Sasane Amol Sasane Sara Maad A Friendly Approach To Complex

This page intentionally left blank
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
Cover image: The chameleons are conformal images of one another.
A FRIENDLY APPROACH TO COMPLEX ANALYSIS

Second Edition
Copyright © 2024 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.
ISBN 978-981-127-280-6 (hardcover)

ISBN 978-981-127-410-7 (paperback)
ISBN 978-981-127-281-3 (ebook for institutions)
ISBN 978-981-127-282-0 (ebook for individuals)
For any available supplementary material, please visit

https://www.worldscientific.com/worldscibooks/10.1142/13321#t=suppl
Desk Editor: Tan Rok Ting
Printed in Singapore
Contents
Overview 1
What is complex analysis? 1
Why study complex analysis? 3
Chapter 1. Complex numbers and their geometry 7

§1.1. The field of complex numbers 7
§1.2. Geometric representation of complex numbers 10
§1.3. Topology of C 15
§1.4. The exponential function and kith 20
Chapter 2. Complex differentiability 29
§2.1. Complex differentiability 30
§2.2. Cauchy-Riemann equations 34
§2.3. Geometric meaning of the complex derivative 41
Chapter 3. Cauchy Integral Theorem 47

§3.1. Definition of the contour integral 47
§3.2. Properties of the contour integral 52
§3.3. Fundamental Theorem of Contour Integration 55
§3.4. The Cauchy Integral Theorem 57
§3.5. Existence of a primitive 63
§3.6. The Cauchy Integral Formula 65
§3.7. Holomorphic functions are infinitely complex differentiable 68
§3.8. Liouville’s Theorem, Fundamental Theorem of Algebra 69
§3.9. Morera’s Theorem 71
§3.10. Appendix 72
iii
iv Contents
Chapter 4. Taylor and Laurent series 77

§4.1. Series 78
§4.2. Power series 79
§4.3. Taylor series 85
§4.4. Classification of zeroes 88
§4.5. The Identity Theorem 91
§4.6. The Maximum Modulus Theorem 93
§4.7. Laurent series 94
§4.8. Classification of singularities 99
§4.9. Residue Theorem 106
Chapter 5. Harmonic functions 113
§5.1. What is a harmonic function? 113
§5.2. Link between harmonic and holomorphic functions 114
§5.3. Consequences of the two way traffic 117
Solutions 123
Solutions to the exercises from the Introduction 123
Solutions to the exercises from Chapter 1 124
Some real analysis background 201
Notes 207
Bibliography 209
Index 211
Overview
We give an overview of what complex analysis is about and why it is important. For
the discussion in the overview, we assume some familiarity with complex numbers,
but in Chapter 1, we will start from scratch again. So the reader should not worry
about being lost in this overview!
What is complex analysis?

In real analysis, calculus is studied rigorously in the setting of real numbers. Thus
concepts such as the convergence of real sequences, continuity of real-valued func-
tions, differentiation and integration are studied. Based on this, one might guess
that complex analysis is all about studying similar concepts in the setting of com-
plex numbers. This is partly true, but it turns out that up to the point of studying
differentiation, there are no new features in complex analysis as compared to the
real analysis counterparts. But the subject of complex analysis departs radically
from real analysis when one studies differentiation. Thus, complex analysis is not
merely about doing analysis in the setting of complex numbers, but rather, much
more specialised:
Complex analysis is the study of ‘complex differentiable’ functions.
In real analysis, a function f : R Ñ R is differentiable at x0 P R if there exists a
real number L such that
f pxq ´ f px0 q
lim “ L,
xÑx0 x ´ x0
that is, for every ǫ ą 0, there is a δ ą 0 such that whenever 0 ă |x ´ x0 | ă δ,
ˇ f pxq ´ f px q ˇ
ˇ 0 ˇ
ˇ ´ Lˇ ă ǫ.
x ´ x0
In other words, given any distance ǫ, we can make the difference quotient f pxq´f
x´x0
px0 q
lie within a distance of ǫ from the real number L for all x sufficiently close to, but
distinct from, x0 .
1
2 Overview
In the same way, f : C Ñ C is complex differentiable at z0 P C if there exists a

complex number L such that
f pzq ´ f pz0 q
lim “ L,
z ´ z0
zÑz0
that is, for every ǫ ą 0, there is a δ ą 0 such that whenever 0 ă |z ´ z0 | ă δ,

ˇ f pzq ´ f pz q ˇ
ˇ 0 ˇ
ˇ ´ Lˇ ă ǫ.
z ´ z0
The only change from the previous definition is that now the distances are measured
with the complex absolute value, and instead of multiplication/division of real
numbers, we now have multiplication/division of complex numbers — so this is a
straightforward looking generalisation.
But we will see that this innocent looking generalisation is actually quite deep,
and the class of complex differentiable functions looks radically different from real
differentiable functions. Here is an instance of this.
" 2
x if x ě 0,
Example 0.1. Let f : R Ñ R be given by f pxq “
´x2 if x ă 0.
f1
f
0 0
" *
2x if x ě 0,
Then f is differentiable everywhere, and f 1 pxq “ “ 2|x|.
´2x if x ă 0.
Indeed, the above expressions for f 1 pxq are immediate when x ‰ 0, and f 1 p0q “ 0
can be seen as follows. For x ‰ 0,
ˇ f pxq ´ f p0q ˇ ˇ f pxq ˇ |x|2
ˇ ˇ ˇ ˇ
ˇ ´ 0ˇ “ ˇ ˇ“ “ |x| “ |x ´ 0|,
x´0 x |x|
and so given ǫ ą 0, taking δ “ ǫ (ą 0), we have that whenever 0 ă |x ´ 0| ă δ,
ˇ f pxq ´ f p0q ˇ
ˇ ˇ
ˇ ´ 0 ˇ “ |x ´ 0| ă δ “ ǫ.
x´0
However, it can be shown that f 1 is not differentiable at 0; see Exercise 0.1. This
is visually obvious since the graph of f 1 has a ‘corner’ at x “ 0.
Summarising, we gave an example1 of an f : R Ñ R, which is differentiable
everywhere in R, but whose derivative f 1 is not differentiable on R.
1Here f 1 only failed to be differentiable at one point (namely, 0), but it can get much worse. There exist
differentiable f : R Ñ R for which f 1 is differentiable nowhere! For example we can
ş start with a continuous
function g : R Ñ R which is nowhere differentiable, and define f by f pxq “ 0x gpξqdξ, so that f 1 “ g
everywhere. Examples of such g can be found for example in [7, Part I,Chap.3, §8].
Exercise 0.1. Let g “ f 1 : R Ñ R be given by gpxq “ 2|x| for all x P R. Prove that g is
not differentiable at 0.
In contrast, we will later learn that if f : C Ñ C is a complex differentiable

function in C, then it is infinitely many times complex differentiable! In particular,
its complex derivative f 1 is also complex differentiable in C. Clearly this is an
unexpected result if all we are used to is real analysis. We will later learn that the
reason this miracle takes place in complex analysis is that complex differentiability
imposes some ‘rigidity’ on the function which enables this phenomenon to occur.
We will also see that this rigidity is a consequence of the special geometric meaning
of multiplication of complex numbers2.
Why study complex analysis?

Although it might seem that complex analysis is just an exotic generalisation of
real analysis, this is not so. Complex analysis is fundamental in all of mathematics.
In fact real analysis is actually inseparable with complex analysis, as we shall see,
and complex analysis plays an important role in the applied sciences as well. Here
is a list of a few reasons to study complex analysis:
Partial Differential Equations (PDE). Associated with a function f : C Ñ C,

there are two real-valued functions u, v : R2 Ñ R, namely the real and imaginary
parts of f : for px, yq P R2 , upx, yq :“ Repf px, yqq and vpx, yq :“ Impf px, yqq.
f px, yq
f vpx, yq
px, yq
upx, yq
It turns out that if f is complex differentiable, then its real and imaginary parts
u, v satisfy an important PDE, called the Laplace equation:
B2 u B2 u
∆u :“ ` 2 “ 0.
Bx2 By
Similarly ∆v “ 0 in R2 as well. The Laplace equation arises in many applied
problems, e.g., in electrostatics, steady-state heat conduction, incompressible fluid
flow, Brownian motion, etc.
2Given a function f : C Ñ C, and a point z P C, we could just think of real differentiability of f : R2 Ñ R2
0
at z0 “ px0 , y0 q P R2 ; see e.g. [1, §8.18]. But this is not the same as complex differentiability at z0 . Instead,
we use the multiplicative structure of C to define the complex derivative, as limit of difference quotients,
just as in the case of functions from R to R. This turns out to be a much stronger notion: We shall see
(p.45) that only when the matrix of the (real analysis) derivative at px0 , y0 q is of a special type (namely
corresponding to a ‘rotation’ followed by a ‘scaling/dilation’) will f be complex differentiable at z0 !
4 Overview
Real analysis. Using complex ş8 analysis, we can calculate some awkward integrals
1
in real analysis, for example ´8 1`x 4 dx. The problem is set in the reals, but one
can solve it using complex analysis.
Moreover, sometimes complex analysis helps to clarify some matters in real
1
analysis. Here is an example. Let f pxq :“ 1´x 2 , x P Rzt´1, 1u. Then f has a
‘singularity’ at x “ ˘1, i.e. it is not defined there. It is, however defined in the
interval p´1, 1q. The geometric series 1`x2 `x4 `x6 `. . . converges for |x2 | ă 1, or
1
equivalently for |x| ă 1, and 1 ` x2 ` x4 ` x6 ` ¨ ¨ ¨ “ 1´x 2 “ f pxq for all x P p´1, 1q.
From the formula for f , it is not a surprise that the power series expansion of f is
valid only for x P p´1, 1q, since f itself has singularities at x “ 1 and at x “ ´1.
1 2 4 6
Let gpxq :“ 1`x 2 , x P R. The geometric series 1 ´ x ` x ´ x ` ´ . . . converges for
1
| ´ x2 | ă 1, or equivalently for |x| ă 1, and 1 ´ x2 ` x4 ´ x6 ` ´ ¨ ¨ ¨ “ 1`x 2 “ gpxq
for all x P p´1, 1q. So the power series expansion of g is again valid only for
x P p´1, 1q, despite there being no obvious reason from the formula for g for the
series to break down at x “ ´1 and x “ 1. The mystery will be resolved later:
1 1
Consider the complex extensions of the functions, f pzq “ 1´z 2 and gpzq “ 1`z 2
(whose restriction to R are the given functions f and g, respectively). Then g has
singularities at z “ ˘i, and we will see that the power series expansion is valid in
the biggest disk centred at z “ 0 which does not contain any singularity of g.
i
f g
´1 0 1 0
í
Applications. Many tools used for solving problems in applications, such as the
Fourier/Laplace/z-transform, rely on complex function theory. These tools in turn
are useful for example to solve differential equations which arise from applications.
Thus complex analysis plays an important role e.g. in physics and engineering.
Analytic number theory. Perhaps surprisingly, many questions about the nat-
ural numbers can be answered using complex analytic tools. For example, consider
the Prime Number Theorem, which gives an asymptotic estimate on πpnq for large
n, where πpnq :“ number of primes ď n:
πpnq
lim n “ 1.
nÑ8 log n
A proof of the Prime Number Theorem can be given using complex analytic com-
putations with a certain complex differentiable function called the Riemann zeta
function ζ. Associated with ζ is also a famous unsolved problem in analytic num-
ber theory, namely the Riemann Hypothesis, saying that all the ‘nontrivial’ zeros
of the Riemann zeta function lie on the line Re s “ 12 in the complex plane. We
will meet the Riemann zeta function in Exercise 4.6.
What will we learn in complex analysis. The central object of study in this
course is: Holomorphic functions in a domain, that is, complex differentiable func-
tions f : D Ñ C, where D is a ‘domain’ (the precise meaning of what we mean by a
domain will be given in Subsection 1.3.4). The bulk of the course is in Chapters 2,
3, 4, where the following three ‘lanterns’ shed light on holomorphic functions in a
domain:
Cauchy-Riemann Cauchy Integral

equations Theorem
Holomorphic
functions
in a domain
Taylor and Laurent series
The core content of the book can be summarised in the following Main Theorem3:
Main Theorem. Let D be an open path connected set and let f : D Ñ C.
Then the following are equivalent:
(1) f is complex differentiable in D.
(2) f is infinitely many times complex differentiable in D.
(3) u :“ Re f, v :“ Im f are real differentiable in D, and moreover, we have the
Cauchy-Riemann equations Bu Bv Bu Bv
Bx “ By , and By “ ´ Bx in D.
(4) For each simply connected domain S Ă D, there exists an F : S Ñ C, complex
differentiable in S, such that F 1 “ f in S.
(5) f is continuous on D, and for
ş all piecewise smooth closed paths γ in each simply
connected domain S Ă D, γ f pzqdz “ 0.
(6) If z0 P D and r ą 0 is such that Dpz0 , rq :“ tz P C : |z ´ z0 | ă ru Ă D, then
8 f pnq pz q
ř
f pzq “ n!
0
pz ´ z0 qn for all z P Dpz0 , rq.
n“0
Complex analysis is not complex analysis! It is not very complicated, and

there is not much messy analysis. Once core properties of complex differentiable
functions are established, it is ‘softer’ than real analysis, and there are fewer cum-
bersome ǫ-δ estimates. The Main Theorem above tells us that the subject is radi-
cally different from real analysis. Indeed, a real-valued differentiable function on an
open interval pa, bq need not have a continuous derivative. In contrast, a complex
3Do not worry about the unfamiliar terms/notation here. That is what we will learn, besides the proof!
6 Overview
differentiable function on an open subset of C is infinitely many times complex

differentiable! This happens because the geometric meaning of complex multipli-
cation implies that complex differentiable functions behave in a controlled manner
locally infinitesimally. This controlled behaviour makes these functions rigid and
we will see this in Section 2.3. Nevertheless there are enough of them to make the
subject nontrivial and interesting!
Prerequisites. We assume familiarity with multivariable calculus and linear al-
gebra, e.g. at the level of [1, Chapters 1,2,8]. Even if the student is unfamiliar
with proofs, we hope that at least the definitions and statements of results are
familiar. For example, for a function f : Rn Ñ Rm , the reader knows the meaning
of continuity at a point x P Rn , and of real differentiability and the real analysis
derivative. Also, we assume knowledge of the fact that if a function has continuous
partial derivatives in an open subset U Ă Rn , then it is real differentiable in U ,
and the fact if a function has continuous partial derivatives up to order 2, then the
mixed partial derivative with respect to x and y can be taken in any order. At the
end of the book, an appendix is included, where we have summarised some of the
relevant results from real analysis.
Title, first edition versus the second edition. The title of the book is meant
to indicate that we aim to cover the bare bones of the subject with minimal pre-
requisites. It does not mean that it is merely a book with ‘drill’ kind of material.
In fact it contains many challenging exercises. As the solutions to all the exercises
are included in the book, we hope that the book will be useful for self-study.
The first edition was a much revised version of lecture notes written when
the second author gave a cour for third year students of the BSc programme in
Mathematics and/with Economics. In the second edition we have tried to correct
all the mistakes we could find, and also added many exercises, some of which cover
auxiliary topics that were omitted in the first edition. We have added an appendix
to Chapter 3 containing a detailed proof of the Cauchy Integral Theorem, and a
chapter called ‘Some real analysis background’, recalling the needed real analysis.
Acknowledgements. This book relies heavily on some of the sources mentioned
in the bibliography. At some instances detailed references are given in a chapter
called ‘Notes’, but no claim to originality is made in case there is a missing reference.
Thanks are due to Raul Hindov, Raymond Mortini, Adam Ostaszewski, Rudolf
Rupp and Gabriel Winberg for useful comments. Finally, it is a pleasure to thank
the World Scientific team, especially executive editor Rochelle Kronzek for her
support, enthusiasm and prompt help.
Sara Maad Sasane and Amol Sasane
Chapter 1
Complex numbers and their geometry
In this chapter, we set the stage for doing complex analysis. First, we introduce the
set of complex numbers, and their arithmetic, making C into a field, ‘extending’
the usual field of real numbers.
Points in C can be depicted in the plane R2 , and we will see that the arithmetic
in C has geometric meaning in the plane. This correspondence between C and
points in the plane also allows one to endow C with the usual Euclidean topology
of the plane.
We will study a fundamental function in complex analysis, namely the exponen-
tial function (and some elementary functions related to the exponential function,
namely trigonometric functions and the logarithm). Thus besides polynomial and
rational functions, these will serve as important examples of complex differentiable
functions in certain domains.
1.1. The field of complex numbers

By definition, a complex ? number is an ordered pair of real numbers. For example,
p1, 0q, p0, 1q, p0, 0q, p´ 34 , 2q are all complex numbers. The set R ˆ R “ R2 of all
complex numbers is denoted by C. Thus C “ tz “ px, yq : x P R and y P Ru. For
z “ px, yq P C, where x, y P R, the real number x is called the real part of z, and y
is called the imaginary part of z. The operations of addition ‘`’ and multiplication
‘ ¨ ’ on C are defined by:
px1 , y1 q`px2 , y2 q “ px1 ` x2 , y1 ` y2 q,
px1 , y1 q ¨ px2 , y2 q “ px1 x2 ´ y1 y2 , x1 y2 ` x2 y1 q,
for complex numbers px1 , y1 q, px2 , y2 q. With these operations, C is a field, that is,
(F1) pC, `q is an ‘Abelian group’,
(F2) pCzt0u, ¨q is an Abelian group, and
(F3) the distributive law holds: for a, b, c P C, pa ` bq ¨ c “ a ¨ c ` b ¨ c.
7
8 1. Complex numbers and their geometry
(F1) means that ` : C ˆ C Ñ C is associative and commutative, that there exists

an ‘identity element’ p0, 0q, such that px, yq ` p0, 0q “ px, yq “ p0, 0q ` px, yq for all
px, yq P C, and that each element px, yq P C has a corresponding ‘additive inverse’
p´x, ýq: px, yq ` p´x, ýq “ p0, 0q “ p´x, ýq ` px, yq.
Similarly, in (F2), the multiplicative identity is p1, 0q, and the multiplicative
x ý
inverse of a complex number px, yq P Cztp0, 0qu is given by p x2 `y 2 , x2 `y 2 q.
x ý
Exercise 1.1. Check that p x2 `y 2 , x2 `y 2 q is indeed the inverse of px, yq P Cztp0, 0qu.
Proposition 1.1. pC, `, ¨q is a field.
R is ‘contained’ in C. We can embed R inside C, and view R as a ‘subfield’ of C,

that is, the map x ÞÑ px, 0q sending the real number x to the complex number px, 0q
is an injective field homomorphism. This means that the operations of addition
and multiplication are preserved by this map, and distinct real numbers are sent
to distinct complex numbers.
R C
x Þ Ñ px, 0q
x1 ` x2 ÞÑ px1 ` x2 , 0q “ px1 , 0q ` px2 , 0q
x1 ¨ x2 Þ Ñ px1 ¨ x2 , 0q “ px1 , 0q ¨ px2 , 0q
1 ÞÑ p1, 0q
0 ÞÑ p0, 0q
So we can view real numbers as?if they are complex numbers, via this identification.
?
For example, the real number 2 can be viewed as the complex number p 2, 0q. If
this makes one uneasy, one should note that we have been doing such identifications
right from elementary school, where for instance, we identified integers with rational
numbers, for example, Z Q 3 “ 31 P Q, and we didn’t lose sleep over it!
An advantage of working with C is that while in R there is no solution x P R to
the equation x2 ` 1 “ 0, now with complex numbers we have with x :“ p0, 1q P C
that x2 ` 1 “ p0, 1q ¨ p0, 1q ` p1, 0q “ p´1, 0q ` p1, 0q “ p0, 0q “ 0, where we have
made the usual identification of the real numbers 1 and 0 with their corresponding
complex numbers p1, 0q and p0, 0q. If we give a special symbol, say i, to the number
p0, 1q, then the above says that i2 ` 1 “ 0. Henceforth, for the complex number
px, yq, where x, y are real, we write x ` yi, since
px, yq “ lo
px, on ` lo
omo0q py,
omo on ¨ lo
0q p0,
omo on “ x ` yi.
1q
”x ”y ”i
As complex multiplication is commutative, yi “ iy, and so x ` yi “ x ` iy.

For z P Czt0u, we will denote its inverse z ´1 by 1z , and if w P C, then w
z :“ w 1z .
Exercise 1.2. Let θ P p´ π2 , π2 q. Express 1ì tan θ
1í tan θ in the form x ` yi, where x, y P R.
1.1. The field of complex numbers 9
Historical development of complex numbers. Contrary to popular belief,

historically, it wasn’t the need for solving quadratic equations, but rather cubic
equations, that led mathematicians to take complex numbers seriously. The gist
of this is as follows. Around the 16th century, one viewed solving equations like
x2 ` bx ` c “ 0 as the geometric problem of finding the intersection point of the
parabola y “ x2 with the line y “ ´bx ´ c. Based on this geometric interpretation,
it was easy to dismiss the lack of solvability in reals of a quadratic such as x2 `1 “ 0,
since that just reflected the geometric fact that parabola y “ x2 did not meet the
line y “ ´1. See the picture on the left below.
0 0
´1
Meanwhile, Cardano (1501-1576) gave a formula for solving the cubic equation
x3 “ 3px ` 2q, namely,
b a b a
3 3
x“ q` q 2 ´ p3 ` q´ q 2 ´ p3 .
For example, one can check that for
? the?equation x3 “ 6x`6, with p “ 2 and q “ 3,
3 3
this yields one solution to be x “ 4` 2. Using the Intermediate Value Theorem,
it can be shown that the cubic y “ x3 always intersects the line y “ 3px ` 2q. See
the picture on the right above. For the equation x3 “ 15x`4, when p “ 5 and q “ 2,
we have q 2 ´ p3 “ ´121 ă 0, and so Cardano’s formula fails with real arithmetic.
But we do have a real root, namely x “ 4: 43 “ 64 “ 60 ` 4 “ 15 ¨ 4 ` 4. Three
decades after the appearance of Cardano’s work, Bombelli suggested that with the
use of complex arithmetic, Cardano’s formula gives the desired real root. So we
? ? ?
may ask if x “ 3 2 ` 11i` 3 2 ´ 11i “ 4. As p2ìq3 “ 2`11i and p2íq3 “ 2´11i,
the above does work, with appropriate choices of cube roots. Thus Bombelli’s work
established that even for real problems, complex arithmetic can be relevant. From
then on, complex numbers entered mainstream mathematics.
Exercise 1.3. A field F is called ordered if there is a subset P Ă F, called the set of
positive elements of F, satisfying the following:
(P1) For all x, y P P , x ` y P P .
(P2) For all x, y P P , x ¨ y P P .
(P3) For each x P F, exactly one of the following is true: 1˝ x “ 0. 2˝ x P P . 3˝ ´x P P .
The field R of real numbers is ordered, since P :“ p0, 8q is a set of positive elements of R.
Given a set P of positive elements in an ordered field F, elements of F can be compared
via the ‘order relation’ ą in F, defined by setting y ą x for x, y P F if y ´ x P P .
Show that C is not an ordered field. Hint: Consider x :“ i, and first look at x ¨ x.
1.2. Geometric representation of complex numbers

Since C “ R2 , we can identify complex numbers with points in the plane.
y px, yq
p0, 0q x
?
Exercise 1.4. Locate 0, 1, ´ 32 , i, ´ 2i, cos π3 ` i sin π3 in the complex plane.
So we can identify C as a set with the set of points in the plane R2 . Do the field
operations in C have some geometric meaning in the plane? We’ll now see that
addition in C corresponds to ‘vector addition’ in the plane, and multiplication in
C too has a special geometric meaning in the plane, as explained below.
Geometric meaning of complex addition. With the identification of complex
numbers with points in the plane, complex addition corresponds to addition of
vectors in R2 . By addition of vectors, we mean the usual way of combining two
vectors, that is, by completing the parallelogram formed by the line segments
joining p0, 0q to each of the two complex numbers as sides, and then taking the
endpoint of the diagonal from p0, 0q as the sum of the two given complex numbers.
z1 `z2 px1 `x2 , ˚ q p ˚ , y1 `y2 q

z2 x2
x1
y2
z1 y1
0
The middle picture shows that addition of z1 and z2 as vectors in the plane yields
the correct x-coordinate of their sum as complex numbers by looking at the two
congruent shaded triangles. Similarly, from the rightmost picture, we see that the
y-coordinate is also correct.
Geometric meaning of complex multiplication. We will now see the spe-

cial geometric meaning of complex multiplication. In order to do this, it will be
convenient to use polar coordinates. Let the point px, yq P R2 ztp0, 0qu have polar
coordinates r ą 0 and θ P p´π, πs. This means that the Euclidean distance of px, yq
to p0, 0q is r (ą 0), and the ray joining p0, 0q to px, yq makes an angle (in the coun-
terclockwise fashion) θ with the positive real axis (the x-axis). If px, yq “ p0, 0q,
then we set r “ 0 and θ “ 0.
1.2. Geometric representation of complex numbers 11
y px, yq
θ
0 x
In the right-angled triangle shown above, x “ r cos θ, y “ r sin θ. Using polar

coordinates, we have x ` yi “ r cos θ ` pr sin θqi “ rpcos θ ` i sin θq. Now we give
the geometric interpretation of complex multiplication. For two complex numbers
expressed in polar coordinates as z1 “ r1 pcos θ1 ` i sin θ1 q, z2 “ r2 pcos θ2 ` i sin θ2 q,
z1 ¨ z2 “ r1 pcos θ1 ` i sin θ1 q ¨ r2 pcos θ2 ` i sin θ2 q
“ r1 r2p pcos θ1 q cos θ2 ´ psin θ1 q sin θ2 ` ippcos θ1 q sin θ2 ` pcos θ2 q sin θ1 qqq
“ r1 r2 pcospθ1 `θ2 q ` i sinpθ1 `θ2 qq,
using the trigonometric identities for angle addition. Thus z1 ¨ z2 has polar coor-
dinates pr1 r2 , θ1 ` θ2 q. The angles z1 and z2 make with the positive real axis are
added in order to get the angle z1 ¨ z2 makes with the positive real axis, and the
distances to the origin are multiplied to get the distance of z1 ¨ z2 to the origin.
z1 z2
r1 r2 θ1 ` θ2 z2
z1
θ2
θ1
0
As a special case, consider multiplication by cos α ` i sin α, which is at a distance

of 1 from the origin. If z P C, then z ¨ pcos α ` i sin αq is obtained by rotating the
line joining 0 to z anticlockwise through an angle of α. In particular, multiplying
z by i “ 0 ` i ¨ 1 “ cos π2 ` i sin π2 produces a counterclockwise rotation of 90˝ .
zpcos α ` i sin αq iz
α
z z
θ
0 0
Exercise 1.5. In the complex plane, let the points corresponding to 0, z, w, zw, 1 be la-
belled O, P, Q, R, S. Show that the triangle ∆P OS is similar to ∆ROQ.
De Moivre’s formula and nth roots. de Moivre’s formula states that for all
n P N, and θ P R, pcos θ ` i sin θqn “ cospnθq ` i sinpnθq.
Exercise 1.6. Recover the identity cosp3θq “ 4pcos θq3 ´ 3 cos θ using de Moivre’s formula.
Exercise 1.7. Without expanding, express p1 ` iq10 in the form x ` iy, x, y P R.
Exercise 1.8. Considering p2 ` iqp3 ` iq, show that π
4 “ tan´1 1
2 ` tan´1 31 .
Exercise 1.9. Gaussian integers are complex numbers of the form m ` in, where m, n
are integers. Thus they are integral lattice points in the complex plane. Show that it is
impossible to draw an equilateral triangle such that all vertices are Gaussian
? integers.
Hint: Rotation of one of the sides should give the other. Recall that 3 R Q.
De Moivre’s formula gives a way of finding the nth roots of a complex number z,
i.e., complex numbers w satisfying wn “ z.
Proposition 1.2. Let z “ rpcos θ ` i sin θq, r ą 0, θ P r0, 2πq, n P N, w P C. Then
?
wn “ z if and only if w P t n rpcosp nθ ` k 2π θ 2π
n q ` i sinp n ` k n qq : k “ 0, 1, ¨ ¨ ¨ , n ´ 1u.
?
Proof. (‘If part’:) If w P t n rpcosp nθ `k 2π θ 2π
n qì sinp n `k n qq : k “ 0, ¨ ¨ ¨ , n´1u, then
n ? n θ 2π θ 2π
w “ p rq pcospnp n `k n qqì sinpnp n `k n qqq “ rpcospθ`2πkq`sinpθ`2πkqq “ z .
n
(‘Only if part’:) Let w “ ρpcos α ` i sin αq, where ρ ě 0 and α P R. As wn “ z, we

have ρn pcospnαq ` i sinpnαqq “ wn “ z “ rpcos θ ` i sin θq. Equating the distance
?
to the origin, ρn “ r. Hence ρ “ n r, as ρ, r ě 0. The angle wn makes with the
positive real axis is nα P t¨ ¨ ¨ , θ ´ 4π, θ ´ 2π, θ, θ ` 2π, θ ` 4π, θ ` 6π, ¨ ¨ ¨ u,
since the angle made by a nonzero z with the positive real axis is unique only up
to integral multiples of 2π, i.e., instead of θ, we can use θ ` 2πk for any k P Z.
z
θ θ`2π etc.
So α P t nθ ` 2π θ θ 2π θ 2π θ 2π
n k : k P Zu. For α P t n , n ` n , n ` 2 n , ¨ ¨ ¨ , n ` pn ´ 1q n u, we
get distinct w. (Indeed, for any integer k, the division algorithm for integers gives
q, r P Z such that k “ nq ` r and 0 ď r ă n, so that 2π 2π
n k “ n r ` 2πq.)
In particular, if z “ 1, we get the nth roots of unity, which are located at the vertices
of an n-sided regular polygon inscribed in a circle. For example, the following
picture depicts the six 6th roots of unity.
1.2. Geometric representation of complex numbers 13
Exercise 1.10. Find all w P C such that w4 “ ´1. Depict these in the complex plane.
Exercise 1.11. Find all z P C such that z 6 ´ z 3 ´ 2 “ 0.
Exercise 1.12. If a, b, c are real numbers such that a2 `b2 `c2 “ ab`bc`ca, then they must
be equal. Indeed, doubling both sides and rearranging gives pa´bq2 `pbćq2 `pcáq2 “ 0,
and since each summand is nonnegative, each must then be 0. Show that if a, b, c are
complex numbers such that a2 ` b2 ` c2 “ ab ` bc ` ca, then they must lie on the vertices
of an equilateral triangle in the complex plane. Explain the real case result in light of this.
Hint: Compute ppb ´ aqω`pb ´ cqqppb ´ aqω 2`pb ´ cqq, ω being a nonreal cube root of unity.
` ˘
Exercise 1.13. The Binomial Theorem says that if a, b P C, n P N, and nk :“ k!pn´kq! n!
` ˘
ř n k n´k
n
for k “ 0, 1, ¨ ¨ ¨ , n, then pa ` bqn “ k a b . A proof can be given using induction.
k“0
`3n˘ `3n˘ `3n˘ `3n˘ 23n `2p´1qn
Show that 0 ` 3 ` 6 ` ¨ ¨ ¨ ` 3n “ 3 . Hint: If ω denotes a nonreal cube
root of unity, find p1 ` 1q3n ` p1 ` ωq3n ` p1 ` ω 2 q3n .
Exercise 1.14. Show, using the geometry of complex numbers, that the line segments
joining the centres of opposite external squares described on sides of an arbitrary convex
quadrilateral are perpendicular and have equal lengths.
Exercise 1.15. Let p be a polynomial of degree d P N, i.e., there are c0 , ¨ ¨ ¨ , cd P C, with
cd ‰ 0, such that for all z P C, ppzq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d . Let z0 P C. Show that there
exists a polynomial q of degree d ´ 1 such that for all z P C, ppzq “ qpzqpz ´ z0 q ` ppz0 q. In
particular, if ppz0 q “ 0, then ppzq “ qpzqpz ´ z0 q. Conclude that if n P N, and z1 , ¨ ¨ ¨ , zn
are the n distinct nth roots of unity, then z n ´ 1 “ pz ´ z1 q ¨ ¨ ¨ pz ´ zn q.
Exercise 1.16. Show that if d P N Y t0u, and c0 , ¨ ¨ ¨ , cd P C are such that for all z P C,
c0 ` c1 z ` ¨ ¨ ¨ ` cd z d “ 0, then c0 “ ¨ ¨ ¨ “ cd “ 0.
Exercise 1.17 (Division Algorithm). Let h be a polynomial of degree d P N. Show that
for any polynomial p, there exist unique polynomials q (the quotient) and r (the remainder)
such that for all z P C, ppzq “ qpzqhpzq ` rpzq, where the degree of r is strictly less than d.
(We define the degree of the zero polynomial to be 0.)
Absolute value and complex conjugate. The absolute value or modulus a |z|
of the complex number z “ x ` iy, where x, y P R, is defined by |z| “ x2 ` y 2 .
By Pythagoras’s Theorem, |z| is the distance of the complex number z to 0 in the
complex plane. See the picture on the left below. By expressing z1 , z2 P C in terms
of polar coordinates, or by a direct calculation, it is clear that |z1 z2 | “ |z1 ||z2 |. If
z ‰ 0, then 1 “ |1| “ |z z1 | “ |z|| z1 | and so | 1z | “ |z|
1
, and for w P C, | wz | “ |w|
|z| .
Exercise 1.18. Verify |z1 z2 | “ |z1 ||z2 | by expressing z1 , z2 using Cartesian coordinates.
z “ x ` iy
iy
z “ x ` iy
x
|z|
|y|
0
|x| íy
z “ x ´ iy
The complex conjugate z of z “ x` iy where x, y P R, is z :“ x ´ iy. In the complex

plane, z is the reflection in the real axis of z. With this geometric interpretation, it
can be seen that for all z1 , z2 P C, z1 ` z2 “ z1 ` z2 and z1 z2 “ z1 z2 . The following
are easy to check: z “ z, zz “ |z|2 , Re z “ z`z z´z
2 , Im z “ 2i .
Exercise 1.19. Verify that the four equalities above.
Exercise 1.20. Prove that for all z P C, |z| “ |z|, |Re z| ď |z|, and |Im z| ď |z|. Give
geometric interpretations of each.
zá
Exercise 1.21. If a, z P C satisfy |a| ă 1 and |z| ď 1, then prove that | 1áz | ď 1.
Exercise 1.22. Let z, w P C be distinct complex numbers. Show that the ‘convex combi-
nation of z, w’, namely, L :“ tp1 ´ tqz ` tw : t P r0, 1su is the line segment joining z and w.
Hint: p1 ´ tqz ` tw “ z ` tpw ´ zq.
Exercise 1.23. Let p be a polynomial given by ppzq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d , where d P N,
c0 , c1 , . . . , cd P R and cd ‰ 0. Show that if w P C such that ppwq “ 0, then ppwq “ 0.
Exercise 1.24. Show that the area of the triangle formed by 0, a, b P C is | Impabq
2 |.
Exercise 1.25. On a visit to Cornell University, Richard Feynman gave a colloquium
on the history of the Feynman-Kac formula. After the talk, Kai Lai Chung entertained
Feynman to dinner, at which much of the discussion centred on the following result:
Theorem. For a triangle in the plane, if each vertex is joined to
the point one-third along the opposite side (measured anticlockwise),
then the area of the inner triangle formed by these lines is exactly
one-seventh of the area of the initial triangle.
A
B1
Y
C1 Z
X
B A1 C
Feynman could not believe that the ratio of the areas of the triangles was 71 since this had
nothing to do with the number three. He spent most of the evening trying to disprove it,
but finally proved it in the special case when the triangle was equilateral. Use the result
of the previous exercise to prove it in general.
„ 
1 z1 z1
Exercise 1.26. Prove for all z1 , z2 , z3 P C, i det 1 z2 z2 is real.
1 z3 z3
Exercise 1.27. Let z, w P Czt0u. Show that |z `w| “ | |w| |z|

|z| z ` |w| w|. What is the geometric
interpretation?
Exercise 1.28. Show that S 1 “ tz P C : |z| “ 1u is an Abelian group with respect to
complex multiplication. For any n P N, show that S 1 has a subgroup of order n. Does S 1
have an infinite subgroup which is countable?
Exercise 1.29. Let n P N be such that n ě 2, and z1 , ¨ ¨ ¨ , zn be the n distinct nth roots
of unity. Prove that z1 ` ¨ ¨ ¨ ` zn “ 0.
1.3. Topology of C 15
1.3. Topology of C
The concepts in ordinary calculus in the setting of R, like convergence of sequences,
or continuity and differentiability of functions, all rely on the notion of closeness of
points in R. For example, when we talk about the convergence of a real sequence
pan qnPN to its limit L P R, we mean that given any positive ǫ, there is a large
enough index N such that beyond that index, the corresponding terms an all have
a distance to L which is at most ǫ. This ‘distance of an to L’ is taken as |an ´ L|,
and this is the length of the line segment joining the numbers an and L on the real
number line. In order to do calculus with complex numbers, we need a notion of
distance dpz1 , z2 q between complex numbers z1 and z2 .
1.3.1. Metric on C. Since C is just R2 , we use the usual Euclidean distance in R2

a Thus, for complex numbers z1 “ x1 ` iy1 and z2 “ x2 ` iy2 , we
as the metric in C.
have dpz1 , z2 q “ px1 ´ x2 q2 ` py1 ´ y2 q2 “ |z1 ´ z2 |. By Pythagoras’s Theorem,
this is the length of the line segment joining the points px1 , y1 q, px2 , y2 q in R2 .
z2 “ px2 , y2 q
dpz1 , z2 q :“ |z1 ´ z2 |
|x1 ´ x2 |
z1 “ px1 , y1 q
|y1 ´ y2 |
Using the geometric meaning of addition of complex numbers, and the well-known
result from Euclidean geometry that the sum of the lengths of any two sides of a
triangle is at least as big as the length of the third side, we obtain the following
triangle inequality for the absolute value: |z1 ` z2 | ď |z1 | ` |z2 | for z1 , z2 P C. The
triangle inequality can also be verified analytically by using the Cauchy-Schwarz
inequality for x1 , x2 , y1 , y2 P R: px21 ` y12 qpx22 ` y22 q ě px1 x2 ` y1 y2 q2 , or noting
|z1 ` z2 |2 “ pz1 ` z2 qpz1 ` z2 q “ |z1 |2 ` |z2 |2 ` 2 Repz1 z2 q ď |z1 |2 ` |z2 |2 ` 2|z1 ||z2 |.
z1 `z2
z2 |
z2 |z2 |
`
|z 1
z1
|z1 |
0
Exercise 1.30. Show that for all z1 , z2 P C, |z1 ´ z2 | ě ||z1 | ´ |z2 ||.
Exercise 1.31. Let a P C, and r ą 0. Show that C :“ tz P C : |z|2 ´ 2 Repazq ` |a|2 “ r2 u
is a circle centred at a with radius r.
Exercise 1.32. Show that for all z1 , z2 P C, |z1 `z2 |2 ` |z1 ´z2 |2 “ 2p|z1 |2 `|z2 |2 q. What
is the geometric interpretation of this equality?
Exercise 1.33. Consider a circle C of radius r ą 0 and centre 0. For any point z on C,
determine |z ` r|2 ` |z ´ r|2 , and use the converse of the Pythagoras Theorem to conclude
that the angle subtended by the diameter at the point z is a right angle.
Exercise 1.34. Sketch the following sets in the complex plane:
(1) tz P C : |z ´ p1 ´ iq| “ 2u.
(2) tz P C : |z ´ p1 ´ iq| ă 2u.
(3) tz P C : 1 ă |z ´ p1 ´ iq| ă 2u.
(4) tz P C : Repz ´ p1 ´ iqq “ 3u.
(5) tz P C : |Impz ´ p1 ´ iqq| ă 3u.
(6) tz P C : |z ´ p1 ´ iq| “ |z ´ p1 ` iq|u.
(7) tz P C : |z ´ p1 ´ iq| ` |z ´ p1 ` iq| “ 2u.
(8) tz P C : |z ´ p1 ´ iq| ` |z ´ p1 ` iq| ă 3u.
Exercise 1.35 (Cosine Formula). Let z, w be distinct nonzero complex numbers. Show
2
`BO2 ÁB 2
that if A, O, B are the points corresponding to z, 0, w, then cos =AOB “ AO2pAOqpBOq .
Exercise 1.36 (Riemann sphere). Let S 2 :“ tpx, y, zq P R3 : x2 ` y 2 ` z 2 “ 1u be the

unit sphere in R3 , and denote the ‘north pole’ by n :“ p0, 0, 1q. For each p P S 2 ztnu,
define the ‘stereographic’ projection of p “ px, y, zq to be the point ϕppq “ pu, vq where
the ray starting from n, and passing through p, meets the px, yq-plane (thought of as the
complex plane). See the picture below. From the two similar triangles shown there, we
obtain p1 ´ zq : 1 “ x : u and p1 ´ zq : 1 “ y : v. Thus we obtain a map ϕ, given by
1
S 2 ztnu Q px, y, zq ÞÑ ϕpx, y, zq :“ 1´z px, yq P R2 . From the geometric construction of ϕ, it
2
is clear that ϕ : S ztnu Ñ C is a bijection.
1´z
px, y, zq
pu, vq
2 2
2u 2v u `v ´1
Show that R2 Q pu, vq ÞÑ p u2 `v 2
2 `1 , u2 `v 2 `1 , u2 `v 2 `1 q P S ztnu is the inverse of ϕ.
2 2 2
x `y 1´z
Hint: u2 ` v 2 “ p1´zq2 “ p1´zq2 .
As ϕ and ϕ´1 are continuous, the complex plane is homeomorphic to S 2 ztnu (where the
latter is given the Euclidean metric d2 on R3 ). Via the bijective correspondence ϕ, we
can equip the complex plane with another metric dc , called the chordal metric, defined by
1
|
dc pw, w1 q “ d2 pϕ´1 w, ϕ´1 w1 q for all w, w1 P C. Show that dc pw, w1 q “ ? 2|w´w
2 1 2
ď 2.
p1`|w| qp1`|w | q
Hint: Use the Cosine Formula in the triangles formed by n, ϕ´1 w, ϕ´1 w1 and by n, w, w1 .
If w, w1 P Czt0u, then show that dc pw, w1 q “ dc p w1 , w11 q.
1.3.2. Open discs, open sets, closed sets and compact sets. To talk about
sets of points near a given point, we introduce the following. An open ball/disc
Dpz0 , rq with centre z0 and radius r ą 0 is the set Dpz0 , rq :“ tz P C : |z ´ z0 | ă ru.
Dpz0 , rq
0000000
1111111
r
z0
A subset U of C is called open if for every z P U , there exists an rz ą 0 such that

Dpz, rz q Ă U . In other words, no matter what point we choose in U , there is always
some ‘room’ around that point comprising points only from U . For example, it can
be checked that an open disc Dpz0 , rq is an open set. So using the adjective open
in the name for Dpz0 , rq makes sense. Here are some more examples of open sets.
The right half-plane H :“ tz P C : Re z ą 0u, and for 0 ă r ă R, the annulus
A :“ tz P C : r ă |z| ă Ru, are both open sets. It is also convenient to give a
special name to sets whose complement is open, and these are called closed sets.
One can also give a characterisation of closed sets in terms of sequential convergence
(discussed in the following subsection): A set F Ă C is closed if and only if for
every sequence pzn qnPN in F which is convergent in C to L, the limit L P F . A
subset S of C is called bounded if there exists an M ą 0 such that for all z P S,
|z| ď M . Thus S is contained in a big enough disc in the complex plane. A subset
K Ă C is called compact if it is both closed and bounded. We will often use the
known fact from real analysis that a real-valued continuous function on a compact
set possesses a maximiser and a minimiser.
1.3.3. Convergence and continuity. A sequence pzn qnPN of complex numbers

is said to be convergent with limit L P C if for every ǫ ą 0, there exists an index
N P N such that for every n ą N , we have |zn ´ L| ă ǫ. Using the triangle
inequality, it can be seen that for a convergent sequence the limit is unique, and
we write lim zn “ L, or that ‘zn Ñ L as n Ñ 8’.
nÑ8
Example 1.1. If z P C and |z| ă 1, then pz n qnPN converges to 0: |z n ´0| “ ||z|n ´0|,
and as |z| ă 1, the real sequence p|z|n qnPN converges1 to 0.
Exercise 1.37. Let p be the polynomial given by ppzq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d for all z P C,
where d P N, c0 , ¨ ¨ ¨ , cd P C, cd ‰ 0. Show that there exist m, M, R ą 0 such that for all
z P C such that |z| ą R, M |z|d ě |ppzq| ě m|z|d .
Exercise 1.38. Show that a sequence pzn qnPN in C is convergent to L P C if and only if
pRe zn qnPN and pIm zn qnPN are convergent in R to Re L and Im L, respectively.
Exercise 1.39. Show that a sequence pzn qnPN in C is convergent to L P C if and only if
pzn qnPN converges to L.
1This can be seen as follows. Let h “ 1 ´1 ą 0. By the Binomial Theorem, p1`hqn “ 1`nh`¨ ¨ ¨`hn ą nh.
|z|
1 1
So for all n P N, nh
ą p1`hqn
“ |z|n ě 0, and the claim follows from the Sandwich Theorem.
Exercise 1.40. Prove that C is complete, that is, every Cauchy sequence in C converges
in C. (A sequence pzn qnPN is called a Cauchy sequence if for every ǫ ą 0, there is an index
N P N such that for all indices m, n ą N , we have |zn ´ zm | ă ǫ.)
Exercise 1.41. A sequence pzn qnPN in C is said to tend to 8 if for all R ą 0, there exists
an N P N such that for all n P N satisfying n ą N , we have |zn | ą R.
(1) Show that pnin qnPN tends to 8.
(2) Recall Exercise 1.36. Show pzn qnPN tends to 8 if and only if lim d2 pϕ´1 zn , nq “ 0.
nÑ8
It is sometimes convenient to adjoin a point called infinity, denoted by 8, to the complex
plane to form the extended complex plane C p “ C Y t8u, and extend ϕ : S 2 ztnu Ñ C to a
bijection ϕ : S 2 Ñ C p by setting ϕpnq “ 8. If we use this bijection ϕ and the Euclidean
metric on S 2 to obtain a metric on Cp (which is the chordal distance dc we had seen earlier
a
in Exercise 1.36, together with the extra definition dc pz, 8q :“ d2 pϕ´1 z, nq “ 2{ 1 ` |z|2
p dc q is complete and compact.
for all z P C), then it can be shown that the metric space pC,
This space is called the one-point compactification of C.
Let S be a subset of C, z0 P S and f : S Ñ Y , where Y “ C or Y “ R. Then

f is said to be continuous at z0 if for every ǫ ą 0, there exists a δ ą 0 such that
whenever z P S satisfies |z ´ z0 | ă δ, there holds that |f pzq ´ f pz0 q| ă ǫ. The
function f is said to be continuous (on S) if for every z P S, f is continuous at z.
One can also give a characterisation of continuity at a point in terms of convergent
sequences: f is continuous at z0 P S if and only if for every sequence pzn qnPN in S
convergent to z0 , the sequence pf pzn qqnPN is convergent to f pz0 q.
Example 1.2. Complex conjugation, i.e., the map z ÞÑ z : C Ñ C, is continuous.
Indeed, |z ´ z0 | “ |z ´ z0 | “ |z ´ z0 | for all z, z0 P C. Thus complex conjugation
is continuous at each z0 P C. As z0 P C was arbitrary, ¨ is a continuous mapping.
This is geometrically obvious, since complex conjugation is just reflection in the
real axis, and so the image stays close to the reflected point if we are close to
the point. Since pzq “ z for all z P C, complex conjugation is its own inverse.
So complex conjugation is invertible with a continuous inverse. Hence complex
conjugation is a homeomorphism (that is, a continuous bijective mapping with a
continuous inverse) from C to C.
Exercise 1.42. Prove that the maps z ÞÑ Re z and z ÞÑ |z| from C to R are continuous.
Exercise 1.43. Let C be any circle of radius r, and P1 , ¨ ¨ ¨ , Pn be the vertices of a regular
n-gon inscribed in C. Show that the product of the diagonal lengths passing through P1 is
nrn´1 , i.e., pP1 P2 qpP1 P3 q ¨ ¨ ¨ pP1 Pn q “ nrn´1 . Hint: Consider the unit circle first. There is
no loss of generality in assuming P1 , ¨ ¨ ¨ , Pn are the nth roots of unity, say 1, z1 , ¨ ¨ ¨ , zn´1 .
Use continuity to show pz ´ z1 q ¨ ¨ ¨ pz ´ zn´1 q “ z n´1 ` ¨ ¨ ¨ ` z ` 1 for all z P C.
Exercise 1.44. Let S Ă C, z0 P S, f is a real/complex-valued function on S. Show
that f is continuous at z0 if and only if for every sequence pzn qnPN in S convergent to z0 ,
pf pzn qqnPN is convergent to f pz0 q.
Exercise 1.45. Let S Ă C, z0 P S and f : S Ñ C. For each z P S, let upzq :“ Re f pzq and
vpzq :“ Im f pzq. Then u, v : S Ñ R. Show that f is continuous at z0 if and only if u and
v are both continuous at z0 .
1.3.4. Domains. In the rest of the book, the notion of a path-connected open set
will play an important role. We will prove results about our central object of study,
namely functions f : D Ñ C that are complex differentiable at every point of the
set D (Ă C), and for the validity of many of these theorems, we will need D to
be a ‘nice’ subset of C. Sets which satisfy this ‘niceness’ assumption, stipulated
precisely below, will be what we call a ‘domain’. An open subset of C which is
path-connected is called a domain. We already know what ‘open’ means. Now let
us explain what we mean by ‘path-connectedness’. A path (or curve) in C is a
continuous function γ : ra, bs Ñ C.
C
γpbq
γ
a b
γpaq
A stepwise path is a path γ : ra, bs Ñ C such that there exists some integer n ě 0
and points t0 , ¨ ¨ ¨ , tn`1 such that t0 “ a ă t1 ă ¨ ¨ ¨ ă tn ă tn`1 “ b, and for each
k “ 0, 1, ¨ ¨ ¨ , n, the restriction γ : rtk , tk`1 s Ñ C is a path with either a constant
real part or a constant imaginary part. See the picture on the left below.
U
σ
w
γ
z
An open set U Ă C is called path-connected if for every z, w P U , there is a stepwise

path γ : ra, bs Ñ C such that γpaq “ z, γpbq “ w, and for all t P ra, bs, γptq P U .
Remark 1.1. In the definition of path-connected open sets, the restriction that
the paths should be stepwise, can be relaxed: That is, if we look at those open sets
in which any two points can be joined merely by a path, then this class of open
sets coincides with our path-connected sets2. As this is an unnecessary diversion
for the path we follow, we will use our definition with stepwise paths. ˚
2The idea behind this is as follows. Let U be an open set, and let γ : ra, bs Ñ C be a path such that for
all t P ra, bs, γptq P U . We want a stepwise path σ joining γpaq and γpbq. As U is open, for each t P r0, 1s,
γptq P U belongs to an open disc Dt Ă U . The compact set K :“ tγptq : t P ra, bsu has the open cover
tDt : t P ra, bsu, which has a finite subcover tDt1 , ¨ ¨ ¨ , Dtn u. It is then believable that we can construct a
stepwise path by staying within the region formed by these discs. See the picture on the right above. A
detailed proof can be found e.g. in [20, Ch. 2].
Example 1.3. The entire plane C is clearly a domain. The open unit disc, namely
D :“ tz P C : |z| ă 1u, is a domain. More generally, if z0 P C and r ą 0, the disc
Dpz0 , rq :“ tz P C : |z ´ z0 | ă ru is a domain. For 0 ă r ă R, the annulus
A :“ tz P C : r ă |z ´ z0 | ă Ru is a domain. Extreme cases of annuli, namely,
‚ the ‘punctured disc’ D˚ pz0 , Rq :“ tz P C : 0 ă |z ´ z0 | ă Ru “ Dpz0 , Rqztz0 u,
‚ the exterior of a disc, De pz0 , rq :“ tz P C : r ă |z ´ z0 |u, and
‚ the ‘punctured plane’ tz P C : 0 ă |z ´ z0 |u “ Cztz0 u
are domains. The right half-plane H :“ tz P C : Repzq ą 0u is a domain.
A H
D
The set S :“ tz P C : |z| ‰ 1u “: CzT is not a domain. Although S is open, it

is not path-connected. Indeed, there is no path joining 0 to 2, and staying within
S: If there was such a path γ, then by the Intermediate Value Theorem applied
to the continuous function ra, bs Q t ÞÑ |γptq| P R, as |γpaq| “ 0 ă 1 ă 2 “ |γpbq|,
there exists a t˚ P ra, bs such that |γpt˚ q| “ 1, giving γpt˚ q R S, a contradiction.
Exercise 1.46. Show that the set tz P C : pRe zqpIm zq ą 1u is open, but not a domain.
r :“ tz P C : z P Du is a domain.
Exercise 1.47. If D is a domain, then show that D
1.4. The exponential function and kith

In this last section, we discuss some basic complex functions.
zÑ
Þ exp z (exponential function)
zÑÞ sin z, cos z (trigonometric functions)
zÑ Þ Log z (logarithm function).
Later we will see that the first two are complex differentiable in the entire complex
plane, while the logarithm is complex differentiable in the domain formed by delet-
ing the ray corresponding to the real interval p´8, 0s. Of course, we will also see
that polynomials and rational functions are complex differentiable (the latter in the
domain where they are defined). But the above are frequently arising important
maps. They are the counterparts to the familiar special/non-rational functions
from calculus, to which they reduce to, when restricted to the real axis. In other
words, when we restrict our functions to the argument z “ x P R, then we get the
usual real-valued functions
x ÞÑ ex ,
x ÞÑ sin x, cos x,
x ÞÑ log x.
So our definitions provide extensions of the usual real-valued counterparts.
1.4. The exponential function and kith 21
We will see that these extensions have new and interesting properties in the complex
domain that are not possessed by them when the argument is only allowed to be
real. Let us begin with the exponential function.
1.4.1. The exponential exp z.

Definition 1.1 (Complex exponential). For z “ x ` iy P C, where x, y are real,
we define the complex exponential exp z of z by exp z “ ex pcos y ` i sin yq.
When y “ 0, the right hand side is simply ex . So our definition extends the usual
exponential function pR Qq x ÞÑ ex pP Rq. Why do we define the complex extension
in this way? After all, z ÞÑ eRe z also gives an extension of the real exponential. So
why not use this simpler definition instead? We define our exp function in the way
we do, because we will see that this is the unique extension of the real exponential
to the whole complex plane having the property that the extension is complex
differentiable everywhere; see Example 4.10 on page 92. In fact, just as for the real
d x d
counterpart, where dx e “ ex for all x P R, we will see that dz exp z “ exp z for all
z P C. So eventually we will learn that our complicated definition is actually quite
natural. Right now, let us check the following elementary properties.
Proposition 1.3.
(1) exp 0 “ e0 pcos 0 ` i sin 0q “ p1qp1 ` i0q “ 1.
(2) For all z1 , z2 P C, exppz1 `z2 q “ pexp z1 qpexp z2 q.
(3) For all z P C, exp z ‰ 0, and pexp zq´1 “ expp´zq.
(4) For all z P C, exppz ` 2πiq “ exp z.
(5) For all z P C, | exp z| “ eRe z .
Proof. (2) If z1 “ x1 ` iy1 and z2 “ x2 ` iy2 , where x1 , x2 , y1 , y2 P R, then

exppz1 `z2 q “ expppx1 ` x2 q ` ipy1 ` y2 qq
“ ex1 `x2 pcospy1 ` y2 q ` i sinpy1 ` y2 qq
“ ex1 ex2 pcos y1 cos y2 ´ sin y1 sin y2 ` ipsin y1 cos y2 ` cos y1 sin y2 qq
“ ex1 pcos y1 ` i sin y1 qex2 pcos y2 ` i sin y2 q “ pexp z1 qpexp z2 q.
(3) From the previous part, we see that 1 “ exp 0 “ exppz ´ zq “ pexp zqpexpp´zqq,
showing that exp z ‰ 0, and pexp zq´1 “ expp´zq. (Thus exp maps C to the
‘punctured’ plane Czt0u.)
(4) We have
exppz ` 2πiq “ pexp zqpexpp2πiqq
“ pexp zqe0 pcosp2πq ` i sinp2πqq
“ pexp zq1p1 ` i0q “ exp z.
(This shows that exp is ‘periodic in the y-direction’ with a period of 2π.)
z3
w2
z2 exp zi
w1 2π
z1 exp ex
w0 0
exp wi
z0
w´1
This phenomenon is not present in the x-direction, where as in the real setting,
the function x ÞÑ exppx ` iy0 q (with y0 P R fixed) is one-to-one.
a a
(5) For x, y P R, |ex cos y ` iex sin y| “ e2x ppcos yq2 ` psin yq2 q “ e2x p1q “ ex .
It follows that | exppx ` iyq| “ ex . Thus exp maps vertical lines in the complex
plane, i.e., all points having a common real part, into circles, i.e., points having
the same absolute value (which is the distance to the origin).
Proposition 1.3(3) implies exp is not one-to-one; rather it is periodic with period
2πi. The picture below shows the effect of the map z ÞÑ exp z on horizontal (fixed
imaginary part y) and vertical lines (fixed real part x).
´8 `8 exp ‘e`8 ’
‘e´8 ’
e¨ pcos θ0 ` i sin θ0 q
¨ ìθ0
exp θ0
x0 ì ¨
ex0 pcos ¨ ` i sin ¨q
exp ex0
x0
2π
exp
3π
2
π
2
´1 0 1 2
In the picture, we see that exp preserves the angle between the curves considered
in its domain: The horizontal and vertical lines, which are mutually perpendicu-
lar, are mapped to circles and radial rays, which are also mutually perpendicular.
We will see that this is no coincidence, and this property of preservation of an-
gles between curves in the domain together with their ‘orientation’ is something
which is possessed by all complex differentiable functions in domains, and is called
‘conformality’; see page 44.
For z “ iy, where y P R, exppiyq “ cos y ` i sin y. This is called Euler’s formula.
If z P C has polar coordinates pr, θq, then z “ rpcos θ ` i sin θq “ r exppiθq.
Exercise 1.48. Compute exp z for the following values of z: 3 ` πi, i 9π
2 .
Exercise 1.49. Find all z P C that satisfy exp z “ πi.

Exercise 1.50. Sketch the path r0, 2πs Q t ÞÑ exppitq P C.
Exercise 1.51. Describe D “ tz P C : | exppízq| ă 1u. Is D a domain?
Exercise 1.52. Describe the image of the line y “ x in R2 “ C, under the exponential
map z “ x ` iy ÞÑ exp z. Proceed as follows: Start with the parametric form x “ t, y “ t,
and get an expression for the image curve in parametric form. Sketch this curve, explaining
what happens as t Ñ 8, and what happens as t Ñ ´8.
Exercise 1.53. Find the absolute value and the real and imaginary parts of exppz 2 q and
of exp z1 in terms of the real and imaginary parts x, y of z “ x ` iy.
Exercise 1.54. Show that for all z P C, exp z “ exp z.
Exercise 1.55. Let a P R. Given ǫ ą 0, consider the square region Sǫ in the complex
plane, defined by Sǫ :“ tpx, yq P C : |x ´ a| ă ǫ, |y| ă ǫu. Let exp Sǫ denote the image of
Sǫ under exp. Sketch Sǫ and exp Sǫ in the complex plane. Let ρpǫq denote the ratio of the
area of exp Sǫ to the area of Sǫ . Determine the limiting value of ρpǫq as ǫ Ñ 0.
1.4.2. Trigonometric functions. We now extend the real trigonometric func-

tions of a real variable to complex trigonometric functions of a complex variable. By
the Euler formula, for x P R, exppixq “ cos xì sin x and exppíxq “ cos xí sin x,
which give
exppixq ` exppíxq exppixq ´ exppíxq
cos x “ and sin x “ .
2 2i
This prompts the following definitions: For z P C, we define
exppizq ` exppízq exppizq ´ exppízq
cos z “ and sin z “ .
2 2i
Clearly these definitions give extensions of the usual real trigonometric functions
because when we put z “ x P R, we get cos z “ cos x and sin z “ sin x, as shown
above (using Euler’s formula). Several trigonometric identities continue to hold in
the complex setting. For instance, cospz1 ` z2 q “ pcos z1 qpcos z2 q ´ psin z1 qpsin z2 q
for all z1 , z2 P C: Indeed, using properties of exp,
pcos z1 qpcos z2 q ´ psin z1 qpsin z2 q
pexppiz1 qèxppíz1 qq pexppiz2 qèxppíz2 qq
“ 2 2 ´ pexppiz1 qéxppíz
2i
1 qq pexppiz2 qéxppíz2 qq
2i
2 exppipz1 `z2 qq`2 exppípz1 `z2 qq
“ 4 “ cospz1 ` z2 q.
Exercise 1.56. Show that sinpz1 ` z2 q “ psin z1 qpcos z2 q ` pcos z1 qpsin z2 q for all z1 , z2 P C.
For all z P C, psin zq2 ` pcos zq2 “ 1, since
psin zq2 ` pcos zq2 “ p exppizqéxppízq

2i q2 ` p exppizqèxppízq
2 q2
expp2izq´2èxpp´2izq expp2izq`2èxpp´2izq
“ ´4 ` 4 “ 1.
However, as opposed to the real trigonometric functions which satisfy | sin x| ď 1
and | cos x| ď 1 for real x, z ÞÑ sin z and z ÞÑ cos z are not bounded. Indeed, for
z “ iy, where y is real, we have
exppipiyqqèxppípiyqq exppýqèxp y eý èy
cospiyq “ 2 “ 2 “ 2 ,
eý éy
and so cospiyq Ñ `8 as y Ñ ˘8. Similarly, as sinpiyq “ 2i , | sinpiyq| Ñ 8 as
y Ñ ˘8.
We will see later that z ÞÑ cos z, sin z are complex differentiable everywhere in
the complex plane.
y ý y ý
Exercise 1.57. For y P R, let cosh y :“ e è 2 and sinh y :“ e é 2 . Show that
pcosh yq2 ´ psinh yq2 “ 1 for all y P R. For z “ x ` iy, where x, y are real, show that
cos z “ pcos xqpcosh yq ´ ipsin xqpsinh yq and | cos z|2 “ pcosh yq2 ´ psin xq2 .
Exercise 1.58. We know that the equation cos x “ 3 has no real solution x. However,
show that there are complex z that satisfy cos z “ 3, and find them all.
Exercise 1.59. Determine all z P C such that cos z ` i sin z “ e.
1.4.3. Logarithm function. In the real setting, given a positive y, log y P R is

the unique real number such that elog y “ y. Thus log : p0, 8q Ñ R serves as the
inverse of the function x ÞÑ ex : R Ñ p0, 8q.
ex
log y 0
In the complex case, we know that exp : C Ñ Czt0u, and we now wonder if there is
a ‘complex logarithm function’ mapping Czt0u to C that serves as an inverse to the
complex exponential function. Given a z ‰ 0, we seek a complex number w such
that exp w “ z, and we would like to call w the ‘complex logarithm of z’. However,
the exponential function exp is 2π-periodic in the y-direction, and so the moment
we find one w such that ew “ z, we know that there are infinitely many others,
since exppw ` 2πinq “ exp w “ z for all n P Z. Given this infinite choice, which w
is the complex logarithm of z? We remedy this problem of nonuniqueness by just
choosing a w that lies in a fixed particular horizontal strip of width 2π. We will show
that all possible nonzero complex numbers can be obtained as the exp of something
lying in the fixed strip, and now for the purpose of defining the complex logarithm,
we choose (somewhat arbitrarily), the strip S :“ tz P C : ´π ă Impzq ď πu. The
strip S :“ R ˆ p´π, πs is mapped by exp onto Czt0u.
Czt0u
exp
π
S
´π Log
This will give, as we shall show below, a unique w in the strip such that exp w “ z,
and we will call this unique w the ‘principal logarithm of z’, denoted by Log z.
To do this, we first introduce the notion of the principal argument of a nonzero
complex number.
Definition 1.2 (Principal argument).
If z P Czt0u, then the principal argument of z is the unique number Arg z P p´π, πs
such that z “ |z|pcospArg zq ` i sinpArg zqq. If z “ 0, then we define Arg z “ 0.
To see the claimed uniqueness above, we note that if α, β P p´π, πs are such that
|z|pcos αì sin αq “ |z|pcos βì sin βq, then cos α “ cos β and sin α “ sin β, showing
α ´ β “ 2πn for some n P Z. But as α, β P p´π, πs, it follows that |α ´ β| ă 2π, so
that 2π|n| “ |α ´ β| ă 2π, giving |n| “ 0, and hence n “ 0.
For example, Arg 3 “ 0, Argp´1q “ π, Arg i “ π2 , Argpíq “ ´ π2 . If we start at
a point on the positive real axis in the complex plane, and go around anticlockwise
in a circle, then there is a sudden jump in the value of the principal argument as
we cross the negative real axis: On the negative real axis, the value of the principal
argument is π, while just below the negative real axis, the principal argument is
close to ´π.
π π
Exercise 1.60. Depict tz P C : z ‰ 0, 4 ă |Arg z| ă 3u in the complex plane.
Now we are ready to define the principal logarithm of nonzero complex numbers.
Definition 1.3 (Principal Logarithm).
If z P Czt0u, then the principal logarithm Log z of z is Log z “ log |z| ` iArg z.
For example, Log píq “ log | ´ i| ` iArgpíq “ log 1 ´ π2 i “ 0 ´ π2 i “ ´ π2 i. We

now show that exp : S Ñ Czt0u and Log : Czt0u Ñ S are inverses of each other.
In particular, exp : S Ñ Czt0u and Log : Czt0u Ñ S are both bijections. Since
exppLog wq “ elog |w|pcospArg wq ` i sinpArg wqq “ |w|pcospArg wq ` i sinpArg wqq “ w
for all w P Czt0u, we have exp ˝Log “ IdCzt0u , where IdCzt0u denotes the identity
map on Czt0u, Czt0u Q w ÞÑ w.
If z P S, and z “ x ` iy, where x P R and y P p´π, πs, then we have that
Logpexp zq “ Logpex pcos y ` i sin yqq “ log ex ` iy “ x ` iy “ z. Here we used the
fact that Argpex pcos y ` i sin yqq “ y, since y P p´π, πs. Hence Log ˝ exp “ IdS ,
where IdS is the identity map on S, S Q z ÞÑ z.
Of course, had we chosen to define the principal argument θ of a nonzero
z “ |z| exppiθq to lie in a different interval pa, 2π ` as or ra, 2π ` aq for some other
a, we would have obtained a different well-defined notion of the logarithm (which
would also be equally legitimate). But in this book, the principal logarithm of z
will always mean log |z| ` iArg z, with the principal argument Arg z P p´π, πs.
Continuity of Log on Czp´8, 0s. If I Ă R is an interval, then the set I ˆt0u Ă C

will often be denoted simply by I. Intuitively, Arg : Czt0u Ñ p´π, πs is not
continuous at any point of the negative real axis p´8, 0q. Thus Log will not be
continuous at these points either (since Log z “ |z| ` iArg z, and z ÞÑ |z| is
continuous everywhere). We show below that at each point of p´8, 0q Ă C, Log is
not continuous.
Let z P p´8, 0q. Consider the sequence p|z| expp´πi ` ni qqnPN in Czt0u, which
converges to z “ ´|z|:
lim |z| exppip´π ` n1 qq “ lim ´|z|pcos n1 ` i sin n1 q “ ´|z|p1 ` i0q “ ´|z|.
nÑ8 nÑ8
Also, Logp|z| exppip´π ` n1 qqq “ log |z| ` ip´π ` n1 q. Thus
lim Logp|z| exppip´π` n1 qqq “ log |z| ´ iπ, whereas
nÑ8
Logp lim |z| exppip´π` n1 qqq “ Logp´|z|q “ log |z| ` iπ.
nÑ8
Using Exercise 1.44, it follows that Log is not continuous at z P p´8, 0q.
On the other hand, Log is continuous on the smaller set Czp´8, 0s. This is because
the principal argument Arg is continuous on Czp´8, 0s, as argued below. Let z0 be
any complex number not in p´8, 0s. Then Arg z0 P p´π, πq. There is some room
around z0 not touching the negative real axis, and we can find a small enough r
such that the disc Dpz0 , rq does not touch the ray p´8, 0s. Thus, given an ǫ ą 0,
by shrinking r further if necessary (so that the disc Dpz0 , rq is contained in the
angular wedge formed by the rays emanating from the origin making angles of
Arg z0 ´ ǫ and Arg z0 ` ǫ with the positive real axis), we can ensure that the points
z in Dpz0 , rq satisfy |Arg z ´ Arg z0 | ă ǫ.
Arg
z0
π Arg z0 ´π
As both z ÞÑ log |z| and Arg are continuous in Czp´8, 0s, it follows that Log is
continuous there too. Using the continuity of Log on Czp´8, 0s, we will see later
on that Log is complex differentiable in Czp´8, 0s.
The principal value of ab for a P Czt0u and b P C. Recall that if a P Czt0u,

and n P N, then
an :“ a ¨ ¨ ¨ a (n times)
“ pexp Log aq ¨ ¨ ¨ pexp Log aq “ exppLog a ` ¨ ¨ ¨ ` Log aq
“ exppn Log aq.
Recall also that if a ą 0 and b P R, then
ab :“ eb log a
“ exppb log aq (since b log a P R)
“ exppb Log aq (since a ą 0).
These considerations motivate defining ab :“ exppb Log aq for those complex number
a, b for which the right-hand side make sense, that is, a P Czt0u and b P C.
Definition 1.4. For a P Czt0u, b P C, the principal value of ab is ab :“ exppb Log aq.
For example, the principal value of ii is

π
exppi Log iq “ exppiplog |i| ` iArg iqq “ exppip0 ` i π2 qq “ expp´ π2 q “ e´ 2 .
Taking a :“ e, for all complex numbers b :“ z, we have
ez “ exppzLog eq “ exppzplog e ` iArg eqq “ exppzp1 ` i0qq “ exp z.
Hence if z P C, then swapping the notations exp z and ez is allowed.
Exercise 1.61. Find Log p1 ` iq.

Exercise 1.62. Find Logp´1q and Log 1. Show that Logpz 2 q isn’t always equal to 2 Logz.
If n P N, then characterise the set of all z P Czt0u for which we have Logpz n q “ n Log z.
Exercise 1.63. Find the image of tw P C : 1 ă |w| ă eu under the principal logarithm.
Exercise 1.64. Find the real and imaginary parts of the principal value of p1 ` iq1í .
Exercise 1.65. If w P Czt0u and n P N, then show that wń “ p w1 qn . Show that for all
m P Z, and t P R, peit qm “ eimt .
Exercise 1.66 (Möbius transformations). Let a, b, c, d P C be such that ad ´ bc ‰ 0. We
define the complex-valued function f by f pzq “ az`b cz`d for all z P C such that cz ` d ‰ 0.
The constants c and d cannot both be zero since ad ´ bc ‰ 0. If c “ 0, then f is defined
on C, while if c ‰ 0, then f is defined in Czt´ dc u. It will be convenient to extend f to the
extended complex plane C p “ C Y t8u. If c “ 0, then we set f p8q “ 8, while if c ‰ 0, then
p ÑC
we set f p´ dc q “ 8 and f p8q “ ac . Such an f : C p is called a Möbius transformation.
(1) Show that f is a bijection, with the inverse given by the Möbius transformation g,
where gpwq “ ´dw`b
cwá . Note that p´dqpáq ´ bc “ ad ´ bc ‰ 0.
(2) Prove that the composition of Möbius transformations is also a Möbius transformation.
1
az`b z`b1
For simiplicity, just consider the case of transformations f pzq “ cz`d , gpzq “ ac1 z`d 1,
where c, c1 are both nonzero. (The set of all Möbius transformations forms a group
with composition of maps taken as the group operation.)
(3) Describe the fate of circles and lines under the following Möbius transformations:
(Translation) f pzq “ z ` b (Dilation) f pzq “ az pa ą 0q
(Rotation) f pzq “ eiθ z pθ P Rq (Inversion) f pzq “ z1 .
For the image under inversion of a line not passing through 0, it can be assumed (why?)
that the line is L “ tz0 `tiz0 : t P Ru, where z0 ‰ 0. For such a line, show that the image
lies on a circle with centre 2z10 and radius 2|z10 | .
az`b
(If c ‰ 0, then cz`d “ ac ` cpcz`dq
bcád
. So every Möbius transformation is the composition
of the special ones described above. Imagining straight lines as circles of infinite radius,
define a sircle be a line or a circle. Möbius transformations map sircles to sircles.)
Chapter 2
Complex differentiability
In this chapter we will learn three main things:
‚ The definition of complex differentiability, i.e., for f : U Ñ C, where U is an open

subset C, and z0 P U , we will learn the meaning of the statement ‘f is complex
differentiable at z0 with complex derivative f 1 pz0 q’.
‚ The Cauchy-Riemann equations, Bu Bv Bu Bv

Bx “ By and By “ ´ Bx , are partial differential
equations that are satisfied by the the real and imaginary parts u, v of a complex
differentiable function f : U Ñ C wherever it is complex differentiable.
f vpx, yq f px, yq
px, yq
U
upx, yq
Vice versa, if the Cauchy-Riemann equations are satisfied at z0 P U and u, v are

real differentiable there, then f “ u ` iv is complex differentiable at z0 .
‚ The geometric meaning of the complex derivative f 1 pz0 q: infinitesimally the map
f is an amplification by |f pz0 q| together with a twist (a counterclockwise rotation)
through Argpf 1 pz0 qq.
The central result in this chapter is that a function is complex differentiable at a

point if and only if its real and imaginary parts are real differentiable, and they
satisfy the Cauchy-Riemann equations.
29
30 2. Complex differentiability
2.1. Complex differentiability

We can define the notion of the limit of a complex-valued function just as in
ordinary calculus. Let U be an open subset of C, h : U ztz0 u Ñ C, and z0 P U .
If there exists an L P C such that for all ǫ ą 0, there exists a δ ą 0 such that
whenever z P U satisfies 0 ă |z ´ z0 | ă δ, we have |hpzq ´ L| ă ǫ, then we say that
the limit of h as z Ñ z0 exists, and equals L. We then write lim hpzq “ L.
zÑz0
Exercise 2.1. Let U be an open subset of C, z0 P U , and h1 , h2 : U ztz0 u Ñ C be such

that lim hi pzq exist for i “ 1, 2. Let h1 ` h2 , respectively h1 h2 denote the pointwise sum,
zÑz0
respectively product, of h1 and h2 . Show that lim ph1 ` h2 qpzq “ lim h1 pzq ` lim h2 pzq,
zÑz0 zÑz0 zÑz0
and lim ph1 h2 qpzq “ p lim h1 pzqqp lim h2 pzqq.
zÑz0 zÑz0 zÑz0
For an open set U Ă C and an f : U Ñ C, it is clear that f is continuous at z0 P U

if and only if lim f pzq “ f pz0 q.
zÑz0
Exercise 2.2. Let U Ă C be open, z0 P U , f : U ztz0 u Ñ C be such that f pzq ‰ 0 for

all z P U ztz0u and lim f pzq ‰ 0. Define f1 : U Ñ C by p f1 qpzq “ f pzq
1
for all z P U ztz0 u.
zÑz0
Show that lim p f1 qpzq “ 1
lim f pzq .
zÑz0 zÑz0
Definition 2.1 (Complex differentiable, holomorphic, entire).

