Complex Analysis

Summary of the lecture Complex Analysis by

Prof. Özlem Imamoglu

Andreas Compagnoni

Contributions by:
Alberto De Stefani
Simeon Barbey
Viktor Kusar
Noah Talarico

ETH Zürich - December 23, 2024

Preface

This work follows quite accurately the course taught by Professor O.Imamoglu at
ETH Zürich in the years 2023 and 2024, based once again on the book by Stein and
Shakarchi [SS10], properly cited during the proceedings. The material has then under-
gone some changes in the notation and also some slight changes in the reformulation of
a few concept. It does not contain nevertheless any change in the content itself, which
is a good thing, given the title on the title-page.

In case the reader shall notice any imprecision, mistake, typo or similar, we kindly
encourage the reader to report them by sending an e-mail to:

acompagnoni@ethz.ch

specifying “Complex Analysis - ” in the object, immediately followed by the topic under
discussion. Please note that “relatively long” times of response ought to be expected.

Two quick remarks about the notation:

(i) It is possible that the reader might still see some parts of this work being written
in red. These are my comments or remarks about the content: they might of
interest (expecially to understand the notation), but are not necessarily content
covered by the Professor in class. This also applies to the Appendix B.

(ii) To avoid misinterpretation of notation, when possible, hence when a line ends with
a mathematical symbol with nothing following (except a displayed mathematical
formula), the last punctuation symbol of the above mentioned line is omitted.

The webpage of the lecture is available here:

Funktionentheorie/Complex Analysis Autumn 2024

We shall conclude by showing the very last blackboard of the course, to keep in mind
while reading.

iii
iv

I wish the reader an interesting view.

Sincerely,
Andreas Compagnoni
Contents

0 Introduction 1

1 Preliminaries to Complex Analysis 3

1.1 The complex numbers and the complex plane . . . . . . . . . . . . . . 3
1.2 Holomorphic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.1 Definition and basic properties . . . . . . . . . . . . . . . . . . . 12
1.2.2 Cauchy-Riemann Equations . . . . . . . . . . . . . . . . . . . . 17
1.2.3 Power series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 Complex Line Integrals (Integrals along curves) . . . . . . . . . . . . . 26

2 Cauchy’s Theorem and its applications 35

2.1 Goursat’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2 Local existence of primitives and Cauchy’s theorem in a disc . . . . . . 41
2.3 Cauchy’s Integral Formulas . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4 Sequences of holomorphic functions . . . . . . . . . . . . . . . . . . . . 71
2.5 Holomorphic functions defined in terms of integrals . . . . . . . . . . . 78

3 Meromorphic functions and Residue Formula 81

3.1 Zeroes and poles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2 Meromorphic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.3 Applications of the Residue Theorem . . . . . . . . . . . . . . . . . . . 115
3.4 Homotopy and simply connected domains . . . . . . . . . . . . . . . . 122
3.5 The Homotopy Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.6 Complex Logarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.7 Winding numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
3.8 Cauchy Integral Formula . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3.9 Conformal maps and the Riemann mapping Theorem . . . . . . . . . . 152

A The analytic continuation of the Riemann Zeta Function 179

B Other results 185

B.1 Substitution rule in complex line integration . . . . . . . . . . . . . . . 185
B.2 L’Hôpital’s Rule in C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
B.3 Properties of the Residue . . . . . . . . . . . . . . . . . . . . . . . . . . 187

v
vi CONTENTS

B.4 A winding number algorithm . . . . . . . . . . . . . . . . . . . . . . . . 189

References 193
Chapter 0

Introduction

Our goal this semester is to study functions

f :C→C

defined on the complex plane C, or on an open set Ω of C.

We will see that the study of Complex Function Theory is not simply the study of
functions on R2 : in fact, the theory of functions of one real variable is in many ways
more complicated than the theory of functions of a complex variable.

To give on idea of what I mean let’s try to compare and contrast:

1. It is not too difficult to find a function of a real variable that is in Dn (R) but not
in D∞ (R). Consider

x sin x12 , x ∈ R \ {0}

 2

f : R → R, x 7→ f (x) =
0 , x ∈ {0}

The derivative of f exists for every x ∈ R, including x = 0, with f ′ (0) = 0.

Hence, f is differentiable, but its derivative is not continuous, therefore it’s not
differentiable twice.
By integrating f as many times as one likes, one can obtain a function h, that is
differentiable that many times, but not infinitely differentiable.
In contrast: we will see that if f : C → C is differentiable once, then it is
differentiable infinitely many times.

2. There are functions f : R → R that are infinitely many times differentiable, whose
Taylor series does not represent f , i.e. f is not analytic. E.g.

exp −1


2 , x ∈ R \ {0}
f : R → R, x 7→ f (x) = x
0 , x ∈ {0}

1
2 CHAPTER 0. INTRODUCTION

Then f is infinitely differentiable. Unfortunately, at x = 0 all derivatives are zero.

Hence, its Taylor series is identically zero and cannot represent f
In contrast: if f : C → C is a function of a complex variable, which is differen-
tiable, then f is analytic, i.e. it can be represented by a power series (differentiable
= analytic).

3. There are plenty of C ∞ (R) functions of a real variable that are bounded, e.g.
sin(x), cos(x)
In contrast: we will see that if f : C → C is differentiable and bounded, then it
is constant (Liouville’s Theorem 2.8)

4. For two functions of a real variable f, g, they both can ”agree” (be equal) on an
open set without being equal.
In contrast: if f, g : C → C are two differentiable functions which coincide on
an arbitrarily small disc (or even a convergent sequence (zn )n∈N ), then f = g
(Analytic continuation principle 2.10)
Remark 0.1. The power of Complex Function Theory comes from this ”robustness”
or rigidity. It is a field in which, in some sense, Analysis, Geometry and Algebra come
together.
This, we will see, allows one to prove Theorems that a priori seem to have nothing to
do with complex numbers.
Example 0.1. 1. The integral
Z ∞ Z ∞
√
2 2 2π
cos(t )dt = sin(t )dt =
0 0 4

2. Let π(x) := #{p ∈ P : p ≤ x} with P denoting the set of prime numbers. Then
x
π(x) ≈x→∞
log(x)
Result know as the Prime Number Theorem A.5, as
π(x)
lim x =1
x→∞
log(x)

3. If f ∈ C[X] is a non-zero polynomial, then f has a zero in C (Fundamental

Theorem of Algebra 2.5, not valid in R for instance).

4. Let r4 (n) := #{(m1 , ..., m4 ) ∈ Z4 : 4l=1 m2l = n}, then r4 (n) = 8 d|n d
P P
4∤d

Before we start with the definition of differentiability of a function of a complex variable,

we recall the definitions and basic properties of complex numbers.
Chapter 1

Preliminaries to Complex Analysis

1.1 The complex numbers and the complex plane

Definition 1.1. The set of complex numbers is

C := {x + iy : x, y ∈ R and i2 = −1}

We consider R ⊆ C using the following fact:

∃ιR : R → C, r 7→ r + i · 0

Definition 1.2. For z = x + iy ∈ C we define

Re(z) := x

as the real part of z and

Im(z) := y

3
4 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

as the imaginary part of z. Moreover, we define the complex conjugate of z as

z̄ := x − iy

Proposition 1.1. For z ∈ C it holds that:

z+z̄
(i) Re(z) = 2
z−z̄
(ii) Im(z) = 2i

(iii) z ∈ R ⇐⇒ z = z̄

(iv) z ∈ iR ⇐⇒ z = −z̄

2x x+iy+x−iy z+z̄
Proof. (i) Re(z) = x = 2
= 2
= 2

2iy x+iy−x+iy z−z̄

(ii) Im(z) = y = 2i
= 2i
= 2i

(iii) We prove both directions of the equivalence:

=⇒: Assume z ∈ R, then z = x + i0 = x − i0 = z̄

⇐=: Assume z = z̄, then z − z̄ = 0 = x + iy − (x − iy) = 2iy. This means that
y = 0 and hence that z ∈ R

(iv) We prove once again both directions of the equivalence:

=⇒: Assume z ∈ iR, then z = 0 + iy = 0 − (−iy) = −z̄

⇐=: Assume z = −z̄, then z + z̄ = 0 = 2x. This means that x = 0 and
consequently that z ∈ iR

Definition 1.3. For z ∈ C, if z = Re(z), then z is said to be (purely) real, whereas

if z = Im(z), then z is said to be purely imaginary.

Algebraic Structure of C
Complex numbers can also be represented as ordered pairs of real numbers in R2 , so
for z ∈ C we have that
z∼= (x, y)
1.1. THE COMPLEX NUMBERS AND THE COMPLEX PLANE 5

and where for another complex number w ∈ C we have that z = w with w ∼

= (u, v) ⇐⇒
x = u and y = v. Moreover, we defined the following operations:

Addition in C: if z = x + iy and w = u + iv, then

z + w := (x + u) + i(y + v)

or as pairs in R2
z+w ∼
= (x + u, y + v)

Multiplication in C: if z = x + iy and w = u + iv, then

z · w := (x + iy) · (u + iv) = xu + i(xv + yu) + i2 yv = (xu − yv) + i(xv + yu)

or as pairs in R2
z·w ∼ = (xu − yv, xv + yu)
Note that i ∼
= (0, 1) and (0, 1) · (0, 1) = (−1, 0) ∼
= −1, as i2 = −1

R2 with these two operations +, · becomes a field, i.e. (R, +, ·) satisfies the following:

• (R2 , +) is an abelian group with additive identity 0 ∼

= (0, 0)

• R2 \ {(0, 0)}, · is an abelian group with multiplicative identity 1 ∼

= (1, 0)

• The (commutative) distributive law holds:

∀z1 , z2 , z3 ∈ C : z1 (z2 + z3 ) = z1 z2 + z1 z3 = (z2 + z3 )z1

Therefore, complex numbers form a 2-dimensional commutative algebra over R: in this

case “∼
=” can be substituted with with “=”. Interchanging hence the meaning of R2
between field structure and the simple usual real vector space, we can prevent any
abuse of notation, leaving then any other accentuation of the difference between these
structures only to possibly increase the clarity in the proceedings.

Definition 1.4. The real number

√ p
|z| := z z̄ = x2 + y 2 ∈ R

is called the norm, modulus or absolute value of z ∈ C

Additive inverse of z ∈ C:
−z ∼
= (−x, −y)
6 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Multiplicative inverse of z ∈ C:
 
z̄ x −y
z −1
= 2 ∼= , 2
|z| x + y x + y2
2 2

Polar coordinates representation of complex numbers

We also have the following polar coordinates representation of complex numbers:

z = x + iy = r cos(θ) + i sin(θ) = reiθ

where r ≥ 0, θ ∈ R with |z| = r, x = r cos(θ) and y = r sin(θ)

The polar representation is not unique unless z ̸= 0 and we restrict θ ∈ (−π, π] (or any
other interval of length 2π).

Definition 1.5. The number θ ∈ R is called the argument of z ∈ C, which is

defined uniquely up to a multiple of 2π and is denoted by

arg(z) := θ ∈ R : z ∈ |z|eiθ


From this we define then

Definition 1.6. The argument of z ∈ C chosen in the interval (−π, π] is called the
principal argument of z ∈ C and denoted by Arg(z) ∈ arg(z)

π
It holds that: Arg(i) = 2
and ∀c ∈ (0, +∞) : Arg(−c) = π

Remark 1.1. No assignment of argument is made to 0 ∈ C; therefore we often consider

C∗ = C \ {0}, also known as the group of units of C
1.1. THE COMPLEX NUMBERS AND THE COMPLEX PLANE 7

Proposition 1.2. Let z = x + iy ∈ C∗ , then

  
y


 arcsin |z|
, x ∈ R≥0
  
y
Arg(z) = π − arcsin |z| , x ∈ R<0 , y ∈ R≥0
  
 −π − arcsin y

, x, y ∈ R<0

|z|

Proof. Here is enough to observe that for t ∈ [−1, 1], arcsin is the unique number
u ∈ −π , π such that sin(u) = t. By using the various mirroring in the unit circle
2 2
(indeed we divided by |z|), we obtain the result.

Proposition 1.3. For z ∈ C, it holds that

arg(z −1 ) = − arg(z)
arg(zw) = arg(z) + arg(w)

Proof. The proofs of these equalities are direct

 
−1 −1 −1 z̄
iθ
θ ∈ R : 2 = |z −1 |eiθ

arg(z ) = θ ∈ R : z = |z |e = =
|z|
 
z̄ 1 iθ
= θ ∈ R : z̄ = |z|eiθ =

= θ∈R: 2 = e
|z| |z|
= θ ∈ R : z = |z|e−iθ = −θ ∈ R : z = |z|eiθ =
 

= − θ ∈ R : z = |z|eiθ = − arg(z)


where θ is defined with respect to z −1 and where we used that |z −1 | = 1

|z|
as consequence
of a simple computation. Moreover, we have that
arg(zw) = θ ∈ R : zw = |zw|eiθ , θ = θz + θw =


= θz + θw ∈ R : zw = |zw|ei(θz +θw ) =


= θz + θw ∈ R : zw = |z|eiθz |w|eiθw =


= arg(z) + arg(w)

Remark 1.2. Despite these results, it is not always the case that for z ∈ C we have
Arg(z −1 ) = − Arg(z)
Arg(zw) = Arg(z) + Arg(w)
as shown in the following example.
8 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Example 1.1.
It holds that Arg −1

2
= π ̸= − Arg(−2) and also that π = Arg(−1) =
Arg (−i)(−i) ̸= Arg(−i) + Arg(−i) = − π2 − π2 = −π, hence the properties mentioned
above do not hold for Arg

Matrix representation of complex numbers

This representation is obtained by associating a 2 × 2 matrix to a complex number
written in standard form; some properties of the operations between matrices translate
directly into the ones in C, as exposed below.
   
a −b c −d
For z = a + ib ∈ C, let Z = while for w = c + di, let W = , then
b a d c
    
a −b c −d ac − bd −(bc + ad)
ZW = =
b a d c bc + ad ac − bd

On the other hand, we have zw = (ac − bd) + i(bc + ad). The multiplication in C hence
corresponds to the multiplication of the respective matrices in R2×2

We can represent any z ∈ C with the matrix

     
Re(z) −Im(z) 1 0 0 −i
Z= = Re(z) + Im(z)
Im(z) Re(z) 0 1 i 0
   2  
0 −1 0 −1 1 0
where i ∈ C corresponds to and =−
1 0 1 0 0 1
iθ
In polar form z = re , the corresponding matrix is
    
cos(θ) − sin(θ) r 0 cos(θ) − sin(θ)
Z=r =
sin(θ) cos(θ) 0 r sin(θ) cos(θ)

Topological results

Proposition 1.4. The following properties hold for the complex norm | · |:

(i) ∀z ∈ C : |z| = 0 ⇐⇒ z = 0

(ii) ∀z1 , z2 ∈ C : |z1 | − |z2 | ≤ |z1 − z2 | ≤ |z1 | + |z2 |

(iii) ∀z1 , z2 ∈ C : |z1 z2 | = |z1 ||z2 |

(iv) ∀z ∈ C : |z̄| = |z|

(v) ∀z ∈ C : |Re(z)| ≤ |z| and |Im(z)| ≤ |z|

1.1. THE COMPLEX NUMBERS AND THE COMPLEX PLANE 9

Proof. (i) It is very convenient to prove both implications directly:

√
⇐=: This direction is trivial, consider z = 0, then |z| = 02 + 02 = 0
p
=⇒: 0 = x2 + y 2 = |z| ⇐⇒ x2 + y 2 = 0 with both x2 , y 2 ≥ 0. This implies that
x, y = 0 and with it that z = 0

(ii) First, we prove the triangular inequality:

1
∀z1 , z2 ∈ C : |z1 − z2 | = |z1 |2 + |z2 |2 + 2(x1 x2 + y1 y2 ) 2 ≤

1
≤ |z1 |2 + |z2 |2 + 2(|z1 ||z2 | 2 =

2  21
= |z1 | + |z2 | =
= |z1 | + |z2 |

in which we used the Cauchy-Schwarz Inequality in C. Applying then this

initial result, we obtain

|z1 − z2 + z2 | ≤ |z1 − z2 | + |z2 |

⇐⇒ |z1 | − |z2 | ≤ |z1 − z2 |

and since this holds for all z1 , z2 ∈ C, then by swapping z1 and z2 we obtain the
same inequality in absolute value, namely

|z1 | − |z2 | ≤ |z1 − z2 |

√ √ √ √
(iii) Let z1 , z2 ∈ C, then |z1 z2 | =
z1 z2 z¯1 z¯2 = z1 z¯1 z2 z¯2 = z1 z¯1 z2 z¯2 = |z1 ||z2 |
p
(iv) For z ∈ C we have |z̄| = |x − iy| = x2 + y 2 = |x + iy| = |z|

(v) For z ∈ C we have

|Re(z)|2 = x2 ≤ x2 + y 2 = |z|2
|Im(z)|2 = y 2 ≤ x2 + y 2 = |z|2
√
Applying the · function, which is a monotonically increasing function, we obtain
the wished result.

Proposition 1.5. If z = reiθ ∈ C∗ and w = seiν ∈ C∗ , then

zw = rsei(θ+ν) ∈ C∗
10 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Proof. This result follows from simple multiplications rules:

zw = reiθ seiν = rsei(θ+ν)

Since both r, s are non-zero then also their product is not, hence zw ∈ C∗
Next, we recall some definitions that we need from Topology and Analysis.

Definition 1.7. We denote the open disc of radius r > 0 centred at z with
Dr (z) or D(z, r) and the closed disc of radius r ≥ 0 centred at z with Dr (z) or
D(z, r). They are both defined as follows

Dr (z) := {w ∈ C : |w − z| < r}
Dr (z) := {w ∈ C : |w − z| ≤ r}

The boundary of Dr (z) is the circle

Cr (z) := ∂Dr (z) := Dr (z) \ Dr (z) = {w ∈ C : |w − z| = r}

Remark 1.3. If r > 0, then ∀z ∈ C : Dr (z) = Dr (z), hence in this case the closed disc
is equal to the closure of the open disc.

Definition 1.8 (Open set). A subset U ⊆ C is open, if

∀z ∈ U ∃r > 0 : Dr (z) ⊆ U

The set of all such open subsets of C is called the standard topology of open
sets of C and denoted by OC

E.g. ∅, C, Dr (z) and H := {z ∈ C : Im(z) > 0}

Definition 1.9 (Closed set). A subset U ⊆ C is closed, if C \ U is open.

E.g. ∅, C, Dr (z), Cr (z) and R

Proposition 1.6. The following equivalence holds:

 
N∗
U is closed ⇐⇒ ∀(zn )n∈N∗ ∈ U : lim zn = z =⇒ z ∈ U
n→∞
1.1. THE COMPLEX NUMBERS AND THE COMPLEX PLANE 11

Definition 1.10. A subset K ⊆ C is compact, if it is closed and bounded, i.e. if it

is closed and if ∃M > 0∀z ∈ K : |z| < M

∗
Proposition 1.7. K ⊆ C is compact, if and only if every sequence (zn )n∈N∗ ∈ U N
has a subsequence that converges to a point in U

E.g. ∅, Dr (z), Cr (z) and [a, b] × [c, d]

Definition 1.11. A subset A ⊆ C is called disconnected, if

∃U, V ∈ OC : (U ∩ V = ∅) and (A ∩ U ̸= ∅) and (A ∩ V ̸= ∅) and (A ⊆ U ∪ V )

A subset A ⊆ C is called connected, if it is not disconnected.

Moreover, a connected open non-empty set ∅ = ̸ U ⊆ C is called a region or domain.

We mention here that in any euclidean space: a connected open set is automatically
open and path-connected and vice versa. Any two distinct points z0 , z1 in an open
connected set A ⊆ C can be connected by a polygonal path lying in A

E.g. ∅, C, Dr (z), Dr (z), Cr (z) and R are connected, whereas Z, Q and R ∪ D1 (2i) are
disconnected.

Definition 1.12 (Convergence of a complex sequence). A sequence (zn )n∈N∗ = (xn +

∗
iyn )n∈N∗ ∈ CN converges to z = x + iy in C, if one of the following equivalent
conditions holds:

(i) limn→∞ xn = x and limn→∞ yn = y in R

(ii) limn→∞ |zn − z| = 0 in R

12 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

(iii) ∀ε > 0∃N ∈ N∗ ∀m, n ≥ N : |zm − zn | < ε, i.e. (zn )n∈N∗ is a complex Cauchy-
sequence.

Definition 1.13 (Limits). Let U ⊆ C be an open subset and f ∈ CU any function.

For z0 ∈ U and w0 ∈ C we have

lim f (z) = w0
z→z0
z∈U

if one of the following equivalent conditions holds:

(i) ∀ε > 0∃δ > 0∀z ∈ U : |z − z0 | < δ =⇒ |f (z) − w0 | < ε

∗
(ii) If (zn )n∈N∗ ∈ U N is a sequence with limn→∞ zn = z0 , then limn→∞ f (zn ) = w0

Definition 1.14 (Continuity of a function). A function f ∈ CU is continuous on

U , if and only if
∀z0 ∈ U : lim f (z) = f (z0 )
z→z0

that is, if and only if

∗
∀(zn )n∈N∗ ∈ U N : lim zn = z0 =⇒ lim f (zn ) = f (z0 )
n→∞ n→∞

The set of all continuous functions on an set U is denoted by C 0 (U ) or C 0 (U ; C)

1.2 Holomorphic Functions

1.2.1 Definition and basic properties
This is a central notion for the rest of the class.

Definition 1.15 (Holomorphic function). Let Ω ⊆ C an open set and f ∈ CΩ a

complex valued function on Ω, then

• f is called holomorphic at z0 ∈ Ω, if

f (z) − f (z0 ) f (z0 + h) − f (z0 )

∃f ′ (z0 ) := z→z
lim = lim ∈C
0
z̸=z0
z − z0 h→0 h
h̸=0
1.2. HOLOMORPHIC FUNCTIONS 13

Here h ∈ C, z0 + h ∈ Ω (so that the quotient is well defined).

If the limit exists, we denote it with f ′ (z0 ) and we call it the derivative of f
at z0

• f is called holomorphic on Ω, if ∀z0 ∈ Ω : f is holomorphic at z0 ∈ Ω

• If f is holomorphic on all of C, then it is called entire.

Remark 1.4. Regular or complex differentiable are other words used for holomorphic.
Example 1.2. Let f ∈ CC , f (z) = z. Then f is entire, since
f (z0 + h) − f (z0 ) z0 + h − z0 h
lim = lim = lim = 1
h→0 h h→0 h h→0 h

Hence, it holds that ∀z ∈ C : f ′ (z) = 1

Definition 1.16. The set of all holomorphic functions on Ω is defined as follows

H (Ω) := {f ∈ CΩ : f is holomorphic on Ω}

As in the case of real variables, we have

Proposition 1.8. [SS10, Proposition I.2.2]

(i) The set H (Ω) is a C-vector space. More precisely, if f, g ∈ H (Ω), then

∀α, β ∈ C : αf + βg ∈ H (Ω)

and
∀α, β ∈ C : (αf + βg)′ = αf ′ + βg ′
(The zero-function 0 ∈ CΩ is the zero element of the vector space).

(ii) If f, g ∈ H (Ω), then f g ∈ H (Ω) and

(f g)′ = f ′ g + f g ′

f
(iii) If g(z0 ) ̸= 0, then g
is holomorphic at z0 and
 ′
f f ′ (z0 )g(z0 ) − f (z0 )g ′ (z0 )
(z0 ) =
g g 2 (z0 )

Moreover, if g ∈ H (Ω) and ∀z ∈ C : g(z) ̸= 0, then f

g
∈ H (Ω)
14 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

(iv) If f ∈ U Ω and g ∈ CU are both holomorphic, then g ◦ f ∈ CΩ is holomorphic

and
∀z ∈ Ω : (g ◦ f )′ (z) = g ′ f (z) f ′ (z)

Proof. The claims of this Propositions are very similar to the case of Real Analysis;
therefore here we will only show the fourth.

(iv) Let z0 ∈ Ω and w0 := f (z0 ) ∈ U . Consider F ∈ CΩ and G ∈ CU defined by

 f (z)−f (z0 )
z−z0
, z ̸= z0
F (z) :=
f ′ (z0 ) , z = z0
 g(w)−g(w0 )
w−w0
, w ̸= w0
G(z) :=
g ′ (w0 ) , w = w0

Since
f (z) − f (z0 )
lim F (z) = lim = f ′ (z0 ) = F (z0 )
z→z0 z→z0 z − z0
we have that F is continuous at z0
Similarly, G is continuous at w0 . Hence, since f is differentiable at z0 and hence
continuous at z0 , G ◦ f is continuous at z0

For z ∈ Ω \ {z0 } we have

(g ◦ f )(z) − (g ◦ f )(z0 ) g f (z) − g f (z0 )
= =
z − z0 z − z0
(
)
g f (z) −g(w0 ) f (z)−f (z0 )
= f (z)−w0 z−z0
, f (z) ̸= w0 =
0 , f (z) = w0


= G f (z) F (z)
w0 −f (z)
Note that if f (z) = w0 , then F (z) = z−z0
=0

Hence, we finally obtain

(g ◦ f )(z) − (g ◦ f )(z0 )
(g ◦ f )′ (z0 ) = lim =
z→z0 z − z0


= lim G f (z) F (z) =
z→z0


= G f (z0 ) F (z0 ) =
= G(w0 )F (z0 ) =
= g ′ (w0 )f ′ (z0 ) =
= g ′ f (z0 ) f ′ (z0 )


1.2. HOLOMORPHIC FUNCTIONS 15

since G ◦ f and F are continuous at z0

Remark 1.5. Note that if f ∈ CΩ is complex differentiable at z0 ∈ Ω, then there exists

a complex number c ∈ C such that for all z ∈ C

f (z) = f (z0 ) + c(z − z0 ) + E(z, z0 )

with E ∈ CΩ satisfying
E(z, z0 )
lim =0
z→z0 z − z0
Here we have that c = f ′ (z0 )

Example 1.3. 1. Example 1.2 and Proposition 1.8 applied repeatedly show that any
polynomial p ∈ C[X] is complex differentiable at every point z ∈ C

For p(z) = z n with n ∈ N we have that in z0 ∈ C:

n−1
z n − z0n (z − z0 ) n−1 n−1−k k
P
k=0 z z0 X
′
p (z0 ) = lim = lim = lim z n−1−k z0k = nz0n−1
z→z0 z − z0 z→z0 z − z0 z→z0
k=0

2. Important non-example: Let f (z) = z̄, then

f (z0 + h) − f (z0 ) z0 + h − z̄0 z̄0 + h̄ − z̄0 h̄

= = =
h h h h
For h = t and t ∈ R this limit is 1
For h = it and t ∈ R this limit is −1

Hence limh→0 f (z0 +h)−f

h
(z0 )
does not exists for any z0 ∈ C and consequently f (z) =
z̄ is not holomorphic at any point in C. This procedure can be used more generally
to disprove the existence of a limit, as done here.

Note that f (z) = z̄ as a function R2 → R2 is given by

F : R2 → R2
(x, y) 7→ (x, −y)

Hence it is a linear function and is differentiable with

 
1 0
DF (x0 , y0 ) =
0 −1
16 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Recall: A function

F : R2 → R2


(x, y) 7→ u(x, y), v(x, y)

is differentiable at a point P0 = (x0 , y0 ) if it exists a linear map dF : R2 → R2 such

that
F (P ) − F (P0 ) − dF (P − P0 )
lim =0
P →P0 ∥P − P0 ∥
P ̸=P0

or equivalently

F (P ) − F (P0 ) = dF (P − P0 ) + Ψ(P − P0 )∥P − P0 ∥

0 P →P
with ∥Ψ(P − P0 )∥ −−−→ 0

The linear map dF : R2 → R2 is unique and called the differential of F at P0 . In the

standard basis of R2 , dF is represented by the Jacobian Matrix of F , namely DF :
∂u ∂u
!
∂x ∂y
DF (x, y) = ∂v ∂v
∂x ∂y

Recall: We can view C as a 1-dimensional vector space over C with basis BC = {1} or
as a 2-dimensional real vector space with basis BR = {1, i}

A map T : C → C is C-linear if
T (z) = λz
for some λ ∈ C. On the other hand, a map T : C → C is R-linear if

T (z) = T (x + iy) = xT (1) + yT (i) = λz + µz̄

with
1
λ = (T (1) − iT (i))
2
1
µ = (T (1) + iT (i))
2
z+z̄ z−z̄
using that x = 2
and y = 2i

Hence, every C-linear map T : C → C is R-linear, but a R-linear map T : C → C is

C-linear only if µ = 0, i.e. T (i) = iT (1) (quick and fun to verify).

If T (1) = a + ib and T (i) = c + id for a, b, c, d ∈ R, then T (i) = iT (1) =⇒ b = −c and

a=d
1.2. HOLOMORPHIC FUNCTIONS 17
 
2 x
If we identify C with R with z = x + iy = , since every R-linear map R2 → R2 is
y
given by a 2 × 2 real matrix
    
x a c x
7→
y b d y
| {z }
=:A

such a map is also C-linear if A is of the form

 
a −b
A=
b a

Remark 1.6. Note that in Example 1.3, the function f (z) = z̄ as map R2 → R2 is
differentiable with Jacobian equal to
 
1 0
0 −1
but which is not of the above form.
Our next goal is to see how this Linear Algebra fact about R-linear versus C-linear maps
is reflected in the case of a linear function f : C → C and of its complex differentiability.

1.2.2 Cauchy-Riemann Equations

Let f ∈ CC be holomorphic at z0 . If f (x + iy) = u + iv, via some linking R-linear
isomorphism we can also view f as a map from R2 to R2 , such as

f˜ : R2 → R2


(x, y) 7→ u(x, y), v(x, y)

As specified in 1.1, we can identify the two spaces R2 and C as equal, hence f˜ ∼ = f =⇒
˜
f =f
This said, the derivative limz→z0 f (z)−f
z−z0
(z0 )
exists independently of how z → z0

In particular, we can have z tending to z0 along the line z = x + iy0 by letting x → x0 ,

hence
f (x + iy0 ) − f (x0 + iy0 )
f ′ (z0 ) = lim =
x→x0 x − x0
f˜(x, y0 ) − f˜(x0 , y0 )
= lim =
x→x0 x − x0
u(x, y0 ) − u(x0 , y0 ) v(x, y0 ) − v(x0 , y0 )
= lim + i lim =
x→x0 x − x0 x→x0 x − x0
= ux (x0 , y0 ) + ivx (x0 , y0 )
18 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

We can conclude from this that the usual partial derivatives ux (z0 ), vx (z0 ) exist and
hence also the partial derivative fx (z0 ) = ux (z0 ) + ivx (z0 ) exists and that

f ′ (z0 ) = ux (z0 ) + ivx (z0 ) = fx (z0 )

On the other hand approaching z0 = x0 + iy0 along z = x0 + iy with y → y0 gives

f (x0 + iy) − f (x0 + iy0 )
f ′ (z0 ) = lim =
y→y0 y − y0
f˜(x0 , y) − f˜(x0 , y0 )
= lim =
y→y0 y − y0
v(x0 , y) − v(x0 , y0 ) u(x0 , y) − u(x0 , y0 )
= lim − i lim =
x→x0 y − y0 y→y0 y − y0
= uy (x0 , y0 ) − ivy (x0 , y0 ) =
= vy (z0 ) − iuy (z0 )
We obtain that the partial derivatives uy (z0 ), vy (z0 ) also exist together with fy (z0 ) =
uy (z0 ) + ivy (z0 ) and that

f ′ (z0 ) = vy (z0 ) − iuy (z0 ) = −ify (z0 )

By pulling together all previous results we obtain the Cauchy-Riemann Equations

(CR)
ux (z0 ) = vy (z0 )
uy (z0 ) = −vx (z0 )
If we introduce two differential operators
 
∂ 1 ∂ ∂
:= −i
∂z 2 ∂x ∂y
 
∂ 1 ∂ ∂
:= +i
∂ z̄ 2 ∂x ∂y
What we have shown can be summarised in

Proposition 1.9. [SS10, Proposition I.2.3] If f is holomorphic at z0 , then

∂f
(z0 ) = 0
∂ z̄
and
∂f ∂u
f ′ (z0 ) = (z0 ) = 2 (z0 )
∂z ∂z
If we write f (z) = f˜(x, y), then f˜ : R2 → R2 , (x, y) 7→ u(x, y), v(x, y) is complex


1.2. HOLOMORPHIC FUNCTIONS 19

differentiable with
     
ux uy ux uy ux −vx
Df˜(x0 , y0 ) = = =
vx vy −uy ux vx ux

and

˜
 u u 2
det Df (x0 , y0 ) = x y = f ′ (z0 ) = u2x (z0 ) + u2y (z0 ) = u2x (z0 ) + vx2 (z0 )
vx vy

Proof. Using the Cauchy-Riemann Equations we have

 
∂f 1 ∂ ∂ 1 1

= +i (u + iv) = (ux + ivx + iuy − vy ) = (ux − vy ) + i(vx + uy ) = 0
∂ z̄ 2 ∂x ∂y 2 2
By summing together the Cauchy-Riemann Equations and dividing by 2, one can obtain
1
f ′ (z0 ) = fx (z0 ) − ify (z0 )


2
These equations also give
∂f 1 ∂u
(z0 ) = ux + ivx = 2 (ux − iuy ) = 2 (z0 )
∂z 2 ∂z
If z0 = x0 + iy ∈ U and h = h1 + ih2 ∈ C, then for f being holomorphic at z0 means
that
f (z0 + h) = f (z0 ) + f ′ (z0 )h + hE(h)
with limh→0 E(h) = 0. If f ′ (z0 ) = a + ib, then
f ′ (z0 )h = (a + ib)(h1 + ih2 ) = ah1 − bh2 + i(bh1 + ah2 )
Hence, if we write f˜(x, y) = f (z) for h̃ = (h1 , h2 )T , we have that
  
a −b h1
f˜ (x0 , y0 ) + (h1 , h2 ) − f˜(x0 , y0 ) −


b a h2 |h̃|→0
−−−→ 0
|h̃|
This means that f˜ : R2 → R2 is differentiable with a C-linear differential
       
˜ h1 a −b h1 ux uy h1
Df (x0 , y0 ) = =
h2 b a h2 vx vy h2
Using a = ux = vy and b = vx = −uy we get
   
˜ a −b ux uy
Df (x0 , y0 ) = =
b a vx vy
Consequently, computing the determinant results in
 
2
det Df˜(x0 , y0 ) = u2x + vx2 = f ′ (z0 )
20 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Remark 1.7. Recall that we can represent any complex number z = a + bi with a 2 × 2
real matrix:  
a −b
a + ib ←→
b a
If f (z) = u(z) + iv(z) is complex differentiable at z0 , then f ′ (z0 ) = ux (z0 ) + ivx (z0 ) has
matrix representation  
′ ux (z0 ) −vx (z0 )
f (z0 ) ←→
vx (z0 ) ux (z0 )
On the other hand, the corresponding function

f˜ : R2 → R2 , (x, y) 7→ u(x, y), v(x, y)

has a Jacobian matrix  

u u
Df˜(x, y) = x y
vx vy
Comparing these two matrices, we get exactly the Cauchy-Riemann equations:

ux = vy
vx = −uy

If one remembers the general form of the Jacobian of a function g : R2 → R2 and the
matrix representation of a complex number, then can remember the Cauchy-Riemann
equations.

