MA270 - Analysis 3

©2024 - Jose Rodrigo

April 15, 2025

Contents

Contents i

1 Introduction 2
1.1 Review of limits of sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Review of continuity and differentiability . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Review of integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Sequences and Series of Functions 7

2.1 Pointwise convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Uniform convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Series of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 A continuous, nowhere differentiable function . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5 Space filling curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.6 Absolute Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Basic results about Rn 20

3.1 Notation in Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 The Euclidean norm and inner product . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Convergence in Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4 Subsequences and the Bolzano-Weierstrass theorem . . . . . . . . . . . . . . . . . . . . . 24
3.5 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5.1 Definitions of continuity and continuous limit . . . . . . . . . . . . . . . . . . . . 25
3.5.2 Separate continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5.3 Basic properties of continuous functions . . . . . . . . . . . . . . . . . . . . . . . 27
3.5.4 Constructing continuous functions of several variables from continuous real valued
functions of a single real variable. . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5.5 Caution with taking limits in dimension ¥ 2 . . . . . . . . . . . . . . . . . . . . . 29

4 Rudiments of topology of Rn and Continuity 32

4.1 Closed and open subsets of Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Continuity and topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.1 Continuity in terms of open sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.2 Continuity and sequential compactness . . . . . . . . . . . . . . . . . . . . . . . . 35

5 The space of linear maps and matrices 37

5.1 Two norms on the space of linear maps and matrices . . . . . . . . . . . . . . . . . . . . 38
5.1.1 Comparison of the two norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.1.2 Properties of the operator norm . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Convergence and continuity in LpRn , Rk q . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2.1 Continuity of functions involving matrices or linear maps . . . . . . . . . . . . . . 40
5.3 The space GLpn, Rq € LpRn , Rk q of invertible linear transformations . . . . . . . . . . . 41

i
 CONTENTS

6 The Derivative 43
6.1 Directional derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.1.1 Directional derivative and continuity . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.2 The (Fréchet) Derivative as an affine linear approximation . . . . . . . . . . . . . . . . . 44
6.2.1 Affine linear approximation in the 1-variable case . . . . . . . . . . . . . . . . . . 44
6.2.2 The (Fréchet) Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.2.3 Differentiability of components of vector-valued functions . . . . . . . . . . . . . . 45
6.2.4 Relation between the derivative and directional derivative . . . . . . . . . . . . . . 45
6.3 Partial derivatives, gradient and Jacobian matrix . . . . . . . . . . . . . . . . . . . . . . . 46
6.3.1 Algebraic rules for partial derivatives. . . . . . . . . . . . . . . . . . . . . . . . . 46
6.3.2 Gradient and Jacobian matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.3.3 Why so many different notations for the same thing?! . . . . . . . . . . . . . . . . 48
6.4 The Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4.1 Jacobian form of chain rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.4.2 Calculating with the chain rule and gradient . . . . . . . . . . . . . . . . . . . . . 51
6.5 Continuity of partial derivatives implies differentiability . . . . . . . . . . . . . . . . . . . 52
6.5.1 The space of continuously differentiable functions . . . . . . . . . . . . . . . . . . 53
6.6 Proof of Theorem 6.25 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

7 Complex Analysis 56
7.1 Review of basic facts about C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
7.2 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.2.1 The exponential and the circular functions . . . . . . . . . . . . . . . . . . . . . . 64
7.2.2 Argument and Log . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.3 Complex integration, contour integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.3.1 Links with Green’s and Gauss’ Theorems . . . . . . . . . . . . . . . . . . . . . . . 72
7.3.2 Consequences of Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.3.3 Applications of Cauchy’s formula to evaluate integrals in R . . . . . . . . . . . . . 82

Bibliography 88

ANALYSIS III 1
Chapter 1

Introduction

MA270 - Analysis 3 is a self-contained module. Only material lectured in class will be examined. Non-
examinable will be clearly marked.
The course covers the following topics:
• Pointwise and uniform convergence (sequences and series of functions).
• Differentiation.
• Complex valued functions.

The module does not follow any specific source. There references at the end of the notes cover most
of the topics in the module. I would be happy to supply a list of references that can be used to expand
any of the Chapters in the notes.
The course relies heavily on material covered in MA141 - Analysis 1 and MA139 - Analysis 2. Below is
a list of the most relevant topics from each of them.
MA141 - Analysis 1
• Section 1.6 The triangle inequality
• Chapter 2 Sequences and convergence (specially Section 2.6 Subsequences and The Bolzano-Weierstrass
Theorem and Section 2.7 Cauchy sequences)
• Chapter 3 Summation of series (specially 3.1, 3.2, 3.5, 3.6)
• Chapters 4 Continuity (specially 4.1 – 4.7)
• Section 5.4 (will not be assumed as it was part of the supplementary material)

MA139 - Analysis 2
• Section: Basic properties of power series
• Section: The continuity of power series
• Section: The derivative
• Section: The diferentiability of power series
• The radius of convergence formula (page 93)
• The Riemann integral, construction and basic properties
• Uniform continuity (page 70)

The notes for those modules are available in the Moodle page of MA270 and students are encouraged
to review the material when relevant. Below is a brief summary of those results.

2
 CHAPTER 1. INTRODUCTION

1.1 Review of limits of sequences

In MA141 you have studied the convergence of sequences of real numbers (Chapter 2). Here is a quick
recap.

Definition 1.1. A sequence pan q converges to a limit l P R if for every ε ¡ 0 there exists N P N such that
|an
l| ε for every n ¥ N.

In this case we write an Ñ l as n Ñ 8.

The following results were covered in MA141:

• Uniqueness of limits (Lemma 2.3): A sequence can have at most one limit.

• The shift rule (Lemma 2.5): For any fixed k P N, an Ñ l as n Ñ 8 if and only if an k Ñ l as
n Ñ 8. (This effectively means that for the question of convergence, we can disregard the first k
terms of the sequence (for any fixed k we like).

• Convergent sequences are bounded (Lemma 2.6): Any convergent sequence is bounded.

• (Lemma 2.7): If an Ñ a then |an| Ñ |a|.

• The basic algebra of limits (Lemma 2.9): Suppose an Ñ a and bn Ñ b. Then
(i) an bn Ña b;
(ii) an bn Ñ ab;
(iii) if b  0 then an {bn Ñ a{b.

• Limits and inequalities (Lemma 2.10): If an¤ bn for all n, an Ñ a and bn Ñ b then a ¤ b.
• Sandwich rule (Lemma 2.12): an ¤ bn ¤ cn with an Ñ l and cn Ñ l implies that bn Ñ l.

1.2 Review of continuity and differentiability

We review the notion of continuity (Chapter 4 in MA141) and uniform continuity (covered in page 70 in
MA139).

Definition 1.2. Given f : Ω € R Ñ R, we say that f is continuous at x P Ω if for every ε ¡ 0 there exists
δ  δ px, εq ¡ 0 such that

y P Ω and |x
y| δ ùñ |f pyq
f pxq| ε. (1.1)

The key point to note from the definition above is that given a function f , ε ¡ 0 and a point x there
exists δ, but δ can depend on ε and x (and of course f ).

Definition 1.3. Given f : Ω Ñ R, we say that it is uniformly continuous if for every ε ¡ 0 there exists
δ  δ pεq ¡ 0 such that
x, y P Ω and |x
y | δ ùñ |f py q
f pxq| ε. (1.2)

The key point here is that δ can be chosen independently of x.

In the case in which Ω  ra, bs we have the following result.

Theorem 1.4. Let f : ra, bs Ñ R be a continuous function. Then it is uniformly continuous.

ANALYSIS III 3
 CHAPTER 1. INTRODUCTION

Before we prove the result let’s consider a couple of examples in which the closed, bounded interval
ra, bs is replaced by an unbounded or an open domain.
Consider f pxq  ex , defined in R. Clearly this is a continuous function, but not uniformly continuous.
Indeed, since f grows faster and faster for larger x it is possible to find arbitrarily small intervals in which
f changes by at least ε. This example shows that the result in Theorem 1.4 is not necessarily true for
unbounded domains.
We can also consider g pxq  x1 on p0, 1q. Just as in the previous example, near zero, the function g
grows to infinity faster and faster as we approach the origin, making it impossible to find δ independent of
x that satisfies (1.2).
This result, in much more generality, not just for closed intervals on R will be proven in MA260 Norms,
Metrics and Topologies. The key point is that the domain of f is a compact set (which in this case is
equivalent to closed and bounded).

Proof of Theorem 1.4. We will argue by contradiction. That would mean that there exist ε ¡ 0 and xn , yn
such that |xn
yn | ¤ n1 but |f pxn q
f pyn q| ¡ ε.
The sequences txn u and tyn u are bounded, as they are in ra, bs, and therefore we can apply Bolzano–
Weierstrass1 to obtain convergent subsequences txnk u8 8
k1 to x and tynk uk1 to y. Notice that

|x
yn | ¤ |x
xn | |xn
yn | ¤ |x
xn |
k k k k k
1
nk
ÝÝÝÝÝÝÝ
kÑ8
Ñ 0,
which implies that x  y. However we know that |f pxnk q
f pynk q| ¡ ε for all k. Since f is continuous,
taking limits as k goes to infinity we obtain 0  |f pxq
f pxq| ¡ ε, which is a contradiction.

For completeness we reproduce the definition of derivative from MA139

Definition 1.5. Suppose f : I Ñ R is defined on an open interval I and c P I. We say that f is

differentiable at c if
f pc hq
f pcq
lim
hÑ0 h
exists. If so we call the limit f 1 pcq.

Also from that module

Lemma 1.6. Suppose I is an open interval, f : I Ñ R and c P I. Then f is differentiable at c if and only
if there exists a number A and a function ε with the properties that for all x

f pxq
f pcq  Apx
cq εpxqpx
cq,

εpcq  0 and ε is continuous at c: (εpxq Ñ 0 as x Ñ c). If that happens A  f 1 pcq.

1.3 Review of integration

To construct the integral on an interval ra, bs we consider partitions of the interval.
In practice, a partition of the interval ra, bs is determined by a collection of points txi uni0 , for some n
such that
a  x0 x1 . . . xn
1 xn  b,
which yields the collection of intervals Ij  rxj
1, xj s, for j  1, . . . , n. Given a partition of P 
tI1, . . . , Inu of I  ra, bs we denote
M  supf m  inf f
I
Mk  supf mk  inf
I
f.
I Ik k
1 n
Theorem: Bolzano–Weierstrass. In R every bounded sequence has a convergent subsequence. See the Appendix for a
sketch of a proof

ANALYSIS III 4
 CHAPTER 1. INTRODUCTION

Definition 1.7. Given f : ra, bs Ñ R and a partition P  tI1, . . . , Inu of ra, bs we define the upper
Riemann sum of f with respect to P as
ņ
U pf, P q : Mk |Ik |,
k 1 
and the lower Riemann sum of f with respect to P as
ņ
Lpf, P q : mk |Ik |.
k 1 
Figure 1.1 shows the intuitive idea for calculating an integral, displaying the Lower and Upper Riemann
sums, for a uniform partition with 10 intervals (for f pxq  x2 ).
x2
x2
1.

0.8

0.6

0.4

0.2

0. x 0 x
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Figure 1.1: Lower (left) and Upper (right) Riemann sum of f

We will denote by P the set of all partitions of ra, bs.

Definition 1.8. Given f : ra, bs Ñ R, bounded, we define the upper Riemann integral of f by

U pf q : inf U pf, P q.
P
P P

We define the lower Riemann integral of f by

Lpf q : sup Lpf, P q.

P
P P

Definition 1.9. Given f : ra, bs Ñ R bounded we say that it is Riemann integrable if and only if Lpf q 

b
b
U pf q, and define its Riemann integral, denoted by a f pxqdx or a f , by
 b
f pxqdx : Lpf q  U pf q.
a

The following result will also prove useful in showing that a function is integrable.
Theorem 1.10. Let f : ra, bs Ñ R be a bounded function. Then f is integrable if and only if for every
ε ¡ 0 there exists a partition P of ra, bs such that

U pf, P q
Lpf, P q ε

Theorem 1.11. Let f : ra, bs Ñ R be a continuous function. Then it is Riemann integrable.

Theorem 1.12. Left f, g : ra, bs Ñ R be Riemann integrable functions, and c P R. Then f g and cf
are Riemann integrable and we have
 b  b  b  b  b
cf c f, pf gq  f g.
a a a a a

ANALYSIS III 5
 CHAPTER 1. INTRODUCTION

Theorem 1.13. Let f, g : ra, bs Ñ R be Riemann integrable functions such that f ¤ g. Then
 b  b
f ¤ g.
a a

Theorem 1.14. Let f : ra, bs Ñ R be an integrable function. Then |f | is integrable and we have
  
 b  b

 f  ¤ |f |.
a a

The Fundamental Theorem of Calculus explores the relationship between integration and differentiation,
and how under sufficient conditions they can be understood as inverse operations. The first result we
consider is when the integral of a derivative is the original function.

Theorem 1.15. Let F : ra, bs Ñ R be a continuous function that is differentiable on pa, bq with F 1  f.
Assume that f : ra, bs Ñ R is an integrable function. Then
 b
f pxqdx  F pbq
F paq.
a

Theorem 1.16. Let f : ra, bs Ñ R be an integrable function and define the function F : ra, bs Ñ R by
 x
F pxq : f ptqdt.
a

Then F is continuous of ra, bs. Additionally if f is continuous at c P ra, bs then F 1pcq  f pcq, with the
derivatives at a and b understood as one-sided derivatives.

ANALYSIS III 6
Chapter 2

Sequences and Series of Functions

In this Chapter we will consider sequences and series of functions and aspects relating to pointwise and
uniform convergence and its interactions with continuity, integrability and differentiability questions.

2.1 Pointwise convergence

We will consider sequences of functions fn : Ω Ñ R from a fixed domain Ω. Here we do not make any
assumptions about Ω, i.e. being open or closed, bounded or unbounded for example. While most examples
will be in one dimension, unless otherwise noted they apply to higher dimensions. We start by defining
pointwise convergence.
Definition 2.1. Let pfn q8
n1 be a sequence of functions, with fn : Ω Ñ R. We say that pfn q or fn
converges pointwise to f : Ω Ñ R if and only if for every x P Ω we have limnÑ8 fn pxq  f pxq. We will
denote pointwise convergence by fn Ñ f .
Example 2.2. Consider the sequence pfn q given by fn : r0, 1s Ñ R, fn pxq  x1{n .
1.0
f20
0.8

f4 f3 f2 f1
0.6

0.4

0.2

0.0 0.2 0.4 0.6 0.8 1.0

Figure 2.1: The sequence fn for n  1, 2, 3, 4 and 20.

Notice that fn p0q  0 for every n, but that for every x P p0, 1s we have limnÑ8 x1{n  1. As a result
the limit of the sequence pfn q is #
x  0,
f pxq 
0
1 x P p0, 1s.
Remark 2.3. Notice that the above example shows that the pointwise limit of a sequence of continuous
functions need not be continuous. It also produces a counterexample for the commutativity of the limits.
We have
lim lim fn pxq  lim lim fn pxq,
nÑ8 xÑ0 x Ñ0 nÑ8
as the left-hand side equals zero, while the right-hand side equals one.

7
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Pointwise convergence clearly does not preserve continuity. It can also be very non-uniform, in the sense
that while fn pxq Ñ 0 for every x we may have supx |fn pxq
f pxq| Ñ C ¡ 0 or even supx |fn pxq
f pxq| Ñ
8 as n goes to infinity, as shown in the next examples.
Example 2.4. Consider the sequences
$ $
'
&2nx x P r0, 2n 1
q ' 2
&2n x x P r0, 2n 1
q
gn pxq 
2npx
n1 q x P r 2n , n q
1 1 hn pxq 
2n2px
n1 q x P r 2n , n q
1 1
' '
x P r n1 , 1s x P r n1 , 1s.
% %
0 0

It is easy to see that gn and hn are continuous and converge to the function f  0. However, for every n
we have gn p1{p2nqq  1 (with that being the maximum of gn ) and therefore

sup |gn pxq
0|  1.

Pr s
x 0,1

The situation is worse for the sequence phn q, known as the Witch’s hat. Indeed hn p1{p2nqq  n, which
shows that while hn Ñ 0 we have
sup |hn pxq
0| Ñ 8.
Pr s
x 0,1

Pointwise convergence and integrability do not interact
as one
would hope. Indeed, even if we assume
that the pointwise limit is integrable we may not have lim fn  lim fn .
Example 2.5. Consider fn pxq  χrn,n 1q pxq, where χI is the indicator of the set I, i.e., takes value 1 if
x P I and zero otherwise. Clearly fn converges pointwise to f  0. However,
 
1 fn  f  0.
Here we are considering the integral as an improper integral over the real line. Notice that since every
function is 0 outside a bounded interval, this is straight forward. We can think of this, as “the mass scaping
to infinity” (along the x axis).
Another example of this phenomena,
can be found by considering gn pxq  nχp0,1{nq pxq we also have
that gn converges to 0, while having gn  1 for every n. We can think of this as “pointwise convergence
allowing the mass to go to infinity” (along the y axis this time). The Witch’s hat above also provides a
similar example, in this case with continuous functions.
Example 2.6. Another sequence that will play a role in several modules this year is fn pxq  sinpn xq.
This sequence is connected to Fourier series and will be discussed in MA265 for example. Notice that for
x  kπ with k P Z the limit exists and equals 0. If x  p{q π with p{q R Z then there is no limit. Indeed
sinpnqxq  0 while sinpp2nq 1qxq  sinpxq  0. If x is an irrational multiple of π, then the rest of the
division of nx by 2π is dense in r0, 2π s and there is no limit.
Despite the fact that sinpnxq does not have a limit for most x, you will see in MA265 that for every
integrable function f  π
f pxq sinpnxqdx Ñ 0 as n Ñ 8.

π
This result, known as the Riemann–Lebesgue Lemma, suggests that sinpnxq goes to zero in some sense
(known as the weak sense, which will be covered in Measure Theory, Functional Analysis and Fourier
Analysis). We can also consider the sequence gn pxq  cospnnxq . As cosine is a bounded function it is easy
to see that gn converges pointwise to 0. Since gn are smooth we can also consider gn1 pxq 
sinpnxq.
This tells us that even for smooth functions, having gn converge pointwise to g does not imply that gn1
converges to g 1 even if g is smooth.


The final example we consider is one of a sequence pfn q such that pfn
f qdx converges to zero, but
where fn does not converge pointwise to f .

ANALYSIS III 8
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Example 2.7. We will consider functions defined on r0, 1s. Let

f0 pxq  χr0,1s pxq,

f1 pxq  χr0,1{2s pxq, f2 pxq  χr1{2,1s pxq,

f3 pxq  χr0,1{4s pxq, f4 pxq  χr1{4,1{2s pxq, f5 pxq  χr1{2,3{4s pxq, f6 pxq  χr3{4,1s pxq.
Notice that each function is an indicator of an interval, and that in each group above the intervals sweep
r0, 1s. When we move to the next block the length of the corresponding intervals gets divided by 2 and
therefore we consider twice as many functions for each group. While the process is clear from the list
writing a formula for fn is annoying to say the least. You can check that the following works. For an index
k¸1
ķ
nPr 2l , 2l s, k  1, 2, . . .

l 0 
l 0

we set fn as the indicator of the interval

 °k
1 °k
1 
n

l 0 2l n

l 0 2l 1
, .
2k 2k


Since the length of the intervals tends to zero it is clear that fn Ñ 0, but since the intervals keep sweeping
the entire interval r0, 1s the sequence fn does not converge to zero (or any other function for that matter).
This is contrary to the intuition that if the area between f and fn is going to zero the functions fn must
be approaching, and therefore converging to f .

2.2 Uniform convergence

We now consider the notion of uniform convergence.

Definition 2.8. Let fn : Ω Ñ R be a sequence of functions. We say that pfn q converges uniformly to
f : Ω Ñ R if and only if for every ε ¡ 0 there exists N pεq such that |fn pxq
f pxq| ε for every x P Ω
and for all n ¡ N pεq.

The key different with pointwise convergence is that N depends only on ε and not on x. For pointwise
convergence we first froze x and consider the convergence of fn pxq to f pxq. We will denote uniform
convergence by fn Ñ f .
As before we are not making any assumption on Ω. In order to simplify the presentation we introduce
the notation
}f }8  sup |f pxq|.
P
x Ω

With this notation we have

fn Ñ f ðñ @ε ¡ 0, DN pεq such that }fn
f }8 ε @n ¡ N pεq.

Remark 2.9. Clearly uniform convergence implies pointwise convergence. The converse is of course false,
as we have seen in Remark 2.3.

Definition 2.10. A sequence pfn q of functions in Ω is called uniformly Cauchy if and only if for every ε ¡ 0
there exists N pεq such that }fn
fm }8 ε for all n, m ¡ N pεq.

Theorem 2.11. A sequence pfn q is uniformly convergent if and only if it is uniformly Cauchy.

ANALYSIS III 9
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Proof. Assume that pfn q is uniformly convergent to f , i.e. for every ε there exists N such that }fn
f }8
ε{2 for all n ¡ N . Then, for all m, n ¡ N

}fn
fm}8 ¤ }fn
f f
fm}8 ¤ }fn
f }8 }fm
f }8 ¤ ε{2 ε{2  ε.

For the converse, assume pfn q is uniformly Cauchy. That means that for every x, fn pxq is a Cauchy
sequence in R and therefore convergent. That means there exists f pxq such that fn pxq converges to f pxq
at least pointwise. Now, we know that given ε ¡ 0 there exists N pεq ¡ 0 such that |fn pxq
fm pxq| ε
for every x and all n, m ¡ N pεq. That is

fm pxq
ε fn pxq fm pxq ε for all x, and all n, m ¡ N pεq.

As the left-hand side holds for all m ¡ N pεq we can take limits as m goes to infinity. We find

f pxq
ε ¤ fn pxq ¤ f pxq ε for all x, and all n ¡ N pεq.

from which it follows that

|f pxq
fnpxq| 2ε for all x, and all n ¡ N pεq,

which proves the result.

Remark 2.12. While this topic will be discussed in more depth in MA260 Norms, Metrics and Topologies
it is worth noting that }  }8 is a norm in the space of bounded functions in Ω ( we make no assumptions
about it being open, closed, bounded or unbounded). }}8 is referred to as the supremum norm, or uniform
norm. Recall that by norm we mean that it satisfies

1. }f }8¥ 0, with }f }8  0 if and only if f  0,

2. }λf }8  |λ|}f }8 , for all λ P R, and

3. }f g }8 ¤ }f }8 }g }8 .

Theorem 2.13. Let pfn q be a sequence of continuous functions in Ω that converges uniformly to f : Ω Ñ R.
Then f is continuous.

Proof. First notice that the uniform convergence implies that given any ε ¡ 0 there exists N ¡ 0 such
that }fn
f }8 ε{3 for all n ¡ N . In order to show that f is continuous at x0 P Ω we need to show
that given ε there exists δ  δ pεq such that for all x P px0
δ, x0 δ q X Ω we have |f pxq
f px0 q| ε.
With N as above, we choose n ¡ N , fixed from now own. Since fn is continuous at x0 we know that
there exists δ  δ pεq such that for all x P px0
δ, x0 δ q X Ω we have |fn pxq
fn px0 q| ε{3.
We estimate |f pxq
f px0 q| using the triangle inequality

|f pxq
f px0q|  |f pxq
fnpxq fn pxq
fn px0 q fn px0 q
f px0 q|

¤ |f pxq
fnpxq| |fnpxq
fnpx0q| |fnpx0q
f px0q|

¤ }fn
f }8 |fnpxq
fnpx0q| }fn
f }8 ¤ 3ε 3ε 3ε ,
for n ¡ N and x P px0
δ, x0 δ q X Ω, with N and δ chosen as above. This completes the proof.

We will denote the space of bounded, continuous functions with the uniform norm by pCb ; }  }8 q.

Theorem 2.14. pCb ; }  }8 q is a complete space, i.e. every Cauchy sequence converges to a continuous
bounded function.

ANALYSIS III 10
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Proof. We need to show that if pfn q is Cauchy in the space, then there is a limit, and that the limit is
bounded and continuous. First notice that a Cauchy sequence in pCb ; }  }8 q is, by definition, a uniformly
Cauchy sequence. Theorem 2.11 implies that the sequence is convergent and since all the functions are
continuous Theorem 2.13 implies the limit is continuous.
To see that it is bounded, notice that for every x P Ω
|f pxq| ¤ |f pxq
fnpxq| |fnpxq|
for every n. Since fn converges uniformly to f there exists n large enough |fn pxq
f pxq| 1. For that
n, since fn is bounded we have |fn | ¤ M . These two inequalities lead to |f pxq| ¤ M 1 for every x P Ω,
proving the boundedness of f .

Remark 2.15. We could consider the interaction of uniform convergence and differentiation or integration.
Consider for example fn pxq  sinpnn xq . The sequence pfn q converges to f  0 uniformly. Indeed
2


p
 sin n2 x
 q
0 ¤ 1 @x.
 n  n
Clearly all the functions fn are smooth. The derivatives are given by fn1 pxq  n cospn2 xq. It is easy to see
that the sequence pfn1 q does not converge uniformly (or pointwise). This example shows that while fn Ñ f
we may not have fn1 Ñ f 1 or even fn1 Ñ f 1 .
To explore integrability, we consider gn pxq  2n1
χr
n,ns . Recall that strictly speaking we have not
defined Riemann
integration in R, but rather improper integration, via a limiting procedure. It is clear
however that gn 
1 for every n.
The sequence gn converges uniformly to g  0 as we have |gn
0| ¤
1{p2nq, and so lim gn  1  0  g. We reiterate that strictly speaking
gn are
not Riemann integrable
and we will prove that in fact, on a bounded interval fn Ñ f does imply fn Ñ f .
Theorem 2.16. Lef pfn q, fn : ra, bs Ñ R be a sequence in Riemann

integrable

functions that converges
uniformly to f : ra, bs Ñ R. Then f is Riemann integrable and fn Ñ f .
Proof. First we need to show that f is Riemann integrable, that is show that for every ε ¡ 0 there exists
a partition P of ra, bs such that
U pf, P q
Lpf, P q ε.
Now, since fn Ñ f we know that for any ε ¡ 0 there exists N such that }fn
f }8 ε{p4pb
aqq for
n ¡ N . For a fixed n ¡ N since fn is integrable we know that given ε ¡ 0 there exists a partition P such
that
U pfn , P q
Lpfn , P q
ε
.
2
Now, for that P
¸ ¸
U pf, P q
Lpf, P q  rsup f
inf f s|Ik |  rsuppf
fn fn q
inf pf
fn fn qs|Ik |
Ik I k Ik Ik
 
¸
¤ }f
fn}8 sup fn }f
fn}8
inf
I
fn |Ik |
Ik k
¸ ¸
 2 }f
fn}8|Ik | rsup fn
inf
I
fn s|Ik |
Ik k

¤ 2}f
fn}8pb
aq U pfn, P q
Lpfn, P q

¤ 2 4pb ε
aq pb
aq 2ε  ε.



