KCL Year1 Calculus Lecture Notes

Download as pdf or txt
Download as pdf or txt
You are on page 1of 136

4CCM111A Calculus I Compact Lecture Notes

Lecturer: Dr. Paul P. Cook ([email protected])


Room S4.09, Department of Mathematics, Kings College London

Based on lecture notes written by


Professor A. C. C. Coolen.
Department of Mathematics, Kings College London

Version of August 2014

1 Introduction
1.1 A bit of history ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1 Birth of modern science and of calculus
Stage I, 15001630: from speculation to science ... . . . . . . . . . . . .
1.1.2 Birth of modern science and of calculus
Stage II, 16301680: science is written in the language of mathematics!
1.1.3 Birth of modern science and of calculus
Stage III, around 1680: how to speak the language of mathematics! . .
1.2 Style of the course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Revision of some elementary mathematics . . . . . . . . . . . . . . . . . . . .
1.3.1 Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.2 Powers of real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.3 Solving quadratic equations . . . . . . . . . . . . . . . . . . . . . . . .
1.3.4 Functions, inverse functions, and graphs . . . . . . . . . . . . . . . . .
1.3.5 Exponential function, logarithm, laws for logarithms . . . . . . . . . .
1.3.6 Trigonometric functions . . . . . . . . . . . . . . . . . . . . . . . . . .

5
5

.
.
.
.
.
.
.
.
.

9
12
13
13
15
16
16
18
20

2 Proof by induction

22

3 Complex numbers
3.1 Introduction and definition . . . . . . . . . . . . . . . . . .
3.2 Elementary properties of complex numbers . . . . . . . . .
3.3 Absolute value and division . . . . . . . . . . . . . . . . .
3.4 The complex plane (Argand diagram) . . . . . . . . . . . .
3.4.1 Complex numbers as points in a plane . . . . . . .
3.4.2 Polar coordinates . . . . . . . . . . . . . . . . . . .
3.4.3 The exponential form of numbers on the unit circle
3.5 Complex numbers in exponential notation . . . . . . . . .
3.5.1 Definition and general properties . . . . . . . . . .
3.5.2 Multiplication and division in exponential notation
3.5.3 The argument of a complex number . . . . . . . . .
3.6 De Moivres Theorem . . . . . . . . . . . . . . . . . . . . .
3.6.1 Statement and proof . . . . . . . . . . . . . . . . .
3.6.2 Applications . . . . . . . . . . . . . . . . . . . . . .
3.7 Complex equations . . . . . . . . . . . . . . . . . . . . . .

26
26
27
28
29
29
30
32
34
34
35
36
38
38
39
40

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

3
4 Trigonometric and hyperbolic functions
4.1 Definitions of trigonometric functions . . . . . . . . . . . . . .
4.1.1 Definition of sine and cosine . . . . . . . . . . . . . . .
4.1.2 Elementary values . . . . . . . . . . . . . . . . . . . .
4.1.3 Related functions . . . . . . . . . . . . . . . . . . . . .
4.1.4 Inverse trigonometric functions . . . . . . . . . . . . .
4.2 Elementary properties of trigonometric functions . . . . . . . .
4.2.1 Symmetry properties . . . . . . . . . . . . . . . . . . .
4.2.2 Addition formulae . . . . . . . . . . . . . . . . . . . . .
4.2.3 Applications of addition formulae . . . . . . . . . . . .
4.2.4 The tan(/2) formulae . . . . . . . . . . . . . . . . . .
4.3 Definitions of hyperbolic functions . . . . . . . . . . . . . . . .
4.3.1 Definition of hyperbolic sine and hyperbolic cosine . . .
4.3.2 General properties and special values . . . . . . . . . .
4.3.3 Connection with trigonometric functions . . . . . . . .
4.3.4 Applications of connection with trigonometric functions
4.3.5 Inverse hyperbolic functions . . . . . . . . . . . . . . .
5 Functions, limits and differentiation
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
5.1.1 Rate of change, tangent of a curve . . . . . . .
5.1.2 Finding tangents and velocities why we need
5.2 The limit . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 Left and right limits . . . . . . . . . . . . . .
5.2.2 Asymptotics - limits involving infinity . . . . .
5.2.3 When left/right limits exists and are identical
5.2.4 Rules for limits of composite expressions . . .
5.2.5 Examples . . . . . . . . . . . . . . . . . . . .
5.3 Differentiation . . . . . . . . . . . . . . . . . . . . . .
5.3.1 Derivatives of functions . . . . . . . . . . . . .
5.3.2 Rules for derivatives of composite expressions
5.3.3 Derivatives of implicit functions . . . . . . . .
5.3.4 Applications of derivative: sketching graphs .
6 Integration
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . .
6.1.1 Area under a curve . . . . . . . . . . . . . . .
6.1.2 Examples of integrals calculated via staircases
6.1.3 Fundamental theorems of calculus: integration

. . . .
. . . .
limits
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
vs differentiation

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

42
42
42
45
45
46
49
49
51
52
54
55
55
56
58
58
59

.
.
.
.
.
.
.
.
.
.
.
.
.
.

63
63
63
64
67
67
68
69
70
71
73
73
74
77
80

.
.
.
.

81
81
81
84
89

6.2

6.3

6.1.4 Indefinite and definite integrals, and other conventions . . . .


Techniques of integration . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.1 List of elementary integrals and general methods for reduction
6.2.2 Examples: integration by substitution . . . . . . . . . . . . . .
6.2.3 Examples: integration by parts . . . . . . . . . . . . . . . . .
6.2.4 Further tricks: recursion formulae . . . . . . . . . . . . . . . .
6.2.5 Further tricks: differentiation with respect to a parameter . .
6.2.6 Further tricks: partial fractions . . . . . . . . . . . . . . . . .
Some simple applications . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.1 Calculation of surface areas . . . . . . . . . . . . . . . . . . .
6.3.2 Calculation of volumes of revolution . . . . . . . . . . . . . . .
6.3.3 Calculation of the length of curves . . . . . . . . . . . . . . .

7 Taylors theorem and series


7.1 Introduction to series and questions of convergence . . .
7.1.1 Series notation and elementary properties . . .
7.1.2 Series convergence criteria . . . . . . . . . . . .
7.1.3 Power series notation and elementary properties
7.2 Taylors theorem . . . . . . . . . . . . . . . . . . . . . .
7.2.1 Expression for the coefficients of power series . . .
7.2.2 Taylor series around x = 0 . . . . . . . . . . . . .
7.2.3 Taylor series around x = a . . . . . . . . . . . . .
7.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.1 Series expansions for standard functions . . . . .
7.3.2 Indirect methods for finding Taylor series . . . . .
7.4 LHopitals rule . . . . . . . . . . . . . . . . . . . . . . .
8 Exercises

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

91
92
93
95
97
99
101
103
107
107
108
110

.
.
.
.
.
.
.
.
.
.
.
.

113
. 113
. 113
. 114
. 115
. 118
. 118
. 120
. 121
. 122
. 122
. 123
. 125
126

5
1. Introduction
The motivation for calculus and analysis begin with the study of the cosmos, so we commence
with
1.1. A bit of history ...
1.1.1. Birth of modern science and of calculus
Stage I, 15001630: from speculation to science ...

Ptolemy of Alexandria , 2nd century AD:


Style of the ancient Greeks: no experiments,
just logical thought and elegance
published Almagest (summary of astronomy,
based on 500 years of Greek astronomical and
cosmological thinking)
earth is centre of the universe
complicated model of spheres carrying
heavenly bodies, moving themselves in circles

Nicolaus Copernicus , 14731543:


problems with the motion of the moon ...
published De Revolutionibus, sun-centred
universe, with moon orbiting around the earth
Catholic Church:
put De Revolutionibus on the
Index of banned books
(stayed on the Index until 1835!)

6
Tycho Brahe , 15461601:

The genius observer...


First systematic and comprehensive measurement
of the trajectories of the moon, the planets,
the comets, and the stars,
over many years and with unrivaled precision!
Compiled huge amounts of data
Did not himself believe Copernicus ideas ...
(lost his nose while a student in a duel in 1566)

Johannes Kepler , 15711630:


The genius in analyzing data ...
Believed Copernicus, but could not observe
anything himself (poor eyesight ...)
Developed further models of sun-centered
universe, with spheres within spheres
Became Brahes assistant in 1599,
discovered quantitative laws,
based on Tycho Brahes data
published Astronomia Nova in 1609,
Harmonice Mundi in 1619,
Epitome of Copernican Astronomy
(3 volumes) 16181621

7
Keplers First Law (1605):
the orbit of each planet is an ellipse, with the sun at one of the two foci

Keplers Second Law (1602):


a line joining the sun to an orbiting planet sweeps out equal areas in equal times

Keplers Third Law (1618):


the square of a planets orbit period is proportional to cube of its distance to the sun

8
1.1.2. Birth of modern science and of calculus
Stage II, 16301680: science is written in the language of mathematics!

Galileo Galilei , 15641642:


the wrangler, loved arguments ...
the first real scientist:
(i) state a hypothesis,
(ii) devise an experiment to test it,
(iii) carry out the experiment,
(iv) accept or reject the hypothesis
always worried about money (sisters dowries ...)
worked on inventions to get rich
(thermometer, calculator)
interested in movement of objects
constructed improved telescope in 1609: new observations all supported Copernicus ...
published Dialogue on the Two Chief World Systems in 1632
(Salviati vs Simplicio, with Sagredo as impartial commentator)
it was suggested that Pope Urban VIII was the simpleton ...
1633: show trial by the Inquisition, Galileo (69 and fearing torture):
I abjure, curse and detest my errors
published Discourses and Mathematical Demonstrations Concerning Two New Sciences
(first modern scientific textbook), smuggled out of Italy, published in 1638

Ren
e Descartes , 15961650:
1637: Discours de la Methode pour bien conduire
la raison et chercher la Verite dans les Sciences
invented Cartesian coordinates:
each position in space represented by three numbers
introduced letters x, y, z to denote
unknown quantities in mathematical problems
published Principia Philosophiae (1644)

9
1.1.3. Birth of modern science and of calculus
Stage III, around 1680: how to speak the language of mathematics!
(i) state a hypothesis,
(ii) devise an experiment to test it,
(iii) carry out the experiment,
(iv) accept or reject the hypothesis
The problem in making it work in practice:
To test hypotheses on forces and movements of objects, one needs to be
able to calculate the trajectories that would be caused by the assumed forces ...
1673
Christiaan Huygens: outward force on
object in circular orbit of radius R
is proportional to R2
1674
Robert Hooke: object that feels no force
will move along a straight line
(Newtons first law of motion ...)

Huygens

Hooke

Halley

Wren

1684 at the Royal Society ...


Edmond Halley, Christopher Wren, Robert Hooke
hypothesis: the sun attracts planets
at distance R with a force proportional to R2
Is it possible to derive the observed motion
of the planets from this inverse square law?
1684 somewhat later ... Halley visits Isaac Newton
according to Newtons friend De Moivre:
Dr Halley came to visit him in Cambridge, after they had been some time together the
Dr asked him what he thought the Curve would be that would be described by the Planets
supposing the force of attraction towards the Sun to be reciprocal to the square of their
distance from it. Sir Isaac replied immediately it would be an Ellipsis, the Dr struck with
joy & amazement asked him how he knew it, why saith he, I have calculated it, whereupon
Dr Halley asked him for the calculation without any further delay, Sr Isaac looked among
his papers, but could not find it, but he promised him to renew it, & then send it him.

10
The two parents of Calculus:

Isaac Newton , 16421727:


developed mechanics, calculus, theory of light
before the age of 30 ...
then spent 20 years of his life on alchemy ...
1687: publishes
Philosophiae Naturalis Principia Mathematica
1704: published Opticks
brilliant but obsessive and nasty piece of work ...
great re-writer of history (in his own favour ...)
e.g. Hooke
(no references in Principia or Opticks!
... by standing on the shoulders of Giants ...
move of the Royal Society and a missing portrait)
or Leibniz
(independent commission of the Royal Society)

Gottfried Leibniz , 16461716:


invented calculus independently of Newton
(although slightly later)
Leibniz notation more transparent
it is in fact what we use today!

11
Newton and his successors established the principles and the mode of work
for all quantitative sciences (physics, biology, economics, etc):
science: no longer descriptive, but aimed at finding the (usually mathematical) laws
underlying the observed phenomena
ones degree of understanding of an area of science is measured by the extent to which one
can predict new phenomena from the discovered laws
Galileos principles define the procedure for finding the underlying laws.
They now (i.e. after Newton and Leibniz) take the form:
(i) state a hypothesis,
(ii) devise an experiment to test it,
(iii) calculate the predicted outcome of the experiment from the hypothesis
(iv) carry out the experiment,
(v) accept or reject the hypothesis
When there are several distinct hypotheses, that are all consistent with the available data:
select the simplest hypothesis (Occams Razor)

Side effects of the scientific revolution:


industrial revolution
mechanistic view of the universe: nature is governed by differential equations
(i) solution depends only on initial conditions
(ii) no free will
(iii) no divine intervention required to keep the world going ...
of course that all changed around 1920 ...

12
1.2. Style of the course
Description of contents:
Calculus: the related areas of differentiation, integration, sequences and series, that are
united in their reliance on the idea of limits.
Additional topics: complex numbers and trigonometric functions. This simplifies
the discussion of the classical functions of calculus (i.e. trigonometric, hyperbolic,
exponential, logarithmic) and their relations.
Relation between calculus and analysis:
Calculus:
intuitive and operational ideas, no emphasis on strict step-by-step logical derivation
e.g. derivative as limit of a ratio, integral as limit of a sum
initially (Newton, Leibniz) without rigorous definition of limit.
Analysis:
logical, rigorous proofs of the intuitive ideas of calculus.
stage 1 (calculus): find a method to crack the problem
stage 2 (analysis): determine carefully why and when the method works
(this order of developing maths continues today: see e.g. path integrals in particle physics)
Rationale behind this division of work:
Hard to prove a theorem without being already familiar with the unproven (but strongly
believed in) result.
Hard to understand the need for the rigorous style of analysis until one has sufficient
experience with calculus to realize the need to prove theorems, and to appreciate the
beauty and elegance of such a logical formal treatment.
Too much initial attention to the details of proofs while learning a subject often conceals
the relative simplicity of the result.
Variation in approaches to calculus:
there are different but mathematically equivalent routes via which to develop calculus,
e.g. alternative definitions of trigonometric functions:
(i) introduce series,
(ii) define sin and cos as power series
(iii) proof of the properties of sin cos through study of the series
since all are equivalent, we will jump between definitions, dependent on the problem
you should, however, never be satisfied by results without any derivation; we will try to
give as many (full or partial) proofs as feasible within the time constraints of the course

13
1.3. Revision of some elementary mathematics
1.3.1. Numbers
First we start with some terminology and definitions:

definition :

set :

nonordered collection of objects (elements)

e.g. : S = {a, b, c, d}
no ordering : {a, b, c, d} = {b, a, d, c, } etc.
set membership : a S

(a belongs to the set S)

definition :

={}

definition :

IN = {1, 2, 3, . . .}

definition :

ZZ = {. . . , 3, 2, 1, 0, 1, 2, 3, . . .}

definition :

ZZ+ = {1, 2, 3, . . .} = IN

(positive integers)

definition :

ZZ = {1, 2, 3, . . .}

(negative integers, China 1500 BC)

definition :

Q
| = set of all numbers of the form p/q with p, q ZZ
1
27
27
1
Q
| ,
/ ZZ

Q
| ,
ZZ
e.g.
2
2
3
3

definition :

IR = set of all rational and irrational numbers (real numbers)

e.g. IR,
/Q
|
2 IR, 2
/Q
|
64 IR, 64 IN

definition :

subsets of sets : A B if and only if every a A obeys a B


note :

(the empty set)


(natural numbers)
(integer numbers)

(rational numbers)

IN ZZ Q
| IR

Logical symbols: (AND), (OR)


S1 S2 : both statements S1 and S2 are true
S1 S2 : either S1 is true, or S2 is true, or both are true
Logical consequences: S1 S2 means if statement S1 is true then also statement S2 is true

14
Every element x IR can be represented by a point on the number line:

3/2

-4

-3

-2

-1

The set IR is an ordered set. Let x, y IR, then the ordering symbols are defined as:
x < y : x is smaller than y ,

i.e. x to the left of y on the number line

x > y : x is larger than y ,

i.e. x to the right of y on the number line

x y : x is smaller than or equal to y


x y : x is larger than or equal to y
Note (should be obvious):
x < y x > y x, y

(i.e. no such x, y IR exist)

x < y x y x, y

(i.e. no such x, y IR exist)

x > y x y x, y

(i.e. no such x, y IR exist)

xy xy x=y

Interval: segment of the number line


Closed interval: line segment that includes both end points
[a, b] = {x IR | a x b}
Open interval: line segment that does not include end points
(a, b) = {x IR | a < x < b}
Semi-open (or semi-closed) interval, exactly one endpoint is included
[a, b) = {x IR | a x < b}

or

(a, b] = {x IR | a < x b}

Unions and intersections:


union of A and B :

A B = {x | x A x B}

intersection of A and B :

A B = {x | x A x B}

Example:
(5, 4) [2, 5] = (5, 5]

(5, 4) [2, 5] = [2, 4)

15
1.3.2. Powers of real numbers
In an expression of the form xn we call x the base and n the power.
First define natural powers of real numbers. Let n IN and x IR, x 6= 0:
definition x0 = 1
definition xn = x.x. . . . .x

(nfold product, for n > 0)

Generalize to integer powers, by giving a definition for negative powers. Let n IN, n > 0:
1
(nfold product in denominator)
definition xn =
x.x. . . . .x

Three basic laws of manipulation, let x IR and n, m ZZ:


first law :

xm .xn = xn+m

second law :

xm /xn = xmn

third law :

(xm )n = xmn

(i) prove these laws from the definitions, by checking the different case for the signs of m and n
(ii) note that (ii) can be derived from (i) and (iii)
Generalize to fractional powers of positive real numbers x IR+ , by giving a definition for
powers of the form 1/n with n IN, n > 0:

definition x1/n = n x (nth root)

Here n x is the number y IR+ with the property that y n = x.


Verify that the above laws of manipulation still hold in the case of fractional powers.
For example, let m, n, ` IN+ :

a`/n = ( n a)`
aq+`/n = aq .a`/n = aq .( n a)`
Generalize to real powers of positive real numbers x IR+ . Each real number y IR can
be approximated to arbitrary accuracy by fractions n/m, where m IN+ , n ZZ. One then
defines xy similarly by substituting for y this fraction approximation. Let m IN and n ZZ:
n
is the best approximation of y by a fraction with denominator m
if
m

then xn/m = ( m x)n is our associated approximation of xy


Negative real powers are again defined via xy = 1/xy , and our laws of manipulation still hold!

16
1.3.3. Solving quadratic equations
Quadratic equations are equations of the following form (or can be reduced to this form), where
x is the unknown quantity to be determined, and a, b, c IR (the coefficients) are given:
ax2 + bx + c = 0
Assume a 6= 0, otherwise equation reduces to a linear one.
Note: solutions x IR do not always exist!
Methods for solution:
solution by factorization:

find d, f, g, h IR such that

ax2 + bx + c = (dx + f )(gx + h)


(not always possible!)
New problem involves two linear expressions: (dx + f )(gx + h) = 0.
Solutions: x = f /d and x = h/g
solution by completing a square:

find d, f IR such that

ax2 + bx + c = a[(x + d)2 f ]


(always possible!)

New problem is solved using square root: (x + d)2 = f , so x + d = f , so x = d f .


No solution x IR exists if f < 0. This leads to a general solution in the form of the
familiar formula:

b b2 4ac
x=
2a
Solutions x IR exist only if b2 4ac.
Try both methods on the following quadratic equations:
x2 + 3x 10 = 0

6x2 + 5x = 4

x2 8x = 0

x2 = 4x 5

1.3.4. Functions, inverse functions, and graphs


A function f is a rule (or map) that assigns a unique output number f (x) to each input
number x. The set of input values x for which the function is defined is called the domain of f .
The full set of output values that the function can generate when we choose values of x from
its domain is called the range of f . Domain D and range R are indicated in our notation via
f : D R.

17
Example 1:
let the description of a function f be: take any input x IR, add 2 to this input number
We write
f : IR IR

f (x) = x + 2

Example 2:
let the description of a function g be: take any input x [1, 1] and square it, then subtract
7
We write
g : [1, 1] [7, 6]

g(x) = x2 7

Note:
a function is allowed to take different forms on different intervals, e.g.
Example 3:
f : [0, ) [0, 12] (14, 16) {9}

3x
f (x) = 2x + 6

for x [0, 4]
for x (4, 5)
for x 5

definition:
The inverse f 1 of a function f : D R is defined by the following:
f 1 : R D

f 1 (f (x)) = x for all x D

In words: f 1 restores the original number x after the action of the function f .
Claim:
f 1 can not not exist if there exist two different numbers x1 , x2 D with f (x1 ) = f (x2 )
Intuitively: how would f 1 select from the two candidates x1 , x2 which one to restore?
Formal proof: call f (x1 ) = y and substitute x1 and x2 into the above definition of f 1 , one
then finds the simultaneous requirements
f 1 (f (x1 )) = x1 f 1 (y) = x1
f 1 (f (x2 )) = x2 f 1 (y) = x2
The assumption of the existence of f 1 would thus lead to x1 = x2 , in contradiction with the
starting point x1 6= x2 . Hence f 1 cannot exist.
definition:
A function f : D R is invertible if and only if f (x1 ) 6= f (x2 ) for any two values x1 , x2 D
with x1 6= x2

18
How to find the inverse of a given function f ?
(i) write y = f (x)
(ii) rearrange this formula to find x expressed as a (new) function of y i.e. x = g(y))
(iii) then y = f (x) = f (g(y)) so g(y) = f 1 (y).
Work out the inverse functions for
f (x) = x + 4 for x IR,
f (x) = 6x + 4 for x IR and

f (x) = x for x IR+ .


1.3.5. Exponential function, logarithm, laws for logarithms
definition:
the exponential function is f (x) = ex , where e is a special irrational number
(exponential growth)
Properties:
(i) ex > 0 for all x IR
(ii) ex increases monotonically
(iii) from the value 0 as x , via e0 = 1 at x = 0, to unbounded growth as x
Define n! = n(n 1)(n 2) . . . .3.2.1 (n factorial)
Equivalent expressions for e = 2.71828182 . . . (more about this later):
1
1
1
1
1
e = 1 + + + + + ...
e = n
lim (1 + )n
1! 2! 3! 4!
n
x
x
Similarly, exponential decay is described by f (x) = e = 1/e
(same shape of curve, just exchange x x)
Let a IR+ :
definition:
the logarithmic function to the base a, written as loga (x), is the inverse of the function f (x) = ax
in words: loga (y) gives the power to which I must raise a to get y
definition:
the natural logarithmic is defined as the logarithmic function to the base e, i.e. ln(x) = loge (x)
in words: ln(y) gives the power to which I must raise e to get y
Properties (direct consequences of the concept of inverse):
aloga (y) = y

loga (ax ) = x

eln(y) = y

ln(ex ) = x

19
Manipulation identities for logarithms:
switching base :

loga (x) =

logb (x)
logb (a)

(1)

products, fractions :

loga (x.y) = loga (x) + loga (y)

(2)

loga (x/y) = loga (x) loga (y)

(3)

loga (xy ) = y loga (x)

(4)

powers :

Proof of (1):
Strategy: we prove the equivalent statement loga (x). logb (a) = logb (x)
using the manipulation identities for powers
Show that the left-hand side (LHS) of the latter equation has the property bLHS = x:
bLHS = bloga (x). logb (a) = (blogb (a) )loga (x) = aloga (x) = x
By the definition of logb (x) this implies that LHS= logb (x), which is exactly the right-hand
side. This completes the proof.
Proof of (2):
Strategy: we use the manipulation identities for powers
We show that the right-hand side (RHS) of (2) has the property aRHS = xy:
aRHS = aloga (x)+loga (y) = aloga (x) .aloga (y) = x.y
By the definition of loga (x) this implies that RHS= loga (xy), which is exactly the left-hand
side of (2). This completes the proof.
Proof of (3):
Strategy: we use the manipulation identities for powers
We show that the right-hand side (RHS) of (3) has the property aRHS = x/y:
aRHS = aloga (x)loga (y) = aloga (x) .a loga (y) = x/y
By the definition of loga (x) this implies that RHS= loga (x/y), which is exactly the left-hand
side of (3). This completes the proof.
Proof of (4):
Strategy: we use the various manipulation identities for powers
We show that the right-hand side (RHS) of (4) has the property aRHS = xy :
aRHS = ay loga (x) = (aloga (x) )y = xy
By the definition of loga (x) this implies that RHS= loga (xy ), which is exactly the left-hand
side of (4). This completes the proof.