The previous Proposition 1.9 shows that f holomorphic implies that ∂f

∂ z̄
= 0 using the
Cauchy-Riemann Equations, hence that f satisfies the same Cauchy-Riemann Equa-
tions. We also have the following partial converse.

Theorem 1.1. [SS10, Theorem I.2.4] Suppose f = u + iv ∈ CΩ for an open set

Ω ⊆ C. If u, v ∈ C 1 (Ω; R) and satisfy the Cauchy-Riemann Equations (CR) in Ω,
then f ∈ H (Ω) and
df ∂f
f ′ (z) = (z) = (z)
dz ∂z

Proof. Let z0 = x0 + iy0 ∈ Ω and h = h1 + ih2 ∈ C. Having u, v ∈ C 1 (Ω; R) implies

that
u(z0 + h) − u(z0 ) = ∂x u(z0 )h1 + ∂y u(z0 )h2 + |h|ε1 (h)
h→0
with ε1 (h) −−→ 0. Similarly,

v(z0 + h) − v(z0 ) = ∂x v(z0 )h1 + ∂y v(z0 )h2 + |h|ε2 (h)

1.2. HOLOMORPHIC FUNCTIONS 21

h→0
with ε2 (h) −−→ 0. By then summing these two together we obtain

f (z0 + h) − f (z0 ) = (u + iv)(z0 + h) − (u + iv)z0 =

= (∂x u + i∂x v)h1 + (∂y u + i∂y v)h2 + |h|ε(h)

h→0
where ε(h) = (ε1 + ε2 )(h) −−→ 0. Using now the Cauchy-Riemann Equations we can
reform the previous expansion in what follows

f (z0 + h) − f (z0 ) = (∂x u − i∂y u)h1 + (∂y u + i∂x u)h2 + |h|ε(h) =

= (∂x u − i∂y u)(h1 + ih2 ) + |h|ε(h)

h→0
Hence f (z0 + h) − f (z0 ) = (∂x u − i∂y u)h + |h|ε(h) where ε(h) −−→ 0

This says that

f (z0 + h) − f (z0 ) h→0
−−→ ∂x u − i∂y u
h
Hence f ′ (z0 ) exists and is equal to ∂x u − i∂y u = 2∂z u = ∂z f

Example 1.4. Let f ∈ CC such that z 7→ f (z) = x2 + y 2 + 2ixy, considering that

u(z) = x2 + y 2 ∈ R and v(z) = 2xy ∈ R. We analyse the partial derivatives to see
whether and where they satisfy the Cauchy-Riemann equations:

∂x u(z) = 2x ∂x v(z) = 2y
∂y u(z) = 2y ∂y v(z) = 2x

It holds that for all z ∈ C we have ∂x u = 2x = ∂y v, while ∂y u = 2y = −2y = −∂x v

is only true if y = 0. f therefore satisfies the Cauchy-Riemann equations only when
Im(z) = 0. Hence, f is holomorphic only for points on the real axis. For these points

f ′ (x0 ) = ∂x u(x0 ) + i∂x v(x0 ) = 2x0

Remark 1.8. Many books distinguish between complex differentiability at a point and
holomorphicity at a point as follows:

• Complex differentiability at a point is given, if the limit in 1.15 exists at that

point.

• Holomorphicity at a point is instead given instead when the limit in 1.15 exists in
a neighbourhood of that point.

[SS10] does not make such distinction (and in this course we will not need it).
22 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

A quick summary
Let Ω be an open subset of C, then
1. For f ∈ CΩ with f (z) = u(z) + iv(z) we have that if f is holomorphic on Ω, then
u, v satisfy the Cauchy-Riemann Equations ux = vy and uy = −vx . Moreover it
holds that
f ′ (z) = ux (z) − iuy (z)

2. If u, v ∈ C 1 (Ω; C) and satisfy the Cauchy-Riemann Equations, then f = u + iv is

holomorphic.

3. If we write f˜(x, y) = f (z), for f ∈ H (Ω) and identifying C with R2 , then

f˜ : R2 → R2 is differentiable with
 
˜ ux −vx 
˜

2
Df (x0 , y0 ) = and det Df (x0 , y0 ) = f ′ (z0 )
vx ux
 
a −b
Remark 1.9. A matrix of the form defines a linear map
b a

L : R2 → R2
    
x a −b x
7→
y b a y

which preserves angles and orientation, i.e. it is a rotation and a dilation. If a + ib ̸= 0

with a + ib = |a + ib|eiθ , then it is a rotation by the angle θ and a dilation by |a + ib|
Our next result gives important examples of holomorphic functions.

1.2.3 Power series

Recall: A (complex-) power series is a series of the form
∞
X
an z n
n=0

with an ∈ C for all n ∈ N and z ∈ C

P∞
Theorem 1.2. [SS10, Theorem I.2.5] Let n=0 an z n be a power series. Then ∃R ∈
[0, +∞] such that

(i) if |z| < R, the series converges absolutely.

(ii) if |z| > R, the series diverges.

1.2. HOLOMORPHIC FUNCTIONS 23

1 1
Moreover, with the convention that 0
:= ∞ and ∞
:= 0, R is given by

1 1
= lim sup |an | n
R n→∞

in this case R is called the radius of convergence and

DR (0) := {z ∈ C : |z| < R}

is called disc of convergence (or region of convergence).

Proof. Exercise (as in Real Analysis).

Example 1.5 (Exponential function). An important example of a power series is the
complex exponential function
∞
z
X zn
e :=
n=0
n!
This series converges absolutely for all z ∈ C. Also,
∞
X |z|n
z
|e | ≤ = e|z| < ∞
n=0
n!
Hence, ez convergence uniformly on compact subsets of C
The following Theorem shows that ez in particular and power series in general give
examples of holomorphic functions in their disc of convergence.

Theorem 1.3. [SS10, Theorem I.2.6] The power series f (z) = ∞ n

P

an z defines a
n=0
holomorphic function in its disc of convergence, i.e. f ∈ H DR (0) , and
∞
X
′
f (z) = nan z n−1
n=1

Moreover, f ′ has the same radius of convergence as f , i.e. f ′ ∈ H DR (0)

Proof. Let R be the radius of convergence of f , since

1
lim n n = 1
n→∞

then it holds that1

1 1
lim sup |nan | n = 1 · lim sup |an | n = R
n→∞ n→∞
1
If {an }n∈N and {bn }n∈N are two sequences of non-negative real numbers and b := limn→∞ bn is
the limit of the second, then lim supn→∞ an bb = (lim supn→∞ an )(lim supn→∞ bn ). In general, lim sup
is submultiplicative for sequences of non-negative real numbers.
24 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Hence , ∞ n−1
P
n=1 nan z has the same radius of convergence. Repeated applications of
this same argument show that the sum
∞
X n!
an z n−k
n=k+1
(n − k − 1)!

has radius of convergence R for any k ∈ N

f (z+h)−f (z) P∞ h→0
We now want to show that h
− n=1 nan z n−1 −−→ 0 in the following way.

Let now z ∈ DR (0) ⊆ C and choose δ > 0 such that |z| + δ < R, e.g. one can take
δ = R−|z|
2
, with h ∈ C such that |h| < δ, then

∞ ∞ 
an (z + h)n − an z n

f (z + h) − f (z) X n−1
X
n−1
− nan z = − nan z ≤
h n=1 n=1
h
∞ n  
! !
X 1 X n
≤ |an | hk z n−k − z n − nz n−1 =
h k Bin
n=1 k=0
∞ n  
X X n
= |an | hk−1 z n−k ≤
k Bin
n=2 k=2
∞ n  
X X n
≤ |an | |h|k−1 |z|n−k ≤
k Bin
n=2 k=2
1 ∞ n  
(∗ ) X X n−2
≤ |an |n(n − 1) |h|k−2 |z|n−k |h| ≤
k − 2 Bin
n=2 k=2
∞
X
n−2
≤ |an |n(n − 1) |z| + |h| |h| ≤
n=2
∞ n−2
(∗2 )

X R + |z|
≤ |h| |an |n(n − 1)
n=2
2
| {z }
independent of h

     
1 n n n−1 n(n−1) n − 2
using in the (∗ )-step that ∀k ≥ 2 : = k = k(k−1) ≤
k Bin k − 1 Bin k − 2 Bin
 
n−2
n(n − 1) and using lastly in the (∗2 )-step that |h| < R−|z|
k − 2 Bin 2

We therefore obtain
∞  n−2
X R + |z| h→0
|h| |an |n(n − 1) −−→ 0
n=2
2
1.2. HOLOMORPHIC FUNCTIONS 25

and so we finally get that

∞
f (z + h) − f (z) h→0 X
−−→ nan z n−1
h n=1

which concludes the proof.

Example 1.6. 1. The exponential function
exp : C → C
∞
X zn
z 7→ exp(z) :=
n=0
n!

converges on all of C and as such is holomorphic on all of C, moreover we can

easily extract
∞ ∞ ∞
′
X nz n−1 X z n−1 X zn
exp (z) = = = = exp(z)
n=1
n! n=1
(n − 1)! n=0 n!

2. Trigonometric functions
∞
eiz − e−iz X (−1)n z 2n+1
sin(z) := =
2i n=0
(2n + 1)!
∞
eiz + e−iz X (−1)n z 2n
cos(z) := =
2 n=0
(2n)!

with 2 2
ei − e−i e2 − 1
 
sin(i) = =i
2i 2e
and 2 2
ei + e−i e−1 + e e2 + 1
cos(i) = = =
2 2 2e
For the records, these functions are not bounded.
3. The series ∞
X zn
n=1
n2
P∞ 1
has convergence radius 1, i.e. it converges for all z ∈ D1 (0), since n=1 n2 <∞
4. The geometric series
∞
X
zn
n=0

converges for z ∈ D1 (0)

26 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

5. The series ∞
X zn
(−1)n−1
n=1
n
converges for all z ∈ D1 (0). Moreover, for z = 1 the series converges (Leibniz’
criteria), while for z = −1 the series diverges (Harmonic series).

1.3 Complex Line Integrals (Integrals along curves)

We start by recalling the main definitions and properties of curves.

Definition 1.17. • A parametrised

curve in C is a continuous function γ :
0
[a, b] → C, i.e. γ ∈ C [a, b] , where [a, b] is a closed interval in R

• A smooth curve is a curve

γ : [a, b] → C
t 7→ γ(t) = x(t) + iy(t)

such that its derivative

γ ′ (t) = x′ (t) + iy ′ (t)
exists for all t ∈ [a, b], γ ∈ C 1 [a, b] and γ ′ (t) ̸= 0 for all t ∈ [a, b]

Here, we consider
γ(a + h) − γ(a)
γ ′ (a) := lim
h↘0 h
and
γ(b + h) − γ(b)
γ ′ (b) := lim
h↗0 h
as the right and left derivatives respectively.

• A piecewise smooth curve
is a curve γ : [a, b] → C such that γ is continuous

on [a, b], i.e. γ ∈ C 0 [a, b] , and exist points

a = a0 < a1 < ... < an = b

such that γ is smooth on each interval [ak , ak+1 ]

• A closed curve is a curve γ : [a, b] → C with γ(a) = γ(b)

• A curve is simple if it’s not self intersecting, i.e. γ(t) ̸= γ(s) unless s = t or
s = a and t = b
1.3. COMPLEX LINE INTEGRALS (INTEGRALS ALONG CURVES) 27

• γ̃ is called reparametrisation of γ : [a, b] → C if
there exists a continuously

differentiable bijective function σ ∈ C 1 [c, d]; [a, b] with ∀t ∈ [c, d] : σ ′ (t) >
0 and with γ̃ = γ ◦ σ (the condition σ ′ > 0 means that the orientation is
preserved), i.e. γ and γ̃ represent the same geometric object with different
parametrisation.

Remark 1.10. For us in this course the curves will always be piecewise smooth. From
now on when we say “a curve” we mean “a piecewise smooth one”, even if we forget to
write it.

Remark 1.11. We will often work with a particular parametrisation, since most im-
portant notions will be independent of parametrisation (for example path integrals).
Because of this independence, we often describe curves by drawing them as geometric
objects in the plane.

There are two elementary methods to modify or combine paths in order to obtain new
paths.

Definition 1.18. • If γ : [a, b] → C, t 7→ γ(t) is a path, the reverse path γ − is

the path

γ − : [a, b] → C
t 7→ γ − (t) := γ(b + a − t)

i.e. γ − (t) = γ(a + b − t)

• If γ1 : [a1 , b1 ] → C and γ2 : [a2 , b2 ] → C are two paths such that γ1 (b1 ) = γ2 (a2 ),
then the concatenation or sum of the paths γ1 , γ2 is a path

γ1 ⊎ γ2 : [a1 , b1 + b2 − a2 ] → C

γ1 (t) , t ∈ [a1 , b1 ]
t 7→ (γ1 ⊎ γ2 )(t) :=
γ2 (t − b1 + a2 ) , t ∈ [b1 , b1 + b2 − a2 ]
28 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Example 1.7. 1. Given two points z1 , z2 ∈ C, the path

γ : [0, 1] → C
t 7→ (1 − t)z1 + tz2

is the (standard) parametrisation of the line segment between z1 and z2

2. Consider the path

γ : [0, 4] → C


 t , t ∈ [0, 1]
1 + i(t − 1) , t ∈ [1, 2]

t 7→ γ(t) :=

 (3 − t) + i , t ∈ [2, 3]
i(4 − t) , t ∈ [3, 4]


as shown in the picture

1.3. COMPLEX LINE INTEGRALS (INTEGRALS ALONG CURVES) 29

It is clear that γ is the sum of four different paths, namely

γ1 : [0, 1] → C, t 7→ γ1 (t) = t
γ2 : [0, 1] → C, t 7→ γ2 (t) = 1 + it
γ3 : [0, 1] → C, t 7→ γ3 (t) = i + (1 − t)
γ4 : [0, 1] → C, t 7→ γ4 (t) = i(1 − t)

3. A circle with center at z0 and radius r has a parametrisation of the form

γ : [0, 2π] → C
t 7→ z0 + reit

4. Consider also the path

γ : [0, 1] → C
t 7→ t + it2

To define the complex line integrals we recall that a continuous function g : [a, b] → R
Rb
is Riemann integrable, i.e. a g(t)dt exists.
30 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Definition 1.19. For a complex valued function g ∈ C[a,b] we can define the integral
Z b Z b Z b
g(t)dt := u(t)dt + i v(t)dt
a a a

where ∀t ∈ [a, b] : g(t) = u(t) + iv(t)

Definition 1.20. Suppose γ : [a, b] → C is a smooth path and f ∈ CΩ is a complex

valued function, which is defined and continuous on γ. We define the integral of f
along γ by Z Z b
f γ(t) γ ′ (t)dt


f (z)dz :=
γ a

Since g(t) = f γ(t) γ ′ (t) ∈ C[a,b] is continuous on [a, b], the integral on the right is

meaningful, as long as we show that it is independent of the parametrisation of γ

Let γ̃ : [c, d] → C be another parametrisation
of im(γ), such that γ̃(s) = (γ ◦ σ)(s) for
some σ : [c, d] → [a, b] with σ ∈ C 1 [c, d], [a, b] and σ ′ (s) > 0. Then we have
Z Z d Z d 

′

f γ σ(s) γ ′ σ(s) σ ′ (s)ds


f (z)dz = f γ̃(s) γ̃ (s)ds =
γ̃ c c
′
By letting t = σ(s) and consequently dt = σ (s)ds, we obtain
Z d  Z b Z

 ′
′
′
f γ σ(s) γ σ(s) σ (s)ds = f γ(t) γ (t)dt = f (z)dz
c a γ

The following properties of path integrals follow from the properties of the Riemann
integral.

Proposition 1.10. [SS10, Proposition I.3.1] Let f, g ∈ C 0 (Ω; C), γ, γ1 , γ2 piecewise

smooth curves in Ω and a, b ∈ C. Then

(i) The path integral is linear, i.e.

Z Z Z


af (z) + bg(z) dz = a f (z)dz + b g(z)dz
γ γ γ

(ii) If γ − is the curve γ with reverse orientation, then

Z Z
f (z)dz = − f (z)dz
γ− γ
1.3. COMPLEX LINE INTEGRALS (INTEGRALS ALONG CURVES) 31

(iii) The path of integration can be split

Z Z Z
f (z)dz = f (z)dz + f (z)dz
γ1 ⊎γ2 γ1 γ2

(iv) The following estimate holds

Z
f (z)dz ≤ Lγ sup f (z)
γ z∈im(γ)

where for a partition a0 ≤ ... ≤ an of the interval [a, b] we have

n−1 Z
X ak+1
Lγ = γ ′ (t) dt
k=0 ak

Proof. (i) Follows from the linearity of the Riemann integral.

(ii) If γ : [a, b] → C, then γ − : [a, b] → C, t 7→ γ(b+a−t) with (γ − )′ (t) = −γ ′ (b+a−t).

Hence, we have
Z Z b Z a

′
f γ(u) γ ′ (u)du =


f (z)dz = − f γ(b + a − t) γ (b + a − t)dt =
γ− a b
Z b Z
f γ(u) γ ′ (u)du = −


=− f (z)dz
a γ

where we used that u = b + a − t and consequently that du = −dt

(iii) Exercise.

(iv) Consider the following steps:

Z n−1 Z
X ak+1 n−1
X Z ak+1

′
γ ′ (t) dt ≤


f (z)dz = f γ(t) γ (t)dt ≤ f γ(t)
γ k=0 ak k=0 ak
! n−1 Z

X ak+1
≤ sup f γ(t) γ ′ (t) dt
t∈[a,b] k=0 ak

Remark 1.12. If S ⊆ C is a set that can be described as the image of a pathR γ in C,

we will then often denote the integral of a function f along this path with S f dz
32 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Definition 1.21 (Primitive). Let f ∈ CΩ . A primitive of f on Ω is a function

F ∈ H (Ω) such that
∀z ∈ Ω : F ′ (z) = f (z)

The existence of a primitive gives

Theorem 1.4. [SS10, Theorem I.3.2] Let f ∈ C 0 (Ω; C) be a continuous function

on an open set Ω ⊆ C. If f has a primitive F in Ω and γ is a curve which begins at
z1 and ends at z2 , i.e. γ : [a, b] → C with im(γ) ⊆ Ω, γ(a) = z1 and γ(b) = z2 , then
Z
f (z)dz = F (z2 ) − F (z1 )
γ

An immediate Corollary is

Corollary 1.1. [SS10, Corollary I.3.3] If γ is a closed curve (i.e. γ(a) = γ(b)) in an
open set Ω, f ∈ C 0 (Ω; C) and has a primitive in Ω, then
Z
f (z)dz = 0
γ

Proof of Theorem 1.4. Let F = U (x, y) + iV (x, y) and γ : [a, b] → C. We first assume
that γ is smooth. We define a function

G : [a, b] → C



t 7→ F γ(t) = F x(t), y(t)

and write γ(t) = x(t) + iy(t)

We need to check the compatibility of the real derivative of G and the complex deriva-
tive of F .


We have that G ∈ C 0 [a, b]; C is a continuous function, hence

′


′
G (t) = U x(t), y(t) + iV x(t), y(t) =
 
= Ux x(t), y(t) x′ (t) + Uy x(t), y(t) y ′ (t) +



 
+ i Vx x(t), y(t) x′ (t) + Vy x(t), y(t) y ′ (t)



1.3. COMPLEX LINE INTEGRALS (INTEGRALS ALONG CURVES) 33

by using the Chain Rule from Vector Analysis [EW22]. Now, applying the Cauchy-
Riemann Equations for F we get

′
′ 
= Ux x(t), y(t) x (t) − Vx x(t), y(t) y (t) +
 
+ i Vx x(t), y(t) x′ (t) + Ux x(t), y(t) y ′ (t) =





= Ux x(t), y(t) + iVx x(t), y(t) x′ (t)+




 ′
+ − Vx x(t), y(t) + iUx x(t), y(t) y (t) =




= Ux x(t), y(t) + iVx x(t), y(t) x(t) + iy(t) =
= F ′ γ(t) γ ′ (t) =

= f γ(t) γ ′ (t) , since F is a primitive of f on Ω

Hence, we finally get

Z Z b Z b

′ (∗)
f (z)dz = f γ(t) γ (t)dt = G′ (t)dt = G(b) − G(a) =
γ a a



= F γ(b) − F γ(a) = F (z2 ) − F (z1 )

where in the step denoted by (∗) we used the Fundamental Theorem of Analysis on G
[EW22].
If γ is piecewise smooth, then there is a dissection of [a, b] of the form
[
[a, b] = [aℓ−1 , aℓ ]
ℓ∈{1,...,n}

with a =: a0 and b =: an in accordance with the curve; this means such that for
ℓ ∈ {1, ..., n} we can define γℓ := γ|[aℓ−1 ,aℓ ] and dissect the curve as follows
n
]
γ= γℓ
ℓ=1

Then
Z n Z
X n
X
f (z)dz = f (z)dz = F (aℓ ) − F (aℓ−1 ) = F (an ) − F (a0 ) = F (b) − F (a)
γ ℓ=1 γℓ ℓ=1

using the newly obtained result for the smooth case in the third step and acknowledging
the telescopic character of the sum in the second last one.
Another Corollary of Theorem 1.4 is the following:
34 CHAPTER 1. PRELIMINARIES TO COMPLEX ANALYSIS

Corollary 1.2. [SS10, Corollary I.3.4] Let f ∈ H (Ω) on an open and connected
subset Ω ⊆ C. If f ′ = 0, then f is constant.

Proof. We want to show that for any two points z, w ∈ Ω, it holds f (z) = f (w)

Since Ω is open and connected, there is a (polygonal) path γ : [0, 1] → Ω connecting

the two points z, w ∈ Ω, i.e. such that γ(0) = z and γ(1) = w

Since f is holomorphic on Ω, f is clearly a primitive of f ′ . We can hence use the

Theorem 1.4 on f ′ to obtain
Z
f ′ (z)dz = f γ(1) − f γ(0) = f (w) − f (z)



γ

But given that f ′ = 0, the integral on the left is equal to zero and therefore f (z) = f (w)
The arbitrariness of the choice of z, w ∈ Ω concludes the proof.

Example 1.8. An important example is the function

f : C∗ → C
1
z 7→
z
f has no primitive on C∗ . To see this let γ parametrise the circle centred at 0 and of
radius 1, namely C1 (0), i.e. γ : [0, 2π] → C, t 7→ eit , then
Z Z 2π Z 2π Z 2π
it

it 1 it
f (z)dz = f e ie dt = ie dt = i dt = 2πi ̸= 0
γ 0 0 eit 0
R
Example 1.9. What is γ z 2 dz, if γ : [0, 1] → C, t 7→ t + πit2 ?

z3
Since F (z) = 3
is a primitive of z 2 , using Theorem 1.4 we have that

(1 + πi)3
Z
z 2 dz = F γ(1) − F γ(0) =


γ 3
or Z Z 1
2
z dz = (t + πit2 )2 (1 + 2πit)dt = ...
γ 0

which is much longer.

Chapter 2

Cauchy’s Theorem and its

applications

Cauchy’s Theorem is at the heart of Complex Analysis: it “roughly” says that if f is

holomorphic in an open set Ω and im(γ) ⊆ Ω is a closed (not necessarily simple) curve,
whose “interior” is contained in Ω, then
Z
f (z)dz = 0
γ

Cauchy’s Theorem, as we will see, has many applications, e.g. Liouville’s Theorem,
which in return gives a proof of Fundamental Theorem of Algebra.

The interior of a path is not easy to define for a general curve. We will work around
this difficulty by first proving Cauchy’s Theorem for curves, whose interior is easy to
define, namely for triangles and rectangles (Goursat’s Theorem).

We then use Goursat’s Theorem to show that a holomorphic function on an open disc
has a primitive in that disc. This then will give us a Corollary: Cauchy’s Theorem on
a disc.

We first need the following Proposition about nested compact sets, but before that we
define a useful notational tool to approach this type of scenarios, namely the diameter
of a set.

Definition 2.1. Let S ⊆ C, then we define the diameter of S ⊆ C as

diam(S) := sup |a − b|
a,b∈S

35
36 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Proposition 2.1. [SS10, Proposition I.1.4] If

∆(1) ⊇ ∆(2) ⊇ ... ⊇ ∆(n) ⊇ ...

is a sequence of non-empty compact sets in C with the property that

 
n→∞
diam ∆(n) −−−→ 0

Then
∃!z0 ∈ C∀n ∈ N∗ : z0 ∈ ∆(n)

Proof. Choose zn any point in ∆(n) . For any n, m ∈ N∗ such that n ≥ m ≥ 1 we have
 
|zn − zm | ≤ diam ∆(m) < +∞
 
n→∞
Since diam ∆(n) −−−→ 0, this says that (zn )n∈N∗ ∈ n∈N∗ ∆(n) is a Cauchy sequence,
Q

hence it converges to a limit z0 ∈ C. Then for m ≥ 1, note that ∀n ≥ m : zn ∈ ∆(m)

since zn ∈ ∆(n) ⊆ ∆(m) . Moreover, we have that ∆(m) is compact, in particular closed
and hence limn→∞ zn = z0 is also in ∆(m) for all m ≥ 1 and z0 is unique; since if
z0 , z0′ ∈ ∆(n) for n ∈ N∗ and z0 ̸= z0′ , then |z0 − z0′ | > 0, which contradicts the
 all 
n→∞
shrinking diam ∆(n) −−−→ 0

2.1 Goursat’s Theorem

Theorem 2.1 (Goursat’s Theorem). [SS10, Theorem II.1.1] Let Ω ⊆ C be open
and T a path with shape of a triangle such that im(T ) ⊆ Ω, whose interior is also
contained in Ω. Let f ∈ H (Ω), then
Z
f (z)dz = 0
T

Proof. Note that a triangle is a closed curve, which is the union of three line segments.
If T has three corners at z1 , z2 and z3 , then we will write
3
]
T = αl =: ⟨z1 , z2 , z3 ⟩
l=1
U
(Note that the sign does NOT represent a “flipped gravestone” ∼ Greta from Lo-
carno1 )
1
Locarno is a southern Swiss town and municipality in the district Locarno (of which it is the
2.1. GOURSAT’S THEOREM 37

with
α1 : [0, 1] → C, t 7→ α1 (t) = z1 + (t − 0)(z2 − z1 )
α2 : [0, 1] → C, t 7→ α2 (t) = z2 + (t − 1)(z3 − z2 )
α3 : [0, 1] → C, t 7→ α3 (t) = z3 + (t − 2)(z1 − z3 )
We now define the simplex
∆ := {z ∈ C : z = t1 z1 + t2 z2 + t3 z3 and 0 ≤ t1 , t2 , t3 and t1 + t2 + t3 = 1}
such that it is the smallest convex set containing z1 , z2 , z3 . We hence have that im(T ) ⊆
∆, in fact im(T ) = ∂∆

We will inductively construct a sequence of triangular paths, so we define for all n ∈ N

(n) (n) (n)
T (n) = ⟨z1 , z2 , z3 ⟩
as follows:
(0) (0) (0) U3 (0)
1. Let T (0) = ⟨z1 , z2 , z3 ⟩ = l=1 αl
(n) (n) (n)
2. Assume that T (n) is defined as T (n) = ⟨z1 , z2 , z3 ⟩, then T (n+1) is one of the
the four triangular paths:
* +
(n) (n) (n) (n)
(n+1) z1 + z2 (n) z2 + z3
T1 := , z2 ,
2 2
* +
(n) (n) (n) (n)
(n+1) z2 + z3 (n) z + z1
T2 := , z3 , 3
2 2
* +
(n) (n) (n) (n)
(n+1) z3 + z1 (n) z + z2
T3 := , z1 , 1
2 2
* +
(n) (n) (n) (n) (n) (n)
(n+1) z1 + z2 z + z z + z
T4 := , 2 3
, 3 1
2 2 2

capital), located on the northern shore of Lake Maggiore at its northeastern tip in the canton of Ticino
at the southern foot of the Swiss Alps. It has a population of about 16,000 (proper), and about 56,000
for the agglomeration of the same name including Ascona besides other municipalities.
38 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

i.e. we are at each step partitioning ∆ using lines parallel to the sides and passing
through their midpoints.
(n)
The triangular paths Tk with k ∈ {1, ..., 4} are all entirely contained in ∆ and we
have that
Z 4 Z
X
f (z)dz = f (z)dz
(n+1)
T (n) k=1 Tk

therefore we get
Z Z
f (z)dz ≤ 4 max f (z)dz
k∈{1,...,4} (n+1)
T (n) Tk

(n+1)
We choose T (n+1) as one of Tk with k ∈ {1, ..., 4} so that
Z Z
f (z)dz ≤ 4 f (z)dz
T (n) T (n+1)

and it then follows that

Z Z
n
f (z)dz ≤ 4 f (z)dz
T T (n)

We also have that the closed filled triangles ∆(n) , defined with respect to their T (n) in
a similar fashion as ∆ with T , are nested compact sets

C ⊇ ∆ = ∆(0) ⊇ ∆(1) ⊇ ∆(2) ⊇ ...

Moreover, if dn and Pn are the diameter and the perimeter of ∆(n) respectively, then
for all n ∈ N we have
d0 d P0 P
dn = n
=: n and Pn = n
=: n
2 2 2 2
Hence  
n→∞
diam ∆(n) −−−→ 0
We can now apply the result anticipated by Proposition 2.1. In order to do so, though,
we remember that f is holomorphic at z0 ∈ ∆(n) ⊆ Ω, hence it holds that

f (z) = f (z0 ) + f ′ (z0 )(z − z0 ) + E(z)

2.1. GOURSAT’S THEOREM 39

|E(z)|
with limz→z0 |z−z0 |
= 0. It is clear that E(z) is continuous at z0 and that limz→z0 E(z) =
0
R R
Let ε > 0, we will show that I := T
f (z)dz ≤ εP d, which will show that T
f (z)dz = 0

To this end, we use the above equation for

Z Z Z
′


f (z)dz = f (z0 ) + f (z0 )(z − z0 ) dz + E(z)dz
T (n) T (n) T (n)

Using Corollary 1.1, the first integral on the right is zero, since g(z) = f (z0 )+f ′ (z0 )(z −
2
z0 ) has a primitive (take G(z) = f (z0 )z + f ′ (z0 ) (z−z2 0 ) as such) and since the curve is
closed.

Hence, we can reduce to the whole problem to the estimate

Z Z Z
n n
f (z)dz ≤ 4 f (z)dz ≤ 4 E(z)dz
T T (n) T (n)

By letting Z
In := E(z)dz
T (n)

because of the initial assumption on the asymptotic behaviour of E when approaching

z0 , for the given ε > 0 we choose an open disc Dδ (z0 ) ⊆ Ω such that

∀z ∈ Dδ (z0 ) : |E(z)| ≤ ε|z − z0 |

n→∞
Because dn −−−→ 0, there exists an index N ∈ N such that ∀n ≥ N : dn < δ. We also
have that z0 ∈ ∆(n) for all n ∈ N and that ∀n ≥ N : |z − z0 | ≤ dn < δ

Hence, we conclude that ∀n ≥ N : ∆(n) ⊆ Dδ (z0 ) and from it we get

Z Z
n n n
|I| ≤ 4 |In | = 4 E(z)dz ≤ 4 |E(z)||dz| ≤
T (n) T (n)
Z
n P d
≤4 ε |z − z0 ||dz| ≤ 4n εPn dn = 4n ε n n = εP d
T (n) 2 2
Since ε > 0 was arbitrary, we obtain that I = 0 and conclude the proof.
As a Corollary we get

Corollary 2.1. [SS10, Corollary II.1.2] If f ∈ H (Ω) is holomorphic in an open

set Ω ⊆ C that contains a path R with the shape of a rectangle and such that
40 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

im(R) ⊆ Ω as its interior, then

Z
f (z)dz = 0
R

Proof. This follows immediately from the Goursat’s Theorem 2.1 and by dividing the
rectangle R into two triangles T1 , T2 as shown in the picture.