To see that fn Ñ f , notice that
      
 b b   b  b b

 fn
f  ¤ 
 fn
f  ¤ |fn
f | ¤ }f
fn}8  }fn
f }8pb
aq.
a a a a a

Clearly the right hand side goes to zero as n goes to infinity by the uniform convergence of pfn q to f .

ANALYSIS III 11
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Now that we have understood the implications of uniform converngence in moving limits inside an
integral (on finite intervals!), we revisit differentation. We start by setting up some useful notation.
We will use C k pa, bq to denote functions that are k times continuously differentiable on pa, bq, and
C 8 pa, bq for functions that are infinitely differentiable on pa, bq.
We have seen examples of sequences pfn q that are differentiable, with pfn q converging uniformly to f
but for which fn1 does not converge to f 1 . In fact it is easy to construct examples of C 1 functions that
converge uniformly for which f 1 does not exist. Consider
1{2
fn pxq  x2 1{n .

They are clearly C 1 as the x2 1{n never vanishes for fixed n. pfn q converges uniformly to f pxq  |x|,
which is not smooth at the origin. To see this notice that if
1{2
A : x2 1{n
| x|
then
A¤ p|x| ?
1{ nq2
1{2

|x| ¤ ?1n ,
and the uniform convergence follows.
The following result will prove rather useful.

Theorem 2.17. Let pfn q be a sequence of C 1 functions on ra, bs (understood as a one-sided derivative).
Assume fn Ñ f in the pointwise sense and that fn1 converges uniformly to g. Then f is C 1 and g  f 1 or
fn1 Ñ f 1 .

Proof. Since fn1 Ñ g, Theorem 2.16 yields

 x  x  x
g py qdy  lim f 1 py qdy  lim fn1 py qdy,
a a nÑ8 n nÑ8 a

which by the FTC yields

 x
g py qdy  nlim
a Ñ8rfn pxq
fn paqs  f pxq
f paq.
Notice that since g is continuous this means that f is continuous. While the Theorem does not assume
that g is continuous, that 1

x is a consequence of the uniform convergence of fn to g, since fn are C . Now,
1

the FTC implies that a g is differentiable with derivative g. Since

 x
g  f px q
f pa q
a

we obtain that f is differentiable and g  f 1.

2.3 Series of functions

In this section we consider series of functions, i.e., we study
8̧
fk pxq,

k 1

with fk : Ω Ñ R. We begin by establishing the notion of pointwise convergence and uniform convergence
for a series.

ANALYSIS III 12
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Definition 2.18. Let pfk q be a sequence of functions fk : Ω Ñ R. Let pSn q be the sequence of partial
sums, with Sn : Ω Ñ R defined by
ņ
Sn pxq  fk pxq.

k 1
Then the series
8̧
fk pxq

k 1
converges pointwise to S : Ω Ñ R in Ω if Sn Ñ S pointwise in Ω and it converges uniformly to S in Ω if
Sn Ñ S uniformly on Ω.
Theorem
°n
2.19. Let pfk q, with fk : ra, b°
s Ñ R, be a sequence of integrable functions. Assume that
Sn  k1 fk converges uniformly. Then 8 k1 fk is Riemann integrable and
 8̧ 8̧ 
fk  fk .
k 1 k 1 
Proof. Sn is a finite sum of integrable functions and therefore integrable (by additivity). Since Sn converges
uniformly Theorem 2.16 implies that S is integrable and moreover
 

n
lim
Ñ8 Sn  lim Sn .
nÑ8

°n
°8
Since Sn  
k 1  k1 fk we obtain the result.
fk and limnÑ8 Sn
°
, with fk : ra, bs Ñ R, be a sequence of C 1 functions such that 8
Theorem 2.20. Let pfk q° k1 fk converges
8 f 1 converges uniformly. Then
pointwise. Assume that k1 k

8̧ 1 8̧
fk pxq  fk1 pxq,

k 1 k 1 
that is, the series is differentiable and can be differentiated term-by-term.
Proof. The proof is a simple consequence of Theorem [Link] results says (changing the °n
notation) that
1 1 1 1 1 1
if Sn is C , Sn Ñ S, Sn Ñ g then S P C and S  g (or°Sn Ñ S ). If we define Sn  k1 fk then, it
is C 1 , since each fk is C 1 ; it converges pointwise to S  8 1
k1 fk and finally S converges uniformly, to g
1 1
say. Then S is C and Sn Ñ S . This means
1


8̧ 1
lim S 1  S 1  fk pxq
nÑ8 n
k 1 
but since Sn1  p°nk1 fk q1  °nk1 fk1 we obtain the result, namely
8̧ 
8̧ 1
f 1 px q 
k fk pxq .

k 1 
k 1

Theorem 2.21 (The Weierstrass M-test). Let pfk q be a sequence of functions fk°: Ω Ñ R, and assume
that for every k there exists Mk ¡ 0 such that |fk pxq| ¤ Mk for every x P Ω and 8
k 1 M k 8. Then
8̧
fk
k 1 
converges uniformly on Ω.

ANALYSIS III 13
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

°n
Proof. Notice that it suffices to show that Sn : 
k 1 fk pxq is uniformly Cauchy. Now since °8k1 Mk
8, given ε ¡ 0 there exists N such that
ņ
Mk ε for all m, n ¡ N.

k m 1
Now
   
 ņ m̧   ņ  ņ ņ
|Snpxq
Smpxq|  


fk pxq
pq
fk x 




pq¤
fk x 

|fk | ¤ Mk ¤ ε,
k 1  
k 1 
k m 1 
k m 1 
k m 1
for every x. Therefore Sn is uniformly Cauchy and the proof is complete.

2.4 A continuous, nowhere differentiable function

In 1872 Weierstrass showed that there exist continuous functions that are nowhere differentiable. Standard
examples are constructed using Fourier Series. For example
8̧
f px q  ak cosp2πbk xq
k 0 
for any 0 a 1 b with ab ¡ 1.
We will construct an example based on the sawtooth function. Consider
#
x
txu x ¤ txu 1
ϕpxq  2
1
x txu x ¡ txu 1
2.
The function ϕ is equal to the distance function from x to Z.
We define, for n  0, 1, . . .
fn pxq  n ϕp4n xq.
1
4
°
We will show that f pxq  8 n0 fn is continuous but nowhere differentiable. Notice that

0 ¤ fn ¤ 41n ϕ ¤ 21 41n ,
and that by the Weierstrass M-test we have the uniform convergence of the series. Since each fn is
continuous, and the convergence is uniform we have that f is C 0 .
Given x P R we will choose the sign of hn   4n1 1 in such a way that the points 4n x and 4n px hn q
both belong to the same interval of length 1/2, r k2 , k 2 1 s for some k P Z. We make this choice of sign for
hn because on each of these intervals r k2 , k 2 1 s, the function ϕ has constant slope 1 or
1.
Consider the incremental quotient
fn px hn q
fn pxq ϕp4n x 4n hn q
ϕp4n xq
hn
 4n hn
 1.
Moreover, if m n the graph offm also has slope 1 on the interval in which x and x hn belong to.
Therefore
fm px hn q
fm pxq ϕp4m x 4m hn q
ϕp4m xq
ϵm :   1.
hn 4m hn
However for m ¥ n 1 we have (since 4m x 4m hn
4m x  4m hn  4m
n
1 P Z)

fm px hn q
fn pxq  m pϕp4m x 4m hn q
ϕp4m xqq  0.

1
4
Therefore
f px hn q
f px q fm px hn q
fm pxq
ņ ņ
An :   ϵm .
hn m 0  hn 
m 0
Therefore An is an even integer if n is odd and an odd integer if n is even. Hence there is no limit as n
goes to infinity. Since hn goes to zero that proves that f is not differentiable.

ANALYSIS III 14
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

2.5 Space filling curves

In this section we will show the existence of surjective, continuous maps γ : r0, 1s Ñ r0, 1s  r0, 1s by
looking at Hilbert’s curve. This is just one example of what is known as space-filling curves.
The curve is constructed as the limit of an iterative construction, similarly to the Devil’s staircase.
We start with the unit square, and divide it up into 4 equal squares. On the left of Figure 2.2 the curve
in blue is the building block of the first iterate. It connects the centres of the 4 squares using 3 straight
segments. Any choice would yield a similar curve, but once we have chosen one we will use the same
order (in this case starting in the bottom-left square and moving clockwise) in the rest of the iterates. We
extend the curve with the two green segments to join it to the boundary. The total length of the curve
is 2. We construct γ1 : r0, 1s Ñ r0, 1s  r0, 1s running along the described curve at uniform speed. That
means r0, 1{4s is mapped to the section of the curve in the bottom left square, r1{4, 1{2s to the top left,
r1{2, 3{4s to the top tight and r3{4, 1s to the bottom right.

Figure 2.2: First iterate in the construction of the Hilbert curve

In order to construct the second iterate, we want to bisect every square in our previous grid. In this
case we will have 16 squares. In each block of 4 we copy a scaled down version of the building block in
the previous iteration (see left-hand side of Figure 2.2), but changing the orientation as indicated in the
left-hand side of Figure 2.3. To complete the curve we join the pieces using the 3 segments in red and
connect the curve to the boundary using the two segments in green (see RHS of Figure (2.3)).
Note that the length of this curve is double what it was in the previous iteration. We construct
γ2 : r0, 1s Ñ r0, 1s  r0, 1s by running the curve at uniform speed. Notice that the curve moves along the
16 squares, and in each the length of the curve is 1/4.

Figure 2.3: First iterate in the construction of the Hilbert curve

To construct the third iterate we proceed as before. Now we use the curve from the previous iteration
(without the green segments), and place 4 scaled down copies in the new grid, performing the same

ANALYSIS III 15
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

rotations used in the previous iterations. See the left hand side of Figure 2.4. As before we join the 4
curves with the segments in red and link the curve to the boundary with the segments in green.

Figure 2.4: First iterate in the construction of the Hilbert curve

In this iterate there are 64 squares, and the length of the curve has doubled again to 8.
In this fashion we can construct γn : r0, 1s Ñ r0, 1s  r0, 1s which is continuous and runs through the
grid with 22n squares, in particular runs through the centres of all those squares. If we divide r0, 1s in 4k
intervals, in each of them we run through one square.
If we show that γ is a continuous curve we obtain the result. Since the image of a compact set by
a continuous map is compact the range of γ must be the unit square (notice that γn runs through the
centres of all the squares, and therefore we can approximate any point in r0, 1s  r0, 1s with points in the
image of γn ).
In order to prove that γ is continuous notice that if we consider an interval I € r0, 1s of length |I | 41k
then γn pI q is contained in at most two squares of sides 1{2k provided n ¥ k.
Therefore the Euclidean distance between γn psq and γn ptq is control by
?
|γnpsq
γnptq| ¤ 5
2k
,

the furthest possible distance between two adjacent squares of side 21k .
Given t, s we find k such that

4 k
1
1
|t
s| ¤ k .
4
1

Now for n ¥ k ? ? ?
|γnptq
γnpsq| ¤ 2k  ? k 1 ¤ 2 5|t
s|1{2.
5 1 5
4
Therefore, taking limits as n goes to infinity we find

|γ ptq
γ psq| ¤ c|t
s|1{2,

?
with c  6. This proves that γ is continuous.
We remark that the curve cannot be injective. Indeed, if it was a simple curve (i.e. non-self-intersecting),
the Jordan curve Theorem would imply that the curve divides the plane into an interior and an exterior
region, so that any curve joining a point from the interior with one from the exterior would cross the curve.
To make this a bit more precise we include a formal definition and the main theorem.
Definition 2.22. A Jordan curve C in R2 is the image of an injective, continuous map of a circle, φ :
S1 Ñ R2 .
Theorem 2.23. Let C be a Jordan curve in R2 . Then the complement R2 zC consists of exactly two
connected components. One of the components is bounded (the interior) and the other is unbounded (the
exterior) and the curve is the boundary of each component.

ANALYSIS III 16
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

Notice that the curve cannot be differentiable or have a notion of tangent that makes it possible to
define the left-hand side and the right-hand side, which would correpond the two connected components
of the theorem.

2.6 Absolute Continuity

Definition 2.24. Let I be in an interval in R. A function f : I Ñ R is increasing (strictly increasing)
if f pxq ¤ f py q (f pxq f py q) whenever x, y P I and x y. Similarly f : I Ñ R is decreasing (strictly
decreasing) if f pxq ¥ f py q (f pxq ¡ f py q) whenever x, y P I and x y.

Theorem 2.25. Let f : ra, bs Ñ R be an increasing function. Then f is differentiable almost everywhere.
Moreover  b
f 1 pxqdx ¤ f pbq
f paq.
a

Notice that an analogous result (reversing the inequality) is true for decreasing functions.

Definition 2.26. Given f : ra, bs Ñ R the total variation of f over ra, bs is

ņ
V f : supt |f pxiq
f pxi
1q|u,

i 1

where the supremum is taken over all possible partitions a  x0 x1  xn
1 xn  b of ra, bs. A
function f is of bounded variation if V f is finite.

Theorem 2.27. A function f : ra, bs Ñ R is of bounded variation if and only if f is the difference of two
monotone functions on ra, bs.

Proof. We define
ņ
Tf pxq : supt |f pxj q
f pxj
1q|u,

j 1

where the supremum is taken over all n P N and all partitions a  x0 x1    xn  x of the interval
ra, xs. Notice that if we add a point to any such partition the sum in the definition is made bigger, and
therefore we have
Tf pxq ¤ Tf py q x y,
i.e. Tf is increasing. We claim that Tf f and Tf
f are increasing. This will prove the result as
f  12 rTf fs

1
2
rTf
f s.
To prove the claim, let x y and ε ¡ 0. Choose a partition a  x0 x1 ... xn  x such that
ņ
|f pxj q
f pxj
1q| ¥ Tf pxq
ε.
j 1 
Then
ņ
|f pxj q
f pxj
1q| |f pyq
f pxq|

j 1

is an approximation for Tf py q, and it is less than or equal to Tf py q. Now, since x y, with the partition
above for ra, xs
ņ
Tf py q f py q ¥ |f pxj q
f pxj
1q| |f pyq
f pxq| f py q

j 1

ANALYSIS III 17
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

ņ
¥ |f pxj q
f pxj
1q| |lf pyq
f pxq|jh f pyq
f pxnq f pxq ¥ Tf pxq
ε f p x q.

j 1
¥0
Since ε ¡ 0 is arbitrary we obtain the result for Tf f . Notice that the results is the same for Tf
f.
Namely
ņ
Tf py q
f py q ¥ |f pxj q
f pxj
1q| |f pyq
f pxq|
f pyq
j 1
ņ
¥ |f pxj q
f pxj
1q| |lf pyq
f pxq|jh
f pyq f pxq
f pxq ¥ Tf pxq
ε
f pxq.
n

j 1
¥0

As a consequence, if f is of bounded variation on ra, bs then f 1 pxq exists for almost every x P ra, bs.

Definition 2.28. A function f : ra, bs Ñ R is absolutely continuous if for every ε ¡ 0 there exists δ ¡ 0
such that
ņ
|f pbiq
f paiq| ε

i 1

for every n and every disjoint collection of intervals pa1 , b1 q, . . . , pan , bn q with
ņ
bi
ai δ.

i 1

Notice that since we can take n  1 in the definition above, functions that are absolutely continuous
are continuous, and that since δ cannot depend on x they are also uniformly continuous. However, if we
consider
f pxq  x sin
1
x
on r
1, 1s, the function if continuous (and therefore uniformly continuous) but it is not absolutely contin-
uous.
1.0

0.5

-1.0 -0.5 0.5 1.0

-0.5

-1.0

Figure 2.5: Graph of x sinp1{xq.

This can be seen by showing that it is not of bounded variation, by carefully choosing partitions where
sinp1{xq equals +1 and -1 at the endpoints.

Theorem 2.29. Let f : ra, bs Ñ R be continuous and nondecreasing. The following three statements are
equivalent:

1. f is absolutely continuous on ra, bs.

ANALYSIS III 18
 CHAPTER 2. SEQUENCES AND SERIES OF FUNCTIONS

2. f maps sets of measure 0 to sets of measure 0.

3. f is differentiable almost everywhere on ra, bs, f 1 P L1 (i.e. f 1 is integrable) and

 x
f 1 ptqdt  f pxq
f paq.
a

ANALYSIS III 19
Chapter 3

Basic results about Rn

3.1 Notation in Rn

The main vector spaces that we shall consider in this module are Rn , n P N. Thus, by a vector x P Rn
we mean the n-tuple px1 , . . . , xn q, xi P R, 1 ¤ i ¤ n.
For ease of writing, a vector x P Rn will be written as a row vector x  px1 , . . . , xn q, xi P R, 1 ¤ i ¤ n.
However, in calculations vectors will be written as column vectors

x1
x
 .. 
 . .
xn

This is so that, if A : Rn Ñ Rk is a linear map represented by the matrix


a11 . . . a1n
A
 .. .. 
 . .
ak1 . . . akn

with respect to the standard bases of Rn and Rk , then the vector y : Ax is obtained by multiplying the
column vector x by the matrix A on the left. In index notation, if y  py1 , . . . , yk q, then
ņ
yi  aij xj , i P t1, . . . , k u.

j 1

A vector-valued function f (with values in Rk ) of the variables x1 , . . . , xn is denoted by f : U Ñ Rk where

U € Rn is domain of the function f , i.e., the subset in which the independent variables x  px1 , . . . , xn q
lie.1 Thus
f pxq is shorthand for pf1 px1 , . . . , xn q, . . . , fk px1 , . . . , xn qq
and, 
f1 px1 , . . . , xn q
for calculations, f pxq  ..
 
 . .
fk px1 , . . . , xn q
We normally use a, b and c at the start of the latin alphabet to denote real numbers (i.e. scalars) and
we shall use letters like x, y, p, q, u, v and w in the second half of the latin alphabet to denote vectors.
Thus we shall write ax for the vector pax1 , . . . , axn q without always spelling out that a P R and x P Rn .
In two and three dimensions, we will often be written f px, y q and f px, y, z q. Finally, 0 will denote both
the zero vector (though we shall occasionally write it as p0, . . . , 0q) and the zero scalar!
1
A real valued function f : U Ñ R will be referred to as a scalar function.
20
 CHAPTER 3. BASIC RESULTS ABOUT Rn

3.2 The Euclidean norm and inner product

The Euclidean norm, or length, or magnitude of x  px1 , . . . , xn q P Rn is denoted by }x} and is defined by

ņ {
1 2
}x} : x2i . (3.1)

i 1

The direction of a nonzero vector x is defined to be the unit vector }x} . The obvious relation
x

x  }x} x  0,
x
}x} ,
is the mathematical statement of the informal definition of a (nonzero) vector as a quantity that has both
magnitude and direction2 .
The Euclidean distance between x and y in Rn is defined as }x
y }.
The Euclidean inner product x  y, also called the dot product and scalar product, of x, y P Rn is
defined as
ņ
x  y : x i yi .

?x  x.
i 1

Other notations for x  y include px, y q and xx, y y. Evidently, }x} 

The Cauchy-Schwarz inequality states that

|x  y| ¤ }x}}y}.
It follows from  2
0 ¤ }y }2 x
px  y qy   }y}4}x}2
px  yq2}y}2.
Definition 3.1 (Angle between two nonzero vectors). The Cauchy-Schwarz inequality implies that, if x
and y are both nonzero, then there exists unique θ P r0, π s such that

xy  }x} }y} cos θ;

θ is then defined to be the angle between x and y.

Proposition 3.2 (The triangle inequality). @ x, y P Rn,

}x y} ¤ }x} }y}. (3.2)

Proof.

}x y }2  px y q  px y q  }x}2 2x  y }y}2 ¤ }x}2 2}x}}y } }y}2  p}x} }y}q2.

Replacing x by x
z and y by z
y we get:

}x
y} ¤ }x
z} }z
y}, (3.3)

which corresponds to the familiar fact that the distance from x to y is less than or equal to the sum of the
distances from x to z and z to y. Regarding x, y and z as the vertices of a triangle, (3.3) says that the
length of the edge joining x and y is less than or equal to the sum of the lengths of the edges joining x to
z and z to y; this is the usual triangle inequality.
2
Note that the zero vector does not have a direction.

ANALYSIS III 21
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Exercise 3.1. Prove that, for all x, y P Rn,

 
} }
}y}  ¤ }x
y}.

 x (3.4)

Proposition 3.3.

(i) }ax}  |a| }x} @ a P R, x P Rn.

(ii) }x} ¥ 0 @ x P Rn and }x}  0 ô x  0.

Warning 3.4. It is true that y  x if, and only if, }x
y }  0. However, }x}  }y } does not imply that
y  x. (Think of points on the unit circle.)

Remark 3.5 (}  } satisfies the definition of a norm). A norm is a non-negative valued function }  } :
X Ñ R on a real vector space X which satisfies

(i) }x} ¥ 0 @ x P X and }x}  0 ô x  0.

(ii) For every a P R and x P X we have }ax}  |a| }x}.

(iii) The triangle inequality (3.2): for every x, y P X we have }x y } ¤ }x} }y }.

Norms are discussed more fully in MA260 Norms, Metric Spaces and Topologies.

3.3 Convergence in Rn
Definition 3.6. A sequence pxj q of vectors in Rn converges to x P Rn if

@ ε ¡ 0, D N P N such that, j ¥ N ñ ∥xj
x∥ ε.

Proposition 3.7 (Uniqueness of limits). Let pxj q be a sequence in Rn . If it converges to both x and x̃,
then x  x̃.

Proof. If we assume, by contradiction that x  x̃, then ε : 21 }x
x̃} ¡ 0. Since xj converges to x,

D N1 P N such that
j ¥ N1 ñ }xj
x} ε. (3.5)
Similarly, since xj also converges to x̃, D N2 P N such that
j ¥ N2 ñ }xj
x̃} ε. (3.6)

Then, for j ¥ maxtN1, N2u we have:

2ε  }x
x̃} ¤ }x
xj } }xj
x̃} 2ε.

This is of course a contradiction and therefore x  x̃.3

The notation when we consider each of the coordinates of one of the elements xj in the sequence can
get a bit awkward. We will donte the i-th coordinate of xj P Rn by xj,i .

Proposition 3.8 (Componentwise Convergence). A sequence pxj q of vectors in Rn converges to x0 P Rn if,

and only if, for each i P t1, . . . , nu, lim xj,i  x0,i , where xj  pxj,1 , . . . , xj,n q and x0  px0,1 ,    , x0,n q.
j Ñ8
3
Note how the triangle inequality is crucial for proving the uniqueness of the limit of a sequence.

ANALYSIS III 22
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Proof that convergence implies componentwise convergence. Note that

 1{2
ņ
@ i P t1, . . . nu, |x0,i
xj,i| ¤ }x0
xj }  px0,i
xj,iq2 .

i 1

Now by definition of convergence, given ε ¡ 0, D N P N such that

j ¥ N ñ }x 0
x j } ε
and therefore
¥ N ñ |x0,i
xj,i|
j ε @ i P t1, . . . nu,
i.e., limj Ñ8 xj,i  x0,i for every i P t1, . . . nu.

Proof that componentwise convergence implies convergence. Given ε ¡ 0 and i P t1, . . . , nu, D Ni P N
such that j ¥ Ni ñ |x0,i
xj,i | ε. Set N : maxtN1 , . . . , Nn u. Then
 {
j ¥ N ñ }x0
xj } :
ņ
px0,i
xj,iq 2
1 2
?n ε,

i 1

i.e., limj Ñ8 xj  x.
Remark 3.9. Proposition 3.8 allows us to reduce questions of convergence of a vector-valued sequence to
the corresponding (more familiar) questions of convergence of a sequence of real numbers.

The norm }  } that we defined in (3.1) corresponds to the standard notion of distance we are used to.
However we could have defined other alternative norms.

Definition 3.10 (Max-norm }  }8). The max-norm, which is denoted by }  }8, is defined by
}x}8 : maxt|x1|, . . . , |xn|u, x  px1, . . . , xnq. (3.7)

The following definition provides yet another norm on Rn .

Definition 3.11 (The ‘Manhattan or taxi cab norm’ }  }1).

}x}1 : |x1|    |xn|, x  px1, . . . , xnq. (3.8)

In fact there is a full family of norms }x}p for 1 ¤ p 8 given by


ņ {
1 p
}x}p : |xi|
p
.
i 1
The Euclidean norm (3.1) corresponds to p  2, and }  }8 corresponds to taking the limit as p Ñ 8.
These norms will be considered on MA260 Norms, Metrics and Topologies where the usefulness of the new
notions of distance associated to these norms will also be revealed.

Exercise 3.2 (Comparison of the Euclidean norm with }  }8 and }  }1). Prove that
}x}8 ¤ }x} ¤ ?n }x}8 (3.9)

}x} ¤ }x} ¤ ?n }x}.

and that
1 (3.10)

Furthermore, verify that }}8 and }}1 satisfy the triangle inequality and the relations stated in Remark
3.5 for a norm; indeed, }  }1 and }  }8 are actually norms.

ANALYSIS III 23
 CHAPTER 3. BASIC RESULTS ABOUT Rn

This exercise shows that in Definition 3.6 we could have equivalently used }  }1 or }  }8 instead of }  }
to define the limit of a sequence.
Since Proposition 3.8 reduces convergence to componentwise convergence the we have the following
result.

Proposition 3.12 (Sequence sum rule). If pxj q converges to x, pyj q converges to y, with x, y, xi , yi P Rn
and a, b P R then
lim paxj byj q  ax by.
j Ñ8
Exercise 3.3 (Sequence product rules). Let paj q be a sequence of real numbers that converges to a and
let pxj q and pyj q be sequences of vectors in Rn that converge to x and y respectively. Prove that

(i) the sequence of vectors paj xj q converges to ax and

(ii) the sequence of real numbers xj  yj converges to x  y.

Definition 3.13 (Boundedness of a sequence). A sequence pxj q is bounded if there D M ¡ 0 such that
}xj || ¤ M for every j P N.
Proposition 3.14 (Boundedness of a convergent sequence). If pxj q converges to x, then pxj q is bounded.

It is possible to prove this proposition by using componentwise convergence and the boundedness of
real sequences. A more direct proof is based on the following lemma.

Lemma 3.15. If pxj q converges to x then the sequence of real numbers }xj } converges to }x}.

Proof. Given ε ¡ 0, D N P N such that j ¥ N ñ }xj
x} ε. It follows from the reverse triangle
inequality that
for j ¥ N, | }xj }
}x} | ¤ }xj
x} ε.

Proof of boundedness of a convergent sequence. By the lemma, the convergence of pxj q implies the con-
vergence of }xj }. The boundedness of a convergent sequence of real numbers then implies that }xj } is
bounded and therefore, by definition, pxj q is bounded.

Remark 3.16. Note that the converse of Lemma (3.15) does not hold, not even when n  1. (Use a
mathematical software package to plot the sequence pcos n, sin nq in the plane for a demonstration of how
badly the converse of Lemma (3.15) can fail.)

