the largest value of f in the x direction (around the point (a, b)), that
is, the single-variable function g( x) = f ( x, b) has a local maximum at x = a. So we know that
f f
g (a) = 0. Since g ( x) = x ( x, b), then x (a, b) = 0. Similarly, f (a, b) is the largest value of f
f
near (a, b) in the y direction and so y
(a, b) = 0. We thus have the following theorem:
f f
Theorem 2.5. Let f ( x, y) be a real-valued function such that both x (a, b) and y (a, b) exist.
Then a necessary condition for f ( x, y) to have a local maximum or minimum at (a, b) is that
f (a, b) = 0.
Note: Theorem 2.5 can be extended to apply to functions of three or more variables.
A point (a, b) where f (a, b) = 0 is called a critical point for the function f ( x, y). So given
f
a function f ( x, y), to find the critical points of f you have to solve the equations x ( x, y) = 0
f
and y ( x, y) = 0 simultaneously for ( x, y). Similar to the single-variable case, the necessary
condition that f (a, b) = 0 is not always sufficient to guarantee that a critical point is a local
maximum or minimum.
f
Example 2.18. The function f ( x, y) = x y has a critical point at (0, 0): x
= y = 0 y = 0, and
f
y
= x = 0 x = 0, so (0, 0) is the only critical point. But clearly f does not have a local
maximum or minimum at (0, 0) since any disk around (0, 0) contains points ( x, y) where the
values of x and y have the same sign (so that f ( x, y) = x y > 0 = f (0, 0)) and different signs (so
that f ( x, y) = x y < 0 = f (0, 0)). In fact, along the path y = x in R2 , f ( x, y) = x2 , which has a
local minimum at (0, 0), while along the path y = x we have f ( x, y) = x2 , which has a local
maximum at (0, 0). So (0, 0) is an example of a saddle point, i.e. it is a local maximum in one
direction and a local minimum in another direction. The graph of f ( x, y) is shown in Figure
2.5.1, which is a hyperbolic paraboloid.
100
50
0
z
-50
-10
-100-10 -5
-5 0 x
0 5 5
y 10 10
Figure 2.5.1 f (x, y) = x y, saddle point at (0, 0)
The following theorem gives sufficient conditions for a critical point to be a local maximum
or minimum of a smooth function (i.e. a function whose partial derivatives of all orders exist
and are continuous), which we will not prove here.6
Theorem 2.6. Let f ( x, y) be a smooth real-valued function, with a critical point at (a, b) (i.e.
f (a, b) = 0). Define
2 f 2 f 2 f 2
D = 2 (a, b) 2 (a, b) (a, b)
x y y x
Then
2 f
(a) if D > 0 and x2
(a, b) > 0, then f has a local minimum at (a, b)
2 f
(b) if D > 0 and x2
(a, b) < 0, then f has a local maximum at (a, b)
(c) if D < 0, then f has neither a local minimum nor a local maximum at (a, b)
(d) if D = 0, then the test fails.

If condition (c) holds, then (a, b) is a saddle point. Note that the assumption that f ( x, y) is
smooth means that 2
f 2 f

x2 ( a, b ) ( a, b )
y x

D=
2 f 2
f
(a, b) (a, b)
2
x y y
2 2
2 f 2 f 2 f 2 f f 2 f
since y x = x y . Also, if D > 0 then x2 (a, b) y2 (a, b) = D + y x (a, b) > 0, and so x2 (a, b)
2 f
and y2
(a, b) have the same sign. This means that in parts (a) and (b) of the theorem one
2 f 2 f
can replace x2 (a, b) by y2 (a, b) if desired.
Example 2.19. Find all local maxima and minima of f ( x, y) = x2 + x y + y2 3 x.

Solution: First find the critical points, i.e. where f = 0. Since
f f
= 2x + y 3 and = x + 2y
x y
then the critical points ( x, y) are the common solutions of the equations
2x + y 3 = 0
x + 2y =0
which has the unique solution ( x, y) = (2, 1). So (2, 1) is the only critical point.
To use Theorem 2.6, we need the second-order partial derivatives:
2 f 2 f 2 f
=2 , =2 , =1
x2 y2 y x
and so
2 f 2 f 2 f 2
D = 2
(2, 1) 2
(2, 1) (2, 1) = (2)(2) 12 = 3 > 0
x y y x
2 f
and x2
(2, 1) = 2 > 0. Thus, (2, 1) is a local minimum.
Example 2.20. Find all local maxima and minima of f ( x, y) = x y x3 y2 .

f f
= y 3 x2 and = x 2y
x y
y 3 x2 = 0
x 2y = 0
The first equation yields y = 3 x2 , substituting that into the second equation yields x 6 x2 = 0,
2 1
which has the solutions x = 0 and x = 16 . So x = 0 y = 3(0) = 0 and x = 61 y = 3 61 = 12 .
1 1
So the critical points are ( x, y) = (0, 0) and ( x, y) = 6 , 12 .
2 f 2 f 2 f
= 6 x , = 2 , =1
x2 y2 y x
So
2 f 2 f 2 f 2
D = 2
(0, 0) 2
(0, 0) (0, 0) = (6(0))(2) 12 = 1 < 0
x y y x
and thus (0, 0) is a saddle point. Also,
2 f 1 2 f 1 2 f 1 1 2 1
D = 1
6 , 12
1
6 12 y x 6 , 12
, = (6 )(2) 12 = 1 > 0
x2 y2 6
2 f 1 1 1 1

and ,
x2 6 12
= 1 < 0. Thus, 6 , 12 is a local maximum.
Example 2.21. Find all local maxima and minima of f ( x, y) = ( x 2)4 + ( x 2 y)2 .
f f
= 4( x 2)3 + 2( x 2 y) and = 4( x 2 y)
x y
4( x 2)3 + 2( x 2 y) = 0
4( x 2 y) = 0
The second equation yields x = 2 y, substituting that into the first equation yields 4(2 y 2)3 =
0, which has the solution y = 1, and so x = 2(1) = 2. Thus, (2, 1) is the only critical point.
2 f 2 f 2 f
= 12( x 2)2 + 2 , =8 , = 4
x2 y2 y x
So
2 f 2 f 2 f 2
D = 2
(2, 1) 2
(2, 1) (2, 1) = (2)(8) (4)2 = 0
x y y x
and so the test fails. What can be done in this situation? Sometimes it is possible to examine
the function to see directly the nature of a critical point. In our case, we see that f ( x, y) 0
for all ( x, y), since f ( x, y) is the sum of fourth and second powers of numbers and hence must
be nonnegative. But we also see that f (2, 1) = 0. Thus f ( x, y) 0 = f (2, 1) for all ( x, y), and
hence (2, 1) is in fact a global minimum for f .
2
+ y2 )
Example 2.22. Find all local maxima and minima of f ( x, y) = ( x2 + y2 ) e( x .
f 2
+ y2 )
= 2 x(1 ( x2 + y2 )) e( x
x
f 2
+ y2 )
= 2 y(1 ( x2 + y2 )) e( x
y
then the critical points are (0, 0) and all points ( x, y) on the unit circle x2 + y2 = 1.
2 f 2
+ y2 )
= 2[1 ( x2 + y2 ) 2 x2 2 x2 (1 ( x2 + y2 ))] e( x
x2
2 f 2
+ y2 )
= 2[1 ( x2 + y2 ) 2 y2 2 y2 (1 ( x2 + y2 ))] e( x
y2
2 f 2
+ y2 )
= 4 x y[2 ( x2 + y2 )] e( x
y x
2 f
At (0, 0), we have D = 4 > 0 and x2 (0, 0) = 2 > 0, so (0, 0) is a local minimum. However, for
points ( x, y) on the unit circle x2 + y2 = 1, we have
D = (4 x2 e1 )(4 y2 e1 ) (4 x ye1 )2 = 0
and so the test fails. If we look at the graph of f ( x, y), as shown in Figure 2.5.2, it looks like
we might have a local maximum for ( x, y) on the unit circle x2 + y2 = 1. If we switch to using
polar coordinates ( r, ) instead of ( x, y) in R2 , where r 2 = x2 + y2 , then we see that we can write
2 2
f ( x, y) as a function g( r ) of the variable r alone: g( r ) = r 2 er . Then g ( r ) = 2 r (1 r 2 ) er ,
so it has a critical point at r = 1, and we can check that g (1) = 4 e1 < 0, so the Second
Derivative Test from single-variable calculus says that r = 1 is a local maximum. But r = 1
corresponds to the unit circle x2 + y2 = 1. Thus, the points ( x, y) on the unit circle x2 + y2 = 1
are local maximum points for f .
0.4
0.35
0.3
0.25
z 0.2
0.15 -3
0.1 -2
0.05
0 -1
-3
-2 0
-1 x
1
0
1 2
y 2
3 3
2 + y2 )
Figure 2.5.2 f (x, y) = (x2 + y2 )e( x
Exercises
A
For Exercises 1-10, find all local maxima and minima of the function f ( x, y).
1. f ( x, y) = x3 3 x + y2 2. f ( x, y) = x3 12 x + y2 + 8 y
3. f ( x, y) = x3 3 x + y3 3 y 4. f ( x, y) = x3 + 3 x2 + y3 3 y2
5. f ( x, y) = 2 x3 + 6 x y + 3 y2 6. f ( x, y) = 2 x3 6 x y + y2
p
7. f ( x, y) = x2 + y2 8. f ( x, y) = x + 2 y
9. f ( x, y) = 4 x2 4 x y + 2 y2 + 10 x 6 y 10. f ( x, y) = 4 x2 + 4 x y 2 y2 + 16 x 12 y
B
11. For a rectangular solid of volume 1000 cubic meters, find the dimensions that will min-
imize the surface area. (Hint: Use the volume condition to write the surface area as a
function of just two variables.)
12. Prove that if (a, b) is a local maximum or local minimum point for a smooth function
f ( x, y), then the tangent plane to the surface z = f ( x, y) at the point (a, b, f (a, b)) is parallel
to the x y-plane. (Hint: Use Theorem 2.5.)
C
13. Find three positive numbers x, y, z whose sum is 10 such that x2 y2 z is a maximum.
2.6 Unconstrained Optimization: Numerical Methods 89
2.6 Unconstrained Optimization: Numerical Methods

The types of problems that we solved in the previous section were examples of unconstrained
optimization problems. That is, we tried to find local (and perhaps even global) maximum
and minimum points of real-valued functions f ( x, y), where the points ( x, y) could be any
points in the domain of f . The method we used required us to find the critical points of f ,
which meant having to solve the equation f = 0, which in general is a system of two equa-
tions in two unknowns ( x and y). While this was relatively simple for the examples we did,
in general this will not be the case. If the equations involve polynomials in x and y of degree
three or higher, or complicated expressions involving trigonometric, exponential, or loga-
rithmic functions, then solving even one such equation, let alone two, could be impossible by
elementary means.7
For example, if one of the equations that had to be solved was
x3 + 9 x 2 = 0 ,
you may have a hard time getting the exact solutions. Trial and error would not help much,
3 p 3 p
p p
8
especially since the only real solution turns out to be 28 + 1 28 1. In a situation
such as this, the only choice may be to find a solution using some numerical method which
gives a sequence of numbers which converge to the actual solution. For example, Newtons
method for solving equations f ( x) = 0, which you probably learned in single-variable calcu-
lus. In this section we will describe another method of Newton for finding critical points of
real-valued functions of two variables.
Let f ( x, y) be a smooth real-valued function, and define
2 f 2 f 2 f 2
D ( x, y) = 2
( x, y) 2
( x, y) ( x, y) .
x y y x
Newtons algorithm: Pick an initial point ( x0 , y0 ). For n = 0, 1, 2, 3, . . . , define:

2 2
f 2 f f 2 f
y2 ( xn , yn ) (x , y )
x y n n x2 ( xn , yn ) (x , y )
x y n n

f f f f
y ( xn , yn ) (x , y )
x n n
x ( xn , yn ) (x , y )
y n n
x n +1 = x n , yn+1 = yn (2.14)
D ( xn , yn ) D ( xn , yn )
Then the sequence of points ( xn , yn )

n=1 converges to a critical point. If there are several
critical points, then you will have to try different initial points to find them.
7 This is also a problem for the equivalent method (the Second Derivative Test) in single-variable calculus,
though one that is not usually emphasized.

8 There are also two nonreal, complex number solutions. Cubic polynomial equations in one variable can be
solved using Cardans formulas, which are not quite as simple as the familiar quadratic formula. See U SPENSKY
for more details. There are formulas for solving polynomial equations of degree 4, but it can be proved that there
is no general formula for solving equations for polynomials of degree five or higher.
Example 2.23. Find all local maxima and minima of f ( x, y) = x3 x y x + x y3 y4 .

Solution: First calculate the necessary partial derivatives:
f f
= 3 x 2 y 1 + y3 , = x + 3 x y2 4 y3
x y
2 f 2 f 2 f
= 6x , = 6 x y 12 y2 , = 1 + 3 y 2
x2 y2 y x
Notice that solving f = 0 would involve solving two third-degree polynomial equations in x
and y, which in this case can not be done easily.
We need to pick an initial point ( x0 , y0 ) for our algorithm. Looking at the graph of z = f ( x, y)
over a large region may help (see Figure 2.6.1 below), though it may be hard to tell where
the critical points are.
50000
0
-50000
-100000
-150000
-200000 -20
-250000 -15
-300000 -10
-350000 -5
-20
-15 0
-10 5 x
-5
0 10
5
y 10 15
15
20 20
Figure 2.6.1 f (x, y) = x3 x y x + x y3 y4 for 20 x 20 and 20 y 20
Notice in the formulas (2.14) that we divide by D , so we should pick an initial point where
D is not zero. And we can see that D (0, 0) = (0)(0) (1)2 = 1 6= 0, so take (0, 0) as our initial
point. Since it may take a large number of iterations of Newtons algorithm to be sure that
we are close enough to the actual critical point, and since the computations are quite tedious,
we will let a computer do the computing. For this, we will write a simple program, using
the Java programming language, which will take a given initial point as a parameter and
then perform 100 iterations of Newtons algorithm. In each iteration the new point will be
printed, so that we can see if there is convergence. The full code is shown in Listing 2.1.
//Program to find the critical points of f(x,y)=x^3-xy-x+xy^3-y^4

public class newton {
public static void main(String[] args) {
//Get the initial point (x,y) as command-line parameters
double x = Double.parseDouble(args[0]); //Initial x value
double y = Double.parseDouble(args[1]); //Initial y value
System.out.println("Initial point: (" + x + "," + y + ")");
//Go through 100 iterations of Newtons algorithm
for (int n=1; n<=100; n++) {
double D = fxx(x,y)*fyy(x,y) - Math.pow(fxy(x,y),2);
double xn = x; double yn = y; //The current x and y values
if (D == 0) { //We can not divide by 0
System.out.println("Error: D = 0 at iteration n = " + n);
System.exit(0); //End the program
} else { //Calculate the new values for x and y
x = xn - (fyy(xn,yn)*fx(xn,yn) - fxy(xn,yn)*fy(xn,yn))/D;
y = yn - (fxx(xn,yn)*fy(xn,yn) - fxy(xn,yn)*fx(xn,yn))/D;
System.out.println("n = " + n + ": (" + x + "," + y + ")");
}
}
}
//Below are the parts specific to the function f
//The first partial derivative of f wrt x: 3x^2-y-1+y^3
public static double fx(double x, double y) {
return 3*Math.pow(x,2) - y - 1 + Math.pow(y,3);
}
//The first partial derivative of f wrt y: -x+3xy^2-4y^3
public static double fy(double x, double y) {
return -x + 3*x*Math.pow(y,2) - 4*Math.pow(y,3);
}
//The second partial derivative of f wrt x: 6x
public static double fxx(double x, double y) {
return 6*x;
}
//The second partial derivative of f wrt y: 6xy-12y^2
public static double fyy(double x, double y) {
return 6*x*y - 12*Math.pow(y,2);
}
//The mixed second partial derivative of f wrt x and y: -1+3y^2
public static double fxy(double x, double y) {
return -1 + 3*Math.pow(y,2);
}
}
Listing 2.1 Program listing for newton.java

To use this program, you should first save the code in Listing 2.1 in a plain text file called
newton.java. You will need the Java Development Kit9 to compile the code. In the directory
where newton.java is saved, run this command at a command prompt to compile the code:
javac newton.java
Then run the program with the initial point (0, 0) with this command:
java newton 0 0
Below is the output of the program using (0, 0) as the initial point, truncated to show the
first 10 lines and the last 5 lines:
java newton 0 0
Initial point: (0.0,0.0)
n = 1: (0.0,-1.0)
n = 2: (1.0,-0.5)
n = 3: (0.6065857885615251,-0.44194107452339687)
n = 4: (0.484506572966545,-0.405341511995805)
n = 5: (0.47123972682634485,-0.3966334583092305)
n = 6: (0.47113558510349535,-0.39636450001936047)
n = 7: (0.4711356343449705,-0.3963643379632247)
n = 8: (0.4711356343449874,-0.39636433796318005)
n = 9: (0.4711356343449874,-0.39636433796318005)
n = 10: (0.4711356343449874,-0.39636433796318005)
...
n = 96: (0.4711356343449874,-0.39636433796318005)
n = 97: (0.4711356343449874,-0.39636433796318005)
n = 98: (0.4711356343449874,-0.39636433796318005)
n = 99: (0.4711356343449874,-0.39636433796318005)
n = 100: (0.4711356343449874,-0.39636433796318005)
As you can see, we appear to have converged fairly quickly (after only 8 iterations) to
what appears to be an actual critical point (up to Javas level of precision), namely the point
(0.4711356343449874, 0.39636433796318005). It is easy to confirm that f = 0 at this
f f
point, either by evaluating x and y at the point ourselves or by modifying our program to
also print the values of the partial derivatives at the point. It turns out that both partial
derivatives are indeed close enough to zero to be considered zero:
f
(0.4711356343449874, 0.39636433796318005) = 4.85722573273506 1017
x
f
(0.4711356343449874, 0.39636433796318005) = 8.326672684688674 1017
y
We also have D (0.4711356343449874, 0.39636433796318005) = 8.776075636032301 < 0,
so by Theorem 2.6 we know that (0.4711356343449874, 0.39636433796318005) is a saddle
point.
9 Available for free at http://www.oracle.com/technetwork/java/javase/downloads/
Since f consists of cubic polynomials, it seems likely that there may be three critical
points. The computer program makes experimenting with other initial points easy, and
trying different values does indeed lead to different sequences which converge:
java newton -1 -1
Initial point: (-1.0,-1.0)
n = 1: (-0.5,-0.5)
n = 2: (-0.49295774647887325,-0.08450704225352113)
n = 3: (-0.1855674752461383,-1.2047647348546167)
n = 4: (-0.4540060574531383,-0.8643989895639324)
n = 5: (-0.3672160534444,-0.5426077421319053)
n = 6: (-0.4794622222856417,-0.24529117721011612)
n = 7: (0.11570743992954591,-2.4319791238981274)
n = 8: (-0.05837851765533317,-1.6536079835854451)
n = 9: (-0.129841298650007,-1.121516233310142)
n = 10: (-1.004453014967208,-0.9206128022529645)
n = 11: (-0.5161209914612475,-0.4176293491131443)
n = 12: (-0.5788664043863884,0.2918236503332734)
n = 13: (-0.6985177124230715,0.49848120123515316)
n = 14: (-0.6733618916578702,0.4345777963475479)
n = 15: (-0.6704392913413444,0.4252025996474051)
n = 16: (-0.6703832679150286,0.4250147307973365)
n = 17: (-0.6703832459238701,0.42501465652421205)
n = 18: (-0.6703832459238667,0.4250146565242004)
n = 19: (-0.6703832459238667,0.42501465652420045)
n = 20: (-0.6703832459238667,0.42501465652420045)
...
n = 98: (-0.6703832459238667,0.42501465652420045)
n = 99: (-0.6703832459238667,0.42501465652420045)
n = 100: (-0.6703832459238667,0.42501465652420045)
f f
Again, it is easy to confirm that both x and y vanish at the point
(0.6703832459238667, 0.42501465652420045), which means it is a critical point. And
D (0.6703832459238667, 0.42501465652420045) = 15.3853578526055 > 0
2 f
(0.6703832459238667, 0.42501465652420045) = 4.0222994755432 < 0
x2
so we know that (0.6703832459238667, 0.42501465652420045) is a local maximum. An
idea of what the graph of f looks like near that point is shown in Figure 2.6.2, which does
suggest a local maximum around that point.
Finally, running the computer program with the initial point (5, 5) yields the critical
point (7.540962756992551, 5.595509445899435), with D < 0 at that point, which makes
it a saddle point.
(0.67, 0.42, 0.57)
0.6
0.4
0.2
0
z -0.2
-0.4 -1
-0.6
-0.8 -0.8
-1 -0.6
0
0.2 -0.4 x
0.4
0.6 -0.2
y 0.8
1 0
Figure 2.6.2 f (x, y) = x3 x y x + x y3 y4 for 1 x 0 and 0 y 1
We can summarize our findings for the function f ( x, y) = x3 x y x + x y3 y4 :
(0.4711356343449874, 0.39636433796318005) : saddle point

(0.6703832459238667, 0.42501465652420045) : local maximum
(7.540962756992551, 5.595509445899435) : saddle point
The derivation of Newtons algorithm, and the proof that it converges (given a reason-
able choice for the initial point) requires techniques beyond the scope of this text. See
R ALSTON and R ABINOWITZ for more detail and for discussion of other numerical methods.
Our description of Newtons algorithm is the special two-variable case of a more general
algorithm that can be applied to functions of n 2 variables.
In the case of functions which have a global maximum or minimum, Newtons algorithm
can be used to find those points. In general, global maxima and minima tend to be more
interesting than local versions, at least in practical applications. A maximization problem
can always be turned into a minimization problem (why?), so a large number of methods
have been developed to find the global minimum of functions of any number of variables.
This field of study is called nonlinear programming. Many of these methods are based on the
steepest descent technique, which is based on an idea that we discussed in Section 2.4. Recall
that the negative gradient f gives the direction of the fastest rate of decrease of a function
f . The crux of the steepest descent idea, then, is that starting from some initial point, you
move a certain amount in the direction of f at that point. Wherever that takes you
becomes your new point, and you then just keep repeating that procedure until eventually
(hopefully) you reach the point where f has its smallest value. There is a pure steepest
descent method, and a multitude of variations on it that improve the rate of convergence,
ease of calculation, etc. In fact, Newtons algorithm can be interpreted as a modified steepest
descent method. For more discussion of this, and of nonlinear programming in general, see
B AZARAA, S HERALI and S HETTY.
Exercises
C
1. Recall Example 2.21 from the previous section, where we showed that the point (2, 1) was
a global minimum for the function f ( x, y) = ( x 2)4 + ( x 2 y)2 . Notice that our computer
program can be modified fairly easily to use this function (just change the return values
in the fx, fy, fxx, fyy and fxy function definitions to use the appropriate partial derivative).
Either modify that program or write one of your own in a programming language of your
choice to show that Newtons algorithm does lead to the point (2, 1). First use the initial
point (0, 3), then use the initial point (3, 2), and compare the results. Make sure that your
program attempts to do 100 iterations of the algorithm. Did anything strange happen
when your program ran? If so, how do you explain it? (Hint: Something strange should
happen.)
2. There is a version of Newtons algorithm for solving a system of two equations
f 1 ( x, y) = 0 and f 2 ( x, y) = 0 ,
where f 1 ( x, y) and f 2 ( x, y) are smooth real-valued functions:

Pick an initial point ( x0 , y0 ). For n = 0, 1, 2, 3, . . . , define:

f 1 ( xn , yn ) f 2 ( xn , yn ) f 1 ( xn , yn ) f 2 ( xn , yn )

f1
( xn , yn ) f 2 ( xn , yn )
f1
( xn , yn ) f 2 ( xn , yn )
y y x x
x n +1 = x n , yn+1 = yn + , where
D ( xn , yn ) D ( xn , yn )
f1 f2 f1 f2
D ( xn , yn ) = ( xn , yn ) ( xn , yn ) ( xn , yn ) ( xn , yn ) .
x y y x
Then the sequence of points ( xn , yn )

n=1 converges to a solution. Write a computer program
that uses this algorithm to find approximate solutions to the system of equations
f 1 ( x, y) = sin( x y) x y = 0 and f 2 ( x, y) = e2 x 2 x + 3 y = 0 .
Show that you get two different solutions when using (0, 0) and (1, 1) for the initial point
( x0 , y0 ).
2.7 Constrained Optimization: Lagrange Multipliers

In Sections 2.5 and 2.6 we were concerned with finding maxima and minima of functions
without any constraints on the variables (other than being in the domain of the function).
What would we do if there were constraints on the variables? The following example illus-
trates a simple case of this type of problem.
Example 2.24. For a rectangle whose perimeter is 20 m, find the dimensions that will max-
imize the area.
Solution: The area A of a rectangle with width x and height y is A = x y. The perimeter P of
the rectangle is then given by the formula P = 2 x + 2 y. Since we are given that the perimeter
P = 20, this problem can be stated as:
Maximize : f ( x, y) = x y
given : 2 x + 2 y = 20
The reader is probably familiar with a simple method, using single-variable calculus, for
solving this problem. Since we must have 2 x + 2 y = 20, then we can solve for, say, y in
terms of x using that equation. This gives y = 10 x, which we then substitute into f to get
f ( x, y) = x y = x(10 x) = 10 x x2 . This is now a function of x alone, so we now just have to
maximize the function f ( x) = 10 x x2 on the interval [0, 10]. Since f ( x) = 10 2 x = 0 x = 5
and f (5) = 2 < 0, then the Second Derivative Test tells us that x = 5 is a local maximum
for f , and hence x = 5 must be the global maximum on the interval [0, 10] (since f = 0 at
the endpoints of the interval). So since y = 10 x = 5, then the maximum area occurs for a
rectangle whose width and height both are 5 m.
Notice in the above example that the ease of the solution depended on being able to solve
for one variable in terms of the other in the equation 2 x + 2 y = 20. But what if that were not
possible (which is often the case)? In this section we will use a general method, called the
Lagrange multiplier method10 , for solving constrained optimization problems:
Maximize (or minimize) : f ( x, y) (or f ( x, y, z))

given : g( x, y) = c (or g( x, y, z) = c) for some constant c
The equation g( x, y) = c is called the constraint equation, and we say that x and y are con-
strained by g( x, y) = c. Points ( x, y) which are maxima or minima of f ( x, y) with the con-
dition that they satisfy the constraint equation g( x, y) = c are called constrained maximum
or constrained minimum points, respectively. Similar definitions hold for functions of three
variables.
The Lagrange multiplier method for solving such problems can now be stated:
10 Named after the French mathematician Joseph Louis Lagrange (1736-1813).
2.7 Constrained Optimization: Lagrange Multipliers 97
Theorem 2.7. Let f ( x, y) and g( x, y) be smooth functions, and suppose that c is a scalar
constant such that g( x, y) 6= 0 for all ( x, y) that satisfy the equation g( x, y) = c. Then to
solve the constrained optimization problem
Maximize (or minimize) : f ( x, y)

given : g( x, y) = c ,
find the points ( x, y) that solve the equation f ( x, y) = g( x, y) for some constant (the
number is called the Lagrange multiplier). If there is a constrained maximum or mini-
mum, then it must be such a point.
A rigorous proof of the above theorem requires use of the Implicit Function Theorem,
which is beyond the scope of this text.11 Note that the theorem only gives a necessary con-
dition for a point to be a constrained maximum or minimum. Whether a point ( x, y) that
satisfies f ( x, y) = g( x, y) for some actually is a constrained maximum or minimum can
sometimes be determined by the nature of the problem itself. For instance, in Example 2.24
it was clear that there had to be a global maximum.
So how can you tell when a point that satisfies the condition in Theorem 2.7 really is a
constrained maximum or minimum? The answer is that it depends on the constraint func-
tion g( x, y), together with any implicit constraints. It can be shown12 that if the constraint
equation g( x, y) = c (plus any hidden constraints) describes a bounded set B in R2 , then the
constrained maximum or minimum of f ( x, y) will occur either at a point ( x, y) satisfying
f ( x, y) = g( x, y) or at a boundary point of the set B.
In Example 2.24 the constraint equation 2 x + 2 y = 20 describes a line in R2 , which by itself
is not bounded. However, there are hidden constraints, due to the nature of the problem,
namely 0 x, y 10, which cause that line to be restricted to a line segment in R2 (including
the endpoints of that line segment), which is bounded.
Example 2.25. For a rectangle whose perimeter is 20 m, use the Lagrange multiplier method
to find the dimensions that will maximize the area.
Solution: As we saw in Example 2.24, with x and y representing the width and height,
respectively, of the rectangle, this problem can be stated as:
given : g( x, y) = 2 x + 2 y = 20
Then solving the equation f ( x, y) = g( x, y) for some means solving the equations
11 See T AYLOR and M ANN, 6.8 for more detail.