20
5

tan(x)

ex

sin(x)

3
1

ln(x)

cos(x)

x/

-1
-2
-1
-3
-4
-5

-2
-5

-4

-3

-2

-1

-2

-1

1.3.6. Trigonometric functions


Many equivalent definitions possible (more follows in this course).
definition:
Consider rotations around the origin. 1 radian is the magnitude of a rotation angle such that
it cuts a segment of the unit circle of length 1.
Consequence: going round once implies an angle of 2, i.e. 2 radians = 3600
(since circumference of a radius-R circle equals 2R)
definition:
Consider a half-line with its one end-point in the origin. Choose it initially to lie along the
positive x-axis, and then rotate it anti-clockwise around the origin; call the rotation angle
(measured in radians). Find the coordinates (X, Y ) of the point where the half-line intersects
the unit circle: now call cos() = X, sin() = Y .
definition: tan() = sin()/ cos()
Consequences:
(i) cos2 (x) + sin2 (x) = 1 for all x IR
(definition of the unit circle!)
(ii) cos(x) and sin(x) are periodic, with period 2
(since 2 rotation gives a complete turn)

21
(iii) special values of sin(x) and cos(x) follow immediately,
e.g. for x = 0, 4 , 2 , 3
,...
4
(iv) more general than definition in terms of ratios of sides in triangles
(the latter do come out for x [0, 2 ], here we have a definition for any value of x)
(v) zero points:
sin(x) = 0 for x = n with n ZZ, cos(x) = 0 for x =

(vi) tan(x) = 0 for x = n with n ZZ, doesnt exist for x =

+ n with n ZZ

+ n with n ZZ

22
2. Proof by induction
Induction is a method that allows us to prove an infinite number of statements, by proving just
two statements: the basis step and the induction step.
The basic idea is that if, by assumming a statement Sn (where n IN) is true, we can
show that it follows that statement Sn+1 is also true AND if we can prove that a single (simple)
statement S1 is true then we have proved that all statements Sn for n > 0 are true. The two
distinct parts of a proof by induction are
(i) that a statement is true for n = 1 (the basis stape) and
(ii) that if Sn is true then Sn+1 is true for all n (the induction step)
N.B. the two parts of the proof can be shown in any order.
N.B.2 if we had not been able to prove S1 is true in the basis step, but we had been able to
prove S2 is true, then we would have a proof by induction of the statements {S2 , S3 , S4 , . . .}.
definition: the summation symbol
n
X

(n, m ZZ, n m)

ak = am + am+1 + am+2 + . . . + an1 + an

k=m

definition: binomial coefficients


n
k

n!
k!(n k)!

with the convention

0! = 1

meaning:
the number of distinct ways to select k elements from a set of n elements
(where permutations are not counted separately)
Example 1:
Let n IN. Use induction to prove that
n
X

1
k = n(n + 1)
2
k=1
Proof:
We subtract the two sides and define An = nk=1 k 21 n(n + 1).
We must now prove that An = 0 for all n IN
P

(i) the basis:


P
for n = 1 one has A1 = 1k=1 k 12 1.2 = 1 1 = 0 (so claim is true for n = 1)
(ii) induction step:

23
now suppose that An = 0 for some n IN, i.e.
It follows that
n+1
X
1
An+1 =
k (n + 1)(n + 2)
2
k=1

Pn

k=1

k = 12 n(n + 1)

n
X

1
(next use An = 0 !)
k + (n + 1) (n + 1)(n + 2)
2
k=1
1
1
= n(n + 1) + (n + 1) (n + 1)(n + 2)
2
2
1
1
= (n + 1)[ n + 1 (n + 2)] = 0
2
2
We have shown: if An = 0 then also An+1 = 0. Knowing already that A1 = 0 (the basis), this
completes the proof that An = 0 for all n IN.
=

Example 2:
use induction to prove Newtons binomial formula for all integer n 0:
n

(a + b) =

n
X
k=0

n
k

ank bk

Proof:
We subtract the two sides and define An = (a + b)n
We must now prove that An = 0 for all integer n 0

Pn

n nk k
b .
k=0 ( k )a

(i) the basis:


P
for n = 0 one has A0 = (a + b)0 0k=0 ( 0k )ak bk = 1 ( 00 ) = 0, so claim is true for n = 0
(ii) induction step:
P
now suppose that An = 0 for some n IN, i.e. (a + b)n = nk=0 ( nk )ank bk
It follows that
An+1 = (a + b)n+1

n+1
X

( n+1
)an+1k bk
k

k=0

= (a + b)(a + b)n

n+1
X

( n+1
)an+1k bk
k

(next use An = 0)

k=0

=
=

( n
X

n+1
X
(a + b)
( nk )ank bk
( n+1
)an+1k bk
k
k=0
k=0
n
n
n+1
X
X
X
( nk )an+1k bk +
( nk )ank bk+1
( n+1
)an+1k bk
k
k=0
k=0
k=0

In the middle term we substitute k = ` 1, so ` = 1, . . . , n + 1. Then we separate terms with


b0 and with bn+1 from the rest. This gives
An+1 =

n
X

( nk )an+1k bk +

k=0

n+1
X

n+1
X

`=1

k=0

( n`1 )an+1` b`

( n+1
)an+1k bk
k

24
n+1n1 n+1
= ( n0 )an+10 b0 ( 0n+1 )an+10 b0 + ( nn+11 )an+1n1 bn+1 ( n+1
b
n+1 )a

n
X

an+1k bk ( nk ) + ( nk1 ) ( n+1


)
k

k=1

n
X

an+1k bk ( nk ) + ( nk1 ) ( n+1


)
k

k=1

Finally we must work out the combinatorial terms, for k 1, . . . , n}:


n!
n!
(n + 1)!
+

(n k)!k! (n k + 1)!(k 1)! (n + 1 k)!k!


"
#
1
1
n+1
n!
+

=
(k 1)!(n k)! k n k + 1 k(n + 1 k)
"
#
n!
(n + 1 k) + k (n + 1)
=
=0
(k 1)!(n k)!
k(n + 1 k)

( nk ) + ( nk1 ) ( n+1
)=
k

Insertion into our previous intermediate result gives An+1 = 0.


Thus we have shown: if An = 0 then also An+1 = 0. Knowing already that A0 = 0 (the basis),
this completes the proof that An = 0 for all integer n 0.
Example 3:
Let x IR, x 6= 1 and n IN. Use induction to prove the following (geometric series)
n1
X

xk =

k=0

1 xn
1x

Proof:
n1 k
x
We subtract the two sides and define An = k=0
We must now prove that An = 0 for all integer n IN

(i) the basis:


P
for n = 1 one has A1 = 0k=0 xk

1x
1x

1xn
.
1x

= 1 1 = 0, so claim is true for n = 1

(ii) induction step:


P
k
now suppose that An = 0 for some n IN, i.e. n1
k=0 x =
An+1 =

n
X
k=0

xk

1xn
.
1x

It follows that

X
1 xn+1 n1
1 xn+1
=
xk + xn
1x
1x
k=0

(next use An = 0)

1 xn
1 xn+1
n
=
+x
=
1x
1x
1 xn + (1 x)xn (1 xn+1 )
=
=0
1x
Thus we have shown: if An = 0 then also An+1 = 0. Knowing already that A1 = 0 (the basis),
this completes the proof that An = 0 for all n IN.
Exercises:

25
Similarly, prove the following statements for n IN by induction:
n
X

1
k = n(n + 1)(2n + 1)
6
k=1
2

n
X

1
k 3 = n2 (n + 1)2
4
k=1

26
3. Complex numbers
3.1. Introduction and definition
definition:
The number i is defined as a solution of the equation z 2 + 1 = 0
(note: this eqn had no solutions z IR)
definition:
The set C
| of complex numbers consists of all expressions of the form a + ib with a, b IR,
C
| = {a + ib | a, b IR}
definition:
Addition and multiplication of numbers in C
| is defined as follows. Let a, b, c, d IR:
(a + bi) + (c + di) = (a + c) + (b + d)i
(a + bi)(c + di)

= (ac bd) + (ad + bc)i

(i.e. calculate as if with real numbers, and put i2 = 1)


note:
i2 = 1 can alternatively be taken as a consequence of the multiplication definition
definition: let a, b IR
The real part of z = a + ib is defined as Re(z) = a
The imaginary part of z = a + ib (where a, b IR) is defined as Im(z) = b
definition: let a, b IR
The complex conjugate of a complex number z = a + ib is defined as z = a ib
(i.e. obtained from z by replacing i by i)
z and z are called a complex conjugate pair
Notes:
(i) also z = i is a solution of z 2 + 1 = 0
proof: z 2 + 1 = (i)2 + 1 = (1)2 i2 + 1 = i2 + 1 = 0

(ii) sometimes i is written as i = 1


(iii) sometimes z is written as z ?
(iv) unlike IR, it is impossible to order the numbers of C
| in terms of larger and smaller
(just postulate i < 0 or i > 0 and see what happens ...)

27
3.2. Elementary properties of complex numbers
Every quadratic equation az 2 + bz + c = 0 can be solved in C
| , giving the solutions

b b2 4ac
z =
2a


2
(where for b 4ac < 0 one has b2 4ac = 1 4ac b2 = i 4ac b2 )
proof:


a(z z+ )(z z ) = a z 2 z(z+ + z ) + z+ z

!
!
!
b + b2 4ac
b b2 4ac
b + b2 4ac b b2 4ac
2
+
+a
= az az
2a
2a
2a
2a





1
1
= az 2 z (2b) +
b + b2 4ac b b2 4ac
2
4a


1  2
b ( b2 4ac)2
= az 2 + bz +
4a

1  2
b b2 + 4ac = az 2 + bz + c
= az 2 + bz +
4a
It now follows immediately that az 2 + bz + c = 0 for z = z . This completes the proof.
Every n-th order polynomial with real-valued or complex coefficients,
P (z) = z n + an1 z n1 + . . . + a1 z + a0 , can be factorized into linear factors to give
P (z) = (z z1 )(z z2 ) . . . (z zn1 )(z zn )
with n complex numbers z1 . . . zn (the zeros or roots of the polynomial)
(the proof will not be given here)
For all z, w C
|:
For all z, w C
|:

z + w = z + w (proof in tutorials)
zw = z.w
(proof in tutorials)

For every z C
| : zz IR, with zz 0 and where zz = 0 if and only if z = 0
(proof in tutorials)
If z is a root of a polynomial P (z) with real coefficients, then also z will be a root
proof:
We know that P (z) = 0, i.e. z n + an1 z n1 + . . . + a1 z + a0 = 0.
Now, since z + w = z + w and zw = z.w, also
0 = z n + an1 z n1 + . . . + a1 z + a0
0 = z n + an1 z n1 + . . . + a1 z + a0
0 = z n + an1 z n1 + . . . + a1 z + a0
Thus P (z) = 0. This completes the proof.

28
3.3. Absolute value and division
definition:

The absolute value (or modulus) of a complex number is defined as |z| = z.z
Consequences:
If z = a + ib with a = Re(z) and b = Im(z), then |z| =
proof:
|z| =

z.z =

(a + ib)(a ib) =

a2 + b 2

a2 (ib)2 =

a2 + b 2

(z) = z
proof: let z = a + ib,
(z) = (a ib) = a + i(b) = a i(b) = a + ib = z
|Re(z)| |z| and |Im(z)| |z|
(proof in tutorials)
|z.w| = |z||w|, |z| = |z|
(proof in tutorials)
The above definition of |z| obeys the triangular inequality |z + w| |z| + |w|
proof:
|z + w|2 = (z + w)(z + w) = (z + w)(z + w)
= z.z + z.w + w.z + w.w
= |z|2 + 2Re(z.w) + |w|2
2

(due to z + z = 2Re(z))

|z| + 2|z.w| + |w|

(due to |Re(z)| |z|)

= |z|2 + 2|z||w| + |w|2

(due to |zw| = |z||w| and |w| = |w|)

= (|z| + |w|)2
Taking the square root of both sides then completes the proof.
The property |z| IR makes it easy to work out the ratio of complex numbers, and write it in
the standard form v = Re(v) + iIm(v). Method: multiply numerator and denominator by the
complex conjugate of the denominator.
Let z = a + ib and w = c + id, with a, b, c, d IR and w 6= 0:
a + ib c id
(a + ib)(c id)
a + ib
=
=
c + id
c + id c id
|c + id|2
ac + ibc iad + bd
ac + bd
bc ad
=
= 2
+i 2
2
2
2
c +d
c +d
c + d2
In particular: 1/i = i,
1/z = z/|z|2

29
3.4. The complex plane (Argand diagram)
3.4.1. Complex numbers as points in a plane
underlying idea:
there is a one-to-one correspondence between complex numbers z C
|
2
and points in the ordinary plane IR , namely
z = a + ib C
|

(a, b) IR2

where a is taken as the x-coordinate and b is taken as the y-coordinate.


one-to-one:
(i) with every z C
| corresponds exactly one point (a, b) in the plane
(ii) with every point (a, b) in the plane corresponds exactly one z C
|
examples:
(a, b) IR2 :

z C
| :
1 = 1 + 0.i

(1, 0)

i = 0 + 1.i

(0, 1)

1 = 1 + 0.i

(1, 0)

i = 0 1.i

(0, 1)

2 + 3i

(2, 3)

1 i 2
1 + 2i
3
3 i
2
2 + i

(1, 2)

(1, 2)

3
(3, )
2
(2, )

definition:
The complex plane (or Argand diagram) is defined as the plane in which complex numbers
z = a + ib (with a, b IR) are represented by points with coordinates (a, b) = (Re(z), Im(z)).
The horizontal axis is called the real axis, and the vertical axis the imaginary axis.
notes:
(i) The real axis is the set of all real numbers,
(as it contains all z = a + ib C
| for which b = 0, i.e. of the form z = a)
(ii) The imaginary axis is the set of all purely imaginary numbers,
(it contains all z = a + ib C
| for which a = 0, i.e. of the form z = ib)

30
Im(z)
4i 6
2 + i

3i

2+3i

1 + 2i

2i
i

-4

-3

-2

-1

Re(z)

-i

1i 2

-2i

3 3i
2

-3i
-4i

3.4.2. Polar coordinates


Recall the definition of the trigonometric functions:
(i) each point on the unit circle around the origin,
with Cartesian coordinates (x, y) such that x2 + y 2 = 1,
can be written in the form (x, y) = (cos(), sin())
(ii) Here denotes the angle with the x-axis of a half-line through the origin and (x, y)
This can be generalized easily to any circle around the origin:
(i) each point on the circle of radius r around the origin,
with Cartesian coordinates (x, y) such that x2 + y 2 = r2 ,
can be written in the form (x, y) = (r cos(), r sin())
(ii) Here denotes the angle with the x-axis of a half-line through the origin and (x, y)

31
y-axis
4 6
@
@
@




@
@





@


@
@





( 4 2, 4 2) =

@
3
(cos( 3 ), sin( 3 ))
2

( 3, 1)



1

@

@

-4

-3

-2

= 2(cos( 6 ), sin( 6 ))

-1

x-axis

-1
-2
-3
-4
underlying idea:
We can represent each point in the plane with Cartesian coordinates (x, y)
alternatively by so-called polar coordinates (r, ).
The same is then also true for each complex number z C
|.
examples:
polar coordinates:
(r, )

Cartesian coordinates:
(x, y) = (r cos(), r sin())

complex number:
z = r cos + ir sin

(r, ) = (1, 0)
(r, ) = (1, 2 )
(r, ) = (1, )
(r, ) = (1, 2 )
(r, ) = ( 32 , 3
)
4

(r, ) = (2, 6 )

(x, y) = (1, 0)
(x, y) = (0, 1)
(x, y) = (1, 0)
(x, y) = (0, 1)

(x, y) = ( 43 2, 34 2)

(x, y) = ( 3, 1)

z
z
z
z
z
z

= 1 + 0.i = 1
= 0 + 1.i = i
= 1 + 0.i = 1
= 0 1.i = i

= 34 2 + 43 2i

= 3 + 1.i = 3 + i

Notes:
(i) each complex number can thus be written as z = r(cos() + i sin()),
with r 0 and IR

32
(ii) due to periodicity of sin() and cos():
for all n IN also r(cos( + 2n) + i sin( + 2n) = z
(iii) If z = r(cos() + i sin()) then r = |z|
Proof:
|z|2 = z = r2 (cos() + i sin()(cos() i sin()
= r2 (cos2 () i2 sin2 ()) = r2
3.4.3. The exponential form of numbers on the unit circle
definition:
The unit circle in the complex plane is the set {z C
| | Re2 (z) + Im2 (z) = 1}
Alternatively, using polar coordinates: {z C
| | z = cos() + i sin() for some IR}
We now proceed, for numbers on the unit circle, to one of the main statements in complex
number theory. It impacts on all complex numbers. Whether it is a definition or a theorem
depends on ones starting point. We have so far only defined exponential functions with real
arguments, so here it has the status of an expansion of the definition of the exponential function:
definition:
ei = cos() + i sin()
rationale:
Let us show why this is the natural extension of the exponential function. Define f () = cos()+
d
i sin(), then (recall definition of derivative from school, and remember that d
cos() = sin()
d
and d sin() = cos()):
d
d
d
f () =
cos() + i sin() = sin() + i cos()
d
d
d 

= i cos() + i sin() = iei = if ()
f (0)

=1

Whereas for real numbers a we would have had, with f () = eb :


d
d
f () = eb = beb = bf ()
d
d
f (0) = 1
We see that the two situations (real exponentials versus complex exponentials) connect if and
only if we choose b = i, i.e. if we define f () = cos() + i sin() = ei
Some examples:

33
Im(z)
1.5

1.0

z = ei0 = cos(0) + i sin(0) = 1 + 0.i = 1

0.5

Re(z)

0.0

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Im(z)
1.5

1.0

0.5

Re(z)

0.0

z = ei/2 = cos( 2 ) + i sin( 2 ) = 0 + 1.i = i

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Im(z)
1.5

1.0

z = ei = cos() + i sin() = 1 + 0.i = 1

0.5

Re(z)

0.0

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Im(z)
1.5

1.0

0.5

Re(z)

0.0

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

z = e3i/2 = cos( 3
) + i sin( 3
) = 0 1.i = i
2
2

34
Im(z)
1.5

1.0

) + i sin( 2
) = 21 +
z = e2i/3 = cos( 2
3
3

0.5

Re(z)

0.0

1
2

3i

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Im(z)
1.5

1.0

z = ei/4 = cos( 4 ) + i sin( 4 ) =

0.5

Re(z)

0.0

1
2

-0.5

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

3.5. Complex numbers in exponential notation


3.5.1. Definition and general properties
Upon combining the polar coordinate representation z = r cos() + ir sin()
with the new identity ei = cos() + i sin():
claim:
Each complex number can be written in the form z = rei
with r, IR and r 0
Notes:
(i) |z| = r
proof: |z| =

zz =

rei .rei =

r2 = r

(ii) Re(z) = Re(r cos() + ir sin()) = r cos()


Im(z) = Im(r cos() + ir sin()) = r sin()

2+

1
2

2i

35
(iii) z = rei
proof: z = r cos() + i sin() = r cos() i sin() = rei
(iv) 1/z = 1r ei
proof: multiply numerator and denominator by z:
1
1
1 rei
rei
1
= i = i i = 2 = ei
z
re
re re
r
r
(v) 1/z = 1r ei
proof: multiply numerator and denominator by z:
1
1
1 rei
rei
1
= i = i i = 2 = ei
z
re
re re
r
r
(vi) The angle in z = rei is as yet not uniquely defined,
changing + 2n with n ZZ leaves one with the same number z
proof:
rei(+2n) = rei(+2n) = rei e2ni
= rei (cos(2n) + i sin(2n)) = rei .1 = rei
3.5.2. Multiplication and division in exponential notation
claim:
Multiplication of two complex numbers z = rei and w = ei ,
with real r, 0 and real , , implies
(i) multiplication of the absolute values, and
(ii) addition of the arguments
proof:
z.w = rei .ei = r ei+i = r ei(+)
claim:
Division of two complex numbers z = rei and w = ei ,
with real r, 0 and real , , implies
(i) division of the absolute values, and
(ii) subtraction of the arguments
proof:
z/w =

rei
r
r
= ei ei = ei()
i
e

36
Examples:
1
2ei/4
1
3e3i/2 . ei/2
2

1
= ei/4
2
3
= ei
2

6ei/6
3ei/4

= 2ei(1/41/6) = 2ei/12

i.rei

= ei/2 .rei = rei(+/2)

In a nutshell:
adding or subtracting complex numbers is easier in standard notation z = a + ib
multiplying or dividing is easier in exponential notation z = rei
3.5.3. The argument of a complex number
motivation
The angle in z = rei is not unique, we could add multiples of 2
This also makes it impossible to define ln(z) in C
| as the inverse of ez
0
(see conditions for inverse: demand ez 6= ez if z 6= z 0 )
definition:
The argument arg(z) of a complex number z is the angle such that
(i) z = rei with r, IR and r 0
(ii) <
Notes:
One always has, by construction: z = |z| eiarg(z)
The new condition that (, ] removes the previous ambiguity,
leaving only one unique angle = arg(z) to represent z
definition:
The natural logarithm ln(z) of a complex number z C
| can now be defined as follows


ln(z) = ln |z|eiarg(z) = ln(|z|) + ln eiarg(z) = ln(|z|) + i arg(z)


A common task is to write a complex number from standard into exponential form,
i.e. to find r = |z| and arg(z) when z = a + ib is given

37
Three-step method for finding |z| and arg(z) when z is given:
(i) calculate |z| = r using r2 = zz
(ii) calculate ei using ei = z/r,
and find all solutions using ei = cos() + i sin() (draw diagram)
(iii) determine which one obeys < : this must be arg(z)
Examples:
z=i:

r2 = zz = i.(i) = i2 = 1 hence r =

1=1

ei = z/r = i/1 = i, hence


cos() + i sin() = i

cos() = 0 and sin() = 1

solutions : = /2 + 2n (n ZZ)
(, ] :
z = 3 + 3i :

= /2 (i.e. n = 0), thus arg(z) = /2

r2 = (3 + 3i)(3 3i) = 92 + 92 = 18 hence r =

18 = 3 2

1
1
i
ei = z/r = (3 + 3i) = + , hence
3 2
2
2
1
i
1
1
cos() + i sin() = +
cos() = , sin() =
2
2
2
2
solutions : = 3/4 + 2n (n ZZ)
(, ] :

z = 3i:

= 3/4 (i.e. n = 0), thus arg(z) = 3/4

r2 = ( 3 i)( 3 + i) = 3 + 1 = 4 hence r = 4 = 2
1
1
3 i, hence
2
2

3 i
3
1
cos() + i sin() =

cos() =
, sin() =
2
2
2
2
solutions : = 7/6 + 2n (n ZZ)
ei = z/r =

(, ] :
z = e2i/3 :

= 5/6 (i.e. n = 1), thus arg(z) = 5/6

r2 = (e2i/3 )(e2i/3 ) = 1, hence r =

1=1

1 i 3
e = z/r = e
= cos(2/3) i sin(2/3) =
, hence
2
2

1 i 3
3
1
cos() + i sin() =
cos() = , sin() =
2
2
2
2
i

2i/3

38
solutions : = /3 + 2n (n ZZ)
(, ] : = /3 (i.e. n = 0), thus arg(z) = /3

common pitfalls
(see last example):
If z = e2i/3 this does not imply that r = |z| = 1 and arg(z) = 2/3
note that always |z| 0
If you arrive at ei = e2i/3 this does not imply that there has been a mistake,
note that 1 = ei , so we may write e2i/3 = ei e2i/3 = e5i/3
3.6. De Moivres Theorem
3.6.1. Statement and proof
theorem:
For all IR and n IN:


n

cos() + i sin()

= cos(n) + i sin(n)

proof 1: (via induction)


We will use cos(a+b)
= cos(a)cos(b)sin(a) sin(b) and sin(a+b) = sin(a) cos(b)+cos(a) sin(b).

n
Define An = cos() + i sin() cos(n) i sin(n).
We have to prove that An = 0 for all n IN.
(i) Induction basis: A1 = cos() + i sin() cos() i sin() = 0, so claim is true for n = 1
(ii) Induction step. We assume that An = 0 for some n IN. Now


An+1 = cos() + i sin()


= (cos() + i sin()

n+1



cos(n + ) i sin(n + ) use An = 0 :




cos(n) + i sin(n) cos(n + ) i sin(n + )

= cos() cos(n) sin() sin(n) + i sin() cos(n) + i cos() sin(n)




cos(n) cos() sin(n) sin() i sin(n) cos() + cos(n) sin()


=0
If An = 0 for some n, then also An+1 = 0. Hence, in combination with the basis,
we have now shown that An = 0 for all n IN. This completes the proof.
proof 2:


n

cos() + i sin()