Proposition 3.17 (Completeness of Rn ). Let pxj q be a Cauchy sequence in Rn , that is, @ ε ¡ 0, D N P N

such that j, k ¥ N ñ }xj
xk } ε. Then pxj q converges to some x P Rn .

Sketch proof. Show that each component xj,i , 1 ¤ i ¤ n, is a Cauchy sequence of real numbers. Then
use the completeness of R and componentwise convergence of xj .

3.4 Subsequences and the Bolzano-Weierstrass theorem

The Bolzano-Weierstrass theorem is one of the most important theorems about sequences of real numbers.
It states that every bounded sequence of real numbers has a convergent subsequence. It generalises
immediately to sequences in Rn .

Theorem 3.18 (Bolzano-Weierstrass for a bounded sequence of vectors). A bounded sequence pxj q in Rn
has a convergent subsequence pxjℓ q.

ANALYSIS III 24
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Sketch of the proof. The proof of the Bolzano-Weierstrass in Chapter 3 in MA141 is done in R. The
argument below is a complete proof in two dimensions, which can be esily generalised to any dimension
(primarily by choosing better notation).
Let xj  pxj,1 , . . . , xj,n q be a bounded sequence in Rn . Then xj,1 is a bounded sequence in R and
therefore, by the Bolzano-Weierstrass Theorem it has a convergent subsequence xjk ,1 which converges to
x1 P R. Since we are only interested in finding a subsequence, we can consider the following sequence,
indexed by k, pxjk ,1 , . . . , xjk ,n q. So far we have constructed a subsequence of the original for which the
first coordinate is a convergent sequence.
Consider now the sequence xjk ,2 . The sequence is of course bounded and therefore, by Bolzano-
Weierstrass, it has a subsequence xjkl ,2 which converges to x2 P R. Notice that since xjk ,1 is convergent,
so is xjkl ,1 . Therefore, if we consider the sequence indexed by l, pxjkl ,1 , . . . , xjkl ,n q, we now have convergent
sequences in the first two components.
It is hopefully clear that, aside from running out letters (and having to resort to cleverer notation), we
can repeat this procedure n times, iteratively constructing subsequences to ensure that every component
is convergent.

3.5 Continuity
3.5.1 Definitions of continuity and continuous limit
We define continuity following the results in year 1 (see Definition 1.2). The only changes are in the
dimension of the domain and the target of the function. We consider U € Rn , p P U and a function
f : U Ñ Rk .

Definition 3.19 (ε-δ Definition of Continuity). Given f : U € Rn Ñ Rk we say that f is continuous at p

if,
@ ε ¡ 0, D δ ¡ 0 such that, for x, p P U , }x
p} δ ñ }f pxq
f ppq} ε.
Notice that the two norms }} in the definition above corresponds to norms in different spaces, namely
Rn and Rk , but we dot make a distinction in the notation.

Definition 3.20 (Sequential Definition of Continuity). f : U € Rn Ñ Rk is continuous at p if, for every

sequence pxj q in U which converge to p, pf pxj qq converges to f ppq.

Exercise 3.4. Check that these two definitions are equivalent

Hint: The argument is the same as that given in First Year Analysis. That the ε-δ definition implies
the sequential definition is straightforward. The converse proceeds by proving the contrapositive, i.e., one
assumes the failure of the ε-δ definition and then one constructs a sequence xj in U which converges to p
but for which f pxj q does not converge to f ppq.
We say that f is continuous, without specifying a particular point, if it is continuous at all points of its
domain. If we wish to emphasize the domain U on which f is continuous, then we say that f is continuous
on U .

Notation 3.21. The space of functions continuous on U with values in Rk is denoted by C pU, Rk q or
C 0 pU, Rk q.4 When k  1, we simply write C pU q or C 0 pU q.

Definition 3.22 (Continuous limit). f : U Ñ Rk has a (continuous) limit at p P U if there exists q P Rk

such that
@ ε ¡ 0, D δ ¡ 0 such that, x P U and 0 }x
p} δ ñ }f pxq
q} ε.
We then write limxÑp f pxq  q.
4
The superscript will later denote the number of derivatives.

ANALYSIS III 25
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Just as for limits of sequences, continuous limits are unique. It is also clear that f is continuous at p if,
and only if, limxÑp f pxq  f ppq. Notice that the definition of continuous limit tacitly assumes that there
exist points in U , different from x, which are arbitrarily close to x.

3.5.2 Separate continuity

In this subsubsection, we shall restrict ourselves to the case n  2 and to functions defined on all of
R2 . The generalisation to higher dimensions is straightforward but the presentation is much easier in two
dimensions using x and y as variables.
Given a real valued function f px, y q, we consider two families of functions tg y : R Ñ RuyPR and
th : R Ñ RuxPR defined by
x

g y pxq : f px, y q : hx py q. (3.11)

Thus g s is the restriction of f to the horizontal line y  s and ht is the restriction of f to the vertical line
x  t.

Definition 3.23 (Separate continuity). A function f : R2 Ñ R is separately continuous at px0 , y0 q if g y0

is continuous at x0 as a function of x and hx0 is continuous at y0 as a function of y.

Two natural questions arise:

(i) Does continuity imply separate continuity?

(ii) Does separate continuity imply continuity?

Exercise 3.5. Prove that continuity implies separate continuity.

As for question (ii), the example below shows that separate continuity does not imply continuity.

Example 3.24. Define f : R2 Ñ R by

f px, y q : 1 , if xy  0, f px, y q : 0 , if xy  0.
g 0 pxq  0 for every x P R and h0 py q  0 for every y P R. In particular, both g 0 and h0 are continuous
at 0 and therefore, f is separately continuous at p0, 0q.
However, f is not continuous at p0, 0q because limpx,yqÑp0,0q f px, y q does not exist. We can estab-
lish this by finding two sequences paj , bj q and pαj , βj q both of which converge to p0, 0q but for which
limj Ñ8 f paj , bj q  limj Ñ8 f pαj , βj q; we also require paj , bj q  p0, 0q and pαj , βj q  p0, 0q @ j P N. So
take, for example, aj  αj  βj  1{j and bj  0. Then f paj , bj q  0 and f pαj , βj q  1 @ j and therefore
limj Ñ8 f paj , bj q  0  1  limj Ñ8 f pαj , βj q. By uniqueness of limits, if limpx,yqÑp0,0q f px, y q were to
exist, limj Ñ8 f paj , bj q and limj Ñ8 f pαj , βj q would have to have the same value. Since they do not, we
conclude that limpx,yqÑp0,0q f px, y q does not exist.

The following are easy to verify (left as exercises):

• f is continuous at all points px, y q such that xy  0,

• f is not separately continuous at points px, 0q such that x  0 (because hx is then not continuous
at 0) and similarly,

• f is not separately continuous at points p0, y q such that y  0.

ANALYSIS III 26
 CHAPTER 3. BASIC RESULTS ABOUT Rn

3.5.3 Basic properties of continuous functions

Throughout this subsubsection, U € Rn , p P U and a, b P R.
The following results are basic properties of continuous functions with effectively the same proof as in
one dimension.
Proposition 3.25 (The sum of continuous functions is continuous). If f, g : U Ñ Rk are both continuous
at p then, af bg is continuous at p.
The proof of this is just an application of the sum rule for limits of sequences of vectors.
Proposition 3.26 (The product of a continuous scalar (real) valued function with a continuous vector-val-
 f : UÑ R and g : U Ñ R are both continuous at p then, f g is continuous
ued function is continuous). If k

at p where pf g qpxq : f pxq g pxq .

The proof of this is just an application of the product rule for a convergent sequence of real numbers
and a convergent sequence of vectors.
Exercise 3.6. Suppose that f : U Ñ Rk is continuous at p. Prove that if f ppq  0 then there exists δ ¡0
such that }f pxq} ¡ 21 }f ppq} @ x P U for which }x
p} δ.
Exercise 3.7. Suppose that f : U Ñ R is continuous at p P U and f pxq  0 @ x P U . Prove that 1{f is
continuous at p.
Corollary 3.27. Suppose that f : U Ñ R and g : U Ñ Rk are both continuous at p and that f pxq 
0 @ x P U . Then g {f is continuous at p.
Proposition 3.28 (The composition of continuous functions is continuous). If U € Rn , V € Rk , f : U Ñ
Rk is continuous at p P U, f pU q € V, g : V Ñ Rm is continuous at f ppq P V , then g  f : U Ñ Rm is
continuous at p.
The proof of this is just an application of the sequential definition of continuity.
Proposition 3.29 (Componentwise continuity). Recall that f : U Ñ Rk can be written as
px1, . . . , xnq  x ÞÑ f pxq  pf1px1, . . . , xnq, . . . , fk px1, . . . , xnqq, x P U.
f is continuous at p if, and only if, @ i P t1, . . . , k u, fi : U Ñ R is continuous at p.
Remark 3.30. The proposition above says that f is continuous if, and only if, all of its component functions
are continuous. The proof is a straightforward application of the sequential definition of continuity and the
equivalence of convergence and componentwise convergence for sequences in Rn .
This proposition suggests that most (but not all5 ) of the features related to the continuity (and, as we
shall see, differentiability) of functions f : Rn Ñ Rk that are different from those of functions f : R Ñ R
arise when n ¡ 1; whether k is greater than 1 is less significant.

3.5.4 Constructing continuous functions of several variables from continuous real valued
functions of a single real variable.
A function like f px, y q  looks continuous on R2 ztp0, 0qu, but how do we prove it without resorting
xy
y2x2
to the ε-δ definition or the sequential definition of continuity? We know g pxq  x and hpxq  x2 are
continuous as functions of the single real variable x. However, what we need to know is that the functions
γ : R2 Ñ R and η : R2 Ñ R defined by
γ px, y q : x, η px, y q : x2
are continuous as functions of two variables. That is precisely the content of the next proposition, which
follows from the following easy lemma.
5
See, for instance, Corollary 4.25.

ANALYSIS III 27
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Lemma 3.31. Write Rn ℓ as Rn ` Rℓ , that is

Rn ℓ
 tpx, yq : x P Rn, y P Rℓu.
Denote by π1 and π2 the two projections of Rn ℓ onto Rn and Rℓ respectively:

π1 px, y q : x, π2 px, y q : y, x P Rn , y P Rℓ .
Then π1 and π2 are continuous.
Proof. Fix px0 , y0 q P Rn ℓ and, given ε ¡ 0, choose δ ε. Then
}px, yq
px0, y0q} δ ñ }π1px, yq
π1px0, y0q}  }x
x0} ¤ }px, yq
px0, y0q} ε,

that is, π1 is continuous. The continuity of π2 is proved similarly.

Proposition 3.32. Consider E € R, a P E and a function g : E Ñ R. For i P t1, . . . , nu, define

πi : Rn Ñ R by
πi px1 , . . . , xi , . . . , xn q : xi
and let Ui : πi
1 pE q : tpx1 , . . . , xn q P Rn : xi P E u. Define f : Ui Ñ R by f px1 , . . . , xn q : g pxi q, that
is, f pxq  g pπi pxqq  g  πi pxq. Suppose that g is continuous at a. Then f is continuous at all points of
πi
1 tau  tpx1 , . . . , xn q P Rn : xi  au.
Proof. By Lemma 3.31, π is continuous on Rn and therefore, by the continuity of composition of continuous
functions, f  g  πi is continuous on πi
1 tau.

We can use Proposition 3.32 and the results in section 3.5.3 to prove the continuity of f px, y q 
xy
x2 y2
on R2 ztp0, 0qu as follows. Consider the four functions, each defined on R2 by

γ px, y q : x, η px, y q : x2 , σ px, y q : y, τ px, y q : y 2 .

Proposition 3.32 tells us that the continuity of these four functions follows from the continuity (proved in
First Year Analysis) of g ptq  t and hptq  t2 as functions of the single real variable t. Now

f px, y q 
pγ px, yqqpσpx, yqq
pηpx, yqq pτ px, yqq
and therefore, the continuity of f on R2 ztp0, 0qu follows from the continuity of the product,
sum and
quotient of continuous functions at points where the denominator does not vanish.
A similar approach can be followed for most functions given by explicit formulas. However, the continuity
of a function at points where the function is given special values (not by a formula) has to be investigated
by separate arguments.
The following two examples are intended to clarify what is meant by ‘natural domain of definition’ of
a function defined by an expression involving familiar continuous functions. The natural domain of

x2 sinpy q
F px, y q 
ex
cosh y

is R2 ztplogpcoshpy qq, y q : y P Ru and F is continuous on this set. Similarly,


logpx y q a
f px, y, z q : , arccospy q 1 pcospxez qq2
sin z

is continuous on tpx, y, z q P R3 : x y ¡ 0,
1 ¤ y ¤ 1, z  nπ, n P Zu.

Proposition 3.32 is, in a sense, ‘obvious’ and you need not quote it explicitly when appealing to the
continuity of functions given by expressions similar to the ones above.

ANALYSIS III 28
 CHAPTER 3. BASIC RESULTS ABOUT Rn

3.5.5 Caution with taking limits in dimension ¥2

If a P R then a can be approached from only two directions, namely left and right. So, limxÑa f pxq exists
if the right-hand limit limxÑa f pxq and the left-hand limit limxÑa
f pxq both exist and are equal.
The situation is much more complicated in higher dimensions. It suffices to illustrate the issue by
considering some of the many different ways of approaching p0, 0q in R2 . We may, for instance, approach
p0, 0q along any line ax by  0. We can also follow more complicated paths from a point in R2ztp0, 0qu to
p0, 0q. For example, we can proceed along px, x2q, i.e. along the parabola y  x2. Indeed, we can approach
p0, 0q along the graph of any continuous function ψpxq for which ψp0q  0. This still does not exhaust all
possibilities because, for instance, we may approach p0, 0q along a spiral like pt cosp1{tq, t sinp1{tq, t ¡ 0.
So, by Proposition 3.28, if f : R2 Ñ R is continuous at p0, 0q then limtÑ0 f pφptq, ψ ptqq would have to
exist for any pair of functions φ, ψ : R Ñ R that are continuous at 0 and equal to 0 there. This should
make it clear that continuity is much more restrictive than separate continuity and indeed, more restrictive
than continuity along lines, which we shall now define.

Definition 3.33 (continuity along lines, also called linear continuity). A function f : Rn Ñ Rk is continuous
along lines (also referred to as linearly continuous) at x0 if the restriction f L of f to the line L passing
through x0 is continuous for every such line L.

The line L through x0 in the direction of v P Rn is parameterised by rv ptq : x0 tv. Therefore f is

continuous along lines at x0 if f  rv is continuous at t  0 for every choice of v P Rn . In particular,

lim f px0 tv q  f px0 q @ v P Rn ,

tÑ0
that is, limtÑ0 f px0 tv q is independent of v. We have seen above that continuity implies continuity along
lines.
In the next example we will exhibit a function which is separately continuous at all points of R2
but which is not continuous along lines through p0, 0q.

Example 3.34. Define f : R2 Ñ R by

f px, y q  2 if px, y q  p0, 0q, f p0, 0q : 0.
xy
,
x y2

For y  0, g y (defined by (3.11)) is a continuous function of x and g 0 pxq  0 @ x P R. Therefore, g y

is continuous for any choice of y P R. By similar reasoning, hx is continuous for any choice of x P R. This
shows that f is separately continuous at all points of R2 . However, note that
,
f pt, tq  21 ,/
.
f pt, 2tq  25 , @ t P Rzt0u.
/
and f pt,
tq 
1-
2

Therefore, limpx,yqÑp0,0q f px, y q depends on the line in R2 along which we approach p0, 0q. In particular, it
is not possible to assign any value to f at p0, 0q that would make it continuous along lines through p0, 0q.

Remark 3.35. One may be tempted to think that a separately continuous function fails to be continuous
only at isolated points. This is not the case. For a fixed pa, bq P R2 define fpa,bq : R2 Ñ R by fpa,bq px, y q :
f px
a, y
bq where f is as in Example 3.34. Let pan , bn q be an enumeration of Q  Q, i.e., of all points
in R2 both of whose coordinates are rational. Then define F : R2 Ñ R by
8̧
F px, y q : 2
n fpan ,bn q px, y q.

n 0

It can be show that F is separately continuous, but discontinuous precisely on Q  Q. Conceptually, the
sum in the definition of F disperses the discontinuity of f to all the rational points in R2 .

ANALYSIS III 29
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Remark 3.36. (This is not examinable.) One may further ask whether a separately continuous function
can fail to be continuous everywhere. This turns out to be not possible, as was shown by René-Louis Baire
in his PhD thesis, published in Annali di Mathematica Pura ed Applicata, vol. 3 (1899) pp. 1-122. In this
foundational paper, Baire introduced what has become known as Baire category of a set, which is useful in
quantifying whether certain properties are generic. For instance, among continuous functions, differentiable
functions are non-generic, that is, atypical. W. H. and G. C. Young furnished an example in 1910 of a
function f px, y q which is separately continuous but which has an uncountable number of discontinuities.
You can read about the history of this topic in The Genesis of Separate versus Joint Continuity by Zbigniew
Piotrowski in Tatra Mountains Math. Publ. 8 (1996), pp. 113-126. A survey that includes more recent
developments can be found in A continuous tale on continuous and separately continuous functions by
Krzysztof Chris Ciesielski and David Miller in Real Analysis Exchange, Vol. 41(1), 2016, pp. 1-36.

The next example exhibits a function f : R2 Ñ R which is continuous along lines through p0, 0q
but which is not continuous at p0, 0q.
Example 3.37. Define f : R2 Ñ R by f px, y q  1 if 0 y x2 and f px, y q  0 otherwise. Show that
limtÑ0 f ptv q  0  f p0, 0q @ v P R2 . However, show also that f is discontinuous at p0, 0q.
In the diagram below, the grey shaded region E is the set on which f  0, i.e.,
E : tpx, y q : f px, y q  0u  tpx, y q : y ¤ 0 or y ¥ x2 u.

Figure 3.1: Diagram of the function

Given v P R2 , D τ ¡ 0 such that, |t| τ ñ tv P E. (If v  pa, bq, take τ  b{a2 if ab  0 and τ  8
if ab  0.) In other words, f ptv q  0 @ t P p
τ, τ q. It follows that limtÑ0 f ptv q  0  f p0, 0q @ v P R2 ,
as claimed.
Finally, limxÑ0 f px, 12 x2 q  1  0  f p0, 0q which shows that f is discontinuous at p0, 0q.
In the next example, we consider the following question: Suppose we demand that f : R2 Ñ R be
continuous along any line in R2 and not just the ones that pass through a chosen point. Would f then
have to be continuous? Remarkably, this is still not the case, as demonstrated by the following example.
Example 3.38. Define f : R2 Ñ R by
x2 y
f px, y q  , if px, y q  p0, 0q, f p0, 0q : 0.
x4 y 2
For each k P R, define gk : R Ñ R to be the restriction of f to the line y  kx through the origin in
R , i.e.,
2

g k pxq : f px, kxq  @ x P R.

kx
x2 k2

ANALYSIS III 30
 CHAPTER 3. BASIC RESULTS ABOUT Rn

Also define g 8 : R Ñ R to be the restriction of f to the y-axis, i.e.,

g 8 pxq : f p0, y q  0 @ y P R.
We see that, for any choice of k P R Y 8, g k is continuous.
We now consider the restriction of f to the parabola y  x2 . This is given by φ : R Ñ R where

x2 x2
φpxq : f px, x2 q   21 , if x  0, φp0q  f p0, 0q  0.
x 4 p x 2 q2

Thus φ is not continuous. It follows from Propositon 3.28 that, since g pxq : px, x2 q is continuous, f
cannot be continuous at p0, 0q because, if it were, then φ  f  g would also have to be continuous, which
it is not. Indeed, we have also shown that limpx,yqÑp0,0q f px, y q does not exist.

ANALYSIS III 31
Chapter 4

Rudiments of topology of Rn and

Continuity

4.1 Closed and open subsets of Rn

Definition 4.1 (Closed set). X € Rn is defined to be closed if, whenever xj is a sequence of points in X
which converges to x P Rn then the limit x also belongs to X.

Definition 4.2 (Open set). U € Rn is defined to be open if, @ x P U, D ε ¡ 0 such that y P Rn and
}y
x} ε ñ y P U .
By convention, the empty set is defined to be both open and closed.

Proposition 4.3. A set is open if, and only if, its complement is closed.

Proof. Suppose that U € Rn is open and that U c is not empty. In order to show that U c is closed, we
consider a sequence xj in U c which converges to x P Rn . We have to show that x lies in U c . If it does
not, then, since U is open, D ε ¡ 0 such that }y
x} ε ñ y P U . But limj Ñ8 xj  x and therefore,
D N P N such that
j ¥ N ñ } xj
x} ε ñ xj P U
which contradicts the assumption that xj R U @ j P N.
For the converse, let X be a closed subset of Rn whose complement X c is nonempty. To prove that
X (which we assume to be nonempty) is open, we have to show that, given y P X c D ε ¡ 0 such that
c

}x
y} ¥ ε @ x P X. (ε is allowed to depend on y but not on x P X). If this were not the case, then, we
could find y P X c and a sequence xj in X such that }xj
y } ¤ 1{j. But then y  limj Ñ8 xj and, since
X is closed, y must belong to X, contrary to the assumption that y R X.

Remark 4.4. Most textbooks first define an open set and then define a closed set to be the complement
of an open set. Of course, these textbooks then have to prove that a closed set satisfies Definition 4.1.

The definition of open set motivates the following definition.

Definition 4.5 (Open (Euclidean) ball). The open ball of radius r ¡ 0 centred at a P Rn is denoted by
Bpa, rq or Br paq and is defined by

Br paq  Bpa, rq : tx P Rn : }x
a} r u.

We abbreviate Br p0q to Br and B1 to just B.

The definition of an open set can now be rephrased as

U € Rn is open if, @ x P U, D ε ¡ 0 such that Bεpxq € U .

32
 CHAPTER 4. RUDIMENTS OF TOPOLOGY OF Rn AND CONTINUITY

This is the definition of an open set that is given in most textbooks. The following proposition justifies the
use of the adjective ‘open’ in the definition of an open ball.

Proposition 4.6. An open ball is open, i.e., it satisfies the definition of an open set.

Proof. For each y P Bpa, rq we need to find ρy ¡ 0 so that the open ball Bpy, ρy q € Bpa, rq.
To this end, set ρy  r
}y
a}. Then, since }y
a} r, we have ρy ¡ 0 and, for x P Bpy, ρy q,

}x
a} ¤ }x
y} }y
a} ρy }y
a}  r,
i.e., the open ball Bpy, r
}y
a}q € Bpa, rq as required.

Definition 4.7 (Closed ball). The closed ball of radius r ¡ 0 centred at a P Rn is denoted by Bpa, rq or
Br paq and is defined by
Br paq  Bpa, rq : tx P Rn : }x
a} ¤ ru.
We abbreviate Br p0q to Br and B1 to just B.

The following proposition justifies the use of the adjective ‘closed’ in the definition of a closed ball.

Proposition 4.8. A closed ball is closed, i.e., it satisfies the definition of a closed set.

Sketch proof. This proposition can be proved in at least two ways. One way is to prove that the complement
of a closed ball is open. Another way is to prove that, if xj is a sequence in Bpa, rq which converges to x,
then }x
a} ¤ r, i.e., x P Bpa, rq.

Proposition 4.9 (An arbitrary union of open sets

”
is open). If Uλ is open for all λ P Λ, where Λ is an
indexing set (which could be uncountable), then λPΛ Uλ is open.
”
Proof. If p P λPΛ Uλ then D λ P Λ such ” that p P Uλ . ”But Uλ is open and therefore, D ε ¡ 0 such
that Bpp, εq € Uλ . In particular, Bpp, εq € λPΛ Uλ , i.e., λPΛ Uλ is open.

Definition 4.10 (ε-neighbourhood). Let E be any subset of Rn . Given ε ¡ 0, the ε-neighbourhood

N pE, εq of E is defined by ¤
N pE, εq : Bpx, εq.
P
x E

By the previous proposition, N pE, εq is open.

Example 4.11. Let E : Q Xr0, 1s € R. Enumerate the rational numbers in E by a sequence x1 , x2 , . . . .

Given ε ¡ 0, let
8
¤
O : px j
2
j ε , x j 2
j εq.

j 1
°
Then O is open. The sum of the lengths of the intervals that make up O is ε 8j 1 2
1
j  2ε. Therefore,

2 , O cannot contain all the irrationals between zero and 1. This example shows how complicated
1
if ε
open sets can be.

Proposition 4.12 (The finite intersection “m

of open sets is open).
If U1 , U2 , . . . , Um are all open, then j 1 Uj is also open.
“m
Proof. If p P 
j 1 Uj then

D ε1 ¡ 0 such that Bpp, ε1q € U1,

... ,
D εm ¡ 0 such that Bpp, εmq € Um.
“ “m
Set ε : mintε1 , . . . , εm u ¡ 0. Then Bpp, εq € mj 1 Uj , i.e., j 1 Uj is open.

ANALYSIS III 33
 CHAPTER 4. RUDIMENTS OF TOPOLOGY OF Rn AND CONTINUITY

Corollary 4.13. An arbitrary intersection of closed sets is closed and the finite union of closed sets is
closed.

Sketch proof. Consider the complements of the relevant closed sets and apply the preceding propositions
together with de Morgan’s laws on complements, unions and intersections.

Remark 4.14. Note that a subset of Rn may be neither open, nor closed. For example r0, 1q in R or

tpx, yq x2 y2 ¤ 1u X tpx, yq |y ¡ 0u.
Note, too, that H and Rn are both open and closed.
Terminology. The collection of open subsets of Rn is called a topology of Rn . Other topologies are
discussed in modules on MA260 Metric Spaces and Topology.

4.2 Continuity and topology

4.2.1 Continuity in terms of open sets
The ε-δ definition of continuity at p for a function f : U Ñ Rk , p P U € Rn can be phrased as

@ ε ¡ 0 D δ ¡ 0 such that f pBpp, δq X U € Bpf ppq, εq. (4.1)

Equivalently, 
@ ε ¡ 0 D δ ¡ 0 such that Bpp, δq X U € f
1 Bpf ppq, εq . (4.2)
Informally and pictorially, f is continuous at p if f pxq can be guaranteed to stay near f ppq (ε-near) by
requiring x to stay sufficiently close (δ-close) to p in U .

Theorem 4.15 (Continuity via open sets and closed sets). The following statements are equivalent.

(i) f : Rn Ñ Rk is continuous at all points of Rn.

(ii) for all open subsets V of Rk , f
1 pV q is open.

(iii) for all closed subsets F of Rk , f
1 pF q is closed.

Proof. Suppose that f is continuous at all points of Rn and let V € Rk be open. Then, for each
p P f
1 pV q, D ε ¡ 0 such that Bpf ppq, εq € V . By continuity of f at p as stated in (4.2), D δ ppq ¡ 0 such
that Bpp, δ ppqq € f
1 Bpf ppq, εq € f
1 pV q, which shows that f
1 pV q is open. (We have assumed that
f
1 pV q is not empty; if it is, it would still be an open set.)
Conversely, given p P Rn and ε ¡ 0, the ball Bpf ppq, εq is open and therefore, it follows
 (by assumption)
that f
1 Bpf ppq, εq is open. In particular, D δ ¡ 0 such that Bpp, δ q € f
1 Bpf ppq, εq , which is precisely
the statement of the ε-δ definition of continuity as expressed by (4.2).
Finally, the equivalence of (ii) and (iii) follows from f
1 pV c q  pf
1pV qqc and the fact that a set is
closed if, and only if, its complement is open.