Therefore, we obtain
Z Z Z
f (z)dz = f (z)dz + f (z)dz = 0
R T1 T2

For future results, for example for the deduction of Cauchy’s Integral Formula, a minor
extension of this result is useful.

Theorem 2.2 (Goursat’s Theorem stronger version). If a function f ∈ C 0 (Ω) is

continuous in an open set Ω and f |Ω\{z0 } ∈ H (Ω \ {z0 }) for some z0 ∈ Ω, then
Z
f (z)dz = 0
R

for every closed rectangle R ⊆ Ω, i.e. such that ∂R = R

Proof. Fix a closed triangle, in the topological sense, R ⊆ Ω. We assume that z0 ∈ R,

otherwise the conclusion follows from the first version above as the integral is 0

Given a positive integer n ∈ N we subdivide R into n2 congruent rectangles, such that

∂R = R
2.2. LOCAL EXISTENCE OF PRIMITIVES AND CAUCHY’S THEOREM IN A DISC41

Once again it follows that

Z n X
X n Z
f (z)dz = f (z)dz
R l=1 j=1 Rlj

with ∂Rlj = Rlj . If z0 ∈

/ Rlj , then
Z
f (z)dz = 0
Rlj

by the first version of Corollary 2.1. If instead z0 ∈ Rlj , then

Z
L
f (z)dz ≤ M PRlj = M
Rlj n

in which L is the length of the perimeter of R, PRlj = Ln is the length of the perimeter
of Rlj and M = maxz∈R |f (z)| = ∥f ∥∞,R is the maximum of the continuous function
|f | on the compact set R

By construction, the point z0 cannot belong to more than four subrectangles: the point
in which touches most subrectangles is a common vertices of all the four. Hence we can
finally consider the following estimate
Z X Z X Z L n→∞
f (z)dz = f (z)dz ≤ f (z)dz ≤ 4M −−−→ 0
R z0 ∈Rlj ∂Rlj z0 ∈Rlj Rlj n

2.2 Local existence of primitives and Cauchy’s the-

orem in a disc
To prove the Cauchy’s Theorem in a disc, we will need the local existence of primitives.
We therefore have the following Theorem:
42 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Theorem 2.3. [SS10, Theorem II.2.1] A holomorphic function in an open disc D

has a primitive in that disc.

We are going to prove the following version which assumes that f is continuous in D
and that its integral along rectangles whose sides parallel to the coordinate axes vanish.
Which then we are going to use to give a slightly stronger form of Cauchy’s Theorem.

Theorem 2.4. Let D be an open disc in C and f ∈ C 0 (D) with the property that
Z
f (z)dz = 0
R

for every closed rectangle R ⊆ D with ∂R = R in D and whose sides are parallel
to the coordinate axes. Then f has a primitive in D

Before we prove Theorem 2.4, note that we have as a Corollary.

Theorem 2.5 (Cauchy’s Theorem for a disc). [SS10, Theorem II.2.2] Suppose D ⊆
C is an open disc in C and f ∈ H (D), or more generally f ∈ C 0 (D) with f |D\{z0 } ∈
H (D \ {z0 }) for some z0 ∈ D. Then
Z
f (z)dz = 0
γ

for every closed piecewise smooth path γ in D

Proof. Suppose f ∈ C 0 (D) with f |D\{z0 } ∈ H (D \ {z0 }) for some z0 ∈ D, then by

Goursat’s Theorem 2.2 Z
f (z)dz = 0
R

for every closed rectangle R ⊆ D with ∂R = R (including the ones, whose sides are
parallel to axes). By Theorem
R 2.4 f has a primitive in D and by Theorem 1.4 and
Corollary 1.1 we have that γ f (z)dz = 0 for every piecewise smooth path γ in D

Proof of Theorem 2.4. Let f ∈ C 0 (D) be continuous on the disc D and let z0 = x0 +iy0
be the center of D
For an arbitrary point z = x+iy ∈ D such that z ̸= z0 , let z1 := x+iy0 and z2 := x0 +iy
2.2. LOCAL EXISTENCE OF PRIMITIVES AND CAUCHY’S THEOREM IN A DISC43

By assumption
Z Z Z Z
f (w)dw + f (w)dw + f (w)dw + f (w)dw = 0
[z0 ,z1 ] [z1 ,z] [z,z2 ] [z2 ,z0 ]

where we denote the path along the line segment linking two points p1 , p2 ∈ D, from
p1 to p2 , with [p1 , p2 ]
R R
This sum represent either R
f (w)dw or − R
f (w)dw, depending on the location of z

We define F ∈ CD as follows, for z ∈ D let

Z Z
F (z) := f (w)dw + f (w)dw
[z0 ,z2 ] [z2 ,z]

which is, by the above assumption

Z Z
= f (w)dw + f (w)dw
[z0 ,z1 ] [z1 ,z]

Parametrizing the line segments we have

Z y Z x
F (z) = i f (x0 + it)dt + f (t + iy)dt
y0 x0

and Z x Z y
F (z) = f (t + iy0 )dt + i f (x + it)dt
x0 y0

Using both the above results and the Fundamental Theorem of Analysis [EW22]:
Z x
d
g(t)dt = g(x)
dx a


if g ∈ C 0 Dr (a)∩R; C with g(t) = f (t+iy), then we have that Fx (z) = f (x+iy) = f (z).
Similarly, since Z y
d
h(t)dt = h(y)
dy a
44 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS


if h ∈ C 0 Dr (a) ∩ R; C with h(t) = f (x + it), we have Fy (z) = if (x + iy) = if (z)
(for both parts we used that the first integral in the equations are independent of x, y
respectively). Hence, it follows that Fx , Fy exist and are continuous, so F ∈ C 1 (D)
At this point, since Fx (z) = f (z) and Fy (z) = if (z), if we write F (z) = u + iv, then
this gives
f (z) = Fx (z) = ux + ivx = −iFy (z) = −i(uy + ivy ) = vy − iuy
Hence ux = vy and vx = −uy and so the the Cauchy-Riemann equations hold.

Finally, we know that F ∈ C 1 (D) and that F satisfies that Cauchy Riemann Equations,
therefore by Theorem 1.1 F ∈ H (Ω) and
∂F
F ′ (z) = (z) = f (z)
∂x
hence F is a primitive of f

Corollary 2.2. [SS10, Corollary II.2.3] Let f ∈ H (Ω) for Ω ⊆ C an open set
containing a circle C and its interior, then
Z
f (z)dz = 0
C

R
Sketch of a proof. Simply consider F (z) := γz
f (w)dw as in the following illustration.

Remark 2.1. Corollary 2.2 is in fact valid whenever we can define the interior of a
contour unambiguously and construct polygonal paths in an open neighbourhood of both
the contour and its interior. In [SS10] these contours are called toy contours.
Remark 2.2. Note that Cauchy’s Theorem 2.5 does not say anything about integrals
of functions over arbitrary open sets and arbitrary closed curves. Indeed recall Example
1.8.
2.2. LOCAL EXISTENCE OF PRIMITIVES AND CAUCHY’S THEOREM IN A DISC45

Some applications of Cauchy’s Theorem

We can use Cauchy’s Theorem for a disc to calculate some integrals.

Example 2.1. We can show by parametrizing the circle that for all r > 0 we have:
Z
1
dz = 2πi
Cr (z0 ) z − z0

Indeed, the circle of center z0 and radius r has parametrization σ(t) = z0 + reit for
t ∈ [0, 2π] and hence
Z Z 2π
1 1
dz = it
ireit dt = 2πi
|w−z0 |=r z − z0 0 re

Now, using Cauchy’s Theorem 2.5 we can also show that

Z
1
dz = 2πi
R z − z0

for any rectangle R with center at z0

R 1 1
Note that R z−z 0
dz is not zero, since z−z 0
is not continuous at z0 and hence Cauchy’s
Theorem 2.5 does not apply directly. We can though use it as follows:

Let Cr (z0 ) be the circle that circumscribes the rectangle R and

4
! 4
!
] ]
R = im γk and Cr (z0 ) = im βk
k=1 k=1

For each k ∈ {1, ..., 4} we choose an open disc Dk , so that the trajectory of the closed
path γk ⊎ βk− is in Dk and so that f (z) = z−z
1
0
is holomorphic in Dk
46 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Now we apply Cauchy’s Theorem 2.5 to 1

z−z0
in disc Dk to γk ⊎ βk− in Dk to get
Z
f (z)dz = 0
γk ⊎βk−

From this follows that Z Z

1 1
dz = dz
γk z − z0 βk z − z0
But then, considering then all paths at the same time, we get
Z Z
1 1
dz = U dz
U4
k=1 γk
z − z0 4
k=1 βk
z − z0
and hence Z Z
1 1
dz = dz = 2πi
R z − z0 Cr (z0 ) z − z0
Example 2.2 (Fresnel integrals). Fresnel proved the following identity:
Z ∞ Z ∞ √
2π
cos(x2 )dx = sin(x2 )dx =
0 0 4
2
To show this we first note that the map eix has real and imaginary parts cos(x2 ) and
sin(x2 ). If we can prove that
Z ∞ √
ix2 2π
e dx = (1 + i)
0 4
2
then we are done. We are naturally led to define f (z) = eiz , which is holomorphic in
all of C as follows.

where
γ1 : [0, R] → C, t 7→ γ1 (t) = t
h πi
γ2 : 0, → C, t 7→ γ2 (t) = Reit
4
π
γ3 : [0, R] → C, t 7→ γ3 (t) = tei 4
2.2. LOCAL EXISTENCE OF PRIMITIVES AND CAUCHY’S THEOREM IN A DISC47

We have that for γ := γ1 ⊎ γ2 ⊎ γ3−

Z
2
eiz dz = 0
γ

This separates into all its components as follows:

Z R Z Z Z Z R
iz 2 iz 2 iz 2 i π4 2
it
e dt = − e dz + e dz = − e dz + e e−t dt
0 γ2 γ3 γ2 0

Claim. 2
π 1 − e−R
Z

iz 2 R→+∞
e dz ≤ −−−−→ 0
γ2 4R
Proof of the claim. Expanding the integral we have
Z Z Z π Z π
4 4
iz 2 iz 2 i(Reit )2 it 2 e2it )
e dz ≤ e |dz| = e iRe dt = R ei(R eit dt =
γ2 γ2 0 0
Z π
4 2 (cos(2t)+i sin(2t))
=R ei(R eit dt =
0
Z π
4 2 2
=R eiR cos(2t) e−R sin(2t)
eit dt =
0 | {z } |{z}
=1 =1
Z π Z π
4 R R→∞ 4
= dt −−−→ 0dt = 0
0 eR2 sin(2t) 0

Using that the last integrand in bounded for R ∈ [0, +∞) and therefore applying the
Dominant Convergence Theorem [Da 24] to swap the limit and the integral.
Therefore we have
Z R
√ Z ∞ √
it2 2 −t2 2π
lim e dt = (1 + i) e dt = (1 + i)
R→∞ 0 2 0 2 2
R∞ √
2 π
since from [EW22] we have that 0
e−t dt = 2

Remark 2.3. Note that the use of compact paths to solve this kind of integrals in a
very useful tool. One first solves the compact case and then by letting some parameter
approach its limit case of interest, if the integrand is nice enough, the complexity of the
problem can be considerably reduced.
48 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

2.3 Cauchy’s Integral Formulas

We are going to prove Cauchy’s Integral Formula (CIF), from which we will deduce
many properties of holomorphic functions.

Theorem 2.6 (Cauchy’s Integral Formula). [SS10, Theorem II.4.1] Suppose f ∈

H (Ω) is holomorphic in an open set Ω that strictly contains the closure of a disc
D, i.e. D ⊂ Ω. If C = ∂D denotes the boundary circle of this disc D with positive
orientation, i.e. counterclockwise, then
Z
1 f (w)
∀z ∈ D : f (z) = dw
2πi C w − z

Remark 2.4. Note that Cauchy’s Integral Formula says that the values of f on D are
determined by their boundary values on the circle C
Proof. Let z0 ∈ Ω, since Ω is open, there exist r > 0 such that Dr (z0 ) ⊂ Ω, still since
Ω is open, ∀z ∈ ∂Dr (z0 ) : z ∈ Ω and so ∃ε > 0 : Dr+ε (z0 ) ⊂ Ω. We obtain from this
that Cr (z0 ) ⊂ Dr+ε (z0 ). We let z ∈ Dr (z0 ) and define
 f (w)−f (z)
, w ̸= z
g : Dr+ε (z0 ) → C, w 7→ g(w) = w−z
′
f (z) ,w=z


Then it holds that g ∈ C 0 Dr+ε (z0 ) and away from z we also have that g is holomor-
phic, i.e. g|Dr+ε (z0 )\{z} ∈ H Dr+ε (z0 ) \ {z}

By Cauchy’s Theorem 2.5 applied to g, we have

Z
g(w)dw = 0
Cr (z0 )

i.e.
f (w) − f (z)
Z
dw = 0
Cr (z0 ) w−z
Note that on Cr (z0 ) it holds that w ̸= z, since z ∈ Dr (z0 ). Hence
Z Z
f (w) 1
dw = f (z) dw
Cr (z0 ) w − z Cr (z0 ) w − z

To finish the proof we only lack the following claim:

Claim. We state that the integral has the following form
Z
1
dw = 2πi
Cr (z0 ) w − z
2.3. CAUCHY’S INTEGRAL FORMULAS 49

Proof of the Claim. Consider the following parametrization of Cr (z0 )

γ : [0, 2π] → C
t 7→ γ(t) = z0 + reit

Though, Cr (z0 ) also has the following parametrization

γ̃ : [0, 2π] → C
s 7→ γ̃(s) = z + ρ(s)eit


where ρ : [0, 2π] → R, s 7→ ρ(s) := γ t(s) − z = |γ̃(s) − z| with σ : [0, 2π] →
[0, 2π], s 7→ t(s) as reparametrization of the control variable. Clearly ρ is smooth,
moreover
γ̃ ′ (s) = ρ′ (s)eis + iρ(s)eis
We proceed now to develop the integral form above:
Z 2π ′
ρ (s)eis + iρ(s)eis
Z
1
dw = ds =
Cr (z0 ) w − z 0 ρ(s)eis
Z 2π ′ Z 2π
ρ (s)
= ds +i ds =
0 ρ(s) 0
| {z }
real integral

s=2π
= ln ρ(s) s=0 +2πi =
| {z }
=0
= 2πi

since ρ(0) = ρ(2π)

We conclude illustrating some more the new parametrization we introduced.

Here t changes with s and ρ = ρ(s) = γ t(s) − z . As mentioned before

γ̃(s) = z + ρ(s)eit
50 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

is the new parametrisation, with

σ : [0, 2π] → [0, 2π]
s 7→ t(s)
as the change of variable. Hence we have γ̃ = γ ◦ σ
Before we give important theoretical applications of Cauchy’s Theorem 2.5 and the
Cauchy’s Integral Formula 2.6, we’ll look at one more example of contour shifting,
which helps us to evaluate certain integrals.

R ∞ f ∈ C , which is Riemann inte-

Definition 2.2 (Fourier Transform). For a function R

grable on every [a, b] and which has converging −∞ f (t) dt, its Fourier Transform
Z ∞
fˆ(ξ) := f (x)e−2πiξx dx
−∞

is well defined for all ξ ∈ R

2
Example 2.3. We’ll show that e−πx is its own Fourier Transform. We want to show
that if f (x) = e−πx , then fˆ(ξ) = e−πξ
2 2

We want to show that Z ∞

−πξ 2 2
e = e−πx e−2πiξx dx
−∞
or equivalently Z ∞
2 2
1= e−πx e−2πiξx eπξ dx
−∞
If ξ = 0, this gives Z ∞
2
e−πx dx = 1
−∞
which we know from Real Analysis (see [EW22]).
2
We first suppose that ξ > 0 and let f (z) = e−πz , then f (z) is entire and in particular
holomorphic in the piecewise smooth contour γR = γ1 ⊎ γ2 ⊎ γ3 ⊎ γ4 as in the picture.
2.3. CAUCHY’S INTEGRAL FORMULAS 51

Where γ1 is the path on the real axis, γ2 is the right vertical path, γ3 is the upper
horizontal one and γ4 is the left vertical one. Hence, using Cauchy’s Theorem 2.5
Z
f (z)dz = 0
γR

Note that on γ1 :
Z Z R
2
f (z)dz = e−πx dx
γ1 −R

while on γ3 :
Z Z −R Z R Z R
−π(x+iξ)2 −π(x2 +2πixξ) πξ 2 πξ 2 2
f (z)dz = e dx = − e e dx = −e e−πx e−2πixξ dx
γ3 R −R −R

As R → ∞, the first integral over γ1 equals 1, while the integral over γ3 gives
Z R
πξ 2 2
−e e−πx e−2πixξ dx
−R

On the other vertical side on the right we have

Z Z ξ Z ξ
2 +2iRy−y 2 )
f (z)dz = f (R + iξ)idy = e−π(R idy
γ2 0 0

For a fixed ξ, the integral can be bounded using Proposition 1.10 with
Z
2 2 2
f (z)dz ≤ ξ sup e−πR e−πiRy eπy ≤ Ce−πR
γ2 y∈[0,ξ]

where C is a constant dependent on ξ. A similar bound holds for γ4 . Hence, as R → ∞,

both integrals vanish to 0 and we obtain that
Z 4 Z Z ∞
πξ 2 2
X
0 = lim f (z)dz = lim f (z)dz = 1 + 0 − e e−πx e−2πixξ dx + 0
R→∞ γR R→∞ γk −∞
k=1

We therefore obtain that Z ∞

−πξ 2 2
e = e−πx e−2πixξ dx
−∞

Next we are going to see that Cauchy’s theorem 2.5 and the Cauchy’s Integral Formula
2.6 will imply fundamental properties of holomorphic functions.

Namely, we are going to see that they are enough to prove:

52 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

1. If Ω ⊆ C is open and f ∈ H (Ω), then f ′ ∈ H (Ω). Hence, f is infinitely often

differentiable.
Moreover, if z0 ∈ Ω and r > 0 such that Dr (z0 ) ⊂ Ω, then f has a power series
expansion at z0 , namely
∞
X
∀z ∈ Dr (z0 ) : f (z) = an (z − z0 )n
n=0


i.e. f ∈ C ω Dr (z0 ) , namely f is analytic in Dr (z0 )

2. If f is entire, i.e. f ∈ H (C), then f is constant.

3. The Fundamental Theorem of Algebra holds, i.e. any polynomial p(z) ∈ C[z] of
degree n ∈ N, has n roots in C (counted with multiplicity).

4. If f, g ∈ H (Ω) and f (z) = g(z) for all z in some sequence of distinct points with
a limit point in Ω, then ∀z ∈ Ω : f (z) = g(z). In particular, if f, g agree on an
open set U of Ω, then they agree on all of Ω

We start with the following Theorem:

Theorem 2.7. [SS10, Theorem II.4.4] Suppose that Ω ⊆ C is an open set and
f ∈ H (Ω). Let z0 ∈ Ω and r > 0 be such that Dr (z0 ) ⊂ Ω. Then f has a power
series extension at z0
∞
X
∀z ∈ Dr (z0 ) : f (z) = an (z − z0 )n
n=0

and the coefficients an are give by the formula

f [n] (z0 )
Z
1 f (w)
∀n ∈ N : an = = dw
n! 2πi Cr (z0 ) (w − z0 )n+1

Moreover, the convergence of the series is absolute and uniform on Dr (z0 )

Proof. Let z0 and r be defined as above, so that Dr (z0 ) ⊂ Ω. Fix s ∈ (0, r) and let
Cs (z0 ) be the circle of radius s with center z0 . By setting γ a path in Dr (z0 ) such that
im(γ) = Cs (z0 ) ⊂ Ω and using Theorem 2.6, we obtain
Z
1 f (w)
f (z) = dw
2πi γ w − z

for all z ∈ Ds (z0 )

2.3. CAUCHY’S INTEGRAL FORMULAS 53

The trick is to write

!
1 1 1 1
= = z−z0
w−z (w − z0 ) − (z − z0 ) w − z0 1− w−z0

Since we are integrating on γ, for w ∈ im(γ) (and for z ∈ Ds (z0 )) we have for the term
in the last fraction that
z − z0 |z − z0 |
= <1
w − z0 s
This means that we can rewrite the whole last fraction as a geometric series (see [EW22])
∞  n
1 X z − z0
z−z0 =
1 − w−z 0 n=0
w − z0

for w ∈ im(γ) and z ∈ Ds (z0 ). The convergence of the series is uniform, since the
bound |z−z
s
0| z−z0
for w−z 0
is independent of w ∈ im(γ). Hence we can interchange the series
and the integral (again, see [EW22]), obtaining
∞
!
(z − z0 )n
Z
1 X
f (z) = f (w) n+1
dw =
2πi γ n=0
(w − z0 )
∞ Z
n 1 f (w)
X
= (z − z0 ) dw =
n=0
2πi γ (w − z0 )n+1
∞
X
= an (z − z0 )n
n=0

where we defined the an ’s as

Z
1 f (w)
an := dw
2πi γ (w − z0 )n+1

Hence, we have that in Ds (z0 ) the function f is the sum of the power series
∞
X
an (z − z0 )n
n=0
54 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

We have seen that series in their disc of convergence are differentiable with derivatives
given by termwise differentiation, as stated in Theorem 1.3. Hence, for all z ∈ Ds (z0 )
we have
X∞
′
f (z) = nan (z − z0 )n−1 =
n=1
∞
X
= (k + 1)ak+1 (z − z0 )k
k=0

Being a power series, f ′ (z) is also holomorphic in Ds (z0 ), i.e. f ′ |Ds (z0 ) ∈ H Ds (z0 ) .

Inductively, we get that f is differentiable infinitely many times for z ∈ Ds (z0 ) and
evaluating it and its derivatives at z = z0 gives
a0 = f (z0 )
a1 = f ′ (z0 )
..
.
n!an = f [n] (z0 )
[n]
Hence an = f n!(z0 ) is independent of s and we have ∞ n
P
n=0 an (z − z0 ) converging for all
z ∈ Ds (z0 ), this for any arbitrary s ∈ (0, r). By taking the limit for s → r we notice
that γ expands to Cr (z0 ), which is still in Ω: the above calculations and step are still
valid, especially the uniform convergence of the series, also in this case is granted, as
|z − z0 | < r. Viewing this as (constant for each z ∈ Ds (z0 )) continuous function of s,
we can extend by continuity to r. This concludes the proof.

Remark 2.5. In all these Theorems with Ω ⊆ C open by assumption, it always follows
by definition of open set in C that
∃r > 0 : Dr (z0 ) ⊂ Ω
Moreover, the above proof of Theorem 2.7 gives a method for determine the radius of
convergence of the series expansion of any f ∈ H (Ω), as done above by expanding s
to r: we start by a fully contained Ds (z0 ) ⊂ Ω for some s ∈ (0, +∞) and progressively
increase it until we meet ∂Ω, thus external boundary of the set or some singularity
points. As long as Ds (z0 ) ⊂ Ω, the success of the operation is granted.

Remark 2.6. Note that the proof also gives that if f is complex differentiable at z0 ,
then in facts it is complex differentiable infinitely many times at z0 . It follows that for
every n ∈ N
f [n] (z0 )
Z
1 f (w)
an = dw =
2πi Cs (z0 ) (w − z0 )n+1 n!
In fact we have
2.3. CAUCHY’S INTEGRAL FORMULAS 55

Corollary 2.3 (Cauchy’s Integral Formula for derivatives). [SS10, Corollary II.4.2]
If f ∈ H (Ω), then f is infinitely often complex differentiable in Ω and in particular

∀n ∈ N : f [n] ∈ H (Ω)

Moreover, if z0 ∈ Ω and r > 0 are such that Dr (z0 ) ⊂ Ω, then

Z
[n] n! f (w)
∀n ∈ N∀z ∈ Dr (z0 ) : f (z) = dw
2πi Cr (z0 ) (w − z)n+1

Proof. The fact that f ′ is complex differentiable (or holomorphic) follows from the fact
that, since Ω is open, ∀z0 ∈ Ω∃r > 0 : Dr (z0 ) ⊂ Ω. By Theorem 2.7, f has a power
series expansion there, which is therefore holomorphic in Dr (z0 ). Since power series are
infinitely often complex differentiable in their disc of convergence, we have that f is
complex differentiable infinitely often at z0 . Since z0 was arbitrary, f in infinitely often
complex differentiable in Ω

To prove this, by induction on n ∈ N, note that the base case with n = 0 is simply the
Cauchy’s Integral Formula in Theorem 2.6. We are now going to show the induction
step, therefore suppose that
(n − 1)!
Z
[n−1] f (w)
f (z) = n
dw
2πi Cr (z0 ) (w − z0 )

for any z ∈ Dr (z0 ). For h ∈ C small enough, so that z + h and z are both away from
Cr (z0 ), we have
!
[n−1] [n−1]
(z + h) − f (n − 1)!
Z
f (z) f (w) 1 1
=
n − dw
h 2πi Cr (z0 ) h w − (z + h) (w − z)n
We then use the equality (in C)
n−1
!
X
an − bn = (a − b) an−1−k bk
k=0

1 1 h→0
with a = w−(z+h) and b = w−z and take the limit as h → 0: note that a−b h
1
−−→ (w−z)2
Pn−1 n−1−k k h→0 n
and k=0 a b −−→ (w−z)n−1 , to get that

f [n−1] (z + h) − f [n−1] (z) (n − 1)!

Z
1 n
lim = f (w) dw =
h→0 h 2πi Cr (z0 ) (w − z) (w − z)n−1
2
Z
n! f (w)
= dw
2πi Cr (z0 ) (w − z0 )n+1
56 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Remark 2.7. Theorem 2.7 says that a holomorphic function f can be locally developed
as a power series around each point of the P definition domain Ω. Explicitly, for each
z0 ∈ Ω, ∃Dr (z0 ) ⊂ Ω and a power series ∞ n
n=0 an (z − z0 ) , which converges for all
z ∈ Dr (z0 ) and represents the function f in Dr (z0 )

Due to this power series expansion we have that holomorphic functions are exactly the
functions which are everywhere representable as a power series (with a positive radius
of convergence). Recall that any power series represents a holomorphic function in their
disc of convergence. This is why we have the words ”holomorphic” and ”analytic” used
interchangeably in various sources.

Note that the power series ∞ n

P
n=0 an (z − z0 ) might not represent f (z) in all of Ω, but
it represents it at least in a disc whose radius is the distance from the point to the
boundary of Ω

Corollary 2.4 (Cauchy Inequality). [SS10, Corollary II.4.3] With the assumptions
as in Theorem 2.3, we have for every n ∈ N that

n!∥f ∥∞,Cr (z0 )

f [n] (z0 ) ≤
rn
where ∥f ∥∞,Cr (z0 ) := supz∈Cr (z0 ) f (z)

Proof. By Corollary 2.3 we have that

Z
[n] n! f (w)
f (z0 ) = dw =
2πi Cr (z0 ) (w − z0 )n+1
Z 2π
n! f (z0 + reiθ ) iθ
= ire dθ ≤
2π 0 (reiθ )n+1
n! 2π f (z0 + reiθ )
Z
≤ dθ ≤
2π 0 rn
n! ∥f ∥∞,Cr (z0 )
≤
2π rn
2.3. CAUCHY’S INTEGRAL FORMULAS 57

An immediate Corollary of these results is the remarkable Liouville’s Theorem.

Theorem 2.8 (Liouville’s Theorem). [SS10, Theorem II.4.5] If f ∈ H (C) and if f

is bounded, i.e. f ∈ B(C), then f is constant.

Proof. Since C is connected, it is enough to show that f ′ = 0 (Corollary 1.2).

Let z0 ∈ C, then for all r > 0 we have Dr (z0 ) ⊂ C and since f is holomorphic on all of
C, we have by using Cauchy’s Inequality 2.4.
∥f ∥∞,Cr (z0 )
f ′ (z0 ) ≤
r
By assumption f is bounded, i.e. ∃M ≥ 0∀z ∈ C : |f (z)| < M , hence
M
∀r > 0 : f ′ (z0 ) <
r
By letting r → ∞ we get f ′ (z0 ) = 0 and since z0 was arbitrary, we get that ∀z ∈ C :
f ′ (z) = 0 and hence f is constant.
Remark 2.8. The assumption that in Liouville’s Theorem 2.8 f has to be holomorphic
1
in all of C is essential, e.g. let Ω = {z ∈ C : Re(z) > 0} and f (z) = 1+z on it.

Corollary
Pn 2.5 (Fundamental Theorem of Algebra). Every polynomial p(z) = a0 +
k
k=1 ak z with deg(p) = n ≥ 1 has precisely n roots in C, counted with multiplicity.
If these roots are w1 , ..., wn (with possible repetitions), then
n
Y
p(z) = an (z − wk )
k=1

Proof. We first show that p(z) has a root in C for deg(p) ≥ 1

By contradiction, suppose that there is no such root and that consequently the function
1
Q(z) = p(z) ∈ H (C)

If it held Q ∈ B(C), then it would be a constant by Liouville’s Theorem 2.8, which

Pn wouldk
then contradict that p(z) is not constant, i.e. deg(p) ≥ 1 with p(z) = a0 + k=1 ak z
where an ̸= 0 and ∀k ∈ {0, ..., n} : ak ∈ C
Claim. Q is bounded, i.e. Q ∈ B(C)
58 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Proof of the Claim. For z ̸= 0 we have

n
X
|p(z)| = a0 + ak z k ≥
k=1
n−1
X
≥ |an ||z|n − |ak ||z|k ≥
k=1
n−1
!
X
≥ |z|n |an | − |ak ||z|k−n
k=1

|z|→∞ |z|→∞
Hence p(z) −−−−→ ∞ and consequently Q(z) −−−−→ 0, from this result we deduce
that ∃r > 0 : Q(z) ≤ 1 whenever |z| ≥ r (Q is a continuous function, i.e. Q ∈ C 0 (C)),
but Q is continuous and hence bounded on the compact set Dr (0), say Q < m for
some m ∈ R

Choose M := max{m, 1}, then

∀z ∈ C : Q(z) ≤ M
Hence Q is constant by Liouville’s Theorem 2.8
This contradicts the assumption that p is non-constant and hence proves the existence
of one solution by contradiction.

Hence p has a root, say w1 ∈ C. Then by writing z = (z − w1 ) + w1 we have

n
X
k
p(z) = a0 + ak (z − w1 ) + w1 =
k=1
n
X
= b0 + bk (z − w1 )k
k=1

using the Binomial Theorem [EW22], new coefficients bn−1 , ..., b0 and bn = an . Since
p(w1 ) = 0, we must then have b0 = 0

Hence !
n
X
p(z) = (z − w1 ) bk (z − w1 )k−1 = (z − w1 )p̃(z)
k=1

where p̃ is a polynomial with deg(p̃) = n − 1. By induction on the degree of the

polynomial we get the result.
Next we discuss the principle of analytic continuation (of identities), which states that:
if Ω is open and connected, f ∈ H (Ω) and f vanishes on an infinite set Z of distinct
points with a limit point z0 ∈ Ω \ Z , then f = 0.
2.3. CAUCHY’S INTEGRAL FORMULAS 59

Remark 2.9. 1. Holomorphic functions can have infinitely many zeroes, but we we
are going to see that these zeroes are isolated, i.e. for each zero z0 of a holomorphic
function f there exists a neighbourhood of z0 with no other zero.
E.g. cos(z) and sin(z) have respectively zeroes for z = (2k + 1) π2 and z = πk, with
k∈Z

2. There are holomorphic functions with no zeroes.

E.g. the constant function c or the complex exponential ez are examples of func-
tions with no zeroes.
We start by the definition of a limit point.

Definition 2.3. An element z0 ∈ C is a limit point of a set Ω, if

N∗
∃(zn )n∈N∗ ∈ Ω \ {z0 } : lim zn = z0
n→∞

Hence, it holds that ∀ε > 0 : Ω ∩ Ḋε (z0 ) ̸= ∅ and that ∀n ∈ N∗ : zn ̸= z0

Example 2.4. If Ω = [−1, 1] ∪ {2i}, then the zn ̸= z0 condition avoids the case 2i is a
limit point of Ω, since otherwise we could take ∀n ∈ N∗ : zn = 2i
We next define the order of zero of f at z0

Definition 2.4. Let Ω ⊆ C be open, f ∈ H (Ω) and z0 ∈ Ω, then the order of

zero of f at z0 or order of vanishing of f at z0 , denoted by ordz0 (f ) or nz0 (f )
or νz0 (f ), is either ∞, if ∀k ≥ 0 : f [k] (z0 ) = 0 or it is the smallest integer k ∈ N such
that f (z0 ) = f ′ (z0 ) = ... = f [k−1] (z0 ) = 0 and f [k] (z0 ) ̸= 0. If f (z0 ) ̸= 0, then k = 0.
Therefore, we define it as follows:
 
min{k ∈ N : f [k] (z0 ) ̸= 0} , {k ∈ N : f [k] (z0 ) ̸= 0} = ̸ ∅
ordz0 (f ) := ∈ N ∪ {∞}
∞ , {k ∈ N : f [k] (z0 ) ̸= 0} = ∅

We have the following result

Proposition 2.2. [SS10, Theorem III.1.1] Let Ω ⊆ C be open, f ∈ H (Ω) and

z0 ∈ Ω.