Let U be an open subset of C, f : U Ñ C and z0 P U . Then f is said to be complex
differentiable at z0 if there exists a complex number L such that lim f pzq´f
z´z0
pz0 q
“ L,
zÑz0
i.e., for every ǫ ą 0, there exists a δ ą 0 such that whenever z P U satisfies
0 ă |z ´ z0 | ă δ, we have | f pzq´f pz0 q
z´z0 ´ L| ă ǫ. This (unique limit) L is denoted by
1 df
f pz0 q or dz pz0 q, and is called the complex derivative of f at z0 .
We say f : U Ñ C is holomorphic1 in U if for all z P U , f is complex differentiable
at z. We denote by OpU q the set of all holomorphic functions in U .
A function that is holomorphic in C is called entire, that is, the domain of f is
understood to be the whole of C and moreover, f is holomorphic in C.
Example 2.1. Consider the function f : C Ñ C defined by f pzq “ z 2 for all z P C.
z 2 ´z 2
We will now show that f is entire. We have f pzq´f z´z0
pz0 q
“ z´z00 “ z ` z0 « 2z0 for
1
z near z0 , and so we guess ˇ thatˇ f pz0 q “ˇ 2z0 . Let us show this now. For z ‰ z0 ,
f pzq´f pz0 q ˇ ˇ z 2 ´z02 ˇ
we have | z´z0 ´ 2z0 ˇ “ ˇ z´z0 ´ 2z0 ˇ “ |z ` z0 ´ 2z0 | “ |z ´ z0 |. So the left-
hand side can be made as small as we please when z is close enough to z0 . Let
ǫ ą 0. Set δ :“ ǫ ą 0. Then whenever z P C satisfies 0 ă |z ´ z0 | ă δ, we have
| f pzq´f
z´z0
pz0 q
´ 2z0 | “ |z ´ z0 | ă δ “ ǫ. Hence f 1 pz0 q “ 2z0 . As z0 P C was arbitrary,
df
f is holomorphic in C, i.e., f is entire. We have f 1 pzq “ dz pzq “ 2z for all z P C,
d 2
and we often we simply write dz z “ 2z (z P C).
1‘Holomorphic’ is derived from the Greek ‘holos’ meaning ‘entire’, and ‘morphe’ meaning ‘form’.
2.1. Complex differentiability 31
Here’s an example of a map which at each z P C is not complex differentiable at z.
Example 2.2. Let g : C Ñ C be defined by gpzq “ z (z P C). We show that g

is complex differentiable nowhere. Suppose g is complex differentiable at z0 P C.
Let ǫ :“ 21 ą 0. Then there exists δ ą 0 such that whenever z P C satisfies
0 ă |z ´ z0 | ă δ, we have
ˇ gpzq ´ gpz q ˇ ˇz ´ z ˇ
ˇ 0 ˇ ˇ 0 ˇ
ˇ ´ g 1 pz0 qˇ “ ˇ ´ g 1 pz0 qˇ ă ǫ.
z ´ z0 z ´ z0
So whenever z is in the punctured disc of radius δ with centre z0 , we are guaran-
teed that this inequality holds. We will now make special choices of the z as the
midpoints on the horizontal and vertical radii shown in the figure, and show that
the special cases of the inequality above yield that g 1 pz0 q must lie in discs of radius
1
2 with centres at ´1 and 1. But these discs have an empty intersection, and this
will be our contradiction. We give the details below.
z0 ´1 0 1
Taking z “ z0 ` 2δ , we have 0 ă |z ´ z0 | ă δ, and so

ˇz ´ z ˇ ˇδ ˇ
ˇ 0 ˇ ˇ ˇ
ˇ ´ g 1 pz0 qˇ “ ˇ 2δ ´ g 1 pz0 qˇ “ |1 ´ g1 pz0 q| ă ǫ. (2.1)
z ´ z0 2
Also, taking z “ z0 ` i 2δ , we have 0 ă |z ´ z0 | ă δ, and so

ˇz ´ z ˇ ˇ í δ ˇ
ˇ 0 ˇ ˇ ˇ
ˇ ´ g1 pz0 qˇ “ ˇ δ 2 ´ g1 pz0 qˇ “ | ´ 1 ´ g1 pz0 q| “ |1 ` g1 pz0 q| ă ǫ. (2.2)
z ´ z0 i2
It follows from (2.1) and (2.2) that
1
2 “ |1 ´ g 1 pz0 q ` 1 ` g1 pz0 q| ď |1 ´ g1 pz0 q| ` |1 ` g1 pz0 q| ă ǫ ` ǫ “ 2ǫ “ 2 “ 1,
2
a contradiction. So g is not complex differentiable at z0 .
Exercise 2.3. Show that f : C Ñ C defined by f pzq “ |z|2 for z P C, is complex
differentiable at 0 and that f 1 p0q “ 0. We will see later (in Exercise 2.12) that f is not
complex differentiable at any nonzero complex number.
Exercise 2.4. Let D be a domain, and f : D Ñ C be holomorphic in D. By Exercise 1.47
r :“ tz P C : z P Du is a domain as well. Define fr : D
D r Ñ C by frpzq “ f pzq for all z P D.
r
r r
Prove that f is holomorphic in D.
1
Exercise 2.5. Let f pzq “ z for all z P Czt0u. Prove that f 1 pzq “ ´ z12 for all z P Czt0u.
The following reformulation of complex differentiability is useful to prove ele-

mentary facts about complex differentiation. Roughly speaking, the result says
that for a complex differentiable function f with complex derivative L at z0 ,
f pzq ´ f pz0 q ´ Lpz ´ z0 q goes to 0 ‘faster than z ´ z0 ’.
Lemma 2.1. Let U be an open set in C, z0 P U, f : U Ñ C, and L P C.
Then the following are equivalent:
(1) f is complex differentiable at z0 with f 1 pz0 q “ L.
(2) There exist r ą 0, and h : Dpz0 , rqpĂ U q Ñ C, such that
(a) f pzq “ f pz0 q ` pL ` hpzqqpz ´ z0 q for all z P Dpz0 , rq and
(b) lim hpzq “ 0.
zÑz0
Proof. (2) ñ (1): Upon rearranging, we have lim p f pzq´f

z´z0
pz0 q
´ Lq “ lim hpzq “ 0.
zÑz0 zÑz0
So f is complex differentiable at z0 , and f 1 pz0 q “ L.
(1) ñ (2): Let f be complex differentiable at z0 , and L :“ f 1 pz0 q. As U is open,

and since f is complex differentiable at z0 with f 1 pz0 q “ L, there exists a δ1 ą 0
such that Dpz0 , δ1 q Ă U and whenever z P Dpz0 , δ1 qztz0 u, | f pzq´f
z´z0
pz0 q
´ L| ă 1. Set
r :“ δ1 , and define h : Dpz0 , rq Ñ C by
#
f pzq´f pz0 q
z´z0 ´ L if z ‰ z0 ,
hpzq “
0 if z “ z0 .
Then f pzq “ f pz0 q ` pL ` hpzqqpz ´ z0 q for all z P Dpz0 , rq. If ǫ ą 0, then there is
a δ ą 0 (which can be chosen smaller than r) such that whenever 0 ă |z ´ z0 | ă δ,
we have p|hpzq´0| “q | f pzq´f
z´z0
pz0 q
´ L| ă ǫ. So lim hpzq “ 0.
zÑz0
Exercise 2.6. Let U Ă C be open. Use Lemma 2.1 to show that if f : U Ñ C is complex
differentiable at z0 P U , then f is continuous at z0 .
Proposition 2.2. Let U be an open subset of C. Let f, g : U Ñ C be complex
differentiable functions at z0 P U . Then:
(1) f ` g is complex differentiable at z0 and pf ` gq1 pz0 q “ f 1 pz0 q ` g1 pz0 q.
pHere f ` g : U Ñ C is defined by pf ` gqpzq “ f pzq ` gpzq for all z P U .q
(2) If α P C, then α¨f is complex differentiable and pα¨f q1 pz0 q “ α f 1 pz0 q.
pHere α¨f : U Ñ C is defined by pα¨f qpzq “ α f pzq for all z P U .q
(3) f g is complex differentiable at z0 , and pf gq1 pz0 q “ f 1 pz0 qgpz0 q`f pz0 qg1 pz0 q.
pHere f g : U Ñ C is defined by pf gqpzq “ f pzqgpzq for all z P U .q
Remark 2.1. Let U be an open subset of C. It follows from the above that OpU q
is a complex vector space with pointwise operations. The third statement above
shows that also the pointwise product of two holomorphic functions is holomorphic,
and so OpU q also has the structure of a complex algebra (that is, it is a complex
vector space with a bilinear product OpU q ˆ OpU q Q pf, gq ÞÑ f g P OpU q). ˚
2.1. Complex differentiability 33
Example 2.3. It is easy to see that if f pzq :“ z (z P C), then f 1 pzq “ 1. Moreover,
if gpzq :“ 1 (z P C), then g1 pzq “ 0. Using the rule for complex differentiation of
a pointwise product of holomorphic functions, induction gives that for all n P N,
d n
z ÞÑ z n is entire, and dz z “ nz n´1 . In particular, all polynomials are entire.
Exercise 2.7. Use Lemma 2.1 to prove Proposition 2.2.
Exercise 2.8. Is OpCq a finite-dimensional vector space (with pointwise operations)?
Exercise 2.9. Let U Ă C be open, and let f P OpU q be such that f pzq ‰ 0 for z P U .
1
pzq
Define f1 : U Ñ C by p f1 qpzq “ f pzq
1
for all z P U . Prove p f1 q1 pzq “ ´ pffpzqq 2 for all z P U .
d m
Exercise 2.10. Show that in Czt0u, for each m P Z, dz z “ mz m´1 .
Just as we have the chain rule for the derivative of composition of functions in the
real setting, there is an analogous chain rule for holomorphic maps.
Proposition 2.3 (Chain rule).
If ‚ Uf , Ug are open subsets of C,
‚ f : Uf Ñ C is complex differentiable at z0 P Uf , f pUf q Ă Ug ,
‚ g : Ug Ñ C is complex differentiable at f pz0 q,
then their composition2 is complex differentiable at z0 and pg ˝f q1 pz0 q “ g1 pf pz0 qqf 1 pz0 q.
Uf Ug
f f pUf q g C
Proof. By the complex differentiability of f at z0 , and of g at f pz0 q, there exist

rf , rg ą 0, and functions hf , hg such that f pzq ´ f pz0 q “ pf 1 pz0 q ` hf pzqqpz ´ z0 q
for all z P Dpz0 , rf q Ă Uf , gpwq ´ gpf pz0 qq “ pg 1 pf pz0 qq ` hg pwqqpw ´ f pz0 qq for all
w P Dpf pz0 q, rg q Ă Ug , lim hf pzq “ 0, and lim hg pwq “ 0. Set
zÑz0 wÑf pz0 q
"
r hg pwq if w P Dpf pz0 q, rg qztf pz0 qu,
hg pwq :“
0 if w “ f pz0 q.
1 r
Then gpwq´gpf pz0 qq “ pg pf pz0 qq` hg pwqqpw´f pz0 qq for all w P Dpf pz0 q, rg q Ă Ug ,
and lim r
hg pwq “ 0. As f is continuous at z0 , there exists a δ P p0, rf q such that
wÑf pz0 q
if 0 ă |z ´ z0 | ă δ, then f pzq P Dpf pz0 q, rg q. So for all z P Dpz0 , δq, we have
pg ˝ f qpzq ´ pg ˝ f qpz0 q “ pg 1 pf pz0 qq ` r hg pf pzqqqpf 1 pz0 q ` hf pzqqpz ´ z0 q
“ pg 1 pf pz0 qqf 1 pz0 q ` ϕpzqqpz ´ z0 q,
where ϕpzq “ f 1 pz0 qr
hg pf pzqq ` g1 pf pz0 qqhf pzq ` r
hg pf pzqqhf pzq for all z P Dpz0 , δq.
2The composition g ˝ f : U Ñ C is defined by pg ˝ f qpzq “ gpf pzqq for all z P U .
f f
We claim that lim ϕpzq “ 0. As lim hf pzq “ 0, it is enough to show lim r

hg pf pzqq “ 0.
zÑz0 zÑz0 zÑz0
Let ǫ ą 0. Since r hg pwq Ñ 0 as w Ñ f pz0 q, there exists a δr P p0, rg q such that

r
for all w P Dpf pz0 q, δqztf pz0 qu, |r
hg pwq| “ |hg pwq| ă ǫ. Clearly if w “ f pz0 q, then
|r
hg pwq| “ |0| “ 0 ă ǫ too. So for all w P Dpf pz0 q, δq,r |rhg pwq| ă ǫ. By the continuity
of f at z0 , there exists a δ1 P p0, δq such that whenever z P Dpz0 , δ1 q, we have
r i.e., f pzq P Dpf pz0 q, δq. So for all z P Dpz0 , δ1 q, |r
|f pzq ´ f pz0 q| ă δ, hg pf pzqq| ă ǫ.
The claim follows from Lemma 2.1.
d 1
Example 2.4. From Exercise 2.5, dz z “ ´ z12 in Czt0u.
i
f g
í
Uf “ Cztí, iu Ug “ Czt0u C
With f pzq “ 1`z 2 for z P Uf “ Cztí, iu, and gpwq “ w1 for w P Ug “ Czt0u, we have
1 1 2z
f pUf q Ă Ug . By the Chain rule, ddz 1`z 2 “ ´ p1`z 2 q2 2z “ ´ p1`z 2 q2 in Cztí, iu.
Exercise 2.11. Assuming that exp is entire and that exp1 z “ exp z (shown later), prove
that z ÞÑ expp´ 1`z
1´z q is holomorphic in D :“ tz P C : |z| ă 1u, and find its derivative.
2.2. Cauchy-Riemann equations

We now prove the main result in this chapter, which says roughly that a function
f “ u ` iv is holomorphic if and only if its real and imaginary parts u, v (viewed
as real-valued functions living in an open subset of R2 ) satisfy a pair of partial
differential equations, called the Cauchy-Riemann equations.
Let U be an open subset of C, and let f : U Ñ C be a function. Then taking
any point px, yq P U , we have f px ` iyq P C, and we can look at the real part upx, yq
of f px ` iyq, and the imaginary part vpx, yq of f px ` iyq.
U
f vpx, yq f px, yq
px, yq upx, yq
If one changes the point px, yq, then f pxìyq changes, and so do upx, yq and vpx, yq.
In this manner, associated with f , we obtain two real-valued functions
u : U Ñ R, U Q px, yq ÞÑ Repf px ` iyqq “: upx, yq,
v : U Ñ R, U Q px, yq ÞÑ Impf px ` iyqq “: vpx, yq.
2.2. Cauchy-Riemann equations 35
Our first result in this section is the necessity of the Cauchy-Riemann equations
for complex differentiability, and we will prove this result in Theorem 2.4 below.
The result says that if f is complex differentiable at z0 “ px0 , y0 q P U , then
Bu Bv Bu Bv
Bx px0 , y0 q “ By px0 , y0 q and By px0 , y0 q “ ´ Bx px0 , y0 q ,
and these two equations are called the Cauchy-Riemann equations. If these equa-
tions are not satisfied by the real and imaginary part of some complex-valued
function at some z0 , then the function cannot be complex differentiable at z0 . Let
us revisit Example 2.2. We saw (by ‘brute force’, using the ǫ-δ definition of complex
differentiability), that z ÞÑ z is not complex differentiable anywhere in the complex
plane. Now we will use the Cauchy-Riemann equations to show the same result.
Example 2.5. Define g : C Ñ C by gpzq “ z for all z P C. Writing z “ x ` iy,
where x, y P R, we have that upx, yq “ Repgpx ` iyqq “ Repx ´ iyq “ x and
vpx, yq “ Impgpx ` iyqq “ Impx ´ iyq “ ý. Thus Bu Bv
Bx px, yq “ 1 ‰ ´ 1 “ By px, yq,
showing that the Cauchy-Riemann equations cannot hold at any point in C. So we
recover our previous observation that g is complex differentiable nowhere.
Before proving the necessity of the Cauchy-Riemann equations for complex differ-
entiability, let us also mention the second important result we will show in this
section, namely the sufficiency of the Cauchy-Riemann equations for complex dif-
ferentiability. Suppose that U is an open subset of C, z0 “ px0 , y0 q P U , the real
and imaginary parts u, v of f : U Ñ C are real differentiable at px0 , y0 q, and the
Cauchy-Riemann equations are satisfied by u and v at px0 , y0 q. Then f is complex
differentiable at z0 , and moreover, f 1 pz0 q “ Bu Bv
Bx px0 , y0 q ` i Bx px0 , y0 q. This important
result will enable us to establish the holomorphicity of important functions without
having to go through the rigmarole of verifying the ǫ-δ definition. As an example,
we revisit Example 2.1.
Example 2.6. Define f : C Ñ C by f pzq “ z 2 for all z P C. Writing z “ x ` iy,
where x, y P R, we have that upx, yq “ Repf pxìyqq “ Repx2 ý 2 `2xyiq “ x2 ý 2 ,
and vpx, yq “ Impf px ` iyqq “ Impx2 ´ y 2 ` 2xyiq “ 2xy. As u, v are polynomials,
they are real differentiable, and moreover the partial derivatives satisfy the Cauchy-
Riemann equations: Bu Bv Bu Bv
Bx px, yq “ 2x “ By px, yq, and By px, yq “ ´2y “ ´ Bx px, yq.
We recover that f is entire and f 1 pzq “ Bu Bv
Bx px, yqì Bx px, yq “ 2x`2yi “ 2z in C.
Theorem 2.4 (C-R necessity). Let U Ă C be open, z0 “ px0 , y0 q P U, f : U Ñ C.
Define u, v : U Ñ R by upx, yq “ Re f pxìyq and vpx, yq “ Im f pxìyq, px, yq P U .
If f is complex differentiable at z0 , then u, v are real differentiable at px0 , y0 q, and
Bu Bv Bu Bv
Bx px0 , y0 q “ By px0 , y0 q and By px0 , y0 q “ ´ Bx px0 , y0 q.
Moreover, f 1 pz0 q “ Bu Bv
Bx px0 , y0 q ` i Bx px0 , y0 q.
Proof. (The idea of the proof is that we let px, yq tend to px0 , y0 q first keeping y
fixed at y0 , and then keeping x fixed at x0 , and look at what this gives us.)
px0 , yq
px0 , y0 q px, y0 q
Let ǫ ą 0. Then there is δ ą 0 for all z P U satisfying 0 ă |z ´ z0 | ă δ,

| f pzq´f
z´z0
pz0 q
´ f 1 pz0 q| ă ǫ. p‹q
Bu
Step 1. We will show that Bx px0 , y0 q exists and equals Re f 1 pz0 q.
Let z :“ x ` iy0 , where x P R is such that 0 ă |x ´ x0 | ă δ. Then z ´ z0 “ x ´ x0 ,
and so 0 ă |z ´ z0 | “ |x ´ x0 | ă δ. Thus
qúpx0 ,y0 q
| upx,y0x´x 0
´ Re f 1 pz0 q| “ |Rep f pxìy0x´x
q´f px0 ìy0 q
0
q ´ Re f 1 pz0 q|
“ |Rep f pzq´f
z´z0
pz0 q
q ´ Re f 1 pz0 q|
ď | f pzq´f
z´z0
pz0 q
´ f 1 pz0 q| ă ǫ,
Bu upx,y0 qúpx0 ,y0 q
using (‹). Thus the partial derivative Bx px0 , y0 q “ lim x´x0 “ Re f 1 pz0 q.
xÑx0
Bv
Step 2. We show that Bx px0 , y0 q “ Im f 1 pz0 q.
Proceeding in a manner similar to Step 1, we have with the same notation that
q´vpx0 ,y0 q
| vpx,y0x´x 0
´ Im f 1 pz0 q| “ |Imp f pxìy0x´x
q´f px0 ìy0 q
0
q ´ Im f 1 pz0 q|
“ |Imp f pzq´f
z´z0
pz0 q
q ´ Im f 1 pz0 q|
ď | f pzq´f
z´z0
pz0 q
´ f 1 pz0 q| ă ǫ.
q´vpx0 ,y0 q
So Bv
lim vpx,y0x´x
Bx px0 , y0 q “ xÑx 0
“ Im f 1 pz0 q. Thus f 1pz0 q “ Bu Bv
Bx px0 , y0 qì Bx px0 , y0 q.
0
Bu
Step 3. We show that By px0 , y0 q “ Ím f 1 pz0 q.
Now let z :“ x0 ìy, where y P R is such that 0 ă |yý0 | ă δ. Then z´z0 “ ipyý0 q
and so 0 ă |z ´ z0 | “ |y ´ y0 | ă δ. For a, b P R, Repa ` ibq “ Impipa ` ibqq. Hence
0 ,y0 q
| upx0 ,yqúpx
yý0 ` Im f 1 pz0 q| “ |Imp ipf pzq´f
yý0
pz0 qq
q ` Im f 1 pz0 qq|
“ |Imp´ f pzq´f
z´z0
pz0 q
` f 1 pz0 qq|
ď | f pzq´f
z´z0
pz0 q
´ f 1 pz0 q| ă ǫ.
Bu upx0 ,yqúpx0 ,y0 q
Thus the partial derivative By px0 , y0 q “ lim yý0 “ Ím f 1 pz0 q.
yÑy0
Bv
Recall from Step 2 that Bx px0 , y0 q “ Im f 1 pz0 q, and so, we have established one of
the two Cauchy-Riemann equations, namely Bu Bv
By px0 , y0 q “ ´ Bx px0 , y0 q.
Bv
Step 4. We show that By px0 , y0 q “ Re f 1 pz0 q.
We proceed as in Step 3. For a, b P R, Impa ` ibq “ ´Repipa ` ibqq. Hence
0 ,y0 q
| vpx0 ,yq´vpx
yý0 ´ Re f 1 pz0 q| “ | ´ Repi f pzq´f
yý0
pz0 q
q ´ Re f 1 pz0 q|
ď | ´ i f pzq´f
yý0
pz0 q
´ f 1 pz0 q| “ | f pzq´f
z´z0
pz0 q
´ f 1 pz0 q| ă ǫ.
0 ,y0 q
Thus Bv
lim vpx0 ,yq´vpx
By px0 , y0 q “ yÑy yý0 “ Re f 1 pz0 q. From Steps 1 and 4, we obtain
0
Bu Bv
Bx px0 , y0 q “ By px0 , y0 q. So we have got both Cauchy-Riemann equations.
Finally, we show that u, v are real differentiable (as real-valued functions of two
real variables) at px0 , y0 q. For z “ px, yq satisfying 0 ă |z ´ z0 | ă δ,
“ ‰
|upx, yq ´ upx0 , y0 q ´ r Bu
Bx
px0 ,y0 q Bu
By
px0 ,y0 q s x´x
yý0 |
0
}px, yq ´ px0 , y0 q}2

|upx, yq ´ upx0 , y0 q ´ p Bu Bv
Bx px0 , y0 qqpx ´ x0 q ` p Bx px0 , y0 qqpy ´ y0 q|
“
}px, yq ´ px0 , y0 q}2
|Repf pzq ´ f pz0 q ´ f 1 pz0 qpz ´ z0 qq|
“ ă ǫ.
|z ´ z0 |
So u is real differentiable at px0 , y0 q. Similarly, v is real differentiable at px0 , y0 q.
We will see later on that in fact the real and imaginary parts of a holomorphic
function are infinitely many times real differentiable.
Exercise 2.12. Consider Exercise 2.3 again. Show that f is not complex differentiable at
any point of the open set Czt0u.
One also has the following converse to Theorem 2.4. This is a very useful result to
check the holomorphicity of functions.
Theorem 2.5 (C-R sufficiency). Let U Ă C be open, z0 “ px0 , y0 q P U, f : U Ñ C.
Define u, v : U Ñ R by upx, yq “ Re f pxìyq and vpx, yq “ Im f pxìyq, px, yq P U .
If u, v are real differentiable at px0 , y0 q,
Bu Bv Bu Bv
Bx px0 , y0 q “ By px0 , y0 q, and By px0 , y0 q “ ´ Bx px0 , y0 q,
then f is complex differentiable at z0 , and f 1 pz0 q “ Bu Bv
Bx px0 , y0 q ` i Bx px0 , y0 q.
Proof. Let A :“ Bu Bv Bu Bv
Bx px0 , y0 q “ By px0 , y0 q, B :“ ´ By px0 , y0 q “ Bx px0 , y0 q. Let ǫ ą 0.
As u is real differentiable at px0 , y0 q, there exists δ1 ą 0 such that if px, yq P U
satisfies 0 ă }px, yq ´ px0 , y0 q}2 ă δ1 , then
“ 0‰ “ 0‰
|upx, yqúpx0 , y0 q´rA ´B s x´x yý0 | |upx, yqúpx0 , y0 q´r BuBx
px0 ,y0 q Bu
By
px0 ,y0 q s x´x
yý0 | ǫ
“ ă .
}px, yq ´ px0 , y0 q}2 }px, yq ´ px0 , y0 q}2 2
Similarly, let δ2 ą 0 be such that if px, yq P U satisfies 0 ă }px, yq´px0 , y0 q}2 ă δ2 ,
“ 0‰ “ 0‰
|vpx, yq´vpx0 , y0 q´rB As x´xyý0 |
Bv
|vpx, yq´vpx0 , y0 q´r Bx px0 ,y0 q Bv
By
px0 ,y0 q s x´x
yý0 | ǫ
“ ă .
}px, yq ´ px0 , y0 q}2 }px, yq ´ px0 , y0 q}2 2
Let z0 :“ x0 ` iy0 . Set δ :“ mintδ1 , δ2 u ą 0. If z “ x ` iy (x, y P R), satisfies

0 ă |z ´ z0 | ă δ, then we have
ˇ f pzq´f pz q ´ Bu Bv ¯ˇ |f pzq´f pz q´pAìBqpz´z q|
ˇ 0 ˇ 0 0
ˇ ´ px0 , y0 q ` i px0 , y0 q ˇ “
z ´ z0 Bx Bx |z ´ z0 |
|Repf pzq´f pz0 q´pAìBqpz´z0 qq| |Impf pzq´f pz0 q´pAìBqpz´z0 qq|
ď `
|z ´ z0 | |z ´ z0 |
|upx, yqúpx0 , y0 q´pApx´x0 q´Bpyý0 qq| |pvpx, yq´vpx0 , y0 q´pBpx´x0 qÀpyý0 qq|
“ `
|px ´ x0 q ` ipy ´ y0 q| |px ´ x0 q ` ipy ´ y0 q|
“ x´x0 ‰ “ ‰
|upx, yq ´ upx0 , y0 q ´ r A ´B s yý0 | |vpx, yq ´ vpx0 , y0 q ´ r B A s x´x
yý0 |
0
“ `
}px, yq ´ px0 , y0 q}2 }px, yq ´ px0 , y0 q}2
ǫ ǫ
ă ` “ ǫ.
2 2
Bu Bv
Thus f is complex differentiable at z0 and f 1 pz0 q “ Bx px0 , y0 q ` i Bx px0 , y0 q.
If u, v : U Ñ R have the partial derivatives Bu Bu Bv Bv

Bx , By , Bx , By in U , which moreover
are continuous on U , then u, v are real differentiable in U .
Example 2.7 (exp, sin, cos are entire). For x, y P R,
upx, yq “ Re exppx ` iyq “ Repex pcos y ` i sin yqq “ ex cos y,
vpx, yq “ Im exppx ` iyq “ Impex pcos y ` i sin yqq “ ex sin y.
Thus Bu x Bv Bu x Bv
Bx px, yq “ e cos y “ By px, yq, and By px, yq “ é sin y “ ´ Bx px, yq. So the
partial derivatives are continuous, and the Cauchy-Riemann equations hold in C.
Thus exp is entire, and exp1 z “ Bu Bv x x
Bx px, yq ` i Bx px, yq “ e cos y ` ie sin y “ exp z.
iz íz iz íz
By Propositions 2.3 and 2.2, also sin z “ e é 2i and cos z “ e è2 are entire,
d ieiz ´píqeíz eiz èíz d ieiz `píqeíz
dz sin z “ 2i “ 2 “ cos z, and dz cos z “ 2 “ ´ sin z.
x ý
Example 2.8. Let u, v be defined by upx, yq “ x2 `y 2
, vpx, yq “ x2 `y 2
for all
px, yq P R2 ztp0, 0qu. Then we have
Bu 1px2 ` y 2 q ´ x2x y 2 ´ x2
“ 2 2 2
“ 2 ,
Bx px ` y q px ` y 2 q2
Bu ´x2y ´2xy
“ 2 “ 2 ,
By px ` y 2 q2 px ` y 2 q2
Bv 2xy Bv y 2 ´ x2
“ 2 2 2
, and “ 2 .
Bx px ` y q By px ` y 2 q2
Clearly px, yq ÞÑ px2 ` y 2 q2 , y 2 ´ x2 , ˘2xy are continuous in R2 and px2 ` y 2 q2

is pointwise nonzero in R2 ztp0, 0qu. So each of the above partial derivatives is
continuous in R2 ztp0, 0qu. Thus u, v are real differentiable in R2 ztp0, 0qu. Also the
Cauchy-Riemann equations hold. Hence f :“ u ` iv is holomorphic in Czt0u.
In fact, f is inversion, Czt0u Q z ÞÑ 1z : For px, yq P R2 ztp0, 0qu, setting z “ x ` iy,
x ý xíy z z 1
we have f pzq “ upx, yq ` ivpx, yq “ x2 `y 2 ` i x2 `y 2 “ x2 `y 2 “ |z|2 “ zz “ z .
Example 2.9 (Holomorphicity of Log in Czp´8, 0s). The principal logarithm

is holomorphic in the open set3 Czp´8, 0s. This can be done using the Cauchy-
Riemann equations; see Exercise 2.13. A different argument is given below.
The principal logarithm is defined in the bigger set Czt0u, but it is not contin-
uous in this bigger set (because at each negative real number, it is discontinuous).
In the smaller set Czp´8, 0s, the principal logarithm is continuous. We will now
use this continuity to show that Log is in fact holomorphic in Czp´8, 0s, and that
d 1
dz Log z “ z for z P Czp´8, 0s. If z, z0 P Czp´8, 0s are distinct, then Log z ‰ Log z0
(otherwise z “ exppLog zq “ exppLog z0 q “ z0 , a contradiction). Let ǫ ą 0. Set
2
ǫ1 :“ mint |z20 | , |z02| ǫu ą 0. Since exp is complex differentiable at w0 :“ Log z0 , there
is a δ1 ą 0 such that whenever w satisfies 0 ă |w ´ w0 | “ |w ´ Log z0 | ă δ1 ,
| exp w´w
wéxp w0
0
exp w´z0
´ exp w0 | “ | w´Log z0 ´ z0 | ă ǫ1 . But by the continuity and injectivity
of Log in Czp´8, 0s, there exists a δ ą 0 such that whenever 0 ă |z ´ z0 | ă δ, we
have 0 ă |Log z ´ Log z0 | ă δ1 . Thus with w :“ Log z, and 0 ă |z ´ z0 | ă δ, we have
0 ă |w ´ w0 | ă δ1 , and so | Log z´z z´Log z0 ´ z0 | ă ǫ1 . By the reverse triangle inequality,
0
|z0 |
|z0 | ´ | Log z´Log z0 | ď | Log z´Log z0 ´ z0 | ă ǫ1 , and so | Log z´z
z´z0 z´z0
z´Log z0 | ě |z0 | ´ ǫ1 ě 2 . Thus
0
whenever 0 ă |z ´ z0 | ă δ, we have
| Log z´z
z´Log z0
0
´ 1
z0 | “ |pz0 ´ z´z0
Log z´Log z0 q
1
z´z0
1
z0 |
Log z´Log z0
z´z0 1 1 1 1
“ |z0 ´ Log z´Log z0 | | z´z0 |z0 | ă ǫ1 |z0 | |z0 | ď ǫ.
Log z´Log z0
| 2
Thus Log is holomorphic in Czp´8, 0s and moreover, d

dz Log z “ z1 .
´1
Exercise 2.13. Recall that cos´1 : p´1, 1q Ñ p0, πq has the derivative d
dy cos´1 y “ a ,
1 ´ y2
1
and that sin´1 : p´1, 1q Ñ p´ π2 , π2 q has the derivative d
dy sin´1 y “ a .
1 ´ y2
$ x
’
’ cos´1 a if y ą 0,
’
’ x2 ` y 2
’
& y
For z “ x ` iy P Czp´8, 0s, where x, y P R, Arg z “ sin´1 a if x ą 0,
’
’ x2 ` y 2
’
’ ´1 x
’
%´ cos a 2 if y ă 0.
x ` y2
Use the Cauchy-Riemann equations to show that Log is holomorphic in Czp´8, 0s, and
find its complex derivative.
Exercise 2.14. Let U, V Ă C be open, f : U Ñ V , g : V Ñ U be continuous maps such
that f ˝ g “ idV and g ˝ f “ idU . (For a set X, idX denotes the identity map X Q x ÞÑ x.)
Show that if f is complex differentiable at z0 P U and f 1 pz0 q ‰ 0, then g is complex
1
differentiable at f pz0 q and g 1 pf pz0 qq “ f 1 pz 0q
. Use this to show Log1 z “ z1 , z P Czp´8, 0s.
The Cauchy-Riemann equations can also be used to prove some interesting facts, for
example the following one, which highlights the ‘rigidity’ of holomorphic functions
alluded to earlier. We will use this later to prove the ‘Maximum Modulus Theorem’
(Theorem 4.14).
3Here the ‘interval notation’ p´8, 0s’ actually means the set p´8, 0s ˆ t0u in the complex plane, i.e., the
ray in the complex plane comprising the nonpositive reals.
Example 2.10 (f P OpDpz0 , rqq with constant |f | is constant). Let z0 P C and

r ą 0. We will show using the Cauchy-Riemann equations that if f P OpDpz0 , rqq
and there exists a c P R such that |f pzq| “ c for all z P Dpz0 , rq, then f is constant.
r f
z0 0
With u :“ Re f and v :“ Im f , c2 “ |f |2 “ u2 ` v 2 . Differentiating, u Bu Bv

Bx ` v Bx “ 0
Bu Bv Bv Bu Bv Bu
and u By `v By “ 0. Using Bx “ ´ By in the first equation and By “ Bx in the second,
u Bu Bu
Bx ´ v By “ 0, p˚q
u Bu
By ` v Bu
Bx “ 0. p˚˚q
Bu
For eliminating By , u(˚)`v(˚˚) gives: pu2 ` v 2 q Bu 2 Bu
Bx “ 0, i.e., c Bx “ 0.
Bu
For eliminating Bx , ´v(˚)ù(˚˚) gives: pu2 ` v 2 q Bu 2 Bu
By “ 0, i.e., c By “ 0.
(If c “ 0, then u2 ` v 2 “ c2 “ 0, and so u “ v “ 0, giving f “ 0 in Dpz0 , rq.)

If c ‰ 0, then from the above Bu Bu
Bx “ By “ 0. The Cauchy-Riemann equations
Bv Bv
yield Bx “ By “ 0. By the Mean Value Theorem, for x ‰ x0 , and y ‰ y0 , there
exists a ξ in the open interval with endpoints x, x0 , and there exists an η in the
qúpx0 ,y0 q
open interval with endpoints y, y0 such that upx,y0x´x 0
“ Bu
Bx pξ, y0 q “ 0, and
upx,yqúpx,y0 q
yý0 “ Bu
By px, ηq “ 0. So upx, yq “ upx0 , y0 q for all px, yq P Dpz0 , rq. Thus u
is constant in Dpz0 , rq.
px, yq
px0 , y0 q
px, y0 q
Similarly, v is constant in D. Consequently, f “ u ` iv is constant in D.

3
Exercise 2.15. Show that z ÞÑ z is entire using the Cauchy-Riemann equations.
Exercise 2.16. Show that z ÞÑ Re z is complex differentiable nowhere.
Exercise 2.17. Let D Ă C be a domain. Show that if f : D Ñ C is holomorphic in D,
with the property that f pzq P R for all z P D, then f is constant in D.
Exercise 2.18. Let D Ă C be a domain. Show that if f : D Ñ C is holomorphic in D,
with the property that f 1 pzq “ 0 for all z P D, then f is constant in D.
Exercise 2.19. Let f : C Ñ C be entire. Set u :“ Re f , v :“ Im f . Show that if there
exists a real differentiable h : R Ñ R such that u “ h ˝ v, then f is constant in C.
Exercise 2.20. Let k P R be a fixed, and let f be defined by f pzq “ x2 ´ y 2 ` kxyi for
z “ x ` iy, x, y P R. Show that f is entire if and only if k “ 2.
2.3. Geometric meaning of the complex derivative 41
2.3. Geometric meaning of the complex derivative

Recall the geometric meaning of the derivative f 1 px0 q of a function f : R Ñ R at a
point x0 P R: f 1 px0 q is the slope of the tangent to the graph of f at x0 .
q
f
1 px 0
f
p e“
slo
x0
f pxq´f px0 q
lim x´x0 “ f 1 px0 q implies f pxq´f
x´x0
px0 q
« f 1 px0 q for x near x0 , i.e.,
xÑx0
f pxq ´ f px0 q « f 1 px0 qpx ´ x0 q.

This means that locally around x0 , f pxq ´ f px0 q looks like the action of the linear
map h ÞÑ f 1 px0 qh : R Ñ R on x ´ x0 . Visually this means that near x0 , there is
very little difference between the (tangent) line with slope f 1 px0 q passing through
px0 , f px0 qq and the graph of f . That is, if we zoom into the graph of the function
around the point px0 , f px0 qq, then the graph looks like a straight line.
We pose an analogous question for a complex-valued function map f : U Ñ C
defined on an open set U , that happens to be complex differentiable at a point z0 :
What is the geometric meaning of the complex number f 1 pz0 q?
We cannot draw a graph of f , because z as well as f pzq belong to C “ R2 , and
so pz, f pzqq would be a point in R2 ˆ R2 “ R4 . But we can draw a copy of U in
the plane on the left-hand side, and a copy of C on the right-hand side, with f
mapping points from U on the left to points on the right, as shown below.
U C
f f pzq
z “ px, yq
We will show that the complex number f 1 pz0 q describes the action of the complex
differentiable function locally infinitesimally around z0 by an anticlockwise rotation
through the angle Arg f 1 pz0 q, along with a scaling/magnification by |f 1 pz0 q|.
f pzq´f pz0 q f pzq´f pz0 q
lim z´z0 “ f 1 pz0 q, implies z´z0 « f 1 pz0 q for z near z0 , i.e.,
zÑz0
f pzq ´ f pz0 q « f 1 pz0 qpz ´ z0 q.

But from the geometric meaning of complex multiplication, when we multiply

z ´ z0 by f 1 pz0 q, z ´ z0 gets rotated anticlockwise through the angle Arg f 1 pz0 q,
and the length of z ´ z0 gets multiplied by the length of f 1 pz0 q, namely we get a
magnification by the factor |f 1 pz0 q|. See the picture below.
f pzq
f pr
zq
zr
z f
z0 f pz0 q
?
Suppose that f 1 pz0 q “ 3 ` i, so that |f 1 pz0 q| “ 2 and Argpf 1 pz0 qq “ π6 . First look
at z ´ z0 shown in the domain U as the solid line segment between z and z0 . On
the right-hand side, we have shown a translated version of this line segment as a
dashed line, emanating from f pz0 q. In order to find out where f pzq is, we just use
the fact that f pzq ´ f pz0 q is approximately equal to f 1 pz0 q multiplied by z ´ z0 . So
the solid line joining f pz0 q to f pzq on the right-hand side is obtained by rotating
the rightmost dashed line anticlockwise through an angle of Arg f 1 pz0 q (assumed to
be 30˝ in this picture), and magnifying the length of the dashed line by |f 1 pz0 q| “ 2.
If want to find the image of another point zr near z0 , we repeat the same procedure.
Namely we first look at the line segment joining z0 to zr, which is the solid line on
the left. We have shown a translated version of this as a dashed line, emanating
from f pz0 q, in the picture on right-hand side. To find the location of f pr z q, we
first rotate the leftmost dashed line anticlockwise through the argument of f 1 pz0 q,
that is 30˝ , and magnify the length of the leftmost dashed line by |f 1 pz0 q| “ 2. In
this manner, we obtain the solid line on the right-hand side joining f pz0 q to f pr zq
(approximately!). So, locally, the action of f is as follows. Imagine the domain
as a rubber sheet, and look at a point z0 on this rubber sheet. Tear out a small
portion of this rubber sheet around z0 . Then f takes the point z0 on this rubber
sheet to a point f pz0 q somewhere in the complex plane. If we want to know how
the rest of the points on our little torn rubber sheet are mapped by f , one follows
this procedure. We place our rubber sheet such that z0 on our rubber sheet is lying
over the point f pz0 q in the complex plane. (Imagine pinning it on the plane with
the pin passing through the point marked z0 on our little torn rubber sheet.) Then
we stretch out our rubber sheet about the point z0 by a factor of |f 1 pz0 q|, and then
rotate this stretched rubber sheet anticlockwise by an angle of Arg f 1 pz0 q around
the point z0 on the rubber sheet. The images of the points on the original little
torn rubber sheet neighbourhood of z0 in the domain are then approximately given
by their corresponding new positions on the stretched and rotated rubber sheet
pinned at f pz0 q.
In order to stress this geometric interpretation, let us revisit Example 2.1 yet
again, where we considered the squaring map z ÞÑ z 2 .
Example 2.11. Suppose we assume complex differentiability of the squaring map

f , z ÞÑ z 2 , at a point z0 P C. Let us then show that the complex derivative of
the squaring map at z0 must be 2z0 by figuring out geometrically the approximate
local amplification and rotation produced by the squaring map at z0 .
We first ask: What is the local rotation produced? To find this out, consider
a point z close to z0 along the ray joining 0 to z0 . See the picture below, which
shows the effect of the squaring map: The angle is doubled, and the distance to 0
is squared. Thus z 2 lies on the ray joining 0 to z02 , which makes an angle 2Arg z0
with the positive real axis. Hence the line segment joining z02 to z 2 is obtained by
rotating the line segment joining z0 to z anticlockwise through the angle Arg z0 .
Consequently, Arg f 1 pz0 q “ Arg z0 .
z2
θ
θ
z02
z
z0 z ÞÑ z 2 θ
θ “ Arg z0 θ
0 0
Next we ask: What is the local amplification produced? To find this out, consider
a point z close to z0 which is at the same distance from 0 as z0 , but it makes a
slightly bigger angle θ ` dθ with the positive real axis. See the following picture.
Since dθ is tiny, the length |z ´ z0 | is approximately |z0 | dθ, while the length
|z 2 ´ z02 | is approximately |z0 |2 2dθ. So the magnification factor |f 1 pz0 q| is equal to
|z0 |2 2dθ
|z0 | dθ “ 2|z0 |.
z2
z02
|z0 |2
2dθ
z z ÞÑ z 2
|z| dθ z0 θ
θ “ Arg z0 θ
0 0
So f 1pz0 q “ |f 1 pz0 q|pcos Arg f 1 pz0 qì sin Arg f 1 pz0 qq “ 2|z0 |pcos Arg z0 ì sin Arg z0 q “ 2z0 .
Thus by investigating the local behaviour of the squaring map f near z0 , we could
find out the complex derivative f 1 pz0 q.
Example 2.12 (Complex conjugation is complex differentiable nowhere).
z
z0
z ÞÑ z
R R
z0
z
Suppose that z ÞÑ z is complex differentiable at z0 . Then the local behaviour of

the map around z0 should be a rotation followed by an amplification. Consider
the point z near z0 obtained by a tiny horizontal translation. From the picture,
by looking at the images z0 and z, we see that since z ´ z0 “ z ´ z0 , no rotation
is produced. On the other hand, if we look at w which is near z0 obtained by a
tiny vertical displacement, then from the picture, we see that w ´ z0 “ ´pz ´ z0 q,
and so there is a rotation through 180˝ . But this means that locally the map is
not a rotation (because if it were, all infinitesimal vectors emanating at z0 would
be rotated by a same fixed amount).
Exercise 2.21. We know that the power function z ÞÑ z n , n P N, is entire. Find its
complex derivative by investigating its local behaviour.
Exercise 2.22. We know that the inversion map f P OpCzt0uq, where f pzq “ z1 , z P Czt0u.
Find the complex derivative of f at z P Czt0u by investigating its local behaviour.
Exercise 2.23. We know that the exponential function z ÞÑ exp z is entire. Find its
complex derivative by investigating its local behaviour. Hint: Displace a point z0 vertically
(respectively horizontally) to find the local amplification (respectively rotation).
Exercise 2.24. We know that Log P OpCzp´8, 0sq. Find the complex derivative of Log
at a point w P Czp´8, 0s by investigating its local behaviour.
Exercise 2.25. Give a visual argument to show that the map z ÞÑ Re z is not complex
differentiable anywhere in C.
Exercise 2.26. In Exercise 1.55, show that the limiting value we calculated there equals
the number | exp1 pa ` i0q|2 . Is this expected?
Conformality. The picture on page 23 shows the action of the entire mapping
exp. Just as in the domain, the images of the vertical and horizontal lines under exp
are mutually perpendicular. This is a manifestation of the ‘conformality’ property
of holomorphic functions, that is, of the preservation, under the mapping action,
of the angles between curves in the domain. Let us now see why holomorphic
functions possess this property, based on what we have learnt about the local
action of complex differentiable functions.
T2 ϕ
f θ
θ ϕ
p T1
f ppq
Let f : U Ñ C be holomorphic in U . Imagine two smooth curves intersecting

at a point p P U . Since the curves are smooth, they have tangents at p, say T1
and T2 . Near the point p, there is very little difference between the curve and its
tangent line at p, so we may assume that the curves are replaced by their tangent
lines. These tangent lines make a certain angle. Now let us look at what f does to
these lines. Each of the curves is mapped to new curves in C by f intersecting at
f ppq, and these new curves are smooth too, possessing tangent lines at f ppq. But
since the infinitesimal local action of f around p is rotation counterclockwise by
θ :“ Arg f 1 ppq followed by magnification, the new tangent lines are obtained by just
rotating counterclockwise the old tangent lines, and magnifying the image. Thus it
is obvious that the angle will be the same. Hence the conformality of holomorphic
maps is not a mystery anymore.
Complex versus real differentiability. Let U Ă C be open, and f : U Ñ C

be complex differentiable at z0 “ px0 , y0 q P U . If u :“ Re f , v :“ Im f , then
u, v : U Ñ R are real differentiable at px0 , y0 q, and the Cauchy-Riemann equa-
tions hold, i.e., a :“ Bu Bv Bu Bv
Bx px0 , y0 q “ By px0 , y0 q and b :“ ´ By px0 , y0 q “ Bx px0 , y0 q. Thus
f
px, yq ÞÑ pupx, yq, vpx, yqq : U Ñ R2 is real differentiable, and its real derivative is
the linear transformation R2 Q x ÞÑ Ax P R2 , where the matrix A is given by
« Bu Bu
ff „ 
Bx px0 , y0 q By px0 , y0 q a ´b
A :“ Bv “ .
Bv b a
Bx px0 , y0 q By px0 , y0 q
?
Set r :“ |f 1 pz0 q| “ | Bu Bv
Bx px0 , y0 q ` i Bx px0 , y0 q| “ |a ` ib| “ a2 ` b2 . If θ :“ Arg f 1 pz0 q,
Bu 1
then a “ Bx px0 , y0 q “ Re f pz0 q “ r cos θ, and similarly, b “ r sin θ. Thus the real
derivative is the linear transformation
„  „ 
a ´b cos θ ´ sin θ
A“ “r .
b a sin θ cos θ
But we recognise this as the linear transformation describing a counterclockwise
rotation about the origin through θ, followed by a dilation/magnification by r.
“ θ ´ sin θ ‰
Exercise 2.27. Show that if θ P R, and R “ cossin θ
2
cos θ , then the map R Q x ÞÑ Rx P R
2
is a counterclockwise rotation about the origin through θ.
So, if f is complex differentiable at a point z0 , then it is real differentiable (as a map

from U Ă R2 to R2 ), but what distinguishes complex differentiability from mere
real differentiability is that the real derivative for a complex differentiable mapping
is not just any linear transformation, but a special one: It is a counterclockwise
rotation through an angle θ P r0, 2πq, followed by a scaling by r ě 0.
Exercise 2.28 (The BBz -operator). The two Cauchy-Riemann equations can be written as
B
a single equation via the operator Bz . Let U Ă C be open. If ϕ : U Ñ R is real differentiable
Bϕ
at z0 P U , then Bz pz0 q :“ 2 p Bx pz0 q ´ i Bϕ
1 Bϕ Bϕ 1 Bϕ Bϕ
By pz0 qq and Bz pz0 q :“ 2 p Bx pz0 q ` i By pz0 qq. Suppose
f : U Ñ C is such that u :“ Re f , v :“ Im f are real-differentiable at z0 P U . Then define
Bf Bu Bv Bf Bu Bv
Bz pz0 q :“ Bz pz0 q ` i Bz pz0 q and Bz pz0 q :“ Bz pz0 q ` i Bz pz0 q.
Bf
(1) Show that f is complex differentiable at z0 if and only if Bz pz0 q “ 0.
Bf 1
(2) Show that if f is complex differentiable at z0 , then Bz 0 “ f pz0 q.
pz q
2
Bz B|z|
(3) Determine Bz and Bz .
Bz 2 Bz 2
(4) Determine Bz and Bz .
So holomorphic functions are thought of as ‘functions of z, z that are independent of z’.
Chapter 3
Cauchy Integral Theorem
Having become familiar with complex differentiation, we now turn to integration.

We will learn an important result called the Cauchy Integral Theorem, after intro-
ducing the ‘contour integral’. The Cauchy Integral Theorem is important because
it will lead to a deeper understanding of holomorphic functions. For example, using
this, we will show the fact that holomorphic functions are infinitely many times
complex differentiable. In this chapter we will learn the following:
‚ The contour integral and its properties.
‚ The Fundamental Theorem of Contour Integration.
‚ The Cauchy Integral Theorem.
‚ The following consequences of the Cauchy Integral Theorem:
(a) Existence of a primitive,
(b) Infinite complex differentiability of holomorphic functions,
(c) Liouville’s Theorem and the Fundamental Theorem of Algebra,
(d) Morera’s Theorem.
3.1. Definition of the contour integral
şb
In ordinary calculus, given a continuous function f : ra, bs Ñ R, a f pxqdx has a
clear meaning. Now suppose we wish to generalise this in the complex şw setting:
Given z, w complex numbers, want to give meaning to something like z f pζq dζ.
Then a first question is: How do we get from z to w? In R, if a ă b, then there
is just one way of going from the real number a to the real number b, and so our
starting data in the real case is: a, b P R, and a continuous function f : ra, bs Ñ R.
C
w
a b ?
R
z
But now z and w are points in the complex plane, and so there are many possible
connecting paths along which we could integrate.
47
48 3. Cauchy Integral Theorem
So in the complex setting, besides specifying the end points z and w, we also specify
şb
the path γ taken to go from z to w, and we will replace ş the integral a f pxqdx in
the real case by an expression which looks like this: γ f pzqdz. We call such an
expression a ‘contour’ integral, for the computation of which we need the following
data:
‚ A domain D (Ă C), and z, w P D.
‚ A continuous function f : D Ñ C.
‚ A smooth path γ : ra, bs Ñ D joining z to w.
Note that we need not merely a path, but a smooth path, defined below. A path γ :
ra, bs Ñ D is a continuous function. Decompose γ into its real and imaginary parts:
γptq “ xptq` iyptq for all t P ra, bs, where x, y : ra, bs Ñ R. Then x, y are continuous
functions. The path γ is smooth if x, y are continuously real differentiable (that is,
the derivatives x1 , y 1 : ra, bs Ñ R exist, and they are also continuous).
Example 3.1. Define γ : r0, 1s Ñ C by γptq “ tp1 ` iq, t P r0, 1s. The real and
imaginary parts x, y : r0, 1s Ñ R of γ are given by xptq “ t, yptq “ t for all t P r0, 1s.
Since x, y are continuously real differentiable on r0, 1s, γ is a smooth path.
1ì
Define the two paths γ1 , γ2 : r0, 2πs Ñ R by γ1 ptq “ exppitq and γ2 ptq “ expp2itq,
t P r0, 2πs. The real and imaginary parts of γ1 , γ2 are cos t, sin t, cosp2tq, sinp2tq,
each of which is continuously real differentiable, and so γ1 , γ2 are smooth paths.
γ1 γ2
0 0
Although the ranges of γ1 and γ2 are equal, i.e.,

tγ1 ptq : t P r0, 2πsu “ tγ2 ptq : t P r0, 2πsu “ tz P C : |z| “ 1u (unit circle with center 0),
γ1 and γ2 are deemed to be different paths, because the functions are not the same:
For example, γ1 pπq “ ´1 ‰ 1 “ γ2 pπq.
Remark 3.1. It is convenient to refer to the range tγptq : t P ra, bsu of a path
γ : ra, bs Ñ C as the path itself. With this usage, a path becomes a concrete
geometric object (as opposed to being a map), such as a circle or a line segment
in the complex plane and hence can be easily visualised. The difficulty with this
abuse of terminology is that several different paths can have the same image, and
so it causes ambiguity. ˚
3.1. Definition of the contour integral 49
Definition 3.1 (Contour integral). Given

‚ a domain D,
‚ a continuous function f : D Ñ C (with u :“ Re f , v :“ Im f ), and
‚ a smooth path γ : ra, bs Ñ D (with x :“ Re γ, y :“ Im γ : ra, bs Ñ R),
ş
we define the contour integral γ f pzqdz by
ş şb 1 ptqdt
γ f pzqdz :“ a f pγptqqγ
şb
:“ q 1 1
a p upγptqq ` ivpγptqqq px ptq ` iy ptqqdt
şb 1 1
şb 1
:“ q
a p upγptqq x ptq ´ vpγptqq y ptqq dt ` i a p vpγptqq x ptq ` upγptqq y 1 ptqqqdt.
The last two integrals above are the usual Riemann integrals of real-valued contin-
uous functions. The contour integral can be interpreted geometrically as follows.
The term γ 1 ptqdt “ x1 ptqdt ` iy 1 ptqdt can be viewed as an infinitesimal incremental
piece of the contour. We multiply this by the (almost constant) value f pγptqq of
f on this incremental piece. Finally, we add up all these contributions along the
şb
contour to get the total as the integral a f pγptqqγ 1 ptqdt.
D
f pγptqq
γpbq
f
γ 1 ptqdt iy 1 ptqdt
x1 ptqdt
γpaq
Example 3.2. Let D “ C, and γ : r0, 1s Ñ D be the smooth path given by

γptq “ tp1ìq for all t P r0, 1s. Then γ 1 ptq “ 1 ` i, t P r0, 1s. If f “ pz ÞÑ zq, then
ş ş1 ş1 ş1 2 2
γ f pzq dz “ 0 tp1 ` iqp1 ` iq dt “ 0 tp1 ´ iqp1 ` iq dt “ 0 tp1 ´ i q dt
ş1 ş1 2
“ 0 tp1 ` 1q dt “ 2 0 t dt “ 2 t2 |10 “ 1.
Exercise 3.1. Consider the three paths γ1 , γ2 , γ3 : r0, 2πs Ñ C defined by γ1 ptq “ exppitq,
γ2 ptq “ expp2itq, γ3 ptq “ exppítq,
ş for tş P r0,ş2πs. Show that their images are the same,
but the three contour integrals γ1 1z dz, γ2 z1 , γ3 1z dz are all different.
Exercise 3.2. Let f be holomorphic in a domain and let γ : ra, bs Ñ D be a smooth path.
d
Show that dt f pγptqq “ f 1 pγptqq γ 1 ptq for all t P ra, bs.
Exercise 3.3. Prove ş or disprove:ş For every entire function f and every smooth curve
γ : r0, 1s Ñ C, Re p γ f pzq dzq “ γ Re f pzq dz.
Equivalent paths give equal integrals. We will often assume that our smooth
paths are parametrised by r0, 1s, rather than some more general interval ra, bs. Let
us explain why we may assume this. Suppose that γ : ra, bs Ñ C and γ r : rc, ds Ñ C,
are two smooth paths, such that there is a continuously differentiable function
ϕ : rc, ds Ñ ra, bs such that a “ ϕpcq, b “ ϕpdq, and γrptq “ γpϕptqq for t P rc, ds.
We call such smooth paths γ, γr ‘equivalent’. (Imagine going from γpaq “ γ rpcq to
γpbq “ γrpdq along the same route, but with possibly different speeds.)
γpbq “ γ
rpdq
a ϕptq b γ
ϕ γ ptq “ γpϕptqq
r
γpaq “ γ
rpcq
c t d r
γ
By the chain rule, γ r1 ptq “ γ 1 pϕptqq ϕ1 ptq for all t P rc, ds, and so
ş şd şd
γr f pzq dz “ c f pr γ 1 ptq dt “ c f pγpϕptqqqγ 1 pϕptqqϕ1 ptq dt
γ ptqqr
τ “ϕptq ş b ş
“ a f pγpτ qqγ 1 pτ q dτ “ γ f pzq dz.
In particular, given any γ : ra, bs Ñ C, define ϕ : r0, 1s Ñ ra, bs by ϕptq “ p1 ´ tqa ` tb
for all t P r0, 1s. Then ϕ is smooth, and ϕp0q “ a, ϕp1q ş “ b. So with
ş c :“ 0, d :“ 1
in the above, and γ r “ γ ˝ ϕ, γr f pzqdz “ γ f pzqdz. Hence
r : r0, 1s Ñ C defined by γ
there is no loss of generality (when it comes to statements about contour integrals)
in assuming that the smooth path is parametrised by r0, 1s.
Contour integrals along piecewise smooth paths. We extend the definition

above to paths with ‘corners’. A path γ : ra, bs Ñ C is called piecewise smooth
if there exist points c0 :“ a ă c1 ă ¨ ¨ ¨ ă cn ă b “: cn`1 (for some n P N) such
that the restriction of γ to rck , ck`1 s is smooth for all k P t0, 1, ¨ ¨ ¨ , nu. For such a
ş n şc
ř
piecewise path γ, we define γ f pzq dz :“ ck
k`1
f pγptqqγ 1 ptq dt.
k“0
"
t if t P r0, 1s
Example 3.3. Let γ rptq “
r be the path given by γ
1`pt´1qi if t P p1, 2s.
1ì
r
γ
Then
ş ş1 ş2
r
γ z dz “ 0 t 1 dt ` 1 p1`pt´1qiq i dt
ş1 ş2
“ 0 t dt ` 1 p1´pt´1qiq i dt
1
ş2 1 4´1
“ 2 ` 1 pi`pt´1qq dt “ 2 ` i ` 2 ´ p2 ´ 1q “ 1` i.
3.1. Definition of the contour integral 51
In Examples 3.2 and 3.3, we found that the contour integrals of the (nonholo-
şmorphic) function z şÞÑ z along the paths γ, γr joining 0 and 1 ` i are different:
γ z dz “ 1 ‰ 1` i “ γ
r z dz.
1ì
γ r
γ
Thus the integral depends on the path for the nonholomorphic integrand z ÞÑ z.
This is not strange, because from the definition of the contour integral of course we
expect the value of the contour integral to depend on the route chosen. The main
goal in this chapter is to show that the contour integrals of a function f along two
paths joining z, w P C is the same provided that f is holomorphic everywhere in
the region between the paths. This result (called the Cauchy Integral Theorem) is
fundamental in complex analysis because many further results about holomorphic
functions follow from it. First let us check that the result holds in a simple example.
Example 3.4. Let γ, γ r be the paths considered in Examples 3.2 and 3.3. Instead
of the nonholomorphic map z ÞÑ z, we take the entire function z ÞÑ z. Then
ş ş1 ş1
γ z dz “ 0 p1 ` iqt p1 ` iq dt “ 0 2it dt “ i, and
ş ş1 ş2 ş1 ş2 1 1
r z dz “ 0 t 1 dt ` 1 p1`pt´1qiqi dt “ 0 t dt ` 1 pi´pt´1qq dt “ 2 ì ´ 2 “ i.
γ
ş ş
This time the integrals are the same for γ and γ r: γ z dz “ i “ γr z dz.
Exercise 3.4. Integrate the following over γ, where γptq “ 2 exppitq for all t P r0, 2πs:
z ` z, z 2 ´ 2z ` 3, C Q z ÞÑ xy (where z “ x ` iy, x, y P R).
ş
Exercise 3.5. Evaluate γ Re z dz, where γ is given by:
‚ The line segment from 0 to 1ì.
‚ The short circular arc with centre i and radius 1 joining 0 to 1ì.
‚ The part of the parabola y “ x2 from x “ 0 to x “ 1 joining 0 to 1 ` i.
An important integral. Let us now calculate a simple, but important contour

integral, which will appear repeatedly. We fix a useful convention: Throughout, a
circular path with centre z0 and radius r ą 0 traversed in the anticlockwise direction
will mean the path C : r0, 2πs Ñ C, where Cptq “ z0 ` reit for all t P r0, 2πs, or
the (reparametrised) path C : r0, 1s Ñ C, where Cptq “ z0 ` re2πit for all t P r0, 1s.
(Thus the circle is with centre z0 and radius r traversed once counterclockwise.)
C
r
z0
Theorem 3.1. Let C be a circular path with centre z0"and radius r ą 0 traversed
ş 2πi if n “ ´1,
in the anticlockwise direction. Then C pz ´ z0 qn dz “
0 if n ‰ ´1.
Proof. We have Cptq “ z0 ` reit “ z0 ` r cos t ` ir sin t, t P r0, 2πs, and so

C 1 ptq “ ´r sin t ` ir cos t “ irpcos t ` i sin tq “ ireit , t P r0, 2πs.
ş ş ş2π ş2π
1˝ If n “ ´1, then C pz ´ z0 qn dz “ C pz ´ z0 q´1 dz “ 0 re1it ireit dt “ 0 i dt “ 2πi.
2˝ If n ‰ ´1, then
ş n
ş2π n nit it ş2π n`1 ipn`1qt
C pz ´ z0 q dz “ 0 r e ire dt “ 0 ir e dt
n`1
ş2π n`1
ş2π
“ ´r 0 sinppn`1qtqdt ` ir 0 cosppn`1qtqdt “ 0`0 “ 0.
The claim now also holds for the reparameterised path, r0, 1s Q t ÞÑ z0 ` re2πit .
We will see later that this has significant consequences. For instance, suppose that
a function f has an ‘expansion in terms of integral powers of z’, in an annulus
A :“ tz P C : r ă |z ´ z0 | ă Ru with centre z0 , inner radius r, and outer radius R:
ř
f pzq “ an pz ´ z0 qn for all z P A. p‹q
nPZ
We will give a precise meaning to the (infinite) ‘sum’ later, but for now, imagine a
finite sum (so that all but finitely many an s are zeros). Multiplying both sides by
ř
pz ´ z0 q´pm`1q for some m P Z, we obtain pz´zf pzq
0q
m`1 “ an pz ´ z0 qn´m´1 , and so
nPZ
1
ş f pzq ř ş
2πi C pz´z0 qm`1 dz “ an C pz ´ z0 qn´m´1 dz “ am .
nPZ
Here, we assumed that the sum passes through integration over C, which for finite
sums, follows from the definition of the integral, and we will see this in the next
section. When the sum is not finite, we will make precise the details later. The
upshot of it all is that the coefficients are expressible in terms of a contour integral,
and we will see later (Chapter 4) that any function holomorphic in an annulus A
will have such a (‘Laurent series’) expansion (‹).
Exercise 3.6. Let C be the circular path with centre
` ˘ 0 and radius 1 traversed once in the
ş p1`zqn
anticlockwise direction. Show that for 0 ď k ď n, nk “ 2πi
1
C z k`1 dz.
3.2. Properties of the contour integral

In this section we will show some useful properties of the contour integral. The
next result follows in a straightforward manner from the definition of contour in-
tegration, and is left as an exercise.
Proposition 3.2. Let D be a domain in C and γ : ra, bs Ñ D be a piecewise
smooth path. Then the following hold:
ş ş ş
(1) For all continuous f, g : D Ñ C, γ pf `gqpzq dz “ γ f pzq dz ` γ gpzq dz.
ş ş
(2) For all continuous f : D Ñ C and all α P C, γ pαf qpzq dz “ α γ f pzq dz.
3.2. Properties of the contour integral 53
Let CpD; Cq denote the vector space over C of all complex-valued continuous func-
tions on D with pointwise operations. Proposition 3.2 implies that each piece-
wise şsmooth path γ in D induces a linear transformation from CpD; Cq to C:
f ÞÑ γ f pzq dz : CpD; Cq Ñ C.
Exercise 3.7. Prove Proposition 3.2.
Opposite paths. Given a smooth path γ : ra, bs Ñ D in a domain D, its opposite

path, ´γ : ra, bs Ñ D, is defined by p´γqptq “ γpa ` b ´ tq, t P ra, bs. Then
p´γqpaq “ γpbq and p´γqpbq “ γpaq, and so ´γ starts where γ ends, and ends at
the starting point of γ while traversing the same set of points as the range of γ,
but in the opposite direction.
γ ´γ
Why do we denote the opposite path by ´ γ?
Proposition 3.3. If γ : ra, bs Ñ

ş D be a smoothşpath in a domain D, and f : D Ñ C
is a continuous function, then ´γ f pzqdz “ ´ γ f pzqdz.
Proof. We have
ş şb 1
şb 1
´γ f pzq dz “ a f pp´γqptqqp´γq ptq dt “ a f pγpa ` b ´ tqqpγ pa ` b ´ tqqp´1q dt
τ “a`b´t ş a 1
şb 1
ş
“ b f pγpτ qqγ pτ q dτ “ ´ a f pγpτ qqγ pτ q dτ “ ´ γ f pzqdz.
Exercise 3.8. If γ : ra, bs Ñ D is a smooth path in a domain D, then show that ´p´γq “ γ.
Concatenation of paths. Let γ1 : ra1 , b1 s Ñ D and γ2 : ra2 , b2 s Ñ D be two

paths in a domain D, such that γ1 pb1 q “ γ2 pa2 q (so that γ2 starts where γ1 ends).
Define the concatenation γ1 `γ2 : ra1 , b1 `b2 á2 s of γ1 and γ2 by:
"
γ1 ptq for a1 ď t ď b1 ,
pγ1 `γ2 qptq “
γ2 pt´b1 à2 q for b1 ď t ď b1 `b2 á2 .
γ1
γ2
γ1 `γ2
Proposition 3.4. Let γ1 : ra1 , b1 s Ñş D, γ2 : ra2 , b2 s şÑ D be two paths

ş in a domain
D such that γ1 pb1 q “ γ2 pa2 q. Then γ1 `γ2 f pzq dz “ γ1 f pzq dz ` γ2 f pzq dz.
Proof. We use the substitution τ “ t ´ b1 ` a2 to obtain the fourth equality below:

ş şb1 `b2 á2
γ1 `γ2 f pzq dz “ a1 f ppγ1 `γ2 qptqqpγ1 `γ2 q1 ptqdt
şb1 şb `b á
“ a1 f ppγ1 `γ2 qptqqpγ1 `γ2 q1 ptqdt ` b11 2 2 f ppγ1 `γ2 qptqqpγ1 `γ2 q1 ptqdt
şb şb `b á
“ a11 f pγ1 ptqqγ11 ptq dt ` b11 2 2 f pγ2 pt´b1 à2 qqγ21 pt´b1 à2 q dt
ş şb ş ş
“ γ1 f pzq dz ` a22 f pγ2 pτ qqγ21 pτ q dτ “ γ1 f pzq dz ` γ2 f pzq dz.
Exercise 3.9. If γ : ra, bs şÑ D is a smooth path in a domain D, and f : D Ñ C is
continuous, then show that γ`p´γq f pzqdz “ 0.
A useful estimate. We now prove a useful inequality for the size of the contour
integral in terms of the size of |f | along the contour, and the length of the contour.
Proposition 3.5 (ML inequality). Let γ : ra, bs Ñ D be a smooth path in a domain
D, and f : D Ñ C is a continuous function, then
ş
| γ f pzq dz| ď p max |f pγptqq| q p length of γ q .
tPra,bs
şb a
If x :“ Re γ, y :“ Im γ : ra, bs Ñ R, then the length of γ is a px1 ptqq2 ` py 1 ptqq2 dt:
The length
a of the path γ is the a sum of the incremental arc lengths ds, where
ds “ px1 ptqdtq2 ` py 1 ptqdtq2 “ px1 ptqq2 ` py 1 ptqq2 dt.
γpbq
ds y 1 ptqdt
x1 ptqdt
γpaq
şb şb
Proof. First, for a path ϕ : ra, bs Ñ C, we will prove | a ϕptq dt| ď a |ϕptq| dt.
şb şb şb şb
Here a ϕptq dt :“ a Repϕptqq dt ` i a Impϕptqq dt. Let a ϕptq dt “ reiθ for some
r ě 0 and θ P p´π, πs. Then
şb şb şb
| a ϕptq dt| “ r “ eíθ reiθ “ eíθ a ϕptq dt “ a eíθ ϕptq dt
şb şb
“ a Repeíθ ϕptqq dt ` i a Impeíθ ϕptqq dt.
But the left-hand side is real, and so the integral of the imaginary part on the
right-hand side must be zero. Consequently,
şb şb şb şb şb
| a ϕptqdt| “ a Repeíθ ϕptqqdt ď a |Repeíθ ϕptqq|dt ď a |eíθ ϕptq|dt “ a |ϕptq|dt.
The proposition now follows, since with ϕptq :“ f pγptqqγ 1 ptq for all t P ra, bs,
ş şb şb şb
| γ f pzq dz| “ | a f pγptqqγ 1 ptq dt| ď a |f pγptqqγ 1 ptq| dt “ a |f pγptqq||γ 1 ptq| dt
şb
ď p max |f pγptqq|qq a |γ 1 ptq| dt.
tPra,bs
If γptq “ xptq ` iyptq, where x, y are real-valued, then

şb 1 şb a
a |γ ptq| dt “ a px1 ptqq2 ` py 1 ptqq2 dt “ length of γ.
3.3. Fundamental Theorem of Contour Integration 55
By applying the ML inequality to each smooth section of a piecewise smooth path,

it follows that the ML inequality also holds for a piecewise smooth path.
şb şb
Exercise 3.10. For a path ϕ : ra, bs Ñ C and z P C, show that a z ϕptq dt “ z a ϕptq dt.
Exercise 3.11. Calculate
ş the upper bound given by the ML inequality on the absolute
value of the integral γ z 2 dz, where γ is the straight line path from 0 to 1ì. Also, compute
the integral and find its absolute value.
` ˘
Exercise 3.12. Using the calculation done in Exercise 3.6, deduce that 2n n ď 4 .
n
ş 1 ?
Exercise 3.13. Let γ be the straight line path from i to 1. Prove that | γ z4 dz| ď 4 2.
3.3. Fundamental Theorem of Contour Integration

Let us recall the Fundamental Theorem of Calculus in the real setting:
Theorem 3.6 (Fundamental Theorem of Calculus).
şb
If f : ra, bs Ñ R is continuously real differentiable, then a f 1 pxqdx “ f pbq ´ f paq.
This is an important result, because it facilitates the computation of integrals.

Indeed, knowing that a function is the derivative of something, it is easy to calculate
d x2
şb b2 á2
its integral. For example, since x “ dx 2 , we have a x dx “ 2 .
Analogously, if f P OpDq where D is aş domain, and f 1 is continuous on D,
then the calculation of the contour integral γ f 1 pzq dz is easy, since similar to the
Fundamental Theorem of Calculus in the real setting, we have the following.
Theorem 3.7 (Fundamental Theorem of Contour Integration).
If D is a domain, f P OpDq such
ş that f 1 is continuous on D, and γ : ra, bs Ñ D is
1
a piecewise smooth path, then γ f pzq dz “ f pγpbqq ´ f pγpaqq.
How does this theorem help? One can now calculate some contour integrals very
easily (just like in ordinary calculus). Here is an example.
d z 2
Example 3.5. As “ z in C, we have that for any piecewise smooth path γ
ş dz 2 2 2 2
joining 0 to 1 ` i, γ z dz “ p1ìq
2 ´ 02 “ 1`2iì
2 “ 1`2i´1
2 “ i. In particular, we
recover the answer obtained in Example 3.4.
From the above theorem, if f possesses an ‘antiderivative’ or ‘primitive’ F in D

(i.e., F P OpDq
ş and F 1 “ f ), then for any path
ş γ joining the points z to w, we
have that γ f pzq dz “ F pwq ´ F pzq, so that γ f pzq dz is independent of the path
γ taken to go from z to w.
Example 3.6. There is no function F : C Ñ C such that F 1 pzq “ z for all z P C.
Indeed, the calculations in Examples 3.2 and 3.3 show that the contour integral
along paths joining 0 to 1 ` i does depend on the path chosen.
Proof of Theorem 3.7. Let z “ x ` iy P D, x, y P R. Define U, V, u, v : D Ñ R

by f px ` iyq “ U px, yq ` iV px, yq and f 1 px ` iyq “ upx, yq ` ivpx, yq. We obtain
upx, yq ` ivpx, yq “ f 1 px ` iyq “ BU BV BV BU
Bx px, yq ` i Bx px, yq “ By px, yq ´ i By px, yq, where
we have used the Cauchy-Riemann equations. Set γptq “ xptq ` iyptq (t P ra, bs),
where x, y are real-valued. It is enough to show the theorem for smooth paths. So
d BU 1 BU 1
dt U pxptq, yptqq “ Bx pxptq, yptqq x ptq ` By pxptq, yptqq y ptq
“ upxptq, yptqq x1 ptq ´ vpxptq, yptqq y1 ptq,
d BV 1 BV 1
dt V pxptq, yptqq “ Bx pxptq, yptqq x ptq ` By pxptq, yptqq y ptq
“ vpxptq, yptqq x1 ptq ` upxptq, yptqq y1 ptq,
by the chain rule. Thus
ş 1 şb 1 1
şb
p q 1 1
γ f pzq dz “ a f pγptqqγ ptqdt “ a upxptq, yptqq ` ivpxptq, yptqqq px ptq ` iy ptqqdt
şb d şb d
“ a dt U pxptq, yptqqdt ` i a dt V pxptq, yptqqdt
“ U pxpbq, ypbqq ´ U pxpaq, ypaqq ` ipp V pxpbq, ypbqq ´ V pxpaq, ypaqqqq
“ f pγpbqq ´ f pγpaqq.
Exercise 3.14. Use the Cauchy-Riemann equations to show that z ÞÑ z has no primitive
in C.
Exercise 3.15 (Integration by Parts Formula). Let f, g be holomorphic functions defined
in a domain D, such that f 1 , g 1 are şcontinuous in D. Let γ be a piecewise smooth
ş path in
D from w P D to z P D. Show that γ f pζqg 1 pζq dζ “ f pzqgpzq ´ f pwqgpwq ´ γ f 1 pζqgpζq dζ.
ş iz
For r ą 0, let Sr ptq :“ Reit , t P r0, πs. Let a P Rzt0u. Show that lim | Sr zze
2 à2 dz| “ 0.
rÑ8
ş
Exercise 3.16. Evaluate γ cos z dz, where γ is any path joining í to i.
Exercise 3.17. Let D “ tz P C : |z| ă 1u. Let f P OpDq be such that f 1 is continuous on
D, and for all z P D, |f 1 pzq| ď 1. Prove that for all z, w P D, |f pzq´f pwq| ď |z ´w|.
ş1 1
Exercise 3.18. Use the Fundamental Theorem of Contour Integration to find 0 1ít dt.
A path γ : ra, bs Ñ C is closed if γpaq “ γpbq.
γpaq “ γpbq
Corollary 3.8. If D is a domain, f P OpDq is such thatşf 1 is continuous on D,

and γ : ra, bs Ñ D is a closed piecewise smooth path, then γ f 1 pzq dz “ 0.
ş
Proof. γ f 1 pzq dz “ f pγpbqq ´ f pγpaqq “ 0, since γpbq “ γpaq.
d z m
Example 3.7. For m P Zzt0u, dz m´1 in D :“ Czt0u. Also, D Q z ÞÑ z m´1
m “z ş
is continuous. So for any closed path γ in D, γ z m´1 dz “ 0.
What if m “ 0? As Log1 z “ 1z for z P Dr :“ Czpp´8, 0s ˆ t0uq, for any closed path
ş 1
r
r in D,
γ 1
dz “ 0. But does not have a primitive in D; see Exercise 3.21.
r z
γ z
3.4. The Cauchy Integral Theorem 57
Exerciseş 3.19. Use the Fundamental Theorem of Contour Integration to determine the
value of γ exp z dz, where γ is a path joining 0 and a ` ib. Equate the answer obtained
with the contour integral along the straight line from 0 to a ` ib, and hence deduce that
ş 1 ax a a
0
e cospbxq dx “ ape cos ab´1q`be
2 `b2
sin b
.
Exercise 3.20. Applying the Fundamental Theorem of Contour Integration to exp z and
ş2π
integrating round a circular path, show that for all r ą 0, 0 er cos θ cospr sin θ ` θq dθ “ 0.
1
Exercise 3.21. Show that z has no primitive in the punctured plane Czt0u.
3.4. The Cauchy Integral Theorem

Theorem 3.9 (Cauchy Integral Theorem).
Let ‚ D be a domain in C,
‚ f : D Ñ C be holomorphic in D, and
‚ γ0 , γ1 : r0, 1s Ñ D be two closed, piecewise smooth, D-homotopic paths.
ş ş
Then γ0 f pzq dz “ γ1 f pzq dz.
Before we go further, let us try to understand the statement. Firstly, the two paths
in D are closed. Secondly, what is meant by saying that the two closed paths are
‘D-homotopic’ ? Intuitively, this means the following. See the picture on the left
below, where the two paths are depicted in the domain D “ C. Imagine a rubber
band placed along γ0 . For γ0 to be D-homotopic to γ1 , it should be possible to
deform this rubber band so as to get γ1 , with the condition that each intermediate
position of the rubber band lies in D. Clearly this is not possible sometimes, for
example if the domain has holes. See for example the picture on the right below,
where we expect the two paths in the domain D “ Czt0u to be not D-homotopic.
γ1 D“C D “ Czt0u
γ1
γ0
γ0
obstruction
intermediate position
of the rubber band
Definition 3.2. Let D be a domain in C and γ0 , γ1 : r0, 1s Ñ D be closed paths.

Then γ0 is D-homotopic to γ1 if there is a continuous H : r0, 1s ˆ r0, 1s Ñ D such
that the following hold:
(H1) For all t P r0, 1s, Hpt, 0q “ γ0 ptq.
(H2) For all t P r0, 1s, Hpt, 1q “ γ1 ptq.
(H3) For all τ P r0, 1s, Hp0, τ q “ Hp1, τ q.
The map H may be thought of as describing a family of closed paths from r0, 1s to
D, parameterised by the τ -variable, thought of as ‘time’. The closed path Hp¨, τ q
represents the intermediate position of the ‘rubber band’ at time τ . Initially, for
τ “ 0, Hp¨, 0q is the path γ0 , while finally, when the time τ “ 1, Hp¨, 1q is γ1 . This
is the content of (H1) and (H2). The requirement (H3) says that at time τ , the
intermediate path Hp¨, τ q is closed too. Continuity of H means that the rubber
band never breaks, and the deformation takes place continuously.
Hp¨, 0q “ γ0
Hp¨, τ q
Hp¨, 1q “ γ1
Example 3.8. Let D “ C, and γ0 , γ1 : r0, 1s Ñ C be the two circular paths given
by γ0 “ 4e2πit , and γ1 “ 2i ` e2πit , for all t P r0, 1s. Then γ0 is C-homotopic to γ1 :
We define H by taking a ‘convex combination’ of the points γ0 ptq and γ1 ptq.
Define H : r0, 1s ˆ r0, 1s Ñ C by Hpt, τ q “ p1 ´ τ qγ0 ptq ` τ γ1 ptq for all 0 ď t, τ ď 1.

Then H is continuous, and moreover,
(H1) for all t P r0, 1s, Hpt, 0q “ γ0 ptq,
(H2) for all t P r0, 1s, Hpt, 1q “ γ1 ptq, and
(H3) for each τ P r0, 1s, Hp0, τ q “ p1 ´ τ q4 ` τ p2i ` 1q “ Hp1, τ q.
Hence (H1), (H2), (H3) are satisfied and so γ0 is C-homotopic to γ1 .
On the other hand, the same two paths are not Czt0u-homotopic. Why? If they
were Czt0u-homotopic, then by the Cauchy Integral Theorem,
ş the contour integral
ş
of z1 P OpCzt0uq for the two paths would be equal. But γ0 1z dz “ 2πi ‰ 0 “ γ1 z1 dz,
where the last equality is due to the Fundamental Theorem of Contour Integration,
by observing that 1z has the primitive Log in Czppp´8, 0s ˆ t0uqq.
Exercise 3.22. Let a, b, r ą 0. Define C, E : r0, 1s Ñ Czt0u by Cptq “ re2πit and
Eptq “ a cosp2πtq ` ib sinp2πtq for all t P r0, 1s. Show that C is Czt0u-homotopic to E.
Exercise 3.23. Let D be a domain in C. Show that D-homotopy is an equivalence relation
on the set of all closed paths in D. In particular, we can say ‘γ0 , γ1 are D-homotopic’ instead
of ‘γ0 is D-homotopic to γ1 ’.
r
Exercise 3.24. Let r ą 0, Cptq “ re2πit , Cptq “ ´2 ` e2πit for all t P r0, 1s. Prove that
r are not Czt0u-homotopic.
C, C
Proof.(of Theorem 3.9.) We will make the simplifying assumption that the real
and imaginary parts of the homotopy H are twice continuously differentiable. This
smoothness condition can be omitted (see §3.10), but the proof is technical. The
assumption of twice continuous differentiability is mild, and is used below to ex-
2 B2 H
change the order of partial differentiation: BBτ H Bt “ BtBτ . ş(See e.g. [1, §8.23].)
Let γτ :“ Hp¨, τ q be the path at time τ . Define Ipτ q :“ γτ f pzq dz for all τ P r0, 1s.
dI
(We differentiate underşthe integral sign with respect ş to τ to obtain dτ ” 0, showing
that I is constant. So γ0 f pzq dz “ Ip0q “ Ip1q “ γ1 f pzq dz.) We have
dI d
ş d
ş1 BH
dτ pτ q “ dτ γτ f pzq dz “ dτ 0 f pHpt, τ qq Bt pt, τ q dt
ş1 B
“ 0 Bτ p f pHpt, τ qq BH q
Bt pt, τ qq dt
p˚q ş1 2
“ 0 p f 1 pHpt, τ qq BH BH B H q
Bτ pt, τ q Bt pt, τ q ` f pHpt, τ qq Bτ Bt pt, τ qq dt
ş1 2
“ 0 p f 1 pHpt, τ qq BH BH B H q
Bt pt, τ q Bτ pt, τ q ` f pHpt, τ qq BtBτ pt, τ qq dt
p‹q ş1d BH
“ q
0 dt p f pHpt, τ qq Bτ pt, τ qq dt
“ f pHp1, τ qq BH BH
Bτ p1, τ q ´ f pHp0, τ qq Bτ p0, τ q
(using the Fundamental
Theorem of Calculus)
Hp0,σq´Hp0,τ q
“ f pHp1, τ qq BH
Bτ p1, τ q ´ f pHp1, τ qq lim σ´τ
σÑτ
Hp1,σq´Hp1,τ q
Bτ p1, τ q ´ f pHp1, τ qq lim σ´τ
σÑτ
Bτ p1, τ q ´
BH
f pHp1, τ qq Bτ p1, τ q “ 0.
ş ş
So r0, 1s Q τ ÞÑ Ipτ q is constant. In particular, γ1 f pzqdz “ Ip1q “ Ip0q “ γ2 f pzqdz.
This completes the proof. But where was the holomorphicity of f used? To
get the equalities labelled by (˚) and (‹), for e.g., in (˚), we used the fact that
B 1 BH
Bτ pf pHpt, τ qqq “ f pHpt, τ qq Bτ pt, τ q. We justify this now.
d
Claim: If ϕ is a smooth path, then dτ f pϕpτ qq “ f 1 pϕpτ qqϕ1 pτ q.
Let u :“ Re f , v :“ Im f , x :“ Re ϕ, y :“ Im ϕ. Then
d d q
dτ f pϕpτ qq “
dτ p upxpτ q, ypτ qq ` ivpxpτ q, ypτ qqq
“ Bu 1 Bu 1 p Bv 1 Bv 1
Bx pϕpτ qqx pτ q ` By pϕpτ qqy pτ q ` ip Bx pϕpτ qqx pτ q ` By pϕpτ qqy pτ qq
q
Bu 1 Bv 1 Bv 1 Bu 1
“ Bx pϕpτ qqx pτ q´ Bx pϕpτ qqy pτ q ` ipp Bx pϕpτ qqx pτ q ` Bx pϕpτ qqy pτ qqq
“ p Bu Bv q 1 1 1 1
Bx pϕpτ qq ` i Bx pϕpτ qqq px pτ q ` iy pτ qq “ f pϕpτ qqϕ pτ q.
We also assumed that f 1 is continuous when differentiating under the integral sign.
Again, the result holds without this assumption, and the complete proof of the
Cauchy Integral Theorem is given in §3.10.
Exercise 3.25. Let a ą 0, b ą 0. Define E by Eptq “ a cos t ` ib sin t for all t P r0, 2πs.
ş ş2π
Considering E z1 dz, show that 0 a2 pcos θq2 `b
1 2π
2 psin θq2 dθ “ ab .
Exercise 3.26. We have seen that if C is theş circular path with centre 0 and radius 1
traversed in the anticlockwise direction, then C z1 dz “ 2πi. Now consider the path S,
comprising the four line segments which are the sides of the square with vertices ˘1 ˘ i,
traversed anticlockwise. Draw a picture şto convince yourself that S is Czt0u-homotopic to
C. Evaluate parametrically the integral S 1z dz, and confirm that the answer is indeed 2πi.
Special case: simply connected domains. If D is a domain and p P D, then

consider the ‘degenerate path’ γp : r0, 1s Ñ D given by γp ptq “ p for all ş t P r0, 1s.
As γp p0q “ p “ γp p1q, γp is closed. If f : D Ñ C is continuous, what is γp f pzq dz?
ş ş1
0, because γp1 ” 0, which yields γp f pzq dz “ 0 f pγp ptqqγp1 ptq dt “ 0.
In light of this, an important special case of the Cauchy Integral Theorem is
obtained when a closed path γ is D-homotopic to a point (that is, the constant
path γp ptq “ p for all t P r0, 1s). In this case we say that γ is D-contractible.
Imagine placing a rubber band along γ, and then shrinking it to a point such that
each intermediate position of the rubber band is in D. For a D-contractible path
γ (which is D-homotopic to ş a point p şP D), and for any f P OpDq, the Cauchy
Integral Theorem implies γ f pzqdz “ γp f pzqdz “ 0. A domain in which every
closed path is D-contractible is called simply connected. We have the following
corollary of the Cauchy Integral Theorem.
Corollary 3.10. If D is a simply

ş connected domain, γ is a closed piecewise smooth
path in D, and f P OpDq, then γ f pzq dz “ 0.
The domains C, Dpz0 , rq, Czpp´8, 0s ˆ t0uq, H :“ tz P C : Re z ą 0u are all simply

connected. For example if we take any p in the domains in first two cases, then the
homotopy given by Hpt, τ q :“ p1´τ qγptq`τ p for all t, τ P r0, 1s does the job. In the
case of D “ Czpp´8, 0s ˆ t0uq, or H, given any γ : r0, 1s Ñ D, we first choose any
real p ą 0 for example p “ 1, and then use the same H as above. In the former case,
if Impγptqq ‰ 0, then ImpHpt, τ qq “ p1 ´ τ qImpγptqq ‰ 0 for all τ P r0, 1q, showing
that in this case Hpt, τ q P D, while if Impγptqq “ 0, then as γptq P D, we must have
Repγptqq ą 0, which yields RepHpt, τ qq “ p1 ´ τ qRepγptqq ` τ 1 ą 0, so that again
Hpt, τ q P D. None of the above domains have any ‘holes’ in them. On the other
hand, domains with holes are not simply connected. For example, the punctured
complex plane Czt0u is not simply connected. For example consider the circular
path C with centre 0 and any ş radius r ą 0 traversed once in the anticlockwise
direction. We have seen that C z1 dz “ 2πi. But if C were Czt0u-contractible to a
ş
point in the punctured plane, we should have C z1 dz “ 0 by the Cauchy Integral
Theorem. So this means that C is not Czt0u-contractible to a point in Czt0u and
so Czt0u is not simply connected. Similarly, it can be shown that the annulus
tz P C : 1 ă |z| ă 2u is not simply connected. The obstruction of the hole can be
thought of as a nail or a pillar emanating from the plane, which prevents a rubber
band encircling it from being shrunk to a point in the domain while always staying
in the plane.
simply connected not simply connected
C D
Czpp´8, 0sˆt0uq Czt0u
şExample 3.9. For any closed path γ, as exp ş is entire, and as C is simply connected,
γ exp z dz “ 0. For any entire function f , γ f pzq dz “ 0, for any closed path γ.
It can happen for a domain which is not simply connected, and an f P OpDq that
ş 1
γ f pzqş dz “ 0 for all closed paths γ in D. For example, let D “ Czt0u, f :“ z 2 .
1 1
Then γ z 2 dz “ 0 for all closed paths γ in the punctured plane (as z 2 possesses a
primitive in Czt0u): dz d
p´ 1z q “ z12 .
Corollary 3.11. If D is a simply connected domain, f P OpDq, γ : ra, bs Ñ D and

r : rc, ds Ñ D are two smooth paths such that they have the
γ ş same startş and end
points, that is, γpaq “ γ rpdq “: w, then γ f pzq dz “ γr f pzq dz.
rpcq “: z and γpbq “ γ
γpbq “ γ
rpdq
r
γ
γpaq “ γ
rpcq
Proof. As γ ´ γ
r is closed, it follows from the Cauchy Integral Theorem that
ş ş ş
0 “ γ´rγ f pzq dz “ γ f pzq dz ´ γr f pzq dz.
Exercise 3.27. Integrate the following over the circular path given by |z| “ 3 traversed
1
in the anticlockwise direction: Logpz ´ 4iq, z´1 , principal value of iz´3 .
Exercise 3.28. Define γ0 , γ1 : r0, 1s Ñ Czt0u by γ0 ptq “ e2πit , γ1 ptq “ e´2πit for all t P r0, 1s.
Prove or disprove: γ0 is Czt0u-homotopic to γ1 .
Exercise 3.29 (Winding number of a curve).