= (ei )n = ein = cos(n) + i sin(n)

39
3.6.2. Applications
First application:
easy derivation of identities for trigonometric functions of multiple angles
n = 2:
cos(2) + i sin(2) = [cos() + i sin()]2 = cos2 () sin2 () + 2i sin() cos()
Real and imaginary parts on both sides must be equal, so
cos(2) = cos2 () sin2 ()
sin(2) = 2 sin() cos()
n = 3:
cos(3) + i sin(3) = [cos() + i sin()]3


= cos2 () sin2 () + 2i sin() cos() (cos() + i sin())


= cos3 () cos() sin2 () + 2i sin() cos2 ()


+ i cos2 () sin() sin3 () + 2i cos() sin2 ()

= cos3 () 3 cos() sin2 () + 3i sin() cos2 () i sin3 ()


Real and imaginary parts on both sides must be equal, so
cos(3) = cos3 () 3 cos() sin2 ()
sin(3) = 3 sin() cos2 () sin3 ()
Using sin2 () + cos2 () = 1, this can also be written as
h

cos(3) = cos() 4 cos2 () 3


sin(3) = sin() 4 cos2 () 1

Second application:
finding the roots of unity, i.e. the n solutions of z n = 1 (integer n)
We know that |z| = 1, due to 1 = |z n |n = |z|n , so we may put z = ei
For any integer m = 0, 1, 2, . . . we may write 1 = cos(2m) + i sin(2m)
Our equation then becomes (ei )n = cos(2m) + i sin(2m), or ein = e2mi
It follows that for every integer m we have a solution = 2m/n,
i.e. a complex root z = e2im/n
Note, finally: for m n we will generate solutions already found earlier
Hence: the n solutions of z n = 1 are given by
z = e2im/n

for m = 0, 1, 2, . . . , n 1

40
For example:
Im(z)

Im(z)

Im(z)

1.5

1.5

1.5

1.0

1.0

1.0

0.5

0.5

0.5

Re(z)

0.0

Re(z)

0.0

-0.5

-0.5

-0.5

-1.0

-1.0

-1.0

-1.5
-1.5

-1.0

-0.5

0.0

0.5

1.0

-1.5
-1.5

1.5

-1.0

z2 = 1

-0.5

0.0

0.5

1.0

-1.5
-1.5

1.5

Im(z)

Im(z)

1.0

1.0

1.0

0.5

0.5

Re(z)

0.0

0.0

-0.5

-1.0

-1.0

-1.0

1.0

1.5

-1.5
-1.5

z5 = 1

1.0

1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Re(z)

0.0

-0.5

0.5

0.5

0.5

Re(z)

-0.5

0.0

0.0

Im(z)
1.5

-0.5

-0.5

z4 = 1

1.5

-1.0

-1.0

z3 = 1

1.5

-1.5
-1.5

Re(z)

0.0

-1.5
-1.5

-1.0

z6 = 1

-0.5

0.0

0.5

1.0

z7 = 1

3.7. Complex equations


Complex equations are equations that can be reduced to the general form F (z, z) = 0,
where F denotes some function of z and z. Note: F can involve |z|, since |z|2 = zz.
Solving a complex equation means finding all z C
| with the property F (z, z) = 0.
We have already encountered examples:
(i) zz 1 = 0:
Here the solution set is the unit circle in the complex plane (i.e. infinitely many)
(ii) z n 1 = 0:
Here the solution set is a discrete set of n points (see previous section)
Note:
the solution sets in the complex plane of complex equations
can be more diverse than those of real equations, or than the previous examples

1.5

41
Lines in the complex plane:
these are found as solutions of linear equations, for example uz+vz+w = 0 with u, v, w C
|
has solutions which are lines (as well as other solutions). Check this and identify the
condition for the solution sets to be lines in the complex plane.
Discrete sets of points that are not arranged around the origin:
Im(z)

Im(z)

5.0

4.0

3.0
1
2.0

Re(z)

0
1.0
-1
0.0

Re(z)
-2

-1.0

-2.0
-2.0

-3
-1.0

0.0

1.0

2.0

3.0

4.0

5.0

(z 2 i)6 64 = 0

-3

-2

-1

z 5 2(1 + i)z 4 + (1 + i)z 3


+(i 1)z 2 (2 + i)z + i + 3 = 0
Im(z)

Ellipses:
these are solutions of equations of the type
|z u| + |z w| R = 0
with u, w C
| and R IR+
(u and w will be the foci of the ellipse)

Re(z)

-1

-2
-2

-1

|z 1| + |z + 1| 3 = 0

But also unions of isolated points


and curves

Im(z)
3

And more ...

Re(z)

-1

-2

-3
-3

-2

-1

(zz 1)(z 3 2iz 2 2(2i + 1)z 20 8i) = 0

42
4. Trigonometric and hyperbolic functions
4.1. Definitions of trigonometric functions
4.1.1. Definition of sine and cosine
We need only define sin() and cos(), since all other trigonometric functions are simply
combinations of these elementary two.
There are different but mathematically equivalent options:
Option I. geometric definition:
Consider a half-line with its one end-point in the origin. Choose it first to lie along the positive
x-axis, then rotate it anti-clockwise around the origin; call the rotation angle (in radians).
Find the coordinates (X(), Y ()) of the point where the half-line intersects the unit circle.
Now define
tan()
2
cos() = X()
sin() = Y ()
1

sin()

Option II. definition via differential equations:


We could also define trigonometric functions
as the solutions of the following equations,
with specific initial values:

cos()
/

-1

d
sin() = cos()
d
d
cos() = sin()
d
cos(0) = 1, sin(0) = 0

-2
-2

-1

Option III. analytic definition:


Here we start by defining the function ez for any z C
| via a series,
and subsequently define trigonometric functions via
1
1
sin(z) = (eiz eiz )
cos(z) = (eiz + eiz )
2
2i
(note: this will also generalize trigonometric functions to complex arguments!)
The exponential function takes the form

X
zn
z
z2 z3 z4
ez =
=1+ +
+
+
+ ...
1! 2!
3!
4!
n=0 n!

43
(note: one defines 00 = 1). We turn later in detail to existence and convergence questions for
infinite series. For now we assume (rightly, as will turn out) that this infinite sum is always
finite and well-behaved.
One then finds

X
(1)k z 2k+1
z3 z5 z7
sin(z) =
= z
+

+ ...
3!
5!
7!
k=0 (2k + 1)!

z2 z4 z6
(1)k z 2k
= 1
+

+ ...
cos(z) =
(2k)!
2!
4!
6!
k=0
proof:
Let us start with sin(z):
sin(z) =

n n
(iz)n
1 X
(iz)n
1X
i z 1
1 X

=
(1 (1)n )
2i n=0 n!
2i n=0 n!
i n=0 n! 2

Note:
(i) that 21 (1 (1)n ) = 0 for even n, and 12 (1 (1)n ) = 1 for odd n
hence in the sum we only retain term with odd n
(ii) that the odd values of n can be written as n = 2k + 1, with k = 0, 1, 2, 3, . . .
sin(z) =

2k+1 2k+1
1 X
1X
in z n
i
z
=
i n odd n!
i k=0 (2k + 1)!

X
(1)k z 2k+1
(i2 )k z 2k+1
=
=
k=0 (2k + 1)!
k=0 (2k + 1)!

Next we turn to cos(z):


cos(z) =

n n
X
1X
(iz)n
1 X
(iz)n
i z 1
+
=
(1 + (1)n )
2 n=0 n!
2i n=0 n!
n!
2
n=0

Note:
(i) that 12 (1 + (1)n ) = 1 for even n, and 12 (1 + (1)n ) = 0 for odd n
hence in the sum we only retain term with even n
(ii) that the even values of n can be written as n = 2k, with k = 0, 1, 2, 3, . . .
cos(z) =
=

2k 2k
X
in z n
i z
=
even n!
k=0 (2k)!

X
(i2 )k z 2k
(1)k z 2k
=
(2k)!
k=0 (2k)!
k=0

44
0 2 4 6

x
sin(x)

-1

-2

1 3 5 7

-3
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10

Figure 1. Building sin(x) as a power series, by taking more and more terms in the summation.
PN
Dashed: sin(x). Solid: f (x) = k=0 (1)k x2k+1 /(2k + 1)! for different choices of N (values of
N are indicated in italics). As N increases: f (x) starts to resemble sin(x) more and more, but
f (x) is only fully identical to sin(x) for N .

2 4 6

cos(x)

-1

-2

1 3 5 7

-3
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10

Figure 2. Building cos(x) as a power series, by taking more and more terms in the summation.
PN
Dashed: cos(x). Solid: f (x) = k=0 (1)k x2k /(2k)! for different choices of N (values of N are
indicated in italics). As N increases: f (x) starts to resemble cos(x) more and more, but f (x)
is only fully identical to sin(x) for N .

45
4.1.2. Elementary values
These can all be extracted from either (i) suitably chosen triangles, and/or (ii) projections of
special points on the unit circle onto (x, y)-axes, or (iii) simple transformations to reduce to
one of the previous two classes:
= 0, inspect projections of point on unit circle,
cos(0) = 1, sin(0) = 0
= /4, inspect inspect projections of point on unit circle, use sin2 () + cos2 () = 1

cos(/4) = sin(/4) = 1/ 2
= /6: cut a triangle with three equal sides in half and pick a suitable corner,

sin(/6) = 12 , cos(/6) = 12 3
= /3: cut a triangle with three equal sides in half and pick a suitable corner,

sin(/3) = 12 3, cos(/3) = 1/2


= /2: inspect projections of point on unit circle,
cos(/2) = 0, sin(/2) = 1
4.1.3. Related functions
These are just short-hands for frequently occurring combinations of sine and cosine:
tangent: tan() = sin()/ cos()
(i) not defined for = /2 + n with n ZZ, where cos() = 0
(ii) tan( + ) = tan() for all IR
since sin( + ) = sin() and cos( + ) = cos()
cotangent: cot() = cos()/ sin()
(i) not defined for = n with n ZZ, where sin() = 0
(ii) cot( + ) = cot() for all IR
since sin( + ) = sin() and cos( + ) = cos()
secant: sec() = 1/ cos()
(i) not defined for = /2 + n with n ZZ, where cos() = 0
(ii) sec( + 2) = sec() for all IR
since cos( + 2) = cos()
cosecant: cosec() = 1/ sin()
(i) not defined for = n with n ZZ, where sin() = 0
(ii) cosec( + 2) = cosec() for all IR
since sin( + 2) = sin()

46
4.1.4. Inverse trigonometric functions
Recall our definitions:
The inverse f 1 of a function f : D R
is defined by:
f 1 : R D
f

tan()

sin()

(f (x)) = x for all x D

cos()

f (f 1 (x)) = x for all x R


/

A function f : D R is invertible
if and only if:

-1

f (x1 ) 6= f (x2 ) for any two x1 , x2 D


with x1 6= x2

-2
-2

-1

Inspect graphs of trigonometric functions:


(i) problem:
there are many 1 , 2 IR with 1 6= 2 such that sin(1 ) = sin(2 ) ...
(same is true for cosine and tangent)
(ii) hence:
one can only define inverse trigonometric functions by limiting their domains
to sets where any two distinct angles will give different function values!
definition of the inverse of sin():

arcsin(x)

need interval D that satisfies


(i) sin(1 ) 6= sin(2 ) for all 1 , 2 D with 1 6= 2
(ii) the range corresponding to D covers all possible values of sine, R = [1, 1]
Answer: D = [/2, /2], so
arcsin : [1, 1] [/2, /2]
arcsin(sin()) = for all [/2, /2]
sin(arcsin(x)) = x for all x [1, 1]
arcsin(x) in words:
gives the angle [/2, /2] such that sin() = x

47

2.0

arcsin(x)

/2

1.5

1.0

sin(x)

0.5

0.0

/2

-0.5

sin(x)

-1.0

-1.5

arcsin(x)
-2.0
-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Special values:

arcsin(0) = 0, arcsin(1/ 2) = /4, arcsin( 12 ) = /6, arcsin( 12 3) = /3, arcsin(1) = /4


negative values via: arcsin(x) = arcsin(x)
definition of the inverse of cos():

arccos(x)

need interval D that satisfies


(i) cos(1 ) 6= cos(2 ) for all 1 , 2 D with 1 6= 2
(ii) the range corresponding to D covers all possible values of cosine: R = [1, 1]
Answer: D = [0, ], so
arccos : [1, 1] [0, ]
arccos(cos()) = for all [0, ]
cos(arccos(x)) = x for all x [1, 1]
arccos(x) in words:
gives the angle [0, ] such that cos() = x

48

3.5
3.0

arccos(x)

2.5
2.0
1.5
1.0

cos(x)

0.5

0.0

arccos(x)
-0.5

cos(x)

-1.0
-1.5
-1.5 -1.0 -0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Special values:

arccos(0) = /2, arccos( 21 ) = /3, arccos(1/ 2) = /4, arccos( 21 3) = /6, arccos(1) = 0


negative values of x via: arccos(x) = arccos(x)
definition of the inverse of tan():

arctan(x)

need interval D that satisfies


(i) tan(1 ) 6= tan(2 ) for all 1 , 2 D with 1 6= 2
(ii) the range corresponding to D covers all possible values of tangent, R = IR
Answer: D = (/2, /2), so
arctan : IR (/2, /2)
arctan(tan()) = for all (/2, /2)
tan(arctan(x)) = x for all x IR
arctan(x) in words:
gives the angle (/2, /2) such that tan() = x

49

4.0

tan(x)

3.5
3.0
2.5

arctan(x)

2.0

/2

1.5
1.0
0.5
0.0

/2

-0.5

-1.0
-1.5

arctan(x)

-2.0
-2.5
-3.0
-3.5

tan(x)

-4.0
-4.0-3.5-3.0-2.5-2.0-1.5-1.0-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Special values:

arctan(0) = 0, arctan(1/ 3) = /6, arctan(1) = /4, arctan( 3) = /3


negative values via: arctan(x) = arctan(x)

4.2. Elementary properties of trigonometric functions


4.2.1. Symmetry properties
The unit circle is invariant under:
(i) reflection in the y-axis, i.e. (x, y) (x, y)
(ii) reflection in the x-axis, i.e. (x, y) (x, y)
(iii) rotation by , i.e. (x, y) (x, y)
(iv) reflection in the line x = y, i.e. (x, y) (y, x)
Symmetries have implications for the values of trigonometric functions
(note the geometric definition of sine and cosine):

50
y

reflection in x-axis, i.e. :

We see that:
sin() = sin()
cos() = cos()

-1

-1

reflection in y-axis, i.e. :

We see that:
sin( ) = sin()
cos( ) = cos()

-1

-1

rotation by , i.e. + :

We see that:
sin( + ) = sin()
cos( + ) = cos()

-1

-1

reflection in line x = y, i.e. /2 :

We see that:
sin(/2 ) = cos()
cos(/2 ) = sin()

-1

-1

51
4.2.2. Addition formulae
Trigonometric functions of sums or differences of angles:
claim: for all , IR one has
sin( + ) = sin() cos() + cos() sin()
cos( + ) = cos() cos() sin() sin()
proofs:
subtract left- and right-hand sides of the two identities,
and show that the result is zero
use definitions in terms of exponentials:
LHS1 RHS1 = sin( + ) sin() cos() cos() sin()
1
1
1
= (ei(+) ei(+) ) (ei ei )(ei + ei ) (ei + ei )(ei ei )
2i
4i
4i

1 i(+)
1 i(+)
1  i(+)
= e
e

e
+ ei() ei() ei(+
2i
2i
4i

1  i(+)

e
ei() + ei() ei(+)
4i

1
1  i(+)
1 i(+)
ei(+)
2e
2ei(+ = 0
= e
2i
2i
4i
LHS2 RHS2 = cos( + ) cos() cos() + sin() sin()
1
1
1
= (ei(+) + ei(+) (ei + ei )(ei + ei ) (ei ei )(ei ei )
2
4
4

1 i(+) 1 i(+) 1  i(+)
i()
i()
+ e

= e
e
+e
+e
+ ei(+)
2
2
4

1  i(+)
i()

e
e
ei() + ei(+)
4

1 i(+) 1 i(+) 1  i(+)
+ e

= e
2e
+ 2ei(+) = 0
2
2
4
This completes the proofs.
From the formulae for sine and cosine follow also:
(dont memorize, but derive when needed!)
sin() cos() + cos() sin()
sin( + )
=
cos( + )
cos() cos() sin() sin()
sin()/ cos() + sin()/ cos()
=
1 sin() sin()/ cos() cos()
tan() + tan()
=
1 tan() tan()

tan( + ) =

52
observation:
we rely on the property ez+w = ez .ew
interesting to prove this property
from the series representation of ez alone!
use Newtons binomial formula and (
ez+w =

am )(

n bn )

n,m

am b n :

n  
X
(z + w)n
1 X
n k nk
=
z w
n!
n=0
n=0 n! k=0 k

Inspect these summations closer:


we have ultimately n = 0, 1, 2, . . . and k = 0, 1, 2, . . . ,,
but we restrict ourselves to those combinations (n, k) with k n
Hence we may also write :
X

X
1
n!
1  n  k nk X
z k wnk
z w
=
ez+w =
k
n!
n!
k!(n

k)!
k=0 n=k
k=0 n=k
X

1
z k wnk
k!(n

k)!
k=0 n=k
Finally, switch from the index n to ` = n k, so ` = 0, 1, 2, . . .:
!
!
X

X
X
X w`
1 1 k `
zk
z+w
z w =
= ez .ew
e
=
k=0 `=0 k! `!
k=0 k!
`=1 `!
=

4.2.3. Applications of addition formulae


writing products of trigonometric functions as sums:

1
cos() cos() =
cos( + ) + cos( )
2

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

1
sin( + ) + sin( )
sin() cos() =
2
proofs: trivial
just insert the appropriate addition formulae in the right-hand sides
recovering formulae for double angles:
just choose = in the addition formulae
sin(2) = 2 sin() cos()
cos(2) = cos2 () sin2 ()
tan(2) = 2 tan()/[1 tan2 ()]

53
half-angle formulae:
1
1
1
1
cos() + cos() = cos( ( + ) + ( )) + cos( ( + ) ( ))
2
2
2
2


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


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


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


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


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


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


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


1
1
1
1
sin( ( + )) cos( ( )) cos( ( + )) sin( ( ))
2
2
2
2
1
1
= 2 sin( ( )) cos( ( + ))
2
2
Claim: one can always write linear combinations a cos() + b sin() in the form c sin( + )
with some suitable c, IR and with c 0
(lets disregard the trivial case a = b = 0)
Proof & construction in three steps:

54
(i) First note that
c sin( + ) = c sin() cos() + c cos() sin()
Hence we seek c and such that
a/c = sin()

b/c = cos()

Use sin2 () + cos2 () = 1: a2 + b2 = c2 , so c =

a2 + b 2

(ii) Next: find from


a/c = sin()

b/c = cos()

hence tan() = a/b, so we find = arctan(a/b) + n with n ZZ


(note: arctan() must be in (/2, /2), but there is no reason why should be there!)
(iii) Finally: determine n from a/c = sin() and b/c = cos()
Just check the quadrant of the solution in the plane, by inspecting signs.
Note: arctan() (/2, /2) is always in quadrant 1 or 4:
b/c = 0 :

arctan() doesn0 t exist, here cos() = 0 so


= /2 + 2n (n ZZ) if a > 0
= 3/2 + 2n (n ZZ) if a < 0

b/c > 0 :

quadrant 1 or 4, so = arctan(a/b) + 2n (n ZZ)

b/c < 0 :

quadrant 2 or 3, so = arctan(a/b) + + 2n (n ZZ)

4.2.4. The tan(/2) formulae


Objective:
write all trigonometric functions in terms of t = tan( 21 )
(useful later in integrals)
tan() =

2 tan( 12 )
2t
=
2 1
1 t2
1 tan ( 2 )

1
1
1
2
cos() = cos2 ( ) sin2 ( ) = 2 cos2 ( ) 1 =
1
2
2
2
cos2 ( 12 )
2
2
=
1=
1
2 1
1
1
2 1
2
2
tan ( 2 ) + 1
[sin ( 2 ) + cos ( 2 )] cos ( 2 )
2
1 + t2
1 t2
=

=
1 + t2 1 + t2
1 + t2
sin() = cos() tan() =

1 t2 2t
2t
=
2
2
1+t 1t
1 + t2

55
4.3. Definitions of hyperbolic functions
4.3.1. Definition of hyperbolic sine and hyperbolic cosine
Option I. definition via differential equations:
We can define the hyperbolic sine sinh(z) and hyperbolic cosine cosh(z) as the solutions of the
following equations, with specific initial values:
d
d
sinh(z) = cosh(z),
cosh(z) = sinh(z),
cosh(0) = 1,
sinh(0) = 0
dz
dz
(note:
difference with previous eqns defining sine and cosine
is only in a minus sign in the second eqn!)
cosh(x)

Option II. direct analytic definition

cosh(x)

Having already defined e


P
n
by the series ez =
n=0 z /n!
we define hyperbolic functions via

sinh(x)

tanh(x)
1
cosh(z) = (ez + ez )
2
1
sinh(z) = (ez ez )
2

0
-1

tanh(x)
-2

(this also generalizes


hyperbolic functions
to complex arguments)

-3

sinh(x)

-4
-5
-5

-4

-3

-2

Related functions:
hyperbolic tangent :

tanh(z) = sinh(z)/ cosh(z)

hyperbolic cotangent :

coth(z) = cosh(z)/ sinh(z)

hyperbolic secant :

sech(z) = 1/ cosh(z)

hyperbolic cosecant :

cosech(z) = 1/ sinh(z)

-1

56
4.3.2. General properties and special values
Properties involving both sinh and cosh:
One immediately confirms from the direct analytic definition:
d
d
sinh(z) = cosh(z)
cosh(z) = sinh(z)
dz
dz
For any z C
| : cosh2 (z) sinh2 (z) = 1
proof:
cosh2 (z) sinh2 (z) =

1

(ez + ez )

2

1

2

(ez ez )

2
2


1  2z
1
= (e + 2 + e2z ) e2z 2 + e2z )
4
4

1  2z
= (e + 2 + e2z e2z + 2 e2z ) = 1
4

Consequence:
if for z IR the two are regarded as coordinates (X, Y ) in a plane,
i.e. X = cosh(z) and Y = sinh(z), then the possible points (X, Y )
define the branches of a hyperbole X 2 Y 2 = 1 (hence the name!)

Properties of sinh:
sinh(z) = sinh(z)
proof: follows directly from definition,
1
1
sinh(z) = (ez ez ) = (ez ez ) = sinh(z)
2
2
for z IR: sinh(z) increases monotonically
proof: differentiate the analytic definition,
d az
e = aeaz ,
using dz
d
1
sinh(z) = cosh(z) = (ez + ez ) > 0
dz
2
consider z IR:
as z : ez 0 and ez = 1/ez , hence: sinh(z) = 21 (ez ez )
at z = 0:
sinh(0) = 12 (e0 e0 ) = 0
as z : ez and ez = 1/ez 0, hence: sinh(z) = 21 (ez ez )

57
Properties of cosh:
cosh(z) = cosh(z)
proof: follows directly from definition,
1
1
cosh(z) = (ez + ez ) = (ez + ez ) = cosh(z)
2
2
for z IR+ : cosh(z) increases monotonically
for z IR : cosh(z) decreases monotonically
proof: differentiate the analytic definition,
d az
e = aeaz ,
using dz
d
cosh(z) = sinh(z)
dz

>0
=0

<0

for z > 0
for z = 0
for z < 0

consider z IR:
as z : ez 0 and ez = 1/ez , hence: cosh(z) = 12 (ez + ez )
at z = 0:
cosh(0) = 21 (e0 + e0 ) = 0 = 1
as z : ez and ez = 1/ez 0, hence: cosh(z) = 21 (ez + ez )
Properties of tanh:
tanh(z) = tanh(z)
proof: follows directly from definition,
tanh(z) = sinh(z)/ cosh(z) = sinh(z)/ cosh(z) = tanh(z)
for z IR:

tanh(z) increases monotonically

proof: differentiate the definition,


d
d
using dz
sinh(z) = cosh(z) and dz
sinh(z) = cosh(z),
d
d
cosh(z) dz
sinh(z) sinh(z) dz
cosh(z)
2
cosh (z)
2
2
cosh (z) sinh (z)
1
=
=
>0
2
cosh (z)
cosh2 (z)

d
d
tanh(z) =
dz
dz

sinh(z)
cosh(z)

consider z IR, rewrite tanh(z) in two distinct ways:


z

sinh(z)
e e
tanh(z) =
= z
=
cosh(z)
e + ez

(1 e2z )/(1 + e2z ) so tanh(z) 1


if z
0
for
z=0
2z
2z
(e 1)/(e + 1) so tanh(z) 1 if z

58
4.3.3. Connection with trigonometric functions
More than just similarity between trigonometric and hyperbolic functions,
When defined for complex numbers they can be expressed in terms of each other!
Let x IR:
trigonometric functions are hyperbolic functions of imaginary arguments:
sin(x) = i sinh(ix)
cos(x) = cosh(ix)
tan(x) = i tanh(ix)
proofs:
just write RHS in terms of exponentials ...
hyperbolic functions are trigonometric functions of imaginary arguments:
sinh(x) = i sin(ix)
cosh(x) = cos(ix)
tanh(x) = i tan(ix)
proofs:
just write RHS in terms of exponentials ...