(i) If ordz0 (f ) = ∞, then ∃r > 0 : Dr (z0 ) ⊂ Ω and

∀r > 0 : Dr (z0 ) ⊂ Ω =⇒ ∀z ∈ Dr (z0 ) : f (z) = 0

i.e. f is locally zero around z0 .

60 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

(ii) If ordz0 (f ) ̸= ∞, then ∃r > 0 : Dr (z0 ) ⊂ Ω in which

∃!h ∈ H Dr (z0 ) ∃!n ∈ N∀z ∈ Dr (z0 ) : f (z) = (z − z0 )n h(z)

where h(z0 ) ̸= 0 and n = ordz0 (f )

(iii) For any f, g ∈ H (Ω) we have

ordz0 (f + g) ≥ min{ordz0 (f ), ordz0 (g)}

ordz0 (f g) = ordz0 (f ) + ordz0 (g)

Proof. (i) Let f be holomorphic in Ω with Ω open. By Theorem 2.7 and using that
Ω is open, ∃r > 0 is such that Dr (z0 ) ⊂ Ω and on it we have
∞
X f [n] (z0 )
∀z ∈ Dr (z0 ) : f (z) = (z − z0 )n
n=0
n!

Since ordz0 (f ) = ∞, it holds that ∀n ≥ 0 : f [n] (z0 ) = 0. Hence ∀z ∈ Dr (z0 ) :

f (z) = 0
(ii) If ordz0 (f ) ̸= ∞, then by definition
∃k ∈ N : f (z0 ) = ... = f [k−1] (z0 ) = 0 and f [k] (z0 ) ̸= 0
Again, using Theorem 2.7 and the fact that Ω is open, ∃r > 0 is such that
Dr (z0 ) ⊂ Ω, then for all z ∈ Dr (z0 ) we have the power series representation
∞
f [k] (z0 ) k
X f [n] (z0 )
f (z) = (z − z0 ) + (z − z0 )n =
k! n=k+1
n!
 
∞
k
 f [k] (z0 ) X f [m+k] (z0 ) 
= (z − z0 )  + (z − z0 )m 
=
| {z k! } m=1 (m + k)!
̸=0
| {z }
̸=0
∞
!
[m+k]
X f (z0 )
= (z − z0 )k (z − z0 )m
m=0
(m + k)!

Hence, if we define
∞
X f [m+k] (z0 )
∀z ∈ Dr (z0 ) : h(z) := (z − z0 )m
m=0
(m + k)!

Then h ∈ H Dr (z0 ) , since it is given by a convergent power series and also

f [k] (z0 )
h(z0 ) = k!
̸= 0, as (z − z0 )m = 0 if and only if m > 0
2.3. CAUCHY’S INTEGRAL FORMULAS 61

Note that, since h ∈ H Dr (z0 ) , it is also continuous there and since h(z0 ) ̸= 0,

it holds that
∃ε ∈ (0, r)∀z ∈ Dε (z0 ) : h(z) ̸= 0
Moreover, h and n are unique, since if we assumed not, hence

f (z) = (z − z0 )n h(z) = (z − z0 )m g(z)

with f, g holomorphic and h(z0 ) ̸= 0, g(z0 ) ̸= 0, then if m > n we would get

f (z) = (z − z0 )n (z − z0 )m−n g(z) =

= (z − z0 )n h(z)

and for z ̸= z0 that

h(z) = (z − z0 )m−n g(z)
Now, taking the limit on both sides as z → z0 gives h(z0 ) = 0, which is a
contradiction, therefore m = n. Then h(z) = g(z)
(iii) Note that for any k ∈ N

f [k] (z0 ) + g [k] (z0 ) = (f + g)[k] (z0 )

Hence, if f [k] (z0 ) = 0 = g [k] (z0 ), then also (g + f )[k] (z0 ) = 0. This implies that
ordz0 (f + g) ≥ min{ordz0 (f ), ordz0 (g)}
By part (ii), instead, we write

f (z) = (z − z0 )ordz0 (f ) h1 (z)

g(z) = (z − z0 )ordz0 (g) h2 (z)

with ∀z ∈ Dr (z0 ) : h1 (z) ̸= 0 and h2 (z) ̸= 0, then

f g = (z − z0 )ordz0 (f )+ordz0 (g) h1 (z)h2 (z)

with (h1 h2 )(z0 ) ̸= 0 From this, using the power series expansion of f g or the
uniqueness of n and h in part (ii) we get

ordz0 (f ) + ordz0 (g) = ordz0 (f g)

As Corollary we get that the zeroes of an holomorphic function are isolated. More
precisely we have

Theorem 2.9. Let Ω ⊆ C be open, f ∈ H (Ω) and z0 ∈ Ω. Assume f (z0 ) = 0 , i.e.

ordz0 (f ) ≥ 1. If ordz0 (f ) ̸= ∞, then

∃δ > 0∀z ∈ Ḋδ (z0 ) : f (z) ̸= 0

62 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Proof. Using Proposition 2.2 we have that ∃r > 0 : Dr (z0 ) ⊂ Ω in which we write
f |Dr (z0 ) (z) = (z − z0 )n h(z) with n = ordz0 (f ) and h(z0 ) ̸= 0. Let z ∈ Ḋr (z0 ), then

f (z) = 0 ⇐⇒ h(z) = 0

since (z − z0 )n ̸= 0 for z ̸= z0
Being h(z0 ) ̸= 0 and h is continuous on Dr (z0 ), we have that

∃δ ∈ (0, r] : |z − z0 | < δ =⇒ h(z) ̸= 0

By the equivalence previously stated we have that

∀z ∈ Ḋδ (z0 ) : f (z) ̸= 0

Now we can state the Principle of Analytic Continuation2 .

Theorem 2.10 (Principle of Analytic Continuation). [SS10, Theorem II.4.8] Let

Ω ⊆ C be a region and let f ∈ H (Ω). Let Z ⊂ Ω be an infinite set with a limit
/ Z . Then
point z0 ∈ Ω, but z0 ∈
 
∀z ∈ Z : f (z) = 0 =⇒ f = 0

Before we give the proof, we record the following immediate Corollary.

Corollary 2.6 (Identity Theorem). [SS10, Corollary II.4.9] Let Ω ⊆ C be a region

̸ U ⊆ Ω be one of its open subsets. Suppose f, g ∈ H (Ω), then
and ∅ =



∀z ∈ U : f (z) = g(z) =⇒ ∀z ∈ Ω : f (z) = g(z)
∗
This holds more in general for a sequence (zn )n∈N∗ ∈ Z N of distinct points with

2
It is useful to this purpose to recall the definition of a region in C, see 1.11.
2.3. CAUCHY’S INTEGRAL FORMULAS 63

/Z
limit point limn→∞ zn =: z0 ∈ Ω such that z0 ∈

Proof. Apply Theorem 2.10 to f − g to obtain this result.

Note that if U ⊆ Ω is open and non-empty, then ∃r > 0∃z0 ∈ U : Dr (z0 ) ⊆ U and the
r
sequence {z0 + n+1 }∞ r ∞
n=1 in Dr (z0 ) ⊆ U has a limit point z0 ∈ Ω \ {z0 + n+1 }n=1

Remark 2.10. 1. The reason this result is called Principle of Analytic Continua-
tion 2.10 is the following:

If f ∈ H (Ω) with a region Ω and Ω ⊆ Ω̃ is again a region, then there is at most

one f˜ ∈ H (Ω̃) such that ∀z ∈ Ω : f (z) = f˜(z). When such a function f˜ exists,
we say that f has analytic continuation f˜ to Ω̃

Note that if g ∈ H (Ω̃) is such that ∀z ∈ Ω : g(z) = f (z). Then ∀z ∈ Ω :

f˜(z) − g(z) = 0, hence ∀z ∈ Ω̃ : f˜(z) − g(z) = 0 by the above Theorem 2.10.
Therefore f˜ is unique.

2. The assumption that Ω is connected is essential, since if Ω = Ω1 ∪Ω2 with Ω1 ̸= ∅,

Ω2 ̸= ∅ and Ω1 ∩ Ω2 = ∅, then one can define f, g : Ω → C by f |Ω1 = 1 and
f |Ω2 = 0 and g = 0. Then even though f |Ω2 = g|Ω2 coincide in Ω2 , f and g do
not coincide in Ω

3. The condition that the limit point of zeroes is in Ω is also crucial, as shown in
the following Example 2.5.
Example 2.5. Take Ω = C∗ and

f :Ω→C
iπ −iπ
π  ez −e z
z 7→ sin =
z 2i
π −e−π
It holds that f ∈ H (C∗ ) and f ̸= 0, since already f (i) = e 2i
̸= 0
Consider a sequence of zeroes given by { n1 }n∈N∗ with limn→∞ 1
n
= 0, as f ( n1 ) = sin(πn) =
64 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

0 for all n ∈ N∗ . The limit point of zeroes is not in Ω, causing the result. This example
shows that the zeroes can converge to a boundary point. Note that we do have that the
zeroes { n1 }n∈N∗ are isolated.
Another interesting example regarding the above concept of continuation is given here
at Example 2.6.
We are now going to prove the following theorem, which proves Theorem 2.10.

Theorem 2.11 (Principle of Analytic Continuation alternative version). Let Ω be

a region, f ∈ H (Ω). Then the following are equivalent:

(i) f = 0

(ii) ∃a ∈ Ω∀n ∈ N : f [n] (a) = 0

(iii) Zf := {z ∈ Ω : f (z) = 0} has a limit point in Ω

An immediate Corollary of Theorem 2.11 is

Corollary 2.7 (Identity Theorem alternative version). Let Ω be a region in C,

f, g ∈ H (Ω). Then the following are equivalent:

(i) f = g

(ii) ∃a ∈ Ω∀n ∈ N : f [n] (a) = g [n] (a)

(iii) {z ∈ Ω : f (z) = g(z)} has a limit point in Ω

Proof of Theorem 2.11. Clearly (i)=⇒(iii), since by assumption {z ∈ Ω : f (z) = 0} =

Ω. In the following, we will prove (iii)=⇒(ii)=⇒(i):

(iii)=⇒(ii): Let Zf := {z ∈ Ω : f (z) = 0}. By assumption Zf has a limit point

a ∈ Ω. Since Ω is open, let r > 0 be a radius such that Dr (a) ⊂ Ω. We also have
that f is continuous and using that a is a limit point in Zf , i.e.

N∗
∃(zn )n∈N∗ ∈ Zf \ {a} : lim zn = a
n→∞

by continuity of f we have that

 
0 = lim f (zn ) = f lim zn = f (a)
n→∞ n→∞
2.3. CAUCHY’S INTEGRAL FORMULAS 65

Claim. It holds that ∀n ∈ N : f [n] (a) = 0

Proof of the Claim. Suppose on the contrary that

∃n ∈ N∗ ∀l ∈ {0, ..., n − 1} : f [l] (a) = 0 but f [n] (a) ̸= 0
Then as in the
proof of Theorem 2.7, since f is analytic in Dr (a) ⊂ Ω, i.e.
f ∈ C ω Dr (a) , expanding f in a power series there
∞
X
∀z ∈ Dr (a) : f (z) = ak (z − a)k
k=n

we have that by Proposition 2.2

f (z) = (z − a)n g(z)


with g(a) ̸= 0 and g is analytic in Dr (a), i.e. g ∈ C ω Dr (a)
Since g is continuous, ∃ε ∈ (0, r) : Dε (a) ⊂ Dr (a) such that ∀z ∈ Dε (a) : g(z) ̸= 0,
then
f (z) = (z − a)n · g(z)
| {z } |{z}
̸=0 on Ḋε (a) ̸=0 on Dε (a)

Therefore f (z) ̸= 0 on Ḋε (a). Hence Zf ∩ Ḋε (a) = ∅, but this says that a is not
a limit point of Zf . Hence ∀n ∈ N : f [n] (a) = 0


(ii)=⇒(i): Let A := z ∈ Ω : ∀n ∈ N : f [n] (a) = 0 . By assumption a ∈ A,
hence A ̸= ∅. We will show that A = Ω via the characterisation of connectedness,
hence that f = 0

Recall: for an open subset Ω ⊆ C, connected means that the only both open and
closed subsets of Ω are ∅ and Ω (It is not possible to find two disjoint non-empty
open sets Ω1 , Ω2 such that Ω = Ω1 ∪ Ω2 ).

Since A ̸= ∅, if we can show that A is both open and closed, then A = Ω. Hence:
A is open: To see this, let c ∈ A and let r > 0 such that Dr (c) ⊂ Ω. Then
by Theorem 2.7
∞
X
∀z ∈ Dr (c) : f (z) = an (z − c)n
n=0
with
f [n] (c)
an = =0
n!
since c ∈ A, hence f |Dr (c) = 0. This means that Dr (c) ⊆ A. Hence, for an
arbitrary c ∈ A, we found a neighbourhood Dr (c) ⊆ A, which shows that A
is open.
66 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

A is closed: We want to show that if {zk }k∈N∗ is a sequence of points in A

such that limk→∞ zk = c ∈ Ω, then c ∈ A, i.e. A contains all its limit points.
Let c ∈ Ω be a limit point of a sequence {zk }k∈N∗ in A. Then for any k ∈ N∗
and n ∈ N, it holds that
f [n] (zk ) = 0
by definition of the set A. We though have that f [n] is continuous, hence
 
0 = lim f [n] (zk ) = f [n] lim zk = f [n] (c)
k→∞ k→∞

Since n ∈ N was arbitrary, we obtain that f [n] (c) = 0 for every n ∈ N

and therefore c ∈ A. It follows from this that A is closed in Ω, given that
arbitrariness of c ∈ A

Remark 2.11. 1. (a) The Identity Theorem 2.7 makes it clear that the real func-
tions
sin, cos, exp : R → R
can be uniquely extended to complex numbers, via their form on the real line.
E.g. sin(z) and cos(z) are entire functions. For any z = x ∈ R we have that
sin2 (x) + cos2 (x) = 1. Thus we define f ∈ CC , z 7→ f (z) = sin2 (z) + cos2 (z)
and g ∈ CC , z 7→ 1. Since f and g agree on the real line, they have to agree
on all of C, from which

∀z ∈ C : sin2 (z) + cos2 (z) = 1

(b) The functional equations can also be transferred from reals to complex num-
bers.
E.g. from
∀x, y ∈ R : exp(x + y) = exp(x) exp(y)
we first conclude

∀y ∈ R∀z ∈ C : exp(z + y) = exp(z) exp(y)

and the another application of the Identity Theorem gives

∀z, w ∈ C : exp(z + w) = exp(z) exp(w)

General case: Let Ω ⊆ C be a region which contains a set U ⊂ Ω, which itself

contains a sequence of points with limit point also in U . Let F (z, w) be a function
defined for z, w ∈ Ω such that F (z, w) is analytic in z for any w and vice versa.
If F (z, w) = 0 whenever z, w ∈ U , then ∀z, w ∈ Ω : F (z, w) = 0
2.3. CAUCHY’S INTEGRAL FORMULAS 67

2. The geometric series for z ∈ D1 (0) is given by:

∞
X
g(z) = zn
n=0
1
has analytic continuation to C \ {1} given by f (z) = 1−z

P∞ 1
We have g : Ω1 → C, z 7→ n=0 z n and f : Ω2 ⊇ Ω1 → C, z 7→ 1−z
. Therefore, if
z ∈ D1 (0), we then have
1
f (z) = g(z) =
1−z
Example 2.6. Let Ω = D1 (0), then it holds that
∞
X
∀z ∈ Ω : f (z) = zn
n=0

f converges on D1 (0) and defines a holomorphic function there. Note thatPfor z = 1,

f (z) does not converge, hence for any ε > 0, we cannot define f (z) as ∞ n
n=0 z on
D1+ε (0), since any such disc contains z = 1

1
Let Ω̃ = C \ {1}, then Ω ⊆ Ω̃ and F (z) = 1−z is defined on all of Ω̃ and it agrees with
P∞ n 1
n=0 z whenever z ∈ D1 (0), so F (z) = 1−z is the analytic continuation of f to C\{1}

P∞
Warning: This does not say that F represents n=0 z n on the complement C \ D1 (0).

Note that in the Identity Theorem 2.7 we have two holomorphic functions defined on
the same set Ω. Here we have instead:
f : D1 (0) → C F : C \ {1} → C
∞
X 1
z 7→ zn z 7→
n=0
1−z
68 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Remark 2.12. Not every holomorphic function f : Ω → C can be extended to f˜ : Ω̃ →

C, with Ω ⊆ Ω̃

Example 2.7. On D1 (0) it holds that

∞
X
f (z) = z n!
n=0

n! n
converges by comparison to the geometric series, since |z|P∞< |z| . When we look at
n!
series of holomorphic functions, for any ε > 0 we see that n=0 z converges absolutely
and uniformly on compact subsets of D1−ε (0), but f cannot be extended anywhere beyond
D1 (0)

Here is another Corollary of the Identity Theorem 2.7:

Theorem 2.12. Let f, g ∈ H (Ω) with Ω open and connected. Then

f g = 0 =⇒ f = 0 or g = 0

Proof. Suppose without loss of generality f ̸= 0, we want to show that g = 0 (otherwise

by renaming of the functions, it falls under this same case).

Since f ̸= 0, so ∃a ∈ Ω : f (a) ̸= 0. By continuity of f , it holds that

∃ε > 0∀z ∈ Dε (a) ⊆ Ω : f (z) ̸= 0

The assumption
∀z ∈ Ω : f (z)g(z) = 0
then implies that ∀z ∈ Dε (a) : g(z) = 0. But then

g|Dε (a) = 0|Dε (a) = 0

Using the Identity Theorem 2.7 applied to g : Ω → C and the zero function 0 : Ω →
C, w 7→ 0 gives
g|Ω = 0|Ω = 0

Remark 2.13. The analytic functions on a non-empty open subset Ω ⊆ C, namely

H (Ω), form a commutative ring with 1. This since the sum and product of holomor-
phic functions are holomorphic.
2.3. CAUCHY’S INTEGRAL FORMULAS 69

Recall: (R, +, ·) is a ring with 1

• (R, +) is an abelian group.

• R is a monoid under “·”, i.e.

– ∀a, b, c ∈ R : (ab)c = a(bc)

– ∃1R ∈ R∀a ∈ R : a · 1R = 1R · a

• “·” distributes over “+”, i.e.

– ∀a, b, c ∈ R : a(b + c) = ab + ac
– ∀a, b, c ∈ R : (b + c)a = ba + ca

The last Theorem 2.12 says that if Ω is open and connected, then the ring of analytic
functions on it has no zero divisor, hence is an integral domain3 .
Our next application is the Morera’s Theorem, which is a converse to Goursat’s Theo-
rem 2.1.

Recall: Goursat’s Theorem 2.1 says the following: let f : Ω → C for Ω open be a
holomorphic function and let T in Ω be a triangle, whose interior is also contained in
Ω, then Z
f (z)dz = 0
T

Theorem 2.13 (Morera’s Theorem). [SS10, Theorem II.5.1] Let Ω ⊆ C open and
f ∈ C 0 (Ω; C). Assume that for any open disc D with D ⊂ Ω and any triangular
path T such that im(T ) ⊆ D we have that
Z
f (z)dz = 0
T

Then f ∈ H (Ω), i.e. f is holomorphic on Ω

Proof. Let z0 ∈ Ω and r > 0 such that Dr (z0 ) ⊂ Ω. For z ∈ Dr (z0 ) we define
Z
F (z) := f (w)dw
γz :=ℓ[z0 ,z]

3
In algebra, an integral domain is a non-zero commutative ring in which the product of any two
non-zero elements is non-zero. Integral domains are generalizations of the ring of integers and provide
a natural setting for studying divisibility. In an integral domain, every non-zero element a has the
cancellation property, that is, if a ̸= 0, an equality ab = ac implies b = c
70 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

where γz : [0, 1] → C, z 7→ (1 − t)z0 + zt is the line segment joining z0 to z. Then for a

small enough h ∈ C such that z + h ∈ Dr (z0 ), it holds that
Z
F (z + h) − F (z) = f (w)dw
σ:=ℓ[z,z+h]

where σ is the linear path linking z and z + h. Since by assumption

Z
f (w)dw = 0
T

for any T in Dr (z0 ), in particular for im(T ) = ⟨z0 , z, z +h⟩, we state the following claim.

Claim. Using continuity of f at z one can show that

F (z + h) − F (z)
lim = f (z)
h→0 h
Proof of the Claim. Consider the difference
Z


F (z + h) − F (z) = f (w) − f (z) + f (z) dw =
ℓ[z,z+h]
Z Z


= f (z) dw + f (w) − f (z) dw
ℓ[z,z+h] ℓ[z,z+h]
| {z }
=h

and the estimate obtained from 1.10 (iv), namely

Z


f (w) − f (z) dw ≤ sup |f (w) − f (z)| h
ℓ[z,z+h] w∈ℓ[z,z+h]

It follows that
F (z + h) − F (z)
− f (z) ≤ sup |f (w) − f (z)|
h w∈ℓ[z,z+h]
2.4. SEQUENCES OF HOLOMORPHIC FUNCTIONS 71

Since f is continuous at z, it holds that

h→0
sup |f (w) − f (z)| −−→ 0
w∈ℓ[z,z+h]

and thus finally

 
F (z + h) − F (z) h
lim = lim f (z) + lim sup |f (w) − f (z)| = f (z)
h→0 h h→0 h h→0 w∈ℓ
[z,z+h]

so F is holomorphic on Dr (z0 ), but then F ′ is also holomorphic on Dr (z0 ), by Theorem

2.3. Since F ′ = f , it follows that f is holomorphic on Dr (z0 ) as above, but then f is
holomorphic on all of Ω, as z0 ∈ Ω was arbitrary.

2.4 Sequences of holomorphic functions

It is known from Real Analysis that pointwise convergence of a sequence of functions
leads to pathologies, such as the pointwise limit of a sequence of continuous functions
not being necessarily continuous.
To avoid this we used a stronger form of convergence: the uniform convergence. For
example, the limit of a uniform convergent sequence of continuous functions is contin-
uous.
We also have that uniformly convergent sequences of integrable functions converges to
an integrable function.
Hence, uniform convergence of sequences of functions has better stability properties.
Uniformly convergent sequences of differentiable functions do not necessarily have dif-
ferentiable limits, though.
In this regard we are going to see that sequences of complex functions have much better
stability properties. As in the real case, the uniform limit of a sequence of continuous
functions is continuous and similarly line integrals of a uniformly convergent sequence
of functions converge to the line integral of the limit function.

N∗
Definition 2.5. A sequence (fn )n∈N∗ ∈ CΩ of functions defined on an open set
Ω ⊆ C is called uniformly convergent in Ω to the limit limn→∞ fn = f ∈ CΩ , if

∀ε > 0∃N ∈ N∀n ≥ N ∀z ∈ Ω : f (z) − fn (z) < ε


⇐⇒ lim sup f (z) − fn (z) : z ∈ Ω = 0
n→∞

Alternatively, such a sequence is denoted as fn ⇒ f or simply fn ⇒ f

n→∞
72 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Note that in this definition, N does not depend on z, rather only on ε

It turns out that we only need local uniform convergence or equivalently uniform con-
vergence on compact subsets.

N∗
Definition 2.6. Let Ω ⊆ C be open and (fn )n∈N∗ ∈ CΩ a sequence of functions.
Ω

N∗
(fn )n∈N∗ ∈ C is called locally uniformly convergent or compactly con-
vergent or uniformly convergent on compact sets, if the following equivalent
conditions are satisfied:
 
(i) ∀a ∈ Ω∃ε > 0 : Dε (a) ⊆ Ω and fn |Dε (a) converges uniformly in Dε (a)
n∈N∗

(ii) ∀K ⊆ Ω compact : ( fn |K )n∈N∗ converges uniformly in K

Remark 2.14. Note that the previous conditions are equivalent.

(i)=⇒(ii): since K is covered by finitely many discs in (i).
(i)=⇒(ii): since Ω is open, for all a ∈ Ω there is a closed disc D, i.e. compact, such
that a ∈ D ⊂ Ω

Remark 2.15. Note that since continuity is a local property even in the case of real
valued functions, local uniform convergence of continuous functions will imply continuity
of the limit function.

Hence, similarly to the real case one can show that:

N∗
Proposition 2.3. Let Ω ⊆ C be open and (fn )n∈N∗ ∈ C 0 (Ω; C) a sequence
of continuous functions. If (fn )n∈N∗ converge uniformly to a function f in every
compact subset of Ω, then f ∈ C 0 (Ω; C)

With this result in mind we continue to the main Theorem that we have in this section.

N∗
Theorem 2.14. [SS10, Theorem II.5.2] Let Ω ⊆ C be open and (fn )n∈N∗ ∈ H (Ω)
a sequence of holomorphic functions. If (fn )n∈N∗ converge uniformly to a function
f in every compact subset of Ω, then f ∈ H (Ω)

Proof. Since each of the fn ’s is holomorphic, they are all also continuous, hence by the
above Proposition 2.3 their limit f is also continuous.
2.4. SEQUENCES OF HOLOMORPHIC FUNCTIONS 73

To show that f is also holomorphic we will use Morera’s Theorem 2.13 and the fact
that the set described by any triangle T is compact.

By Morera’s Theorem 2.13, since f is already continuous, to show that f is holomorphic,

it is enough to show that Z
f (w)dw = 0
T
for any open disc D such that T ⊆ D ⊆ Ω and where T is a triangle contained in D

Let D = Dr (z0 ) ⊆ Ω an open disc in Ω for a z0 and a r > 0. Let T be any triangle with
inside contained in D. By Goursat’s Theorem 2.1, we have
Z
∀n ≥ 1 : fn (w)dw = 0
T

since fn (z) → f (z) uniformly on compact sets. Being the set described by T compact,
we have that fn (z) → f (z) uniformly on it, so
Z Z Z
fn (z)dz − f (z)dz ≤ |fn (z) − f (z)| |dz| ≤
T T T
n→∞
≤ sup |fn (z) − f (z)| LT −−−→ 0 · LT = 0
z∈T

for LT ∈ R the length of the perimeter of T and since (fn )n∈N∗ converges uniformly on
the set delimited by T . We hence have that
Z Z
lim fn (z)dz = f (z)dz
n→∞ T T
| {z }
=0

and therefore, we obtain Z

f (z)dz = 0
T

and so finally f ∈ H (Ω), using Morera’s Theorem 2.13 to conclude the proof.
We can extend the previous result to the following generalisation.

N∗
Theorem 2.15. [SS10, Theorem II.5.3] Let Ω ⊆ C be open and (fn )n∈N∗ ∈ H (Ω)
a sequence of holomorphic functions such that fn ⇒ f , i.e. fn converges uniformly
to f , on every compact subset K ⊆ Ω.

N∗
Then (fn′ )n∈N∗ ∈ H (Ω) converges uniformly to limn→∞ fn′ = f ′ on every com-
pact subset of Ω
74 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

Proof. Let z0 ∈ Ω and r > 0 such that Dr (z0 ) ⊂ Ω, then (fn )n∈N∗ converges uniformly
to f also on Dr (z0 ) as subset of Ω

Let s > r such that Ds (z0 ) ⊂ Ω and σ := r+s 2

∈ (r, s). We have then by the Cauchy
Integral Formula for derivatives 2.3 that
Z
′ 1 f (w)
f (z) = dw
2πi Cσ (z0 ) (w − z)2

for every z ∈ Dr (z0 ) ⊂ Dσ (z0 ). Hence for z ∈ Dr (z0 ) we have

fn (w) − f (w) fn (w) − f (w)

Z
′ ′ 1 1
|fn (z) − f (z)| = 2
dw ≤ (2πσ) sup
2πi Cσ (z0 ) (w − z) 2π w∈Cσ (z0 ) (w − z)2

To bound the denominator for both w ∈ Cσ (z0 ) and z ∈ Dr (z0 ), it holds that

|w − z| = |w − z0 + z0 − z| = w − z0 − (z − z0 ) ≥ |w − z0 | − |z − z0 | = |σ − r|

Hence, we conclude that

σ n→∞
|fn′ (z) − f ′ (z)| ≤ sup |fn (w) − f (w)| −−−→ 0
(σ − r)2 w∈Cσ (z0 )
2.4. SEQUENCES OF HOLOMORPHIC FUNCTIONS 75

and since fn ⇒ f on the compact set Cσ (z0 ), we have that fn′ ⇒ f ′ on Dr (z0 )

Since by definition every compact set is contained in a union of finitely many such discs,
the result follows.

These Theorems are often used to prove holomorphicity of functions defined by infinite
series.

N∗
Corollary 2.8. Let Ω ⊆ C be open and (fn )n∈N∗ ∈ H (Ω) , if
∞
X
∀z ∈ Ω : F (z) = fn (z)
n=1

we then let SN (z) = N

P
n=1 fn (z) and so SN is holomorphic. If the sequence (SN )N ∈N∗
converges uniformly on compact subsets of Ω, then limN →∞ SN (z) = F (z) is also
holomorphic.

Proof. This Corollary is a direct consequence of the previous Theorem 2.15 applied on
the sequence (SN )N ∈N∗ defined as follows: for each N ∈ N∗
N
X
SN : Ω → C, z 7→ SN (z) = fn (z)
n=1

For series of functions, we have also the following useful Theorem of Weierstrass, called
Weierstrass M-test.

N∗
Theorem 2.16 (Weierstrass M -test). Let (fn )n∈N∗ ∈ CΩ be a sequence of
functions and ∅ =
̸ U ⊆ Ω. Suppose that

N∗   X
∃(Mn )n∈N∗ ∈ R≥0 : ∀n ∈ N∗ ∀z ∈ U : fn (z) ≤ Mn and Mn < ∞
n∈N∗
P
Then n∈N∗ fn converges absolutely and uniformly on U

Proof. Let z ∈ Ω and consider N ∈ N∗ , then

N
X N
X
|fn (z)| ≤ Mn
n=1 n=1
76 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

by then taking the limit on both sides one gets

∞
X N
X N
X ∞
X
|fn (z)| = lim |fn (z)| ≤ lim Mn = Mn < ∞
N →∞ N →∞
n=1 n=1 n=1 n=1


Hence, |fn | n∈N∗ convergence absolutely and hence also in the normal sense, but being
(Mn )n∈N∗ independent of z, then also uniformly. The arbitrariness of z and N give the
result.
Example 2.8. For z = x + iy ∈ C with x, y ∈ R and n ∈ N the function z 7→ nz :=
exp(z log(n)) is an analytic function on C

|nz | = exp (x + iy) log(n) = exp x log(n) = nx

Then we have

Proposition 2.4. [SS10, Proposition VI.2.1] The series that represents the Rie-
mann Zeta function ∞
X 1
ζ(z) :=
n=1
nz
converges absolutely and uniformly on every half plane

Uδ := {z ∈ C : Re(z) ≥ 1 + δ} with δ > 0

and is holomorphic in {z ∈ C : Re(z) > 1}, i.e. ζ ∈ H {z ∈ C : Re(z) > 1}

Proof. For each δ > 0 it holds that, given

Re(z) = σ ≥ 1 + δ > 1
the series ζ(z) is uniformly bounded by
∞
X 1
1+δ
<∞
n=1
n

since
∞ ∞ ∞ ∞
X 1 X 1 X 1 X 1
|ζ(z)| = z
≤ z
≤ Re(z)
≤ 1+δ
<∞
n=1
n n=1
|n | n=1
n n=1
n
Applying the previous Weierstrass M -test 2.16 we obtain the result.
Hence ∞ 1
P
n=1 nz converges uniformly on every half plane as ∀δ > 0 : Re(z) ≥ 1 + δ > 1
and therefore defines a holomorphic function if Re(z) > 1 (every compact subset of
{z ∈ C : Re(z) > 1} is contained in such a half plane in which Re(z) ≥ 1 + δ).
2.4. SEQUENCES OF HOLOMORPHIC FUNCTIONS 77

Definition 2.7. For z ∈ H := {z ∈ C : Im(z) > 0} ⊂ C (note that H is open), we

define the Theta function

θ:H→C
∞
2πin2 z 2z
X X
z 7→ θ(z) := e =1+2 e2πin
n∈Z n=1

A basic result for it is the following.

Proposition 2.5. For all z ∈ H, the function θ(z) is well defined, i.e. the series
converges and defines a holomorphic function there.

Proof. We are going to show that θ converges uniformly on any subset of the form
Hδ := {z ∈ C : Im(z) ≥ δ} ⊂ C with δ > 0. Since any compact subset of H is
contained in such a set, by Theorem 2.14 this will imply the result.