Remark 4.16. If f : Rn Ñ Rk is continuous, then the image of an open set need not be open. For instance,
consider the constant map f pxq  0 @ x P Rn .
Ñ Rk is continuous, then the2 image of a closed set need not be closed. For instance, consider
If f : Rn
the map f : R Ñ R given by f pxq  2 @ x P R. Then f pRq  r0, 1q which is neither a closed subset
x
x 1
nor an open subset of R.

Example 4.17 (Open and closed sets via Theorem 4.15). Show that

ANALYSIS III 34
 CHAPTER 4. RUDIMENTS OF TOPOLOGY OF Rn AND CONTINUITY

(i) the unit sphere S n
1 : tx P Rn : }x}  1u is a closed subset of Rn .

 
(ii) The set E : tpx, y q P R2 : xy sinp1{xq cosp1{y q ¡
1u is an open subset of R2.
Solution.
(i) Define f P C pRn q by f pxq  }x}. Then S n
1  f
1 pt1uq which, by part (iii) of Theorem 4.15, is a
closed subset of Rn because t1u is a closed subset of R.
(ii) Define f : R2Ñ R by
 
f px, y q : xy sinp1{xq cosp1{y q if xy  0, f px, y q  0 if xy  0.

Then f is continuous and E  f
1 p
1, 8q which, by part (ii) of Theorem 4.15, is an open subset of
R2 because p
1, 8q is an open subset of R.
Remark 4.18. (NON-EXAMINABLE) Close inspection of the proof of Theorem 4.15 will reveal that if
U € Rn is open then f : U Ñ Rk is continuous at all points of Rn if, and only if, for all open subsets V of
Rk , f
1 pV q is open. However, it is no longer true that the preimage of a closed set is necessarily closed,
that is, statement (iii) of Theorem 4.15 no longer applies.
Similarly, if U € Rn is closed then f : U Ñ Rk is continuous at all points of Rn if, and only if, for all
closed subsets F of Rk , f
1 pF q is closed. However, it is no longer true that the preimage of an open set
is necessarily open, that is, statement (ii) of Theorem 4.15 no longer applies.
The extension of Theorem 4.15 to functions f : U Ñ Rk , where U is an arbitrary subset of Rn , requires
the notion of sets that are open/closed relative to U . We will not explore these notions in this module.

4.2.2 Continuity and sequential compactness

Definition 4.19 (Sequentially compact subset). K € Rn is sequentially compact if every sequence xj in
K has a convergent subsequence xjℓ whose limit is in K.
€ Rn is bounded if D M ¡ 0 such that }x} ¤ M @ x P X.
Definition 4.20. X
Theorem 4.21. K € Rn is sequentially compact if, and only if, K is closed and bounded.
Proof. Suppose that K is sequentially compact. To prove that K is closed, we consider a sequence xj in
K which converges to x P Rn and then we have to show that x P K. By the sequential compactness of
K, xj has a subsequence xjℓ whose limit is in K. But x  limj Ñ8 xj  limℓÑ8 xjℓ P K. The proof that
K is closed is complete.
To prove that K is bounded, assume, for a contradiction, that it is unbounded. Then there exists a
sequence xj in K such that }xj } ¥ j @ j P N. By the sequential compactness of K, xj has a subsequence
xjℓ whose limit is in K. In particular, xjℓ is bounded, i.e., D M ¡ 0 such that }xjℓ } ¤ M @ ℓ P N. But by
definition of subsequence, jℓ ¥ ℓ and, by the way the sequence xj was chosen, }xjℓ } ¥ jℓ . Therefore,

M ¥ }xj } ¥ jℓ ¥ ℓ @ ℓ P N.
ℓ

This clearly cannot hold and we conclude that K must be bounded.

We now assume that K is closed and bounded and prove that it is sequentially compact. So consider
an arbitrary sequence xj in K. Since K is bounded, xj must be bounded and, by the Bolzano-Weierstrass
theorem, it has a convergent subsequence xjℓ whose limit x must be in K, since K is closed. The proof
that K is sequentially compact is complete.

Theorem 4.21 is important because it enables us to determine easily whether a set is sequentially
compact. For instance, the theorem asserts that a closed ball Bpa, rq is sequentially compact without
having to check whether all its sequences contain a convergent subsequence! Similarly, we can assert that
the sphere Sn
1 pa, rq : tx P Rn : }x
a}  ru is sequentially compact; it is clearly bounded and we
showed above (for a  0 and r  1, but the proof is virtually identical) that it is closed.

ANALYSIS III 35
 CHAPTER 4. RUDIMENTS OF TOPOLOGY OF Rn AND CONTINUITY

Theorem 4.22 (Continuity preserves sequential compactness). If f : K Ñ Rk is continuous and K is

sequentially compact then f pK q is also sequentially compact.

Proof. Let yj be a sequence in f pK q. Then, for each j P N, D xj P K such that f pxj q  yj . By the
sequential compactness of K, there exists a convergent subsequence xjℓ of xj such that limℓÑ8 xjℓ 
x P K. By continuity of f at x, limℓÑ8 yjℓ  limℓÑ8 f pxjℓ q  f pxq P f pK q, i.e., f pK q is sequentially
compact.

Theorem 4.23 (Extreme Value Theorem). Let K € Rn be sequentially compact and let f : K Ñ R be
continuous. Then D x , x P K such that

f p x  q ¤ f px q ¤ f px  q @ x P K.
This theorem asserts that a continuous real valued function on a sequentially compact space attains
its extreme values, i.e., max and min. This theorem was proved in First Year Analysis in the case that K is
a closed, bounded interval. It is one of the most important theorems of elementary mathematical analysis
because, for instance, it is used in the proof of Rolle’s Theorem which, in turn, is used in the proof of
Taylor’s theorem.

Proof of Extreme Value Theorem. By the previous theorem and Theorem 4.21, f pK q € R must be closed
and bounded. Therefore, M : sup f pK q and m : inf f pK q are both finite because f pK q is bounded. By
definition of sup and inf, there exist sequences aj , bj P f pK q such that limj Ñ8 aj  m and limj Ñ8 bj  M .
But f pK q is closed and therefore, m, M P f pK q, i.e., D x , x P K such that

f px  q  m ¤ f p x q ¤ M  f pxq @ x P K.

Remark 4.24. The notion of supremum cannot be extended from R to Rk , k ¥ 2. That is why we had
to restrict ourselves to scalar functions in the preceding theorem. The best we can do for vector-valued
functions is stated in the following corollary.

Corollary 4.25. Let K € Rn be sequentially compact and let f : K Ñ Rk be continuous. Then D x, x P
K such that
}f pxq} ¤ }f pxq} ¤ }f pxq} @ x P K.
Proof. Since x ÞÑ }x} : Rk Ñ R is continuous (by the triangle inequality | }x}
}y } | ¤ }x
y }), the map
x ÞÑ }f pxq} : K Ñ R is continuous. The result now follows from the theorem on extreme values.

ANALYSIS III 36
Chapter 5

The space of linear maps and matrices

Notation 5.1.

(i) The space of linear maps, i.e., tA : Rn Ñ Rk | A is linearu, shall be denoted by LpRn, Rk q and
LpRn , Rn q will be abbreviated to LpRn q.1

(ii) The space of k  n matrices with real entries shall be denoted by Rk,n .2

To a matrix 
a11 . . . a1n
paij q   ..
 .
..  P Rk,n
.
ak1 . . . akn

we associate A P LpRn , Rk q defined by

  
x1 a11 . . . a1n x1
R Q x   . ÞÑ Ax :  . .  . PR .
 ..   .. .
.   .. 
n k
(5.1)
xn ak1 . . . akn xn

Let tv1 , . . . , vn u and tw1 , . . . , wk u be the standard bases of Rn and Rk respectively, i.e.,

vj  p0, . . . , 0, 1Ò , 0, . . . , 0q P Rn, wi  p0, . . . , 0, 1Ò , 0, . . . , 0q P Rk .

j th position among n entries ith position among k entries

Then, 
a1j ķ
Avj   .. 
 .  aij wi , j P t1, . . . , nu, (5.2)
akj 
i 1

and therefore, paij q is the matrix representation of A with respect to the standard bases on Rn and Rk .
It is sometimes useful to express this association of paij q with A as defined above more formally as a
map µ : LpRn , Rk q Ñ Rk,n , i.e.,

µpAq : paij q, where A and paij q are related by (5.2). (5.3)

It is easy to verify that µ is a linear isomorphism. We will use standard bases on Rn and Rk throughout
(unless otherwise stated), and so we will switch between the linear map A and the associated matrix
µpAq  paij q without explicit mention.
1
Other notations in use are HomR pRn , Rk q and EndpRn q.
2
Other notations in use are Rkn , M pk  n, Rq, Mkn pRq and Mkn pRq.

37
 CHAPTER 5. THE SPACE OF LINEAR MAPS AND MATRICES

It is easy to check that the identification


a11 . . . a1n
 .. ..  ÐÑ pa , . . . , a , a , . . . , a , . . . , a , . . . , a q (5.4)
 . . 11 1n 21 2n k1 kn
ak1 . . . akn
between Rk,n and Rnk is a linear isomorphism. It follows that
dimpLpRn , Rk qq  dimpRk,n q  nk.

5.1 Two norms on the space of linear maps and matrices

We shall be discussing the continuity of maps like
f : Rn Ñ LpRk , Rℓq, f : L pR n q Ñ R and f : LpRn q Ñ LpRn q.
As we have seen, this necessitates defining a notion of distance (norm) on LpRn , Rk q, or equivalently, a
norm on Rk,n . The first such notion that comes to mind is to use the identification (5.4) and define the
so-called Frobenius norm }  }F by

ķ ņ {
1 2

}paij q}F : a2ij . (5.5)

 
i 1j 1

This is fine, but we shall make more use of the operator norm (defined below) as it turns out to be more
convenient.3
The operator norm arises from studying how large }Ax} can get relative to }x} as x ranges over Rn .
We can do this using (5.1) and the Cauchy-Schwarz inequality:
   
ķ ņ 2 ķ ņ ņ ķ ņ
}Ax} 
2
aij xj ¤ a2ij x2j  a2ij }x}2  }paij q}2F }x}2.

i 1 
j 1 
i 1 
j 1 
j 1  
i 1j 1

In particular, for x  0 we have

}Ax}2 ¤ }pa q}2 . (5.6)
zt0u }x}2
sup ij F
P
x Rn

This makes possible the following definition.

Definition 5.2. The operator norm of A P LpRn , Rk q, denoted by ~A~ or }A}op , is defined by

}A}op : }Ax} . (5.7)

zt0u }x}
sup
P
x Rn

In practice, (5.7) is often used in the form

}Ax} ¤ }A}op }x} @ x P Rn. (5.8)
Also in practice, one may somehow establish that A P LpRn , Rk q satisfies
}Ax} ¤ M }x} @ x P Rn for some M ¡ 0.
Then (5.7) implies that }A}op ¤ M .    
} Ax}  1



 1


Ax 
 
The expression
 }x} can be equivalently written as 
}x} Ax and, by the linearity of A, 
}x}
   x 
}x} . Since  }x}   1, we have the following equivalent definition of }A}op:
A x   


}A}op : sup }Ax}. (5.9)

}x}1
 
3
Observe that }µpAq}2F  trace pµpAqqT pµpAqq  trace pµpAqqpµpAqqT , where the superscript T denotes transpose
and the trace of a matrix is the sum of its diagonal entries.

ANALYSIS III 38
 CHAPTER 5. THE SPACE OF LINEAR MAPS AND MATRICES

Observe that the supremum in (5.9) is taken over a sequentially compact set (namely, the unit sphere

in Rn ). We will see in Proposition 5.5 below that this can be an advantage of (5.9) over (5.7).
Sn 1

5.1.1 Comparison of the two norms

Recalling from (5.3) that µpAq  paij q we can rewrite (5.6) as

}A}op ¤ }µpAq}2F . (5.10)

From (5.2) we see that

ņ ķ ņ ņ
}µpAq}  }paij q} :
2
F
2
F a2ij  }Avj } ¤ }A}
2 2
op }vj }2  n}A}2op. (5.11)
 
j 1i 1 
j 1 
j 1

Combining (5.11) and (5.10) gives the comparison

?1n }µpAq}F ¤ }A}op ¤ }µpAq}F (5.12)

5.1.2 Properties of the operator norm

The properties of }  } in the next proposition justify calling it a norm.
Proposition 5.3. In (i), (ii) and (iii) below, A, B P LpRn , Rk q and a P R.

(i) }A}op ¥ 0 and }A}op  0 ô A  0.

(ii) }aA}op  |a| }A}op .

(iii) Triangle inequality. }A B }op ¤ }A}op }B }op .

Proof. The first two items are elementary and the proofs are left to the reader. For the third item,

}pA B qx}  }Ax Bx} ¤ }Ax} }Bx} ¤ p}A}op }B }opq}x}

and therefore

}A B }op  }pA B q x}
¤ }A}op }B }op.
x
sup
P zt0u
Rn }x}

Proposition 5.4 (Composition bound). A P LpRn, Rk q and B P LpRk , Rmq ñ BA P LpRn, Rmq and
}BA}op ¤ }B }op }A}op.
Proof. }pBAqpxq}  }B pAxq} ¤ }B }op}Ax} ¤ }B }op }A}op }x} and therefore, }BA}op ¤ }B }op }A}op.

Proposition 5.5. A P LpRn , Rk q is injective if, and only if, D α ¡ 0 such that }Ax} ¥ α}x} @ x P Rn.
(Note that k does not have to be equal to n.)

Proof. If Ax  0 and }Ax} ¥ α}x} for some α ¡ 0 then x  0, i.e., A is injective.

The converse is proved by establishing the contrapositive, i.e., suppose that there is a sequence xj
in Rn zt0u such that }Axj }{}xj } Ñ 0 as j Ñ 8. Set uj : xj {}xj }. Then }uj }  1 @ j P N and
Auj Ñ 0 as j Ñ 8. Since S n
1 is sequentially compact (recall S n
1  t}x}  1u), there exists a
subsequence ujℓ which converges to u P S n
1 . But the map x ÞÑ }Ax} is continuous and therefore,
}Au}  limjÑ8 }Auj }  0, i.e., u P kerpAq. It follows that A is not injective.
ANALYSIS III 39
 CHAPTER 5. THE SPACE OF LINEAR MAPS AND MATRICES

Remark 5.6. Proposition 5.5 can be regarded as a quantitative measure of injectivity. An injective linear
map has to keep a nonzero vector x away from zero. The larger the value of α in the inequality }Ax} ¥ α}x}
the more the linear map A pushes x away from zero.
A better way of saying this is to consider a perturbation of A by a matrix B to get the matrix A B.
We then have the following

Proposition 5.7. Suppose that A, B P LpRn, Rk q satisfy

}Ax} ¥ α}x} for some α ¡ 0 (5.13)

and }B }op α. Then A B is still injective.

Proof.
}pA B qx} ¥ }Ax}
}Bx} ¥ α}x}
}B }op }x}  δ }x},
where δ : α
}B }op ¡ 0. Therefore pA B qx  0 ñ x  0, which proves that A B is injective.

This proposition can be interpreted as saying that if (5.13) holds then the open ball BpA, αq €
LpRn , Rk q4 is contained in the set of injective linear transformations in LpRn , Rk q.
So, a larger value of α in (5.13) indicates that A is able to withstand perturbations by ‘larger’ (as
measured by the operator norm) linear transformations while maintaining injectivity.

We shall revisit Proposition 5.7 in the context of invertible matrices; see Proposition 5.13.

5.2 Convergence and continuity in LpRn , Rk q

These are defined in exactly the same way as for sequences in Rn and functions f : U Ñ Rk , U € Rn .
For instance, a sequence pAj qj PN of linear transformations in LpRn , Rk q converges to A P LpRn , Rk q if,
@ ε ¡ 0, D N P N such that j ¥ N ñ }Aj
A}op ε.
Similarly, for r ¡ 0, BpA, rq : tB P LpRn , Rk q : }B
A}op ru.
A moment’s thought will reveal that, because of (5.12), we could also use }}F instead of the operator
norm to define these notions. Recall that }  }F on Rk,n is the same as }  } on Rnk via (5.4) and (5.5).
Therefore, the completeness5 of Rnk that was established in Proposition 3.17, immediately implies the
completeness of Rk,n with respect to }  }F . The completeness of LpRn , Rk q with respect to }  }op then
follows from (5.12).
Property (i) in Proposition (5.3) and the triangle inequality are precisely the properties that are needed
to prove the usual properties of limits of sequences and continuous limits, such as uniqueness of limit, sum
rule, boundedness of convergent sequences, etc..

5.2.1 Continuity of functions involving matrices or linear maps

A function f : U Ñ Rk,n is continuous at x P U if @ ε ¡ 0, D δ ¡ 0 such that }y
x} δ ñ
}f pyq
f pxq}F ε. As above, since }  }F on R is the same as }  } on R via (5.4) and (5.5), we can
k,n nk

use Proposition 3.29 to assert that


a11 pxq . . . a1n pxq
x ÞÑ .. ..  : U Ñ Rk,n
f 
 . .
ak1 pxq . . . akn pxq

is continuous at x if, and only if, @ i P t1, . . . , k u, j P t1, . . . , nu, x ÞÑ aij pxq is continuous at x.
A function F : U Ñ LpRn , Rk q is continuous at x P U if @ ε ¡ 0, D δ ¡ 0 such that }y
x} δ ñ
}F pyq
F pxq}op ε.
4
This ball is taken with respect to the operator norm on LpRn , Rk q; see §5.2.
5
A space X is complete if every Cauchy sequence in X converges to an element of X.

ANALYSIS III 40
 CHAPTER 5. THE SPACE OF LINEAR MAPS AND MATRICES

Remark 5.8. Because of (5.12), we see that F : U Ñ LpRn, Rk q is continuous at x P U if, and only if,
µpF q : U € Rn Ñ Rk,n is continuous at x P U .

This remark is very useful because it provides a practical way of checking the continuity of
F : U Ñ LpRn , Rk q. Namely, we simply have to check whether all the matrix entries of the
matrix representation µpF q (with respect to the standard bases on Rn and Rk ) are continuous.

The continuity of f : Rk,n Ñ Rℓ and of F : Rk,n Ñ Rℓ,m is defined similarly by identifying pRk,n , }}F q
with pRnk , }  }q, as in the next proposition and the example below it.
Proposition 5.9 (Continuity of the determinant function). The map ∆ : Rn,n Ñ R defined by ∆paij q :
detpaij q is continuous with respect to the norm }  }F on Rn,n .
Proof. The determinant is simply a polynomial of degree n in its n2 variables

a11 , . . . , a1n , a21 , . . . , a2n , . . . , an1 , . . . , ann . (5.14)

 
Therefore, its continuity follows from the identifications (5.4) and (5.5) of Rn,n , }  }F with Rn , }  }
2

2
and the usual continuity of polynomials6 on Rn .

Example 5.10. Define F : R2,2 Ñ R2,2 by F pAq  A2. If

 
a2
A 
a b bc ab bd
, then A2
c d ac cd bc d2

and therefore, F can be viewed as φ : R4 Ñ R4 defined by

φpa, b, c, dq : pa2 bc, ab bd, ac cd, bc d2 q,

which is clearly continuous and therefore F is continuous.

5.3 The space GLpn, Rq € LpRn , Rk q of invertible linear transformations

It is clear that if a linear map A : Rn Ñ Rk is a bijection, then n  k and ker A  t0u. A remarkable
consequence of the rank-nullity theorem is that the converse is also true, i.e., a linear map A : Rn Ñ Rk is
a bijection (whose inverse is then linear) if k  n and ker A  t0u. This statement sounds more dramatic
when stated as: the system Ax  y of n linear equations in n variables is guaranteed a solution for all
y P Rn if, and only if, should the solution exist, it must be unique, i.e., uniqueness of a solution (that is
not yet known to exist!) in this setting guarantees its existence!
Definition 5.11. The general linear group over the real numbers is denoted by GLpn, Rq and is defined
by
GLpn, Rq : tA P LpRn q : A is invertibleu,
with the group operation being composition of linear maps.
In terms of matrices,
GLpn, Rq : tpaij q P Rn,n : detpaij q  0u.
This is the space of nonsingular matrices with matrix multiplication as the group operation. It is easy to
check that GLpn, Rq satisfies the group axioms and that it is infinite. Note that GLp1, Rq is just the set
of nonzero real numbers with multiplication.
6
See Proposition 3.32 and use the algebraic properties of continuous functions. Each term of ∆paij q is, in fact, linear
in each of the n2 variables (5.14)and this is what makes ∆ an example of a multilinear map. The special case of the
determinant of a 2  2 matrix ac db may help to clarify matters. This determinant is then just the function on R4 defined by
∆pa, b, c, dq : ad
bc, which is easily seen to be continuous by Proposition 3.32 and the algebraic properties of continuous
functions.

ANALYSIS III 41
 CHAPTER 5. THE SPACE OF LINEAR MAPS AND MATRICES

Proposition 5.12. The space GLpn, Rq of nonsingular matrices is an open subset of Rn,n .

Proof. This follows immediately from

• recognising that GLpn, Rq  ∆
1 pRzt0uq,

• the continuity of ∆ from Proposition 5.9 and

• the openness of Rzt0u and the theorem on continuity via open sets.

The openness of GLpn, Rq means that invertibility of a linear transformation in LpRn q is a stable
property. In other words, an invertible linear transformation can be perturbed a little and it remains
invertible. The next proposition addresses the question of by how much, in terms of }A
1 }, an invertible
linear transformation A can be perturbed and maintain its invertibility.

Proposition 5.13 (Size of an open ball in GLpn, Rq). Given A P GLpn, Rq, set α : 1{}A
1 }op . If
B P LpRn q and }B
A}op α then B is invertible, i.e., the open ball tB P LpRn q : }B
A}op αu €
GLpn, Rq. Furthermore,

}B
A}op α ñ }B
1}op ¤ α
}B1
A} . (5.15)

op

Proof. x  A
1 pAxq ñ }x} ¤ }A
1}op }Ax}, i.e., }Ax} ¥ α}x} @ x P Rn. Therefore, if x  0 and
}B
A}op α,
}Bx}  }Bx
Ax Ax} ¥ }Ax}
}pB
Aqx}
¥ pα
}B
A}opq}x} ¡ 0, (5.16)

i.e., Bx  0. Therefore, kerpB q  t0u and B P GLpn, Rq. (Compare with the proof of Proposition 5.7.)
Finally, on replacing x by B
1 x in (5.16), we see that

}x}  }B pB
1xq} ¥ pα
}B
A}opq}B
1x},

i.e.,
}B
1x} ¤ α
}B1
A} }x} @ x P Rn,
op

which establishes (5.15).

Proposition 5.14. The map A ÞÑ A
1 : GLpn, Rq Ñ GLpn, Rq is continuous.

Proof. We need to show that, if B is close to A in GLpn, Rq then B
1 is close to A
1 . So, we consider
A
1
B
1  A
1 BB
1
A
1 AB
1  A
1 pB
AqB
1 . (This is the analogue for matrices of the

a , a, b P Rzt0u.) Therefore,
equality a1
1b  bab

}A
1
B
1}op ¤ }A
1}op }B
A}op }B
1}op.

As above, set α : 1{}A
1 }op and, given ε ¡ 0, set δ : mint 21 α, εu. Then, from (5.15), we deduce that

}B
A}op δ ñ }B
1}op ¤ α2 ñ }A
1
B
1}op ¤ α2ε2 ,

i.e., we have verified the ε-δ definition of continuity of A ÞÑ A
1 for A P GLpn, Rq.

ANALYSIS III 42
Chapter 6

The Derivative

From now on, the domain U € Rn of a function f : U Ñ Rk shall be an open subset of Rn , unless
otherwise stated. In particular, this means that when p P U and a limit like limxÑp is considered, x is
allowed to approach p from any direction.

6.1 Directional derivative

The rate of change of a function of two or more variables depends on the direction in which that change
is measured. For example, f px, y q  x increases as we move to the right along a line parallel to the x-axis
but it does not change at all when we move vertically along a line parallel to the y-axis. So we introduce
the notion of directional derivative in order to take this dependence on direction into account.
Given v P Rn , the line Lx,v passing through x P Rn in the direction of v is parameterised by rptq 
x tv, t P R. Since U is open, D τ ¡ 0 such that x tv P U @ t P p
τ, τ q. The restriction gx,v of f to
this segment of Lx,v is defined by

gx,v ptq : f px tv q P Rk , |t| τ.

Since gx,v ptq  pg1 ptq, . . . , gk ptqq is a function of a single real real variable, we can differentiate it component
by component in the usual way.1

Definition 6.1. The directional derivative Bv f pxq is defined by



Bv f p xq : d
gx,v ptq
dt t0


 d
f px tv q (6.1)
dt 
t 0
f px tv q
f pxq
 tlim
Ñ0 t
. (6.2)

Example 6.2. Calculate Bv f px, yq for the function f px, yq : x2
y2 in the direction of v  pa, bq.

Solution. f px, yq tpa, bq  px taq2
py tbq2  x2 2tax t2a2
y2
2tby
t2b2. Therefore
f px tv q  2ax 2ta2
2by
2tb2 ,
d
dt 

Bv f px, yq  d
f px tv q  2ax
2by.
dt t0
1
In the definition 3.33 of linear continuity, gx,v was denoted by f L .

43
 CHAPTER 6. THE DERIVATIVE

6.1.1 Directional derivative and continuity

A function of one variable that is differentiable must, in particular, be continuous. So it is reasonable to
ask the following question.
Suppose that Bv f pxq exists for all v P Rn, does it follow that f is continuous at x?
Somewhat surprisingly, the answer is no!

Example 6.3. As in Example 3.37, let f px, y q  1 if 0 y x2 and f px, y q  0 otherwise. Show that
Bv f p0, 0q exists for all v P R2 even though f is not continuous at p0, 0q!
Solution. As in Example 3.37, given v P R2 , D τ ¡ 0 such that f ptvq  0 @ t P p
τ, τ q. It follows that
Bv f p0, 0q  lim f ptvq
tf p0,0q  0 @ v P R2.
Ñ0
t
So, we need a definition of derivative that is more restrictive than partial and directional derivatives
just as continuity is more restrictive than separate and linear continuity. The key idea, which we are about
to describe, is to regard the derivative as providing a ‘best’ affine linear approximation of a function. In this
process we move away from the ‘kinematic’ notion of derivative as rate of change and adopt, instead, a
‘mapping’ viewpoint of functions in which a nonlinear map is approximated by an affine linear one. Hence
the need to fully understand linear maps as in a module on Linear Algebra.