4.3.4. Applications of connection with trigonometric functions


All previous identities for trigonometric functions
whose derivation did not rely on argument being real (e.g. addition formulae),
translate via the above into identities for hyperbolic functions!
Addition formulae:
sin( + ) = sin() cos() + cos() sin()
cos( + ) = cos() cos() sin() sin()
tan() + tan()
tan( + ) =
1 tan() tan()
give:
sinh( + ) = i sin(i) cos(i) i cos(i) sin(i)
= sinh() cosh() + cosh() sinh()
cosh( + ) = cos(i) cos(i) sin(i) sin(i)

59
= cosh() cosh() + sinh() sinh()
tanh( + ) =

i tan(i) i tan(i)
tanh() + tanh()
=
1 tan(i) tan(i)
1 + tanh() tanh()

Formulae for double angles:


sin(2) = 2 sin() cos()
cos(2) = cos2 () sin2 ()
tan(2) = 2 tan()/[1 tan2 ()]
give:
sinh(2) = 2i sin(i) cos(i) = 2 sinh() cosh()
cosh(2) = cos2 (i) sin2 (i) = cosh2 () + sinh2 ()
tanh(2) = 2i tan(i)/[1 tan2 (i)] = 2 tanh()/[1 + tanh2 ()]

4.3.5. Inverse hyperbolic functions


Recall our definitions:
The inverse f 1 of a function f : D R
is defined by:

cosh(x)

cosh(x)

f 1 : R D

f 1 (f (x)) = x for all x D

f (f 1 (x)) = x for all x R

sinh(x)
tanh(x)
x

0
-1

A function f : D R is invertible
if and only if:

tanh(x)
-2
-3

f (x1 ) 6= f (x2 ) for any two x1 , x2 D


with x1 6= x2

sinh(x)

-4
-5
-5

-4

-3

-2

-1

Inspect graphs of hyperbolic functions:


(i) there are generally two x1 , x2 IR with x1 6= x2 such that cosh(x1 ) = cosh(x2 ),
namely x1 = x2
(in contrast to sinh : IR IR and tanh : IR (1, 1), which are invertible)
(ii) hence: we must limit the domain of cosh(x) to a set where
any two distinct angles will give different function values

60
definition of the inverse of sinh(x):

arcsinh(y)

arcsinh : IR IR
arcsinh(sinh(x)) = x for all x IR
sinh(arcsinh(y)) = y for all y IR
arcsinh(y) in words:
gives the value x IR such that sinh(x) = y
In fact: we can produce a simple formula:
1
y = sinh(x) : y = (ex ex ) ex ex 2y = 0 e2x 2yex 1 = 0
2
q
q
1
(ex )2 2y(ex ) 1 = 0 ex = (2y 4y 2 + 4 = y y 2 + 1
2
2
2
x
x
since e > 0: e = y + y + 1, so x = ln(y + y + 1)
hence:
arcsinh(y) = ln(y +

y 2 + 1)

for all y IR

sinh(x)
3

arcsinh(x)

-1

arcsinh(x)-2
-3

sinh(x)
-4
-4

-3

-2

-1

61
definition of the inverse of tanh(x):

arctanh(y)

arctanh : (1, 1) IR
arctanh(tanh(x)) = x for all x IR
tanh(arctanh(y)) = y for all y (1, 1)
arctanh(y) in words:
gives the value x IR such that tanh(x) = y
In fact: we can produce a simple formula:
y = tanh(x) : y =

ex ex
ex + ex

y(ex + ex ) = ex ex y(e2x + 1) = e2x 1

1 + y = e2x (1 y) e2x =

1+y
1y

2x = ln[(1 + y)/(1 y)]

hence:
1+y
1
arctanh(y) = ln
2
1y

for all y (1, 1)

arctanh(x)
3

tanh(x)
1

-1

tanh(x)
-2

-3

arctanh(x)
-4
-4

-3

-2

-1

62
definition of the inverse of cosh(x):

arccosh(y)

need interval D that satisfies


(i) cosh(x1 ) 6= cosh(x2 ) for all x1 , x2 D with x1 6= x2
(ii) the range corresponding to D covers all possible values of cosh, R = [1, )
Answer: D = [0, ), so
arccosh : [1, ) [0, )
arccosh(cosh(x)) = x for all x [0, )
cosh(arccosh(y)) = y for all y [1, )
arccosh(y) in words: gives the value x [0, ) such that cosh(x) = y
In fact: we can produce a simple formula:
1
y = cosh(x) : y = (ex + ex ) ex + ex 2y = 0 e2x 2yex + 1 = 0
2
q
q
1
(ex )2 2y(ex ) + 1 = 0 ex = (2y 4y 2 4 = y y 2 1
2
2

x
x
since x [0, ): e 1, and hence e = y + y 1, so x = ln(y + y 2 1)
hence:
arccosh(y) = ln(y +

y 2 1)

for all y [1, )

cosh(x)
3

arccosh(x)

1
-1

-2

-3

-4
-4

-3

-2

-1

63
5. Functions, limits and differentiation
5.1. Introduction
5.1.1. Rate of change, tangent of a curve
limits resolved two long-standing problems:
(i) mechanics: how to define and find the instantaneous rate of change of a quantity
(ii) geometry: how to define and find the tangent to arbitrary curves in arbitrary points
Consider time-dependent quantity x(t), t IR denotes time
(e.g. position of a particle moving along a line)
rate of change over interval t [t1 , t2 ]: average velocity v during the interval
v=

x(t2 ) x(t1 )
change in x
=
time taken
t2 t1

observation in (t, x) graph:


v is slope
of line through the points
(t1 , x(t1 )) and (t2 , x(t2 ))
instantaneous speed at time t1 :
value of v when t2 t1

(t1 , x(t1 ))
4

x(t)
result: tangent at curve x(t)
at the point t = t1

(t2 , x(t2 ))
2

problems (i) (mechanics)


and (ii) (geometry)
are the same!

v at t1
v interval
0
0

So far: only ideas, definitions and pictures ...


Calculus: find formulas for the instantaneous velocities (or tangents),
when the curves x(t) are given

64
note:
not all curves are written as x(t) or f (x) or y(x) ...
(liberate yourself from name conventions!)

5.1.2. Finding tangents and velocities why we need limits


Calculation would seem obvious, e.g.
Fermats calculation of instantaneous slope of function f (x) at value x:
work out formula for average slope during interval [x, x + h]
f (x + h) f (x)
h
put h = 0 in the result
slope =

Often this simple recipe works ...


f (x) = ax + b:
f (x + h) f (x)
[a(x + h) + b] [ax + b]
ah
=
=
=a
h
h
h
slope = a
slope =

Put h = 0:

f (x) = ax2 + bx + c:
f (x + h) f (x)
[a(x + h)2 + b(x + h) + c] [ax2 + bx + c]
=
h
h
a(x2 + 2xh + h2 ) + bx + bh + c ax2 bx c
=
h
2
a(2xh + h ) + bh
= 2ax + ah + b
=
h
slope = 2ax + b
slope =

Put h = 0:

f (x) = axn , n IN+ :


use binomial formula,
n  
f (x + h) f (x)
a(x + h)n axn
a X
n nk k
slope =
=
=
x h xn
h
h
h k=0 k

=
Put h = 0:

n  
n  
n1
X
X n 
aX
n nk k
n nk k1
x h =a
x h
=a
xn`1 h`
k
`+1
h k=1 k
k=1
`=0

slope = a

n

xn1 = a

n!
1!(n1)!

xn1 = anxn1

65
But equally often it doesnt ...
f (x) = ax :
ah 1
f (x + h) f (x)
ax+h ax
slope =
=
= ax
h
h
h
Putting h = 0 gives:

slope = ax ((a0 1)/0) = ax (0/0) ... ??

f (x) = sin(x):
f (x + h) f (x)
sin(x + h) sin(x)
=
h
h
sin(x) cos(h) + cos(x) sin(h) sin(x)
=
h!
!
cos(h) 1
sin(h)
= sin(x)
+ cos(x)
h
h
Putting h = 0 gives: slope = sin(x)(0/0) + cos(x)(0/0) ... ??
slope =

f (x) = cos(x):
cos(x + h) cos(x)
f (x + h) f (x)
=
h
h
cos(x) cos(h) sin(x) sin(h) cos(x)
=
h!
!
cos(h) 1
sin(h)
= cos(x)
sin(x)
h
h
Putting h = 0 gives: slope = cos(x)(0/0) sin(x)(0/0) ... ??
slope =

The problem:
One cannot set h = 0 in expressions like (ah 1)/h or (cos(h) 1)/h or sin(h)/h
The solution: (Newton, Leibniz)
The correct thing to do is to take h smaller and smaller, and investigate whether
the quantity [f (x + h) f (x)]/h then approaches a well-defined value for h 0.
If so: that value will be the slope at point x, to be called the derivative of f (x) at x

Notes:
(i) Newton & Leibniz had the concept, the intuitive definition of limit
but exact mathematical definition of limit was due to Cauchy, much later ...
(ii) notation for derivative:
df
f (x + h) f (x)
f 0 (x) =
= lim
dx h0
h

66
We saw:
(i) derivatives of functions that are sums of powers are easy to find
(ii) ergo: another important use of power series!

ex

sin(x) =
cos(x) =

xn
n=0 n!

1
= 1 + x + x2 + . . .
2

(1)n x2n+1
1
1 5
= x x3 +
x + ...
6
120
n=0 (2n + 1)!

(1)n x2n
(2n)!
n=0

1
1
= 1 x2 + x4 + . . .
2
24

Remember stumbling blocks on previous page:


1
lim (ah 1)
h0 h



1  ln(ah )
1  h ln a
e
1 = lim
e
1
h0 h
h0 h


1
1
1 + h ln a + (h ln a)2 + . . . 1
= lim
h0 h
2


1
= lim ln a + h(ln a)2 + . . . = ln a
h0
2

= lim

1
1
1
1
lim (cos(h) 1) = lim
1 h2 + h4 + . . . 1
h0 h
h0 h
2
24


1
1 3
= lim h + h + . . . = 0
h0
2
24

1
sin(h)
h0 h
lim

1
1
1 5
h h3 +
h + ...
h0 h
6
120


1
1 4
= lim 1 h2 +
h + ... = 1
h0
6
120


= lim

Hence
d x
a
dx

1
= ax lim (ah 1) = ax ln(a)
h0 h

d
1
1
sin(x) = sin(x) lim (cos(h) 1) + cos(x) lim sin(h) = cos(x)
h0 h
h0 h
dx
d
1
1
cos(x) = cos(x) lim (cos(h) 1) sin(x) lim sin(h) = sin(x)
h0 h
h0 h
dx

67
5.2. The limit
Limit in words:
the output value f (x) to which a function tends (if at all)
as x approaches a specific input value x0
Some simple abbreviations and symbols:
: there exists

: for all

: if and only if

5.2.1. Left and right limits


The proper definition ...
The right limit: approach x0 from the right,
notation: limxx+0 f (x) or limxx0 f (x)
definition
lim f (x) = L ( > 0)( > 0) : |f (x)L| < whenever x (x0 , x0 +)

xx0

in words:
lim f (x) = L for all > 0 there exists a > 0 such that
|f (x)L| < whenever x (x0 , x0 +)

xx0

in sloppy words:
one may get f (x) as close as one wishes to the value L
simply by lowering x sufficiently close to x0
The left limit: approach x0 from the left,
notation: limxx0 f (x) or limxx0 f (x)
definition
lim f (x) = L ( > 0)( > 0) : |f (x)L| < whenever x (x0 , x0 )

xx0

in words:
lim f (x) = L for all > 0 there exists a > 0 such that
|f (x)L| < whenever x (x0 , x0 )

xx0

in sloppy words:
one may get f (x) as close as one wishes to the value L
simply by raising x sufficiently close to x0

68
Notes:
(i) these limits need not always exist
(ii) if they do, then limxx0 f (x) might be different from limxx0 f (x)
(iii) limxx0 f (x) and limxx0 f (x) could exist even if f (x0 ) does not exist
Examples (draw pictures!):

f (x) = 1/x :

neither lim f (x) nor lim f (x) exist,


x0

x0

f (0) does not exist


f (x) = 1/x :

lim f (x) = lim f (x) = f (1) = 1

f (x) = tanh(1/x) :

lim f (x) = 1, lim f (x) = 1,

x1

x0

x1

x0

f (0) does not exist


(

f (x) =

x/|x|
0

for
for

x 6= 0
:
x=0

lim f (x) = 1, lim f (x) = 1,


x0

x0

f (0) = 0

x
f (x) =

x1

for
for
for

x>0
x=0 :
x<0

lim f (x) = 0, lim f (x) does not exist,


x0

x0

f (0) =

5.2.2. Asymptotics - limits involving infinity


Approach (always from the left!),
notation: limx f (x)
definition
lim f (x) = L ( > 0)(X > 0) : |f (x)L| < whenever x > X

in sloppy words:
one may get f (x) as close as one wishes to the value L
simply by making x larger and larger

69
Approach (always from the right!),
notation: limx f (x)
definition
lim f (x) = L ( > 0)(X < 0) : |f (x)L| < whenever x < X

in sloppy words:
one may get f (x) as close as one wishes to the value L
simply by making x smaller and smaller
5.2.3. When left/right limits exists and are identical
Approach x0 from either side,
notation: limxx0 f (x)
definition (version 1)
lim f (x) = L

xx0

lim f (x) = lim f (x) = L


x0

x0

definition (version 2)
lim f (x) = L ( > 0)( > 0) : |f (x)L| < whenever |xx0 | <

xx0

in sloppy words:
one may get f (x) as close as one wishes to the value L
simply by taking x sufficiently close to x0 , from either side
the concept of continuity of a function
definition
A function f is continuous at the point x0 if
limxx0 f (x) exists and limxx0 f (x) = f (x0 )
continuous functions are those for which the graph can always be drawn
without lifting ones pen from the paper
Mathematical expressions may involve multiple limits,
notation conventions:
lim lim f (x, y) = lim

lim lim f (x, y) = lim

xx0 yy0
yy0 xx0

yy0

xx0
yy0

lim f (x, y)

lim f (x, y)

xx0

70
Note: the order in which limits are taken matters! (jargon: limits do not commute)
e.g.
f (x, y) = x2 y exy :

lim lim f (x, y) = lim (x2 ex1 ) = 1/e

x0 y1

x0

lim lim f (x, y) = lim (ey ) = 1/e

y1 x0

f (x, y) = (2x y)/(x + 3y) :

y1

lim lim f (x, y) = lim (2x/x) = 2

x0 y0

x0

lim lim f (x, y) = lim (y/3y) = 1/3

y0 x0

y0

lim lim f (x, y) = lim (1 1) = 0

f (x, y) = 1 + tanh(x + y) :

x y

lim lim f (x, y) = lim (1 + 1) = 2

y x

5.2.4. Rules for limits of composite expressions


(i) If limxx0 f (x) = a, limxx0 g(x) = b:


lim f (x) + g(x) = a + b


xx
0

lim f (x)g(x) = ab
xx
0

if b 6= 0 :

lim f (x)/g(x) = a/b


xx
0

if b = 0 and a 6= 0 :

lim f (x)/g(x)
xx

does not exist

(ii) If limxx0 f (x) = a, limya g(y) = b, g(a) = b:


lim g(f (x)) = b

xx0

(iii) determine limits via pinching or sandwiching:


Let there be a > 0 such that
(x, |xx0 | < ) : f (x) g(x) h(x)
limxx0 f (x) = limxx0 h(x) = a

lim g(x) = a

xx0

71
5.2.5. Examples
0.5

x sin(x1 )

(i) limx0 x sin(x ) = 0


proof: via sandwiching,
since always sin(. . .) [1, 1],
(x IR) :

|x| x sin(x1 ) |x|

0.0

Clearly: limx0 |x| = 0 and limx0 (|x|) = 0,


hence limx0 x sin(x1 ) = 0
-0.5
-0.5

(ii) limx0 x

0.0

tan(x) = 1

proof:
use limx0 x1 sin(x) = 1 and limx0 cos(x) = cos(0) = 1
!
!
sin(x)
sin(x)
1
1
x tan(x) =
=
x cos(x)
x
cos(x)
Hence
!
!
sin(x)
1
1
lim x tan(x) = lim
lim
= 1.1 = 1
x0
x0
x0 cos(x)
x
(iii) limx0 x1 ln(1 + x) = 1
proof:
use suitable substitution, e.g. x = ey 1 with y 0
as well as limx0 x1 (ex 1) = 1


x1 ln(1 + x) = (ey 1)1 ln(ey ) = (ey 1)/y

1

Hence


lim x1 ln(1 + x) = lim (ey 1)/y

x0

y0

1

= 11 = 1

(iv) limx x ln(1 + x1 ) = 1


proof:
substitute x = 1/y with y 0, and convert into limit (iii)
1
x ln(1 + ) = y 1 ln(1 + y)
x
Hence
1
lim x ln(1 + ) = lim y 1 ln(1 + y) = 1
x
y0
x

0.5

72
(v) limx0 x ln(x) = 0
proof:
more tricky, substitute x = ey with y and think ...
x ln(x) = yey
Proof of limy yey = 0, in three steps without using power series:
(a) claim: ez > z for all z > 0.
d
f (z) = ez 1 > 0 for z > 0
proof: consider f (z) = ez z for z 0: f (0) = 1, dz
so f increases monotonically for z > 0, starting at f (0) = 1
Thus ez > z for all z > 0
(b) now choose z = 21 y: ey/2 > y/2, so also ey > (y/2)2 = 14 y 2
equivalently: yey < y/( 14 y 2 ) = 4/y
(c) since also yey 0 for y > 0, we can proceed by sandwiching:
y>0:

0 yey 4/y

Since limy (4/y) = 0, we have proven limy yey = 0


and hence also the original statement.
Alternative version of the proof that limy yey = 0, using power series:
n
y
`
(a) since ey =
n=0 y /n!, one has e > y /`! for any ` IN and any y > 0
(b) hence we know for y > 0 that 0 yey < `! y 1`
(c) take any ` > 1 and proceed by sandwiching ...

(vi) limx x1 ln(x) = 0


proof:
substitute x = 1/y with y 0, and convert into limit (v)
lim x1 ln(x) = lim y ln(y) = 0

y0

This completes the proof.


(vii) limx xa ex = 0 for any a IR
proof:
for a 0 the claim is trivial, so we concentrate on a > 0
we generalize the ideas used in proving (v)
Proof without using power series:

73
(a) we know that ez > z for all z > 0 (was demonstrated under (iv))
(b) now choose z = x/2a: ex/2a > x/2a, so also ex > (x/2a)2a = (2a)2a x2a
equivalently: xa ex < xa /(2a)2a x2a = (2a)2a xa
(c) we can proceed by sandwiching:
x>0:

0 xa ex (2a)2a xa

Since limx xa = 0, we have proven limx xa ex = 0


This completes the proof.
Alternative version using power series:
n
x
`
(a) since ex =
n=0 x /n!, one has e > x /`! for any ` IN and any x > 0
(b) hence we know for x > 0 that 0 xa ex < `! xa`
(c) take any ` > a and proceed by sandwiching ...

5.3. Differentiation
5.3.1. Derivatives of functions
Recall definition of derivative of function f for continuous functions:
(notation: df /dx, or f 0 (x), or dy/dx with y = f (x))
f (x + h) f (x)
f 0 (x) = lim
h0
h
calculation of f 0 (x) from scratch:
first simplify (f (x + h) f (x))/h if possible
if putting h = 0 in the formula is allowed (Fermat)
then the result will be f 0 (x). Done.
If putting h = 0 is not allowed, we must determine the limit:
(i) decompose your expression into parts that have known limits,
then use the rules for limits of composite expressions,
or
(ii) substituting suitable power series for the difficult parts,
and simplify further until you can put h = 0
Note:
the formulas are used when you prove that a limit takes a value L
(they do not help in finding the candidate L in the first place ...)

74
Elementary derivatives found earlier:
d
d
sin(x) = cos(x)
cos(x) = sin(x)
dx
dx
d x
a = ax ln(a)
dx

d n
x = nxn1 (n ZZ+ )
dx

Further examples:

d
dx

ln(x) = 1/x for x IR+

proof:


ln x(1 + h/x) ln(x)

ln(x + h) ln(x)
d
ln(x) = lim
= lim
h0
h0
dx
h
h
ln(x) + ln(1 + h/x) ln(x)
ln(1 + h/x)
= lim
= lim
substitute y = h/x
h0
h0
h
h
ln(1 + y)
1
ln(1 + y)
1
= lim
= lim
=
use limit (iii) of section 4.25
y0
xy
x y0
y
x

d a
x
dx

= axa1 for a IR, x 6= 0


(so far only proven for a ZZ+ )
proof:
d a
(x + h)a xa
xa (1 + h/x)a xa
x = lim
= lim
h0
h0
dx
h
h
a
(1
+
h/x)

1
= xa lim
= xa lim h1 [ea ln(1+h/x) 1] substitute y = h/x
h0
h0
h


ay y 1 ln(1+y)

= xa lim (xy)1 [ea ln(1+y) 1] = axa1 lim (ay)1 [e


y0

y0

= ax

a1

lim z [e

z limy0 y 1 ln(1+y)

z0

1]

= axa1 lim z 1 [ez 1] = axa1


z0

5.3.2. Rules for derivatives of composite expressions


Objective: efficiency
(we dont want to calculate derivatives always from scratch, via the limit)
(i) The sum rule:
y = f (x) + g(x) :

dy
= f 0 (x) + g 0 (x)
dx

1]

75
proof:
f (x + h) + g(x + h) f (x) g(x)
d
[f (x) + g(x)] = lim
h0
dx
h)
(
(
)
f (x + h) f (x)
g(x + h) g(x)
= lim
+ lim
h0
h0
h
h
= f 0 (x) + g 0 (x)

(ii) The product rule:


y = f (x)g(x) :

dy
= f 0 (x)g(x) + f (x)g 0 (x)
dx

proof:
d
f (x + h)g(x + h) f (x)g(x)
[f (x)g(x)] = lim
h0
dx
h
[f (x + h) f (x)]g(x + h) + [g(x + h) g(x)]f (x)
= lim
h0
h)
(
)
(
g(x + h) g(x)
f (x + h) f (x)
+ lim f (x)
= lim g(x + h)
h0
h0
h
h
= g(x)f 0 (x) + f (x)g 0 (x)

(iii) The chain rule:


y = f (g(x)) :

dy
= f 0 (g(x))g 0 (x)
dx

proof:
d
f (g(x + h)) f (g(x))
f (g(x)) = lim
h0
dx
h
f (g(x + h)) f (g(x)) g(x + h) g(x)
= lim
h0
g(x + h) g(x)
h
(
)(
)
f (g(x + h)) f (g(x))
g(x + h) g(x)
lim
= lim
h0
h0
g(x + h) g(x)
h
f (g(x + h)) f (g(x))
= g 0 (x) lim
substitute = g(x + h) g(x)
h0
g(x + h) g(x)
f (g(x) + ) f (g(x))
= g 0 (x) lim
= g 0 (x)f 0 (g(x))
0

(some subtleties with this version,


leave that to analysis)

76
(iv) The quotient rule:
f 0 (x)g(x) f (x)g 0 (x)
dy
=
dx
[g(x)]2

y = f (x)/g(x) :

proof:
write f (x)g 1 (x), use product rule, chain rule, and

d 1
x
dx

= x2

i
d h
d h 1 i
f (x)g 1 (x) = f 0 (x)g 1 (x) + f (x)
g (x)
dx
dx
h

= f 0 (x)g 1 (x) + f (x)g 0 (x) 1/g 2 (x) =

f 0 (x)g(x) f (x)g 0 (x)


g 2 (x)

Examples:
"

"

i
d h 2
d
d 2
x sin(x) = x2
sin(x) + sin(x)
x
dx
dx
dx
= x2 cos(x) + 2x sin(x)

d
f (ax)
dx
d xn
e
dx

= af 0 (ax)