12 Again, see T AYLOR and M ANN.
f g f g
= and = , namely:
x x y y
y = 2 ,
x = 2
The general idea is to solve for in both equations, then set those expressions equal (since
they both equal ) to solve for x and y. Doing this we get
y x
= = x= y ,
2 2
so now substitute either of the expressions for x or y into the constraint equation to solve for
x and y:
20 = g( x, y) = 2 x + 2 y = 2 x + 2 x = 4 x x = 5 y = 5
There must be a maximum area, since the minimum area is 0 and f (5, 5) = 25 > 0, so the
point (5, 5) that we found (called a constrained critical point) must be the constrained maxi-
mum.
The maximum area occurs for a rectangle whose width and height both are 5 m.
Example 2.26. Find the points on the circle x2 + y2 = 80 which are closest to and farthest
from the point (1, 2).
Solution: The distance d from any point ( x, y) to the point (1, 2) is
q
d= ( x 1)2 + ( y 2)2 ,
and minimizing the distance is equivalent to minimizing the square of the distance. Thus
the problem can be stated as:
Maximize (and minimize) : f ( x, y) = ( x 1)2 + ( y 2)2

given : g( x, y) = x2 + y2 = 80
Solving f ( x, y) = g( x, y) means solving the following equations:
2( x 1) = 2 x ,
2( y 2) = 2 y
Note that x 6= 0 since otherwise we would get 2 = 0 in the first equation. Similarly, y 6= 0.
So we can solve both equations for as follows:
x1 y2
= = x y y = x y 2x y = 2x
x y
2.7 Constrained Optimization: Lagrange Multipliers 99
y
Substituting this into g( x, y) = x2 + y2 = 80 yields 5 x2 = 80, x2 + y2 = 80
(4, 8)
so x = 4. So the two constrained critical points are (4, 8) and
(4, 8). Since f (4, 8) = 45 and f (4, 8) = 125, and since there
must be points on the circle closest to and farthest from (1, 2), (1, 2) x
then it must be the case that (4, 8) is the point on the circle clos- 0
est to (1, 2) and (4, 8) is the farthest from (1, 2) (see Figure
2.7.1).
(4, 8)
Notice that since the constraint equation x2 + y2 = 80 describes
a circle, which is a bounded set in R2 , then we were guaranteed
Figure 2.7.1
that the constrained critical points we found were indeed the
constrained maximum and minimum.
The Lagrange multiplier method can be extended to functions of three variables.
Example 2.27.
Maximize (and minimize) : f ( x, y, z) = x + z

given : g( x, y, z) = x2 + y2 + z2 = 1
Solution: Solve the equation f ( x, y, z) = g( x, y, z):
1 = 2 x
0 = 2 y
1 = 2 z
The first equation implies 6= 0 (otherwise we would have 1 = 0), so we can divide by in the
second equation to get y = 0 and we can divide by in the first and third equations to get
x = 21 = z. Substituting these expressions into the constraint equation g( x, y, z) = x2 + y2 +

z2 = 1 yields the constrained critical points p1 , 0, p1 and p 1 1
, 0, p . Since f p1 , 0, p1 >
2 2 2 2 2 2

1 1 2 2 2
f p , 0, p , and since the constraint equation x + y + z = 1 describes a sphere (which is
2 2

bounded) in R3 , then p1 , 0, p1 is the constrained maximum point and p 1 1
, 0, p is the
2 2 2 2
constrained minimum point.
So far we have not attached any significance to the value of the Lagrange multiplier . We
needed only to find the constrained critical points, but made no use of its value. It turns
out that gives an approximation of the change in the value of the function f ( x, y) that we
wish to maximize or minimize, when the constant c in the constraint equation g( x, y) = c is
changed by 1.
For example, in Example 2.25 we showed that the constrained optimization problem
given : g( x, y) = 2 x + 2 y = 20
had the solution ( x, y) = (5, 5), and that = x/2 = y/2. Thus, = 2.5. In a similar fashion we
could show that the constrained optimization problem
given : g( x, y) = 2 x + 2 y = 21
has the solution ( x, y) = (5.25, 5.25). So we see that the value of f ( x, y) at the constrained
maximum increased from f (5, 5) = 25 to f (5.25, 5.25) = 27.5625, i.e. it increased by 2.5625
when we increased the value of c in the constraint equation g( x, y) = c from c = 20 to c = 21.
Notice that = 2.5 is close to 2.5625, that is,
f = f (new max. pt) f (old max. pt) .
Finally, note that solving the equation f ( x, y) = g( x, y) means having to solve a system
of two (possibly nonlinear) equations in three unknowns, which as we have seen before,
may not be possible to do. And the 3-variable case can get even more complicated. All of
this somewhat restricts the usefulness of Lagranges method to relatively simple functions.
Luckily there are many numerical methods for solving constrained optimization problems,
though we will not discuss them here.13
Exercises
A
1. Find the constrained maxima and minima of f ( x, y) = 2 x + y given that x2 + y2 = 4.
2. Find the constrained maxima and minima of f ( x, y) = x y given that x2 + 3 y2 = 6.
3. Find the points on the circle x2 + y2 = 100 which are closest to and farthest from the point
(2, 3).
B
4. Find the constrained maxima and minima of f ( x, y, z) = x + y2 + 2 z given that 4 x2 + 9 y2
36 z2 = 36.
5. Find the volume of the largest rectangular parallelepiped that can be inscribed in the
ellipsoid
x 2 y2 z 2
+ + =1 .
a2 b 2 c 2
13 See B AZARAA, S HERALI and S HETTY.

3 Multiple Integrals
3.1 Double Integrals

In single-variable calculus, differentiation and integration are thought of as inverse opera-
tions. For instance, to integrate a function f ( x) it is necessary to find the antiderivative of f ,
that is, another function F ( x) whose derivative is f ( x). Is there a similar way of defining in-
tegration of real-valued functions of two or more variables? The answer is yes, as we will see
shortly. Recall also that the definite integral of a nonnegative function f ( x) 0 represented
the area under the curve y = f ( x). As we will now see, the double integral of a nonnegative
real-valued function f ( x, y) 0 represents the volume under the surface z = f ( x, y).
Let f ( x, y) be a continuous function such that f ( x, y) 0 for all ( x, y) on the rectangle
R = {( x, y) : a x b, c y d } in R2 . We will often write this as R = [a, b] [ c, d ]. For any
number x in the interval [a, b], slice the surface z = f ( x, y) with the plane x = x parallel to
the yz-plane. Then the trace of the surface in that plane is the curve f ( x, y), where x is
fixed and only y varies. The area A under that curve (i.e. the area of the region between the
curve and the x y-plane) as y varies over the interval [ c, d ] then depends only on the value of
x. So using the variable x instead of x, let A ( x) be that area (see Figure 3.1.1).
z
z = f (x, y)
c d y
0 A(x)
a
x
b
x R
Figure 3.1.1 The area A(x) varies with x

Rd
Then A ( x) = c f ( x, y) d y since we are treating x as fixed, and only y varies. This makes
sense since for a fixed x the function f ( x, y) is a continuous function of y over the interval
[ c, d ], so we know that the area under the curve is the definite integral. The area A ( x) is a
function of x, so by the slice or cross-section method from single-variable calculus we know
that the volume V of the solid under the surface z = f ( x, y) but above the x y-plane over the
101
102 CHAPTER 3. MULTIPLE INTEGRALS
rectangle R is the integral over [a, b] of that cross-sectional area A ( x):
Zb Zb Zd
V = A ( x) dx = f ( x, y) d y dx (3.1)
a a c
We will always refer to this volume as the volume under the surface. The above expression
uses what are called iterated integrals. First the function f ( x, y) is integrated as a func-
tion of y, treating the variable x as a constant (this is called integrating with respect to y).
That is what occurs in the inner integral between the square brackets in equation (3.1).
This is the first iterated integral. Once that integration is performed, the result is then an
expression involving only x, which can then be integrated with respect to x. That is what
occurs in the outer integral above (the second iterated integral). The final result is then
a number (the volume). This process of going through two iterations of integrals is called
double integration, and the last expression in equation (3.1) is called a double integral.
Notice that integrating f ( x, y) with respect to y is the inverse operation of taking the
partial derivative of f ( x, y) with respect to y. Also, we could just as easily have taken the
area of cross-sections under the surface which were parallel to the xz-plane, which would
then depend only on the variable y, so that the volume V would be
Zd Zb
V = f ( x, y) dx d y . (3.2)
c a
It turns out that in general1 the order of the iterated integrals does not matter. Also, we will
usually discard the brackets and simply write
Zd Zb
V = f ( x, y) dx d y , (3.3)
c a
where it is understood that the fact that dx is written before d y means that the function
f ( x, y) is first integrated with respect to x using the inner limits of integration a and b,
and then the resulting function is integrated with respect to y using the outer limits of
integration c and d . This order of integration can be changed if it is more convenient.
Example 3.1. Find the volume V under the plane z = 8 x + 6 y over the rectangle R = [0, 1]
[0, 2].
1 due to Fubinis Theorem. See Ch. 18 in T AYLOR and M ANN.

3.1 Double Integrals 103
Solution: We see that f ( x, y) = 8 x + 6 y 0 for 0 x 1 and 0 y 2, so:

Z2 Z1
V = (8 x + 6 y) dx d y
0 0
Z2 x =1
4 x2 + 6 x y

= dy
0 x =0
Z2
= (4 + 6 y) d y
0
2
= 4 y + 3 y2

0
= 20
Suppose we had switched the order of integration. We can verify that we still get the same
answer:
Z1 Z2
V = (8 x + 6 y) d y dx
0 0
Z1 y=2
8 x y + 3 y2

= dx
0 y=0
Z1
= (16 x + 12) dx
0
1
= 8 x2 + 12 x

0
= 20
Example 3.2. Find the volume V under the surface z = e x+ y over the rectangle R = [2, 3]
[1, 2].
Solution: We know that f ( x, y) = e x+ y > 0 for all ( x, y), so
Z2 Z3
V = e x+ y dx d y
1 2
Z2 x =3
e x+ y

= dy
1 x =2
Z2
= ( e y+3 e y+2 ) d y
1
2
= e y+3 e y+2

1
= e5 e4 ( e4 e3 ) = e5 2 e4 + e3
Rb
Recall that for a general function f ( x), the integral a f ( x) dx represents the difference of
the area below the curve y = f ( x) but above the x-axis when f ( x) 0, and the area above the
curve but below the x-axis when f ( x) 0. Similarly, the double integral of any continuous
function f ( x, y) represents the difference of the volume below the surface z = f ( x, y) but above
the x y-plane when f ( x, y) 0, and the volume above the surface but below the x y-plane when
f ( x, y) 0. Thus, our method of double integration by means of iterated integrals can be
used to evaluate the double integral of any continuous function over a rectangle, regardless
of whether f ( x, y) 0 or not.
Z2 Z
Example 3.3. Evaluate sin( x + y) dx d y.
0 0
Solution: Note that f ( x, y) = sin( x + y) is both positive and negative over the rectangle [0, ]
[0, 2]. We can still evaluate the double integral:
Z2 Z Z2 x =

sin( x + y) dx d y = cos( x + y) dy
0 0 0 x =0
Z2
= ( cos( y + ) + cos y) d y
0
2

= sin( y + ) + sin y = sin 3 + sin 2 ( sin + sin 0)
0
= 0
Exercises
A
For Exercises 1-4, find the volume under the surface z = f ( x, y) over the rectangle R .
1. f ( x, y) = 4 x y, R = [0, 1] [0, 1] 2. f ( x, y) = e x+ y , R = [0, 1] [1, 1]
3. f ( x, y) = x3 + y2 , R = [0, 1] [0, 1] 4. f ( x, y) = x4 + x y + y3 , R = [1, 2] [0, 2]
For Exercises 5-12, evaluate the given double integral.

Z1 Z2 Z1 Z2
2
5. (1 y) x dx d y 6. x( x + y) dx d y
0 1 0 0
Z2 Z1 Z2 Z1
7. ( x + 2) dx d y 8. x( x y + sin x) dx d y
0 0 1 1
Z/2 Z1 Z Z/2
9. x y cos( x2 y) dx d y 10. sin x cos( y ) dx d y
0 0 0 0
Z2 Z4 Z1 Z2
11. x y dx d y 12. 1 dx d y
0 1 1 1
Rd Rb
13. Let M be a constant. Show that c a M dx d y = M ( d c)( b a).
3.2 Double Integrals Over a General Region 105
3.2 Double Integrals Over a General Region

In the previous section we got an idea of what a double integral over a rectangle represents.
We can now define the double integral of a real-valued function f ( x, y) over more general
regions in R2 .
Suppose that we have a region R in the x y-plane that is bounded on the left by the vertical
line x = a, bounded on the right by the vertical line x = b (where a < b), bounded below by
a curve y = g 1 ( x), and bounded above by a curve y = g 2 ( x), as in Figure 3.2.1(a). We will
assume that g 1 ( x) and g 2 ( x) do not intersect on the open interval (a, b) (they could intersect
at the endpoints x = a and x = b, though).
y y
y = g 2 (x)
d
R
x = h 1 (y)
x = h 2 (y)
y = g 1 (x) c
R
x x
0 a b 0
R R g ( x) R Rh ( y)
(a) Vertical slice: ab g 2( x) f ( x, y) d y dx (b) Horizontal slice: cd h 2( y) f ( x, y) dx d y
1 1
Figure 3.2.1 Double integral over a nonrectangular region R
Then using the slice method from the previous

section, the double integral of a real-valued
function f ( x, y) over the region R , denoted by f ( x, y) d A , is given by
R
Zb Z g 2 ( x)
f ( x, y) d A = f ( x, y) d y dx (3.4)
a g 1 ( x)
R
This means that we take vertical slices in the region R between the curves y = g 1 ( x) and
y = g 2 ( x). The symbol d A is sometimes called an area element or infinitesimal, with the A
signifying area. Note that f ( x, y) is first integrated with respect to y, with functions of x as
the limits of integration. This makes sense since the result of the first iterated integral will
have to be a function of x alone, which then allows us to take the second iterated integral
with respect to x.
Similarly, if we have a region R in the x y-plane that is bounded on the left by a curve
x = h 1 ( y), bounded on the right by a curve x = h 2 ( y), bounded below by the horizontal line
y = c, and bounded above by the horizontal line y = d (where c < d ), as in Figure 3.2.1(b)
(assuming that h 1 ( y) and h 2 ( y) do not intersect on the open interval ( c, d )), then taking
horizontal slices gives
Zd Zh2 ( y)
f ( x, y) d A = f ( x, y) dx d y (3.5)
c h 1 ( y)
R
Notice that these definitions include the case when the region R is a rectangle. Also, if

f ( x, y) 0 for all ( x, y) in the region R , then f ( x, y) d A is the volume under the surface
R
z = f ( x, y) over the region R .
Example 3.4. Find the volume V under the plane z = 8 x + 6 y over the region R = {( x, y) : 0
x 1, 0 y 2 x2 }.
y
Solution: The region R is shown in Figure 3.2.2. Using vertical slices we
y = 2x2
get:

V = (8 x + 6 y) d A R
R x
Z1 "Z2 x2 #
0 1
= (8 x + 6 y) d y dx
0 0 Figure 3.2.2
Z1 y=2 x2
2
= 8x y + 3 y dx
0 y=0
Z1
= (16 x3 + 12 x4 ) dx
0
1
= 4 x4 + 12
5 x 5
= 4 + 12 5 =
32
5 = 6.4
0
y
We get the same answer using horizontal slices (see Figure 3.2.3): 2

V = (8 x + 6 y) d A p
x= y/2
R
R
Z2 Z1
x
= p (8 x + 6 y) dx d y
0 y/2 0 1
Z2 x =1
Figure 3.2.3
4 x2 + 6 x y p

= dy
0 x= y/2
Z2 Z2 p
p
= (4 + 6 y (2 y + p6 y y )) d y = (4 + 4 y 3 2 y3/2 ) d y
0 2 0
p 2 p p
= 4 y + 2 y2 6 5 2 y5/2 = 8 + 8 6 25 32 = 16 48 32

0 5 = 5 = 6.4
Example 3.5. Find the volume V of the solid bounded by the three coordinate planes and
the plane 2 x + y + 4 z = 4.
y
4
z
y = 2x + 4
(0, 0, 1) 2x + y + 4z = 4
y R
0 (0, 4, 0)
x
x (2, 0, 0) 0 2
(a) (b)
Figure 3.2.4
Solution: The solid is shown in Figure 3.2.4(a) with a typical vertical slice. The volume V

is given by f ( x, y) d A , where f ( x, y) = z = 14 (4 2 x y) and the region R , shown in Figure
R
3.2.4(b), is R = {( x, y) : 0 x 2, 0 y 2 x + 4}. Using vertical slices in R gives

1
V = 4 (4 2 x y) d A
R
Z2 Z2 x+4
1
= 4 (4 2 x
y) d y dx
0 0
Z2 y=2 x+4
1 2
= 8 (4 2 x y) dx
0 y=0
Z2
1 2
= 8 (4 2 x) dx
0
2
1
(4 2 x)3 = 64 4

= 48 48 = 3
0
For a general region R , which may not be one of the types of regions we have considered so

far, the double integral f ( x, y) d A is defined as follows. Assume that f ( x, y) is a nonnega-
R
tive real-valued function and that R is a bounded region in R2 , so it can be enclosed in some
rectangle [a, b] [ c, d ]. Then divide that rectangle into a grid of subrectangles. Only consider
the subrectangles that are enclosed completely within the region R , as shown by the shaded
subrectangles in Figure 3.2.5(a). In any such subrectangle [ xi , xi+1 ] [ y j , y j+1 ], pick a point
( xi , y j ). Then the volume under the surface z = f ( x, y) over that subrectangle is approxi-
mately f ( xi , y j ) xi y j , where xi = xi+1 xi , y j = y j+1 y j , and f ( xi , y j ) is the height and
xi y j is the base area of a parallelepiped, as shown in Figure 3.2.5(b). Then the total vol-
ume under the surface is approximately the sum of the volumes of all such parallelepipeds,
namely XX
f ( x i , y j ) x i y j , (3.6)
j i
where the summation occurs over the indices of the subrectangles inside R . If we take
smaller and smaller subrectangles, so that the length of the largest diagonal of the subrect-
angles goes to 0, then the subrectangles begin to fill more and more of the region R , and so
the above sum approaches the actual volume under the surface z = f ( x, y) over the region R .

We then define f ( x, y) d A as the limit of that double summation (the limit is taken over all
R
subdivisions of the rectangle [a, b] [ c, d ] as the largest diagonal of the subrectangles goes
to 0).
y
d
z
z = f (x, y)
yj
xi f (x i , y j )
y j +1
(x i , y j )
y j y j +1 y
yj
0
c xi
x x i +1
0 a (x i , y j ) R
x i x i +1 b x
(a) Subrectangles inside the region R (b) Parallelepiped over a subrectangle,
with volume f ( x i , y j ) x i y j
Figure 3.2.5 Double integral over a general region R
A similar definition can be made for a function f ( x, y) that is not necessarily always non-
negative: just replace each mention of volume by the negative volume in the description
above when f ( x, y) < 0. In the case of a region of the type shown in Figure 3.2.1, using the def-

inition of the Riemann integral from single-variable calculus, our definition of f ( x, y) d A
R
reduces to a sequence of two iterated integrals.
Finally, the region R does not have to be bounded. We can evaluate improper double
integrals (i.e. over an unbounded region, or over a region which contains points where the
function f ( x, y) is not defined) as a sequence of iterated improper single-variable integrals.
Z Z1/ x2
Example 3.6. Evaluate 2 y d y dx.
1 0
Solution:
Z Z1/ x2 Z y=1/ x2
2
2 y d y dx = y dx
1 0 y=0
Z1
x4 dx = 13 x3 = 0 ( 13 ) = 1

= 3
1 1
Exercises
A
For Exercises 1-6, evaluate the given double integral.
Z1 Z1 Z Z y
2 2. sin x dx d y
1. p 24 x y d y dx
0 x 0 0
Z2 Zln x Z2 Z2 y
2
3. 4 x d y dx 4. e y dx d y
1 0
Z/2 Z y Z0 Z
0
2
+ y2 )
5. cos x sin y dx d y 6. x ye( x dx d y
0 0 0 0
Z2 Z y Z1 Z x2
7. 1 dx d y 8. 2 d y dx
0 0 0 0
9. Find the volume V of the solid bounded by the three coordinate planes and the plane
x + y + z = 1.
10. Find the volume V of the solid bounded by the three coordinate planes and the plane
3 x + 2 y + 5 z = 6.
B

11. Explain why the double integral 1 d A gives the area of the region R . For simplicity,
R
you can assume that R is a region of the type shown in Figure 3.2.1(a).
C
12. Prove that the volume of a tetrahedron with mutually perpendic- c
ular adjacent sides of lengths a, b, and c, as in Figure 3.2.6, is abc
6 .
b
(Hint: Mimic Example 3.5, and recall from a
Section 1.5 how three noncollinear points determine a plane.)
13. Show how Exercise 12 can be used to solve Exercise 10. Figure 3.2.6
3.3 Triple Integrals

Our definition of a double integral of a real-valued function f ( x, y) over a region R in R2 can
be extended to define a triple integral of a real-valued function f ( x, y, z) over a solid S in R3 .
We simply proceed as before: the solid S can be enclosed in some rectangular parallelepiped,
which is then divided into subparallelepipeds. In each subparallelepiped inside S , with sides
of lengths x, y and z, pick a point ( x , y , z ). Then define the triple integral of f ( x, y, z)

over S , denoted by f ( x, y, z) dV , by
S
XXX
f ( x, y, z) dV = lim f ( x , y , z ) x y z , (3.7)
S
where the limit is over all divisions of the rectangular parallelepiped enclosing S into sub-
parallelepipeds whose largest diagonal is going to 0, and the triple summation is over all the
subparallelepipeds inside S . It can be shown that this limit does not depend on the choice
of the rectangular parallelepiped enclosing S . The symbol dV is often called the volume
element.
Physically, what does the triple integral represent? We saw that a double integral could
be thought of as the volume under a two-dimensional surface. It turns out that the triple
integral simply generalizes this idea: it can be thought of as representing the hypervolume
under a three-dimensional hypersurface w = f ( x, y, z) whose graph lies in R4 . In general,
the word volume is often used as a general term to signify the same concept for any n-
dimensional object (e.g. length in R1 , area in R2 ). It may be hard to get a grasp on the concept
of the volume of a four-dimensional object, but at least we now know how to calculate that
volume!
In the case where S is a rectangular parallelepiped [ x1 , x2 ] [ y1 , y2 ] [ z1 , z2 ], that is, S =
{( x, y, z) : x1 x x2 , y1 y y2 , z1 z z2 }, the triple integral is a sequence of three iterated
integrals, namely
Z z2 Z y2 Z x2
f ( x, y, z) dV = f ( x, y, z) dx d y dz , (3.8)
z1 y1 x1
S
where the order of integration does not matter. This is the simplest case.
A more complicated case is where S is a solid which is bounded below by a surface z =
g 1 ( x, y), bounded above by a surface z = g 2 ( x, y), y is bounded between two curves h 1 ( x) and
h 2 ( x), and x varies between a and b. Then
Zb Zh2 ( x) Z g 2 ( x,y)
f ( x, y, z) dV = f ( x, y, z) dz d y dx . (3.9)
a h 1 ( x) g 1 ( x,y)
S
Notice in this case that the first iterated integral will result in a function of x and y (since its
limits of integration are functions of x and y), which then leaves you with a double integral of
3.3 Triple Integrals 111
a type that we learned how to evaluate in Section 3.2. There are, of course, many variations
on this case (for example, changing the roles of the variables x, y, z), so as you can probably
tell, triple integrals can be quite tricky. At this point, just learning how to evaluate a triple
integral, regardless of what it represents, is the most important thing. We will see some
other ways in which triple integrals are used later in the text.
Z3 Z2 Z1
Example 3.7. Evaluate ( x y + z) dx d y dz.
0 0 0
Solution:
Z3 Z2 Z1 Z3 Z2 x =1
1 2
( x y + z) dx d y dz = 2 x y + xz d y dz
0 0 0 0 0 x =0
Z3 Z2
1
= 2 y+ z
d y dz
0 0
Z3 y=2
1 2
= 4 y + yz dz
y=0
0
Z3
= (1 + 2 z) dz
0
3
= z + z2 = 12

0
Z1 Z1 x Z2 x y
Example 3.8. Evaluate ( x + y + z) dz d y dx.
0 0 0
Solution:
Z1 Z1 x Z2 x y Z1 Z1 x z=2 x y
1 2
( x + y + z) dz d y dx = ( x + y) z + 2 z z =0 d y dx
0 0 0 0 0
Z1 Z1 x

= ( x + y)(2 x y) + 21 (2 x y)2 d y dx
0 0
Z1 Z1 x

= 2 12 x2 x y 21 y2 d y dx
0 0
Z1 y=1 x
2 y 12 x2 y x y 21 x y2 16 y3

= dx
0 y=0
Z1
11 3

= 2 x + 16 x dx
6
0
1
= 6 x x2 + 24
11 1 4
x = 78
0
Note that the volume V of a solid in R3 is given by

V = 1 dV . (3.10)
S
Since the function being integrated is the constant 1, then the above triple integral reduces
to a double integral of the types that we considered in the previous section if the solid is
bounded above by some surface z = f ( x, y) and bounded below by the x y-plane z = 0. There
are many other possibilities. For example, the solid could be bounded below and above by
surfaces z = g 1 ( x, y) and z = g 2 ( x, y), respectively, with y bounded between two curves h 1 ( x)
and h 2 ( x), and x varies between a and b. Then
Zb Zh2 ( x) Z g 2 ( x,y) Z b Z h 2 ( x)
V= 1 dV = 1 dz d y dx = ( g 2 ( x, y) g 1 ( x, y)) d y dx
a h 1 ( x) g 1 ( x,y) a h 1 ( x)
S
just like in equation (3.9). See Exercise 10 for an example.
Exercises
A
For Exercises 1-8, evaluate the given triple integral.
Z3 Z2 Z1 Z1 Z x Z y
1. x yz dx d y dz 2. x yz dz d y dx
Z0 Z0 x Z0 x y 0 0 0
Z1 Z z Z y
x2 sin z dz d y dx
2
3. 4. ze y dx d y dz
0 0 0 0 0 0
Z e Z y Z1/ y Z2 Z y2 Z z2
5. x2 z dx dz d y 6. yz dx dz d y
1 0 0 1 0 0
Z2 Z4 Z3 Z1 Z1 x Z1 x y
7. 1 dx d y dz 8. 1 dz d y dx
1 2 0 0 0 0
R z2 R y2 Rx2
9. Let M be a constant. Show that z1 y1 x1 M dx d y dz = M ( z2 z1 )( y2 y1 )( x2 x1 ).
B
10. Find the volume V of the solid S bounded by the three coordinate planes, bounded above
by the plane x + y + z = 2, and bounded below by the plane z = x + y.
C
Zb Z z Z y Zb
( b x )2
11. Show that f ( x) dx d y dz = 2 f ( x) dx. (Hint: Think of how changing the
a a a a
order of integration in the triple integral changes the limits of integration.)
3.4 Numerical Approximation of Multiple Integrals 113
3.4 Numerical Approximation of Multiple Integrals

As you have seen, calculating multiple integrals is tricky even for simple functions and
regions. For complicated functions, it may not be possible to evaluate one of the iterated in-
tegrals in a simple closed form. Luckily there are numerical methods for approximating the
value of a multiple integral. The method we will discuss is called the Monte Carlo method.
The idea behind it is based on the concept of the average value of a function, which you
learned in single-variable calculus. Recall that for a continuous function f ( x), the average
value f of f over an interval [a, b] is defined as
Zb
1
f = f ( x) dx . (3.11)
ba a
The quantity b a is the length of the interval [a, b], which can be thought of as the
volume of the interval. Applying the same reasoning to functions of two or three variables,
we define the average value of f ( x, y) over a region R to be

1
f = f ( x, y) d A , (3.12)
A (R )
R
where A (R ) is the area of the region R , and we define the average value of f ( x, y, z) over a
solid S to be
1
f = f ( x, y, z) dV , (3.13)
V (S )
S
where V (S ) is the volume of the solid S . Thus, for example, we have

f ( x, y) d A = A (R ) f . (3.14)
R
The average value of f ( x, y) over R can be thought of as representing the sum of all the
values of f divided by the number of points in R . Unfortunately there are an infinite number
(in fact, uncountably many) points in any region, i.e. they can not be listed in a discrete
sequence. But what if we took a very large number N of random points in the region R
(which can be generated by a computer) and then took the average of the values of f for
those points, and used that average as the value of f? This is exactly what the Monte Carlo
method does. So in formula (3.14) the approximation we get is
s
f 2 ( f)2

f ( x, y) d A A (R ) f A (R ) , (3.15)
N
R
where PN PN 2
i =1 f ( xi , yi ) i =1 ( f ( x i , yi ))
f = and f2 = , (3.16)
N N
with the sums taken over the N random points ( x1 , y1 ), . . ., ( xN , yN ). The error term in
formula (3.15) does not really provide hard bounds on the approximation. It represents a
single standard deviation from the expected value of the integral. That is, it provides a likely
bound on the error. Due to its use of random points, the Monte Carlo method is an example
of a probabilistic method (as opposed to deterministic methods such as Newtons method,
which use a specific formula for generating points).
For example, we can use formula (3.15) to approximate the volume V under the plane
z = 8 x + 6 y over the rectangle R = [0, 1] [0, 2]. In Example 3.1 in Section 3.1, we showed
that the actual volume is 20. Below is a code listing (montecarlo.java) for a Java program
that calculates the volume, using a number of points N that is passed on the command line
as a parameter.
//Program to approximate the double integral of f(x,y)=8x+6y

//over the rectangle [0,1]x[0,2].
public class montecarlo {
//Get the number N of random points as a command-line parameter
int N = Integer.parseInt(args[0]);
double x = 0; //x-coordinate of a random point
double y = 0; //y-coordinate of a random point
double f = 0.0; //Value of f at a random point
double mf = 0.0; //Mean of the values of f
double mf2 = 0.0; //Mean of the values of f^2
for (int i=0;i<N;i++) { //Get the random coordinates
x = Math.random(); //x is between 0 and 1
y = 2 * Math.random(); //y is between 0 and 2
f = 8*x + 6*y; //Value of the function
mf = mf + f; //Add to the sum of the f values
mf2 = mf2 + f*f; //Add to the sum of the f^2 values
}
mf = mf/N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f^2 values
System.out.println("N = " + N + ": integral = " + vol()*mf + " +/- "
+ vol()*Math.sqrt((mf2 - Math.pow(mf,2))/N)); //Print the result
}
//The volume of the rectangle [0,1]x[0,2]
public static double vol() {
return 1*2;
}
}
Listing 3.1 Program listing for montecarlo.java
The results of running this program with various numbers of random points (e.g. java
montecarlo 100) are shown below:
3.4 Numerical Approximation of Multiple Integrals 115
N = 10: 19.36543087722646 +/- 2.7346060413546147

N = 100: 21.334419561385353 +/- 0.7547037194998519
N = 1000: 19.807662237526227 +/- 0.26701709691370235
N = 10000: 20.080975812043256 +/- 0.08378816229769506
N = 100000: 20.009403854556716 +/- 0.026346782289498317
N = 1000000: 20.000866994982314 +/- 0.008321168748642816
As you can see, the approximation is fairly good. As N , it can be shown p that the
Monte Carlo approximation converges to the actual volume (on the order of O ( N ), in com-
putational complexity terminology).
In the above example the region R was a rectangle. To use the Monte Carlo method for
a nonrectangular (bounded) region R , only a slight modification is needed. Pick a rectangle
R that encloses R , and generate random points in that rectangle as before. Then use those
points in the calculation of f only if they are inside R . There is no need to calculate the area
of R for formula (3.15) in this case, since the exclusion of points not inside R allows you to
use the area of the rectangle R instead, similar to before.
For instance, in Example 3.4 we showed that the volume under the surface z = 8 x + 6 y
over the nonrectangular region R = {( x, y) : 0 x 1, 0 y 2 x2 } is 6.4. Since the rectangle
R = [0, 1] [0, 2] contains R , we can use the same program as before, with the only change
being a check to see if y < 2 x2 for a random point ( x, y) in [0, 1] [0, 2]. Listing 3.2 below
contains the code (montecarlo2.java):
//Program to approximate the double integral of f(x,y)=8x+6y over the