Let δ > 0, for any z ∈ Hδ with z := x + iy and y ≥ δ > 0 it holds that

2 2 2 2y
∀n ∈ N : |e2πin z | = |e2πin x | ·|e−2πn y | = e−2πn ≤ e−2πny
| {z }
=1

Since y ≥ δ, it holds that e−2πny ≤ e−2πnδ < 1. Hence

∞ ∞
2πin2 z
X X
e ≤ e−2πnδ < ∞
n=0 n=0
| {z }
geometric series

P∞ 2πin2 z
Hence, n=0 e converges uniformly on Hδ for any δ > 0 and therefore also on
every of their compact subset. Thus it defines a holomorphic function on H, as argued
above.

Example 2.9 (Fourier analysis and the Theta function). The Fourier Transform serves
as important tool to Number Theory, in this settings we can think of the Theta function
as such a transform of a certain function. The uniqueness of the coefficients of this
transform allows the Theta function to identify the squares, for instance. So,

∞ ∞  1 , m=0
2πin2 z 2πin2 z
X X
1+2 e = α(m)e =⇒ α(m) = 2 , m is a square
0 , else

n=1 n=0
78 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

A more remarkable result is obtained when looking at, for example

∞
2 2 2 2
X X X X X
θ4 (z) = e2πi(n1 +n2 +n3 +n4 )z = β(m)e2πimz
n1 ∈Z n2 ∈Z n3 ∈Z n4 ∈Z m=1

Here is interesting that

β(m) = #{n1 , n2 , n3 , n4 ) ∈ Z4 : (n21 + n22 + n23 + n24 = m}
Remark 2.16. We are going to come back to ζ(s) and θ(z) in Chapter 4: P∞there, we
1
are going to use θ(z) to show that ζ(s) (which is defined as the series n=1 ns for
Re(s) > 1) has an analytic continuation to C \ {1}. There we are also going to show
the relation between these two functions allows the following identity:
1 ∞
z  Z
−z
z dt
π2Γ ζ(z) = θ(it) − 1 t 2
2 2 0 t
for Re(z) > 1. One can show that in fact the RHS makes sense for any z ∈ C \
{0, 1}, this gives the analytic continuation of ζ. The latter identity is known as Mellin
transform.

2.5 Holomorphic functions defined in terms of inte-

grals
Finally, we also have a similar Theorem for functions defined in terms of integrals, i.e.
similar to the Theorems for functions defined in terms of infinite series.

Many special functions in mathematics are defined in terms of integrals of the type
Z b
f (z) := F (z, t)dt
a

or as limits of such integrals.

Example 2.10 (Γ representation as limit). Consider in this sense
Z ε
Γ(z) := lim e−t tz−1 dt
ε→∞ 1
ε

We have the following Theorem

Theorem 2.17. [SS10, Theorem II.5.4] Let Ω ⊆ C be open and I = [a, b] ⊂ R a

compact interval. Let F : Ω × I → C be a function with the following properties:

(i) F ∈ C 0 (Ω × I)

(ii) ∀t0 ∈ I : ft0 (z) := F (z, t0 ) ∈ H (Ω)

2.5. HOLOMORPHIC FUNCTIONS DEFINED IN TERMS OF INTEGRALS 79

Then the function f defined by the following integral is holomorphic on Ω, hence

Z b
f (z) := F (z, t)dt ∈ H (Ω)
a

Proof. The idea is to use the Riemann sums to approximate the integral: let
n−1  
b−aX b−a
fn (z) := F z, a + j
n j=0 n

then fn (z) is a finite sum of holomorphic functions, hence holomorphic.

We want to show that (fn )n∈N∗ converges to f uniformly on compact subsets. Then
using Theorem 2.15 we can conclude that f is holomorphic.

Let K ⊆ Ω be compact. We use that a continuous function F : Ω × I → C, when

restricted to the compact set K × I, is uniformly continuous.

Hence
∀ε > 0∃δ > 0∀(z1 , t1 ), (z2 , t2 ) ∈ K × I :

ε
|z1 − z2 | < δ and |t1 − t2 | < δ =⇒ |F (z1 , t1 ) − F (z2 , t2 )| <
b−a
Let now n ∈ N∗ be such that b−a
n
< δ, then let z ∈ K and consider
!
n−1 Z a+(j+1) b−a  
X n b−a
fn (z) − f (z) = F z, a + j − F (z, t) dt
j=0 a+j b−a
n
n | {z }
| {z } dependent of t
independent of t

using
Z b n−1 Z
X a+(j+1) b−a
n
f (z) = F (z, t)dt = F (z, t)dt
a j=0 a+j b−a
n

and
n−1  
b−aX b−a
fn (z) = F z, a + j =
n j=0 n
n−1 Z a+(j+1) b−a  
X n b−a
= F z, a + j dt =
b−a
j=0 a+j n
n
n−1
F z, a + j b−a


X
n
= b−a
j=0 n
80 CHAPTER 2. CAUCHY’S THEOREM AND ITS APPLICATIONS

since the integrand is independent of t in the second last equality.

h i
Now, for t ∈ a + j b−a b−a
, it holds that t − a + j b−a b−a


n
, a + (j + 1) n n
< n
< δ and
since the “z-arguments” are equal, we also have 0 = |z − z| < δ. Hence
 
b−a ε
F z, a + j − F (z, t) <
n b−a

and finally
n−1
ε Xb−a ε b−a
|fn (z) − f (z)| ≤ = n =ε
b − a j=0 n b−a n

which gives the uniform convergence of fn to f on K, since z was arbitrary in that set.
Being K arbitrary we obtain by Theorem 2.15 that f is holomorphic.

Remark 2.17. With some more work one can also show that f ′ is given by
Z b Z b Z b
′ d ′ d
∀z ∈ Ω : f (z) = F (z, t)dt = F (z, t)dt = F ′ (z, t) dt
dz a a dz a | {z }
=ft′ (z)

In other words, we can interchange integration and differentiation.

Remark 2.18. Many special functions that appear as solutions of differential equa-
tion, for example Bessel functions, have integral representations,e.g. Jn (z) is defined as
solution of Bessel’s complex differential equations:

d2 f (z) df (z)
z2 2
+z + (z 2 − n2 )f (z) = 0
dz dz
For n ∈ Z, it holds that
Z π
1
Jn (z) = eiz sin(t) e−int dt
2π −π
Z π
1
and J0 (z) = eiz sin(t) dt
2π −π

with F (z, t) = eiz sin(t) is continuous on C × R. ForReach t ∈ [−π, π] it holds ft (z) =

π
eiz sin(t) : C → C is holomorphic. Hence the function −π eiz sin(t) dt is holomorphic on C
and by the previous Theorem 2.17, it results that
Z π
′ 1
eiz sin(t) i sin(t) dt


J0 (z) =
2π −π
Chapter 3

Meromorphic functions and

Residue Formula

The goal is to extend Cauchy’s Theorem 2.5 and the Cauchy Integral Formula 2.6 from
holomorphic functions to functions which might have singularities.

Recall: Cauchy’s Theorem 2.5 states that for any f ∈ H (Ω), any closed curve γ such
that the image of the curve and its interior are both contained in Ω, we have
Z
f (z)dz = 0
γ

The Cauchy Integral Formula 2.6 states instead that for any z ∈ D ⊂ Ω, with D a disc,
such that ∂D =: C and for any f |D ∈ H (D), we have
Z
1 f (w)
f (z) = dw
2πi C w−z

To this end, we are first going to look at isolated singularities of a function f . We will
see that there are three prototypes for these:

• removable singularities

• poles

• essential singularities
1
given respectively by the following functions at z = 0: sin(z)
z
, z1 , e z . For instance, for the
first function it holds that
∞
sin(z) X (−1)n z 2n
=
z n=0
(2n + 1)!

81
82 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

shows that z = 0 is a “removable” singularity. One can view the RHS1 as an analytic
continuation to all C of the LHS2 = sin(z)
z
. For the second function we have

1
lim =∞
z→0 z
1
whereas for the third function |e z | oscillates. In fact,
1 x↘0
• if z → 0 on positive real numbers, then |e x | −−→ ∞
1 x↗0
• if z → 0 on negative real numbers, then |e x | −−→ 0

These three examples of singularities are what we call a removable singularity, a pole
and an essential singularity respectively.

We will prove a generalisation of Cauchy’s Theorem 2.5 to functions that are holomor-
phic except for finitely many isolated points. This will lead us to the

Residue Formula: if f ∈ H (Ω) for an open Ω containing a circle C and its interior,
except finitely many points z1 , ..., zn with n ∈ N inside C, then
Z N
X
f (z)dz = 2πi Reszk (f )
C k=1
| {z }
a−1

where we will also see that f in a neighbourhood of z0 has the form

n
X a−ℓ
f (z) = G(z) +
ℓ=1
(z − z0 )ℓ

where G ∈ H Dr (z0 )

This Theorem, like Cauchy’s Theorem 2.5 can be used to evaluate many real integrals
and complex line integrals. It will also lead to many theoretical results just like Cauchy’s
Theorem 2.5 did.

Argument principle: which allows us to count the number of zeroes (and poles) of
holomorphic (meromorphic) functions inside closed curves.

Rouché’s Theorem: a holomorphic function can be perturbed slightly without chang-

ing the number of its zeroes.
Let f, g ∈ H (Ω) with Ω open and containing C and its interior. If ∀z ∈ C : |f (z)| ≥
1
Short form for: “Right Hand Side”
2
Short form for: “Left Hand Side”
3.1. ZEROES AND POLES 83

|g(z)|, then f and f + g have the same number of zeroes inside C

Open mapping Theorem: let f ∈ H (Ω) be non-constant in a region Ω, then f is

an open map, i.e. image of an open set is open.

Maximum modulus principle: if f is non-constant on Ω open and connected, with

compact closure Ω, and f is continuous on Ω, then

sup |f (z)| ≤ sup |f (z)|

z∈Ω z∈Ω\Ω

Another way to say this is the following

max |f (z)| ≤ max |f (z)|

z∈Ω z∈Ω\Ω
| {z }
exists because Ω is compact

3.1 Zeroes and poles

We start with a definition of singularities

Definition 3.1. Let Ω ⊆ C be open and let f ∈ H (Ω). The point z0 ∈ C \ Ω is

called a (possible) isolated singularity of f if
 
∃r > 0 : f |Ḋr (z0 ) ∈ H Ḋr (z0 )

or if
∃Uz0 ∈ OCΩ : f |Uz ∈ H Uz0 \ {z0 }


0 \{z0 }

where OCΩ is the induced topology from C

Note that in the above Definition 3.1 the term ”possible” means that the singularity in
question at that point can possibly be removed.

Example 3.1. Let tan z1 be defined in C without some countable set has singularities

at
2 2 2
, , ,...,0
π 3π 5π
Note that 0 is not an isolated singularity of f . The points
2 3 2
, ,...,
π 2π (2k + 1)π

for k ∈ N are instead isolated.

84 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Example 3.2. Consider the function

f : C∗ = C \ {0} → C
z 7→ z

z = 0 is an isolated singularity of f , because f is not defined there, but f can in fact be

extended to all of C by defining f (0) = 0. In this case, z = 0 is a “removable singular-
ity” of f

On the other hand, we have that

f : C∗ = C \ {0} → C
1
z 7→
z
has a singularity at z = 0, which cannot be removed.

Definition 3.2.
Let Ω ⊆ C be open. An isolated singularity z0 ∈ C of a function

f ∈ H Ω \ {z0 } is called removable if f is holomorphically extendable to all
of Ω, i.e.
∃F ∈ H (Ω)∀z ∈ Ω \ {z0 } : F (z) = f (z)

We have the following Theorem of Riemann, sometimes called Riemann continuation

Theorem.

Theorem 3.1 (Riemann continuation Theorem). Let Ω ⊆ C be open and non-

with z0 ∈ Ω. Then the following assertions for a function f ∈
empty together
H Ω \ {z0 } are equivalent:

(i) f is holomorphically extendable to Ω

(ii) f is continuously extendable to Ω

 
(iii) f is bounded in a neighbourhood of z0 , i.e. ∃r > 0 : f |Ḋr (z0 ) ∈ B Ḋr (z0 )

(iv) limz→z0 (z − z0 )f (z) = 0

Proof. Exercise.
If a point satisfies one of these equivalent conditions, we call it a removable singularity.
As a consequence of Riemann’s continuation Theorem we have Riemann’s Theorem
on removable singularities.
3.1. ZEROES AND POLES 85

Theorem 3.2 (Riemann’s Theorem on removable
singularities). [SS10, Theorem III.3.1]

Let Ω ⊆ C be open, z0 ∈ Ω and f ∈ H Ω \ {z0 } , if f |Ḋr (z0 ) ∈ B Ḋr (z0 ) for some

Dr (z0 ) ⊂ Ω with r > 0, then z0 is a removable singularity of f , i.e.

∃F ∈ H (Ω)∀z ∈ Ω \ {z0 } : F (z) = f (z)

Note that this represents (iii)=⇒(i) in the previous Theorem which states (iii)⇐⇒(i)

Example 3.3. 1. Let f (z) = sin(z)

z
for z ̸= 0, f has a removable singularity at z = 0.
We can see this either using (iv) of the Riemann’s continuation Theorem 3.1

lim zf (z) = lim sin(z) = 0

z→z0 z→z0

or using the extension of L’Hôpital’s Rule to C [see Appendix B.2].

sin(z) cos(z)
lim = lim =1
z→z0 z z→z 0 1
sin(z)
which
 imply  that f (z) = z is bounded in a neighbourhood of 0, i.e. f ∈
B Ḋr (0) . In fact, it holds that

sin(z)
lim =1
z→0
z̸=0
z

Finally, let ε = 21 , then ∃r > 0∀|z| ∈ (0, r) it holds that

sin(z) 1
−1 <
z 2
 
and therefore f ∈ B Ḋr (0) . One can also use that on C

∞
X (−1)n z 2n+1
sin(z) =
n=0
(2n + 1)!

and hence it follows that on C

∞
X (−1)n z 2n
F (z) =
n=0
(2n + 1)!

sin(z)
is the holomorphic extension of z
to all of C
86 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

2. Another such example is: if f (z) = z

ez −1
on C∗ , then f has removable singularity
at z = 0, since
z 1
lim f (z) = lim = lim z = 1
z→0 z→0 ez − 1 z→0 e
by the same extension of L’Hôpital’s rule  as before. Hence f is bounded in a
neighbourhood of 0, i.e. ∃r > 0 : f ∈ B Ḋr (0)

If f ∈ H Ω \ {z0 } does not have a removable singularity at z0 , then f is not bounded

near z0 . We can then ask whether its unboundedness is similar to the unboundedness
of (z−z1 0 )n , i.e. we can ask whether (z − z0 )n f (z) is bounded near z0 for n sufficiently
large. If such an n ∈ N exists, then z0 is called a pole of f , let us hence first define this
once for all.

Definition 3.3 (Pole). Let z0 ∈ C and f ∈ H Ω \ {z0 } , if

 
∃n ∈ N∗ ∃r > 0 : Ḋr (z0 ) ⊂ Ω and (z − z0 )n f |Ḋr (z0 ) ∈ B Ḋr (z0 )

then z0 is called a pole of f and the natural number

n  o
∗ n
m := min n ∈ N : ∃r > 0 : (z − z0 ) f ∈ B Ḋr (z0 ) ∈ N∗

is called the order of the pole of f at z0 . Poles of first order are called simple
poles.

Remark 3.1. To compute the order of pole of f at z0 , one can can compute the order
of zero of f1 at z0 , as shown later.
1
Example 3.4. The function f (z) = (z−z0 )m
for z ∈ C \ {z0 } has a unique pole of order
m at z = z0
We will see soon that poles arise from reciprocals of holomorphic functions with zeroes.

Before we make this more precise, let us recall that zeroes of holomorphic functions are
isolated and we have the following Proposition for their behaviour near a zero.

Recall (Proposition 2.2): Suppose f ∈ H (Ω) in Ω open connected, f has a zero at a

point z0 ∈ Ω and does not vanish identically on Ω. Then
∃r > 0∃!g ∈ H Dr (z0 ) \ { 0|Dr (z0 ) }∃!n ∈ N∗ : Dr (z0 ) ⊂ Ω and

∀z ∈ Dr (z0 ) : f (z) = (z − z0 )n g(z) and n = min{n ∈ N : f [n] (z0 ) ̸= 0}

We remark that g is a unique non-vanishing holomorphic function in H Dr (z0 )

The analogous Theorem for the poles is the following:

3.1. ZEROES AND POLES 87

Theorem 3.3. Let Ω ⊆ C, z0 ∈ Ω and f ∈ H Ω \ {z0 } . For m ∈ N∗ the following

statements are equivalent:

(i) f has a pole of order m at z0 , i.e. (z − z0 )m f is bounded near z0 and m is the

smallest such integer.

(ii) ∃r > 0 : Dr (z0 ) ⊂ Ω and ∃g ∈ H Dr (z0 ) : g(z0 ) ̸= 0 and

g(z)
∀z ∈ Ḋr (z0 ) : f (z) =
(z − z0 )m

(iii) ∃r > 0 : Dr (z0 ) ⊂ Ω and ∃h ∈ H Dr (z0 ) ∀z ∈ Ḋr (z0 ) : h(z) ̸= 0, moreover h

has a zero of order m at z0 and such that

1
∀z ∈ Ḋr (z0 ) : f (z) =
h(z)

Proof. (i)=⇒(ii): The fact that f has a pole of order m at z0 means that (z − z0 )m f (z)
is bounded near z0 and that m is minimal, thus ∃r > 0 : Dr (z0 ) ⊂ Ω on which
f in bounded.
The Riemann’s Theorem on removable singularities 3.2 states that
∃g ∈ H Dr (z0 ) such that g(z) = (z − z0 )m f (z) whenever z0 ̸= z ∈ Dr (z0 )

If g(z0 ) = 0, then it would imply by the previous Proposition 2.2 that

g(z) = (z − z0 )g̃(z)

where g̃ ∈ H Dr (z0 ) . Consequently, this would give that for z ∈ Dr (z0 )

g̃(z) = (z − z0 )m−1 f (z)

is bounded near z0 and this would contradict the minimality of m. Hence g(z0 ) ̸= 0
and together with it we get that f (z) = (z − z0 )−m g(z) for z ∈ Ḋr (z0 )

(ii)=⇒(iii): Suppose that ∃r > 0 as in (ii) and that ∃g ∈ H Dr (z0 ) such that

g(z0 ) ̸= 0 and f (z) = (z − z0 )−m g(z) for z ∈ Ḋr (z0 ). Since g(z0 ) ̸= 0 and g is contin-
uous, then ∃r > 0 : ∀z ∈ Dr (z0 ) : g(z) ̸= 0, meaning that r can be chosen adequately
since the beginning.

Let for z ∈ Dr (z0 )

(z − z0 )m 1
h(z) := =
g(z) f (z)
88 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

then h(z) ̸= 0 for z ∈ Ḋr (z0 ) and therefore h ∈ H Dr (z0 ) and with h(z0 ) = 0. Also,

we have that in Ḋr (z0 )

1
= (z − z0 )−m g(z) = f (z)
h(z)
1
Note that h has a zero of order m at z0 , since h(z) = (z − z0 )m g(z) and ∀z ∈ Dr (z0 ) :
1
g(z)
̸= 0

(iii)=⇒(i): Suppose that ∃r > 0 : Dr (z0 ) ⊂ Ω and ∃h ∈ H Dr (z0 ) such that h(z) ̸= 0

in Ḋr (z0 ). Suppose also that h has a zero of order m at z0 and

1
∀z ∈ Ḋr (z0 ) : f (z) =
h(z)

Since h has a zero of order m at z0 , then by Proposition 2.2 ∃k ∈ H Dr (z0 ) such that

h(z) = (z − z0 )m k(z) and

∃s > 0∀z ∈ Ds (z0 ) ⊆ Dr (z0 ) : k(z) ̸= 0

1
Since k is holomorphic and non-vanishing, k
is holomorphic on Ds (z0 ). But then

1 1
∀z ∈ Ḋs (z0 ) : f (z) = = (z − z0 )−m
h(z) k(z)

1
would imply that (z − z0 )m f (z) = k(z) is holomorphic on Ḋs (z0 ) and has the holo-
1
morphic extension k on Ds (z0 ) (given that k1 is holomorphic on Ds (z0 ) since k ̸= 0 on
Ds (z0 )).

By the Riemann’s continuation Theorem 3.1 we have that

(z − z0 )m f (z)

is bounded in a neighbourhood of z0 . Moreover, it holds that

  
m−1 1 1
(z − z0 ) f (z) =
k(z) z − z0

is not bounded since

1 1 z→z0
̸= 0 and −−−→ ∞
k(z0 ) z − z0
Hence, m ∈ N∗ is minimal and f has a pole of order m at z0

Example 3.5. Consider the following examples:

3.1. ZEROES AND POLES 89

1
1. The function f (z) = ez −1
has a pole of order 1 at z = 0, since
∞
! ∞ ∞
!
1 X zn X zn X z n−1
= ez − 1 = −1= =z 1+
f (z) n=0
n! n=1
n! n=1
n!

1
has a zero of order 1 at z = 0 (note that ez −1
has simple poles at z = 2πik for
k ∈ Z).
z
2. The function f (z) = z 2 −1
has poles of order 1 at z = ±1, since
 
−1 z
f (z) = (z − 1)
z+1
z
and h(z) = z+1
is holomorphic and non-vanishing on D 1 (1). Similarly,
2

 
−1 z
f (z) = (z + 1)
z−1
z
and h̃(z) = z−1
is holomorphic and non-vanishing on D 1 (−1)
2

The next Theorem is the analogue of the power series expansion of a holomorphic func-
tion.

Recall: if f ∈ H (Ω) and z0 ∈ Ω such that Dr (z0 ) ⊂ Ω. Then by Theorem 2.7

∞
X
∀z ∈ Dr (z0 ) : f (z) = an (z − z0 )n
n=0

For functions with poles we have

90 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Theorem 3.4. [SS10, Theorem III.1.3] If f ∈ H Ω \ {z0 } has a pole of order n

at z0 , then
n
X a−k
f (z) = G(z) +
k=1
(z − z0 )k

where G is a holomorphic function in a neighbourhood of z0 , i.e. G ∈ H Dr (z0 )

for some r > 0

Proof. f has a pole of order n at z0 , hence ∃r > 0 such that D ˙ (z ) ⊂ Ω \ {z } and we

r 0 0
can write
∀z ∈ Ḋr (z0 ) : f (z) = (z − z0 )−n g(z)
with g ∈ H Dr (z0 ) and with g(z0 ) ̸= 0 with Theorem 3.3

We expand g in a power series by Theorem 2.7, hence

∞
X g [k] (z0 )
∀z ∈ Dr (z0 ) : g(z) = (z − z0 )k
k=0
k!

It follows that for any z ∈ Ḋr (z0 ) we can write

1
f (z) = g(z) =
(z − z0 )n
∞
1 X g [k] (z0 )
= (z − z0 )k =
(z − z0 )n k=0
k!
n−1
! ∞
X g [k] (z0 ) X g [k] (z0 )
= + (z − z0 )k−n =
k=0
k!(z − z0 )n−k k=n
k!
| {z }
=G(z)
n
!
X g [n−ℓ] (z0 )
= + G(z)
ℓ=1
(n − ℓ)!(z − z0 )ℓ

g [n−k] (z0 )
where ∀k ∈ {1, ..., n} : a−k := (n−k)!
P∞
Remark 3.2. The function f (z) := −n an (z − a)k is a special case of a Laurent
series (see Serie 9 and Serie 10).

Definition 3.4. • The number a−1 , i.e. the coefficient of (z − z0 )−1 in Theorem
3.4, is called the residue of f at the pole z0 , denoted by

Resz0 (f ) := a−1
3.1. ZEROES AND POLES 91

• The function n
X a−j
Pz0 (f, z) := Pzf0 (z) :=
j=1
(z − z0 )j

in Theorem 3.4 is called the principal part of f at the pole z0 , where

g [j] (z0 )
∀j ∈ {1, ..., n} : a−j :=
j!
as in latter proof.

Remark 3.3. If f has a pole of order 1 at z0 , i.e. a simple pole, then

Resz0 (f ) = lim (z − z0 )f (z)

z→z0

Since if f has a simple pole at z0 , then

a−1
f (z) = + g(z)
z − z0

with g ∈ H Dr (z0 ) . Hence, it follows that

(z − z0 )f (z) = a−1 + (z − z0 )g(z)

lim (z − z0 )f (z) = a−1 + lim (z − z0 )g(z)
z→z0 z→z0

lim (z − z0 )f (z) = a−1 = Resz0 (f )

z→z0

Conversely, if limz→z0 (z −z0 )f (z) exists and is non-zero, then (z −z0 )f (z) is bounded in
some neighbourhood of z0 . Hence, z0 is a pole of f of order at most 1 by our definition
of pole.

If the limit exists and is equal to 0, then it means that f has a removable singularity at
z0

More generally we have the following Theorem:

Theorem 3.5. [SS10, Theorem III.1.4] If f ∈ H Ω \ {z0 } has a pole of order n

at z0 , then
1 dn−1
(z − z0 )n f (z)


Resz0 (f ) = lim n−1
z→z0 (n − 1)! dz

Proof. Let
f (z) = Pzf0 (z) + G(z)
92 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

a−j
and G ∈ H (Dr (z0 )). Then it holds that
Pn
for z ∈ Dr (z0 ) with Pzf0 (z) = j=1 (z−z0 )j

n
X
n n
(z − z0 ) f (z) = G(z)(z − z0 ) + a−j (z − z0 )n−j
j=1

differentiating n − 1 times gives

dn−1 n

dn−1 n


(z − z0 ) f (z) = G(z)(z − z0 ) + (n − 1)!a−1
dz n−1 dz n−1
and lastly
dn−1
(z − z0 )n f (z) = (n − 1)!a−1 + lim (z − z0 )G̃(z)


lim n−1
z→z0 dz z→z0
| {z }
=0

for some G̃ ∈ H (Dr (z0 )) by the General Leibniz Rule [EW22]. Hence, the final result
is obtained by some final algebraic manipulations, so
1 dn−1 n


Resz0 (f ) = a−1 = lim (z − z0 ) f (z)
z→z0 (n − 1)! dz n−1

Example 3.6. Consider the examples:

1. For the function f (z) = z21+1 we have a simple pole at z = i
 
1 1 1 1
Resi 2
= lim(z − i) 2 = lim =
z +1 z→i z +1 z→i z+1 2i

1
2. The function f (z) = has two poles of order 2 at z = ±i
(z 2 +1)2
   
1 d 2 1 d −2
Res±i (f ) = lim (z ± i) 2 = lim =
z→±i (2 − 1)! dz (z + 1)2 z→±i dz (z ± i)2
−2 ±1
= lim 3
=
z→±i (z ± i) 4i

Remark 3.4. The following one is a useful tool to calculate residues of simple poles.

Lemma 3.1. If f, g ∈ H (Ω) with z0 ∈ Ω, f (z0 ) ̸= 0 and g has a simple zero at z0 ,

then fg has a simple pole at z0 and
 
f (z) f (z0 )
Resz0 =
g(z) g ′ (z0 )
3.1. ZEROES AND POLES 93

Proof. It is clear that if g has a simple zero at z0 , then g(z) = (z − z0 )g̃(z) where
g̃(z0 ) ̸= 0, non-zero and holomorphic in some Dr (z0 ) ⊂ Ω for some r > 0
f (z) f (z)
= (z − z0 )−1
g(z) g̃(z)
where fg̃(z)
(z)
∈ H Dr (z0 ) and non-zero at z0 . So fg(z)
(z)


has a simple pole at z0 . We now
f
apply Theorem 3.5 to g , so
 
f f (z) (z − z0 )
Resz0 = lim (z − z0 ) = lim f (z) =
g z→z 0 g(z) |{z} z→z 0 g(z) − g(z0 )
g(z0 )=0

(z − z0 ) f (z0 )
= f (z0 ) lim = ′
z→z0 g(z) − g(z0 ) g (z0 )
since g is holomorphic at z0
Example 3.7. 1. Let g(z) = z21+1 with a simple pole at z = i, then
   
1 1 1 1
Resi = Resi = =
z2 + 1 g(z) g ′ (i) 2i
 3 
2. We want to determine Resi z2z+1 , hence we can either use partial fraction ex-
pansion as
z3 z 1 1 1 1
2
=z− 2 =z− −
z +1 z +1 2z −i 2z +i
 3 
and get Resi z2z+1 = − 21 , or use the above Lemma 3.1
 3 
z f (i) i3 1
Resi 2
= ′
= =−
z +1 g (i) 2i 2
with f (z) = z 3 and g(z) = z 2 + 1 (notably faster in this case).
Remark 3.5. Note that, if f (z) = Pzf0 (z) + G(z) for z ∈ Dr (z0 ), where Pzf0 is the
principal part of f at z0 and G is a holomorphic function, and C(z0 ) is any circle
centred at z0 and contained in Dr (z0 ), then
Z Z n
f
X a−k
Pz0 (z)dz = k
dz = 2πia−1
C(z0 ) C(z0 ) k=1 (z − z0 )

since3 Z 
1 0 , n ̸= 1
dz =
C(z0 ) (z − z0 )n 2πi , n = 1
3
Using the Cauchy’s Integral Formula 2.6 with the constant function 1|D(z0 ) : D(z0 ) → C, z 7→ 1
and C = ∂D
94 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

By Cauchy’s Theorem 2.5 we also know that if C(z0 ) ⊂ Dr (z0 ), then

Z
G(z)dz = 0
C(z0 )

Hence, we have that Z

f (z)dz = 2πia−1
C(z0 )

In fact, we have the following general formula

Theorem 3.6 (The Residue Formula). [SS10, Theorem III.2.1] Let Ω ⊆ C be

open and let S := {z0 , ..., zn } ⊆ Ω be a finite set. Suppose f ∈ H (Ω \ S), i.e.
holomorphic except for poles at z0 , ..., zn ∈ S. Let γ be any circle contained in Ω
with counterclockwise orientation and such that im(γ) ∩ S = ∅. Let D be the open
disc bounded by γ. Then
Z X
f (z)dz = 2πi Resz (f )
γ z∈S∩D

Before we give the proof we will look at simple examples which use this formula to
calculate integrals.

Example 3.8. 1. Let γ be the circle such that |z| = 3 and consider the integral

Z    
dz 1 1
= 2πi Resi + 2πi Res−i =
γ (z 2 + 1)2 (z 2 + 1)2 (z 2 + 1)2
 
1 −1
= 2πi + =0
4i 4i
3.1. ZEROES AND POLES 95

2. The second example is

z3 z3 z3
Z    
2 2
dz = 2πi Resi + 2πi Res−i =
C3 (0) (z + 1) (z 2 + 1)2 (z 2 + 1)2
 
−1 −1
= 2πi + = −2πi
2 2

3. The third example is Z

dz
=0
C 1 (1) (z 2 + 1)2
2
1 1
since there is no pole of (z 2 +1)2
inside the circle given by |z − 1| = 2

4. The fourth example is

ez ez ez
Z    
2 2
dz = 2πi Res1 + 2πi Res−1
C2 (0) (z − 1) (z 2 − 1)2 (z 2 − 1)2
Evaluating separately the two residues we obtain
ez
 
f (1) e
Res1 2 2
= ′ =
(z − 1) g (1) 2
where f (z) = ez and g(z) = z 2 − 1 and using the Lemma 3.1, also
ez
 
f (−1) −1
Res1 = =
(z 2 − 1)2 g ′ (−1) 2e
by the same procedure on the other pole. Hence, we finally obtain
ez
Z
2 − 1)2
dz = πi(e − e−1 )
C2 (0) (z
We can now proceed to the proof of the formula.
Proof of Theorem 3.6. Let us first assume that f is holomorphic on an open set Ω
containing a circle and its interior, except for a single pole at z0 inside γ. Let D be the
disc bounded by γ
96 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

By Theorem 3.4
f (z) = Pzf0 (z) + G(z)
where G is holomorphic in a neighbourhood Dr (z0 ) of z0 and
n
X a−k
Pzf0 (z) =
k=1
(z − z0 )k

is the principal part of f at z0 , note that Pzf0 (z) is holomorphic in all of C \ {z0 }.
Another way to say this is that the function f (z) − Pzf0 (z) extends holomorphically to
Ω, as for the function

f (z) − Pzf0 (z) , z ∈ Ω \ {z0 }
g : Ω → C, z 7→ g(z) =
G(z0 ) , z ∈ Dr (z0 )

is the holomorphic extension of f (z) − Pzf0 (z) to Ω

We then obtain that

Z Z
f (z) − Pzf0 (z) dz = 0


g(z)dz =
γ
Zγ Z
f (z)dz = Pzf0 (z)dz
γ γ
R
and we are left to prove that γ Pzf0 (z)dz = 2πia−1 . This follows from the Cauchy
Integral Formula 2.6 applied to the constant function F : Ω → C, z 7→ F (z) = 1

Recall: The Cauchy Integral Formula for derivatives 2.3. Let C = ∂D be any circle
whose interior D is contained in Ω. Then for F ∈ H (Ω) and any z ∈ D it holds that
Z
[n] n! F (w)
F (z) = dw
2πi C (w − z)n+1

hence,
2πi dz n−1
Z 
dz 0 , n−1≥1
= (1) =
γ (z − z0 )n n! dz n−1 2πi , n − 1 = 0
and so we get Z
f (z)dz = 2πia−1 = 2πiz0 Res(f )
γ

For the general case consider f holomorphic in Ω except for finitely many points z0 , ..., zn
and let Pzfk be the principal part at zk for k ∈ {0, ..., n} =: I, which is holomorphic in
C \ {zk }
3.1. ZEROES AND POLES 97

Hence, define in Ω \ S the following function

X
g : Ω \ S → C, z 7→ g(z) := f (z) − Pzfk (z)
zk ∈S
k∈I

then for g ∈ H (Ω \ S) and in fact g can be extended holomorphically to all of Ω. To

see this let z0 ∈ S, r > 0 such that Dr (z0 ) ⊂ D, Ḋr (z0 ) ∩ S = ∅ and f (z) − Pzf0 (z) is
holomorphic in Dr (z0 )

Then for z ∈ Ḋr (z0 )

X
g(z) = f (z) − Pzf0 (z) + Pzfk (z)
zk ∈S\{z0 }
| {z }
extends holomorphically to Dr (z0 ) k∈I\{0}
| {z }
holomorphic in Dr (z0 )

This gives an extension of g to (Ω \ S) ∪ {z0 } = Ω \ {z1 , ..., zn }, we can do this for each
zk ∈ S with k ∈ I to get a holomorphic extension of g to all Ω. By Cauchy’s Theorem
2.5 we obtain that Z
g(z)dz = 0
γ
which in return gives Z XZ
f (z)dz = Pzfk (z)dz
γ zk ∈S γ
k∈I

and if z̃ ∈ S ∩ D, then as before (for ℓ = ordz̃ (f ))

Z Z X ℓ
f a−j
Pz̃ (z)dz = j
dz = 2πia−1 = 2πi Resz̃ (f )
γ γ j=1 (z − z̃)

If z̃ ∈ S, but not inside D, then

Z
Pz̃f (z)dz = 0
γ
98 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

since then Pz̃f is holomorphic inside the disc.