=e

xn

d n
n
x = nxn1 ex
dx

1
sinh(x)
d
ln(cosh(x)) = cosh0 (x).
=
= tanh(x)
dx
cosh(x)
cosh(x)
d x
x
dx

d ln(xx )
d x ln(x)
e
=
e
dx
dx
"
#
d
d
= ex ln(x) [x ln(x)] = ex ln(x) 1. ln(x) + x ln(x)
dx
dx


x
= xx [1 + ln(x)]
= ex ln(x) ln(x) +
x
=

d
d
f (g(h(x))) = f 0 (g(h(x)))
g(h(x))
dx
dx
d
= f 0 (g(h(x))) g 0 (h(x))
h(x)
dx
= f 0 (g(h(x))) g 0 (h(x)) h0 (x)

77
5.3.3. Derivatives of implicit functions
Above methods apply only when we can write an
explicit formula for the function f (x) we wish to differentiate.
But: some functions are defined by a property, without a formula ...
A. functions defined as solutions of equations for points in the plane
function y(x): solution of an equation of the type F (x, y) = 0
(for a domain D where each x D is associated with only one y)
How to find y 0 (x) ?
method:
(i) differentiate the equation, using chain rule
(ii) then try to solve the result for dy/dx
y
10

Example 1:

y + sin(y) x = 0

8
6

F (x, y) = y + sin(y) x,
so y(x) is solution of y + sin(y) x = 0
(i) differentiate F (x, y):
dy
dy
+ cos(y) 1 = 0
dx
dx
(ii) solve for dy/dx:
dy
(1 + cos(y)) = 1
dx

4
2

0
-2
-4
-6
-8

dy
1
=
dx
1 + cos(y)

-10
-10

-8

-6

-4

-2

10

y
Example 2:

1.2

y tanh(xy) = 0

1.0

F (x, y) = y tanh(xy),
so y(x) is solution of y tanh(xy) = 0
(i) differentiate F (x, y):
dy
d
tanh0 (xy) (xy) = 0
dx
dx
!
dy
1
dy
x +y =0

dx cosh2 (xy)
dx
(ii) solve for dy/dx:
!
dy
x
y
1

=0
2
dx
cosh (xy)
cosh2 (xy)

0.8

0.6

0.4

0.2

0.0
0

dy
y
=
2
dx
cosh (xy) x

78
B. functions defined as inverse of another given function
function f 1 (x): solution of the equation f 1 (f (x)) = x for all x, with given f
Suppose we have no formula for f 1 (x), e.g. arcsin(x)
d 1
f (x) ?
How to find dx
method:
(i) differentiate the equivalent equation f (f 1 (x)) = x, using chain rule
d 1
(ii) then solve the result for dx
f (x)
This can be done generally:
d
f (f 1 (x)) = 1
dx
so
1
d 1
f (x) = 0 1
dx
f (f (x))

d 1
f (x) f 0 (f 1 (x)) = 1
dx

Example 1:
let f (x) = ex , so f 1 (x) = ln(x) (with x > 0)
d
1
1
1
ln(x) = 0
= ln(x) =
dx
f (ln(x))
e
x

Example 2:
let f (x) = sin(x), so f 1 (x) = arcsin(x) (with x [ 2 , 2 ])
d
1
1
1
arcsin(x) = 0
=
=q
dx
f (arcsin(x))
cos(arcsin(x))
1 sin2 (arcsin(x))
1
=
1 x2
Example 3:
let f (x) = tan(x), so f 1 (x) = arctan(x)
d
1
arctan(x) = 0
= cos2 (arctan(x))
dx
f (arctan(x))
cos2 (arctan(x))
1
=
=
2
2
2
tan (arctan(x)) + 1
sin (arctan(x)) + cos (arctan(x))
1
=
1 + x2

79
C. functions defined parametrically
Given two functions x(t) and y(t), varying t traces out a curve in the x-y plane
This curve defines a function y(x) implicitly
(for those t IR where each x is associated with only one y)
How to find y 0 (x) ?
method:
dy/dt
dy
(i) calculate x0 (t) = dx/dt and y 0 (t) = dy/dt, then work out dx
= dx/dt
(ii) if possible, use formulas for x(t) and y(t) to eliminate t from your result

Example 1:
for t IR:

y
1.0

x(t) = t + cos(t)
y(t) = ln(cosh(sin(t)))

(i) differentiate x(t) and y(t):

0.5

0.0

x (t) = 1 sin(t)
y 0 (t) = cos(t) tanh(sin(t))
-0.5

so

-8

-4

dy
cos(t) tanh(sin(t))
=
dx
1 sin(t)
(ii) cannot simplify further

Example 2:
for t IR:

y
20

15

x(t) = et

10

y(t) = tan(t)

(i) differentiate x(t) and y(t):


x0 (t) = et
y 0 (t) = 1/ cos2 (t)
so

-5

-10

-15

-20
-10

10

dy
e
=
dx
cos2 (t)
(ii) simplify by removing t:
dy
et [sin2 (t) + cos2 (t)]
tan2 (t) + 1
1 + y2
=
=
=
dx
cos2 (t)
et
x

20

30

40

50

60

70

80

90

100

80
Finally ...
dy
Are we at all allowed to put dx
=
(even if these derivatives exist)

dy/dt
dx/dt

Proof 1:
(i) Assume that y can indeed be written as a function (as yet unknown) of x: y(t) = f (x(t))
We aim to calculate dy/dx = f 0 (x)
(ii) Apply chain rule to y(t) = f (x(t)):
0

y (t) = f (x(t))x (t)


Hence f 0 (x) =

dy/dt
,
dx/dt

y 0 (t)
= f 0 (x(t))
0
x (t)

as claimed.

Proof 2:
Via the original definition:
x0 (t) = limh0 h1 [x(t + h) x(t)]
y 0 (t) = limh0 h1 [y(t + h) y(t)]

y(t + h) y(t)
y 0 (t)
= lim
0
h0
x (t)
x(t + h) x(t)

In the limit we switch from h to the new variable z = x(t + h) x(t).


Since h 0, also z 0.
If we write y(t) = f (x(t)) for some unknown function f ,
then y(t + h) = f (x(t + h)) = f (x(t) + z):
f (x(t) + z) f (x(t))
y(t + h) y(t)
= lim
= f 0 (x(t))
lim
z0
h0 x(t + h) x(t)
z
Hence
y 0 (t)
= f 0 (x(t))
0
x (t)

5.3.4. Applications of derivative: sketching graphs


standard procedure for sketching the graph of a function f (x)
(a reminder should be secondary school knowledge)
(i) Determine values of x for which f is defined, and those for which it isnt
(ii) Find all stationary points of f (i.e. those x where f 0 (x) = 0, with zero slope)
(iii) Determine the nature of the stationary points (local minimum, local maximum, or neither)
(iv) Find the points where f (x) = 0, and calculate f 0 (x) at these points
(v) State whether f is even, f (x) = f (x) for all x, or odd, f (x) = f (x) for all x, or neither
(vi) Sketch the graph of f over an appropriate range of x

81
6. Integration
6.1. Introduction
6.1.1. Area under a curve
We define the integral (including notation) in terms of areas:
definition
The integral ab f (x)dx is the total area in the (x, y) plane between the graph of y = f (x) and
the x-axis, counted positively for f (x) > 0 (area above the x-axis) and negatively for f (x) < 0
(area below the x-axis).
R

How to calculate

Rb
a

f (x)dx ?

The area is easy to calculate


for functions that are built
of steps only (staircases):

1.2
1.0

f (x)

0.8

Denote jump points as x` ,


with ` ZZ and
x1 < x2 < x3 < . . .

0.6

If x [x` , x`+1 ) : f (x) = f`

0.0

0.4

f`

0.2

x
x` x`+1

-0.2

Contribution to integral
from interval [x` , x`+1 ):

-0.4

area of rectangle, f` (x`+1 x` )


also sign is correct!

-0.8

-0.6

-1.0
-1.2

Total integral:
add up contributions of
all intervals between a and b
(let x1 = a, xL = b)

Z b
a

f (x)dx =

L1
X
`=1

f` (x`+1 x` )

82
Notes:
(i) previous formula is exact only for staircases
(ii) but we can approximate most functions to arbitrary accuracy
using staircases with smaller and smaller steps ... (limits!!)

What if f (x) is not a staircase?

1.2
1.0

Rb

Find the integral a f (x)dx


as the limit of the integral
for a suitable staircase,
where all steps go to zero

f (x)

0.8
0.6
0.4

f`

0.2

0.0

Not as simple as it sounds:


(i) many staircase possible ...
(ii) what is a suitable staircase?
(iii) result ought not to depend on
your choice of staircase!
(analysis gets involved)

x` x`+1

-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
0

Formal definition of integral (analysis) therefore involves


(i) consider all possible staircases where each step touches the curve of f (x)
(ii) find area underneath each staircase in the limit width of the widest step goes to zero
(iii) check whether all these limits for different staircases are identical

Result in a nutshell:
If f (x) is a continuous function on [a, b], then the integral
Rb

Rb
a

f (x)dx exists

If the integral a f (x)dx exists, it will be equal to the limit of any staircase approximation,
where each step touches the curve of f (x), as the width of the widest step goes to zero
Consequence:
to calculate the integral of a continuous function
we can choose the most convenient staircase approximation
and then find its limit

83
Direct route (nor relying on analysis results): sandwich method using staircases
Step (i):
build staircase functions f (x) such that
f (x) f (x) f+ (x) for all x [a, b]
e.g.
f+ (x) = f`+ = maxx[x` ,x`+1 ) f (x)

x [x` , x`+1 ) :

f (x) = f` = minx[x` ,x`+1 ) f (x)

(let x1 = a, xL = b)

1.2

1.2

1.0

1.0

staircase: f (x)

0.8

1.2
1.0

f (x)

0.8

0.6

0.6

0.6

0.4

0.4

0.4

0.2

0.2

0.2

0.0

0.0

0.0

-0.2

-0.2

-0.2

-0.4

-0.4

-0.4

-0.6

-0.6

-0.6

-0.8

-0.8

-0.8

-1.0

-1.0

-1.0

-1.2

-1.2
0

staircase: f+ (x)

0.8

-1.2

we are now sure that:


Z b
a

f (x)dx

Z b

f (x)dx

Z b
a

f+ (x)dx

i.e.
L1
X

Z b

f` (x`+1 x` )

f (x)dx

`=1

L1
X

f`+ (x`+1 x` )

`=1

Step (ii):
take the limit where x` x`+1 0 for all `,
PL1
P
+
in the two bounding areas A = `=1
f` (x`+1 x` ) and A+ = L1
`=1 f` (x`+1 x` )
lim

step widths 0

Z b

f (x)dx

lim

step widths 0

A+

Step (iii):
Conclusion:
if limstep widths

A = limstep

widths 0

A+ = A, then:

Rb
a

f (x)dx = A

84
6.1.2. Examples of integrals calculated via staircases
Example 1:
A=

Z b

cos(x)dx,

with b

(so on [0, b]: cos(x) decreases monotonically, i.e. if x0 > x then cos(x0 ) < cos(x))
Method: sandwich with staircases
use tutorial exercise 35 (let m ZZ):
n
X

6= 2m :

cos(k) =

k=0

1cos(n+)cos()+cos(n)
2 2 cos()

Step (i):
Build staircase functions f (x) such that
f (x) cos(x) f+ (x) for all x [0, b]
e.g.
f+ (x) = f`+ = maxx[x` ,x`+1 ) cos(x) = cos(x` )

x [x` , x`+1 ) :

f (x) = f` = minx[x` ,x`+1 ) cos(x) = cos(x`+1 )

Choose steps of equal size,


with x1 = 0 and xL = b:
x` = (`1)h, with h =

b
:
L1

x1 = 0, x2 = h, x3 = 2h, . . . xL = (L1)h = b

Upper bound to A:
A+

L1
X

f`+ (x`+1 x` ) = h

`=1
L2
X

cos(x` ) = h

`=1

L1
X

cos((` 1)h)

`=1

1cos((L1)h)cos(h)+cos((L2)h)
2 2 cos(h)
k=0
1cos(b)cos(h)+cos(b h)
=h
2 2 cos(h)
1cos(b)cos(h)+cos(b) cos(h)+sin(b) sin(h)
=h
2 2 cos(h)
[1cos(b)][1cos(h)]+sin(b) sin(h)
=h
2 2 cos(h)


h
1
h sin(h)
=
1cos(b) + sin(b)
2
2
1cos(h)
=h

eliminate L :

L1
X

cos(kh) = h

Lower bound to A:
A =

L1
X
`=1

f` (x`+1 x` ) = h

L1
X
`=1

cos(x`+1 ) = h

L1
X
`=1

cos(`h)

85
=h

L
X

cos((k 1)h) = h

"L1
X

k=2

cos((k 1)h) + cos((L1)h) cos(0)

k=1

= A+ + h cos((L1)h) h = A+ + h[cos(b) 1]

h
h sin(h)
1
= h[cos(b) 1] + 1cos(b) + sin(b)
2
2
1cos(h)
We now know that A

Rb
0

cos(x)dx A+ .

Step (ii):
take the limit h 0 in the bounds A+ and A
(


h
1
h sin(h)
lim A+ = lim
1cos(b) + sin(b)
h0
h0 2
2
1cos(h)
(
)
1
h sin(h)
1
h sin(h)
= lim sin(b)
= sin(b) lim
h0 2
h0 2 1cos2 (h/2)+sin2 (h/2)
1cos(h)
(
)
(
)
h sin(h)
sin(h) (h/2)2
= sin(b) lim
= sin(b) lim
h0 4 sin2 (h/2)
h0
h sin2 (h/2)
!2

sin(h)
h/2
= = sin(b) lim
lim
= sin(b)
h0
h0 sin(h/2)
h
lim A = lim {A+ + h[cos(b) 1]} = lim A+ = sin(b)

h0

h0

h0

We conclude
lim A A lim A+

h0

h0

sin(b) A sin(b) A = sin(b)

Example 2:
A=

Z b

cos(x)dx,

with arbitrary b > 0

(cos(x) need no longer decrease monotonically, so sandwich method becomes messy)


Method: use most convenient approximating staircase
(i) cos(x) is a continuous function on any interval
(ii) both f (x) and f+ (x) in example 1 are suitable approximating staircases to cos(x)
(each step in each staircase touches the curve, and all step sizes go to zero)
Hence:
A = lim A = lim A+ = sin(b)
h0

h0

86
Example 3:
A=

Z b

with b /2

sin(x)dx,

(so on [0, b]: sin(x) increases monotonically, i.e. if x0 > x then sin(x0 ) > sin(x))
Method: sandwich with staircases
we use tutorial exercise 39 (let ` ZZ):
n
X

6= 2` :

sin(k) =

k=0

sin(n+)+sin()+sin(n)
2 2 cos()

Step (i):
Build staircase functions f (x) such that
f (x) sin(x) f+ (x) for all x [0, b]
e.g.
f+ (x) = f`+ = maxx[x` ,x`+1 ) sin(x) = sin(x`+1 )

x [x` , x`+1 ) :

f (x) = f` = minx[x` ,x`+1 ) sin(x) = sin(x` )

Choose steps of equal size,


with x1 = 0 and xL = b:
x` = (`1)h, with h =

b
:
L1

x1 = 0, x2 = h, x3 = 2h, . . . xL = (L1)h = b

Upper bound to A:
A+

L1
X

f`+ (x`+1 x` ) = h

sin(x`+1 ) = h

L1
X

sin(`h)

`=1

`=1

`=1
L1
X

sin(Lh)+sin(h)+sin((L1)h)
sin(kh) = h
2 2 cos(h)
k=0
sin(b + h)+sin(h)+sin(b)
=h
2 2 cos(h)
sin(b) cos(h)cos(b) sin(h)+sin(h)+sin(b)
=h
2 2 cos(h)
[1cos(b)] sin(h) + [1cos(h)] sin(b)
=h
2 2 cos(h)
h
1
h sin(h)
= sin(b) + [1cos(b)]
2
2
1 cos(h)
= =h

eliminate L :

L1
X

Lower bound to A:
A =

L1
X

f` (x`+1 x` ) = h

`=1
L2
X

=h

k=0

L1
X
`=1

sin(kh) = h

L1
X
k=1

sin(x` ) = h

L1
X

sin((`1)h)

`=1

sin(kh) h sin((L1)h) = A+ h sin(b)

87
We now know that A

Rb
0

sin(x)dx A+ .

Step (ii):
take the limit h 0 in the bounds A+ and A
(
)
h
1
h sin(h)
lim A+ = lim
sin(b) + [1cos(b)]
h0
h0 2
2
1 cos(h)
(
)
h sin(h)
1
h sin(h)
1
= [1cos(b)] lim
= lim [1cos(b)]
h0 2 1cos2 (h/2)+sin2 (h/2)
h0 2
1 cos(h)
)
(
)
(
sin(h) (h/2)2
h sin(h)
= [1cos(b)] lim
= [1cos(b)] lim
h0
h0 4 sin2 (h/2)
h sin2 (h/2)
!

!2

sin(h)
h/2
= = [1cos(b)] lim
lim
h0
h0 sin(h/2)
h
lim A = lim {A+ h sin(b)} = lim A+ = 1cos(b)

h0

h0

= 1cos(b)

h0

We conclude
lim A A lim A+ 1cos(b) A 1cos(b) A = 1cos(b)

h0

h0

Example 4:
A=

Z b

sin(x)dx,

with arbitrary b

(sin(x) need no longer increase monotonically, so sandwich method becomes more messy)
Method: use most convenient approximating staircase
(i) sin(x) is a continuous function on any interval
(ii) both f (x) and f+ (x) in example 3 are suitable approximating staircases to sin(x)
(each step in each staircase touches the curve, and all step sizes go to zero)
Hence:
A = lim A = lim A+ = 1 cos(b)
h0

h0

Example 5:
A=

Z b

xn dx,

with a < b and n ZZ+

(so we know the function increases monotonically)


Method: sandwich with staircases
we use tutorial exercise 39 (let ` ZZ):
n
X
1 z n+1
z 6= 1 :
zk =
1z
k=0

88
Step (i):
Build staircase functions f (x) such that
f (x) xn f+ (x) for all x [a, b]
e.g.
x [x` , x`+1 ) :

f+ (x) = f`+ = maxx[x` ,x`+1 ) xn = xn`+1


f (x) = f` = minx[x` ,x`+1 ) xn = xn`

Choose steps of non-equal size, in a so-called geometric progression:


with x1 = 0 and xL = b:
x` = ah`1 , with b = ahL1 :

x1 = a, x2 = ah, x3 = ah2 , . . . xL = ahL1 = b

Note:
(i) h > 1 (since we require x`+1 > x` always)
(ii) ln(b/a) = (L1) ln(h), so L 1 = [ln(b) ln(a)]/ ln(h)
(iii) here x`+1 x` = ah` ah`1 = ah`1 (h 1)
(iv) largest step size: xL xL1 = ahL2 (h 1) = ahL1 (1 h1 ) = b(1 h1 )
(v) limit of zero step sizes: h 1
Upper bound to A:
A+ =

L1
X

f`+ (x`+1

x` ) =

= an+1 (h 1)

L1
X

1) = a(h 1)

h`n+`1 = an+1 (1 h1 )

`=0

n+1 n

`=1
L2
X

n+1 k

n+1 n

n+1 n

k=0
L2
X

n+1 k

n+1 n

h (h 1)
h (h 1)

[hn+1 ]` = an+1 (1 h1 )
[h
[h

] =a

] =a

k=0

[ah` ]n h`1

h`(n+1)

`=1
L2
X

=a

L1
X
`=1

L1
X

`=1
L1
X
1

= an+1 (1 h )

eliminate L :

xn`+1 ah`1 (h

`=1

`=1

=a

L1
X

[hn+1 ]`+1

1 h(n+1)(L1)
h (h 1)
1 hn+1
1 (b/a)n+1
h (h 1)
1 hn+1

Lower bound to A:
A =

L1
X
`=1
n

=h

f` (x`+1 x` ) =

L1
X
`=1

xn` ah`1 (h 1) = a(h 1)

L1
X

[ah`1 ]n h`1

`=1

A+

We now know that A

Rb n
x dx
a

A+ .

Step (ii):
take the limit h 1 in the bounds A+ and A
(
!)


1 (b/a)n+1
hn (h 1)
n+1 n
n+1
n+1
lim A+ = lim a h (h 1)
=
a

b
lim
h1
h1
h1 1 hn+1
1 hn+1

89
!

h1
lim n+1
= b
a
lim h
substitute h = ey with y 0
h1 h
h1
1



ey 1
= bn+1 an+1 lim (n+1)y
y0 e
1
!
y


y
e 1 (n+1)y
n+1
n+1
= b
a
lim
y0
y e(n+1)y 1 (n+1)y
!
!


1  n+1
ey 1
(n+1)y
1  n+1
n+1
=
b
a
lim
lim (n+1)y
=
b
an+1
y0
y0 e
n+1
y
1
n+1


n+1

n+1



lim A = lim hn A+ = lim A+ =


h1

h1

h1


1  n+1
b
an+1
n+1

We conclude
lim A A lim A+
h1

hence A =

1
n+1

h1

1  n+1 n+1 
1  n+1 n+1 
b a
A
b a
n+1
n+1

bn+1 an+1 .

6.1.3. Fundamental theorems of calculus: integration vs differentiation


Any sane person would like to find a more efficient way to calculate integrals
than via the (tedious) construction and analysis of bounding staircases ...
Furthermore: we need explicit expressions for the relevant sums to do it ...
help is at hand:
First fundamental theorem of calculus:
If a function f : [a, b] IR is continuous on the interval [a, b],
and F : [a, b] IR is another function such that
F 0 (x) = f (x) for all x [a, b], then
Z b

f (x)dx = F (b) F (a)

Second fundamental theorem of calculus:


If a function f : [a, b] IR is continuous on the interval [a, b],
and F : [a, b] IR is another function defined by
F (x) =

Z x

f (t)dt

for all x [a, b]

then F 0 (x) = f (x) for all x [a, b].

90
Consequence: a new way to calculate integrals!
Need to find a function F (x) such that F 0 (x) = f (x)
But where do these theorems come from?
Sketch of proofs:

Theorem II:
If a function f : [a, b] IR is continuous on the interval [a, b],
and F : [a, b] IR is another function defined by
F (x) =

Z x

f (t)dt

for all x [a, b]

then F 0 (x) = f (x) for all x [a, b].


sketch of proof:
(for differentiable f , for simplicity; proof can be done more generally!)
Let C = maxx[a,b] |f 0 (x)| (maximum slope in absolute sense):
#

"

Z x
dF (x)
F (x + h) F (x)
1 Z x+h
f (t)dt
f (t)dt
= lim
= lim
h0
h0 h
dx
h
a
a
1 Z x+h
= lim
f (t)dt
h0 h x
Easy to bound:
on t [x, x + h] one has f (x) Ch f (t) f (x) + Ch, so
Z x+h

[f (x) Ch]dt

f (t)dt

h[f (x) Ch]

Z x+h

Z x+h

Z x+h

[f (x) + Ch]dt

f (t)dt h[f (x) + Ch]

1 Z x+h
lim[f (x) Ch] lim
f (t)dt lim[f (x) + Ch]
h0
h0 h x
h0
Hence
1 Z x+h
F (x) = lim
f (t)dt = f (x)
h0 h x
0

why was this a sketch rather than formal proof?