//region bounded by x=0, x=1, y=0, and y=2x^2
public class montecarlo2 {
//Get the number N of random points as a command-line parameter
int N = Integer.parseInt(args[0]);
double x = 0; //x-coordinate of a random point
double y = 0; //y-coordinate of a random point
double f = 0.0; //Value of f at a random point
double mf = 0.0; //Mean of the values of f
double mf2 = 0.0; //Mean of the values of f^2
for (int i=0;i<N;i++) { //Get the random coordinates
x = Math.random(); //x is between 0 and 1
y = 2 * Math.random(); //y is between 0 and 2
if (y < 2*Math.pow(x,2)) { //The point is in the region
f = 8*x + 6*y; //Value of the function
mf = mf + f; //Add to the sum of the f values
mf2 = mf2 + f*f; //Add to the sum of the f^2 values
}
}
mf = mf/N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f^2 values
System.out.println("N = " + N + ": integral = " + vol()*mf +
" +/- " + vol()*Math.sqrt((mf2 - Math.pow(mf,2))/N));

}
//The volume of the rectangle [0,1]x[0,2]
public static double vol() {
return 1*2;
}
}
Listing 3.2 Program listing for montecarlo2.java
The results of running the program with various numbers of random points (e.g. java
montecarlo2 1000) are shown below:
N = 10: integral = 6.95747529014894 +/- 2.9185131565120592

N = 100: integral = 6.3149056229650355 +/- 0.9549009662159909
N = 1000: integral = 6.477032813858756 +/- 0.31916837260973624
N = 10000: integral = 6.349975080015089 +/- 0.10040086346895105
N = 100000: integral = 6.440184132811864 +/- 0.03200476870881392
N = 1000000: integral = 6.417050897922222 +/- 0.01009454409789472
To use the Monte Carlo method to evaluate triple integrals, you will need to generate
random triples ( x, y, z) in a parallelepiped, instead of random pairs ( x, y) in a rectangle, and
use the volume of the parallelepiped instead of the area of a rectangle in formula (3.15) (see
Exercise 2). For a more detailed discussion of numerical integration methods, see P RESS et
al.
Exercises
C
1. Write a program that uses the Monte Carlo method to approximate the double integral
xy
e d A , where R = [0, 1] [0, 1]. Show the program output for N = 10, 100, 1000, 10000,
R
100000 and 1000000 random points.
2. Write a program that uses the Monte Carlo method to approximate the triple integral
x yz
e dV , where S = [0, 1] [0, 1] [0, 1]. Show the program output for N = 10, 100,
S
1000, 10000, 100000 and 1000000 random points.
3. Repeat Exercise 1 with the region R = {( x, y) : 1 x 1, 0 y x2 }.
4. Repeat Exercise 2 with the solid S = {( x, y, z) : 0 x 1, 0 y 1, 0 z 1 x y}.
5. Use the Monte Carlo method to approximate the volume of a sphere of radius 1.
x2 y2 2
6. Use the Monte Carlo method to approximate the volume of the ellipsoid 9 + 4 + z1 = 1.
3.5 Change of Variables in Multiple Integrals 117
3.5 Change of Variables in Multiple Integrals

Given the difficulty of evaluating multiple integrals, the reader may be wondering if it is
possible to simplify those integrals using a suitable substitution for the variables. The an-
swer is yes, though it is a bit more complicated than the substitution method which you
learned in single-variable calculus.
Recall that if you are given, for example, the definite integral
Z2 p
x3 x2 1 dx ,
1
then you would make the substitution
u = x2 1 x2 = u + 1
du = 2 x dx
which changes the limits of integration
x=1 u=0
x=2 u=3
so that we get
Z2 p Z2 p
x3 x2 1 dx = 1 2
2 x 2x x2 1 dx
1 1
Z3
1 p
= 2 ( u + 1) u du
0
Z3
= 1
2 u3/2 + u1/2 du , which can be easily integrated to give
0
p
14 3
= 5 .
Let us take a different look at what happened when we did that substitution, which will give
some motivation for how substitution works in multiple integrals. First, we let u = x2 1.
On the interval of integration [1, 2], the function x 7 x2 1 is strictly increasing (and maps
[1, 2] onto [0, 3]) and hence has an inverse function (defined on the interval [0, 3]). That is,
on [0, 3] we can define x as a function of u, namely
p
x = g( u) = u+1 .
p
Then substituting that expression for x into the function f ( x) = x3 x2 1 gives
p
f ( x) = f ( g( u)) = ( u + 1)3/2 u ,
and we see that
dx
= g ( u) dx = g ( u) du
du
dx = 21 ( u + 1)1/2 du ,
so since
g(0) = 1 0 = g1 (1)
g(3) = 2 3 = g1 (2)
then performing the substitution as we did earlier gives
Z2 Z2 p
f ( x) dx = x3 x2 1 dx
1 1
Z3
1 p
= 2 ( u + 1) u du , which can be written as
0
Z3
p
= ( u + 1)3/2 u 21 ( u + 1)1/2 du , which means
0
Z2 Z g1 (2)
f ( x) dx = f ( g( u)) g ( u) du .
1 g1 (1)
In general, if x = g( u) is a one-to-one, differentiable function from an interval [ c, d ] (which

you can think of as being on the u-axis) onto an interval [a, b] (on the x-axis), which means
that g ( u) 6= 0 on the interval ( c, d ), so that a = g( c) and b = g( d ), then c = g1 (a) and d =
g1 ( b), and
Zb Z g 1 ( b )
f ( x) dx = f ( g( u)) g ( u) du . (3.17)
a g 1 ( a )
This is called the change of variable formula for integrals of single-variable functions, and it
is what you were implicitly using when doing integration by substitution. This formula turns
out to be a special case of a more general formula which can be used to evaluate multiple
integrals. We will state the formulas for double and triple integrals involving real-valued
functions of two and three variables, respectively. We will assume that all the functions
involved are continuously differentiable and that the regions and solids involved all have
reasonable boundaries. The proof of the following theorem is beyond the scope of the text.2
2 See T AYLOR and M ANN, 15.32 and 15.62 for all the details.
Theorem 3.1. Change of Variables Formula for Multiple Integrals

Let x = x( u, v) and y = y( u, v) define a one-to-one mapping of a region R in the uv-plane onto
a region R in the x y-plane such that the determinant

x x

u v
J ( u, v) =
(3.18)
y y

u v
is never 0 in R . Then

f ( x, y) d A ( x, y) = f ( x( u, v), y( u, v)) | J ( u, v) | d A ( u, v) . (3.19)
R R
We use the notation d A ( x, y) and d A ( u, v) to denote the area element in the ( x, y) and ( u, v)
coordinates, respectively.
Similarly, if x = x( u, v, w), y = y( u, v, w) and z = z( u, v, w) define a one-to-one mapping of
a solid S in uvw-space onto a solid S in x yz-space such that the determinant

x x x

u v w

y y y
J ( u, v, w) = (3.20)
u v w

z z z

u v w
is never 0 in S , then

f ( x, y, z) dV ( x, y, z) = f ( x( u, v, w), y( u, v, w), z( u, v, w)) | J ( u, v, w) | dV ( u, v, w) . (3.21)
S S
The determinant J ( u, v) in formula (3.18) is called the Jacobian of x and y with respect
to u and v, and is sometimes written as
( x, y)
J ( u, v) = . (3.22)
( u, v)
Similarly, the Jacobian J ( u, v, w) of three variables is sometimes written as
( x, y, z)
J ( u, v, w) = . (3.23)
( u, v, w)
Notice that formula (3.19) is saying that d A ( x, y) = | J ( u, v) | d A ( u, v), which you can think of
as a two-variable version of the relation dx = g ( u) du in the single-variable case.
The following example shows how the change of variables formula is used.
x y
Example 3.9. Evaluate e x+ y d A , where R = {( x, y) : x 0, y 0, x + y 1}.
R
Solution: First, note that evaluating this double integral without using substitution is prob-
ably impossible, at least in a closed form. By looking at the numerator and denominator of
the exponent of e, we will try the substitution u = x y and v = x + y. To use the change of
variables formula (3.19), we need to write both x and y in terms of u and v. So solving for
x and y gives x = 21 ( u + v) and y = 12 (v u). In Figure 3.5.1 below, we see how the mapping
x = x( u, v) = 21 ( u + v), y = y( u, v) = 21 (v u) maps the region R onto R in a one-to-one manner.
y v
x = 21 (u + v) 1
1
1
y= 2 (v u) R
x+ y =1
u = v u=v
R x u
0 1 1 0 1
Figure 3.5.1 The regions R and R
Now we see that

x x 1
1
u v 2 2 1 1 1
J ( u, v) = = = | J ( u, v) | = = ,
y y 1 1 2 2 2
2 2
u v
so using horizontal slices in R , we have

x y
e x+ y d A = f ( x( u, v), y( u, v)) | J ( u, v) | d A
R R
Z1 Z v
u
1
= ev 2 du dv
0 v
Z1 u=v
v uv
= 2 e dv
0 u=v
Z1
v 1
= 2 ( e e ) dv
0
2
v 1 1 1 e2 1
( e e 1 ) =

= e =
4 0 4 e 4e
The change of variables formula can be used to evaluate double integrals in polar coordi-
nates. Letting
x = x( r, ) = r cos and y = y( r, ) = r sin ,
we have

x x
cos
r r sin
= r cos2 + r sin2 = r | J ( u, v) | = | r | = r ,

J ( u, v) = =
y y sin r cos

r
so we have the following formula:
Double Integral in Polar Coordinates

f ( x, y) dx d y = f ( r cos , r sin ) r dr d , (3.24)
R R
where the mapping x = r cos , y = r sin maps the region R in the r -plane onto the
region R in the x y-plane in a one-to-one manner.
Example 3.10. Find the volume V inside the paraboloid z = x2 + y2 for 0 z 1.
Solution: Using vertical slices, we see that z

x 2 + y2 = 1
1
V = (1 z) d A = (1 ( x2 + y2 )) d A ,
R R
where R = {( x, y) : x2 + y2 1} is the unit disk in R2 (see

Figure 3.5.2). In polar coordinates ( r, ) we know that
x2 + y2 = r 2 and that the unit disk R is the set R = {( r, ) :
y
0 r 1, 0 2}. Thus,
0
Z2 Z1 x
V = (1 r 2 ) r dr d Figure 3.5.2 z = x 2 + y2
0 0
Z2 Z1
= ( r r 3 ) dr d
0 0
Z2 r =1
r2 r4
= 2 4 d
0 r =0
Z2
1
= 4 d
0

=
2
p
Example 3.11. Find the volume V inside the cone z = x2 + y2 for 0 z 1.
Solution: Using vertical slices, we see that z x 2 + y2 = 1

q 1
V = (1 z) d A = 2 2
1 x + y dA ,
R R
where R = {( x, y) : x2 + y2 1} is the unit disk in R2

y
(see Figure
p 3.5.3). In polar coordinates ( r, ) we know 0
that x2 + y2 = r and that the unit disk R is the set x
R = {( r, ) : 0 r 1, 0 2}. Thus, p
Figure 3.5.3 z= x 2 + y2
Z2 Z1
V = (1 r ) r dr d
0 0
Z2 Z1
= ( r r 2 ) dr d
0 0
Z2 r =1
r2 r3
= 2 3 r =0 d
0
Z2
1
= 6 d
0

=
3
In a similar fashion, it can be shown (see Exercises 5-6) that triple integrals in cylindrical
and spherical coordinates take the following forms:
Triple Integral in Cylindrical Coordinates

f ( x, y, z) dx d y dz = f ( r cos , r sin , z) r dr d dz , (3.25)
S S
where the mapping x = r cos , y = r sin , z = z maps the solid S in r z-space onto the
solid S in x yz-space in a one-to-one manner.
Triple Integral in Spherical Coordinates

f ( x, y, z) dx d y dz = f ( sin cos , sin sin , cos ) 2 sin d d d , (3.26)
S S
where the mapping x = sin cos , y = sin sin , z = cos maps the solid S in -
space onto the solid S in x yz-space in a one-to-one manner.
Example 3.12. For a > 0, find the volume V inside the sphere S = x2 + y2 + z2 = a2 .
Solution: We see that S is the set = a in spherical coordinates, so
Z2 Z Za
V = 1 dV = 1 2 sin d d d
0 0 0
S
Z2 Z 3 Z2 Z
= a a3
= sin d d
sin d d =
0 0 3 =0 0 0 3
Z2 3 = Z2 3
a 2a 4 a 3
= cos d = d = .
0 3 = 0 0 3 3
Exercises
A
1. Find the volume V inside the paraboloid z = x2 + y2 for 0 z 4.
p
2. Find the volume V inside the cone z = x2 + y2 for 0 z 3.
B
3. Find the volume V of the solid inside both x2 + y2 + z2 = 4 and x2 + y2 = 1.
p
4. Find the volume V inside both the sphere x2 + y2 + z2 = 1 and the cone z = x 2 + y2 .
5. Prove formula (3.25). 6. Prove formula (3.26).

x+ y x y
7. Evaluate sin 2 cos 2 d A , where R is the triangle with vertices (0, 0), (2, 0) and
R
(1, 1). (Hint: Use the change of variables u = ( x + y)/2, v = ( x y)/2.)
8. Find the volume of the solid bounded by z = x2 + y2 and z2 = 4( x2 + y2 ).

x2 y2
9. Find the volume inside the elliptic cylinder a2
+ b2 = 1 for 0 z 2.
C
2 y2 2
10. Show that the volume inside the ellipsoid ax2 + b2 + zc2 = 1 is 43abc . (Hint: Use the change
of variables x = au, y = bv, z = cw, then consider Example 3.12.)
11. Show that the Beta function, defined by

Z1
B( x, y) = t x1 (1 t) y1 dt , for x > 0, y > 0,
0
satisfies the relation B( y, x) = B( x, y) for x > 0, y > 0.
12. Using the substitution t = u/( u + 1), show that the Beta function can be written as
Z
u x 1
B( x, y) = x+ y
du , for x > 0, y > 0.
0 ( u + 1)
3.6 Application: Center of Mass

y
Recall from single-variable calculus that for a region
R = {( x, y) : a x b, 0 y f ( x)} in R2 that represents y = f (x)
a thin, flat plate (see Figure 3.6.1), where f ( x) is a con-
tinuous function on [a, b], the center of mass of R has R
coordinates ( x, y ) given by y )
( x, x
0 a b
My Mx
x = and y = ,
M M Figure 3.6.1 Center of mass of R
where
Zb Zb Zb
( f ( x))2
Mx = dx , My = x f ( x) dx , M= f ( x) dx , (3.27)
a 2 a a
assuming that R has uniform density, i.e the mass of R is uniformly distributed over the
region. In this case the area M of the region is considered the mass of R (the density is
constant, and taken as 1 for simplicity).
In the general case where the density of a region (or lamina) R is a continuous function
= ( x, y) of the coordinates ( x, y) of points inside R (where R can be any region in R2 ) the
coordinates ( x, y ) of the center of mass of R are given by
My Mx
x = and y = , (3.28)
M M
where
My = x( x, y) d A , Mx = y( x, y) d A , M= ( x, y) d A , (3.29)
R R R
The quantities M x and M y are called the moments (or first moments) of the region R about
the x-axis and y-axis, respectively. The quantity M is the mass of the region R . To see this,
think of taking a small rectangle inside R with dimensions x and y close to 0. The mass
of that rectangle is approximately ( x , y ) x y, for some point ( x , y ) in that rectangle.
Then the mass of R is the limit of the sums of the masses of all such rectangles inside R as

the diagonals of the rectangles approach 0, which is the double integral ( x, y) d A .
R
Note that the formulas in (3.27) represent a special case when ( x, y) = 1 throughout R in
the formulas in (3.29).
Example 3.13. Find the center of mass of the region R = {( x, y) : 0 x 1, 0 y 2 x2 }, if the

density function at ( x, y) is ( x, y) = x + y.
3.6 Application: Center of Mass 125
y
Solution: The region R is shown in Figure 3.6.2. We have
y = 2x2

M = ( x, y) d A
R
R
Z1 Z2 x2 x
= ( x + y) d y dx 0 1
0 0
Z1 y=2 x2 ! Figure 3.6.2
y2
= xy+ dx
0 2 y=0
Z1
= (2 x3 + 2 x4 ) dx
0
4 1
x 2 x5 9
= + =
2 5 0 10
and

Mx = y( x, y) d A My = x( x, y) d A
R R
Z1 Z2 x2 Z1 Z2 x2
= y( x + y) d y dx = x( x + y) d y dx
0 0 0 0
Z1 y=2 x2 ! Z1 2 y=2 x
2!
x y2 y3 x y
= + dx = x2 y + dx
0 2 3 y=0 0 2 y=0
Z1 6 Z1
8x
= (2 x5 +
) dx = (2 x4 + 2 x5 ) dx
0 3 0
1 1
x6 8 x7 5 2 x5 x6 11
= + = = + = ,
3 21 7 0 5 3 0 15
y ) is given by
so the center of mass ( x,
My 11/15 22 Mx 5/7 50
x = = = , y = = = .
M 9/10 27 M 9/10 63
Note how this center of mass is a little further towards the upper corner of the region R than

y ) = 34 , 53
when the density is uniform (it is easy to use the formulas in (3.27) to show that ( x,
in that case). This makes sense since the density function ( x, y) = x + y increases as ( x, y)
approaches that upper corner, where there is quite a bit of area.
In the special case where the density function ( x, y) is a constant function on the region
y ) is called the centroid of R .
R , the center of mass ( x,
The formulas for the center of mass of a region in R2 can be generalized to a solid S in R3 .
Let S be a solid with a continuous mass density function ( x, y, z) at any point ( x, y, z) in S .
y , z ), where
Then the center of mass of S has coordinates ( x,
M yz M xz Mx y
x = , y = , z = , (3.30)
M M M
where

M yz = x( x, y, z) dV , M xz = y( x, y, z) dV , Mx y = z( x, y, z) dV , (3.31)
S S S

M = ( x, y, z) dV . (3.32)
S
In this case, M yz , M xz and M x y are called the moments (or first moments) of S around the
yz-plane, xz-plane and x y-plane, respectively. Also, M is the mass of S .
Example 3.14. Find the center of mass of the solid S = {( x, y, z) : z 0, x2 + y2 + z2 a2 }, if

the density function at ( x, y, z) is ( x, y, z) = 1.
Solution: The solid S is just the upper hemisphere inside the sphere z
a
of radius a centered at the origin (see Figure 3.6.3). So since the
density function is a constant and S is symmetric about the z-axis,
y , z)
( x,
then it is clear that x = 0 and y = 0, so we need only find z . We have y
0 a
M = ( x, y, z) dV = 1 dV = V olume(S ).
S S
x
Figure 3.6.3
But since the volume of S is half the volume of the sphere of radius
3 3
a, which we know by Example 3.12 is 43a , then M = 23a . And

Mx y = z( x, y, z) dV
S

= z dV , which in spherical coordinates is
S
Z2 Z/2 Za
= ( cos ) 2 sin d d d
0 0 0
Z2 Z/2 Za
= sin cos 3 d d d
0 0 0
Z2 Z/2
a4
= 4 sin cos d d
0 0
3.6 Application: Center of Mass 127
Z2 Z/2
a4
Mx y = 8 sin 2 d d (since sin 2 = 2 sin cos )
0 0
Z2 =/2
a4
= 16 cos 2 d
0 = 0
Z2
a4
= 8 d
0
a4
= ,
4
so
a4
Mx y 4 3a
z = = = .
M 2a3 8
3

y , z ) = 0, 0, 38a .
Thus, the center of mass of S is ( x,
Exercises
A
For Exercises 1-5, find the center of mass of the region R with the given density function
( x, y).
1. R = {( x, y) : 0 x 2, 0 y 4 }, ( x, y) = 2 y
2. R = {( x, y) : 0 x 1, 0 y x2 }, ( x, y) = x + y
3. R = {( x, y) : y 0, x2 + y2 a2 }, ( x, y) = 1
p
4. R = {( x, y) : y 0, x 0, 1 x2 + y2 4 }, ( x, y) = x 2 + y2
5. R = {( x, y) : y 0, x2 + y2 1 }, ( x, y) = y
B
For Exercises 6-10, find the center of mass of the solid S with the given density function
( x, y, z).
6. S = {( x, y, z) : 0 x 1, 0 y 1, 0 z 1 }, ( x, y, z) = x yz
7. S = {( x, y, z) : z 0, x2 + y2 + z2 a2 }, ( x, y, z) = x2 + y2 + z2
8. S = {( x, y, z) : x 0, y 0, z 0, x2 + y2 + z2 a2 }, ( x, y, z) = 1
9. S = {( x, y, z) : 0 x 1, 0 y 1, 0 z 1 }, ( x, y, z) = x2 + y2 + z2
10. S = {( x, y, z) : 0 x 1, 0 y 1, 0 z 1 x y}, ( x, y, z) = 1
3.7 Application: Probability and Expected Value
In this section we will briefly discuss some applications of multiple integrals in the field of
probability theory. In particular we will see ways in which multiple integrals can be used to
calculate probabilities and expected values.
Probability
Suppose that you have a standard six-sided (fair) die, and you let a variable X represent
the value rolled. Then the probability of rolling a 3, written as P ( X = 3), is 16 , since there
are six sides on the die and each one is equally likely to be rolled, and hence in particular
the 3 has a one out of six chance of being rolled. Likewise the probability of rolling at most a
3, written as P ( X 3), is 36 = 21 , since of the six numbers on the die, there are three equally
likely numbers (1, 2, and 3) that are less than or equal to 3. Note that P ( X 3) = P ( X =
1) + P ( X = 2) + P ( X = 3). We call X a discrete random variable on the sample space (or
probability space) consisting of all possible outcomes. In our case, = {1, 2, 3, 4, 5, 6}. An
event A is a subset of the sample space. For example, in the case of the die, the event X 3
is the set {1, 2, 3}.
Now let X be a variable representing a random real number in the interval (0, 1). Note
that the set of all real numbers between 0 and 1 is not a discrete (or countable) set of values,
i.e. it can not be put into a one-to-one correspondence with the set of positive integers.3 In
this case, for any real number x in (0, 1), it makes no sense to consider P ( X = x) since it must
be 0 (why?). Instead, we consider the probability P ( X x), which is given by P ( X x) = x.
The reasoning is this: the interval (0, 1) has length 1, and for x in (0, 1) the interval (0, x)
has length x. So since X represents a random number in (0, 1), and hence is uniformly
distributed over (0, 1), then
length of (0, x) x
P ( X x) = = = x.
length of (0, 1) 1
We call X a continuous random variable on the sample space = (0, 1). An event A is a
subset of the sample space. For example, in our case the event X x is the set (0, x).
In the case of a discrete random variable, we saw how the probability of an event was the
sum of the probabilities of the individual outcomes comprising that event (e.g. P ( X 3) =
P ( X = 1) + P ( X = 2) + P ( X = 3) in the die example). For a continuous random variable, the
probability of an event will instead be the integral of a function, which we will now describe.
Let X be a continuous real-valued random variable on a sample space in R. For sim-
3 For a proof see p. 9-10 in K AMKE , E., Theory of Sets, New York: Dover, 1950.
3.7 Application: Probability and Expected Value 129
plicity, let = (a, b). Define the distribution function F of X as
F ( x) = P ( X x) , for < x < (3.33)

1, for x b

= P ( X x), for a < x < b (3.34)

0, for x a .
Suppose that there is a nonnegative, continuous real-valued function f on R such that

Zx
F ( x) = f ( y) d y , for < x < , (3.35)

and Z
f ( x) dx = 1 . (3.36)

Then we call f the probability density function (or p.d.f. for short) for X . We thus have
Zx
P ( X x) = f ( y) d y , for a < x < b . (3.37)
a
Also, by the Fundamental Theorem of Calculus, we have
F ( x) = f ( x) , for < x < . (3.38)
Example 3.15. Let X represent a randomly selected real number in the interval (0, 1). We
say that X has the uniform distribution on (0, 1), with distribution function

1, for x 1

F ( x) = P ( X x) = x, for 0 < x < 1 (3.39)

0, for x 0 ,
and probability density function

(
1, for 0 < x < 1
f ( x) = F ( x) = (3.40)
0, elsewhere.
In general, if X represents a randomly selected real number in an interval (a, b), then X has
the uniform distribution function

1, for x b

x
F ( x) = P ( X x) = ba , for a < x < b (3.41)

0, for x a ,
and probability density function

(
1
ba , for a < x < b
f ( x) = F ( x) = (3.42)
0, elsewhere.
Example 3.16. A famous distribution function is given by the standard normal distribution,
whose probability density function f is
1 2
f ( x) = p e x /2 , for < x < . (3.43)
2
This is often called a bell curve, and is used widely in statistics. Since we are claiming that
f is a p.d.f., we should have Z
1 2
p e x /2 dx = 1 (3.44)
2
by formula (3.36), which is equivalent to

Z p
2
e x /2 dx = 2 . (3.45)

We can use a double integral in polar coordinates to verify this integral. First,
Z Z Z Z
( x2 + y2 )/2 y2 /2 x2 /2
e dx d y = e e dx d y

Z Z
2 2
/2 /2
= e x dx e y dy

Z 2
2
/2
= e x dx

since the same function is being integrated twice in the middle equation, just with different
variables. But using polar coordinates, we see that
Z Z Z2 Z
2
+ y2 )/2 2
e ( x dx d y = er /2
r dr d
0 0
Z2 r=
r 2 /2

= e d
0 r =0
Z2 Z2
= (0 ( e0 )) d = 1 d = 2 ,
0 0
and so
Z 2
x2 /2
e dx = 2 , and hence

Z p
2
/2
e x dx = 2 .

In addition to individual random variables, we can consider jointly distributed random

variables. For this, we will let X , Y and Z be three real-valued continuous random variables
defined on the same sample space in R (the discussion for two random variables is similar).
Then the joint distribution function F of X , Y and Z is given by
F ( x, y, z) = P ( X x, Y y, Z z) , for < x, y, z < . (3.46)
If there is a nonnegative, continuous real-valued function f on R3 such that

Zz Z y Zx
F ( x, y, z) = f ( u, v, w) du dv dw , for < x, y, z < (3.47)

and Z Z Z
f ( x, y, z) dx d y dz = 1 , (3.48)

then we call f the joint probability density function (or joint p.d.f. for short) for X , Y and Z .
In general, for a 1 < b 1 , a 2 < b 2 , a 3 < b 3 , we have
Zb3 Zb2 Zb1
P (a 1 < X b 1 , a 2 < Y b 2 , a 3 < Z b 3 ) = f ( x, y, z) dx d y dz , (3.49)
a3 a2 a1
with the and < symbols interchangeable in any combination. A triple integral, then, can
be thought of as representing a probability (for a function f which is a p.d.f.).
Example 3.17. Let a, b, and c be real numbers selected randomly from the interval (0, 1).
What is the probability that the equation ax2 + bx + c = 0 has at least one real solution x?
Solution: We know by the quadratic formula that there is at least c

one real solution if b2 4ac 0. So we need to calculate P ( b2 4ac
1
0). We will use three jointly distributed random variables to do this.
1
First, since 0 < a, b, c < 1, we have c= 4a
p p
b2 4ac 0 0 < 4ac b2 < 1 0 < 2 a c b < 1 , R1 R2
a
0 1 1
where the last relation holds for all 0 < a, c < 1 such that 4
Figure 3.7.1 Region

1
0 < 4ac < 1 0 < c < . R = R1 R2
4a
Considering a, b and c as real variables, the region R in the ac-plane where the above
relation holds is given by R = {(a, c) : 0 < a < 1, 0 < c < 1, 0 < c < 41a }, which we can see is a
union of two regions R 1 and R 2 , as in Figure 3.7.1 above.
Now let X , Y and Z be continuous random variables, each representing a randomly se-
lected real number from the interval (0, 1) (think of X , Y and Z representing a, b and c,
respectively). Then, similar to how we showed that f ( x) = 1 is the p.d.f. of the uniform dis-
tribution on (0, 1), it can be shown that f ( x, y, z) = 1 for x, y, z in (0, 1)
(0 elsewhere) is the joint p.d.f. of X , Y and Z . Now,
p p
P ( b2 4ac 0) = P ((a, c) R, 2 a c b < 1) ,
p p
so this probability is the triple integral of f (a, b, c) = 1 as b varies from 2 a c to 1 and as
(a, c) varies over the region R . Since R can be divided into two regions R 1 and R 2 , then the
required triple integral can be split into a sum of two triple integrals, using vertical slices in
R:
Z1/4 Z1 Z1 Z1 Z1/4a Z1
2
P ( b 4ac 0) = p p 1 db dc da + p p 1 db dc da
| 0 {z 0} 2 a c | 1/4 {z0 } 2 a c
R1 R2
Z1/4 Z1 Z1 Z1/4a
p p p p
= (1 2 a c) dc da + (1 2 a c) dc da
0 0 1/4 0
Z1/4 c =1 Z1 c=1/4a
p p
c 43 a c3/2 c 34 a c3/2

= da + da
0 c =0 1/4 c =0
Z1/4 Z1
p
= 1 43 a da + 1
12a da
0 1/4
1/4 1
8 3/2 1
= a a + ln a
9 0 12
1/4
1 1 1 1 5 1
= + 0 ln = + ln 4
4 9 12 4 36 12
5 + 3 ln 4
P ( b2 4ac 0) = 0.2544
36
In other words, the equation ax2 + bx + c = 0 has about a 25% chance of being solved!
Expected Value
The expected value E X of a random variable X can be thought of as the average value
of X as it varies over its sample space. If X is a discrete random variable, then
X
EX = x P ( X = x) , (3.50)
x
with the sum being taken over all elements x of the sample space. For example, if X repre-
sents the number rolled on a six-sided die, then
6
X 6
X 1
EX = x P ( X = x) = x = 3.5 (3.51)
x =1 x =1 6
is the expected value of X , which is the average of the integers 1 6.