Suppose γ : r0, 1s Ñ C is a closed smooth path that does not pass through the origin 0.
1
ş 1 1
ş 1 γ 1 ptq
γ z dz “ 2πi 0 γptq dt.
Define the winding number wpγq of γ (about 0) by wpγq :“ 2πi
(1) Use the observation rexpp2πiaq “ 1 ô a P Zs, to show wpγq P Z proceeding as follows.
ş t 1 psq
We define the function ϕ : r0, 1s Ñ C by setting ϕptq “ exp 0 γγpsq ds for all t P r0, 1s.
To show wpγq P Z, it suffices to show that ϕp1q “ 1. To this end, calculate ϕ1 ptq, and
use this expression to show ϕ{γ is constant in r0, 1s. Conclude ϕp1q “ 1.
(2) Calculate the winding number of Γ1 : r0, 1s Ñ C given by Γ1 ptq “ expp2πitq (t P r0, 1s).
(3) Prove that if γ1 , γ2 : r0, 1s Ñ C are smooth closed paths not passing through 0, and
γ1 ¨γ2 is their pointwise product, then wpγ1 ¨γ2 q “ wpγ1 q ` wpγ2 q.
(4) Let m P N. Find the winding number of Γm , where Γm ptq :“ expp2πimtq (t P r0, 1s).
(5) Show that the winding number map γ ÞÑ wpγq is ‘locally constant’: If γ0 : r0, 1s Ñ Czt0u
is a smooth closed path, then there is a δ ą 0 such that for every smooth closed path
γ : r0, 1s Ñ Czt0u such that }γ ´ γ0 }8 :“ maxt|γptq ´ γ0 ptq| : t P r0, 1su ă δ, wpγq “ wpγ0 q.
(In other words, if the set of curves is equipped with the uniform topology, and Z
with the discrete topology, then γ ÞÑ wpγq is continuous.)
3.4.1. What happens with nonholomorphic functions? We now highlight

the fact that the Cauchy Integral Theorem may fail if one drops the assumption
of holomorphicity of f . Let us see what happens when we consider our favourite
nonholomorphic function, the complex conjugation map z ÞÑ z. We will show that
rather than the integral around the closed loop γ being 0, the contour integral of
z around γ yields the area enclosed by γ, which is of course very much dependent
on γ, and two C-homotopic paths can enclose widely different areas (take two
concentric circles with different radii). We will only give a plausibility argument
by resorting to a specific picture, e.g. as shown in the following figure.
γ pxptq, yptqq
pxpaq, ypaqq
=
pxpbq, ypbqq
For the smooth path γ : ra, bs Ñ C, with x :“ Re γ and y :“ Im γ, we have

ş şb 1 1
γ z dz “ a pxptq ´ iyptqqpx ptq ` iy ptqq dt
şb 1 1 1
“ p
a xptqx ptq ` yptqy ptq ` ipxptqy ptq ´ yptqx1 ptqqqq dt
şb d pxptqq2 `pyptqq2 şb 1 ptq
“ a dt 2 dt ` i a pxptqy ´ yptqx1 ptqq dt
pxpbqq2 ´pxpaqq2 `pypbqq2 ´pypaqq2 şb 1 ptq
“ 2 ` i a pxptqy ´ yptqx1 ptqq dt
şb
“ i a pxptqy 1 ptq ´ yptqx1 ptqq dt.
3.5. Existence of a primitive 63
` ´
y 1 ptqdt ý 1 ptqdt
xptq
` yptq
´x1 ptqdt
x1 ptqdt
şb
From the top two pictures, we see that a xptqy 1 ptqdt “ (Area enclosed by γ).
şb
From the bottom two pictures, we see that a x1 ptqyptqdt “ ´(Area enclosed by γ).
ş şb
Thus γ z dz “ i a pxptqy 1 ptq ´ yptqx1 ptqqdt “ 2i ¨ (Area enclosed by γ).
Exercise 3.30. Suppose a coin of radius r rolls around a fixed bigger coin of radius R.
Then the path traced by a point on the rim of the rolling coin is called an epicycloid,
and it is a closed curve if R “ nr, for some n P N. With the centre of the fixed coin at
the origin, show that the epicycloid is given by zptq “ rppn ` 1qeit ´ eipn`1qt q, t P r0, 2πs.
By evaluating the integral of z along the epicycloid, show that the area enclosed by the
epicycloid is equal to πr2 pn`1qpn`2q.
Next, we will show the following consequences of the Cauchy Integral Theorem:
‚ In simply connected domains every holomorphic function possesses a primitive.
‚ Holomorphic functions are infinitely many times complex differentiable.
‚ Bounded entire functions are constants (Liouville’s Theorem), and a proof of
the Fundamental Theorem of Algebra using Liouville’s Theorem.
‚ Morera’s Theorem (which is a sort of a converse to the Cauchy Integral Theorem).
3.5. Existence of a primitive
On a simply connected domain, every holomorphic function f is the derivative of
some holomorphic function, that is, f possesses a primitive.
Theorem 3.12. If D is a simply connected domain and f P OpDq, then there
exists an F P OpDq such that F 1 “ f in D.
ş
Proof. Fix a point p P D. Define F : D Ñ C by F pzq “ γz f pζq dζ (z P D), where
γz is any smooth path in D joining p to z. Then F is well-defined, i.e., F pzq does
not depend on the path joining p to z. If γ is another smooth path in D that joins
p to z, then γz ´ γ is a closed smooth path ş in the simplyş connected domain
ş D.
The Cauchy Integral Theorem implies 0 “ γz ´γ f pζq dζ “ γz f pζq dζ ´ γ f pζq dζ,
ş ş
so that γz f pζq dζ “ γ f pζq dζ.
γ
w
rz, ws
´γ z γw
z
γz
γz
p p
Next, we will show F P OpDq and that F 1 “ f in D. Since f P OpDq, f is

continuous in D, and so given a z P D and an ǫ ą 0, there exists a δ ą 0 such
that if w P D and |w ´ z| ă δ, then |f pwq ´ f pzq| ă ǫ. If w P D is such that
ş ş
0 ă |w ´ z| ă δ, then F pwq´F w´z
pzq 1
“ w´z p γw f pζq dζ ´ γz f pζq dζ q . If rz, ws is
a straight line path joining z to w, then the concatenation of γz with the con-
catenation
ş of rz, ws with ´γşw is a closed path.
ş By the Cauchy
ş Integral Theorem,
0 “ γz `rz,ws´γw f pζq dζ “ γz f pζq dζ ` rz,ws f pζq dζ ´ γw f pζq dζ. The Funda-
ş ş
mental Theorem of Contour Integration gives rz,ws 1 dζ “ rz,ws ζ 1 dζ “ w ´ z. So
F pwq´F pzq 1
ş 1
ş 1
ş
w´z ´ f pzq “ w´z rz,ws f pζq dζ ´ w´z rz,ws f pzq dζ “ w´z rz,ws pf pζq ´ f pzqq dζ .
Thus
ş ş
| F pwq´F
w´z
pzq 1
´ f pzq| “ | w´z rz,ws pf pζq ´ f pzqq dζ| “ 1
|w´z| | rz,ws pf pζq ´ f pzqq dζ|
1 1
ď |w´z| pζPrz,ws
max |f pζq ´ f pzq|q(length of rz, ws) ď |w´z| ǫ|w ´ z| “ ǫ.
Thus F 1 pzq “ f pzq. As z P D was arbitrary, F P OpDq and F 1 “ f in D.

Remark 3.2. A primitive for a holomorphic function f in a simply connected
domain is unique up to a constant. Indeed, if F, Fr are both primitives for F , then
F 1 “ f “ Fr1 in D, and so dz
d
pF ´ Frq “ F 1 ´ Fr 1 “ f ´ f “ 0 in D. By Exercise 2.18,
there is a constant C such that F ´ Fr “ C in D. So F “ Fr ` C in D. ˚
Example 3.10. expp´z 2 q is entire. So there exists an F , which is also entire,
such that for all z P C, F 1 pzq “ expp´z 2 q. But one ‘cannot express F in terms of
ş 2
elementary functions’; see e.g. [23]. One primitive is given by Frpzq “ γz e´ζ dζ
for z P C, whereşγz is the straight line path joining 0 to z. Then in particular, for
x 2
real x, Frpxq “ 0 e´ξ dξ, and it turns out that this (and so any other primitive
too) cannot be expressed in terms of elementary functions.
Exercise 3.31. If D is a domain, and f P OpDq is such that there is no F P OpDq such
that F 1 “ f in D, then D is not simply connected. Give an example of such a D and f .
3.6. The Cauchy Integral Formula 65
3.6. The Cauchy Integral Formula

The Cauchy Integral Formula, roughly speaking, says that if γ is a closed path, and
f is holomorphic inside γ, then the value of f at any point inside γ is determined by
the values of f on γ. This illustrates the ‘rigidity’ of holomorphic functions. In the
next chapter, we will study a more general Cauchy Integral Formula, expressing
all the derivatives of f at any point inside γ in terms of the values of the function
on γ. The result below is the ‘n “ 0 case’ of the more general result to follow.
Theorem 3.13 (The Cauchy Integral Formula for circular paths).
Let D be a domain, f P OpDq, r ą 0, z0 P D, ∆ :“ tz P C : |z ´ z0 | ď ru Ă D, and
1
ş f pζq
Cptq “ z0 ` re2πit , t P r0, 1s. Then for all z P Dpz0 , rq, f pzq “ 2πi C ζ´z dζ.
C
z0 r
w ∆
In order to prove this result, we will first prove the following technical fact, which
will also prove to be useful later on. In this result we essentially prove the Cauchy
Integral Formula for the centre z0 of the disc ∆.
Proposition 3.14. Let D be a domain, z0 P D, f : D Ñ C be holomorphic in
Dztz0 u, and continuous on D, r ą 0, ∆ :“ tz P C : |z ´ z0 | ď ru Ă D, and
1
ş f pζq
Cptq “ z0 ` re2πit , t P r0, 1s. Then f pz0 q “ 2πi C ζ´z0 dζ.
Proof. Let ǫ ą 0. Since f is continuous at z0 , there exists a δ P p0, rq such that

whenever |ζ ´ z0 | ď δ, |f pζq ´ f pz0 q| ă ǫ. Consider the circular path Cδ , with
centre z0 and radius δ traversed in the anticlockwise direction. Then Cδ and C
are Dztz0 u-homotopic (e.g., take the homotopy map H as a convex combination
of C, Cδ : Hp¨, τ q :“ p1 ´ τ qCp¨q ` τ Cδ p¨q, τ P r0, 1s).
C
Cδ
z0
ş f pζq ş f pζq
By the Cauchy Integral Theorem, C ζ´z 0
dζ “ Cδ ζ´z 0
dζ. Hence,
1
ş f pζq 1
ş f pζq 1
ş 1
| 2πi C ζ´z0 dζ ´ f pz0 q| “ | 2πi Cδ ζ´z0 dζ ´ f pz0 q 2πi Cδ ζ´z0 dζ|
1
ş f pζq´f pz0 q
“ | 2πi Cδ ζ´z0 dζ|
1 |f pζq´f pz0 q| 1 ǫ
ď 2π p max |ζ´z0 | q 2πδ ă 2π δ 2πδ “ ǫ.
ζPC δ
Since ǫ ą 0 was arbitrary, the claim follows.

Proof. (of Theorem 3.13.) Let z be such that |z ´ z0 | ă r. Choose a δ ą 0 small

enough so that the circular path Cδ with centre z and radius δ is contained in the
interior of C. We will now show that C and Cδ are Dztzu-homotopic.
z
Cδ
z0
We have Cptq “ z0 ` re2πit ,

and Cδ ptq :“ z ` δe2πit , t P r0, 1s. Define the map
H : r0, 1s ˆ r0, 1s Ñ C by Hpt, τ q “ p1 ´ τ qCptq ` τ Cδ ptq for all t, τ P r0, 1s. Then
H is continuous. For all t P r0, 1s, Hpt, 0q “ Cptq and Hpt, 1q “ Cδ ptq. For all
τ P r0, 1s, Hp0, τ q “ p1 ´ τ qCp0q ` τ Cδ p0q “ p1 ´ τ qCp1q ` τ Cδ p1q “ Hp1, τ q. Now
we show that Hpt, τ q P Dztzu for all t, τ P r0, 1s. We have
|Hpt, τ q ´ z0 | “ |p1 ´ τ qCptq ` τ Cδ ptq ´ z0 |
“ |p1 ´ τ qCptq ` τ Cδ ptq ´ pp1 ´ τ qz0 ` τ z0 q|
ď p1 ´ τ q|Cptq ´ z0 | ` τ |Cδ ptq ´ z0 |
ď p1 ´ τ qr ` τ |Cδ ptq ´ z ` z ´ z0 |
ď p1 ´ τ qr ` τ pδ ` |z ´ z0 |q
ď p1 ´ τ qr ` τ r “ r.
Thus Hpt, τ q P ∆ Ă D. Also,
|Hpt, τ q ´ z| “ |p1 ´ τ qpz0 ` re2πit q ` τ pz ` δe2πit q ´ z|
“ |re2πit ` p1 ´ τ qpz0 ´ zq ` τ pδ ´ rqe2πit |
ě r ´ |p1 ´ τ qpz0 ´ zq ` τ pδ ´ rqe2πit |
ě r ´ pp1 ´ τ q|z0 ´ z| ` τ pr ´ δqq
“ p1 ´ τ qpr ´ |z0 ´ z|q ` τ δ
ě p1 ´ τ qδ ` τ δ “ δ ą 0.
f p¨q
So C and Cδ are Dztzu-homotopic. Since ¨´z P OpDztzuq, by the Cauchy Integral
1
ş f pζq 1
ş f pζq
Theorem, the second equality holds: f pzq “ 2πi Cδ ζ´z dζ “ 2πi C ζ´z dζ.
Exercise 3.32.
ş Let 0 ă a ă 1, and
ş 2π C be the unit circle with centre 0 traversed anticlockwise.
i 1
Show that C pzáqpaz´1q dz “ 0 1à2 ´2a cos t dt.
ş 2π 1 2π
Use the Cauchy Integral Formula to deduce 0 1à2 ´2a cos t dt “ 1á2 .
1
Exercise 3.33. Does z ÞÑ zp1´z 2 q have a primitive in tz P C : 0 ă |z| ă 1u?
Exercise 3.34. Let f be entire. Define the path C byş Cpθq :“ eiθ , for θ P r0, 2πs. Let
1 1´ζz
z P C such that |z| ă 1. Prove that p1 ´ |z|2 qf pzq “ 2πi C f pζq ζ´z dζ.
ş
1 2π
Deduce from the above that for all z P D, p1 ´ |z|2 q|f pzq| ď 2π iθ
0 |f pe q|dθ.
3.6. The Cauchy Integral Formula 67
Exercise 3.35. Fill in the blanks.

ş
(1) γ exp z
z´1 dz “ , where γ is the circle |z| “ 2 traversed in the anticlockwise direction.
ş z2 `1
(2) γ z2 ´1 dz “ , where γ is the circle |z´1| “ 1 traversed in the anticlockwise direction.
ş z2 `1
(3) γ z2 ´1 dz “ , where γ is the circle |zí| “ 1 traversed in the anticlockwise direction.
ş z2 `1
(4) γ z2 ´1 dz “ , where γ is the circle |z`1| “ 1 traversed in the anticlockwise direction.
ş z2 `1
(5) γ z2 ´1 dz “ , where γ is the circle |z| “ 3 traversed in the anticlockwise direction.
Corollary 3.15 (Cauchy’s Integral Formula for general paths).

Let D be a domain, f P OpDq, z0 P D, γ be a closed path in D which is Dztz0 u-
homotopic to a circular path C centred at z0 , such that C and its interior is con-
1
ş f pζq
tained in D. Then f pz0 q “ 2πi γ ζ´z0 dζ.
1
ş f pζq
Proof. By the Cauchy Integral Formula for circular paths, f pz0 q “ 2πi C ζ´z0 dζ.
As γ, C are Dztz0 u-homotopic, it follows from the Cauchy Integral Theorem that
1
ş f pζq 1
ş f pζq
2πi C ζ´z0 dζ “ 2πi γ ζ´z0 dζ.
This result highlights the ‘rigidity’ associated with holomorphic functions men-
tioned earlier. By this we mean that their highly structured nature (everywhere
locally infinitesimally a rotation followed by a magnification) enables one to pin
down their precise behaviour from very limited information. That is, even if we
know the effect of a holomorphic function in a small portion of the plane (for ex-
ample the values along a closed path), its values can be inferred at other far away
points in a unique manner. The picture below illustrates this in the case of the
Cauchy Integral Formula, where knowing the values of f on the curve γ enables
one to determine the values at all points in the shaded region.
γ
z0
exppizq
Exercise 3.36. Define F by F pzq “ z 2 `1 for all z P Czti, íu. Let R ą 1.
(1) Let σ be the closed path comprising the line segment r´R, Rs of the real axis from ´R
ş by the semicircle S of radius R in the upper half plane from R to ´R.
to R, followed
Show that σ F pzqdz “ πe .
(2) Prove that | exppizq| ď 1 for z in the upper half plane, and conclude that for all large
enough |z|, |F pzq| ď |z|2 2 .
ş ş
(3) Show that lim S F pzq dz “ 0, and so lim r´R,Rs F pzq dz “ πe .
RÑ8 RÑ8
ş 8 cos x şR cos x π
(4) Parametrising r´R, Rs in terms of x, show that ´8 1`x 2 dx :“ lim ´R 1`x2 dx “ e .
RÑ8
ş 2π
Exercise 3.37. Evaluate 0 ecos θ cospsin θq dθ. Hint: Consider exppexppiθqq.
3.7. Holomorphic functions are infinitely complex differentiable

In this section it is shown that holomorphic functions are infinitely many times
complex differentiable. Let us contrast this with the situation in Real Analysis.
We had seen in Example 0.1 that the derivative may fail to be real differentiable
at isolated points, and also that there are even more extreme examples of this
phenomenon, namely functions f : R Ñ R which are real differentiable everywhere,
but f 1 is real differentiable nowhere.
Corollary 3.16. If D be a domain, and f P OpDq, then f 1 P OpDq.
We get the chain of implications: f P OpDq ñ f 1 P OpDq ñ f 2 P OpDq ñ ¨ ¨ ¨ . So,
if f P OpDq, then f is infinitely many times complex differentiable in D.
Here is a plan of how we will show this. From the Cauchy Integral Formula,
1
ş f pζq
f pzq “ 2πi C ζ´z dζ, where C is a circle centred at z with radius r such that the
circle C and its interior are contained in D. If we formally differentiate under the
1
ş f pζq
integral sign, we get an expression for the derivative of f : f 1 pzq “ 2πi C pζ´zq2 dζ.
1 pwq´f 1 pzq
Having shown this formula, we will show that lim f z´w exists by using the
wÑz
above expression for the derivative at z and w.
Proof. Let z P D, and r ą 0 be such that ∆ :“ tζ P C : |ζ ´ z| ď ru Ă D. Let

1
ş f pζq
Cptq “ z `re2πit for all t P r0, 1s. We will show that f 1 pzq “ 2πi C pζ´zq2 dζ. Define
#
f pζq´f pzq
ζ´z if z P Dztzu,
gpζq “ 1
f pzq if ζ “ z.
Then g is holomorphic in Dztzu and continuous in D. By Proposition 3.14,
1
ş gpζq 1
ş f pζq´f pzq
f 1 pzq “ gpzq “ 2πi C ζ´z dζ “ 2πi C pζ´zq2 dζ
1
ş f pζq f pzq ş 1
“ 2πi C pζ´zq2 dζ ´ 2πi C pζ´zq2 dζ
1
ş f pζq
“ 2πi C pζ´zq2 dζ ´ 0. p‹q
So we have shown the formula for f 1 pzq at centres z of the closed discs that are
contained in D. For w P Dpz, rq, let δ P p0, rq be such that Dpw, δq Ă Dpz, rq. From
1
ş f pζq 1
ş f pζq
the above, we get f 1 pwq “ 2πi Cδ pζ´wq2 dζ “ 2πi C pζ´wq2 dζ (the last equality
f p¨q
follows from the Cauchy Integral Theorem, since p¨´wq2 P OpDztwuq, and the paths
C, Cδ are Dztwu-homotopic).
C
Cδ
z
z0
3.8. Liouville’s Theorem, Fundamental Theorem of Algebra 69
So for w ‰ z inside C,
f 1 pwq´f 1 pzq 1 1
ş f pζq 1
ş f pζq 1
ş f pζqp2ζ´z´wq
w´z “ w´z p 2πi C pζ´wq2 dζ ´ 2πi C pζ´zq2 dζ q “ 2πi C pζ´wq2 pζ´zq2 dζ.
When w « z, the integrand numerator f pζqp2ζ ´ z ´ wq « 2f pζqpζ ´ zq, while the
1 1 pzq ş f pζq
denominator « pζ ´ zq4 . So we expect that lim f pwq´f w´z
2
“ 2πi C pζ´zq3 dζ. We
wÑz
prove this claim next. For w ‰ z,
f 1 pwq´f 1 pzq 2
ş f pζq 1
ş p3ζ´z´wqf pζq
w´z ´ 2πi C pζ´zq3 dζ “ pw ´ zq 2πi C pζ´wq2 pζ´zq3 dζ. p˚q
If we show that the integral is bounded by some constant for all w close enough to
z, then since the integral is being multiplied by w ´ z, we see that as w Ñ z, we
can make the overall expression as small as we please. To this end, let us consider
a disc with centre z and having radius smaller than r, say r{2, and we will confine
w (which is anyway supposed to be near z) to lie within this disc. As f P OpDq,
in particular, its restriction to the C is continuous, and so is |f | : C Ñ R. But by
the Extreme Value Theorem, M :“ maxζPran C |f pζq| “ maxtPr0,1s |f pCptqq| exists.
Moreover, for ζ P ran C,
|ζ ´ w| ě 2r ,
|3ζ ´ z ´ 2w| “ |3ζ ´ 3z ` 2z ´ 2w| ď 3r ` 2 2r “ 4r, |ζ ´ z|3 “ r 3 .
ş
By the M L-inequality, | C p3ζ´z´2wqf pζq
pζ´wq2 pζ´zq3
dζ| ď 2πr p 4rM
r 2 3 “
q r
32πM
r3
. Using (˚),
2
1 1 pzq ş f pζq wÑz
| f pwq´f
w´z
2
´ 2πi 2 32πM 32M
C pζ´zq3 dζ| ď |w ´ z| 2π r 3 “ r 3 |w ´ z| ÝÑ 0.
Thus f 1 is complex differentiable at z. As z P D was arbitrary, f 1 P OpDq.
pnq
Exercise 3.38. If D is a domain and f P OpDq, show that for all n P N, f has a
continuous complex derivative.
Exercise 3.39. Let U Ă C be open, f P OpU q, u :“ Re f and v :“ Im f . Show that
u, v P C 8 pU q, where C 8 pU q denotes the set of all real-valued functions on U possessing
continuous partial derivatives of all orders on U .
3.8. Liouville’s Theorem, Fundamental Theorem of Algebra

Theorem 3.17 (Liouville’s Theorem). Every bounded entire function is constant.
Let us contrast this with the situation in real analysis. For example R Q x ÞÑ sin x
is real differentiable on R, and is bounded (| sin x| ď 1 for all x P R), but is not
constant. On the other hand, in light of the above Liouville’s Theorem, the entire
(complex) function sin, being nonconstant, must necessarily be unbounded on C!
We had checked this earlier on page 24: | sinpiyq| Ñ 8 as y Ñ ˘8.
Proof. Let M ě 0 be such that for all z P C, |f pzq| ď M . Suppose z P C, R ą 0,

and let CR ptq :“ z ` Re2πit , t P r0, 1s. From (‹) in the proof of Corollary 3.16,
1
ş f pζq
f 1 pzq “ 2πi CR pζ´zq2 dζ. By the M L-inequality,
1
ş f pζq 1 M M
|f 1 pzq| “ | 2πi CR pζ´zq2 dζ| ď 2π R2 2πR “ R .
As R ą 0 was arbitrary, f 1 pzq “ 0. So f 1 pzq “ 0 for all z P C. For zş P C, let r0, zs be

the straight line path joining 0 to z. Then we get f pzq ´ f p0q “ r0,zs f 1 pζq dζ “ 0.
So for all z P C, f pzq “ f p0q, that is, f is constant.
The Fundamental Theorem of Algebra1 can be proved using Liouville’s Theorem.
Corollary 3.18 (Fundamental Theorem of Algebra).

Every polynomial of degree at least 1 has a zero in C.
For a polynomial p : C Ñ C, given by ppzq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d , z P C, where

c0 , c1 , ¨ ¨ ¨ , cd P C, cd ‰ 0, the number d is called the degree of p. A number ζ P C
such that ppζq “ 0 is called a zero of p.
Proof. Let ppzq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d be a polynomial with degree d ě 1, such

that it has no zero in C, i.e., for all z P C, ppzq ‰ 0. Thus p : C Ñ Czt0u. The
function g :“ 1z P OpCzt0uq, and so the composition map f :“ g ˝ p “ 1p , is entire.
Recall Exercise 1.37, which gives an estimate on the growth of a polynomial: There
exist m, R ą 0 such that |ppzq| ě m|z|d ě mRd for all |z| ą R. So |f pzq| ď mR 1
d for
|z| ą R. By Weierstrass’s Theorem, in the compact set tz P C : |z| ď Ru, the real-
valued continuous function |p| has a positive minimum k (as p is never 0). Thus
1 1
|f pzq| ď maxt mR d , k u for all z P C. By Liouville’s Theorem, f must be constant,
1 1
say C. So for all |z| ą R, m|z| d ě |ppzq| “ |C|, giving |C| “ 0 and so C “ 0. Hence
1
for all z P C, ppzq “ C “ 0. Multiplying by ppzq, we get 1 “ 0, a contradiction.
Exercise 3.40. Let f be an entire function such that f is bounded away from 0, that is,
there is a δ ą 0 such that for all z P C, |f pzq| ě δ. Show that f is a constant.
Exercise 3.41. Let w0 P C and r ą 0. Show that an entire function, whose range is
disjoint with the disc tw P C : |w ´ w0 | ă ru, must be a constant.
Exercise 3.42. Let f P OpCq, and periodic in the real and in the imaginary directions,
i.e., there exist positive T1 , T2 in R such that f pzq “ f pz ` T1 q “ f pz ` iT2 q for all z P C.
Prove that f is constant. Give examples to show that periodicity in just the real direction
or just the imaginary direction is not enough.
Exercise 3.43. A classical theme in the theory of entire functions is to try to characterise
the entire function f based on the way |f | grows for large |z|. Here is one instance of this.
(1) Show that if f is entire, and |f pzq| ď | exp z| for all z P C, then in fact f is equal to
c exp z for some complex constant c with |c| ď 1. (Thus if a nonconstant entire function
‘grows’ no faster than the exponential function, it is an exponential function.)
(2) One may be tempted to argue that this cannot be right on the grounds that ‘polynomials
grow slower than the exponential function’, but surely p ‰ exp z. Show that if p is a
polynomial satisfying |ppzq| ď | exp z| for all z P C, then p ” 0. Hint: Take z “ x ă 0.
1Despite its name, there is no purely algebraic proof of the theorem, since it turns out that any proof must
use the completeness of the reals.
3.9. Morera’s Theorem 71
Exercise 3.44. Suppose f is an entire function, and satisfies |f pzq| ą |f 1 pzq| for all z P C.
Prove that there exists a constant C P C such that for all z P C, f 1 pzq “ Cf pzq. Calculate
peĆz f q1 , and hence determine all such f .
Exercise 3.45. Let M P R, and f be an entire function such Re f pzq ď M for all z P C.
Prove that f is a constant function.
Exercise 3.46. Suppose that f is an entire function. Let u :“ Re f and v :“ Im f .
Suppose further that upx, yq ď vpx, yq for all px, yq P C. Prove that f 1 pzq “ 0 for all z P C.
Hint: Consider exppf ` if q.
Exercise 3.47. Let f : C Ñ C be an entire function. Suppose that a1 , a2 are complex
numbers such that a1 ‰ a2 , and such that a1 , a2 are contained in the interior of the circular
path C with radius R ą 0 and centre 0, traversed once in the counterclockwise direction.
ş pzq
(1) Prove that | C pzáf1 qpzá2q
dz| ď pR´|a12πR
|qpR´|a2 |q max |f pzq|.
zPC
1 α β
(2) Find α, β P C such that for all z P C, pzá1 qpzá2 q “ zá1 ` zá2 .
ş pzq ş f pzq ş f pzq
(3) Express C pzáf1 qpzá2q
dz in terms of C zá1 dz and C zá2 dz.
Use the Cauchy Integral Formula to simplify these latter expressions.
(4) Deduce Liouville’s Theorem.
Exercise 3.48. Throughout this exercise, the polynomial p has degree d ě 1. Using
the Division Algorithm and the Fundamental Theorem of Algebra, p can be factored as
ppzq “ Cpz ´ α1 qm1 ¨ ¨ ¨ pz ´ αk qmk , where m1 , ¨ ¨ ¨ , mk P N, m1 ` ¨ ¨ ¨ ` mk “ d, C ‰ 0, and
α1 , ¨ ¨ ¨ , αk are the distinct zeroes of p. A zero α of p is called simple if ppzq “ pz ´ αqqpzq,
where q is a polynomial such that qpαq ‰ 0. Show that α is a simple zero of p if and only
if ppαq “ 0 and p1 pαq ‰ 0. Show that there exists a w P C such that the polynomial pr,
prpzq “ ppzq ` w (z P C), has only simple zeroes.
Exercise 3.49. Let f be an entire function such that f p0q “ 0 and |f pzq ´ ez sin z| ă 3
for all z P C. Find f .
3.9. Morera’s Theorem

The Cauchy Integral Theorem implies the following:
Let D be a domain. If f P OpDq, then for şany disc ∆ Ă D, and
any closed piecewise smooth path γ in ∆, γ f pzq dz “ 0.
Morera’s theorem is a converse to the above: If the contour integral is zero for
some special paths, then f is holomorphic.
Theorem 3.19 (Morera’s Theorem).
Let D be a domain, and f : D Ñ C be continuous. Ifş for every disc ∆ Ă D, and for
every closed rectangular path γ in D, we have that γ f pzqdz “ 0, then f P OpDq.
Proof. Let z0 P D, and let ∆ be a disc with centre z0 such that ∆ Ă D, and γz0 ,z
is the path joining z0 to z by first
ş moving horizontally and then moving vertically.
Define F : ∆ Ñ C by F pzq “ γz ,z f pζq dζ for all z P ∆.
0
∆
z
z0
γz0 ,z
We will show that F is holomorphic in ∆, and its derivative is f . This shows that
f is holomorphic in ∆! Why? Because f (being the derivative of a holomorphic
function) is itself then holomorphic in ∆ by Corollary 3.16.
Let z P ∆, and ǫ ą 0. As f is continuous, there exists δ ą 0 such that if w P ∆
satisfies |w ´ z| ă δ, then |f pwq ´ f pzq| ă ǫ. We have
ş ş
F pwq ´ F pzq “ γz ,w f pζq dζ ´ γz ,z f pζq dζ.
0 0
ş
As the integral of f on closed rectangular paths is zero, F pwq´F pzq “ γz,w f pζq dζ,
where γz,w is the path joining z to w by again first moving horizontally and then
moving vertically. See the picture below which shows one particular case:
ş ş ş ş ş ş ş
F pwq ´ F pzq “ γz ,w f pζq dζ ´ γz ,z f pζq dζ “ ✓✓A ` B ` C ´p✓ ✓
A ` D qf pζq dζ
ş 0
ş ş 0
ş ş ş
“ p B ` C ` ´γz,w ` ´D q ` γz,w f pζq dζ “ p 0qq ` γz,w f pζq dζ.
δz γz,w
w
D
z0
C
A B
Thus for 0 ă |w ´ z| ă δ,
F pwq´F pzq 1
ş 1
ş 1
ş
w´z ´ f pzq “ w´z γz,w f pζq dζ ´ f pzq w´z γz,w 1 dζ “ w´z γz,wpf pζq ´ f pzqq dζ.
Here we have used the Fundamental ş Theorem of Contour Integration for the holo-
morphic function 1 to obtain γz,w 1 dζ “ w ´ z. From the above, and using the fact
that the length of γz,w is |Repw ´ zq| ` |Impw ´ zq| ď 2|w ´ z|, we have
ş
| F pwq´F
w´z
pzq 1
´f pzq| “ | w´z ǫ
γz,wpf pζq ´f pzqq dζ| ď |w´z| p|Repw ´ zq|`|Impw ´ zq|q ă 2ǫ.
This completes the proof.

Exercise 3.50. Let f : C Ñ C be continuous, and f P OpCzpRˆt0uqq. Show that f P OpCq.
3.10. Appendix
In this appendix, we will prove the Cauchy Integral Theorem without the simpli-
fying assumptions used earlier. We first prove the following.
Theorem A (Cauchy-Goursat). ş Let U be an open subset of C, and f P OpU q.
Then for any triangle ∆ Ă U , B∆ f pzq dz “ 0.
3.10. Appendix 73
(Here a triangle ∆ is considered open, i.e., the open region contained within the
three sides constituting its boundary B∆. Also, we denote the counterclockwise
path starting at any vertex traversed once along the boundary also by B∆.)
ş
Proof. f is continuous in U . Let ∆0 Ă U be a triangle. Set I “ B∆0 f pzq dz. Divide
∆0 in 4 triangles ∆10 , ∆20 , ∆30 , ∆40 , using the midpoints of the sides of ∆0 . Then
ş 4 ş
ř
I “ B∆0 f pzq dz “ B∆j
f pzq dz. (Why?)
j“1 0
∆20
∆40
∆30 ∆10
ş |I|
So there is a j such that | B∆j0
f pzq dz| ě 4 (otherwise |I| ă 4 |I|
4 , a contradiction).
Let ∆1 be one of the ∆j0 s with this property. The perimeter of B∆1 is half that of
B∆0 . Repeating this process of dividing a triangle into four smaller triangles, we
get a decreasing sequence of triangles ∆0 Ą ∆1 Ą ∆2 Ą ¨ ¨ ¨ , such that
ş
| B∆k f pzq dz| ě |I|
4k
, and the perimeter ℓpB∆k q “ ℓpB∆
2k
0q
.
Let c P Xkě0 ∆k ‰ 2 H, where ∆k “ ∆k Y B∆k . Since f is complex differentiable
at c, given ǫ ą 0, there exists a δ ą 0 such that whenever |z ´ c| ă δ, we have
|f pzq ´ f pcq ´ f 1 pcqpz ´ cq| ď ǫ|z ´ c|. Take k large enough
ş so that ∆k Ă Dpc, δq. By
the Fundamental Theorem of Contour Integration, B∆k pf pcq ` f 1 pcqpz ´ cqqdz “ 0.
So ş ş
| B∆k f pzq dz| “ | B∆k pf pzq ´ f pcq ´ f 1 pcqpz ´ cqqdz|
ď p sup |f pzq ´ f pcq ´ f 1 pcqpz ´ cq|qq ℓpB∆k q
zPB∆k
ď p sup ǫ|z ´ c|qq ℓpB∆k q
zPB∆k
ď ǫ ℓpB∆k q ℓpB∆k q (Why?)
2
“ ǫ pℓpB∆
4k
0 qq
.
c
z
ş 2
For large enough k, |I|
4k
ď | B∆k f pzq dz| ď ǫ pℓpB∆
4k
0 qq
, that is, |I| ď ǫpℓpB∆0 qq2 . As
ǫ ą 0 was arbitrary, |I| “ 0, and so I “ 0.
Theorem B. Let z0 P C, r ą 0, and f P OpDpz0 , rqq. Then there exists an
F P OpDpz0 , rqq such that F 1 “ f in Dpz0 , rq.
2For example, the centres c of ∆ form a bounded sequence in R2 , and by the Bolzano-Weierstrass
k k
Theorem, this sequence has a convergent subsequence, say pcnk qkPN , with limit, say, c P R2 . For each
n P N Y t0u, there exists a K P N such that for all k ě K, nk ą n, and so pcnk qkěK is a sequence in the
closed set ∆n , showing that its limit c also belongs to ∆n . As n P N was arbitrary, c P Xně0 ∆n .
Proof. For z P Dpzş0 , rq, let rz0 , zs be the straight line segment joining z0 to z.
Define F by F pzq “ rz0 ,zs f pζq dζ for all z P Dpz0 , rq. For w P Dpz0 , rq and w ‰ z,
ş ş ş
we get by Theorem A that rz0 ,zs f pζq dζ ` rz,ws f pζq dζ ` rw,z0 s f pζq dζ “ 0, i.e.,
ş
F pwq ´ F pzq “ rz,ws f pζq dζ. So for such w,
ş
ş | rz,ws pf pζq´f pzqqdζ|
| F pwq´F
w´z
pzq 1
´ f pzq| “ | w´z rz,ws f pζq dζ ´ f pzq| “ |w´z|
|w´z| max |f pζq´f pzq|
ζPrz,ws
ď |w´z| “ max |f pζq ´ f pzq|.
ζPrz,ws
Given ǫ ą 0, let δ P p0, rq be such that whenever |ζ ´ z| ă δ, |f pζq ´ f pzq| ă ǫ.

Then for 0 ă |w ´ z| ă δ, every ζ P rz, ws satisfies |ζ ´ z| ă δ, and so we obtain
| F pwq´F
w´z
pzq
´ f pzq| ď max |f pζq ´ f pzq| ă ǫ. Thus F 1 pzq “ f pzq.
ζPrz,ws
Cauchy Integral Theorem

ş γ0 , γ1 : r0, ş1s Ñ D be smooth, closed, D-homotopic paths, and
Let D be a domain,
f P OpDq. Then γ0 f pzq dz “ γ1 f pzq dz.
Proof. Let I “ r0, 1s and H : I 2 Ñ D be a homotopy taking γ0 to γ1 . Since H is

continuous and I 2 is compact, H is uniformly continuous and HpI 2 q is a compact
subset of D. Define ρpwq “ inf zPCzD |w ´ z| for w P C. For w, w1 P C, and z P CzD,
ρpwq ď |w ´ z| ď |w ´ w1 | ` |w1 ´ z|, giving ρpwq ď |w ´ w1 | ` ρpw1 q. Interchanging
w, w1 , we arrive at |ρpwq´ρpw1 q| ď |w´w1 |, showing that ρ is continuous. We claim
that ρ is pointwise positive on the compact set HpI 2 q. Let w P HpI 2 q. Clearly
ρpwq ě 0. If ρpwq “ 0, then there exists a sequence pzn qnPN in CzD converging to
w, and as CzD is closed, w P CzD Ă CzHpI 2 q, a contradiction. So the minimum
value of the continuous function ρ on the compact set HpI 2 q is positive. Thus
r :“ inf |w´z| ą 0.
wPHpI 2 q,zPCzD
p1, 1q ζj,k`1
ζj`1,k`1
p j`1 k`1
n , n q
Sjk
I2 Pjk
p nj , nk q
ζjk
p0, 0q ζj`1,k
There is an n P N such that if ps ´ s1 q2 ` pt ´ t1 q2 ă n42 , then |Hps, tq ´ Hps1 , t1 q| ă r.

Let ζjk :“ Hp nj , nk q, 0 ď j, k ď n, and?put Sjk “ r nj , j`1 k k`1
n sˆr n , n s, 0 ď j, k ď n´1.
The diameter of the square Sjk is n2 , and so HpSjk q Ă Dpζjk , rq. So if we let
Pjk be the (possibly degenerate) quadrilateral rζjk , ζj`1,k , ζj`1,k`1 , ζj,k`1 , ζjk s, then
because discs are convex, Pjk Ă Dpζjk , rq. But by Theorem A,
ş
Pjk f pzq dz “ 0. (Why?) (‹)
3.10. Appendix 75
ş ş
It can now be shown that γ0 f pzq dz “ γ1 f pzq dz by going up the ladder we have
constructed, one rung at a time, as explained below.
ζ1,k
H
ζ0k
p n0 , nk q p nn , nk q Qk
ζn´1,k
Let Qk be the closed polygon rζ0k , ζ1k , ¨ ¨ ¨ , ζnk s, 0 ď k ď n. We will show that
ş ş ş ş ş
γ0 f pzq dz “ Q0 f pzq dz “ Q1 f pzq dz “ ¨ ¨ ¨ “ Qn f pzq dz “ γ1 f pzq dz.
ş ş
To see that γ0 f pzq dz “ Q0 f pzq dz, observe that if σj ptq “ γ0 ptq for nj ď t ď j`1
n ,
then σj ` rζj`1,0 , ζj0 s is a closed smooth path in the disc Dpζj0, rq Ă D.
σj p j`1 j`1 j`1
n q “ γ0 p n q “ Hp n , 0q “ ζj`1,0
γ0
σj p nj q “ γ0 p nj q “ Hp nj , 0q “ ζj0
Hence by Theorem B and the Fundamental Theorem of Contour Integration,

ş ş ş
σj f pzq dz “ ´ rζj`1,0 ,ζj0 s f pzq dz “ rζj0 ,ζj`1,0 s f pzq dz.
ş ş
Adding both sides of this equation for 0 ď j ď n yields γ0 f pzq dz “ Q0 f pzq dz.
ş ş
Similarly, γ1 f pzq dz “ Q1 f pzq dz.
ş ş
To see that Qk f pzq dz “ Qk`1 f pzq dz, 0 ď k ď n ´ 1, we use (‹). Note that
ř ş
n´1
0“ Pjk f pzq dz. (‹‹)
j“0
ş
Pjk f pzq dz includes the integral over rζj`1,k , ζj`1,k`1 s, which is the negative ş
of the
integral over rζj`1,k`1 , ζj`1,k s, which in turn is a part of the integral Pj`1,k f pzq dz.
ζj`2,k`1
ζj,k`1
ζj`1,k`1
ζj`2,k
ζjk
ζj`1,k
Moreover, ζ0,k “ Hp0, nk q “ Hp1, nk q “ ζn,k , so that rζş0,k`1 , ζ0,k s “ş ´rζn,k , ζn,k`1s.
Taking these cancellations into account, (‹‹) gives 0 “ Qk f pzqdz ´ Qk`1f pzqdz.
ζ2,k`1
21
42
ζ2k
31
32
11
12
22 ζ3k “ ζ0k 1 ζ1k 41

4 2
ζ3,k`1 “ ζ0,k`1 ζ1,k`1
3
Illustrative example with n “ 3, showing the ‘cancellations’ arising in (‹‹):

ş ş ş
0 “ P0k ` P1k ` P2k f pzq dz
ş ş ş ş ş ş ş ş ş ş ş ş
“ 1 `❙❙2 ` 3 `✓✓
4 ` 11 ` ❙
✓✓
21 ` 31 ` ❙
❙ ❙41 ` 12 `��2 ` 32 `❅ �
2 f pzq dz
2 �4❅
ş ş ş ş ş ş ş ş
“ 1 ` 11 ` 12 ` 3 ` 31 ` 32 f pzq dz “ Qk ´ Qk`1 f pzq dz.
Chapter 4
Taylor and Laurent series
In this chapter, firstly, we will learn that a complex power series

8
ř
cn pz ´ z0 qn
n“0
converges in some disc Dpz0 , Rq and diverges for |z ´ z0 | ą R. Also, in the disc
Dpz0 , Rq, it defines a holomorphic function. Vice versa, if f P OpDpz0 , Rqq, then it
has a power series expansion as above, called its ‘Taylor series’ (centred at z0 ). We
will use this to show the general Cauchy Integral Formula, and to understand the
nature of zeroes of holomorphic functions. Moreover, we will prove two additional
fundamental theorems about holomorphic functions: the Identity Theorem and the
Maximum Modulus Theorem.
z0 z0
8
ř
Taylor series: cn pz ´ z0 qn ř
Laurent series: cn pz ´ z0 qn
n“0 nPZ
In the second part of the chapter, we will study ‘double-sided’ series

ř
cn pz ´ z0 qn ,
nPZ
where now negative integer powers of z´z0 also occur. We will show that in general
such a ‘Laurent series’ converges in an annulus A :“ tz P C : r ă |z ´ z0 | ă Ru,
and defines a holomorphic function there. Vice versa, any f P OpAq possesses a
Laurent series expansion as above. We will use this result to classify ‘singularities’
of functions that are holomorphic in a punctured disc Dpz0 , Rqztz0 u, and to learn
‘Residue Calculus’, which in turn can be useful to evaluate some real integrals and
real series, as we will see at the end of this chapter.
77
78 4. Taylor and Laurent series
4.1. Series
Just as with real series, given a sequence pan qnPN of complex numbers, one can
form the sequence psn qnPN of its partial sums, where sn :“ a1 ` ¨ ¨ ¨ ` an for all n P N.
8
ř
‚ The series an converges if psn qnPN converges.
n“1
8
ř 8
ř
‚ If an converges, then the sum of the series is an :“ lim sn .
n“1 n“1 nÑ8
8
ř
‚ The series an diverges if psn qnPN diverges.
n“1
8
ř 8
ř
‚ The series an converges absolutely if the real series |an | converges.
n“1 n“1
From the result in Exercise 1.38, which says that a complex sequence converges if
and only if the sequences of its real and imaginary parts converge, it follows that
8
ř 8
ř 8
ř
an converges ô the real series Re an and Im an converge.
n“1 n“1 n“1
Thus the results from real analysis are useful in testing the convergence of complex
series. For example, it is easy to prove the following two facts, which we leave as
exercises.
ř
8
Exercise 4.1. If an converges, then show that lim an “ 0.
n“1 nÑ8
ř
8 ř
8
Exercise 4.2. If an converges absolutely, then show that an converges.
n“1 n“1
ř
8
Exercise 4.3. If an converges, and ǫ ą 0, then show that there exists an N P N such
n“1
ř
8 ř
8
that an converges and | an | ă ǫ.
n“N `1 n“N `1
ř
8 ř
8
1
Exercise 4.4. Show that if |z| ă 1, then zn converges and that zn “ 1´z .
n“0 n“0
ř
8
1
Exercise 4.5. Show that if |z| ă 1, then nz n´1 “ p1´zq2 .
n“1
1 1 1
Exercise 4.6. Show that 1s ` 2s ` 3s ` ¨ ¨ ¨ converges for all complex s with Re s ą 1.
ř
8
1
Thus s ÞÑ ζpsq :“ ns is a well-defined map in the half-plane ts P C : Re s ą 1u, and is
n“1
called the Riemann zeta function.
The Riemann zeta function is linked with number theory through the Euler Product For-
mula: If p1 , p2 , p3 , ¨ ¨ ¨ is the infinite list of primes in increasing order, then
ś K
1
ζpsq “ lim ´s (Re s ą 1).
KÑ8 k“1 1´pk
Riemann (1826-66) showed that the function ζ can be extended holomorphically to Czt1u.
It can be shown that the function ζ has zeroes at ´2, ´4, ´6, ¨ ¨ ¨ , called ‘trivial zeroes’,
but it also has other zeroes. All the non-trivial zeroes Riemann computed turned out to
lie on the line Re s “ 21 . This led him to formulate the following conjecture (now called the
Riemann Hypothesis), which is a famous unsolved problem in Mathematics: All non-trivial
zeroes of the Riemann zeta function lie on the line Re s “ 21 .
4.2. Power series 79
4.2. Power series

Region of convergence.
Let pcn qnPN be a complex sequence (thought of as a sequence of ‘coefficients’).
8
ř
Then cn z n is called a power series in the complex variable z.
n“0
We imagine putting in specific values of z in the above series. Then for some z P C,
the power series will converge, while for other values of z it may diverge.
8
ř
A fundamental question is: For which z P C does the power series cn z n converge?
n“0
Example 4.1. All polynomial expressions are power series, with only finitely many
nonzero coefficients. Polynomials converge for all z P C.
8
ř
Example 4.2. From Exercise 4.4, the power series zn converges if |z| ă 1.
n“0
8
ř
As |z n ´ 0| “ |z|n ě 1 for all n P N, �p lim z n “ 0q. So z n diverges if |z| ě 1.
nÑ8 n“0
The following result gives a general answer to the fundamental question above.
Theorem 4.1.
Let pcn qnPN be a sequence in C. Then exactly one of the following hold :
8
ř
‚ Either cn z n is absolutely convergent for all z P C,
n“0
‚ or there is a unique nonnegative real number R such that :
8
ř
(1) cn z n is absolutely convergent for all z P C with |z| ă R, and
n“0
8
ř
(2) cn z n is divergent for all z P C with |z| ą R.
n“0
The unique R ą 0 in the above theorem is called the radius of convergence of the
power series, and if the power series converges for all z P C, then we say that the
power series has an infinite radius of convergence, and write ‘R “ 8’.
diverge
R
z0
converge
What happens on the circle |z| “ R? Complex power series may diverge at every
point on the boundary (given by |z| “ R), or diverge on some points of the bound-
ary and converge at other points of the boundary, or converge at all the points on
the boundary. There is no general result answering what happens at each point on
the circle, and the specific power series at hand has to be investigated to find out
the behaviour.
Proof of Theorem 4.1.

8
ř
Let S :“ ty P r0, 8q : Dz P C such that y “ |z| and cn z n convergesu. Then 0 P S.
n“0
Only two cases are possible:
1˝ S is not bounded above. We will show that the radius of convergence is infinite.
Given z P C, there exists a y P S such that |z| ă y. But as y P S, there exists
ř8
a z0 P C such that y “ |z0 | and cn z0n converges. So its nth term tends to 0
n“0
as n Ñ 8, and in particular, the sequence of terms is bounded: |cn z0n | ď M
for all n P N. Then with r :“ |z|z|0 | (ă 1), we have |cn z n | “ |cn z0n |p |z|z|0 | qn ď M r n
8
ř
for all n P N. But M r n converges (r ă 1), and so by the Comparison Test,
n“0
8
ř
cn z n is absolutely convergent. Since z P C was arbitrary, the claim follows.
n“0
2˝ S is bounded above. We will show that the radius of convergence is sup S, i.e.,
8
ř
(a) if |z| ă sup S, then cn z n converges absolutely, and
n“0
8
ř
(b) if |z| ą sup S, then cn z n diverges.
n“0
If z P C and |z| ă sup S, then by the definition of supremum, there exists a

y P S such that |z| ă y. Then we repeat the proof in 1˝ . Since y P S, there
8
ř
exists a z0 P C such that y “ |z0 | and cn z0n converges. So its nth term tends
n“0
to 0 as n Ñ 8, and in particular, the sequence of terms is bounded: |cn z0n | ď M .
Then with r :“ |z|z|0 | (ă 1), we have |cn z n | “ |cn z0n |p |z|z|0 | qn ď M r n for all n P N.
8
ř 8
ř
But M r n converges (as r ă 1), and so cn z n is absolutely convergent.
n“0 n“0
If z P C and |z| ą sup S, then setting y :“ |z|, y R S, and by the definition of S,

8
ř
cn z n diverges (otherwise y P S).
n“0
r
R
R 11
00
00
11
0 z
r have the property described in the theorem and R ă R.

Uniqueness of R: Let R, R r
r 8
ř 8
ř
R`R r
We have R ă r :“ 2 ă R. As r ă R, r n
cn r converges. As R ă r, cn r n
n“1 n“0
diverges, a contradiction.
8
ř 8
ř
Instead of the power series cn z n ‘centred at 0’, consider cn pz ´ z0 qn centred
n“0 n“0
8
ř
at z0 P C. Applying Theorem 4.1 to cn wn , and taking w “ z ´ z0 , we obtain:
n“0
Corollary 4.2.
Let pcn qnPN be a sequence in C, and z0 P C. Then exactly one of the following hold :
8
ř
‚ Either cn pz ´ z0 qn is absolutely convergent for all z P C,
n“0
‚ or there is a unique nonnegative real number R such that :
8
ř
(1) cn pz ´ z0 qn is absolutely convergent for all z P C with |z ´ z0 | ă R.
n“0
8
ř
(2) cn pz ´ z0 qn is divergent for all z P C with |z ´ z0 | ą R.
n“0
ř
8
Exercise 4.7. Let R denote the radius of convergence of cn pz ´ z0 qn . Show that:
n“0
ř
8
‚ If cn pz ´ z0 qn converges for a z˚ P C, then R ě |z˚ ´ z0 |.
n“0
ř
8
‚ If cn pz ´ z0 qn diverges for a z˚ P C, then R ď |z˚ ´ z0 |.
n“0
ř
N ř!
N ř
8
Exercise 4.8. Let N P N and z P C. Show that |z n! | ď |z n |. Prove that z n!
n“1 n“1 n“1
ř
8
converges in D :“ tz P C : |z| ă 1u. Determine the radius of convergence of z n! .
n“1
8
ř
Theorem 4.3. Consider the power series cn z n . If L :“ lim | cn`1
cn | exists, then:
n“0 nÑ8
1
‚ If L ‰ 0, then the radius of convergence is L.
‚ If L “ 0, then the radius of convergence is infinite.
Proof.
1
1˝ Let L ‰ 0. For all nonzero z such that |z| ă L there exists a q ă 1 and an
|cn`1 z n`1 |
N large enough such that |cn z n | “ | cn`1
cn ||z| ď q ă 1 for all n ą N . (This is
nÑ8 L|z|`1
because | cn`1
cn z| ÝÑ L|z| ă 1. So we may take q “ 2 ă 1.) Thus by the
8
ř
Ratio Test, cn z n is absolutely convergent for such z.
n“0
|cn`1 z n`1 |
If |z| ą 1
L, then there is an N large enough such that |cn z n | “ | cn`1
cn ||z| ą 1
nÑ8 8
ř
for all n ą N , as | cn`1
cn z| ÝÑ L|z| ą 1. By the Ratio Test, cn z n diverges.
n“0
|cn`1 z n`1 |
2˝ Let L “ 0. For z P Czt0u, there is a q ă 1 such that |cn z n | “ | cn`1
cn ||z| ď q ă 1
nÑ8
for all n ą N . (This is because | cn`1
cn z| ÝÑ 0|z| “ 0 ă 1. So we may take
8
ř
1
q“ 2 ă 1.) By the Ratio Test, cn z n is absolutely convergent for such z.
n“0
1
pn`1q2 8 n
ř z
Example 4.3. As lim 1 “ 1, n2
converges if |z| ă 1 and diverges if |z| ą 1.
nÑ8 n“1
n2
n 8
ř 8 n
ř
If |z| “ 1, then | zn2 | “ n12 , and as 1
n2 converges, it follows that z
n2 is absolutely
n“1 n“1
convergent. Thus at every point of the circle |z| “ 1, the power series converges.
8
ř
The previous example is in contrast to the geometric series zn, where we had
n“0
convergence at no point of the circle |z| “ 1.
ř
8 a
Exercise 4.9. Consider the power series cn xn . If L :“ lim n |cn | exists, then:
n“0 nÑ8
‚ If L ‰ 0, then the radius of convergence is L1 .

‚ If L “ 0, then the radius of convergence is infinite.
ř
8
zn
Exercise 4.10. Show that nn converges for all z P C.
n“1
ř
8
Exercise 4.11. Show that nn z n converges only when z “ 0.
n“1
ř
8 p´1qn ř
8 ř
8
1 n
Exercise 4.12. Find the radius of convergence of: n z n, n2022 z n , z .n!
n“1 n“0 n“0
ř
8
Exercise 4.13. Let R be the radius of convergence of cn z n . Show that:
a n“0
(1) If p n |cn |qnPN is not bounded, then R “ 0.
a a
(2) If p n |cn |qnPN is bounded, and Mn :“ a supt m |cm | : m ě nu, n P N, then pMn qnPN is
convergent. Set L “ lim Mn“ lim sup n |cn |. If L “ 0, then R “ 8. If L ‰ 0, then R “ L1 .
nÑ8 nÑ8
8
ř 8
ř
Let an z n and bn z n be two power series, with radii of convergence Ra , Rb . Their
n“0 n“0
termwise sum is (absolutely) convergent in the common region of convergence,
8
ř
namely the disc Dp0, mintRa , Rb uq. Their Cauchy product is cn z n , where
n“0
ř
n
cn :“ an´k bk , for n P t0, 1, 2, 3, ¨ ¨ ¨ u.
k“0
Using Exercise 4.14 below, it can be shown that the Cauchy product of two power
series converges in the common region of convergence.
ř
8 ř
8 ř
n
Exercise 4.14. Suppose an and bn are absolutely convergent. Define cn “ an´k bk
n“0 n“0 k“0
ř
8 ř
8 ř
8 ř
8
for nonnegative integers n. Show that cn converges, and cn “ p an qp bn q.
n“0 n“0 n“0 n“0
th
Hint: If An , Bn , Cn denote the n partial sums, show that |C2n ´ An Bn | Ñ 0 as n Ñ 8.
Exercise 4.15. Let A and E be opposite vertices of an octagon ABCDEF GH. A frog
starts at vertex A. From any vertex except E it jumps to one of the two adjacent vertices.
When it reaches E it stops. Let an be the number of distinct paths of exactly n P N jumps
ending at E. To obtain a formula for an in terms of n, proceed as follows. Show that
a2n´1 “ 0 for all n P N. Also show that a2 “ 0 and a4 “ 2. Let bn be the number of
distinct paths of length n from C to E. Prove that for all n ą 1, a2n “ 2a2n´2 ` 2b2n´2
and b2n “ 2b2n´2 ` a2n´2 . Deduce that a2n`2 “ 4a2n ´ 2a2n´2 for all n ą 1. Define
ř
8
Apzq :“ a2n z n , taking a0 :“ 0. Show that p1 ´ 4z ` ?2z 2 qApzq “
?
2z 2 in the disc of
n“0 p2` 2qn´1? ´p2´ 2qn´1
convergence of the power series. Conclude that a2n “ 2
for all n P N.
Explain using the Binomial Theorem why the right-hand side is an integer.
4.2.1. Power series are holomorphic. We have seen that polynomials are
power series with an infinite radius of convergence, that is, they converge in the
whole of C. They are of course also holomorphic there. More generally, a power
8
ř
series f pzq :“ cn z n , which is convergent for |z| ă R, is holomorphic there, and
n“0
8
ř
d
f 1 pzq “ dz pc0 ` c1 z ` c2 z 2 ` ¨ ¨ ¨ q “ c1 ` 2c2 z ` 3c3 z 2 ` ¨ ¨ ¨ “ cn nz n´1
n“1
as expected (differentiating termwise, as with finite sums, namely polynomials).

8
ř
Theorem 4.4. Let R ą 0 and f pzq :“ cn z n converge for |z| ă R.
n“0
8
ř
Then f P OpDp0, Rqq and f 1 pzq “ ncn z n´1 for |z| ă R.
n“1
8
ř
Proof. First we will show that gpzq :“ ncn z n´1 “ c1 `2c2 z `¨ ¨ ¨ `ncn z n´1 `¨ ¨ ¨
n“1
is absolutely convergent for |z| ă R. Fix z and let r satisfy |z| ă r ă R. By
ř8
hypothesis cn r n converges, and so there is some positive number M such that
n“0
|z|
|cn r n | ă M for all n. Let ρ :“ r . Then 0 ď ρ ă 1, and
M nρn´1
|ncn z n´1 | “ |cn r n | 1r n| zr |n´1 ď r .
8
ř 8
ř
1
As nρn´1 converges (to p1´ρq2 ; Exercise 4.5), by the Comparison Test, ncn z n´1
n“1 n“1
is absolutely convergent.
Next, we will show that f 1 pz0 q “ gpz0 q for |z0 | ă R, i.e., lim p f pzq´f
z´z0
pz0 q
´gpz0 qq “ 0.
zÑz0
As before, let r be such that |z0 | ă r ă R. Below, we consider z satisfying |z| ă r.
z0
0
r
R
8
ř
Let ǫ ą 0. As ncn r n´1 converges absolutely, there exists an N P N such that
n“1
8
ř 8
ř
|ncn r n´1 | ă 4ǫ . We have f pzq ´ f pz0 q “ cn pz n ´ z0n q, and so for z ‰ z0 ,
n“N `1 n“1
f pzq´f pz0 q 8
ř z n ´z0n 8
ř
z´z0 “ cn z´z0 “ cn pz n´1 ` z n´2 z0 ` ¨ ¨ ¨ ` zz0n´2 ` z0n´1 q.
n“1 n“1
f pzq´f pz0 q 8
ř
Thus z´z0 ´ gpz0 q “ cn pz n´1 ` z n´2 z0 ` ¨ ¨ ¨ ` z0n´1 ´ nz0n´1 q. We let S1 be
n“1
the sum of the first N terms of this series (that is, from n “ 1 to n “ N ) and S2
be the sum of the remaining terms. Then since |z|, |z0 | ă r, it follows that
8
ř 8
ř
n´1
|S2 | ď ` r n´1 ` ¨ ¨ ¨ ` r n´1 `nr n´1 q “
|cn |prlooooooooooooooomooooooooooooooon 2n|cn |r n´1 ă 2ǫ .
n“N `1 n“N `1
n terms
ř
N
S1 “ cn pz n´1 ` z n´2 z0 ` ¨ ¨ ¨ ` zz0n´2 ` z0n´1 ´ nz0n´1 q is a polynomial in z and so
n“1
ř
N ř
N
lim S1 “ cn pz0n´1 ` z0n´2 z0 ` ¨ ¨ ¨ ` z0 z0n´2 ` z0n´1 ´ nz0n´1 q “ cn pnz0n´1 ´ nz0n´1 q “ 0.
zÑz0 n“1 n“1
So there is a δ ą 0 such that whenever |z ´ z0 | ă δ, we have |S1 | ă 2ǫ . Thus for

|z| ă r and 0 ă |z ´ z0 | ă δ, we have | f pzq´f
z´z0
pz0 q
´ gpz0 q| ď |S1 | ` |S2 | ă 2ǫ ` 2ǫ “ ǫ.
This means that f 1 pz0 q “ gpz0 q, as claimed.
Remark 4.1. If pcn qnPN is a sequence of real numbers, then consider the real power
8
ř
series cn xn . From Real Analysis, we know that such a power series converges in
n“0
an interval of the form p´R, Rq and diverges in Rzr´R, Rs for some R ě 0.
Theorems 4.1 and 4.4 imply that if we replace the real variable x by a complex
variable z, then we can extend the real power series to a holomorphic function in the
disc Dp0, Rq in the complex plane. So we can view real analytic functions (namely
functions of a real variable having a local power series expansion) as restrictions
of holomorphic functions. This again highlights the interplay between the worlds
of real analysis and complex analysis. We have seen a previous instance of this
interaction when we studied the Cauchy-Riemann equations. ˚
8
ř
Corollary 4.5. Let R ą 0 and let f pzq :“ cn z n converge for |z| ă R.
8 n“0
ř
Then for k P N, and |z| ă R, f pkqpzq “ npn´1qpn´2q ¨ ¨ ¨ pn´k`1qcn z n´k . p˚q
n“k
f pnq p0q
In particular, cn “ n! for all n P N Y t0u.
Proof. A repeated application of Theorem 4.4 gives this: For n, k P N Y t0u,
"
npn ´ 1q ¨ ¨ ¨ pn ´ pk ´ 1qqz n´k for 0 ď k ď n,
pz n qpkq “
0 for k ą n.
For the last claim, we have f p0q “ c0 , and for the n P N cases, set z “ 0 in (˚):
8
ř
f pkq p0q “ kpk ´ 1q ¨ ¨ ¨ 1 ck ` z npn ´ 1q ¨ ¨ ¨ pn ´ k ` 1qcn z n´k´1 |z“0 “ k! ck .
n“k`1
There is nothing special about taking power series centred at 0; we have:

8
ř
Corollary 4.6. Let z0 P C, R ą 0. If f pzq :“ cn pz´z0 qn converges for |z´z0 | ă R,
8n“0
ř
then for k P N and |z ´ z0 | ă R, f pkq pzq “ npn ´ 1q ¨ ¨ ¨ pn ´ k ` 1qcn pz ´ z0 qn´k .
n“k
f pnq pz 0q
In particular, cn “ n! , for all n P N Y t0u.
8
ř
Proof. gpwq :“ cn wn converges for w P Dp0, Rq. So g P OpDp0, Rqq, and
n“0
8
ř
g 1 pwq “ ncn wn for all w P Dp0, Rq. Composing g with p¨ ´ z0 q|Dpz0 ,Rq , we have
n“1
that f p¨q “ gp¨ ´ z0 q P OpDpz0 , Rqq, and also it follows by the Chain Rule that
8
ř
f 1 pzq “ g 1 pz ´ z0 q ¨ 1 “ ncn pz ´ z0 qn´1 for all z P Dpz0 , Rq. The general case for
n“1
n ě 2 now follows by a repeated application of this last result.
4.3. Taylor series 85
Remark 4.2 (Uniqueness of coefficients). Suppose z0 P C, R ą 0, and the power

8
ř 8
ř
series cn pz´z0 qn , rcn pz´z0 qn both converge to the same function f in Dpz0 , Rq.
n“0 n“0
f pnq pz0 q
Then from the above, for n ě 0, we have cn “ n! “r
cn . ˚
Exercise 4.16. For |z| ă 1, what is 12 ` 22 z ` 32 z 2 ` 42 z 3 ` ¨ ¨ ¨ ?
ř
8
Exercise 4.17. True or false? All statements refer to power series cn z n .
n“0
(1) The set of points z for which the power series converges equals either the singleton set
t0u or some open disc of finite positive radius or the entire complex plane, but no other
type of set.
(2) If the power series converges for z “ 1, then it converges for all z with |z| ă 1.
(3) If the power series converges for z “ 1, then it converges for all z with |z| “ 1.
(4) If the power series converges for z “ 1, then it converges for z “ ´1.
(5) Some power series converge at all points of an open disc with centre 0 of some positive
radius, and also at certain points on the boundary of the disc (that is the circle bounding
the disc), and at no other points.
(6) There are power series that converge on a set of points which is exactly equal to the
closed disc given by |z| ď 1.
(7) If the power series diverges at z “ i, then it diverges at z “ 1 ` i as well.
Exercise 4.18 (Generalised Binomial Theorem). Let a P p0, 1q. Show that the radius of
convergence of the complex power series 1 ` az ` apa´1q
2! z 2 ` apa´1qpa´2q
3! z 3 ` ¨ ¨ ¨ is 1. Let
f P OpDp0, 1qq be the sum of the above power series. Prove that p1 ` zqf 1 pzq “ af pzq in
Dp0, Rq. Show that p1 ` ¨qa P OpDp0, 1qq. Calculate pp1 ` ¨qá f q1 in Dp0, 1q, and hence
show that f pzq “ p1 ` zqa in Dp0, 1q.
ř
8
p´1qn 2n
Exercise 4.19. Show that the complex power series J0 pzq “ 22n pn!q2 z converges for
n“0
ř
8
p´1qn n
all z P C. (Hint: Consider first 22n pn!q2 w .) J0 is called the Bessel function of order 0.
n“0
Prove that J0 solves the differential equation f 2 pzq ` 1z f 1 pzq ` f pzq “ 0 in Czt0u.
ř
8 ř
8
Exercise 4.20. Let pcn qnPN be a sequence in C such that |cn | converges, but n|cn |
n“0 n“0
ř
8
does not. Show that the radius of convergence of cn z n is equal to 1.
n“0
4.3. Taylor series

8
ř
We have seen that complex power series f :“ cn p¨ ´ z0 qn belong to OpDpz0 , Rqq,
n“0
where R is the radius of convergence of the power series. In this section, we will
show that conversely, if f P OpDpz0 , Rqq, for a z0 P C and R ą 0, then f has a power
8
ř
series expansion, f pzq “ cn pz ´ z0 qn for z P Dpz0 , Rq, where the coefficients cn ,
n“0
n ě 0, can be determined from the f . This power series is called the Taylor series
of f centred at z0 . Thus every holomorphic f defined in a domain D possesses a
power series expansion locally, in a disc around any point z0 P D. En route we will
also prove the general Cauchy Integral Formula, expressing the derivatives f pnq for
all n ě 0 of a holomorphic function f inside a disc from the values of f along its
bounding circle. The cases n “ 0, 1 were seen earlier in Sections 3.6 and 3.7.
Theorem 4.7. Let z0 P C, R ą 0, and f P OpDpz0 , Rqq. For any r P p0, Rq, define
1
ş f pζq
the circular path Cptq :“ z0 `re2πit for all t P r0, 1s, and let cn :“ 2πi C pζ´z0 qn`1 dζ
8
ř
for all integers n ě 0. Then f pzq “ cn pz ´ z0 qn for all z P Dpz0 , Rq.
n“0
r
z0 R
Proof. Let z P Dpz0 , Rq. Initially, let r be such that |z ´ z0 | ă r ă R. Then by

the Cauchy Integral Formula,
1
ş f pζq 1
ş f pζq 1
ş f pζq
f pzq “ 2πi C ζ´z dζ “ 2πi C ζ´z0 `z0 ´z dζ “ 2πi C z´z0 dζ.
pζ´z0 qp1´ ζ´z q
0
z´z0 |z´z0 |
If w :“ ζ´z0 , then |w| “ r ă 1. Thus
1 1 wn
z´z “ 1´w “ 1 ` w ` w2 ` w3 ` ¨ ¨ ¨ ` wn´1 ` 1´w
1´ ζ´z0
0
z´z0 pz´z0 qn´1 pz´z0 qn
“ 1` ζ´z0 ` ¨¨¨ ` pζ´z0 qn´1
` pζ´z0 qn´1 pζ´zq
,
1
ş f pζqpz´z0 qn
and so plugging this in the above, we have, with Rn pzq :“ 2πi C pζ´z0 qn pζ´zq dζ,
1
ş pz´z0 qn´1 pz´z0 q n
f pzq “ 2πi p 1
C f pζqp ζ´z0 ` ¨ ¨ ¨ ` pζ´z0 qn ` pζ´z0 qn pζ´zq q dζ
“ c0 ` c1 pz ´ z0 q ` ¨ ¨ ¨ ` cn´1 pz ´ z0 qn´1 ` Rn pzq.
So we will be done if we show that Rn pzq goes to 0 as n Ñ 8. As |f | is a continuous
real-valued function on the compact set ran C, i.e., there is an M ą 0 such that for
pz´z0 qn |z´z0 | n nÑ8
all ζ P ran C, |f pζq| ă M . Also, for ζ P ran C, | pζ´z 0q
n| “ p r q ÝÑ 0. The term
1
|ζ´z| for ζ P ran C, is bounded by the reciprocal of the ‘distance between the circle
ran C and z’. We have |ζ ´ z| “ |ζ ´ z0 ´ pz ´ z0 q| ě |ζ ´ z0 | ´ |z ´ z0 | “ r ´ |z ´ z0 |.
nÑ8 8
ř
Thus |Rn pzq| ď p |z´z 0| n M
r q r´|z´z0 | ÝÑ 0. Hence cn pz ´ z0 qn converges to f pzq.
n“0
C
z
ζ
z0
4.3. Taylor series 87
1
ş f pζq
We have only shown the series expansion with cn “ 2πi C pζ´z0 qn`1 dζ for r satisfy-
ing |z´z0 | ă r ă R, while the theorem statement said that we could have taken any
ş f pζq
r P p0, Rq. By the Cauchy Integral Theorem, C pζ´z 0q
n`1 dζ is independent of r, and
any value of r P p0, Rq can be chosen here: Indeed, p¨´zf0p¨q

qn`1
P OpDpz0 , Rqztz0 uq,
r
and if, besides C, C is another circular path with centre z0 and some other radius
r are Dpz0 , Rqztz0 u-homotopic.
rr P p0, Rq then C, C
8
ř
From Corollary 4.6, whenever a power series f pzq “ cn pz ´ z0 qn is convergent
n“0
f pnq pz 0q
for all z P Dpz0 , Rq, we have cn “ n! for all integers n ě 0.
Corollary 4.8 (Taylor Series). If D is a domain, f P OpDq, z0 P D, and R ą 0 is
1 f 2 pz0 q
such that Dpz0 , Rq Ă D, then f pzq “ f pz0 q ` f pz 0q
1! pz ´ z0 q `
2
2! pz ´ z0 q ` ¨ ¨ ¨
for all z P Dpz0 , Rq. Also, for any r P p0, Rq, if Cptq :“ z0 ` re 2πit for all t P r0, 1s,
n!
ş f pzq
then f pnq pz0 q “ 2πi C pz´z0 qn`1 dz for all integers n ě 0.
8
ř
Proof. Theorem 4.7 implies that f pzq “ cn pz ´ z0 qn for all z P Dpz0 , Rq, where
n“0
1
ş f pzq
cn “ 2πi C pz´z0 qn`1 dz for all n ě 0, Cptq :“ z0 ` re2πit for all t P r0, 1s, and r is
f pnq pz0 q
any number in p0, Rq. By Corollary 4.6, for all n ě 0, n! “ cn .
Corollary 4.9 (Cauchy Integral Formula). Let D be a domain, f P OpDq, z0 P D,
and ∆ :“ tz P C : |z ´ z0 | ď Ru Ă D. Let Cptq “ z0 ` Re2πit for all t P r0, 1s.
n!
ş f pζq
Then for all z P Dpz0 , Rq, and all integers n ě 0, f pnq pzq “ 2πi C pζ´zqn`1 dζ.
Proof. Let z P ∆. Then there exists a r ą 0 small enough so that Dpz, rq Ă ∆.