(i) restriction to differentiable f can be lifted
(ii) should also do limh0 in dF (x)/dx

91
Theorem I:
If a function f : [a, b] IR is continuous on the interval [a, b],
and F : [a, b] IR is another function such that
F 0 (x) = f (x) for all x [a, b], then
Z b

f (x)dx = F (b) F (a)

proof:
in three steps:
(i) preparation:
R
Since f is continuous on [a, b] the integral ax f (t)dt exists for x [a, b].
R
We define G(x) = ax f (t)dt
Properties of G:
from Theorem II : G0 (x) = f (x) for all x [a, b]
from definition :

G(a) = 0

Clearly (using definition of G!): ab f (t)dt = G(b)


We are given another function F with F 0 (x) = f (x) for all x [a, b]
R

(ii) We next show that now G(x) = F (x) F (a) for all x [a, b]
How to show this? Call the difference between F and G: H(x) = F (x) G(x)
d
H(x) = f (x) f (x) = 0 for all x [a, b]
It follows that dx
Hence the function H(x) has zero slope on [a, b],
i.e. its graph is horizontal, so H(x) = H(a) for all x [a, b]
Consequence: F (x) G(x) = F (a) G(a) for all x [a, b],
Consequence (since G(a) = 0): G(x) = F (x) F (a) for all x [a, b]
(iii) Combine previous two intermediate results:

Rb
a

f (t)dt = G(b) = F (b) F (a)

6.1.4. Indefinite and definite integrals, and other conventions


Some final notation conventions and terminology:
Definite integral: A = ab f (x)dx
(the object we worked with so far, always a number)
definition: see earlier
R

Indefinite integral: F (x) = f (x)dx


(i.e. without any boundary values indicated, always a function)
R

92
definition: any function F such that F 0 (x) = f (x)
Not uniquely defined: the solutions F differ by a constant, since

d
C
dx

=0

Other names: primitive of f , or anti-derivative of f


Doing an integral via Theorem I requires finding the primitive F (x) of f (x).
Often we emphasize this intermediate step, show how the integral was done
and allow the reader to verify that F 0 (x) = f (x), by writing
Z b
a

f (x)dx = [F (x)]ba = F (b) F (a)

So far we defined ab f (x)dx for a b


Generalization to a > b?
R

Logical choice: via Theorem I


(which will then hold for any a, b IR)
a > b : define

Z b

f (x)dx = F (b) F (a) = (F (a) F (b)) =

Z a

f (x)dx

6.2. Techniques of integration


Integration now more or less boils down to this:
when given a function f (x), find a function F (x) (the primitive) such that F 0 (x) = f (x)
Integration is therefore an art rather than a science:
in contrast to differentiation, where you just follow (carefully) a set of clear rules,
integration mostly relies on a mixture of skill, experience, memory and intuition
The basic strategy: divide and conquer
manipulate (break up, simplify) the integral until it has been reduced to
expressions of which you know (i.e. remember) the primitives
So to be a successful integrator you need to
(i) know and practice the formal manipulation rules and other tricks
(ii) memorize a list of basic primitives
(iii) be creative
We must now
(i) agree on the list of elementary integrals that we will consider known
(ii) describe the tools for manipulation and simplification (most based on rules for
differentiation) that we can use to reduce our problem to elementary integrals

93
6.2.1. List of elementary integrals and general methods for reduction
Ten elementary integrals
f (x)

F (x)

xa (a 6= 1)
x1

1
xa+1
a+1
ln |x|

ln(x)

x ln(x) x

ex

ex

cos(x)

sin(x)

sin(x)
1

1 x2
1
1 + x2
1

2
x 1
1

x2 + 1

cos(x)
arcsin(x)
arctan(x)
arccosh(x)
arcsinh(x)

Four general integration rules


R

let F (x) = f (x)dx, G(x) = g(x)dx, etc


Z 

(i) Linearity :

Z 

(ii) Sum rule :

Z 

(iii) Integration by parts :

(iv) Integration by substitution :

cf (x) dx = cF (x)


f (x) + g(x) = F (x) + G(x)




f (x)G(x) dx = F (x)G(x)

f (x)dx =

Z 

f (x(t))

proofs:
(i) consequence of:

d
dx

(ii) consequence of:

d
dx

(iii) consequence of:

d
F (x)G(x) = F 0 (x)G(x)
dx
d
F (x(t)) = x0 (t)F 0 (x(t))
dt

(iv) consequence of:

cF (x) = cF 0 (x)


F (x) + G(x) = F 0 (x) + G0 (x)




+ G0 (x)F (x)

dx 
dt
dt

Z 

g(x)F (x) dx

94
Note:
(iii) and (iv) are usually given as identities for definite integrals,
Z b

Integration by parts :

ib

f (x)G(x) dx = F (x)G(x)

Integration by substitution :

Z b

f (x)dx =

Z t(b) 

f (x(t))

t(a)

Z b

g(x)F (x) dx

dx 
dt
dt

Let us inspect the validity of (iv) more carefully:


Rb

f (x)dx = F (b) F (a), one knows generally that


d Zb
f (x)dx = F 0 (b) = f (b)
db a
Subtract the left-hand and the right-hand side in the claimed equality,
Z b
Z t(b) 
dx 
I(b) = LHS RHS =
f (x)dx
f (x(t))
dt
dt
a
t(a)
Then calculate dI/db, via the chain rule:

d Z t(b) 
dI
d Zb
f (x)dx
f (x(t))x0 (t) dt
=
db
db a
db t(a)

d Z t(b) 
0
f (x(t))x0 (t) dt
= f (b) t (b)
dt(b) t(a)
Since

= f (b) t0 (b) f (x(t))x0 (t)


0

t=t(b)

= f (b) f (b) t (b) x (t(b)) = f (b) f (b) = 0


d
where we used [t0 (b)x0 (t)]t=t(b) = db
x(t(b)) =
we may choose b = a to calculate it, i.e.

LHS RHS = I(a) =

Z a
a

d
(b)
db

f (x)dx

= 1. Thus I(b) is independent of b. Hence

Z t(a) 

f (x(t))

t(a)

dx 
dt = 0 0 = 0
dt

This completes the proof that LHS=RHS for all a and b,


i.e. the claimed identity (iv) is indeed true.

Warning:
changing variables via substitution must involve a one-to-one transformation
(i.e. a unique t for every x, and vice versa)
Example: suppose we are not careful
R1
Clearly 1
dx = 2. Now do the integral via substitution of y = x2 , with dy/dx = 2x
Z 1
Z y=1
dy
dx =
= 0 ...
1
y=1 2x(y)

95
6.2.2. Examples: integration by substitution

sinm (x) cos(x)dx:

substitute t = sin(x), so dt/dx = cos(x) i.e. dt = cos(x)dx


Z
Z
1 m+1
1
sinm (x) cos(x)dx = tm dt =
t
=
sinm+1 (x)
m+1
m+1

R x2
e x

dx:

sunstitute t = x2 , so dt/dx = 2x i.e. dt = 2xdx


Z
1Z t
1
1 2
2
ex x dx =
e dt = et = ex
2
2
2

(1 x2 )1/2 dx:

try to use the relation cos2 () + sin2 () = 1


substitute x = sin() with [/2, /2], so dx/d = cos() i.e. dx = cos()d
Z
Z
Z
dx
cos()d
cos()d Z
q

=
= d = = arcsin(x)
=
cos()
1x2
1sin2 ()

(1 + x2 )1/2 dx:

try to use the relation cosh2 () sinh2 () = 1


substitute x = sinh(), so dx/d = cosh() i.e. dx = cosh()d
Z
Z
Z
dx
cosh()d
cosh()d Z
q

=
=
= d = = arcsinh(x)
cosh()
1+x2
1+sinh2 ()

(1 + x2 )1 dx:

try to use the relation cos2 () + sin2 () = 1


substitute x = tan() with [/2, /2], so dx/d = cos2 () i.e. dx = cos2 ()d
Z
Z
Z
dx
cos2 ()d Z
d
=
=
= d = = arctan(x)
1+x2
1+tan2 ()
cos2 ()+sin2 ()

(1 x2 )1 dx:

try to use the relation cosh2 () sinh2 () = 1


substitute x = tanh(), so dx/d = cosh2 () i.e. dx = cosh2 ()d
Z

Z
Z
dx
cosh2 ()d Z
d
=
=
=
d = = arctanh(x)
1x2
1tanh2 ()
cosh2 ()sinh2 ()

96

R3
2

(x2 + 2x)1/2 dx:

first convert to a form similar to above, using x2 + 2x = (x + 1)2 1


Z 3
Z 3
dx
dx
q

=
2
2
2
x + 2x
(x + 1)2 1
try to use the relation cosh2 () sinh2 () = 1
substitute x + 1 = cosh(), so dx/d = sinh() i.e. dx = sinh()d
Z arccosh(4)
Z 3
Z arccosh(4)
sinh()d
dx
sinh()d
q

=
=
sinh()
arccosh(3)
2
arccosh(3)
x2 + 2x
cosh2 () 1
=

Z arccosh(4)
arccosh(3)

R 1/2
3/2

h iarccosh(4)

d =

arccosh(3)

= arccosh(4) arccosh(3)

(x2 2x)1/2 dx:

first convert to a form similar to above, using x2 2x = 1 (x + 1)2


Z 1/2
Z 1/2
dx
dx
q

=
2
3/2
3/2
x 2x
1(x+1)2
try to use the relation cos2 () + sin2 () = 1
substitute x + 1 = sin() with [/2, /2], so dx/d = cos() i.e. dx = cos()d
Z arcsin(1/2)
Z arcsin(1/2)
Z 1/2
cos()d
cos()d
dx
q

=
=
arcsin(1/2)
arcsin(1/2) cos()
3/2
x2 2x
1sin2 ()
=

Z arcsin(1/2)
arcsin(1/2)

h iarcsin(1/2)

d =

arcsin(1/2)

1
1
= arcsin( ) arcsin( )
2
2

= + =
6
6
3

(6 + 4x 2x2 )1/2 dx:

first convert to a form similar to above, using




6 + 4x 2x2 = 2(x2 2x 3) = 2((x 1)2 4) = 8 ( 12 x 21 )2 1
Z

dx
1 Z
dx
q

=
2
6 + 4x 2x
2 2
1 ( 12 x 12 )2

try to use the relation cos2 () + sin2 () = 1


substitute 21 x 21 = sin() with [/2, /2], so 12 dx/d = cos() i.e. dx = 2 cos()d
Z

dx
1 Z 2 cos()d
1 Z cos()d

=
=
=
cos()
6 + 4x 2x2
2 2
2
2
1 sin2 ()
1
1
1
= arcsin( x )
2
2
2

97

tan(x)dx:

write tan(x) = sin(x)/ cos(x) and substitute cos(x) = t, so dt/dx = sin(x) i.e.
dt = sin(x)dx
Z

tan(x)dx =

Z
dt
sin(x)dx
=
= ln |t| = ln | cos(x)|
cos(x)
t

sin1 (x)dx:

first convert into a form similar to earlier examples, using sin(x) = 2 sin(x/2) cos(x/2).
Then substitute cos(x/2) = t, so dt/dx = 21 sin(x/2) i.e. dt = 21 sin(x/2)dx:
Z

Z
Z
dx
dt
dx
=
=
2
sin(x)
2 sin(x/2) cos(x/2)
sin (x/2)t
Z
Z
dt
dt
=
=
2
t(1cos (x/2))
t(1t2 )

substitute u = t2 , so du/dt = 2t3 i.e. dt = 21 t3 du


Z

Z
dx
dt
1Z
t3 du
1Z
du
1 Z du
=
=
=
=
sin(x)
t(1t2 )
2 t(1u1 )
2 u(1u1 )
2 u1
1
1




1
1
1



1 2
= ln |u1| = ln 2 1 2 = ln 2
2
t
cos (x/2)
1 cos2 (x/2) 1

2
= ln
= ln | tan(x/2)|
2
cos (x/2)

6.2.3. Examples: integration by parts


This method works when the function to be integrated is the product of two factors, one which
does not get more complicated upon repeated differentiation (e.g. trigonometric or exponential
functions), with the other becoming simpler upon repeated differentiation (e.g. powers):

arcsin(x)dx:

first substitute x = sin(), so dx/d = cos() i.e. dx = cos()d, then integrate by parts
Z

arcsin(x)dx =

arcsin(sin()) cos()d =

cos()d

Z
d 
= sin() sin()
d = sin() sin()d
d

= sin() + cos() = x arcsin(x) + 1 x2


Z

98

x2 cos(2x)dx:

integrate by parts twice, to get rid of the annoying factor x2


Z
Z
 d

1
1
2
2
x cos(2x)dx = sin(2x)x
sin(2x)
x2
2
2
dx
Z
1
= sin(2x)x2 x sin(2x)
2
(
)
Z
 d 
1
1
1
2
cos(2x)
x
= sin(2x)x cos(2x)x +
2
2
2
dx
Z
1
1
1
= sin(2x)x2 + cos(2x)x
cos(2x)
2
2
2
1
1
1
= sin(2x)x2 + cos(2x)x sin(2x)
2
2
4

xex dx:

integrate by parts to eliminate the factor x:


Z

(
x

xe dx = e x

Z
d 
x = ex x + ex dx
dx

= (x + 1)ex

Sometimes it is helpful to create a form of two factors where initially there was just one,
d
x), e.g.
by inserting 1 = ( dx

ln(x)dx:
Z

ln(x)dx =

ln(x)

= x ln(x)

Z 

d
d 
x dx = x ln(x) x
ln(x) dx
dx
Z
Z dx

xx1 dx = x ln(x)

dx = x ln(x) x

arctan(x)dx:

d
x), then integrate by parts, then substitute x2 = y, so dy/dx = 2x
first insert 1 = ( dx

Z 

d 
d
x dx = x arctan(x) x
arctan(x) dx
dx
dx
Z
xdx
1 Z dy
= x arctan(x)
= x arctan(x)
1 + x2
2 1+y
1
1
= x arctan(x) ln |1 + y| = x arctan(x) ln(1 + x2 )
2
2

arctan(x)dx =

arctan(x)

99
6.2.4. Further tricks: recursion formulae
Certain families of integrals can be calculated by a series of integrations by parts, leading in a
natural way to so-called recursion formulae. This is best explained directly via examples, one
with a family of definite integrals and one with a family of indefinite ones:
Example 1:
In =

xn ex dx

n ZZ+

Integration by parts (using limx xn ex = 0):


Z
d n
ex xn1 dx = nIn1
x dx = n
0
dx
0
0
Recursion formula: the expression for In in terms of In1 .
Further iteration of the recursion formula: In = n(n1)In1 = . . . = n!I0
Thus we need only calculate I0 to know all In :
h

In = xn ex

I0 =

ex dx = ex

i
0

ex

= 0 (1) = 1

hence : In = n!

Example 2:
In,m (x) =

sinn (x) cosm (x)dx

n, m ZZ

Just using cos2 (x) + sin2 (x) = 1 already gives


In+2,m (x) + In,m+2 =

sinn (x)[sin2 (x) + cos2 (x)] cosm (x)dx = In,m (x)

hence also (just substitute n n 2):


In2,m+2 = In2,m (x) In,m (x)
Integration by parts of In,m (x) gives
In,m (x) =

=
=
=
=

sinn1 (x)
d

cosm+1 (x) dx
m+1
dx
!
n1
m+1
sin (x) cos
(x)
1 Z
d
n1
+
sin (x) cosm+1 (x)dx
m+1
m+1
dx
sinn1 (x) cosm+1 (x)
n1 Z
sinn2 (x) cosm+2 (x)dx
+
m+1
m+1
sinn1 (x) cosm+1 (x)
n1
+
In2,m+2 (x)
m+1
m+1
sinn1 (x) cosm+1 (x)
n1
+
{In2,m (x) In,m (x)}
m+1
m+1
!

100
Solving for In,m (x) then gives
n1
In,m (x) 1 +
m+1


In,m (x)

n1
sinn1 (x) cosm+1 (x)
+
In2,m (x)
m+1
m+1

sinn1 (x) cosm+1 (x)


n1
m+n
=
+
In2,m (x)
m+1
m+1
m+1

Final result:
sinn1 (x) cosm+1 (x)
n1
+
In2,m (x)
m+n
m+n
We need only calculate I1,m (x) and I0,m (x) to know all In,m (x).
The integrals I0,m (x) will be done as a tutorial exercise, the integrals I1,m (x) are:
In,m (x) =

I1,m (x) =

sin(x) cosm (x)dx


Z

y m dy =

put cos(x) = y so dy/dx = sin(x)

1
1 m+1
y
=
cosm+1 (x)
m+1
m+1

Special case of previous calculation:


(choose m = 0 in above result)
In (x) =

sinn (x)dx

n ZZ

Recursion formula:
In (x) =

sinn1 (x) cos(x) n 1


+
In2 (x)
n
n

I0 (x) = x

Note 1:
The result of example 1 is used to generalize the concept of factorials n! to real-valued
(and later even complex) numbers n, by using the integral as a definition:
for all z IR :

z! =

xz ex dx

This prompted the introduction of the so-called gamma function (z) =


Thus (n) = (n 1)! for integer n.
Note 2:
As an alternative in the case of integrals with trigonometric functions

R z1 x
x e dx,
0

101
one could write the latter in complex form, e.g.
Z

sin (x)dx = (2i)

Z 

= (2i)

n 
X

eix eix

m=0

n

use Newton0 s binomial formula

dx

n
(1)nm
m

eix(2mn) dx

if n even :

1
(2i)n

if n odd :

1
(2i)n

Pn

m=0, m6=n/2

Pn

m=0

 n 
n  (1)m ix(2mn)
n/2
e
+
(1)
x
m i(2mn)
n/2

n  (1)m+1 ix(2mn)
e
m i(2mn)

For instance:
Z

sin2 (x)dx =

2
X

1
(2i)2 m=0,
(


m6=1


2

(1)
2
x
eix(2m2)
m i(2m n)
1


1
1 2ix  2  1 2ix  2 
=
e
+
e
x
0 2i
2 2i
1
4


1
1
1
1
1 2ix
e
+ e2ix 2x = sin(2x) + x
=
4 2i
2i
4
2
2

Check against recursion method of example 2:


Z
1
1
1
1
sin2 (x)dx = I2 (x) = sin(x) cos(x) + I0 (x) = sin(x) cos(x) + x
2
2
2
2
1
1
= sin(2x) + x
4
2

6.2.5. Further tricks: differentiation with respect to a parameter


Remember: any dirty trick that leads to a proposal for a primitive is allowed,
provided one verifies correctness of the proposed answer a posteriori via differentiation!
Now consider integrals that involve a further parameter a IR:
I(x, a) =

f (x, a)dx

For sufficiently well-behaved functions it is true that


Z
d Z
d
f (x, a)dx =
f (x, a)dx
da
da
But not always ...
(integral is a limit, derivative is a limit, we know that the order of limits matters!)
Our strategy:

102
(i) assume that moving the derivative d/da inside or outside the integral over x is allowed
(ii) use that as a tool to calculate the integral
(iii) check whether the assumption was correct by explicit differentiation of the result
We illustrate the procedure via examples:
d n
) means differentiate n times with respect to a)
(notation: ( da
Example 1:
I(x, a) =

xn eax dx

 d n 
 d n Z

d n ax
=
eax dx =
e dx =
a1 eax
da
da
da
We get all the integrals we need by differentiating a simple expression.
For instance:
Z
x
1
d  1 ax 
a e
= ( 2 ) eax
xeax dx =
da
a a
Z 

d 2  1 ax 
d
a e
=
da
da
Differentiation confirms that these primitives are
Z

x2 eax dx =

x
1
x2 2x
2
( 2 )eax = ( 2 + 3 ) eax
a a
a
a
a
correct.

Example 2:
Here we exploit the relations
 d 2 1
1
1.2
d 1
=
,
=
,
da x2 a
(x2 a)2
da x2 a
(x2 a)3

d 3 1
1.2.3
=
, etc
da x2 a
(x2 a)4

More generally:


d n1 1
(n1)!
= 2
2
da
x a
(x a)n

Hence, for a 6= 0:
I(x, a) =
=

dx

(x2 a)n
Z (

1  d n1 1
1  d n1 Z dx
dx =
(n1)! da
x2 a
(n1)! da
x2 a

Again we obtain all the (complicated) integrals we need


by differentiating some simple basic ones:
Z

dx
1Z
dx

a>0:
=
put
x
=
y
a
x2 a
a (x/ a)2 1

1 Z dy
1
arctanh(x/ a)

=
= arctanh(y) =
a 1 y2
a
a

103
!
o

1
1 + x/ a
1 n

= ln
= ln( + x) ln( x)
2 a
1 x/
2 a
a<0:

dx
1 Z
dx
q
=
2
x a
|a| (x/ |a|)2 + 1

put x = y |a|
q

arctan(x/ |a|)
dy
1
q arctan(y) =
q
=
1 + y2
|a|
|a|

1 Z

=q

|a|

For instance:
Z

dx
=
2
(x + 1)2

d Z dx
da x2 a

=
a=1

)
d arctan(x/ a)

da
a
a=1

put z = (a)1/2 and use the chain rule:


Z

dx
=
(x2 + 1)2


dz d 
.
z arctan(zx)
da dz
a=1



1
zx 
3/2
(a)
arctan(zx) +
=
2
1+z 2 x2 a=1
x
1
= arctan(x) +
2
2(1+x2 )

Correctness is confirmed by explicit differentiation.

6.2.6. Further tricks: partial fractions


Come into play when integrating
Z

p(x)
dx
q(x)

p(x), q(x) : polynomials

Method is based on the following two facts:


p(x) can always be written as p(x) = s(x)q(x) + r(x) (r(x): the remainder)
s(x), r(x) other polynomials, with r(x) of lower order than q(x)
This gives
Z

Z
Z
p(x)
r(x)
dx = s(x)dx +
q(x)
q(x)

first part : easy!

If q(x) is of the following form, with i , j , j IR (all different),


q(x) =

n
Y

(x + i )ai

i=1

m
Y

(x2 + j x + j )bj

j=1

with

ai , bi ZZ+

104
(remember: ni=1 ki = k1 k2 . . . kn1 kn ) and the order of r(x) is less than that of q(x), then
there always exists constants Aik , Bj` , Cj` IR such that
Q

ai
j
n X
m X
X
r(x) X
Aik
Bj` x + Cj`
=
+
k
2
`
q(x) i=1 k=1 (x + i )
j=1 `=1 (x + j x + j )

(the latter are called partial fractions)


In combination:
our initial integral can always be written as
Z

bj Z
Z
Z
ai
n X
m X
X
X
p(x)
dx
(Bj` x + Cj` )dx
dx = s(x)dx +
Aik
+
k
q(x)
(x + i )
(x2 + j x + j )`
i=1 k=1
j=1 `=1

s(x)dx with s(x) polynomial: easy

(x + )k dx: easy

(Bx + C)/(x2 + x + )` : do-able ...

Our general strategy for integrating a ratio of polynomials:


convert the ratio into the above form

Note 1:
The simplest case of the above is the following
If q(x) is of the following form, with i IR (all different),
q(x) =

n
Y

(x + i )

with i IR

i=1

and the order of r(x) is less than that of q(x), then there are constants Ai IR such that
n
r(x) X
Ai
=
q(x) i=1 x + i

and our initial integral can always be written as


Z

Z
n
X
p(x)
dx = s(x)dx +
Ai ln |x + i |
q(x)
i=1

The constants Aj can be found upon multiplying our initial equation by x + j


followed by setting x = j :
n
X
r(x)
Ai
=
Qn
i=1 (x + i )
i=1 x + i

Qn

i=1,

n
X
r(x)
x + j
= Aj +
Ai
x + i
i6=j (x + i )
i=1, i6=j

105
now put x = j :
r(j )
i=1, i6=j (i j )

Aj = Qn

Note 2:
Let is inspect the not easy but do-able integrals above in more detail
First: write numerator as derivative of quadratic form in denominator
Z
Z
Bx + C
2x
BZ
dx
dx =
dx + C
2
`
2
`
2
(x + x + )
2
(x + x + )
(x + x + )`
Z
Z

2x +
dx
B
1
=
dx + C B
2
`
2
2
(x + x + )
2
(x + x + )`

 
Z
Z
B
d
1
dx
1
=
(x2 + x + )1` + C B
2
dx 1`
2
(x2 + x + )`
Z


1
dx
B
(x2 + x + )1` + C B
=
2(1`)
2
(x2 + x + )`
Second: complete squares in the remaining integral and make an appropriate substitution
1
1
1
1
x2 + x + = (x + )2 + ( 2 ) put y = x + , 2 = | 2 |
2
4
2
4
Z

Z
dx
dy
=
(x2 + x + )`
(y 2 2 )`

The latter integrals are of the form done earlier


(when explaining differentiation with respect to a parameter).
Let us illustrate how this works via examples:
Example 1:
I=

(x2

dx
1)(x + 3)

We recognize that
(i) we integrate a ratio of polynomials method of partial fractions
(ii) order of numerator is less than that of denominator,
i.e. we just have just the remainder r(x)/q(x) (no s(x) needed)
(iii) in fact the simplest case: r(x) = 1 and q(x) = (x + 1)(x 1)(x + 3)
We know that we can always find constants A1 , A2 , A3 such that
A1
A2
A3
1
=
+
+
for all x IR
2
(x 1)(x + 3)
x1 x+1 x+3

106
Find these constants:
x1
A2 (x 1) A3 (x 1)

1)(x + 3)
x+1
x+3
A2 (x 1) A3 (x 1)
1

put x = 1
=
(x + 1)(x + 3)
x+1
x+3
1
1
=
=
(1 + 1)(1 + 3)
8

A1 =

A2 =
=
=
A3 =
=
=

(x2

A1 (x + 1) A3 (x + 1)
x+1

1)(x + 3)
x1
x+3
1
A1 (x + 1) A3 (x + 1)

put x = 1
(x 1)(x + 3)
x1
x+3
1

4
x+3
A1 (x + 3) A2 (x + 3)

2
(x 1)(x + 3)
x1
x+1
A1 (x + 3) A2 (x + 3)
1

put x = 3
2
x 1
x1
x+1
1
8
(x2

Hence
Z (

1
1
1

+
dx
I=
8(x 1) 4(x + 1) 8(x + 3)


1
1
1
1 x2 +2x3
= ln |x 1| ln |x + 1| + ln |x + 3| = ln 2

8
4
8
8 x +2x+1
Example 2:
I=

(x2

xdx
+ 1)(x + 2)

We recognize that
(i) we integrate a ratio of polynomials method of partial fractions
(ii) order of numerator is less than that of denominator (no s(x) needed)
(iii) this one is not of the simplest form q(x) = (x + 1 )(x + 2 )(x + 3 )
We know that we can always find constants A, B, C such that
x
A
Bx + C
=
+
for all x IR
(x2 + 1)(x + 2)
x+2
x2 + 1
Find first constant:
x(x + 2)
(Bx + C)(x + 2)
x
(Bx + C)(x + 2)
A= 2

= 2

2
(x + 1)(x + 2)
x +1
x +1
x2 + 1

107
put x = 2 :

A=

2
5

so we must find constants B, C such that


x
Bx + C
2
= 2

for all x IR
2
(x + 1)(x + 2)
x +1
5(x + 2)
Insert two convenient values for x and solve the two resulting eqns for B and C:
1
3
x = 1 : B C =
x=1: B+C =
5
5
The solution is: B = 2/5 and C = 1/5. Hence


1 Z 2x + 1
2
I=

dx
5
x2 + 1 x + 2
1 Z dx
2 Z dx
1 Z 2xdx
+

=
5 x2 + 1 5 x2 + 1 5 x + 2
1
2
1
= ln |x2 + 1| + arctan(x) ln |x + 2|
5
5
5
6.3. Some simple applications
6.3.1. Calculation of surface areas
The area A of the surface between the x-axis and a curve f (x),
R
taken from x = a to x = b (with the accepted sign conventions) is: A = ab f (x)dx
Area inside a circle:
y

Equation for circle with radius R,


centred in the origin: x2 + y 2 = R2

Upper half of circle: y = R2 x2 , with x [R, R]


Area A between upper half of circle and x-axis:
Z R
A=
R2 x2 dx
R

put x = R cos() so dx = R sin()d


= R

Z 0q

R2 R2 cos2 () sin()d

=R

Z q

R2 R2 cos2 () sin()d

= R2

sin2 ()d = R2

Z 
1
0

1
1
sin(2) =
2
4
0
Since this is exactly half of the
= R2

Acircle = R2

1
cos(2) d
2


1 2
R
2
surface area inside the circle:

108
Area inside an ellipse:
Equation for an ellipse centred at the origin,
with foci on x-axis at x = a
and with sum of distances to the foci given by 2R:
q

(x

a)2

y2

(x + a)2 + y 2 = 2R

Not yet of the required form y = f (x) ...