If X is a real-valued continuous random variable with p.d.f. f , then

Z
EX = x f ( x) dx . (3.52)

For example, if X has the uniform distribution on the interval (0, 1), then its p.d.f. is
(
1, for 0 < x < 1
f ( x) = (3.53)
0, elsewhere,
and so Z Z1
1
EX = x f ( x) dx = .
x dx = (3.54)
0 2
For a pair of jointly distributed, real-valued continuous random variables X and Y with
joint p.d.f. f ( x, y), the expected values of X and Y are given by
Z Z Z Z
EX = x f ( x, y) dx d y and EY = y f ( x, y) dx d y , (3.55)

respectively.
Example 3.18. If you were to pick n > 2 random real numbers from the interval (0, 1), what
are the expected values for the smallest and largest of those numbers?
Solution: Let U1 , . . . ,Un be n continuous random variables, each representing a randomly
selected real number from (0, 1), i.e. each has the uniform distribution on (0, 1). Define
random variables X and Y by
X = min(U1 , . . . ,Un ) and Y = max(U1 , . . . ,Un ) .
Then it can be shown4 that the joint p.d.f. of X and Y is

(
n( n 1)( y x)n2 , for 0 x y 1
f ( x, y) = (3.56)
0, elsewhere.
Thus, the expected value of X is

Z1 Z1
EX = n( n 1) x( y x)n2 d y dx
0 x
Z1 y=1
n 1
= nx( y x) dx
0 y= x
Z1
= nx(1 x)n1 dx , so integration by parts yields
0
1 1
= x(1 x)n (1 x)n+1

n+1 0
1
EX = ,
n+1
4 See Ch. 6 in H OEL, P ORT and S TONE.
and similarly (see Exercise 3) it can be shown that

Z1 Z y
n
EY = n( n 1) y( y x)n2 dx d y = .
0 0 n+1
So, for example, if you were to repeatedly take samples of n = 3 random real numbers from
(0, 1), and each time store the minimum and maximum values in the sample, then the aver-
age of the minimums would approach 14 and the average of the maximums would approach
3
4 as the number of samples grows. It would be relatively simple (see Exercise 4) to write a
computer program to test this.
Exercises
B
R x2
1. Evaluate the integral edx using anything you have learned so far.
R 2 2
2. For > 0 and > 0, evaluate p1 e( x) /2 dx.
2
n
3. Show that EY = n +1 in Example 3.18
C
4. Write a computer program (in the language of your choice) that verifies the results in
Example 3.18 for the case n = 3 by taking large numbers of samples.
5. Repeat Exercise 4 for the case when n = 4.
6. For continuous random variables X , Y with joint p.d.f. f ( x, y), define the second moments
E ( X 2 ) and E (Y 2 ) by
Z Z Z Z
E( X 2) = x2 f ( x, y) dx d y and E (Y 2 ) = y2 f ( x, y) dx d y ,

and the variances Var( X ) and Var(Y ) by
Var( X ) = E ( X 2 ) (E X )2 and Var(Y ) = E (Y 2 ) (EY )2 .
Find Var( X ) and Var(Y ) for X and Y as in Example 3.18.
7. Continuing Exercise 6, the correlation between X and Y is defined as

E ( X Y ) (E X )(EY )
= p ,
Var( X ) Var(Y )
R R
where E ( X Y ) = x y f ( x, y) dx d y. Find for X and Y as in Example 3.18.
(Note: The quantity E ( X Y ) (E X )(EY ) is called the covariance of X and Y .)
8. In Example 3.17 would the answer change if the interval (0, 100) is used instead of (0, 1)?
Explain.
4 Line and Surface Integrals
4.1 Line Integrals
In single-variable calculus you learned how to integrate a real-valued function f ( x) over an
interval [a, b] in R1 . This integral (usually called a Riemann integral) can be thought of as
an integral over a path in R1 , since an interval (or collection of intervals) is really the only
kind of path in R1 . You may also recall that if f ( x) represented the force applied along the
x-axis to an object at position x in [a, b], then the work W done in moving that object from
position x = a to x = b was defined as the integral:
Zb
W = f ( x) dx
a
In this section, we will see how to define the integral of a function (either real-valued or
vector-valued) of two variables over a general path (i.e. a curve) in R2 . This definition will
be motivated by the physical notion of work. We will begin with real-valued functions of two
variables.
In physics, the intuitive idea of work is that
Work = Force Distance .
Suppose that we want to find the total amount W of work done in moving an object along a
curve C in R2 with a smooth parametrization x = x( t), y = y( t), a t b, with a force f ( x, y)
which varies with the position ( x, y) of the object and is applied in the direction of motion
along C (see Figure 4.1.1 below).
y t = ti q
C
s i xi 2 + yi 2
t=a yi
t = t i +1
xi
t=b
x
0
Figure 4.1.1 Curve C : x = x(t), y = y(t) for t in [a, b]
We will assume for now that the function f ( x, y) is continuous and real-valued, so we only
consider the magnitude of the force. Partition the interval [a, b] as follows:
a = t 0 < t 1 < t 2 < < t n1 < t n = b , for some integer n 2
135
136 CHAPTER 4. LINE AND SURFACE INTEGRALS
q a typical subinterval [ t i , t i+1 ] the distance s i traveled

As we can see from Figure 4.1.1, over
along the curve is approximately xi 2 + yi 2 , by the Pythagorean Theorem. Thus, if the
subinterval is small enough then the work done in moving the object along that piece of the
curve is approximately
q
Force Distance f ( xi , yi ) xi 2 + yi 2 , (4.1)
where ( xi , yi ) = ( x( t i ), y( t i )) for some t i in [ t i , t i+1 ], and so
1
nX q
W f ( xi , yi ) xi 2 + yi 2 (4.2)
i =0
is approximately the total amount of work done over the entire curve. But since
s
q 2 2
xi yi
xi 2 + yi 2 = + t i ,
t i t i
where t i = t i+1 t i , then

s
1
nX 2 2
xi yi
W f ( xi , yi ) + t i . (4.3)
i =0 t i t i
Taking the limit of that sum as the length of the largest subinterval goes to 0, the sum over
xi yi
all subintervals becomes the integral from t = a to t = b,
t i and t i become x ( t) and y ( t),
respectively, and f ( xi , yi ) becomes f ( x( t), y( t)), so that
Zb q
W = f ( x( t), y( t)) x ( t)2 + y ( t)2 dt . (4.4)
a
The integral on the right side of the above equation gives us our idea of how to define,
for any real-valued function f ( x, y), the integral of f ( x, y) along the curve C , called a line
integral:
Definition 4.1. For a real-valued function f ( x, y) and a curve C in R2 , parametrized by

x = x( t), y = y( t), a t b, the line integral of f ( x, y) along C with respect to arc length
s is Z Z bq
f ( x, y) ds = f ( x( t), y( t)) x ( t)2 + y ( t)2 dt . (4.5)
C a
The symbol ds is the differential of the arc length function

Zt q
s = s( t) = x ( u)2 + y ( u)2 du , (4.6)
a
4.1 Line Integrals 137
which you may recognize from Section 1.9 as the length of the curve C over the interval [a, t],
for all t in [a, b]. That is,
q

ds = s ( t) dt = x ( t)2 + y ( t)2 dt , (4.7)
by the Fundamental Theorem of Calculus.

R
For a general real-valued function f ( x, y), what does the line integral C f ( x, y) ds rep-
resent? The preceding discussion of ds gives us a clue. You can think of differentials as
infinitesimal lengths. So if you think of f ( x, y) as the height of a picket fence along C , then
f ( x, y) ds can be thought of as approximately the area of a section of that fence over some
R
infinitesimally small section of the curve, and thus the line integral C f ( x, y) ds is the total
area of that picket fence (see Figure 4.1.2).
f ( x, y)
C ds
x
0
Figure 4.1.2 Area of shaded rectangle = height width f (x, y) ds
Example 4.1. Use a line integral to show that the lateral surface area A of a right circular
cylinder of radius r and height h is 2 rh.
Solution: We will use the right circular cylinder with base circle C z
given by x2 + y2 = r 2 and with height h in the positive z direction r
(see Figure 4.1.3). Parametrize C as follows:
x = x( t) = r cos t , y = y( t) = r sin t , 0 t 2 h = f (x, y)
Let f ( x, y) = h for all ( x, y). Then y

0
Z Zb q
A = f ( x, y) ds = f ( x( t), y( t)) x ( t)2 + y ( t)2 dt
x C : x 2 + y2 = r 2
C a
Z2 p Figure 4.1.3
= h ( r sin t)2 + ( r cos t)2 dt
0
Z2 p
= h r sin2 t + cos2 t dt
0
Z2
= rh 1 dt = 2 rh
0
Note in Example 4.1 that if we had traversed the circle C twice, i.e. let t vary from 0 to
4, then we would have gotten an area of 4 rh, i.e. twice the desired area, even though the
curve itself is still the same (namely, a circle of radius r ). Also, notice that we traversed the
circle in the counter-clockwise direction. If we had gone in the clockwise direction, using the
parametrization
x = x( t) = r cos(2 t) , y = y( t) = r sin(2 t) , 0 t 2 , (4.8)
then it is easy to verify (see Exercise 12) that the value of the line integral is unchanged.
In general, it can be shown (see Exercise 15) that reversing the direction in which a curve
R
C is traversed leaves C f ( x, y) ds unchanged, for any f ( x, y). If a curve C has a parametriza-
tion x = x( t), y = y( t), a t b, then denote by C the same curve as C but traversed in the
opposite direction. Then C is parametrized by
x = x( a + b t) , y = y( a + b t ) , atb , (4.9)
and we have Z Z
f ( x, y) ds = f ( x, y) ds . (4.10)
C C
Notice that our definition of the line integral was with respect to the arc length parameter
s. We can also define
Z Zb
f ( x, y) dx = f ( x( t), y( t)) x ( t) dt (4.11)
C a
as the line integral of f ( x, y) along C with respect to x, and
Z Zb
f ( x, y) d y = f ( x( t), y( t)) y ( t) dt (4.12)
C a
as the line integral of f ( x, y) along C with respect to y.

In the derivation of the formula for a line integral, we used the idea of work as force
multiplied by distance. However, we know that force is actually a vector. So it would be
helpful to develop a vector form for a line integral. For this, suppose that we have a function
f( x, y) defined on R2 by
f( x, y) = P ( x, y) i + Q ( x, y) j
for some continuous real-valued functions P ( x, y) and Q ( x, y) on R2 . Such a function f is

called a vector field on R2 . It is defined at points in R2 , and its values are vectors in R2 . For
a curve C with a smooth parametrization x = x( t), y = y( t), a t b, let
r( t) = x( t) i + y( t) j
be the position vector for a point ( x( t), y( t)) on C . Then r ( t) = x ( t) i + y ( t) j and so

Z Z Zb Zb
P ( x, y) dx + Q ( x, y) d y = P ( x( t), y( t)) x ( t) dt + Q ( x( t), y( t)) y ( t) dt
C C a a
Zb
= (P ( x( t), y( t)) x ( t) + Q ( x( t), y( t)) y ( t)) dt
a
Zb
= f( x( t), y( t)) r ( t) dt
a
by definition of f( x, y). Notice that the function f( x( t), y( t)) r ( t) is a real-valued function on
[a, b], so the last integral on the right looks somewhat similar to our earlier definition of a
line integral. This leads us to the following definition:
Definition 4.2. For a vector field f( x, y) = P ( x, y) i + Q ( x, y) j and a curve C with a smooth

parametrization x = x( t), y = y( t), a t b, the line integral of f along C is
Z Z Z
f dr = P ( x, y) dx + Q ( x, y) d y (4.13)
C C C
Zb
= f( x( t), y( t)) r ( t) dt , (4.14)
a
where r( t) = x( t) i + y( t) j is the position vector for points on C .
We use the notation d r = r ( t) dt = dx i + d y j to denote the differential of the vector-valued

function r. The line integral in Definition 4.2 is often called a line integral of a vector field
to distinguish it from the line integral in Definition 4.1 which is called a line integral of a
scalar field. For convenience we will often write
Z Z Z
P ( x, y) dx + Q ( x, y) d y = P ( x, y) dx + Q ( x, y) d y ,
C C C
where it is understood that the line integral along C is being applied to both P and Q . The
quantity P ( x, y) dx + Q ( x, y) d y is known as a differential form. For a real-valued function
F ( x, y), the differential of F is dF = Fx dx + Fy d y. A differential form P ( x, y) dx + Q ( x, y) d y
is called exact if it equals dF for some function F ( x, y).
Recall that if the points on a curve C have position vector r( t) = x( t) i + y( t) j, then r ( t) is a
tangent vector to C at the point ( x( t), y( t)) in the direction of increasing t (which we call the
direction of C ). Since C is a smooth curve, then r ( t) 6= 0 on [a, b] and hence
r ( t)
T( t) =
r ( t)
is the unit tangent vector to C at ( x( t), y( t)). Putting Definitions 4.1 and 4.2 together we get
the following theorem:
Theorem 4.1. For a vector field f( x, y) = P ( x, y) i + Q ( x, y) j and a curve C with a smooth

parametrization x = x( t), y = y( t), a t b and position vector r( t) = x( t) i + y( t) j,
Z Z
f dr = f T ds , (4.15)
C C
r ( t)
where T( t) = k r ( t) k is the unit tangent vector to C at ( x( t), y( t)).
If the vector field f( x, y) represents the force moving an object along a curve C , then the work
W done by this force is Z Z
W = f T ds = f dr . (4.16)
C C
R 2
Example 4.2. Evaluate C (x + y2 ) dx + 2 x y d y, where:
(a) C : x = t , y = 2t , 0t1
(b) C : x = t , y = 2 t2 , 0t1
Solution: Figure 4.1.4 shows both curves. y

(1, 2)
2
(a) Since x ( t) = 1 and y ( t) = 2, then
Z Z1

( x2 + y2 ) dx + 2 x y d y = ( x( t)2 + y( t)2 ) x ( t) + 2 x( t) y( t) y ( t) dt
C 0
Z1 x
2 2
0
= ( t + 4 t )(1) + 2 t(2 t)(2) dt 1
0
Z1 Figure 4.1.4
= 13 t2 dt
0
1
13 t3 13
= =
3 0 3
(b) Since x ( t) = 1 and y ( t) = 4 t, then

Z Z1
2 2

( x + y ) dx + 2 x y d y = ( x( t)2 + y( t)2 ) x ( t) + 2 x( t) y( t) y ( t) dt
C 0
Z1

= ( t2 + 4 t4 )(1) + 2 t(2 t2 )(4 t) dt
0
Z1
= ( t2 + 20 t4 ) dt
0
31
t
5 1 13
= + 4t = + 4 =
3 0 3 3
So in both cases, if the vector field f( x, y) = ( x2 + y2 ) i + 2 x y j represents the force moving an

object from (0, 0) to (1, 2) along the given curve C , then the work done is 13 3 . This may lead
you to think that work (and more generally, the line integral of a vector field) is independent
of the path taken. However, as we will see in the next section, this is not always the case.
Although we defined line integrals over a single smooth curve, if C is a piecewise smooth
curve, that is
C = C1 C2 . . . C n
is the union of smooth curves C 1 , . . . , C n , then we can define
Z Z Z Z
f dr = f d r1 + f d r2 + . . . + f d rn
C C1 C2 Cn
where each ri is the position vector of the curve C i .
R 2
Example 4.3. Evaluate C (x + y2 ) dx + 2 x y d y, where C is the polygonal path from (0, 0) to
(0, 2) to (1, 2).
y
Solution: Write C = C 1 C 2 , where C 1 is the curve given by x = 0, y = t,
2 (1, 2)
0 t 2 and C 2 is the curve given by x = t, y = 2, 0 t 1 (see Figure C2
4.1.5). Then
Z Z C1
( x2 + y2 ) dx + 2 x y d y = ( x2 + y2 ) dx + 2 x y d y x
C C1
Z 0 1
+ ( x2 + y2 ) dx + 2 x y d y
C2 Figure 4.1.5
Z2 Z1
2
= (02 + t2 )(0) + 2(0) t(1) dt + ( t + 4)(1) + 2 t(2)(0) dt
0 0
Z2 Z1
= 0 dt + ( t2 + 4) dt
0 0
1
t3 1 13
= + 4 t = + 4 =
3 0 3 3
Line integral notation varies quite a bit. For example, in physics it is common to see the
Rb
notation a f d l, where it is understood that the limits of integration a and b are for the
underlying parameter t of the curve, and the letter l signifies length. Also, the formulation
R
C f T ds from Theorem 4.1 is often preferred in physics since it emphasizes the idea of
integrating the tangential component f T of f in the direction of T (i.e. in the direction of C ),
which is a useful physical interpretation of line integrals.
Exercises
A R
For Exercises 1-4, calculate C f ( x, y) ds for the given function f ( x, y) and curve C .
1. f ( x, y) = x y; C : x = cos t, y = sin t, 0 t /2
x
2. f ( x, y) = ; C : x = t, y = 0, 0 t 1
x2 + 1
3. f ( x, y) = 2 x + y; C : polygonal path from (0, 0) to (3, 0) to (3, 2)
4. f ( x, y) = x + y2 ; C : path from (2, 0) counterclockwise along the circle x2 + y2 = 4 to the

point (2, 0) and then back to (2, 0) along the x-axis
5. Use a line integral to find the lateral surface area of the part of the cylinder
x2 + y2 = 4 below the plane x + 2 y + z = 6 and above the x y-plane.
R
For Exercises 6-11, calculate C f d r for the given vector field f( x, y) and curve C .
6. f( x, y) = i j; C : x = 3 t, y = 2 t, 0 t 1
7. f( x, y) = y i x j; C : x = cos t, y = sin t, 0 t 2
8. f( x, y) = x i + y j; C : x = cos t, y = sin t, 0 t 2
9. f( x, y) = ( x2 y) i + ( x y2 ) j; C : x = cos t, y = sin t, 0 t 2
10. f( x, y) = x y2 i + x y3 j; C : the polygonal path from (0, 0) to (1, 0) to (0, 1) to (0, 0)
11. f( x, y) = ( x2 + y2 ) i; C : x = 2 + cos t, y = sin t, 0 t 2
B
12. Verify that the value of the line integral in Example 4.1 is unchanged when using the
parametrization of the circle C given in formulas (4.8).
R
13. Show that if f r ( t) at each point r( t) along a smooth curve C , then C f d r = 0.
14. Show that if f points in the same direction as r ( t) at each point r( t) along a smooth
R R
curve C , then C f d r = C k f k ds.
C
R R
15. Prove that C f ( x, y) ds = C f ( x, y) ds. (Hint: Use formulas (4.9).)
16. Let C be a smooth curve with arc length L, and suppose that f( x, y) = P ( x, y) i + Q ( x, y) j
isRa vector
field such that k f( x, y) k RM for all ( x, R
y) on C . Show that
f d r ML. (Hint: Recall that b g( x) dx b | g( x) | dx for Riemann integrals.)
C a a
Rb
17. Prove that the Riemann integral a f ( x) dx is a special case of a line integral.
4.2 Properties of Line Integrals 143
4.2 Properties of Line Integrals

We know from the previous section that for line integrals of real-valued functions (scalar
fields), reversing the direction in which the integral is taken along a curve does not change
the value of the line integral:
Z Z
f ( x, y) ds = f ( x, y) ds (4.17)
C C
For line integrals of vector fields, however, the value does change. To see this, let f( x, y) =
P ( x, y) i + Q ( x, y) j be a vector field, with P and Q continuously differentiable functions. Let
C be a smooth curve parametrized by x = x( t), y = y( t), a t b, with position vector r( t) =
x( t) i + y( t) j (we will usually abbreviate this by saying that C : r( t) = x( t) i + y( t) j is a smooth
curve). We know that the curve C traversed in the opposite direction is parametrized by
x = x(a + b t), y = y(a + b t), a t b. Then
Z Zb
d
P ( x, y) dx = P ( x(a + b t), y(a + b t)) ( x(a + b t)) dt
C a dt
Zb
= P ( x(a + b t), y(a + b t)) ( x (a + b t)) dt (by the Chain Rule)
Zaa
= P ( x( u), y( u)) ( x ( u)) ( du) (by letting u = a + b t)
Zba
= P ( x( u), y( u)) x ( u) du
b
Zb Za Zb
= P ( x( u), y( u)) x ( u) du , since = , so
Z Za b a
P ( x, y) dx = P ( x, y) dx
C C
since we are just using a different letter ( u) for the line integral along C . A similar argument
shows that Z Z
Q ( x, y) d y = Q ( x, y) d y ,
C C
and hence
Z Z Z
f dr = P ( x, y) dx +
Q ( x, y) d y
C C C
Z Z
= P ( x, y) dx + Q ( x, y) d y
C C
Z Z
= P ( x, y) dx + Q ( x, y) d y
Z ZC C
f dr = f dr . (4.18)
C C
The above formula can be interpreted in terms of the work done by a force f( x, y) (treated
as a vector) moving an object along a curve C : the total work performed moving the object
along C from its initial point to its terminal point, and then back to the initial point moving
backwards along the same path, is zero. This is because when force is considered as a vector,
direction is accounted for.
The preceding discussion shows the importance of always taking the direction of the curve
into account when using line integrals of vector fields. For this reason, the curves in line
integrals are sometimes referred to as directed curves or oriented curves.
Recall that our definition of a line integral required that we have a parametrization x =
x( t), y = y( t), a t b for the curve C . But as we know, any curve has infinitely many
parametrizations. So could we get a different value for a line integral using some other
parametrization of C , say, x = x ( u), y = y ( u), c u d ? If so, this would mean that our
definition is not well-defined. Luckily, it turns out that the value of a line integral of a
vector field is unchanged as long as the direction of the curve C is preserved by whatever
parametrization is chosen:
Theorem 4.2. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a vector field, and let C be a smooth curve

parametrized by x = x( t), y = y( t), a t b. Suppose that t = ( u) for c u d , such that
a = ( c), b = ( d ), and ( u) > 0 on the open interval ( c, d ) (i.e. ( u) is strictly increasing on
R
[ c, d ]). Then C f d r has the same value for the parametrizations x = x( t), y = y( t), a t b
and x = x ( u) = x(( u)), y = y ( u) = y(( u)), c u d .
Proof: Since ( u) is strictly increasing and maps [ c, d ] onto [a, b], then we know that t =
( u) has an inverse function u = 1 ( t) defined on [a, b] such that c = 1 (a), d = 1 ( b),
1
and du
dt = ( u) . Also, dt = ( u) du, and by the Chain Rule
d x d dx dt x ( u)
x ( u) = = ( x(( u))) = = x ( t) ( u ) x ( t) =
du du dt du ( u)
so making the susbstitution t = ( u) gives

Zb Z1 (b)
x ( u)
P ( x( t), y( t)) x ( t) dt = P ( x(( u)), y(( u))) ( ( u) du)
a 1 (a) ( u)
Zd
= P ( x ( u), y ( u)) x ( u) du ,
c
R
which shows that C P ( x, y) dx has the same value for both parametrizations. A similar
R
argument shows that C Q ( x, y) d y has the same value for both parametrizations, and hence
R
C f d r has the same value. QED
Notice that the condition ( u) > 0 in Theorem 4.2 means that the two parametrizations
move along C in the same direction. That was not the case with the reverse parametriza-
tion for C : for u = a + b t we have t = ( u) = a + b u ( u) = 1 < 0.
R
Example 4.4. Evaluate the line integral C ( x2 + y2 ) dx + 2 x y d y from Example 4.2, Section
4.1, along the curve C : x = t, y = 2 t2 , 0 t 1, where t = sin u for 0 u /2.
dt
Solution: First, we notice that 0 = sin 0, 1 = sin(/2), and du = cos u > 0 on (0, /2). So by
Theorem 4.2 we know that if C is parametrized by
x = sin u , y = 2 sin2 u , 0 u /2
R
then C ( x2 + y2 ) dx + 2 x y d y should have the same value as we found in Example 4.2, namely
13
3 . And we can indeed verify this:
Z Z/2
2 2

( x + y ) dx + 2 x y d y = (sin2 u + (2 sin2 u)2 ) cos u + 2(sin u)(2 sin2 u)4 sin u cos u du
C 0
Z/2

= sin2 u + 20 sin4 u cos u du
0
/2
sin3 u 5

= + 4 sin u
3 0
1 13
= +4 =
3 3
In other words, the line integral is unchanged whether t or u is the parameter for C .
By a closed curve, we mean a curve C whose initial point and terminal point are the
same, i.e. for C : x = x( t), y = y( t), a t b, we have ( x(a), y(a)) = ( x( b), y( b)).
t=a t=b t=a
t=b

C C
(a) Closed (b) Not closed
Figure 4.2.1 Closed vs nonclosed curves
A simple closed curve is a closed curve which does not intersect itself. Note that any
closed curve can be regarded as a union of simple closed curves (think of the loops in a figure
eight). We use the special notation
I I
f ( x, y) ds and f dr
C C
to denote line integrals of scalar and vector

Z fields, respectively, along closed curves. In some
Z
older texts you may see the notation or to indicate a line integral traversing a closed
curve in a counterclockwise or clockwise direction, respectively.
So far, the examples we have seen of line integrals (e.g. Example 4.2) have had the same
value for different curves joining the initial point to the terminal point. That is, the line
integral has been independent of the path joining the two points. As we mentioned before,
this is not always the case. The following theorem gives a necessary and sufficient condition
for this path independence:
R
Theorem 4.3. In a region R , the line integral C f d r is independent of the path between
H
any two points in R if and only if C f d r = 0 for every closed curve C which is contained in
R.
H
Proof: Suppose that C f d r = 0 for every closed curve C which is contained in R . Let P1
and P2 be two distinct points in R . Let C 1 be a curve in R going from P1 to P2 , and let C 2
be another curve in R going from P1 to P2 , as in Figure 4.2.2.
Then C = C 1 C 2 is a closed curve in R (from P1 to C1
H
P1 ), and so C f d r = 0. Thus,
I
0 = f dr P1 P2
C
Z Z
= f dr + f dr
C1 C 2 C2
Z Z
Figure 4.2.2
= f dr f d r , and so
C1 C2
R R
This proves path independence.
C 1 f d r = C 2 f d r. R
Conversely, suppose that the line integral C f d r is independent of the path between any
two points in R . Let C be a closed curve contained in R . Let P1 and P2 be two distinct points
on C . Let C 1 be a part of the curve C that goes from P1 to P2 , and let C 2 be the remaining
part of C that goes from P1 to P2 , again as in Figure 4.2.2. Then by path independence we
have
Z Z
f dr = f dr
C1 C2
Z Z
f dr f dr = 0
C1 C2
Z Z
f dr + f d r = 0 , so
C1 C 2
I
f dr = 0
C
since C = C 1 C 2 . QED
Clearly, the above theorem does not give a practical way to determine path independence,
since it is impossible to check the line integrals around all possible closed curves in a region.
What it mostly does is give an idea of the way in which line integrals behave, and how seem-
ingly unrelated line integrals can be related (in this case, a specific line integral between
two points and all line integrals around closed curves).
For a more practical method for determining path independence, we first need a version
of the Chain Rule for multivariable functions:
Theorem 4.4. (Chain Rule) If z = f ( x, y) is a continuously differentiable function of x and

y, and both x = x( t) and y = y( t) are differentiable functions of t, then z is a differentiable
function of t, and
dz z dx z d y
= + (4.19)
dt x dt y dt
at all points where the derivatives on the right are defined.
The proof is virtually identical to the proof of Theorem 2.2 from Section 2.4 (which uses the
Mean Value Theorem), so we omit it.1 We will now use this Chain Rule to prove the following
sufficient condition for path independence of line integrals:
Theorem 4.5. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a vector field in some region R , with P and

Q continuously differentiable functions on R . Let C be a smooth curve in R parametrized
by x = x( t), y = y( t), a t b. Suppose that there is a real-valued function F ( x, y) such that
F = f on R . Then Z
f d r = F (B ) F ( A ) , (4.20)
C
where A = ( x(a), y(a)) and B = ( x( b), y( b)) are the endpoints of C . Thus, the line integral is
independent of the path between its endpoints, since it depends only on the values of F at
those endpoints.
R
Proof: By definition of C f d r, we have
Z Zb

f dr = P ( x( t), y( t)) x ( t) + Q ( x( t), y( t)) y ( t) dt
C a
Zb
F dx F d y F F
= + dt (since F = f = P and = Q)
a x dt y dt x y
Zb
= F ( x( t), y( t)) dt (by the Chain Rule in Theorem 4.4)
a
b

= F ( x( t), y( t)) = F (B) F ( A )
a
by the Fundamental Theorem of Calculus. QED

Theorem 4.5 can be thought of as the line integral version of the Fundamental Theorem
of Calculus. A real-valued function F ( x, y) such that F ( x, y) = f( x, y) is called a potential
for f. A conservative vector field is one which has a potential.
R
Example 4.5. Recall from Examples 4.2 and 4.3 in Section 4.1 that the line integral C ( x2 +
y2 ) dx + 2 x y d y was found to have the value 13
3 for three different curves C going from the
point (0, 0) to the point (1, 2). Use Theorem 4.5 to show that this line integral is indeed path
independent.
Solution: We need to find a real-valued function F ( x, y) such that
F F
= x 2 + y2 and = 2x y .
x y
Suppose that Fx = x2 + y2 , Then we must have F ( x, y) = 13 x3 + x y2 + g( y) for some function

g( y). So Fy = 2 x y + g ( y) satisfies the condition Fy = 2 x y if g ( y) = 0, i.e. g( y) = K , where K
is a constant. Since any choice for K will do (why?), we pick K = 0. Thus, a potential F ( x, y)
for f( x, y) = ( x2 + y2 ) i + 2 x y j exists, namely
1 3
F ( x, y) = x + x y2 .
3
R
Hence the line integral C ( x2 + y2 ) dx + 2 x y d y is path independent.
Note that we can also verify that the value of the line integral of f along any curve C going
from (0, 0) to (1, 2) will always be 133 , since by Theorem 4.5
Z
1 1 13
f d r = F (1, 2) F (0, 0) = (1)3 + (1)(2)2 (0 + 0) = + 4 = .
C 3 3 3
A consequence of Theorem 4.5 in the special case where C is a closed curve, so that the
endpoints A and B are the same point, is the following important corollary:
I
Corollary 4.6. If a vector field f has a potential in a region R , then f d r = 0 for any closed
I C
curve C in R (i.e. F d r = 0 for any real-valued function F ( x, y)).
C
I
Example 4.6. Evaluate x dx + y d y for C : x = 2 cos t, y = 3 sin t, 0 t 2.
C
Solution: The vector field f( x, y) = x i + y j has a potential F ( x, y):
F 1 2
= x F ( x, y) = x + g( y) , so
x 2
F 1
= y g ( y) = y g ( y) = y2 + K
y 2
1 2 1 2
for any constant K , so F ( x, y) = x + y is a potential for f( x, y). Thus,
2 2
I I
x dx + y d y = f dr = 0
C C
x2 y2
by Corollary 4.6, since the curve C is closed (it is the ellipse 4 + 9 = 1).
Exercises
A
I
1. Evaluate ( x2 + y2 ) dx + 2 x y d y for C : x = cos t, y = sin t, 0 t 2.
C
Z
2. Evaluate ( x2 + y2 ) dx + 2 x y d y for C : x = cos t, y = sin t, 0 t .
C
3. Is there a potential F ( x, y) for f( x, y) = y i x j? If so, find one.
4. Is there a potential F ( x, y) for f( x, y) = x i y j? If so, find one.
5. Is there a potential F ( x, y) for f( x, y) = x y2 i + x3 y j? If so, find one.

B
6. Let f( x, y) and g( x, y) be vector fields, let a and b be constants, and let C be a curve in R2 .
Show that Z Z Z
(a f b g) d r = a f d r b g d r .
C C C
R
7. Let C be a curve whose arc length is L. Show that C 1 ds = L.
8. Let f ( x, y) and g( x, y) be continuously differentiable real-valued functions in a region R .