8
ř
By Theorem 4.7, f has a power series expansion f pwq “ cn pw´zqn (w P Dpz, rq),
n“0
1
ş f pζq f pnq pzq 2πit (t P r0, 1s), and ρ is any
where cn “ 2πi Cρ pζ´zq n`1 dζ “ n! , C ρ ptq “ z ` ρ e
f p¨q
number in p0, rq. But P OpDztzuq and Cρ is Dztzu-homotopic to C, and
p¨´zqn`1
ş f pζq ş f pζq
so by the Cauchy Integral Theorem, Cρ pζ´zq n`1 dζ “ C pζ´zqn`1 dζ.
8 exppnq p0q
ř
Example 4.4. exp P OpCq. So exp z “ n!zn for all z P C. As exp1 z “ exp z,
n“0 8 n
ř z
exppnq “ exp (n ě 0) and exppnq p0q “ exp 0 “ 1. Thus exp z “ n! (z P C).
n“0
Example 4.5. Log P OpDq, where D denotes the ‘plane with a cut’ given by
Czpp´8, 0s ˆ t0uq. The largest open disc with centre z0 “ 1, which is contained in
n
D, is Dp1, 1q. Since Log1 z “ z1 for z P D, it follows that Logpnq z “ p´1q zpn´1q!
n . In
pnq p´1qn pn´1q!
particular, Log 1 “ 1n . Also Log 1 “ 0. Thus
z´1 pz´1q2 p´1qn pz´1qn
Log z “ 1 ´ 2 ` ¨¨¨ ` n ` ¨ ¨ ¨ for all z P Dp1, 1q,
2 p´1qn wn
w w
or equivalently, Log p1 ` wq “ 1 ´ 2 ` ¨¨¨ ` n ` ¨ ¨ ¨ for all |w| ă 1.
3 5 2 4
Exercise 4.21. Show that sin z “ z´ z3! ` z5! ´ ` ¨ ¨ ¨ , cos z “ 1´ z2! ` z4! ´ ` ¨ ¨ ¨ for all z P C.
Exercise 4.22. Find the Taylor series of the polynomial z 6 ´ z 4 ` z 2 ´ 1 centred at 1.
ř
8
Exercise 4.23. For the f below, find cn , n ě 0, in the Taylor series cn z n of f .
ş 2 n“0 2
‚ f pzq “ r0,zs eζ dζ (z P C), where r0, zs is the straight path from 0 to z. Hint: f 1 pzq “ ez .
z2 1
‚ f pzq “ pz`1q2 (z P Czt´1u). Hint: For |z| ă 1, z`1 “ 1 ´ z ` z2 ´ ` ¨ ¨ ¨ .
ř
8
Exercise 4.24. Let f P OpDp0, 1qq be such that |f pzq| ă 1 for all z P Dp0, 1q. Let cn z n
n“0
be the Taylor series of f centred at 0. Prove that |cn | ď 1 for all n ě 0.
Exercise 4.25. True or false? There exists an r ą 0 and a function f P OpDp0, rqq such
that f pnq p0q “ pn!q2 for all n ě 0.
Exercise 4.26. Determine all entire functions f which satisfy f pz 2 q “ f pzq for all z P C.
Exercise 4.27. Let a0 , a1 , ¨ ¨ ¨ be a sequence of real numbers tending to infinity, and let
f : C Ñ C be an entire function satisfying |f pnq pak q| ď eák for all nonnegative integers k
and n. Prove that f pzq “ c e´z for some constant c P C with |c| ď 1.
Corollary 4.10 (Cauchy’s inequality). Let z0 P C, R ą 0, f P OpDpz0 , Rqq.
Suppose there exists an M ą 0 such that |f pzq| ď M for all z P Dpz0 , Rq. Then
for all n ě 0, |f pnq pz0 q| ď n!M Rn .
2πit
Proof. Let Cptq “ z0 ` re , t P r0, 1s, where r P p0, Rq. By Corollary 4.9,
ş f pzq f pzq
|f pnq pz0 q| “ | 2πi
n! n! n! M n!M
C pz´z0 qn`1 dz| ď 2π max | pz´z0 qn`1 | 2πr “ 2π r n`1 2πr “ r n .
zPC
The claim now follows by passing to the limit as r Õ R.
Exercise 4.28. Let f P OpCq such that there exists an M ą 0 and an integer n ě 0 such
that for all z P C, |f pzq| ď M |z|n . Use Cauchy’s inequality to prove f pn`1q pzq “ 0 for all
z P C, and show that f is a polynomial of degree at most n. What happens when n “ 0?
ş
Exercise 4.29. Evaluate C zsin z
2023 dz, where C is the circular path with centre 0 and radius
1 traversed once in the anticlockwise direction.
4.4. Classification of zeroes

Let f P OpDq, where D is a domain. A point z0 P D is a zero of f if f pz0 q “
0. Theorem 4.11 says that the zeroes of a nonzero f are ‘isolated’. This need
not happen with continuous functions (which are not as ‘rigid’ as holomorphic
functions): Zeroes of nonzero continuous functions need not be isolated.
Example 4.6. Define ϕ : R Ñ R by ϕpxq “ x sin x1 if x ‰ 0, and ϕp0q “ 0. Then ϕ
is continuous at any x ‰ 0. Moreover, |ϕpxq ´ ϕp0q| “ |x sin x1 ´ 0| ď |x| ¨ 1 “ |x ´ 0|
1 1
for all x ‰ 0, and so ϕ is continuous at 0. For all n P N, ϕp nπ q “ nπ sinpnπq “ 0,
2 2 π 2 2
while ϕp p4n`1qπ q “ p4n`1qπ sinp2nπ ` 2 q “ p4n`1qπ 1 “ p4n`1qπ . Define f : C Ñ C
by f pzq “ ϕpRe zq ` iϕpIm zq for all z P C. Then f is continuous. Since we have
f p0q “ ϕp0q ` iϕp0q “ 0, 0 is a zero of f . Also, 0 is not an isolated zero of f , since
1 1
f p nπ q “ ϕp nπ q ` iϕp0q “ 0 for all n P N. Finally, f is not identically zero in any
2 2 2
disc centred at 0: For all n P N, f p p4n`1qπ q “ ϕp p4n`1qπ q ` iϕp0q “ p4n`1qπ ‰ 0.
4.4. Classification of zeroes 89
Example 4.7. exp has no zeroes in C. Indeed, | exp z| “ eRe z ą 0 for all z P C.
?
cos z ´ 3 has infinitely many zeroes in C at 2πn ˘ i logp3 ` 2 2q, n P Z, all of which
lie on a horizontal line, and they are isolated, with a distance of 2π between any
two isolated adjacent zeroes. We had seen this in Exercise 1.58.
The polynomial p, ppzq “ pz ` 1q3 z 9 pz ´ 1q9 , has zeroes at ´1, 0, 1.
If p is a nonzero polynomial such that ppz0 q “ 0, then by the Division Algorithm
there exists a polynomial q (the quotient) such that ppzq “ pz ´ z0 qqpzq (that is,
the remainder is 0). Now we have two possible cases:
1˝ qpz0 q ‰ 0. Then z0 is an isolated zero of p.
2˝ qpz0 q “ 0. Then we repeat the above procedure with p replaced by q.
Eventually, we obtain ppzq “ pz ´ z0 qm qpzq for some m ě 1 and qpz0 q ‰ 0. Then m
is called the order of z0 (as a zero of p). Note that 0 “ p1 pz0 q “ ¨ ¨ ¨ “ ppm´1q pz0 q and
ppmq pz0 q ‰ 0. Theorem 4.11 below says that the same thing holds for holomorphic
functions f (replacing the polynomial p), except that we end up with another
holomorphic function g (instead of the polynomial q). This is not completely
surprising, since power series are analogues of polynomials, and every holomorphic
function has a local power series expansion.
Let D be a domain and f P OpDq. We say z0 P D is a zero of f of order m P N
if 0 “ f pz0 q “ f 1 pz0 q “ ¨ ¨ ¨ “ f pm´1q pz0 q, but f pmq pz0 q ‰ 0.
Theorem 4.11 (Classification of zeroes). Let D be a domain, f P OpDq, and
z0 P D be a zero of f . Then exactly one of the following hold:
1˝ f ” 0.
2˝ There exists an m P N such that z0 is a zero of f of order m, and there exists
a g P OpDq such that gpz0 q ‰ 0 and f pzq “ pz ´ z0 qm gpzq for all z P D.
In 2˝ , as g is continuous and as gpz0 q ‰ 0, there is a small r ą 0 such that g is
never zero in Dpz0 , rq, and so f pzq “ pz ´ z0 qm gpzq ‰ 0 for all z P Dpz0 , rqztz0 u.
Thus z0 is the only zero of f in Dpz0 , rq, that is, z0 is isolated.
Proof. First we will just show the theorem replacing 1˝ by the statement that
there is an R ą 0 such that f pzq “ 0 for all z P Dpz0 , Rq. Later, after showing the
Identity Theorem, we will prove the more general statement.
Let R ą 0 be such that Dpz0 , Rq Ă D. Then f has a Taylor series expansion,
f pzq “ c0 ` c1 pz ´ z0 q ` c2 pz ´ z0 q2 ` ¨ ¨ ¨ for z P Dpz0 , Rq. Since f pz0 q “ 0, c0 “ 0.
Then exactly one of the following hold:
1˝ All the cn are zero. Then f pzq “ 0 for all z P Dpz0 , Rq.
2˝ There is a first integer m ě 1 such that cm ‰ 0. Then c0 “ c1 “ ¨ ¨ ¨ “ cm´1 “ 0.
pnq
As cn “ f n!pz0 q for all integers n ě 0, z0 is a zero of order m. For z P Dpz0 , Rq,
8
ř
f pzq “ cm pz´z0 qm `cm`1 pz´z0 qm`1 ` ¨ ¨ ¨ “ pz´z0 qm cm`k pz´z0 qk . (˚)
k“0
$
& f pzq
pz´z0 qm for z P Dztz0 u,
Define g : D Ñ C by gpzq “ 8
ř
% cm`k pz´z0 qk for z P Dpz0 , Rq.
k“0
From (˚), the two definitions give the same value whenever both are applicable,
and so g is well-defined. Moreover, we have:
‚ g P OpDq. For z P Dztz0 u, g “ p¨´zf0 qm , and since f, p¨´z10 qm P OpDztz0 uq,
we have g P OpDztz0 uq. For z P Dpz0 , Rq, g is given by a power series, and so
g is holomorphic in Dpz0 , Rq.
‚ gpz0 q “ cm ‰ 0, by the definition of m.
‚ f pzq “ pz ´ z0 qm gpzq for z P Dztz0 u follows from the definition of g.
If z “ z0 , then both sides are 0. So for all z P D, f pzq “ pz ´ z0 qm gpzq.
Clearly, if f P OpDq, where D is a domain, and f ” 0 in a neighbourhood of its

zero z0 , then f pnq pz0 q “ 0 for all n ě 0. So z0 does not have a finite order.
Exercise 4.30. Let z0 P C, R ą 0, and f, g P OpDpz0 , Rqq be such that gpz0 q ‰ 0, and
there exists an m P N such that for all z P Dpz0 , Rq, f pzq “ pz ´ z0 qm gpzq. Prove that z0
is a zero of f of order m.
Example 4.8. For each n P Z, nπ is a zero of sin P OpCq. Also, sin is not
identically zero in any neighbourhood of nπ (consider sin |R ). As sin1 z “ cos z and
cos z|z“nπ “ p´1qn ‰ 0, nπ is a zero of sin z of order 1. So for each n P Z, there
exists a gn P OpCq such that gpnπq ‰ 0 and for all z P C, sin z “ pz ´ nπqgn pzq. 1
3 6
exppz 3 q´1 has a zero at 0 since expp03 q´1 “ 1´1 “ 0. As exppz 3 q “ 1` z1! ` z2! `¨ ¨ ¨
3
for all z P C, exppz 3 q ´ 1 “ z 3 gpzq, where g is defined by gpzq :“ 1!1 ` z2! ` ¨ ¨ ¨ . As
g is a power series that converges in C, g is entire. Also, gp0q “ 1 ‰ 0. Thus the
order of 0 as a zero of exppz 3 q ´ 1 is 3 by Exercise 4.30.
Exercise 4.31. Find the order of the zero z0 for the function f in each case:
‚ z0 “ i and f pzq “ p1 ` z 2 q4 .
‚ z0 “ 2nπi, where n is an integer, and f pzq “ exp z ´ 1.
‚ z0 “ 0 and f pzq “ cos z ´ 1 ` 21 psin zq2 .
Exercise 4.32. Let f be holomorphic in a disc that contains a circle C in its interior.
Suppose that in the disc, there is exactly one zero z0 of order 1 of f , which lies in the
1
ş zf 1 pzq
interior of C. Prove that z0 “ 2πi C f pzq
dz.
Exercise 4.33. Let D be a domain and f P OpDq have a zero of order m at z0 P D. Prove
that z0 is a zero of pf p¨qq2 of order 2m, and if m ą 1, then z0 is a zero of f 1 of order m´1.
Exercise 4.34. Let f be an entire function having z0 as a zero of order 1. Define g by
gpzq “ cospf pzqq ´ 1 for all z P C. Show that z0 is a zero of g with order 2.
Exercise 4.35. Let f P OpDp0, 1qq satisfy |f p log1 n q| ď 1
n for all n “ 2, 3, ¨ ¨ ¨ . Show that f
must be identically 0. Hint: First show f p0q “ 0.
8
1In fact, sin z “ z ś p1 ´ z2
q for all z P C. This can be shown using the ‘Weierstrass Factorisation
n“1 n2 π 2
Theorem; see e.g. [3, Chap. VII, §5 and §6].
4.5. The Identity Theorem 91
4.5. The Identity Theorem

In this section, we will learn the Identity Theorem, which once again highlights the
rigidity of holomorphic functions. It says roughly that if a holomorphic function
has an accumulation of zeroes in its domain, then it is identically 0.
Theorem 4.12 (Identity Theorem).
Let D be a domain, f P OpDq, pzn qnPN be a sequence of distinct zeroes of f which
converges to z˚ P D. Then f is identically zero in D.
Proof. First note that z˚ is itself a zero of f because by the continuity of f ,

we have f pz˚ q “ f p lim zn q “ lim f pzn q “ lim 0 “ 0. Moreover, as pzn qnPN is
nÑ8 nÑ8 nÑ8
a sequence of distinct zeroes of f converging to z˚ , it is clear that z˚ is not an
isolated zero of f . So f ” 0 in some disc Dpz˚ , rq, where r ą 0. We want to show
that f ” 0 in D. Let w P D. Then there is a path γ : r0, 1s Ñ D that joins z˚
to w: γp0q “ z˚ and γp1q “ w. Let S :“ tt P r0, 1s : f pγpτ qq “ 0 for 0 ď τ ď tu,
and T :“ sup S (which2 exists: 0 P S ‰ H, and S is bounded above by 1).
Also, T ą 0 since f ˝ γ : r0, 1s Ñ D is continuous, and 0 belongs to the open subset
pf ˝ γq´1 pDpz˚ , Rqq of r0, 1s. Let ptn qnPN be3 a sequence in S such that lim tn “ T .
nÑ8
There are two possible cases:
1˝ T “ 1. By continuity, f pwq “ f pγpT qq “ lim f pγptn qq “ lim 0 “ 0.
nÑ8 nÑ8
2˝ T P p0, 1q. Then f pγpT qq “ lim f pγptn qq “ 0. So γpT q is a zero of f .
nÑ8
γpT q
But γpT q is not an isolated zero4. So there is a ρ ą 0 so that f ” 0 on

DpγpT q, ρq. Then as T belongs to the open set γ ´1 pDpγpT q, ρqq X p0, 1q, there
is an ǫ ą 0 so that T ` ∆t P γ ´1 pDpγpT q, ρqq X p0, 1q for all ∆t ă ǫ. But this
means T ` 2ǫ P S, contradicting T “ sup S. So this case is impossible.
Hence 1˝ holds, and as w was arbitrary, f ” 0 in D.
Example 4.9. Using the definitions of cos, sin : C Ñ C, we had shown that for
all z P C, pcos zq2 ` psin zq2 “ 1. Here is a different proof. Define f P OpCq by
f pzq “ pcos zq2 ` psin zq2 ´ 1 for all z P C. By Pythagoras’ Theorem, for all x P R,
f pxq “ pcos xq2 ` psin xq2 ´ 1 “ 0. The Identity Theorem now implies that f ” 0
in C.
2If we think of t as time while we travel from z (time t “ 0) to w (time t “ 1) along γ, then T is the
˚
largest time such that f has been 0 along the path covered so far.
3This can be constructed by considering T ´ T ă T , n P N, which while failing to be upper bounds of S,
n
guarantee the existence of tn P S so that T ´ T n
ă tn ď T , and by using the Sandwich Theorem.
4Otherwise, there is a δ1 ą 0 so that f is never zero in DpγpT q, δ1 qztγpT qu. This is a contradiction, since
for all n large enough, |γptn q ´ γpT q| ă δ1 , i.e., γptn q P DpγpT q, δ1 q and f pγptn qq “ 0.
The following is equivalent to the Identity Theorem.

Theorem 4.13 (Identity Theorem). Let D be a domain, f, g P OpDq, pzn qnPN be
a sequence of distinct points in D which converges to z˚ P D, and such that for all
n P N, f pzn q “ gpzn q. Then f pzq “ gpzq for all z P D.
Proof. Define h : D Ñ C by hpzq “ f pzq ´ gpzq, z P D, and note that the zn ,

n P N, are zeroes of the holomorphic function h. By Theorem 4.12, h must be
identically zero in D, and so the claim follows.
Vice versa, Theorem 4.12 follows from the above by taking g ” 0.

Example 4.10. Recall that we defined exp : C Ñ C by exp z :“ ex pcos y ` i sin yq
for z “ x ` iy P C, where x, y P R. Then exp is an entire function such that
exp x “ ex for x P R. In other words, exp is an entire extension of the usual real
exponential function. Is there any other entire extension possible? We show that
the answer is no! Suppose g : C Ñ C is entire and gpxq “ ex for all real x. But
then exp x “ gpxq for all x P R. In particular, exp n1 “ gp n1 q, for all n P N, and
1
n Ñ 0 P C as n Ñ 8. By the Identity Theorem, exp z “ gpzq for all z P C. So
there is only one entire function whose restriction to R is ex . This explains the
naturalness of the definition of the complex exponential in Section 1.4.1.
We can now also complete the proof of the result on the classification of zeroes
(Theorem 4.11). We had shown the theorem replacing 1˝ by the statement that
there is an R ą 0 such that f pzq “ 0 for all z P Dpz0 , Rq. But from the Identity
R
Theorem, f ” 0 in D e.g. by taking zn “ z0 ` 2n for all n P N.
Exercise 4.36. Use the Identity Theorem to show cospz`wq “ pcos zqpcos wq´psin zqpsin wq
for all z, w P C, by appealing to the corresponding identity when z, w are real numbers.
Exercise 4.37. Let D be a domain. Then it is easy to check that OpDq is a commutative
ring5 with the pointwise operations: pf ` gqpzq “ f pzq ` gpzq and pf ¨ gqpzq “ f pzqgpzq for
all z P D and all f, g P OpDq. Check that OpDq is an integral domain, that is, a nonzero
ring having no zero divisors (if f ¨ g “ 0 for f, g P OpDq, then either f “ 0 or g “ 0).
If instead of OpDq, we consider the set CpDq of all complex-valued continuous functions on
D, then CpDq is again a commutative ring with pointwise operations. Is CpDq an integral
domain? This shows that continuous functions are not as ‘rigid’ as holomorphic functions.
Exercise 4.38. Does there exist a nonzero entire function with an uncountable set of
Ť
zeroes? Hint: C “ Dp0, nq. A countable union of finite sets is finite or countable.
nPN
Exercise 4.39. Let D :“ tz P C : |z| ă 1u, and f P OpDq be such that f has at least one
zero in Dzt0u, and f pzqf p´zq “ f pz 2 q for all z P D. Prove that if z0 P Dzt0u is a zero of
f , then z02 is a new zero of f in Dzt0u. Determine all possible such f . Hint: Construct a
sequence of zeros and invoke the identity theorem.
5A commutative ring R, is a set R with two laws of composition ` and ¨ such that pR, `q is an Abelian
group, ¨ is associative, commutative and has an identity, and the distributive law holds: for a, b, c P R,
pa ` bq ¨ c “ a ¨ c ` b ¨ c.
4.6. The Maximum Modulus Theorem 93
Exercise 4.40. Let D be a domain and f, g P OpDq. Which of the following conditions
imply f “ g identically in D?
(1) There is a sequence pzn qnPN of distinct terms in D such that f pzn q “ gpzn q for all n P N.
(2) There is a convergent sequence pzn qnPN of distinct points in D with its limit in D such
that f pzn q “ gpzn q for all n P N.
(3) γ is a smooth path in D joining distinct points z, w P D and f “ g on γ.
(4) w P D is such that f pnq pwq “ g pnq pwq for all n ě 0.
Exercise 4.41. Let f : p0, 8q Ñ R be given by f pxq “ sinplog xq for all x ą 0. Find all
x ą 0 such that f pxq “ 0. Does there exist an entire function F which extends f ?
ř
8
Exercise 4.42. Suppose r ą 0, and that g P OpDp0, rqq is such that g pnq p0q converges.
n“0
Prove that g has a unique entire extension.
Exercise 4.43. Let the entire f be such that in every power series (i.e., for every z0 P C)
ř
8
f pzq “ cn pz ´ z0 qn , at least one coefficient is 0. Prove that f is a polynomial.
n“0
Exercise 4.44. Let D :“ Dp0, 1q, and f, g P OpDq be nonzero everywhere in D, and such
f 1p 1 q g1 p 1 q
that f p 1nq “ gp 1nq for all integers n ě 2. Prove that f “ λg for some λ P Czt0u.
n n
Exercise 4.45. Find all entire f satisfying f p n1 q “ n12 for all n P N. Hint: gpzq :“ f pzq´z 2 .
Show that there is no entire function f satisfying f p n12 q “ n1 for all n P N.
Exercise 4.46 (Argument Principle). Let D be a domain, f P OpDq, z0 P D, R ą 0 be
such that tz P C : |z ´ z0 | ď Ru Ă D, Cptq :“ z0 ` Re2πit , t P r0, 1s, and f pCptqq ‰ 0 for
all t P r0, 1s. Let Zpf q :“ tz P Dpz0 , Rq : f pzq “ 0u and mpzq be the order of the zero
1
ş f 1 pzq ř
z P Zpf q of f . Then 2πi C f pzq
dz “ mpzq.
zPZpf q
4.6. The Maximum Modulus Theorem

We now prove the Maximum Modulus Theorem, saying that for a nonconstant
f P OpDq, |f | cannot have a maximiser in the domain D.
Theorem 4.14 (Maximum Modulus Theorem). Let D be a domain, f P OpDq,
z0 P D be such that for all z P D, |f pz0 q| ě |f pzq|. Then f is constant on D.
Proof. Let r ą 0 be such that Dpz0 , 2rq Ă D. Define Cr by Cr ptq “ z0 ` re2πit for
all t P r0, 1s. By the Cauchy Integral Formula,
1
ş f pzq ş
1 1 f pz0 `re2πit q
ş1
f pz0 q “ 2πi Cr z´z0 dz “ 2πi 0 re2πit
2πire2πit dt “ 0 f pz0 ` re2πit qdt.
(The last expression can be viewed as the ‘average’ of the values of f on Cr .) Thus
ş1 ş1 ş1
0 |f pz0 q| dt “ |f pz0 q| “ | 0 f pz0 ` re2πit q dt| ď 0 |f pz0 ` re2πit q| dt. Rearranging,
ş1 2πit q| ´ |f pz q|q 2πit q| ď |f pz q| for all t P r0, 1s,
0 p |f pz0 ` re 0 q dt ě 0. Since |f pz0 ` re 0
ş1
the integrand is pointwise ď 0. So ´ 0 p |f pz0 q| ´ |f pz0 ` re2πit q| q dt “ 0. As
the integrand is continuous and pointwise nonnegative, the above implies that
|f pz0 ` re2πit q| “ |f pz0 q| for all t P r0, 1s. By replacing r by any smaller number, the
same conclusion holds. Thus f maps the disc ∆ :“ tz P C : |z ´ z0 | ď ru into the
circle tw P C : |w| “ |f pz0 q|u. By Example 2.10, f is constant on ∆. The Identity
Theorem shows that f is constant on D.
e ´z
Example 4.11. Let H “ tz P C : Re z ě 0u. Define f : H Ñ C by f pzq “ z`1 ,
z P H. It can be shown }f }8 :“ maxt|f pzq| : z P Hu exists. Without worrying about
the existence of this maximum, let us see how, assuming its existence, the Maximum
Modulus Theorem enables us to calculate its value. Suppose z0 P H is a maximiser
of |f | on H. If Re z0 ą 0, then z0 is a maximiser of |f | on the open right-half plane
1
H˝ :“ tz P C : Re z ą 0u, and since f is not constant (e.g. f p1q “ 2e ‰ 3e12 “ f p2q),
we get a contradiction to the Maximum Modulus Theorem. So z0 P iR, that is,
eíy
z0 “ iy0 for some y0 P R. But |f piyq| “ | iy`1 | “ ? 12 ď ?012 `1 for all y P R. Thus
y `1
}f }8 “ max |f pzq| “ max |f piyq| “ max ? 12 “ ? 1
2
0 `1
“ 1.
zPH yPR yPR y `1
Exercise 4.47. Let D be a domain and let f P OpDq be a nonconstant map. Prove that
there is no maximiser of |f | in D.
Exercise 4.48 (Minimum Modulus Theorem). Let D be a domain and let f P OpDq be
such that there is a z0 P D such that for all z P D, |f pzq| ě |f pz0 q|. Then prove that
f pz0 q “ 0 or f is constant on D.
Exercise 4.49. Consider the function f defined by f pzq “ z 2 ´ 2. Find the maximum
and minimum value of |f pzq| on tz P C : |z| ď 1u.
Exercise 4.50. Prove or disprove: There exists a f P OpDp0, 1qq such that |f pzq| “ e|z|
for all z P Dp0, 1q.
4.7. Laurent series

Laurent series generalise Taylor series. Indeed, while a Taylor series
8
ř
cn pz ´ z0 qn “ c0 ` c1 pz ´ z0 q1 ` c2 pz ´ z0 q2 ` ¨ ¨ ¨
n“0
has nonnegative powers of the term z ´ z0 , and converges in a disc, a Laurent series
is an expression of the following type, which has negative powers of z ´ z0 too:
ř
cn pz´z0 qn “ ¨ ¨ ¨`c´2 pz´z0 q´2 `c´1 pz´z0 q´1 `c0 `c1 pz´z0 q1 `c2 pz´z0 q2 `¨ ¨ ¨ ,
nPZ
We will see that a Laurent series ‘converges’ in an annulus tz P C : r ă |z´z0 | ă Ru

with centre z0 and gives a holomorphic function there, and conversely, if we have a
holomorphic function in an annulus with centre z0 , then the function has a Laurent
2 3
series expansion in the annulus. E.g., for all z P C, exp z “ 1` 1!z ` z2! ` z3! `¨ ¨ ¨ , and
so for z ‰ 0, we have the ‘Laurent series expansion’ exp 1z “ 1` z1 ` 2!1 z12 ` 3!1 3!1 `¨ ¨ ¨ .
Note that exp 1z P OpCzt0uq, which is a degenerate annulus centred at 0 with inner
radius r “ 0 and outer radius R “ `8. Let us first define what we mean by the
ř
convergence of cn pz ´ z0 qn .
nPZ
4.7. Laurent series 95
Definition 4.1. Let Z Q n ÞÑ cn P C be a function. We say that the Laurent series

ř 8
ř 8
ř
cn pz ´ z0 qn converges (for z) if both cń pz ´ z0 qń , cn pz ´ z0 qn converge.
nPZ n“1 n“0
ř ř 8
ř 8
ř
If cn pz´z0 qn converges, we write cn pz´z0 qn “ cń pz´z0 qń ` cn pz´z0 qn ,
nPZ nPZ n“1 n“0
and call this the sum of the Laurent series.
Example 4.12. Let us find out the set of complex numbers z for which the Laurent
series ¨ ¨ ¨ ` 8z13 ` 4z12 ` 2z
1
` 1 ` z ` z 2 ` z 3 ` ¨ ¨ ¨ converges. We have:
‚ 1 ` z ` z 2 ` z 3 ` ¨ ¨ ¨ converges for |z| ă 1, and it diverges for |z| ě 1.
1
‚ 2z ` 4z13 ` 8z15 ` ¨ ¨ ¨ converges for | 2z
1 1
| ă 1 and diverges for | 2z | ě 1, that is, it
1 1
converges for |z| ą 2 and diverges for |z| ď 2 .
So the Laurent series converges when |z| ă 1 and |z| ą 21 , i.e., inside the annulus
A :“ tz P C : 12 ă |z| ă 1u, and diverges when |z| ě 1 or |z| ď 12 , i.e., outside A.
ř
For which z does cn pz ´ z0 qn converge?
nPZ
‚ From Theorem 4.1, there is some nonnegative R (possibly R “ `8) such that
8
ř
cn pz ´ z0 qn converges for |z ´ z0 | ă R and diverges for |z ´ z0 | ą R.
n“0
8
ř
‚ What about the series cń pz ´ z0 qń ? Again by Theorem 4.1, there is some
n“1
8
r ě 0 such that the power series ř cń wn converges for |w| ă R
R r and diverges
n“1
8
r. For z ‰ z0 , set w :“ pz ´ z0 q´1 . Then ř cń pz ´ z0 qń converges for
for |w| ą R
n“1
1 r, i.e. for |z ´ z0 | ą
ăR 1
“: r, and diverges for |z ´ z0 | ă r.
|z´z0 | r
R
Hence the Laurent series converges in the annulus A :“ tz P C : r ă |z ´ z0 | ă Ru

and diverges if either |z ´ z0 | ă r or |z ´ z0 | ą R.
ř
Is cn pz ´ z0 qn holomorphic in the annulus A where it converges?
nPZ
8
ř 8
ř
‚ cn p¨ ´ z0 qn P OpDpz0 , Rqq and so also cn p¨ ´ z0 qn P OpAq.
n“0 n“0
g 8
ř
‚ w ÞÝÑ r Also f “ 1 P OpCztz0 uq.
cń wn is holomorphic in tw P C : |w| ă Ru.
n“1
¨ ´z0
8
ř
So g ˝f is holomorphic in tz P C : |z ´z0 | ą ru, i.e., cń p¨´z0 qń is holomorphic
n“1
in tz P C : r ă |z ´ z0 |u, and so also in the annulus A.

8
ř 8
ř ř
Hence the sum z ÞÑ cn pz ´z0 qn ` cń pz ´z0 qń “ cn pz ´z0 qn is holomorphic
n“0 n“1 nPZ
in the annulus tz P C : r ă |z ´ z0 | ă Ru.
ř
Summarising, a Laurent series cn pz ´ z0 qn converges in a possibly empty annulus
nPZ
ř
A “ tz P C : r ă |z ´ z0 | ă Ru for some r, R, and cn p¨ ´ z0 qn P OpAq.
nPZ
Conversely, a function holomorphic in an annulus has a Laurent series expansion.
Theorem 4.15. Let z0 P C, 0 ď r ă R ď 8, A “ tz P C : r ă |z ´z0 | ă Ru, f P OpAq.

For any ρ P pr, Rq, define the circular path Cptq :“ z0 ` ρ e2πit for all t P r0, 1s.
1
ş f pζq ř
Let cn :“ 2πi C pζ´z0 qn`1 dζ pn P Zq. Then f pzq “ cn pz ´ z0 qn for all z P A.
nPZ
Moreover, the coefficients are unique in such a Laurent series expansion of f in A.
Example 4.13. Let f pzq :“ z 3 exp z1 for all z ‰ 0. Then f is holomorphic in the
annulus A :“ Czt0u “ tz P C : 0 ă |z| ă `8u. By Theorem 4.15, f has a Laurent
ř
series expansion cn z n in Czt0u. For z ‰ 0, since we have exp z1 “ 1` z1 ` 2!z1 2 `¨ ¨ ¨ ,
nPZ
we obtain f pzq “ z 3 exp z1 “ z 3 ` z 2 ` z
2! ` 1
3! ` 1
4!z ` ¨¨¨.
Proof of Theorem 4.15. (Existence.) Fix z P A. Suppose that rr and R r are such
r
that r ă rră |z ´ z0 | ă R ă R. Define circular paths Crr and CRr by Crrptq “ z0 ` rreit
r it for all t P rθ, 2π `θs, and θ :“ pArg zq` π . Let σ : rr
and CRr ptq “ z0 ` Re r ÑA
r , Rs
2
z´z0
be the path σptq “ ti |z´z0 | . (This is a straight line path joining Crr and CRr , and
the multiplication by i produces a rotation by 90˝ , ensuring that σ avoids z.)
CRr
´σ Crr
σ
z
rr
r z0
r
R R
Then CRr ´ σ ´ Crr ` σ is equivalent to a reparameterised path γ : r0, 1s Ñ Aztzu,

p¨q
which is Aztzu-homotopic to a small circle Cδ centred at z. Also, f¨´z P OpAztzuq,
ş f pζq ş f pζq
and so γ ζ´z dζ “ Cδ ζ´z dζ “ f pzq2πi, where the first equality follows from the
Cauchy Integral Theorem, and the second equality follows from the Cauchy Integral
Formula. Since the integral along σ cancels with that along ´σ, we obtain that
1
ş f pζq 1
ş ş ş ş f pζq
f pzq “ 2πi γ ζ´z dζ “ 2πi CRr ´✓ ✓
σ ´ Crr `✓✓
σ ζ´z dζ
1
ş f pζq 1
ş f pζq
“ 2πi C r ζ´z dζ ´ 2πi Crr ζ´z dζ “ pIq ` pIIq,
R
1
ş f pζq 1
ş f pζq
where pIq :“ 2πi C r ζ´z dζ and pIIq “ ´ 2πi Crr ζ´z dζ. We will now show that (I)
R
8
ř 8
ř
equals cn pz ´ z0 qn , while (II) equals cń pz ´ z0 qń , and these put together
n“0 n“1
yield the desired Laurent series expansion of f .
1
ş f pζq 8
ř
Step 1. In this step we will show that 2πi C r ζ´z dζ “ cn pz ´ z0 qn . For ζ P CRr ,
R n“0
f pζq f pζq f pζq f pζq z´z0
we have ζ´z “ ζ´z0 `z0 ´z “ z´z “ pζ´z0 qp1´wq , where w :“ ζ´z0 . So
pζ´z0 qp1´ ζ´z0 q
0
|z´z0 | |z´z0 | 1 wn
|w| “ |ζ´z0 | “ r ă 1, and so 1´w “ 1 ` w ` w2 ` w3 ` ¨ ¨ ¨ ` wn´1 ` 1´w . Thus
R
4.7. Laurent series 97
f pζq f pζq f pζq f pζq n
“ n´1 ` w n q n´1 ` f pζqpz´z0 q .
ζ´z ζ´z0 p1`w `¨ ¨ ¨`w 1´w “ ζ´z0 `¨ ¨ ¨` pζ´z 0q
n pz´z0 q pζ´z0 qn pζ´zq
Hence
1
ş f pζq
2πi CRr ζ´z dζ
1
ş f pζq 1
ş f pζq n´1 1
ş f pζq n
“ 2πi CRr ζ´z0 dζ ` ¨¨¨ ` 2πi CRr pζ´z0 qn dζpz´z0 q ` 2πi CRr pζ´z0 qn pζ´zq dζpz´z0 q
p˚q
“ c0 ` c1 pz ´ z0 q ` ¨ ¨ ¨ ` cn´1 pz ´ z0 qn´1 ` Rn pzq,
1
ş f pζqpz´z0 qn
where Rn pzq :“ 2πi C pζ´z0 qn pζ´zq dζ. To see (˚), note that a reparametrisation of
r
R
CRr is A-homotopic to a circular path Cptq “ z0 `ρe2πit , t P r0, 1s, for any ρ P pr, Rq,
ş f pζq ş f pζq
and so by the Cauchy Integral Theorem, C r pζ´z k dζ “ C pζ´z0 qk dζ “ ck for
R 0q
k “ 1, ¨ ¨ ¨ , n ´ 1. To show finish the proof of Step 1, we will show lim Rn “ 0.
nÑ8
Let M ą 0 be such that for all ζ P CRr , |f pζq| ă M . We have |ζ ´ z0 | “ R r and also
r |z´z0 | n M nÑ8
|ζ ´ z| “ |ζ ´ z0 ´ pz ´ z0 q| ě R ´ |z ´ z0 |. So |Rn pzq| ď p r q r ÝÑ 0.
R R´|z´z0 |
1
ş f pζq
Thus c0 ` c1 pz ´ z0 q ` c2 pz ´ z0 q2 ` ¨ ¨ ¨ “ 2πi γ2 ζ´z dζ.
1
ş f pζq 8
ř
Step 2. We will show that ´ 2πi Crr ζ´z dζ “ cń pz ´ z0 qń . We have
n“1
1
ş f pζq 1
ş f pζq 1
ş f pζq
´ 2πi Crr ζ´z dζ “ 2πi Crr pz´z0 q´pζ´z0 q dζ “ 2πi Crr ζ´z0 dζ.
pz´z0 qp1´ z´z q
0
ζ´z0 rr 1 wn
If w :“ z´z0 , then |w| “ |z´z0 | ă 1, and ζ´z “ 1 ` w ` ¨¨¨ ` wn´1 ` 1´w . So
1´ z´z0
0
1
ş f pζq
´ 2πi Crr ζ´z dζ
1
ş 1 pζ´z0 qn´1 pζ´z0 qn
“ 2πi Crr f pζqp z´z0 ` ¨¨¨ ` pz´z0 qn ` pz´z0 qn pz´ζq qdζ
1
ş 1 1
ş f pζq 1 1
ş f pζqpζ´z0 qn
“ 2πi Crr f pζqdζ ¨ z´z0 ` ¨¨¨ ` 2πi Crr pζ´z0 qń`1 dζ ¨ pz´z0 qn ` 2πi Crr pz´z0 qn pz´ζq dζ
p‹q
“ c´1 pz ´ z0 q´1 ` ¨ ¨ ¨ ` cń pz ´ z0 qń ` R rn pzq,
ş f pζqpζ´z0 qn
where R rn pzq :“ 1
2πi Crr pz´z0 qn pz´ζq dζ. To see (‹), note that a reparameterisation
of Crr is A-homotopic to C (Cptq “ z0 ` ρe2πit , t P r0, 1s, where ρ P pr, Rq). By the
ş f pζq ş f pζq
Cauchy Integral Theorem, Crr pζ´z k dζ “ C pζ´z qk dζ “ ck for k “ ´1, ¨ ¨ ¨ , ń.
0q 0
Let M ą 0 be such that for all ζ P Crr, |f pζq| ă M . We have |ζ ´ z0 | “ rr and
|z ´ ζ| “ |pz ´ z0 q´ pζ ´ z0 q| ě |z ´ z0 |´ rr, and so |R rn pzq| ď p rr qn M rr nÑ8
|z´z | r ÝÑ 0.
|z´z0 |´r
´1 ń 1
ş f pζq 0
Thus c´1 pz ´ z0 q ` ¨ ¨ ¨ ` cń pz ´ z0 q ` ¨ ¨ ¨ “ ´ 2πi Crr ζ´z dζ.
This completes the proof of the existence part of the Laurent expansion.
Uniqueness of coefficients.
Cauchy’s Integral Formula allows us to show that the Laurent expansion is unique,
ř
i.e., if f pzq “ rcn pz ´ z0 qn for r ă |z ´ z0 | ă R, then for all n P Z, r
cn “ cn .
nPZ
d pz´z0 qn`1
ş
If n ‰ ´1, then pz ´ z0 qn “ dz p n`1 q. So C pz ´ z0 qn dz “ 0 (n ‰ ´1), where
ş 1
C is given by Cptq “ z0 ` ρe2πit , t P r0, 1s. But C z´z 0
dz “ 2πi. By term-by-term
integration,
ş f pzq ş ´m´1 ř r ř ş
C pz´z0 qm`1 dz “ C pz´z0 q cn pz´z0 qn dz “ r cn C pz´z0 qn´m´1 dz “ 2πir
cm ,
nPZ nPZ
proving the claim about the uniqueness of coefficients. The term-by-term integra-
tion is justified as follows. We have
ř crm´1 crm
cn pz ´ z0 qn´m´1 “ p ¨ ¨ ¨ `
r pz´z0 q2
q ` z´z 0
`pr
cm`1 ` r
cm`2 pz ´ z0 q ` ¨ ¨ ¨ q
nPZ
crm
“ f1 pzq ` z´z 0
` f 2 pzq.
We need only show that f1 , f2 have a primitive in the annulus, because then
ş ř ş crm
C cn pz ´ z0 qn´m´1 dz “ C pf1 pzq ` z´z
r 0
` f2 pzqqdz “ 0 ` 2πir
cm ` 0 “ 2πir
cm .
nPZ
8
ř 8 c
ř rm`n
We have f2 pzq “ cm`n pz ´ z0 qn´1 for z P Dpz0 , Rq. If F2 pzq :“
r n pz ´ z0 qn
n“1 n“1
for z P Dpz0 , Rq, then ddz F2 pzq “ f2 pzq, and so F2 is a primitive of f2 . Note
that since for all n P N, | crm`n n cm`n pz ´ z0 qn´1 |, it follows by the
n pz ´ z0 q | ď R|r
8 r
ř cm`n
comparison test that n pz ´ z0 qn converges absolutely in Dpz0 , Rq.
n“1
8
ř crmń 8
ř 1
For R ą |z ´ z0 | ą r, f1 pzq “ pz´z0 qn`1
“ cmń wn`1 , where w “
r z´z0 . So
n“1 n“1
8
ř 8 r
ř cmń
cmń wn`1 converges for |w| ă 1r . Let Gpwq :“ ´
r n wn for |w| ă 1r . As
n“1 n“1
8
ř 1
cmń wn`1 is absolutely convergent for
r R ă |w| ă 1r , and since for all n P N,
n“1
8 r
ř
| crmń n cmń wn`1 |, by the comparison test,
n w | ď R|r
cmń
n wn is absolutely con-
n“1
vergent, and being a power series, this holds also for all w P Dp0, 1r q. We have
8
ř 8 c
ř rmń
d 1
dz Gpwq “´ cmń wn´1 . If F1 pzq :“ Gp z´z
r 0
q“´ n pz ´ z0 qń for z P A,
n“1 n“1
8
ř
1
then d
dz F1 pzq “ pG1 p z´z 0
qqp´ pz´z1 0 q2 q “ pz ´ z0 q´2 cmń pz ´ z0 qń`1 “ f1 pzq.
r
n“1
The uniqueness of coefficients is valid only if we consider a particular fixed annulus.

It can happen that the same function has different Laurent expansions, but valid
in different annuli, as shown in the following example.
1
Example 4.14. Consider f defined by f pzq “ zpz´1q , z P Czt0, 1u.
r
A
A
0 1 0 0
4.8. Classification of singularities 99
f is holomorphic in the annulus A :“ tz P C : 0 ă |z| ă 1u. Since |z| ă 1,

1
` ˘
f pzq “ zpz´1q “ ´ z1 1 ` z ` z 2 ` z 3 ` ¨ ¨ ¨ “ ´ 1z ´ 1 ´ z ´ z 2 ´ z 3 ´ ¨ ¨ ¨ .
The Laurent series coefficients are c´2 “ c´3 “ ¨ ¨ ¨ “ 0, c´1 “ c0 “ c1 “ ¨ ¨ ¨ “ ´1.
But f is also holomorphic in the annulus A r :“ tz P C : 1 ă |z|u. So f has a
r
Laurent series expansion in A too. Since |z| ą 1, we have
´ ¯
1 1 1 1 1 1
f pzq “ zpz´1q “ z 2 p1´ 1 “ z2 1 ` z ` z2 ` z3 ` ¨ ¨ ¨
q
“ z12 ` z13 ` ¨ ¨ ¨ .
z
The Laurent series coefficients are rc´2 “ r c´3 “ ¨ ¨ ¨ “ 1, c´1 “ c0 “ c1 “ ¨ ¨ ¨ “ 0.
So the coefficients are different, but this is not surprising, since the annuli for the
Laurent series expansions were different too.
Exercise 4.51. In Example 4.14 find Laurent series expansions for f also in the annuli
r 1 :“ tz P C : 1 ă |z ´ 1|u.
A1 :“ tz P C : 0 ă |z ´ 1| ă 1u and A
Exercise 4.52. Let w P C. Define fw by fw pzq “ expp w2 pz ´ z1 qq for all z P Czt0u. Then
ř
8
fw P OpCzt0uq. So fw has a Laurent series expansion fw pzq “ Jn pwqz n , z P Czt0u,
n“´8
defining the Bessel functions Jn for all n P Z. Show the following:
ş
1 2π
f´w p´zq “ fw pzq, Jn p´wq “ p´1qn Jn pwq, Jn pwq “ 2π 0 exppiw sin θ ´ inθqdθ.
ş 5 ´1
Exercise 4.53. Using the Laurent series theorem, evaluate C z e z2 dz, where C is the
circular path given by Cptq “ eit for all t P r0, 2πs.
4.8. Classification of singularities

If we look at the three functions sinz z , z13 , exp 1z , then we notice that each of them
is not defined at 0. We refer to 0 as a ‘singularity’ of these functions, because they
are not defined at that point. However, the functions are holomorphic in a region
obtained by deleting that point from its neighbourhood. We will see that these
functions behave very differently near their common singularity. In other words,
the ‘nature of the singularity’ differs in each case. We will explain precisely how the
behaviour is different in each case, and this is what we mean by the classification of
singularities. Moreover, we will learn two results, which will allow us to find out the
type of singularity at hand. Of these two characterisation results for singularities,
one result will be in terms of limits, while the other will be in terms of what happens
with Laurent coefficients. We first give the following definition.
Definition 4.2. Let f be a complex-valued function, not defined at z0 P C, such
that f is holomorphic in a punctured disc tz P C : 0 ă |z ´ z0 | ă Ru centred at z0
with a radius R ą 0. Then we call z0 an isolated singularity of f .
sin z 1
Example 4.15. z , z3 , exp 1z , each have an isolated singularity at 0.
1 1
sin z1
has a singularity at 0, but 0 is not an isolated singularity (since e.g. at z “ nπ ,
1
n P N, the function sin z1
is not defined).
Definition 4.3. Let z0 P C be an isolated singularity of f P OpDpz0 , Rqztz0 uq,

R ą 0. Then z0 is called
‚ a removable singularity of f if DF P OpDpz0 , Rqq such that F |Dpz0 ,Rqztz0 u “ f .
‚ a pole of f if lim |f pzq| “ 8
zÑz0
(i.e., @M ą 0 Dδ ą 0 such that whenever 0 ă |z´z0 | ă δ, we have |f pzq| ą M ).
‚ an essential singularity of f if z0 is neither removable nor a pole.
Example 4.16.
sin z
(1) f given by f pzq “ z has a removable singularity at 0: For z ‰ 0,
8 p´1qn
ř
sin z 1 z3 z5 2n
z “ z pz ´ 3! ` 5! ´ ` ¨ ¨ ¨ q “ p2n`1q! z .n“0
The right-hand side, being a power series with an infinite radius of convergence
(why?), defines an entire function F . Since this entire function F coincides with
f in the punctured plane Czt0u, f has a removable singularity at 0. From here,
we obtain lim sinz z “ lim F pzq “ F p0q “ 1.
zÑ0 zÑ0
1 1
(2) The function z3
has a pole at 0, since lim 3 “ `8. (If M ą 0, then taking
zÑ0 |z|
1 1
δ :“ M 1{3
ą 0, whenever 0 ă |z ´ 0| ă δ, we have |f pzq| “ |z|3
ą M .)
1
(3) The function exp has an essential singularity at 0. Indeed,
z
1
‚ 0 is not a removable singularity, because for example lim e x “ `8.
xÑ0`
1
‚ 0 is also not a pole, as lim e x “ 0, and so � p lim |f pzq| “ `8q.
xÑ0´ zÑ0
1
Exercise 4.54. Let f pzq :“ cospe z q for z ‰ 0. Show that 0 is an essential singularity of f .
1
Hint: Consider zn “ logpnπq , n P N.
We will now learn the characterisation result for singularities, in terms of limiting
behaviour. Below � is the symbol for negation, read as ‘it is not the case that’.
Theorem 4.16 (Classification of singularities via limits).
Suppose z0 is an isolated singularity of f . Then
z0 is removable ô lim pz ´ z0 qf pzq “ 0.
zÑz0
z0 is a pole ô ‚ �pp lim pz ´ z0 qf pzq “ 0qq and
zÑz0
‚ Dn P N such that lim pz ´z0 qn`1f pzq “ 0.
zÑz0
pThe smallest such n is called
the order of the pole z0 of f.q
z0 is essential ô @n P N �pp lim pz ´ z0 qn f pzq “ 0qq.
zÑz0
Proof. (1) z0 removable ñ lim pz ´ z0 qf pzq “ 0 :

zÑz0
Let z0 be removable. Let F P OpDpz0 , Rqq, R ą 0, be such that F |Dpz0 ,Rqztz0 u “ f .
Then p¨´z0 qF is continuous at z0 , and its value there is 0. Given ǫ ą 0, there exists
a δ ą 0 such that if |z ´ z0 | ă δ, then |pz ´ z0 qF pzq ´ 0| ă ǫ, and so in particular,
if 0 ă |z ´ z0 | ă δ, then |pz ´ z0 qf pzq ´ 0| ă ǫ. Thus lim pz ´ z0 qf pzq “ 0.
zÑz0
lim pz ´ z0 qf pzq “ 0 ñ z0 is removable :

zÑz0
Let lim pz ´ z0 qf pzq “ 0, and f P OpDpz0 , Rqztz0 uq, where R ą 0. Then f has a
zÑz0
ř
Laurent series expansion f pzq “ cn pz ´ z0 qn , 0 ă |z ´ z0 | ă R, where for n P Z,
nPZ
1
ş f pzq
cn “ 2πi Cr pz´z0 qn`1 dz, and C r ptq “ z0 ` re2πit , t P r0, 1s, for any r P p0, Rq.
We will show cń “ 0 for n P N. Given ǫ ą 0, let r ą 0 be such that |pz´z0 qf pzq| ă ǫ
on Cr . Then, for n P N,
1
ş f pzq 1
ş pz´z0 qf pzq 1 ǫ n´1 ď ǫRn´1 .
|cń | “ | 2πi Cr pz´z0 qń`1 dz| “ | 2πi Cr pz´z0 qń`2 dz| ď 2π r ń`2 2πr “ ǫ r
8
ř
As ǫ ą 0 was arbitrary, cń “ 0 for all n P N. With F pzq :“ cn pz ´ z0 qn ,
n“0
F P OpDpz0 , Rqq, and F “ f in the punctured disc tz P C : 0 ă |z ´ z0 | ă Ru.
! �pp lim pz ´ z0 qf pzq “ 0qq and
zÑz0
(2) z0 is a pole ñ
Dn P N such that lim pz ´ z0 qn`1 f pzq “ 0.
zÑz0
Let z0 be a pole. Then z0 is not removable (otherwise if F P OpDpz0 , Rqq, R ą 0,
is such that F |Dpz0 ,Rqztz0 u “ f , then lim |f pzq| “ lim |F pzq| “ |F pz0 q| ă `8, a
zÑz0 zÑz0
contradiction). By the part (1) above, �pp lim pz ´ z0 qf pzq “ 0qq.
zÑz0
There is some r ą 0 such that |f pzq| ą 1 for all z in the punctured disc Dpz0 , rqztz0 u.
In particular, f ‰ 0 in Dpz0 , rqztz0 u, and so g :“ f1 P OpDpz0 , rqztz0 uq. Since z0
is a pole of f , it follows that lim pz ´ z0 qgpzq “ 0 ¨ 0 “ 0. By part (1) above,
zÑz0
z0 is a removable singularity of g. So there exists a G P OpDpz0 , rqq such that
G|Dpz0 ,rqztz0 u “ g. We have Gpz0 q “ lim Gpzq “ lim gpzq “ 0. Hence z0 is a
zÑz0 zÑz0
zero of G. Since G is not identically zero in a neighbourhood of z0 , by the re-
sult on the classification of zeroes, z0 has some order m P N, and there exists an
H P OpDpz0 , rqq such that Hpz0 q ‰ 0 and Gpzq “ pz ´ z0 qm Hpzq in Dpz0 , rq. So
1 1
for z P Dpz0 , rqztz0 u, we have f pzq “ gpzq “ Gpzq “ pz´z0 q1m Hpzq . Consequently,
z´z0 0
lim pz ´ z0 qm`1 f pzq “ lim “ Hpz0 q “ 0.
zÑz0 zÑz0 Hpzq
�pp lim pz ´ z0 qf pzq “ 0qq and )
zÑz0
ñ z0 is a pole.
Dn P N such that lim pz ´ z0 qn`1 f pzq “ 0
zÑz0
Let m be the smallest such n: lim pz ´z0 qm`1f pzq “ 0, � p lim pz ´z0 qmf pzq “ 0qq.
zÑz0 zÑz0
By part (1), p¨ ´z0 qm f has a removable singularity at z “ z0 . Thus there exists
an F P OpDpz0 , Rqq, R ą 0, such that F |Dpz0 ,Rqztz0 u “ pz ´ z0 qm f . Thus we have
F pzq
F pz0 q “ lim F pzq “ lim pz ´z0 qm f pzq ‰ 0 (definition of m). From f pzq “ pz´z0 qm
zÑz0 zÑz0
|F pzq| 1
for z P Dpz0 , Rqztz0 u, lim |f pzq| “ lim m “ |F pz0 q| lim m “ `8. So
zÑz0 zÑz0 |z´z0 | zÑz0 |z´z0 |
z0 is a pole of f .
(3) This follows from the first two parts. If z0 is an essential singularity, then it is
not removable, and so � p lim pz ´ z0 qf pzq “ 0qq, proving the claim for n “ 1. As
zÑz0
z0 is not a pole, by (2), the claim is true for all other n P N as well.
Conversely, if for all n P N, � p lim pz ´ z0 qn f pzq “ 0 q , then in particular,
zÑz0
from the n “ 1 case, we get by (1) that z0 cannot be a removable singularity. From
the cases for all other natural numbers n ě 2, using (2), we conclude that z0 cannot
be a pole either. So z0 must be an essential singularity of f .
Corollary 4.17. If z0 is an isolated singularity of f and there exists an m P N

such that lim pz ´ z0 qm f pzq “ L ‰ 0, then z0 is a pole of order m of f .
zÑz0
Proof. lim pz ´ z0 qm`1 f pzq “ p lim pz ´ z0 qq lim ppz ´ z0 qm f pzqq “ 0 ¨ L “ 0. For

zÑz0 zÑz0 zÑz0
pN Qq ℓ ă m ` 1 we will show �pp lim pz ´ z0 qℓ f pzq “ 0qq. For ℓ “ m, this is given.
zÑz0
If lim pz ´ z0 qℓ f pzq “ 0 for some ℓ ă m, then we arrive at the contradiction that
zÑz0
0 ‰ L “ lim pz ´ z0 qm f pzq “ p lim pz ´ z0 qm´ℓ q lim ppz ´ z0 qℓ f pzqq “ 0 ¨ 0 “ 0.
zÑz0 zÑz0 zÑz0
Example 4.17. Let us reconsider Example 4.16.

We have lim z sinz z “ lim sin z “ sin 0 “ 0. So 0 is a removable singularity of sin z
z .
zÑ0 zÑ0
As lim z 3 z13 “ 1 ‰ 0, 0 is a pole of order 3 of 1
z3
.
zÑ0
1
1 1 1 1
For x ą 0, e x “ 1 ` 1!x2!x2 ` 3!x3 ` ¨ ¨ ¨ ą n!xn for all n P N. Thus
` for all n P N,
1 1
1
xn e x ą n! and �p lim x e x “ 0q. So for all n P N, �plim z n exp 1z “
n 0q. Thus 0 is
xÑ0` zÑ0
an essential singularity of exp z1 .
Logp1`zq
Exercise 4.55. As Logp1 ` zq P OpDp0, 1qq and z12 P OpCzt0uq, z2 P OpDp0, 1qzt0uq.
Prove that z “ 0 is a pole of order 1 of Logp1`zq
z2 .
Exercise 4.56 (Bernoulli numbers). Determine the set Z of complex zeroes of ez ´ 1. For
each z P Z, determine the order of z as a zero of ez ´ 1. Define f by f pzq “ ezz´1 for
z P CzZ. Show that 0 is an isolated singularity of f , and that it is removable. Prove that
there exist r ą 0 and complex Bn (n ě 0), such that
z ř
8
Bn n
ez ´1 “ z n! for 0 ă |z| ă r.
n“0
Determine B0 . Determine all the coefficients with odd indices, namely B1 , B3 , B5 , ¨ ¨ ¨ .

Hint: Use ezz´1 ´ ep´zq
´z ´1 “ ´z.
1
Exercise 4.57. Let f pzq :“ cos z´1 for all 0 ă |z| ă 2π. Prove that 0 is a pole of f , and
find its order.
We will now learn our second characterisation result for singularities, in terms of
the Laurent series coefficients.
Theorem 4.18 (Classification via Laurent coefficients). Let z0 be an isolated sin-

ř
gularity of f, and f pzq “ cn pz ´ z0 qn for 0 ă |z ´ z0 | ă R, for some R ą 0.
nPZ
Then
z0 is removable ô For all n P N, cń “ 0
z0 is a pole of order m P N ô The index m P N satisfies
‚ c´m ‰ 0 and
‚ for all n ą m, cń “ 0.
z0 is essential ô There are infinitely many
n P N such that cń ‰ 0.
removable
pole
´m
essential
infinitely many nonzero 0
Proof. (1) z0 is removable ñ (for all n P N, cń “ 0).

Let z0 be a removable singularity. Then f has an extension F P OpDpz0 , Rqq. Let
8
ř
the Taylor series expansion of F be given by F pzq “ rcn pz ´ z0 qn for z P Dpz0 , Rq.
8
ř ř n“0
For 0 ă |z ´ z0 | ă R, f pzq “ cn pz ´ z0 qn “
r cn pz ´ z0 qn . By the uniqueness of
n“0 nPZ
the Laurent series expansion, cn “ r
cn for n ě 0, and cn “ 0 for all n ă 0.
(For all n P N, cń “ 0) ñ z0 is removable.
ř 8
ř
Let cn “ 0 for all n ă 0. For 0 ă |z´z0 | ă R, f pzq “ cn pz´z0 qn “ cn pz´z0 qn .
8 nPZ n“0
ř
Define F pzq “ cn pz ´ z0 qn for z P Dpz0 , Rq. As F is given by a power series,
n“0
F P OpDpz0 , Rqq. Moreover F |Dpz0 ,Rqztz0 u “ f . Hence z0 is removable.
c´m c´1 8
ř
(2) z0 is a pole of order m ñ f pzq “ pz´z0 qm ` ¨ ¨ ¨ ` z´z 0
` cn pz ´ z0 qn , c´m ‰ 0.
n“0
Let z0 be a pole of order m. Using the classification of singularities via limits
(Theorem 4.16), lim pz ´ z0 qppz ´ z0 qm f pzqq “ lim pz ´ z0 qm`1 f pzq “ 0, and
zÑz0 zÑz0
so p¨ ´ z0 qm f has a removable singularity at z0 . For z P Dpz0 , Rqztz0 u, we have
ř ř
pz ´ z0 qm f pzq “ pz ´ z0 qm cn pz ´ z0 qn “ cn pz ´ z0 qn`m , and by the previous
nPZ nPZ
part, this last series has all coefficients of negative powers of z ´ z0 equal to 0, that
is, 0 “ c´pm`1q “ c´pm`2q “ ¨ ¨ ¨ . Hence for 0 ă |z ´ z0 | ă R,
pz ´ z0 qm f pzq “ c´m ` c´m`1 pz ´ z0 q ` c´m`2 pz ´ z0 q2 ` ¨ ¨ ¨ ,
c´m c´1 2
and so f pzq “ pz´z 0q
m ` ¨ ¨ ¨ ` pz´z q ` c0 ` c1 pz ´ z0 q ` c2 pz ´ z0 q ` ¨ ¨ ¨ . Moreover,
0
the power series c´m ` c´m`1 pz ´ z0 q ` c´m`2 pz ´ z0 q2 ` ¨ ¨ ¨ is holomorphic in
Dpz0 , Rq, and so it has the limit c´m as z Ñ z0 . Thus
lim pz ´ z0 qm f pzq “ lim pc´m ` c´m`1 pz ´ z0 q ` c´m`2 pz ´ z0 q2 ` ¨ ¨ ¨ q “ c´m .
zÑz0 zÑz0
As z0 is a pole of f order m, � p lim pz ´ z0 qm f pzq “ 0q. So c´m ‰ 0.
zÑz0
c´m c´1 8
ř
f pzq “ pz´z0 qm ` ¨¨¨ ` z´z0 ` c0 ` cn pz ´ z0 qn , c´m ‰ 0 ñ z0 pole of order m.
n“0
Suppose that there is some m P N such that c´m ‰ 0 and cn “ 0 for all n ă ´m.
Then pz ´ z0 qm f pzq “ c´m ` c´m`1 pz ´ z0 q ` c´m`2 pz ´ z0 q2 ` ¨ ¨ ¨ , and since the
right hand side defines a holomorphic function, say h, in Dpz0 , Rq, it follows that
lim pz ´ z0 qm f pzq “ lim hpzq “ hpz0 q “ c´m ‰ 0, and
zÑz0 zÑz0
lim pz ´ z0 qm`1 f pzq “ 0 ¨ c´m “ 0.
zÑz0
Thus z0 is a pole of order m of f , by Corollary 4.17.
(3) This is immediate from the previous two parts and the fact that an essential
singularity is neither a removable singularity nor a pole.
Example 4.18. Let us reconsider Example 4.16.
3 5 7 2 4 6
For z ‰ 0, sinz z “ 1z pz ´ z3! ` z5! ´ z7! ` ´ ¨ ¨ ¨ q “ 1 ´ z3! ` z5! ´ z7! ` ´ ¨ ¨ ¨ . Since there
are no negative powers of z appearing in the Laurent series expansion, it follows
that 0 is a removable singularity of sinz z .
For z ‰ 0, z13 “ ¨ ¨ ¨ ` 0` z13 ` 0` ¨ ¨ ¨ . Thus c´3 “ 1 ‰ 0, while 0 “ c´4 “ c´5 “ ¨ ¨ ¨ .
So 0 is a pole of z13 of order 3.
1
For z ‰ 0, exp 1z “ 1 ` 1!z ` 2!z1 2 ` 3!z1 3 ` ¨ ¨ ¨ . For infinitely many n P N (in fact for
all of them), cń “ n! ‰ 0. Thus 0 is an essential singularity of exp 1z .
1
Also, for z ‰ 0, exp z12 “ 1 ` 1!z1 2 ` 2!z1 4 ` 3!z1 6 ` ¨ ¨ ¨ . For infinitely many n P N,
cń ‰ 0. Thus 0 is an essential singularity of exp z12 .
Exercise 4.58. Let f P OpDq, where D is a domain, and suppose f has a zero z0 P D of
order m P N. Show that f1 is well-defined locally in a punctured disc Dpz0 , rqztz0 u Ă D,
for some r ą 0, and that z0 is a pole of f1 of order m.
Exercise 4.59. Let z0 P C, r ą 0. Suppose f P OpDpz0 , rqztz0 uq is such that f pzq ‰ 0 for
all z P Dpz0 , rqztz0 u, and z0 is a pole of f of order m. Show that f1 P OpDpz0 , rqztz0 uq has
a holomorphic extension g to Dpz0 , rq, and that g has a zero of order m at z0 .
Exercise 4.60. Let r ą 0, and z0 P C be a pole of order m of f P OpDpz0 , rqztz0 uq. Let
ř
f have the Laurent series expansion f pzq “ cn pz ´ z0 qn for 0 ă |z ´ z0 | ă r. Show that
1 dm´1 nPZ
m
c´1 “ pm´1q! lim dzm´1 ppz ´ z0 q f pzqq.
zÑz0
Exercise 4.61. True or false?

(1) If f has a Laurent expansion z ´1 ` c0 ` c1 z ` ¨ ¨ ¨ , convergent in some punctured disc
about the origin, then f has a pole at 0.
(2) A function may have different Laurent series centred at z0 , depending on the annulus
of convergence selected.
(3) If f has an isolated singularity at z0 , then it has a Laurent series centred at z0 and is
convergent in some punctured disc 0 ă |z ´ z0 | ă R.
(4) If a Laurent series for f is convergent in an annulus R1 ă |z ´ z0 | ă R2 is actually a
Taylor series (no negative powers of z ´ z0 ), then it converges (at least) in Dpz0 , R2 q.
Exercise 4.62. Decide the nature of the singularity, if any, at 0 for the following func-
tions. If the function is holomorphic or the singularity is isolated, expand the function in
appropriate powers of z convergent in a punctured disc given by 0 ă |z| ă R.
sin z, sin z1 , sinz z , sin z
z2 ,
1
sin 1
, z sin z1
z
Exercise 4.63. If z0 is a pole of order m P N of f , then show that lim pz ´ z0 qm f pzq ‰ 0.

zÑz0
Exercise 4.64. True or false?

(1) lim | exp 1z | “ `8.
zÑ0
(2) If f has a pole of order m at z0 , then there exists a polynomial p such that f ´ pz´zp0 qm
has a holomorphic extension to a disc around z0 .
(3) If f is not identically zero and holomorphic in a neighbourhood of 0, then there is an
nonnegative integer m such that zfn has a pole at 0 whenever n ą m.
(4) If f, g have poles of order mf , mg respectively at z0 , then their pointwise product f ¨ g
has a pole of order mf ` mg at z0 .
Exercise 4.65. Give an example of a function holomorphic in all of C except for essential
singularities at the two points 0 and 1.
1
Exercise 4.66. The function f given by f pzq “ z´1 clearly does not have a singularity
´1 ´2 ´3
at 0. As it has the Laurent series z ` z ` z ` ¨ ¨ ¨ for |z| ą 1, one might then say
that this series has infinitely many negative powers of z, and fallaciously conclude that the
point 0 is an essential singularity of f . Point out the flaw in this argument.
Exercise 4.67. Prove or disprove: If f and g have a pole and an essential singularity
respectively at the point z0 , then f g has an essential singularity at z0 .
Wild behaviour near essential singularities. We now show a result illustrat-

ing the ‘wild’ behaviour of a function f at its essential singularity z0 . It says that
given any w P C, any ǫ ą 0, and any arbitrarily small punctured disc ∆ with centre
z0 , there is a point z in ∆ such that f pzq lies within a distance ǫ from w. So the
image of any punctured disc centred at the essential singularity is dense in C. Or
in even more descriptive terms, f comes arbitrarily close to any complex value in
every neighbourhood of z0 .
Theorem 4.19 (Casorati-Weierstrass). If z0 is an essential singularity of f, then
@w P C, @δ ą 0, @ǫ ą 0, Dz P C such that 0 ă |z ´ z0 | ă δ and |f pzq ´ w| ă ǫ.
f pzq
δ f
z0 w ǫ
z
Proof. Suppose the statement is false. Then there exist w P C, δ ą 0 and ǫ ą 0

1
such that whenever z P Dpz0 , δqztz0 u, we have |f pzq ´ w| ě ǫ. Let gpzq :“ f pzq´w
1
for all z P Dpz0 , δqztz0 u. Then g P OpDpz0 , δqztz0 uq. As 0 ď |gpzq| “ |f pzq´w| ď 1ǫ
for all z P Dpz0 , δqztz0 u, we obtain lim pz ´ z0 qgpzq “ 0.
zÑz0
So g has a removable singularity at z0 . Denote its extension to Dpz0 , δq by g

again. Let m be the order of the zero of g at z0 . (Set m “ 0 if gpz0 q ‰ 0.) Then
gpzq “ pz ´ z0 qm hpzq, for some h P OpDpz0 , δqq and such that hpz0 q ‰ 0. Then for
0 ă |z ´ z0 | ă δ,
1
pz ´ z0 qm`1 f pzq “ pz ´ z0 qm`1 pf pzq ´ w ` wq “ pz ´ z0 qm`1 gpzq ` pz ´ z0 qm`1 w
z´z0 zÑz 0
“ hpzq ` pz ´ z0 qm`1 w ÝÑ0 hpz0 q ` 0 ¨ w “ 0.
Thus either f has a removable singularity at z0 (when m “ 0) or a pole at z0 (when
m P N). Hence z0 cannot be an essential singularity of f , a contradiction.
Example 4.19. exp 1z has an essential singularity at 0. We will show that it
assumes any given nonzero complex w (“ ρ eiθ , θ P R, ρ ą 0), in any arbitrarily small
1 cos t sin t
neighbourhood of 0. Let z “ reit , r ą 0, t P R, be such that e z “ e r í r “ ρ eiθ .
Equating absolute values, cosr t “ log ρ. Equating arguments, one solution is given
when ´ sinr t “ θ. As pcos tq2 `psin tq2 “ 1, r “ ?plog1ρq2 `θ2 . Also, t “ cos´1 ?ploglog ρ
ρq2 `θ 2
.
As θ can be increased by integral multiplies of 2π, without changing w, r can be
made as small as we please.
This example illustrates a much stronger result than the Casorati-Weierstrass The-
orem, due to Picard, which says that the image of any punctured disc centred at
an essential singularity misses at most one point of C. In our example above, the
exceptional value is w “ 0. A proof of Picard’s Theorem is beyond the scope of
this book, but can be found in [3].
Exercise 4.68. Prove that if f P OpDpz0 , rqztz0 uq, where z0 P C and r ą 0, has an
essential singularity at z0 and w P C, then there exists a sequence pzn qnPN in Dpz0 , rqztz0 u
such that lim zn “ z0 and lim f pzn q “ w.
nÑ8 nÑ8
Exercise 4.69. Prove that if f is entire and not a polynomial, and w P C, then there
exists a complex sequence pzn qnPN such that lim zn “ 8 and lim f pzn q “ w.
nÑ8 nÑ8
4.9. Residue Theorem

Let D be a domain, z0 P D, and f P OpDztz0 uq have an isolated singularity at z0 .
If Dpz0 , Rq Ă D for an R ą 0, and for z P Dpz0 , Rqztz0 u, f has the Laurent series
ř
f pzq “ cn pz ´ z0 qn , then the coefficient c´1 is called the residue of f at z0 , and
nPZ
is denoted by Respf, z0 q. In everyday language, ‘residue’ means something which
is ‘left over’. One can think of c´1 as something which is left over in the following
manner: If r P p0, Rq, and Cptq :“ z0 ` re2πit , t P r0, 1s, then as
"
ş n dz “ 0 for n ‰ ´1
c
C n pz ´ z0 q
2πi c´1 for n “ ´1,
if termwise integration (˚) is allowed below, then
ş ş ř p˚q
C f pzqdz “ C cn pz ´ z0 qn dz “ 2πi c´1 .
nPZ
4.9. Residue Theorem 107
ş ş
In fact, from the Laurent series theorem, 2πi c´1 “ C pz´zf0pzq q´1`1
dz “ C f pzqdz.
ş
Why make a fuss about the residue? 2πi c´1 “ C f pzqdz gives a way of computing
contour integrals via calculating the residue of f at z0 (which amounts to finding
the value of the coefficient c´1 in the Laurent şexpansion ofş f ). So if there is way
of calculating c´1 easily, then we can compute γ f pzqdz “ C f pzqdz “ 2πi c´1 for
any closed path γ which is Dztz0 u-homotopic to C. In Exercise 4.60, there is a way
1 m´1
of calculating c´1 via the formula c´1 “ pm´1q! lim ddz m´1 ppz ´ z0 qm f pzqq, when
zÑz0
z0 is a pole of order m of f . It turns out that some awkward
ş real integrals can be
computed by first relating them to a contour integral γ f pzqdz for an appropriate
holomorphic f in some domain and some path γ, and then using this route via the
residue to evaluate the contour integral, and eventually also the real integral.
ş2π
Example 4.20. We will view the real integral 0 5`31cos θ dθ as a contour integral
iθ íθ z`z ´1 z 2 `1
along a circular path γ as follows. If z “ eiθ , then cos θ “ e è 2 “ 2 “ 2z .
Thus if γpθq :“ eiθ , θ P r0, 2πs, then γ 1 pθqdθ “ ieiθ dθ, and so
ş2π 1
ş 1 1
ş 2i
0 5`3 cos θ dθ “ γ z 2 `1 iz
5`3
dz “ γ ´ p3z`1qpz`3q dz.
2z
2i
Let f pzq :“ ´ p3z`1qpz`3q , z P Czt´3, ´ 13 u. Then f has only two singularities,
which are isolated, and they are poles, at ´ 13 and at ´3, both of order 1. Of
these, only the one at ´ 13 lies inside γ. A reparametrisation of γ is Czt´3, ´ 13 u-
homotopic to the circle C, where Cptq “ ´ 13 ` re2πit , t P r0, 1s, and e.g. r “ 13 . So
ş2π 1
ş ş 1 1 1
0 5`3 cos θ dθ “ γ f pzqdz “ C f pzqdz “ 2πi Respf, ´ 3 q. In Dp´ 3 , rqzt´ 3 u, we have
c´1
f pzq “ z` 1 ` c0 ` ¨ ¨ ¨ . So pz ` 3 qf pzq “ c´1 ` c0 pz ` 3 q ` ¨ ¨ ¨ P OpDp´ 31 , rq. Thus
1 1
3 ş2π
Respf, ´ 13 q “ c´1 “ lim pz ` 31 qf pzq “ lim ´ 3pz`3q
2i
“ ´ 4i . So 0 5`31cos θ dθ “ π2 .
zÑ´ 31 zÑ´ 13
More generally, if f has a finite number of poles in D, then the following holds.
Theorem 4.20 (Residue Theorem). Let D be a domain, p1 , ¨ ¨ ¨ , pK P D for some
K P N, f P OpDztp1 , . . . , pK uq, p1 , . . . , pK be poles of f of order m1 , . . . , mK P N,
respectively. Let R1 , ¨ ¨ ¨ , RK ą 0 be such that the discs Dppk , Rk q Ă D, and they are
mutually disjoint. For k “ 1, ¨ ¨ ¨ K, let Ck be a circular path centred at pk traversed
once in the counterclockwise direction, contained in Dppk , Rk q. Let γ be a closed
path in Dztp1 , . . . , pK u, such that for each k “ 1, . . . , K, a reparametrisation of γ
ş ř
K
is Dztpk u-homotopic to Ck . Then γ f pzqdz “ 2πi Respf, pk q.
k“1
γ
C2
C1 p2
p1
CK
pK
m
řk pkq 8 pkq
ř
Proof. In Dppk , Rk qztpk u, f pzq “ cń pz ´ pk qń ` cn pz ´ pk qn “ rk ` hk ,
n“1 n“0
where we denote the sum with negative powers of z ´ pk by rk , and the sum with
nonnegative powers of z ´ pk by hk , k “ 1, ¨ ¨ ¨ K. Then f ´ rk “ hk P OpDppk , Rk qq,
and rk is a rational function, defined in Cztpk u, having only one singularity, namely
a pole at pk . Set
$
& f pzq ´ pr1 pzq ` ¨ ¨ ¨ ` rK pzqq if z P Dztp1 , ¨ ¨ ¨ , pK u,
gpzq :“ ř
% hk ´ rj pzq if z P Dppk , Rk q, k “ 1, ¨ ¨ ¨ , K.
jPt1,¨¨¨ ,Kuztku
Then g is well-defined as f ´ rk “ hk , k P t1, ¨ ¨ ¨ , Ku. Also, g P OpDq. Firstly,

g P OpDztp1 , ¨ ¨ ¨ , pK uq. Moreover, for any k P t1, ¨ ¨ ¨ , Ku, as hk P OpDppk , Rk qq,
and rj P OpDppk , Rk qq for all j ‰ k, it follows that g P OpDppk , Rk qq. As a
reparametrisation
ş of γ is Dztp ş 1 u-homotopic
ş to the circle C1 , by the Cauchy Integral
Theorem, 0 “ C1 gpzqdz “ γ gpzq dz “ γ pf ´ pr1 ` ¨ ¨ ¨ ` rK qq dz. Thus
ş K ş
ř K ş
p‹q ř p‹‹q ř
K ř
K
γ f dz “ γ rk dz “ Ck rk dz “ 2πi c´1,k “ 2πi Respf, pk q,
k“1 k“1 k“1 k“1
where (‹) follows by the Cauchy Integral Theorem (since for each k “ 1, ¨ ¨ ¨ , K,
rk P OpDztpk uq, and a reparametrisation of γ is Dztpk u-homotopic to Ck ), and
ş řk pkq ş
m pkq
(‹‹) follows since for all k “ 1, ¨ ¨ ¨ K, Ck rk dz “ cń Ck pz´p1 k qn dz “ c´1 .
n“1
ş Log z
Exercise 4.70. Evaluate γ 1èxp z dz along the path γ shown below.
´5 ` 10i 10 ` 10i
5 ` 5i
´5 ` 5i
´5 ´ 2i
10 ´ 2i
´5 ´ 5i 5 ´ 5i
şπ 1
Exercise 4.71. Evaluate ´π 2`cos θ dθ.
ş 2π cos θ
Exercise 4.72. Evaluate 0 5`4 cos θ dθ.
ş2π 2π
ş exp z
Exercise 4.73. Deduce 0 ecos θ cospnθ ´ sin θqdθ “ n! , n P N, by finding C z n`1 dz,
iθ
where Cpθq :“ e , θ P r0, 2πs.
1
Exercise 4.74. Let f have a zero of order 1 at z0 , so that f has a pole of order 1 at z0 .
1
Prove that Resp f1 , z0 q “ f 1 pz 0q
.
Exercise 4.75. Prove that Resp sin1 z , kπq “ p´1qk , k P Z.
ş2π π
`2n˘
Exercise 4.76. Using Residue Calculus, show that 0 psin tq2n dt “ 22n´1 n for all n P N.
Hint: Use the Binomial Theorem for finding the coefficient of z 0 in pz ´ z ´1 q2n .
Exercise 4.77. Let g be an entire function.
(1) Show that the residue of gpzqp1 ` z12 q at 0 is g 1 p0q.
ş 2π
(2) Show that 0 gpeit q cos t dt “ πg 1 p0q. Hint: 2 cos t “ z ` 1z , where z “ eit .
Exercise 4.78. The nth Fibonacci number fn , where n ě 0, is defined by the following
ř
8
recurrence relation: f0 “ 1, f1 “ 1, fn “ fn´1 ` fn´2 for n ě 2. Let F pzq :“ fn z n .
n“0
(1) Prove by induction that fn ď 2n for all n ě 0.

(2) Using the estimate fn ď 2n , deduce that the radius of convergence of F is at least 21 .
1
(3) Using the recurrence relation, show that F pzq “ 1´z´z 2 .
1
(4) Verify that Resp zn`1 p1´z´z 2 q , 0q “ fn .
? ?
(5) Using the Residue Theorem, prove that fn “ ?15 pp 1`2 5 qn`1 ´p 1´2 5 qn`1 q.
ş 1 it
Hint: If IR “ CR zn`1 p1´z´z 2 q dz, CR ptq “ Re , t P r0, 2πs, then show that lim IR “ 0.
RÑ8
1
Exercise 4.79. Consider the rational function f pzq :“ z 2022 ´1 of the complex variable z.
(1) Determine the poles of f and their order.
(2) What is the residue Respf, pq at each of the poles p?
ş
(3) Show that C f pzqdz “ 0, where Cptq :“ 2eit , t P r0, 2πs.
Integral of a real rational function using the Residue Theorem. As we

mentioned earlier, the Residue Theorem can be used to calculate contour integrals,
and sometimes gives an easy way to calculate some real integrals. Let us see how it
can be used to calculate
ş8 the improper integrals of rational functions. Consider an
integral of the type ´8 f pxqdx, where f : R Ñ R is a continuous function. Recall
that such an integral, for which the interval of integration is not finite, is called an
improper integral, defined as
ş8 ş0 şb
´8 f pxqdx “ lim a f pxqdx ` lim 0 f pxqdx,
aÑ´8 bÑ`8
when both the limits on the right hand side exist. In this case,
ş8 ş0 şr
´8 f pxqdx “ rÑ8
lim ´r f pxqdx ` lim 0 f pxqdx
rÑ`8
ş0 şr şr
“ lim p ´r f pxqdx ` 0 f pxqdxqq “ lim ´r f pxqdx.
rÑ8 rÑ`8
şr
We call lim ´r f pxqdx the Cauchy principal value of the integral6.
rÑ`8
We now assume that the integrand f is a real rational function whose denom-
inator is different from 0 for all real x, and the denominator has degree
ş8 at least
two more than the degree of the numerator. Then it can be seen that ´8 f pxqdx
exists as follows. Let f “ pq , where p, q P Rrxs have degrees deg p, deg q, respec-
tively. By the result in Exercise 1.37, there exist constants m, M, R ą 0 such that
1
|ppxq| ď M |x|deg p and |qpxq| ě m|x|deg q for all |x| ą R, giving |f pxq| ď M m x2 .
ş8 1 ş8
As R x2 dx ă 8, it follows by comparison that ´8 |f pxq|dx ă 8, and hence also
ş8 şr
´8 |f pxq|dx
ş8
exists. We will compute ´r f pxqdx using the Residue Theorem, and
determine ´8 f pxqdx by passing to the limit as r Ñ 8.
6Even if the Cauchy principal value exists, ş 8 f pxqdx need not: e.g., ş
lim r xdx “ 0,
şb
xdx “ b2
.
´8 rÑ`8 ´r 0 2
ş
Consider γ f pzqdz, around a path γ “ s ` σ as shown in the following picture,
where s denotes the straight line path r´r, rs Q x ÞÑ spxq :“ x, and σ is the
semicircular path starting at r and ending at ´r in the upper half-plane.
s
´r 0 r
Since f is rational, it has finitely many poles in the upper half plane, and choosing
r large enough, γ encloses all of these poles in its interior. We now apply the
Residue Theorem to f , taking the domain D to be the half-plane above the dotted
line shown in picture, separating the real axis and the poles of f in the lower
half-plane. Then
ş ş şr ř
γ f pzqdz “ σ f pzqdz ` ´r f pxqdx “ 2πi Respf, pk q,
k: Imppk qą0
where the sum consists

şr of terms for all the poles that lie inş the upper half-plane.
ř
From this, we obtain ´r f pxqdx “ 2πi Respf, pk q ´ σ f pzqdz.
k: Imppk qą0
We show that as r increases, the value of the integral over the corresponding
semicircular arc σ approaches 0. Indeed, as the degree of the denominator of f is
at least two more than the degree of the numerator, thereş exist M, r0 large enough
M M Mπ
such that |f pzq| ď |z| 2 for all |z| ą r0 . Hence for r ą r0 , | σ f pzq dz| ď r 2 πr “ r .
ş ş8 ř
Consequently, lim σ f pzqdz “ 0, and so ´8 f pxqdx “ 2πi Respf, pk q.
rÑ`8 k: Imppk qą0
1
ş8 π 1
Example 4.21. We will show that 0 1`x 4 dx “ . With f pzq :“ , f has
?
2 2 1`z 4
four poles of order 1: p1 “ exp 4 , p2 “ exp 4 , p3 “ exp 5πi
πi 3πi
4 , p4 “ 7πi
exp 4 .
?
2i
p2 p1
0 1
p3 p4
The first two of these poles lie in the upper half plane. We have
z´p1
Respf, p1 q “ lim 4 “ lim 1
p1`z 4 q´p1`p4 q
“ d
1
“ 1
4p31
“ ´ p41 ,
zÑp1 1`z zÑp1 1 dz
p1`z 4 q|z“p1
z´p1
z´p2
Respf, p2 q “ lim 4 “ lim 1
p1`z 4 q´p1`p4 q
“ d
1
“ 1
4p32
“ ´ p42 .
zÑp2 1`z zÑp2 2 dz
p1`z 4 q|z“p2
z´p2
ş8 ?
Thus 1
´8 1`x4 dx “ 2πip´ p41 ´ p2
“ 2πip´ 42i q “ ?π2 . As f is an even function
4 q
ş8 1 ş
1 8 1 π
(i.e., f pxq “ f p´xq for all x P R), we obtain 0 1`x 4 dx “ 2 ´8 1`x4 dx “
?
2 2
.
Exercise 4.80. Evaluate the following integrals:

ş8 1
(1) 0 1`x 2 dx.
ş8 1
(2) 0 pa2 `x2 qpb 2 `x2 q dx, where a ą b ą 0.
ş8 1
(3) 0 p1`x 2 q2 dx.
ş 8 1`x2
(4) 0 1`x4 dx.
ş8 eíξx
Exercise 4.81. Use Residue Calculus to determine ´8 1`x2 dx, where ξ P R.
Hint: Consider the two cases ξ ě 0 and ξ ă 0 separately.
Exercise 4.82 (Basel Problem). The Residue Theorem can sometimes be used for sum-
ř
8
1 π2 1 cospπzq
ming series. In this exercise, we will show that n2 “ 6 . Let f pzq :“ z 2 sinpπzq . The
n“1
zeroes of sinpπzq are located at the integers. It follows that f has poles at the integers.
(1) Show that 0 is a pole of f of order 3. Find Respf, 0q.
(2) Show that each n P Zzt0u is a pole of f of order 1. Find Respf, nq.
Consider the square path SN with vertices at pN ` 12 qp˘1 ˘ iq, traversed once in the
counterclockwise direction. For a z “ x ` iy, x, y P R, belonging to the top or bottom sides
of SN , we have the crude estimate |y| ě 12 . Thus
iπz íπz |eiπz |`|eíπz | e´πy èπy
| cospπzq e è
sinpπzq | “ | eiπz éíπz | ď ||eiπz |´|eíπz || “ |e´πy éπy | “ coshpπyq
| sinhpπyq| “ 1
| tanhpπyq| ď 1
tanh π ,
2
1 1
since tanh is strictly increasing (tanh t “ ą 0 for all t P R). Thus for z belonging
pcosh tq2
A 1
to the top or bottom sides of SN , we have |f pzq| ď |z| 2 , where A “ tanh π .
2
(3) If z belongs to the right or left side of SN and y :“ Im z, then using the trigonometric
addition formulae, show that | cospπzq| “ | sinpiπyq| and | sinpπzq| “ | cospiπyq|.
So for a z belonging to the right or left side of SN , if y :“ Im z, then
´πy ´π|y|
πy π|y|
1é´2π|y|
| cospπzq sinpiπyq e é e é
sinpπzq | “ | cospiπyq | “ | e´πy èπy | “ | eπ|y| è´π|y| | “ 1è´2π|y|
ď 1.
Thus for z belonging to the right or left sides of SN , we have |f pzq| ď |z|1 2 .
ş
(4) Show that lim SN f pzq dz “ 0 using the M L-inequality and the estimates above.
N Ñ8
ř
8
1 π2
(5) Deduce that n2 “ 6 .
n“1
Chapter 5
Harmonic functions
In this last chapter, we will study harmonic functions, which are real-valued func-
tions that solve a certain partial differential equation, called the Laplace equation.
We will show that the real and imaginary parts of holomorphic functions are
harmonic, and that the converse holds locally, and even globally on simply con-
nected domains.
Finally, some consequences of the above interplay between harmonic and holo-
morphic functions, in particular in the context of a certain ‘boundary value prob-
lem’ for the Laplace equation, called the Dirichlet problem.
5.1. What is a harmonic function?