Rewrite ellipse equation:
move one term to the right, then square both sides
q

(xa)2 +y 2 = 2R (x+a)2 +y 2
q

(xa)2 +y 2 = 4R2 +(x+a)2 +y 2 4R (x+a)2 +y 2


q

2xa = 4R +2xa4R (x+a)2 +y 2


q

(x+a)2 +y 2 = R+xa/R

(x+a)2 +y 2 = (R+xa/R)2
Giving: y 2 = R2 a2 x2q
(1 a2 /R2 )
Upper half of ellipse: y = R2 a2 x2 (1a2 /R2 ), with x [R, R]
Hence, area A between upper half of ellipse and x-axis:
A=

Z R q

R2 a2 x2 (1a2 /R2 ) dx

R2 a2

Z R

x2
dx
R2

put x = R cos() so dx = R sin()d

Z 0q

1cos2 () sin()d
= R R2 a2
Z

= R R2 a2

1
sin2 ()d = R R2 a2
2
0
( integral already calculated for the circle)
Since this is exactly half of the surface area inside the ellipse:

Aellipse = R R2 a2

6.3.2. Calculation of volumes of revolution

f (x)

Take the graph of a function f (x)


with f (x) 0 for all x [a, b]
Revolve this around the x-axis (see figure)
Result: a solid in three dimensions
What is its volume V ?

109
split the x-axis into small steps of size h
(i.e. split solid into small slices of width h)
e.g. xi = a + (i 1)h,
with i = 1, . . . , L and h = (b a)/(L 1)
let the cross-section of the solid at point xi
have an area equal to A(xi )
then a slice at position xi contributes an
amount to the volume that is approximately
hA(xi ) (if h is sufficiently small)

A(xi )
A
A
A
A
A
A

cross-section of the solid at point xi


is a circle with radius f (xi )
hence its area is A(xi ) = f 2 (xi )
Total volume:
sum all contributions from the L slices:
V

L
X

hf 2 (xi )

i=1

Expression becomes exact for h 0:


V = lim
h0

L
X

hf 2 (xi ) =

i=1

Z b

f 2 (x)dx

Example:
Our solid becomes a sphere
if what we revolve around the x-axis is a circle:

x2 + y 2 = R 2 y = R 2 x2

f (x) =

x [R, R]

Hence
Vsphere =

Z R

f (x)dx =

1
= R x x3
3


Z R

(R2 x2 )dx

R
R

1
4
= 2(R3 R3 ) = R3
3
3

Example:
Our solid becomes a cigar
if what we revolve around the x-axis is an ellipse:

R 2 x2

110

y = R a x (1a /R )

f (x) =

x [R, R]

R2 a2 x2 (1a2 /R2 )

Hence
Vcigar =

Z R

f (x)dx =

Z R

R2 a2 x2 (1a2 /R2 )dx

1
= (R a )x (1a2 /R2 )x3
3
4
= R(R2 a2 )
3


R
R



1
= 2 (R2 a2 )R (R2 a2 )R
3

6.3.3. Calculation of the length of curves


Consider an arbitrary curve in the plane,
described by a function y = f (x)

f (x)

What is the length L of this curve


between, say, x = a and x = b ?
split the x-axis into small steps of size h
e.g. xi = a + (i 1)h,
with i = 1, . . . , K and h = (b a)/(K 1)
write f (xi ) = yi

(xi+1 , yi+1 )
`i

(xi , yi ) (xi+1 , yi )

consider the curve segment between


the points (xi , yi ) and (xi+1 , yi+1 )
Pythagoras theorem in triangle:
the length of this segment is approximately
q

`i

(xi+1 xi )2 + (yi+1 yi )2

(if xi+1 xi sufficiently small)

use xi+1 xi = h and yi = f (xi )


s

`i h 1 +

y

 f (x + h) f (x ) 2
y i 2
i
i
=h 1+
h
h

i+1

Combine:
L = lim

h0

Z b
a

K
X

`i = lim h
h0

i=1

1+

 df 2

dx

dx

K
X
i=1

1+

 f (x

+ h) f (xi ) 2
h

111

Example:
Circumference of a circle with radius R?

Twice length of the curve y = R2 x2 , with x [R, R]:


Lcircle = 2
=2

Z R

Z R

= 2R

Z R

d
1 2
2
2
2
1+
(R x )
dx = 2
dx
R


Z 1
1

1+

2
x
dx
R 2 x2

Z R
x2
R2
1+ 2
dx
=
2
dx put x = Ry
R x2
R 2 x2
R
h
i1

dy


(
)
=
2R
arcsin(y)
=
2R
1
2
2
1 y2

= 2R
Next consider an arbitrary curve in the plane,
described parametrically by functions x(t) and y(t)

(x(t), y(t))

What is the length L of this curve


between, say, t = a and t = b ?
split the t-axis into small steps of size h
e.g. ti = a + (i 1)h,
with i = 1, . . . , K and h = (b a)/(K 1)
write x(ti ) = xi and y(ti ) = yi

(xi+1 , yi+1 )
`i

(xi , yi ) (xi+1 , yi )

consider the curve segment between


the points (xi , yi ) and (xi+1 , yi+1 )
Pythagoras theorem in triangle:
the length of this segment is approximately
q

`i

(xi+1 xi )2 + (yi+1 yi )2

(if ti+1 ti sufficiently small)

use xi = x(ti ) and yi = y(ti )


s

`i h
s

=h

x

xi 2  yi+1 yi 2
+
h
h

i+1

 x(t

+ h) x(ti ) 2  y(ti + h) y(ti ) 2


+
h
h

Combine:
L = lim

h0

K
X
i=1

Z bq

`i = lim h
h0

K
X

 x(t

i=1

(dx/dt)2 + (dy/dt)2 dt

+ h) x(ti ) 2  y(ti + h) y(ti ) 2


+
h
h

112
Example:
Circumference of a circle with radius R?
parametrize x() = R cos(), y() = R sin(),
with [0, 2]:
Lcircle =

Z 2 q

(dx/d)2 + (dy/d)2 d =

Z 2 q

R2 sin2 () + R2 cos2 () d

= 2R
Example:

Length of a spiral with radial velocity v


and angular velocity ?

(x(t), y(t))

parametrize, with t [0, ]:


x(t) = vt cos(t), y(t) = vt sin(t)

t=

dx
= v cos(t) vt sin(t)
dt
dy
= v sin(t) + vt cos(t)
dt

t=0

Hence:
Lspiral =

Z q

=v

(dx/dt)2 + (dy/dt)2 dt

Z rn

cos(t) t sin(t)

o2

+ sin(t) + t cos(t) dt

Z0

1 + 2 t2 dt
put t = u so dt = du/
0
v Z
=
1 + u2 du
put u = sinh(z) so du = cosh(z)dz
0

v Z arcsinh( )  2z
v Z arcsinh( )
2
=
cosh (z)dz =
e + 2 + e2z dz
0
4 0


v
1 2z 1 2z arcsinh( )
=
2z + e e
4
2
2
0
o
v n
2 arcsinh( ) + sinh(2arcsinh( ))
=
4 n
o
v
=
arcsinh( ) + cosh(arcsinh( ))
2



v
v q
=
ln + 2 2 +1 +
1 + sinh2 (arcsinh( ))
2 
2


v
v
=
ln + 2 2 +1 +
1 + 2 2
2
2
=v

where we used
cosh2 (z) sinh2 (z) = 1,

z 2 + 1)

sinh(2arcsinh(z)) = 2 sinh(arcsinh(z)) cosh(arcsinh(z)) = z 1 + z 2


arcsinh(z) = ln(z +

113
7. Taylors theorem and series
7.1. Introduction to series and questions of convergence
7.1.1. Series notation and elementary properties
definition:
A series is an expression of the form
(usually n0 = 0, 1)

n=n0

an

definition:
A partial sum of the series is an expression of the form SN =

PN

n=n0

an

definition
A series
n=n0 an is called convergent if the limit S = limN SN exists. Only then will the
series yield a well-defined finite number. If the series does not converge it is called divergent.
P

Notes:
If a series
n=n0 an converges, then limn an = 0
Proof: limn an = limn (Sn Sn1 ) = S S = 0
P

We have already inspected the following series in exercises:

X
n=1

2n :

SN =

1:

SN =

N
X
n=1
N
X

2n = 12N
1=N

n=1

n=1

lim SN = 1,

series convergent

lim SN does not exist,

series divergent

However, having limn an = 0 is not enough to guarantee that a series converges, the an will
also have to go to zero sufficiently fast as n gets larger. See e.g. the two examples below:
Examples:
The series

n=1

n1 is divergent

Proof:
P
1
Consider the partial sums SN = N
n=1 n , they obey
1
1
1
1
1
1
S2N SN =
+
+ ... +
+
N
=
N +1 N +2
2N 1 2N
2N
2
If we assume that the series converges, then limN SN = S exist. However, taking
the limit limN in the previous inequality would then lead us to the contradiction
0 = S S 12 . We conclude that the series must be divergent.

114
The series

n=1

Proof:
1
Since n(n+1)
=

1
n

1/n(n+1) is convergent

1
,
n+1

the partial sum SN can be written as

N
X

N
X
1
1
1
SN =
=

n+1
n=1 n(n+1)
n=1 n

 



1 1
1 1
1
1

=
+
+ ... +
1 2
2 3
N N +1

 



1 1
1
1 1
1 1
1
=1+ +
= 1
+ +
+ ... + +

2 2
3 3
N N
N +1
N +1
P
It follows that limN SN = 1, so the series converges: n=1 1/n(n+1) = 1.

7.1.2. Series convergence criteria


If we are only interested in whether a series converges/diverges, the start index n0 is irrelevant;
P
hence one often speaks simply about the convergence or divergence of a series n an .
We now mention a number of useful results on the convergence of series, without proof.
The first applies if our an are similar for large n to those of another series that we know.
The others apply if we can calculate either limn |an |1/n or limn |an+1 /an |:
(C1)

bn is convergent and

(C2)
(C3)
(C4)
(C5)

lim |an |1/n < 1

lim |an |1/n > 1

lim |an+1 /an | < 1

lim |an+1 /an | > 1


n

lim |an |/bn exists


n

an is convergent

an is convergent

an is divergent

an is convergent

an is divergent

The tricky series are those with


limn |an |1/n = limn |an+1 /an | = 1
Examples:
The series

1/n2 is convergent

Proof:
P
Let an = 1/n2 , and compare to the convergent series n bn with bn = 1/n(n+1):

1/n2
1
n2 +n
= lim
=
lim
1+
=1
lim
|a
n |/bn = lim
n
n 1/n(n+1)
n n2
n
n
P
It now follows from (C1) above that n 1/n2 is also convergent.

115
The series

n+1
/n
n (1)

is convergent

Proof:
Separate in the partial sum SN the terms with even n from those with odd n. Write
n = 2m for even n, and n = 2m 1 for odd n.
N even : SN =

n=1,

N odd : SN =

N/2
N/2
N/2
N
X
X
X 1
X
1
1
1
1

=
n=1, even n
m=1 2m1
m=1 2m
m=1 2m(2m1)
odd n

N
X

+1)/2
(N 1)/2
N
X
X
1
1 (NX
1
1

n=1, even n
m=1 2m1
m=1 2m
odd n

N
X
n=1,

(N 1)/2

X
m=1

1
1
+
2m(2m1) N

n+1

We see that our present series n (1) /n is convergent if and only if the series n an
P
1
with an = 2n(2n1)
is convergent. We may now use the convergence of n bn with bn = 1/n2
together with statement (C1) to finish the proof:
1

1
1
1/2n(2n1)
=
lim
=
2
n
n
1/n
4 11/2n
4

lim |an |/bn = lim

We conclude:

an converges, and therefore also

n+1
/n
n (1)

converges.

7.1.3. Power series notation and elementary properties


definition:
A power series is an expression of the form S(x) =
i.e. a series where the terms are an = bn xn

n=n0 bn x

Why our interest?


n
(i) we saw that some functions can be written as power series f (x) =
n=0 bn x
(ii) each of these functions had its own unique coefficients {bn }
(iii) power series representations of functions are extremely useful
P
n
(iv) which other functions can perhaps be written as power series f (x) =
n=0 bn x ?
(v) how would we find the right coefficients {bn } for any given function?
(vi) how can we know whether the power series would converge?
(vii) how fast does a power series converge to the original function it represents?

Let us apply the convergence criteria (C2 C5) to power series,


by substitution of an = bn xn
(statements below make sense only if the various limits exist)

116
1/n

Let R1 = lim |bn |


n

Let R2 = n
lim |bn /bn+1 | :

|x| < R1 :
|x| > R1 :

|x| < R2 :
|x| > R2 :

n bn x
n bn x

n bn x
n bn x

converges
diverges
converges
diverges

Clearly, the two limits above must be the same.


This can be shown properly: if the limits exist, then R1 = R2 = R
We are now led in a natural way to the concept of radius of convergence:
definition
n
The radius of convergence R of a power series
n=n0 bn x :
a nonnegative number such that the series converges for all x with |x| < R
and diverges for all x with |x| > R.

It can be calculated either from R = limn |bn |1/n or from R = limn |bn /bn+1 |
if these latter limits exist
Notes:
(i) Exactly at the radius of convergence, i.e. for |x| = R,
there is no general rule (just inspect the series at hand in detail)
(ii) If R = 0 there is apparently no x 6= 0 for which the series converges
(iii) If R = the series converges for all x IR
Examples:
ex =

n=0

xn /n!

Here bn = 1/n!, so
|bn |
(n+1)!
= n
lim
= n
lim (n+1) =
|bn+1 |
n!
Thus the series for ex converges for all x IR (as claimed earlier).
R = n
lim

f (x) =

n=0

xn

Here bn = 1, so R = limn |bn |/|bn+1 | = 1.


P
n
Thus the series f (x) =
n=0 x converges for |x| < 1.
n
N +1
Compare this to our earlier result N
)/(1 x) for x 6= 1.
n=0 x = (1 x
We now see that this gives the series expansion for the function f (x) = 1/(1 x):

X
1
=
xn
|x| < 1
1 x n=0

117
n
Building (1 x)1 =
n=0 x as a power series,
by taking more and more terms in the summation:

65 4 3

10
9
8
7
6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-7
-8
-9
-10

-2

-1

dashed :

1
1x

solid :

SN (x) =

N
X

xn

for different choices of N

n=0

values of N are indicated in italics


As N increases: S(x) starts to resemble (1 x)1 more and more,
but only on the interval (1, 1)
f (x) is only fully identical to (1 x)1 for N ,
within the radius of convergence, i.e. for |x| < 1
sin(x) =

m 2m+1
/(2m+1)!
m=0 (1) x

(be careful! not yet written in standard form with xn )


For n odd: bn = (1)m /(2m+1)! (where n = 2m+1), i.e. bn = (1)(n1)/2 /n!
For n even: bn = 0
(
1/n

R = lim |bn |
n

= lim

0
(n!)1/n

if n even
if n odd

118
It can be shown that (n!)1/n as n , hence our series converges for all x IR.
P
P
A quicker way to prove that R = is based on the inequality |
n=0 |an |:
n=0 an |
|

m 2m+1

(1) x

/(2m+1)!|

m=0

2m+1

|x|

/(2m+1)!

m=0

|x|n /n!

n=0

The last series (the exponential function of |x|) converges for all x IR,
P
m 2m+1
/(2m+1)! converges for all x IR (as claimed earlier).
hence also
m=0 (1) x

7.2. Taylors theorem


We have achieved some understanding of when power series converge,
and next turn to the construction of such
for arbitrary functions.
 series
n
d
(n)
First some further notation: f (x) = dx f (x)
(the n-th derivative of f at the point x)
7.2.1. Expression for the coefficients of power series
We will inspect different routes towards the formula for the coefficients of a power series.
Let us first consider non-pathological functions, where the derivative of the sum equals the sum
of the derivatives. Here we can use simple ideas:
Claim:
If a given function f (x) can be written as a power series,
with some convergence radius R > 0, i.e.
x IR, |x| < R :

f (x) =

bn x n

n=0

and if differentiation and summation in the power series for f can be interchanged
(as for finite sums), then bn = f (n) (0)/n! (if this n-th derivative exists).
Proof:
n
If the function f (x) is indeed identical to the series S(x) =
n=0 bn x for |x| < R,
then also their derivatives must be identical, i.e. f (m) (x) = S (m) (x) for |x| < R. Thus

(m)

(x) =
=

d
dx

!m
X
n=0

bn x =

X
n=0

bn

d
dx

!m

xn

bn n(n1)(n2) . . . (nm+1)x

n=0

nm

X
n=m

bn

n!
xnm
(nm)!

Now choose x = 0: all powers of x in the right-hand side with m > n will vanish, giving
f (m) (0) = bm m!

119
Hence bm = f (m) (0)/m! as claimed. This also shows that under the assumed conditions there
can only be one power series representation of the function f .

Let us now derive this result in another way, where we dont need to interchange summation
and differentiation. It is based on repeated application of the fundamental theorem of calculus:
Z

f (n) (x) = f (n) (0) +

f (n+1) (y)dy

Let f be differentiable as many times as we need:


f (x) = f (0) +

Z x

f (1) (y)dy

put y = xz1

= f (0) + x

Z 1

f (1) (xz1 )dz1

= f (0) + x

Z 1

f (1) (0) +

Z xz1

Z 1 Z xz1

= f (0) + f (1) (0)x + x

= f (0) + f

(1)

(0)x + x

= f (0) + f

(0)x + x

f (2) (y)dy dz1

Z 1
0

(1)

f (2) (y)dy dz1

Z 1
0

Z 1

z1

Z 1 

z1

= f (0) + f (1) (0)x + x2 f (2) (0)

+ x2

= f (0) + f

(1)

2 (2)

(0)x + x f

(0)

+x

Z 1
0

= f (0) + f

(1)

z12

Z xz1 z2

(3)

(y)dy dz2 dz1

dz2 dz1


f (3) (y)dy dz2 dz1

put y = xz1 z2 z3

z1 dz1

Z 1 

z2

(0) +

Z 1

z1
0
1

(xz1 z2 )dz2 dz1

(2)

Z 1 Z xz1 z2

z1

0
1

Z 1

(2)

put y = xz1 z2

Z 1

(3)

(xz1 z2 z3 )dz3 dz2 dz1

Z 1
1
(0)x + f (2) (0)x2 + x3
z12
2
0

Z 1 
0

z2

Z 1

(3)

(xz1 z2 z3 )dz3 dz2 dz1

We observe:
(i) we once more generate the coefficients bn = f (n) (0)/n!
(ii) but now we also find explicit formulas for the difference between the true function f
P
(n)
and the partial sums SN (x) = N
(0)xn /n! of the power series:
n=0 f
f (x)
with e.g.

N
X

f (n) (0) n
x + RN +1 (x)
n!
n=0
R1 (x) = x
R2 (x) = x

Z 1
0
2

f (1) (xz1 )dz1

Z 1
0

Z 1

z1

f
0

(2)

(xz1 z2 )dz2 dz1

120
R3 (x) = x

Z 1
0

z12

Z 1 
0

z2

Z 1

(3)

(xz1 z2 z3 )dz3 dz2 dz1

7.2.2. Taylor series around x = 0


The last result gives us essentially the so-called Taylor series expansion in powers of x of a
differentiable function f , the only difference with the above form is that the remainder term is
written in a nicer way:
definition
The Taylor expansion to order N around x = 0 of a function f is the following (exact)
expression, involving an N -th order polynomial and a remainder term RN +1 (x):
f (x) =

N
X

f (n) (0) n
x + RN +1 (x)
n!
n=0

RN +1 (x) =

1 Z x (N +1)
f
(y)(xy)N dy
N! 0

Proof:
(of the form claimed to be exact for the remainder term)
This is done by induction. We define as always the difference between left- and right-hand side
of the claimed identity,
N
X

f (n) (0) n
1 Z x (N +1)
f
(y)(xy)N dy
x
n!
N
!
0
n=0

AN (x) = f (x)

We aim to prove that AN (x) = 0 for all integer N 0. First we check the basis, i.e. N = 0:
A0 (x) = f (x)

f (0) (0) 0 1 Z x (1)


f (y)(x y)0 dy
x
0! Z
0! 0
x

= f (x) f (0)

ix

f 0 (y)dy = f (x) f (0) f (y)

=0

So the claim is true for N = 0. Next we prove the induction step. We assume that AN (x) = 0
(i.e. the Taylor expansion is exact for the value N ) and prove from this that also AN +1 (x) = 0:
AN +1 (x) = f (x)

N
+1
X
n=0

Z x
f (n) (0) n
1
x
f (N +2) (y)(xy)N +1 dy
n!
(N +1)! 0

N
X

Z x
f (n) (0) n f (N +1) (0) N +1
1
= f (x)
x
x

f (N +2) (y)(xy)N +1 dy
n!
(N +1)!
(N +1)! 0
n=0

1 Z x (N +1)
f (N +1) (0) N +1
f
(y)(xy)N dy
x
N! 0
(N +1)!
Z x
1

f (N +2) (y)(xy)N +1 dy
(N +1)! 0
1 Z x (N +1)
f (N +1) (0) N +1
N
integr by parts : =
f
(y)(xy) dy
x
N! 0
(N +1)!

use AN (x) = 0 : =

121
1

N!
=

("

f (N +1) (y)(xy)N +1
N +1

#x

0
(N +1)

Z x

(N +1)

(y)(xy) dy

f (N +1) (0) N +1
f
(0) N +1
x
0+
x
=0
(N +1)!
(N +1)!

This completes the proof.


Notes:
One can regard the Taylor series as an approximation of f for values of x close to zero,
with the remainder term RN +1 (x) indicating the precise difference between the polynomial
approximation up to order N and the true function f
If the n-th derivative of f is bounded on the interval [R, R], i.e. |f (n) (x)| Cn for all
x [R, R], then |RN +1 (x)| Cn xN +1 /(N +1)!
Proof:
1 Zx
1 Z x (N +1)
|f
(y)|(xy)N dy Cn
(xy)N dy
N! 0
N! 0
h
i
Cn
1
N +1 x
(xy)
=
xN +1
= Cn
0
(N +1)!
(N +1)!

|RN +1 (x)|

For small |x|, the correction is therefore generally much smaller than any of the previous
terms in the approximating polynomial.