Show that I I
f g dr = g f dr
C C
for any closed curve C in R . (Hint: Use Exercise 21 in Section 2.4.)
y
9. Let f( x, y) = x2 + y2
i + x2 +x y2 j for all ( x, y) 6= (0, 0), and C : x = cos t, y = sin t, 0 t 2.
(a) Show that f = F , for F ( x, y) = tan1 ( y/ x).
I
(b) Show that f d r = 2. Does this contradict Corollary 4.6? Explain.
C
C
10. Let g( x) and h( y) be differentiable functions, and let f( x, y) = h( y) i + g( x) j. Can f have a
potential F ( x, y)? If so, find it. You may assume that F would be smooth. (Hint: Consider
the mixed partial derivatives of F .)
4.3 Greens Theorem

We will now see a way of evaluating the line integral of a smooth vector field around a
simple closed curve. A vector field f( x, y) = P ( x, y) i + Q ( x, y) j is smooth if its component
functions P ( x, y) and Q ( x, y) are smooth. We will use Greens Theorem (sometimes called
Greens Theorem in the plane) to relate the line integral around a closed curve with a double
integral over the region inside the curve:
Theorem 4.7. (Greens Theorem) Let R be a region in R2 whose boundary is a simple

closed curve C which is piecewise smooth. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a smooth vector
field defined on both R and C . Then
I
Q P
f dr = dA , (4.21)
C x y
R
where C is traversed so that R is always on the left side of C .

Proof: We will prove the theorem in the case for a simple region R , that is, where the
boundary curve C can be written as C = C 1 C 2 in two distinct ways:
C 1 = the curve y = y1 ( x) from the point X 1 to the point X 2 (4.22)

C 2 = the curve y = y2 ( x) from the point X 2 to the point X 1 , (4.23)
where X 1 and X 2 are the points on C farthest to the left and right, respectively; and
C 1 = the curve x = x1 ( y) from the point Y2 to the point Y1 (4.24)

C 2 = the curve x = x2 ( y) from the point Y1 to the point Y2 , (4.25)
where Y1 and Y2 are the lowest and highest points, respectively, on C . See Figure 4.3.1.
y
y = y2 (x)
d
Y2

X 2 x = x2 (y)
x = x1 (y) X 1 R
C
Y1
c
y = y1 (x)
x
a b
Figure 4.3.1
Integrate P ( x, y) around C using the representation C = C 1 C 2 given by (4.23) and (4.24).

4.3 Greens Theorem 151
Since y = y1 ( x) along C 1 (as x goes from a to b) and y = y2 ( x) along C 2 (as x goes from b to
a), as we see from Figure 4.3.1, then we have
I Z Z
P ( x, y) dx = P ( x, y) dx + P ( x, y) dx
C C1 C2
Zb Za
= P ( x, y1 ( x)) dx + P ( x, y2 ( x)) dx
a b
Zb Zb
= P ( x, y1 ( x)) dx P ( x, y2 ( x)) dx
a a
Zb
= (P ( x, y2 ( x)) P ( x, y1 ( x))) dx
a
Zb y= y2 ( x)

= P ( x, y) dx
a y= y1 ( x)
Zb Z y2 ( x)
P ( x, y)
= d y dx (by the Fundamental Theorem of Calculus)
a y1 ( x) y

P
= dA .
y
R
Likewise, integrate Q ( x, y) around C using the representation C = C 1 C 2 given by (4.25)

and (4.26). Since x = x1 ( y) along C 1 (as y goes from d to c) and x = x2 ( y) along C 2 (as y goes
from c to d ), as we see from Figure 4.3.1, then we have
I Z Z
Q ( x, y) d y = Q ( x, y) d y + Q ( x, y) d y
C C1 C2
Zc Zd
= Q ( x1 ( y), y) d y + Q ( x2 ( y), y) d y
d c
Zd Zd
= Q ( x1 ( y), y) d y + Q ( x2 ( y), y) d y
c c
Zd
= (Q ( x2 ( y), y) Q ( x1 ( y), y)) d y
c
Zd x= x2 ( y)

= Q ( x, y) dy
c x= x1 ( y)
Zd Z x2 ( y)
Q ( x, y)
= dx d y (by the Fundamental Theorem of Calculus)
c x1 ( y) x

Q
= d A , and so
x
R
I I I
f dr = P ( x, y) dx + Q ( x, y) d y
C C C

P Q
= dA + dA
y x
R R

Q P
= dA .
x y
R
QED
Though we proved Greens Theorem only for a simple region R , the theorem can also be
proved for more general regions (say, a union of simple regions).2
H
Example 4.7. Evaluate C ( x2 + y2 ) dx + 2 x y d y, where C is the boundary (traversed counter-
clockwise) of the region R = { ( x, y) : 0 x 1, 2 x2 y 2 x }.
Solution: R is the shaded region in Figure 4.3.2. By Greens Theorem, for y
(1, 2)
2 2
P ( x, y) = x + y and Q ( x, y) = 2 x y, we have 2
I
2 2 Q P
( x + y ) dx + 2 x y d y = dA C
C x y
R
x
= (2 y 2 y) d A = 0dA = 0 . 0 1
R R Figure 4.3.2
We actually already knew that the answer was zero. Recall from Example 4.5 in Section
4.2 that the vector field f( x, y) = ( x2 + y2 ) i + 2 x y j has a potential function F ( x, y) = 13 x3 + x y2 ,
H
and so C f d r = 0 by Corollary 4.6.
Example 4.8. Let f( x, y) = P ( x, y) i + Q ( x, y) j, where

y x
P ( x, y) = and Q ( x, y) = ,
x 2 + y2 x 2 + y2
and let R = { ( x, y) : 0 < x2 + y2 1 }. For the boundary curve C : x2 + y2 = 1, traversed counter-
H
clockwise, it was shown in Exercise 9(b) in Section 4.2 that C f d r = 2. But

Q y2 x 2 P Q P
= 2 = d A = 0dA = 0 .
x ( x + y2 )2 y x y
R R
This would seem to contradict Greens Theorem. However, note that R is not the entire
region enclosed by C , since the point (0, 0) is not contained in R . That is, R has a hole at
the origin, so Greens Theorem does not apply.
2 See T AYLOR and M ANN, 15.31 for a discussion of some of the difficulties involved when the boundary curve
is complicated.
y
If we modify the region R to be the annulus R =
{ ( x, y) : 1/4 x2 + y2 1 } (see Figure 4.3.3), and take 1
the boundary C of R to be C = C 1 C 2 , where C 1 is C1
the unit circle x2 + y2 = 1 traversed counterclockwise

R
and C 2 is the circle x2 + y2 = 1/4 traversed clockwise, 1/2
then it can be shown (see Exercise 8) that C2

x
I 0 1/2 1
f dr = 0 .
C
Q
We would still have x
Py d A = 0, so for this R
R
we would have
I
Q P
Figure 4.3.3 The annulus R
f dr = dA ,
C x y
R
which shows that Greens Theorem holds for the annular region R .
It turns out that Greens Theorem can be extended to multiply connected regions, that is,
regions like the annulus in Example 4.8, which have one or more regions cut out from the
interior, as opposed to discrete points being cut out. For such regions, the outer boundary
and the inner boundaries are traversed so that R is always on the left side.
C1 C1
R1 R1

C2 C3 C2

R2 R2

(a) Region R with one hole (b) Region R with two holes
Figure 4.3.4 Multiply connected regions
The intuitive idea for why Greens Theorem holds for multiply connected regions is shown
in Figure 4.3.4 above. The idea is to cut slits between the boundaries of a multiply con-
nected region R so that R is divided into subregions which do not have any holes. For
example, in Figure 4.3.4(a) the region R is the union of the regions R 1 and R 2 , which are
divided by the slits indicated by the dashed lines. Those slits are part of the boundary of
both R 1 and R 2 , and we traverse then in the manner indicated by the arrows. Notice that
along each slit the boundary of R 1 is traversed in the opposite direction as that of R 2 , which
means that the line integrals of f along those slits cancel each other out. Since R 1 and R 2 do
not have holes in them, then Greens Theorem holds in each subregion, so that
I I
Q P Q P
bdy
f d r = d A and bdy
f d r = dA .
x y x y
of R 1 R1 of R 2 R2
But since the line integrals along the slits cancel out, we have
I I I
f d r = bdy f d r + bdy f d r ,
C 1 C 2
of R 1 of R 2
and so
I
Q P Q P Q P
f dr = dA + dA = dA ,
C 1 C 2 x y x y x y
R1 R2 R
which shows that Greens Theorem holds in the region R . A similar argument shows that
the theorem holds in the region with two holes shown in Figure 4.3.4(b).
We know from Corollary 4.6 that when a smooth vector field f( x, y) = P ( x, y) i + Q ( x, y) j on
a region R (whose boundary is a piecewise smooth, simple closed curve C ) has a potential in
H
R , then C f d r = 0. And if the potential F ( x, y) is smooth in R , then Fx = P and Fy = Q , and
so we know that
2 F 2 F P Q
= = in R .
y x x y y x
P Q
Conversely, if y
= x
in R then
I
Q P
f dr = dA = 0dA = 0 .
C x y
R R
For a simply connected region R (i.e. a region with no holes), the following can be shown:
The following statements are equivalent for a simply connected region R in R2 :
(a) f( x, y) = P ( x, y) i + Q ( x, y) j has a smooth potential F ( x, y) in R

Z
(b) f d r is independent of the path for any curve C in R
C
I
(c) f d r = 0 for every simple closed curve C in R
C
P Q
(d) = in R (in this case, the differential form P dx + Q d y is exact)
y x
Exercises
A
For Exercises 1-4, use Greens Theorem to evaluate the given line integral around the curve
C , traversed counterclockwise.
I
1. ( x2 y2 ) dx + 2 x y d y; C is the boundary of R = { ( x, y) : 0 x 1, 2 x2 y 2 x }
C
I
2. x2 y dx + 2 x y d y; C is the boundary of R = { ( x, y) : 0 x 1, x2 y x }
C
I
3. 2 y dx 3 x d y; C is the circle x2 + y2 = 1
C
I
2 2
4. ( e x + y2 ) dx + ( e y + x2 ) d y; C is the boundary of the triangle with vertices (0, 0), (4, 0)
C
and (0, 4)
5. Is there a potential F ( x, y) for f( x, y) = ( y2 + 3 x2 ) i + 2 x y j? If so, find one.
6. Is there a potential F ( x, y) for f( x, y) = ( x3 cos( x y) + 2 x sin( x y)) i + x2 y cos( x y) j? If so, find

one.
7. Is there a potential F ( x, y) for f( x, y) = (8 x y + 3) i + 4( x2 + y) j? If so, find one.

I
8. Show that for any constants a, b and any closed simple curve C , a dx + b d y = 0.
C
B
H
9. For the vector field f as in Example 4.8, show directly that C f d r = 0, where C is the
boundary of the annulus R = { ( x, y) : 1/4 x2 + y2 1 } traversed so that R is always on
the left.
I
10. Evaluate e x sin y dx + ( y3 + e x cos y) d y, where C is the boundary of the rectangle with
C
vertices (1, 1), (1, 1), (1, 1) and (1, 1), traversed counterclockwise.
C
11. For a region R bounded by a simple closed curve C , show that the area A of R is
I I I
1
A = y dx = xdy = x d y y dx ,
C C 2 C
where C is traversed so that R is always on the left. (Hint: Use Greens Theorem and the

fact that A = 1 d A .)
R
4.4 Surface Integrals and the Divergence Theorem

In Section 4.1 we learned how to integrate along a curve. We will now learn how to perform
integration over a surface in R3 , such as a sphere or a paraboloid. Recall from Section 1.8
how we identified points ( x, y, z) on a curve C in R3 , parametrized by x = x( t), y = y( t), z = z( t),
a t b, with the terminal points of the position vector
r( t) = x( t)i + y( t)j + z( t)k for t in [a, b].
The idea behind a parametrization of a curve is that it transforms a subset of R1 (nor-

mally an interval [a, b]) into a curve in R2 or R3 (see Figure 4.4.1).
z
(x(t), y(t), z(t))
(x(a), y(a), z(a))
x = x(t) C
y = y(t) r(t) (x(b), y(b), z(b))
z = z(t)
R1 y
a t b 0
x
Figure 4.4.1 Parametrization of a curve C in R3
Similar to how we used a parametrization of a curve to define the line integral along the
curve, we will use a parametrization of a surface to define a surface integral. We will use
two variables, u and v, to parametrize a surface in R3 : x = x( u, v), y = y( u, v), z = z( u, v),
for ( u, v) in some region R in R2 (see Figure 4.4.2).
v R2
z

R x = x(u, v)
y = y(u, v)
(u, v) z = z(u, v) r(u, v)
y
0
u x
Figure 4.4.2 Parametrization of a surface in R3
In this case, the position vector of a point on the surface is given by the vector-valued
function
r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k for ( u, v) in R .
4.4 Surface Integrals and the Divergence Theorem 157
r r
Since r( u, v) is a function of two variables, define the partial derivatives u
and v
for
( u, v) in R by
r x y z
( u, v) = ( u, v)i + ( u, v)k , and
( u, v)j +
u u u u
r x y z
( u, v) = ( u, v)i + ( u, v)j + ( u, v)k .
v v v v
The parametrization of can be thought of as transforming a region in R2 (in the uv-

plane) into a 2-dimensional surface in R3 . This parametrization of the surface is sometimes
called a patch, based on the idea of patching the region R onto in the grid-like manner
shown in Figure 4.4.2.
In fact, those gridlines in R lead us to how we will define a surface integral over . Along
the vertical gridlines in R , the variable u is constant. So those lines get mapped to curves on
, and the variable u is constant along the position vector r( u, v). Thus, the tangent vector
to those curves at a point ( u, v) is rv . Similarly, the horizontal gridlines in R get mapped to
curves on whose tangent vectors are ur .
Now take a point ( u, v) in R as, say, the lower left corner of one of the rectangular grid
sections in R , as shown in Figure 4.4.2. Suppose that this rectangle has a small width
and height of u and v, respectively. The corner points of that rectangle are ( u, v), ( u +
u, v), ( u + u, v + v) and ( u, v + v). So the area of that rectangle is A = u v. Then that
rectangle gets mapped by the parametrization onto some section of the surface which,
for u and v small enough, will have a surface area (call it d ) that is very close to the
area of the parallelogram which has adjacent sides r( u + u, v) r( u, v) (corresponding to the
line segment from ( u, v) to ( u + u, v) in R ) and r( u, v + v) r( u, v) (corresponding to the
line segment from ( u, v) to ( u, v + v) in R ). But by combining our usual notion of a partial
derivative (see Definition 2.3 in Section 2.2) with that of the derivative of a vector-valued
function (see Definition 1.12 in Section 1.8) applied to a function of two variables, we have
r r( u + u, v) r( u, v)
, and
u u
r r( u, v + v) r( u, v)
,
v v
and so the surface area element d is approximately

(r( u + u, v) r( u, v)) (r( u, v + v) r( u, v)) ( u r ) (v r ) = r r u v

u v u v
by Theorem 1.13 in Section

1.4.
Thus, the total surface area S of is approximately the sum
of all the quantities ur rv u v, summed over the rectangles in R . Taking the limit of
that sum as the diagonal of the largest rectangle goes to 0 gives

r r
S = u v du dv . (4.26)

R
We will write the double integral on the right using the special notation

r r
d = u v du dv . (4.27)

R
This is a special case of a surface integral over the surface , where the surface area element
d can be thought of as 1 d . Replacing 1 by a general real-valued function f ( x, y, z) defined
in R3 , we have the following:
Definition 4.3. Let be a surface in R3 parametrized by x = x( u, v), y = y( u, v),

z = z( u, v), for ( u, v) in some region R in R2 . Let r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k be the
position vector for any point on , and let f ( x, y, z) be a real-valued function defined on some
subset of R3 that contains . The surface integral of f ( x, y, z) over is

r r
f ( x, y, z) d = f ( x( u, v), y( u, v), z( u, v))
du dv . (4.28)
u v
R
In particular, the surface area S of is

S = 1 d . (4.29)

Example 4.9. A torus T is a surface obtained by revolving a circle of radius a in the yz-plane
around the z-axis, where the circles center is at a distance b from the z-axis (0 < a < b), as
in Figure 4.4.3. Find the surface area of T .
z
z
(y b)2 + z2 = a2
y
a (x, y, z)
u y v
0 a
b x
(a) Circle in the yz-plane (b) Torus T
Figure 4.4.3
Solution: For any point on the circle, the line segment from the center of the circle to that
point makes an angle u with the y-axis in the positive y direction (see Figure 4.4.3(a)). And
as the circle revolves around the z-axis, the line segment from the origin to the center of that
circle sweeps out an angle v with the positive x-axis (see Figure 4.4.3(b)). Thus, the torus
can be parametrized as:
x = ( b + a cos u) cos v , y = ( b + a cos u) sin v , z = a sin u , 0 u 2 , 0 v 2
So for the position vector
r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k

= ( b + a cos u) cos v i + ( b + a cos u) sin v j + a sin u k
we see that
r
= a sin u cos v i a sin u sin v j + a cos u k
u
r
= ( b + a cos u) sin v i + ( b + a cos u) cos v j + 0k ,
v
and so computing the cross product gives
r r
= a( b + a cos u) cos v cos u i a( b + a cos u) sin v cos u j a( b + a cos u) sin u k ,
u v
which has magnitude
r r
u v = a( b + a cos u) .

Thus, the surface area of T is

S = 1 d

Z2 Z2
r r
= u v du dv

0 0
Z2 Z2
= a( b + a cos u) du dv
0 0
Z2 u=2
abu + a2 sin u

= dv
0 u =0
Z2
= 2ab dv
0
= 42 ab
Since ur and rv are tangent to the surface (i.e. lie in the tangent plane to at each point
on ), then their cross product ur rv is perpendicular to the tangent plane to the surface
at each point of . Thus,

f ( x, y, z) d = f ( x( u, v), y( u, v), z( u, v)) k n k d ,
R
r
where n = u
rv . We say that n is a normal vector to .
z
Recall that normal vectors to a plane can point in two opposite
directions. By an outward unit normal vector to a surface ,
we will mean the unit vector that is normal to and points away
from the top (or outer part) of the surface. This is a hazy y
definition, but the picture in Figure 4.4.4 gives a better idea of 0
what outward normal vectors look like, in the case of a sphere.
With this idea in mind, we make the following definition of a
x
surface integral of a 3-dimensional vector field over a surface:
Figure 4.4.4
Definition 4.4. Let be a surface in R3 and let f( x, y, z) = f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k

be a vector field defined on some subset of R3 that contains . The surface integral of f
over is
f d = f n d , (4.30)

where, at any point on , n is the outward unit normal vector to .
Note in the above definition that the dot product inside the integral on the right is a
real-valued function, and hence we can use Definition 4.3 to evaluate the integral.

Example 4.10. Evaluate the surface integral f d , where f( x, y, z) = yzi + xzj + x yk and

is the part of the plane x + y + z = 1 with x 0, y 0, and z 0, with the outward unit normal
n pointing in the positive z direction (see Figure 4.4.5).
z
Solution: Since the vector v = (1, 1, 1) is normal to the plane x + y + z = 1
1
(why?), then dividing v by its length yields the outward unit normal n
vector n = p1 , p1 , p1 . We now need to parametrize . As we can see
3 3 3
from Figure 4.4.5, projecting onto the x y-plane yields a triangular y
0
region R = { ( x, y) : 0 x 1, 0 y 1 x }. Thus, using ( u, v) instead of 1
1 x+ y+ z = 1
( x, y), we see that x
Figure 4.4.5
x = u, y = v, z = 1 ( u + v), for 0 u 1, 0 v 1 u
is a parametrization of over R (since z = 1 ( x + y) on ). So on ,

1 1 1 1
f n = ( yz, xz, x y) p , p , p = p ( yz + xz + x y)
3 3 3 3
1 1
= p (( x + y) z + x y) = p (( u + v)(1 ( u + v)) + uv)
3 3
1
= p (( u + v) ( u + v)2 + uv)
3
for ( u, v) in R , and for r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k = ui + vj + (1 ( u + v))k we have

r r r r p
= (1, 0, 1) (0, 1, 1) = (1, 1, 1) = 3 .
u v u v
Thus, integrating over R using vertical slices (e.g. as indicated by the dashed line in Figure
4.4.5) gives

f d = f n d

r r
= (f( x( u, v), y( u, v), z( u, v)) n)
dv du
u v
R
Z1 Z1u p
1
= p (( u + v) ( u + v)2 + uv) 3 dv du
0 0 3
Z1 v=1u !
( u + v)2 ( u + v)3 uv2
= + du
0 2 3 2 v =0
Z1
1 u 3 u2 5 u3
= + + du
0 6 2 2 6
1
u u2 u3 5 u4 1
= + + = .
6 4 2 24 0 8
Computing surface integrals can often be tedious, especially when the formula for the
outward unit normal vector at each point of changes. The following theorem provides an
easier way in the case when is a closed surface, that is, when encloses a bounded
solid in R3 . For example, spheres, cubes, and ellipsoids are closed surfaces, but planes and
paraboloids are not.
Theorem 4.8. (Divergence Theorem) Let be a closed surface in R3 which bounds a

solid S , and let f( x, y, z) = f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k be a vector field defined on some
subset of R3 that contains . Then

f d = div f dV , (4.31)
S
where
f1 f2 f3
div f = + + (4.32)
x y z
is called the divergence of f.
The proof of the Divergence Theorem is very similar to the proof of Greens Theorem, i.e. it
is first proved for the simple case when the solid S is bounded above by one surface, bounded
below by another surface, and bounded laterally by one or more surfaces. The proof can then
be extended to more general solids.3

Example 4.11. Evaluate f d , where f( x, y, z) = xi + yj + zk and is the unit sphere

x2 + y2 + z2 = 1.
Solution: We see that div f = 1 + 1 + 1 = 3, so

f d = div f dV = 3 dV
S S

4(1)3
= 3 1 dV = 3 vol(S ) = 3 = 4 .
3
S

In physical applications, the surface integral f d is often referred to as the flux of f

through the surface . For example, if f represents the velocity field of a fluid, then the flux
is the net quantity of fluid to flow through the surface per unit time. A positive flux means
there is a net flow out of the surface (i.e. in the direction of the outward unit normal vector
n), while a negative flux indicates a net flow inward (in the direction of n).
The term divergence comes from interpreting div f as a measure of how much a vector
field diverges from a point. This is best seen by using another definition of div f which is
equivalent4 to the definition given by formula (4.32). Namely, for a point ( x, y, z) in R3 ,

1
div f( x, y, z) = lim f d , (4.33)
V 0 V

3 See T AYLOR and M ANN, 15.6 for the details.

4 See S CHEY, p. 36-39, for an intuitive discussion of this.
where V is the volume enclosed by a closed surface around the point ( x, y, z). In the
limit, V 0 means that we take smaller and smaller closed surfaces around ( x, y, z), which
means that the volumes they enclose are going to zero. It can be shown that this limit is
independent of the shapes of those surfaces. Notice that the limit being taken is of the
ratio of the flux through a surface to the volume enclosed by that surface, which gives a
rough measure of the flow leaving a point, as we mentioned. Vector fields which have zero
divergence are often called solenoidal fields.
The following theorem is a simple consequence of formula (4.33).
Theorem 4.9. If the flux of a vector field f is zero through every closed surface containing a
given point, then div f = 0 at that point.
Proof: By formula (4.33), at the given point ( x, y, z) we have

1
div f( x, y, z) = lim f d for closed surfaces containing ( x, y, z), so
V 0 V

1
= lim (0) by our assumption that the flux through each is zero, so
V 0 V
= lim 0
V 0
= 0. QED
Lastly, we note that sometimes the notation

f ( x, y, z) d and f d

is used to denote surface integrals of scalar and vector fields, respectively, over closed sur-
H
faces. Especially in physics texts, it is common to see simply instead of .

Exercises
A
For Exercises 1-4, use the Divergence Theorem to evaluate the surface integral f d of

the given vector field f( x, y, z) over the surface .
1. f( x, y, z) = xi + 2 yj + 3 zk, : x2 + y2 + z2 = 9
2. f( x, y, z) = xi + yj + zk, : boundary of the solid cube S = { ( x, y, z) : 0 x, y, z 1 }
3. f( x, y, z) = x3 i + y3 j + z3 k, : x2 + y2 + z2 = 1
4. f( x, y, z) = 2i + 3j + 5k, : x2 + y2 + z2 = 1
B
5. Show that the flux of any constant vector field through any closed surface is zero.
6. Evaluate the surface integral from Exercise 2 without using the Divergence Theorem, i.e.
using only Definition 4.3, as in Example 4.10. Note that there will be a different outward
unit normal vector to each of the six faces of the cube.

7. Evaluate the surface integral f d , where f( x, y, z) = x2 i + x yj + zk and is the part of

the plane 6 x + 3 y + 2 z = 6 with x 0, y 0, and z 0, with the outward unit normal n
pointing in the positive z direction.
8. Use a surface integral to show that the surface area of a sphere of radius r is 4 r 2 . (Hint:
Use spherical coordinates to parametrize the sphere.)
9. Use a surface integral

p to show that the surface area of a right circular cone of radius R
and height h is R h2 + R 2 . (Hint: Use the parametrization x = r cos , y = r sin , z = Rh r ,
for 0 r R and 0 2.)
x2 y2 2
10. The ellipsoid a2
+ b2 + zc2 = 1 can be parametrized using ellipsoidal coordinates
x = a sin cos , y = b sin sin , z = c cos , for 0 2 and 0 .
Show that the surface area S of the ellipsoid is

Z Z2 q
S = sin a2 b2 cos2 + c2 (a2 sin2 + b2 cos2 ) sin2 d d .
0 0
(Note: The above double integral can not be evaluated by elementary means. For specific
values of a, b and c it can be evaluated using numerical methods. An alternative is to
express the surface area in terms of elliptic integrals.5 )
C
11. Use Definition 4.3 to prove that the surface area S over a region R in R2 of a surface
z = f ( x, y) is given by the formula
r 2 2
f f
S = 1 + x + y d A .
R
(Hint: Think of the parametrization of the surface.)
5 B OWMAN, F., Introduction to Elliptic Functions, with Applications, New York: Dover, 1961, III.7.
4.5 Stokes Theorem 165
4.5 Stokes Theorem

So far the only types of line integrals which we have discussed are those along curves in R2 .
But the definitions and properties which were covered in Sections 4.1 and 4.2 can easily be
extended to include functions of three variables, so that we can now discuss line integrals
along curves in R3 .
Definition 4.5. For a real-valued function f ( x, y, z) and a curve C in R3 , parametrized by

x = x( t), y = y( t), z = z( t), a t b, the line integral of f ( x, y, z) along C with respect to
arc length s is
Z Zb q
f ( x, y, z) ds = f ( x( t), y( t), z( t)) x ( t)2 + y ( t)2 + z ( t)2 dt . (4.34)
C a
The line integral of f ( x, y, z) along C with respect to x is

Z Zb
f ( x, y, z) dx = f ( x( t), y( t), z( t)) x ( t) dt . (4.35)
C a
The line integral of f ( x, y, z) along C with respect to y is

Z Zb
f ( x, y, z) d y = f ( x( t), y( t), z( t)) y ( t) dt . (4.36)
C a
The line integral of f ( x, y, z) along C with respect to z is

Z Zb
f ( x, y, z) dz = f ( x( t), y( t), z( t)) z ( t) dt . (4.37)
C a
R
Similar to the two-variable case, if f ( x, y, z) 0 then the line integral C f ( x, y, z) ds can be
thought of as the total area of the picket fence of height f ( x, y, z) at each point along the
curve C in R3 .
Vector fields in R3 are defined in a similar fashion to those in R2 , which allows us to define
the line integral of a vector field along a curve in R3 .
Definition 4.6. For a vector field f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k and a curve C

in R3 with a smooth parametrization x = x( t), y = y( t), z = z( t), a t b, the line integral
of f along C is
Z Z Z Z
f dr = P ( x, y, z) dx + Q ( x, y, z) d y + R ( x, y, z) dz (4.38)
C C C C
Zb
= f( x( t), y( t), z( t)) r ( t) dt , (4.39)
a
where r( t) = x( t) i + y( t) j + z( t) k is the position vector for points on C .

Similar to the two-variable case, if f( x, y, z) represents the force applied to an object at a

R
point ( x, y, z) then the line integral C f d r represents the work done by that force in moving
the object along the curve C in R3 .
Some of the most important results we will need for line integrals in R3 are stated below
without proof (the proofs are similar to their two-variable equivalents).
Theorem 4.10. For a vector field f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k and a curve

C with a smooth parametrization x = x( t), y = y( t), z = z( t), a t b and position vector
r( t) = x( t) i + y( t) j + z( t) k, Z Z
f dr = f T ds , (4.40)
C C
r ( t)
where T( t) = k r ( t) k is the unit tangent vector to C at ( x( t), y( t), z( t)).
Theorem 4.11. (Chain Rule) If w = f ( x, y, z) is a continuously differentiable function of

x, y, and z, and x = x( t), y = y( t) and z = z( t) are differentiable functions of t, then w is a
differentiable function of t, and
dw w dx w d y w dz
= + + . (4.41)
dt x dt y dt z dt
Also, if x = x( t 1 , t 2 ), y = y( t 1 , t 2 ) and z = z( t 1 , t 2 ) are continuously differentiable function of

( t 1 , t 2 ), then6
w w x w y w z
= + + (4.42)
t1 x t1 y t1 z t1
and
w w x w y w z
= + + . (4.43)
t2 x t2 y t2 z t2
Theorem 4.12. Let f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k be a vector field in some

solid S , with P , Q and R continuously differentiable functions on S . Let C be a smooth
curve in S parametrized by x = x( t), y = y( t), z = z( t), a t b. Suppose that there is a
real-valued function F ( x, y, z) such that F = f on S . Then
Z
f d r = F (B ) F ( A ) , (4.44)
C
where A = ( x(a), y(a), z(a)) and B = ( x( b), y( b), z( b)) are the endpoints of C .
I
Corollary 4.13. If a vector field f has a potential in a solid S , then f d r = 0 for any closed
I C
curve C in S (i.e. F d r = 0 for any real-valued function F ( x, y, z)).
C
6 See T AYLOR and M ANN, 6.5 for a proof.

Example 4.12. Let f ( x, y, z) = z and let C be the curve in R3 parametrized by
x = t sin t , y = t cos t , z=t , 0 t 8 .