Hence, we finally get Z X

f (z)dz = 2πi Resz̃ (f )
γ z̃∈S∩D

Remark 3.6. 1. Another way to prove this is the following: first assume there is
just one pole inside of γ. Consider the following contour Γε,δ

inside of it f is holomorphic and we can show using Cauchy’s Theorem 2.5 that
Z
f (z)dz = 0
Γε,δ

Here we went around the pole z0 with a circle of radius ε > 0. The width of the
corridor is δ > 0
We can then make the width of the corridor narrower by letting δ → 0 and use
continuity of f to show that the two sides of the corridor cancel each other. The
remaining part consists of two curves, the larger circle γ and the small circle
Cε (z0 ) with clockwise orientation; we therefore get
Z Z
f (z)dz + f (z)dz = 0
γ Cε (z0 )

It takes, though, some effort to make this argument rigorous.

2. The best way to understand and generalise the Residue Formula 3.6 (and Cauchy
Integral Formula 2.6) is via homotopy. It is based on the following principle.
Let f be holomorphic on an open set Ω. For example, the space between two
circles.
3.1. ZEROES AND POLES 99

The principle is that if two closed curves can be deformed to each other while
remaining in Ω, then Z Z
f (z)dz = f (z)dz
γ1 γ2

We are going to get back to this soon.

3. Assume γ is not a circle, then consider it a triangle, a polygon or any curve γ,
which has a parametrisation of the form
γ : [a, b] → C \ {z0 }
t 7→ z0 + r(t)eiθ(t)


for some r, θ ∈ C 1 [a, b], R functions such that r(t) > 0, r(a) = r(b), θ(a) = 0
and θ(b) = 2π

Moreover,
r(t) = γ(t) − z0
and θ(t) is a continuous choice of argument along the line segment γ̃(t) = γ(t)−z0 ,
lastly we have that
γ(t) − z0
eiθ(t) =
γ(t) − z0
100 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Then it holds that

γ ′ (t) = r′ (t)eiθ(t) + r(t)eiθ(t) iθ′ (t)

and so we obtain that

b Z b ′ Z b
γ ′ (t)
Z Z
1 dz r (t)
= iθ(t)
dt = dt + i θ′ (t)dt =
2πi γ z − z0 a r(t)e a r(t) a

b b
= log r(t) a + i [θ(t)]a = 0 + 2πi

(This is similar to the parametrisation of a circle using a point inside other than
the centre)

Note that for

Z
dz
=0
γ (z − z0 )n
1 1 1
with n > 1, since (z−z0 )n−1 1−n
is a primitive of (z−z0 )n
in C \ {z0 } and γ is a
closed curve.

Hence, for any such contour γ we have

Z X
f (z)dz = 2πi Resz (f )
γ z∈Uγ ∩F

where Uγ is the set contoured by γ

Before we give more theoretical applications of the Residue Theorem 3.6, let us give
some applications to the evaluation of real integrals.

Example 3.9 (Integrals of rational functions). E.g.

Z ∞
dx
=π
−∞ 1 + x2

This of course can be evaluated easily using arctangent. We though give another proof
of it, using the Residue Theorem 3.6.

Idea: To choose a function f and a closed contour, so that part of the contour leads to
the real integral after taking limits.

1
In this particular case we take f (z) = 1+z 2
as function and γR as the contour path.
3.1. ZEROES AND POLES 101

f has only one pole, at z = i inside γR , hence

Z
1 1
f (z)dz = 2πi Resi (f ) = 2πi lim(z − 1) 2
= 2πi lim =π
γR z→i 1+z z→i z + i

Then we also have that

Z Z R Z
1 1
f (z)dz = dx + dz
γR −R 1 + x2 ΓR 1 + z2

for ΓR as the semicircular part of the closed path. As R → ∞, the first integral gives
Z ∞
1
2
dx
−∞ 1 + x

and similarly, as R → ∞, we see that over the semicircle ΓR the integral goes to zero.
This is because on ΓR : |z 2 + 1| > R2 − 1 and hence z21+1 < R21−1 ≈ R12 . Therefore, we
have that Z
1 1 π R→∞
2
dz < 2 πR ≈ −−−→ 0
ΓR 1 + z R −1 R
Hence finally we conclude that
Z ∞
1
dx = π
−∞ 1 + x2

Example 3.10. The same technique works to calculate the integrals of the form
Z ∞
P (x)
dx
−∞ Q(x)

where P, Q ∈ C[z], Q has no zero in the real axis and deg(Q) ≥ deg(P ) + 2

Note that we need this bound for the degrees of P and Q in order to get
Z
P (z) R→∞
dz −−−→ 0
ΓR Q(z)
102 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

If deg(Q) = n and deg(P ) = m on the semicircle, for R sufficiently large Q satisfies

the inequality |Q(z)| > K|z|n for some K ∈ R≥0 and we can hence bound
P (z) Rm C
< C n = n−m
Q(z) R R
Hence, it holds that Z
P (z) C C
dz ≤ n−m R = n−m−1
ΓR Q(z) R R
For this reason, in order for the control bound to go to zero, we need
n − m − 1 > 0 ⇐⇒ n > m + 1 ⇐⇒ n ≥ m + 2
i.e.

namely
deg(Q) ≥ deg(P ) + 2
Once we get this result, we can proceed with the calculation of the initial integral: we
remember that Z Z R Z
P (z) P (x) P (z)
dz = dx + dz
γR Q(z) −R Q(x) ΓR Q(z)
gives as R → ∞ Z  
P (z) X P
dz = 2πi Resz̃
γR Q(z) Q
z̃∈F ∩int(γR )
P
for F the set of all poles of Q
(therefore we then consider only the ones inside γR ).
Example 3.11. Z ∞
1 π
dx = 3
−∞ (x2 2
+a )2 2a

1
Consider the function f (z) = (z2 +a 2 )2 with poles at ±ai of order 2 each. Without loss

of generality, we assume that a > 0, since this would only switch the poles between
themselves in a manner that would way in the same result as the following.
   
1 d 2

d 1
Resai = lim (z − ai) f (z) = lim =
(z 2 + a2 )2 z→ai dz z→ai dz (z + ai)2
−2 −2 −2 −i
= lim = = =
z→ai (z + ai)3 (2ai)3 8a3 i 4a3
3.1. ZEROES AND POLES 103

As above we conclude Z
1
lim dz = 0
R→∞ ΓR (z 2 + a2 )2
and we get Z ∞
1 π
dx = 3
−∞ (x2 2
+a )2 2a
Example 3.12. The same contour can be used to evaluate integrals of rational functions
times sin(ax), cos(ax), for a ∈ R, i.e. of the form
Z ∞
P (x)
cos(ax)dx
−∞ Q(x)

where P, Q are polynomials in R[x] with deg(Q) ≥ deg(P ) + 2

P (z) iaz P (z)
Take the function f (z) = Q(z) e and not Q(z) cos(az), since cos(az) behaves badly on
the upper half plane. On the imaginary axis for example
et + e−t e2t + 1
cos(it) = =
2 2et
is the hyperbolic cosine, which grows exponentially. Whereas |eiz | = |ei(x+iy) | = e−y ,
which is bounded by 1 in the upper half plane. So, for Im(z) > 0 : |eiz | ≤ 1

E.g. for a > 0 we have Z ∞

1
cos(ax)dx = πe−a
−∞ x2 +1
eiaz
with f (z) = z 2 +1
, which has only one pole on the upper half plane at z = i, so

e−a
 iaz 
e eiaz
Resi = lim =
z2 + 1 z→i z + i 2i
Hence
e−a
Z
f (z)dz = 2πi = πe−a
γR 2i
Since |eiaz | < 1 on the upper half plane, we have that
eiaz 1
2
≤ 2
z +1 R −1
and hence that
eiaz
Z
πR R→∞
2
dz ≤ 2 −−−→ 0
Γr z +1 R −1
and consequently
∞
eiax
Z
∀a > 0 : dx = πe−a
−∞ x2 + 1
104 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA
 
cos(ax) eiax
we now note that x2 +1
= Re x2 +1
, hence by taking the real part of the function we
get Z ∞
cos(ax)
dx = πe−a
−∞ x2 + 1
This also shows that Z ∞
sin(ax)
dx = 0
−∞ x2 + 1
sin(ax)
which can also be directly seen, as x2 +1
is an odd function.
Example 3.13 (Integrals of trigonometric functions). The Residue Theorem 3.6 can
be used to evaluate real integrals of the form
Z 2π

P cos(t), sin(t)

dt
0 Q cos(t), sin(t)
where P, Q are polynomials and where it holds that ∀x, y ∈ R : (x2 + y 2 = 1) =⇒
Q(x, y) ̸= 0 (every pair (x, y) on C1 (0) has Q(x, y) ̸= 0.)

Consider the example of Z 2π

1
∀a > 1 : dθ
0 a + cos(θ)
The idea is to convert it to a contour integral around the unit circle. Expressing cos(θ) =
Re(z) leads to an effective approach: the trigonometric function cos(θ), sin(θ) can also
be written in terms of z on the unit circle C1 (0) as follows
1
eiθ + e−iθ z+ z
cos(θ) = =
2 2
and
eiθ − e−iθ z − z1
sin(θ) = =
2i 2i
z+ 1
Hence, we can write a + cos(θ) = a + 2 z and
Z 2π Z Z
1 1 dz 2 1
dθ = 1 = 2
dz
0 a + cos(θ) C1 (0) a +
z+ z iz i C1 (0) z + 2az + 1
2
1
where the term arises naturally from the construction of the integral. The poles of
iz √
−a ± a2 − 1, but only one of these roots is inside the unit circle,
the integral are at √
namely z0 = −a + a2 − 1
 
1 1 1 1 1
Resz0 2
= lim = lim = = √
z + 2az + 1 z→z0 (z − z1 )(z − z0 ) z→z0 2z + 2a 2z0 + 2a 2 a2 − 1
Therefore, we finally have
Z 2π
1 2 1 2π
dθ = 2πi √ =√
0 a + cos(θ) i 2
2 a −1 a2 − 1
3.1. ZEROES AND POLES 105

We now turn to more theoretical applications of the Residue Theorem 3.6. We start
by giving one more description of an isolated singularity, which is a pole. Namely, we
have the following Proposition.

Proposition 3.1. [SS10, Corollary III.3.2] Suppose f ∈ H Ω \ {z0 } with Ω open

has an isolated singularity at the point z0 , then

z0 is a pole of f ⇐⇒ lim f (z) = ∞

z→z0

Proof. If f has a pole of order k ∈ N∗ at z0 , then by Theorem 3.3 we have that there
exists some r >
0 for which f (z) = (z − z0 )−k h(z) on Ḋr (z0 ) with a bounded function
h ∈ H Dr (z0 ) for which it holds that h(z0 ) ̸= 0. Then
z→z
f (z) = (z − z0 )−k h(z) −−−→
0
∞
z→z
0
since h(z) −−−→ h(z0 ) ̸= 0 and k ≥ 1
z→z
0
Conversely, if f (z) −−−→ ∞, then we can find some r > 0 such that f (z) ≥ 1 (or any

0 ). In particular, f (z) ̸= 0 on Ḋr (z0 )
ε > 0 would also be suited for the case) on Ḋr (z
by Theorem 3.3 and there is a h ∈ H Ω \ {z0 } such that
1
h(z) :=
f (z)

is holomorphic in Ḋr (z0 ) and h(z) ≤ 1 there. Furthermore, it holds that limz→z0 h(z) =
1
limz→z0 f (z) = 0. By the Riemann’s Theorem on removable singularities 3.2, h extends
1
to a holomorphic function h̃ in Dr (z0 ) by defining h̃(z0 ) := limz→z0 f (z)
= 0 and other-
wise h̃ = h|Ḋr (z0 )
1
Therefore, if N ∈ N is the order of zero of h̃ at z0 , then f (z) = h(z)
has a pole of order
N ∈ N at z0
We have seen that an isolated singularity z0 of f is removable if f is bounded near z0 ,
z→z0
at the same time z0 is a pole if f (z) −−−→ ∞

Definition 3.5 (Essential singularity). An isolated singularity that is neither remov-

able nor a pole is called an essential singularity.

1
As we saw in the very beginning, the function e z has a more exotic behaviour. E.g.
1 1
e x → 0 as x ↗ 0 from the negative reals, whereas e x → ∞ as x ↘ 0 from the positive
reals.
106 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

In fact, any function f ∈ CΩ behaves exotically near an essential singularity. More

precisely we have

Theorem 3.7 (Casorati-Weierstrass). Suppose f ∈ H Ḋr (z0 ) and has an essen-

tial singularity at z0

Then the image of Ḋr (z0 ) under f , namely f Ḋr (z0 ) , is dense in C

Remark 3.7. The Casorati-Weierstrass Theorem 3.7 states that the image of a punc-
tured disc Ḋr (z0 ), no matter how small, effectively fills up the whole complex plane
(where z0 is an essential singularity). In fact, a remarkable Theorem of Picard says

Theorem 3.8 (Picard’s Theorem (1879)). If f ∈ H Ḋr (z0 ) and has an essential




singularity at z0 , then C \ f Ḋr (z0 ) contains at most one point.

1
The function f (z) = e z maps each punctured disc centred at z = 0 to C∗ , i.e. it does
not take the value 0, so the “exceptional value” permitted by Picard’s Theorem 3.8
may in fact exist.
Proof of Theorem 3.7. We want to show that for the given r > 0

∀ε > 0∀w ∈ C∃z ∈ Ḋr (z0 ) : f (z) − w < ε

and to do so we argue by contradiction and show that this will forces the singularity at z0
to be either removable or a pole, hence contradicting the assumption that z0 is essential.

Assume on the contrary that the image in question is not dense in C, hence for the
given r > 0
∃δ > 0∃w0 ∈ C∀z ∈ Ḋr (z0 ) : f (z) − w0 ≥ δ
1
Let g : Ḋr (z0 ) → C, z 7→ g(z) := f (z)−w0
, then on Ḋr (z0 ) we have that g is bounded
1
by δ , hence has a removable singularity at z0 by Riemann’s Theorem on removable
singularities 3.2. Hence, there is a holomorphic extension of g, i.e. we can define g at
z0 , so that g becomes holomorphic in Dr (z0 )

1
Since f (z) − w0 ≥ δ and g(w) = f (z)−w 0
, clearly g has no zero in Ḋr (z0 ), hence its
1
reciprocal g has an isolated singularity at z0 ∈ Dr (z0 ). This singularity is either a
removable one or a pole, depending on whether limz→z0 g(z) = 0 or not, respectively.
In turn, this gives that the singularity of f = w0 + g1 at z0 can be at most a pole, giving
the anticipated contradiction. Note that the limit limz→z0 g(z) exists, since g has a
removable singularity at z0
3.2. MEROMORPHIC FUNCTIONS 107

3.2 Meromorphic functions

We now look at functions whose singularities are poles. Since at a pole limz→z0 f (z) =
∞, this suggest that we can add ∞ to the values of functions, including consequently
the poles in their domain of definition. E.g. The function

f : C∗ → C
1
z 7→
z
can be extended to

fˆ : C → C ∪ {∞}
1
z 7→
z

Definition 3.6. • The set

Ĉ := C ∪ {∞}
is called the Extended Complex Plane. Here ∞ represents a point “at
infinity” and is unsigned. In Ĉ, we can supplement the rules in C by

∀z ∈ C : ∞ ± z = z±∞ = ∞
∀z ∈ Ĉ \ {0} : ∞ · z = z·∞ = ∞
z
∀z ∈ C : ∞
= 0
z
∀z ∈ Ĉ \ {0} : 0
= ∞

The expressions ∞ ± ∞, ∞ , 0 and 0 · ∞ are not assigned a meaning in Ĉ

∞ 0
∗
• A sequence (zn )n∈N∗ ∈ CN converges to ∞, i.e. limn→∞ zn = ∞, if

lim |zn | = ∞
n→∞

∗
where |zn | n∈N∗ ∈ RN . Similarly, we say that limz→z0 f (zn ) = ∞, if

lim f (zn ) = ∞
z→z0

Remark 3.8. Ĉ is not a field.

Talking now about functions, we can extend our definition of holomorphic function to
a more general notion.
108 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Definition 3.7 (Meromorphic function). A function f ∈ ĈΩ with Ω ⊆ C open in C

is called meromorphic, if the following conditions are satisfied:

• The set Sf := {z ∈ Ω : f (z) = ∞} = f −1 {∞} has no limit point in Ω, i.e. Sf

is discrete in Ω

• The points in Sf are poles of f

• The restriction of f
to Ω \ Sf = {z ∈ Ω : f (z) ̸= ∞} is holomorphic, i.e.

f |Ω\Sf ∈ H Ω \ Sf

Let M (Ω) denote the set of all meromorphic functions in Ω

Note that the set of poles for meromorphic functions is discrete, as the set of zeroes for
holomorphic functions.

Example 3.14. 1. Let P, Q ∈ C[z] be two polynomials with no common zeroes.

Note that any rational function pq (for p, q rational functions) can be reduced to
P
Q
with no common zeroes. Let
(
P (z)
, Q(z) ̸= 0
f : C → C, z 7→ f (z) := Q(z)
∞ , Q(z) = 0

Then f ∈ M (C), since f is holomorphic outside the finite zeroes’ set of Q(z). If
z0 is a zero of Q, then f has a pole at z0 , as

P (z)
lim f (z) = lim =∞
z→z0 z→z0 Q(z)

having we assumed that P (z0 ) ̸= 0. Hence, z0 is a pole of f

cos(πz)
2. The function f (z) = cot(πz) = sin(πz)
is a meromorphic function in C with Sf = Z
1
3. Let f (z) = z2e−1
z
, then f is meromorphic in C∗ with Sf = {±1}, but not in C, as
z = 0 is an essential singularity.

If we have two functions f, g ∈ M (Ω) with pole sets Sf and Sg , then f +g is holomorphic
in Ω \ (Sf ∪ Sg ) and we can define f + g at points in this set using the usual definition:

∀z ∈ Ω \ (Sf ∪ Sg ) : (f + g)(z) = f (z) + g(z)

So we only have to worry about points z ∈ Sf ∪ Sg and by doing so we can extend

(f + g)|Ω\(Sf ∪Sg ) : Ω \ (Sf ∪ Sg ) → C to a meromorphic function f + g : Ω → Ĉ. We
3.2. MEROMORPHIC FUNCTIONS 109

can do this in the following manner:

If z0 ∈ Sf ∪ Sg , write for all z ∈ Ḋr (z0 ) for a r > 0 as

f (z) = Pzf0 (z) + f˜(z)

g(z) = Pzg0 (z) + g̃(z)

where Pzf0 , Pzg0 are the principal parts of f and g at z0 (one of them can be zero if f or
g does not have a pole at z0 ) and where f˜, g̃ ∈ H Dr (z0 ) . Then

(f + g)(z) = Pzf0 (z) + Pzg0 (z) + f˜(z) + g̃(z)

| {z } | {z }
linear combination of terms of the form 1 holomorphic in Dr (z0 )
(z−z0 )l

so f + g has a pole of order ≥ 1 at z0 and we assign the value ∞ to that point under
f + g, this unless Pzf0 (z) + Pzg0 (z) = 0 (which can happen). Hence, f + g ∈ M (Ω) with
Sf +g ⊆ Sf ∪ Sg

We have proved part (ii) of the following Proposition:

Proposition 3.2. Let Ω ⊆ C be open, then

(i) H (Ω) ⊆ M (Ω)

(ii) If f, g ∈ M (Ω), then ∀a, b ∈ C : af + bg ∈ M (Ω). Hence, M (Ω) is a C-vector

space.

(iii) If f, g ∈ M (Ω), then f g ∈ M (Ω)

(iv) If 0 ̸= f ∈ M (Ω) and the zeroes of f do not have a limit point in Ω, then
1
f
∈ M (Ω)

Proof. (i) Obvious, but note that we identified a holomorphic function f ∈ H (Ω)
with the corresponding function f˜ ∈ M (Ω), where f˜ = i ◦ f and i : C ,→ Ĉ

(ii) The same argument for f + g works with af + bg

(iii) Let f = Pzf0 + f˜ and g = Pzg0 + g̃ with z0 ∈ Sf ∪ Sg , then

f g = (Pzf0 + f˜)(Pzg0 + g̃) = Pzf0g + G

l terms and G ∈ H
1


where Pzf0g is a linear combination of (z−z 0 )
Dr (z0 ) for some
r > 0 by Theorem 3.4 (G is the sum of the respective functions in the case of f
110 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

and g separately), so
∞
! ∞
!
X X
fg = ak (z − z0 )k bl (z − z0 )l =
k=−n l=−m
 
∞
X  X
= ak bN −l  (z − z0 )N


N =−(n+m) k,l
k+l=N

a−1 b−2 b−1 P∞

an (z −z0 )n and g(z) =
P
E.g. if f (z) = f = z−z0
+ n∈N∗ (z−z0 )2
+ z−z 0
+ l=0 bl (z −
z0 )l , then

b−2 a−1 b−2 a0 + a−1 b−1 a−1 b0 + b−1 a0 + b−2 a1

fg = 3 + + +G
(z − z0 ) (z − z0 )2 (z − z0 )

where G is holomorphic in some disc around z0 . Hence, we have that f g has a

pole of order 3

Similarly to the case of f + g, we can define


f (z)g(z) , z ∈ Ω \ (Sf ∪ Sg )
(f g)(z) =
∞ , z ∈ Sf ∪ Sg

Then f g is meromorphic, i.e. f g ∈ M (Ω) with Sf g ⊆ Sf ∪ Sg

(iv) Let f ∈ M (Ω), if z0 ∈ Ω \ Sf and if f (z0 ) ̸= 0, then f1 is holomorphic at z0 , if

z0 ∈ Ω \ Sf and f (z0 ) = 0, then f1 has a pole of order k (equal to the order of
zeroes of f at z0 ) at z0 (of order ≥ 1). If z0 ∈ Sf , then

1 z→z0
−−−→ 0
f (z)

hence f1 is a removable singularity at z0 . So, if the zero of f has no limit point in

Ω, then the poles of f1 have no limit point in Ω and hence f1 ∈ M (Ω)

Remark 3.9. If we assure that f ̸= 0 in an connected component of Ω, then 1

f
∈ M (Ω)

Recall: If f ∈ CΩ , Ω open and connected and f ∈ H (Ω), then the zeroes of f do not
have a limit point in Ω (see Theorem 2.10).

For an open and connected set Ω, the same is true for f ∈ M (Ω)
3.2. MEROMORPHIC FUNCTIONS 111

Proposition 3.3. Let Ω be open and connected, 0 ̸= f ∈ M (Ω) and Zf := {z ∈

Ω : f (z) = 0}, then Zf has no limit point in Ω

∗
Proof. Assume on the contrary that ∃(zn )n∈N∗ ∈ ZfN of distinct points such that
limn→∞ zn = b ∈ Ω

Let Sf be the set of poles of f (recall that it is countable), then f |Ω\Sf ∈ H (Ω \ Sf )

and Ω \ Sf is open, connected and f ̸= 0. Hence, by the above result we have recalled
b∈/ Ω \ Sf

But b ∈
/ Sf either, since if b were a pole of f , then limz→b f (z) = ∞ and it would
mean that for ε > 0 we would have f (z) > 0 for z ∈ Ω with |z − b| < ε. But this is
impossible, since if zn → b, then |zn − b| < ε for n ≥ n0 and f (zn ) = 0
Remark 3.10. Let f ∈ M (Ω) and z0 a pole of f , since Sf has no limit point in Ω,
then it exists a punctured neighbourhood Ḋr (z0 ) of z0 for some r > 0 such that
Ḋr (z0 ) ∩ Sf = ∅
If the order of the pole of
f at z0 is k, then f (z) = (z − z0 )−k g(z) with an analytic
function g(z) ∈ H Dr (z0 ) by Theorem 3.3.

Hence, locally every meromorphic function is the quotient of two holomorphic functions.
Here:
g(z)
f (z) =
(z − z0 )k
It is a non-trivial result that if Ω is non-empty, open and connected (i.e. a region), then
we can do this globally.

For any f ∈ M (Ω) with Ω open and connected, there exist g, h ∈ H (Ω) such that f = g
h

Algebraically, we can state this as follows: recall that if Ω is open and connected, then
H (Ω) has no zero divisor, hence it is an integral domain. Consequentially, it has a
quotient field (or field of fractions):

ng o
Q H (Ω) = : g, h ∈ H (Ω) and h ̸= 0 = M (Ω)
h
This is similar to the construction of Q as field of fractions of the integral domain Z
112 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Definition 3.8. Let Ω ⊆ C open, z0 ∈ Ω and 0 ̸= f ∈ M (Ω). Then define the

valuation (or order) of f at z0 , denoted by ordz0 (f ) or νz0 (f ) to be the integer
k ∈ Z such that:

• If z0 is not a pole of f , i.e. f (z0 ) ̸= ∞, then k ≥ 0 is the order of vanishing

(zero) of f at z0

• If f (z0 ) = ∞, i.e. z0 is a pole, then k ≤ −1 is minus the order of the pole at z0

In particular,

• If ordz0 (f ) > 0, then z0 is a zero.

• If ordz0 (f ) < 0, then z0 is a pole.

• If ordz0 (f ) = 0, then f (z0 ) ̸= 0 and f (z0 ) ̸= ∞ (neither a pole nor a zero).

Combining what we know about the behaviour of functions near zeroes and poles (The-
orem 2.2 and Theorem 3.3), we get:

Proposition 3.4. If 0 ̸= f ∈ M (Ω) and z0 ∈ Ω, then

(i) ordz0 (f ) = k ∈ Z⇐⇒ ∃r > 0∃h ∈ H Dr (z0 ) :


(h(z0 ) ̸= 0) and ∀z ∈ Ḋr (z0 ) : f (z) = (z − z0 )k h(z)
Here we have k < 0 if z0 is a pole, k > 0, if z0 a zero.

(ii) ordz0 (f g) = ordz0 (f ) + ordz0 (g)


(iii) If f + g ̸= 0, then ordz0 (f + g) ≥ min ordz0 (f ), ordz0 (g)

z 2 2
Example 3.15. Let f (z) = (ez −1) 2 , z has zero of order 1 at z = 0, while (e − 1) has

zeroes of order 2 at z = 2πin, for n ∈ Z


2 
ordz0 (f ) = ord0 (z) − ord0 e2 − 1 = 1 − 2 = −1

Hence, f has a pole of order 1 at z = 0. For n ̸= 0 we have


2 
ord2πin (f ) = ord2πin (z) − ord2πin e2 − 1 = 0 − 2 = −2

Therefore, f has a pole of order 2 at 2πin, for n ∈ Z \ {0}

Remark 3.11 (Ĉ and the stereographic Projection). Let

S 2 := (x1 , x2 , x3 ) ∈ R3 : x21 + x22 + x23 = 1 ⊂ R3


3.2. MEROMORPHIC FUNCTIONS 113

Identifying (x1 , x2 , 0) with C we can think of C sitting in R3 as the (x1 , x2 )-plane.

Set N = (0, 0, 1) and define the map
π : S 2 \ {N } → C
as the map that takes p ∈ S 2 \ {N } and maps it to π(p), which is the intersection of C
with the ray in R3 that starts at N and passes through p

π is called the stereographic projection of S 2 \ {N } into C

Explicitly, π is given by
 
x1 x2 x1 x2
π(p) = π(x1 , x2 , x2 ) = , ,0 = + i
1 − x3 1 − x3 1 − x3 1 − x3
Note that the equation of the ray that starts at N and goes through p is:
N + t(p − N ), t≥0
and consequently
π(p) = N + t0 (p − N )
where t0 is unique positive real number, so that (0, 0, 1) + t0 (x1 , x2 , x3 − 1) has third
coordinate equal to 0. Solving this equation for t0 gives the formula for π(p) above.
Defining π(N ) = ∞ gives a bijection
π : S 2 → Ĉ
Conversely, given z ∈ C one checks that
|z|2 − 1
 
−1 2x 2y
π (z) = 2
, 2 , 2 ∈ S 2 \ {N }
|z| + 1 |z| + 1 |z| + 1
and π −1 (∞) := N gives the inverse map. Here we get S 2 homeomorphic to Ĉ, since
both maps are continuous.
114 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Before we study the values of holomorphic functions using the Residue Formula 3.6, let
us mention that we can also talk about meromorphic functions on Ĉ (as opposed to
M (Ω) with Ω ⊆ C).

We have already allowed ∞ as a value of meromorphic functions. We can also allow ∞

in the definition domain and study functions f ∈ ĈΩ̃ , where Ω̃ ⊆ Ĉ

1
If a function f is analytic for large values of z, i.e. |z| > R
for some R > 0, then the
function  
1
g(z) := f
z
is holomorphic in a deleted neighbourhood of 0, i.e. in ḊR (0)

Definition 3.9 (Deleted neighbourhood at ∞). We define the delete neighbour-

hood at ∞ as
ḊR (∞) := z ∈ C : |z| > R−1


This notation is designed to have for R < S that

ḊR (∞) ⊂ ḊS (∞)

Definition 3.10. • For a function f , which is analytic for |z| > R1 for some
R > 0, we say that f has an isolated singularity
1

at ∞, which will be called
removable, a pole or essential, if g(z) = f z has an isolated singularity at
0 (which is removable, a pole or essential respectively).

• A meromorphic function in the complex plane that is either holomorphic at ∞

or has a pole at ∞ is called meromorphic in Ĉ
3.3. APPLICATIONS OF THE RESIDUE THEOREM 115

Example 3.16. 1. An entire function is analytic in ḊR (∞) for every R > 0
E.g. the function f (z) = ez has an isolated singularly at ∞, which is essential,
1
because e z has an essential singularity at 0. Hence, ez is not meromorphic in Ĉ,
but it is meromorphic on C
Pn k
2. The function p(z) ∈ C[z] has
Pn ak a pole at ∞. If p(z) = a 0 + k=1 ak z for some
1


n ∈ N, then p z = a0 + k=1 zk has a pole of order n at 0
3. The function f (z) = tan(z) does not have an isolated singularity at ∞: Each
ḊR (∞) includes poles of f with z = π2 + kπ, k ∈ Z

n
−1 o
1 π


Also note that g(z) = tan z
has singularities S := 2
+ kπ : k ∈ Z , which
1


accumulate at z = 0. Hence, the singularity of tan z
at z = 0 is not isolated.
The following Theorem for meromorphic functions on Ĉ

Theorem 3.9. [SS10, Theorem III.3.4] If f ∈ M (Ĉ), then it is a rational function.

Clearly, each rational function is a meromorphic function on Ĉ. Hence, we have
that  
P
M (Ĉ) = : P, Q ∈ C[z]
Q

Proof. Exercise.