Definition 5.1. Let U be an open subset of R2 . A function u : U Ñ R is called
harmonic if u P C 2 pU q (i.e., it has continuous partial derivatives up to order 2),
2 2
and u satisfies the Laplace equation, i.e., p∆uqpx, yq :“ BBxu2 px, yq ` BByu2 px, yq “ 0 for
all px, yq P U .
Example 5.1. Let U “ R2 . If upx, yq “ x2 ´ y 2 for px, yq P R2 , then Bu
Bx “ 2x,
B2 u Bu B2 u B2 u B2 u
Bx2
“ 2, By “ ´2y, By2 “ ´2. Thus Bx2 px, yq ` By2 px, yq “ 2 ´ 2 “ 0. Since
u P C 2 pR2 q and ∆u “ 0 in R2 , u is harmonic in R2 .
2 2
Example 5.2. If urpx, yq “ x2 `y 2 for px, yq P R2 , then BBxur2 px, yq` BByur2 px, yq “ 2`2 ‰ 0.
u is never 0 in R2 , u
Since ∆r r is not harmonic in any open subset of R2 .
Exercise 5.1. Show that the functions u below are harmonic in the given open set U .
(1) upx, yq “ logpx2 ` y 2 q, U “ R2 ztp0, 0qu.
(2) upx, yq “ ex sin y, U “ R2 .
Exercise 5.2. Show that the set HpU q of all harmonic functions on an open set U forms
a real vector space with pointwise operations. Is the pointwise product of two harmonic
functions also necessarily harmonic?
113
114 5. Harmonic functions
Why study harmonic functions? Harmonic functions are important because

they satisfy the Laplace equation. The Laplace equation is the prototype of an
important class of partial differential equations, namely ‘elliptic equations’, which
is one of the three main classes of second order linear PDEs.
Class of PDE Main example
B2 u B2 u
Elliptic Laplace equation Bx2 ` By 2 “ 0
Bu B2 u
Parabolic Diffusion equation Bt ´ Bx2 “ 0
B2 u 2
Hyperbolic Wave equation Bt2
´ BBxu2 “ 0
The Laplace equation arises in many applications. For example, the heat equation
B2 u B2 u
in the plane is Bu
Bt “ Bx2 ` By 2 , where upx, y, tq denotes the temperature at the point
px, yq belonging in a planar region U Ă R2 at time t P R. Then the steady-state
temperature (i.e., u for which Bu Bt ” 0) satisfies the Laplace equation. The Laplace
equation also has a link with stochastic processes, described very roughly below.
Consider D :“ tz P C : |z| ă 1u, and imagine a particle starting at some point
z P D and undergoing Brownian motion (e.g., a pollen grain in water, bombarded
by many tiny water molecules producing ‘random’ motion). Intuitively, we expect,
since the motion is random, that eventually the particle will leave the boundary
T :“ tz P C : |z| “ 1u of D. Denote by ζz the point on T where the particle first
exits the unit circle T, having started at z P D. So we get a random variable ζz
which lives on the unit circle.
T
D
f lives on T
0110
z 11ζz
00
Now let f : T Ñ R be a given continuous function. Then we can think of f pζz q

as being a real-valued random variable on T. Let us denote its expectation by
Epf pζz qq. This depends on where one starts initially, that is, it depends on z.
Define u : D Ñ R by upzq “ Epf pζz qq for all z P D. It turns out that u is
then harmonic, and in fact it is a solution to the ‘Dirichlet problem’, which is the
boundary value problem, where given f : T Ñ R on the boundary T, we have to
find a function u, solving the Laplace equation in the interior D of T, such that it
has a continuous extension to T, matching with the given data f :
"
∆u “ 0 in D,
u|T “ f.
5.2. Link between harmonic and holomorphic functions

It might appear that harmonic functions should belong just to the realm of real
analysis. In this section, we will now learn about two results, which justifies their
5.2. Link between harmonic and holomorphic functions 115
study in complex analysis. Roughly, a function is harmonic in an open set if and

only if locally, it is the real part of some holomorphic function.
Theorem 5.1. If U Ă C is an open set and f P OpU q, then u :“ Re f and v :“ Im f
are harmonic functions in U .
Vice versa, we will also learn the following converse to this.
Theorem 5.2. If U Ă C is simply connected and u : U Ñ R is harmonic in U,
then there exists a function v : U Ñ R, such that f :“ u ` iv P OpU q.
For the f in the conclusion of the above result, we have Re f “ u, and Im f “ v.
So every harmonic function in a simply connected domain is the real part of some
holomorphic function defined there. Since a disc is simply connected, in particular
every harmonic function is locally the real part of a holomorphic function defined
(at least in that disc). We will see later on in Exercise 5.4 that the assumption of
simply connectedness is not superfluous, and that given a harmonic function in a
non-simply connected domain, there may fail to exist a globally defined holomorphic
function in the whole domain whose real part is the given harmonic function. Before
proving Theorem 5.1, let us revisit Example 5.1.
Example 5.3. In Example 5.1, we had seen that u “ x2 ý 2 is harmonic in U “ R2 .
With z “ px, yq P R2 , u “ Repz 2 q “ Repx2 ´ y 2 ` 2xyiq. As z ÞÑ z 2 is entire,
Theorem 5.1 delivers again the earlier observation that u is harmonic in R2 . From
the calculation above, Theorem 5.1 also shows that v :“ 2xy “ Impz 2 q is harmonic.
B2 u B2 v B2 u B2 v
(Check: Bu Bv
Bx “ 2y, Bx2 “ 0, Bx “ 2x, By 2 “ 0, and so Bx2 ` By 2 “ 0 ` 0 “ 0.)
Also, it follows from Example 5.2 and Theorem 5.1 that for any open subset U of
r :“ x2 ` y 2 is not the real part of any holomorphic function defined in U .
C, u
Proof of Theorem 5.1. We have f px ` iyq “ upx, yq ` ivpx, yq for px, yq P U .

Since f is infinitely many times complex differentiable, u, v have continuous partial
derivatives of all orders (see Exercise 3.39). Using the Cauchy-Riemann equations,
2
B2 u B Bu C-R B Bv u P C B Bv C-R B Bu B2 u
Bx2 “ Bx p Bx q “ Bx p By q “ By p Bx q “ By p´ By q “ ´ By 2 , and so u is harmonic. As
v “ Repíf q, v is harmonic as well.
Now we show Theorem 5.2, which gives a converse to the above result when the
open set in question is a simply connected domain. As mentioned earlier, for more
general domains, it can happen that there are harmonic functions which are not
globally the real part of a holomorphic function; see Exercise 5.4, where we take
‚ U :“ R2 ztp0, 0qu (which is not simply connected) and
‚ u :“ logpx2 ` y 2 q (which is harmonic in U ).
Then there is no holomorphic f in Czt0u such that u “ Re f in U .
Proof of of Theorem 5.2. We will construct a holomorphic f with real part u,

and then v :“ Im f will serve as the required (harmonic) function. Set g “ Bu Bu
Bx í By .
We will prove that g is holomorphic, and then construct a primitive of g, necessarily

holomorphic, which will be (up to an additive constant) the f we seek. To show
that g is holomorphic, we will use the sufficiency of the Cauchy-Riemann equations
(Theorem 2.5). Since u is harmonic, Re g “ Bu Bu
Bx and Im g “ ´ By have continuous
partial derivatives. Moreover, as u satisfies the Laplace equation, the real and
imaginary parts of g satisfy the Cauchy-Riemann equations:
B B Bu B2 u 2
Bx pRe gq “ Bx p Bx q “ Bx2
“ ´ BByu2 “ B Bu
By p´ By q “ B
By pIm gq,
B B Bu B Bu
By pRe gq “ By p Bx q “ Bx p By q “ ´ BBx p´ Bu B
By q “ ´ Bx pIm gq.
So g P OpU q. By Theorem 3.12, as U is simply connected, g has a primitive
G P OpU q. Decompose G “ u
r ` ir
v into its real and imaginary parts u
r, vr. Then
Bu
Bx ´ i Bu 1
By “ g “ G “
Br
u
Bx
Br
v
` i Bx , p˚q
Bpr
uúq
and so Bx “ 0. Hence, by the Fundamental Theorem of Calculus, u
r ú is locally
constant along horizontal lines. Also, (˚) gives
Bu Br
´ By “ v
Bx “ ´ Br
u
By ,
where we used the Cauchy-Riemann equations to obtain the last equality. Hence
Bpr
uúq
By “ 0, showing that u
rú is locally constant along vertical lines as well. Since U
is a domain, any two points in U can be joined by a stepwise path, and so u rú must
be a constant, say C (P R), in U . Consequently, f :“ GĆ “ pr u Ćqìr v “ uìr v.
We have f P OpU q, u “ Re f , v :“ vr “ Im G, and f “ u ` iv.
Definition 5.2. Let U Ă C be open and u : U Ñ R be harmonic in U . A harmonic
conjugate of u is any v : U Ñ R such that f :“ u ` iv P OpU q.
Then as v is the imaginary part of the holomorphic function f , v is harmonic too.

Example 5.4. v :“ 2xy is a harmonic conjugate of the harmonic u :“ x2 ´ y 2
considered in Example 5.1, since f :“ u ` iv “ x2 ´ y 2 ` 2xyi “ px ` iyq2 “ z 2 ,
(where z “ x ` iy, x, y P R), is entire.
Harmonic conjugates are obviously not unique since we can just add a constant to
a harmonic conjugate and get a new harmonic conjugate. In a domain D, this is
the only indeterminacy: If v, vr are both harmonic conjugates to a harmonic u in D,
then u ` iv, u ` ir
v P OpDq, so that iprv ´ vq P OpDq. Hence by the Cauchy-Riemann
Bpr
v ´vq B0 Bpr
v ´vq B0
equations, Bx “ ´ By “ 0 and By “ Bx “ 0, showing that vr ´ v is constant
along horizontal line segments and along vertical line segments. As D is a domain,
it follows that vr ´ v is constant in D, that is, vr “ v ` C for some C P R.
In a simply connected domain, for each harmonic function, Theorem 5.2 guar-
antees the existence of a harmonic conjugate, but it is not particularly useful for
finding a harmonic conjugate (from the proof, we see that it relies on the construc-
tion of a primitive G). A more direct way is to use the Cauchy-Riemann equations,
as shown below.
5.3. Consequences of the two way traffic 117
Example 5.5. Let upx, yq “ ´psin xq sinh y for all px, yq P R2 . Then u P C 2 pR2 q.
2 2 2 2
Also, we have BBxu2 “ psin xq sinh y and BByu2 “ ´psin xq sinh y, so that BBxu2 ` BByu2 “ 0
in R2 . Thus u is harmonic in R2 . As R2 is simply connected, a harmonic conjugate
exists on R2 . If v is a harmonic conjugate, then u ` iv is entire, and so u, v satisfy
the Cauchy-Riemann equations. Thus v satisfies
Bv
Bx “ ´ Bu
By “ psin xq cosh y, p‹q
Bv Bu
By “ Bx “ ´pcos xq sinh y. p‹‹q
Integrating (‹) with respect to x from 0 to x while keeping y fixed, we obtain that
vpx, yq ´ vp0, yq “ ´pcos xqpcosh yq ` cosh y, and so
vpx, yq “ ´pcos xqpcosh yq ` Cpyq, p˚q
where Cpyq :“ vp0, yq ` cosh y. Differentiating with respect to y, we get
p‹‹q Bv p˚q dC
´ pcos xqpsinh yq “ By “ ´pcos xqpsinh yq ` dy pyq.
Consequently dC dy pyq “ 0, and so Cpyq “ C for all y. Thus we could take any
constant C, and in particular we may choose C “ 0. Hence based on the above, we
take v “ ´pcos xqpcosh yq as a candidate for a harmonic conjugate of u. In order
to verify this guess, we may note that u, v are real differentiable in R2 , and the
Cauchy-Riemann equations are satisfied by the pair pu, vq, showing that f :“ uìv
is holomorphic. But instead, let us use an old calculation from Exercise 1.57 to
check directly that f :“ u ` iv is holomorphic by giving f explicitly:
f “ u ` iv “ ´psin xqpsinh yq ´ ipcos xqpcosh yq
“ íppcos xqpcosh yq ´ ipsin xqpsinh yqq “ í cos z,
which is entire.
2
Exercise 5.3. Find harmonic conjugates for the following harmonic functions in R :
ex sin y, x3 ´ 3xy 2 ´ 2y, xp1 ` 2yq.
Exercise 5.4. Show that there is no holomorphic function f defined in Czt0u whose real
part is the harmonic function u defined by upx, yq “ logpx2 ` y 2 q, px, yq P R2 ztp0, 0qu.
Hint: If v is a harmonic conjugate of u, then show that f P OpCzt0uq satisfies f 1 pzq “ 2z .
Exercise 5.5. Is it possible to find a function v : R2 Ñ R so that the function f defined
by f px ` iyq “ x3 ` y 3 ` ivpx, yq, px, yq P R2 , is entire?
Exercise 5.6. Let U Ă C be open, u be harmonic in U , v be a harmonic conjugate of u.
u Bu `v Bv
Suppose that u2 ` v 2 never vanishes in U . Prove that uBx2 `v2Bx is harmonic in U .
Exercise 5.7. Let U Ă C be open, and v : U Ñ R be a harmonic conjugate of a harmonic
function u : U Ñ R. Prove that the product uv is harmonic in U .
5.3. Consequences of the two way traffic

In the previous section, the two results given in Theorems 5.1 and 5.2 show that
there is a two way traffic between the real analysis world of harmonic functions and
the complex analysis world of holomorphic functions, allowing a fruitful interaction

between the two worlds.
Harmonic
functions Holomorphic functions
In the previous three chapters we have learnt many pleasant properties possessed by
holomorphic functions. Let us now use these to derive some important properties
of harmonic functions. In particular, we will now show the following results:
‚ If u is harmonic, then it is C 8 .
‚ The Mean Value Property for harmonic functions.
‚ The Maximum Principle for harmonic functions.
‚ Uniqueness of solutions to the Dirichlet Problem.
Harmonic functions are smooth.

Corollary 5.3. Harmonic functions are infinitely many times real differentiable.
The definition of a harmonic function demands only twice continuous differentia-

bility. The remarkable result here says that thanks to the fact that the Laplace
equation is satisfied, in fact the function has got to be infinitely differentiable. A
result of this type (where some initial assumption of smoothness together with
the satisfaction of some differential equation implies extra regularity) is called a
regularity result in PDE theory.
Proof. Let u be harmonic in an open set U Ă C. Let z0 “ px0 , y0 q P U . Then

there is a r ą 0 such that Dpz0 , rq Ă U . As u|Dpz0 ,rq is harmonic and Dpz0 , rq is
simply connected, there exists an f P OpDpz0 , rqq such that Re f “ u in Dpz0 , rq.
But f is infinitely many times complex differentiable in Dpz0 , rq. Consequently,
u is infinitely many times real differentiable in Dpz0 , rq by Exercise 3.39. As the
choice of z0 P D was arbitrary, the result follows.
Exercise 5.8. Show that all the partial derivatives of a harmonic function are harmonic.
Mean value property. Using the Cauchy Integral Formula, we immediately ob-
tain the following ‘mean value property’ of harmonic functions, which says that
the value of a harmonic function at a point is the average (or mean) of the values
on a circle with that point as the centre.
Theorem 5.4 (Mean-value property of harmonic functions). Let U Ă C be open,
u : U Ñ R be harmonic in U, z0 P U, R ą 0 be such that Dpz0 , Rq Ă U . Then
ş
1 2π it
upz0 q “ 2π 0 upz0 ` re qdt for all r P p0, Rq.
Proof. As Dpz0 , Rq is simply connected, there exists an f P OpDpz0 , Rqq such that
u “ Re f . By the Cauchy Integral Formula, with Cptq :“ z0 ` reit , t P r0, 2πs,
1
ş f pzq ş
1 2π f pz0 `reit q
ş
1 2π
f pz0 q “ 2πi C z´z0 dz “ 2πi 0 reit
ireit dt “ 2π it
0 f pz0 ` re qdt.
Equating real parts, the claim is proved.
Exercise 5.9.
1
Let z0 P C, R ą 0, and u be a harmonic function in Dpz 2
0 , Rq. Show that
upz0 q “ πR 2 Dpz0 ,Rq
u dA, where dA “ dxdy is the area measure in R .
Exercise 5.10 (Harnack’s inequality). Let z0 P C, R ą 0, and u be a pointwise nonnega-

1
tive harmonic function in Dpz0 , Rq. Show that for all z P Dpz0 , Rq, upzq ď |z´z0 | 2 upz0 q.
p1´ R q
Exercise 5.11 (An analogue of Liouville’s theorem). Let u be nonnegative and harmonic
in R2 . Show that u is constant. Show that the condition of ‘nonnegativeness of u’ can
be relaxed to ‘u is bounded above or bounded below’. Hint: Pass to the limit R Ñ 8 in
Exercise 5.10. (See Exercise 5.14 for a different proof using Liouville’s theorem.)
Intuitively, the mean-value property of harmonic functions rules out the possibility
of them having a local maximum or local minimum (unless they are constant). To
see this, suppose z0 is such a would-be (say) local maximum, and take a small
enough r ą 0. The maximum maxC u on C, is at least as big as the mean of u
on C, and by the mean-value property at least as big as upz0 q. But as z0 is a
maximiser, we cannot have a strict inequality, and so maxC u equals the mean of u
on C. This implies u is constant on C. But as this happens with each r ą 0 small
enough, u must be a constant. We give a rigorous proof below.
Maximum Principle. From the Maximum Modulus Theorem (see p. 93), we

obtain the following.
Theorem 5.5 (Maximum Principle). If U Ă C is a simply connected domain,

u : U Ñ R is harmonic in U, z0 P U is such that upz0 q ě upzq for all z P D, then
u is constant in U .
Proof. There exists an f P OpU q such that u “ Re f . If gpzq :“ exppf pzqq (z P U ),

then g P OpU q. We have |gpz0 q| “ | exppf pz0 qq| “ eRe z0 “ eupz0 q ě eupzq “ |gpzq|
for all z P U. By the Maximum Modulus Theorem applied to g, g is constant in U .
Thus |g| is also constant in U , i.e., |g| “ eRe f “ eu is a constant in U . Taking the
(real) logarithm, it follows that u is constant in U .
Exercise 5.12 (Comparison Principle). Let D “ tz P C : |z| ă 1u, T :“ tz P C : |z| “ 1u.

Let u, v : D Y T Ñ R be two continuous functions that are harmonic in D. If upzq ě vpzq
for all z P T, then show that upzq ě vpzq in D too.
Uniqueness of solution for the Dirchlet problem. We will now use the Max-
imum Principle to show the uniqueness of solutions in the Dirichlet problem. Let
D “ tz P C : |z| ă 1u, T :“ tz P C : |z| “ 1u. The Dirchlet problem is the following:
D
Given ϕ : T Ñ R, continuous,
find a u : D Y T Ñ R such that
‚ u is continuous on D Y T,
∆u “ 0 T
‚ u|T “ ϕ,
2
‚ u P C pDq,
‚ ∆u “ 0 in D.
ϕ lives here
ϕ is called the boundary data. The Dirichlet problem arises in applications, for
example in heat conduction. Using the Maximum Principle, we have the following.
Proposition 5.6. The solution to the Dirichlet problem is unique.
Proof. Let u1 , u2 be two distinct solutions corresponding to the boundary data ϕ.
In particular, u1 “ u2 on T. So u1 differs from u2 somewhere inside D. Without
loss of generality, let w P D be such that u1 pwq ą u2 pwq. (Otherwise swap u1 , u2 .)
Then u :“ u1 ´ u2 is such that u ” 0 on T, upwq ą 0, and u is harmonic in D.
Let z0 P D Y T be the maximiser for the real-valued continuous function u on the
compact set D Y T. As u ” 0 on T, and upwq ą 0, we have z0 R T. So z0 P D.
Since upz0 q ě upzq for all z P D, the Maximum Principle implies that u must be
constant in D. As u is continuous on D Y T, and since u is 0 on T, the constant
value of u must be 0 everywhere in D Y T. Hence u1 “ u2 , a contradiction.
ş
1 2π 1´r 2
Remark 5.1. Let upreit q :“ 2π iθ
0 1´2r cospθ´tq`r 2 ϕpe qdθ, 0 ď r ă 1, t P p´π, πs.
This expression for u is called the Poisson Integral Formula, and it can be shown
that it solves the Dirichlet problem with boundary data ϕ. We will not prove this
here, but refer the interested reader to e.g. [10] for a proof. ˚
Exercise 5.13 (Half-plane Dirichlet problems). Given a continuous function ϕ : R Ñ R,
we seek a continuous, real-valued function h defined in tpx, yq P R2 : y ě 0u, such that h is
harmonic in tpx, yq P R2 : y ą 0u and hpx, 0q “ ϕpxq for all x P R.
(1) If ϕ P Rrxs, then show that we can take h given by hpx, yq “ Repϕpx ` iyqq.
1
(2) If ϕpxq “ 1`x 2 , then px, yq ÞÑ Repϕpx ` iyqq is not a solution (as its not defined at i).
i y`1
Show that hpx, yq :“ Re zì “ x2 `py`1q 2 gives a solution.
Exercise 5.14 (An analogue of Liouville’s theorem redux). Let u be a pointwise nonneg-
ative, harmonic function in R2 . Show that u is constant. Hint: If f P OpCq and u “ Re f ,
consider expp´f q. (In Exercise 5.11, we have seen a proof using Harnack’s inequality.)
Exercise 5.15. The regularity of Laplace equation solutions is not completely for free,
i.e., the solution should be at least C 2 to guarantee that it is then C 8 . Here is an example
to show that a discontinuous function may satisfy the Laplace equation! Define u : R2 Ñ R
by up0, 0q “ 0, and for z “ px, yq P Cztp0, 0qu, by upx, yq “ Repexpp´ z14 qq.
(1) Verify that u is discontinuous at p0, 0q.
1
1 ´
(2) Check that upx, 0q “ e´ x4 , up0, yq “ e y4 .
(3) As u is the real part of a holomorphic function in Czt0u, u satisfies the Laplace equation
2 2 2 2
in R2 ztp0, 0qu. Show that BBxu2 p0, 0q and BByu2 p0, 0q exist, and BBxu2 p0, 0q ` BByu2 p0, 0q “ 0.
Exercise 5.16. Let D1 , D2 be domains in C. Let ϕ : D1 Ñ D2 be holomorphic and

ϕpD1 q Ă D2 . Show that if h : D2 Ñ R is harmonic, then h ˝ ϕ : D1 Ñ R is harmonic as
well.
Let ϕ : D1 Ñ D2 be holomorphic and a bijection, and let ϕ´1 : D2 Ñ D1 be holomorphic.
We call such a map ϕ a biholomorphism. Conclude that h : D2 Ñ R is harmonic if and
only if h ˝ ϕ : D1 Ñ R is harmonic.
Thus the existence of a biholomorphism between two domains allows one to transplant
harmonic (or even holomorphic) functions from one domain to the other. This mobility
has the advantage that if D1 is ‘nice’ (like a half plane or a disc), while D2 is complicated,
then problems (like the Dirichlet Problem) in D2 can be solved by first moving over to D1 ,
solving it there, and then transplanting the solution to D2 .
A natural question is then the following: Given domains D1 and D2 , is there a biholomor-
phism between them? An answer is provided by the Riemann Mapping Theorem, a proof
of which is beyond the scope of this introductory course, but can be found e.g. in [3].
Riemann Mapping Theorem. Let D ‰ C be a simply connected domain in C.
Then there exists a biholomorphism ϕ : D Ñ D :“ tz P C : |z| ă 1u.
Thus the above result guarantees a biholomorphism between any two proper simply con-
nected domains (by a passage through D). Unfortunately, the proof does not give a prac-
tical algorithm for finding the biholomorphism. We also note that the condition D ‰ C
is needed, since otherwise ϕ would be a bounded entire function, and hence a constant by
Liouville’s Theorem, a contradiction to its purported bijectivity.
Show that the Möbius transformation ϕ : H Ñ D, where H :“ ts P C : Repsq ą 0u, given
s´1
by ϕpsq “ s`1 , s P H, is a biholomorphism between the right half plane H and the disc D.
Solutions
Solutions to the exercises from the Introduction

Solution to Exercise 0.1. Let the derivative of g at 0 be L. Taking ǫ :“ 1 ą 0,
there exists a δ ą 0 such that whenever 0 ă |x ´ 0| ă δ, we have
ˇ gpxq ´ gp0q ˇ
ˇ ˇ
ˇ ´ Lˇ ă ǫ.
x´0
In particular, with x :“ 2δ , we have 0 ă |x ´ 0| “ 2δ ă δ, and so
ˇ gpxq ´ gp0q ˇ ˇ2δ ´ 0 ˇ
ˇ ˇ ˇ ˇ
ˇ ´ Lˇ “ ˇ δ2 ´ Lˇ “ |2 ´ L| ă ǫ. (6.1)
x´0 2 ´ 0
On the other hand, with x :“ ´ 2δ , we have 0 ă |x ´ 0| “ δ2 ă δ, and so
ˇ gpxq ´ gp0q ˇ ˇ ´2p´ δ q ´ 0 ˇ
ˇ ˇ ˇ 2 ˇ
ˇ ´ Lˇ “ ˇ δ
´ Lˇ “ |2 ` L| ă ǫ. (6.2)
x´0 ´2 ´ 0
From (6.1) and (6.2) it follows, using the triangle inequality for the real absolute
value, that 4 “ |2 ` L ` 2 ´ L| ď |2 ` L| ` |2 ´ L| ă ǫ ` ǫ “ 2ǫ “ 2, a contradiction.
Hence g cannot be real differentiable at 0.
123
124 Solutions
Solutions to the exercises from Chapter 1

Solution to Exercise 1.1. Since px, yq ‰ 0, at least one among x, y is nonzero,
x ý
and so x2 ` y 2 ‰ 0. Thus p x2 `y 2
2 , x2 `y 2 q P R . Moreover,
x ý x ý ý x
px, yq ¨ p x2 `y 2 , x2 `y 2 q “ px x2 `y 2 ´ yp x2 `y 2 q, xp x2 `y 2 q ` y x2 `y 2 q
2 `y2 ´xy`xy
“ p xx2 `y 2, x2 `y 2
q “ p1, 0q.
x ý
Hence for px, yq ‰ p0, 0q, we have px, yq´1 “ p x2 `y 2 , x2 `y 2 q in C.
Solution to Exercise 1.2. Since θ P p´ π2 , π2 q, tan θ P R. We have

sin θ
1 1 tan θ pcos θq2 pcos θq2
1í tan θ “ 12 `ptan θq2 ` i 12 `ptan θq2 “ pcos θq2 `psin θq2 ` i pcoscosθqθ2 `psin θq2
pcos θq2
“ 1 ` i psin θq1 cos θ “ pcos θq2 ` ipsin θq cos θ.
Hence
1ì tan θ
1í tan θ “ p1 ` i tan θqppcos θq2 ` ipsin θq cos θq
sin θ sin θ
“ pcos θq2 ´ cos θ psin θqpcos θq ` ippsin θqpcos θq ` 2
cos θ pcos θq q
“ pcos θq2 ´ psin θq2 ` i2psin θq cos θ “ cosp2θq ` i sinp2θq.
Solution to Exercise 1.3. Let P Ă C be a set of positive elements of C. Then

since i ‰ 0, by (P3), either i P P or (i R P and í P P ). By (P2), we have
´ 1 “ i ¨ i “ píq ¨ píq P P , and so 1 “ p´1q ¨ p´1q P P. p‹q
But 1 ‰ 0, and (‹) contradicts (P3) for x “ 1.
Solution to Exercise 1.4.
cos π3 ` i sin π3
i
´ 23
0 1
?
2
?
´ 2i
Solution to Exercise 1.5. Let z, w have polar coordinates pr, αq and pρ, βq, re-
spectively. Then we have =P OS “ α, =QOS “ β, and =ROS “ α ` β. Thus
1
=ROQ “ =ROS ´ =QOS “ pα ` βq ´ β “ α “ =P OS. Also, OP r OS
OR “ rρ “ ρ “ OQ .
As the ratio of a pair of sides is equal, and the included angle is equal too, ∆P OS
is similar to ∆ROQ.
Solution to Exercise 1.6. For θ P R, pcos θ ` i sin θq3 “ cosp3θq ` i sinp3θq. But
pcos θ ` i sin θq3 “ pcos θ ` i sin θqppcos θq2 ´ psin θq2 ` i2pcos θqpsin θqq
“ pcos θqppcos θq2 ´ psin θq2 q ´ psin θq2pcos θqpsin θq ` ip¨ ¨ ¨ q.
Taking the real parts, we obtain
cosp3θq “ Repcosp3θq ` i sinp3θqq “ Reppcos θ ` i sin θq3 q

“ pcos θqppcos θq2 ´ psin θq2 q ´ 2pcos θqpsin θq2
“ pcos θqppcos θq2 ´ 1 ` pcos θq2 q ´ 2pcos θqp1 ´ pcos θq2 q
“ pcos θq3 ´ cos θ ` pcos θq3 ´ 2 cos θ ` 2pcos θq3
“ 4pcos θq3 ´ 3 cos θ.
? ?
Solution to Exercise 1.7. As 1 ` i “ 2p ?12 ` i ?12 q “ 2pcos π4 ` i sin π4 q,
?
p1 ` iq10 “ p 2q10 pcos π4 ` i sin π4 q10 “ 25 pcosp10 π4 q ` i sinp10 π4 qq
“ 32pcosp2π ` π2 q ` i sinp2π ` π2 qq
π
“ 32pcos 2 ` i sin π2 q “ 32p0 ` i 1q “ 32i.
Solution to Exercise 1.8. The angle made by 2 ` i with the positive real axis is
tan´1 21 , and the angle made by 3 ` i with the positive real axis is tan´1 31 . Thus
the angle made by p2 ` iqp3 ` iq with the positive real axis is tan´1 21 ` tan´1 31 . As
p2 ` iqp3 ` iq “ 6 ´ 1 ` ip2 ` 3q “ 5 ` 5i, the angle made by p2 ` iqp3 ` iq with the
positive real axis is tan´1 55 “ tan´1 1 “ π4 . Consequently, π4 “ tan´1 21 ` tan´1 31 .
Solution to Exercise 1.9. Suppose that the vertices A, B, C of the equilateral

triangle are at the complex numbers zA , zB , zC , and that they are labelled in the
anticlockwise fashion. Since AC “ AB and =CAB “ π3 , we have
zC ´ zA “ pcos π3 ` i sin π3 qpzB ´ zA q. p˚q
We argue by contradiction. Let p, q, m, n P Z be such
? that zC ´ zA “ p ` iq,?and
m ` in. Then (˚) becomes p ` iq “ p 21 ` 23 iqpm ` inq, i.e., p “ m
zB ´ zA “ ? 3
2 ´ 2 n,
and q “ m2 3 ` n2 . Multiplying?the expression for p by ń, the one for q by m, and
adding, we obtain qm ´ pn “ 23 pm2 ` n2 q. But m2 ` n2 ‰ 0 (as zB ‰ zA !), and so
?
3 “ 2pqm´pnq
m2 `n2 P Q, a contradiction.
126 Solutions
Solution to Exercise 1.10. Let w “ ρpcos α ` i sin αq, where ρ ě 0 and α P R.

Then ρ4 pcosp4αq ` i sinp4αqq “ w4 “ ´1 “ 1pcos π ` i sin πq. Hence ρ4 “ 1, and as
ρ ě 0, ρ “ 1. Also, 4α P tπ, π˘2π, π˘4π, ¨ ¨ ¨ u, and so α P t π4 , π4 ˘ π2 , π4 ˘π, ¨ ¨ ¨ u.
So w “ ρpcos α ` i sin αq “ 1pcos α ` i sin αq belongs to the set
´1ì
? , ´1í
tcos π4 ì sin π4 , cos 3π 3π 5π 5π 7π 7π 1ì
4 ì sin 4 , cos 4 ì sin 4 , cos 4 ì sin 4 u “ t 2 ,
?
2
? , 1í
2
? u.
2
The location of the four fourth roots of ´1 in the complex plane is shown below.
´1ì
? 1ì
?
2 2
3π
4
5π
4
π
4
´1í
? 1í
?
7π 2
2 4
Solution to Exercise 1.11. 0 “ z 6 ´z 3 ´2 “ pz 3 q2 ´2z 3 `z 3 ´2 “ pz 3 ´2qpz 3 `1q.

So z 3 “ 2 or z 3 “ ´1. We have z 3 “ 2 if and only if
? ? ? ? ?
3
? ?
3
?
z P t 3 2pcos 2π
3 ì sin 2π
3 q, 3
2pcos 4π
3 ì sin 4π
3 q, 3
2u “ t 3
2p´ 1
2 ì 2 q, 3
2p´ 1
2 ´ i 3
2 q, 2u.
We have z 3 “ ´1 if and only if
? ?
3 3
z P tcos π3 ì sin π3 , cos πì sin π, cos 5π 5π 1
3 ì sin 3 u “ t 2 ì 2 , ´1, 1
2 í 2 u.
So z 6 ´ z 3 ´ 2 “ 0 if and only if rpz 3 “ 2q or pz 3 “ ´1qs, that is, if and only if
? ? ? ? ? ? ?
z P t 3 2p´ 12 ` i 23 q, 3 2p´ 12 ´ i 23 q, 3 2u Y t 12 ` i 23 , ´1, 12 ´ i 23 u
? ? ? ? ? ? ?
“ t 3 2p´ 12 ì 23 q, 3 2p´ 12 í 23 q, 3 2, 12 ì 23 , ´1, 21 í 23 u.
Solution to Exercise 1.12. Suppose that ω P CzR is such that ω 3 “ 1. Then

0 “ ω 3 ´ 1 “ pω ´ 1qpω 2 ` ω ` 1q, and since ω ‰ 1, we have ω 2 ` ω ` 1 “ 0. Hence
ppb ´ aqω`pb ´ cqqppb ´ aqω 2 ` pb ´ cqq “ pb ´ aq2 ω 3 `pb ´ aqpb ´ cqp´1q`pb ´ cq2
“ pb ´ aq2 1`pb ´ aqpb ´ cqp´1q`pb ´ cq2
“ pb ´ aqpb ´ a ´ b ` cq ` pb ´ cq2
“ pb ´ aqpc ´ aq ` pb ´ cq2
“ bc ´ ca ´ ab ` a2 ` b2 ´ 2bc ` c2
“ a2 ` b2 ` c2 ´ ab ´ bc ´ ca “ 0.
Hence pb ´ aqω “ c ´ b or pb ´ aqω 2 “ c ´ b. But the latter case is the same as
pb ´ aqω 3 “ pc ´ bqω, that is, pc ´ bqω “ b ´ a. In either of these cases, it follows
that the distance of b ´ a to the origin is the same as the distance of c ´ b to the
origin, and moreover, the angle between the line segments joining a to b and c to
b is π3 . So the triangle formed by a, b, c is equilateral.
a c
π π
3 3
b c or b a
2π 2π
3 3
If a, b, c are all real, then the equilateral triangle must degenerate to a point r P R,
and so a “ b “ c (“ r). Thus we recover the real case result.

Let ω P CzR satisfy ω 3 “ 1. Then 0 “ 1 ´ ω 3 “ p1 ´ ωqp1 ` ω ` ω 2 q “ 0. As ω ‰ 1,
1 ` ω ` ω 2 “ 0. Also, 1 ` ω 2 ` ω 4 “ 1 ` ω 2 ` ωω 3 “ 1 ` ω 2 ` ω “ 0. We have
3n ` ˘
ř 3n
p1 ` 1q3n ` p1 ` ωq3n ` p1 ` ω 2 q3n “ k 2k
k p1 ` ω ` ω q. But
k“0
$ , $
&1 ` 1 ` 1 if k ” 0 mod 3 . & 3 if k ” 0 mod 3
p1 ` ω k ` ω 2k q “ 1 ` ω ` ω 2 if k ” 1 mod 3 “ 0 if k ” 1 mod 3
% - %
1 ` ω 2 ` ω 4 if k ” 2 mod 3 0 if k ” 2 mod 3.
`3n˘ `3n˘ ` ˘
Thus p1 ` 1q ` p1 ` ωq ` p1 ` ω q “ 3p 0 ` 3 ` ¨ ¨ ¨ ` 3n
3n 3n 2 3n
3n q. But also
p1 ` 1q3n ` p1 ` ωq3n ` p1 ` ω 2 q3n “ 23n ` p´ω 2 q3n ` p´ωq3n
“ 23n ` p´1qn ` p´1qn “ 23n ` 2p´1qn .
Solution to Exercise 1.14. Let A, B, C, D correspond to the complex numbers
a, b, c, d, respectively. Since AB 1 is obtained from AB by rotating AB about A
in an clockwise fashion by 90˝ , B 1 corresponds to a ´ ipb ´ aq. Since P is the
midpoint of BB 1 , P corresponds to bàípbáq2 . Similarly, Q, R, S correspond to
c`bípc´bq d`cípdćq a`dípa´dq
2 , 2 , 2 , respectively.
C
D
S Q
A B
B1
128 Solutions
Denote the complex numbers corresponding to P, Q, R, S, by p, q, r, s, respectively,

ipq ´sq “ ip c`bípc´bq
2 ´ a`dípa´dq
2 q “ c´bá`dìpc`bá´dq
2
´báìpbáq d`cípdćq
“ 2 ` 2 “ ´p ` r.
Hence the distance of q ´ s to the origin is the same as that of p ´ r to the origin,
showing that QS “ P R. Also, as multiplication by i produces a rotation about
the origin by 90˝ , it follows from ipq ´sq “ r ´ p that P R K QS.
Solution to Exercise 1.15. For k P N, pz ´z0 qpz k´1 `z k´2 z0 `¨ ¨ ¨`zz0k´2 `z0k´1 q
when expanded ‘telescopes’ to give z k ´ z0k . We have:
ppzq “ c0 ` c1 z ` c2 z 2 ` ¨ ¨ ¨ ` cd z d
ppz0 q “ c0 ` c1 z0 ` c2 z02 ` ¨ ¨ ¨ ` cd z0d
ppzq ´ ppz0 q “ c1 pz ´ z0 q ` c2 pz 2 ´ z02 q ` ¨ ¨ ¨ ` cd pz d ´ z0d q.
With qpzq :“ c1 ` c2 pz ` z0 q ` ¨ ¨ ¨ ` cd pz d´1 ` z k´2 z0 ` ¨ ¨ ¨ ` zz0k´2 ` z0k´1 q, we have
ppzq´ppz0 q “ pz´z0 qqpzq, i.e., ppzq “ pz´z0 qqpzq`ppz0 q. As cd ‰ 0, the polynomial
q has degree d ´ 1. If ppz0 q “ 0, then ppzq “ pz ´ z0 qqpzq ` 0 “ pz ´ z0 qqpzq.
If ppzq :“ z n ´ 1, then since ppz1 q “ 0, we have ppzq “ pz ´ z1 qq1 pzq for some
polynomial q1 of degree n ´ 1. But 0 “ ppz2 q “ pz2 ´ z1 qq1 pz2 q, and since z2 ‰ z1 ,
q1 pz2 q “ 0, and so q1 pzq “ pz ´ z2 qq2 pzq for some polynomial q2 of degree n ´ 2.
Hence ppzq “ pz ´ z1 qpz ´ z2 qq2 pzq. Proceeding in this manner, we eventually get
ppzq “ pz ´ z1 qpz ´ z2 q ¨ ¨ ¨ pz ´ zn´1 qqn´1 pzq for a polynomial qn´1 of degree 1.
Moreover, qn´1 pzn q “ 0, so that qn´1 pzq “ Cpz ´ zn q for some constant C. But
´1 “ 0n ´ 1 “ pp0q “ Cp0 ´ z1 q ¨ ¨ ¨ p0 ´ zn q “ Cp´1qn z1 ¨ ¨ ¨ zn
ś
n´1
“ Cp´1qn pcosp 2π 2π 2π 2π
n ` k n q ` i sinp n ` k n qq
k“0
“ Cp´1qn pcosp2π`pn´1qπq ` sinp2π`pn´1qπqq “ Cp´1qn p´1qn´1 “ Ć,
and so C “ 1.
Solution to Exercise 1.16. In particular, for all x P R, c0 ` c1 x ` ¨ ¨ ¨ ` cd xd “ 0.

Setting x “ 0, we obtain c0 “ 0. Suppose that for some n P t0, 1, ¨ ¨ ¨ , du we
have shown c0 “ ¨ ¨ ¨ “ cn “ 0. Hence for all x P R, cn`1 xn`1 ` ¨ ¨ ¨ ` cd xd “ 0.
Differentiating n ` 1 times with respect to x, and then setting x “ 0, we obtain
cn`1 pn`1q!`0 “ 0, and so cn`1 “ 0. By induction, it follows that c0 “ ¨ ¨ ¨ “ cd “ 0.
Solution to Exercise 1.17. First we show uniqueness. Let p “ qh ` r “ qrh ` rr

for polynomials q, qr, r, rr, where r, rr each have a degree strictly less than d. Then
pq ´ qrqh “ rr ´ r. If q ‰ qr, then the left-hand side has degree at least d, while the
right-hand side has degree strictly less than d, a contradiction. So q “ qr. This
then yields rr ´ r “ 0h “ 0, and so rr “ r as well.
We use induction on the degree of p to show existence. If p is constant, then with
q :“ 0 and r :“ pp0q, we have ppzq “ pp0q “ 0 ¨ hpzq ` pp0q “ qpzqhpzq ` rpzq.
Suppose the claim is true for all polynomials of degree ď n, for some n P N Y t0u.
Let p be a polynomial of degree n ` 1 (ě 0 ` 1 “ 1). By the Division Algorithm
for integers, n ` 1 “ d ¨ k ` ℓ, for nonnegative integers k, ℓ, where 0 ď ℓ ă d. Let
ppzq “ cn`1 z n`1 ` ¨ ¨ ¨ ` c0 , and hpzq “ ad z d ` ¨ ¨ ¨ ` a0 . Then the coefficient of the
leading term, namely z n`1 , of the polynomial cn`1 akd
hk is cn`1 , and so p ´ cn`1
akd
hk has
degree ď n. By the induction hypothesis, there exist polynomials qr, rr such that
p ´ cn`1
ak
hk “ hr
q ` rr, and such that the degree of rr is strictly less than d.
d
‚ If k “ 0, then q :“ qr, and absorb the constant cn`1

akd
hk in rr, i.e., r :“ rr ` cn`1
ak
hk .
d
‚ If k ą 0, then r :“ rr, and q :“ qr ` cn`1

akd
hk´1 .
In either case, p “ qh ` r, with r having degree ă d, completing the induction step.
Solution to Exercise 1.18. Let z1 “ x1 ` iy1 , z2 “ x2 ` iy2 , where x1 , x2 , y1 , y2

belong to R. Then z1 z2 “ x1 x2 ´ y1 y2 ` ipx1 y2 ` y1 x2 q, and
|z1 z2 |2 “ px1 x2 ´ y1 y2 q2 ` px1 y2 ` y1 x2 q2
✭✭ ✭✭
2x✭
“ x21 x22 ´ ✭ 1 x✭ 1 y✭
2x✭
2 2 2 2 2 2
2 y 1 y 2 ` y 1 y 2 ` x1 y 2 ` ✭ 2 y 1 x2 ` y 1 x2
“ x21 px22 ` y22 q ` y12 py22 ` x22 q “ px21 ` y12 qpx22 ` y22 q “ |z1 |2 |z2 |2 .
Since |z1 |, |z2 |, |z1 z2 | are all nonnegative, it follows that |z1 z2 | “ |z1 | |z2 |.
Solution to Exercise 1.19. Let z “ x ` iy, where x, y P R. Then

pzq “ x ´ iy “ x ´ ipýq “ x ` iy “ z.
Also, zz “ px ` iyqpx ´ iyq “ x2 ` y 2 ` ip´xy ` xyq “ x2 ` y 2 “ |z|2 .
x`�
iy`x´�
Finally, z`z
2 “ 2
iy
“ x “ Re z, and z´z
2i “ ✁xìy´
2i
✁xìy “ 2iy
2i “ y “ Im z.

a a
|z| “ |x ` iy| “ x2 ` y 2 “ x2 ` pýq2 “ |x ´ iy| “ |z|,
? ? a
|Re z| “ |x| “ x2 “ x2 ` 0 ď x2 ` y 2 “ |x ` iy| “ |z|,
a a a
|Im z| “ |y| “ y 2 “ 0 ` y 2 ď x2 ` y 2 “ |x ` iy| “ |z|.
Since z is the reflection of z in the real axis, and as 0 P R, the distance of z to 0
is equal to the distance of z to 0, i.e., |z| “ |z|. The inequalities |Re z| ď |z| and
|Im z| ď |z| express the fact that the length of any side in a right-angled triangle is
at most the length of the hypotenuse.
z
|z|
|Im z|
|Re z|
|z|
z
130 Solutions
Solution to Exercise 1.21. |az| “ |a||z| “ |a||z| ă 1p1q “ 1. So az ‰ 1. We have

zá 2 zá zá pzáq pzáq zzázázàa |z|2 ázáz`|a|2
| 1áz | “ 1áz p 1áz q “ p1ázq p1ázq “ 1ázázàazz “ 1ázáz`|a|2 |z|2
1ázáz`|a|2 |z|2 `|z|2 `|a|2 ´1´|a|2 |z|2 |z|2 `|a|2 ´1´|a|2 |z|2
“ 1ázáz`|a|2 |z|2
“1` 1ázáz`|a|2 |z|2
|z|2 `|a|2 ´1´|a|2 |z|2 p1´|z|2 qp1´|a|2 q
“1` |1áz|2
“1´ |1áz|2
.
p1´|z|2 qp1´|a|2 q zá 2 p1´|z|2 qp1´|a|2 q
|1áz|2
ě 0 as |z| ď 1, |a| ă 1. So | 1áz | “1´ |1áz|2
ď 1 ´ 0 “ 1.
Solution to Exercise 1.22. In the picture below, 0, z, w, w ´ z forms a parallelo-

gram. So when we form the parallelogram determined by 0, z, tpw ´ zq, the fourth
vertex, namely z ` tpw ´ zq, lies on the segment joining z and w (as the line joining
0 and tpw ´ zq is parallel to the line joining z and w). As the t increases from 0 to
1, z ` tpw ´ zq moves from z to w along the segment joining them.
w
z ` tpw ´ zq
z
w´z
tpw ´ zq
0
´z
Solution to Exercise 1.23. As ppwq “ 0, c0 ` c1 w ` ¨ ¨ ¨ ` cd wd “ ppwq “ 0 “ 0.

We have 0 “ c0 ` c1 w ` ¨ ¨ ¨ ` cd wd “ c0 ` c1w ` ¨ ¨ ¨ ` cd wd “ c0 ` c1 w ` ¨ ¨ ¨ ` cd pwqd
(using ck P R, and wk “ pwqk , 1 ď k ď d). So 0 “ c0 ` c1 w ` ¨ ¨ ¨ ` cd pwqd “ ppwq.
Solution to Exercise 1.24. Let a “ |a|pcos αì sin αq, and b “ |b|pcos β ì sin βq,
where α, β P r0, 2πq. Then ab “ |a|pcos α ` i sin αq|b|pcos β ´ i sin βq and so we have
Impabq “ |a| |b|ppcos αqp´ sin βq ` psin αqpcos βqq “ |a| |b| sinpα ´ βq.
A
β a
|a|
β´α
O α
0
|b|
B b
The area of the triangle OAB formed by 0, a, b (where O ” 0, A ” a, B ” b) is

given by 21 OA ¨ OB ¨ sin =AOB “ 21 |a||b|| sinpα ´ βq| “ 12 |Impabq| “ | Impabq
2 |.
Solution to Exercise 1.25. In ∆ABC, the points C 1 , A1 , B 1 on AB, BC, CA,

respectively, are such that AC 1 : C 1 B “ BA1 : A1 C “ CB 1 : B 1 A “ 2 : 1. Join AA1 ,
BB 1 , CC 1 , and label the intersection points in the line segment pairs pCC 1 , AA1 q,
pAA1 , BB 1 q, pBB 1 , CC 1 q as X, Y, Z, respectively. We need the ratio of the areas of
the internal triangle ∆XY Z to that of ∆ABC.
A
B1
Y
C1 Z
X
B A1 C
Let A, B, C correspond to z, 0, w P C, respectively. Then the points C 1 , A1 , B 1

correspond to 3z , 2w3 ,
2z`w 1
3 , respectively. As Y lies on the intersection of AA and
2z`w 2w
BB 1 , for some s, t P r0, 1s, t 3 “ p1 ´ sqz ` s 3 , i.e., p2t ´ 3p1 ´ sqqz “ p2s ´ tqw.
Since 0, z, w are not collinear, it follows that 2t “ 3p1 ´ sq and t “ 2s. Solving,
we get s “ 37 . Hence Y corresponds to the complex number 4z`2w 7 . Noticing the
symmetry of the situation (that is, by ‘shifting the origin’ to z and then to w),
we can read off from the above that X, Z correspond to the complex numbers
4pw´zq`2p´zq 4p´wq`2pz´wq
7 ` z “ 4w`z 2 , and 7 ` w “ w`2z
7 , respectively. Shifting the
origin to say Z, we have that X, Y then correspond to a “ ´z`3w 7 , b “ 2z`w
7 , and
so the area of the internal triangle, ∆XY Z, will be
| 12 Imp p2z`wq
7
p´z`3wq
7
1
q| “ | 12 7p7q Imp´2|z|2 ` 3|w|2 ´ wz ` 6zwq|
1 |Impzwq|
1
“ | 12 7p7q Imp´wz ´ wz ` 7zwq| “ 7 2 .
The area of ∆ABC is |Impzwq|

2 , and so the above shows that the area of ∆XY Z is
one-seventh the area of ∆ABC.
„  „ 
1 z1 z1 1 z1 z1
Solution to Exercise 1.26. For z1 , z2 , z3 P C, w :“ i det 1 z2 z2 “ í det 1 z2 z2 .
1 z3 z3 1 z3 z3
Let Sn denote the set of all permutations on t1, ¨ ¨ ¨ , nu. For M “ rmi,j s P Cnˆn ,
ř ř
det M “ psign σqm1,σp1q ¨ ¨ ¨ mn,σpnq “ psign σqm1,σp1q ¨ ¨ ¨ mn,σpnq “ det M ,
σPSn σPSn
where M is obtained from the matrix M by taking entrywise complex conjugates.
So „  „  „ 
1 z1 z1 1 z1 z1 1 z1 z1
det 1 z z
2 2 “ det 1 z z
2 2 “ ´ det 1 z z
2 2 ,
1 z3 z3 1 z3 z3 1 z3 z3
where the last equality follows by swapping the second and third columns. Thus,
„  „  ´ „  „ 
1 z1 z1 1 z1 z1 1 z1 z1 ¯ 1 z1 z1
i det 1 z2 z2 “ í det 1 z2 z2 “ í ´ det 1 z2 z2 “ i det 1 z2 z2 .
1 z3 z3 1 z3 z3 1 z3 z3 1 z3 z3
So w is its own complex conjugate, and hence it is real.

132 Solutions
Solution to Exercise 1.27. The claim is a consequence of the following:

| |w|
|z| z `
|z|
|w| w|
2 “ p |w|
|z| z `
|z| |w|
|w| wqp |z| z ` |z|
|w| wq
|w|2 2 |z|2 2 |w| |z| |z| |w|
“ |z|2 |z| ` |w|2 |w| ` |z| z |w| w ` |w| w |z| z
“ |w|2 ` |z|2 ` zw ` wz “ pw ` zqpw ` zq “ |w ` z|2 .

As |w|
|z| z is just a scaled version of z, it lies along the ray starting from 0 and passing
through z, and the distance of |w| |z| z from the origin is |w|, which is one of the lengths
|z|
of the sides of the parallelogram drawn with solid lines below. Similarly, |w| w lies
along the ray passing through 0 and w, and is at a distance of |z| from 0, which is
the other length of the side of the parallelogram drawn with solid lines. Also the
included angle between |w| |z|
|z| z and |w| w is the same as the angle between z and w. So
the parallelogram drawn with solid lines is congruent to the parallelogram drawn
with dashed lines, and in particular, the lengths of the corresponding diagonals are
the same, giving |z ` w| “ | |w| |z|
|z| z ` |w| w|.
z`w
|w| |z|
w |z| z` |w| w
|z|
|w| w
|w|
|z| z
z
0
Solution to Exercise 1.28. If z, w P S 1 , then |zw| “ |z||w| “ 1p1q “ 1, and so
zw P S 1 . Complex multiplication is associative and commutative. Also, 1 P S 1 ,
and z1 “ z “ 1z for all z P S 1 . If z P S 1 , then z ‰ 0 and 1z P S 1 too, because
1
| 1z | “ |z| “ 11 “ 1, and so every element in S 1 has a multiplicative inverse in S 1 .
1
Hence S forms an Abelian group.
For n P N, let ζ “ cos 2π 2π m
n ` i sin n . If for some m P N, ζ “ 1, then in particular,
2πm 2πm
cos n “ 1, so that p0 ‰q n “ 2πk for some k P Zzt0u, showing that n divides
m. This means that m ě n. So ζ, ¨ ¨ ¨ , ζ n´1 cannot be 1. Also, ζ n “ 1. Thus the
order of ζ P S 1 is n, and Hn “ t1, ζ, ¨ ¨ ¨ , ζ n´1 u is a subgroup of S 1 of order n.
? ?
Let H :“ tcosp 2πmq ` i sinp 2πmq : m P Zu Ă S 1 , and 1 “ cos 0 ` i sin 0 P H.
Also, H is closed under
? multiplication? because if z, w P ? H, then there ? exist m, n P Z
such that ? z “ cosp 2πmq ` i sinp
? 2πmq and w “ cosp 2πnq ` i sinp 2πnq, giving
zw “ cosp 2πpm ` nqq ` i sinp 2πpm `? nqq P H, as m `?n P Z. Finally, H is closed
under taking inverses,
? since if z “?cosp 2πmq ` i sinp 2πmq P H, where m P Z,
then z ´1 “ cosp ?2πp´mqq ` i sinp? 2πp´mqq P H too. The map ϕ : Z Ñ H given
by Z Q m ÞÑ cosp 2πmqì sinp ?2πmq P H is surjective.? It is?also injective, ?since if
for some distinct m,?n P Z, cosp
? 2πmq ` i sinp 2πmq “ cosp 2πnq ?` i sinp 2πnq,
2k
then it follows that 2πm “ 2πn`2πk for some integer k, giving 2 “ mń P Q,
a contradiction. Thus the map ϕ is a bijection. As Z is countable, so is H. (There
are many other examples of infinite countable subgroups too.)
Solution to Exercise 1.29. Let S “ z1 ` ¨ ¨ ¨ ` zn . With ζ :“ cos 2π 2π

n ` i sin n , the
set tζz1 , ¨ ¨ ¨ , ζzn u “ tz1 , ¨ ¨ ¨ , zn u. Thus S “ z1 ` ¨ ¨ ¨ ` zn “ ζz1 ` ¨ ¨ ¨ ` ζzn “ ζS,
i.e., pζ ´ 1qS “ 0. But as n ě 2, ζ ‰ 1 (otherwise 2π n “ 2πk for some integer k,
giving 1 “ nk for integers n, k with n ě 2, which is absurd). So S “ 0.
Alternatively, by Exercise 1.15, z n ´ 1 “ pz ´ z1 q ¨ ¨ ¨ pz ´ zn q, and by expanding the
right-hand side, z n ´ 1 “ z n ´ pz1 ` ¨ ¨ ¨ ` zn qz n´1 ` ¨ ¨ ¨ ` p´1qn z1 ¨ ¨ ¨ zn . ‘Comparing
coefficients’ of z n´1 on both sides (see Exercise 1.16), z1 ` ¨ ¨ ¨ ` zn “ 0.
Solution to Exercise 1.30. For z1 , z2 P C, |z1 | “ |z1 ´z2 `z2 | ď |z1 ´z2 |`|z2 |. So
|z1 | ´ |z2 | ď |z1 ´ z2 |. As this holds for all z1 , z2 P C, by swapping z1 and z2 , we
also obtain |z2 | ´ |z1 | ď |z2 ´ z1 | “ | ´ pz1 ´ z2 q| “ | ´ 1| |z1 ´ z2 | “ |z1 ´ z2 |. Thus
||z1 | ´ |z2 || ď |z1 ´ z2 |.
Solution to Exercise 1.31. We have z P C if and only if

r2 “ |z|2 ´ 2 Repazq ` |a|2
“ zz ´ paz ` azq ` aa
“ zpz ´ aq ´ apz ´ aq “ pz ´ aqpz ´ aq
“ |z ´ a|2 .
So C is the set of all points z P C, whose distance |z ´ a| to a is r. Thus C is a
circle with centre a and radius r.
Solution to Exercise 1.32. We have

|z1 `z2 |2 `|z1 ´z2 |2 “ pz1 ` z2 qpz1 ` z2 q ` pz1 ´ z2 qpz1 ´ z2 q
“ |z1 |2 `✟ z✟
z1✟ ❍ z1 `|z2 |2 `|z1 |2 ´ z1✟
2 ` z2❍ ❍
✟ ❍ z1 `|z2 |2
✟ z2 ´ z2❍
❍
“ 2p|z1 |2 ` |z2 |2 q.
Consider the parallelogram P with vertices at 0, z1 , z2 , z1 `z2 in the complex plane.
Then |z1 ` z2 | denotes the length of one diagonal of P , while |z1 ´ z2 | is the length
of the other diagonal of P . Also, |z1 |, |z2 | are the lengths of the sides of P . So the
above equality says:
In a parallelogram, the sum of the squares of the lengths of the diagonals
equals the sum of the squares of the lengths of all the sides.
Solution to Exercise 1.33. From Exercise 1.32, with z1 “ z and z2 “ r, we

obtain |z ´ r|2 ` |z ` r|2 “ 2p|z|2 ` |r|2 q “ 2pr 2 ` r 2 q “ 4r 2 “ p2rq2 . Thus in the
triangle formed by z, p´r, 0q, p0, rq, the sum of the squares of the lengths of the
two sides (namely, |z ´ r|2 ` |z ` r|2 ) equals the square of the length of the third
side (i.e., p2rq2 ). By the converse of the Pythagoras Theorem, the triangle formed
by z, p´r, 0q, p0, rq is a right angled triangle. Also, the right angle is opposite the
longest side (hypotenuse), and the longest side in our case is the diameter.
134 Solutions
Solution to Exercise 1.34. (1),(2),(3) are depicted below (left to right):
1í
2
(4) If z “ x ` iy (x, y P R), then Repz ´ p1 ´ iqq “ 3 ô x ´ 1 “ 3, i.e., x “ 4.

See the picture on the left below.
2
0 4
´4
(5) If z “ xìy (x, y P R), then |Impz ´p1íqq| ă 3 ô |y `1| ă 3, i.e., ´4 ă y ă 2.

See the picture on the right above.
(6) tz P C : |z ´ p1 ´ iq| “ |z ´ p1 ` iq|u is the set of all complex numbers whose
distance to 1 ´ i is equal to its distance to 1 ` i. So it is the set of all points
lying on the perpendicular bisector of the line segment joining 1 ´ i to 1 ` i.
Thus the set is the real line R.
1ì
R
z
1í
(7) |z ´ p1 ´ iq| ` |z ´ p1 ` iq| “ 2 means the sum of the distances of z to 1 ` i and

to 1 ´ i is 2. As the distance between 1 ´ i and 1 ` i is 2, z lies on the line
segment joining 1 ´ i and 1 ´ i.
1ì
1ì
1í
E
1í
(8) The set of z such that |z ´ p1 ´ iq| ` |z ´ p1 ` iq| “ 3 lies on an ellipse E with
foci at 1 ` i and 1 ´ i, and so tz P C : |z ´ p1 ´ iq| ` |z ´ p1 ` iq| ă 3u is the
region in the interior of the ellipse E. See the picture on the right above.
Solution to Exercise 1.35. Let z, w make angles α, β, respectively, with the

positive x-axis in the counterclockwise direction. Then
z “ |z|pcos α ` i sin αq “ |z|pcos α ´ i sin αq “ |z|pcosp´αq ` i sinp´αqq.
So zw “ |z|pcosp´αqì sinp´αqq|w|pcos βì sin βq “ |z||w|pcospβ´αqì sinpβ´αqq.
Thus Repzwq “ |z||w| cospβ ´ αq “ pOAqpOBq cos =AOB. Hence
AB 2 “ |z ´ w|2 “ pz ´ wqpz ´ wq “ |z|2 ` |w|2 ´ zw ´ zw

“ |z|2 ` |w|2 ´ 2 Repzwq “ OA2 ` OB 2 ´ 2pOAqpOBq cos =AOB.
z
β
α
´α
Solution to Exercise 1.36. Let pu, vq P R2 . We want p “ px, y, zq P S 2 ztnu such

x y
that ϕppq “ pu, vq, i.e., p 1´z , 1´z q “ pu, vq. For such a triple px, y, zq, we have that
x2 y2 x2 `y 2 1´z 2 1`z u2 `v2 ´1
u2 `v 2 “ p1´zq2 ` p1´zq2 “ p1´zq2 “ p1´zq2 “ 1´z , and so z “ u2 `v2 `1 . Also,
u2 `v2 ´1 2u 2v
x “ up1 ´ zq “ up1 ´ u2 `v2 `1 q“
u2 `v2 `1 and y “ vp1 ´ zq “ u2 `v2 `1 .
2u 2v u2 `v2 ´1
So we define p :“ px, y, zq “ p u2 `v 2 `1 , u2 `v 2 `1 , u2 `v 2 `1 q. Then
2 2 4 `v 4 `1`2u2 v 2 ´2u2 ´2v 2 2 `v 2 `1q2

x2 ` y 2 ` z 2 “ 4u `4v ù pu 2 `v 2 `1q2 “ pu
pu2 `v2 `1q2
“ 1.
2u 2v u `v ´1 2 2
Thus p P S 2 . Also p ‰ n, since otherwise p u2 `v 2 `1 , u2 `v 2 `1 , u2 `v 2 `1 q “ p0, 0, 1q,
`v ´1 2 2
and so from the first two entries, u “ 0 and v “ 0, implying uu2 `v 2 `1 “ ´1 ‰ 1, a
2 1 2u 2v
contradiction. So p P S ztnu. Finally, ϕppq “ u2 `v2 ´1 p u2 `v2 `1 , u2 `v 2 `1 q “ pu, vq.
1´
u2 `v 2 `1
2u 2v u2 `v2 ´1
Hence ϕ´1 pu, vq “ p u2 `v 2 `1 , u2 `v 2 `1 , u2 `v 2 `1 q.
Let w “ pu, vq and w1 “ pu1 , v 1 q. Let A, A1 , a, a1 be the lengths in R3 of the segments

joining n to w, w1 , ϕ´1 w, ϕ´1 w1 , respectively. Then A2 “ u2 ` v 2 ` 1 “ |w|2 ` 1
and A12 “ u12 ` v 12 ` 1 “ |w1 |2 ` 1.
a
Considering the similar triangles shown in the picture on page 16, A “ 1´z
1 , where
u2 `v2 ´1 |w|2 ´1 ´1 2 a1 1´z 1
z “ u2 `v2 `1 “ |w|2 `1 is the z coordinate of ϕ w P S ztnu. Similarly, A 1 “ 1 ,
u12 `v12 ´1 |w 1 |2 ´1 2 2
where z 1 “ u12 `v12 `1
“ |w 1 |2 `1
. These yield a “ ? and a1 “ ? .
1`|w|2 1`|w 1 |2
The triangles formed by n, ϕ´1 w, ϕ´1 w1 and by n, w, w1 share the angle θ between
the ray from n to w, and the ray from n to w1 . The Cosine Formula in each triangle
136 Solutions
gives |w ´ w1 |2 “ A2 ` A12 ´ 2AA1 cos θ and dc pw, w1 q2 “ a2 ` a12 ´ 2aa1 cos θ. Thus
2 À12 ´|w´w 1 |2
dc pw, w1 q2 “ a2 ` a12 ´ 2aa1 A 2AA1
4 4 2 ? 2 `1`|w 1 |2?
|w|2? `1´|w´w 1 |2
“ |w|2 `1 ` |w 1 |2 `1 ´ 2?
|w|2 `1 |w 1 |2 `1 2 |w|2 `1 |w 1 |2 `1
✘2✘ ✘ ❳
❳1 |❳
2 `1q´4p|w|✘ ✘ ❳
“ ✘ ❳ ✘ `1`|w ❳
❳ ❳
4p|w| `1q`4p|w 2 1 |2 `1´|w´w 1 |2 q
p1`|w|2 qp1`|w 1 |2 q
4|w´w 1 |2
“ p1`|w|2 qp1`|w 1 |2 q
.
Taking square roots, the result follows. As dc pw, w1 q is at most the diameter of the
sphere, dc pw, w1 q ď 2. If w, w1 P Czt0u, then
1
1
2| w ´ w11 | 2 |w ´w|
|w||w1 | 2|w´w 1 |
dc pw´1 , w1´1 q “ b 1
“ ?
p|w|2 `1qp|w1 |2 `1q
“? “ dc pw, w1 q.
p1` qp1` 11 2 q p1`|w|2 qp1`|w 1 |2 q
|w|2 |w | |w||w1 |
Solution to Exercise 1.37. For R ě 1, k P t0, 1, ¨ ¨ ¨ , du, and z P C satisfying

|z| ą R, we have |z|k “ |z|k 1 ď |z|k Rd´k ď |z|k |z|d´k “ |z|d . Hence for any R ě 1
and all z P C satisfying |z| ą R, we have
|ppzq| “ |c0 ` c1 z ` ¨ ¨ ¨ ` cd z d | ď |c0 | ` |c1 ||z| ` ¨ ¨ ¨ ` |cd ||z d |
ď |c0 ||z|d ` |c1 ||z|d ` ¨ ¨ ¨ ` |cd ||z|d “ M |z|d ,
where M :“ |c0 | ` |c1 | ` ¨ ¨ ¨ ` |cd | ą 0.
cd´1 c1 c0 |c | |c1 | |c0 |
For z ‰ 0, ppzq “ z d pcd ` z `¨ ¨ ¨` z d´1 ` z d q. As nÑ8 lim p d´1 n `¨ ¨ ¨` nd´1 ` nd q “ 0,
|c | |c1 | |c0 | |cd |
there exists an N large enough so that d´1 N ` ¨ ¨ ¨ ` N d´1 ` N d ă 2 . Hence for
|z| ą R :“ N (ě 1), with m :“ |c2d | , we have

c c1 c0 cd´1 c1 c0
|ppzq| “ |z d ||cd ` d´1 d
z ` ¨ ¨ ¨ ` z d´1 ` z d | ě |z| p|cd | ´ | z ` ¨ ¨ ¨ ` z d´1 ` z d |q
|c | |c1 | |c0 |
ě |z|d p|cd | ´ p d´1|z| ` ¨ ¨ ¨ ` |z|d´1 ` |z|d qq
|c | |c1 | |c0 |
ě |z|d p|cd | ´ p d´1N ` ¨ ¨ ¨ ` N d´1 ` N d qq
ě |z|d p|cd | ´ |c2d | q “ |c2d | |z|d “ m|z|d .
Solution to Exercise 1.38. (‘If’ part) Let pRe zn qnPN , pIm zn qnPN be convergent
in R to Re L, Im L respectively. Given ǫ ą 0, let N P N be such that for all
n ą N , |Re zn ´ Re L| ă ?ǫ2 and |Im zn ´ Im L| ă ?ǫ2 . Then for all n ą N ,
a b
2 2
|zn ´ L| “ pRe zn ´ Re Lq2 ` pIm zn ´ Im Lq2 ă ǫ2 ` ǫ2 “ ǫ.
So pzn qnPN converges to L.
(‘Only if’ part) Let pzn qnPN converge to L. Given ǫ ą 0, let N P N be such that for
all n ą N , |z ´ L| ă ǫ. For n ą N , |Re zn ´ Re L| “ |Repzn ´ Lq| ď |zn ´ L| ă ǫ,
and |Im zn ´ Im L| “ |Impzn ´ Lq| ď |zn ´ L| ă ǫ. Hence pRe zn qnPN , pIm zn qnPN are
convergent in R to Re L, Im L, respectively.
Solution to Exercise 1.39. (‘Only if’ part) Suppose pzn qnPN converges to L.
Then pRe zn qnPN and pIm zn qnPN converge to Re L and Im L, respectively. Hence
pRe zn qnPN and pÍm zn qnPN converge to Re L and Ím L, respectively, that is,
pRe zn qnPN and pIm zn qnPN converge to Re L and Im L, respectively. Consequently
pzn qnPN converges to L.
(‘If’ part) Suppose pzn qnPN converges to L. By the previous part, ppzn qqnPN con-
verges to pLq, that is, pzn qnPN converges to L.
Solution to Exercise 1.40. Let pzn qnPN be a Cauchy sequence in C. Then

|Re zn ´ Re zm | “ |Repzn ´ zm q| ď |zn ´ zm | and
|Im zn ´ Im zm | “ |Impzn ´ zm q| ď |zn ´ zm |
show that the two real sequences pRe zn qnPN and pIm zn qnPN are then also Cauchy
sequences in R. Since R is complete, they are convergent to, say, a, b P R, respec-
tively. But then pzn qnPN converges in C to a ` ib. Hence C is complete.

(1) Given R ą 0, let N P N be such that N ¨ 1 ą R (such an N exists by the
Archimedean Property). For n ą N , |zn | “ |nin | “ n ą N ą R. So pnin qnPN
tends to 8.
(2) Let zn “ xn ` iyn , where xn , yn P R, n P N. From Exercise 1.36,
2 2
xn `yn ´1
2xn 2yn 4 4
d2 pϕ´1 zn , nq2 “ p x2 `y 2 2 2
2 `1 q `p x2 `y 2 `1 q `p x2 `y 2 `1 ´1q “ x2n `yn2 `1 “ |zn |2 `1
.
n n n n n n
(‘Only if’ part) Given ǫ ą 0, let R :“ 2ǫ ą 0. As pzn qnPN tends to 8, there

exists an N P N such that for all n ą N , we have |zn | ą R. So for all n ą N ,
|d2 pϕ´1 zn , nq ´ 0| “ ? 2 2 ă |z2n | ă R2 “ ǫ. So lim d2 pϕ´1 zn , nq “ 0.
|zn | `1 nÑ8
(‘If’ part) Let R ą 0. Set ǫ “ ? 2

ą 0. As the sequence pd2 pϕ´1 zn , nqqnPN
1`R2
converges to 0, there exists an index N P N such that for all n ą N , we have
? 2 2 “ |d2 pϕ´1 zn , nq ´ 0| ă ? 2 2 , i.e., |zn | ą R. So pzn qnPN tends to 8.
|zn | `1 1`R
Solution to Exercise 1.42. Let z0 P C and ǫ ą 0. Set δ “ ǫ ą 0. If |z ´ z0 | ă δ,

then |Re z ´ Re z0 | “ |Repz ´ z0 q| ď |z ´ z0 | ă δ “ ǫ, and by the reverse triangle
inequality (Exercise 1.30), ||z| ´ |z0 || ď |z ´ z0 | ă δ “ ǫ. So z ÞÑ Re z, |z| are
continuous at z0 . As z0 P C was arbitrary, z ÞÑ Re z, |z| are continuous on C.
Solution to Exercise 1.43. First consider the case of the unit circle; r “ 1.
There is no loss of generality in assuming that P1 , ¨ ¨ ¨ , Pn are the n distinct nth
roots of unity, say 1, z1 , ¨ ¨ ¨ , zn´1 , and these are the distinct roots of the polynomial
z n ´ 1. So pz ´ 1qpz n´1 ` z n´2 ` ¨ ¨ ¨ ` z ` 1q “ z n ´ 1 “ pz ´ 1qpz ´ z1 q ¨ ¨ ¨ pz ´ zn´1 q.
Thus for all z ‰ 1, pz ´ z1 q ¨ ¨ ¨ pz ´ zn´1 q “ z n´1 ` z n´2 ` ¨ ¨ ¨ ` z ` 1. As both
sides are continuous functions on C that match everywhere on Czt1u, their values
must also match at 1 (this can be seen e.g. via the sequential characterisation
138 Solutions
of continuity at a point). So p1 ´ z1 q ¨ ¨ ¨ p1 ´ zn´1 q “ 1 ` 11 ` ¨ ¨ ¨ ` 1n´1 “ n.

Hence |1 ´ z1 | ¨ ¨ ¨ |1 ´ zn´1 | “ n. So in the unit circle case, we have shown that
pP1 P2 qpP1 P3 q ¨ ¨ ¨ pP1 Pn q “ n. Next, if we scale the unit circle by r ą 0, then
each diagonal length of the inscribed n-gon gets amplified by a factor of r, and so
pP1 P2 qpP1 P3 q ¨ ¨ ¨ pP1 Pn q “ nr n´1 .
Solution to Exercise 1.44. (‘If’ part) Suppose f is not continuous at z0 , i.e.,

�r@ǫ ą 0 Dδ ą 0 such that @z P S satisfying |z ´ z0 | ă δ, |f pzq ´ f pz0 q| ă ǫs. Thus
Dǫ ą 0 such that @δ ą 0 Dz P S such that |z ´ z0 | ă δ, but |f pzq ´ f pz0 q| ě ǫ.
For n P N, taking δ “ n1 , there exists a zn P S such that |zn ´ z0 | ă δ “ n1 , but
|f pzn q ´ f pz0 q| ě ǫ. The sequence pzn qnPN is contained in S and is convergent with
limit z0 : Indeed, given ε ą 0, by the Archimedean Property, there exists an N P N
such that N ą ε´1 , and for all n ą N , |zn ´ z0 | ă n1 ă N1 ă ε. But the sequence
pf pzn qqnPN does not converge to f pz0 q: E.g. 2ǫ ą 0, and if N P N is such that for
all n ą N , |f pzn q ´ f pz0 q| ă 2ǫ , then ǫ ď |f pzN `1 q ´ f pz0 q| ă 2ǫ , a contradiction.
(‘Only if’ part) Let f be continuous at z0 , and pzn qnPN be a sequence contained in
S with limit z0 . As f is continuous at z0 , given ǫ ą 0, there exists a δ ą 0 such that
for all z P S satisfying |z ´z0 | ă δ, |f pzq´f pz0 q| ă ǫ. As pzn qnPN is convergent with
limit z0 , there exists an N P N such that for all n ą N , |zn ´ z0 | ă δ. Consequently,
for n ą N , |f pzn q ´ f pz0 q| ă ǫ. So pf pzn qqnPN is convergent with limit f pz0 q.
Solution to Exercise 1.45. (‘If’ part) Suppose u and v are continuous at z0 . Let
ǫ ą 0. Then there exists a δ1 ą 0 such that whenever z P S satisfies |z ´ z0 | ă δ1 ,
|upzq ´ upz0 q| ă ǫ{2. Also, there exists a δ2 ą 0 such that whenever z P S satisfies
|z ´ z0 | ă δ2 , |vpzq ´ vpz0 q| ă ǫ{2. So with δ :“ mintδ1 , δ2 u ą 0, if |z ´ z0 | ă δ, then
|f pzq´f pz0 q| “ |upzq` ivpzqúpz0 q´ ivpz0 q| ď |upzqúpz0 q|`|i||vpzq´ vpz0 q| ă 2ǫ ` 2ǫ “ ǫ.
So f is continuous at z0 .
(‘Only if’ part) Let f be continuous at z0 . Let ǫ ą 0. There exists a δ ą 0 such
that if z P S satisfies |z ´ z0 | ă δ, then |f pzq ´ f pz0 q| ă ǫ. So if |z ´ z0 | ă δ, then
|upzq ´ upz0 q| “ |Re f pzq ´ Re f pz0 q| “ |Repf pzq ´ f pz0 qq| ď |f pzq ´ f pz0 q| ă ǫ,
and |vpzq ´ vpz0 q| “ |Im f pzq ´ Im f pz0 q| “ |Impf pzq ´ f pz0 qq| ď |f pzq ´ f pz0 q| ă ǫ.
Thus both u and v are continuous at z0 .
Solution to Exercise 1.46. Let U :“ tz P C : pRe zqpIm zq ą 1u, and define

F :“ AU (the complement of U ). If pzn qnPN is a sequence in F such that pzn qnPN
converges to L in C, then for all n P N, pRe zn qpIm zn q ď 1, and pRe zn qnPN ,
pIm zn qnPN converge respectively to Re L and Im L. So their termwise product
ppRe zn qpIm zn qqnPN converges pRe LqpIm Lq. The inequalities pRe zn qpIm zn q ď 1
(n P N) yield pRe LqpIm Lq ď 1, i.e., L P F . So F is closed, i.e., AF “ U is open.
(Alternatively, as U is the inverse image ψ ´1 pp1, 8qq of the open set p1, 8q Ă R
under the continuous map R2 Q px, yq ÞÑ ψpx, yq :“ xy P R, U is open.)
Next we will show that U is not a domain. Let U be path-connected. Then there is
a (stepwise) path γ : ra, bs Ñ U that joins γpaq “ 2 ` 2i P U to γpbq “ ´2 ´ 2i P U .
ϕ
As C Q z ÞÑ Re z P R is continuous, ra, bs Q t ÞÝÑ Re γptq P R is continuous too.
We have ϕpaq “ Re γpaq “ Rep2`2iq “ 2 and ϕpbq “ Re γpbq “ Rep´2 ´ 2iq “ ´2. As
ϕpaq “ 2 ą 0 ą ´2 “ ϕpbq, by the Intermediate Value Theorem, there is a t˚ P ra, bs
such that 0 “ ϕpt˚ q “ Re γpt˚ q. So pRe γpt˚ qqpIm γpt˚ qq “ p0qpIm γpt˚ qq “ 0 ą 1,
showing that γpt˚ q R U , a contradiction.
Solution to Exercise 1.47. Since D is open, it is clear that its reflection in the
real axis, D, r is also open. Let w1 , w2 P D. r Then w1 , w2 P D. As D is a domain,
there exists a stepwise path γ : ra, bs Ñ C such that γpaq “ w1 , γpbq “ w2 and for
all t P ra, bs, γptq P D. Now define γ r : ra, bs Ñ C by γ rptq “ γptq, t P ra, bs. Then
rpaq “ w1 “ w1 , r
γ γ pbq “ w2 “ w2 , and for all t P ra, bs, γ
rptq “ γptq P D. r As rγ is the
composition of the continuous functions γ and z ÞÑ z, γ r is continuous. Also, since
γ is a stepwise path, there exist points t0 “ a ă t1 ă ¨ ¨ ¨ ă tn ă tn`1 “ b such
that for each k “ 0, 1, ¨ ¨ ¨ , n, the restriction γ|rtk ,tk`1 s has either constant real part
or has a constant imaginary part. Consequently γ r|rtk ,tk`1 s (which has the same
real part as γ|rtk ,tk`1 s and has imaginary part which is minus the imaginary part
of γ|rtk ,tk`1 s ) also has either a constant real part or has a constant imaginary part.
So rγ is a stepwise path too. Hence D r is path-connected.
Solution to Exercise 1.48. expp3 ` πiq “ e3 pcos π ` i sin πq “ e3 p´1 ` i0q “ é3
and exppi 9π π 0 π π
2 q “ exppip4π ` 2 qq “ e pcosp4π ` 2 q ` i sinp4π ` 2 qq “ 1p0 ` i1q “ i.
Solution to Exercise 1.49. With z “ x ` iy (x, y P R), ex pcos y ` i sin yq “ πi.

Taking absolute values, ex “ π, and so x “ log π. Then cos y ` i sin y “ i, so that
sin y “ 1 and cos y “ 0. Thus y “ π2 ` 2πk, k P Z. Hence if exp z “ πi, then
z P tlog π ` ip π2 ` 2πkq, k P Zu.
Vice versa, if z “ log π ` ip π2 ` 2πkq for some k P Z, then
exp z “ elog π pcosp π2 ` 2πkq ` i sinp π2 ` 2πkqq “ πp0 ` i1q “ πi.
Consequently, exp z “ πi if and only if z P tlog π ` ip π2 ` 2πkq, k P Zu.
Solution to Exercise 1.50. Let γptq :“ exppitq, t P r0, 2πs. Then

γptq “ exppitq “ e0 pcos t ` i sin tq “ cos t ` i sin t.
The point pcos t, sin tq lies on a circle of radius 1 and centre p0, 0q, and with in-
creasing t, this point moves anticlockwise. Hence the curve t ÞÑ γptq is the circle
traversed once in the anticlockwise direction, as shown below.
γptq “ cos tì sin t
γp0q “ γp2πq “ 1
140 Solutions
Solution to Exercise 1.51. If z “ x ` iy (x, y P R), then | exppízq| ă 1 if and

only if | exppíx ` yq| ă 1, i.e., ey ă 1, i.e., y ă 0. So D “ tz P C : Im z ă 0u, i.e.,
D is the half-plane below the real axis in the complex plane. As D is open and
path-connected, it is a domain.
Solution to Exercise 1.52. As exppt ` itq “ et pcos t ` i sin tq, the curve is given
by t ÞÑ pet cos t, et sin tq. The curve is a spiral, and as t Œ ´8, et pcos t ` i sin tq
converges to 0, while the curve spirals outwards as t Õ 8.
Solution to Exercise 1.53. If z “ x ` iy, where x, y P R, then

2 ý 2
exppz 2 q “ expppxìyq2 q “ exppx2 ´ y 2 ` 2xyiq “ ex pcosp2xyq ` i sinp2xyqq.
2 2 2 2 2 2
So | exppz 2 q|
“ ex ý ,
Re exppz 2 q
“ ex ý
cosp2xyq, Im exppz 2 q “ ex ý sinp2xyq.
x
ý
1
For z ‰ 0, exp 1z “ exp xìy “ exp xxíy
2 `y 2 “ e
x2 `y 2 pcos
x2 `y 2
` i sin x2ý
`y 2
q. Thus
x x x
y y
| exp 1z | “ e x2 `y2 , Re exp 1z “ e x2 `y2 cos x2 `y 1
2 , Im exp z “ é
x2 `y 2 sin
x2 `y 2
.

exp z “ ex cos y ` iex sin y “ ex cos y ´ iex sin y “ ex pcospýq ` i sinpýqq
“ exppx ´ iyq “ exp z.
Solution to Exercise 1.55. For x, y P R, exppx ` iyq “ ex pcos y ` i sin yq, and
so the image of vertical lines are circles and the image of horizontal lines are rays
from 0. Thus the image of the square is a portion of an annulus, as shown below.
ǫ ea´ǫ ea`ǫ
a´ǫ a a`ǫ ´ǫ
´ǫ
The area of the sector with radius r and opening angle θ is 12 θr 2 , so that the ratio
1
p2ǫqppea`ǫ q2 ´pea´ǫ q2 q 4ǫ
ρpǫq of the areas is ρpǫq “ 2
p2ǫq2
“ e2pa´ǫq e 4ǫ´1 . Hence we obtain
4ǫ x
lim ρpǫq “ lim e2pa´ǫq e 4ǫ´1 “ e2pa´0q de 2a 0 2a
dx |x“0 “ e e “ e .
ǫÑ0 ǫÑ0
Solution to Exercise 1.56. For z1 , z2 P C, we have

psin z1 qpcos z2 q ` pcos z1 qpsin z2 q
pexppiz1 qéxppíz1 qq pexppiz2 qèxppíz2 qq
“ 2i 2 ` pexppiz1 qèxppíz
2
1 qq pexppiz2 qéxppíz2 qq
2i
2 exppipz1 `z2 qq´2 exppípz1 `z2 qq
“ 4i “ sinpz1 ` z2 q.
Solution to Exercise 1.57. For y P R, we have

y èý y éý e2y `2è´2y ´pe2y ´2è´2y q 4
pcosh yq2 ´ psinh yq2 “ p e 2 q2 ´ p e 2 q2 “ 4 “ 4 “ 1.
Let z “ x ` iy, where x, y P R. Then using the addition formula on page 24,
cos z “ cospx ` iyq “ pcos xqpcospiyqq ´ psin xqpsinpiyqq
ý èy ý éy
“ pcos xq e 2 ´ psin xq e 2i “ pcos xqpcosh yq ´ psin xqp´ sinh y
i q
“ pcos xqpcosh yq ´ ipsin xqpsinh yq.
Thus
| cos z|2 “ pcos xq2 pcosh yq2 ` psin xq2 psinh yq2
“ p1 ´ psin xq2 qpcosh yq2 ` psin xq2 ppcosh yq2 ´ 1q
“ pcosh yq2 ´ psin xq2 .
Solution to Exercise 1.58. Let z “ x ` iy, where x, y P R. Then cos z “ 3 gives

pcos xqpcosh yq “ 3 p˚q
psin xqpsinh yq “ 0. p˚˚q
Now sinh y “ 0 if and only if y “ 0. If y “ 0, then z “ x ` iy “ x is real, and
r´1, 1s Q cos x “ cos z “ 3, a contradiction. Thus sinh y ‰ 0, and so (˚˚) implies
y ý
that sin x “ 0. Hence x P tnπ : n P Zu. So cos x “ ˘1. As cosh y “ e è 2 ą 0 for
all y P R, (˚) implies cos x cannot be ´1. So cos x “ 1 and x P t2πn : n P Zu. Then
ey èý y 2
?
y ` 1 “ 0. Thus ey “ 3 ˘ 2 2, and
cosh y “ 3 yields? 2 “ 3, that is,
? pe q ´ 6e9´8 ?
so y “ logp3` 2 2q or y “ logp3´ 2 2q “ log 3`2?2 “ log 3`21?2 “ ´ logp3` 2 2q.
?
Hence z P t2πn ˘ i logp3 ` 2 2q, n P Zu.
?
Vice versa, if z “ 2πn ˘ i logp3 ` 2 2q for some n P Z, then
?
cos z “ pcosp2πnqq coshp˘ logp3 ˘ 2 2qq ´ ipsinp2πnqqpsinh ¨ ¨ ¨ q
? ?
“ 1¨coshp˘ logp3 ˘ 2 2qq ´ i0psinh ¨ ¨ ¨ q “ coshp˘ logp3 ˘ 2 2qq
? ? ? ? ? ?
elogp3`2 2q è´ logp3`2 2q 3`2 2`p3`2 2q´1 3`2 2`3´2 2
“ 2 “ 2 “ 3. “ 2
?
Consequently, cos z “ 3 if and only if z P t2πn ˘ i logp3 ` 2 2q, n P Zu.
142 Solutions
Solution to Exercise 1.59. If cos z ` i sin z “ e, then we obtain

exppizqèxppízq
2 ` i exppizqéxppízq
2i “ e,
giving exppizq “ e. Let z “ x ` iy, where x, y P R. Taking complex absolute values
in exppizq “ e, eý “ | exppý ` ixq| “ | exppipx ` iyqq| “ | exppizq| “ |e| “ e.
So y “ ´1. Thus e exppixq “ eý exppixq “ exppý ` ixq “ exppizq “ e, that
is, exppixq “ 1. Hence cos x ` i sin x “ 1, so that cos x “ 1 and sin x “ 0. Thus
x P 2πZ. So if cos zì sin z “ e, then z P t2πní : n P Zu. Vice versa, if z “ 2πní
(n P Z), then cos z ` i sin z “ exppizq “ exppip2πn ´ iqq “ e1 exppi2πnq “ ep1 ` i0q “ e.