R
Evaluate C f ( x, y, z) ds. (Note: C is called a conical helix. See Figure 4.5.1).
Solution: Since x ( t) = sin t + t cos t, y ( t) = cos t t sin t, and z ( t) = 1, we have
x ( t)2 + y ( t)2 + z ( t)2 = (sin2 t + 2 t sin t cos t + t2 cos2 t) + (cos2 t 2 t sin t cos t + t2 sin2 t) + 1
= t2 (sin2 t + cos2 t) + sin2 t + cos2 t + 1
= t2 + 2 ,
so since f ( x( t), y( t), z( t)) = z( t) = t along the curve C , then

Z Z8 q
f ( x, y, z) ds = f ( x( t), y( t), z( t)) x ( t)2 + y ( t)2 + z ( t)2 dt
C 0
Z8 p
= t t2 + 2 dt
0
8
1 2 3/2
1 2 3/2
p
= ( t + 2) = (64 + 2) 2 2 .
3 0 3
30
25
t = 8
20
15
z
10
5 -25
t=0 -20
-15
0-25 -10
-5
-20 -15 -10 0
-5 5 x
0 5 10 10
15 20 15
y 25 30 2520
Figure 4.5.1 Conical helix C
Example 4.13. Let f( x, y, z) = x i + y j + 2 z k be a vector field in R3 . Using the same curve C

R
from Example 4.12, evaluate C f d r.
x2 y2
Solution: It is easy to see that F ( x, y, z) = 2 + 2 + z2 is a potential for f( x, y, z) (i.e. F = f).
So by Theorem 4.12 we know that

Z
f d r = F (B) F ( A ) , where A = ( x(0), y(0), z(0)) and B = ( x(8), y(8), z(8)), so
C
= F (8 sin 8, 8 cos 8, 8) F (0 sin 0, 0 cos 0, 0)
= F (0, 8, 8) F (0, 0, 0)
(8)2
= 0+ + (8)2 (0 + 0 + 0) = 962 .
2
We will now discuss a generalization of Greens Theorem in R2 to orientable surfaces in

R , called Stokes Theorem. A surface in R3 is orientable if there is a continuous vector
3
field N in R3 such that N is nonzero and normal to (i.e. perpendicular to the tangent plane)
at each point of . We say that such an N is a normal vector field.
For example, the unit sphere x2 + y2 + z2 = 1 is orientable, since the z
N
continuous vector field N( x, y, z) = x i + y j + z k is nonzero and normal
to the sphere at each point. In fact, N( x, y, z) is another normal N
vector field (see Figure 4.5.2). We see in this case that N( x, y, z) is y
what we have called an outward normal vector, and N( x, y, z) is an 0
inward normal vector. These outward and inward normal vec-
tor fields on the sphere correspond to an outer and inner side,
respectively, of the sphere. That is, we say that the sphere is a two- x
sided surface. Roughly, two-sided means orientable. Other ex- Figure 4.5.2
amples of two-sided, and hence orientable, surfaces are cylinders,
paraboloids, ellipsoids, and planes.
You may be wondering what kind of surface would not have two sides. An example is the
Mbius strip, which is constructed by taking a thin rectangle and connecting its ends at
the opposite corners, resulting in a twisted strip (see Figure 4.5.3).
AA

A B
B A
(a) Connect A to A and B to B along the ends (b) Not orientable
Figure 4.5.3 Mbius strip
If you imagine walking along a line down the center of the Mbius strip, as in Figure
4.5.3(b), then you arrive back at the same place from which you started but upside down!
That is, your orientation changed even though your motion was continuous along that center
line. Informally, thinking of your vertical direction as a normal vector field along the strip,
there is a discontinuity at your starting point (and, in fact, at every point) since your vertical
direction takes two different values there. The Mbius strip has only one side, and hence is
nonorientable.7
For an orientable surface which has a boundary curve C , pick a unit normal vector n
such that if you walked along C with your head pointing in the direction of n, then the
surface would be on your left. We say in this situation that n is a positive unit normal vector
and that C is traversed n-positively. We can now state Stokes Theorem:
Theorem 4.14. (Stokes Theorem) Let be an orientable surface in R3 whose boundary

is a simple closed curve C , and let f( x, y, z) = P ( x, y, z)i + Q ( x, y, z)j + R ( x, y, z)k be a smooth
vector field defined on some subset of R3 that contains . Then
I
f dr = (curl f ) n d , (4.45)
C

where
R Q P R Q P
curl f = i + j + k, (4.46)
y z z x x y
n is a positive unit normal vector over , and C is traversed n-positively.
Proof: As the general case is beyond the scope of this text, we will prove the theorem only
for the special case where is the graph of z = z( x, y) for some smooth real-valued function
z( x, y), with ( x, y) varying over a region D in R2 .
Projecting onto the x y-plane, we see that the closed z
: z = z(x, y)
curve C (the boundary curve of ) projects onto a closed n
curve C D which is the boundary curve of D (see Fig-
ure 4.5.4). Assuming that C has a smooth parametriza-
tion, its projection C D in the x y-plane also has a smooth
C
parametrization, say y
0
C D : x = x ( t ) , y = y( t ) , a t b , D (x, y)
x
and so C can be parametrized (in R3 ) as CD
Figure 4.5.4
C : x = x( t) , y = y( t) , z = z( x( t), y( t)) , a t b ,
since the curve C is part of the surface z = z( x, y). Now, by the Chain Rule (Theorem 4.4 in
Section 4.2), for z = z( x( t), y( t)) as a function of t, we know that
z z
z ( t) = x ( t) + y ( t) ,
x y
7 For further discussion of orientability, see ON EILL, IV.7.
and so
I Z
f dr = P ( x, y, z) dx + Q ( x, y, z) d y + R ( x, y, z) dz
C C
Zb
z z
= P x ( t) + Q y ( t) + R x ( t) + y ( t) dt
a x y
Zb
z z
= P +R x ( t) + Q + R y ( t) dt
x y
Za
= P ( x, y) dx + Q ( x, y) d y ,
CD
where
z
P ( x, y) = P ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y) , and
x
z
Q ( x, y) = Q ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y)
y
for ( x, y) in D . Thus, by Greens Theorem applied to the region D , we have

Q P
I
f dr = dA . (4.47)
C x y
D
Thus,

Q z
= Q ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y) , so by the Product Rule we get
x x y

z z
= (Q ( x, y, z( x, y))) + R ( x, y, z( x, y)) ( x, y) + R ( x, y, z( x, y)) ( x, y) .
x x y x y
Now, by formula (4.42) in Theorem 4.11, we have
Q x Q y Q z
(Q ( x, y, z( x, y))) = + +
x x x y x z x
Q Q Q z
= 1 + 0 +
x y z x
Q Q z
= + .
x z x
Similarly,
R R z
(R ( x, y, z( x, y))) = + .
x x z x
Thus,

Q Q Q z R R z z 2 z
= + + + + R ( x, y, z( x, y))
x x z x x z x y x y
2
Q Q z R z R z z z
= + + + + R .
x z x x y z x y x y
In a similar fashion, we can calculate
P P P z R z R z z 2 z
= + + + + R .
y y z y y x z y x y x
So subtracting gives

Q P Q R z R P z Q P
= + + (4.48)
x y z y x x z y x y
2 z 2 z
since x y
= y x
by the smoothness of z = z( x, y). Hence, by equation (4.47),
I
R Q z P R z Q P
f dr = + dA (4.49)
C y z x z x y x y
D
after factoring out a 1 from the terms in the first two products in equation (4.48).
Now, recall from Section 2.3 (see p.76) that the vector N = xz i zy j + k is normal to the
tangent plane to the surface z = z( x, y) at each point of . Thus,
xz i zy j + k
N
n = = r
N 2 2
1 + xz + zy
is in fact a positive unit normal vector to (see Figure 4.5.4). Hence, using the parametriza-
tion r( x, y) = x i + y j + z( x, y) k, for ( x, y) in D , of the surface , we have rx = i + xz k and
r
r z
r r z 2 z 2
y
= j + y
k, and so
x
y
= 1 + x
+ y . So we see that using formula (4.46) for
curl f, we have

r r
(curl f) n d = (curl f ) n
dA
x y
D

R Q P R Q P z z
= i+ j+ k i j+k dA
y z z x x y x y
D

R Q z P R z Q P
= + dA ,
y z x z x y x y
D
which, upon comparing to equation (4.49), proves the Theorem. QED

Note: The condition in Stokes Theorem that the surface have a (continuously vary-
ing) positive unit normal vector n and a boundary curve C traversed n-positively can be
expressed more precisely as follows: if r( t) is the position vector for C and T( t) = r ( t)/k r ( t) k
is the unit tangent vector to C , then the vectors T, n, T n form a right-handed system.
Also, it should be noted that Stokes Theorem holds even when the boundary curve C is
piecewise smooth.
Example 4.14. Verify Stokes Theorem for f( x, y, z) = z i + x j + y k when is the paraboloid

z = x2 + y2 such that z 1 (see Figure 4.5.5).
Solution: The positive unit normal vector to the surface z

C
z = z( x, y) = x2 + y2 is 1 n
xz i zy j + k 2 x i 2 y j + k
n = r = p ,
z 2
z
2 1 + 4 x 2 + 4 y2
1 + x + y
and curl f = (1 0) i + (1 0) j + (1 0) k = i + j + k, so
y
q
0
(curl f ) n = (2 x 2 y + 1)/ 1 + 4 x2 + 4 y2 .
x
Since can be parametrized as r( x, y) = x i + y j + ( x + y ) k for2 2 z = x 2 + y2

Figure 4.5.5
( x, y) in the region D = { ( x, y) : x2 + y2 1 }, then

r r
(curl f ) n d = (curl f ) n
dA
x y
D
q
2 x 2 y + 1
= p 1 + 4 x 2 + 4 y2 d A
1 + 4 x 2 + 4 y2
D

= (2 x 2 y + 1) d A , so switching to polar coordinates gives
D
Z2 Z1
= (2 r cos 2 r sin + 1) r dr d
0 0
Z2 Z1
= (2 r 2 cos 2 r 2 sin + r ) dr d
0 0
Z2 r =1
2r3 2r3 r2
= 3 cos 3 sin + 2 d
0 r =0
Z2
2
= 3 cos 23 sin + 12 d
0
2
= 32 sin + 23 cos + 12 = .

0
The boundary curve C is the unit circle x2 + y2 = 1 laying in the plane z = 1 (see Figure
4.5.5), which can be parametrized as x = cos t, y = sin t, z = 1 for 0 t 2. So
I Z2
f dr = ((1)( sin t) + (cos t)(cos t) + (sin t)(0)) dt
C 0
Z2
1 + cos 2 t 2 1 + cos 2 t
= sin t + dt here we used cos t =
0 2 2
t sin 2 t 2
= cos t + + = .
2 4 0
H
So we see that C f d r = (curl f ) n d , as predicted by Stokes Theorem.

The line integral in the preceding example was far simpler to calculate than the surface
integral, but this will not always be the case.
2 y2
Example 4.15. Let be the elliptic paraboloid z = x4 + 9 for z 1, and let C be its boundary
H
curve. Calculate C f d r for f( x, y, z) = (9 xz + 2 y)i + (2 x + y2 )j + (2 y2 + 2 z)k, where C is
traversed counterclockwise.
Solution: The surface is similar to the one in Example 4.14, except now the boundary curve C
2 y2
is the ellipse x4 + 9 = 1 laying in the plane z = 1. In this case, using Stokes Theorem is easier
than computing the line integral directly. As in Example 4.14, at each point ( x, y, z( x, y)) on
2 y2
the surface z = z( x, y) = x4 + 9 the vector
xz i zy j + k 2y
2x i 9 j + k
n = r = q ,
z 2 z 2 x2 4 y2
1 + x + y 1 + 4 + 9
is a positive unit normal vector to . And calculating the curl of f gives
curl f = (4 y 0)i + (9 x 0)j + (2 2)k = 4 y i + 9 x j + 0 k ,
so
2y
(4 y)( 2x ) + (9 x)( 9 ) + (0)(1) 2x y 2x y + 0
(curl f ) n = q = q = 0,
2
x2 4y x2 4 y2
1+ 4 + 9 1+ 4 + 9
and so by Stokes Theorem
I
f dr = (curl f ) n d = 0 d = 0 .
C

H
In physical applications, for a simple closed curve C the line integral C f d r is often called
the circulation of f around C . For example, if E represents the electrostatic field due to a
H
point charge, then it turns out8 that curl E = 0, which means that the circulation C E d r = 0
by Stokes Theorem. Vector fields which have zero curl are often called irrotational fields.
In fact, the term curl was created by the 19th century Scottish physicist James Clerk
Maxwell in his study of electromagnetism, where it is used extensively. In physics, the
curl is interpreted as a measure of circulation density. This is best seen by using another
definition of curl f which is equivalent9 to the definition given by formula (4.46). Namely, for
a point ( x, y, z) in R3 , I
1
n (curl f )( x, y, z) = lim f dr , (4.50)
S 0 S C
where S is the surface area of a surface containing the point ( x, y, z) and with a simple
closed boundary curve C and positive unit normal vector n at ( x, y, z). In the limit, think of
the curve C shrinking to the point ( x, y, z), which causes , the surface it bounds, to have
smaller and smaller surface area. That ratio of circulation to surface area in the limit is
what makes the curl a rough measure of circulation density (i.e. circulation per unit area).
An idea of how the curl of a vector field is y
related to rotation is shown in Figure 4.5.6.
Suppose we have a vector field f( x, y, z) which
is always parallel to the x y-plane at each f
point ( x, y, z) and that the vectors grow larger
the further the point ( x, y, z) is from the y-
axis. For example, f( x, y, z) = (1 + x2 ) j. Think
of the vector field as representing the flow
of water, and imagine dropping two wheels
with paddles into that water flow, as in Fig-
ure 4.5.6. Since the flow is stronger (i.e. the
x
magnitude of f is larger) as you move away
0
from the y-axis, then such a wheel would ro-
tate counterclockwise if it were dropped to Figure 4.5.6 Curl and rotation
the right of the y-axis, and it would rotate
clockwise if it were dropped to the left of the y-axis. In both cases the curl would be nonzero
(curl f( x, y, z) = 2 x k in our example) and would obey the right-hand rule, that is, curl f( x, y, z)
points in the direction of your thumb as you cup your right hand in the direction of the rota-
tion of the wheel. So the curl points outward (in the positive z-direction) if x > 0 and points
inward (in the negative z-direction) if x < 0. Notice that if all the vectors had the same di-
rection and the same magnitude, then the wheels would not rotate and hence there would
be no curl (which is why such fields are called irrotational, meaning no rotation).
8 See Ch. 2 in R EITZ, M ILFORD and C HRISTY .

9 See S CHEY, p. 78-81, for the derivation.
Finally, by Stokes Theorem, we know that if C is a simple closed curve in some solid
region S in R3 and if f( x, y, z) is a smooth vector field such that curl f = 0 in S , then
I
f dr = (curl f ) n d = 0 n d = 0 d = 0 ,
C

where is any orientable surface inside S whose boundary is C (such a surface is some-
times called a capping surface for C ). So similar to the two-variable case, we have a three-
dimensional version of a result from Section 4.3, for solid regions in R3 which are simply
connected (i.e. regions having no holes):
The following statements are equivalent for a simply connected solid region S in R3 :
(a) f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k has a smooth potential F ( x, y, z) in S

Z
(b) f d r is independent of the path for any curve C in S
C
I
(c) f d r = 0 for every simple closed curve C in S
C
R Q P R Q P
(d) = , = , and = in S (i.e. curl f = 0 in S )
y z z x x y
Part (d) is also a way of saying that the differential form P dx + Q d y + R dz is exact.
Example 4.16. Determine if the vector field f( x, y, z) = x yz i + xz j + x y k has a potential in R3 .

Solution: Since R3 is simply connected, we just need to check whether curl f = 0 throughout
R3 , that is,
R Q P R Q P
= , = , and =
y z z x x y
throughout R3 , where P ( x, y, z) = x yz, Q ( x, y, z) = xz, and R ( x, y, z) = x y. But we see that
P R P R
= xy , =y 6= for some ( x, y, z) in R3 .
z x z x
Thus, f( x, y, z) does not have a potential in R3 .
Exercises
A R
For Exercises 1-3, calculate C f ( x, y, z) ds for the given function f ( x, y, z) and curve C .
1. f ( x, y, z) = z; C : x = cos t, y = sin t, z = t, 0 t 2
x
2. f ( x, y, z) = + y + 2 yz; C : x = t2 , y = t, z = 1, 1 t 2
y
p
C : x = t sin t, y = t cos t, z = 2 3 2 t3/2 , 0 t 1
3. f ( x, y, z) = z2 ;
R
For Exercises 4-9, calculate C f d r for the given vector field f( x, y, z) and curve C .
4. f( x, y, z) = i j + k; C : x = 3 t, y = 2 t, z = t, 0 t 1
5. f( x, y, z) = y i x j + z k; C : x = cos t, y = sin t, z = t, 0 t 2
6. f( x, y, z) = x i + y j + z k; C : x = cos t, y = sin t, z = 2, 0 t 2
7. f( x, y, z) = ( y 2 z) i + x y j + (2 xz + y) k; C : x = t, y = 2 t, z = t2 1, 0 t 1
8. f( x, y, z) = yz i + xz j + x y k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to (1, 2, 0)
9. f( x, y, z) = x y i + ( z x) j + 2 yz k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to (1, 2, 0)

to (1, 2, 2)
For Exercises 10-13, state whether or not the vector field f( x, y, z) has a potential in R3 (you
do not need to find the potential itself).
10. f( x, y, z) = y i x j + z k 11. f( x, y, z) = a i + b j + c k (a, b, c constant)
12. f( x, y, z) = ( x + y) i + x j + z2 k 13. f( x, y, z) = x y i ( x yz2 ) j + y2 z k
B
For Exercises 14-15, verify Stokes Theorem for the given vector field f( x, y, z) and surface .
14. f( x, y, z) = 2 y i x j + z k; : x2 + y2 + z2 = 1, z 0
15. f( x, y, z) = x y i + xz j + yz k; : z = x 2 + y2 , z 1
16. Construct a Mbius strip from a piece of paper, then draw a line down its center (like
the dotted line in Figure 4.5.3(b)). Cut the Mbius strip along that center line completely
around the strip. How many surfaces does this result in? How would you describe them?
Are they orientable?
17. Use Gnuplot (see Appendix C) to plot the Mbius strip parametrized as:
r( u, v) = cos u (1 + v cos u2 ) i + sin u (1 + v cos u2 ) j + v sin u2 k , 0 u 2 , 12 v 1

2
C
18. Let be a closed surface and f( x, y, z) a smooth vector field. Show that

(curl f ) n d = 0. (Hint: Split in half.)

19. Show that Greens Theorem is a special case of Stokes Theorem.

4.6 Gradient, Divergence, Curl and Laplacian 177
4.6 Gradient, Divergence, Curl and Laplacian

In this final section we will establish some relationships between the gradient, divergence
and curl, and we will also introduce a new quantity called the Laplacian. We will then show
how to write these quantities in cylindrical and spherical coordinates.
For a real-valued function f ( x, y, z) on R3 , the gradient f ( x, y, z) is a vector-valued func-
tion on R3 , that is, its value at a point ( x, y, z) is the vector

f f f f f f
f ( x, y, z) = , , = i + j + k
x y z x y z
in R3 , where each of the partial derivatives is evaluated at the point ( x, y, z). So in this way,
you can think of the symbol as being applied to a real-valued function f to produce a
vector f .
It turns out that the divergence and curl can also be expressed in terms of the symbol .
This is done by thinking of as a vector in R3 , namely

= i + j + k. (4.51)
x y z
Here, the symbols x , y and z are to be thought of as partial derivative operators that
will get applied to a real-valued function, say f ( x, y, z), to produce the partial derivatives
f f f f
,
x y
and z . For instance, x applied to f ( x, y, z) produces x .
Is really a vector? Strictly speaking, no, since x , y and z are not actual numbers. But
it helps to think of as a vector, especially with the divergence and curl, as we will soon see.
The process of applying x , y , z to a real-valued function f ( x, y, z) is normally thought of
as multiplying the quantities:

f f f
(f ) = , (f ) = , (f ) =
x x y y z z
For this reason, is often referred to as the del operator, since it operates on functions.
For example, it is often convenient to write the divergence div f as f, since for a vector
field f( x, y, z) = f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k, the dot product of f with (thought of as a
vector) makes sense:

f = i + j + k ( f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k)
x y z

= ( f1) + ( f2) + ( f3)
x y z
f1 f2 f3
= + +
x y z
= div f
We can also write curl f in terms of , namely as f, since for a vector field f( x, y, z) =
P ( x, y, z)i + Q ( x, y, z)j + R ( x, y, z)k, we have:

i j k

f =
x y z

P ( x, y, z) Q ( x, y, z) R ( x, y, z)

R Q R P Q P
= i j + k
y z x z x y

R Q P R Q P
= i + j + k
y z z x x y
= curl f
f f f
For a real-valued function f ( x, y, z), the gradient f ( x, y, z) = x
i + y j + z
k is a vector
field, so we can take its divergence:
div f = f

f f f
= i + j + k i + j + k
x y z x y z

f f f
= + +
x x y y z z
2 f 2 f 2 f
= + +
x2 y2 z2
Note that this is a real-valued function, to which we will give a special name:
Definition 4.7. For a real-valued function f ( x, y, z), the Laplacian of f , denoted by f , is

given by
2 f 2 f 2 f
f ( x, y, z) = f = + + . (4.52)
x2 y2 z2
Often the notation 2 f is used for the Laplacian instead of f , using the convention 2 =
.
Example 4.17. Let r( x, y, z) = x i+ y j+ z k be the position vector field on R3 . Then k r( x, y, z) k2 =

r r = x2 + y2 + z2 is a real-valued function. Find
(a) the gradient of k r k2
(b) the divergence of r
(c) the curl of r
(d) the Laplacian of k r k2

Solution: (a) k r k2 = 2 x i + 2 y j + 2 z k = 2 r

(b) r = x
( x) + y ( y) + z ( z) = 1 + 1 + 1 = 3
(c)
i j k

r = = (0 0) i (0 0) j + (0 0) k = 0
x y z

x y z
2 2 2
(d) k r k2 = x2
( x2 + y2 + z2 ) + y2 ( x2 + y2 + z2 ) + z2 ( x2 + y2 + z2 ) = 2 + 2 + 2 = 6
Note that we could have calculated k r k2 another way, using the notation along with parts
(a) and (b):
k r k2 = k r k2 = 2 r = 2 r = 2(3) = 6
Notice that in Example 4.17 if we take the curl of the gradient of k r k2 we get
(k r k2 ) = 2 r = 2 r = 2 0 = 0 .
The following theorem shows that this will be the case in general:
Theorem 4.15. For any smooth real-valued function f ( x, y, z), ( f ) = 0.
Proof: We see by the smoothness of f that

i j k

( f ) = x y z

f f f

x y z
2 2 2
f 2 f f 2 f f 2 f
= i j + k = 0,
y z z y x z z x x y y x
since the mixed partial derivatives in each component are equal. QED
Corollary 4.16. If a vector field f( x, y, z) has a potential, then curl f = 0.
Another way of stating Theorem 4.15 is that gradients are irrotational. Also, notice that
in Example 4.17 if we take the divergence of the curl of r we trivially get
( r) = 0 = 0 .
The following theorem shows that this will be the case in general:
Theorem 4.17. For any smooth vector field f( x, y, z), ( f) = 0.
The proof is straightforward and left as an exercise for the reader.

Corollary 4.18. The flux of the curl of a smooth vector field f( x, y, z) through any closed
surface is zero.
Proof: Let be a closed surface which bounds a solid S . The flux of f through is

( f ) d = ( f ) dV (by the Divergence Theorem)
S

= 0 dV (by Theorem 4.17)
S
= 0. QED
There is another method for proving Theorem 4.15 which can be useful, and is often used

in physics. Namely, if the surface integral f ( x, y, z) d = 0 for all surfaces in some solid

region (usually all of R3 ), then we must have f ( x, y, z) = 0 throughout that region. The proof
is not trivial, and physicists do not usually bother to prove it. But the result is true, and can
also be applied to double and triple integrals.
For instance, to prove Theorem 4.15, assume that f ( x, y, z) is a smooth real-valued func-
tion on R3 . Let C be a simple closed curve in R3 and let be any capping surface for C (i.e.
is orientable and its boundary is C ). Since f is a vector field, then
I
( ( f )) n d = f dr by Stokes Theorem, so
C

= 0 by Corollary 4.13.
Since the choice of was arbitrary, then we must have ( ( f )) n = 0 throughout R3 , where
n is any unit vector. Using i, j and k in place of n, we see that we must have ( f ) = 0 in
R3 , which completes the proof.
Example 4.18. A system of electric charges has a charge density ( x, y, z) and produces an
electrostatic field E( x, y, z) at points ( x, y, z) in space. Gauss Law states that

E d = 4 dV
S
for any closed surface which encloses the charges, with S being the solid region enclosed
by . Show that E = 4 . This is one of Maxwells Equations.10
10 In Gaussian (or CGS) units.

Solution: By the Divergence Theorem, we have

E dV = E d
S

= 4 dV by Gauss Law, so combining the integrals gives
S

( E 4 ) dV = 0 , so
S
E 4 = 0 since and hence S was arbitrary, so
E = 4 .
Often (especially in physics) it is convenient to use other coordinate systems when dealing
with quantities such as the gradient, divergence, curl and Laplacian. We will present the
formulas for these in cylindrical and spherical coordinates.
Recall from Section 1.7 that a point ( x, y, z) can be represented in cylindrical coordinates
( r, , z), where x = r cos , y = r sin , z = z. At each point ( r, , z), let er , e , e z be unit vectors
in the direction of increasing r , , z, respectively (see Figure 4.6.1). Then er , e , e z form an
orthonormal set of vectors. Note, by the right-hand rule, that e z er = e .
ez
e
z e z e
(x, y, z) (x, y, z)
er
e
z
y z y
0 0
x r x
y (x, y, 0) y (x, y, 0)
x x
Figure 4.6.1 Figure 4.6.2

Orthonormal vectors er , e , e z Orthonormal vectors e , e , e
in cylindrical coordinates in spherical coordinates
Similarly, a point ( x, y, z) can be represented in spherical coordinates ( , , ), where x =

sin cos , y = sin sin , z = cos . At each point ( , , ), let e , e , e be unit vectors
in the direction of increasing , , , respectively (see Figure 4.6.2). Then the vectors e , e ,
e are orthonormal. By the right-hand rule, we see that e e = e .
We can now summarize the expressions for the gradient, divergence, curl and Laplacian
in Cartesian, cylindrical and spherical coordinates in the following tables:
Cartesian ( x, y, z): Scalar function F ; Vector field f = f 1 i + f 2 j + f 3 k
F F F
gradient : F = i + j + k
x y z
f1 f2 f3
divergence : f = + +
x y
z

f3 f1 f3
f2 f2 f1
curl : f = i + j + k
y z z x x y
2 F 2 F 2 F
Laplacian : F = + +
x2 y2 z2
Cylindrical ( r, , z): Scalar function F ; Vector field f = f r er + f e + f z e z
F 1 F F
gradient : F = er + e + ez
r r z
1 1 f fz
divergence : f = (r f r ) + +
r r r z
1 f z f fr fz 1 fr
curl : f = er + e + (r f ) ez
r z z r r r

1 F 1 2 F 2 F
Laplacian : F = r + 2 2 +
r r r r z2
Spherical ( , , ): Scalar function F ; Vector field f = f e + f e + f e
F 1 F 1 F
gradient : F = e + e + e
sin
1 2 1 f 1
divergence : f = 2 ( f ) + + (sin f )
sin sin

1 f 1 f
curl : f = (sin f ) e + ( f ) e
sin

1 f 1
+ ( f ) e
sin

1 F 1 2 F 1 F
Laplacian : F = 2 2 + + sin
2 sin2 2 2 sin
The derivation of the above formulas for cylindrical and spherical coordinates is straight-
forward but extremely tedious. The basic idea is to take the Cartesian equivalent of the
quantity in question and to substitute into that formula using the appropriate coordinate
transformation. As an example, we will derive the formula for the gradient in spherical
coordinates.
Goal: Show that the gradient of a real-valued function F ( , , ) in spherical coordinates is:
F 1 F 1 F
F = e + e + e
sin
Idea: In the Cartesian gradient formula F ( x, y, z) = Fx i + Fy j + Fz k, put the Cartesian ba-

sis vectors i, j, k in terms of the spherical coordinate basis vectors e , e , e and functions of
, and . Then put the partial derivatives Fx , Fy , Fz in terms of F , F ,
F
and functions
of , and .
Step 1: Get formulas for e , e , e in terms of i, j, k.