6.2 The (Fréchet) Derivative as an affine linear approximation

6.2.1 Affine linear approximation in the 1-variable case
Given x P pa, bq € R, the derivative at x, f 1 pxq, of a function f : pa, bq Ñ R is defined by

f px hq
f pxq
f 1 pxq  lim . (6.3)
hÑ0 h
This definition cannot be readily extended to functions f : Rn Ñ Rk because it is not possible to divide
vectors in Rk by vectors in Rn , even when n  k. We can get around this difficulty by rewriting (6.3) as

}f px hq
f pxq
f 1 pxq h}
Ñ0
lim
h }h}  0. (6.4)

Then, rewriting f px hq
f pxq
f 1 pxq h as f px hq
pf pxq f 1 pxq hq, we can interpret (6.4) as saying
that, for ‘small’ h, the (nonlinear) mapping h ÞÑ f px hq, x fixed, is optimally approximated by the
affine linear map h ÞÑ f pxq f 1 pxq h. Observe that this is a mapping viewpoint of the derivative, which
is conceptually different from the hitherto held kinematic viewpoint of rate of change. It is this mapping
viewpoint which we are able to generalise to the notion of derivative of functions of several variables.

6.2.2 The (Fréchet) Derivative

By analogy with (6.4), we make the following definition.

Definition 6.4. f : U Ñ Rk is differentiable at x P U if D A P LpRn, Rk q such that

}f px hq
f pxq
Ah}  0, (6.5)
lim
hÑ0 }h}
As above, rewriting f px hq
f pxq
Ah as f px hq
pf pxq Ahq, we can interpret this definition
as saying that, for ‘small’ h, the (nonlinear) map h ÞÑ f px hq is optimally approximated by the affine
linear map h ÞÑ f pxq Ah.

Exercise 6.1. Show that the linear map A in (6.5), if it exists, is unique. This justifies saying that the
Fréchet derivative provides the optimal linear approximation of f .

ANALYSIS III 44
 CHAPTER 6. THE DERIVATIVE

Remark 6.5. A real number a can also be viewed as the linear map A : R Ñ R defined by Ah  ah.
Indeed a is the 1  1 matrix representation of A with respect to the standard basis 1 of R. Similarly, the
real number f 1 pxq in (6.3) is the 1  1 matrix representation of the linear map h ÞÑ f 1 pxq h in (6.4).

Notation 6.6. If a linear map A that satisfies (6.5) exists, it is called the (Fréchet) derivative of f at x
and it is denoted by Df pxq.

Thus (6.5) can be rewritten as

}f px hq
p f px q Df pxq hq}
 0. (6.6)
h
lim
Ñ0 }h}
Exercise 6.1 justifies calling the affine linear map h ÞÑ f pxq Df pxqh the best affine linear approximation
of the map h ÞÑ f px hq. The ε-δ formulation of (6.6) provides us with a way of quantifying how good
this approximation is. Namely, by definition of continuous limit we have that

@ ε ¡ 0, D δ ¡ 0 so that 0 }h} δ ñ }f px hq
pf pxq

}h}
Df pxq hq}
ε.

Multiplying both sides by }h} and then allowing h  0 we deduce that


@ ε ¡ 0, D δ ¡ 0 so that }h} δ ñ } f px hq
f px q Df pxq h } ¤ ε}h}. (6.7)

We shall refer to (6.7) and (6.6) also as the definition of the derivative Df pxq of f at x. Note that equality
has to be allowed in (6.7) because we have allowed the possibility that h  0, which can be convenient in
many situations.

Proposition 6.7 (Differentiability implies continuity). If f : U Ñ Rk is differentiable at x P U then f is

continuous at x.

Proof. By (6.7) we have that @ ε ¡ 0 D δ ¡ 0 such that


}h} δ ñ }f px hq
f pxq Df pxq h } ¤ ε}h}
ñ }f px hq
f pxq} ¤ }Df pxq h} ε}h}
ñ }f px hq
f pxq} p}Df pxq}op εq}h}.
Set δ : mintδ, ε{p}Df pxq}op εqu. Then |h| δ ñ }h} δ and therefore we can use the above
chain of implications to conclude that

}h} δ ñ }f p x hq
f pxq} p}Df pxq}op εq δ  ε.

We have proved the claim that f is continuous at x.

6.2.3 Differentiability of components of vector-valued functions

Exercise 6.2. Given f : U Ñ Rk , f pxq  pf1 pxq, . . . , fk pxqq, prove that f is differentiable at x P U if, and
only if, for each i P t1, . . . , k u, fi : U Ñ R is differentiable at x.

Remark 6.8. Compare Exercise 6.2 with Proposition 3.29 on componentwise continuity.

6.2.4 Relation between the derivative and directional derivative

Proposition 6.9. If Df pxq exists then Bv f pxq exists for all v P Rn and Bv f pxq  Df pxqv. In particular,
if f is differentiable at x, then Bv f pxq is linear in v, i.e.,

Bav bw f px q  a B v f px q b Bw f pxq @ a, b P R and @ v, w P Rn. (6.8)

ANALYSIS III 45
 CHAPTER 6. THE DERIVATIVE

Proof. If v  0 there is nothing to prove. So, we assume that v  0. Then, replacing h in (6.6) by tv and
removing }  } where that is allowed, we get
f px tv q
f pxq
Df pxqptv q
lim
tÑ0 t}v }
 0. (6.9)

Multiply both sides of (6.9) by }v } and use Df pxqptv q  tDf pxqv by the linearity of Df pxq so as to get
f px tv q
f pxq
lim
tÑ0 t
 Df pxqv, i.e.,Bv f pxq  Df pxqv.

Finally, by linearity of Df pxq,

Bav bw f pxq  Df pxqpav bwq  aDf pxqv bDf pxqw  a Bv f p xq b B w f px q.

Example 6.3 shows that the converse of Proposition 6.9 does not hold.

6.3 Partial derivatives, gradient and Jacobian matrix

Let tv1 , . . . , vn u be the standard basis of Rn , i.e.,
vi  p0, . . . , 0, 1Ò , 0, . . . , 0q P Rn.
ith position among n entries

Definition 6.10. For 1 ¤ i ¤ n, Bv f pxq is called the ith-partial derivative of f : U Ñ Rk at x P U . It is

more simply denoted by Bi f pxq.
i

Since
f px tvi q
f pxq
Bv f pxq  tlim
i
Ñ0 t
f px1 , . . . , xi
1 , xi q
f px1, . . . , xi
1, xi, xi q
 lim t Ñ0
t, xi 1 , . . . xn
t
1 , . . . xn

Bif pxq is calculated by differentiating f px1, . . . , xnq with respect to the ith variable, treating all the other
variables as constant2 . It is therefore also common to write BBxf pxq or BBx f px1 , . . . , xn q instead of Bi f pxq.
I prefer the notation Bi f because it makes clear that the differentiation is meant with respect to the ith
i i

variable, irrespective of whether that variable is xi or yi or ui or . . . .

Bearing in mind that f pxq  pf1 pxq, . . . , fk pxqq we have
Bif pxq  pBif1pxq, . . . , Bifk pxqq, Bif pxq is a vector in Rk .
and so,
If f is a function of a few variables, say two, it is common to write f px, y q instead of f px1 , x2 q, and
to write fx instead of BBfx or B1 f . Similarly, fy is shorthand for BBfy . Similar shorthand applies to functions
of 3 or 4 variables. For 5 variables or more, it is usually more convenient to number the variables, rather
than choose distinct letters!

6.3.1 Algebraic rules for partial derivatives.

Since partial differentiation involves differentiating with respect to a single variable, the usual algebraic
rules of differentiation apply. For instance,
if f, g : U Ñ Rk , then B i pf g q  Bi f Big.
Similarly,
if f : U Ñ R and g : U Ñ Rk then Bipf gq  pBif qg f Bi g.
2
See Example 6.16 further down.

ANALYSIS III 46
 CHAPTER 6. THE DERIVATIVE

6.3.2 Gradient and Jacobian matrix

Definition 6.11. The Jacobian matrix at x, B f pxq, of f : U Ñ Rk , f pxq  pf1pxq, . . . , fk pxqq but written
as a column vector, is defined by
 
B1f1pxq ... Bnf1pxq 
B1 f p xq ... B n f px q Bf1pxq ...
B f px q  

..
.
..
.

 .. ..  

..
.


B1fk pxq ... Bnfk pxq . .

Bfk pxq ...

where B1 f pxq, . . . , Bn f pxq are the vector-valued partial derivatives of the vector-valued function f 3 and
Bf1, . . . , Bfk are the Jacobian 1  n matrices (row vectors) of the scalar-valued functions f1, . . . , fk .
Definition 6.12. The gradient at x, ∇f pxq, of a scalar valued function f : U Ñ R is defined to be the
column vector 
B 1 f px q
∇f pxq :  .
 .. 
.
B n f px q

Thus ∇f pxq is the vector in Rn which is the transpose of the row vector B f pxq, ∇f pxq  B f p xq T .
Remark 6.13. For a scalar valued function f : U Ñ R, Bf pxq represents a linear functional on Rn defined
by
  
Rn Q v ÞÑ B f pxq pv q : B1 f pxq v1 
Bnf pxq vn P R, v  pv1, . . . , vnq.
Using the Euclidean inner product, this linear functional B f pxq is identified with the vector ∇f pxq:
 
B f p x q pv q  ∇f pxq  v.
However, be warned that the distinction between ∇f and B f is often suppressed, even in these notes!
In particular, for typographical reasons, I may occasionally write ∇f as a row vector!

Proposition 6.14. If f : U Ñ Rk is differentiable at x P U and h P Rn then

Df pxqh  B f pxqh. (6.10)

Remark 6.15. It is important to appreciate the difference between the two sides of (6.10). On the left hand
side we have the linear map Df pxq acting on the vector h whereas on the right hand side we have the matrix
Bf pxq multiplying the vector h. In other words, Bf pxq is the matrix representation of Df pxq with respect
to the standard bases on Rn and Rk . More formally, B f pxq  µpDf pxqq where µ : LpRn , Rk q Ñ Rk,n is
defined by (5.3).

Proof of Proposition 6.14. h  h1 v1  hn vn and therefore, by linearity of Df pxq,

ņ ņ
Df pxqh  hi Df pxqvi  hi Bi f pxq  B f pxqh,

i 1 
i 1

where, in the second equality, we have used Proposition 6.9.

Example 6.16. Calculate the Jacobian matrix B f of f : R3 Ñ R2 defined by


f px, y, z q :
3 sin x
ex 2y
, a .
1 y2z4
3
The values of f are vectors in Rk .

ANALYSIS III 47
 CHAPTER 6. THE DERIVATIVE

Solution.  3 3
3x2 ex 2y 2ex 2y 0
Bf px, y, zq   ? cos x
p1 yz 4 sin x

p1 2y 2 z 3 sin x .
1 y2 z4 y 2 z 4 3{2
q y 2 z 4 3{2
q
All the entries of the Jacobian matrix B f are continuous functions and we shall see in §6.5 that this implies
the existence of the derivative Df px, y, z q P LpR3 , R2 q, which is the linear map defined by
 3 3
 
3x2 ex 2y 2ex 2y 0 r r
Df px, y, z qpr, s, tq   s  Bf px, y, zq  s
?1cosyx z
2 4

p1 yz 4 sin x
y 2 z 4 3{2
q
p1 2y 2 z 3 sin x
y 2 z 4 3{2
q t t
.

6.3.3 Why so many different notations for the same thing?!

We have seen in Propositions 6.9 and 6.14 that,
when f is differentiable at x, Df pxqh  Bh f pxq and Df pxqh  B f pxqh.
So why bother with three different ways of writing the same thing?
The reason is that even when Df pxq does not exist, Bh f pxq and B f pxqh may both still exist but
it may happen that they are not equal! (See Example 6.18 below.) In other words, the existence
of the Jacobian matrix B f pxq does not guarantee that the linear map it defines is the derivative
Df pxq, unless Df pxq is known to exist.
The reader would be right to wonder at this stage how to go about calculating the derivative
Df if this cannot be simply done by computing the Jacobian matrix B f . Fortunately, as has
already been pointed out in Example 6.16, it suffices to verify further that all the entries of B f
are continuous for then, by Theorem 6.25, Df exists and its matrix representation with respect
to the standard bases of Rn and Rk is given by B f .
However, there are situations where it may be necessary to to calculate Df directly from the definition
(6.6). This would be the case, for example, at points outside the natural domain of definition of f where
it is assigned a special value.
Remark 6.17. The notions of partial derivative, directional derivative and (Fréchet) derivative are the
differentiable analogues of separate continuity, continuity along lines and continuity.

Example 6.18. Define f : R2 Ñ R by

x3
f px, y q  if px, y q  p0, 0q, f p0, 0q  0.
x2 y2
(i) Show that Bv f p0, 0q exists for all v P R2. In particular, calculate Bf p0, 0q.
(ii) Show that Bv f p0, 0q  B f p0, 0q v and that Bv f p0, 0q is not linear in v.
(iii) Calculate fx px, y q and fy px, y q for px, y q  p0, 0q and show that fx and fy are not continuous at
p0, 0q.
(iv) Explain why Df p0, 0q does not exist.
Solution. We shall let v  pa, bq P R2 throughout.
3
(i) As always, Bp0,0qf p0, 0q  0. If pa, bq  p0, 0q we have f pta, tbq  a2ta b2 @ t P R and therefore,
 
 d ta3  3
Bv f p0, 0q  d
f pta, tbq  dt a2 b2 t0
 a2 a .
dt t0 b2
So, Bv f p0, 0q exists for all v P R2 and, in particular,
the Jacobian matrix B f p0, 0q  pBp1,0q f p0, 0q , Bp0,1q f p0, 0qq  p1, 0q.

ANALYSIS III 48
 CHAPTER 6. THE DERIVATIVE

(ii) Bf p0, 0q pa, bq  a  Bpa,bqf p0, 0q, unless b  0. Bpa,bqf p0, 0q is not linear in pa, bq because a a b 2
3
2 is
not a linear function of a and b.

(iii)
x2 px2 3y 2 q
2x3y
fx px, y q  fy px, y q 
p x2 y 2 q 2 , px2 y2q2 .
We have seen that fx p0, 0q  Bp1,0q f p0, 0q  1 but limyÑ0 fx p0, y q  0. Therefore, fx is not
continuous at p0, 0q. Similarly fy p0, 0q  Bp0,1q f p0, 0q  0 but limxÑ0 fy px, xq 
21 . Therefore, fy
is also not continuous at p0, 0q.

(iv) Bv f p0, 0q is not linear in v and therefore, by Proposition 6.9, Df p0, 0q does not exist. The situation is
similar to that in Example 3.38 of a real valued function of two variables which fails to be continuous
even though its restriction to any line in R2 is continuous. We shall see below that the lack of
differentiability of f at p0, 0q can be understood geometrically as the failure of the graph of f (which
lies in R3 ) to have a tangent plane at p0, 0, 0q.

6.4 The Chain Rule

Theorem 6.19. Let U and V be open subsets of Rn and Rk respectively. Suppose that f : U Ñ Rk is
differentiable at x P U and that f pxq P V . Suppose further that g : V Ñ Rm is differentiable at f pxq.
Then g  f : Rn Ñ Rm is differentiable at x and

Dpg  f qpxq  Dg pf pxqq  Df pxq. (6.11)

The following two lemmas will be useful in the proof of the Chain Rule.

Lemma 6.20. Given f : U Ñ Rk , x P U, r ¡ 0 such that Bpx, rq € U and A P LpRn, Rk q, define

∆x,A f : Bp0, rq Ñ Rk by
#
p
f x h q
f pxq
Ah , if h  0,
∆x,A f phq  }h} (6.12)
0, if h  0.
Then f is differentiable at x with Df pxq  A if, and only if, ∆x,A f is continuous at 0.

Proof. If ∆x,A f is continuous at 0 then limhÑ0 }∆x,A f phq}  } limhÑ0 ∆x,A f phq}  }∆x,A f p0q}  0,
Therefore, (6.5) holds and f is differentiable at x with Df pxq  A.
Conversely, if f is differentiable at x and we set A  Df pxq in (6.12) then, (6.6) asserts that
limhÑ0 }∆x,A f phq}  0. But then limhÑ0 ∆x,A f phq  0  ∆x,A f p0q, which is precisely the statement
that ∆x,A f is continuous at 0.

Notation 6.21. If f is differentiable at x, then we let ∆x f phq denote ∆x,Df pxq f phq.

Lemma 6.22. Let τ ¡ 0 and consider a function δ from the open ball Bτ € Rn to Rk defined by
δ phq : ξ phq η phq, 0 }h} τ, δ p0q : 0,

where, ξ : Bτ zt0u Ñ R is bounded and η : Bτ Ñ Rk is continuous at 0 P Bτ and η p0q  0. Then δ is
continuous at 0 P Bτ .

Proof. By continuity of η at 0, given ε ¡ 0,

D σ P p0, τ q such that }h} σ ñ }ηphq} ε.
By boundedness of ξ, D M ¡ 0 such that }ξ phq} M @ h P Bτ zt0u.
Therefore, 0 }h} σ ñ }δ phq} M ε, i.e., limhÑ0 δ phq  0  δ p0q and this completes the proof of
the lemma.

ANALYSIS III 49
 CHAPTER 6. THE DERIVATIVE

Proof of Chain Rule. As in the proof of Proposition 6.7 we have

f px hq  f pxq Df pxqh ∆x f phq}h}

and
g p f px q k q  g pf pxqq Dg pf pxqqk ∆f pxq g pk q}k } (6.13)

where #
p
f x h q
f pxq
Df pxqh , if h  0,
∆x f phq : }h}
0, if h  0.

and #
p p q kq
gpf pxqq
Dgpf pxqqk , if k  0,
g f x
∆f pxq g pk q : }k }
0,  0. if k

Set k phq : Df pxqh ∆x f phq}h} in (6.13). Then, by linearity of Dg pf pxqq,

g pf p x hqq  g pf pxqq Dg pf pxqqpDf pxqhq

}h}Dgpf pxqqp∆xf phqq }kphq} ∆f pxqgpkphqq.
Therefore,

g p f px hqq
g pf pxqq
Dg pf pxqq  Df pxqh  }h} δ1 phq δ 2 phq

where,

δ1 phq : Dg pf pxqqp∆x f phqq,

and δ2 phq :
}kphq} ∆ gpkphqq, h  0, δ2 p0q : 0.
}h} f pxq
The proof of the Chain Rule will be complete once we prove that

lim }δ1 phq}  0 and lim }δ2 phq}  0.

hÑ0 h Ñ0
We start with δ1 phq.
}δ1phq} ¤ }Dgpf pxqq}op }∆xf phq}
and, by differentiability of f at x, we have limhÑ0 }∆x f phq}  0. It follows immediately that limhÑ0 }δ1 phq} 
0.
We move on to δ2 phq. For h  0, set

}kphq} ¤ }Df pxqh} }∆ f phq} ¤ }Df pxq} }∆ f phq}.

ξ phq :
}h} }h} x op x

The continuity of ∆x f at 0 implies that ξ phq is bounded on Bτ zt0u for some τ ¡ 0. Next set

η phq : ∆f pxq g pk phqq.

k phq is a continuous function of h and k p0q  0. Therefore, by differentiability of g at f pxq and Proposition
6.7, η phq is a continuous function of h and η p0q  0. We may therefore apply Lemma 6.22 to δ2 phq 
ξ phq η phq to conclude that limhÑ0 }δ2 phq}  0.
The proof that g  f is differentiable at x and (6.11) holds is complete.

ANALYSIS III 50
 CHAPTER 6. THE DERIVATIVE

6.4.1 Jacobian form of chain rule

The linear isomorphism µ : LpRn , Rk q Ñ Rk,n defined by (5.3) takes composition of linear transformations
to matrix multiplication. Therefore, under the same hypotheses as for Theorem 6.19 we have

B g  f pxq  Bgpf pxqq  Bf pxq where  stands for matrix multiplication. (6.14)

More explicitly,

B 1 g 1  f px q ... Bn g1  f pxq


..
.
..
.


B1 gm  f pxq ... B n g m  f px q
 
B1g1pf pxqq ... Bk g1pf pxqq B1f1pxq . . . Bnf1pxq
 .. ..  .. .. 
. (6.15)
 . .  . .
B1gmpf pxqq . . . Bk gmpf pxqq B1fk pxq . . . Bnfk pxq
B g  f pxq can be written as BBx gj pf pxq . If we set y  f pxq

The entry in the j th row and ith column of
and we see g as a function of y  py1 , . . . , yk q then the entry in the j th row and rth column of B g pf pxqq
i

can be written as
Bgj pf pxqq. Then, (6.15) can be written as
By r

B g pf pxq  ķ Bgj pf pxqq Byr pxq where, by

Byr pxq we mean
Bfr pxq. (6.16)
B xi j r1
Byr B xi B xi B xi
(6.16) is perhaps more memorable than (6.15) because we can imagine cancelling B yr from the denominator
and numerator in the terms of the sum in (6.16). However, it is important to appreciate the difference
between
B g pf pxq and Bgj pf pxqq. In the first expression, the function g  f is being differentiated
B xi j B yi j

with respect to its ith variable (1 ¤ i ¤ n) and evaluated at x whereas in the second expression it is the
function gj that is being differentiated with respect to its ith variable (1 ¤ i ¤ k) and then evaluated at
y  f pxq.

6.4.2 Calculating with the chain rule and gradient

Given f : Rn Ñ R and g : R Ñ R the ith partial derivative of g  f can be computed using (6.16) with
k  1 and m  1: 
Bi g  f pxq  g1pf pxqq Bif pxq
and therefore,
∇ g  f pxq  g 1 pf pxqq∇f pxq.
In the example that follows, the gradient of a function will be written as a row vector (to make it easier to
write)!

? 6.23. Calculate ∇}x}, x P R zt0u, by applying the chain rule to g  f where f pxq : }x} and
Example n 2

g ptq : t, t ¡ 0.

Solution. g 1 ptq  2? 1
and, since f pxq  x21    x2n , we have that B i f px q  2xi , i.e., ∇f pxq 
pB1f pxq, . . . , Bnf pxqq  p2x1, . . . , 2xnq  2x. Therefore,
t

∇}x}  ∇ g  f pxq  g 1 pf pxqq∇f pxq  a 2x 

1 x
(6.17)
2 }x}2 } x} .
ANALYSIS III 51
 CHAPTER 6. THE DERIVATIVE

The component form of (6.17) is

B }x}  xi .
B xi }x}
Another common application of the chain rule occurs in the calculation of the derivative of f  r where
r : R Ñ Rn is a parameterisation of a path in Rn and f : Rn Ñ R is a scalar function. In this case, if
rptq  px1 ptq, . . . , xn ptqq then

pf  rq1ptq  Bf prptqq dxi

ņ

i1
Bxi dt

 ∇f prptqq  r1ptq. (6.18)

Example 6.24. Fix x P Rn and define r : R Ñ Rn by rptq : tx, i.e., if x  0 then r is a parameterisation
of the line through 0 and x. Given f : Rn Ñ R calculate pf  rq1 ptq in terms of ∇f .

Solution. r1 ptq  x and therefore, by (6.18), pf  rq1 ptq  x  ∇f ptxq . Equivalently,
d
f prptqq  x1
Bf ptxq    x Bf ptxq.
dt Bx1 n
B xn
6.5 Continuity of partial derivatives implies differentiability
As noted in Example 6.16, partial derivatives are easily computed using the familiar rules of differentiation.
Unfortunately, as Example 6.18 shows, the existence of all partial derivatives of a function at all points
does not guarantee its differentiability. However, it is worth noting that in Example 6.18 Df was shown to
not exist at a point where the partial derivatives are discontinuous. The partial derivatives of f in Example
6.3 do not even exist at points of the form px, x2 q, x  0 and fy does not exist at points on the real axis
different from p0, 0q. These examples raise the possibility that continuity of the partial derivatives may
guarantee differentiability. That is the content of the next theorem.
Theorem 6.25. Consider f : U Ñ Rk and suppose there exists Bpx, rq € U such that the Jacobian matrix
Bf pyq exists at all points of Bpx, rq and that Bf is continuous at x. Then f is differentiable at x and
Df pxq h  B f pxq h @ h P Rn .
Remark 6.26. Recall from §5.2.1 that, writing f as pf1 , . . . , fk q, B f is continuous at x if, and only if,
B1f1, . . . , Bnf1, B1f2, . . . , Bnf2, . . . , B1fk , . . . , Bnfk are all continuous at x.
Note, too, that we need to make an assumption on the behaviour of the partial derivatives of f at all
points y sufficiently near x in order to conclude the existence of Df at just x.
Proof. We shall only give the proof in the simplest case n  2, k  1. The proof of the general case is
given in Section 6.6.
For 0 }ph1 , h2 q} r define,
∆f ph1 , h2 q : f px1 h1 , x2 h2 q
f px1 , x2 q
h1 B1 f px1 , x2 q
h2 B2 f px1 , x2 q. (6.19)
We need to show that
∆f ph1 , h2 q
ph1 ,h2 qÑp0,0q }ph1 , h2 q}
lim  0. (6.20)

If we succeed, then we would have proved that Df px1 , x2 q is the linear map from R2 to R defined by
Df px1 , x2 qph1 , h2 q : h1 B1 f px1 , x2 q h2 B2 f px1 , x2 q.
Partial derivatives only provide information along lines parallel to the axes. Therefore, we have to break
f px1 h1 , x2 h2 q
f px1 , x2 q into differences along the axes as follows:
 
f px 1 h1 , x2 h2 q
f px1 , x2 q  f px1 h1 , x2 h2 q
f px1 h1 , x2 q f px 1 h1 , x2 q
f px1 , x2 q
 II I.

ANALYSIS III 52
 CHAPTER 6. THE DERIVATIVE

The second term I can be written in terms of B1 f by applying the mean value theorem to f p, x2 q, x2
fixed. Namely,

D θ1 P p0, 1q such that f px1 h1 , x2 q
f px1 , x2 q  h1 B1 f px1 θ1 h1 , x2 q. (6.21)

Similarly, for the first term II,

D θ2 P p0, 1q such that f px1 h1 , x2 h2 q
f p x 1 h1 , x2 q  h2 B2 f px1 h1 , x2 θ 2 h2 q. (6.22)

Substituting (6.21) and (6.22) in the definition (6.19) of ∆f ph1 , h2 q we get


∆f ph1 , h2 q  h1
B1f px1 θ1h1, x2q
B1f px1, x2q  (6.23)
h2 B2 f px1 h1 , x2 θ2 h2 q
B2 f px1 , x2 q .

By continuity of B1 f and B2 f at px1 , x2 q, given ε ¡ 0, D δ ¡ 0 (which we may, and shall, assume to

be less than r) such that

}ph̃1, h̃2q} δ ñ }B1f px1 h̃1, x2 h̃2q
B1f px1, x2q} ε

(6.24)
and }B2 f px1 h̃1 , x2 h̃2 q
B2 f px1 , x2 q} ε.