? ? ?
Log p1 ` iq “ Logp 2p ?12 ` i ?12 qq “ Log p 2pcos π4 ` i sin π4 qq “ log 2 ` i π4 .

Log p´1q “ Log p1pcos π ` i sin πqq “ log 1 ` iπ “ 0 ` iπ “ iπ,
Log 1 “ Log p1pcos 0 ` i sin 0qq “ log 1 ` i0 “ 0 ` i0 “ 0.
With z “ ´1, Log pz 2 q “ Log pp´1q2 q “ Log 1 “ 0, and 2 Log z “ 2 Log p´1q “ 2πi.
So when z “ ´1, Log pz 2 q “ 0 ‰ 2πi “ 2 Log z.
For z P Czt0u, Logpz n q “ n Log z ô log |z n | ` i Argpz n q “ n log |z| ` inArg z.
But log |z n | “ logp|z|n q “ n log |z|. So Logpz n q “ n Log z ô Argpz n q “ nArg z.
If Logpz n q “ n Log z, then nArg z “ Argpz n q P p´π, πs, so that Arg z P p´ πn , nπ s,
that is, z belongs to the ‘sector’ Σ :“ tz P C : ´ πn ă Arg z ď πn u. Vice versa,
if z P Σ, and θ :“ Arg z P p´ nπ , nπ s, then we have z “ |z|pcos θ ` i sin θq, so that
z n “ |z|n pcos θ ` i sin θqn “ |z n |pcospnθq ` i sinpnθqq, with nθ P p´π, πs, showing
that Argpz n q “ nθ “ nArgz, and Logpz n q “ n Log z.

For w P A “ trpcos θ ` i sin θq : 1 ă r ă e, θ P p´π, πsu, Log w “ log r ` iArg w,
and so RepLog wq “ log r P p0, 1q, while ImpLog wq P p´π, πs. So we have that the
image Log A Ă I :“ tx ` iy : 0 ă x ă 1, ´π ă y ď πu.
π
I
A Log
0 1
´π
We claim I Ă Log A. Let z P I. So x :“ Re z P p0, 1q, y :“ Im z P p´π, πs. Then

w :“ exppx ` iyq P A, since |w| “ ex P p1, eq. Also,
Log w “ Logpex pcos y ` i sin yqq “ plog ex q ` iy “ x ` iy “ z.
Hence Log A “ I.
? ?
Solution to Exercise 1.64. Logp1 ` iq “ Logp 2pcos π4 ` i sin π4 q “ log 2 ` i π4 .
Thus the principal value of p1 ` iq1í is
? ? π ?
exppp1íq Logp1ìqq “ exppp1íqplog 2 ` i π4 qq “ elog 2` 4 exppip π4 ´log 2qq
? π ? π ? ?
“ 2 e 4 p1ìq
? exppí log 2q “ e 4 p1 ` iqpcosplog 2q ´ i sinplog 2qq.
2
1í are given respectively
Thus we have ?
π
that the real and
? imaginary
π ? p1 ` iq
parts of ?
by e 4 pcosplog 2q ` sinplog 2qq and e 4 pcosplog 2q ´ sinplog 2qq.
Solution to Exercise 1.65. Since 1 “ exp 0 “ exppz ´ zq “ pexp zqpexpp´zqq,
1
expp´zq “ exp z for all z P C. We have
1 1 n
wń “ exppń Log wq “ exppnp´Log wqq “ pexpp´Log wqqn “ p exppLog n
wq q “ p w q .
For m P N, peit qm “ pcos t ` i sin tqm “ cospmtq ` i sinpmtq “ eimt .
Let m “ ń, n P N. With w :“ eit “ cos t ` i sin t, we have
1 1
w “ cos tì sin t “ cos t ´ i sin t “ cosp´tq ` i sinp´tq “ eip´tq ,
and so peit qm “ peit qń “ peip´tq qn “ einp´tq “ eimt .
Solution to Exercise 1.66. (1) Suppose first that c ‰ 0. We note that if z P C
az`b
and cz`d “ ac , then we get caz ` bc “ acz ` ad, giving ac ´ bd “ 0, which is a
contradiction. So, if z P C p and f pzq “ a , then z R C, that is, z “ 8.
c
p are such that f pzq “ f pz 1 q P Czt a u, then z, z 1 P C and az`b “ az11 `b .
1˝ If z, z 1 P C c cz`d cz `d
This yields upon simplifying that pad ´ bcqpz ´ z 1 q “ 0, and so z “ z 1 .
p are such that f pzq “ f pz 1 q “ a , then z “ z 1 “ 8.
2˝ If z, z 1 P C c
p and f pzq “ f pz 1 q “ 8, then z “ z 1 “ ´ d .
3˝ If z, z 1 P C c
Thus f is injective when c ‰ 0.
az`b ´dw`b
Next we show surjectivity. (Solving for z in w “ cz`d , we arrive at w “ cwá .)
ap ´dw`b q`b ádw`✚
ab`bcw´✚ ´pad´bcqw
For w P Czt ac u, f p ´dw`b
cwá q “
cwá
“ ´✟✟ ✟
ab
“ “ w.
cp ´dw`b
cwá
q`d cdw`bc`✟cdwád ´pad´bcq
144 Solutions
p is in the range of f ,
Also, f p8q “ ac and f p´ dc q “ 8. Hence every point of C
showing that f is surjective.
To show the inverse is given by gpwq “ ´dw`b a
cwá for w P Czt c u, with the understand-
ing that gp ac q “ 8 and gp8q “ ´d c , it is enough to show f ˝ g “ id, where id is the
identity map on C p (as then g “ id˝g “ pf ´1 ˝f q˝g “ f ´1 ˝pf ˝gq “ f ´1 ˝id “ f ´1 ).
1˝ For w P Czt ac u, we have idpwq “ w “ f p ´dw`b
cwá q “ f pgpwqq “ pf ˝ gqpwq.
2˝ For w “ ac , pf ˝ gqpwq “ f pgp ac qq “ f p8q “ a
c “ w “ idpwq.
3˝ For w “ 8, then pf ˝ gqpwq “ f pgp8qq “ f p ´d
c q“8“w “ idpwq.
This completes the proof when c ‰ 0.
Next we suppose that c “ 0. Then 0 ‰ ad ´ bc “ ad implies that a ‰ 0 and d ‰ 0.
Thus α :“ ad ‰ 0. With β :“ db , we have f pzq “ ad z ` db “ αz ` β P C for all
z P C. Also f p8q “ 8. If z, z 1 P C p are such that f pzq “ f pz 1 q P C, then we have
αz ` β “ αz ` β, and so αpz ´ z q “ 0, giving z “ z 1 (as α ‰ 0). Also, if z, z 1 P C
1 1 p
1 1
and f pzq “ f pz q “ 8, then z “ z “ 8. Thus f is injective.
pw´βq
For any w P C, we have f p w´β
α q“α α ` β “ w. Also f p8q “ 8. Hence every
p belongs to the range of f , showing that f is surjective.
point of C
Finally, we will check that the inverse is given by g, where gpwq :“ ´dw`b
0á “ w
α ´ αβ
for w P C, and gp8q :“ 8.
β pw´βq
1˝ For all w P C, pf ˝ gqpwq “ f p w
α ´ αq “ α α ` β “ w “ idpwq.
2˝ Also, pf ˝ gqp8q “ f p8q “ 8 “ idp8q.
1 1
(2) With the Möbius transformations f pzq “ az`b a z`b
cz`d and gpzq “ c1 z`d1 , we associate
“ ‰ “ 1 1‰
a b a b
the matrices F “ c d and G “ c1 d1 . Then their determinants are nonzero.
1 1
1 1 ap a1 z`b1 q`b 1 1 1
paa `bc qzàb `bd 1
We have pf ˝ gqpzq “ f p ac1 z`d
z`b
1q “
c z`d
1 z`b1 “ pca1 `dc1 qz`pcb1 `dd1 q “
Az`B
Cz`D , where
cp ac1 z`d 1 q`d
“A B‰ ” 1 1 1 1 ı “ ‰“ ‰
aa `bc ab `bd 1 1
H :“ C D :“ ca 1 `dc1 cb1 `dd1 “ ac db ac1 db 1 “ F G.
Now to show that f ˝g is a Möbius transformation, it suffices to show AD´BC ‰ 0.
But AD ´ BC “ det H “ detpF Gq “ pdet F qpdet Gq “ pad ´ bcqpa1 d1 ´ b1 c1 q ‰ 0.
(3) Let C “ tz P C : |z ´ z0 | “ ru be the circle with centre z0 P C and radius r ą 0.

For a translation f pzq “ z ` b, f pCq “ tw P C : |w ´ pz0 ` bq| “ ru, which is a
circle with centre z0 ` b and radius r ą 0.
For a dilation f pzq “ az (a ą 0), f pCq “ tw P C : |w ´ az0 | “ aru, which is a circle
with centre az0 and radius ar ą 0.
For a rotation f pzq “ eiθ z (θ P R), f pCq “ tw P C : |w ´ eiθ z0 | “ ru, which is a
circle with centre eiθ z0 and radius r ą 0.
So for translations, dilations, and rotations, the images of circles are circles.
With inversions, we will now see that the image of a circle is either a circle or a
straight line (which can be viewed as a circle with an infinite radius). If f pzq “ z1 ,
then f pCq X C “ tw P C : | w1 ´ z0 | “ ru “ tw P C : |1 ´ wz0 | “ r|w|u. So:
w P f pCq X C ô 0 “ |1 ´ wz0 |2 ´ r 2 |w|2 ô 0 “ p1 ´ wz0 qp1 ´ w z0 q ´ r 2 |w|2
ô p|z0 |2 ´ r 2 q|w|2 ´ z0 w ´ z0 w ` 1 “ 0.
1˝ |z0 | “ r ą 0. Then w P f pCqX C if and only if 2 Repz0 wq “ 1. If z0 “ px0 , y0 q P R2 ,
we have with w “ px, yq P R2 that w P f pCq if and only if x0 x ´ y0 y “ 21 . Thus
f pCq “ tpx, yq P R2 : x0 x ´ y0 y “ 21 u Y t8u, which is a line in the complex
plane. (Since |z0 | “ r ą 0, px0 , y0 q ‰ p0, 0q. Also, 0 P C, and so 8 P f pCq.)
2˝ |z0 | ą r. Let k ą 0 be such that k2 “ |z0 |2 ´ r 2 ą 0. Also, let a “ kz02 P C.
Then w P f pCq if and only if |w|2 ´ 2Repawq ` |a|2 “ p kr2 q2 . Thus (see e.g.
Exercise 1.31) f pCq “ tw P C : |w ´ a| “ kr2 u, which is a circle in the complex
plane with centre a and radius kr2 .
3˝ |z0 | ă r. Let k ą 0 be such that k2 “ r 2 ´ |z0 |2 ą 0. Also, let a “ ´ kz02 P C.
Then w P f pCq if and only if |w|2 ´ 2Repawq ` |a|2 “ p kr2 q2 . It follows that
f pCq “ tw P C : |w ´ a| “ kr2 u, which is a circle with centre a and radius kr2 .
Next we consider the case of lines. From Exercise 1.22, a line passing through
z0 , z1 , z0 ‰ z1 , is L “ ttz1 ` p1 ´ tqz0 : t P Ru. So if d :“ z1 ´ z0 ‰ 0, then
L “ tz0 ` td : t P Ru is the line passing through z0 and z0 ` d (‰ z0 ).
For a translation f pzq “ z ` b, f pLq “ tz0 ` b ` td : t P Ru, which is a line passing
through z0 ` b and z0 ` b ` d.
For a dilation f pzq “ az (a ą 0), f pLq “ taz0 ` tad : t P Ru, which is a line passing
through az0 and az0 ` ad.
For a rotation f pzq “ eiθ z (θ P R), f pLq “ teiθ z0 ` teiθ d : t P Ru, which is a line
passing through eiθ z0 and eiθ z0 ` eiθ d.
Thus in each of the cases above, the images of lines are lines again.
With inversions f pzq “ 1z , we will now see that the image of a line L is either a line
(if L passes through the origin) or a circle (if L does not pass through the origin).
1˝ If L “ ttd : t P Ru, then using f p8q “ 0 and f p0q “ 8, it can be seen that
f pL Y t8uq “ ts d1 : s P Ru Y t8u. The set ts d1 : s P Ru describes a line passing
through 0 (and is obtained by reflecting L in the real axis).
2˝ If L does not pass through the origin, then there is a nonzero complex number
z0 P L closest to the origin, so that the segment joining 0 to z0 is perpendicular
1
to L. So we may take d “ iz0 , and L “ tz0 ` tiz0 : t P Ru. For w :“ z0 p1ìtq ,
1 1 2 1 2p1ítq 1 1´t2 2t
|w ´ 2z0 | “ 2|z0 | | 1ìt ´ 1| “ 2|z0 | | 1`t2 ´ 1| “ 2|z0 | | 1`t2 ´ i 1`t2 |
1
a 1 1
“ p1 ´ t2 q2 ` 4t2 “ 2|z0 |p1`t 2
2|z0 |p1`t2 q 2 q p1 ` t q “ 2|z | .
0
1 1
So f pL Y t8uq “ tw P C : |w ´ 2z0 | “ 2|z0 | u, which is a circle passing through
the origin.
146 Solutions

Solution to Exercise 2.1. Let L1 , L2 denote the limits as z Ñ z0 of h1 , h2 ,
respectively. Let ǫ ą 0. There exists δ1 ą 0 such that whenever z P U satisfies
0 ă |z ´ z0 | ă δ1 , |h1 pzq ´ L1 | ă ǫ{2. Also, there exists δ2 ą 0 such that whenever
z P U satisfies 0 ă |z´z0 | ă δ2 , |h2 pzq´L2 | ă ǫ{2. Set δ “ mintδ1 , δ2 u ą 0. If z P U
and 0 ă |z ´z0 | ă δ, then |ph1 `h2 qpzq´pL1 `L2 q| ď |h1 pzq´L1 |`|h2 pzq´L2 | ă ǫ.
Next we consider the case of pointwise product. Let ǫ ą 0. There exists δ1 ą 0
such that whenever z P U satisfies 0 ă |z ´ z0 | ă δ1 , |h1 pzq ´ L1 | ă 2p|L2ǫ|`1q ď 2ǫ .
Also, there exists δ2 ą 0 such that whenever z P U satisfies 0 ă |z ´ z0 | ă δ2 ,
ǫ
|h2 pzq ´ L2 | ă ǫ`2|L 1|
. Set δ “ mintδ1 , δ2 u ą 0. If z P U and 0 ă |z ´ z0 | ă δ,
|ph1 h2 qpzq ´ L1 L2 | “ |h1 pzqh2 pzq ´ L2 h1 pzq ` L2 h1 pzq ´ L1 L2 |
“ |h1 pzq||h2 pzq ´ L2 | ` |L2 ||h1 pzq ´ L1 |
ď p|h1 pzq ´ L1 | ` |L1 |q|h2 pzq ´ L2 | ` |L2 ||h1 pzq ´ L1 |
ă p 2ǫ ` |L1 |q ǫ`2|L
ǫ
1|
` |L2 | 2p|L2ǫ|`1q “ ǫ.
Solution to Exercise 2.2. Rough work: Let L be the limit of f as z Ñ z0 . For

1
z, z0 P U , | f pzq ´ L1 | “ |f pzq´L|
|f pzq||L| . For z close to z0 , |f pzq| is close to |L| ą 0, and |f pzq|
stays away from zero as follows: if |f pzq ´ L| ă |L| 2 , then using the reverse triangle
inequality, |L|
2 ą |f pzq ´ L| ě ||f pzq| ´ |L|| ě |L| ´ |f pzq|, giving |f pzq| ą |L|
2 .
2 |L|
Given ǫ ą 0, ǫ1 :“ mint ǫ|L| 2 , 2 u ą 0. Let δ ą 0 be such that whenever z P U
satisfies 0 ă |z ´ z0 | ă δ, we have |f pzq ´ L| ă ǫ1 . Thus, if 0 ă |z ´ z0 | ă δ, then
we have |L| |L|
2 ą |f pzq ´ L| ě ||f pzq| ´ |L|| ě |L| ´ |f pzq|, giving |f pzq| ą 2 . So if
2
1
0 ă |z ´ z0 | ă δ, | f pzq ´ L1 | “ |f|fpzq´L| 1 2 ǫ|L| 2
pzq||L| ă ǫ |L|2 ď 2 |L|2 “ ǫ.
2 2
Solution to Exercise 2.3. For z ‰ 0, f pzq´f z´0
p0q
“ |z|z´0
´0
“ |z|z “ z. So we
guess that f 1 p0q “ 0. Given ǫ ą 0, set δ :“ ǫ ą 0. If 0 ă |z ´ 0| “ |z| ă δ, then
| f pzq´f
z´0
p0q
´0| “ |z| “ |z| ă δ “ ǫ. So f is complex differentiable at 0, and f 1 p0q “ 0.
Solution to Exercise 2.4. Let w0 P D. r Then w0 P D. As f is holomorphic on

D, given ǫ ą 0, there exists a δ ą 0 such that whenever z P D and 0 ă |z ´ w0 | ă δ,
| f pzq´f
z´w0
pw0 q
´ f 1 pw0 q| ă ǫ. p‹q
r satisfies 0 ă |w ´ w0 | ă δ, then firstly, w P D, and moreover we have that
If w P D
0 ă |w ´ w0 | “ |w ´ w0 | “ |w ´ w0 | ă δ. So
r r
| f pwq´ f pw0 q
w´w0 ´ f 1 pw0 q| “ | f pwq´f
w´w0
pw0 q
´ f 1 pw0 q| “ | f pwq´f
w´w0
pw0 q
´ f 1 pw0 q|
“ | f pwq´f
w´w0
pw0 q
´ f 1 pw0 q| ă ǫ (using p‹q).
So fr is complex differentiable at w0 and pfrq1 pw0 q “ f 1 pw0 q. As w0 P D

r was arbitrary,
fr is holomorphic in D.r
Solution to Exercise 2.5. Let z0 P Czt0u. For z, z0 P Czt0u, z ‰ z0 , we have

1 1
´ pz´z0 q |z´z0 |
| zz´zz00 ´ p´ z12 q| “ | ´ zz0 pz´z0 q ` 1
z02
| “|´ 1
zz0 ` 1
z02
| “ |z||z0 |2 . p‹q
0
For z close to z0 , the numerator is small. Also, the denominator term |z| is close
to |z0 | ą 0, and hence can be made to stay away from zero (so that 1{|z| can
be bounded), as follows: If (the to-be-chosen) δ ą 0 is such that δ ď |z0 |{2,
then whenever 0 ă |z ´ z0 | ă δ, we have (using the reverse triangle inequality)
|z0 |{2 ě δ ą |z ´ z0 | ě ||z| ´ |z0 || ě |z0 | ´ |z|, which yields |z| ą |z0 |{2. Then
|z´z0 | 2 1
the right-hand estimate in (‹) can be bounded above: |z||z 2 ă δ |z | |z |2 , and so to
0| 0 0
make this latter combination smaller than ǫ, we can further choose δ ď ǫ|z0 |3 {2. So,
3
given ǫ ą 0, setting δ :“ mint |z20 | , ǫ|z20| u, for all z P Czt0u such that 0 ă |z ´z0 | ă δ,
1 1
´ |z´z0 | ǫ|z0 |3 2
| zz´zz00 ´ p´ z12 q| “ |z||z0 |2
ă δ |z20 | |z10 |2 ď 1
2 |z0 | |z0 |2 “ ǫ.
0
Thus f 1 pz0 q “ ´ z12 .

0
Solution to Exercise 2.6. Since f is complex differentiable at z0 , there exist

r ą 0 and h : Dpz0 , rqpĂ U q Ñ C such that f pzq “ f pz0 q ` pf 1 pz0 q ` hpzqqpz ´ z0 q
for all z P Dpz0 , rq, and lim hpzq “ 0.
zÑz0
Let r 1 P p0, rq be such that for all z P D˚ pz0 , r 1 q :“ tz P C : 0 ă |z ´z0 | ă r 1 u, we have
|hpzq| ă 1. Given ǫ ą 0, set δ “ mint |f 1 pz0ǫ q|`1 , r 1 u. Then whenever 0 ă |z ´ z0 | ă δ,
we have z P Dpz0 , r 1 qpĂ Dpz0 , rq Ă U q and so
|f pzq ´ f pz0 q| “ |f 1 pz0 q ` hpzq||z ´ z0 | ď p|f 1 pz0 q| ` |hpzq|q |f 1 pz0ǫ q|`1
ă p|f 1 pz0 q| ` 1q |f 1 pz0ǫ q|`1 “ ǫ.
Solution to Exercise 2.7. As f, g : U Ñ C are complex differentiable at z0 P U ,

by Lemma 2.1 there exist r ą 0 and hf , hg : Dpz0 , rq Ñ C, such that for z P Dpz0 , rq,
f pzq “ f pz0 q ` pf 1 pz0 q ` hf pzqqpz ´ z0 q, p˚q

gpzq “ gpz0 q ` pg 1 pz0 q ` hg pzqqpz ´ z0 q, p˚˚q
and lim hf pzq “ 0 “ lim hg pzq.
zÑz0 zÑz0
(1) Adding (˚) and (˚˚), pf `gqpzq “ pf `gqpz0 q`ppf 1 pz0 q`g 1 pz0 q`hf `g pzqqqpz ´z0 q,
for z P Dpz0 , rq, where hf `g pzq :“ hf pzq ` hg pzq for all z P Dpz0 , rq. Moreover,
lim hf `g pzq “ lim phf pzq ` hg pzqq “ lim hf pzq ` lim hg pzq “ 0 ` 0 “ 0. By
zÑz0 zÑz0 zÑz0 zÑz0
Lemma 2.1, f ` g is complex differentiable and pf ` gq1 pz0 q “ f 1 pz0 q ` g 1 pz0 q.
(2) Multiplying (˚) by α, pα¨f qpzq “ pα¨f qpz0 q ` p α f 1 pz0 q ` hα¨f pzqqqpz ´ z0 q, for
z P Dpz0 , rq, where hα¨f pzq :“ α hf pzq for all z P Dpz0 , rq. Moreover, we have
lim hα¨f pzq “ lim pα hf pzqq “ α lim hf pzq “ α 0 “ 0. By Lemma 2.1, α¨f is
zÑz0 zÑz0 zÑz0
complex differentiable and pα¨f q1 pz0 q “ α f 1 pz0 q.
148 Solutions
(3) We have pf gqpzq “ pf gqpz0 q ` p f 1 pz0 qgpz0 q ` f pz0 qg 1 pz0 q ` hf g pzqqqpz ´ z0 q,

for z P Dpz0 , rq (obtained by multiplying (˚) and (˚˚)), where hf g is given by
hf g pzq :“ f pz0 qhg pzq ` gpz0 qhf pzq ` pz ´ z0 qpf 1 pz0 q ` hf pzqqpg1 pz0 q ` hg pzqq, for
z P Dpz0 , rq. Then lim hf g pzq “ f pz0 q 0`gpz0 q 0`0 pf 1 pz0 q ` 0qpg1 pz0 q`0q “ 0.
zÑz0
So f g is complex differentiable and pf gq1 pz0 q “ f 1 pz0 qgpz0 q ` f pz0 qg1 pz0 q by
Lemma 2.1.
Solution to Exercise 2.8. No. (Let OpCq be finite-dimensional with dimension
d P N. As OpCq contains the nonzero vector C Q z ÞÑ 1, we have d ‰ 0. The d ` 1
vectors 1, z, ¨ ¨ ¨ , z d P OpCq are linearly dependent. So there exist α0 , ¨ ¨ ¨ , αd P C,
not all zeroes, such that for all z P C, we have α0 ¨1 ` α1 ¨z ` ¨ ¨ ¨ ` αd ¨z d “ 0. Let
k P t0, 1, ¨ ¨ ¨ , du be the smallest index such that αk ‰ 0. Differentiating k times,
and evaluating at 0 P C, 0 ` αk ¨k! ` 0 “ 0, and so αk “ 0, a contradiction.)
Solution to Exercise 2.9. Let z0 P U . As f is complex differentiable at z0 , there
exist r ą 0 and h : Dpz0 , rqpĂ U q Ñ C such that f pzq “ f pz0 q`pf 1 pz0 q`hpzqqpz´z0 q
for all z P Dpz0 , rq, and lim hpzq “ 0. With g :“ f1 ,
zÑz0
1
gpzq“ gpz10 q ` pf 1 pz0 q ` hpzqqpz ´ z0 q,
and so gpz0 q “ gpzq ` p f 1 pz0 q ` hpzqqq gpz0 qgpzqpz ´ z0 q. Rearranging,
gpzq “ gpz0 q ` p ´ f 1 pz0 qgpz0 qgpzq ´ hpzqgpz0 qgpzqqq pz ´ z0 q
f 1 pz0 q f 1 pz0 q 1
f pz0 q
“ gpz0 q ` p ´ pf pz0 qq2
` pf pz0 qq2
´ f pz 0 qf pzq
´ f pzhpzq
0 qf pzq
q pz ´z0 q
f 1 pz0 q
“ gpz0 q ` p ´ pf pz0 qq2
` ϕpzqqq pz ´ z0 q,
f 1 pz
0q f 1 pz
hpzq0q
where ϕpzq :“ pf pz0 qq2 ´ f pz0 qf pzq ´
f pz0 qf pzq for all z P Dpz0 , rq. As f is continuous
✟
✟0 q✟ ´
1 pz ✟
at z0 and lim hpzq “ 0, lim ϕpzq “ pff✟ ✟
0q f 1 pz
´ f pz✟ 0
“ 0. Hence g is
zÑz0 zÑz0 ✟ pz0 qq2 ✟ 0 qf pz0 q f pz0 qf pz0 q
1
complex differentiable at z0 and g 1 pz0 q “ ´ pffpzpz00qqq2 .
Solution to Exercise 2.10. For m ě 0, this was shown in Example 2.3. Let
m “ ń for some n P N. Set f pzq “ z n for z P Czt0u. Consider the map
z ÞÑ z m “ z ń “ z1n “ f pzq
1
. As f P OpCzt0uq and f is pointwise nonzero, f1 is
pzq 1 n´1
holomorphic, and p f1 q1 pzq “ ´ pffpzqq nz 1 1
2 “ ´ pz n q2 “ ń z n`1 “ m z ´m`1 “ mz
m´1 .
Solution to Exercise 2.11. Define f : D Ñ C by f pzq “ ´ 1`z1´z for all z P D, and

g : C Ñ C by gpzq “ exp z, for all z P C. Then f pDq Ă C “ Ug . By the chain rule,
g ˝ f is holomorphic in D, and
1`z d
pg ˝ f q1 pzq “ g 1 pf pzqq f 1 pzq “ expp´ 1´z q dz p´ 1`z
1´z q
“ expp´ 1`z p d 1
1´z qp ´ p1 ` zq dz p 1´z q ´
1 d
1´z dz p1 ` zqqq
p1`zq
“ expp´ 1`z p
1´z qp ´ p1´zq2
´ 1
1´z q “ 2
´ p1´zq2 expp´ 1`z
1´z q.
d 1`z 2 1`z
Thus dz expp´ 1´z q “ ´ p1´zq2 expp´ 1´z q in D.
Solution to Exercise 2.12. With z “ x ` iy, where x, y P R, |z|2 “ x2 ` y 2 . So

if u, v denote the real part and the imaginary part of |z|2 , then u “ x2 ` y 2 and
v “ 0. So Bu Bv Bu Bv
Bx “ 2x, By “ 0, By “ 2y, Bx “ 0. Since z ‰ 0, it follows that x or y
is nonzero, and so at least one of the Cauchy-Riemann equations is violated, that
is, Bu Bv Bu Bv 2
Bx “ 2x ‰ 0 “ By or By “ 2y ‰ 0 “ ´ By . So |z| is not differentiable at any
nonzero complex number.
Solution to Exercise
a 2.13. 1˝ Let y ą 0. If u :“ RepLogq and v :“ ImpLogq
x
then upx, yq “ log x2 ` y 2 “ 12 logpx2 ` y 2 q, and vpx, yq “ cos´1 a 2 2 . So
x `y
1
0´x a 2y
Bu 1 1 x Bv 1 2 x2 ` y 2 x
“ 2x “ 2 , “ ´b “ 2 ,
Bx 2 x2 ` y 2 x ` y2 By 1´ x2 x2 ` y 2 x ` y2
x2 `y 2
a x
1 x2 ` y 2 ´ x a
Bu 1 1 y Bv 1 x2 ` y 2 y
“ 2y “ 2 , “ ´b “´ 2 .
By 2 x2 ` y 2 x ` y2 Bx 1´ x2 x2 ` y 2 x ` y2
x2 `y 2
These partials are continuous on R ˆ p0, 8q, and the Cauchy-Riemann equations
y
are satisfied. Thus Log is holomorphic here, and Log1 z “ x2 `y
x z 1
2 ´ i x2 `y 2 “ zz “ z ,
where z “ px, yq P R ˆ p0, 8q.

1
2˝ Let x ą 0. If u :“ RepLogq and v :“ ImpLogq then upx, yq “ 2 logpx2 ` y 2 q, and
y
vpx, yq “ sin´1 a 2 2 . So
x `y x
0ýa
Bu x Bv 1 x2 ` y 2 y
“ 2 , “b “´ 2 ,
Bx x ` y2 Bx 2
1 ´ x2y`y2 x2 ` y 2 x ` y2
a y
1 x2 ` y 2 ´ y a
Bu y Bv 1 x2 ` y 2 x
“ 2 , “b “ 2 .
By x ` y2 By 2
1 ´ x2y`y2 x2 ` y 2 x ` y2
These partials are continuous on p0, 8q ˆ R, and the Cauchy-Riemann equations

y
x z 1
2 ´ i x2 `y 2 “ zz “ z ,
where where z “ px, yq P p0, 8q ˆ R.
3˝ Let y ă 0. If u :“ RepLogq and v :“aImpLogq then upx, yq “ 12 logpx2 ` y 2 q, and

x
vpx, yq “ ´ cos´1 a 2 2 . Note that y 2 “ ý since y ă 0. We have
x `y y
0´xa
Bu x Bv 1 x2 ` y 2 x
“ 2 , “b “ 2 ,
Bx x ` y2 By 2
1 ´ x2x`y2 x2 ` y 2 x ` y2
a x
1 x2 ` y 2 ´ x a
Bu y Bv 1 x2 ` y 2 y
“ 2 , “b “´ 2 .
By x ` y2 Bx 2
1 ´ x2x`y2 x2 ` y 2 x ` y2
These partials are continuous on R ˆ p´8, 0q, and the Cauchy-Riemann equations
y
x z 1
2 ´ i x2 `y 2 “ zz “ z ,
where z “ px, yq P R ˆ p´8, 0q.

150 Solutions
Solution to Exercise 2.14. Let w0 :“ f pz0 q, w P V ztw0 u, and z :“ gpwq. Then

z0 “ gpf pz0 qq “ gpw0 q. Also, z ‰ z0 (otherwise, w “ f pgpwqq “ f pzq “ f pz0 q “ w0 ).
For all w P V ztw0 u,
idV pwqídV pw0 q f pgpwqq´f pgpw0 qq f pgpwqq´f pgpw0 qq gpwq´gpw0 q
1“ w´w0 “ w´w0 “ gpwq´gpw0 q w´w0 .
By the continuity of g, lim gpwq “ gpw0 q. As f is complex differentiable at z0 ,
wÑw0
f pgpwqq´f pgpw0 qq
lim gpwq´gpw0 q “ f 1 pgpw0 qq “ f 1 pz0 q ‰ 0.
wÑw0
0q
Hence by Exercise 2.2, lim gpwq´gpw
w´w0 “ lim 1
f pgpwqq´f pgpw0 qq “ 1
f 1 pz0 q .
wÑw0 wÑw0 gpwq´gpw0 q
With U “ R ˆ p´π, πq, V “ Czp´8, 0s, f “ exp and g “ Log, f, g are continuous,
and f ˝g “ idV , g˝f “ idU . As f P OpU q, and for all z P U , f 1 pzq “ exp z ‰ 0, we have
for all w P V and with z “ gpwq, that Log1 w “ g1 pwq “ f 11pzq “ exp1 1 1
z “ exppLog wq “ w .
Solution to Exercise 2.15. Let z “ x ` iy, where x, y P R. Then we have

that z 3 “ px ` iyq3 “ x3 ` 3x2 piyq ` 3xpiyq2 ` piyq3 “ x3 ´ 3xy 2 ` ip3x2 y ´ y 3 q.
If u, v denote the real and imaginary parts of z 3 , then upx, yq “ x3 ´ 3xy 2 and
vpx, yq “ 3x2 y ´ y 3 . We have Bu 2 2 Bv Bu
Bx “ 3x ´ 3y “ By , and By “ ´6xy “ ´ Bx ,
Bv
2 2
which are continuous on R . Hence u, v are real differentiable in R . Also, the
Cauchy-Riemann equations are satisfied in R2 . Thus z ÞÑ z 3 is entire.
Solution to Exercise 2.16. If z “ xìy, where x, y P R, then Re z “ Repxìyq “ x.
The real and imaginary parts u, v of Re z are upx, yq “ x and vpx, yq “ 0. Thus
Bu Bv 2
Bx “ 1 ‰ 0 “ By for all px, yq P R . Hence the Cauchy-Riemann equations are
2
satisfied nowhere in R . So at each point of C, Re z is not complex differentiable.
Solution to Exercise 2.17. Let u, v be the real and imaginary parts of f . Then
v “ 0 everywhere in D. Thus Bu Bv Bu Bv
Bx “ By “ 0 and By “ ´ Bx “ 0 everywhere in D. If
the line segment joining px, y0 q P D to px0 , y0 q P D lies in D, then there exists a ξ
qúpx0 ,y0 q
in the open interval with endpoints x, x0 such that upx,y0x´x 0
“ Bu
Bx pξ, y0 q “ 0.
Also, whenever the line segment joining px0 , y0 q P D to px0 , yq P D lies in D, then
upx0 ,yqúpx0 ,y0 q
yý0 “ Bu
By px0 , ηq “ 0 for some η in the open interval with endpoints y, y0 .
So u is constant along horizontal and vertical line segments lying in D. As D is
path-connected, u is constant in D (as any two points in D can be joined by a
stepwise path lying in D). So f “ u ` i0 “ u is constant in D.
f 1 pzq “ Bu Bv Bu Bv
Bx ` i Bx “ 0 gives Bx “ Bx “ 0 in D, and using the Cauchy-Riemann
Bu
equations, also that By p“ ´ Bx q “ 0 and Bv
Bv Bu
By p“ Bx q “ 0 everywhere in D. If the
line segment joining px, y0 q P D to px0 , y0 q P D lies in D, then there exists a ξ in
qúpx0 ,y0 q
the open interval with endpoints x, x0 such that upx,y0x´x 0
“ Bu
Bx pξ, y0 q “ 0.
Also, whenever the line segment joining px0 , y0 q P D to px0 , yq P D lies in D, then
upx0 ,yqúpx0 ,y0 q
yý0 “ Bu
By px0 , ηq “ 0 for some η in the open interval with endpoints y, y0 .
So u is constant along horizontal and vertical line segments lying inside D. As D
is path-connected, u is constant in D (because any two points in D can be joined

by a stepwise path). Similarly v is constant in D. So f “ u ` iv is constant in D.
Solution to Exercise 2.19. By the chain rule, Bu 1 Bv

Bx px, yq “ h pvpx, yqq Bx px, yq and
Bu 1 Bv
By px, yq “ h pvpx, yqq By px, yq. Using these and the Cauchy-Riemann equations,
Bu Bv
By px, yq “ h1 pvpx, yqq By px, yq “ h1 pvpx, yqq Bu
Bx px, yq
` Bv
˘ Bv
1 1
“ h pvpx, yqq h pvpx, yqq Bx px, yq “ ph1 pvpx, yqqq2 Bx px, yq
“ ´ph1 pvpx, yqqq2 Bu
By px, yq,
and so p1` ph1 pvpx, yqqq2 q Bu 1 2 Bu

By px, yq “ 0. As p1` ph pvpx, yqqq q ě 1 ą 0, By px, yq “ 0.
Bv
Using the Cauchy-Riemann equations, we also obtain Bx px, yq “ ´ Bu
By px, yq “ 0.
Bu 1 Bv 1
Thus Bx px, yq “ h pvpx, yqq Bx px, yq “ h pvpx, yqq0 “ 0, and using the Cauchy-
Bv
Riemann equations again, By px, yq “ Bu
Bx px, yq “ 0. So u is constant along horizontal
and along vertical line segments. Since every z P C can be joined to the origin
using a combination of a horizontal and a vertical line segment, u is constant in C.
Similarly v is constant in C. So f “ u ` iv is constant in C.
upx, yq “ x2 ´ y 2 and vpx, yq “ kxy for all px, yq P R2 .
(‘Only if’ part) If f is entire, then the Cauchy-Riemann equations are satisfied in
R2 , and so 2x “ Bu Bv
Bx “ By “ kx for all x, y P R. Taking x “ 1 and any y P R, k “ 2.
(‘If’ part) If k “ 2, then f pzq “ x2 ý 2 `2xyi “ x2 `piyq2 `2xpiyq “ pxìyq2 “ z 2 .
We have seen in Example 2.1 that z ÞÑ z 2 is entire.

zn
qθ
pn ´ 1
z z ÞÑ zn z0n
ndθ
dθ z0 nθ
θ “ Arg z0
0 0
|z ´ z0 | “ |z0 | tan dθ « |z0 |dθ, while |z n ´ z0n | “ |z0 |n tanpndθq « |z0 |n ndθ, and so
|z n ´z n | n ndθ
the magnification produced locally by z ÞÑ z n is |z´z00| « |z|z0 |0 |dθ “ n|z0 |n´1 . Also,
π π
from the picture, the anticlockwise local rotation is 2 ´p 2 ´ pn´1qθq “ pn´1qθ. So
f 1 pz0 q “ n|z0 |n´1 pcosppn´1qθqì sinppn´1qθqq “ np|z0 |pcos θì sin θqqn´1 “ nz0n´1 .
d n
Consequently, dz z “ nz n´1 in C.
152 Solutions
Solution to Exercise 2.22. Let z P Czt0u. For simplicity, we suppose z is in the

first quadrant and |z| ą 1. Displace z radially outwards through a small distance
to obtain zr. Let Arg z “ θ “ Arg zr. Then z “ |z|pcos θ ` i sin θq, and so
1 1 1 1
z “ |z|pcos θì sin θq “ |z| pcos θ ´ i sin θq “ |z| pcosp´θq ` i sinp´θqq.
Thus Arg 1z “ ´θ “ Arg 1zr . We note that | 1zr | ă | 1z |. So the local anticlockwise
rotation produced is π´2θ.
zr zr
z
z
dθ
θ
0 ´θ
π´2θ
1 0 1
dθ 1
1 z
zr 1
z 1
zr
Next, displace z0 counterclockwise along a circular arc with centre 0 and radius
|z0 | through an angle dθ to obtain zr. Then it is clear that z1r is obtained from 1z
by moving clockwise through an angle of dθ along a circular arc with radius | z1 |
1
centred at 0. So the local magnification is |z| dθ{p|z|dθq “ |z|1 2 .
d 1 1 ´1
Consequently, dz p z q “ |z|2
pcospπ´2θq ` i sinpπ´2θqq “ |z|2 pcos θì sin θq2
“ ´ z12 .

z exp z exp z
ipy`δq
δ exp z0 y
exp z0 z0 z
iy z0 exp
δ iy
δ exp
y y
ex δ
From the left picture above, the magnification is δ “ ex . From the right picture,
the anticlockwise local rotation produced is y. Thus the complex derivative at z0
must be ex pcos y ` i sin yq “ exppx ` iyq. Consequently, exp1 z “ exp z for all z P C.
Solution to Exercise 2.24. Let w P Czp´8, 0s. Displace w radially outwards

r If Arg w “ θ “ Arg w,
through a distance δ to obtain w. r then Log wr is a horizontal
δ
rightward displacement of Log w (through a distance logp1 ` |w| q). So the local
anticlockwise rotation produced is 2π´θ.
Log w1
w1 r
w
dθ
θ
w Log Log w
r
dθ
Log w
θ
0 0
Next, displace w counterclockwise along a circular arc with centre 0 and radius |w|
through an angle dθ to obtain w1 . Then it is clear that Log w1 is obtained from
Log w by a vertical displacement through a distance dθ. So the local magnification
dθ
is |w|dθ 1
“ |w| . Thus Log1 w “ |w|
1 1
pcosp2π´θq ` i sinp2π´θqq “ |w|pcos θì 1
sin θq “ w .
Solution to Exercise 2.25. Let z0 P C. Move z0 along the line with slope 1
through a distance of δ to get the point z. Similarly move z0 horizontally to the
left by a distance δ to get the point zr. Suppose that the map ζ ÞÑ Re ζ is complex
differentiable at z0 . In the following picture, by considering the images of z, zr
under the mapping Re ¨, we get conflicting values of the local rotation produced as
being 45˝ and 0˝ , i.e., π4 “ ArgpRe1 z0 q “ 0, a contradiction. So the map cannot
be complex differentiable at z0 . As the choice of z0 P C was arbitrary, the map is
complex differentiable nowhere.
z
δ
zr
δ z0 Re ¨
No Clockwise
rotation 45˝ rotation
Re zr Re z0 Re z
Solution to Exercise 2.26. We have | exp1 paì0q|2 “ | exppaì0q|2 “ |ea |2 “ e2a .

Thus we do have that lim ρpǫq “ e2a “ | exp1 pa ` i0q|2 . This is expected, since for
ǫÑ0
small ǫ, the image under exp of the square region Sǫ with a side length 2ǫ centred
at a will be approximately a square region with a side length 2ǫ| exp1 pa ` 0iq|, and
4ǫ2 | exp1 paì0q|2
so the ratio of the areas is 4ǫ2 “ | exp1 pa ` i0q|2 “ |ea |2 “ e2a .
2
Solution to Exercise 2.27. a Let 0 :“ p0, 0q. For x “ px, yq P R zt0u, its distance
to the origin 0 is }x}2 “ x2 ` y 2 “: r ą 0. Denote by α the angle made by
the ray joining 0 to x with the positive real axis in the counterclockwise direction.
Then x “ r cos α and y “ r sin α. Thus
“ θ ´ sin θ ‰ r cos α ” ı ” ı
rpcos θ cos α´sin θ sin αq r cospθ`αq
Rx “ cossin θ cos θ r r sin α s “ rpsin θ cos α`cos θ sin αq “ r sinpθ`αq .
Hence the angle made by the ray joining 0 to Rx with the positive real axis in the
counterclockwise direction is θ ` α, and the distance of Rx to 0 is
a
}Rx}2 “ r pcospθ ` αqq2 ` psinpθ ` αqq2 “ r “ }x}2 (the distance of x to 0).
Rx
x
r
θ
α
0
Thus Rx is obtained by rotating x about the origin 0 counterclockwise through θ.

154 Solutions

(1) We have
Bf Bu Bv 1 Bu Bu 1 Bv Bv
Bz pz0 q “ Bz pz0 q ` i Bz pz0 q “ 2 p Bx pz0 q ` i By pz0 qq ` i 2 p Bx pz0 q ` i By pz0 qq
1 Bu Bv 1 Bu Bv
“ 2 p Bx pz0 q ´ By pz0 qq ` i 2 p By pz0 q ` Bx pz0 qq.
So Bf
Bz pz0 q “ 0 if and only if the Cauchy-Riemann equations hold at z0 . As it
is given that u and v are real differentiable at z0 , Theorems 2.5 and 2.4 imply
that f is complex differentiable at z0 if and only if Bf
Bz pz0 q “ 0.
(2) We have
Bf Bu Bv 1 Bu Bu 1 Bv Bv
Bz pz0 q “ Bz pz0 qì Bz pz0 q “ 2 p Bx pz0 qí By pz0 qqì 2 p Bx pz0 qí By pz0 qq
1 Bu Bv 1 Bu Bv Bu Bv 1
“ 2 p Bx pz0 q` By pz0 qqì 2 p´ By pz0 q` Bx pz0 qq “ Bx pz0 qì Bx pz0 q “ f pz0 q.
(3) For the map R2 Q px, yq “ z ÞÑ z, we have

1 Bpýq
Bz
Bz “ 21 p Bx Bx
Bx ` i By q ` i 2 p Bx ` i Bpýq
By q“ 1
2 ´ i 21 i “ 1 ‰ 0
everywhere in R2 . (So z is complex differentiable nowhere in C.)
For the map R2 Q px, yq “ z ÞÑ |z|2 “ zz “ x2 ` y 2 , we have
B|z|2 Bpzzq Bpx2 `y 2 q 2 `y 2 q 2 `y 2 q
Bz “ Bz “ Bz “ 12 p Bpx
Bx ` i Bpx
By q “ 21 p2x ` i2yq “ x ` iy “ z.
So |z|2 is complex differentiable only at 0.
(4) R2 Q px, yq “ z ÞÑ f pzq “ z 2 “ x2 ´ y 2 ` ip2xyq is entire, because
1 Bpx2 ý 2 q 2 ý 2 q
Bz 2
Bz “ 2 p Bx ` i Bpx
By q ` i 21 p Bp2xyq
Bx ` i Bp2xyq
By q
1 1
“ 2 p2x ´ i2yq ` i 2 p2y ` i2xq “ 0.
Moreover,
Bz 2 1 Bpx2 ý 2 q Bpx2 ý 2 q
Bz “ 2 p Bx ´ i By q ` i 12 p Bp2xyq
Bx ´ i Bp2xyq
By q
1 1
“ 2 p2x ` i2yq ` i 2 p2y ´ i2xq “ 2px ` iyq “ 2z.

Solution to Exercise 3.1. For t P r0, 2πs, we have γ1 ptq “ exppitq “ cos tì sin t,
γ2 ptq “ expp2itq “ cosp2tq ` i sinp2tq, γ3 ptq “ exppítq “ cos t ´ i sin t, and in each
case, pRe γk ptqq2 ` pIm γk ptqq2 “ 1, k “ 1, 2, 3, t P r0, 2πs. So the range of γk is
contained in the unit circle T with centre 0. Also, if z “ exppiθq with θ P r0, 2πq,
then z “ γ1 pθq “ γ2 pθ{2q “ γ3 p2π ´ θq, and so every z P T belongs to the range of
each of γ1 , γ2 , γ3 . We have γ11 ptq “ ´ sin tì cos t “ ipcos t ` i sin tq “ i exppitq, and
similarly, γ21 ptq “ 2i expp2itq, γ31 ptq “ í exppítq for all t P r0, 2πs. Thus
ş 1 ş2π 1
γ1 z dz “ 0 exppitq i exppitq dt “ 2πi,
ş 1 ş2π 1
γ2 z dz “ 0 expp2itq 2i expp2itq dt “ 4πi,
ş 1 ş2π 1
γ3 z dz “ 0 exppítq píq exppítq dt “ ´2πi.
Solution to Exercise 3.2. Let γptq “ xptq ` iyptq, t P ra, bs, where x, y are
real-valued. Also, let u, v be the real and imaginary parts of f , respectively. Then
f 1 pγptqqγ 1 ptq
“ p Bu Bv q 1 1
Bx pxptq, yptqq ` i Bx pxptq, yptqqq px ptq ` iy ptqq
Bu 1 Bv 1 p Bu 1 Bv 1 q
“ Bx pxptq, yptqqx ptq ´ Bx pxptq, yptqqy ptq ` ip Bx pxptq, yptqqy ptq ` Bx pxptq, yptqqx ptqq
Bu 1 Bu 1 p Bv 1 Bv 1 q
“ Bx pxptq, yptqqx ptq ` By pxptq, yptqqy ptq ` ip By pxptq, yptqqy ptq ` Bx pxptq, yptqqx ptqq
(using the Cauchy-Riemann equations)
d d
“ dt upxptq, yptqq ` i dt vpxptq, yptqq (using the Chain Rule)
d q d
“ dt p upxptq, yptqq ` ivpxptq, yptqqq “ dt f pγptqq.
Solutionş to Exercise 3.3. With ş u :“ Reşf , v :“ Im f , andş x :“ Re γ, y :“ Im γ,

1 1
we have 0 pux1 ´vy 1 q dt “ Re p γ f pzq dzq “ γ Repf pzqq dz “ 0 upx1 ìy 1 q dt, giving,
ş1 1
ş 1 1
0 ´vy dt “ i 0 uy dt, and this will not hold in general. To construct a concrete
counterexample,
ş take f ” i andşγptq “ it, t P r0, 1s. Then f is entire, γ is smooth,
and Re p γ f pzq dzq “ ´1 ‰ 0 “ γ Re f pzq dz.

ş ş2π ş 2π
‚ γ pz ` zqdz “ 0 p2eit ` 2eít q2ieit dt “ 4i 0 pe2it ` 1qdt “ 4i 0 ` 4i 2π “ 8πi.
ş ş2π ş2π
‚ γ pz 2 ´2z`3qdz “ 0 p4e2it ´4eit `3q2ieit dt “ 0 ip8e3it ´8e2it `6eit qdt “ 0.
‚ The 2π-periodicity of psinp2¨qq cos is used to get the first equality in the last line.
ş ş2π ş 2π
γ xy dz “ 0 2pcos tq2psin tq 2ipcos tì sin tqdt “ 4i 0 psinp2tqqpcos tì sin tqdt
şπ ş2π
“ 4i ´π psinp2tqq cos t dt ´ 2 0 pcos t ´ cosp3tqqdt “ 0.
looooooomooooooon
odd function
156 Solutions

ş ş1 1ì
‚ γptq “ p1 ` iqt, t P r0, 1s. So γ Re z dz “ 0 tp1 ` iqdt “ 2 .
‚ γptq “ i ` exppitq, t P r´ π2 , 0s.
So
ş ş0 ş0 2
p q
γ Re z dz “ ´ π2 pcos tqi exppitqdt “ ´ π2 ipcos tq ´ pcos tqpsin tqq dt
ş0
“ ´ π pi cosp2tq`1
2 ´ sinp2tq 1 π 1 1 π
2 qdt “ 0 ` i 2 2 ` 2 “ 2 ` i 4 .
2
ş ş1 1
‚ γptq “ t ` it2 , t P r0, 1s. So γ Re z dz “ 0 tp1 ` 2itq dt “ 2 ` 2i 13 “ 1
2 ` 32 i.
n ` ˘
ř n ℓ
Solution to Exercise 3.6. By the Binomial Theorem, p1 ` zqn “ ℓ z .
ℓ“0
n ` ˘
ř n ℓ´k´1
n
For 0 ď k ď n, and z P ran C, we have p1`zq
z k`1
“ ℓ z . Consequently,
ℓ“0
1
ş p1`zqn
1
ş ` ˘
ř n ℓ´k´1
n `
ř n 1
n ˘ ş `
ℓ´k´1 dz “ n .
˘
2πi C z k`1 dz “ 2πi C ℓ z dz “ ℓ 2πi C z k
ℓ“0 ℓ“0

(1) Let uf , vf , ug , vg : ra, bs Ñ R be such that f pγptqqγ 1 ptq “ uf ptq ` ivf ptq and
gpγptqqγ 1 ptq “ ug ptq ` ivg ptq for all t P ra, bs. Then
ş şb 1
şb 1 1 q
γ pf ` gqpzq dz “ a pf ` gqpγptqqγ ptq dt “ a p f pγptqqγ ptq ` gpγptqqγ ptqq dt
şb şb
“ a puf ptq ` ug ptqq dt ` i a pvf ptq ` vg ptqq dt
şb şb şb şb
“ a uf ptqdt ` a ug ptqdt ` i a vf ptqdt ` i a vg ptqdt
ş ş
“ γ f pzq dz ` γ gpzq dz.
(2) Let α “ p ` iq where p, q P R, and let u, v : ra, bs Ñ R be such that for all
t P ra, bs, f pγptqqγ 1 ptq “ uptq ` ivptq. Then
ş şb
γ pα¨f qpzq dz “ a pp ` iqqpuptq ` ivptqq dt
şb şb
“ a pp uptq ´ q vptqq dt ` i a pp vptq ` q uptqq dt
şb şb şb şb
“ ppp a uptq dt ` i a vptq dtqq ` iq p a uptq dt ` i a vptq dtqq
şb şb ş
“ pp ` iqqpp a uptq dt ` i a vptq dtqq “ α γ f pzq dz.
Solution to Exercise 3.8. ´p´γq : ra, bs Ñ C is given by

p´p´γqqptq “ p´γqpa ` b ´ tq “ γpa ` b ´ pa ` b ´ tqq “ γptq, t P ra, bs,
and so ´p´γq “ γ. (This is also obvious visually.)
Solution to Exercise 3.9. We do have that γpbq “ p´γqpaq, and so the two paths
γ and ´γ can be concatenated, and we have
ş ş ş ş ş
γ`p´γq f pzq dz “ γ f pzq dz ` ´γ f pzq dz “ γ f pzq dz ´ γ f pzq dz “ 0.
Solution to Exercise 3.10. Let z “ α ` iβ, α, β P R, and ϕptq “ xptq ` iyptq,

where x, y : ra, bs Ñ R. Then
şb şb
a z ϕptq dt “ a pα ` iβqpxptq ` iyptqq dt
şb
“ a p α xptq ´ βyptq ` ipα yptq ` βxptqqqq dt
şb şb
“ a pα xptq ´ βyptqq dt ` i a pα yptq ` βxptqq dt
şb şb şb şb
“ α a xptq dt ´ β a yptq dt ` ipp α a yptq dt ` β a xptq dtqq
şb şb şb
“ pα ` iβqp a xptq dt ` i a yptq dtq “ z a ϕptq dt.
ş1 ? ? γ : r0, 1s Ñ C by γptq “ p1 ` iqt for t P r0, 1s.
3.11. Define
The length of γ is 0 12 ` 12 dt “ 2. (Also evident by Pythagoras’s Theorem).
1ì
γ
1
0 1
Also, |pγptqq2 | “ |t ` it|2 “ 2t2 , and so max |pγptqq2 | “ 2p12 q “ 2. Hence

tPr0,1s
ş 2 2
?
| γ z dz| ď p max |pγptqq |q plength of γq “ 2 2.
tPr0,1s
ş ş1 ş1 ş ? 3 ?
p1ìq3
As γ z 2 dz “ 0 pt ` itq2 p1ìqdt “ 0 p1` iq3 t2 dt “ 3 , | γ z 2 dz| “ p 32q “ 2 3 2 .

`2n˘ `2n˘ ş p1`zq2n p1`zq2n 2n
1
n “ | n | “ | 2πi C z n`1 dz| ď
1
2π p max | z n`1 | q 2π 1 “ max |1`z|
1
|z|“1 |z|“1
ď p1 ` 1q2n “ 22n “ 4n .
Solution to Exercise 3.13. Set γptq “ p1´tqi`t, t P r0, 1s. By the ML inequality,
ş ?
| γ z14 dz| ď plength of γq max | pγptqq
1
4| “ 2 max pp1´tq12 `t2 q2 .
t Pr0,1s t Pr0,1s
2
But for a, b P R, a2 ` b2 ě pa`bq
2 (since pa ´ bq2 ě 0). So with a “ p1 ´ tq, b “ t,
2 ş ? ?
we obtain p1 ´ tq2 ` t2 ě p1´t`tq
2 “ 21 . So | γ z14 dz| ď 2 max pp1´tq12 `t2 q2 ď 4 2.
t Pr0,1s
Solution to Exercise 3.14. Write z “ x ` iy, where x, y P R. Let f “ u ` iv

be a primitive of the map z ÞÑ z in C, where u :“ Re f and v :“ Im f . Then
Bu Bv Bv Bu
Bx ` i Bx “ By ´ i By “ f 1 “ z “ x ´ iy in R2 . Fix x0 P R. Then for px, yq P R2 ,
şx Bv şx
vpx, yq´vpx0 , yq “ x0 Bx pξ, yq dξ “ x0 ´ y dξ “ ´ xy`x0 y. So vpx, yq “ ´ xy`ϕpyq,
Bv
where ϕpyq :“ vpx0 , yq ` x0 y. Hence x “ By “ ´ x ` ϕ1 pyq, that is, ϕ1 pyq “ 2x for
all x P R, which is impossible (otherwise, e.g., 2 “ 2p1q “ ϕ1 pyq “ 2p0q “ 0).
158 Solutions
Solution to Exercise 3.15. The continuous map D Q ζ ÞÑ f pζqg1 pζq ` f 1 pζqgpζq

has a primitive f g P OpDq, since pf gq1 “ f g1 ` f 1 g. By the Fundamental Theorem
of Contour Integration,
ş 1 1
ş 1
q
γ p f pζqg pζq ` f pζqgpζqq dζ “ γ pf gq pζq dζ “ f pzqgpzq ´ f pwqgpwq,
and a rearrangement proves the claim.
ş iz eiz ´r
ş 2 ´z 2 eiz
For r ą |a|, integrating by parts, Sr zze z
2 à2 dz “ z 2 à2 i |r ´ Sr pza2 à 2 q2 i dz.
From Exercise 1.37, there exists an r ą 0 and constants C1 , C2 such that for all
z C1 a2 ´z 2 C2
z P C such that |z| ě r, | z 2 à2 | ď |z| and | pz 2 à2 q2 | ď |z|2 . For z P ran Sr ,
Im z ě 0, and so |eiz | “ eÍm z ď e0 “ 1. It follows from these estimates that
z eiz ´r
ş a2 ´z 2 eiz C2
lim | z 2 à 2 i |r | “ 0. Using the M L-inequality | S pz 2 à2 q2 i dz| ď r 2 πr, and
r
rÑ8
ş 2 ´z 2 eiz ş zeiz
so lim | Sr pza2 à 2 q2 i dz| “ 0. Consequently, lim | S z 2 à2 dz| “ 0.
r
rÑ8 rÑ8
Solution to Exercise 3.16. cos is entire, and in particular continuous. Also cos
has a primitive in C: sin1 z “ cos z. Thus by the Fundamental Theorem of Contour
ş ´1
Integration γ cos z dz “ sin i ´ sinpíq “ 2 exppiiqéxppíiq
1
2i “ e ié “ pe ´ 1e qi.
Solution to Exercise 3.17. By the Fundamental Theorem of Contour ş Integra-

tion, if γ is the straight line path joining z, w P D, then f pzq´f pwq “ γ f 1 pζq dζ. By
ş
the ML-inequality, |f pzq ´ f pwq| “ | γ f 1 pζq dζ| ď sup |f 1 pζq|plength of γq ď 1 |z ´ w|.
ζPran γ
Solution to Exercise 3.18. Let D :“ Czpp´8, 0s ˆ t0uq and γ : r0, 1s Ñ D be

the straight line path joining 1 to 1 ´ i, given by γptq “ p1 ´ tq1 ` tp1 ´ iq “ 1 ´ it
for all t P r0, 1s. As Log P OpDq and Log1 z “ 1z (which is continuous on D), it
follows from the Fundamental Theorem of Contour Integration that
ş1 1 1
ş 1 ş 1
0 1ít dt “ í γ z dz “ i γ Log z dz “ ipLogp1 ´ iq ´ Log 1q
?
“ ipplog 2q ´ i π4 ´ 0q “ π4 ` i 21 log 2.
Solution to Exercise 3.19. Since exp1 z “ exp z in C, we have

ş a a a
γ exp z dz “ exppaìbq ´ exp 0 “ e pcos b ` i sin bq ´ 1 “ e cos b ´ 1 ` ie sin b,
for a path γ joining 0 to a ` ib. If γpxq “ pa ` ibqx for all x P r0, 1s, then
ş ş1 ş1 ax
γ exp z dz “ 0 expppa ` ibqxq pa ` ibq dx “ 0 e pcospbxq ` i sinpbxqqpa ` ibq dx.
ş ş1
Hence pa ´ ibq γ exp z dz “ 0 eax pcospbxq ` i sinpbxqqpa2 ` b2 q dx. Thus
ş1 ş
pa2 ` b2 q 0 eax cospbxq dx “ Repppa ´ ibq γ exp z dz q
“ Repppa ´ ibqpea cos b ´ 1 ` iea sin bqqq
“ apea cos b ´ 1q ` bea sin b.
ş1 apea cos b´1q`bea sin b
Hence 0 eax cospbxqdx “ a2 `b2 .
Solution to Exercise 3.20. Let C be the circular path with centre at 0 and radius
r ą 0 traversed in the anticlockwise direction, i.e., Cpθq “ reiθ for all θ P r0, 2πs.
By the Fundamental Theorem of Contour Integration,
ş ş2π ş2π
0 “ C exp z dz “ 0 er cos θìr sin θ rieiθ dθ “ 0 er cos θ rieipr sin θ`θq dθ.
ş2π
Equating real and imaginary parts, in particular, 0 er cos θ cospr sin θ ` θq dθ “ 0.
Solution to Exercise 3.21. Suppose f P OpCzt0uq is such that f 1 “ z1 in Czt0u.
Let C be a circular path with a positive radius centred at 0 traversed in the
anticlockwise direction.
ş By the Fundamental
ş Theorem
ş of Contour Integration,
since C is closed, C f 1 pzq dz “ 0. But C f 1 pzq dz “ C z1 dz “ 2πi, a contradiction.
Solution to Exercise 3.22. E, C are both closed, since Ep0q “ a “ Ep1q and
Cp0q “ r “ Cp1q. Define H : r0, 1sˆr0, 1s Ñ Czt0u by Hpt, τ q “ p1´τ qCptq`τ Eptq,
for all t, τ P r0, 1s. To see that |Hpt, τ q| ‰ 0, setting m “ minta, b, ru ą 0,
|Hpt, τ q|2 “ pp1´τ qr`τ aq2 pcosp2πtqq2 `pp1´τ qr`τ bq2 psinp2πtqq2
ě pp1´τ qm`τ mq2 pcosp2πtqq2 `pp1´τ qm`τ mq2 psinp2πtqq2 “ m2 ą 0.
Then C is Czt0u-homotopic to E, since H is continuous, and moreover:
Hpt, 0q “ Cptq for all t P r0, 1s,
Hpt, 1q “ Eptq for all t P r0, 1s,
Hp0, τ q “ p1´τ qCp0q`τ Ep0q “ p1´τ qCp1q`τ Ep1q “ Hp1, τ q for all τ P r0, 1s.

(ER1) Let γ : r0, 1s Ñ D be a closed path. Define H : r0, 1s ˆ r0, 1s Ñ D by
Hpt, τ q “ γptq, t, τ P r0, 1s. Then H is continuous, and
Hpt, 0q “ γptq for all t P r0, 1s,
Hpt, 1q “ γptq for all t P r0, 1s,
Hp0, τ q “ γp0q “ γp1q “ Hp1, τ q for all τ P r0, 1s.
Hence γ is D-homotopic to itself. So the relation is reflexive.
(ER2) Let γ0 , γ1 : r0, 1s Ñ D be closed paths such that γ0 is D-homotopic to γ1 .
Then there exists a continuous H : r0, 1s ˆ r0, 1s Ñ D such that
Hpt, 0q “ γ0 ptq for all t P r0, 1s,
Hpt, 1q “ γ1 ptq for all t P r0, 1s,
Hp0, τ q “ Hp1, τ q for all τ P r0, 1s.
r : r0, 1s ˆ r0, 1s Ñ D be defined by Hpt,
Let H r τ q “ Hpt, 1 ´ τ q for all t, τ P r0, 1s.
r
Then H is continuous and
r 0q “ Hpt, 1q “ γ1 ptq for all t P r0, 1s,
Hpt,
r 1q “ Hpt, 0q “ γ0 ptq for all t P r0, 1s,
Hpt,
r τ q “ Hp0, 1 ´ τ q “ Hp1, 1 ´ τ q “ Hp1,
Hp0, r τ q for all τ P r0, 1s.
Thus γ1 is D-homotopic to γ0 . Hence the relation is symmetric.
160 Solutions
(ER3) Let γ0 , γ1 , γ2 be closed paths such that γ0 is D-homotopic to γ1 , and γ1 is

D-homotopic to γ2 . There are continuous H, K : r0, 1s ˆ r0, 1s Ñ D such that
Hpt, 0q “ γ0 ptq for all t P r0, 1s, Kpt, 0q “ γ1 ptq for all t P r0, 1s,
Hpt, 1q “ γ1 ptq for all t P r0, 1s, Kpt, 1q “ γ2 ptq for all t P r0, 1s,
Hp0, τ q “ Hp1, τ q for all τ P r0, 1s, Kp0, τ q “ Kp1, τ q for all τ P r0, 1s.
" *
Hpt, 2τ q if τ P r0, 21 s
Define L : r0, 1s ˆ r0, 1s Ñ D by Lpt, τ q “ .
Kpt, 2pτ ´ 21 qq if τ P p 12 , 1s
Then Lpt, 0q “ Hpt, 0q “ γ0 ptq and Lpt, 1q “ Kpt, 1q “ γ2 ptq for all t P r0, 1s. Also,
for 0 ď τ ď 21 , Lp0, τ q “ Hp0, 2τ q “ Hp1, 2τ q “ Lp1, τ q, and for 21 ă τ ď 1,
Lp0, τ q “ Kp0, 2pτ ´ 21 qq “ Kp1, 2pτ ´ 21 qq “ Lp1, τ q. For the continuity of L, it is
enough to show ‘left continuity in the τ -variable’ at τ “ 12 , as H, K are continuous.
If pptn , τn qqnPN is a sequence in r0, 1s ˆ p 21 , 1s that converges to pt0 , 12 q, then
lim Lptn , τn q “ lim Kptn , 2pτn ´ 12 qq “ Kpt0 , 0q “ γ1 pt0 q “ Hpt0 , 1q “ Lpt0 , 12 q “ Lp lim ptn , τn qq.
nÑ8 nÑ8 nÑ8
So L is continuous, and γ0 is D-homotopic to γ2 . Thus the relation is transitive.
As D-homotopy is reflexive, symmetric and transitive, it is an equivalence relation.
Solution to Exercise 3.24. The ‘point’ path γ : r0, 1s Ñ Czt0u, γptq :“ ´2,
r are Czt0u-homotopic: If H : r0, 1s ˆ r0, 1s Ñ Czt0u
t P r0, 1s, is closed. Also, γ, C
r ` τ p´2q “ ´2 ` p1 ´ τ qe2πit , τ, t P r0, 1s, then
is defined by Hpt, τ q “ p1 ´ τ qCptq
r
|Hpt, τ q ` 2| ď 1, so that |Hpt, τ q| ě 1 ą 0, and for all t, τ P r0, 1s, Hpt, 0q “ Cptq,
Hpt, 1q “ ´2 “ γptq and Hp0, τ q “ Hp1, τ q. Also, H is continuous.
If C, Cr are Czt0u-homotopic, then by the transitivity of Czt0u-homotopy, C, γ
are Czt0u-homotopic, and the Cauchy Integral Theorem with f :“ z1 P OpCzt0uq
ş ş ş1 1
implies that 2πi “ C z1 dz “ γ z1 dz “ 0 ´2 0 dt “ 0, a contradiction.
Solution to Exercise 3.25. For a circular path C : r0, 1s Ñ Czt0u ş with centre
0 and radius r ą 0, traversed once in the anticlockwise direction, C z1 dz “ 2πi.
Define E r : r0, 1s Ñ C by Eptq r “ a cosp2πtq ` ib sinp2πtq for all t P r0, 1s. The
r
elliptic path E is Czt0u-homotopic to C, as shown in Exercise r is a
ş 3.22,
1
and
ş E
1
reparamterisation of E. So by the Cauchy Integral Theorem, Er z dz “ C z dz “
2πi, and so
ş ş ş2π
2πi “ Er z1 dz “ E z1 dz “ 0 a cos θìb 1
sin θ pá sin θ ` ib cos θq dθ
ş2π pá sin θìb cos θqpa cos θíb sin θq
“ 0 a2 pcos θq2 `b2 psin θq2
dθ
ş2π pb2 á2 qpcos θqpsin θqìabppcos θq2 `psin θq2 q
“ 0 a2 pcos θq2 `b2 psin θq2
dθ
ş2π pb2 á2 qpcos θqpsin θqìabp1q
“ 0 a2 pcos θq2 `b2 psin θq2 dθ.
ş2π 1 2π
Equating the imaginary parts, 0 a2 pcos θq2 `b2 psin θq2 dθ “ ab .
Solution to Exercise 3.26. C is Czt0u-homotopic to S.

´1ì γ2 1ì
C S
0
γ3
γ1
´1í γ4 1í
For t P r0, 1s, define

γ1 ptq :“ p1 ´ tqp1 ´ iq ` tp1 ` iq “ 1 ` ip2t ´ 1q,
γ2 ptq :“ p1 ´ tqp1 ` iq ` tp´1 ` iq “ p1 ´ 2tq ` i,
γ3 ptq :“ p1 ´ tqp´1 ` iq ` tp´1 ´ iq “ ´1 ` ip1 ´ 2tq,
γ4 ptq :“ p1 ´ tqp´1 ´ iq ` tp1 ´ iq “ 2t ´ 1 ´ i.
ş ş ş ş ş
Then S 1z dz “ γ1 1z dz ` γ2 1z dz ` γ3 z1 dz ` γ4 1z dz. We have
ş 1 ş1 2i
ş1 2ip1íp2t´1qq ş1 1
ş 1 2t´1
γ1 z dz “ 0 1ìp2t´1q dt “ 0 1`p2t´1q2 dt “ 2i 0 1`p2t´1q2 dt`2 0 1`p2t´1q2 dt
u“2t´1 ş 1 1
ş1 u ´1 1´tan´1 p´1qq`0 “ ip π ´p´ π qq “ i π .
“ i ´1 1ù 2 du` ´1 1ù2 du “ iptan 4 4 2
ş1 2 π
ş1 2t´1
Noting that 0 1`p2t´1q2 dt “ 2 and 0 1`p2t´1q2 dt “ 0, we obtain
ş 1 ş1 ´2 ş1 ´2p´p2t´1qíq
γ2 z dz “ 0 1´2tì dt “ 0 1`p2t´1q2
dt “ 0 ` p´1qpíq π2 “ i π2 ,
ş 1 ş1 ´2i
ş1 ´2ip´1ìp2t´1qq
γ3 z dz “ 0 ´1ìp1´2tq dt “ 0 1`p2t´1q2 dt “ íp´1q π2 ` 0 “ i π2 ,
ş 1 ş1 2 ş1 2pp2t´1qìq π π
γ4 z dz “ 0 2t´1í dt “ 0 1`p2t´1q2 dt “ 0 ` i 2 “ i 2 .
ş
Thus S 1z dz “ 4i π2 “ 2πi, as expected.
4i
r C
C
ş
(1) Logp¨ ´ 4iq P OpCztr`4i : r ď 0uq. So C Logpz ´ 4iq dz “ 0.
ş
r is the circle with centre 1 and any radius r ą 0, then r
(2) If C 1
dz “ 2πi.
C z´1
As 1 r are Czt1u-homotopic, it follows from the
P OpCzt1uq, and since C, C
z´1 ş 1 ş 1
Cauchy Integral Theorem that C z´1 dz “ Cr z´1 dz “ 2πi.
162 Solutions
(3) iz´3 “ expppz´3qLog iq “ expppz´3qplog 1 ` i π2 qq “ expppz´3qpi π2 qq exppi π2 pz´3qq,

ş “ z´3
and so z ÞÑ iz´3 is entire. By the Cauchy Integral Theorem, C i dz “ 0.

ş 1 ş1 1 ş
2πit dt “ 1 2πi dt “ 2πi,
γ0 z dz “ 0 e2πit 2πie 0
ş 1 ş1 1 ´2πit
ş1
γ1 z dz “ 0 e´2πit p´2πie q dt “ ´ 0 2πi dt “ ´2πi.
If γ0 is Czt0u-homotopic to γş1 , then asş 1z P OpCzt0uq, it follows by the Cauchy

Integral Theorem that 2πi “ γ0 z1 dz “ γ1 z1 dz “ ´2πi, a contradiction. Thus γ0
is not Czt0u-homotopic to γ1 .
şt 1 psq3.29. şt γ 1 psq γ 1 ptq
(1) ϕ1 ptq “ pexp 0 γγpsq dsq ddt 0 1
γpsq ds “ ϕptq γptq . Hence ϕ γ ´ ϕγ “
1 0, and1
ϕ1 γ´ϕγ 1
d ϕ
dt γ “ “ 0.
γ2 “ 0
γ2 Thus ϕp0q ϕp1q
γp0q “ γp1q . But γp0q “ γp1q (since γ is closed).
ş0 1 psq
So ϕp1q “ ϕp0q “ exp 0 γγpsq ds “ exp 0 “ 1. Hence wpγq P Z.
ş 1
1 1 Γ1 ptq
ş
1 1 2πi expp2πitq 1
(2) We have wpΓ1 q “ 2πi 0 Γ1 ptq dt “ 2πi 0 expp2πitq dt “ 2πi 2πi “ 1.
(3) If for all t P r0, 1s, γ1 ptq ‰ 0 and γ2 ptq ‰ 0, then also pγ1 ¨ γ2 qptq ‰ 0 for
all t P r0, 1s, and so γ1 ¨ γ2 does not pass through the origin. Also, γ1 ¨ γ2 is
closed since pγ1 ¨ γ2 qp0q “ γ1 p0qγ2 p0q “ γ1 p1qγ2 p1q “ pγ1 ¨ γ2 qp1q. We have
pγ1 ¨γ2 q1 ptq “ γ11 ptqγ2 ptq ` γ1 ptqγ21 ptq, t P r0, 1s. So
ş
1 1 pγ1¨γ2 q1 ptq
ş 1 1
1 1 γ1 ptqγ2 ptq`γ1 ptqγ2 ptq
wpγ1 ¨γ2 q “ 2πi 0 pγ1¨γ2 qptq dt “ 2πi 0 γ1 ptqγ2 ptq dt
ş 1
1 1 γ1 ptq✟ ✟
γ2 ptq ş
1 1✟ ✟ 1
γ1 ptqγ2 ptq
“ 2πi ✟ dt ` 2πi 0 γ ptqγ
0 γ ptqγ ptq ✟ ptq dt “ wpγ1 q ` wpγ2 q.
1 ✟2 ✟1 2
(4) Γm “ Γ1 ¨ ¨ ¨ Γ1 (m times). So wpΓm q “ wpΓ1 q ` ¨ ¨ ¨ ` wpΓ1 q “ m wpΓ1 q “ m 1 “ m.
(5) Define the map ϕ : r0, 1s Ñ R by ϕptq “ |γ0 ptq|, t P r0, 1s, giving the distance
of γ0 ptq from 0. Being a real-valued continuous function on a compact interval,
it has a minimum value d0 , and d0 ą 0 since γ0 does not pass through 0.
Take δ “ d0 {2 ą 0. Suppose that γ is a smooth closed path such that we have
}γ ´ γ0 }8 :“ maxtPr0,1s |γptq ´ γ0 ptq| ă δ. We will show γ is Czt0u-homotopic
to γ0 . Define H : r0, 1s ˆ r0, 1s Ñ Czt0u by Hpt, τ q “ p1 ´ τ qγ0 ptq ` τ γptq,
t, τ P r0, 1s. Then H is continuous. Also, Hpt, 0q “ γ0 ptq and Hpt, 1q “ γptq for
all t P r0, 1s, and Hp0, τ q “ p1´τ qγ0 p0q`τ γp0q “ p1´τ qγ0 p1q`τ γp1q “ Hp1, τ q
for all τ P r0, 1s. Finally, Hpt, τ q is never 0, because being a convex combination
1Here we use the product and quotient rules for differentiation of complex-valued maps of a real variable.
These follow easily by mimicking the analogous proofs in the context of real-valued functions. Firstly, if
ϕ : I Ñ C is differentiable on an interval I, then ϕ1 ptq :“ pRe ϕq1 ptq ` ipIm ϕq1 ptq “ lim ϕpt`hq´ϕptq
h
. If ϕptq ‰ 0
hÑ0
1
d 1 1 ϕ ptq
(t P I), then dt ϕ
“ lim p 1 ´ 1
q “ ´ pϕptqq 2 . Similarly, the product rule can be shown.
hÑ0 h ϕpt`hq ϕptq
of γ0 ptq and γptq, if p1 ´ τ qγ0 ptq ` τ γptq “ 0 for some t, τ , then we arrive at a
contradiction: 1 d20 ą τ |γ0 ptq´ γptq| “ |γ0 ptq´ pp1´ τ qγ0 ptq` τ γptqq| “ |γ0 ptq| ě d0 .
ě d0
γptq γ0 ptq
0
d0
ă 2
Using the Cauchy Integral Theorem, we get

ş
1 1 γ 1 ptq 1
ş 1 1
ş 1 ş 1
1 1 γ0 ptq
wpγq “ 2πi 0 γptq dt “ 2πi γ z dz “ 2πi γ0 z dz “ 2πi 0 γ0 ptq dt “ wpγ0 q.
Solution to Exercise 3.30. The length of the arc on the big circle subtended by
=Q1 OQ is tpnrq. As the small coin rolls without slipping, the angle made by O1 P
with OO 1 is ptnrq{r “ n t.
P1
O1 πńt
r P
nt
Q1
O nr Q
As O1 ” pnr ` rqeit ,
and since O1 P
is obtained from O1 P 1 “ reit by a clockwise
rotation through π ńt, P ” pnr`rqeit èípπńtq reit “ pn`1qreit `p´1q r epn`1qit .
The area enclosed by the epicycloid γ is
1
ş ş
1 2π it pn`1qit qrppn ` 1qieit ´ pn ` 1qiepn`1qit q dt
2i γ z dz “ 2i 0 rppn ` 1qe ´ e
ş
1 2π 2 ít ´ e´pn`1qit qpn ` 1qipeit ´ epn`1qit q dt
“ 2i 0 r ppn ` 1qe
2 ş 2π
“ pn`1qr
2 0 ppn ` 1q ´ pn ` 1qe
int ´ eínt ` 1q dt
pn`1qr 2
“ 2 ppn ` 1q2π ` 0 ` 0 ` 2πq “ πr 2 pn ` 1qpn ` 2q.
Solution to Exercise 3.31. For example, let f pzq “ z1 for z P D :“ Czt0u. Then
f does not have a primitive in D. (See Example 3.7 and Exercise 3.21.)
Solution to Exercise 3.32. Let Cptq “ eit for all t P r0, 2πs. Then
ş i
ş2π i it
ş2π éit
C pzáqpaz´1q dz “ 0 peit áqpaeit ´1q ie dt “ 0 peit áqpaéít qeit dt
ş2π 1
ş2π 1
“ 0 peit áqpe ít áq dt “ 0 |eit á|2
dt
ş2π ş 2π
“ 0 ppcos tqáq12 `psin tq2 dt “ 0 1´2a cos 1
tà2 dt.
164 Solutions
i
Since z ÞÑ is holomorphic in a disc containing the unit circle C (as 0 ă a ă 1),
az´1
1
ş az´1
i
i i
by the Cauchy Integral Formula, 2πi γ zá dz “ az´1 |z“a “ a2 ´1 .
ş2π 1
ş i
i 2π
So 0 1´2a cos tà2
dt “ γ az´1
zá dz “ 2πi a2 ´1 “ 1á2 .
Solution to Exercise 3.33. Suppose F is a primitive. Consider the closed circu-

lar path γ given by |z ´ 0| “ 12 traversed
ş once anticlockwise.
ş 1 By the Fundamental
1
Theorem of Contour Integration, γ zp1´z 2 q dz “ γ F pzq dz “ 0, since γ is closed.
1
On the other hand, by the Cauchy Integral Formula, as 1´z 2 P OpDp0, 1qq, we get
ş 1
ş 1´z2
1
1
γ zp1´z 2 q dz “ γ z´0 dz “ 2πi 1´z 2 |z“0 “ 2πi,
1
a contradiction. So zp1´z 2 q does not have a primitive in tz P C : 0 ă |z| ă 1u.
Solution to Exercise 3.34. As |z| ă 1, and ζ ÞÑ f pζqp1 ´ ζzq is entire, by the

Cauchy integral formula, we obtain
1
ş 1´ζz
p1´|z|2 qf pzq “ p1 ´ zzqf pzq “ f pζqp1´ζzq|ζ“z “ 2πi C f pζq ζ´z dζ.
Since |z| ă 1,
1
ş 1´ζz ş2π iθ 1éiθ z iθ
p1´|z|2 q|f pzq| “ |p1´|z|2 qf pzq| “ | 2πi C f pζq ζ´z dζ| “ | 0 f pe q eiθ ´z ie dθ|
ş
1 2π iθ 1éiθ z iθ
ď 2π 0 |f pe q eiθ ´z ie |dθ.
The result now follows by observing that for θ P r0, 2πs,
iθ peíθ ´zq| iθ
ie | “ |f peiθ q| |e ´z|
|ieiθ | “ |f peiθ q| 1|e|eiθ ´z|
iθ
|f peiθ q 1é
eiθ ´z
z iθ
|eiθ ´z|
1 “ |f peiθ q|.
´1 1 ´1 1 ´1 1 ´1 1 3 1 2
p2q p3q p4q p5q p1q
ş exp z
(1) γ z´1 dz “ 2πi exp z|z“1 “ 2πi exp 1 “ 2πie.
ş ş z 2 `1
z 2 `1 `1 `1 2 2
(2) γ z 2 ´1 dz “ dz “ 2πi zz`1
z`1
γ z´1 |z“1 “ 2πi 11`1 “ 2πi.
ş z 2 `1 2 ? ?
(3) γ z 2 ´1 dz “ 0, since e.g. zz 2 `1
´1
P OpDpi, 2qq, Dpi, 2q is simply connected,
?
and γ is a closed path in Dpi, 2q.
2 `1
ş 2 ş zz´1 z 2 `1 p´1q2 `1
(4) γ zz 2 `1
´1
dz “ γ z´p´1q dz “ 2πi z´1 |z“´1 “ 2πi ´1´1 “ ´2πi.
ş z 2 `1
ş z 2 `1 1 1
ş z 2 `1 ş z 2 `1
(5) γ z 2 ´1 dz “ γ 2 p z´1 ´ z`1 q dz “ γ z´1 dz ´ γ z´p´1q dz
2 2
2 2
“ 2πi z 2`1 |z“1 ´2πi z 2`1 |z“´1 “ 2πip1q ´ 2πip1q “ 0.

iR
S
i r´R, Rs
´R R
í
(1) By the Cauchy Integral Formula, we have

ş ş exppizq ş exppizq
zì exppizq exppi iq e´1 π
σ F pzqdz “ 2
σ z `1 dz “ σ zí dz “ 2πi zì |z“i “ 2πi iì “ 2πi 2i “ e .
(2) Let z “ xìy, with x, y P R, y ě 0. Then | exppizq| “ | exppýìxq| “ eý ď 1.