7.2.3. Taylor series around x = a


One can generalize the above Taylor series in order to have an expansion of functions around
some other value x = a, in terms of powers in the deviation x a.
definition
The Taylor expansion to order N around x = a of a function f is the following (exact)
expression,
involving an N -th order polynomial and a remainder term RN +1 (x):
N
X

f (n) (a)
f (x) =
(xa)n + RN +1 (x)
n!
n=0

1 Z x (N +1)
RN +1 (x) =
f
(y)(xy)N dy
N! a

Proof:
This can be done by a simple change of variables. Write x = a + u, define g(u) = f (a + u), and

122
apply the previous Taylor expansion to g(u):
g(u) =

N
X

g (n) (0) n
u + RN +1 (u)
n!
n=0

RN +1 (u) =

1 Z u (N +1)
g
(z)(uz)N dz
N! 0

d n
d n
Use the fact that g (n) (u) = ( du
) g(u) = ( du
) f (a + u) = f (n) (a + u). Thus g (n) (0) = f (n) (a).
Insertion into the above expansion, together with g(u) = f (a + u) and u = x a, then gives:

f (n) (a)
(xa)n + RN +1
n!
n=0
Z xa
1
=
f (N +1) (a + z)(xaz)N dz
N! 0
1 Z x (N +1)
=
f
(y)(xy)N dy
N! a

f (a + u) =
RN +1

N
X

put z = y a

Notes:
One can regard the generalized Taylor series as an approximation of f for values of x
close to a, with the remainder term RN +1 (x) indicating the precise difference between the
polynomial approximation up to order N and the true function f
If the n-th derivative of f is bounded on the interval [a R, a + R], i.e. |f (n) (x)| Cn for
all x [a R, a + R], then |RN +1 (x)| Cn (x a)N +1 /(N +1)!
Proof:
1 Z x (N +1)
1 Zx
|f
(y)|(xy)N dy Cn
(xy)N dy
N! a
N! a
h
ix
1
Cn
(xa)N +1
= Cn
(xy)N +1 =
a
(N +1)!
(N +1)!

|RN +1 (x)|

For small values of |xa|, the correction is therefore generally much smaller than any of
the previous terms in the approximating polynomial.
7.3. Examples
7.3.1. Series expansions for standard functions
We can now derive the power series for arbitrary functions
(including the ones we simply stated in the past):
f (x) = ex :
d n x
) e = ex , so f (n) (0) = e0 = 1. Hence
Derivatives: f (n) (x) = ( dx

ex =

N
X

xn
+ RN +1 (x)
n=0 n!

RN +1 (x) =

1 Zx y
e (xy)N dy
N! 0

123
So if limN RN +1 (x) = 0, which is here true for all x IR:
ex =

1
1
xn
= 1 + x + x2 + x3 + . . .
2
6
n=0 n!

f (x) = sin(x):
Derivatives: f (2m) (x) = (1)m sin(x) and f (2m+1) (x) = (1)m cos(x), so f (2m) (0) = 0 and
f (2m+1) (0) = (1)m . Hence
(N 1)/2

sin(x)

m=0

(1)m x2m+1
+ RN +1 (x)
(2m+1)!

RN +1 (x) =

R
(1)` 0x cos(y)(xy)N dy
N!
R
1
(1)`+1 0x sin(y)(xy)N dy
N!

if

N = 2`

if

N = 2` + 1

So if limN RN +1 (x) = 0, which is here easy to prove for all x IR, since sin(y) [1, 1]
and cos(y) [1, 1]:
sin(x) =

1
1 5
(1)m x2m+1
= x x3 +
x + ...
3
120
m=0 (2m+1)!

f (x) = ln(1 x):


Derivatives: f (1) (x) = (1 x)1 , f (2) (x) = (1 x)2 , f (3) (x) = 2(1 x)3 , f (4) (x) =
2.3(1x)4 . Further iteration: f (n) (x) = (n1)!(1x)n for n 1 (this one can also
prove by induction). So f (n) (0) = (n1)! for n 1, and f (0) = 0. Hence
N
X

xn
ln(1x) =
+ RN +1 (x)
n=1 n

RN +1 (x) =

Z x
0

(xy)N
dy
(1y)N +1

So if limN RN +1 (x) = 0, which is here true for all |x| 1:


ln(1x) =

xn
1
1
1
= x x2 x3 x4 + . . .
2
3
4
n=1 n

7.3.2. Indirect methods for finding Taylor series


If you only need the first few terms of a Taylor series, combine & manipulate series you know!
(add, multiply, change signs, divide, differentiate, integrate, etc)
but keep track of which powers you keep on board and which you dont, for consistency
Examples:
ex

1
1
= 1 x + x2 x3 + . . .
2
6

124
1
1
1
ln(1 + x) = x x2 + x3 x4 + . . .
2
3
4
1
1+x
1
cos(x)

d
d
1
1
1
ln(1+x) =
x x2 + x3 x4 + . . .
dx
dx
2
3
4
2
3
= 1 x + x x + ...


1
1


=
1 4
1 4
1
+ 24 x + . . .
x + ...
1 + 12 x2 + 24

 
2
1
1
1
1
= 1 x2 + x4 + . . . + x2 + x4 + . . . + . . .
2
24
2
24
1 2
1 4
1 4
= 1 + x x + ... + x + ...
2
24
4
1 2
5 4
= 1 + x + x + ...
2
24
=

1 2
x
2

1 2

1 3

1 4

(1 + x) = e ln(1+x) = e(x 2 x + 3 x 4 x +...)




2
1
1
1
1 
1
1
1
= 1 + x x2 + x3 x4 + . . . + 2 x x2 + x3 x4 + . . . + . . .
2
3
4
2
2
3
4
1 2 2
1 2
= 1 + x x + . . . + x + . . .
2
2
1
= 1 + x (1)x2 + . . .
2

1
3
1
1
= (1 + x) 2 = 1 x + x2 + . . .
2
8
1+x

tan(x)

1
x 61 x3 + 120
x5 + . . .
sin(x)
=
1 4
cos(x)
1 12 x2 + 24
x + ...



1 3
1 5
1 2
5 4
= x x +
x + ... 1 + x + x + ...
6
120
2
24
1 3
5 5 1 3
1 5
1 5
=x+ x + x x x +
x + ...
2
24
6
12
120
1
2
= x + x3 + x5 + . . .
3
15

125
7.4. LHopitals rule
We saw earlier how power series can be used to calculate nontrivial limits,
for example:
lim

x0

lim

x0

1
3

2 5
2 4
1 + 13 x2 + 15
x + 31 x3 + 15
x + ...
x + ...
=
lim
=2
1
1
1
1
2
x0
x0 1 + x x + . . . 1
8x + . . .
2
8
2

tan(x)
1+x1

= lim

tan(3x) sin(x)
= lim
x0
2x3

1
3

3x + 13 (3x)3 + . . . x 61 x3 + . . .
3

x + 3x x +
x0
2x3

= lim

1 3
x
6

2x3
+ ...

= lim

19 3
x
6

x0

+ ...
19
=
3
2x
12

Let us finally use power series to inspect more generally what can be said
about limits of fractions:
f (0) + xf 0 (0) + 21 x2 f 00 (0) + . . .
f (x)
= lim
lim
x0 g(0) + xg 0 (0) + 1 x2 g 00 (0) + . . .
x0 g(x)
2
Thus, if f (0) = g(0) = 0 (the nontrivial limits, where Fermats simple rule fails):
xf 0 (0) + 12 x2 f 00 (0) + . . .
f (x)
lim
= lim 0
x0 g(x)
x0 xg (0) + 1 x2 g 00 (0) + . . .
2
f 0 (0) + 12 xf 00 (0) + . . .
f 0 (0)
=
x0 g 0 (0) + 1 xg 00 (0) + . . .
g 0 (0)
2

= lim

This result is called lHopitals rule.


Notes:
(i) Never forget that lHopitals rule applies only when f (0) = g(0) = 0
(ii) In contrast to using power series directly,the rule works only if g 0 (0) 6= 0
Examples of the use of lHopitals rule:
h

lim

x0

a 1
x

sin(x)
x0
x
lim

7x

lim

x0

ln(a)ex ln(a)

= lim

1 sin(x/3)
=
ln(1 + x)

x=0

x0

= ln(a)

sin0 (0)
= cos(0) = 1
1
h

7e7x

cos(x/3)

[(1 +

x)1 ]

x=0

x=0

=7

126
8. Exercises
CALCULUS 1 TUTORIAL EXERCISES I
? 1. Use the induction method to prove the following statements:
n
X

(n IN) :

k4 =

k=1

1
n(n + 1)(2n + 1)(3n2 + 3n 1)
30

n + 1 2n (n + 1)!

(n IN) :

(n IN, n 4) : n2 2n n!
2. Express the following three complex numbers in the form a + ib, with a, b IR:
(3 + i) (2 + 6i)

(1 + i)(1 + 2i)

(2 i)2

3. Let z and z be a conjugate pair of complex numbers. Prove the following three identities:
1
1
Re(z) = (z + z)
Im(z) = (z z)
(z) = z
2
2i
4. Prove the following two statements:
(z, w C
| ): z+w =z+w
(z, w C
| ) : z.w = z.w
5. z and z be a conjugate pair of complex numbers. Prove the following claims:
zz IR

zz 0

Prove that z.z = 0 if and only if z = 0.


? 6. Let z, w C
| . Prove the following four statements:
|Re(z)| |z|

|Im(z)| |z|

|z.w| = |z|.|w|

|z| = |z|

7. Let z, w C
| , with w 6= 0. prove the following statements:
z/w = z/w

|z/w| = |z|/|w|

8. Express the following five complex numbers in the form a + ib, with a, b IR:
i3

i4

i5

i63

(i)10

? 9. Express the following five complex numbers in the form a + ib, with a, b IR:

1
2+i
1i
(1+i 3)3
(1 i 3)3
23i
1+3i
1+i
3
10. Find all solutions of the equation z 1 = 0. Hint: write z 3 1 as the product of (z 1)
and another factor.

127
CALCULUS 1 TUTORIAL EXERCISES II
11. Simplify the four complex numbers (1 + i)4 , (1 i)4 , (1 + i)4 , and (1 i)4 . Use your
results to find all solutions of the equation z 4 + 1 = 0.
12. Let z = 1 + 2i. Calculate the following seven complex numbers, and plot them in the
complex plane (i.e. on an Argand diagram):
z, 1/z, 1/z, z + z, z z, zz, and z/z.
13. What are the modulus and the principal values of the arguments of the following complex
numbers?

(a)
1+i
(b) 1 i
(c) 2 2i (d) 1 + 3i

(f) 1 i
(g) e2i/3
(h) 2ei/4
(e) 1 3i
(i)

e3i/4

(j)

e2+i/2

? 14. Let z = rei , with r, IR and r 0. Find all values of r and such that z 5 = 1. Plot
the corresponding numbers in the complex plane.
? 15. Let z = rei , with r, IR and r 0. Find all values of r and such that z 6 = 64. Plot
the corresponding numbers in the complex plane.
? 16. Find all solutions z C
| of the equation (z 1)5 = 1. Plot the corresponding numbers in
the complex plane.
? 17. Find all solutions z C
| of the equation (z i)6 = 64. Plot the corresponding numbers in
the complex plane.
18. Let a, b and u denote real numbers. Find the real and the imaginary parts of the following
complex numbers:
(i) ea+ib

(ii)

eua eibu
a + ib

(iii) ee

19. Sketch the following sets in the complex plane:


A = {z C
| | |z i|2 = 4}
B = {z C
| | (z i)2 = 4}
C = {z C
| | 1 < |z + 1| 3}

iu

128
CALCULUS 1 TUTORIAL EXERCISES III
n
k
= 1 2n for all n IN. Use this result to determine
? 20. Prove by induction that
k=1 2
P
P
the value of the series k=1 2k . Give an example of a series of the form nk=1 ak , with
P
finite values of the sum for any finite n, but such that the summation
k=1 ak does not
exist.

n
21. Show that for x IR the series representation ez =
n=0 z /n! of the exponential function
has the following properties (here and in subsequent exercises you will be allowed to
interchange differentiation and summation):

d ax
e = aeax
dx

e0 = 1

? 22. Use the series representations of the trigonometric functions, i.e.


sin(z) =

k=0

(1)k z 2k+1
(2k+1)!

and cos(z) =

the following properties for IR:


d
sin() = cos(),
d
d
cos() = sin(),
d

k=0

(1)k z 2k
,
(2k)!

to derive

sin(0) = 0
cos(0) = 1

23. Give the values of cos() and sin() for the following choices of the angle:
= 0, 6 , 4 , 3 , 2 , 2
, 3
, 5
, , 5
, 3
, 5
3
4
6
4
2 3
Give your values as fractions in terms of 2, 3 etc.
(i.e. not as decimals as given by a calculator)

? 24. Show that cos(/12) =


2 + 3.
(Hint: use the formula for cos(2) in terms of cos())
1
2

25. Find all solutions IR (if any) of the equation cot() + tan() = , for the following
three cases:
(i) = 1

(ii) = 2

(iii) = 4

26. Let , IR. Rewrite each of the following expressions as a constant times a product of
trigonometric functions:
(i) sin( ) + sin( + )
(ii) cos( ) cos( + )

129
CALCULUS 1 TUTORIAL EXERCISES IV
27. Express the following combinations of functions in the form c sin( + ), where c and
are real constants which you have to find (assume || < /2):

(a) sin() + cos()


(b) 3 sin() cos() (c) sin() tan() cos()
28. Find the values of
(a) arcsin(sin(5/6))

(b) arcsin(sin(7/6))

(c) arccos(cos(7/6))

(d) arccos(cos(11/6))
? 29. Show that the series representations of sinh and cosh are:
sinh(z) =

z 2k+1
k=0 (2k + 1)!

cosh(z) =

z 2k
k=0 (2k)!

30. Use the above series representations of the hyperbolic functions (rather than the definition
in terms of exponentials) to re-derive the following for x IR:
d
sinh(x) = cosh(x),
dx
d
cosh(x) = sinh(x),
dx

? 31. Show that

d
dx

sinh(0) = 0
cosh(0) = 1

tanh(x) = 1 tanh2 (x).

32. Simplify the following complex numbers in which , IR to the standard form a + ib,
with a, b IR:
(a) cosh(i)

(b) sinh(i/3)

(c) tanh(i/6)

(d) sin( + i)

(e) cos( i)

(f) cosh()

? 33. Find formulae that express the hyperbolic functions sinh(x), cosh(x) and tanh(x) as ratios
of polynomials of t = tanh(x/2).
34. Calculate the following three derivatives:
(a)

d
dx

arccosh(x)

(b)

d
dx

arcsinh(x)

(c)

d
dx

arctanh(x)

130
CALCULUS 1 TUTORIAL EXERCISES V
? 35. Let z C
| , z 6= 1. Prove the following statement by induction:
n
X
k=0

zk =

1 z n+1
1z

Now choose z = ei with IR, and use the identity that you have just proven to find
P
P
formulas for the two sums nk=0 sin(k) and nk=0 cos(k). What if is a multiple of 2?
Check the correctness of your formulas for the simplest cases n = 0 and n = 1.
? 36. Use the explicit formula for the function arcsinh : IR IR, namely arcsinh(y) =

ln(y + y 2 + 1), to show that


arcsinh(sinh(x)) = x for all x IR
sinh(arcsinh(y)) = y

for all y IR

37. Use the explicit formula for the function arccosh : [1, ) [0, ), namely arccosh(y) =

ln(y + y 2 1), to show that


arccosh(cosh(x)) = x for all x [0, )
cosh(arccosh(y)) = y

for all y [1, )

38. Use the explicit formula for the function arctanh : (1, 1) IR, namely arctanh(y) =
1
ln[(1 + y)/(1 y)], to show that
2
arctanh(tanh(x)) = x for all x IR
tanh(arctanh(y)) = y

for all y (1, 1)

131
CALCULUS 1 TUTORIAL EXERCISES VI
n

39. Let n IN, and calculate the limit limn 1 + n1 .


Hint: substitute n1 = x and try to use limits calculated in the lectures.
40. Calculate the following limits, if they exist, without using power series:
(a)
(c)

1 cos(x)
x0
x2
sin(x2 )
lim
x0
x
lim

(b)
(d)

lim

x0

sin(7x)
tan(2x)

lim cosh(x)

x0

41. Calculate the following limits, if they exist, using the power series representations of the
hyperbolic functions:
(a)
(c)

sinh(x)
x
1 cosh(x)
lim
x0
x2
lim

x0

(b)
(d)

1 cosh(x)
x
tanh(x)
lim
x0
x
lim

x0

42. Calculate the following limits by making clever substitutions:


(a)
(c)

sin(x)
x
arcsin(3x)
lim
x0
tan(5x)
lim

(b)
(d)

arcsin(x)
x0
x
arctanh(x)
lim
x0
x
lim

? 43. Calculate the following limits, if they exist, using any suitable method:
(a)
(c)

sin(x)
x
lim x(e1/x 1)
lim

x
x

ln(x)
x2 + 1

(e)

lim
x

(g)

lim e1/x

(i)
(k)

x0

lim

a b
(a > b > 0)
x0
x
lim e1/x
lim

(f)
(h)

x0

(d)

cos(x)
x2
x
lim x

(b)

(j)
(l)

lim

x
x0

xx 1
x0 x ln(x)
lim xsin(x)
x0

(x + 1) ln(x)
x0
sin(x)
lim xe1/x
x
lim

132
CALCULUS 1 TUTORIAL EXERCISES VII
? 44. Let fk (x) : IR IR denote n arbitrary functions, with k = 1, 2, . . . , n, and let
Qn
k=1 fk (x) = f1 (x).f2 (x). . . . .fn1 (x).fn (x). Prove by induction the following generalized
version of the product rule:
+

for all n ZZ :

d
dx

n
Y

fk (x) =

k=1

n
X
f`0 (x)
`=1

f` (x)

n
Y

fk (x)

k=1

? 45. Now prove the previous generalized product rule directly (i.e. without induction), for the
special case where fk (x) > 0 for all x IR and all k. Hint: write f (x) = eg(x) with
g(x) = ln(f (x)), and use the chain rule.
46. calculate the following derivatives:
(a)

d x2
e
dx

(b)

d
ln(tan(x))
dx

(c)

d
(x ln(x) x)
dx

(d)

d
arcsin(x2 )
dx

(e)

d
arctan(ex )
dx

(f)

d sin(x)
e
dx

(g)

d
xearctan(x)
dx

(h)

d
ln |arcsin(x)|
dx

47. calculate the following derivatives:


d
1x
arcsin
dx
1+x


(a)
(c)
(e)
(g)

d
ln |x + x2 a2 |
dx


d 1
arccos 1 x2
dx x
d
ln (cosh(x) + sinh(x))
dx

(b)
(d)

d sin(x)
3
dx

d
1+x
dx
x

(f)

d x sin(x)
x
dx

(h)

q

d
2
2
ln
(1 + x )/(1 x )
dx

133
CALCULUS 1 TUTORIAL EXERCISES VIII
? 48. Each of the following is an equation that determines y as an implicit function of x. Find
in all cases an expression for dy/dx.
(a) x2 + y 2 = 1

(b) y 3 + x3 = 1

(c)

sinh(x) + cosh(y) = 1

(d) xy ex+y = 2

(e)

sin(y) + y = x3

(f) y 2 + x(x 1)(x + 1) = 0

49. Each of the following are a pair of equations which determine x and y in terms of a
parameter t IR, which defines a function y(x) implicitly. Find in all cases an expression
for dy/dx.
(a) x = cos(t), y = sin(t)

(b) x = cosh(t), y = sinh(t)

(c)

x = t + sin(t), y = cos(t)

(d) x = t3 , y = t2

(e)

x = e2t , y = tanh(t)

(f) x = et cos(t), y = et sin(t)

? 50. Calculate the integral A = 0b eax dx (with a > 0 and b > 0) using the sandwich method,
with suitably constructed bounding staircase functions, similar to how this was done in
R
the lectures for integrals such as 0b cos(x)dx and others. Hint: use staircases with steps of
equal size h, and evaluate the summations that show up using the formula in exercise 39,
i.e.
n
X
1 z n+1
z 6= 1 :
zk =
1z
k=0
R

51. Find a primitive for each of the following functions, i.e. a function F (x) such that
F 0 (x) = f (x). Prove your claims.
(a) f (x) = 1/(1 + x)
(c)

f (x) = 1/(2x + 1)3

f (x) = 1/(1 x2 )

(g) f (x) = 1/ 1 + x2

(e)

(b) f (x) = 1/(1 x)2


(d) f (x) = 1/(1 + x2 )

(f) f (x) = 1/ 1 x2
1 2

(h) f (x) = xe 2 x

134
CALCULUS 1 TUTORIAL EXERCISES IX
52. Calculate the following indefinite integrals using the method of substitution:
Z
ex dx
(a)
put x = ln(y)
x
1
+
e
Z
1 x2 dx
put x = sin(y)
(b)
xdx
1 x2
Z
dx

(d)
2
a + x2
Z
2x3 dx
(e)
1+x
Z
xdx

(f)
1+ x
(c)

put x2 = 1 y
put x = at
put x = t 1
put x = t2

53. Calculate the following indefinite integrals using the method of substitution:
Z
Z
2xdx
(a)
cos(sin(x)) cos(x)dx (b)
1 + x4
Z
Z
x
dx
e dx
(d)
(c)

2x
e 1
2 x 1x
Z
Z
xdx
x2 dx

(e)
(f)
1 x2
1 x2
Z
Z
dx
xdx
(g)
(h)
2
3/2
(1 x )
(1 x2 )3/2
Z
Z
dx
x2 dx
(j)
(i)
(1 x2 )3/2
cos(x)
? 54. Calculate the following indefinite integrals using integration by parts:
(a)
(c)
(e)

Z
Z
Z

x sinh(x)dx

(b)

arcsin(x)dx

(d)

2

ln(x) dx

(f)

(1 + x2 )ex dx

x3 sin(x)dx

x arctan(x)dx

135
CALCULUS 1 TUTORIAL EXERCISES X
? 55. Define the indefinite integrals In (x) for n ZZ+ as follows:
In (x) =

dx
(1 + x2 )n

Use integration by parts to derive the following recursion formula:


x
2n3
In1 (x)
In (x) =
+
2(n1)(1+x2 )n1 2n2
Use the recursion formula to find the indefinite integral (1 + x2 )2 dx.
R

56. Define the indefinite integrals In (x) for n ZZ as In (x) = cosn (x)dx. Use integration by
parts to derive the following recursion formula:
1
n1
In (x) = sin(x) cosn1 (x) +
In2 (x)
n
n
R
Use the recursion formula to find the indefinite integral cos4 (x)dx.
R

? 57. (i) Define the definite integrals In (a) for n ZZ+ and a IR as In (a) =
x2n eax .
Assume that differentiation with respect to a can be moved from outside to inside
the integral. Use repeated differentiation with respect to a to find a formula that
expresses In (a) in terms of I0 (a).

R y2
(ii) Show that I0 (a) = C/ a, where C =
e dy. From now on you may use (without

proof) that C = .

R 2 x2 /2
R 4 x2 /2
(iii) Use the previous results to show that
xe
= 2 and that
xe
=

3 2.

58. Use the method of partial fractions to calculate the following integrals:
Z
Z
dx
xdx
(a)
(b)
2
x +x6
(1 x)2 (1 + x)
(c)

dx
2
2x + x 1

(d)

12x2

xdx
7x 12

59. Calculate the length L of the curve in the plane described by the function y = ln(cos(x)),
between the points x = /4 and x = /4. Hint: use the result of an earlier exercise to
deal with the integral.

136
CALCULUS 1 TUTORIAL EXERCISES XI
60. Find out for the following series whether they are convergent or divergent, using the various
results stated and/or derived in the lectures or otherwise:
(a)

(c)

n3

(1)n n2

(b)

(d)

1/ n

n en

( > 0)

? 61. Find the radii of convergence for the following powers series:
(a)

(c)

xn /n3

(1)n xn /n2

(b)

(d)

xn / n

xn n en

( > 0)

62. Derive the Taylor expansions for the following functions, up to order N = 3 for the first
three functions and up to order N for the last one, and give in each case an exact expression
for the remainder term:
(a) f (x) = tan(x)
(c)

f (x) = x3

(b) f (x) = tanh(x)

(d) f (x) = 1+x esin(x)

? 63. Write the following two functions as power series of the form
for each the associated radius of convergence:
(a) f (x) = 1/(1x)2

n=0 bn x

, and determine

(b) f (x) = arctan(x)

Substitute x = 1 into your result under (b) (why is it not obvious that this is allowed?)
and derive an expression for the number as a series (the so-called Leibniz series).
64. Find the first three terms (i.e. up to order x3 ) of the Taylor expansions for the following
two functions, by combining and/or manipulation the Taylor expansions of other functions
that you know:


1+2x
(a) f (x) = tanh(x)
(b) f (x) = ln
12x

You might also like