3.3 Applications of the Residue Theorem

The fist application is called the Argument Principle: it uses the Residue Theorem 3.6
′
applied to ff , the logarithmic derivative of f , to count the zeroes and the poles of f
inside a curve.
To this end we first note the following simple Lemma:
116 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

f′
Lemma 3.2. Let Ω ⊆ C be open and connected, also 0 ̸= f ∈ M (Ω), then f
∈
′
M (Ω), the logarithmic derivative of f , is also meromorphic in Ω. Moreover, ff
has poles of order 1 at all z0 ∈ Ω, for which ordz0 (f ) ̸= 0, i.e. either z0 is a zero or a
pole of f  ′
f
∀z0 ∈ Zf ∪ Sf : Resz0 = ordz0 (f )
f

Proof. Since f ̸= 0, Ω open and connected, being f ∈ M (Ω), the zeroes of f do not
have a limit point in Ω, and f1 ∈ M (Ω)

Clearly, f ′ ∈ M (Ω \ Sf ), where Sf is the set of poles of f . If z0 ∈ Sf is a pole of order

n ∈ N∗ of f , then ∃r > 0 by Theorem 3.3 such that

∀z ∈ Ḋr (z0 ) : f (z) = (z − z0 )−n h(z)

where h ∈ H Dr (z0 ) and h(z0 ) ̸= 0. Then, for z ∈ Ḋr (z0 ) we have

−n h′ (z)
f ′ (z) = h(z) + =
(z − z0 )n+1 (z − z0 )n
1
= h′ (z)(z − z0 ) − nh(z)


| {z } (z − z0 )n+1
=:h̃(z)

where h̃ ∈ H Dr (z0 ) and h̃(z0 ) = −nh(z0 ) ̸= 0 by definition. Hence, for all z ∈ Ḋr (z0 )

f ′ (z) = (z − z0 )−(n+1) h̃(z)

Hence, f ′ has a pole of order n + 1 at z0 (Similarly, if f has a zero of order n at z0 ,

′
then f ′ has a zero of order n − 1 at z0 ). Hence, f ′ ∈ M (Ω) and so is ff ∈ M (Ω)

So, for any z ∈ Ω we have



f′
 −(n + 1) − (−n) = −1 , ordz0 (f ) = −n

′
ordz0 = ordz0 (f ) − ordz0 (f ) = (n − 1) − n = −1 , ordz0 (f ) = n
f 
≥0 , otherwise


f′
Hence, f
has a pole of order 1 at the points where ordz0 (f ) ̸= 0

We can also calculate the residue using

∀z ∈ Ḋr (z0 ) : f (z) = (z − z0 )n g(z)

3.3. APPLICATIONS OF THE RESIDUE THEOREM 117

where g ∈ H Dr (z0 ) , g(z) ̸= 0 for all z ∈ Dr (z0 ) and n = ordz0 (f ). Hence, we have:

n > 0 if z0 is a zero and n < 0 if z0 is a pole of f

Finally, for z ∈ Dr (z0 ) we have

f ′ (z) = n(z − z0 )n−1 g(z) + (z − z0 )n g ′ (z)

and
f ′ (z) n(z − z0 )n−1 g(z) + (z − z0 )n g ′ (z)
∀z ∈ Dr (z0 ) : = =
f (z) (z − z0 )n g(z)
n(z − z0 )n−1 g(z) (z − z0 )n g ′ (z) n g ′ (z)
= n
+ n
= +
(z − z0 ) g(z) (z − z0 ) g(z) (z − z0 ) g(z)
| {z }

=G(z)∈H Dr (z0 )

Hence
f′
 
Resz0 = n = ordz0 (f )
f

Lemma 3.2 immediately gives, using the Residue Theorem 3.6, the following result:

Theorem 3.10 (Argument Principle). [SS10, Theorem III.4.1] Let Ω ⊆ C open and
connected, f ∈ M (Ω) and let im(γ) ⊆ Ω be a circle (or any other curve such that
the Residue Formula 3.6 holds). If f has no zeros or poles on γ, then
 
Z ′
1 f (z) X X X
dz =  ordz0 (f ) + ordz0 (f ) = ordz0 (f )
2πi γ f (z)
z0 ∈Zf ∩int(γ) z0 ∈Sf ∩int(γ) z0 ∈int(γ)
ordz0 (f )̸=0

where Zf is the set of zeroes of f and Sf is the set of poles of f

 ′
f
Proof. This follows from the previous Lemma 3.2 and Resz0 f
= ordz0 (f )

f ′ (z) f′
Z  
1 X
dz = Reszk = Z −P
2π γ f (z) f
zk ∈int(γ)

where Z is the number of zeroes of f inside γ counted with multiplicity (to express the
order of zero) and P is the number of poles of f inside γ counted with multiplicity (to
express the order of pole).
118 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Corollary 3.1. Let f ∈ C[z] be a polynomial. Choose R > 0 large enough, so that
all zeroes of f are inside DR (0), then

f ′ (z)
Z
dz = deg(f )
CR (0) f (z)

We have the following Corollary of the Argument Principle Theorem 3.10, which says
that a holomorphic function, when perturbed slightly, does not change its number of
zeroes.

Theorem 3.11 (Rouché’s Theorem). [SS10, Theorem III.4.3] Suppose f, g ∈ H (Ω)

for an open set Ω ⊆ C, which contains a circle C and its interior. If

∀z ∈ C : f (z) > g(z)

Then f and f + g have the same number of zeroes inside of C

Proof. For t ∈ [0, 1], define ft (z) = f (z) + tg(z) so that f0 (z) = f (z) and f1 (z) =
(f + g)(z). Note that for z ∈ C it results that

ft (z) = f (z) + tg(z) ≥ f (z) − t g(z) ≥ f (z) − t g(z) >

> g(z) − t g(z) = (1 − t) g(z) ≥ 0

given the assumption. Hence, we have that ∀z ∈ C : ft (z) > 0 and therefore ft has
no zero in C
Note that if we can show that
Z ′
1 ft (z)
nt := dz
2πi C ft (z)
is a constant, this will show that f and f +g have the same number of zeroes inside of C

By Argument Principle 3.10 applied to ft (z) (which by the above consideration has no
zero in C), we have that Z ′
1 ft (z)
nt = dz
2πi C ft (z)
counts the number of zeroes of ft inside of C, in particular it is integer valued.
f ′ (z)
Note that nt is a continuous function of t, since ftt (z) is jointly continuous on [0, 1] for
all z ∈ C, as both ft′ (z) and ft (z) are continuous and ft (z) ̸= 0 for all z ∈ C
3.3. APPLICATIONS OF THE RESIDUE THEOREM 119

Recall: from Real Analysis we know that, if

h : [a, b] × [c, d] → R
Rd
is continuous on [a, b] × [c, d], then F (t) := c h(t, x)dx is continuous on [a, b].
Using this gives that Z ′
1 ft (z)
dz
2πi C ft (z)
is continuous and since nt is also integer valued, it must be a constant (otherwise the
Intermediate Value Theorem [EW22] gives the existence of t0 ∈ [0, 1] such that nt0 is
not integral). Hence, n0 is the number of zeroes inside of f and n1 is the number of
zeroes inside of f + g, so finally
Z ′
(f + g)′ (z)
Z
1 f (z) 1
dz = n0 = n1 = dz
2πi C f (z) 2πi C (f + g)(z)

Example 3.17. We use Rouché’s Theorem 3.11 to show that the polynomial

p(z) = z 6 + 8z 4 + z 3 + 2z + 3

has four zeroes inside of the unit circle C1 (0)

The idea is to write p = Big + Small on C1 (0), such that Big(z) = 8z 4 = f (z) and
Small(z) = z 6 + z 3 + 2z + 3 = g(z), so: g(z) = |z 6 + z 3 + 2z + 3| < |8z 4 | = f (z) on
C1 (0), so with |z| = 1

|z 6 + z 3 + 2z + 3| ≤ |z|6 + |z|3 + 2|z| + 3 = 1 + 1 + 2 + 3 = 7 < 8 = 8|z|4

Hence, by Rouché’s Theorem 3.11, f (z) = 8z 4 and (f + g)(z) = p(z) have the same
number of zeroes inside the unit circle. f has four zeroes (counted with multiplicity)
and so does p
Example 3.18. Rouché’s Theorem 3.11 can also be used to give a “nice4 ” or simple
proof of the Fundamental
P Theorem of Algebra.
d d−1 k
Let p(z) = z + k=1 ak z + a0 for which, if |z| is large enough, the term |z|d
Pd−1
dominates. Choose R large enough, so that for f (z) = z d and g(z) = k=1 ak z k + a0
we have:
f (z) > g(z)
on CR (0) and hence p = f + g and f have the same number of zeroes inside CR (0). We
obtain that f has d zeroes inside CR (0) and so does p
Rouché’s Theorem also leads us to two other important Theorems:

4
(subjectively) “nice” means “objectively nice”, if said from the Professor.
120 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Theorem 3.12 (Open mapping Theorem). [SS10, Theorem III.4.4] Let Ω ⊆ C be

an open and connected set. Let f ∈ H (Ω) such that f is non-constant, then f is an
open map, i.e. sends open sets to open sets in the standard topology of C, namely
form OCΩ to OC

Proof. Let z0 ∈ U ⊆ Ω with U ∈ OCΩ and f (z0 ) = w0 . We want to show that a

neighbourhood of w0 is also contained in f (U ), i.e. if w is near w0 , then show that
∃z ∈ U : w = f (z), i.e. w ∈ f (U ). If we interpret z as a zeroes of f (z) − w, we have
also
f (z) − w = f (z) − w0 + w0 − w
| {z } | {z }
f˜ g̃

Let r > 0 such that Dr (z0 ) ⊂ U and such that ∀z ∈ D˙ r (z0 ) : f (z) − w0 ̸= 0; this is
allowed, since the zeroes of f˜(z) := f (z) − w0 are isolated. In particular, we have that
f (z) − w0 ̸= 0 for z on the circle Cr (z0 )

Since Cr (z0 ) is compact and f (z) − w0 ̸= 0 on Cr (z0 ), we can find a δ > 0 such that

f (z) − w0 ≥ δ

for all z ∈ Cr (z0 ). Let now w ∈ Dδ (w0 ) and define

F : Ω → C, w 7→ F (z) := f (z) − w = f (z) − w0 + w0 − w

| {z } | {z }
=:f˜ =:g̃

We want to show that F has a zero inside the circle
Cr (z0 ). This will show that
∃z ∈ Dr (z0 ) : f (z) = w and hence that w ∈ f Dr (z0 )

We now apply Rouché’s Theorem 3.11 to f˜, g̃ on the circle Cr (z0 ), where we have that

f˜ ≥ δ and |g̃| < δ

3.3. APPLICATIONS OF THE RESIDUE THEOREM 121

Hence, on the circle Cr (z0 ), we have

|f˜| > |g̃|

and so f˜ and F̃ = f˜+ g̃ have the same number of zeroes inside Dr (z0 ). Since f˜ = f −w0

conclude that ∃z ∈ Dr (z0 ) : F (z) = 0, i.e.
has a zero inside Dr (z0 ), namely z0 , we must
∃z ∈ Dr (z0 ) : f (z) = w, so w ∈ f Dr (z0 ) as wanted.

Remark 3.12. This Theorem 3.12 states, for example, that if f ∈ H (Dr (z0 )) and non-
constant for some z0 ∈ C, then it is not possible to that f (z) ∈ R for all z ∈ Dr (z0 ),
since any subset of R is not open in C

Theorem 3.13 (Maximum modulus principle). [SS10, Theorem III.4.5][SS10, Corol-

lary III.4.6] Let Ω ⊆ C be open and connected and f ∈ H (Ω) non-constant, then

∄z0 ∈ Ω∀z ∈ Ω : f (z) ≤ f (z0 )

i.e. f cannot attain its maximum in Ω. In particular, if Ω is bounded and f ∈ C 0 Ω ,
then
max f (z) = max f (z)
z∈Ω z∈∂Ω

Example 3.19 (Necessity of the boundedness of Ω). The assumption that Ω is bounded,
hence compact, is crucial, as shown in this example.

Let Ω = z ∈ C : Im(z) ∈ − π2 , π2


be open in C and connected, it is not hard to see
that the set Ω is not bounded. Consider the function f (z) = exp (ez ) on Ω, then
π

f |∂Ω (z) = exp ex±i 2 = exp(±iex )

Hence, for z ∈ ∂Ω we have f (z) = 1, but

f (x) = exp(ex ) −−−−−→ ∞

R∋x→∞

which does not exists in R

122 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Proof of Theorem 3.13. We first note that ∃ maxz∈Ω f (z) , since Ω is a compact set
and f ∈ C 0 (Ω), as shown in [EW22].
Let f ∈ H (Ω) and non-constant. Suppose also on the contrary that f attains a
maximum at z0 ∈ Ω. By the Open mapping Theorem 3.12, if f is an open map, by
letting D = Dr (z0 ) ⊆ Ω, then f (D) is open in C and contains f (z0 )

Hence f (D) contains a disc D̃ around f (z0 ) and therefore there are points z ∈ D such
that
f (z0 ) < f (z)
which contradicts the assumption that |f | attains its maximum at z0 (in any disc in C
one can find such points).

If Ω is bounded and f is non-constant and continuous, then f (z) attains its maximum
on Ω, since it is a continuous function on a compact set, as shown in [EW22]. By the
first part of the Theorem, this point where f attains its maximum cannot be inside Ω.
Hence, it has to be on the boundary Ω \ Ω = ∂Ω

3.4 Homotopy and simply connected domains

The key to understand the general form of Cauchy’s Theorem 2.5 is the idea that if
f ∈ H (Ω) and if we “continuously deform” γ0 to γ1 , while staying in Ω and keeping
the endpoints fixed if the paths are not closed, then
Z Z
f (z)dz = f (z)dz
γ0 γ1

for the two paths γ0 : [a, b] → Ω and γ1 [a, b] → Ω, either closed or with fixed endpoints.
3.4. HOMOTOPY AND SIMPLY CONNECTED DOMAINS 123

Not closed curves of this type curves are called homotopic with fixed endpoints: this
means that for each s ∈ [0, 1] it exists a curve γs in Ω parameterised by γs (t), hence
γs : [a, b] → Ω such that

γs (a) = γ0 (a) = γ1 (a)

γs (b) = γ0 (b) = γ1 (b)

and such that at s = 0 we have γs (t)|s=0 = γ0 (t) and at s = 1 we have γs (t)|s=1 = γ1 (t).
All this should be done continuously.

Definition 3.11 (Homotopy). Let Ω ⊆ C be open.

• Let γ0 : [a, b] → Ω and γ1 : [a, b] → Ω be two curves such that γ0 (a) = γ1 (a)
and γ0 (b) = γ1 (b), i.e. they have the same endpoints.
We says that γ0 is homotopic to γ1 in Ω with fixed endpoints, and denote
it by γ0 ∼Ω γ1 rel ∂[a, b], if

∃H ∈ C 0 [a, b] × [0, 1]; Ω , (t, s) 7→ H(t, s) =: γs (t)

such that

– ∀t ∈ [a, b] : H(t, 0) = γ0 (t) and H(t, 1) = γ1 (t)

– ∀s ∈ [0, 1] : H(t, s) = γs (t) ∈ C 0 [a, b]; Ω is a piecewise smooth curve and
∀s ∈ [0, 1] : H(a, s) = γ0 (a) = γ1 (a) and H(b, s) = γ0 (b) = γ1 (b), i.e. γs (t)
has the same endpoints as γ0 , γ1

• Similarly, let γ0 : [a, b] → Ω and γ1 : [a, b] → Ω be two closed curves, we say

that γ0 is homotopic to γ1 in Ω, denoted as γ0 ∼Ω γ1 , if

∃H ∈ C 0 [a, b] × [0, 1]; Ω , (t, s) 7→ H(t, s) =: γs (t)

such that

– ∀t ∈ [a, b] : H(t, 0) = γ0 (t) and H(t, 1) = γ1 (t)

– ∀s ∈ [0, 1] : H(t, s) =: γs (t) ∈ C 0 [a, b]; Ω is a piecewise smooth curve in
Ω and ∀s ∈ [0, 1] : H(a, s) = H(b, s), i.e. γs (t) is a closed curve in Ω for
every s ∈ [0, 1]

Note that if clear enough, the notation for fixed endpoints, namely “rel ∂[a, b]” will here
be omitted.
Example 3.20. 1. If Ω = C, then any two closed curves γ0 , γ1 are homotopic. In
particular, every closed curve is homotopic to the constant curve σc : [a, b] →
C, t 7→ c for every c ∈ C
124 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Consider the function

H : [a, b] × [0, 1] → C
(t, s) 7→ (1 − s)γ0 (t) + sγ1 (t)

H is a combination of continuous functions, hence continuous. Moreover it holds

that:

H(t, 0) = γ0 (t)
H(t, 1) = γ1 (t)
H(a, s) = (1 − s)γ0 (a) + sγ1 (a)
H(b, s) = (1 − s)γ0 (b) + sγ1 (b)

Since γ0 (a) = γ0 (b), γ1 (a) = γ1 (b) and ∀s ∈ [0, 1] : H(a, s) = H(b, s). Hence
γs : [a, b] → C are all closed curves.

Note that geometrically H is defined using the line segment between γ0 (t) and γ1 (t)
for each fixed t ∈ [a, b]. Hence, s ∈ [0, 1] varies over the line segment between
γ0 (t) and γ1 (t) for each fixed t ∈ [a, b].

For the constant curve σc , we can take the homotopy between σc and γ as

H : [a, b] × [0, 1] → C
(t, s) 7→ (1 − s)c + sγ(t)

Note that the same definition we used for closed curves γ0 , γ1 also gives a homotopy
with fixed points, if γ0 : [a, b] → Ω and γ1 : [a, b] → Ω are two curves with fixed
endpoints, i.e. with γ0 (a) = γ1 (a) and γ0 (b) = γ1 (b)

2. Note that the same formula for the homotopy works for any convex Ω ⊆ C. I.e.
if we have two curves γ0 , γ1 either closed or with a fixed endpoints in a convex set
3.4. HOMOTOPY AND SIMPLY CONNECTED DOMAINS 125

Ω, then since for a convex set the line segment between any two points is also in
the set, the function defined by

H : [a, b] × [0, 1] → C
(t, s) 7→ (1 − s)γ0 (t) + sγ1 (t)

gives a homotopy in Ω.
In particular, this works for Ω with the form of a disc.

3. An example of two curves which are not homotopic in Ω: if we take Ω := C∗ and

the curves

γ0 : [0, π] → Ω
t 7→ eit

and

γ1 : [0, π] → Ω
t 7→ e−it

Then γ0 and γ1 are not homotopic in Ω

We will see a simple proof of this when we will see the homotopy version of
Cauchy’s Theorem 2.5. Intuitively, to deform γ0 to γ1 we have to go through 0,
which is not in Ω though.

4. The set Ω = C \ (−∞, 0] is not convex and therefore we cannot use the previ-
ous formula, but still any two closed curves γ0 , γ1 in Ω are homotopic, i.e. we
R deform γR0 to γ1 in the following way. Then for f ∈ H (Ω) we have that
can
γ0
f (z)dz = γ1 f (z)dz
126 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

The idea is to choose any point on the real line, say c ∈ R, and the constant curve

σc : [a, b] → Ω
t 7→ c

We first deform γ0 to c and then c to γ1 , so

c + (1 − 2s) γ0 (t) − c
, s ∈ 0, 21 

 
H(t, s) :=
c + (2s − 1) γ1 (t) − c , s ∈ 21 , 1]

H is continuous, the only point to check is s = 12 , hence

 
1

H t, = c = lim1 c + (2s − 1) γ1 (t) − c
2 s→ 2

To see that the image of H is contained

 in  Ω for all t ∈ [a, b] and s ∈ [0, 1], check
for example that if t ∈ [a, b] and s ∈ 0, 21 , then in case that H(t, s) ∈
/ Ω, it means
≤0
that for some t, s the value of H(t, s) ∈ R , i.e.


c + (1 − 2s) γ0 (t) − c ≤ 0


⇐⇒ (1 − 2s) γ0 (t) − c ≤ −c
   
−c 1 2s
⇐⇒ γ0 (t) ≤ +c=c 1+ =c ≤ −c
1 − 2s 2s − 1 2s − 1
But 0 ≤ s ≤ 12 =⇒ 2s ≥ 0 and 2s − 1 ≤ 0. Hence, γ0 (t) ∈ (−∞, 0] but at the
same time γ0 (t) ∈ Ω and this cannot happen. For 21 < s ≤ 1 is similar.
Remark 3.13. If γ0 homotopic to γ1 in Ω (either closed or with fixed endpoints), we
write γ0 ∼Ω γ1 and simply write γ0 ∼ γ1 if Ω is fixed and clear.

Then ∼ is an equivalence relation:

• The curve γ0 is homotopic with itself via H(t, s) = γ0 (t)
• If γ0 ∼ γ1 with H(t, s), then γ1 ∼ γ0 with H̃(t, s) := H(t, 1 − s)
• If γ0 ∼ γ1 with F (t, s) and γ1 ∼ γ2 with G(t, s), then define
, s ∈ 0, 12 
  
F (t, 2s)
H(t, s) :=
G(t, 2s − 1) , s ∈ 12 , 1
that gives a homotopy between γ0 ∼ γ2

3.5 The Homotopy Theorem

We can now state the Homotopy Theorem.
3.5. THE HOMOTOPY THEOREM 127

Theorem 3.14 (Homotopy Theorem). [SS10, Theorem III.5.1] Let Ω ⊆ C be open.

Let γ0 , γ1 be two curves in Ω that are

(i) either γ0 , γ1 closed curves and homotopic

(ii) or γ0 , γ1 have the same endpoints and are homotopic with fixed endpoints.

Then, for f ∈ H (Ω) we have that

Z Z
f (z)dz = f (z)dz
γ0 γ1

Example 3.21. 1. Let Ω = DR (z0 ) and R > 0 and let im γ0 = Cr (z0 ) with r ∈
(0, R) as in the following picture.

Then γ0 can be deformed into the point z0 by dilation (which can be be thought as
the constant curve for which ∀t ∈ [a, b] : γ1 (t) = z0 ), so for f ∈ H (Ω)
Z Z
f (z)dz = f (z)dz = 0
Cr (z0 ) σz0
128 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

since ∀t ∈ [a, b] : γ1′ (t) = 0

Consider the homotopy

H : [0, 2π] × [0, 1] → DR (z0 )

(t, s) 7→ (1 − s)eit + sz0

In fact, in DR (z0 ) any closed curve γ is homotopic to a constant curve. Hence,

we get Z
f (z)dz = 0
γ

2. Let Ω = C∗ and

γ0 : [0, π] → Ω γ1 : [0, π] → Ω
t 7→ e it
t 7→ e−it

The two paths are not homotopic with fixed endpoints, since if they were, then we
would get that for z1 ∈ H (Ω) it would hold that
Z Z
1 1
dz = dz
γ0 z γ1 z

and therefore Z Z Z
1 1 1
dz − dz = dz = 0
γ0 z γ1 z C1 (0) z
but Z
1
dz = 2πi ̸= 0
C1 (0) z
We now look at the proof of the Homotopy Theorem 3.14. We will look at the case of
closed curves (in [SS10] one finds instead the version where the endpoints of a curve
are fixed).
Proof of the Theorem 3.14. 1. A simpler version of the proof is given with the ex-
tra assumption that the homotopy H(t, s) has continuous second order partial
derivatives and
∂ 2H ∂ 2H
∀(t, s) ∈ [a, b] × [0, 1] : (t, s) = (t, s)
∂s∂t ∂t∂s
For this we first recall from Real Analysis:

Let H : [a, b] × [0, 1] → Ω ⊆ C, (t, s) 7→ H(t, s) =: γs (t) be an homotopy between

the two paths γ0 , γ1 and let h : [a, b] × [0, 1] → R be defined as here below.
3.5. THE HOMOTOPY THEOREM 129

Recall: Let h : [a, b] × [0, 1] → R, (t, s) 7→ h(t, s)

Suppose that ∂h
∂s
exists and is continuous, if we define

G : [0, 1] → R
Z b
s 7→ G(s) := h(t, s)dt
a

then we get that G is differentiable with

Z b
′ ∂h
G (s) = (t, s)dt
a ∂s

We apply this to the real and imaginary part of the following:

Z b Z b Z

∂H
′
I(s) := f H(t, s) (t, s) dt = f γs (t) γs (t)dt = f (z)dz
a | {z ∂t } a γs

=h(t,s)

Also, note that

Z Z
I(0) = f (z)dz and I(1) = f (z)dz
γ0 γ1

We want to show that I(0) = I(1) by showing that I(s) is constant. So consider
Z b  
′ ∂ ∂H
I (s) = (f ◦ H)(t, s) (t, s) dt =
a ∂s ∂t
Z b 
′ ∂H ∂H ∂ ∂H
= (f ◦ H)(t, s) (t, s) (t, s) + (f ◦ H)(t, s) (t, s) dt
a ∂s ∂t ∂s ∂t

and note that what is inside the round parenthesis is also equal by assumption to
 
∂ ∂H
(f ◦ H)(t, s) (t, s)
∂t ∂s

Hence, we have that

Z b 
′ ∂ ∂H
I (s) = (f ◦ H)(t, s) (t, s) dt =
a ∂t ∂s
 t=b

∂H
= f H(t, s) (t, s) =
∂s t=a

∂H
∂H
= f H(b, s) (b, s) − f H(a, s) (a, s) = 0
∂s ∂s
130 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Since H is a homotopy of closed curves for which it holds that

∀s ∈ [0, 1] : γs (a) = H(a, s) = H(b, s) = γs (b)

and also that

∂H ∂H
∀s ∈ [0, 1] : (a, s) = (b, s)
∂s ∂s
Finally, we have that ∀ ∈ [0, 1] : I ′ (s) = 0 and therefore I is constant. In
particular, we obtain Z Z
f (z)dz = f (z)dz
γ0 γ1

2. For the general proof the idea is the following: if we make a small deformation
of one of the curves γs (t), say γ0 (t) to γ 1 (t), so that if we look at a small piece
N
around a point of γ0 (t), say t ∈ (tm , tm+1 ) for m ∈ {0, ..., N − 1} and N the
cardinality of the dissection of [a, b] and of [0, 1], then we can show that these are
contained in a small disc in Ω

Let t ∈ (tm , tm+1 ) for m ∈ {0, ..., M − 1} and M the cardinality of the dissection
of [a, b]

H(t, 0) = γ0 (t)
 
1
H t, = γ 1 (t)
N N

we can apply Cauchy’s Theorem 2.5 in a disc to get

Z Z
f (z)dz = f (z)dz
(m) (m) ′
γ0 γ 1 ⊎σ⊎σ
N
3.5. THE HOMOTOPY THEOREM 131

Now, we move over to the whole curve γ0 , γ 1 using small discs contained in Ω to
N
get Z Z
f (z)dz = f (z)dz
γ0 γ 1
N

To make this idea more precise, we need these two facts:

(a) If K = im(H) = H ([a, b] × [c, d]) ⊆ Ω, then K is compact (since H is

continuous and [a, b] × [0, 1] is compact, see [EW22]).
(b) A continuous function on a compact set is uniformly continuous.

We then have that (a) implies the following Lemma:

Lemma 3.3. ∃ε > 0∀z ∈ K ⊆ Ω : Dε (z) ⊆ Ω

Proof. Assume on the contrary that no such ε exists. Then ∀n ∈ N∗ ∃zn ∈ K :

D 1 (zn ) ⊈ Ω, i.e. ∃wn ∈ C \ Ω : |wn − zn | < n1 for any n ≥ 1. This way we get a
n
∗
sequence (zn )n∈N∗ ∈ K N , where K is compact, hence (zn )n∈N∗ has a subsequence
(znk )k∈N∗ which converges to limk→∞ znk = z. Since K is also closed, we have
z ∈ K. Now, because of |wn − zn | < n1 for any n ≥ 1, we have for any k ∈ N∗

1
|wnk − znk | <
nk
so (wnk )k∈N∗ also converges to z. (wnk )k∈N∗ is also in C \ Ω, so its limit point is
z ∈ C \ Ω and this is a contradiction.

This Lemma 3.3 together with (b) will allow us to find the small discs that are
contained in Ω. This is because we can divide the rectangle [a, b] × [c, d] into small
rectangles such that the images of these small rectangles are contained in small
132 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

discs of radius ε

More precisely, let (tm )m∈{0,...,N } ∈ [a, b]{0,...,N } and (sn )n∈{0,...,N } ∈ [0, 1]{0,...,N }
represent the dissections of the two intervals and ε > 0. Since H is uniformly
continuous on [a, b] × [c, d], then it exists an N ∈ N∗ such that

H(t, s) − H(tm , sn ) < ε

whenever
2
(t, s) − (tm , sn ) <
N
b−a n
for zm,n = H(tm , sn ) with tm := a + N
m and sn := N
for m, n ∈ {0, ..., N }

Let Qm,n := [tm , tm+1 ] × [s , s ] for m, n ∈ {0, ..., N − 1}, since the diameter of
√ n n+1
2
Qm,n is diam(Qm,n ) = N < N2 , it follows that

H(Qm,n ) ⊆ Dε (zm,n )

where z
m,n := H(tm , sn ) and xm,n := (tm , sn ), with z0,n = H a, Nn

and zN,n =
H b, Nn ; we note that in this case z0,n = zN,n
3.5. THE HOMOTOPY THEOREM 133

Using induction on n ∈ N ∩ {0, ..., N }, we want to show now that

Z Z
f (z)dz = f (z)dz
γn γ0
N

This is clearly true for n = 0, as N0 = 0, hence we consider n ≥ 1 and assume

that it holds Z Z
f (z)dz = f (z)dz
γ n−1 γ0
N

It is enough to show that

Z Z
f (z)dz = f (z)dz
γ n−1 γn
N N

n−1 n



where γ n−1 = H t, N
and γ Nn = H t, N
N

For each m ∈ {0, ..., N }, let

(m) (m)
γ n−1 := γ n−1 and γ n := γ Nn [tm ,tm+1 ]
N N [tm ,tm+1 ] N
134 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

and let σm := ℓ[zm,n−1 ,zm,n ] be line segment between zm,n−1 and zm,n , thus σm+1 =
ℓ[zm+1,n−1 ,zm+1,n ] be the line segment between zm+1,n−1 and zm+1,n ; as special case
we consider σ0 = [z0,n−1 , z0,n ] and σN = [zN,n−1 , zN,n ]

Now we apply Cauchy’s Theorem 2.5 in the disc Dε (zm,n−1 ) and obtain
Z Z Z Z
f (z)dz − f (z)dz − f (z)dz + f (z)dz = 0
(m) (m)
γn σm+1 γ n−1 σm
N N

Summing over m to get the full γ Nn , we have that

Z N
X −1 Z
f (z)dz = f (z)dz =
(m)
γ n−1 m=0 γ n−1
N N
 
N
X −1 Z N
X −1 Z Z 
= f (z)dz  + f (z)dz − f (z)dz =
(m)
m=0 γn m=0 σm σm+1
N
Z Z Z
= f (z)dz + f (z)dz − f (z)dz =
γn σ0 σN
N | {z }
=0
Z
= f (z)dz
γn
N

We have σ0 = σN , since γ n−1 (a) = H a, n−1 = H b, n−1

N N
= γ n−1 (b) and simi-
N N
larly γ Nn (a) = γ Nn (b), i.e. the curves γs are closed. Hence, we proved by induction
that the integral over γ0 and over γ1 are equal, which concludes the proof.

In spaces like C, C\(−∞, 0] or any convex set, we observed that any two closed curves, or
any two curves with the same endpoints, are homotopic. This leads us to the following
definition.
3.5. THE HOMOTOPY THEOREM 135

Definition 3.12. An open set Ω ⊆ C is called simply connected, if it is connected

and any two curves with the same endpoints are homotopic in Ω

E.g. C, C \ (−∞, 0] and Dr (z0 ) are simply connected, while C∗ is not.

As Corollary of the Homotopy Theorem 3.14 we have that

Theorem 3.15. [SS10, Theorem III.5.2] Any holomorphic function f ∈ H (Ω) on

a simply connected region Ω ⊆ C has a primitive. In particular, we have that
Z
f (z)dz = 0
γ

for any closed curve γ in Ω and that any two primitives differ by a constant.

Proof. Let Ω ⊆ C be simply connected and fix z0 ∈ Ω, we define F ∈ CΩ such that for
any z ∈ Ω
Z
F (z) := f (w)dw
γz

where γz is a curve connecting z0 to z. Note that this is a well-defined function, since

using the Homotopy Theorem 3.14 and Ω simply connected, any two curves γz , γ̃z
between z0 and z satisfy γz ∼ γ̃z ; therefore they give the same value of
Z Z
f (z)dz = f (z)dz
γz γ̃z

Let z ∈ Ω,
if we choose h small enough, so that the image of the line segment
im ℓ[z,z+h] ⊆ Ω, then
Z
F (z + h) − F (z) = f (w)dw
ℓ[z,z+h]
136 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Arguing as in the proof of Theorem 2.3 or using continuity of f as below, we get that

F (z + h) − F (z)
lim = f (z)
h→0 h
which shows that F is a primitive of f in Ω, as z was arbitrary. I.e.
Z


F (z + h) − F (z) = f (w) − f (z) + f (z) dw =
ℓ[z,z+h]
Z Z


= f (z) dw + f (w) − f (z) dw
ℓ[z,z+h] ℓ[z,z+h]
| {z }
=h

so Z


f (w) − f (z) dw ≤ h sup
f (w) − f (z)
ℓ[z,z+h] w∈im ℓ[z,z+h]

hence
F (z + h) − F (z)
− f (z) ≤ sup
f (w) − f (z)
h
w∈im ℓ[z,z+h]

h→0
Being f is of C 0 - class, it implies that sup
f (w) − f (z) −−→ 0
w∈im ℓ[z,z+h]

3.6 Complex Logarithm

For z ∈ C∗ we want to define the logarithm of z = reiθ and we want it to be the inverse
function of the exponential function, i.e. w = log(z) if ew = z. A natural candidate is

log(z) = log(r) + iθ

where log(r) is the usual logarithm log : R+ → R of the positive real number r. The
problem is that this is not single valued, as θ is only unique up to an integer multiple
of 2π. Indeed, the argument is multivalued.