Now }ph1 , θ2 h2 q} }ph1, h2q} and }pθ1h1, 0q} }ph1, h2q} and therefore, if }ph1, h2q} δ then, using
(6.24) in (6.23) with ph̃1 , h̃2 q  pθ1 h1 , 0q in the first term and ph̃1 , h̃2 q  ph1 , θ2 h2 q in the second term,
yields ?
}∆f ph1, h2q} εp|h1| |h2|q ¤ ε 2 }ph1, h2q}.
In other words, we have established (6.20) and completed the proof of Theorem 6.25.

6.5.1 The space of continuously differentiable functions

Definition 6.27. Suppose that f : U Ñ Rk is differentiable on U . Then f is said to be continuously
differentiable at p P U if the map x ÞÑ Df pxq : U Ñ LpRn , Rk q is continuous at p. More explicitly,

@ ε ¡ 0, D δ ¡ 0 such that }x
p} δ ñ }Df pxq
Df ppq}op ε.

Proposition 6.28. f : U Ñ Rk is continuously differentiable on U if, and only if, Bf : U Ñ Rk,n is

continuous on U .

Proof. Recall that LpRn , Rk q and Rk,n are identified via the map µ : LpRn , Rk q Ñ Rk,n defined by (5.3)
and that, by Proposition 6.14, if Df pxq exists, then B f pxq  µpDf pxqq. Furthermore, by the discussion in
§5.2.1, the continuity at p of x ÞÑ Df pxq : U Ñ LpRn , Rk q implies the continuity at p of x ÞÑ B f pxq : U Ñ
Rk,n .
Conversely, if B f is continuous on U then Theorem 6.25 assures us that Df exists at all points of U
and that B f  µ  Df . We can then appeal again to the discussion in §5.2.1 to assert that the continuity
at p of B f implies the continuity at p of Df .

This proposition is useful because it provides us with a practical way of checking continuous
differentiability, namely, we simply have to compute all the first order partial derivatives Bi fj
of f  pf1 , . . . , fk q and verify that they are all continuous. This means that most functions
that one can explicitly write down in terms of polynomials, exponential, logarithm, etc. are
continuously differentiable on their ‘natural’ domain of definition.

Notation 6.29.
C 1 pU, Rk q : tf : U Ñ Rk | Bf : U Ñ Rk,n is continuousu.
C 1 pU q : C 1 pU, Rq.

ANALYSIS III 53
 CHAPTER 6. THE DERIVATIVE

6.6 Proof of Theorem 6.25

Discussion of proof strategy. We shall be establishing the existence of Df pxq by verifying (6.7). Now
this requires us to consider the difference f px hq
f pxq and we only have information on the partial
derivatives of f . In other words, we do not have information on how f changes as we go from x to x h
along the straight line joining these two points; we only have information on how f changes when we move
along the ‘axes’. So we have to go from x  px1 , . . . , xn q to x h  px1 h1 , x2 h2 , x3 h3 , . . . , xn
1
hn
1 , xn hn q as follows:

px1, x2, x3, . . . , xn
1, xnq Ñ px1 h1, x2, x3, . . . , xn
1, xnq Ñ
px1 h1, x2 h2, x3, . . . , xn
1, xnq Ñ    Ñ px1 h1, x2 h2, x3 h3, . . . , xn
1 hn
1, xnq
Ñ px1 h1, x2 h2, x3 h3, . . . , xn
1 hn
1, xn hnq.
It shall be therefore convenient to introduce the following notation,

r0 : p0, 0, 0, . . . , 0, 0q,
r1 : ph1 , 0, 0, . . . , 0, 0q,
r2 : ph1 , h2 , 0, . . . , 0, 0q,
.. ..
. . (6.25)
rn
1 : ph1 , h2 , h3 , . . . , hn
1 , 0q,
rn : ph1 , h2 , h3 , . . . , hn
1 , hn q

so that then we can write

x  x0 Ñx r1 Ñ  Ñ x rn
1 Ñx rn x h. (6.26)

When proving the differentiability of a function f : U Ñ Rk directly from the definition, i.e., using
(6.7), one has to propose a candidate for Df pxq. In light of (6.10), Df pxq would have to be the map in
LpRn , Rk q defined by
  
h1 B1f1pxq . . . Bnf1pxq h1
h   . ÞÑ  ...  Bf pxq  h.
 ..   ..   .. 
.  .
hn B1fk pxq ... Bnfk pxq hn

We now embark on the proof proper.

Proof. By Exercise 6.2, it suffices to prove the theorem when k  1, i.e., for f : U Ñ R and then
Bf pxq  pB1f pxq, . . . , Bnf pxqq.
By continuity of B f at x, given ε ¡ 0, D δ ¡ 0 such that η P Rn , }η } δ ñ }B f px η q
B f pxq} ε.
In particular,
}η} δ ñ }Bif px ηq
Bif pxq} ε @ i P t1, . . . , nu. (6.27)
For h P Rn , }h} δ, define r0 , . . . , rn by (6.25). Then, for i P t1, . . . , nu, ri  ri
1 hi vi where
tv1, . . . , vnu is the standard basis of Rn. So, taking the cue from (6.26), we express f px hq
f pxq by
f px hq
f pxq  f px rn q
f px rn
1 q f px rn
1 q
f px rn
2 q
 f px r2 q
f px r1 q f px r1 q
f px r0 q. (6.28)

For each i P t1, . . . , nu we can estimate f px ri q
f px ri
1 q by the Mean Value Theorem as follows.
Define gi : r0, hi s Ñ R by gi ptq : f px ri
1 tvi q. Then D θi P p0, 1q such that

f px ri q
f px ri
1 q  g phi q
g p0q  g 1 pθi hi q phi
0q  hi Bi f px ηi q

ANALYSIS III 54
 CHAPTER 6. THE DERIVATIVE

where
ηi : ri
1 θi hi vi  ph1, h2, . . . , hi
1, θihi, 0 . . . , 0q.
b
Now }ηi }  h21 ... h2i
1 θi2 h2i ¤ }h} δ. Therefore, by (6.27) we have

}f px ri q
f px ri
1 q
hi Bi f pxq}  }hi } }Bi f px ηi q
Bi f pxq} ¤ }hi }ε.

Summing up over i and using (6.28) yields:

ņ ņ
}f px hq
f px q
hi Bi f pxq} ¤ }f px ri q
f px ri
1 q
hi Bi f pxq}

i 1 
i 1

|hi| ¤ ε?n}h},
ņ
¤ε
i 1 
i.e., f is differentiable at x and Df pxq is the linear map
ņ
h  ph1 , . . . , hn q ÞÑ hi B i f px q : R n Ñ R.

i 1

ANALYSIS III 55
Chapter 7

Complex Analysis

This part of the course is an introduction to complex analysis. The main topics will be complex differ-
entiability, power series and contour integrals. Basic notions and properties for complex numbers were
introduction in Year 1 and we only provide a quick review here.

7.1 Review of basic facts about C

The field of complex numbers is given by

C  tz x iy, x, y P R u,
with i2 
1. For z  x iy as above we say that x is the real part of z, denoted by x  Re z and that
y is the imaginary part of z, denoted by y  Im z. By |z | we denote the modulus (or norm) of z, given by
a
x2 y 2 . We denote by z̄ the complex conjugate of z. That is, if z  x iy then z̄  x
iy. It is easy
to see that

1. z̄¯  z,

2. z w  z̄ w̄,

3. zw  z̄w̄,
4. |z|2  zz̄ and |z̄|  |z|.
Notice that we can identify C with R2 , simply by identifying z  x iy with px, y q. In this way |z |
corresponds to the Euclidean norm in R2 . We will not use }  }, the notation that we used for the norm in
R2 .
The notions of convergence, open and closed for C are identical to those in the plane (see results in
Sections 3.5 and 4.1).

Definition 7.1. We say that pzn q8n1 € C converges to z if and only if |zn
z | tends to zero as n goes to
8. That is, if for every ε ¡ 0 there exists N ¡ 0 such that |zn
z| ε for all n ¡ N .
Definition 7.2. We say that Ω € C is open if and only for every x P Ω there exits r ¡ 0 such that
Br pxq  tz P C |z
x| ru € Ω. We say that Ω is closed if and only if Ωc is open.

Definition 7.3. A set K € C is sequentially compact if and only if for every sequence pzj qj PN € K has a
convergent subsequence pzj plq qlPN whose limit is in K.

Now, maps in f : Ω € C Ñ C, are given by a pair of real-functions f pz q  upz q iv pz q, the real part u
of f and the imaginary part v of f . We can think of those two functions as functions of z or as functions
in R2 of x and y, the real and imaginary part of z. This means we can also think of f as a function from
Ω € R2 to R2 .

56
 CHAPTER 7. COMPLEX ANALYSIS

Definition 7.4. Given f : Ω € C Ñ C we say that it is continuous at z0 P Ω if and only if for every ε ¡ 0
there exists δ such that |z
z0 | δ, with z P Ω implies that |f pz q
f pz0 q| ε.

Notice that the notion of continuity coincides with the one defined for maps from R2 to R2 . We will
now consider the notion of differentiability, where the two notions differ very significantly.
Recall from Definition 6.4 that a function f : Rn Ñ Rk is differentiable at a point p if and only if there
exists a linear map Df ppq P LpRn ; Rk q such that

}f pp hq
f ppq
Df ppqh}
 0. (7.1)
lim
hÑ0 }h}
The reason for introducing that definition arose from the fact that when k ¡ 1 we have no notion of
division for the quantity we would like to study

f pp hq
f ppq
lim
Ñ0
h h
as division by h P Rn , n ¡ 1 is not well defined. However, in C we do have a notion of multiplication and
therefore we can use that quotient to define differentiability.

Definition 7.5. Let Ω € C be an open set and z P Ω. We say that f is complex differentiable at z if and
only if the limit
f pz hq
f pz q
lim (7.2)
hÑ0 h
exists. We denote the limit by f 1 pz q.

In contrast to what happened in the real valued case, where the derivative was a linear map from Rn to
Rk , which in our case would mean from R2 to R2 , corresponding to a 2 by 2 matrix, in the complex case we
obtain a complex number. Before studying how to reconcile this difference, we look at the consequences
of the definition for the real and imaginary part of f . Let’s write h  ∆x i∆y, and f pz q  upz q iv pz q,
which we can also think of as f px, y q  upx, y q iv px, y q. Then the quotient in the definition of complex
derivative can be rewritten as
f pz hq
f pz q
h
 upx ∆x, y ∆y q
upx, y q irv px
∆x i∆y
∆x, y ∆y q
v px, y qs
.

We could consider multiple ways of sending ∆x i∆y to zero, obtaining the same answer if the limit exists.
We will consider the two obvious options, sending ∆x first to zero followed by ∆y, and the reverse, ∆y
first followed by ∆x. We find

upx ∆x, y ∆y q
upx, y q irv px ∆x, y ∆y q
v px, y qs

lim lim
∆y Ñ0∆xÑ0 ∆x i∆y

 ∆ylimÑ0 upx, y ∆y q
upx, y q irv px, y

i∆y
∆y q
v px, y qs

 1i ∆ylimÑ0 upx, y ∆y q
upx, y q

∆y
i
v px, y ∆y q
v px, y q
∆y
 
 1 Bu px, yq B
i px, y q  vy px, y q
iuy px, y q,
v
i By By
while
upx ∆x, y ∆y q
upx, y q irv px ∆x, y ∆y q
v px, y qs
lim lim
∆x Ñ0∆yÑ0 ∆x i∆y

 ∆xlimÑ0 upx ∆x, y q
upx, y q irv px

∆x
∆x, y q
v px, y qs

ANALYSIS III 57
 CHAPTER 7. COMPLEX ANALYSIS

 ∆xlimÑ0 upx ∆x,∆xy q
upx, y q

i
v px ∆x, y q
v px, y q
∆x
 
 BBux px, yq i BBxv px, yq  uxpx, yq ivxpx, yq.
This immediately means that at the very least we need to demand some relationships between the partial
derivatives of u and v to hold in order to have a complex differential. Namely

ux  vy uy 
vx (7.3)

These equations are known as the Cauchy–Riemann equations. These are clearly necessary conditions, but
at this point in no way guarantee that a complex differential would exists if satisfied.
By considering two simple examples, it is easy to see that the notion of complex differential is highly
restrictive as functions that are obviously smooth when considered as a map from R2 to R2 are not actually
complex differentiable. First we consider f pz q  z. Notice that f 1 pz q exists and equals 1. Indeed
h
z
f 1 pz q  lim  1.
z
h Ñ0 h
However if we consider g pz q  z̄, we obtain a function that is not complex differentiable. We have

g pz hq
g pz q h̄
z̄
lim
hÑ0 h
 hlim
Ñ0
z̄
h
 hlim h̄
Ñ0 h
,

a limit that does not exist. (Consider for example the limits obtained by taking h along the real or the
imaginary axis.) The function g does not satisfy the Cauchy–Riemann equations. We have g pz q  x
iy,
and therefore
ux  1, vy 
1, uy  0, vx  0.
When considering g as a function from R2 to R2 we have g px, y q  px,
y q we clearly have a differentiable
function, as all components are smooth functions. (The existence of continuous partial derivatives suffices
to obtain differentiability, see Theorem 6.25.)
Definition 7.6. We say that f : Ω Ñ C is analytic (or holomorphic) in a neighbourhood U of z if it is
complex differentiable everywhere in U . We say that f is entire if it is analytic in the whole of C.
A function can be differentiable at one point, but not necessarily analytic. Consider as an example
the function f pz q  |z |2 . We will show that the function is complex differentiable at 0, but that it is not
analytic, as it is not complex differentiable outside the origin. Notice that f pz q  x2 y 2 , and u  x2 y 2
and v  0. When computing the Cauchy–Riemann equations we find

ux  2x uy  2y, vx  vy  0.
The Cauchy–Riemann equations mean 2x  0 and 2y  0, which is only satisfied at the origin. Now, to
check that f is complex differentiable at the origin

|z h|2
|z |2 
  |hh|  h̄ ÝÝÝÝÝÝÝ
2
Ñ 0,
h  hÑ0
z 0 
proving that f is complex differentiable at the origin with derivative 0.
We will now revisit the Cauchy–Riemann equations and connect complex differentiability with the
dependence of the function on z̄. Consider f pz q as given by upx, y q iv px, y q. Using the fact that x  z 2 z̄

z̄ we can rewrite the function back in terms of z and z̄. Now, we could consider the derivative
and y  z 2i
of f with respect to z̄. Applying the chain rule we would obtain
Bu  u 1
u 1 Bv  v 1
v 1
Bz̄ x 2 y 2i Bz̄ x 2 y 2i
ANALYSIS III 58
 CHAPTER 7. COMPLEX ANALYSIS

Therefore  
Bf  u ivz̄  ux
1

uy 2i1 1

vy 2i1 ,
Bz̄ z̄ 2
i vx
2
which we can simplify to
Bf  1 ru
v s 1
rvx uy s .
Bz̄ 2 x y 2
i

Notice that if the function is complex differentiable, it satisfies the Cauchy–Riemann equations and therefore
the expression above is identically zero. In this sense we say that if a function is complex differentiable,
then
Bf  0.
Bz̄
This illustrates why f pz q  z̄ or g pz q  |z |2  zz̄ were not complex differentiable.
Now that we know that the Cauchy–Riemann equations need to be satisfied for a function to be complex
differentiable we can identify the complex plane with a subspace of 2  2 matrices. This identification will
allow us to connect directly complex differentiability with the standard notion of differentiability discussed
in Chapter 6. We have already identified a ib with the point in R2 given by pa, bq. We can also identify
it with the matrix 
a
b
.
b a
Note that which factor of b contains a minus sign is just a convention. Notice that the determinant of that
matrix equals |a ib|2 , and that therefore the matrix is invertible unless a ib  0. This identification
preserves the basic operations we have for complex numbers, for example summation and multiplication.
That is it is possible to perform the operation pa ibq pc idq as complex numbers or as the sum of
the two corresponding matrices, with the results agreeing (modulo the identification). For the product we
have
pa ibqpc idq  pac
bdq ipbc adq
and   
a
b c
d  ac
bd
pbc adq
ac
bd
,
b a d c bc ad
proving the result. Sometimes it is useful to consider a hybrid of both identification, the one as a matrix,
and the one as a point (or vector) in R2 . For example, for the product of two complex numbers that we
have just considered, we could identify it with
 
a
b c
.
b a d

The answer is the vector 

ac
bd
,
bc ad

ac
bd
pbc adq
which corresponds to the right complex number pac
bdq ipbc adq and to the matrix .
bc ad ac
bd
We are now ready to connect complex differentiation with Cauchy–Riemann and differentiation for
functions in R2 .

Theorem 7.7. Let f : Ω € C Ñ C with Ω open. f is complex differentiable at z  a ib P Ω if and only

if f , when considered as map from Ω € R2 to R2 has a differential at the point pa, bq that satisfies the
Cauchy–Riemann equation.

Before we prove this result, we emphasize that some books will replace the right-hand side by asking
that the Cauchy–Riemann equations are satisfied and that all partial derivatives are continuous. Notice
that this last condition implies the existence of a differential.

ANALYSIS III 59
 CHAPTER 7. COMPLEX ANALYSIS

Proof. Assume that f is complex differentiable at z  a ib. Then we have

f pz hq
f p z q
h
lim
Ñ0 h
 f 1 pz q,
which we can rewrite as  
f z p hq
f pz q
f 1 pz qh 
lim 
hÑ0 h   0. (7.4)

In order to prove that f is differentiable as a map in R2 we need to find a linear map Df that satisfies
(7.1), which translates in finding a 2  2 matrix. Notice that (7.4) suggest that f 1 pz q P C should be the
map. Indeed if we identify f 1 pz q with the corresponding matrix, and think of f 1 pz qh not as a product of
two complex numbers but as a matrix acting on the vector h then we have in fact proven that f has a
differential. Since we already know that all complex differentiable functions satisfy the Cauchy–Riemann
equations we have completed that implication.
For the reverse, assuming that we have a differential, that means that we have a 2  2 matrix which is
given by 
Df ppa, bqq 
ux uy
vx vy
and that satisfies
|f ppa, bq hq
f ppa, bqq
Df ppa, bqqh|
 0.
h
lim
Ñ0 |h|
Since the Cauchy–Riemann equations are satisfied we know that this matrix does in fact have the form

ux
v x ,
vx ux

meaning that we could identify it with a complex number as before. We could therefore identify Df h with
the product of the complex numbers f 1 pz q  ux ivx and h. Identifying pa, bq with z we obtain
|f pz hq
f pz q
f 1 pz qh|
h
lim
Ñ0 |h| 0
which implies that
f pz hq
f pz q
lim
hÑ0 h
exists and equals f 1 pz q, completing the proof.

As a consequence of the above result, since we can connect complex differentials with differentials as
maps from R2 to R2 , we have the following results:
Theorem 7.8. Lef f, g : Ω € C Ñ C be complex differentiable functions. Then (asumming g  0 in the
third expression) we have that the familiar expressions
 1 1 1
pf g q1  f 1 g1 pf g q1  f 1 g f g1
f
g
 f g g
2 f g pf pgqq1  f 1pgqg1
apply to the complex-valued case as well. For the final expression one needs to assume that the composition
makes sense, i.e. the range of g is contained in the domain of f .
We conclude this section by proving that f pz q  z n is complex differentiable for every n P N. Using
Theorem 7.7 it suffices to show that it has a differential at every point and that it satisifies the Cauchy–
Riemann equations. Notice that since it is a polynomial (once expanded in terms of x and y and considered
as map from R2 to R2 we trivially have that it has a differential). To see that it satisfies the Cauchy–
Riemann equations, notice that (and similarly for v)

ux  pRef qx  Repfxq.
ANALYSIS III 60
 CHAPTER 7. COMPLEX ANALYSIS

Therefore
fx  ux ivx  n px iy qn
1 fy  uy ivy  n px iy qn
1 i.
Without computing what ux , vx , uy , vy are, notice that it follows from the expression above that

uy ivy  ipux ivx q,

which implies that ux  vy and uy 
vx, which are the Cauchy–Riemann equations.

7.2 Power Series

°
We want to focus on the study of power series, i.e. expressions of the form 8 n0 an z . We begin by
n

reviewing (in a very utilitarian way) some basic ideas of series for complex numbers covered in year 1.
°8
Definition 7.9. The series 
n 0 an , with an P C is convergent if and only if the sequence SN  °Nn0 an
is convergent in C.
°8
Definition
°8
7.10. The series 
n 0 an , with an P C is absolutely convergent if and only if the series
n0 |an | is convergent.
°8
The geometric series is convergent if and only if |z |

n 0z
n 1, and sums up to 1{p1
z q (with
partial sums SN  p1
z Nq{p
zq ). We review a couple of the convergence tests from year 1.
1 1
°
Theorem 7.11 (Ratio Test). Consider 8 n0 an and assume that an  0 for all n. Then

1. If lim sup |a|nan |1 |

°8
1 then 
n 0 an is convergent.

|an 1 | ¥ 1 for all n ¡ N then °8n0 an is divergent.

2. If |an |
In particular if lim |a|nan |1 | exists, and equals L we have convergence for L 1 and divergence for L ¡ 1.
(The test is inconclusive if L  1.)
°8
Theorem 7.12 (Root Test). Consider 
n 0 an . Then
°8
1. If lim sup |an |1{n 
1 then n 0 an
converges.
°
2. If lim sup |an |1{n ¡ 1 then 8n0 an diverges.

The proofs of these results are obtained by comparison with the geometric series and will not be covered
in these notes.
We will focus on studying expressions of the form
8̧ 8̧
an z n or a n pz
z 0 qn ,

n 0 
n 0

P C. The first observation is the existence of a radius of convergence.

with an , z

Theorem 7.13. Given pan q8

n0 there exists R P r0, 8s such that

8̧
an z n
n 0 
converges for all |z | R and diverges for |z | ¡ R. (As we will see in the proof R  1
| |1{n .)
lim sup an

ANALYSIS III 61
 CHAPTER 7. COMPLEX ANALYSIS

Proof. We consider z given, but fixed, and apply the root test to the series given by the sequence pan z n q8
n0 .
We know that the corresponding series is convergent if

lim sup |an z n |1{n

is less than 1 and divergent if it is greater than 1. But that implies covergence if

|z| 1
lim sup |an |1{n
and divergence when
|z| ¡ lim sup1|a |1{n ,
n
proving the result.

A simple application of the ratio tests yields the following result:

Theorem 7.14. Let an  0 for all n ¥ N , and assume that lim |a|a n 1 | °8 n
| exists. Then n0 an z has radius
of convergence R  lim |a|nan |1 | .
n

Next we will show that within the radius of convergence a power series is actually differentiable, and
that we can in fact compute the derivative term-by-term. More precisely:
°
Theorem 7.15. Assume 8
°8 n0 an z has radius of convergence R. Then for |z |
n R the function f pz q 
n
n0 an z is differentiable and
8̧
f 1 pz q  nan z n
1 .
 n 1

Proof. First we will show that the power

°8
series for f 1 pz q °
does have the same radius of convergence. Notice
that the radius of convergence of n1 nan z n
1 and 8 8
n° 1 nan z (i.e. where we have multiplied the
n

expression by z) is the same. To see this notice that if n1 nan z n
1 is convergent for |z | R and
divergent for |z | ¡ R then the same will apply to the second series. Therefore we just need to consider

lim sup |nan |1{n  lim n1{n lim sup |an|1{n  lim sup |an|1{n,
° °
which shows that the radius of convergence is the same as for 8
n0 an z . Notice that the series
n 8 n pn

n2
1qan z n
2 also has the same radius of convergence (we will need this result in our estimate below, even
though we never formally compute second derivatives).
Next, notice that for k P N we have

wk
z k
w
z
 wk
1 w k
2 z  wz k
2 z k
1 . (7.5)

Now, in order to prove that f is complex differentiable and compute its derivative we study

f pz hq
f pz q
8̧

nan z n
1 .
h 
n 1

We denote by w z  w
z), and substitute the expression for f in terms of a series to

h (and so h
find
8̧  wn
z n
an


nzn
1 .
n0
w z
We look more carefully at the term in brackets. Using (7.5) we find (taking k  n)
wn
z n
w
z

nzn
1  wn
1 wn
2z    wzn
2 zn
1
nzn
1
ANALYSIS III 62
 CHAPTER 7. COMPLEX ANALYSIS

 wn
1
zn
1 rwn
2
zn
2sz    pw
zqzn
2

 n
1 

z n
1 wn
2
z n
2 w
z n
2
 pw
zq w
z
w
w
z
z 
w
z
z . (7.6)

Now, for |z | r R and |w| r R we have

 k 
 w
zk  k
1
 w
z   |w
  wk
2 z    wz k
2 z k
1 | krk
1

and therefore  k
w

zk zn
k
1 ¤ krk
1rn
k
1 ¤ krn
2,
 w
z 

which substituted in (7.6) yields

 
¤ |w
z| |pn
1qrn
2 pn
2qrn
2    2rn
2 rn
2 |

¤ |w
z|rn
2 12 npn
1q.

We have shown that
 

p hq
f pz q
8̧  8̧

nan z
¤ |w
z| 21 npn
1q|an |rn
2 ¤ M |h|,
f z n 1
 
 
h 
n 1 
n 0

which goes to zero as h goes to zero. Notice that in the last inequality we have used that the series
°8 n
2 is finite, since we observed that the radius of convergence of the corresponding
n0 npn
1q|an |r
power series was also R.

We have the following simple consequence of the Theorem above, which allows us to compute the
coefficients an in terms of derivatives of f .
°
Corollary
°8
7.16. Let 8 n
n0 an z be a power series with radius of convergence R ¡ 0. Then f pz q 
n0 an z is infinitely differentiable for |z |
n R and moreover

f pnq p0q  an n!, n  0, 1, 2, . . .

Proof. The result is trivial for f p0q, as it clearly equals a0 . A simple induction argument using the formula
for the derivative of f in the previous Theorem yields the desired result.
°
Theorem 7.17. Let 8 n0 an z be a power series with radius of convergence R ¡ 0. Then for every r
n R
the sequence of functions
ķ
fk : an z n

n 0

converges uniformly in |z | ¤ r.

Proof. We show the result by proving that pfk q is uniformly Cauchy in |z | ¤ r. We have (assuming that
j ¤ k)
ķ ķ 8̧
|fk pzq
fj pzq|  | an z n | ¤ |an|rn ¤ |an|rn.

n j 1 
n j 1 
n j 1
°8
Since by assumption n0 |an |rn is finite, given any ε ¡ 0 we can choose N large enough to make
|fk pzq
fj pzq| ε for all j, k ¡ N , concluding the proof. (This proof is essentially an application of the
Weierstrass M-test that we covered a few weeks ago.)

ANALYSIS III 63
 CHAPTER 7. COMPLEX ANALYSIS

7.2.1 The exponential and the circular functions

Many of these functions should have appeared in year I, though perhaps only in the real-valued case.

Definition 7.18. We define the following power series for z PC.