Thus |F pzq| “ | exppizq|
|z 2 `1|
ď |z 21`1| . But |z 2 | ´ | ´ 1| ď |z 2 ´ p´1q| “ |z 2 ` 1|, and
1 1
so if |z| ą 1, then |F pzq| ď |z 2 `1|
ď
Finally, for |z| ą 1, we have that
|z|2 ´1
.
1 2
?
ď
|z|2 ´1
if|z|2
ď |z|2 2|z|2
´ 2, i.e., 2
|z | ě 2,
i.e., |z| ě 2.
? ş ş
(3) For R ě 2, | S F pzq dz| ď 2πR max |F pzq| ď 2πR R22 . So lim S F pzqdz “ 0.
ş ş şzPS ş RÑ8
As r´R,Rs F pzqdz “ σ F pzqdz ´ S F pzqdz “ πe ´ S F pzqdz, we obtain that
ş ş
lim r´R,Rs F pzqdz “ πe ´ lim S F pzqdz “ πe ´ 0 “ πe .
RÑ8 RÑ8
(4) Using the parameterisation x ÞÑ x for the straight line path r´R, Rs, we have
ş ş R exppixq şR cos x şR sin x şR cos x
r´R,Rs F pzq dz “ ´R x2 `1 1 dx “ ´R x2 `1 dx ` i ´R x2 `1 dx “ ´R x2 `1 dx ` 0,
where we have used the fact that xsin
2
x
is an odd function to get the last equality.
şR cos x ş `1
Hence lim ´R x2 `1 dx “ lim S F pzqdz “ πe .
RÑ8 RÑ8
Solution to Exercise 3.37. Let Cpθq ş :“ exppiθq for all θ P r0, 2πs. As exp P OpCq,
1 exp z
by the Cauchy Integral Formula, 2πi C z´0 dz “ exp z|z“0 “ exp 0 “ 1. But
ş exp z ş2π exppexppiθqq ş2π
C z´0 dz “ 0 exppiθq i exppiθq dθ “
0 i exppexppiθqq dθ
ş2π ş2π cos θ
“ i exppcos θ ` i sin θq dθ “ 0 ie
0 pcospsin θq ` i sinpsin θqq dθ
ş2π cos θ ş2π cos θ
“ ´ 0 e sinpsin θq dθ ` i 0 e cospsin θq dθ
ş2π
Hence 0 ecos θ cospsin θq dθ “ 2π.
Solution to Exercise 3.38. If f is holomorphic, then f pnq is holomorphic too, and

so is its derivative f pn`1q . But f pn`1q , being complex differentiable, is in particular
continuous. So f pnq has a continuous complex derivative.
166 Solutions
Solution to Exercise 3.39. For n P N, let C n pU q denote the set of all real-valued
functions on U possessing continuous partial derivatives of order ď n on U . We use
induction to show that u P C n pU q for all n P N, hence proving the claim. Consider
the following sequence of statements Sn , n P N:
Sn : For every holomorphic function on U , its real part belongs to C n pU q.
We show that S1 is true. Let f P OpU q. We want to show that u :“ Re f P C 1 pU q.
As f “ u ` iv P OpU q, f 1 P OpU q. Also, f 1 “ Bu Bv Bu Bu
Bx ` i Bx “ Bx ´ i By . It follows from
1
Theorem 2.4 that for the holomorphic function f , its real and imaginary parts are
real differentiable, and in particular, continuous, everywhere in U . So Bu Bu
Bx , By are
continuous on U . Thus u P C 1 pU q. This shows S1 is true.
Suppose Sn is true. Let f P OpU q. We want to show that u :“ Re f P C n`1 pU q
in order to complete the induction step. As f 1 P OpU q and f 1 “ Bu Bu
Bx ´ i By , it follows
from the validity of Sn that Bu 1 n Bu 1
Bx “ Repf q P C pU q, and By “ Repif q P C pU q.
n
n
We already know that u “ Re f P C pU q. To show that u P C n`1 pU q, we need to
show that the partial derivatives of order n ` 1 of u are continuous. But any such
partial derivative will be an nth order partial derivative of Bu Bu
Bx or of By , which we
have established is continuous. Thus u P C n`1 pU q. Hence Sn`1 is true, completing
the induction step. By the induction principle, Sn is true for all n P N.
So if f P OpU q, then u :“ Re f P C 8 pU q. Also, v :“ Im f “ Repíf q P C 8 pU q.
Solution to Exercise 3.40. Since for all z P C, |f pzq| ě δ ą 0, in particular

f pzq ‰ 0 for all z P C, and so f1 is entire. For all z P C, | f pzq
1
| ď 1δ .
By Liouville’s Theorem, f1 is constant, say C (‰ 0, as f is pointwise nonzero).
Thus f ” C1 . Hence f is a constant.
Solution to Exercise 3.41. Let g be defined by gpzq :“ f pzq ´ w0 for z P C.

Then g is entire and |gpzq| ě r ą 0 for all z P C (Why?). By Exercise 3.40, g is
constant. So f “ g ` w0 must be constant too.
Solution to Exercise 3.42. Let K :“ tpx, yq P R2 : 0 ď x ď T1 , 0 ď y ď T2 u.

Then K is compact. (Why?) By the Weierstrass Theorem, the continuous function
K Q px, yq ÞÑ |f px ` iyq| P R assumes a maximum value M on K. Given any
x, y P R “ YnPZ rnT1 , pn ` 1qT1 q “ YmPZ rmT1 , pm ` 1qT1 q, there exist integers n, m
such that x ` iy “ x0 ` nT1 ` ipy0 ` mT2 q for some x0 P r0, T1 q and y0 P r0, T2 q. By
the periodicity of f , f px ` iyq “ f px0 ` nT1 ` ipy0 ` mT2 qq “ f px0 ` iy0 q P f pKq.
So for all x, y P R, |f px ` iyq| ď M . Hence f is bounded on C, and by Liouville’s
Theorem, must be constant.
The exponential function is 2π-periodic the imaginary direction (since we have
exppz ` 2πiq “ pexp zqpexpp2πiqq “ pexp zqp1q “ exp z, but exp is not constant.
By Exercise 1.56, sinpz ` 2πq “ psin zqpcosp2πqq ` psinp2πqqpcos zq “ sin z, and so
sin is 2π-periodic the real direction, but not constant.

(1) Define g by gpzq “ pexpp´zqqf pzq for all z P C. Then g is entire. Moreover,
since |f pzq| ď | exp z|, by rearranging, we get |gpzq| “ |pexpp´zqqf pzq| ď 1
for all z P C. So by Liouville’s Theorem, g is constant with value say, c,
i.e., gpzq “ pexpp´zqqf pzq “ c, giving f pzq “ c exp z for all z P C. Also, since
|gpzq| ď 1, we obtain |c| ď 1.
(2) By Exercise 1.37, if p is a polynomial of degree d ě 1, then there exist m, R ą 0
such that |ppzq| ě m|z|d for all |z| ą R. Thus for real z “ x ă ´R ă 0, we
have |z| “ ´x ą R, and so m|x|d ď |ppxq| ď |ex | “ ex ď 1 (because x ă 0).
Hence |x|d ď 1{m for all x ă ´R, a contradiction. So p is a constant, say equal
to c0 . But then |ppzq| ď | exp z| again gives with z “ x ă 0 that |c0 | ď |ex | “ ex
for all x ă 0, and so |c0 | “ 0, i.e., c0 “ 0. Consequently, p “ c0 “ 0.
Solution to Exercise 3.44. For all z P C, |f pzq| ą |f 1 pzq| ě 0, and so f is never
1
0. Thus g :“ ff is well-defined, and is entire. Also from the given inequality, we
see that |g| ă 1 everywhere in C, that is, g is bounded. By Liouville’s theorem, g
must be a constant, say C. Hence f 1 “ Cf in C.
We have peĆz f q1 “ ĆeĆz f ` eĆz f 1 “ eĆz pf 1 ´ Cf q “ eĆz 0 “ 0. As C is
domain and eĆz f is entire, it follows that it must be a constant there, say k P C.
This k must be nonzero, since we know that f is nowhere 0 in C. So f pzq “ keCz
with k ‰ 0. But |f 1 | ă |f | gives |kCeCz | ă |keCz |, showing that |C| ă 1. So f
must have the form f pzq “ keCz , with k P Czt0u and C P D :“ tz P C : |z| ă 1u.
Vice versa, if f has this form, then f is entire, and moreover, as |f | “ |keCz | ą 0,
we have for all z P C that |f 1 pzq| “ |kCeCz | “ |C||keCz | ă 1 |f pzq| “ |f pzq|.
Solution to Exercise 3.45. The function ef is entire, being the composition of
the entire functions exp and f . We have |ef pzq | “ eRe f pzq ď eM . By Liouville’s
Theorem, ef is a constant function. Thus pef q1 ” 0. By the Chain Rule, we now
obtain 0 “ pef q1 “ ef f 1 . As ef pzq ‰ 0 for all z P C, it follows from the above that
f 1 pzq “ 0. As f 1 ” 0, we conclude (e.g. by C-R equations or the Fundamental
Theorem of Contour Integration) that f is a constant function.
Solution to Exercise 3.46. We have that if “ ipu ` ivq “ ´v ` iu, and so
g :“ f ìf “ pu´vqìpu`vq. But g is entire too, and so is eg (being the composition
of the entire functions exp and g). We have |egpzq | “ eRe gpzq “ eu´v ď e0 “ 1. By
Liouville’s Theorem, eg is a constant function. Thus peg q1 ” 0. By the Chain Rule,
we now obtain 0 “ peg q1 “ eg g1 . As egpzq ‰ 0 for all z P C, it follows from the above
that g1 pzq “ 0. But g1 “ p1 ` iqf 1 . So f 1 ” 0.
(1) For z P C, |z ´ a1 | ě |z| ´ |a1 | “ R ´ |a1 | and |z ´ a2 | ě |z| ´ |a2 | “ R ´ |a2 |.
Thus by the M L-inequality, with M :“ maxzPC |f pzq|,
ş pzq
| C pzáf1 qpzá2q
dz| ď maxzPC |zá|f1pzq| M
||zá2 | 2πR ď pR´|a1 |qpR´|a2 |q 2πR.
168 Solutions
1 1 zá2 ´pzá1 q a1 á2
(2) For z P C, zá1 ´ zá2 “ pzá1 qpzá2 q “ pzá1 qpzá2 q . As a1 ‰ a2 , we obtain
1 1 1 1 1
pzá1 qpzá2 q “ a1 á2 p zá1 ´ zá2 q. Hence α :“ ´β :“ a1 á2 .
ş f pzq ş 1 f pzq f pzq 1
ş f pzq ş f pzq
(3) C pzá1 qpzá2 q dz “ C a1 á2 p zá1 ´ zá2 qdz “ a1 á2 p C zá1 dz ´ C zá2 dzq.
Consider a small disc ∆1 with centre a1 and radius r1 ą 0 with boundary C1 .

f p¨q
Then C and C1 are Czta1 u-homotopic, and g :“ ¨á P OpCzta1 uq. Thus by
ş f pzq ş f pzq 1
the Cauchy Integral Theorem, C zá1 dz “ C1 zá1 dz. By the Cauchy Integral
1
ş f pzq ş f pzq
Formula, 2πi dz “ f pa1 q. Thus C zá dz “ 2πif pa1 q. Similarly, we
ş f pzq C1 zá1 ş 1
pzq
have C zá2 dz “ 2πif pa2 q. Consequently, C pzáf1 qpzá 2q
dz “ 2πipf pa1 q´f pa2 qq
a1 á2 .
(4) Let f be entire and bounded. Let M ą 0 be such that |f pzq| ď M for all z P C.
Let a1 , a2 be any two distinct points in C. Let Cptq “ Reit , t P r0, 2πs, with
R ą 0 large enough so that a1 , a2 lie in the interior of C. By the above,
á2 | ş
|f pa1 q ´ f pa2 q| “ |a12π
á2 | 2πipf pa1 q´f pa2 qq
| a1 á2 | “ |a12π pzq
| C pzáf1 qpzá2q
dz|
|a1 á2 | M
ď 2π pR´|a1 |qpR´|a2 |q 2πR.
As R can be made as large as we please and since pR´|a2πRM 1 |qpR´|a2 |q

Ñ 0 as R Ñ 8,
it follows that |f pa1 q ´ f pa2 q| “ 0, and so f pa1 q “ f pa2 q. Hence f is constant.
Solution to Exercise 3.48. (‘If part’) Let ppαq “ 0 and p1 pαq ‰ 0. So α is a zero
of p. Suppose that α is not a simple zero of p. Then ppzq “ pz ´ αqm qpzq, where
the integer m ą 1, and q is a polynomial. Thus p1 “ mpz ´ αqm´1 q ` pz ´ αqm q 1 ,
and so p1 pαq “ 0, a contradiction. So α is a simple zero of p.
(‘Only if part’) Let α be a simple zero of p. Then ppαq “ 0 and ppzq “ pz ´ αqqpzq,
where q is a polynomial such that qpαq ‰ 0. Thus p1 “ q ` pz ´ αqq 1 , and so
p1 pαq “ qpαq ` 0 “ qpαq ‰ 0.
If the degree of p is 1, then ppzq “ az ` b, where a ‰ 0, and so taking w “ 0, we
have that pr “ p ` w “ p ` 0 “ p has the only zero at ´ ab , which is a simple zero.
Next suppose that the degree of p is ą 1. Then p1 has degree d ´ 1 ě 1. By the
Fundamental Theorem of Algebra and the Division Algorithm, p1 has d ´ 1 zeroes
(possibly with repetition), say β1 , ¨ ¨ ¨ , βd´1 . Let w P Czt´ppβ1 q, ¨ ¨ ¨ , ´ppβd´1 qu.
We claim pr`w has only simple zeroes. Indeed, if α is a zero of pr, then prpαq “ 0, i.e.,
ppαq ` w “ 0, i.e., ppαq “ ´w R tppβ1 q, ¨ ¨ ¨ , ppβd´1 qu. Hence α R tβ1 , ¨ ¨ ¨ , βd´1 u,
and so pr 1 pαq “ p1 pαq ‰ 0. By the previous part of the exercise, we conclude that
α is a simple zero of pr. Consequently, all the zeroes of pr are simple.
Solution to Exercise 3.49. As g :“ f ´ ez sin z is entire and bounded, by Liou-

ville’s Theorem, g is a constant, say c. In particular, c “ f p0q´ e0 sin 0 “ 0´ 0 “ 0.
Thus g ” c “ 0, and so f pzq “ ez sin z for all z P C.
Solution to Exercise 3.50. ş By Morera’s Theorem, it is enough to show that for

every rectangular path R, R f pzqş dz “ 0. If the rectangular path R lies entirely
within D :“ CzpR ˆ t0uq, then R f pzq dz “ 0 since f P OpDq. We will show
the result for rectangles with a side on the real axis. From here the result follows
also for rectangles going across the real axis (since it can be decomposed into two
rectangles, each of which has a common side lying on the real axis). Consider a
rectangle in the closed upper half-plane with the bottom side s as ra, bs Ă R, and
sptq :“ t, t P ra, bs. (The lower half-plane case is dealt analogously.) Let h be the
height of the rectangle R. Displacing s upwards through a distance nh , we get the
straight line path sn , sn ptq “ t ` ih
n , t P ra, bs. Call the rectangle
ş obtained from R
by this side displacement as Rn . Thus Rn Ă D, and so Rn f pzq dz “ 0. As f is
continuous, it is uniformly continuous on compact sets. Thus
ş şb ih nÑ8 ş b ş
sn f pzq dz “ a f pt` n q 1 dt ÝÑ a f ptq 1 dt “ s f pzq dz.
Also the contributions of the small vertical segments va,n , vb,n at a, b, each of height
h
n converge to 0 as n Ñ 8: For e.g.,
ş ş1 iht ih
ş
1 1 iht nÑ8 ş1
va,n f pzq dz “ 0 f pa` n q n dt “ n 0 f pa` n qih dt ÝÑ 0 0 f paqih dt “ 0.
Hence ş ş ş ş ş ş
R f pzq dz “ s ´ sn ` vb,n ´ va,n ` Rn f pzq dz
ş ş ş ş
“ s ´ sn
v `
´ va,n f pzq dz ` 0
ş b,n ş
ş ş
“ lim p s ´ sn ` vb,n ´ va,n f pzq dzqq “ 0.
nÑ8
170 Solutions

8
ř 8
ř
Solution to Exercise 4.1. As an converges, so do the real series Re an and
n“1 n“1
8
ř
Im an . Hence lim Re an “ 0 and lim Im an “ 0. Thus lim an “ 0 too.
n“1 nÑ8 nÑ8 nÑ8
8
ř
Solution to Exercise 4.2. As |an |converges, and for all n P N, |Re an | ď |an | and
n“1
8
ř 8
ř
|Im an | ď |an |, it follows that |Re an | and |Im an | converge by the Comparison
n“1 n“1
8
ř ř8
Test. Thus the two real series Re an and Im an are absolutely convergent and
n“1 n“1
hence convergent (using the real analysis result that a real series which is absolutely
8
ř
convergent is convergent). So an converges.
n“1
8
ř
Solution to Exercise 4.3. Let sn :“ a1 ` ¨ ¨ ¨ ` an , n P N, and L :“ an . Then
n“1
psn qnPN converges to L. Let N P N be such that |sN ´ L| ă ǫ. The subsequence
psN `k qkPN of psn qnPN converges to L too. Thus psN `k ´sN qkPN converges to L´sN ,
8
ř
i.e., paN `1 ` ¨ ¨ ¨ ` aN `k qkPN converges to L ´ sN . In other words, an “ L ´ s N .
8 n“N `1
ř
Moreover, | an | “ |L ´ sN | “ |sN ´ L| ă ǫ.
n“N `1
Solution to Exercise 4.4. The nth partial sum is sn :“ 1 ` z ` ¨ ¨ ¨ ` z n´1 . Thus

zsn “ z ` z 2 ` ¨ ¨ ¨ ` z n´1 ` z n , and so p1 ´ zqsn “ 1 ´ z n . Since |z| ă 1, z ‰ 1 and
n
so 1 ´ z ‰ 0. So sn “ 1 ` z ` ¨ ¨ ¨ ` z n´1 “ 1´z 1´z .
1´z n 1´0
As |z| ă 1, lim z n “ 0 (Example 1.1, p. 17). Thus lim sn “ lim “ 1´z ,
nÑ8 nÑ8 nÑ8 1´z
8
ř 8
ř 1
and so zn converges, with the sum zn “ lim sn “ 1´z .
n“0 n“0 nÑ8
Solution to Exercise 4.5. Let σn :“ 1 ` 2z ` 3z 2 ` ¨ ¨ ¨ ` pn´1qz n´2 ` nz n´1 ,

n P N. Then zσn “ z ` 2z 2 ` ¨ ¨ ¨ ` pn ´ 1qz n´1 ` nz n . Subtracting, we get
n
p1 ´ zqσn “ σn ´ zσn “ 1 ` z ` z 2 ` ¨ ¨ ¨ ` z n´1 ´ nz n “ 1´z n
1´z ´ nz . Thus
1´z n nz n 1´z n`1
σn “ p1´zq2 ´ 1´z . (We can also obtain this by differentiating 1`z`¨ ¨ ¨`z n “ 1´z
1
with respect to z.) If we set r :“ |z|, then 0 ď r ă 1 and so r “ 1`h where
1 n
`n
˘ ` ˘
n 2
` ˘
n n
` ˘
n 2 npn´1q 2
h :“ r ´1 ą 0. We have p1`hq “ 1` 1 h` 2 h `¨ ¨ ¨` n h ě 2 h “ 2 h .
n 2 2
Hence 0 ď nr n “ p1`hqn ď n npn´1qh2 “ pn´1qh2 , and so by the Sandwich Theorem,
1´z nz n n 1´0 0 1
lim nr n “ 0. Consequently, lim σn “ lim p p1´zq2 ´ 1´z q “ p1´zq2
´ 1´z “ p1´zq2
.
nÑ8 nÑ8 nÑ8
| n1s | “ | expps1Log nq | “ | expps1log nq | “ 1
eReps log nq
“ 1
eplog nqRe s
“ 1
pelog n qRe s
“ 1
nRe s
.
8
ř 8
ř
1 1
Recall that np converges if p ą 1. Hence if Re s ą 1, then nRe s
converges.
n“1 n“1
8
ř 1
Thus ns converges absolutely for Re s ą 1, and so it converges for Re s ą 1.
n“1
8
ř
Solution to Exercise 4.7. Suppose that cn pz ´ z0 qn converges for a z˚ P C.
n“0
8
ř
If R ă |z˚ ´z0 |, then cn pz ´z0 qn diverges at z˚ , a contradiction. So R ě |z˚ ´z0 |.
n“0
8
ř
Suppose that cn pz ´ z0 qn diverges for a z˚ P C. If R ą |z˚ ´ z0 |, then the series
n“0
8
ř
cn pz ´ z0 qn converges at z˚ , a contradiction. Thus R ď |z˚ ´ z0 |.
n“0
Solution to Exercise 4.8. Each term |z n! |, n “ 1, ¨ ¨ ¨ , N , appears in the list of

terms |z n |, n “ 1, ¨ ¨ ¨ , N !. As |z k | ě 0 for all k P N, it follows immediately that
|z 1 | ` |z 2! | ` 0 ` 0 ` 0 ` |z 3! | ` ¨ ¨ ¨ ` |z N ! | ď |z 1 | ` |z 2 | ` ¨ ¨ ¨ ` |z N ! |.
8
ř |z|
For |z| ă 1, the geometric series converges: |z|n “ 1´|z| . So the increasing real
n“1
ř
N ř
N ř!
N ř!
N 8
ř |z|
sequence p |z n! |qN PN is bounded: |z n! | ď |z n | “ |z|n ď |z|n “ 1´|z| .
n“1 n“1 n“1 n“1 n“1
ř
N 8
ř
Thus p |z n! |qN PN converges, that is, z n! converges absolutely. As absolutely
n“1 n“1
8
ř
convergent complex series are convergent in C, for z P D, z n! converges in C.
n“1
We show that the radius of convergence R “ 1. From the above, it is enough to
8
ř 8
ř
show that z n! diverges if |z| ą 1. Suppose for a z with |z| ą 1, z n! converges.
n“1 n“1
If cn denote the coefficients of the power series, then pcn z n qnPN converges to 0.
Hence the subsequence pcn! z n! qnPN “ pz n! qnPN also converges to 0. So there exists
an N large enough so that for all n ą N , |z n! | ă 1. But as |z| ą 1, this leads to
the contradiction that for n ą N , 1 ă |z|n! “ |z n! | ă 1.
1˝ Let L ‰ 0. For allaz such that a |z| ă L1 , there exists a q ă 1 and an N large
enough such that n |cn z n | “ n |cn | |z| ď q ă 1 for all n ą N . This is because
a nÑ8
n
|cn | |z| ÝÑ L|z| ă 1. (E.g., take q “ L|z|`1
2 ă 1.) So by the Root Test, the
power series is absolutely convergent for such z.
a a
If |z| ą L1 , then there exists an N P N such that n |cn z n | “ n |cn | |z| ą 1 for
a nÑ8
all n ą N . This is because n |cn | |z| ÝÑ L|z| ą 1. So again by the Root Test,
the power series diverges.
a a
˝
2 Let L “ 0. For z P C, there exists a q ă 1 such that n |cn z n | “ n |cn | |z| ď q ă 1
a nÑ8
for all n ą N . This is because n |cn | |z| ÝÑ 0|z| “ 0 ă 1. (E.g., q “ 12 ă 1.)
So again by the Root Test, the power series is absolutely convergent for such z.
b
Solution to Exercise 4.10. We have lim n n1n “ lim n1 “ 0. So the radius of
nÑ8 nÑ8
8 n
ř z
convergence of nn is infinite, and the power series converges for all z P C.
n“1
Solution to Exercise 4.11. For z “ 0, the series converges with sum 0. For
1
z ‰ 0, |z| ‰ 0, and let N P N be such that N ą |z| . For n ą N , |nz| ą N |z| ą 1,
8
ř
giving |nn z n ´ 0| “ |nz|n ą 1n “ 1, showing �p lim nn z n “ 0q. So nn z n diverges.
nÑ8 n“1
172 Solutions
Solution ton`1
Exercise 4.12.
p´1q 8 p´1qn
n ř
As lim | n`1
p´1qn | “ lim “ 1, the radius of convergence of zn is 1.
nÑ8 n nÑ8 n`1 n“1
n
pn`1q2022 8
ř
As lim | n2022
| “ lim p1` n1 q2022 “ 1, the radius of convergence of n2022 z n is 1.
nÑ8 nÑ8 n“0
1 8
1 ř 1 n
As lim | pn`1q!
1 | “ lim “ 0, the radius of convergence of z is infinite.
nÑ8 n! nÑ8 n`1 n“0
n!
Solutionato Exercise 4.13. 8

ř
(1) Let p n |cn |qnPN be unbounded. Suppose for some z P Czt0u, cn z n converges.
n“0
Then lim cn z n “ 0, and so pcn z n qnPN is bounded. Let M ą 0 be such that
nÑ8 a 1
for all n P N, |cn z n | ď M . Thus n |cn | ď |z| ?
n
M
, a contradiction. So for all
8
ř 8
ř
z P Czt0u, cn z n diverges, i.e., the radius of convergence of cn z n is 0.
n“0 n“0
a a
(2) We have: M1 “ supt|c1 |, a |c2 |, a 3
|c3 |, ¨ ¨ ¨ u,
M2 “ supt |c2 |, a|c3 |, ¨ ¨ ¨ u,
3
M3 “ supt 3
|c3 |, ¨ ¨ ¨ u,
¨¨¨ .
pMn qnPN converges to L :“ inf Mn (as it’s decreasing and bounded below by 0).
nPN
If z “ 0, then convergence of the power series is obvious.
1 1
Let z P Czt0u satisfy |z| ą L. As there is a gap between |z| and L, there
1
exists an α such that L “ inf Mn ă α ă |z| . Then there exists an N P N such
nPN a 1
that for all n ą N , Mn ď MN ă α. Thus for all n ą N , n |cn | ă α ă |z| , i.e.,
a 8
ř
n
|cn z n | ă α|z| “: r ă 1. By the Root Test, cn z n is absolutely convergent.
n“0
Hence we have shown:
8
ř
‚ If L ą 0, then cn z n is absolutely convergent for all z P Dp0, Lq.
n“0
8
ř
‚ If L “ 0, then cn z n is absolutely convergent for all z P C.
n“0
1 1
It remains to show that if L ą 0, then we have divergencea L . If |z| ą L ,
for |z| ą a
1 1
then |z| ă L “ inf Mn . So for all n P N, |z| ă Mn “ supt |cn |,
n n`1
|cn`1 |, ¨ ¨ ¨ u,
nPN
and in particular, there exists an mn ą n such that |cmn z mn | ą 1. So for all
n P N, there exists an mn ą n such that |cmn z mn | ą 1. Thus it is not the case
8
ř
that lim cn z n “ 0, and hence cn z n diverges.
nÑ8 n“0
ř
n ř
n ř
n
Solution to Exercise 4.14. For n ě 0, let An , Bn , Cn be ak , bk , ck . Then:
k“0 k“0 k“0
C2n ´ An Bn “ a0 pbn`1 `¨ ¨ ¨` b2n q ` a1 pbn`1 `¨ ¨ ¨` b2n´1 q `¨ ¨ ¨` an´1 pbn`1 q
` b0 pan`1 `¨ ¨ ¨` a2n q`b1 pan`1 `¨ ¨ ¨` a2n´1 q`¨ ¨ ¨`bn´1 pan`1 q.
See the following picture.
b2n
bn ¨¨¨
bn´1
.. ..
. .
b1
b0 ¨¨¨
a0 a1 an´1 an a2n
An Bn
c0 c1 cn´1 cn c2n
So
|C2n ´ An Bn | ď |a0 |p|bn`1 | `. . . q ` |a1 |p|bn`1 | `. . . q `. . .` |an´1 |p|bn`1 | `. . . q
` |b0 |p|an`1 | `. . . q`|b1 |p|an`1 | `. . . q`. . .`|bn´1 |p|an`1 | `. . . q
8
ř 8
ř 8
ř 8
ř nÑ8
ď p |ak |q |bk | ` p |bk |q |ak | ÝÑ 0.
k“0 k“n`1 k“0 k“n`1
8
ř 8
ř
Thus lim C2n “ lim An Bn “ lim An lim Bn “ p an qp bn q. Similarly, one can
nÑ8 nÑ8 nÑ8 nÑ8 n“0 n“0
nÑ8 8
ř 8
ř
show that |C2n`1 Án Bn | ÝÑ 0, and so lim C2n`1 “ lim An Bn “ p an qp bn q.
nÑ8 nÑ8 n“0 n“0
8
ř 8
ř 8
ř
Consequently, cn “ lim Cn “ p an qp bn q.
n“0 nÑ8 n“0 n“0
Solution to Exercise 4.15. In one jump, the ‘edge-distance’ (i.e., smallest num-
ber of edges to E) changes by 1. As the initial edge-distance is 4, which is even,
an odd number of jumps cannot reduce it to 0. Thus a2n´1 “ 0 for all n P N.
F E
G D
H C
A B
In two jumps, the frog can only stay at A, or reach C or G, and so a2 “ 0. In four
jumps, the frog can reach E either following the path ABCDE or AHGF E, and
so a4 “ 2.
174 Solutions
bn is also the number of paths of length n from C to E. For a path of 2n jumps

starting from A, the first two jumps can land the frog back at A in two ways
(ABA or AHA), or at C in one way (ABC), or at G in one way (AHG). Thus
a2n “ 2a2n´2 ` 2b2n´2 for all n ą 1. Next, for a path of 2n ą 2 jumps starting from
C, the first two jumps can land the frog back at C in two ways (CDC or CBC),
or at A in one way (CBA). Thus b2n “ 2b2n´2 ` a2n´2 for all n ą 1.
From a2n “ 2a2n´2 `2b2n´2 , we obtain b2n´2 “ a2n ´2a
2
2n´2
. So b2n “ a2n`22´2a2n .
Substituting these expressions for b2n´2 and b2n in b2n “ 2b2n´2 ` a2n´2 , we get
a2n`2 “ 4a2n ´ 2a2n´2 for all n ą 1.
We have Apzq “ a0 ` a2 z ` a4 z 2 ` a6 z 3 ` ¨ ¨ ¨ “ 2z 2 ` a6 z 3 ` ¨ ¨ ¨ . For all
n ą 1, a2n`2 “ 4a2n ´ 2a2n´2 , and so we get a2n`4 “ 4a2n`2 ´ 2a2n for all n P N.
Multiplying by z n`2 and adding, we obtain Apzq ´ 2z 2 “ 4zApzq ´ 2z 2 Apzq, ? and
so p1 ´ 4z ` 2z 2 qApzq “ 2z 2 , i.e., p1 ´ αzqp1 ´ βzqApzq “ 2z 2 , where α “ 2 ´ 2,
? řn
β “ 2 ` 2. Hence for |z| small enough, and with cn “ αn´k β k ,
k“0
8
ř
2z 2
Apzq “ p1´αzqp1´βzq “ 2z 2 p1 ` αz ` α2 z 2 ` ¨ ¨ ¨ qp1 ` βz ` β 2 z 2 ` ¨ ¨ ¨ q “ 2z 2 cn z n ,
n“0
‘Comparing coefficients’ (justified by Corollary 4.5; see Remark 4.2), for n ě 2,

ř
n´2
a2n “ 2cn´2 “ 2 αn´2´k β k “ 2pαn´2 ` αn´3 β ` ¨ ¨ ¨ ` αβ n´3 ` β n´2 q β´α
β´α
k“0 ? ?
n´1 ´αn´1 p2` 2qn´1?
´p2´ 2qn´1
“ 2β β´α “ 2
.
If n “ 1, then the right-hand side is 0, which matches a2 “ 0. So the formula holds
for all n P N.
As a2n is the number of paths, a2n is a nonnegative integer, and so the right-
hand side is necessarily a nonnegative integer. But this also follows from the
Binomial Theorem, since
? ? ř `n´1˘? k n´1´k
n´1 ř `n´1˘
n´1 ?
k 2k 2n´1´k
p2 ` 2qn´1 ´ p2 ´ 2qn´1 “ k 2 2 ´ k p´1q
k“0 k“0
ř `n´1˘
n´1 ?
k q 2k 2n´1´k
“ k p1 ´ p´1q
k“0
As 1´p´1qk is 0 for even k, the only terms that survive are the ones with an odd k.
If k` “ 2m˘ ` 1mď?n ń´1´p2m`1q
1, m a nonnegative integer, then
? the corresponding summand
n´1
is 2m`1 p2q2 22 , which is equal to 2 times a nonnegative integer.
?
So when the sum of such terms is divided by 2, we get a nonnegative integer.
1
Solution to Exercise 4.16. f pzq :“ 1 ` 2z ` 3z 2 ` 4z 3 ` ¨ ¨ ¨ “ p1´zq2 for |z| ă 1.
2 3 4 z
Multiplying by z, we get zf pzq “ gpzq :“ z ` 2z ` 3z ` 4z ` ¨ ¨ ¨ “ p1´zq 2
2 3 4
for |z| ă 1. As gpzq :“ z ` 2z ` 3z ` 4z ` ¨ ¨ ¨ converges for |z| ă 1, g is
holomorphic in the disc Dp0, 1q, with g 1 pzq “ 1 ` 22 z ` 32 z 2 ` 42 z 3 ` ¨ ¨ ¨ for
1 2 1´z`2z 1`z
z
|z| ă 1. But g 1 pzq “ ddz p1´zq2 “ 1 p1´zq2 ` z p1´zq3 “ p1´zq3
“ p1´zq 3 . Hence
1`z
1 ` 22 z ` 32 z 2 ` 42 z 3 ` ¨ ¨ ¨ “ p1´zq3 for |z| ă 1.

8 n
ř z
(1) False. For example, tz P C : n2
convergesu “ tz P C : |z| ď 1u.
n“1
(2) True.
8 p´1qn
ř
(3) False. For example n zn converges for z “ 1, but diverges for z “ ´1.
n“1
(4) False. See the example in (3).
(5) True. Same example as in (3).
8 n
ř z
(6) True. For example, consider n2
.
n“1 ?
(7) True. The radius of convergence R satisfies R ď |i| “ 1, and |1` i| “ 2 ą 1 “ R.
Solution to Exercise 4.18. As a P p0, 1q, none of the coefficients is 0. We have

apa´1q¨¨¨pańq a
´1
lim | cn`1 pn`1q! ań 0´1
cn | “ lim | apa´1q¨¨¨pa´pn´1qq | “ lim | n`1 | “ lim | 1` 1 | “ | 1`0 | “ 1.
n
nÑ8 nÑ8 n! nÑ8 nÑ8 n
1 1
So the radius of convergence is c “ 1 “ 1.
lim | n`1
c
|
nÑ8 n
As R “ 1, Dp0, Rq “ tz P C : |z| ă 1u. By termwise differentiation, for |z| ă 1,

a
f 1 pzq “ 0` 1! 1` apa´1q
2! 2z `
apa´1qpa´2q
3! 3z 2 `¨ ¨ ¨ “ a` apa´1q
1! z `
apa´1qpa´2q 2
2! z `¨ ¨ ¨ .
a apa´1q apa´1qpa´2q
Multiplying by z, zf 1 pzq “ 1! z ` 2! 2z `
2
3! 3z 3 ` ¨ ¨ ¨ . So for all |z| ă 1,
apa´1q apa´1q¨¨¨pańq n
p1 ` zqf 1 pzq “ a ` 1! z ` ¨ ¨ ¨ ` n! z ` ¨¨¨
a apa´1q¨¨¨pa´pn´1qq n ` ¨¨¨
` 1! z ` ¨ ¨ ¨ ` n! nz
apa´1q¨¨¨pa´pn´1qq
“ a ` a2 z ` ¨ ¨ ¨ ` n! pa ´✚ nqz n ` ¨ ¨ ¨
n `✚
a2 pa´1q¨¨¨pa´pn´1qq n
“ a ` a2 z ` ¨ ¨ ¨ ` n! z ` ¨¨¨
apa´1q¨¨¨pa´pn´1qq n
“ ap1 ` az ` ¨ ¨ ¨ ` n!` ¨ ¨ ¨ q “ af pzq.
z
For z ‰ ´1, we have p1 ` zqa “ exppaLogp1 ` zqq. As Log P Czpp´8, 0s ˆ t0uq, we
have Logp1 ` ¨q P OpCzpp´8, ´1s ˆ t0uq. In particular, Logp1 ` ¨q P OpDp0, 1qq.
Thus p1 ` zqa P OpDp0, 1qq. For z P Dp0, 1q,
pp1 ` ¨qá f q1 pzq “ pexppáLogp1 ` ¨qqq1 pzqf pzq ` exppáLogp1 ` zqqf 1 pzq
pzq
á
“ exppáLogp1 ` zqq 1`z 1f pzq ` exppáLogp1 ` zqq af
1`z
pzq
á
“ exppáLogp1 ` zqqp 1`z f pzq ` af
1`z q “ 0.
By Exercise 2.18, p1 ` ¨qá f p¨q is constant in Dp0, 1q. Hence for all z P Dp0, 1q,
p1 ` zqá f pzq “ p1 ` 0qá f p0q “ 1. Thus f pzq “ p1 ` zqa for all z P Dp0, 1q.
8
ř p´1qn
Solution to Exercise 4.19. Consider first cn wn , where cn “ 22n pn!q2
. We have
n“0
p´1qn`1
nÑ8 8 p´1qn
ř
2pn`1q ppn`1q!q2 22n ¨n!¨n!
| cn`1
cn | “ |
2
p´1qn | “ 22n ¨22 ¨pn`1q!¨pn`1q! “ 1
4pn`1q2 ÝÑ 0. So 22n pn!q2 w
n
22n pn!q2 n“0
8 p´1qn
ř
converges everywhere. For z P C, putting w “ z 2 , J0 pzq “ 22n pn!q2
z 2n converges.
n“0
176 Solutions
Power series can be differentiated termwise inside the disc of their convergence. So
8 p´1qn
ř 8
2n´1 , J 2 pzq “ ř p´1q 2np2n´1qz 2n´2 , and
n
for z P Czt0u, J01 pzq “ 22n pn!q2 2nz 0 22n pn!q2
n“1 n“1
8 p´1qn
ř 8 p´1qn´1
ř
1 1
J02 pzq ` z J0 pzq “ 22n pn!q2
2npp2n´1q ` 1qz 2n´2 “ p´1q 22n pn!q2
2np2nqz 2pn´1q
n“1 n“1
8
ř p´1qn´1 8
ř p´1qm
“ p´1q 2pn´1q “ p´1q z 2m “ ´J0 pzq.
2pn´1q ppn´1q!q2 z 22m pm!q2
n“1 2 m“0
8
ř
Solution to Exercise 4.20. As cn z n is absolutely convergent for z “ 1, it is
n“0
convergent for z “ 1, and the radius of convergence R ě 1. If R ą 1, then with
8
ř 8
ř
f pzq :“ cn z n for z P Dp0, Rq, we have f P OpDp0, Rqq, and f 1 pzq “ ncn z n´1 in
n“0 n“1
8
ř
Dp0, Rq. So ncn z n´1 is absolutely convergent for all z P Dp0, Rq. In particular,
n“1 8
ř
as R ą 1, with z “ 1, we get n|cn | converges, a contradiction. Thus R “ 1.
n“1
d2n d2n`1
Solution to Exercise 4.21. As dz 2n sin z “ p´1qn sin z, n
dz 2n`1 sin z “ p´1q cos z,
8
ř 1 dn 3 5
sin 0 “ 0 and cos 0 “ 1, we have sin z “ n! p dz n sin zq|z“0 z n “ z ´ z3! ` z5! ´ ` ¨ ¨ ¨ .
n“0
z2 z4
Similarly, cos z “ 1 ´ 2! ` 4! ´ ` ¨ ¨ ¨ .
8 n
ř 8
exppizqèxppízq i n ř p´1qn in n
Alternatives: As cos z “ 2 “ 12 p z `n! z q, n! and i2n “ p´1qn,
n“0 n“0
1 z2 iz 3 z4 iz 5 z6
cos z “ 2 p 1 ` iz ´ 2! ´ 3! ` 4! ` 5! ´ 6! ` ¨¨¨
z2 iz 3 z4 iz 5 z6
`1 ´ iz ´ 2! ` 3! ` 4! ´ 5! ´ 6! ` ¨¨¨ q
z2 z4 z6
“ 1 ´ ` ´ ` ´¨¨¨ . 2! 4! 6!
Or termwise differentiate the power series for sin:
8 p´1qn
ř 8
2n “ ř p´1q z 2n “ 1 ´
n
z2 z4 z6
cos z “ sin1 z “ p2n`1q! p2n ` 1qz p2nq! 2! ` 4! ´ 6! ` ´¨¨¨.
n“0 n“0
Solution to Exercise 4.22. Let ppzq “ z 6 ´ z 4 ` z 2 ´ 1, z P C. Then

p1 pzq “ 6z 5 ´ 4z 3 ` 2z p3 pzq “ 120z 3 ´ 24z pp5q pzq “ 720z pp7q pzq “ 0
2 4
p pzq “ 30z ´ 12z ` 2 2 p4q 2
p pzq “ 360z ´ 24 p6q
p pzq “ 720 ¨¨¨ .
p1 p1q p2 p1q p3 p1q
pp4q p1q pp5q p1q pp6q p1q
So pp1q “ 0, 1! “ 4, 2! “ 10, 3! “ 16,
4! “ 14, 5! “ 6, 6! “ 1. So @z P C,
p1 p1q pp6q p1q
z 6 ´z 4 `z 2 ´1 “ pp1q ` 1! pz ´ 1q ` ¨ ¨ ¨ ` 6! pz ´ 1q ` 0 6
“ 4pz´1q`10pz´1q2 `16pz´1q3 `14pz´1q4 `6pz´1q5 `pz´1q6 .

Or, use Binomial Theorem: ppzq “ ppz ´ 1q ` 1q6 ´ ppz ´ 1q ` 1q4 ` pz ´ 1 ` 1q2 ´ 1.
(1) z ÞÑ exppz 2 qş has a primitive, ş say g, in the simply connected domain C.
Thus f pzq “ r0,zs exppζ 2 qdζ “ r0,zs g 1 pζqdζ “ gpzq ´ gp0q. So f P OpCq and
8
ř 1 2n 1 d2n 1
f 1 pzq “ g 1 pzq “ exppz 2 q “ z .n! Thus for all n ě 0, 1
p2nq! dz 2n f pzq|z“0 “ n!
n“0
and 1 d2n`1 1
p2n`1q! dz 2n`1 f pzq|z“0 “ 0, i.e., f p2n`1q p0q “ p2nq!
n! and f
p2n`2q p0q “ 0. Also,
ş 8 f pmq p0q
ř 8 p2n`1q p0q
ř 8
ř z 2n`1
f p0q “ r0,0s exppζ
2 qdζ “ 0. So f pzq “ m!zm “ f z 2n`1 “ p2n`1q! p2n`1qpn!q .
m“0 n“0 n“0
1
(2) For |z| ă 1, z`1 “ 1 ´ z ` z 2 ´ z 3 ` z 4 ´ ` ¨ ¨ ¨ . As power series are holomorphic
in the region of convergence with complex derivative obtained by termwise
1 d 1 2 3
differentiation, ´ pz`1q2 “ dz z`1 “ ´1 ` 2z ´ 3z ` 4z ´ ` ¨ ¨ ¨ for |z| ă 1.
8
ř
z2
Multiplying by ´z 2 gives pz`1q2
“ z 2 ´ 2z 3 ` 3z 4 ´ ` ¨ ¨ ¨ “ p´1qn pn ´ 1qz n
n“2
for |z| ă 1. So we have c0 “ c1 “ 0 and cn “ p´1qn pn ´ 1q for n ě 2.
Solution to Exercise 4.24. Let Cr ptq “ re2πit , t P r0, 1s. Then for all integers
1
ş f pζq
n ě 0, we have cn “ 2πi Cr ζ n`1 dζ. Consequently, by the M L-inequality, we get
2πr |f pζq| 1 1 1
|cn | ď 2π max n`1 ď r rn`1 “ rn . Passing to the limit as r Õ 1, |cn | ď lim n “ 1.
ζPCr r rÑ1´ r
Solution to Exercise 4.25. Suppose f P OpDp0, rqq is such that f pnq p0q “ pn!q2
8 f pnq p0q
ř 8 pn!q2
ř 8
ř
for all n ě 0. Then for z P Dp0, rq, f pzq “ n! zn “ n! zn “ n!z n .
n“0 n“0 n“0
8
ř
r
Setting z “ 2 P Dp0, rq, n!p 2r qn converges. If N P N is such that N 2r ą 1, then for
n“0
all n ą N , n!p 2r qn “ N !pN `1q ¨ ¨ ¨ pN `pnŃ qqp 2r qN p 2r qnŃ ą N !p 2r qN 1nŃ , showing
that �pp lim n!p 2r qn “ 0qq, a contradiction. Hence, there is no f P OpDp0, rqq such
nÑ8
that f pnq p0q “ pn!q2 for all n ě 0.
8
ř
Solution to Exercise 4.26. Let the Taylor series of f be given by f pzq “ cn z n .
n“0
The series is (absolutely) convergent for all z P C since f is entire. As f pz 2 q “ f pzq,
we have c0 ` c1 z 2 ` c2 z 4 ` ¨ ¨ ¨ “ c0 ` c1 z ` c2 z 2 ` ¨ ¨ ¨ , that is, for all z P C,
c1 z ` pc2 ´ c1 qz 2 ` ¨ ¨ ¨ ` c2n`1 z 2n`1 ` pc2n`2 ´ cn`1 qz 2n`2 ` ¨ ¨ ¨ “ 0. Since the
right-hand side is the zero function, which has the Taylor expansion given by the
left-hand side, it follows that the kth coefficient of the left-hand side power series
pkq
is zero (being 0 k!p0q ) for all integers k ě 0. So we get 0 “ c1 “ c3 “ c5 “ ¨ ¨ ¨
and c2 ´ c1 “ 0, c4 ´ c2 “ 0, ¨ ¨ ¨ , c2n`2 ´ cn`1 “ 0, ¨ ¨ ¨ , so that all the coefficients
cn “ 0 for n P N. Thus f pzq “ c0 for all z P C, that is, f is a constant. Conversely,
if f is constant, it satisfies the given relation and is entire.
Solution to Exercise 4.27. For δ ą 0, let Sk :“ tz P C : |zák |`Repzák q ď δu.

Then Sk is a ‘parabolic sector’: Writing z ´ ak “ x ` iy, the above qualifying
2
inequality becomes x ď 2δ ´ y2δ , and a picture of Sk is displayed below.
2
Sk xák “ 2δ ´ y2δ
ak x
8 f pnq pa q
ř
The Taylor series centered at ak of f P OpCq is f pzq “ k
n! pz ´ ak qn (z P C).
n“0
178 Solutions
In particular,
8 f pnq pa q
ř 8 |f pnq pa q|
ř
|f pzq| “ | k
n! pz ´ ak qn | ď k
n! |z ´ ak |n
n“0 n“0
8 á
ř 8
ř
e k 1
ď n! |z ´ ak |n “ eák n! |z ´ ak |n “ eák e|zák | .
n“0 n“0
Thus for z P Sk , | fepzq

´z | ď e
Re z eák e|zák | ď eδ .
As lim ak “ 8, the sectors Sk (which are all congruent, being shifted-versions of

kÑ8 8
Ť
each other) together cover C, i.e., C “ Sk . Hence
k“0
8
Ť
for z P C “ Sk , | fepzq δ
´z | ď e . p‹q
k“0
By Liouville’s theorem, the bounded entire function z ÞÑ fepzq ´z is a constant. So
there exists a c P C such that f pzq “ c e´z (z P C). Substituting this expression for
f in (‹) yields |c| ď eδ . As δ ą 0 was arbitrary, we obtain |c| ď inf δą0 eδ “ e0 “ 1.
Solution to Exercise 4.28. Let z P C. Let R ą |z|. Then

pn`1q! pn`1q!
|f pn`1q pzq| ď max |f pzq|
Rn`1 |z|ďR
ď max M |z|n “ pn`1q!
Rn`1 |z|ďR Rn`1
M Rn “ pn`1q!M
R .
But the choice of R ą |z| was arbitrary, and so f pn`1q pzq “ 0. Since z P C
was
arbitrary, we obtain that f pn`1q ” 0 in C. By Taylor’s Theorem, for all z P C,
8 f pkq p0q
ř řn f pkq p0q
f pzq “ k k
k! pz ´ 0q “ k! z . So f is a polynomial of degree at most n.
k“0 k“0
If n “ 0, then f is a bounded entire function, and our conclusion obtained above

says that f is constant. So the special case when n “ 0 is Liouville’s Theorem.
ş
Solution to Exercise 4.29. C zsin z
2023 dz “ 0: By the Cauchy Integral Formula,
2022!
ş sin z d2022 2022
2πi C pz´0q2022`1 dz “ dz 2022 sin z|z“0 “ p´1q
2 sin z|z“0 “ 0.
Solution to Exercise 4.30. We have f pz0 q “ 0m gpz0 q “ 0, showing that z0

is a zero of f . We claim that z0 is an isolated zero of f . By the continuity of
g at z0 , there exists a r P p0, Rq, such that gpzq ‰ 0 for z P Dpz0 , rq. From
the factorisation f pzq “ pz ´ z0 qm gpzq for all z P Dpz0 , Rq, we obtain f pzq ‰ 0
for all z P Dpz0 , rqztz0 u. So z0 is an isolated zero, i.e., we are in case 2˝ of the
Theorem on Classification of Zeroes. Hence there exists an m r P N (which is the
order of z0 as a zero of f ), and there exists a gr P OpDpz0 , Rqq, such that grpz0 q ‰ 0
and f pzq “ pz ´ z0 qm rr
g pzq for all z P Dpz0 , Rq. Then for z P Dpz0 , Rq, we have
r
pz ´ z0 q grpzq “ pz ´ z0 qm gpzq. We show that this implies m
m r “ m. For if m r ą m,
r
then for all z P Dpz0 , Rqztz0 u, pz´z0 qm´m g pzq “ gpzq, and we get the contradiction
r
r
0 ‰ gpz0 q “ lim gpzq “ lim pz ´ z0 qm´m grpzq “ 0 ¨ grpz0 q “ 0.
zÑz0 zÑz0
Similarly, m ą mr also yields a contradiction. Consequently, m “ m, r and so z0 is
a zero of order m “ m.r (Also, then g ” r g on Dpz0 , Rqztz0 u, and since they are
continuous, also on Dpz0 , Rq.)

(1) f pzq “ p1 ` z 2 q4 “ ppz ´ iqpz ` iqq4 “ pz ´ iq4 pz ` iq4 for all z P C. So with
gpzq :“ pz ` iq4 , g is entire, gpiq “ p2iq4 “ 16 ‰ 0 and f pzq “ pz ´ iq4 gpzq.
Thus i is a zero of f of order 4.
(2) For n P Z, f p2nπiq “ 1 ´ 1 “ 0, and f 1 p2nπiq “ exp z|z“2nπi “ 1 ‰ 0. So 2nπi
is a zero of f of order 1.
(3) f p0q “ cos 0 ´ 1 ` 12 psin 0q2 “ 1 ´ 1 ` 12 p0q2 “ 0, and we have
1 p1ćosp2zqq
f pzq “ cos z ´ 1 ` 12 psin zq2 “ cos z ´ 1 ` 2 2
“ cos z ´ 43 ´ 14 cosp2zq
2 4 6 2 4 26 z 6
“ p1 ´ z2! ` z4! ´ z6! ` ´ ¨ ¨ ¨ q ´ 43 ´ 14 p1 ´ 4z2! ` 16z 4! ´ 6! ` ´¨¨¨q
“ p1´ 43 ´ 14 q ` p´ 2!
1
` 14 2!4 qz 2 ` p 4!1 ´ 14 16
4! qz 4 ` ¨¨¨
“ 0 ` 0z 2 ´ 4!3 z 4 ` ¨ ¨ ¨ ,
and so z0 is a zero of order 4.
Solution to Exercise 4.32. In the disc, there is only one zero of f , which is at
z0 , and so for z in the disc such that z ‰ z0 , we have f pzq ‰ 0. By the result on
the classification of zeroes, there exists a function g that is holomorphic in the disc,
gpz0 q ‰ 0, and f pzq “ pz ´ z0 qgpzq for all z in the disc. It follows that g is nonzero
at each point of C. Thus
1
1
ş zf 1 pzq 1
ş zp1 gpzq`pz´z0 q g1 pzqq 1
ş zpgpzq`pz´z
gpzq
0 q g pzqq
2πi C f pzq dz “ 2πi C pz´z0 qgpzq dz “ 2πi C z´z0 dz

zpgpzq`pz´z0 q g 1 pzqq
“ gpzq |z“z0 (Cauchy Integral Formula)
z0 pgpz0 q`0 g 1 pz0 qq
“ gpz0 q “ z0 .
Solution to Exercise 4.33. By the result on the classification of zeroes, there

exists a g P OpDq such that f pzq “ pz ´ z0 qm gpzq, and gpz0 q ‰ 0. Thus with
Gpzq :“ pgpzqq2 , we have pf pzqq2 “ pz ´ z0 q2m Gpzq. Clearly pf pz0 qq2 “ 0, and
G P OpDq, with Gpz0 q “ pgpz0 qq2 ‰ 0. So z0 is a zero of z ÞÑ pf pzqq2 of order 2m.
We have f 1 pzq “ mpz ´ z0 qm´1 gpzq ` pz ´ z0 qm g1 pzq “ pz ´ z0 qm´1 g1 pzq, where
g1 pzq :“ mgpzq ` pz ´ z0 qg 1 pzq. As m ą 1, f 1 pz0 q “ 0 g1 pz0 q “ 0. Clearly, g1 P OpDq
and g1 pz0 q “ mgpz0 q ` 0 g1 pz0 q “ mgpz0 q ‰ 0. Thus z0 is a zero of f 1 of order m ´ 1.
Solution to Exercise 4.34. We have gpz0 q “ cospf pz0 qq ´ 1 “ cos 0 ´ 1 “ 0.

By the classification of zeroes theorem, there exists an entire function h such that
f pzq “ pz ´ z0 qhpzq for all z P C, and hpz0 q ‰ 0. Moreover,
g 1 pzq “ p ´ sinppz ´ z0 qhqqq p hpzq ` pz ´ z0 qh1 pzqqq,
g 2 pzq “ ´pp cosppz ´ z0 qhqqq phpzq ` pz ´ z0 qh1 pzqqq 2
`pp ´ sinppz ´ z0 qhqqq p h1 pzq ` h1 pzq ` pz ´ z0 qh2 pzqqq.
Thus g 1 pz0 q “ 0, g2 pz0 q “ ´1phpz0 qq2 `0 ‰ 0. So z0 is a zero of g of order 2.
180 Solutions
Solution to Exercise 4.35. We have |f p log1 n q| ď 1

n for n “ 2, 3, ¨ ¨ ¨ . As f is
continuous at 0, by passing to the limit as n Ñ 8 in the above, we get |f p0q| ď 0.
Thus f p0q “ 0. Suppose f is not identically 0 in Dp0, 1q. By the theorem on the
classification of zeroes, 0 is a zero of f of some order m P N, and there exists a
1
g P OpDp0, 1qq such that f “ z m g in D, and gp0q ‰ 0. Setting z “ log n , for integers
n ě 2, we get plog1nqm |gp log1 n q| “ |f p log1 n q| ď n1 , which gives upon rearranging that
m
|gp log1 n q| ď plognnq for all integers n ě 2. Passing to the limit as n Ñ 8, we obtain,
m
thanks to the continuity of g at 0, that |gp0q| ď lim plognnq “ 0, and so gp0q “ 0,
nÑ8
a contradiction. Hence f ” 0 in Dp0, 1q.
plog nqm n1{m
Proof that lim n “ 0: It is enough to show lim logn1{m “ 0, since then
nÑ8 nÑ8
m
lim plognnq “ lim pm logn1{m
1{m 1{m
n
qm “ p lim m logn1{m
n
qm “ pm ¨ 0qm “ 0.
nÑ8 nÑ8 nÑ8
log n1{m 1
şn1{m 1 1
şn1{m 1 1
?
We have 0 ď n1{m “
n1{m 1 t dt ď n1{m 1
?
t
dt “ n1{m 2p n1{m ´ 1q, and so
log n1{m
by the Sandwich Theorem, lim n1{m “ 0.
nÑ8
Solution to Exercise 4.36. We know that for all x, y P R,

cospx ` yq “ pcos xqpcos yq ´ psin xqpsin yq. p˚q
Let y P R. Define fy P OpCq by fy pzq :“ cospz ` yq ´ ppcos zqpcos yq ´ psin zqpsin yqq
for all z P C. We have fy pxq “ 0 for all x P R, thanks to (˚), and so by the Identity
Theorem, fy pzq “ 0 for all z P C, that is,
cospz ` yq “ pcos zqpcos yq ´ psin zqpsin yq for all z P C. p˚˚q
But the choice of y P R was arbitrary, and so (˚˚) holds for all y P R. Fix z P C.
Define gz P OpCq by gz pwq :“ cospz ` wq ´ ppcos zqpcos wq ´ psin zqpsin wqq for all
w P C. Then gz pyq “ 0 for all y P R by (˚˚). Another application of the Identity
Theorem yields gz pwq “ 0 for all w P C. Hence
cospz`wq “ pcos zqpcos wq´psin zqpsin wq for all w P C. p˚˚˚q
As z P C was arbitrary, (˚˚˚) holds for all z, w P C.
Solution to Exercise 4.37. Let f, g P OpDq be such that pf ¨gqpzq “ f pzqgpzq “ 0

for all z P D. Suppose that there exists a z0 P D such that f pz0 q ‰ 0. By the
continuity of f , there exists a δ ą 0 such that f pzq ‰ 0 whenever |z ´ z0 | ă δ.
The equation f pzqgpzq “ 0 (z P D) then implies gpzq “ 0 for |z ´ z0 | ă δ. By the
Identity Theorem, g ” 0 in D. So OpDq has no zero divisors.
We now show that CpDq is not an integral domain. Let z0 P D and let δ ą 0 be
such that the disc Dpz0 , δq Ă D. Define" the continuous function ϕ : R Ñ R by
0 if t ď 0,
ϕptq “
t if t ą 0.
Define f, g by f pzq “ ϕpRepz ´ z0 qq and gpzq “ ϕp´Repz ´ z0 qq for all z P D.
ϕ
gą0 f ą0
t
0
Being the composition of continuous functions, f, g P CpDq. Also f pzq ą 0 for all z
in the right half of ∆, and so f ‰ 0 in CpDq. Similarly gpzq ą 0 for all z belonging
to the left half of ∆, and so g ‰ 0 in CpDq. Nevertheless, f ¨ g “ 0.
Solution to Exercise 4.38. Let f P OpCq be such that Zpf q “ tz P C : f pzq “ 0u
Ť
is uncountable. For n P N, let Sn :“ Zpf q X Dp0, nq. Then we have Sn “ Zpf q.
nPN
If each Sn is finite, then their countable union, Zpf q, which we know is infinite,
must be countable, which is not true. Thus there exists an N P N such that SN is
an infinite set. Take any sequence pzn qnPN of distinct elements from SN . Then for
all n P N, |zn | ď N , that is, the sequence pzn qnPN is bounded. Using the Bolzano-
Weierstrass Theorem, there exists a subsequence pznk qkPN which is convergent, with
limit say z˚ P C. As f pznk q “ 0 for all k P N, it follows by the Identity Theorem
that f ” 0. So there is no nonzero entire function with an uncountable zero set.
Solution to Exercise 4.39. Suppose that z0 P Dzt0u is a zero of f . Then we
have |z02 | “ |z0 ||z0 | ă 1 |z0 |. So z02 ‰ z0 . Also 0 ă |z0 |2 ă 1, i.e., z02 P Dzt0u. Finally,
f pz02 q “ f pz0 qf p´z0 q “ 0 f p´z0 q “ 0.
It is given that f has some zero z0 P Dzt0u. By the above, each term in the
sequence z0 , z02 , z04 , z08 , ¨ ¨ ¨ is a zero of f . Moreover, these terms are distinct: Using
n N n N n n N n
z02 P Dzt0u, for N ą n, |z02 | “ |z02 ||z02 ´2 | ă |z02 | ¨ 1, so that z02 ‰ z02 .
Also, since |z0 | ă 1, the geometric sequence z0 , z02 , z03 , ¨ ¨ ¨ converges to 0, so that
n
its subsequence pz02 qnPN also converges to 0 P D. By the identity theorem, since f
is holomorphic in the domain D, we conclude that f ” 0 in D.
Vice versa, f ” 0 is holomorphic, satisfies f pzqf p´zq “ f pz 2 q in D, and is not
zero-free in Dzt0u.
(1) No. Take D “ C, f “ exp, g “ 1. Then f p2πinq “ expp2πinq “ 1 “ gp2πinq
for all n P N, but f ‰ g (e.g., because f piπq “ ´1 ‰ 1 “ gpiπq).
(2) Yes.
(3) Yes. Let γptq “ xptq ` iyptq for all t P ra, bs, where x, y : ra, bs Ñ R. Take
t0 P pa, bq so that x1 pt0 q or y 1 pt0 q is nonzero. (If they are both always 0, then
z “ w, a contradiction.) Let x1 pt0 q ą 0 (the other cases are handled similarly).
Then x1 ptq ą 0 in a neighbourhood of t0 . So x is strictly increasing there. Take
tn “ t0 ` n1 , n ě N , with N large enough so that t0 ` N1 P ra, bs. Set zn “ γptn q.
Then pzn qněN is a sequence of distinct points (as the real parts are distinct),
which converges to γpt0 q. By the Identity Theorem, f “ g in D.
(4) Yes. By the Taylor expansion in a small disc Dpw, δq Ă D around w, f “ g in
Dpw, δq, and so by the Identity Theorem, f “ g in D.
182 Solutions
Solution to Exercise 4.41. sinplog xq “ 0 if and only if log x P tnπ : n P Zu, i.e.,
x P tenπ : n P Zu. Let F be an entire extension of f . Then F peńπ q “ f peńπ q “ 0
for every n P N. As ex is strictly increasing, the points eńπ , n P N, are distinct.
Also, eńπ Ñ 0 P C as n Ñ 8. By the Identity Theorem, F ” 0. But this is
π π π
absurd since, e.g., F pe 2 q “ f pe 2 q “ sinplog e 2 q “ sin π2 “ 1 ‰ 0.
8
ř 8
ř
g pnq p0q n
Solution to Exercise 4.42. For z P Dp0, rq, gpzq “ n! z . As g pnq p0q con-
b n“0 n“0
8
ř
g pnq p0q g pnq p0q
verges, gpnq p0q Ñ 0 as n Ñ 8. Thus nÑ8 | “ 0. So has an infinite ra-
n
lim | n! n! zn
n“0
dius of convergence, and defines an entire function that coincides with g on Dp0, rq
(as both have the same power series there). Also, such an entire function has to
be unique by the Identity Theorem (as any two such coincide with g on Dp0, rq).
Solution to Exercise 4.43. Let K “ tz P C : |z| ď 1u. For each z P K, there
8
ř
exists a smallest integer npzq ě 0 such that for all w P C, f pwq “ cn pzqpw ´ zqn
n“0
f pnpzqq pzq
and cnpzq pzq “ 0. Hence pnpzqq! “ 0, and so f pnpzqq pzq “ 0. Let ϕ : K Ñ N Y t0u
be defined by ϕpzq “ npzq. Since K is uncountable, while N Y t0u is countable,
there exists an N such that ϕ´1 pN q is infinite. Let pzn qnPN be a sequence of
distinct points in ϕ´1 pN q Ă K. In particular, |zn | ď 1 for all n P N, and so by the
Bolzano-Weierstrass Theorem, it follows that pzn qnPN has a convergent subsequence
pznk qkPN with limit, say z˚ P K. As f pN q pznk q “ 0 for all k P N, by the Identity
Theorem (applied to f pN q P OpCq), we have f pN q “ 0 in C. By Taylor’s Theorem,
8 f pnq p0q
ř ´1 f pnq p0q
Nř
f pzq “ n! zn “ n! z n , for all z P C, and so f is a polynomial.
n“0 n“0
f 1 g´f g 1
Solution to Exercise 4.44. We have p fg q1 “ g2
, and so for all integers n ě 2,
1 1 1 1
f 1p n qgp n q´f p n qg 1 p n q
p fg q1 p n1 q “ 1 2
pgp n qq
0
pgp n qq
“
1 2 “ 0. By the Identity Theorem applied to
the function p fg q1 P OpDq, we obtain p fg q1 ” 0 in D. By the Fundamental Theorem

p0q
of Contour Integration, if λ :“ fgp0q P Czt0u, and r0, zs is the line segment from 0
f pzq ş f 1 ş
to z P D, then gpzq ´ λ “ r0,zs p g q pζq dζ “ r0,zs 0 dζ “ 0. So f pzq “ λgpzq, z P D.
Solution to Exercise 4.45. g given by gpzq “ f pzq ´ z 2 (z P C) is entire, and

gp n1 q “ f p n1 q ´ n12 “ 0. As p n1 qnPN is a sequence of distinct zeroes of g in the domain
C, such that n1 Ñ 0 P C, the Identity Theorem implies g ” 0. So f pzq “ z 2 for all
z P C. Vice versa, if f pzq “ z 2 (z P C), then f P OpCq and f p n1 q “ n12 for all n P N.
Define h by hpzq “ f pz 2 q ´ z for all z P C. The composition f ˝ pz ÞÑ z 2 q is entire.
So h is entire. We have hp n1 q “ f p n12 q ´ n1 “ 0. As p n1 qnPN is a sequence of distinct
zeroes of h in the domain C, such that n1 Ñ 0 P C, h ” 0 by the Identity Theorem.
Thus f pz 2 q “ z, z P C. Taking z “ 1, f p1q “ f p12 q “ 1. Taking z “ ´1, we obtain
f p1q “ f pp´1q2 q “ ´1, a contradiction to f p1q “ 1 obtained above.
Solution to Exercise 4.46. As K :“ tz P C : |z ´ z0 | ď Ru is compact, the

Identity Theorem implies that K has only finitely many distinct zeroes of f . (Oth-
erwise, there exists a sequence of distinct zeroes of f in K, which will possess a
convergent subsequence in the compact set K, and so f ” 0 in D, a contradiction
to f being nonzero on the range of C Ă D.) So Zpf q is a finite set.
ř f1
If Zpf q “ H, then by definition the empty sum mpzq “ 0. Furthermore, as f
ş f 1 zPZpf q
is holomorphic in a neighbourhood of K, C f dz “ 0, as wanted.
Next, let H ‰ Zpf q “ tζ1 , ¨ ¨ ¨ , ζk u for some k P N. By a repeated application of
the theorem on the classification of zeroes, f “ pz ´ ζ1 qmpζ1 q ¨ ¨ ¨ pz ´ ζk qmpζk q g, for
some g P OpDq such that g is never zero in K. For all z P ran C,
ř
k mpζ qf
f 1 pzq “ ℓ
z´ζℓ ` fg g1 .
ℓ“1
ş f1 ř
k
1
ş ş g1 p˚q ř
k
Hence C f dz “ mpζℓ q
C z´ζℓ dz ` C g dz “ 2πi mpζℓ q ` 0.
ℓ“1 ℓ“1
1 ş 1
Justification of (˚): Since gg is holomorphic in a neighbourhood of K, C gg dz “ 0,
ş 1 1
while C z´ζ ℓ
dz “ 2πi (as ¨´ζ ℓ
P OpCztζℓ uq, and since C is Cztζℓ u-homotopic to
a little circular path centred at ζℓ with a positive radius traversed once in the
counterclockwise direction).
Solution to Exercise 4.47. Let z0 P D be such that |f pz0 q| ě |f pzq| for all z P D.
By the Maximum Modulus Theorem, f is constant in D, a contradiction.
Solution to Exercise 4.48. Let f pz0 q ‰ 0. Then |f pz0 q| ą 0 and so for all z P D,
|f pzq| ě |f pz0 q| ą 0, implying that for all z P D, f pzq ‰ 0. Define g “ f1 P OpDq.
Then |gpz0 q| “ |f pz10 q| ě |f pzq|
1
“ |gpzq| for all z P D, and so by the Maximum
Modulus Theorem, g is constant. But then f is constant too.
Solution to Exercise 4.49. Let z0 be a maximiser, which exists since |f | is
continuous and K :“ tz P C : |z| ď 1u is compact. But z0 cannot be in the interior
of K: Indeed, if |z0 | ă 1, then by the Maximum Modulus Theorem (applied to
f on D :“ tz P C : |z| ă 1u), f would be constant in D, which is not true (e.g.
f p0q “ 0 ‰ 41 ´ 2 “ f p 21 q). Hence z0 P T :“ tz P C : |z| “ 1u. Referring to the
picture below, we see that max |f pzq| “ max |f pzq| “ max |e2it ´ 2| “ | ´ 1 ´ 2| “ 3.
zPK |z|“1 tPr0,2πq
T
e2it
1
´1 0 2
3
1
Similarly, if z1 is a minimiser, z1 cannot be in the interior of K: Indeed, z12 ´ 2 ‰ 0,

and so by the Minimum Modulus Theorem, f would be a constant, a contradiction.
So z1 P T. Hence min |f pzq| “ min |f pzq| “ min |e2it ´ 2| “ |1 ´ 2| “ 1.
zPK |z|“1 tPr0,2πq
184 Solutions
Solution to Exercise 4.50. For z P Dp0, 1q, |f pzq| “ e|z| ą 0, and so f is never
0 on Dp0, 1q. Consider g :“ f1 . Then as f P OpDp0, 1qq, g P OpDp0, 1qq too. For
1 1
z P Dp0, 1q, |z| ě 0, and so |gpzq| “ |f pzq| “ e´|z| ď e0 “ e´|0| “ |f p0q| “ |gp0q|,
and so, by the Maximum Modulus Theorem, g must be a constant. Then f is also
1
a constant. But |f p0q| “ e|0| “ 1 ‰ e 2 “ |f p 12 q|. So we arrive at a contradiction.
Thus there is no f P OpDp0, 1qq such that |f pzq| “ e|z| for all z P Dp0, 1q.
Solution to Exercise 4.51. For z P A1 :“ tz P C : 0 ă |z ´ 1| ă 1u, we have
1 1 1 2 3
zpz´1q “ pz´1`1qpz´1q “ z´1 p1 ´ pz ´ 1q ` pz ´ 1q ´ pz ´ 1q ` ´ ¨ ¨ ¨ q
1
“ ´ 1 ` pz ´ 1q ´ pz ´ 1q2 ` pz ´ 1q3 ´ ` ¨ ¨ ¨ .
z´1
r 1 :“ tz P C : 1 ă |z ´ 1|u,
On the other hand, for z P A
1 1 1 1 1 1
zpz´1q “ pz´1`1qpz´1q “ pz´1q2 p1` 1 q “ pz´1q2 p1 ´ z´1 ` pz´1q2 ´ ` ¨ ¨ ¨ q
z´1
1 1 1 1
“ pz´1q2 ´ pz´1q3 ` pz´1q4 ´ pz´1q5 ` ´¨¨¨ .
Solution to Exercise 4.52. f´w p´zq “ expp ´w 1 w 1
2 p´z´ ´z qq “ expp 2 pz´ z qq “ fw pzq.
8
ř 8
ř
For all z P Czt0u, Jn p´wqp´zqn “ f´w p´zq “ fw pzq “ Jn pwqz n . By the
n“´8 n“´8
uniqueness of coefficients, Jn p´wqp´1qn “ Jn pwq, i.e., Jn p´wq “ p´1qn Jn pwq.
Let C be circular path Cpθq “ eiθ , θ P r0, 2πs. For all n P Z,
1
ş fw pzq ş w iθ
1 2π expp 2 pe é
íθ qq
Jn pwq “ 2πi C pz´0qn`1 dz “ 2πi 0 peiθ qn`1
ieiθ dθ
ş
1 2π exppwi sin θq
ş
1 2π
“ 2π 0 exppinθq dθ “ 2π 0 exppiw sin θ ´ inθq dθ.
1
Solution to Exercise 4.53. e´ z2 P OpCzt0uq, and has the Laurent series
1 8 p´1qn
ř
e´ z2 “ 1 1 1 1 1 1
n! z 2n “ 1 ´ z 2 ` 2! z 4 ´ 3! z 6 ` ´ ¨ ¨ ¨ .
n“0
ş ´ 1 ş 1
The coefficient of z ´6 1
is ´ 3! “ ´ 61 . So ´ 16 “ 1 e z2
2πi C pz´0q´6`1 dz “ 1
2πi C z 5 e´ z2 dz,
ş 5 ´1
giving C z e z2 dz “ ´ πi
3.
Solution to Exercise 4.54. f P OpCzt0uq, and so 0 is an isolated singularity.
We will first show that 0 is not a removable singularity of f by showing it is not
1
the case that lim f pzq exists. If zn :“ logpnπq , n P N, then zn Ñ 0 as n Ñ 8. But
zÑ0
f pzn q “ cospe1{zn q “ cospelogpnπq q “ cospnπq “ p´1qn .
So it is not the case that lim f pzq exists. So 0 is not a removable singularity of f .
zÑ0
We will show 0 is not a pole by showing it is not the case that lim |f pzq| “ `8 :
zÑ0
For all n P N, |f pzn q| “ 1, showing that it cannot be the case that lim |f pzq| “ 8.
zÑ0
So the isolated singularity 0 is essential, as it is neither removable nor a pole.
Solution to Exercise 4.55. If D :“ Czpp´8, 0s ˆ t0uq, then Log P OpDq, and
Log1 z “ z1 . Hence lim Logp1`zq
z “ lim Logp1`zq´Log
p1`zq´1
1
“ Log1 1 “ 11 “ 1. Thus
zÑ0 zÑ0
lim z Logp1`zq
z2
“ 1 ‰ 0. Hence z “ 0 is a pole of order 1 of Logp1`zq
z2
.
zÑ0
Solution to Exercise 4.56. Let z P Z. Write z “ x ` iy, where x, y P R. Then

ez ´ 1 “ 0 implies ex pcos y ` i sin yq “ 1. Taking complex absolute values, ex “ 1,
and so x “ 0. Thus 1pcos yì sin yq “ 1, giving cos y “ 1 and sin y “ 0. So y P 2πZ.
Vice versa, if z “ 0 ` i2πn for some n P Z, then ez “ e0ì2πn “ 1p1 ` i0q “ 1, so
that z P Z. Consequently, Z “ 2πiZ.
For z “ 2πin P Z, n P Z, pez ´ 1q1 |z“2πin “ ez |z“2πin “ 1 ‰ 0, and so each zero in
Z has order 1 as a zero of ez ´ 1.
The numerator z of ezz´1 is entire. Since the denominator ez ´ 1 is nonzero in the
punctured disc Dp0, 2πqzt0u of radius 2π, and is holomorphic there, 0 is an isolated
1 1
singularity of f . We have lim f pzq “ lim ezz´1 “ lim ez é0 “ pez q1 |
z“0
“ e10 “ 1. Thus
zÑ0 zÑ0 zÑ0 z´0
lim pz ´ 0q1f pzq “ 0 ¨ 1 “ 0. By Theorem 4.16, f has a removable singularity at 0.
zÑ0
Let F be the holomorphic extension of f to the disc Dp0, 2πq. By Taylor’s theorem,
F has a power series expansion in Dp0, 2πq, with coefficients, say, cn , n ě 0. In
8
ř
particular, for 0 ă |z| ă 2π “: r, f pzq “ F pzq “ cn z n .
n“0
We define Bn :“ n!cn . Hence B0 “ 1c0 “ F p0q “ lim F pzq “ lim f pzq “ 1.

zÑ0 zÑ0
Using the identity ezz´1 ´ ep´zq

´z ´1 “ ´z, and the series expansion for 0 ă |z| ă r,
8
ř 8
ř
Bn n
´z “ z ´ Bn p´zqn “ 0 ` 2 B1 z 1!
` 0 ` 2 B3!3 z 3 ` 0 ` ¨ ¨ ¨ .
n“0 n! n“0 n!
By the uniqueness of coefficients in the Laurent series expansion, it follows that
B1 “ ´ 12 and the rest are zeroes, that is, 0 “ B3 “ B5 “ ¨ ¨ ¨ .
Solution to Exercise 4.57. For all z P C, we have

z2 z4
cos z ´ 1 “ p1 ´ 2! ` 4! ´ ` ¨ ¨ ¨ q ´ 1 “ z 2 p´ 2!1 ` 1 2
4! z ´ ` ¨ ¨ ¨ q.
1 1 2
If h :“ `
´ 2! ´ ` ¨ ¨ ¨ , then as h is defined by a power series with an infinite
4! z
radius of convergence, h is entire. Also hp0q “ ´ 12 ‰ 0. Thus h is nonzero in
1
a disc centred at 0. So 0 is an isolated singularity of cos 1z´1 “ z 2 hpzq . Moreover
lim z 2 cos 1z´1 “ lim 1
“ ´2 ‰ 0, and lim z 3 cos 1z´1 “ 0p´2q “ 0. By the theorem
zÑ0 zÑ0 hpzq zÑ0
on classification of singularities via limits, 0 is a pole of f , and its order is 2.
Solution to Exercise 4.58. By Theorem 4.11 (classification of zeroes), we have

f pzq “ pz ´ z0 qm gpzq, z P D, where g P OpDq and gpz0 q ‰ 0. Hence g is nonzero
in a disc Dpz0 , rq Ă D for some r ą 0. Then f pzq “ pz ´ z0 qm gpzq ‰ 0 for all
z P Dpz0 , rqztz0 u, and so f1 is well-defined in Dpz0 , rqztz0 u. As g1 P OpDpz0 , rqq, it
8
ř
1 1
has a Taylor expansion in Dpz0 , rq: gpzq “ cn pz ´ z0 qn , and c0 “ gpz0 q ‰ 0. Thus
n“0
8
ř cm´1 8
ř
1
f pzq “ pz´z01qm gpzq “ pz´z10 qm cn pz ´ z0 qn “ c0
pz´z0 qm `¨¨¨ ` z´z0 ` cm`n pz ´ z0 qn
n“0 n“0
for 0 ă |z ´ z0 | ă r. So f1 has a pole of order m at z0 . (Alternatively, use the
1
classification of singularities via limits, noting that lim pz ´ z0 qm f pzq “ gpz10 q ‰ 0.)
zÑz0
186 Solutions
Solution to Exercise 4.59. The function f has a Laurent series expansion in

c´m c´1
Dpz0 , rqztz0 u: f pzq “ pz´z 0q
m ` ¨ ¨ ¨ ` z´z ` c0 ` c1 pz ´ z0 q ` ¨ ¨ ¨ , where c´m ‰ 0.
0
Hence pz ´ z0 qm f “ c´m ` c´m`1 pz ´ z0 q ` ¨ ¨ ¨ for all z P Dpz0 , rqztz0 u. So
h :“ c´m ` c´m`1 pz ´ z0 q ` ¨ ¨ ¨ P OpDpz0 , rqq, and hpz0 q “ c´m ‰ 0. Since for
all z P Dpz0 , rqztz0 u, hpzq “ pz ´ z0 qm f pzq, and as f pzq ‰ 0, also hpzq ‰ 0. Thus
hpzq ‰ 0 for z P Dpz0 , rq. So h1 P OpDpz0 , rqq. Define g :“ pz´z0 qm h1 P OpDpz0 , rqq.
Then g|Dpz0 ,rqztz0 u “ f1 , and by Exercise 4.30, z0 is a zero of g of order m .
Solution to Exercise 4.60. By the classification of singularities via Laurent series

coefficients, cn “ 0 for all n ă ´m. So for all z P Dpz0 , rqztz0 u,
c´m c´1 8
ř
f pzq “ pz´z 0q
m ` ¨ ¨ ¨ ` z´z `
0
cn pz ´ z0 qn .
n“0
Thus pz ´ z0 qm f pzq “ c´m ` c´m`1 pz ´ z0 q ` ¨ ¨ ¨ for all z P Dpz0 , rqztz0 u. Let
g P OpDpz0 , rqq be defined by gpzq :“ c´m `c´m`1 pz´z0 q`¨ ¨ ¨`c´1 pz´z0 qm´1 `¨ ¨ ¨
1 dm´1 g
for all z P Dpz0 , rq. By Taylor’s Theorem, c´1 “ pm´1q! dz m´1 pz0 q. But the function
g pm´1q P OpDpz0 , rqq and in particular, it is continuous at z0 . So
g pm´1q pz0 q “ lim g pm´1q pzq.
zÑz0
dm´1
For z P Dpz0 , rqztz0 u, gpzq “ pz ´ z0 qm f pzq. So gpm´1q pzq “ dz m´1
ppz ´ z0 qm f pzqq.
g pm´1q pz0 q 1 1 m´1
Hence c´1 “ pm´1q! “ lim gpm´1q pzq
pm´1q! zÑz “ lim ddz m´1 ppz ´ z0 qm f pzqq.
pm´1q! zÑz
0 0