We can see from Figure 4.6.2 that the unit vector e in the direction at a general point
( , , ) is e = k rr k , where r = x i + y j + z k is the position vector of the point in Cartesian
coordinates. Thus,
r xi+ yj+ zk
e = = p ,
krk x 2 + y2 + z 2
p
so using x = sin cos , y = sin sin , z = cos , and = x2 + y2 + z2 , we get:
e = sin cos i + sin sin j + cos k
Now, since the angle is measured in the x y-plane, then the unit vector e in the
direction must be parallel to the x y-plane. That is, e is of the form a i + b j + 0 k. To figure
out what a and b are, note that since e e , then in particular e e when e is in the
x y-plane. That occurs when the angle is /2. Putting = /2 into the formula for e gives
e = cos i + sin j + 0 k, and we see that a vector perpendicular to that is sin i + cos j + 0 k.
Since this vector is also a unit vector and points in the (positive) direction, it must be e :
e = sin i + cos j + 0 k
Lastly, since e = e e , we get:
e = cos cos i + cos sin j sin k
Step 2: Use the three formulas from Step 1 to solve for i, j, k in terms of e , e , e .
This comes down to solving a system of three equations in three unknowns. There are
many ways of doing this, but we will do it by combining the formulas for e and e to
eliminate k, which will give us an equation involving just i and j. This, with the formula for
e , will then leave us with a system of two equations in two unknowns (i and j), which we
will use to solve first for j then for i. Lastly, we will solve for k.
First, note that
sin e + cos e = cos i + sin j
so that
sin (sin e + cos e ) + cos e = (sin2 + cos2 )j = j ,
and so:
j = sin sin e + cos e + cos sin e
Likewise, we see that
cos (sin e + cos e ) sin e = (cos2 + sin2 )i = i ,
and so:
i = sin cos e sin e + cos cos e
Lastly, we see that:

k = cos e sin e
Step 3: Get formulas for F , F , F

in terms of F F F
, , .
x y z
By the Chain Rule, we have
F F x F y F z
= + + ,
x y z
F F x F y F z
= + + ,
x y z
F F x F y F z
= + + ,
x y z
which yields:
F F F F
= sin cos + sin sin + cos
x y z
F F F
= sin sin + sin cos
x y
F F F F
= cos cos + cos sin sin
x y z
Step 4: Use the three formulas from Step 3 to solve for Fx , Fy , Fz in terms of F , F ,
F
.
Again, this involves solving a system of three equations in three unknowns. Using a
similar process of elimination as in Step 2, we get:

F 1 F F F
= sin2 cos sin + sin cos cos
x sin

F 1 2 F F F
= sin sin + cos + sin cos sin
y sin

F 1 F F
= cos sin
z
F F F
Step 5: Substitute the formulas for i, j, k from Step 2 and the formulas for , ,
x y z
from
F F F
Step 4 into the Cartesian gradient formula F ( x, y, z) = i + j + k.
x y z
Doing this last step is perhaps the most tedious, since it involves simplifying 3 3 + 3 3 +
2 2 = 22 terms! Namely,

1 2 F F F
F = sin cos sin + sin cos cos (sin cos e sin e
sin
+ cos cos e )

1 F F F
+ sin2 sin + cos + sin cos sin (sin sin e + cos e
sin
+ cos sin e )

1 F F
+ cos sin (cos e sin e ) ,

which we see has 8 terms involving e , 6 terms involving e , and 8 terms involving e . But
the algebra is straightforward and yields the desired result:
F 1 F 1 F
F = e + e + e X
sin
Example 4.19. In Example 4.17 we showed that k r k2 = 2 r and k r k2 = 6, where r( x, y, z) =

x i + y j + z k in Cartesian coordinates. Verify that we get the same answers if we switch to
spherical coordinates.
Solution: Since k r k2 = x2 + y2 + z2 = 2 in spherical coordinates, let F ( , , ) = 2 (so that
F ( , , ) = k r k2 ). The gradient of F in spherical coordinates is
F 1 F 1 F
F = e + e + e
sin
1 1
= 2 e + (0) e + (0) e
sin
r
= 2 e = 2 , as we showed earlier, so
krk
r
= 2 = 2 r , as expected. And the Laplacian is

1 2 F 1 2 F 1 F
F = + + sin
2 2 sin2 2 2 sin
1 2 1 1
= 2
( 2 ) + 2 (0) + 2 sin (0)
sin sin
1
= (2 3 ) + 0 + 0
2
1
= (6 2 ) = 6 , as expected.
2
Exercises
A
For Exercises 1-6, find the Laplacian of the function f ( x, y, z) in Cartesian coordinates.
1. f ( x, y, z) = x + y + z 2. f ( x, y, z) = x5 3. f ( x, y, z) = ( x2 + y2 + z2 )3/2
2
y2 z2
4. f ( x, y, z) = e x+ y+ z 5. f ( x, y, z) = x3 + y3 + z3 6. f ( x, y, z) = e x
7. Find the Laplacian of the function in Exercise 3 in spherical coordinates.
8. Find the Laplacian of the function in Exercise 6 in spherical coordinates.

z
9. Let f ( x, y, z) = in Cartesian coordinates. Find f in cylindrical coordinates.
x 2 + y2
10. For f( r, , z) = r er + z sin e + rz e z in cylindrical coordinates, find div f and curl f.
11. For f( , , ) = e + cos e + e in spherical coordinates, find div f and curl f.
B
For Exercises 12-23, prove the given formula ( r = k r k is the length of the position vector field
r( x, y, z) = x i + y j + z k).
12. (1/ r ) = r/ r 3 13. (1/ r ) = 0 14. (r/ r 3 ) = 0 15. (ln r ) = r/ r 2
16. div (F + G) = div F + div G 17. curl (F + G) = curl F + curl G
18. div ( f F) = f div F + F f 19. div (F G) = G curl F F curl G
20. div ( f g) = 0 21. curl ( f F) = f curl F + ( f ) F
22. curl (curl F) = (div F) F 23. ( f g) = f g + g f + 2( f g)
C
24. Prove Theorem 4.17.
25. Derive the gradient formula in cylindrical coordinates: F = F

r
er + 1r F

e + Fz e z
26. Use f = u v in the Divergence Theorem to prove:

(a) Greens first identity: ( u v + ( u) (v)) dV = ( u v) d
S

(b) Greens second identity: ( u v v u) dV = ( u v v u) d
S
27. Suppose that u = 0 (i.e. u is harmonic) over R3 . Define the normal derivative un of u
over a closed surface with outward unit normal vector n by un = D n u = n u. Show
u
that n
d = 0. (Hint: Use Greens second identity.)

Bibliography
Abbott, E.A., Flatland, 7th edition. New York: Dover Publications, Inc., 1952
Classic tale about a creature living in a 2-dimensional world who encounters a higher-
dimensional creature, with lots of humor thrown in.
Anton, H. and C. Rorres, Elementary Linear Algebra: Applications Version, 8th edition. New
York: John Wiley & Sons, 2000
Standard treatment of elementary linear algebra.
Bazaraa, M.S., H.D. Sherali and C.M. Shetty, Nonlinear Programming: Theory and Algo-
rithms, 2nd edition. New York: John Wiley & Sons, 1993
Thorough treatment of nonlinear optimization.
Farin, G., Curves and Surfaces for Computer Aided Geometric Design: A Practical Guide,
2nd edition. San Diego, CA: Academic Press, 1990
An intermediate-level book on curve and surface design.
Hecht, E., Optics, 2nd edition. Reading, MA: Addison-Wesley Publishing Co., 1987
An intermediate-level book on optics, covering a wide range of topics.
Hoel, P.G., S.C. Port and C.J. Stone, Introduction to Probability Theory, Boston, MA:
Houghton Mifflin Co., 1971
An excellent introduction to elementary, calculus-based probability theory. Lots of good exer-
cises.
Jackson, J.D., Classical Electrodynamics, 2nd edition. New York: John Wiley & Sons, 1975
An advanced book on electromagnetism, famous for being intimidating. Most of the mathemat-
ics will be understandable after reading the present book.
Marion, J.B., Classical Dynamics of Particles and Systems, 2nd edition. New York: Academic
Press, 1970
Standard intermediate-level treatment of classical mechanics. Very thorough.
ONeill, B., Elementary Differential Geometry, New York: Academic Press, 1966
Intermediate-level book on differential geometry, with a modern approach based on differential
forms.
187
188 Bibliography
Pogorelov, A.V., Analytical Geometry, Moscow: Mir Publishers, 1980

An intermediate/advanced book on analytic geometry.
Press, W.H., S.A. Teukolsky, W.T. Vetterling and B.P. Flannery, Numerical Recipes in FOR-
TRAN: The Art of Scientific Computing, 2nd edition. Cambridge, UK: Cambridge Uni-
versity Press, 1992
An excellent source of information on numerical methods for solving a wide variety of problems.
Though all the examples are in the FORTRAN programming language, the code is clear enough
to implement in the language of your choice.
Protter, M.H. and C.B. Morrey, Analytic Geometry, 2nd edition. Reading, MA: Addison-
Wesley Publishing Co., 1975
Thorough treatment of elementary analytic geometry, with a rigor not found in most recent
books.
Ralston, A. and P. Rabinowitz, A First Course in Numerical Analysis, 2nd edition. New York:
McGraw-Hill, 1978
Standard treatment of elementary numerical analysis.
Reitz, J.R., F.J. Milford and R.W. Christy, Foundations of Electromagnetic Theory, 3rd edi-
tion. Reading, MA: Addison-Wesley Publishing Co., 1979
Intermediate text on electromagnetism.
Schey, H.M., Div, Grad, Curl, and All That: An Informal Text on Vector Calculus, New York:
W.W. Norton & Co., 1973
Very intuitive approach to the subject, from a physicists viewpoint. Highly recommended.
Taylor, A.E. and W.R. Mann, Advanced Calculus, 2nd edition. New York: John Wiley & Sons,
1972
Excellent treatment of n-dimensional calculus. A good book to study after the present book.
Many intriguing exercises.
Uspensky, J.V., Theory of Equations, New York: McGraw-Hill, 1948

A classic on the subject, discussing many interesting topics.
Weinberger, H.F., A First Course in Partial Differential Equations, New York: John Wiley &
Sons, 1965
A good introduction to the vast subject of partial differential equations.
Welchons, A.M. and W.R. Krickenberger, Solid Geometry, Boston, MA: Ginn & Co., 1936
A very thorough treatment of 3-dimensional geometry from an elementary perspective, in-
cludes many topics which (sadly) do not seem to be taught anymore.
Appendix A
Answers and Hints to Selected Exercises
Chapter 1 y = 1, z = 3 + t (c) x 2 = z 3, y = 1
5. x = 1 + 2 t, y = 2 + 7 t, z = 3 + 8 t
Section 1.1 (p. 8) 7. 7.65 9. (1, 2, 3)
p p p 11. 4 x 4 y + 3 z 10 = 0
1. (a)p 5 (b) 5 (c) 17 (d) 1
(e) 2 17 2. Yes 3. No 13. x 2 y z + 2 = 0 p
15. 11 x 24 y + 21 z 26 = 0 17. 9/ 35
19. x = 5 t, y = 2 + 3 t, z = 7 t 21. (10, 2, 1)
Section 1.2 (p. 14)
(4, 4, 3) (b) (2,p6, 1)

1. (a) p Section 1.6 (p. 46)
(c) p1 , p5 , p2 (d) 241 (e) 241
30 30 30
(f) (14, 6, 8) (g) (7, 3, 4) 1. radius: 1, center: (2, 3, 5) 3. radius: 5,
(h) (1, 6, 1) (i) (2, 4, 2) (j) No. center: (1, 1, 1) 5. No intersection. p
3. No. k v k + k w k is larger. 7. circle x2 + y2 = 4 in the planes z = 5
y y
9. lines ax = b , z = 0 and ax = b , z = 0

Section 1.3 (p. 18) 13. 22ac , 22bc , 0
1. 10 3. 73.4 5. 90 7. 0
9. Yes, since v w Section 1.7 (p. 50)
p = 0.
p
11. | v w | = 0 < 21 p
p p5 = k vpk k w k 1. (a) (4, 3 , 1) (b) ( 17, 3 , 1.816)
13. k v + w k = 26 < 21 + 5 = k v k + k w k p 11 p
3. (a) (2 7, 6 , 0) (b) (2 7, 116 , 2 )
15. Hint: use Definition 1.6.
5. (a) r 2 + z2 = 25 (b) = 5
24. Hint: See Theorem 1.10(c).
7. (a) r 2 + 9 z2 = 36 (b) 2 (1 + 8 cos2 ) = 36
10. (a, , a cot ) 12. Hint: Use the distance
Section 1.4 (p. 29) formula for Cartesian coordinates.
1. (5, 23, 24)
p 3. (8, 4, 5) 5. 0
7. 16.72 9. 4 5 11. 9 13. 0 Section 1.8 (p. 57)
and (8, 10, 2) 15. 14
1. f ( t) = (1, 2 t, 3 t2 ); x = 1 + t, y = z = 1
3. f ( t) = (2 sin 2 t, 2 cos 2 t, 1); x = 1,
Section 1.5 (p. 39)
y = 2 t, z = t 5. v( t) = (1, 1 cos t, sin t),
1. (a) (2, 3, 2) + t(5, 4, 3) (b) x = 2 + 5 t, a( t) = (0, sin t, cos t)
y3
y = 3 + 4 t, z = 2 3 t (c) x5 2 = 4 = z+32 9. (a) Line parallel to c (b) Half-line paral-
3. (a) (2, 1, 3) + t(1, 0, 1) (b) x = 2 + t, lel to c (c) Hint: Think of the
189
190 Appendix A: Answers and Hints to Selected Exercises
2 f
functions as position vectors. x y
= 21 x( x2 + y + 4)3/2
15. Hint: Theorem 1.16 2 f 2 f
21. x2
= y2 e x y , y2
= x2 e x y ,
2 f 2 f
x y
= (1 + x y) e xy
+1 23. x2
= 12 x2 ,
Section 1.9 (p. 63) 2 f 2 f 2 f
p y2
= 0, x y = 0 25. x2
= x 2 ,
3 5 3/2
1. 2 3. 2(5 8) 5. Replace 2 f 2 f
2
2/3 . y 2 = y , x y
=0
27 s+16
t by 2 4 9 6. Hint: Use
Theorem 1.20(e), Example 1.37, and Section 2.3 (p. 77)
Theorem 1.16 7. Hint: Use Exercise 6.
1. 2 x + 3 y z 3 = 0 3. 2 x + y z 2p= 0
9. Hint: Use f ( t) = k f( t) kT, differentiate 1 4 11
that to get f ( t), put those expressions into 5.p x + 2 y = z 7. 2 ( x 1) + 9 ( y 2) + 12 ( z
2 11
f ( t) f ( t), then write T ( t) in terms of 3 ) = 0 9. 3 x + 4 y 5 z = 0
1
N( t). 11. T( t) = p ( sin t, cos t, 1), N( t) =
2
( cos t, sin t, 0), B( t) = p1 (sin t, cos t, 1), Section 2.4 (p. 82)
2
( t) = 1/2 x y
1. (2 x, 2 y) 3. ( p ,p 2 2 )
x2 + y2 +4 x + y +4
p x yz), xz cos( x yz), x y cos( x yz))

5. (1/ x, 1/ y) 7. ( yz cos(
Chapter 2 9. (2 x, 2 y, 2 z) 11. 2 2 13. p1
p 3
15. 3 cos(1) 17. increase: (45, 20),
Section 2.1 (p. 70) decrease: (45, 20)
1. domain: R2 , range: [1, ) 3. domain:

{( x, y) : x2 + y2 4}, range: [0, )
Section 2.5 (p. 88)
5. domain: R3 , range: [1, 1] 7. 1 1. local min. (1, 0); saddle pt. (1, 0)
9. does not exist 11. 2 13. 2 15. 0 3. local min. (1, 1); local max. (1, 1); sad-
17. does not exist dle pts. (1, 1), (1, 1) 5. local min. (1, 1);
saddle pt. (0, 0) 7. local min. (0, 0)
Section 2.2 (p. 74) 9. local min. (1, 1/2) 11. width = height =
depth=10 13. x = y = 4, z = 2
f f f
1. x = 2 x, y = 2 y 3. x = x( x2 + y + 4)1/2 ,
f f f
y
= 12 ( x2 + y + 4)1/2 5. x
= ye x y + y, y
= Section 2.6 (p. 95)
f f
xe x y + x 7. = 4 x3 , y = 0
x 2. ( x0 , y0 ) = (0, 0) : (0.2858, 0.3998);
f f
9. x = x( x2 + y2 )1/2 , y = y( x2 + y2 )1/2 ( x0 , y0 ) = (1, 1) : (1.03256, 1.94037)
f
11. x = 23x ( x2 + y + 4)2/3 ,
f f 2 2
= 13 ( x2 + y + 4)2/3 13. x = 2 xe( x + y ) ,
Section 2.7 (p. 100)
y
f 2 2 f 4 p
y
= 2 ye( x + y ) 15. x = y cos( x y), 1. min. p , 2 ; max. p4 , p2
5 5 5 5
f 2 f 2 f
= x cos( x y) 17. x2 = 2, y2 = 2, 3. min. p20 , p30 ; max. p20 , p30
y
13 13 p13 13

2 f 2 f
x y
= 0 19. x2 = ( y + 4)( x2 + y + 4)3/2 , 4. min. 9
p , 0, p2 ; max. 89 , 459 , 41
5 5
2 f 8abc
y2
= 41 ( x2 + y + 4)3/2 , 5. p
3 3
191
Chapter 3 Chapter 4
Section 3.1 (p. 104) Section 4.1 (p. 142)
7 7 1 1. 1/2 3. 23 5. 24 7. 2 9. 2
1. 1 3. 12 5. 6 7. 5 9. 2 11. 15
11. 0
1 1
1. 1 3. 8 ln 2 3 5. 4 6. 4 7. 2 9. 6 2 y2
10. 56 1. 0 3. No 4. Yes. F ( x, y) = x2 2
5. No 9. (b) No. Hint: Think of how F is
defined. 10. Yes. F ( x, y) = ax y + bx + c y + d
1. 92 3. (2 cos(2 ) + 4 2)/4 5. 1
6 7. 6 Section 4.3 (p. 155)
10. 13
1. 16/15 3. 5 5. Yes. F ( x, y) = x y2 + x3
7. Yes. F ( x, y) = 4 x2 y + 2 y2 + 3 x
1. The values should converge to 1.318. Section 4.4 (p. 163)
(Hint: In Java the exponential function e x 1. 216 2. 3 3. 12/5 7. 15/4
can be obtained with Math.exp(x). Other
languages have similar functions, otherwise
use e = 2.7182818284590455 in your pro- p p p
gram.) 1. 2 2 2 2. (17 17 5 5)/3 3. 2/5
2. 1.146 3. 0.705 4. 0.168 4. 2 5. 2( 1) 7. 67/15 9. 6
11. Yes 13. No 19. Hint: Think of how
Section 3.5 (p. 123) a vector field f( x, y) = P ( x, y) i + Q ( x, y) j in R2
can be extended in a natural way to be a vec-
4
1. 8 3. 3/2
3 (8 3 ) 7. 1 sin2 2 9. 2ab tor field in R3 .
Section 3.6 (p. 127) Section 4.6 (p. 186)

p
1. (1, 8/3) 3. (0, 34a ) 5. (0, 3/16) 1. 0 3. 12 x2 + y2 + z2 5. 6( x + y + z)
2
7. (0, 0, 5a/12) 9. (7/12, 7/12, 7/12) 7. 12 8. (4 2 6) e 9. 2r3z er + r12 e z
2 sin
11. div f = sin + cot ;
curl f = cot cos e + 2e 2 cos e
25. Hint: Start by showing that er = cos i +
p n 1
1. 2. 1 6. Both are ( n+1)2 ( n+2)
7. n sin j, e = sin i + cos j, e z = k.
Appendix B
We will prove the right-hand rule for the cross product of two vectors in R3 .
For any vectors v and w in R3 , define a new vector, n(v, w), as follows:
1. If v and w are nonzero and not parallel, and is the angle between them, then n(v, w) is
the vector in R3 such that:
(a) the magnitude of n(v, w) is k v k k w k sin ,
(b) n(v, w) is perpendicular to the plane containing v and w, and
(c) v, w, n(v, w) form a right-handed system.
2. If v and w are nonzero and parallel, then n(v, w) = 0.
3. If either v or w is 0, then n(v, w) = 0.
The goal is to show that n(v, w) = v w for all v, w in R3 , which would prove the right-hand
rule for the cross product (by part 1(c) of our definition). To do this, we will perform the
following steps:
Step 1: Show that n(v, w) = v w if v and w are any two of the basis vectors i, j, k.
This was already shown in Example 1.11 in Section 1.4. X
Step 2: Show that n(av, bw) = ab(v w) for any scalars a, b if v and w are any two of the
basis vectors i, j, k.
If either a = 0 or b = 0 then n(av, bw) = 0 = ab(v w), so the result holds. So assume that
a 6= 0 and b 6= 0. Let v and w be any two of the basis vectors i, j, k. For example, we will show
that the result holds for v = i and w = k (the other possibilities follow in a similar fashion).
For av = ai and bw = bk, the angle between av and bw is 90 . Hence the magnitude
of n(av, bw), by definition, is k ai k k bk k sin 90 = | ab |. Also, by definition, n(av, bw) is per-
pendicular to the plane containing ai and bk, namely, the xz-plane. Thus, n(av, bw) must
be a scalar multiple of j. Since its magnitude is | ab |, then n(av, bw) must be either | ab |j or
| ab |j.
There are four possibilities for the combinations of signs for a and b. We will consider the
case when a > 0 and b > 0 (the other three possibilities are handled similarly).
192
193
In this case, n(av, bw) must be either abj or abj. Now, since i, j, k form a right-handed
system, then i, k, j form a left-handed system, and so i, k, j form a right-handed system.
Thus, ai, bk, abj form a right-handed system (since a > 0, b > 0, and ab > 0). So since, by
definition, ai, bk, n(ai, bk) form a right-handed system, and since n(ai, bk) has to be either
abj or abj, this means that we must have n(ai, bk) = abj.
But we know that ai bk = ab(i k) = ab(j) = abj. Therefore, n(ai, bk) = ab(i k),
which is what we needed to show.
n(av, bw) = ab(v w) X
Step 3: Show that n(u, v + w) = n(u, v) + n(u, w) for any vectors u, v, w.

If u = 0 then the result holds trivially since n(u, v + w), n(u, v) and n(u, w) are all the zero
vector. If v = 0, then the result follows easily since n(u, v + w) = n(u, 0 + w) = n(u, w) =
0 + n(u, w) = n(u, 0) = n(u, w) = n(u, v) + n(u, w). A similar argument shows that the result
holds if w = 0.
So now assume that u, v and w are all nonzero vectors. We will describe a geometric
construction of n(u, v), which is shown in the figure below. Let P be a plane perpendicular
to u. Multiply the vector v by the positive scalar k u k, then project the vector k u k v straight
down onto the plane P . You can think of this projection vector (denoted by pro j P k u k v) as
the shadow of the vector k u k v on the plane P , with the light source directly overhead the
terminal point of k u k v. If is the angle between u and v, then we see that pro j P k u k v has
magnitude k u k k v k sin , which is the magnitude of n(u, v). So rotating pro j P k u k v by 90
in a counter-clockwise direction in the plane P gives a vector whose magnitude is the same
as that of n(u, v) and which is perpendicular to pro j P k u k v (and hence perpendicular to v).
Since this vector is in P then it is also perpendicular to u. And we can see that u, v and
this vector form a right-handed system. Hence this vector must be n(u, v). Note that this
holds even if u v, since in that case = 0 and so sin = 0 which means that n(u, v) has
magnitude 0, which is what we would expect.
kuk v
u

v
pro j P k u k v
n(u, v)
Now apply this same geometric construction to get n(u, w) and n(u, v + w). Since k u k (v +
w) is the sum of the vectors k u k v and k u k w, then the projection vector pro j P k u k (v + w) is
the sum of the projection vectors pro j P k u k v and pro j P k u k w (to see this, using the shadow
194 Appendix B: Proof of the Right-Hand Rule for the Cross Product
analogy again and the parallelogram rule for vector addition, think of how projecting a
parallelogram onto a plane gives you a parallelogram in that plane). So then rotating all
three projection vectors by 90 in a counter-clockwise direction in the plane P preserves that
sum (see the figure below), which means that n(u, v + w) = n(u, v) + n(u, w). X
kuk v
k u k (v + w) u

v
v+w
pro j P k u k v
pro j P k u k (v + w) kuk w w
pro j P k u k w
n(u, v) n(u, w)
P
n(u, v + w)
Step 4: Show that n(w, v) = n(v, w) for any vectors v, w.

If v and w are nonzero and parallel, or if either is 0, then n(w, v) = 0 = n(v, w), so the result
holds. So assume that v and w are nonzero and not parallel. Then n(w, v) has magnitude
k w k k v k sin , which is the same as the magnitude of n(v, w), and hence is the same as the
magnitude of n(v, w). By definition, n(v, w) is perpendicular to the plane containing w and
v, and hence so is n(v, w). Also, v, w, n(v, w) form a right-handed system, and so w, v,
n(v, w) form a left-handed system, and hence w, v, n(v, w) form a right-handed system.
Thus, we have shown that n(v, w) is a vector with the same magnitude as n(w, v) and is
perpendicular to the plane containing w and v, and that w, v, n(v, w) form a right-handed
system. So by definition this means that n(v, w) must be n(w, v). X
Step 5: Show that n(v, w) = v w for all vectors v, w.

Write v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k. Then by Steps 3 and 4, we have
195
n(v, w) = n(v1 i + v2 j + v3 k, w1 i + w2 j + w3 k)
= n(v1 i + v2 j + v3 k, w1 i) + n(v1 i + v2 j + v3 k, w2 j + w3 k)
= n(v1 i + v2 j + v3 k, w1 i) + n(v1 i + v2 j + v3 k, w2 j) + n(v1 i + v2 j + v3 k, w3 k)
= n(w1 i, v1 i + v2 j + v3 k) + n(w2 j, v1 i + v2 j + v3 k) + n(w3 k, v1 i + v2 j + v3 k).
We can use Steps 1 and 2 to evaluate the three terms on the right side of the last equation
above:
n(w1 i, v1 i + v2 j + v3 k) = n(w1 i, v1 i) + n(w1 i, v2 j) + n(w1 i, v3 k)

= v1 w1 n(i, i) + v2 w1 n(i, j) + v3 w1 n(i, k)
= v1 w1 (i i) + v2 w1 (i j) + v3 w1 (i k)
= v1 w1 0 + v2 w1 k + v3 w1 (j)
n(w1 i, v1 i + v2 j + v3 k) = v2 w1 k + v3 w1 j
Similarly, we can calculate
n(w2 j, v1 i + v2 j + v3 k) = v1 w2 k v3 w2 i
and
n(w3 j, v1 i + v2 j + v3 k) = v1 w3 j + v2 w3 i .
Thus, putting it all together, we have
n(v, w) = v2 w1 k + v3 w1 j + v1 w2 k v3 w2 i v1 w3 j + v2 w3 i
= (v2 w3 v3 w2 )i + (v3 w1 v1 w3 )j + (v1 w2 v2 w1 )k
= v w by definition of the cross product.
n(v, w) = v w for all vectors v, w. X

So since v, w, n(v, w) form a right-handed system, then v, w, v w form a right-handed
system, which completes the proof.
Appendix C
3D Graphing with Gnuplot
Gnuplot is a free, open-source software package for producing a variety of graphs. Versions
are available for many operating systems. Below is a very brief tutorial on how to use
Gnuplot to graph functions of several variables.
INSTALLATION
1. Go to http://www.gnuplot.info/download.html and follow the links to download the lat-

est version for your operating system. For Windows, you should get the Zip file with a
name such as gp420win32.zip, which is version 4.2.0. All the examples we will discuss
require at least version 4.2.0.
2. Install the downloaded file. For example, in Windows you would unzip the Zip file you
downloaded in Step 1 into some folder (use the Use folder names option if extracting
with WinZip).
RUNNING GNUPLOT
1. In Windows, run wgnuplot.exe from the folder (or bin folder) where you installed Gnu-
plot. In Linux, just type gnuplot in a terminal window.
2. You should now get a Gnuplot terminal with a gnuplot> command prompt. In Windows
this will appear in a new window, while in Linux it will appear in the terminal window
where the gnuplot command was run. For Windows, if the font is unreadable you can
change it by right-clicking on the text part of the Gnuplot window and selecting the
Choose Font.. option. For example, the font Courier, style Regular, size 12 is
usually a good choice (that choice can be saved for future sessions by right-clicking in the
Gnuplot window again and selecting the option to update wgnuplot.ini).
3. At the gnuplot> command prompt you can now run graphing commands, which we will
now describe.
GRAPHING FUNCTIONS
The usual way to create 3D graphs in Gnuplot is with the splot command:
splot <range> <comma-separated list of functions>
196
197
For a function z = f ( x, y), <range> is the range of x and y values (and optionally the range
of z values) over which to plot. To specify an x range and a y range, use an expression of the
form [a : b][ c : d ], for some numbers a < b and c < d . This will cause the graph to be plotted
for a x b and c y d .
Function definitions use the x and y variables in combination with mathematical operators,
listed below:
Symbol Operation Example Result

+ Addition 2+3 5
Subtraction 32 1
* Multiplication 2*3 6
/ Division 4/2 2
3
** Power 2**3 2 =8
exp( x) ex exp(2) e2
log( x) ln x log(2) ln 2
sin( x) sin x sin(pi/2) 1
cos( x) cos x cos(pi) 1
tan( x) tan x tan(pi/4) 1
Example C.1. To graph the function z = 2 x2 + y2 from x = 1 to x = 1 and from y = 2 to

y = 2, type this at the gnuplot> prompt:
splot [-1:1][-2:2] 2*x**2 + y**2
The result is shown below:

2 x 2 + y 2
7
6
5
4
3
2
1
0
2
1.5
1
0.5
-1 0
-0.5 -0.5
0 -1
0.5 -1.5
1 -2
198 Appendix C: 3D Graphing with Gnuplot
Note that we had to type 2*x**2 to multiply 2 times x2 . For clarity, parentheses can be used
to make sure the operations are being performed in the correct order:
splot [-1:1][-2:2] 2*(x**2) + y**2
In the above example, to also plot the function z = e x+ y on the same graph, put a comma
after the first function then append the new function:
splot [-1:1][-2:2] 2*(x**2) + y**2, exp(x+y)
By default, the x-axis and y-axis are not shown in the graph. To display the axes, use this
command before the splot command:
set zeroaxis
Also, by default the x- and y-axes are switched from their usual position. To show the axes
with the orientation which we have used throughout the text, use this command:
set view 60,120,1,1
Also, to label the axes, use these commands:
set xlabel "x"

set ylabel "y"
set zlabel "z"
To show the level curves of the surface z = f ( x, y) on both the surface and projected onto the
x y-plane, use this command:
set contour both
The default mesh size for the grid on the surface is 10 units. To get more of a colored/shaded
surface, increase the mesh size (to, say, 25) like this:
set isosamples 25
Putting all this together, we get the following graph with these commands:
set zeroaxis
set view 60,120,1,1
set xlabel "x"
set ylabel "y"
set zlabel "z"
set contour both
set isosamples 25
splot [-1:1][-2:2] 2*(x**2) + y**2, exp(x+y)
199
2 x 2 + y 2
6
5
4
25 3
20 2
1
15
exp( x + y)
10 20
z
5 15
0 10
5
-1
-0.5
-2 -1.5 0
-1 -0.5 x
0 0.5 0.5
y 1 1.5
2 1
The numbers listed below the functions in the key in the upper right corner of the graph
are the levels of the level curves of the corresponding surface. That is, they are the num-
bers c such that f ( x, y) = c. Because of the large number of level curves, the key was put
outside the graph with the set key outside command. If you do not want the function key
displayed, it can be turned off with this command: unset key
PARAMETRIC FUNCTIONS
Gnuplot has the ability to graph surfaces given in various parametric forms. For example,
for a surface parametrized in cylindrical coordinates
x = r cos , y = r sin , z=z
you would do the following:
set mapping cylindrical

set parametric
splot [a:b][c:d] v*cos(u),v*sin(u),f(u,v)
where the variable u represents , with a u b, the variable v represents r , with c v d ,

and z = f ( u, v) is some function of u and v.
Example C.2. The graph of the helicoid z = in Example 1.34 from Section 1.7 (p. 49) was
created using the following commands:
200 Appendix C: 3D Graphing with Gnuplot
set mapping cylindrical

set parametric
set view 60,120,1,1
set xyplane 0
set xlabel "x"
set ylabel "y"
set zlabel "z"
unset key
set isosamples 15
splot [0:4*pi][0:2] v*cos(u),v*sin(u),u
The command set xyplane 0 moves the z-axis so that z = 0 aligns with the x y-plane (which
is not the default in Gnuplot). Looking at the graph, you will see that r varies from 0 to 2,
and varies from 0 to 4.
PRINTING AND SAVING