8̧ 1
ez :  zn, (7.7)
n0
n!
8̧ p
1qn
cospz q :  2n
(7.8)

n 0
p2nq! z ,

8̧
coshpz q : 
1 2n
(7.9)
n0
p 2n q !
z ,

8̧ p
1qn
sinpz q :  z 2n 1
(7.10)
n0
p 2n 1 q !
,

8̧
sinhpz q : 
1
z 2n 1
(7.11)
n0
p2n 1 q!
.

The ratio test shows (Exercise) that the radius of convergence of all of the series above is R  8.
Notice that using Theorem 7.15 we can prove well known identities like pez q1  ez . Indeed

8̧ 1 1 8̧ 8̧ 1
p e q1 
z
z n
 n n
1
z  z n  ez .
n0 n1 n0
n! n! n!

In fact, we can easily relate all the circular functions to the exponential.

Proposition 7.19. The following identities hold for all z P C:

e
iz
iz
 e
2ie ,
iz iz
cos z e 2
, sin z

e
z
z
 e
2 e .
z z
cosh z e 2
, sinh z

Proof. We only prove the first one. The others are very similar and are left as an Exercise.
 
eiz e
iz
8̧ 1 8̧ 1
 1
pizqn p
izq n

n0 n0
2 2 n! n!

8̧ in p
iqn  8̧ p
1qk
 1
zn 
p q z 2k ,
n0
2 n! k 0
2k !
where we have used that #
in
p
iq  n 2piqn  2p
1qn{2 n even
.
0 n odd

There are additional relationships between sine and cosine and their hyperbolic counterparts. Notice
that we have

cospiz q  coshpz q coshpiz q  cospz q sinpiz q  i sinhpz q sinhpiz q  i sinpz q,

which shows that sine and cosine are unbounded functions in the complex plane. Just consider z  iy for
y P R together with the fact that the real valued sinh and cosh grow exponentially at infinity.

ANALYSIS III 64
 CHAPTER 7. COMPLEX ANALYSIS

Theorem 7.20. The exponential function ez satisfies the following properties

1. ez w  ez ew for all z, w P C.
2. ez  0 for all z P C.

3. ez  1 if and only if z  2kπi for k P Z, and as a result ez w  ez if and only if w  2kπi, k P Z.

Notice that in particular we have shown ez 2kπi  ez for all k P Z, so in this sense the exponential
is periodic in the imaginary variable.

4. ez 
1 if and only if z  p2k 1qπi for k P Z.

Proof. We present the direct proof of part 1, without using more advanced tools from complex analysis
that would reduce the heavy computational nature.
 
¸ 1 8̧ 1 8̧ ¸ z n wk
z w
e e  z n
wk 
n 0  n! k 0  k! l 0 n,k
n! k!
n k l 
8̧ 1 l j l
j
ļ  8̧ 1
 z w  pz w ql  ez w
.
l 0 j 0
l! j l 0
l!

For part 2, notice that ez e
z  1, proving that ez  0. For part 3, denoting z  x iy we find
ez  ex eiy  ex pcos y i sin y q,

which equals 1 if and only if ex  1 and cos y i sin y  1. These happen if and only if x  0 and y  2πk,
k P Z. Similarly for part 4.

7.2.2 Argument and Log

Every complex number z P Czt0u can be written in the form z  |z |eiθ , where θ is the angle that the
vector z forms with the x axis, measured counter-clockwise. Of course that angle is not unique (but rather
up to factors of 2π. Notice that for z  0 there is no natural way to choose an angle.
y

1.5

r
1.0

θ
0.5

x
-1.5 -1.0 -0.5 0.5 1.0 1.5

-0.5

-1.0

-1.5

Figure 7.1: Polar representation of a complex number

We can define the (multivalued) function, for z  0,

argpz q  tθ P R : z  |z |eiθ u. (7.12)

It is not a function as such, as the image is not uniquely defined, and if θ P argpzq then so is θ 2kπ. The
following are easily verified properties of arg.

ANALYSIS III 65
 CHAPTER 7. COMPLEX ANALYSIS

Proposition 7.21.

1. argpαz q  argpz q for all α ¡ 0.

2. argpαz q  argpz q π  tθ π, for θ P argpzqu for all α 0,

3. argpz̄ q 
argpz q  t
θ, for θ P argpz qu,

4. argp1{z q 
argpz q,

5. argpzwq  argpz q argpwq  tθ ϕ, with θ P argpz q, ϕ P argpwqu.

The ambiguity of the argument function can be solved by defining the principal value Arg of the arg
function to take values in p
π, π s. That is for any z P C we have Argpz q P p
π, π s.
Notice that it is impossible to define the Arg function continuously in the entire plane. In particular
as we approach any point in the negative real axis, if we do it from above the Arg function will yield π,
while if we do it from below it will
π. Observe that if we had made any other choice for the range of Arg
there would always be a half-line where we have the same issue, the difference between the values of the
argument when approaching from opposite sides is always 2π.
We want to define the logarithm by analogy of what happens in R. In the real valued case we say (here
w, z P R)
w  logpz q if and only if ew  z.
If we could extend this for w, z P C, since we know that ew  ew 2πik for any k P Z we would have that if
w  logpz q the so is w 2πik. Therefore we would have that logpz q is a multivalued function (just like it
happened before with argpz q, the argument function).
Let’s write z  |z |ei argpz q and w  logpz q  u iv. We have

eu iv
 eueiv  z  |z|ei argpzq,
and therefore comparing the two expressions in polar form we must have

eu  |z| and eiv  ei argpzq.

That means that u  log |z |, with this logarithm being the real logarithm. We will denote by Log the
logarithm in R to distinguish it from the complex valued we want to define. We define the multivalued
function
logpz q  Log|z | i argpz q. (7.13)
In terms of the Arg function we have

logpz q  Log|z | iArgpz q 2πik for k P Z.

For example if we compute the complex logarithm of 1 we have

logp1q  Log|1| iArgp1q  2πik

2πik for k P Z.

Notice that the definition above makes sense provided that z  0, where the real logarithm is not defined.
We can now compute logarithms of negative numbers.

logp
1q  Log|
1| iArgp
1q 2πik  iπ 2πik for k P Z.

The complex logarithm we have just defined obeys many of the properties that we know for the real
logarithm, with the caveat that we have to take care of the multi-valuedness of the function. For example

logpzwq  log z log w.

ANALYSIS III 66
 CHAPTER 7. COMPLEX ANALYSIS

To prove this result, notice that since Log|zw|  Logp|z||w|q  Log|z| Log|w| and that argpzwq 
argpz q argpwq we have

logpzwq  Log|zw| i argpzwq  Log|z | Log|w| i argpz q i argpwq  log z log w.

This equality needs to be understood modulo 2πi, that is there exists k P Z such that
logpzwq
logpz q
logpwq  2πik.

Similarly we have
logpz {wq  logpz q
logpwq.

If we want to consider the (complex) differentiability of the log we have to deal with the multi-valuedness
of the arg function. Indeed if we consider the incremental quotient

logpz ∆z q
log z
∆z
we need to make sure that as we approach z both logs approach the same value. We know that this cannot
be done continuously in the entire plane, and that we need to remove a semi-line arising from the origin.
For example if we consider Cztx ¤ 0u we can consider the principal branch of the logarithm, which by
an abuse of notation we denote by Log, just like the real logarithm, by

Logpz q  Log|z | iArgpz q.

This function, defined on Cztx ¤ 0u is single valued. If we consider points of the form z  x  iε, for
x 0 and ε ¡ 0 small, we find
lim Logpx  iεq  Log|x|  iπ,
ε Ñ0
showing that the function could not be extended continuously along tx 0u. This half-line is called a
branch cut. It is possible to compute the derivative of Log directly from the definition, or in terms of its
inverse. However, for practical purposes, once we know it is differentiable, from the identity

eLogz z
we find
eLogz pLogz q1 1
from which it follows that pLogz q1  1{z.
Once we have defined the notion of logarithm it is possible to consider defining complex powers of
complex numbers. Given α P C, and z  0 we define the α-th power of z by

z α : elogpz q  eαLog|z| p q.
α αi arg z

The multi-valuedness of arg means that the same is true for z α . If we rewrite the above as

zα  eαLog|z| p q  eαLog|z|
αi arg z pq
αiArg z 2παki
 eαLogpzqe2παki
for k P Z the multi-valuedness becomes more evident. The number of α powers, whether it is one, finitely
many or infinitely many will depend on α.
Indeed if α is an integer for example then e2παki  1, which means that in fact there is only one value
of z α . If α is rational, say α  p{q, with p, q coprime, then z α will have finitely many powers. It is easy
to see that for α  p{q (with p, q coprime, and q P N)

e2παki  e2παpk q i q

ANALYSIS III 67
 CHAPTER 7. COMPLEX ANALYSIS

and therefore z α will take q different values

eαLogpz q e2παki , k  0, 1, . . . , q
1.
In the case of an irrational α it will actually take infinitely many values.
In the rational case the result obtained above is consistent with what we know about finding roots of
polynomials. If we consider, for q P N the equation z q  1 we know it should have q roots which correspond
to
z  11{q .
Now, using the expressions above we find

11{q  eLogp1q{q e2πik{q  e2πik{q k  0, 1, . . . , q
1.

7.3 Complex integration, contour integrals

For a function f : ra, bs Ñ C we define
 b  b  b
f ptqdt  Re f ptqdt i Im f ptqdt. (7.14)
a a a
This definition means that we reduce integrating a complex-valued function to integrating two real-valued
functions, and can therefore use every result we know from before, such as the Fundamental Theorem of
Calculus to compute each integral.
It is easy to see that for every f, g : ra, bs Ñ C and every α, β P C we have
 β  b  b
rαf βg sdt  α f ptqdt β g ptqdt.
a a a

b
b
b
That a pf g qdt  a f dt a gdt follows immediately from the definition. We show the more tedious

b
b
a αf ptqdt  α a f ptqdt. We have (suppressing the limits of integration and dt for simplicity)
   
α f α Repf q i Impf q

     
 Repαq Repf q
Impαq Impf q i Impαq Repf q Repαq Impf q
  
 Repαq Repf q
Impαq Impf q i Impαq Repf q Repαq Impf q
  
 Repαf q i Impαf q  pαf q.
Notice that in this case  
b b
f ptqdt  f ptqdt. (7.15)
a a
Indeed
 b  b  b  b  b  b
f ptqdt  Re f ptqdt
i Im f ptqdt  Re f ptqdt i Im f ptqdt  f ptqdt.
a a a a a a

We also have the following estimate (which we will use repeteadly below)
  
 b  b

 pq ¤
f t dt |f ptq|dt. (7.16)
a a

ANALYSIS III 68
 CHAPTER 7. COMPLEX ANALYSIS

b 

f ptqdt   pq
 b 
To prove this result, assume that a Reiθ , where R  a f t dt. As a result of this
representation R also equals
 b  b
R  e
iθ f ptqdt  e
iθ f ptqdt.
a a
Now, if we write e
iθ f ptq  u iv, with u and v real valued. Then we must have
 b  b
R udt vdt  0.
a a
 
Notice that u  Rere
iθ f ptqs ¤ e
iθ f ptq ¤ |f ptq|. This implies
 b  b
R uptqdt ¤ |f ptq|dt.
a a


But since R equals  a f ptqdt we are done.
 b 

The definition above is a natural choice for integrating functions from R to C, with a far less obvious
choice for integrating a function from C to C. Instead, we want to study integrals of complex valued-valued
functions along curves, that is, expressions of the form

f dz,
Γ

where Γ is curve in the complex plane. To define a curve in C, consider a function γ : ra, bs Ñ C, given
by γ ptq  xptq iy ptq. We will ask that the curve γ be C 1 . The primary reason is that we want to
have a well defined tangent at every point of the curve (which is also integrable). We say that the curve
Γ  γ pra, bsq € C is parametrised by the map γ.
Definition 7.22. Given a function f : Ω € C Ñ C along the path Γ € Ω € C parametrised by
γ : ra, bs Ñ C the integral of f over Γ is given by
  b  b  b
f dz  f pγ ptqqγ 1 ptqdt  Repf pγ ptqqγ 1 ptqqdt i Impf pγ ptqqγ 1 ptqqdt.
Γ a a a

Notice that we are not making any regularity assumptions on f , just that the integrals are well defined.
Sometimes
we will
consider more than
one parametrisation of a curve Γ, say γ1 and γ2 and will use the
notation γ1 f and γ2 f in addition to Γ .
On many occasions we want to consider curves that are not C 1 but perhaps just piece-wise C 1 . For
example a square. In this case we can think of Γ as a union of n curves Γj , each one C 1 , and parametrised
in the right direction, so that connected in the right order they describe the entire curve Γ. We can define
 ņ 
f dz : f dz.
Γ 
j 1 Γj

It is straight forward from the definition (details are left as an Exercise) that given a curve Γ, and two
functions f, g : C Ñ C and α, β P C we have
  
pαf pzq βg pz qqdz α f pz qdz β g pz qdz.
Γ Γ Γ

If we allow for γ 1 ptq not to exists at finitely many points, this can be defined as a single integral, with
clearly both formulations being equivalent.
Example 7.23. Let f : C Ñ C be given by f pz q  f px iy q  x4 iy 4 and the curve joining the origin in
a straight line to the point 1 i, parametrized by γ : r0, 1s Ñ C, γ ptq  p1 iqt. Notice that γ 1 ptq  1 i
and so we have   1  1
f pt it qp1 iqdt  2it4dt  52 i.
4 4
Γ 0 0

ANALYSIS III 69
 CHAPTER 7. COMPLEX ANALYSIS

In the next Lemma we want to show that Γf depends only on the orientation of the parametrisation
of the curve. More precisely
Lemma 7.24. Let Γ be a curve in C, parametrised by γ : ra, bs Ñ C, that is γ pra, bsq  Γ. Given
f : Ω € C Ñ C and Γ € Ω we have:
1. if γ
represents the parametrisation of γ in the opposite direction, then
 
f 
f.
γ
γ

If a curve Γ has attached a sense of direction we will call it a directed curve. In this case we will
denote by
Γ the same curve swept in the opposite direction. Without the need to specify the
parametrisation we can reformulate the above result by
 
f dz 
f dz.
Γ
Γ
2. If γ̃ : rã, b̃s Ñ C is another parametrisation of Γ that preserves the orientation then
 
f  f.
γ̃ γ

We refer to this fact as reparametrisation invariance. [In practise, with the regularity we are demanding
on the curves, this means that there exists ϕ : rã, b̃s Ñ ra, bs, bijective and increasing, such that
γ̃  γ pϕq.]
Proof. 1. Notice that if γ : ra, bs Ñ C parametrises the curve in one direction then γ
is given by
γ
: ra, bs Ñ C with γ
ptq  γ pa b
tq. Therefore
  
b 1 b
f  f pγ
ptqq γ
ptqdt  f p γ pa b
tqqp
γ 1 pa b
tqqdt
γ
a a
 a  b 
 f pγ psqqp
γ 1 psqqp
1qds 
f pγ psqqγ 1 psqds 
f.
b a γ

2. The proof is very similar to part one.

  b̃  b̃  b 
f  f pγ̃ ptqqγ̃ 1 ptqdt  f pγ pϕptqqqγ 1 pϕptqqϕ1 ptqdt  f pγ psqqγ 1 psqds  f,
γ̃ ã ã a γ

where we have made the change of variables ϕptq  s and therefore ϕ1 ptqdt  ds

Consider the function f pz q  1 as a complex-valued function and a curve γ : ra, bs Ñ C. Then

  b
f dz  γ 1 ptqdt.
γ a

Here γ 1 ptq is a complex valued number and the integral
will be a complex

1 number. For example if we take
γ just like in Example 7.23 we have γ 1 ptq  1 i and γ f dz  0 p1 iqdt  1 i. This is because we
are considering dz as complex valued, given by γ 1 ptqdt.
We could consider defining the integral
  b  ba
|dz| : |γ 1ptq|dt  px1ptqq2 py1ptqq2dt  lpγ q,
γ a a

where γ : ra, bs Ñ C is given by γ ptq  xptq iy ptq, and lpγ q stands for the length of the curve γ.

ANALYSIS III 70
 CHAPTER 7. COMPLEX ANALYSIS

Similarly for f : C Ñ C we can define

  b
|f ||dz| : |f pγ ptq||γ 1ptq|dt.
γ a


Notice that γ |f ||dz| ¥ 0 and that we have
  
 
 f dz 
  ¤ |f ||dz|.
γ γ

To show this notice that (using (7.16))

     
   b  b
 f dz 
   
 f pγ ptqqγ 1 ptqdt ¤  |f pγ ptq||γ 1ptq|dt  |f ||dz|.
γ a a γ

We can further estimate the right-hand side

 
|f ||dz| ¤ max
z PΓ
|f pzq| |dz|  max
z PΓ
|f pzq|lpγ q.
γ γ

Therefore we obtain  
 
 f dz 
  ¤ max
z PΓ
|f pzq|lpγ q.
γ

Definition 7.25. Given f : C Ñ C and a curve γ : ra, bs Ñ C we define

  b
f dz̄ : f pγ ptqqγ 1 ptqdt.
γ a

Observe that in general 

f pz qdz
γ
is not equal to 
f pz qdz,
γ
unlike when we considered functions f : ra, bs Ñ C; see (7.15). Instead we have
  b  b  b 
f pz qdz  f pγ ptqqγ 1 ptqdt  f pγ ptqqγ 1 ptqdt  f pγ ptqq γ 1 ptqdt  f pz qqdz̄.
γ a a a γ

We compute a few more examples of integrals along curves.

Example 7.26. Integrate f pz q  z̄ (the definition does not require functions to be analytic) along the
circle of centred at 1 i of radius 2 (oriented counterclockwise).
First we describe the curve γ. Notice that 2eit for t P r0, 2π q describes a circle or raidus two centred
at the origin and with the required orientation. Therefore γ ptq  p1 iq 2eit for t P r0, 2π q. We have
γ 1 ptq  2ieit . Therefore the integral becomes
  2π  2π  2π
f pz qdz  pp1 iq 2eit q2ieit dt  2p1
iqi eit dt 4i  8πi,
γ 0 0 0

2π
since 0 eit dt  0. Indeed
 2π  2π  2π
e dt 
it
cos tdt i sin tdt  0.
0 0 0

In fact  2π
eint dt  0, for all n  0.
0

ANALYSIS III 71
 CHAPTER 7. COMPLEX ANALYSIS

Example 7.27. Integrate f pz q  z along the circle of centred at 1 i of radius 2 (oriented counterclock-
wise). As before γ ptq  p1 iq 2eit for t P r0, 2π q. We have γ 1 ptq  2ieit . Therefore the integral
becomes
   
2π  2π 2π
f pz qdz  p1 iq 2e 2ie dt  2p1
it it
iqi it
e dt 4ieit dt  0,
γ 0 0 0

2π
using that 0 eint dt  0, for all n  0.

Theorem 7.28. Assume that F : Ω € C Ñ C is analytic (Ω open) and set f pz q  dF

dz , with f continuous.
Let γ : ra, bs Ñ Ω be a C 1 curve. Then

f dz  F pγ pbqq
F pγ paqq.
γ

Proof. We have
  b  b  b
 f pγ ptqqγ 1 ptqdt  pγ ptqqγ 1ptqdt  F pγ ptqqdt  F pγ pbqq
F pγ paqq.
dF d
f dz
γ a a dz a dt

We remark that there are no assumptions made about Ω other than it is open. That is, all we need
for the result to be true, is that f is analytic in an open neighborhood of the curve. The notion of simply
connected (for a domain) will be defined later, but we emphasize that there is no such requirement on Ω
above result to be true.

7.3.1 Links with Green’s and Gauss’ Theorems

We want to connect the notion of contour integral with the notions introduced in MA144. These results
will be covered extensively in MA263 Multivariable Analysis.
The line integral we have just defined has many similarities the with notion of tangential line integral
introduced in MA144 for a vectorfield F. There the definition read
  β
F  dr : Fprptqq 
dr
dt,
C α dt
where C is a curve parametrised by r : rα, β s Ñ with rpαq  p and rpβ q Rn  q. For closed curves that
integral is usually referred as circulation.
Another integral arising in MA144 is the flux integral, which is given by

F  ndt.
C

Here n represents the normal, with the following convention. If the curve C is parametrised by rptq 
pxptq, yptqq, and r1ptq  px1ptq, y1ptqq has the same direction of the tangent, we choose nptq to have the
same direction as
r1 ptqK  py1ptq,
x1ptqq.
We don’t assume the parametrisation γ to be arc-length so r1 might not have norm 1. When considering
the curves determining the boundary of a regular domain we will consider them as positively oriented. That
is, choose the orientation so that the corresponding n as defined above corresponds (i.e. has the same
direction as) to the outward normal.
The following results (considered here only for two dimensions) correspond to Green’s and Gauss’
Theorems. For a positively oriented regular domain Ω we have
 
curl Fdxdy  F  dr
Ω BΩ
ANALYSIS III 72
 CHAPTER 7. COMPLEX ANALYSIS

and  
divFdxdy  F  ndt.
Ω BΩ


Now let’s consider our contour integral γ f pz qdz for a function f  u iv and a curve γ ptq 
γ1 ptq iγ2 ptq. We have
  b
f pz qdz  rupγ ptqq iv pγ ptqqsrγ11 ptq iγ21 ptqsdt
γ a

 b  b
 upγ ptqqγ 1 ptq
v pγ ptqqγ 1 ptqdt
1 2 i upγ ptqqγ21 ptq v pγ ptqqγ11 ptqdt
a a
 b  b
 pu,
vq  pγ11 , γ21 qdt i pu,
vq  pγ21 ,
γ11 qdt
a a
 
 pu,
vq  dr i pu,
vq  ndt,
γ γ

and so if we define the vector field f  pu,
vq, we have just shown that

f dz  circulationpf q i fluxpf q.
γ

Using the above expression, together with Green’s and Gauss’ Theorem we can prove the following
result.

Theorem 7.29 (Cauchy’s Theorem). Let f : Ω Ñ C be an analytic function, with Ω an open, simply
connected domain. Let γ be a C 1 closed curve in Ω. Then

f pz qdz  0.
γ

Before we prove the result we define simply connected. Loosely speaking means that the domain
contains no holes. A set of more formal definitions is as follows.

Definition 7.30. A set Ω € C is connected if it cannot be expressed as the union of non-empty open sets
Ω1 and Ω2 such that Ω1 X Ω2  H. An open, connected set Ω € C is called simply connected if every
closed curve in Ω can be continuously deformed to a point.

Proof. The proof presented here assumes that the curve a simple, regular curve and that f 1 is continuous.
If the domain is simply connected, the region inside the curve does not have any holes, and f is analytic
in it. We know 
f dz  circulationpf q i fluxpf q
γ
 
 curl f dxdy i divf dxdy.
Ω Ω

We claim that both terms are actually 0, because curl f  divf  0. Since f  pu,
vq we have
divf  ux
vy curl f 
vx
uy

but since f u iv is analytic it satisfies the Cauchy–Riemann equations,

ux  vy vx 
uy
which imply the result.

ANALYSIS III 73
 CHAPTER 7. COMPLEX ANALYSIS

Notice that Cauchy’s Theorem applies to Example 7.27, where the function is analytic, but obviously
not to Example 7.26, where the function is not analytic.
Cauchy’s Theorem works for more general curves. Consider the shaded region Ω in Figure 7.2. If we
think of its boundary as a one curve Γ, even though it is formed by two separate curves we have

f dz  0,
Γ

provided that Γ is oriented positively. That means that the exterior curve, that we denote by γ1 needs to
be oriented counter-clockwise, while the interior curve, denoted by γ2 has to be oriented clockwise.

Figure 7.2: Region bound by two positively oriented curves

An equivalent formulation of this fact, which will be extremely useful is known as the deformation of
contour Theorem.

Theorem 7.31. Let Ω € C be a region bounded by two simple curves γ1 (the exterior curve) and γ2 (the
interior). Assume they are oriented positively, and let f be an analytic function in Ω Y γ1 Y γ2 . Then
 
f dz f dz  0.
γ1 γ2

If we denote by γ2
the anti-clockwise parametrization of γ2 , then the result can be rephrased as
 
f dz  f dz,
γ1 γ2

that is the integral is the same along both curves when both are parametrised counter-clockwise.

Proof. The proof is based on creating two new contours of integration, the boundaries of two simply
connected regions where f is analytic so that we can apply Cauchy’s Theorem 7.29.
To achieve this we add two new curves to the previous picture, now in yellow in Figure 7.3. They join
the points A (in γ1) with D (in γ2 ) and the points B (in γ1) with C (in γ2 ). The two curves we want to
consider are denoted by ρ and η. Each one of them is piecewise C 1 and formed by four sections. Each one
of these curves is oriented positively with respect to the region they enclose, that is, they are both oriented
counter-clockwise.
By Cauchy’s Theorem
    
f dz  f dz f dz f dz f dz  0, (7.17)
ρ ρ1 ρ2 ρ3 ρ4
    
f dz  f dz f dz f dz f dz  0. (7.18)
η η1 η2 η3 η4

ANALYSIS III 74
 CHAPTER 7. COMPLEX ANALYSIS

Figure 7.3: two positively oriented curves

We observe that η1 and ρ4 correspond to the same curve but with parametrisations in opposite directions.
Similarly for η3 and ρ2 . Therefore
   
f dz f dz 0 f dz f dz  0.
η1 ρ4 η3 ρ2

Adding (7.17) and (7.18) and using the above identities we find
   
f dz f dz f dz f dz 0
ρ1 ρ3 η2 η4

Also notice that ρ1 and η4 together build γ1 , while ρ3 and η2 build γ2 . Therefore, the above equality can
be rewritten as  
f dz f dz  0.
γ1 γ2

Since  
f dz 
f dz
γ2 γ2

we obtain  
f dz  f dz
γ1 γ2

as required.

We now compute one of the fundamental contour integrals. We will show that
 #
n 
1,
pz
aqndz  2πi
(7.19)
BBr paq 0 n  1,

where B Br paq denotes the boundary of the ball of radius r, parametrised counter-clockwise (i.e. positively
oriented with respect to Br paq).
Observe that the result is uniform with respect to r. That is a natural consequence of Theorem 7.31,
given than the functions we are integrating only fail to be analytic at one point (at most, depending on
n). In fact we could have chosen any curve that wraps around a once and obtain the same result.

ANALYSIS III 75
 CHAPTER 7. COMPLEX ANALYSIS

Now, to compute the integral above, notice that we can parametrise the curve as γ ptq a reit , for
t P r0, 2π q. Therefore we have (since γ 1 ptq  ireit )
  2π  2π
pz
aq n
dz  pre qit n
ire dt  ir
it n 1
eipn q dt.
1 t
BBr paq 0 0

Notice that in the case n 
1 that expression equals 2πi. When n 
1 notice that we obtain 0, since
for all k  0 we have  
1 ikt 2π
2π
e dt  e  
 0.
ikt 1 1
0 k 0 k k
We restate, in the notation that will be most convenient for the next few results, the fundamental integral
above in the case n 
1, noting that the result does not depend on r. We have

1
 2πi.
BBr pzq w
z
dw

Definition 7.32. Given a simple closed C 1 curve γ we denote by I pγ q the interior region to γ. We denote
by Opγ q the exterior region to γ.
Notice that by the deformation of contours Theorem we have
 

w
z
1
 1
 2πi (7.20)
BBr pzq w
z
dw dw
γ

for every z P I pγ q and every r sufficiently small so that Br pzq € I pγ q.