(1) True, since c´1 “ 1 ‰ 0, and c´2 “ c´3 “ ¨ ¨ ¨ “ 0.
(2) True.
(3) True.
(4) True.
3 5
(1) sin z does not have a singularity at 0, and for z P C, sin z “ z ´ z3! ` z5! ´ ` ¨ ¨ ¨ .
(2) sin z1 has an essential singularity at 0: For z ‰ 0, sin 1z “ ¨ ¨ ¨ ´ ` 5!z1 5 ´ 3!z1 3 ` 1z .
sin z 1 2 1 4
(3) As z “1´ 3! z ` 5! z ´ ` ¨ ¨ ¨ for z ‰ 0, 0 is a removable singularity.
sin z sin z 1 z 3 5
(4) z2
has a pole of order 1 at 0, since for z ‰ 0, z2
“ z ´ 3! ` z5! ´ z7! ` ´ ¨ ¨ ¨ .
1
(5) sin z1
does not have an isolated singularity at 0 (see Example 4.15).
(6) As z sin 1z “ ¨ ¨ ¨ ´ ` 5!z1 4 ´ 1
3!z 2
` 1 for z ‰ 0, 0 is an essential singularity.
Solution to Exercise 4.63. Let R ą 0 be such that f P OpDpz0 , Rqztz0 uq. By
the theorem on classification of singularities via Laurent series coefficients, f has a
c´m c´1 8
ř
Laurent series f pzq “ pz´z 0q
m ` ¨ ¨ ¨ ` z´z `
0
cn pz ´ z0 qn in Dpz0 , Rqztz0 u, where
n“0
c´m ‰ 0. Thus, pz ´z0 qm f pzq “ c´m `c´m`1 pz ´z0 q`¨ ¨ ¨ for all z P Dpz0 , Rqztz0 u.
Define g P OpDpz0 , Rqq by gpzq “ c´m ` c´m`1 pz ´ z0 q ` ¨ ¨ ¨ ` c´1 pz ´ z0 qm´1 ` ¨ ¨ ¨
for all z P Dpz0 , rq. Then lim pz ´ z0 qm f pzq “ lim gpzq “ gpz0 q “ c´m ‰ 0.
zÑz0 zÑz0

1 1
(1) False. lim |e x | “ lim e x “ 0, and so �plim | exp 1z | “ `8q.
xÑ0´ xÑ0´ zÑ0
(2) True. There exists an R ą 0 such that for 0 ă |z ´ z0 | ă R
c´m c´m`1 c´1 8
ř
f pzq “ pz´z 0q
m ` pz´z qm´1 ` ¨ ¨ ¨ ` z´z `
0 0
cn pz ´ z0 qn ,
n“0
and so with p :“ c´m ` c´m`1 pz ´ z0 q ` ¨ ¨ ¨ ` c´1 pz ´ z0 qm´1 , we have for
ppzq 8
ř
0 ă |z ´ z0 | ă R that f pzq ´ pz´z 0q
m “ cn pz ´ z0 qn .
n“0
(3) True. Let the order of 0 as a zero of f be m P N. Set m “ 0 if f p0q ‰ 0. Then

there exists a holomorphic function g such that f pzq “ z m gpzq and gp0q ‰ 0.
z m gpzq
For n ą m, and z ‰ 0, fzpzq
n “ zn “ zgpzq
n´m . Using gp0q ‰ 0 and n ą m,
lim | fzpzq
n | “ lim
|gpzq|
n´m “ |gp0q| lim 1
n´m “ `8.
zÑ0 zÑ0 |z| zÑ0 |z|
(4) True. In some punctured disc Dpz0 , rqztz0 u, r ą 0, f, g are nonzero, and by
Exercise 4.59, there exist hf , hg P OpDpz0 , rqq such that hf pz0 q ‰ 0, hg pz0 q ‰ 0,
1 1
and for all z P Dpz0 , rqztz0 u, f pzq “ pz ´ z0 qmf hf pzq, gpzq “ pz ´ z0 qmg hg pzq.
1
So hf pz0 qhg pz0 q ‰ 0, and @z P Dpz0 , rqztz0 u, f pzqgpzq “ pz´z0 qmf `mg hf pzqhg pzq.
Thus f g has a pole of order mf ` mg at z0 , by Exercise 4.58. (Alternatively,
by Exercise 4.63, lim pz ´ z0 qmf f pzq “: α ‰ 0 and lim pz ´ z0 qmg gpzq “: β ‰ 0,
zÑz0 zÑz0
giving lim pz ´ z0 qmf `mg f pzqgpzq “ αβ ‰ 0, and so by Corollary 4.17, z0 is a
zÑz0
pole of f g of order mf ` mg .)
Solution to Exercise 4.65. Define f by f pzq “ pexp z1 q ` exp z´1 1
, z P Czt0, 1u.
1
Then f P OpCzt0, 1uq. The function exp z´1 is holomorphic in the neighbourhood
Dp0, 1q of z “ 0, while the function exp z1 has an essential singularity at 0. Thus,
considering the Laurent series expansion of their sum f in the punctured disc
Dp0, 1qzt0u, f has an essential singularity at 0 (because the negatively indexed
ř 1
coefficients in f “ cn z n are given by cń “ n! ‰ 0, n P N).
nPZ
Similarly, as exp 1z is holomorphic in the neighbourhood Dp1, 1q of 1, and as
1
exp z´1 has an essential singularity at 1, f has an essential singularity at z “ 1 (as
ř 1
the negatively indexed coefficients in f “ cn pz ´ 1qn are cń “ n! ‰ 0, n P N).
nPZ
Solution to Exercise 4.66. If z0 is an isolated singularity of a function g with

ř
the Laurent series expansion gpzq “ cn pz ´ z0 qn for 0 ă |z ´ z0 | ă R, where
nPZ
R ą 0, and there are infinitely many indices n ă 0 such that cn ‰ 0, then z0 is an
essential singularity of g.
However, for the given f , the annulus for the Laurent expansion z ´1 `z ´2 `z ´3 `¨ ¨ ¨
is given by |z| ą 1. The correct annulus to consider for deciding the nature of the
singularity at z “ 0 is of the form 0 ă |z| ă R for some R ą 0. In fact, for |z| ă 1
1
we have f pzq “ ´ 1´z “ ´p1 ` z ` z 2 ` z 3 ` ¨ ¨ ¨ q, showing that f is holomorphic
for |z| ă 1, and f does not have a singularity at z “ 0.
188 Solutions
Solution to Exercise 4.67. z0 is an isolated singularity of f g: As f, g have an iso-

lated singularity at z0 , f P OpDpz0 , rf qztz0 uq for an rf ą 0, and g P OpDpz0 , rg qztz0 uq
for an rg ą 0, and so f g P OpDpz0 , rqztz0 uq where r :“ mintrf , rg u.
If f g has a removable singularity or a pole at z0 , then there exists an m P N such
that lim pz ´ z0 qm f pzqgpzq “ 0 . Since f has a pole at z0 , say of order ℓ P N, f is
zÑz0 8
ř
c´ℓ c´ℓ`1 c´1
nonzero near z0 . Also, f pzq “ pz´z0 qℓ
` pz´z0 qℓ´1
` ¨¨¨ ` z´z0 ` cn pz ´ z0 qn , for
n“0
0 ă |z ´ z0 | ă rf , and c´ℓ ‰ 0. So for z ‰ z0 , but near z0 , we have
1 zÑz 1
pz ´ z0 qm gpzq “ pz´z0 qℓ f pzq looomooon pz ´ z0 qm f pzqgpzq ÝÑ0
pz ´ z0 qℓ loooooooooomoooooooooon c´ℓ 0 ¨ 0 “ 0.
Ñ0 Ñ0
So g has a pole at z0 or a removable singularity at z0 , a contradiction. Consequently,
f g has an essential singularity at z0 .
Solution to Exercise 4.68. For n P N, set ǫ :“ n1 “: δ (ą 0). By the Casorati-

Weierstrass Theorem, there exists a zn in the punctured disc Dpz0 , δqztz0 u, such
that |f pzn q ´ w| ă ǫ. So for all n P N, |zn ´ z0 | ă δ “ n1 and |f pzn q ´ w| ă ǫ “ n1 .
So pzn qnPN converges to z0 , and pf pzn qqnPN converges to w.
8
ř
Solution to Exercise 4.69. Let the Taylor series of f be f pzq “ cn z n , z P C.
n“0
As f is not a polynomial, infinitely many cn are nonzero. Define g P OpCzt0uq by
8
ř
gpζq “ f p 1ζ q for all ζ P Czt0u . Then gpζq “ cn ζ ń for all ζ P Czt0u, and so g
n“0
has an essential singularity at 0. By Exercise 4.68, there exists a sequence pζn qnPN
in Czt0u which converges to 0, while pgpζn qqnPN converges to w. If zn :“ ζ1n , n P N,
then pzn qnPN converges to 8, and pf pzn qqnPN “ pgpζn qqnPN converges to w.
Solution to Exercise 4.70. 1èxp z “ 0 ô z P tπi`2πni : n P Zu “: Z. Each zero

z P Z of 1 ` exp has order 1, since p1 ` expq1 pπi ` 2πniq “ exppπiq “ ´1 ‰ 0. The
Log z
function f defined by f pzq :“ 1èxp q
z is holomorphic in p Czpp´8, 0s ˆ t0uqq zZ, and
has singularities only at the points of Z, each of which is an isolated singularity,
and in fact each is a pole of order 1 (easily checked using the classification of
singularities via limits). Of these poles, exactly two lie inside the given path γ:
´πi and 3πi.
3πi
γ1
´πi
γ2
ş ş ş
We have γ f pzq dz “ γ1 f pzq dz ` γ2 f pzq dz “ 2πipRespf, 3πiq ´ Respf, ´πiqq.
Log z c´1,3πi
Writing 1èxp z “ z´3πi ` h3πi , where h3πi P OpDp3πi, rqq for some r ą 0, we have
pz´3πiqLog z
c´1,3πi “ lim 1èxp z “ lim z´3πi
Log z “ exp11p3πiq lim Log z
zÑ3πi zÑ3πi exp zéxpp3πiq zÑ3πi
1
“ expp3πiq Logp3πiq “ ´plog |3πi| ` i π2 q “ ´ log 3 ´ log π ´ i π2 .
Log z c´1,´πi
Writing 1èxp z “ z´p´πiq ` h´πi , where h´πi P OpDp´πi, rrqq for some rr ą 0,
pz´p´πiqqLog z
c´1,´πi “ lim 1èxp z “ lim exp z´p´πiq 1
zéxpp´πiq Log z “ exp1 p´πiq zÑ´πi
lim Log z
zÑ´πi zÑ´πi
1 π π
“ expp´πiq Logp´πiq “ ´plog | ´ πi| ` ip´ 2 qq “ ´ log π ` i 2 .
ş Log z
So γ 1èxp z dz “ 2πip´ log 3 ´ log π ´ i π2 ` log π ´ i π2 q “ 2π 2 ´ p2π log 3qi.
Solution to Exercise 4.71. Let Cptq “ eit , t P r´π, πs. Then

şπ 1
şπ 1 ieiθ
ş 1 1
ş 1 1
´π 2`cos θ dθ “ ´π 2` eiθ èíθ ieiθ dθ “ C 2` z`1{z iz dz “ C 2` z 2 `1 iz dz
ş 2z
2
1 2
ş 1
2
2
ş 1
2z
“ C 4z`z 2 `1 iz dz “ i C z 2 `4z`1 dz “ i C pz´p qpz´p q dz,

1 2
? ?
where p1 :“ ´2 ` 3, p2 :“ ´2 ´ 3. As |p1 | ă 1 ă |p2 |, by the Residue Theorem,
şπ 1 2 1 1 4π 2π
´π 2`cos θ dθ “ i 2πi Resp pz´p1 qpz´p2 q , p1 q “ 4π lim pz´p1 q pz´p1 qpz´p2 q “ p1 ´p2 “ 3 .
?
zÑp1
Solution to Exercise 4.72. Let γpθq :“ eiθ for θ P r0, 2πq. Then
ş2π cos θ ş z` 1
z ş ş
2 z 2 `1
0 5`4 cos θ dθ “ γ
2 1
z` 1 iz
dz “ γ 2izp2zz 2`1
`5z`2q dz “ γ 2izp2z`1qpz`2q dz.
5`4 z
2
z 2 `1
Let f pzq :“ 2izp2z`1qpz`2q . Then f has three poles, at 0, ´ 12 , ´2, and each is of
order 1 (which can be seen using the classification of singularities via limits). Of
these, the poles at 0 and ´ 21 lie inside γ. So by the Residue Theorem,
ş2π cos θ 1
0 5`4 cos θ dθ “ 2πipRespf, 0q ` Respf, ´ 2 qq
zpz 2 `1q pz` 21 qpz 2 `1q
“ 2πiplim ` lim 2izp2z`1qpz`2q q
zÑ0 2izp2z`1qpz`2q zÑ´ 21
1 1¨ 5 1 5
“ 2πip 2i¨1¨2 ` 2i¨p´ 14q¨2¨ 3 q “ 2πip 4i ´ 12i q “ ´ π3 .
2 2

2 n
For z ‰ 0, exp z
z n`1
1
“ z n`1 1
p1 ` 1!z ` z2! ` ¨ ¨ ¨ ` zn! ` ¨ ¨ ¨ q “ z n`1 ` ¨ ¨ ¨ ` n!1 1
z ` ¨¨¨.
exp z 1 exp z
Thus Resp z n`1 , 0q “ n! . Also, 0 is a pole of order n ` 1 of z n`1 . By the Residue
ş
Theorem, C zexp z exp z 2πi
n`1 dz “ 2πi Resp z n`1 , 0q “ n! . Hence
2πi
ş2π ecos θì sin θ iθ ş2π
n! “ 0 peiθ qn`1
ie dθ “ i 0 ecos θ ei sin θ eínθ dθ
ş2π
“ i 0 ecos θ pcospnθ ´ sin θq ´ i sinpnθ ´ sin θqqdθ.
ş2π
Equating the imaginary parts, we obtain 0 ecos θ cospnθ ´ sin θqdθ “ 2π n! .
190 Solutions
Solution to Exercise 4.74. For z in a punctured disc Dpz0 , rqztz0 u, r ą 0,

f pzq ‰ 0 and f pzq “ pz ´ z0 qhpzq for some h P OpDpz0 , rqq such that hpz0 q ‰ 0.
From f “ pz ´ z0 qh, we have f 1 pzq “ hpzq ` pz ´ z0 qh1 pzq, and so f 1 pz0 q “ hpz0 q. As
1 1 1 1
h P OpDpz0 , rqq, hpzq “ d0 `d1 pz ´z0 q`¨ ¨ ¨ for z P Dpz0 , rq with d0 “ hpz0 q “ f 1 pz0 q .
1 1 d0
So f pzq “ z´z 0
pd0 ` d1 pz ´ z0 q ` ¨ ¨ ¨ q “ z´z0 ` d1 ` ¨ ¨ ¨ for z P Dpz0 , rqztz0 u. Hence
1 1
Resp f , z0 q “ d0 “ f 1 pz 0q
.
Solution to Exercise 4.75. The set of zeroes of sin is tkπ : k P Zu, and each zero
has order 1. By Exercise 4.74, Resp sin1 z , kπq “ sin1 z|1 1
“ cospkπq 1
“ p´1q k
k “ p´1q .
z“kπ
Solution to Exercise 4.76. Let C be given by Cptq “ eit for all t P r0, 2πs. Then
ş2π ş2π eit éít 2n 1 it ş ´1
2n 1 1
ş 1 2n 1
2n
0 psin tq dt “ 0 p 2i q ieit ie dt “ C p z´z
2i q iz dz “ 22n p´1qn i C pz´ z q z dz.
Define the function f P OpCzt0uq by f pzq :“ pz ´ 1z q2n 1z , z P Czt0u. By the
` ˘
Binomial Theorem, f pzq “ pz 2n ` ¨ ¨ ¨ ` p´1qn 2n 1 1
` ¨ ¨ ¨ ` z 2n q z , z P Czt0u. So
n
`n2n˘
f has a pole at z “ 0. Also, Respf, 0q “ p´1q n . The Residue Theorem gives
ş ş2π 2n 1
` ˘
n 2n “ π
`2n˘
C f pzqdz “ 2πi Respf, 0q. So 0 psin tq dt “ 22n p´1qn i 2πip´1q n 22n´1 n
.
(1) For z P Czt0u, we have
g 2 p0q 2 g 1 p0q g 2 p0q
gpzqp1` z12 q “ pgp0q ` g1 p0qz ` 2! z ` ¨ ¨ ¨ q ` p gp0q
z2 ` z ` 2! ` ¨ ¨ ¨ q,
and so Respgpzqp1` z12 q, 0q “ coefficient of z ´1 “ g 1 p0q.
(2) Let Cptq “ eit , t P r0, 2πs. Then by the Residue Theorem,
ş2π ş2π eit èít 1
ş 1 z` z1
0 gpeit q cos t dt “ 0 gpe q
it
2 ie it ieit dt “
C iz dz
gpzq 2
1
ş 1 1 1
“ 2i C gpzqp1` z 2 q dz “ 2i 2πi Respgpzqp1 ` z 2 q, 0q
1 1 1
“ 2i 2πi g p0q “ πg p0q.
Alternatively, one could use the Cauchy Integral Formula:
ş2π it 1
ş 1 1
ş ş gpzq
0 gpe q cos t dt “ 2i C gpzqp1` z 2 q dz “ 2i p C gpzq dz ` C z2
dzq
1
“ 2i p0 ` 2πi g1 p0qq “ πg 1 p0q.
(1) f0 “ 1 ď 20 “ 1, f1 “ 1 ď 21 “ 2. If for some n ě 1, fm ď 2m for all m ď n, then
fn`1 “ fn ` fn´1 ď 2n ` 2n´1 “ 2n´1 3 ă 2n´1 4 “ 2n`1 .
a a ?
(2) For |z| ă 21 , n |fn z n | “ n |fn | |z| ď n 2n |z| “ 2|z| ă 1 for all n P N. By the Root
8
ř
Test, |fn z n | converges if |z| ă 21 . So the radius of convergence of F is ě 12 .
n“0
(3) For |z| ă 21 , zF pzq “ f0 z ` f1 z 2 ` f2 z 3 ` ¨ ¨ ¨ and z 2 F pzq “ f0 z 2 ` f1 z 3 ` ¨ ¨ ¨ .
Thus zF pzq`z 2 F pzq “ 1 z ` pf1 `f0 qz 2 ` pf2 `f1 qz 3 ` ¨ ¨ ¨
“ f1 z `f2 z 2 `f3 z 3 ` ¨ ¨ ¨
“ pf0 ` f1 z ` f2 z 2 ` f3 z 3 ` ¨ ¨ ¨ q ´ f0 “ F pzq ´ 1.
So for |z| ă 12 , 1 “ F pzq´zF pzq´z 2 F pzq “ p1´z´z 2 qF pzq, and F pzq “ 1
1´z´z 2 .
F pzq f0 `¨¨¨`fn´1 z n´1 `fn z n `fn`1 z n`1 `¨¨¨
(4) For |z| ă 21 , 1
z n`1 p1´z´z 2 q
“ z n`1
“ z n`1
f0 f1 fn
“ z n`1 ` z n ` ¨ ¨ ¨ ` z ` fn`1 ` fn`2 z ` ¨ ¨ ¨ ,
1
and so Resp z n`1 p1´z´z 2 q , 0q “ fn .
(5) For |z| “ R ą 2 : |1 ´ z ´ z 2 | ě |z 2 ` z| ´ 1 “ |z||z ` 1| ´ 1 “ R|z ` 1| ´ 1
ě Rp|z|´1q´1 “ RpR´1q´1 “ R2 ´R´1
ą 2R´R´1 “ R´1 ą 2 ´ 1 ą 0.
it
ş 1 1 1 RÑ8
With CR ptq :“ Re , t P r0, 2πs, | CR z n`1 p1´z´z 2 q dz| ď Rn`1 R2 ´R´1 2πR ÝÑ 0.
?
1 ´1` 5
If Gpzq :“ z n`1 p1´z´z 2 q , then G has a pole at 0 of order n`1, a pole at 2
?
of order 1, and a pole? at ´1´2 5 of order
?
1. The Residue Theorem yields
´1` 5 ´1´ 5 1
ş
RespG, 0q ` RespG, 2 q ` RespG, 2 q “ 2πi CR Gpzqdz for R ą 2. Letting
? ? ş
´1` 5 ´1´ 5 1
R Ñ 8, RespG, 0q ` RespG, 2 q ` RespG, 2 q “ lim 2πi CR Gpzqdz “ 0,
? ? RÑ0 ?
i.e., fn “
?
RespG, 0q “ ´RespG, ´1`2 5
q ´ RespG, ´1´2 5
5`1
q. If τ “
2 , then we
? n`1
5´1
get τ1 “ 2 . So RespG, ´1`2 5 q “ lim pz´ τ1 q z n`1 p1´z´z
1
2q “ 1 ?
p τ1 qn`1 p´ 5q
“ ´ τ ?5 .
zÑ τ1
? ?
Also, RespG, ´1´2 5 q “ ´1´?51 n`1 ? “ p 1´2 5 qn`1 p ?15 q. Hence we obtain that
p 2 q 5
? ? ? ?
fn “ ?15 p 1`2 5 qn`1 ´ ?15 p 1´2 5 qn`1 “ ?15 pp 1`2 5 qn`1 ´ p 1´2 5 qn`1 q.

2π
(1) The poles of f are at the zeroes of qpzq :“ z 2022 ´ 1, at pk :“ eip0` 2022 kq ,
k “ 0, 1, ¨ ¨ ¨ , 2021. These pk are distinct, lying at the vertices of a regular
2022-gon inscribed in the unit circle, with a vertex at 1. We have
1 1
lim pz ´ pk qf pzq “ lim z2022 ´p 2022 “ dz 2022 “ 2022p1 2021
zÑpk zÑpk k dz
|z“pk k
z´pk
pk pk pk
“ 2022p2022
“ 2022¨1 “ 2022 ‰ 0.
k
So lim pz ´ pk q2 f pzq “ 0. This pk is a pole of f or order 1, k “ 0, ¨ ¨ ¨ , 2021.

zÑpk
(2) Near each pole pk (in a punctured disc Dppk , rqztpk u, rk ą 0), f has a Laurent
pkq
c´1 pkq pkq pkq
series f pzq “ z´pk ` c0 ` ¨ ¨ ¨ . So pz ´ pk qf pzq “ c´1 ` c0 pz ´ pk q ` ¨ ¨ ¨ .
pkq pk
Hence Respf, pk q “ c´1 “ lim pz ´ pk qf pzq “ 2022 by the calculation above.
zÑpk
ş ř
2021
2πi ř
2021
(3) By the Residue Theorem, C f pzqdz “ 2πi Respf, pk q “ 2022 pk . But the
k“0 k“0
even number of points pk , k “ 0, ¨ ¨ ¨ , 2021, are symmetrically arranged on the
unit circle (with p0 “ ´p1011 , p1 “ ´p1012 , ¨ ¨ ¨ , p1010 “ ´p2021 ), and so they
ś
2021
all add up to 0. (Alternatively, z 2022 ´ 1 “ pz ´ pk q, and as the coefficient of
k“0
z 2021 on the left-hand side is 0, the sum of the roots pk is 0. See Exercise 1.29.)
192 Solutions

1
(1) Let f1 pzq “ 1`z 2 . Then f1 has poles at i and í, both of order 1. Hence
ş8 1 1 zí 1 1 π
0 1`x2 dx “ 2 2πi Respf1 , iq “ πi lim 1`z 2 “ πi lim zì “ πi 2i “ 2 .
zÑi zÑi
1
(2) Let f2 pzq :“ pa2 `z 2 qpb2 `z 2 q . Then f2 has poles at ai, ái, bi, ´bi, all of order 1.
ş8 1 1
As f2 is even, 0 pa2 `x2 qpb 2 `x2 q dx “ 2 2πipRespf2 , aiq ` Respf2 , biqq
“ πip pb2 á12 q2ai ` pa2 ´b12 q2bi q “ 2abpa`bq

π
.
(3) Let f3 pzq “ p1`z1 2 q2 . Then f3 has poles at i and í, both of order 2. We have
ş8 1 1
0 p1`x2 q2 dx “ 2 2πi Respf3 , iq “ 1! zÑi
πi d
lim dz ppz ´ iq2 pzíq21pzìq2 q
´2 ´2 π
“ πi lim pzìq 3 “ πi ´8i “ 4 .
zÑi
2 πi 3πi 5πi 7πi
(4) Let f4 pzq “ 1`z
1`z 4 . Then f4 has poles at p1 “ e 4 , p2 “ e 4 , p3 “ e 4 , p4 “ e 4 ,
?
all of order 1. We have p1 `p2 “ 2i and p1 p2 “ ´1. By the Residue Theorem,
ş8 1`x2 1 1`p21 1`p22
0 1`x4 dx “ 2 2πipRespf4 , p1 q ` Respf4 , p2 qq “ πip 4p3 ` 4p3 q 1 2
“ p1
πip 4p 1
1
p2
4 ` 4p ` 4p4 ` 4p q
1
2
“ πip´ p1 `p
4
2
` 1
4p1 ` 1
4p2 q
1 2
? ?
2i 2i π
“ πip´ 4 ` 4p´1q q “ ?2 .
íξz ξ íξz e´ξ
Solution to Exercise 4.81. lim pz íq e1`z 2 “ e2i ‰ 0, lim pz ` iq e1`z 2 “ ´2i ‰ 0.
zÑi zÑí
eíξz
So by Corollary 4.17, f pzq :“ 1`z 2
has two poles, at i and at í, both of order 1.
1˝ ξ ă 0. Let γ` be a path comprising a semicircle σ` centred at 0 of radius r ą 0
in the upper half-plane, together with a line segment path from ´r to r, as on
ş şr ş ξ
p. 110. Then σ` f pzqdz` ´r f pxqdx “ γ` f pzqdz “ 2πi Respf, iq “ 2πi e2i “ πeξ .
it it
For r ą 1 and t P r0, πs, |1 ` r 2 e2it | ě r 2 ´ 1, |eíξe | “ eRepíξre q “ eξr sin t ď 1.
ş íξz
it
|eíξre | 1 rÑ8
Thus | σ` e1`z 2 dz| ď πr max |1`r 2 e2it | ď πr r 2 ´1 ÝÑ 0. So if ξ ă 0, then
tPr0,πs
ş8 eíξx
şr
´8 1`x2 dx “ lim f pxqdx “ πeξ .
rÑ8 ´r
2˝ ξ ě 0. Let γ´ be a path comprising a semicircle σ´ centred at 0 of radius
r ą 0 starting at ´r and ending at r in the lower half-plane, together with a
line segment path from r to ´r. Then
ş şr ş e´ξ ´ξ
σ´ f pzqdz ´ ´r f pxqdx “ γ´ f pzqdz “ 2πi Respf, íq “ 2πi ´2i “ ´πe .
it it
For r ą 1 and t P r´π, 0s, |1`r 2 e2it | ě r 2 ´1, |eíξre | “ eRepíξre q “ eξr sin t ď 1.
ş íξz
it
|eíξre | 1 rÑ8
Thus | σ´ e1`z 2 dz| ď πr max |1`r 2 e2it | ď πr r 2 ´1 ÝÑ 0. So if ξ ě 0, then
tPr´π,0s
ş8 eíξx
şr
´8 1`x2 dx “ lim f pxqdx “ ´p´πe´ξ q “ πe´ξ .
rÑ8 ´r
ş8 eíξx
Summarising, ´8 1`x2 dx “ πe´|ξ| .

(1) lim z 3 f pzq “ lim z 3 z 2cospπzq 1
sinpπzq “ lim cospπzq sinpπzq´sin 0
1
π “ pcos 0q sin11 0 π1 “ 1
π ‰ 0.
zÑ0 zÑ0 zÑ0 πz´0
So 0 is a pole of f of order 3.
cospπzq c´3 c´2 c´1
Let z 2 sinpπzq
“ z3
` z2
` z ` c0 ` c1 z ` ¨ ¨ ¨ in Dp0, 1qzt0u. Then
c´3 “ lim z 3 z 2cospπzq
sinpπzq
“ 1
π (as found above).
zÑ0
Thus z 2 p z 2cospπzq
sinpπzq
´ 1
πz 3
q “ c´2 ` c´1 z ` ¨ ¨ ¨ in Dp0, 1qzt0u, so that
πz cospπzq´sinpπzq
c´2 “ lim z 2 p z 2cospπzq

sinpπzq
´ 1
πz 3
q “ lim πz cospπzq´sinpπzq
πz sinpπzq “ lim pπzq2
sinpπzq .
zÑ0 zÑ0 zÑ0 πz
For z P Dp0, 1qzt0u,

2 4 3 5
πz cospπzq´sinpπzq πzp1´ pπzq ` pπzq ´¨¨¨ q´pπz´ pπzq ` pπzq ´¨¨¨ q pπzq3
pπzq2
“ 2! 4!
pπzq2
3! 5!
“ ´ πz
3 ` 30 ` ¨¨¨ .
Thus lim pπzq2
“ 0, and so
zÑ0
πz cospπzq´sinpπzq πz cospπzq´sinpπzq
lim
pπzq2 zÑ0 pπzq2 0
c´2 “ lim sinpπzq “ “ 1 “ 0.
zÑ0 πz
lim sinpπzq
πz
zÑ0
Hence zp z 2cospπzq
sinpπzq
´ 1
πz 3
q “ c´1 ` c0 z ` ¨ ¨ ¨ in Dp0, 1qzt0u, so that
c´1 “ lim zp z 2cospπzq
sinpπzq
´ 1
πz 3
q “ lim π2 z3
sinpπzq .
zÑ0 zÑ0 πz
πz cospπzq´sinpπzq π3 z2
For z P Dp0, 1qzt0u, as found above, π2 z3
“ ´ π3 ` 30 ` ¨¨¨ .
Thus lim πz cospπzq´sinpπzq
π2 z3
“ ´ π3 . So
zÑ0
´ π3
Respf, 0q “ c´1 “ lim π2 z3
sinpπzq “ 1 “ ´ π3 .
zÑ0 πz
(2) For n P Zzt0u,

lim pz ´ nqf pzq “ lim pz ´ nq z 2cospπzq
sinpπzq
“ lim cospπzq
z2
1
sinpπzq´sinpnπq
zÑn zÑn zÑn zń
cospπnq 1 p´1qn 1 1
“ n2 d
sinpπzq|z“n
“ n2 πp´1qn “ n2 π ‰ 0.
dz
So n is a pole of f of order 1.
cospπzq c´1
Let z 2 sinpπzq
“ zń ` c0 ` c1 pz ´ nq ` ¨ ¨ ¨ in Dpn, 1qztnu. Then
c´1 “ lim pz ´ nq z 2cospπzq
sinpπzq “
1
n2 π (as found above).
zÑn
(3) We have | cospπzq| “ | sinpiπyq| because
cospπzq “ cospπp˘pN ` 12 q ` iyqq
“ cosp˘pπN ` π2 qq cospiπyq ´ sinp˘pπN ` π2 qq sinpiπyq
“ 0 ´ ˘p´1qN sinpiπyq.
194 Solutions
Similarly,
sinpπzq “ sinpπp˘pN ` 12 q ` iyqq
“ sinp˘pπN ` π2 qq cospiπyq ` cosp˘pπN ` π2 qq sinpiπyq
“ ˘p´1qN cospiπyq ` 0,
and so | sinpπzq| “ | cospiπyq|.
(4) The length of each side of the square is pN ` 21 q|p1 ` iq ´ p1 ´ iq| “ 2N ` 1.
M
So the length of SN is 4p2N ` 1q. Let M :“ maxtA, 1u. Then |f pzq| ď |z| 2 for
all z P SN . Also, for each z P SN , |z| ě maxt|Re z|, |Im z|u ě N ` 12 ą N . So

ş M M N Ñ8
| SN f pzq dz| ď 4p2N ` 1q max |z| 2 ď 4p2N ` 1q N 2 ÝÑ 0.
zPSN
(5) By the Residue Theorem,

ş ř ř
N
1
SN f pzq dz “ 2πi Respf, nq “ 2πip´ π3 ` 2 n2 π
q.
|n|ďN n“1
8
ř 1 π2
Passing to the limit as N Ñ 8, and rearranging, we get n2
“ 6 .
n“1

Bu 2x B2 u 2py 2 ´x2 q
(1) We have for px, yq P R2 ztp0, 0qu that Bx “ x2 `y 2Bx2 “ px2 `y 2 q2 . Similarly,
and
2y B2 u 2px2 ý 2 q
using the symmetry in x and y, Bu By “ x2 `y 2 and By 2 “ px2 `y 2 q2 . Consequently
2 ´x2 q 2px2 ý 2 q
B2 u
` BByu2 “ 2py
2
Bx2 px2 `y 2 q2
` px2 `y 2 q2
“ 0. Since u is C 2 pR2 ztp0, 0quq and ∆u “ 0 in
2
R ztp0, 0qu, u is harmonic there.
B2 u B2 u
(2) We have Bu x x Bu x x
Bx “ e sin y, Bx2 “ e sin y, and By “ e cos y, By 2 “ e p´ sin yq in R2 .
2 2
So BBxu2 ` BByu2 “ ex sin y ` ex p´ sin yq “ 0. Since u is C 2 pR2 q and ∆u “ 0 in R2 ,
u is harmonic in R2 .
Solution to Exercise 5.2. Let V be the real vector space of all real-valued func-
tions defined on U , with pointwise operations. We will show that HpU q is a subspace
of this vector space V , and hence a vector space with pointwise operations:
(1) The constant function 0 assuming value 0 everywhere on U belongs to HpU q.
2 2
Indeed 0 P C 2 pU q and BBx02 ` BBy02 “ 0 ` 0 “ 0.
(2) If u, v P HpU q, then
B2 pu`vq 2
B pu`vq B2 u 2
B v 2
B u B v 2 B u 2
B u 2 B v 2
B v 2
Bx2
` By 2 “ Bx2
` Bx 2 ` By 2 ` By 2 “ p Bx2 ` By 2 q ` p Bx2 ` By 2 q “ 0 ` 0 “ 0.
B2 pαüq B2 pαüq 2 2 2 2
(3) If α P R, u P HpU q, then Bx2
` By2 “ α BBxu2 `α BByu2 “ αp BBxu2 ` BByu2 q “ α ¨ 0 “ 0.
The pointwise product of two harmonic functions need not be harmonic. We will
show that every open set U admits a harmonic function u such that u ¨ u “ u2
is not harmonic in U . Any u P C 2 pU q such that �p∇u ” 0q will work, as we
will see. Thus, we may take u to be a linear function, upx, yq “ ax ` by, where
pa, bq P R2 ztp0, 0qu. Then Bu Bu
Bx “ a, By “ b, and ∆u “ 0 ` 0 “ 0. So u P HpU q. But
Bpu2 q B2 pu2 q B2 pu2 q
2 B2 u
Bx “ 2u Bu
Bx , Bx2
“ 2p Bu 2 B u
“ 2p Bu
Bx q ` 2u Bx2 , and similarly,
By 2
2
By q ` 2u By 2 .
Thus ∆pu2 q “ 2}∇u}22 ` 2u∆u “ 2}∇u}22 ` 2u 0 “ 2}∇u}22 , which is not identically
zero in U . (For the linear u “ ax ` by, }∇u}22 “ a2 ` b2 ‰ 0 for all px, yq P U .)

(1) Let u “ ex sin y. We seek a v such that u ` iv is holomorphic. So the Cauchy-
Bv
Riemann equations must be satisfied. Hence Bx “ ´ Bu x
By “ é cos y. If we
x
keep y fixed, we obtain by integrating that v “ é cos y ` Cpyq, for some
Cpyq. As v and éx cos y are differentiable with respect to y, so is C. Thus
Bv
ex sin y ` C 1 pyq “ By “ Bu x 1
Bx “ e sin y. So C pyq “ 0, giving Cpyq “ K. Taking
x
K “ 0, we try v :“ é cos y. Then
uìv “ ex sin yìpéx cos yq “ íex pcos yì sin yq “ í exppxìyq “ í exp z,
where z “ x ` iy. Hence u ` iv “ í exp z, which is entire. Thus v “ éx cos y
is a harmonic conjugate for u :“ ex sin y.
196 Solutions
(2) Let u “ x3 ´ 3xy 2 ´ 2y. We seek a v such that u ` iv is holomorphic. So

the Cauchy-Riemann equations must be satisfied. In particular, we have that
Bv Bu
Bx “ ´ By “ 6xy ` 2. Fixing y, and integrating with respect to x, we obtain
2
v “ 6 x2 y ` 2x ` Cpyq “ 3x2 y ` 2x ` Cpyq, for some Cpyq. Differentiating with
Bv
respect to y, 3x2 ` C 1 pyq “ By “ Bu 2 2 1 2
Bx “ 3x ´ 3y and so C pyq “ ´3y , giving
3
Cpyq “ ´3 y3 ` C “ ý 3 ` C. Taking C “ 0, we try v “ 3x2 y ` 2x ´ y 3 . Thus
u ` iv “ x3 ´ 3xy 2 ´ 2y ` ip3x2 y ` 2x ´ y 3 q
“ x3 ` 3xpiyq2 ` 3x2 piyq ` piyq3 ´ 2y ` i2x
“ px ` iyq3 ` 2ipx ` iyq “ z 2 ` 2iz
for z “ x ` iy. So u ` iv “ z 2 ` 2iz, which is entire. Hence v :“ 3x2 y ` 2x ´ y 3
is a harmonic conjugate of u :“ x3 ´ 3xy 2 ´ 2y.
(3) Let u :“ xp1 ` 2yq. We seek a v such that u ` iv is holomorphic. So the
Bv
Cauchy-Riemann equations must be satisfied. Hence Bx “ ´ Bu
By “ ´2x. Fixing
2
y, and integrating, we get v “ ´2 x2 ` Cpyq “ ´x2 ` Cpyq, for some Cpyq. Thus
Bv y2
C 1 pyq “ By “ Bu 2
Bx “ 1 ` 2y. So Cpyq “ y ` 2 2 ` C “ y ` y ` C. Taking C “ 0,
2 2
we try v :“ ´x ` y ` y . Then
u ` iv “ xp1 ` 2yq ` ip´x2 ` y ` y 2 q “ x ` iy ` 2xy ` ipy 2 ´ x2 q
“ x ` iy ´ ippx2 ´ y 2 q ` i2xyq “ x ` iy ´ ipx ` iyq2 “ z ´ iz 2
for z “ x ` iy. So u ` iv “ z ´ iz 2 is holomorphic, and v :“ ´x2 ` y ` y 2 is a
harmonic conjugate of u :“ xp1 ` 2yq.
Solution to Exercise 5.4. Let v be a harmonic conjugate of u. Then f :“ u ` iv
is holomorphic in Czt0u. For z “ x ` iy P Czt0u, x, y P R, we have
f 1 pzq “ Bu
Bx
Bv
` i Bx “ Bu
Bx ´ i Bu
By “
2x
x2 `y 2
´ i x22y
`y 2
“ 2 xxíy z 2
2 `y 2 “ 2 zz “ z .
Thus z2 has aş primitive f in Czt0u, and so if Cptq :“ eit , t P r0, 2πs, then we
obtain 4πi “ C z2 dz “ f p1q ´ f p1q “ 0, which is absurd. Hence u has no harmonic
conjugate in Czt0u.
Solution to Exercise 5.5. Set u :“ x3 ` y 3 . If f were holomorphic, then u would
2 2
be harmonic. But BBxu2 ` BByu2 “ Bx
B B
p3x2 q ` By p3y 2 q “ 6x ` 6y “ 6px ` yq ‰ 0 for
x ‰ ý. Hence the answer is ‘no’.
Solution to Exercise 5.6. As v is a harmonic conjugate of u, f :“ uìv P OpU q.

Then f 1 P OpU q, and f 1 “ Bu Bv 2 2 2
Bx ` i Bx . As 0 ‰ u ` v “ pu ` ivqpu ´ ivq “ f f “ |f | ,
1 1
we have f pzq ‰ 0 for all z P U . Then f f P OpU q, and its real part is harmonic in
1
fq Repp Bu Bv
ì Bx qpuívqq u Bu Bv
`v Bx
U . So Repf 1 f1 q “ Repf
|f |2
“ Bx
u2 `v2
“ Bx
u2 `v2
is harmonic in U .
Solution to Exercise 5.7. As v is a harmonic conjugate of u, f :“ uìv P OpU q.
2 2 2 2 2
Thus f2 P OpU q too. As f2 “ u ´v 2 ` iuv, uv “ Im f2 , and being the imaginary
part of a holomorphic function, uv is harmonic in U .
Solution to Exercise 5.8. Let U Ă C be open. It suffices to show that if u is

harmonic in U , then so are Bu Bu 8
Bx and By . We know that u P C pU q. We have
B2 Bu B2 Bu B B2 u B B2 u B B2 u B B2 u
p q
Bx2 Bx
` p q
By 2 Bx
“ Bx p Bx2 q ` By p ByBx q “ Bx p´ By 2 q ` By p BxBy q
B B2 u B B B B B2 u B2 u B
“ Bx p´ By 2 q ` Bx p By p By uqq “ Bx p´ By 2 ` By 2
q “ Bx p0q “ 0.
B2 Bu B2 Bu B B2 u B B2 u B B2 u B2 u B
Similarly, p q
Bx2 By
` p q
By 2 By
“ By p Bx2 q ` By p By 2 q “ By p Bx2 ` By 2
q “ By p0q “ 0.
Solution to Exercise 5.9. By the mean-value property, for all r P p0, Rq, we
ş2π
have 2πupz0 q “ 0 upz0 `reit qdt. Multiplying by r, and integrating from r “ 0 to R,
şR şR ş2π
πR2 upz0 q “ 2πupz0 q 0 rdr “ 0 0 upz0 ` reit qdtrdr “ Dpz0 ,Rq u dA.
Solution to Exercise 5.10. Let z P Dpz0 , Rq. Then Dpz, R´|z ´z0 |q Ă Dpz0 , Rq.
Using Exercise 5.9, and the nonnegativity of u, we have
1
1

upzq “ πpR´|z´z 0 |q2 Dpz,R´|z´z 0 |q u dA ď πpR´|z´z 0 |q2 Dpz0 ,Rq u dA
πR2 1

“ πpR´|z´z 2
0 |q πR
2 Dpz0 ,Rq u dA
πR2 1
“ πpR´|z´z0 |q2
upz0 q “ |z´z0 | 2 upz0 q.
p1´ R
q
Solution to Exercise 5.11. Using Exercise 5.10, we have for any two points
z, z0 P C and any R ą |z ´ z0 | that
1
upzq ď |z´z0 | 2 upz0 q.
p1´ R
q
Passing to the limit as R Ñ 8, we obtain upzq ď upz0 q. As z, z0 P C were arbitrary,
we can swap their roles to obtain also upz0 q ď upzq. Hence upzq “ upz0 q for all
z P C, that is u is constant.
If u is simply bounded below, say by m, then u ´ m ě 0 everywhere, and it is
also clearly harmonic. So u ´ m is constant, and consequently so is u.
On the other hand, if u is bounded above, then ú is bounded below, and
moreover harmonic, showing that ú is constant. Thus u is constant as well.
Solution to Exercise 5.12. The function ϕ :“ vú is continuous on the compact

set D Y T, and so it has a maximiser, say z0 .
If z0 P D, then z0 is a maximiser of the harmonic function ϕ on the simply
connected domain D, and so by the Maximum Principle, ϕ is a constant in D. By
continuity, ϕ is constant, taking value say C, on D Y T. Since the values of ϕ|T are
pointwise ď 0, it follows that C ď 0. So for all z P D, vpzq ´ upzq “ ϕpzq “ C ď 0,
i.e., upzq ě vpzq for all z P D.
If z0 P T, then for all z P D, vpzq ´ upzq “ ϕpzq ď ϕpz0 q “ vpz0 q ´ upz0 q ď 0,
and so upzq ě vpzq for all z P D.
198 Solutions

(1) Let ϕpxq “ c0 ` c1 x ` ¨ ¨ ¨ ` cd xd , where d P N and c0 , ¨ ¨ ¨ , cd P R. Then
ϕpzq :“ ϕpx ` iyq “ c0 ` c1 z ` ¨ ¨ ¨ ` cd z d is entire, and so h :“ Repϕpx ` iyqq
is harmonic. Moreover for all x P R, hpx, 0q “ Repϕpxì0qq “ Repϕpxqq “ ϕpxq.
i
(2) We have zì is holomorphic in the upper half-plane, and so its real part is
Repipxíp0`1qq 1
harmonic there. Moreover, hpx, 0q “ x2 `p0`1q2 “ x2 `1 “ ϕpxq for all x P R.
Solution to Exercise 5.14. As C is simply connected, there exists an f P OpCq
such that u “ Re f . Then expp´f q is entire too. For all z P C, upzq ě 0, and so
| expp´f pzqq| “ e´Re f pzq “ eúpzq ď e0 “ 1. By Liouville’s Theorem, expp´f q is a
constant. Hence | expp´f q| is constant too, that is, eú is constant. Consequently,
the real logarithm logpeú q “ ú is constant, and so u is constant as well.
í4θ
(1) For z “ reiθ (r ą 0, θ P R), expp´ z14 q “ expp´ e r4 q. Taking r “ n1 , n P N,
and 4θ “ ´π, i.e., with zn :“ n1 exppí π4 q “: xn ` iyn , xn , yn P R, we have
4
upxn , yn q “ exppń4 eiπ q “ en . Now as n Ñ 8, we have pxn , yn q Ñ p0, 0q, but
�pp lim upxn , yn q “ 0qq, showing that u is not continuous at p0, 0q.
nÑ8
1
1 ´ 1
1 1 ´
(2) upx, 0q “ expp´ px`0iq4q “ e , up0, yq “ expp´ p0`yiq
x4
4 q “ expp´ 1 y 4 q “ e . y4
´ 14 ´ 14
Bu upx,0qúp0,0q e ´0 e
Bx p0, 0q “ lim “ lim “ lim “ 0.
x x
(3) We have x´0 x
xÑ0 xÑ0 xÑ0 x
1
1 1 1 2 1
(For the last equality, note that for x ‰ 0, e x4 “1 ` x4 ` 2! p x4 q ` ¨¨¨ ą x4
´ 1
Bu
and so 0 ď | e x | ď |x|3 .) Similarly,
x4
By p0, 0q “ 0. Thus
1 1
2
B u
Bu
px, 0q ´ Bu
p0, 0q d ´ x4
e ´0 e´ x4 4 ´ 1
4e x4
Bx2
p0, 0q “ lim Bx x ´ 0Bx “ lim dx x “ lim x x5
“ lim 6 “0
xÑ0 xÑ0 xÑ0 xÑ0 x
´ 1
“ 1 ` x14 ` 2!1 p x14 q2 ` ¨ ¨ ¨ ą 1
for x ‰ 0, which gives 0 ď | ex6 | ď 2|x|2 ).
1 x4
(as e x4
2x8
B2 u B2 u B2 u
Similarly, By 2 p0, 0q “ 0. Hence Bx2 p0, 0q ` By 2 p0, 0q “ 0 ` 0 “ 0.

(1) Let z0 P D1 . Then ϕpz0 q P D2 . Let Ω be a disc with centre ϕpz0 q and radius
ǫ ą 0 small enough so that Ω Ă D2 . Since Ω is simply connected, there exists
an H P OpΩq such that h|Ω “ Re H. Then ϕ´1 Ω Ă D1 is an open set in R2
containing z0 . The composition of ϕ|ϕ´1 pΩq : ϕ´1 pΩq Ñ Ω and H : Ω Ñ C is
holomorphic, and so RepH ˝ ϕ|ϕ´1 pΩq q is harmonic in ϕ´1 pΩq. For z P ϕ´1 pΩq,
pRepH ˝ ϕ|ϕ´1 pΩq qqpzq “ pRe Hqpϕpzqq “ hpϕpzqq “ ph ˝ ϕqpzq. Hence ph ˝ ϕq|ϕ´1 pΩq
is harmonic in ϕ´1 pΩq. As z0 P D1 was arbitrary, h ˝ ϕ is harmonic in D1 .
(2) If h : D2 Ñ R is harmonic, then by the first part, h ˝ ϕ : D1 Ñ R is harmonic.
If h ˝ ϕ : D1 Ñ R is harmonic, then as ϕ´1 : D2 Ñ D1 is holomorphic, by the
first part, ph ˝ ϕq ˝ ϕ´1 : D2 Ñ R is harmonic. But
ph ˝ ϕq ˝ ϕ´1 “ h ˝ pϕ ˝ ϕ´1 q “ h ˝ idD2 “ h,
where idD2 is the identity map D2 Q z ÞÑ z P D2 . So h : D2 Ñ R is harmonic.
(3) By the triangle inequality in ∆P O1 B (see the picture below), for s (” P ) in H,

|s ` 1| “ P A “ P O 1 ` O1 A “ P O1 ` O 1 B ą P B “ |s ´ 1|,
where we have used the fact that O 1 is on the perpendicular bisector of AB
to get the third equality. It is evident that P A ą P B holds when P lies on the
right of 0 along the line joining A to B. So ϕpsq P D for all s P H.
P p” sq
O1
A 0 B
p” ´1q p” 1q
¨´1 ¨´1
ϕ is holomorphic: ¨`1 P OpCzt´1uq, and as H Ă Czt´1u, we have ¨`1 P OpHq.
1`z
Define ψ : D Ñ H by ψpsq “ z P D. (This expression for ψ, which is a
1´z ,
candidate for ϕ , is obtained by solving for s in the equation z “ ϕpsq “ s´1
´1
s`1 .)
z
1`
z
´1 ´
z P 1 θ
´ 1
´1 z
A 0 B
The angle subtended by the diameter AB at any point of the circle is 90˝ .
So for any P (” z) in D, =AP B ą 90˝ . So
Repψpzqq “ Re 1`z
1´z “ |ψpzq| cos θ “ |ψpzq| cospπ ´ =AP Bq ą 0.
Thus ψpzq P H for all z P D. This can also be seen analytically, since if z P D,
then z “ reiθ for an r P r0, 1q and θ P R, and so
“ Repp1`r cos θìr sin θqp1´r cos θìr sin θqq
1`z iθ 2
Re 1´z “ Re 1`re
1´reiθ p1´r cos θq2 `r 2 psin θq2
1´r
“ p1´r cos θq2 `r 2 psin θq2 ą 0.
1`¨
Also, ψ P OpDq, since 1´¨ P OpCzt1uq, and D Ă Czt1u.
s´1
1` s`1 s`1`s´1 2s
For s P H, pψ ˝ ϕqpsq “ 1´ s´1
“ s`1´s`1 “ 2 “ s “ idH psq, where idH is
s`1
the identity map on H. This shows that ϕ is injective. Also, for z P D, we have
1`z
1´z
´1 1`z´1`z 2z
pϕ ˝ ψqpzq “ 1`z
`1
“ 1`z`1´z “ 2 “ z “ idD pzq, where idD is the identity
1´z
map on D. This shows that ϕ is surjective. So ϕ is a bijection and ϕ´1 “ ψ.
Some real analysis background
We recall here a few of the real analysis results we use in the book. Proofs of results listed
here, or other preliminaries from real analysis needed in the book can be found in most
undergraduate real analysis textbooks, for example [1], [14], [16].
One variable real analysis. Let S be a subset of R. A real number u is said to be an

upper bound of S if for all x P S, x ď u. A number u˚ P R is called a least upper bound
of S (or a supremum of S) if u˚ is an upper bound of S, and for every upper bound u
of S, u˚ ď u. We denote the supremum of S by sup S. If sup S P S, then we call u˚ the
maximum of S.
A real number ℓ is said to be a lower bound of S if for all x P S, ℓ ď x. A number
ℓ˚ P R is called a greatest lower bound of S (or an infimum of S) if ℓ˚ is a lower bound
of S, and for every lower bound ℓ of S, ℓ ď ℓ˚ . We denote the infimum of S by inf S. If
inf S P S, then we call ℓ˚ the minimum of S.
Least Upper Bound Property of R: If S Ă R is such that S ‰ H and S has an upper
bound, then the supremum/least upper bound sup S of S exists.
Archimedean Property of R: If x, y P R and x ą 0, then there exists an n P N such
that nx ą y.
An interval is a set consisting of all the real numbers between two given real numbers,
or of all the real numbers on one side or the other of a given number. So an interval is a
set of any of the following forms, where a, b P R and a ă b :
pa, bq “ tx P R : a ă x ă bu a b
ra, bs “ tx P R : a ď x ď bu
a b
pa, bs “ tx P R : a ă x ď bu
a b
ra, bq “ tx P R : a ď x ă bu
a b
pa, 8q “ tx P R : a ă xu a
ra, 8q “ tx P R : a ď xu
a
p´8, bq “ tx P R : x ă bu
b
p´8, bs “ tx P R : x ď bu
b
p´8, 8q “ R
201
202 Some real analysis background
Cauchy-Schwarz inequality: If n P N, and x1 , ¨ ¨ ¨ , xn , y1 , ¨ ¨ ¨ , yn are real numbers, then

px1 y1 ` ¨ ¨ ¨ ` xn yn q2 ď px21 ` ¨ ¨ ¨ ` x2n qpy12 ` ¨ ¨ ¨ ` yn2 q.
A real sequence is a function f : N Ñ R. We use the notation pan qnPN for the sequence
N Q n ÞÑ an P R. A sequence pan qnPN is said to be convergent with limit L (P R) if for every
ǫ ą 0, there exists an N P N such that for all n P N with n ą N , we have |an ´ L| ă ǫ. Then
we write lim an “ L. If there is no L P R such that lim an “ L, then pan qnPN is called
nÑ8 nÑ8
divergent. A real sequence is monotone if it is increasing (for all n ą N P N, an`1 ě an ) or
if it is decreasing (for all n ą N P N, an`1 ď an ). A real sequence is bounded if there exists
an M ą 0 such that for all n P N, |an | ď M . Every monotone and bounded real sequence
is convergent.
Sandwich Theorem: Let pan qnPN , pbn qnPN be real sequences that converge with the same
limit, that is, lim an “ lim bn . Let pcn qnPN be a real sequence, and let N P N be such
nÑ8 nÑ8
that for all n ą N P N, an ď cn ď bn . Then pcn qnPN is also convergent with the same limit,
that is, lim an “ lim cn “ lim bn .
nÑ8 nÑ8 nÑ8
Let pan qnPN be a sequence and let n1 ă n2 ă n3 ă ¨ ¨ ¨ be a strictly increasing sequence of
natural numbers. Then pank qkPN is called a subsequence of pan qnPN . Any subsequence of a
convergent real sequence is convergent with the same limit.
Bolzano-Weierstrass Theorem: Every bounded real sequence has a convergent subse-
quence.
A real sequence pan qnPN is Cauchy if for every ǫ ą 0, there exists an N P N such that for
all m, n P N such that m, n ą N , we have |an ´ am | ă ǫ. Every convergent sequence is
Cauchy, and a consequence of the least upper bound property of R is that every Cauchy
sequence is convergent.
For a real sequence pan qnPN , if sn :“ a1 `¨ ¨ ¨àn , n P N, then psn qnPN is called the sequence
ř
8
of partial sums of pan qnPN . If psn qnPN converges, we say that ‘the series an converges’,
ř
8 n“1
and we write the sum of the series an “ lim sn . If psn qnPN does not converge we say
ř
8 n“1 nÑ8
that ‘the series an diverges’.
n“1
ř
8 ř
8
1
The real geometric series rn converges if and only if |r| ă 1, and its sum is rn “ 1´r .
n“0 n“0
ř
8
1 ř
8
1
The harmonic series n diverges. In fact for s P R, ns converges if and only if s ą 1.
n“1 n“1
ř
8 ř
8
The real series an is said to converge absolutely if |an | converges. If a real series
n“1 n“1
converges absolutely, then it converges.
Leibniz Alternating Series Theorem: Let pan qnPN be a real, decreasing sequence such
ř
8
that an ě 0 for all n P N, that converges to 0. Then the series p´1qn an converges.
n“1
There are three important tests for the convergence of a real series:
‚ The comparison test (where we compare with a series whose convergence status is known).
‚ The ratio test (where we look at the behaviour of the ratio aan`1 of terms).
a n
‚ The root test (where we look at the behaviour of n

|an |).
We have summarised these in the following table.
Comparison Ratio Root
Absolute a
ð |an | ď cn for all large n; | aan`1 | ďră1 n
|an | ď r ă 1
convergence n
ř
8
cn converges. for all large n. for all large n.
n“1
a
Divergence ð an ě dn ě 0 for all large n; | aan`1
n
|ě1 n
|an | ě 1
ř
8
dn diverges. for all large n. infinitely often.
n“1
ř
8
Real power series: Let pcn qně0 be a real sequence. Then either cn xn is absolutely
n“0
convergent for all x P R (the power series is then said to have an infinite radius of con-
vergence), or there exists a unique r ě 0 (called the radius of convergence of the power
ř
8 ř
8
series) such that cn xn is absolutely convergent for x P p´r, rq and cn xn diverges for
n“0 n“0
x P Rzr´r, rs.
Let I be an interval in R, c P I and f : I Ñ R. The function f is continuous at c if for every
ǫ ą 0, there exists a δ ą 0 such that for all x P I satisfying |x ´ c| ă δ, |f pxq ´ f pcq| ă ǫ.
The function f is continuous (on I) if for every x P I, f is continuous at x.
Intermediate Value Theorem: If f : ra, bs Ñ R is continuous and y P R lies between
f paq and f pbq, pthat is, f paq ď y ď f pbq or f pbq ď y ď f paqq, then there exists a c P ra, bs
such that f pcq “ y.
Extreme Value Theorem: If f : ra, bs Ñ R is continuous, then
(1) The set S, where S :“ tf pxq : x P ra, bsu “: f pra, bsq “ range of f , is bounded.
(2) sup S and inf S exist.
(3) sup S and inf S are attained, i.e., there exist c, d P ra, bs such that f pcq “ sup S “ max S
and f pdq “ inf S “ min S.
If I Ă R is an interval, f : I Ñ R, and c P I, then f is differentiable at c if there exists an
L P R such that lim f pxq´f
xć
pcq
“ L, that is, for every ǫ ą 0, there exists a δ ą 0 such that
xÑc
whenever x P I satisfies 0 ă |x ´ c| ă δ, we have | f pxq´fxć
pcq
´ L| ă ǫ. The number L is
1 df
unique, and we denote this unique number by f pcq or by dx pcq, and call it the derivative
of f at c. If f is differentiable at every x P I, then f is called differentiable on I.
If f is differentiable at c P I, then f is continuous at c. One may wonder how badly be-
haved continuous functions might be with respect to the notion of differentiability. It turns
out that there are functions that are continuous everywhere, but differentiable nowhere.
For example, the blancmange function b is continuous on R, but not differentiable at any
x P R, and is constructed as follows. We start from the sawtooth function f1 , and construct
1
f2 , f3 , ¨ ¨ ¨ by setting fn pxq “ 2n´1 f1 p2n´1 xq, x P R, n ě 2. The blancmange function b is
ř
8
obtained by adding these: bpxq “ fn pxq, x P R.
n“1
1
2 f1
Mean Value Theorem: Let f : ra, bs Ñ R be continuous on ra, bs and differentiable on

pa, bq. Then there is a point c P pa, bq such that f pbq´f
bá
paq
“ f 1 pcq.
The Riemann integral. A partition P of ra, bs is a finite set P “ tx0 , x1 , . . . , xn´1 , xn u

such that a “ x0 ă x1 ă ¨ ¨ ¨ ă xn´1 ă xn “ b. The set of all partitions P of ra, bs will be
denoted by P. Let f : ra, bs Ñ R be a bounded function and let P be a partition of ra, bs.
ř
n´1
An upper sum U pf, P q of f associated with P is U pf, P q :“ p sup f pxqqpxk`1 ´ xk q.
k“0 xPrxk ,xk`1 s
ř
n´1
A lower sum Lpf, P q of f associated with P is Lpf, P q :“ p inf f pxqqpxk`1 ´ xk q.
k“0 xPrxk ,xk`1 s
A bounded function f : ra, bs Ñ R is called Riemann integrable if sup LP pf q “ inf UP pf q.
P PP P PP
şb
We denote this common value by a f pxqdx, and call it the Riemann integral of f . All
functions continuous on ra, bs are Riemann integrable. Let f, g be Riemann integrable on
ra, bs and α P R. Then the following hold:
şb şb şb
‚ a f pxq ` gpxqdx “ a f pxqdx ` a gpxqdx.
şb şb
‚ a α ¨ f pxqdx “ α a f pxqdx.
şb
‚ If for all x P ra, bs, f pxq ě 0, a f pxqdx ě 0.
şb şb
‚ If for all x P ra, bs, f pxq ď gpxq, then a f pxqdx ď a gpxqdx.
şb şb
‚ |f | is also Riemann integrable and | a f pxqdx| ď a |f pxq|dx.
şb
‚ If f is continuous, f ě 0 on ra, bs, and a f pxqdx “ 0, then f ” 0 on ra, bs.
Fundamental Theorem of Calculus: Let f : ra, bs Ñ R be continuously differentiable

şb
on ra, bs. Then a f 1 pxqdx “ f pbq ´ f paq.
m
a In R , we use the Euclidean metric, d2 px, yq “ }x ´ y}2
Multivariable real analysis.
m 2 2 for v “ pv , ¨ ¨ ¨ , v q P Rm .
for x, y P R . Here }v}2 :“ v1 ` ¨ ¨ ¨ ` vm 1 m
A sequence pxn qnPN in Rm is said to converge to x P Rm if for every ǫ ą 0, there exists
an N P N such that for all n ą N , }xn ´ x}2 ă ǫ. A sequence pxn qnPN is convergent
to x P Rm if and only if converges componentwise to x, that is, for all k P t1, ¨ ¨ ¨ , mu,
pkq p1q pmq
pxn qnPN converges to pxpkq qnPN . Here x “: pxp1q , ¨ ¨ ¨ , xpmq q and xn “: pxn , ¨ ¨ ¨ , xn q.
A set U Ă Rm is open if for every x P U , there exists an r ą 0 such that the ball
Bpx, rq :“ ty P Rm : }y ´ x}2 ă ru Ă U . A set F Ă Rm is closed if Rm zF is open.
Any finite intersection of open sets is open, and arbitrary unions of open sets are open.
Taking complements, it follows that a finite union of closed sets is closed, and arbitrary
intersections of closed sets are closed. A set F Ă Rm is closed if and only if for every
convergent sequence pxn qnPN such that xn P F (n P N), we have that lim xn P F . A set
nÑ8
K Ă Rm is compact if every sequence in K has a convergent subsequence with its limit
m
belonging to K. A set K Ă R is compact if and only if K is closed and bounded.
Let Rn , Rm be equipped with the Euclidean distance. Let S Ă Rn , and c P S. Then f is
continuous at c if for every ǫ ą 0, there exists a δ ą 0 such that whenever x P S satisfies
}x ´ c}2 ă δ, we have }f pxq ´ f pcq}2 ă ǫ. Also, f is said to be continuous on S if for
each x P S, f is continuous at x. The function f : S Ñ Rm is continuous on S if and only
if for every V open in Rn , f ´1 pV q :“ tx P S : f pxq P V u is open in S, that is, for each
x P f ´1 pV q, there exists an r ą 0 such that ty P S : }y ´ x}2 ă ru Ă f ´1 pV q. The function
f : S Ñ Rm is continuous at c P S if and only if for every sequence pxk qkPN in S such
that pxk qkPN converges to c, pf pxk qqkPN converges to f pcq. If f : S Ñ Rm is continuous
at c P S, T Ă Rm is such that f pSq Ă T , and if g : T Ñ Rk is continuous at f pcq P T ,

then the composition map f ˝ g : S Ñ Rk , defined by pf ˝ gqpxq :“ f pgpxqq (x P S), is
continuous at c P S. If S Ă Rn and T Ă Rm , then S, T are said to be homeomorphic if
there exist continuous maps f : S Ñ Rm and g : T Ñ Rn such that f pSq Ă T , gpT q Ă S,
f ˝ g “ idT and g ˝ f “ idS . Here, for a set X, idX denotes the identity map, given by
X Q x ÞÑ x “: idX pxq. A function f : S Ñ Rm is continuous if and only if each of its
components f1 , ¨ ¨ ¨ , fm : S Ñ R are continuous. Here f pxq “: pf1 pxq, ¨ ¨ ¨ , fm pxqq for all
x P S. A function f : S Ñ Rn is uniformly continuous if for every ǫ ą 0, there exists a
δ ą 0 such that for all x, y P S satisfying }x ´ y}2 ă δ, we have }f pxq ´ f pyq}2 ă ǫ. If
K Ă Rn is compact and f : K Ñ Rm is continuous, then f is uniformly continuous.
Weierstrass’s Theorem: If K Ă Rn is a compact set and f : K Ñ Rm is a continuous
function on K, then f pKq is a compact subset of Rm , and there exist xmin , xmax P K such
that f pxmin q “ inftf pxq : x P Ku and f pxmax q “ suptf pxq : x P Ku.
Let U Ă Rn be open, and c P U . A function f : U Ñ Rm is preal q differentiable at c if there
exists a linear transformation L : Rn Ñ Rm such that lim }f pxq´f}xć}
pcq´Lpxćq}2
2
“ 0, that is,
xÑc
for every ǫ ą 0, there exists a δ ą 0 such that whenever x P U satisfies 0 ă }x ´ c}2 ă δ, we
have }f pxq´f}xć}
pcq´Lpxćq}2
2
ă ǫ. Then (the unique such linear transformation) L is called the
derivative of f at c, and we write f 1 pcq “ L. Let f1 , . . . , fm denote the components of f and
Bfi f pc ,¨¨¨ ,cj´1 ,xj ,cj`1 ,¨¨¨ ,cn q´fi pc1 ,¨¨¨ ,cj´1 ,cj ,cj`1 ,¨¨¨ ,cn q
let c “ pc1 , ¨ ¨ ¨ , cn q. If Bx pcq :“ lim i 1 xj ćj ex-
j xj Ñcj
Bfi
ists, then we call Bxj pcq the pi, jqth partial derivative f at c. If f : U Ñ Rm is differentiable
Bfi
at c, then all the partial derivatives of f at c, namely, Bx j
pcq (i “ 1, ¨ ¨ ¨ , m; j “ 1, ¨ ¨ ¨ , n)
1 1
exist, and the matrix rf pcqs of the linear transformation f pcq with respect to the standard
bases for Rn and Rm is given by
» Bf1 Bf1 fi
Bx1 pcq ¨¨¨ Bxn pcq
rf 1 pcqs “ – ..
.
..
.
fl.
Bfm Bfm
Bxj pcq ¨¨¨ Bxn pcq
By the standard basis for Rn , we mean the set te1 , ¨ ¨ ¨ , en u, where for k P t1, ¨ ¨ ¨ , nu,
ek P Rn has all components zero, except for the k th component, which is equal to 1.
Bfi
If all the partial derivatives exist at each x P U and the maps x ÞÑ Bx j
pxq : U Ñ R
(i “ 1, . . . , m; j “ 1, . . . , n) are continuous on U , then f is differentiable at each x P U .
Schwarz Theorem: If U Ă R2 , f : U Ñ R, and at each px, yq P U , the partial derivatives
Bf Bf B2 f B Bf B2 f B Bf
Bx px, yq, By px, yq, BxBy px, yq :“ Bx By px, yq, ByBx px, yq :“ By px, yq Bx px, yq exist, and define
B2 f B2 f
continuous functions on U , then BxBy px, yq “ ByBx px, yq for all px, yq P U .
Some relevant set theory. A relation R on a set S is a subset of the Cartesian product
S ˆ S :“ tpa, bq : a, b P Su. If pa, bq P R, then we write aRb. A relation R on a set S is called
an equivalence relation if it is reflexive (that is, for all a P S, aRa), symmetric (that is, if
aRb, then bRa), and transitive (that is, if aRb and bRc, then aRc). If R is an equivalence
relation of a set S, then the equivalence class of a, denoted by ras, is defined to be the set
ras “ tb P S : aRbu.
A set S is finite if it is empty or there exists an n P N and a bijection f : t1, ¨ ¨ ¨ , nu Ñ S. If
no such bijection exists, then S is called infinite. If S is an infinite set, then S is countable
if there is a bijective map from N onto S. If S is infinite and not countable, then it is called
uncountable. Every infinite subset of a countable set is countable. If A, B are countable,
then AˆB is also countable. A countable union of countable sets is countable. A countable
union of finite sets is finite or countable. The sets N, Z, Q are examples of countable sets,
while R and any interval in R are examples of uncountable sets.
Notes
Chapter 1: The remark on the historical development of complex numbers is based on

[11]. Exercise 1.2 is taken from [17]. Exercises 1.8, 1.8, 1.14, 1.24 are taken from [11].
Exercises 1.34 and 1.52 are taken from [2].
Chapter 2: The section on the geometric meaning of the complex derivative follows [11].
Exercises 2.21 and 2.23 are taken from [11].
Chapter 3: The proof of Theorem 3.9 follows [2]. Exercises 3.4, 3.5, 3.15, 3.27, 3.36 are
taken from [2]. Exercises 3.12, 3.19, 3.26, 3.25, 3.30, 3.32 are taken from [11]. Exercise 3.29
is taken from [15]. Exercises 3.33, 3.43, 3.47 are taken from [5]. Exercise 3.37 is taken
from [9]. The proof of the Cauchy Integral Theorem given in the Appendix is based on [3].
Chapter 4: Section 4.8 follows [2]. Exercise 4.15 is based on problem 6 of the International
Mathematics Olympiad, 1979. Exercises 4.17, 4.40, 4.61, 4.64, 4.65, 4.66, 4.67, 4.68 are
taken from [5]. Exercise 4.23 is taken from [22]. Exercises 4.27, 4.35 are taken from [24].
Exercises 4.28, 4.43 are taken from [15]. Exercise 4.49 is taken from [2]. Exercise 4.70 is
taken from [12].
Chapter 5: Exercises 5.4, 5.5 and the proof of Theorem 5.2 is taken from [2]. Exercise 5.13
is based on [6].
207
Bibliography
[1] T . Apostol. Calculus II. 2nd edition. John Wiley, 1969.

[2] M. Beck, G. Marchesi, D. Pixton, and L. Sabalka. A first course in complex analysis.
http://math.sfsu.edu/beck/papers/complex.pdf , 2008.
[3] J. Conway. Functions of one complex variable I. 2nd edition. Springer, 1978.
[4] S. Fisher. Complex variables. 2nd edition. Dover, 1999.
[5] F. Flanigan. Complex variables. Dover, 1972.
[6] F. Flanigan. Classroom notes: Some half-plane Dirichlet problems: A bare hands approach.
American Mathematical Monthly , 80:59-61 , no. 1, 1973.
[7] B. Gelbaum and J. Olmsted. Counterexamples in analysis. Dover , 1964.
[8] J. Gilman , I. Kra, and R. Rodriguez. Complex analysis. In the spirit of Lipman Bers. Springer,
2007.
[9] J. Howie. Complex analysis. Springer, 2003.
[10] D. Minda. The Dirichlet problem for a disk. American Math ematical Monthly, 97:220-223,
no. 3, 1990.
[11] T. Needham. Visual complex analysis. Oxford University Press, 1997.
[12] R. Ash, and W. Novinger. Complex analysis. 2nd edition. Dover, 2007.
[13] R. Remmert. Theory of complex functions , Springer , 1991.
[14] W. Rudin. Principles of mathematical analysis. 3'·d edition. McGraw-Hill, 1976.
[15] W. Rud in. Real and complex analysis. 3rd edition. McGraw-Hill, 1987.
[16] A . Sasane. The how and why of one variable calculus. Wiley, 2015 .
[17] A. Shastri. Basic complex analysis of one variable. Macmillan Publishers India , 2011.
[18] W. Shaw. Complex analysis with MATHEMATICA, Cambridge University Press, 2006.
[19] J. Shurman. Course materials for Mathematics 311: Complex analysis.
http://people.reed.edu/-jerry/311/mats . html , 2009.
[20] I. Stewart and D. Tall. Complex Analysis. 2nd edition. Cambridge University Press, 2018.
[21] D. Tall. Functions of a complex variable. Dover , 1970.
[22] L. Volkovyskil, G . Lunts, and I. Aramanovich . A collection of pr-oblems on complex analysis.
Dover, 1991.
[23] Wikipedia page on Liouville's theorem in Differential algebra.
[24] F. Wikstrom. Funktionsteori. Ovningsbok. Studentlitteratur, 2016.
209
Index
D-contractible curve, 60 complex plane, 10

C, 7 concatenation of paths, 53
conjugate, 14
absolute value, 13 continuity of a function at a point, 203
absolutely convergent series, 78 continuous function, 203
Argument Principle, 93 contour integral, 49
convergent sequence, 202
Basel problem, 111 convergent series, 78, 202
Bernoulli numbers, 102 Cosine Formula, 16
Bessel function, 85, 99 curve, 19
biholomorphism, 121
Binomial Theorem, 13 de Moivre’s formula, 12
blancmange function, 203 degree of a polynomial, 70
Bombelli, 9 Dirichlet problem, 120
boundary data, 120 divergent sequence, 202
divergent series, 78, 202
Cardano, 9 Division Algorithm, 13
Casorati-Weierstrass Theorem, 105 domain, 19
Cauchy Integral Formula, 65, 87
Cauchy Integral Theorem, 57 elliptic PDE, 114
Cauchy principal value, 109 entire function, 30
Cauchy product, 82 epicycloid, 63
Cauchy sequence, 18 essential singularity, 100
Cauchy’s inequality, 88 Euler Product Formula, 78
Cauchy-Riemann equations, 35 Euler’s formula, 23
chain rule, 33 exponential, 21
chordal metric, 16 extended complex plane, 18
classification of zeroes, 89
closed path, 56 Fibonacci number, 109
commutative ring, 92 field, 8
compact set, 204 Fundamental Theorem of Algebra, 70
Comparison Principle, 119 Fundamental Theorem of Contour
complete, 18 Integration, 55
complex differentiable function, 30
complex exponential, 21 Gaussian integer, 12
complex number, 7 greatest upper bound, 201
211
212 Index
harmonic conjugate, 116 removable singularity, 100

harmonic function, 113 Residue Theorem, 107
Harnack’s inequality, 119 Riemann Hypothesis, 4, 78
holomorphic function, 30 Riemann Mapping Theorem, 121
homotopic paths, 57 Riemann sphere, 16
Riemann zeta function, 4, 78
Identity Theorem, 92 roots of a complex number, 12
imaginary part, 7
improper integral, 109 sawtooth function, 203
infimum, 201 sequence, 202
infinity, 18 simply connected domain, 60
integral domain, 92 smooth path, 48
Integration by Parts Formula, 56 stepwise path, 19
interval, 201 supremum, 201
isolated singularity, 99
Taylor series, 85, 87
Laplace equation, 113 transitive relation, 205
Laurent series, 94 triangle inequality, 15
least upper bound, 201 trigonometric functions, 24
Liouville’s Theorem, 69 upper bound, 201
Log, 25
logarithm function, 25 winding number, 62
lower bound, 201
zero divisor, 92
Möbius transformation, 28, 121 zero of a polynomial, 70
Maximum Modulus Theorem, 93
Mean Value Property, 118
Minimum Modulus Theorem, 94
ML inequality, 54
modulus, 13
Morera’s Theorem, 71
one-point compactification, 18
order of a pole, 100
order of a zero of a function, 89
ordered field, 9
partial derivative, 205

partial sums, 78
path, 19
path-connected, 19
Picard’s Theorem, 106
Poisson Integral Formula, 120
pole, 100
power series, 79
Prime Number Theorem, 4
primitive, 63
principal argument, 25
principal value, 27
radius of convergence, 79
real analytic function, 84
real part, 7

Sasane Amol Sasane Sara Maad A Friendly Approach To Complex

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sasane Amol Sasane Sara Maad A Friendly Approach To Complex

Uploaded by

Copyright:

Available Formats

This page intentionally left blank

British Library Cataloguing-in-Publication Data

Cover image: The chameleons are conformal images of one another.

A FRIENDLY APPROACH TO COMPLEX ANALYSIS

ISBN 978-981-127-280-6 (hardcover)

For any available supplementary material, please visit

Desk Editor: Tan Rok Ting

Chapter 1. Complex numbers and their geometry 7

Chapter 3. Cauchy Integral Theorem 47

Chapter 4. Taylor and Laurent series 77

What is complex analysis?

In the same way, f : C Ñ C is complex differentiable at z0 P C if there exists a

that is, for every ǫ ą 0, there is a δ ą 0 such that whenever 0 ă |z ´ z0 | ă δ,

In contrast, we will later learn that if f : C Ñ C is a complex differentiable

Why study complex analysis?

Partial Differential Equations (PDE). Associated with a function f : C Ñ C,

Cauchy-Riemann Cauchy Integral

Taylor and Laurent series

Complex analysis is not complex analysis! It is not very complicated, and

differentiable function on an open subset of C is infinitely many times complex

Complex numbers and their geometry

1.1. The field of complex numbers

(F1) means that ` : C ˆ C Ñ C is associative and commutative, that there exists

Proposition 1.1. pC, `, ¨q is a field.

R is ‘contained’ in C. We can embed R inside C, and view R as a ‘subfield’ of C,

As complex multiplication is commutative, yi “ iy, and so x ` yi “ x ` iy.

Historical development of complex numbers. Contrary to popular belief,

1.2. Geometric representation of complex numbers

z1 `z2 px1 `x2 , ˚ q p ˚ , y1 `y2 q

Geometric meaning of complex multiplication. We will now see the spe-

In the right-angled triangle shown above, x “ r cos θ, y “ r sin θ. Using polar

As a special case, consider multiplication by cos α ` i sin α, which is at a distance

(‘Only if part’:) Let w “ ρpcos α ` i sin αq, where ρ ě 0 and α P R. As wn “ z, we

The complex conjugate z of z “ x` iy where x, y P R, is z :“ x ´ iy. In the complex

Exercise 1.27. Let z, w P Czt0u. Show that |z `w| “ | |w| |z|

1.3.1. Metric on C. Since C is just R2 , we use the usual Euclidean distance in R2

Exercise 1.36 (Riemann sphere). Let S 2 :“ tpx, y, zq P R3 : x2 ` y 2 ` z 2 “ 1u be the

A subset U of C is called open if for every z P U , there exists an rz ą 0 such that

1.3.3. Convergence and continuity. A sequence pzn qnPN of complex numbers

Let S be a subset of C, z0 P S and f : S Ñ Y , where Y “ C or Y “ R. Then

An open set U Ă C is called path-connected if for every z, w P U , there is a stepwise

The set S :“ tz P C : |z| ‰ 1u “: CzT is not a domain. Although S is open, it

1.4. The exponential function and kith

1.4.1. The exponential exp z.

Proof. (2) If z1 “ x1 ` iy1 and z2 “ x2 ` iy2 , where x1 , x2 , y1 , y2 P R, then

Exercise 1.49. Find all z P C that satisfy exp z “ πi.

1.4.2. Trigonometric functions. We now extend the real trigonometric func-

For all z P C, psin zq2 ` pcos zq2 “ 1, since

psin zq2 ` pcos zq2 “ p exppizq´expp´izq

1.4.3. Logarithm function. In the real setting, given a positive y, log y P R is

For example, Log p´iq “ log | ´ i| ` iArgp´iq “ log 1 ´ π2 i “ 0 ´ π2 i “ ´ π2 i. We

Continuity of Log on Czp´8, 0s. If I Ă R is an interval, then the set I ˆt0u Ă C

The principal value of ab for a P Czt0u and b P C. Recall that if a P Czt0u,

For example, the principal value of ii is

Exercise 1.61. Find Log p1 ` iq.

In this chapter we will learn three main things:

‚ The definition of complex differentiability, i.e., for f : U Ñ C, where U is an open

‚ The Cauchy-Riemann equations, Bu Bv Bu Bv

Vice versa, if the Cauchy-Riemann equations are satisfied at z0 P U and u, v are

The central result in this chapter is that a function is complex differentiable at a

2.1. Complex differentiability

Exercise 2.1. Let U be an open subset of C, z0 P U , and h1 , h2 : U ztz0 u Ñ C be such

For an open set U Ă C and an f : U Ñ C, it is clear that f is continuous at z0 P U

Similarly, v is constant in D. Consequently, f “ u ` iv is constant in D.