In Windows, to print a graph from Gnuplot right-click on the titlebar of the graphs window,
select Options and then the Print.. option. If that does not work on your version of
Gnuplot, then go to the File menu on the main Gnuplot menubar, select Output Device ...,
and enter pdf in the Terminal type? textfield, hit OK. That will allow you to print the graph
as a PDF file.
To save a graph, say, as a PNG file, go to the File menu on the main Gnuplot menubar,
select Output Device ..., and enter png in the Terminal type? textfield, hit OK. Then, in the
File menu again, select the Output ... option and enter a filename (say, graph.png) in the
Output filename? textfield, hit OK. Now run your splot command again and you should see
a file called graph.png in the current directory (usually the directory where wgnuplot.exe is
located, though you can change that setting using the Change Directory ... option in the
File menu).
In Linux, to save the graph as a file called graph.png, you would issue the following com-
mands:
set terminal png
set output graph.png
and then run your splot command. There are many terminal types (which determine the
output format). Run the command set terminal to see all the possible types. In Linux,
the postscript terminal type is popular, since the print quality is high and there are many
PostScript viewers available.
To quit Gnuplot, type quit at the gnuplot> command prompt.
GNU Free Documentation License
Version 1.2, November 2002
Copyright 2000,2001,2002 Free Software Foundation, Inc.
51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Everyone is permitted to copy and distribute verbatim copies of this license document, but
changing it is not allowed.
Preamble
The purpose of this License is to make a manual, textbook, or other functional and useful
document "free" in the sense of freedom: to assure everyone the effective freedom to copy
and redistribute it, with or without modifying it, either commercially or noncommercially.
Secondarily, this License preserves for the author and publisher a way to get credit for their
work, while not being considered responsible for modifications made by others.
This License is a kind of "copyleft", which means that derivative works of the document
must themselves be free in the same sense. It complements the GNU General Public License,
which is a copyleft license designed for free software.
We have designed this License in order to use it for manuals for free software, because free
software needs free documentation: a free program should come with manuals providing the
same freedoms that the software does. But this License is not limited to software manuals; it
can be used for any textual work, regardless of subject matter or whether it is published as a
printed book. We recommend this License principally for works whose purpose is instruction
or reference.
1. APPLICABILITY AND DEFINITIONS

This License applies to any manual or other work, in any medium, that contains a notice
placed by the copyright holder saying it can be distributed under the terms of this License.
Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that
work under the conditions stated herein. The "Document", below, refers to any such manual
or work. Any member of the public is a licensee, and is addressed as "you". You accept
the license if you copy, modify or distribute the work in a way requiring permission under
copyright law.
A "Modified Version" of the Document means any work containing the Document or a
portion of it, either copied verbatim, or with modifications and/or translated into another
language.
201
202 GNU Free Documentation License
A "Secondary Section" is a named appendix or a front-matter section of the Document

that deals exclusively with the relationship of the publishers or authors of the Document
to the Documents overall subject (or to related matters) and contains nothing that could
fall directly within that overall subject. (Thus, if the Document is in part a textbook of
mathematics, a Secondary Section may not explain any mathematics.) The relationship
could be a matter of historical connection with the subject or with related matters, or of
legal, commercial, philosophical, ethical or political position regarding them.
The "Invariant Sections" are certain Secondary Sections whose titles are designated,
as being those of Invariant Sections, in the notice that says that the Document is released
under this License. If a section does not fit the above definition of Secondary then it is not
allowed to be designated as Invariant. The Document may contain zero Invariant Sections.
If the Document does not identify any Invariant Sections then there are none.
The "Cover Texts" are certain short passages of text that are listed, as Front-Cover Texts
or Back-Cover Texts, in the notice that says that the Document is released under this Li-
cense. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most
25 words.
A "Transparent" copy of the Document means a machine-readable copy, represented in
a format whose specification is available to the general public, that is suitable for revising
the document straightforwardly with generic text editors or (for images composed of pixels)
generic paint programs or (for drawings) some widely available drawing editor, and that
is suitable for input to text formatters or for automatic translation to a variety of formats
suitable for input to text formatters. A copy made in an otherwise Transparent file format
whose markup, or absence of markup, has been arranged to thwart or discourage subsequent
modification by readers is not Transparent. An image format is not Transparent if used for
any substantial amount of text. A copy that is not "Transparent" is called "Opaque".
Examples of suitable formats for Transparent copies include plain ASCII without markup,
Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD,
and standard-conforming simple HTML, PostScript or PDF designed for human modifica-
tion. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats
include proprietary formats that can be read and edited only by proprietary word proces-
sors, SGML or XML for which the DTD and/or processing tools are not generally available,
and the machine-generated HTML, PostScript or PDF produced by some word processors for
output purposes only.
The "Title Page" means, for a printed book, the title page itself, plus such following pages
as are needed to hold, legibly, the material this License requires to appear in the title page.
For works in formats which do not have any title page as such, "Title Page" means the text
near the most prominent appearance of the works title, preceding the beginning of the body
of the text.
A section "Entitled XYZ" means a named subunit of the Document whose title either
is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in an-
other language. (Here XYZ stands for a specific section name mentioned below, such as
"Acknowledgments", "Dedications", "Endorsements", or "History".) To "Preserve the
203
Title" of such a section when you modify the Document means that it remains a section
"Entitled XYZ" according to this definition.
The Document may include Warranty Disclaimers next to the notice which states that
this License applies to the Document. These Warranty Disclaimers are considered to be
included by reference in this License, but only as regards disclaiming warranties: any other
implication that these Warranty Disclaimers may have is void and has no effect on the
meaning of this License.
2. VERBATIM COPYING
You may copy and distribute the Document in any medium, either commercially or non-
commercially, provided that this License, the copyright notices, and the license notice say-
ing this License applies to the Document are reproduced in all copies, and that you add no
other conditions whatsoever to those of this License. You may not use technical measures to
obstruct or control the reading or further copying of the copies you make or distribute. How-
ever, you may accept compensation in exchange for copies. If you distribute a large enough
number of copies you must also follow the conditions in section 3.
You may also lend copies, under the same conditions stated above, and you may publicly
display copies.
3. COPYING IN QUANTITY
If you publish printed copies (or copies in media that commonly have printed covers) of
the Document, numbering more than 100, and the Documents license notice requires Cover
Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover
Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both
covers must also clearly and legibly identify you as the publisher of these copies. The front
cover must present the full title with all words of the title equally prominent and visible.
You may add other material on the covers in addition. Copying with changes limited to the
covers, as long as they preserve the title of the Document and satisfy these conditions, can
be treated as verbatim copying in other respects.
If the required texts for either cover are too voluminous to fit legibly, you should put the
first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto
adjacent pages.
If you publish or distribute Opaque copies of the Document numbering more than 100,
you must either include a machine-readable Transparent copy along with each Opaque copy,
or state in or with each Opaque copy a computer-network location from which the general
network-using public has access to download using public-standard network protocols a com-
plete Transparent copy of the Document, free of added material. If you use the latter option,
you must take reasonably prudent steps, when you begin distribution of Opaque copies in
quantity, to ensure that this Transparent copy will remain thus accessible at the stated lo-
cation until at least one year after the last time you distribute an Opaque copy (directly or
through your agents or retailers) of that edition to the public.
It is requested, but not required, that you contact the authors of the Document well before
redistributing any large number of copies, to give them a chance to provide you with an
updated version of the Document.
4. MODIFICATIONS
You may copy and distribute a Modified Version of the Document under the conditions of
sections 2 and 3 above, provided that you release the Modified Version under precisely this
License, with the Modified Version filling the role of the Document, thus licensing distribu-
tion and modification of the Modified Version to whoever possesses a copy of it. In addition,
you must do these things in the Modified Version:
A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document,
and from those of previous versions (which should, if there were any, be listed in the
History section of the Document). You may use the same title as a previous version if the
original publisher of that version gives permission.
B. List on the Title Page, as authors, one or more persons or entities responsible for au-
thorship of the modifications in the Modified Version, together with at least five of the
principal authors of the Document (all of its principal authors, if it has fewer than five),
unless they release you from this requirement.
C. State on the Title page the name of the publisher of the Modified Version, as the pub-
lisher.
D. Preserve all the copyright notices of the Document.
E. Add an appropriate copyright notice for your modifications adjacent to the other copy-
right notices.
F. Include, immediately after the copyright notices, a license notice giving the public per-
mission to use the Modified Version under the terms of this License, in the form shown
in the Addendum below.
G. Preserve in that license notice the full lists of Invariant Sections and required Cover
Texts given in the Documents license notice.
H. Include an unaltered copy of this License.
I. Preserve the section Entitled "History", Preserve its Title, and add to it an item stating
at least the title, year, new authors, and publisher of the Modified Version as given on the
Title Page. If there is no section Entitled "History" in the Document, create one stating
the title, year, authors, and publisher of the Document as given on its Title Page, then
add an item describing the Modified Version as stated in the previous sentence.
205
J. Preserve the network location, if any, given in the Document for public access to a Trans-
parent copy of the Document, and likewise the network locations given in the Document
for previous versions it was based on. These may be placed in the "History" section. You
may omit a network location for a work that was published at least four years before the
Document itself, or if the original publisher of the version it refers to gives permission.
K. For any section Entitled "Acknowledgments" or "Dedications", Preserve the Title of the
section, and preserve in the section all the substance and tone of each of the contributor
acknowledgments and/or dedications given therein.
L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their
titles. Section numbers or the equivalent are not considered part of the section titles.
M. Delete any section Entitled "Endorsements". Such a section may not be included in the
Modified Version.
N. Do not retitle any existing section to be Entitled "Endorsements" or to conflict in title

with any Invariant Section.
O. Preserve any Warranty Disclaimers.
If the Modified Version includes new front-matter sections or appendices that qualify as
Secondary Sections and contain no material copied from the Document, you may at your
option designate some or all of these sections as invariant. To do this, add their titles to
the list of Invariant Sections in the Modified Versions license notice. These titles must be
distinct from any other section titles.
You may add a section Entitled "Endorsements", provided it contains nothing but endorse-
ments of your Modified Version by various partiesfor example, statements of peer review
or that the text has been approved by an organization as the authoritative definition of a
standard.
You may add a passage of up to five words as a Front-Cover Text, and a passage of up to
25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version.
Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or
through arrangements made by) any one entity. If the Document already includes a cover
text for the same cover, previously added by you or by arrangement made by the same entity
you are acting on behalf of, you may not add another; but you may replace the old one, on
explicit permission from the previous publisher that added the old one.
The author(s) and publisher(s) of the Document do not by this License give permission to
use their names for publicity for or to assert or imply endorsement of any Modified Version.
5. COMBINING DOCUMENTS
You may combine the Document with other documents released under this License, under
the terms defined in section 4 above for modified versions, provided that you include in the
combination all of the Invariant Sections of all of the original documents, unmodified, and
list them all as Invariant Sections of your combined work in its license notice, and that you
preserve all their Warranty Disclaimers.
The combined work need only contain one copy of this License, and multiple identical In-
variant Sections may be replaced with a single copy. If there are multiple Invariant Sections
with the same name but different contents, make the title of each such section unique by
adding at the end of it, in parentheses, the name of the original author or publisher of that
section if known, or else a unique number. Make the same adjustment to the section titles
in the list of Invariant Sections in the license notice of the combined work.
In the combination, you must combine any sections Entitled "History" in the various orig-
inal documents, forming one section Entitled "History"; likewise combine any sections En-
titled "Acknowledgments", and any sections Entitled "Dedications". You must delete all
sections Entitled "Endorsements".
6. COLLECTIONS OF DOCUMENTS
You may make a collection consisting of the Document and other documents released un-
der this License, and replace the individual copies of this License in the various documents
with a single copy that is included in the collection, provided that you follow the rules of this
License for verbatim copying of each of the documents in all other respects.
You may extract a single document from such a collection, and distribute it individually
under this License, provided you insert a copy of this License into the extracted document,
and follow this License in all other respects regarding verbatim copying of that document.
7. AGGREGATION WITH INDEPENDENT WORKS

A compilation of the Document or its derivatives with other separate and independent
documents or works, in or on a volume of a storage or distribution medium, is called an "ag-
gregate" if the copyright resulting from the compilation is not used to limit the legal rights
of the compilations users beyond what the individual works permit. When the Document
is included in an aggregate, this License does not apply to the other works in the aggregate
which are not themselves derivative works of the Document.
If the Cover Text requirement of section 3 is applicable to these copies of the Document,
then if the Document is less than one half of the entire aggregate, the Documents Cover
Texts may be placed on covers that bracket the Document within the aggregate, or the elec-
tronic equivalent of covers if the Document is in electronic form. Otherwise they must ap-
pear on printed covers that bracket the whole aggregate.
8. TRANSLATION
Translation is considered a kind of modification, so you may distribute translations of the
Document under the terms of section 4. Replacing Invariant Sections with translations re-
quires special permission from their copyright holders, but you may include translations of
207
some or all Invariant Sections in addition to the original versions of these Invariant Sections.
You may include a translation of this License, and all the license notices in the Document,
and any Warranty Disclaimers, provided that you also include the original English version
of this License and the original versions of those notices and disclaimers. In case of a dis-
agreement between the translation and the original version of this License or a notice or
disclaimer, the original version will prevail.
If a section in the Document is Entitled "Acknowledgments", "Dedications", or "History",
the requirement (section 4) to Preserve its Title (section 1) will typically require changing
the actual title.
9. TERMINATION
You may not copy, modify, sublicense, or distribute the Document except as expressly pro-
vided for under this License. Any other attempt to copy, modify, sublicense or distribute the
Document is void, and will automatically terminate your rights under this License. How-
ever, parties who have received copies, or rights, from you under this License will not have
their licenses terminated so long as such parties remain in full compliance.
10. FUTURE REVISIONS OF THIS LICENSE
The Free Software Foundation may publish new, revised versions of the GNU Free Doc-
umentation License from time to time. Such new versions will be similar in spirit to the
present version, but may differ in detail to address new problems or concerns. See http://
www.gnu.org/copyleft/.
Each version of the License is given a distinguishing version number. If the Document
specifies that a particular numbered version of this License "or any later version" applies to
it, you have the option of following the terms and conditions either of that specified version or
of any later version that has been published (not as a draft) by the Free Software Foundation.
If the Document does not specify a version number of this License, you may choose any
version ever published (not as a draft) by the Free Software Foundation.
ADDENDUM: How to use this License for your documents
To use this License in a document you have written, include a copy of the License in the
document and put the following copyright and license notices just after the title page:
Copyright YEAR YOUR NAME. Permission is granted to copy, distribute and/or

modify this document under the terms of the GNU Free Documentation License, Ver-
sion 1.2 or any later version published by the Free Software Foundation; with no
Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the
license is included in the section entitled "GNU Free Documentation License".
If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the
"with...Texts." line with this:
with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts
being LIST, and with the Back-Cover Texts being LIST.
If you have Invariant Sections without Cover Texts, or some other combination of the
three, merge those two alternatives to suit the situation.
If your document contains nontrivial examples of program code, we recommend releas-
ing these examples in parallel under your choice of free software license, such as the GNU
General Public License, to permit their use in free software.
History
This section contains the revision history of the book. For persons making modifications to
the book, please record the pertinent information here, following the format in the first item
below.
1. VERSION: 1.0
Date: 2008-01-04
Author(s): Michael Corral
Title: Vector Calculus
Modification(s): Initial version
209
Index
Symbols annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 arc length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
M x , M y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 area element . . . . . . . . . . . . . . . . . . . . . . . . . . 105
M x y , M xz , M yz . . . . . . . . . . . . . . . . . . . . . . . . 126 average value . . . . . . . . . . . . . . . . . . . . . . . . . 113
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
R2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 B
R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Bzier curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .124 Beta function. . . . . . . . . . . . . . . . . . . . . . . . . .123
y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
z . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .126 C
( x, y) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 capping surface . . . . . . . . . . . . . . . . . . . . . . . 175
( x, y, z) Cauchy-Schwarz Inequality . . . . . . . . . . . . 17
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
( u, v, w) center of mass . . . . . . . . . . . . . . . . . . . . . . . . . 124
f centroid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
x Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . 60, 147
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
S
change of variable . . . . . . . . . . . . . . . . 117, 119
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 circulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . .174
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 closed curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

RR closed surface . . . . . . . . . . . . . . . . . . . . . . . . . 161
C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136, 139 collinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
C 1 , C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
conical helix . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 177
conservative field . . . . . . . . . . . . . . . . . . . . . 148
2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
constrained critical point . . . . . . . . . . . . . . . 96
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 continuity. . . . . . . . . . . . . . . . . . . . . . . . . . .52, 69
H
C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
continuously differentiable . . . . . . . . . 59, 80
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
D v f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Cartesian . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
er , e , e z , e , e . . . . . . . . . . . . . . . . . . . . . . . 181 curvilinear . . . . . . . . . . . . . . . . . . . . . . . . . 47
d r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 cylindrical . . . . . . . . . . . . . . . . . . . . 47, 182
i, j, k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 ellipsoidal. . . . . . . . . . . . . . . . . . . . . . . . .164
left-handed . . . . . . . . . . . . . . . . . . . . . . . . . . 2
A polar . . . . . . . . . . . . . . . . . . . . . . . . . . 47, 121
acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 55 rectangular. . . . . . . . . . . . . . . . . . . . . . . . . .1
angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 right-handed . . . . . . . . . . . . . . . . . . . . . . . . 2
210
Index 211
spherical . . . . . . . . . . . . . . . . . . . . . . 47, 182 Euclidean space . . . . . . . . . . . . . . . . . . . . . . . . . 1

coplanar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 exact differential form . . . . . . 139, 154, 175
correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 expected value . . . . . . . . . . . . . . . . . . . . . . . . 132
covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 extreme point . . . . . . . . . . . . . . . . . . . . . . . . . . 83
critical point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
cross product . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 F
curl. . . . . . . . . . . . . . . . . . . . . . . . . .169, 178, 182 flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
cylinder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42 function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
continuous . . . . . . . . . . . . . . . . . . . . . . . . . 69
D scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 vector-valued . . . . . . . . . . . . . . . . . . . . . . 51
derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
directional . . . . . . . . . . . . . . . . . . . . . . . . . 78
G
Gaussian blur . . . . . . . . . . . . . . . . . . . . . . . . . . 70
mixed partial . . . . . . . . . . . . . . . . . . . . . . 73
global maximum . . . . . . . . . . . . . . . . . . . . . . . 83
partial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
global minimum . . . . . . . . . . . . . . . . . . . . . . . . 83
vector-valued function . . . . . . . . . . . . . 52
gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . 80, 182
determinant . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Greens identities . . . . . . . . . . . . . . . . . . . . . 186
differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Greens Theorem . . . . . . . . . . . . . . . . . . . . . . 150
differential form . . . . . . . . . . . . . . . . . . . . . . 139
directed curve . . . . . . . . . . . . . . . . . . . . . . . . . 144 H
direction angles . . . . . . . . . . . . . . . . . . . . . . . . 19 harmonic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
direction cosines . . . . . . . . . . . . . . . . . . . . . . . . 19 helicoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
directional derivative. . . . . . . . . . . . . . . . . . .78 helix . . . . . . . . . . . . . . . . . . . . . . . . . . . 51, 59, 167
distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 hyperbolic paraboloid . . . . . . . . . . . . . . . . . . 44
between points . . . . . . . . . . . . . . . . . . . 6, 7 hyperboloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
from point to line . . . . . . . . . . . . . . . . . . 33 one sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
point to plane . . . . . . . . . . . . . . 37, 41, 42 two sheets. . . . . . . . . . . . . . . . . . . . . . . . . .43
distribution function . . . . . . . . . . . . . . . . . . 129 hypersurface . . . . . . . . . . . . . . . . . . . . . . . . . . 110
joint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .131 hypervolume . . . . . . . . . . . . . . . . . . . . . . . . . . 110
normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
divergence . . . . . . . . . . . . . . . . . . 162, 177, 182 I
Divergence Theorem . . . . . . . . . . . . . . . . . . 162 improper integral . . . . . . . . . . . . . . . . . . . . . 108
dot product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
double integral . . . . . . . . . . . . . . . . . . . 102, 105 double . . . . . . . . . . . . . . . . . . . . . . . 102, 105
polar coordinates . . . . . . . . . . . . . . . . . 121 improper . . . . . . . . . . . . . . . . . . . . . . . . . . 108
doubly ruled surface . . . . . . . . . . . . . . . . . . . . 45 iterated . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
multiple . . . . . . . . . . . . . . . . . . . . . . . . . . 101
E surface . . . . . . . . . . . . . . . . . . . . . . . 156, 158
ellipsoid . . . . . . . . . . . . . . . . . . . . . . 43, 123, 164 triple. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .110
elliptic cone. . . . . . . . . . . . . . . . . . . . . . . . . . . . .45 irrotational . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
elliptic paraboloid . . . . . . . . . . . . . . . . . . . . . . 44 iterated integral . . . . . . . . . . . . . . . . . . . . . . 102
212 Index
J O
Jacobi identity. . . . . . . . . . . . . . . . . . . . . . . . . .30 orientable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Jacobian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .119 orthonormal vectors . . . . . . . . . . . . . . . . . . . . 64
joint distribution . . . . . . . . . . . . . . . . . . . . . . 131 outward normal . . . . . . . . . . . . . . . . . . . . . . . 160
L P
Lagrange multiplier . . . . . . . . . . . . . . . . . . . . 96 paraboloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
lamina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 elliptic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Laplacian . . . . . . . . . . . . . . . . . . . . . . . . 178, 182 hyperbolic . . . . . . . . . . . . . . . . . . . . . . 44, 84
level curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 of revolution . . . . . . . . . . . . . . . . . . . . . . . 44
limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 parallelepiped . . . . . . . . . . . . . . . . . . . . . . . . . . 24
vector-valued function . . . . . . . . . . . . . 52 volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 parameter . . . . . . . . . . . . . . . . . . . . . . . . . . 31, 60
intersection of planes . . . . . . . . . . . . . . 38 parametrization . . . . . . . . . . . . . . . . . . . . . . . . 60
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 partial derivative . . . . . . . . . . . . . . . . . . . . . . . 71
parametric representation . . . . . . . . . 31 partial differential equation . . . . . . . . . . . . 74
perpendicular . . . . . . . . . . . . . . . . . . . . . . 34 path independence . . . . . . . . . . 146, 154, 175
skew . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 piecewise smooth curve . . . . . . . . . . . . . . . 141
symmetric representation . . . . . . . . . 32 plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
through two points . . . . . . . . . . . . . . . . . 33
coordinate . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
vector representation . . . . . . . . . . . . . . 31
Euclidean . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
line integral . . . . . . . . . . . . . . . . . . . . . . 136, 139
in space . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
local maximum . . . . . . . . . . . . . . . . . . . . . . . . . 83
line of intersection . . . . . . . . . . . . . . . . . 38
local minimum . . . . . . . . . . . . . . . . . . . . . . . . . 83
normal form . . . . . . . . . . . . . . . . . . . . . . . 35
M normal vector . . . . . . . . . . . . . . . . . . . . . . 35
mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 point-normal form . . . . . . . . . . . . . . . . . 35
matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 tangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
mixed partial derivative . . . . . . . . . . . . . . . . 73 through three points . . . . . . . . . . . . . . . 36
Mbius strip . . . . . . . . . . . . . . . . . . . . . . . . . . 168 position vector . . . . . . . . . . . . . . . . . 54, 55, 139
moment . . . . . . . . . . . . . . . . . . . . . . . . . . 124, 126 potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Monte Carlo method . . . . . . . . . . . . . . . . . . 113 probability density function . . . . . . . . . . . 129
moving frame fields . . . . . . . . . . . . . . . . . . . . 62 projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
multiple integral . . . . . . . . . . . . . . . . . . . . . . 101
multiply connected . . . . . . . . . . . . . . . . . . . . 153 Q
quadric surface . . . . . . . . . . . . . . . . . . . . . . . . . 43
N
n-positive direction . . . . . . . . . . . . . . . . . . . 169 R
Newtons algorithm . . . . . . . . . . . . . . . . . . . . 89 random variable . . . . . . . . . . . . . . . . . . . . . . 128
normal derivative . . . . . . . . . . . . . . . . . . . . . 186 Riemann integral . . . . . . . . . . . . . . . . . . . . . 135
normal to a curve. . . . . . . . . . . . . . . . . . . . . . .81 right-hand rule . . . . . . . . . . . . . . . . . . . . 21, 192
normal vector field . . . . . . . . . . . . . . . . . . . . 168 ruled surface . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Index 213
S V
saddle point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
sample space . . . . . . . . . . . . . . . . . . . . . . . . . . 128 vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
combination . . . . . . . . . . . . . . . . . . . . . . . . 12 angle between. . . . . . . . . . . . . . . . . . . . . .15
scalar function . . . . . . . . . . . . . . . . . . . . . . . . . 53 basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
scalar triple product . . . . . . . . . . . . . . . . . . . . 25 components . . . . . . . . . . . . . . . . . . . . . . . . 13
Second Derivative Test . . . . . . . . . . . . . . . . . 84 direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . .3
second moment. . . . . . . . . . . . . . . . . . . . . . . .134 magnitude . . . . . . . . . . . . . . . . . . . . . . . . 3, 7
second-degree equation. . . . . . . . . . . . . . . . .43 normal . . . . . . . . . . . . . . . . . . . . . . . . 35, 160
simple closed curve . . . . . . . . . . . . . . . . . . . 145 normalized . . . . . . . . . . . . . . . . . . . . . . . . . 12
simply connected . . . . . . . . . . . . . . . . . 154, 175 parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
smooth function . . . . . . . . . . . . . . . . . . . . 59, 84 perpendicular . . . . . . . . . . . . . . . . . . 16, 17
solenoidal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 positive unit normal . . . . . . . . . . . . . . 169
span . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 principal normal N . . . . . . . . . . . . . . . . 64
sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 scalar multiplication . . . . . . . . . . . . . . . . 9
spherical spiral . . . . . . . . . . . . . . . . . . . . . . . . . 54 subtraction . . . . . . . . . . . . . . . . . . . . . . . . . 10
standard normal distribution . . . . . . . . . 130 tangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
steepest descent . . . . . . . . . . . . . . . . . . . . . . . . 95 translation . . . . . . . . . . . . . . . . . . . . . . . . 5, 9
stereographic projection . . . . . . . . . . . . . . . . 46 unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Stokes Theorem . . . . . . . . . . . . . . . . . 168, 169 unit binormal B . . . . . . . . . . . . . . . . . . . . 64
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 unit tangent T . . . . . . . . . . . . . . . . . . . . . 64
doubly ruled . . . . . . . . . . . . . . . . . . . . . . . 45 zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3, 4
orientable . . . . . . . . . . . . . . . . . . . . . . . . . 168 vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
ruled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
two-sided . . . . . . . . . . . . . . . . . . . . . . . . . 168 smooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
surface integral . . . . . . . . . . . . . . . . . . 156, 158 vector triple product. . . . . . . . . . . . . . . . . . . . 25
velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 55
T volume element . . . . . . . . . . . . . . . . . . . . . . . 110
tangent plane . . . . . . . . . . . . . . . . . . . . . . . . . . 75
torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 W
trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . 74
triangle inequality . . . . . . . . . . . . . . . . . . . . . 18 work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135, 166
triple integral . . . . . . . . . . . . . . . . . . . . . . . . . 110
Z
cylindrical coordinates . . . . . . . . . . . .122 zenith angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
spherical coordinates . . . . . . . . . . . . . 122
U
uniform density . . . . . . . . . . . . . . . . . . . . . . . 124
uniform distribution . . . . . . . . . . . . . . . . . . 129
uniformly distributed . . . . . . . . . . . . . . . . . 128
unit disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Vector Calculus - Corral PDF

Uploaded by

Copyright:

Available Formats

Vector Calculus - Corral PDF

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

Vector Calculus - Corral PDF

Uploaded by

Copyright:

Available Formats

What is the author's background and qualifications?

What is the author's background and qualifications?

What topics does this book cover?

What topics does this book cover?

1

Copyright 2008 Michael Corral.

January 2008 M ICHAEL C ORRAL

1 Vectors in Euclidean Space 1

2 Functions of Several Variables 65

3 Multiple Integrals 101

4 Line and Surface Integrals 135

4.4 Surface Integrals and the Divergence Theorem . . . . . . . . . . . . . . . . . . . . 156

Appendix A: Answers and Hints to Selected Exercises 189

Appendix C: 3D Graphing with Gnuplot 196

GNU Free Documentation License 201

Figure 1.1.1 Figure 1.1.2

The coordinate system shown in Figure 1.1.1 is known as a right-handed coordinate

Figure 1.1.3 Right-handed coordinate system

yolk without cracking the shell!

Figure 1.1.4 Vectors in different dimensions

is not really used. See A NTON and R ORRES.

Figure 1.1.6 Correspondence between points and vectors

Unless otherwise stated, when we refer to vectors as v = (a, b) in R2 or v = (a, b, c) in R3 ,

Recall the distance formula for points in the Euclidean plane:

For points P = ( x1 , y1 ), Q = ( x2 , y2 ) in R2 , the distance d between P and Q is:

By this formula, we have the following result:

For a vector v = (a, b) in R2 , the magnitude of v is:

To calculate the magnitude of vectors in R3 , we need a distance formula for points in

Theorem 1.1. The distance d between points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) in R3 is:

The proof will use the following result:

Theorem 1.2. For a vector v = (a, b, c) in R3 , the magnitude of v is:

Proof: There are four cases to consider: p p

Case 3: exactly one of a, b, c is 0. Without loss of generality, we assume that a = 0, b 6= 0

Example 1.3. Calculate the following:

5. Let v = (a, b, c) and w = (3a, 3 b, 3 c) be vectors in R3 . Show that k w k = 3 k v k.

1.2 Vector Algebra

Intuitively, adding w to v means tacking on w to the end of v (see Figure 1.2.2).

Figure 1.2.2 Adding vectors v and w

Figure 1.2.3 Subtracting vectors v and w

Figure 1.2.4 Geometric vector algebra

(b) Without loss of generality, we assume that y

Theorem 1.4. Let v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 , let k be a scalar. Then

The following theorem summarizes the basic laws of vector algebra.

Theorem 1.5. For any vectors u, v, w, and scalars k, l , we have

Proof: (a) We already presented a geometric proof of this in Figure 1.2.4(a).

Let u = ( u 1 , u 2 , u 3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 . Then

(c) We already discussed this on p.10.

k( l v) = k( lv1 , lv2 , lv3 ) by Theorem 1.4(a)

(f) and (g): Left as exercises for the reader. QED

Figure 1.2.7 Basis vectors in different dimensions

When a vector v = (a, b, c) is written as v = a i + b j + c k, we say that v is in component

v = v1 i + v2 j + v3 k, k a scalar = kv = kv1 i + kv2 j + kv3 k

Example 1.4. Let v = (2, 1, 1) and w = (3, 4, 2) in R3 .

(b) Find 3v + 2w.

(c) Write v and w in component form.

(d) Find the vector u such that u + v = w.

(e) Find the vector u such that u + v + w = 0.

(f) Find the vector u such that 2u + i 2 j = k.

Figure 1.2.8 Proof of Theorem 1.2: d = k w v k