Theorem 7.33. Let γ : ra, bs Ñ C be a positively oriented simple closed C 1 curve. Assume that f is
analytic in γ and on the interior of γ, I pγ q. Then

f pw q
f pz q  for all z P I pγ q.
1
(7.21)
2πi γ w
z
dw

Proof. Fix z P I pγ q, and choose r small enough so that Br pz q € I pγ q. By the deformation of contours
theorem we have  
f pw q f pw q
dw 
1 1
2πi γ w
z 2πi BB pz q w
z
dw,
r

reducing the problem to considering γ as a B Br pz q. Observe that the integral is the same for every r
sufficiently small, and later on we will exploit this fact by talking limits as r tends to zero. For now, we
have
  
f pw q f pz q f pw q
f pz q
dw  dw : I II.
1 1 1
2πi BBr pz q w
z 2πi BBr pz q w
z w
z
dw
2πi BBr pz q
Notice that the first integral I equals f pz q. Indeed, using (7.20)
 
f pz q
1
w
z
 f pzq 2πi
1 1
 f pz q.
BBr pzq w
z
dw dw
2πi BBr pzq
All that remains to is to show that II  0. Notice that since f is analytic in I pγ q, given any ε ¡ 0 we can
find r sufficiently small so that
|f pwq
f pzq| ¤ ε for all w P B Br pz q.
We parametrise B Br pz q counterclockwise by γ ptq  z reit for t P r0, 2π q. We have γ 1 ptq  ireit and
therefore  
|II | 

 1


f w p q
f pzq dw ¤  1  2π f pz reit q
f pz q it 


 2πi BBr pz q w
z   2πi 0 reit

ire dt

2π 
¤ 2π1 |f pz reit q
f pz q|dt ¤ ε.
0
Since ε is arbitrary we obtain the desired result.

ANALYSIS III 76
 CHAPTER 7. COMPLEX ANALYSIS

Remark 7.34. The formula


f pw q
f pz q  P I pγ q
1
w
z
dw for all z
2πi γ

has remarkable consequences for analytic functions. First notice that it claims that we can recover the
value of f at any point by integration along a curve around that point (provided the curve is sufficiently
regular, positively oriented, and contained in I pγ q). This is a very significant difference with respect to
smooth functions in R2 for example.
Notice that since the curve γ is a compact set, for any point z P I pγ q the expression w
z found in
the denominator in Cauchy’s formula is bounded away from zero, suggesting that we can differentiate the
formula with respect to z to obtain

f pw q
f 1 pz q 
1
2πi γ pw
zq2 dw.
Of course we need to justify moving the derivative inside the integral sign. We assumed that f was analytic,
which means that f 1 pz q exists. The expression above would produce a formula for it, a way to compute it.
The key observation is that without assuming that f has more derivatives it seems that the right hand side
can be differentiated arbitrarily many times, which would suggest that f has infinitely many derivatives.
This is indeed the case as we will show in the next Theorem.

Theorem 7.35. Let γ : ra, bs Ñ C be a positively oriented simple closed C 1 curve. Assume that f is
analytic in γ and on the interior of γ, I pγ q. Then f pnq pz q exists for all n P N and the derivative is given by

f pw q
f pnq pz q  P I pγ q .
n!
(7.22)
2πi γ pw
zqpn 1 q dw for all z

Proof. Notice that Theorem 7.33 would correspond to the case n  0 in the current Theorem. In order to
prove the result for n  1 we consider the incremental quotient, and use (7.21) to obtain
   
f pz hq
f pz q f pw q f pw q
h
 1 1
h 2πi w
z
h
dw

1
2πi w
z
dw .
γ γ

By the deformation of contours Theorem we can choose γ as B B2r pz q, with B2r pz q € I pγ q . We have,
operating on the right-hand side

f pz hq
f p z q f pw q
 1
p

hqpw
z q
dw
h 2πi BB2r pzq w z
   
f pw q
 1
pw
zq2 dw
1
f pw q
1
pw
z
hqpw
zq
pw
z q2
1
dw
2πi BB2r pzq 2πi BB2r pzq
   
f pw q hf pwq
 2πi
1
pw
zq2 dw
1
pw
z
hqpw
zq2 dw.
BB2r pzq 2πi BB2r pzq
To conclude the proof all that we need to do is show that the limit of the last integral as h tends to zero
is zero, that is (ignoring factors of 2πi)
  
hf pwq
lim
hÑ0 BB2r pzq pw
z
hqpw
zq2 dw  0,
and recall that we are able to choose r arbitrarily small without affecting the value of the integrals above.
First we choose |h| r so that for all w P B B2r pz q we have

|w
z
h| ¥ |w
z|
|h| ¡ r.
ANALYSIS III 77
 CHAPTER 7. COMPLEX ANALYSIS

Here we have used the reverse triangle inequality in the first case, and the fact that |w
z |  2r for points
w P B B2r pz q. Choosing γ ptq  z 2reit for t P r0, 2π q, we have γ 1 ptq  2rieit , and therefore |γ 1 ptq| ¤ 2r.
Since f is analytic, in particular it is continuous and therefore there exists M ¡ 0 such that |f pwq| ¤ M
for all w P B B2r pz q. Using these facts we have
    

hf pwq  2π
¤ 2rdt  2 h,
  hM πM

 BB2r pz q w p
z
hqpw
zq 2
dw
 0 rp2rq 2 r

which goes to zero as h goes to zero, proving the result for n  1. The general case is proven by induction.
If we assume the result for n  1, 2,    , k
1 we want to prove it for n  k. That is, in particular we
assume
f pk
1q pz q 
pk
1q!  f pwq dw for all z P I pγ q.
γ pw
z q
2πi pk q
We write the corresponding incremental quotient, just as before

f pk
1q pz hq
f pk
1q pz q
1 q! 
 h1 pk 2πi f pw q

pk
1q!  f pwq dw .
γ pw
z
hq γ pw
z q
k
dw k
h 2πi

By the deformation of contours Theorem we can choose γ as B B2r pz q, with B2r pz q € I pγ q. We have,
operating on the right-hand side

f pk
1q pz hq
f pk
1q pz q p k
1 q! f pwqrpw
z qk
pw
z
hqk s
 pw
z
hqk pw
zqk dw
h 2πih BB2r pzq

f pw q
 2πi
k!
pw
z qp k q dw
BB2r pzq 1
 
pk
1 q!  f pw q
rpw
z qk
p w
z
hqk s

pw
zqpk 1q dw
k
2πi BB2r pzq hpw
z
hqk pw
z qk
  
 2πi
k! f pw q pk
1 q!  f pw q
p w
z qk 1
pw
z
hqk pw
z q
khpw
z
hqk
BB2r pzq pw
z q k 1
dw
2πi B B pz q
2r
hpw
z
hqk pw
z qk 1
dw.
(7.23)
As before, all that remains is to show that the last integral tends to zero as h tends to zero. We choose h
and the parametrisation as above. The result will follow if we show that

p
zqk 1
pw
z
hqk pw
zq
khpw
z
hqk  ¤ C |h|,
 w

 h 

where the constant might depend on r. This is the case because, as before |f | ¤M and |w
z
h| ¥
|w
z|
|h| ¡ r implies  
 
 1 ¤ 1
p
 w
z
hq pw
zq p2rqk rk .
k k 

In order to prove (7.23), notice that the binomial formula implies

ķ 
pw
z
hq  k k
pw
zqk
j p
hqj

j 0
j

and therefore
pw
zqk 1
pw
z
hqk pw
zq
khpw
z
hqk
ķ  ķ 

k
pw
z q k 1 j
p
hq
kh
j k
pw
zqk
j p
hqj

j 2
j 
j 1
j

which is of order h2 , proving the result.

ANALYSIS III 78
 CHAPTER 7. COMPLEX ANALYSIS

7.3.2 Consequences of Cauchy’s Theorem

Theorem 7.36 (Taylor Series Expansion). Let f be an analytic function on BR paq for a P C, R ¡ 0.
There there exist unique constants cn , n P N such that
8̧
f pz q  cn pz
aqn for all z P BR paq.

n 0

Moreover, the coefficients cn are given by

 pnq
f pw q
cn  1
2πi pw
aqn 1
dw  f n!paq ,
γ

where γ is any positively oriented simple closed curve (piece-wise C 1 ) that is contained in BR paq with
a P I pγ q.

Proof. Given some z P BR paq we will take γ to be B Br paq (positively oriented), for r small enough so that
|z
a| r R. We can use the Theorem of deformation of contours to prove the integrals over all curves
γ as above are the same. Cauchy’s formula (7.21) gives

f pw q
f pz q 
1
(7.24)
w
z
dw.
2πi BBr paq
Notice that since |w
a|  r and we have chosen r so that |z
a| r we have |z
a| |w
a| for all
w P B Br paq. As a result
|z
a| 1
|w
a|
and we can use the geometric series expansion to obtain
8̧ 
z
a n

w
z
 w
a

1 1 1


 w
a 1
w
a
.
1
z a
w a
n 0
Inserting this expression in (7.24) we obtain
 8̧ 
z
a n
f pz q  f pw q
1 1
w
a n0 w
a
dw.
2πi BBr paq
For w P B Br paq the series converges absolutely (Weierstrass M-test), and therefore we can exchange the
order of the summation and integration to obtain
8̧ 1  f pw q
8̧
f pz q  dwpz
aq 
n
cn pz
aqn ,
n0
2πi B B r paq p w
a q n 1
n0

obtaining
°
the desired result. It remains to show that the coefficients are unique. Now, assume that
f pz q  8k0 bk pz
aq for some bk P C. We have
k


f pw q
 8̧
pw
aqn dw  bk pw
aqk
1
pw
a qn dw
BBr paq 1
BBr paq k0 1

8̧ 
 bk pw
aqk
n
1dw  2πibn,

k 0 BBr paq
where we have used the fundamental integrals, together with the fact that we can commute the summation
and integration. This proves that bn  cn , concluding the proof.

ANALYSIS III 79
 CHAPTER 7. COMPLEX ANALYSIS

Example 7.37. We consider an example of a Taylor series. We consider the function p1 z qa for a P C
and |z | 1.
When we consider logarithms we noticed that z n is well defined for n P N, but not for any a, without
making any specific choice of the argument function. In this case
ņ 
p1 zq n
 n k
k
z ,

k 0
which is a polynomial of order n, and equals the Taylor series expansion centred at the origin. This series
converges for every z P C, not just |z | 1. However, we defined

f pz q  p1 z qa : eaLogp1 z q

having made a choice of the argument function defining the logarithm, which meant creating a branch cut
where the function was not defined. Choosing the argument in p
π, π q, and since our function is translated
(not z a ) we obtained a function that is not defined for z P p
8,
1s.
We want to show that in fact a binomial expansion is possible for all a P C. We know by Taylor’s
Theorem 7.36 that we have a Taylor expansion. To compute we need to work out the derivatives of p1 z qa .
Using the definition we have (for (a R N))

q 1  eaLogp1 zq apLogp1


eaLogp1 z qq1  p1 z qa  ap1 z qa
1 .
z a
1 z
Notice that since by induction we have

dk
aLogp1 q
e z
 apa
1q    pa
k 1qp1 z qa
k .
dz k
Therefore we obtain the Taylor series (centred at 0)
8̧ apa
1q    pa
k 1q
zk .

k 0
k!

Notice that the radius of convergence of this series is 1, as we know there are issues for z P p
8,
1s.
The binomial coefficient, for integer values n and k is

n
k
 pn
n!kq!k!  npn
1q  k! pn
k 1q

and so extending the defition to n P C we obtain

8̧ apa
1q    pa
k 1q
8̧ 
p1 zq a
 k!
z k
 a k
k
z .

k 0 k 0 
We can obtain similar expansions centred at different points
8̧ apa
1q    pa
k 1q
p1 zq a
 p1 z0 qa
k pz
z0 qk ,

k 0
k!

which would naturally a radius of convergence R equal to the distance from the point z0 to the half line
tx ¤
1u, where we have made a brach cut for the Log function. You may ignore the issue of the radius
of convergence for this series for the exam.

The following result is also a direct consequence of Cauchy’s formula.

Theorem 7.38 (Liouville’s Theorem). Let f : C Ñ C be an analytic, bounded function. Then f is

constant.

ANALYSIS III 80
 CHAPTER 7. COMPLEX ANALYSIS

Proof. Assume that |f pz q| ¤ M for all z P C. Let a  b be two points in C. Choose R large enough so
that 2 maxt|a|, |b|u R. That means that if we consider w P B BR p0q, that is |w|  R then

|w
a| ¡ R2 |w
b| ¡ R2 .
Since f is analytic in C we can use Cauchy’s formula to compute f paq and f pbq using B BR p0q as the curve
γ (of course positively oriented!). We have
 
f pw q f pw q
f paq
f pbq  dw

1 1
w
a w
b
dw
2πi BBR p0q 2πi BBR p0q
  
 2πi
1
f pw q
1

1
b
 a2πi f pw q
w
a w
b p
aqpw
bq
dw dw.
BBR p0q B BR p0q w
Therefore 
|a
b| M
|f paq
f pbq| ¤ 2π R2{4 1|dw| 
|a
b|4M ,
BBr p0q R


as BBR p0 1|dw| is just the length of the curve, which equals 2πR. Notice that since R is arbitrary (provided
that it is big enough, as indicated above) we can send R to infinity, showing that |f paq
f pbq|  0 for
any a and b in C, therefore proving that the function is constant.

A fundamental consequence of Liouville’s Theorem is Fundamental Theorem of Algebra.

Theorem 7.39 (Fundamental Theorem of Algebra). Every non-constant polynomial p on C has a root,
that is, there exists a P C such that ppaq  0.

Proof. We will prove the result by contradiction. Assume that |ppz q|  0 for ever z P C. Define f : C Ñ C
by f pz q  pp1z q . Now, since p does not vanish, the function f is analytic in all of C, since it is the
composition of two holomorphic functions (1{z is holomophic outside the origin).
°
Notice that if we assume ppz q  nk0 ck z k , with cn  0 (n ¡ 0), then at infinity the polynomial
behaves like cn z n , as that is the highest power. That means |ppz q| goes to infinity as z goes to infinity,
and satisfies |ppz q| ¡ 1 for all |z | ¡ R for some R ¡ 0. As a result the function f pz q  pp1z q is bounded
in C. It is less than 1 for all |z | ¡ R based on our analysis of p, and it is bounded on the compact set
|z| ¤ R since it is continuous.
Liouville’s Theorem implies that f is in fact constant, which would force p to be constant, which is a
contradiction.

Theorem 7.40. Let fn : Ω Ñ C be a sequence of analytic functions on an open set Ω. If fn converges

uniformly to f , then f is analytic.

Recall that for a function to be analytic at one point we require that the function be differentiable in
a neighbourhood of the point, and therefore the assumption on Ω being open is natural. Being analytic is
a local property, and requiring that the uniform convergence holds only on compact sets would suffice.

Proof. Let z P Ω. Choose r ¡ 0 sufficiently small so that Br pz q € Ω. Since fn is analytic in Ω we can

apply Cauchy’s formula to obtain

fn pwq
fn pz q 
1
w
z
dw.
2πi BBr pzq
Taking limits as n goes to infinity, and assuming that we can move the limit inside the integral we would
obtain 
f pw q
f pz q 
1
2πi BBr pz q w
z
dw.

ANALYSIS III 81
 CHAPTER 7. COMPLEX ANALYSIS

We have seen before that this implies that f is differentiable (in fact infinitely differentiable) and obtained
an expression for its derivative (see Theorem 7.35). So the only thing left is to justify moving the limit
inside the integral. Notice that this is really a one dimensional integral and we can apply the results learnt
earlier in the year. Taking γ ptq  z reit for t P r0, 2π q, we have γ 1 ptq  ireit and so
  
fn pwq 2π
fn pz reit q it 2π

w
z
dw  ire dt  i fn pz reit qdt. (7.25)
BBr pzq 0 reit 0

For fixed z, as a function of t we have that fn pz reit q converges uniformly to f pz reit q and applying
Theorem 2.16 we can move the limit inside the integral, obtaining
  
fn pwq 2π 2π

n
lim
Ñ8 BBr pzq w
z
dw  nlim
Ñ8 i fn pz re qdt  i
it
f pz reit qdt.
0 0

Notice that we have (reading the expression (7.25) backwards, now for f instead of for fn )
 
2π
f pw q
f pz re qdt 
it
w
z
i dw,
0 BBr pzq
obtaining the result.

7.3.3 Applications of Cauchy’s formula to evaluate integrals in R

We present various examples that illustrate a more general theory (of residues) for computing integrals of
functions over R.
Consider for example  8
1
dx.

8 1 x2
The idea is to consider the contours γ1 and γ2 in Figure 7.4.

Figure 7.4: Contours

γ1 is formed by the segment joining
R and R, together with the half circle or radius R. The contour γ2
is a circle centred at i and of radius r 1. To understand the choice of curves, notice that we can rewrite
the integral as  8
1

8 px
iqpx iq
dx.

Notice that in the region enclosed by γ2 the function (the integrand extended to a function on C)

f pz q :
1
pz
iqpz iq
ANALYSIS III 82
 CHAPTER 7. COMPLEX ANALYSIS

is analytic except for at z  i. By the deformation of contours Theorem we know that

 
f pz qdz  f pz qdz,
γ1 γ2

since the two curves have the same orientation. Now

  R 
f pz qdz  f pz qdz f pz qdz.
γ1
R arc
We parametrise the arc by Reit for t P r0, π q. We have
   π
f pz qdz  
1 1
2
dz Rieit dt.
arc arc 1 z 0 1 R2 e2it
Therefore   
  π

 pq
f z dz 

1 ¤ R2
R
dt  π
R2
R

1.
arc 0


As R tends to infinity the arc f pz qdz equals zero. Therefore
 8  
f pz qdz  f pz qdz  f pz qdz.

8 γ1 γ2

Now  
f pz qdz  1 1
iz
i
dz.
γ2 BBr piq z
Recall that by Cauchy’s formula if g pz q is analytic in the interior of a positively oriented curve then

g pz q  2πigpaq.
1
z
a
dz
γ

Therefore, taking g pz q  1
z i we obtain

1
iz
i
1
dz  2πi 2i1  π,
BBr piq z
which yields  8
dx  π.
1

8 1 x2

As a second example, let’s compute  8 1

dx.
8 1 x4
Notice that the function
1
1 z4
has four singularities at the points

eπi{4 e3πi{4 e
πi{4 e
3πi{4 ,

and so if we choose a contour similar to the one above (an expanding semi-circle) there will be two
singularities in the interior. We obtain the picture 7.5.
Now γ1 is built out of the line joining
R and R, together with the semi-circle or radius R centred at
0. γ2 and γ3 correspond to circles centred at e3πi{4 and eπi{4 , oriented clock-wise (positively with respect
to both the blue-shaded and yellow-shaded regions). Notice that with those orientations we have
  

1
1
z 4
dz
1
1
z 4
dz
1
1
z4
dz  0.
γ1 γ2 γ3

ANALYSIS III 83
 CHAPTER 7. COMPLEX ANALYSIS

Figure 7.5: Contours

We start by considering the integral over γ1 . We have

  R 
1
dz  1
dz
1
dz.
γ1 1 z 4

R 1 z 4
arc 1 z4

We will show that the integral over the arc goes to zero as R goes to infinity. Indeed
  
 


1
dz
z4 
 ¤ R4
1

1 |dz |  4
πR
R
1
,
arc 1 arc


which goes to zero as R goes to infinity. Above we have used that arc |dz | length(arc) πR.
We now consider the integral over γ2 . We have (denoting by γ2
the anti-clockwise parametrisation of
the circle)  

1
1
z 4
dz 

p z
e
1
iπ {4 qpz
e3iπ {4 qpz
e
iπ {4 qpz
e
3iπ {4 q
dz
γ2 γ2

g pz q


2πigpe3iπ{4q,
γ2
pz
e
3iπ {4 q
dz

where the above defines g as

g pz q 
1
pz
eiπ{4qpz
e
iπ{4qpz
e
3iπ{4q ,
and we have used Cauchy’s formula as g is analytic inside the curve γ2
. Now

g pe3iπ{4 q   p
?2qp
?21 ?2iqp?2iq .

1
p {
e3iπ 4
eiπ 4{ qp e3iπ 4 {
e
iπ{4qpe3iπ{4
e
3iπ{4q
Now we consider the integral over γ3
 
1
dz 
pz

qpz
{ {
1
qpz
e
iπ{4qpz
e
3iπ{4q dz
γ3 1 z4 γ3
eiπ 4 e3iπ 4



pz
hpeziπq {4q dz 
2πihpeiπ{4q,
γ 3

where the above defines h as

hpz q 
1
pz
e3iπ{4qpz
e
iπ{4qpz
e
3iπ{4q ,
ANALYSIS III 84
 CHAPTER 7. COMPLEX ANALYSIS

and we have used Cauchy’s formula as h is analytic inside the curve γ3
. Now

hpeiπ{4 q   ?2p?2iqp1?2 ?2iq .

1
p {
eiπ 4
e3iπ 4{ qp eiπ 4 {
e
iπ{4qpeiπ{4
e
3iπ{4q
Since we have  8  
1
dz 
1
d
1
dz

8 1 z 4
γ2 1 z 4
γ3 1 z4
we obtain
 8 ?
dz  2πi ? ? ? ? 2πi ? ? ? ? 
1 1 1 π 2

8 1 z 4 p
2qp
2 2iqp 2iq 2p 2iqp 2 2iq 2
.

In addition to being able to integrate quotients involving polynomials, we can integration some trigono-
metric functions. For example  8
cosp3xq
dx.

8 4 x2
We can rewrite this integral as  8 e3iz
Re

8 pz
2iqpz 2iq
dz,

and we can actually drop the Re part as the imaginary part will be an odd integrand and it will vanish.
We consider the contours (notice they are both oriented counter-clockwise)

Figure 7.6: Contours

As before,  
e3iz e3iz
pz
2iqpz 2iq
dz
pz
2iqpz 2iq
dz.
γ1 γ2
Also   
R
e3iz e3iz
e3iz
pz
2iqpz 2iq
 2iq pz
2iqpz |dz|.
arc pz
2iqpz 2iq
dz dz
γ1
R
We consider first the integral over the arc (half circle of radius R). We have, for R ¡¡ 4
     
 e3iz  e3iz  e
3 Im z
 dz  ¤ |dz| ¤ |dz| ¤ R2πR
4 ÝÝÝÑ
arc pz
2iqpz q arc |z 4| 2
4
 0,
2i 2
arc R RÑ8

where we have used that along the arc, Im z ¥ 0 and so e
3 Im z ¤ 1. Now for γ2 (remember it is oriented
anti-clockwise)  
e3iz g pz q
dz  dz  2πig p2iq,
γ2 pz
2iqpz 2iq γ2 z
2i

ANALYSIS III 85
 CHAPTER 7. COMPLEX ANALYSIS

where
e3iz
g pz q 
dz
z 2i
and we have used Cauchy’s formula since g is analytic inside γ2 . We have

e
6
g p2iq  .
4i
Therefore
 8  
2πig p2iq  e
6 .
e3iz e3iz e3iz
pz
2iqpz 2iq
dz pz
2iqpz 2iq
 pz
2iqpz 2iq
 π

8 γ1 γ2 2

We use a similar approach for

 8 x sin x 
1 8 zeiz
dx  dz.

8 1 x2 i
8 1 z 2

Notice the real part of the integral vanishes as the integrand is odd (hence dividing the i). We consider
the following contours of integration We consider the contours (notice they are both oriented counter-
clockwise) By Cauchy’s Theorem since the integrand is analytic in the region between the curves we

Figure 7.7: Contours

have  
zeiz zeiz
1 z2
dz  1 z2
dz.
γ1 γ2

Now for γ2  
zeiz hpz q
pz
iqpz iq

dz
pz
iq dz  2πihpiq,
γ2 γ2

for
zeiz
hpz q  ,
z i
and so 
zeiz
1
pz
iqpz iq
dz
2πi
ie
2i
 π
e
i.
γ2

We now consider the integral over γ1 . We look at the integral along the arc.
    
   π
Reit e
R sin t Ri cos t  π
e
R sin t dt
zeiz  R2
¤
 it 
R2
1
 2
dz   Rie dt
arc 1 z 0 1 R2 e2it 0

ANALYSIS III 86
 CHAPTER 7. COMPLEX ANALYSIS


π{2
 π  π 2{  {
π 2
¤2 e
R sin t dt  4 e
R sin t dt ¤ 4 e
R2t{π dt 
4 e
R2t{π 
π
0 0 0 2R 0


R 2π  
e
R
1 ÝÝÝÑ 0.
RÑ8
Therefore   
8 x sin x zeiz zeiz
dx  1
i γ1 pz
iqpz iq
dz 1
pz
iqpz iq
dz  πe .

8 1 x2 i γ2

As the final example we consider an integrand that has a singularity along the natural path of integration
 8 sin x
dx.

8 x
If we complexify the integrand, we are left with
 8 eiz
1
dz,
i
8 z
since the real part of the integrand is odd. Since the denominator vanishes for a point in the real axis we
need to modify the contours we consider.
The contour is formed of 4 curves, and since the integrand is analytic we know that
   
eiz eiz eiz eiz
z
dz
z
dz
z
dz
z
dz  0.
γ1 γ2 γ3 γ4

Now for γ1

Figure 7.8: Contours

 
eiz π
eiR cos t
R sin t
z
dz  Reit
iReit dt
γ1 0

and so   
 eiz  π

 dz  ¤ eR sin t dt ÝÝÝÑ 0
γ1 z 0 R Ñ8
as we have seen before. As for γ3 , since it is oriented clock-wise
   π it  π
eiε cos t e
ε sin t dt ÝÝÝÑ
πi.
eiz eiz eiεe
dz 
dz 
iεeit dt 
i
Ñ0
γ3 z γ3
z 0 εeit 0 ε

Therefore  
8 sin x 1 8 eiz
dx  dz

8 x i
8 z
      
eiz eiz eiz 1 eiz
 εlim lim
Ñ0 RÑ8 i
1
z
dz
z
dz  εlim lim
Ñ0 RÑ8

1i z
dz

i z
dz  π.
γ2 γ4 γ1 γ3

ANALYSIS III 87
Bibliography

[1] Rudin, W. Principles of mathematical analysis. Third edition. International Series in Pure and Applied
Mathematics. McGraw-Hill Book Co., 1976.

[2] Lebl, J. Basic Analysis: Introduction to Real Analysis. Available to download at

[Link]

[3] Ross, K. Elementary Analysis: the theory of calculus. Springer 2013

[4] Abbott, S. Understanding analysis. Springer, 2001.

[5] Gelbaum, B.R., & Olmsted, J.M.H. Counterexamples in Analysis, Holden Day Inc, 1964.

[6] Pugh, Charles C. Real Mathematical Analysis, Springer, 2015.

88