Example 3.22. Let z = 1 ∈ C, it holds that e0 = 1, but also that for any w ∈ 2πiZ
the equality ew = 1 holds.

We want a holomorphic function ℓ ∈ H (Ω), which satisfies

exp ◦ℓ = id

throughout its domain of definition Ω

3.6. COMPLEX LOGARITHM 137

Definition 3.13. Let Ω ⊂ C be open. A (fixed) branch of the logarithm on Ω,

denoted by logΩ , is a function in H (Ω) such that


∀z ∈ Ω : exp logΩ (z) = z

If Ω is clear from the text, sometimes this function is also denoted by log

Note that any function f ∈ H (Ω) that meets that condition is a branch of logarithm
on Ω, but when one is fixed, then it it denoted by logΩ

Remark 3.14. 1. Since exp(z) ̸= 0 for all z ∈ C, in order for logΩ to exist we need
that 0 ∈
/Ω

C
2. If Ω = C∗ , even though exp ∈ C∗ is surjective, there is no branch of
the
logarithm on Ω. Indeed, if there were a f ∈ H (Ω) such that ∀z ∈ Ω : exp f (z) =
z, then differentiating on both sides would give

f ′ (z) exp f (z) = 1

| {z }
=z

for all z ∈ Ω, which then gives f ′ (z) = 1

z
for all z ∈ Ω, i.e. 1
z
has a primitive in
C∗ , which would imply that Z
1
dz = 0
C1 (0) z
which we know it is not.

3. If Ω is open and connected and ℓ = logΩ ∈ CΩ is a logarithm, then ℓ̃ is also a

logarithm on Ω if and only if
ℓ̃ = ℓ + 2πin


for some
n ∈ Z. Indeed, if ℓ̃ is a logarithm function, then exp ℓ̃(z) = z and
exp ℓ(z) = z


Hence, for all z ∈ Ω we have exp ℓ̃(z) − ℓ(z) = 1, so for all z ∈ Ω we also have

ℓ̃(z) − ℓ(z) ∈ 2πiZ

ℓ̃(z)−ℓ(z)
i.e. 2πi
is a constant integer valued function on Ω, which is connected. Hence,
its image under ℓ̃(z)−ℓ(z)
2πi
is connected and is also a subset of Z, therefore it is a
single point n ∈ Z. Conversely, if ℓ̃ = ℓ + 2πin for some n ∈ Z, then




exp ℓ̃(z) = exp ℓ(z) exp(2πin) = exp ℓ(z) = z
138 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Theorem 3.16. [SS10, Theorem III.6.1] Let Ω ⊂ C∗ be a simply connected set,

of the logarithm on Ω, i.e. a function F ∈ H (Ω) such
then there exists a branch
that ∀z ∈ Ω : exp F (z) = z

/ Ω, then z1 ∈ H (Ω) and since Ω is simply connected, we have that

Proof. Since 0 ∈
1
then z has a primitive in Ω that we call f (z)

Let G : CΩ such that z 7→ G(z) := z exp − f (z) , since f ′ (z) = z1 we have that

G′ (z) = −zf ′ (z) exp − f (z) + exp − f (z) = − exp − f (z) + exp − f (z) = 0

Since Ω is connected, it follows that G(z) = ze−f (z) = a (constant) necessarily, for some
a ∈ C. Moreover, since exp ̸= 0, z ̸= 0, also a ̸= 0, so we obtain that it exists b ∈ C
such that a = exp(b) and therefore, by algebraic manipulations on G, that

z
exp f (z) =
a
Hence, let F ∈ CΩ such that z 7→ F (z) := f (z) + b, then




exp F (z) = exp f (z) + b = exp f (z) exp(b) = z
| {z } | {z }
= az =a

and so F (z) is a branch of the logarithm on Ω

Definition 3.14. Let C− := C\(−∞, 0]. The principal branch of the logarithm
on C− is the unique function logC− ∈ H (C− ) such that logC− (1) = 0. This particular
logC− is also denoted by Log (with capital L) or LogC− .

Proposition 3.5. If z = reiθ ∈ C− with r > 0 and θ ∈ (−π, π), then the principal
branch of the logarithm is given by the formula


LogC− (z) = log(r) + iθ = log |z| + i Arg(z)

The common notation to use for the Principal branch is Log

Proof. Let f ∈ H (Ω) be such that

Z
1
f (z) := dw
γz w
3.6. COMPLEX LOGARITHM 139

−
hence, a primitive of z1 by Theorem
R 13.15, where we take a path γz in C which starts
at 1 and ends at z. Note that γ1 w dw = 0, hence logC− (1) = 0. This implies that
this branch of logarithm is equal to the principal one (we are going to use the usual
notation for this branch from now on).

If z = reiθ with r < 1, take the path γz that goes on the real line from 1 to r, then on
the circular arc to z, so
1 −θ
−ire−it
Z Z
dx
Log(z) = − + dt = log(r) +iθ
r x 0 re−it | {z }
| {z } | {z } the real one
on the x-axis on the arc z = re−it for t ∈ (0, −θ)

Instead, if r > 1, we take the path

and so similar calculations give the result (Exercise).

Notation: From now on, if not specified, log = Log

Remark 3.15. 1. The identity

Log(z) + Log(w) = Log(zw)

140 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

does not hold in general for all z, w, zw ∈ C− , but if w = reiα , z = seiβ and
zw = sreiθ with α, β, θ ∈ (−π, π), then exists γ ∈ {−2π, 0, 2π} such that

θ =α+β+γ

Then

Log(zw) = log(rs) + iθ =
= log(r) + log(s) + i(α + β + γ) =



= log(r) + iα + log(s) + iβ + iγ =
= Log(w) + Log(z) + iγ

In particular,

Log(z) + Log(w) = Log(zw) ⇐⇒ γ = 0 ⇐⇒ α + β ∈ (−π, π)

The condition is satisfied whenever Re(w) > 0 and Re(z) > 0, hence in the real
positive half plane H.

2. For the principal branch of the logarithm, Log, we have the following Taylor ex-
pansion
∞
X (−1)n−1
∀z ∈ D1 (1) : Log(z) = (z − 1)n
n=1
n

To see this, we differentiate both sides: on the left we have z1 , while on the right
we have ∞ ∞
X
n−1 n−1
X 1 1
(−1) (z − 1) = (1 − z)n = =
n=1 n=0
1 − (1 − z) z
(−1)n−1
Hence, log and ∞ (z − 1)n only differ by a constant. Looking at z = 1
P
n=1 n
we obtain that the constant is 0 and with it the result.

3. We have
3.6. COMPLEX LOGARITHM 141

The image of a punctured circle around 0

z ∈ C− : |z| < r and Arg(z) ∈ (−π, π)



is the vertical interval


z ∈ C : Re(w) = log |z| and Im(w) ∈ (−π, π)

where if r < 1, then Re(w) < 0 or if r > 1, then Re(w) > 0


 image of z ∈ C : Arg(z) = θ , a ray from 0 to ∞, is the horizontal line
The
w ∈ C : Im(w) = θ
4. We can define a holomorphic branch of the logarithm on any cut plane of the form
 
Ωα := C \ z ∈ C : Arg(z) = α ∪ {0}

for α ∈ [−π, π], such that

w = logΩα (z) = log |z| + iθ , with θ ∈ (α, α + 2π)

and consequently
θ = Arg(z) + α

5. Let Ω ⊆ C∗ be simply connected and logΩ ∈ CΩ a branch of the logarithm. Let

α ∈ C and z ∈ Ω. We define

z α := exp α logΩ (z) =: [z α ]

Note that this definition depends on the choice of the branch of the logarithm logΩ :
if we choose another branch of the logarithm, call it ℓ, as ℓ = logΩ +2πik for some
k ∈ Z, then 

[z α ]ℓ = exp α logΩ (z) + 2πik = z α e2πikα
If we set Ω = C− , if we choose the principal branch of the logarithm with Log(1) =
0 and α = m1 for m ∈ N, then
1 1
z m = e m Log(z)
142 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

satisfies
 m m
1 1 1
Y
zm = e m Log(z) = em m Log(z) = eLog(z) = z
j=1

Example 3.23. Let Log be the principal branch of the logarithm on C− , then
1 1
z 2 = e 2 Log(z)

1
Note that for z ∈ R+ , the value of z 2 is the usual positive square root.

If we choose another branch of the logarithm in C− , for some k ∈ Z, e.g.

logC− ,k (z) := log(r) + i(θ + 2πk)

then
 
1 1 1 θ+ikπ 1 θ 1 θ
h 1i
z = exp
2 logC− ,k = r 2 ei 2 = r 2 ei 2 eikπ = r 2 ei 2 (−1)k = (−1)k z 2
2

i.e. infinitely many choices

h fori the branch of the logarithm give only two different choices
√ 1
for z (where we wrote z 2 for the principal branch of the square root).

For the choice logC− ,1 (z) = log(r) + i(θ + 2π) we have (always with r > 0 and Arg(z) ∈
(π, 3π))
3.6. COMPLEX LOGARITHM 143

Definition 3.15 (Logarithm of a function). Let Ω ⊆ C open and f ∈ H (Ω), a

function g ∈ H (Ω) such that
f (z) = eg(z)
is called a logarithm of f and (if fixed) is denoted by logΩ (f ) or log(f ) in the sense
of composition, i.e. logΩ (f ) = logΩ ◦f

Remark 3.16. The logarithm of a function is in general not a branch of logarithm on

any subset Ω ⊆ C

Finally, we have that if f ∈ H (Ω) on a simply connected region Ω ad f is non-vanishing

in all of Ω, then f has a logarithm in Ω,

Theorem 3.17. [SS10, Theorem III.6.2] Let Ω ⊆ C be a simply connected region.

If f ∈ H (Ω) non-vanishing in all of Ω, then ∃g ∈ H (Ω), called logarithm of f , i.e.
log(f ), such that
f (z) = eg(z)

f′
Proof. Exercise. Define g a primitive of f

Corollary 3.2. If f ∈ H (Ω), non-vanishing in all of a simply connected region

Ω ⊆ C, then f has a square root in Ω, i.e.

∃h ∈ H (Ω) : h2 (z) = f (z)

Proof. Let    
1 1
h(z) = exp log(f ) = exp g(z)
2 2
from Theorem 3.17, then
h2 = exp g(z) = f (z)

Before we move to conformal maps in the next section, we mention that there are
various ways to look at simply connected regions. This is taken up in the book in the
Appendix B [SS10].
144 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

We have seen that if Ω is simply connected (i.e. such that any two curves in Ω with
same endpoints are homotopic), then for all closed curve γ in Ω and for all f ∈ H (Ω)
we have Z
f (z)dz = 0
γ

Definition 3.16. A region Ω ⊆ C is called holomorphically simply connected,

if for all closed curve γ in Ω and all f ∈ H (Ω)
Z
f (z)dz = 0
γ

Clearly, we have with Cauchy’s Theorem 2.5 or the Homotopy Theorem 3.14 that

Ω simply connected =⇒ Ω holomorphically simply connected

In fact, the converse is also true, as we have

Theorem 3.18. Let Ω be a region, then

Ω holomorphically simply connected ⇐⇒ Ω simply connected

The other direction

Ω holomorphically simply connected =⇒ Ω simply connected

uses the Riemann Mapping Theorem (which we will see soon).

For bounded regions we also have

Theorem 3.19. If Ω is a bounded region in C, then

Ω is simply connected ⇐⇒ C \ Ω is connected

The proof of the direction: Ω bounded and simply connected ⇒ C \ Ω connected, uses
the notion of winding numbers, which we are going to briefly discuss as next, since it
also leads to the natural generalisation of the Residue Theorem 3.6.

Remark 3.17. In the above Theorem 3.19, the assumption that Ω is bounded in C is
important, since the infinite strip is simply connected and unbounded, its complement
has 2 components. However, if the complement is taken in Ĉ = C∪{∞}, the conclusion
holds if Ω is bounded or not.
3.7. WINDING NUMBERS 145

3.7 Winding numbers

See Appendix B in [SS10].

We have seen that for f ∈ H (Ω), for some Ω ⊆ C simply connected, if γ1 , γ2 are two
closed curves such that γ1 ∼Ω γ2 , then by Theorem 3.14
Z Z
f (z)dz = f (z)dz
γ1 γ2

We want to generalise it to f ∈ M (Ω), hence we want to understand the integral

Z
f dz
γ

for some γ in Ω and f ∈ M (Ω)

Recall: If f ∈ M (Ω), z0 ∈ Ω, im(γ) = ∂Dr (z0 ) and Dr (z0 ) ⊂ Ω, then

Z X
f (z)dz = 2πi Resw (f )
γ w∈(Sf ∩int(γ))

with ∀t ∈ [0, 2π] : γ(t) = z0 + r(t)eiθ(t) in Ω ⊆ C for Ω open and for some functions r, θ
of class C 1 such that ∀t ∈ [0, 2π] : r(t) > 0 and r(0) = r(2π), θ(0) = θ(2π)
The same proof we gave for the Residue Formula 3.6 for a circle works also here for
Z X
f (z)dz = 2πi Resw (f )
γ w∈(Sf ∩int(γ))

The Homotopy Theorem 3.14 gives the following first generalisation of the Residue
Theorem 3.6.

Proposition 3.6. Let Ω ⊆ C be open and let f ∈ M (Ω). Let V := Ω \ Sf so that

f ∈ H (V ). Then

(i) If γ1 , γ2 are two closed curves in V ⊆ Ω, which are homotopic in V , then

Z Z
f (z)dz = f (z)dz
γ1 γ2

(ii) In particular, if γ2 is a circle (with counterclockwise orientation), then

Z Z X
f (z)dz = f (z)dz = 2πi Resw (f )
γ1 γ2 w∈Sf ∩int(γ2 )
146 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Proof. (i) This is a special case of the Homotopy Theorem 3.14, since f ∈ H (V )
and γ1 ∼V γ2
(ii) Follows from the previous point and the Residue Formula 3.6.

To look at more general curves, we first introduce the winding number of a curve.

Definition 3.17. [SS10, Appendix B p.347] Let z0 ∈ C and γ a piecewise smooth

closed curve in C, such that z0 ∈/ im(γ). The winding number of γ around z0 is
defined as Z
1 dz
wγ (z0 ) := indγ (z0 ) := ∈Z
2πi γ z − z0
The winding number is also called the index of γ around z0 and denoted by
indγ (z0 )

Why is this called the winding number?

Remark 3.18. To get a feeling for why this is called the winding number:
1. If γ(t) = z0 + reit for t ∈ [0, 2πn] and for n ∈ N, i.e. the circle with center z0
traced n times counterclockwise. Then
Z 2πn Z 2πn
ireit
Z
1 dz 1 1
wγ (z0 ) = = dt = dt = n ∈ N
2πi γ z − z0 2πi 0 reit 2π 0

2. On the other hand, if γ(t) = z0 + reit for t ∈ [0, 2π], but we are looking at a point
z0 ̸= z1 ∈ C
Z   
1 dz X 1 0 , z1 ∈/ int(γ)
wγ (z1 ) = = Resz̃ =
2πi γ z − z1 z − z1 1 , z1 ∈ int(γ)
z̃∈Sf ∩int(γ)
3.7. WINDING NUMBERS 147

1 dz
R
So at least, when γ is a circle, then the integral 2πi γ z−z1
indeed tells us whether
γ wraps around z1 or not.
“To get an intuition”: For a general smooth γ : [a, b] → C with γ(a) = γ(b) and z0
inside the path, the following imprecise and really not completely correct argument
might give an insight as to why it is called winding number.
Z b
γ ′ (t)
Z
dz
wγ (z0 ) = = dt
γ z − z0 a γ(t) − z0

From Real Analysis we might be tempted to write this last integral as

b
log γ(t) − z0
a

But of course this is not correct, because γ(t) − z0 is complex valued and
if γ wraps
around a point z0 , then we cannot define an analytic branch of log γ(t)−z0 on C\{z0 }

If we think of Log(z) = log |z| + i Arg(z) and recall that the difficulty in defining the
logarithm comes from choosing the correct value of the Arg(z), we can look at
Z
1


dz = log γ(b) − z0 − log γ(a) − z0 =
γ z − z0



= log γ(b) − z0 + i Arg γ(b) − z0 − log γ(a) − z0 + i Arg γ(a) − z0 =



= i Arg γ(b) − z0 − Arg γ(a) − z0


The ambiguity in defining Arg γ(t) − z0 for t = a, t = b must be an integral multiple
of 2π and this integer counts the number of times γ wraps around z0
148 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

We have indeed the following Proposition, which shows that ωγ (z) is always an inte-
ger.

Proposition 3.7. [SS10, Proposition B.1.3] Let γ be a closed curve in C and Ω =

C \ im(γ), which is open. Then the map

wγ : Ω → C
Z
1 du
z 7→
2πi γ u − z

takes values in Z and is continuous. Hence it is constant on any connected subset of

Ω. Moreover wγ (z) = 0, if |z| is large enough.

Proof. Suppose γ : [a, b] → C is a parametrization of the curve and let

G : [a, b] → C
t
γ ′ (s)
Z
t 7→ G(t) := ds
a γ(s) − z
Note that G(b) = 2πiwγ (z) and G(a) = 0

The Fundamental Theorem of Analysis [EW22] tells us that G is continuous (except

possibly at finitely many points) and differentiable on (a, b) (except at those already
mentioned points) with
γ ′ (t)
G′ (t) =
γ(t) − z

−G(t) [a,b]
Let H(t) = γ(t) − z e in C , then

H ′ (t) = γ ′ (t)e−G(t) − γ(t) − z G′ (t) e−G(t) = 0

| {z }
=γ ′ (t)

by the latter result. Hence, H is constant by Corollary 1.2 and so H(t) = γ(t) −

−G(t)
z e = c for some c ∈ C. We have that ∀t ∈ [a, b] : γ(t) − z = ceG(t) and

eG(a) = γ(a) − z = γ(b) − z = ceG(b)

c = c |{z}
=1

From this it follows that eG(b) = 1 and so G(b) ∈ 2πiZ (c ̸= 0, since γ(t) ̸= z and
e−G(t) ̸= 0).

Since G(b) = 2πiwγ (z), this shows that wγ (z) is integer valued: being
Z b
1 γ ′ (s)
wγ (z) = ds
2πi a γ(s) − z
3.7. WINDING NUMBERS 149

the integral a continuous function, it is a continuous function of z ∈ Ω \ im(γ). Being

also integer valued, wγ (z) is constant in any open connected subset of Ω \ im(γ)

Finally, if M := maxt∈[a,b] γ(t) (it exists, since [a, b] is compact and γ piecewise of class
C 0 ) and |z| > M , then
Z
1 dw 1 L(γ)
wγ (z) = ≤
2π γ w − z 2π |z| − M

where L(γ) denotes the length of γ. Since

|w − z| > |z| − |w| ≥ |z| − M

we then have
1 L(γ) |z|→∞
wγ (z) ≤ −−−−→ 0
2π |z| − M
Hence |wγ (z)| < 1 once |z| is large enough, but being an integer means that wγ (z) = 0,
if |z| is large enough.
We can now give the general Residue Formula.

Theorem 3.20 (Generalised Residue Formula). Let Ω ⊆ C be open and simply

connected, f ∈ M (Ω) and V = Ω \ Sf . Let γ be a closed curve in V . Then we have
Z X
f (z)dz = 2πi wγ (z) Resz (f )
γ z∈Sf

Proof. For any z0 ∈ Sf , let Pzf0 be the principal part of f at z0

N (z0 )
X a−j (z0 )
Pzf0 (z) =
j=1
(z − z0 )j

for some a−j (z0 ) ∈ C and with N (z0 ) = − ordz0 (f ) (since f has a pole at z0 , the order
of that point is negative).
Case 1: Let Sf be finite, then
X
f˜ := f − Pzf0
z0 ∈Sf

has removable singularities at z0 ∈ Sf , hence has a holomorphic extension to Ω

˜
R
Hence, we have that γ f dz = 0, as Ω is simply connected.
150 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Hence
 
Z Z X X Z
f dz =  Pzf0 (z) dz = Pzf0 (z)dz
γ γ z0 ∈Sf z0 ∈Sf γ

dz 1 −1 1
R
Recall: γ (z−z0 )j
= 0 if j ̸= 1, since (z−z0 )j
has primitive (z−z0 )j−1 j−1
and γ is
closed.

So
 
Z Z X a−1 (z0 )
f (z)dz =   dz =
γ γ z0 ∈Sf
z − z0

X Z a−1 (z0 )
= dz =
z0 ∈Sf γ z − z0
Z
X dz
= 2πia−1 (z0 ) =
z0 ∈Sf γ z − z0
| {z }
=wγ (z0 )
X
= 2πia−1 (z0 ) wγ (z0 ) =
z0 ∈Sf
X
= 2πi Resz0 (f )wγ (z0 )
z0 ∈Sf

Case 2: Sf is infinite. Pick R > 0 such that wγ (z) = 0 if |z|≥ R and γ is homo-


topic (so that for the homotopy in question H it hold that im H [a, b]×[0, 1] ⊆
DR (0)) to the constant cure in Ω ∩ DR (0) (since Ω is simply connected γ ∼Ω σ
(constant curve), which only involves a bounded set). Then Sf ∩ DR (0) is finite,
since Sf is a discrete set.

Let
X
f˜ := f |Ω∩DR (0) − Pzf0 ∈ H Ω ∩ DR (0)


Ω∩DR (0)
z0 ∈Sf
|z0 |<R

f˜ = 0 since γ is homotopic to the constant curve in Ω∩DR (0)

R
we have then that γ
3.8. CAUCHY INTEGRAL FORMULA 151

Hence
Z X Z
f dz = Pzf0 dz =
γ z0 ∈Sf γ
|z0 |<R
X
= 2πi Resz0 (f )wγ (z0 ) =
z0 ∈Sf
|z0 |<R
X
= 2πi Resz0 (f )wγ (z0 )
z0 ∈Sf

Since for |z0 | ≥ R, wγ (z0 ) = 0

3.8 Cauchy Integral Formula

Corollary 3.3. Let Ω be open and simply connected, f ∈ H (Ω) and γ a closed
cure in Ω, then
Z
1 f (w)
∀z ∈ Ω \ im(γ) : dw = f (z)wγ (z)
2πi w−z

Proof. This is the generalised Residue Theorem 3.20 applied to the function fw−z
(w)
=
g(w), which is meromorphic in Ω and has a simple pole at w = z and residue f (z)

Example 3.24. Let f be meromorphic except for poles at z = a, b, c, d, then

Z
f (z)dz = 2 Resa (f ) + 2 Resb (f ) + 1 Resc (f )
γ
152 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

3.9 Conformal maps and the Riemann mapping The-

orem
See Chapter 8 in [SS10].

Motivating questions:
1. Given two open sets U, V ⊆ C, when does there exists a holomorphic bijection
between them, i.e. when is there a bijective f ∈ V U such that f ∈ H (U, V )?
We are going to see that the inverse map f −1 ∈ U V is automatically also holo-
morphic (compare open sets using holomorphic functions).

2. Given an open set Ω ⊆ C, what conditions guarantee that there is a holomorphic

bijection from Ω to D (where D is the unit disc)?
Why D? D has a very nice geometric structure and we developed most properties
of holomorphic functions for D first. If there is a holomorphic bijection between
Ω and D we can hope to transfer questions about holomorphic functions on Ω to
holomorphic functions on D
Remark 3.19. 1. We’ll start by examples of simple maps and show for example
that there is a holomorphic bijection between D and the upper half plane H.
We can then compose simple maps to get more examples of holomorphic bijections.

2. We will then prove Schwarz’s Lemma, which says any f ∈ RD such that f (0) = 0
must satisfy

(a) ∀z ∈ D : f (z) ≤ |z|

(b) If for some z0 ̸= 0 we have |f (z0 )| = |z0 |, then f is a rotation.
(c) |f ′ (0)| ≤ 1 and if equality holds, then f is a rotation.

3. Schwarz’s Lemma will then give us all holomorphic bijections of D to itself.

4. Then we will get to Riemann Mapping Theorem, which says that if Ω ̸= C or

Ω ̸= ∅ and is simply connected, then there is a holomorphic bijection between Ω
and D
More precisely, for any z0 ∈ Ω there exists a unique f ∈ DΩ such that f (z0 ) = 0
and f ′ (z0 ) > 0
Remark 3.20. The Riemann Mapping Theorem says that there are only three kinds of
simply connected domains in C (up to holomorphic bijections) ∅, C and D

Note that there can be no holomorphic bijection f ∈ DC between C and D, since in that
case f would be bounded and entire; hence by Liouville’s Theorem 2.8 f is constant.
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 153

Note that for Ω to be connected is also a necessary condition, since D is connected. The
same is true for simply connected, since if f ∈ V U is a holomorphic bijection with U
simply connected, then so is V

Definition 3.18. Let U, V ⊆ C be two open sets in C.

• An injective map f ∈ H (U, V ) is called a conformal map from U to V .

These are denoted by M onH (U, V )

• If f is bijective, then it is called a conformal equivalence or biholomorphic,

or a holomorphic isomorphism and U and V are said to be conformally
equivalent. The set of such functions is denoted by IsomH (U, V )

• If U = V , a conformal equivalence is called a (conformal) automorphism

and the the of all such (conformal) automorphisms is denoted by AutH (D).

Remark 3.21. Note that there is a small difference in the definition of conformal
compared to the book [SS10]: in the book f is taken to be bijective.
Also, from now on, unless otherwise specified, we are going to assume that U, V ⊆ C
and that they are open in C

Proposition 3.8. [SS10, Proposition VIII.1.1] If f ∈ V U is conformal (i.e. is holo-

morphic and injective), then

∀z ∈ U : f ′ (z) ̸= 0

The inverse of f , which is defined on the image of f , is holomorphic, i.e. f ∈

im(f )U ⊆ V U is a conformal equivalence and f −1 ∈ U im(f ) is also a conformal
equivalence.

Proof. Suppose f is injective and holomorphic, but on the contrary ∃z0 ∈ U : f ′ (z0 ) = 0
We want to show that f cannot be injective. Let h : U → V, z 7→ h(z) := f (z) − f (z0 ),
this implies that h (z0 ) = 0 and h′ (z0 ) = 0

• If k = ordz0 f (z) − f (z0 ) , then by our assumption k ≥ 2

• If k = ∞, then ∀z ∈ U : f (z) − f (z0 ) = 0, hence f is constant and cannot be

injective.

Therefore, we can assume that k < ∞, by Theorem 2.7 we have that ∃r > 0 such that
for all z ∈ Dr (z0 )
154 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

f (k) (z0 )
f (z) − f (z0 ) = (z − z0 )k + G(z) (z − z0 )k+1 =
k!
= a (z − z0 )k + G(z) (z − z0 )k+1
k
where f k!
(z0 )
=: a ̸= 0 and z ∈ Dr (z0 ). Since the zeroes of f ′ are isolated, we can also
choose r > 0 such that f ′ (z) ̸= 0 for z ∈ Ḋr (z0 )

The idea is to use Rouché’s Theorem 3.11 to show that for any w ∈ C

gw : U → V, z 7→ gw (z) = f (z) − f (z0 ) − w

has the same number of zeroes as a (z − z0 )k − w in some disc around z0

Since a (z − z0 )k = w has k solutions, we will have that gw (z) = f (z) − f (z0 ) − w

has k zeroes for z sufficiently close to z0 . Denote these zeroes of gw by z1 , ..., zk . If
w ̸= 0, then those zeroes are not equal to z0 (if for some k we had that zk = z0 , then
0 = gw (z0 ) = f (z0 ) − f (z0 ) + w = w ̸= 0).

Since f ′ (z) ̸= 0 for z ∈ Ḋr (z0 ), we have that gw′ (z) = f ′ (z) ̸= 0 for z ∈ Ḋr (z0 )

Hence, each zero has order 1 and they are distinct, but that means that there exist k
distinct points z1 , z2 , ..., zk such that f (zi ) = f (z0 ) + w, i.e. f is not injective.

To show that in some neighbourhood of z0 the function gw (z) = f (z) − f (z0 ) − w has
k zeroes, we write the following expansion for z ∈ Dr (z0 )

f (z) − f (z0 ) − w = a (z − z0 )k + G(z) (z − z0 )k+1 − w

 
= a (z − z0 ) − w + G(z) (z − z0 )k+1
k

We apply Rouché’s Theorem 3.11 as follows: let c := sup|z−z0 |= r2 |G(z)|, c exists since

k
G is continuous. Pick s ∈ 0, min{ 2r , 1} and assume that |w| < |a| 2s . On Cs (z0 ),

by inverse triangular inequality one has

 s k  s k
k k
a (z − z0 ) − w ≥ |a|s − |a| ≥ |a|
2 2
and that
G(z) (z − z0 )k+1 ≤ csk+1

k |a|
So, if |a| 2s > csk+1 , i.e. s < c2k , then we can apply Rouché’s Theorem 3.11 to get

that g(z) = f (z)−f (z0 )−w has the same number of zeroes in Ds (z0 ) as a (z − z0 )k −w,
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 155
 n o
s k |a| r


for |w| < |a| 2
and s < min , ,1
c2k 2
as wanted.

Note that if w = reiθ , then the zeroes of a (z − z0 )k − w are at {zn }k−1 n=0 , where
1
1 θ+2πn
 
zn − z0 = wa k ei( k ) for n ∈ {0, ..., k − 1}, but then |zn − z0 | = |w|
k
|a|
< 2s < s.
Hence, all k roots of a (z − z0 )k − w are inside Ds (z0 )

The rest is straightforward: f ∈ f (U )U is clearly bijective. Without loss of generality,

assume that f (U ) = V . The inverse function f −1 ∈ U V is continuous, since f ∈ V U is
an open map.
Let w0 ∈ V and w ∈ V close enough to w0 . We write w = f (z) and w0 = f (z0 ). If
w ̸= w0 , then we have
f −1 (w) − f −1 (w0 ) z − z0 1
= = f (z)−f (z0 )
w − w0 f (z) − f (z0 )
z−z0

Since f ′ (z0 ) ̸= 0 and f −1 is continuous, we have

f −1 (w) − f −1 (w0 ) 1 1
lim = lim =
w→w0 w − w0 z→z0 f (z)−f (z0 ) f′ (z0 )
z−z0

Hence f −1 ∈ H (V ) with V = im(f )

Remark 3.22. 1. Proposition 3.8 says that if f ∈ V U is a conformal equivalence,
then f −1 ∈ U V is automatically a conformal equivalence.

2. The conformal equivalence is an equivalence relation:

• u ∼c u, since id : U → U, u 7→ u as the identity map is bijective and

holomorphic
• If U ∼c V with f ∈ V U , then V ∼c U with f −1 ∈ U V
• If U ∼c V and V ∼c W with f ∈ V U and g ∈ W V respectively, then
g ◦ f ∈ W U gives a conformal equivalence between U and W

3. Conformal equivalence allows to transfer the holomorphic functions on one set to

the holomorphic functions on the other set.

Corollary 3.4. If f ∈ V U is a conformal equivalence, then the map

T : H (V ) → H (U )
ϕ 7→ ϕ ◦ f

where ϕ ∈ CV is a holomorphic function on V , is a liner isomorphism of vector

156 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

spaces with inverse

T −1 : H (U ) → H (V )
φ 7→ φ ◦ f −1

where φ ∈ CU is a holomorphic function on U , i.e. T is an isomorphism of vector

spaces, so it holds that

T (aϕ1 + bϕ2 ) = aT (ϕ1 ) + bT (ϕ2 )

for a, b ∈ C and ϕ1 , ϕ2 ∈ H (V )

Example 3.25 (The disc and the upper half plane). Let H := {z ∈ C : Im(z) > 0}
be the upper half plane and D := D1 (0) = {z ∈ C : |z| < 1} be the unit disc. Then the
map
f :H→D
z−i
z 7→
z+i
is a conformal equivalence, so
1+w
f −1 (w) = i
1−w

This example shows that the property that a set is bounded is not preserved under
conform equivalence.
Proof of the Example 3.25. First note that for any z ∈ H it holds:
z−i
|f (z)| = <1
z+i
since the distance from z to i is shorter than the distance from z to −i, which is in the
lower half plane.
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 157

Moreover, f is clearly holomorphic, since ∀z ∈ H : z + i ̸= 0. Similarly, the map

g : D → H, z 7→ g(w) := i 1+w
1−w
is holomorphic.

Lastly, to see that g(w) ∈ H for any w ∈ D, we look at

1+w
− i 1+w





i 1−w 1−w
im g(w) = =
 2i   
1 1 + w 1 + w̄ 1 (1 − w̄)(1 + w) + (1 − w)(1 + w̄)
= + =
2 1 − w 1 − w̄ 2 |1 − w|2
1 − |w|2
= > 0 , since |w| < 1
|1 + w|2

Hence g indeed goes from D to H. Finally a direct calculation verifies that f (g(w)) = w
and (g ◦ f )(z) = z and so g = f −1

Note that the map f from Example 3.25 takes the real line to the boundary of the disc
with f (0) = −1, f (1) = −i and f (”∞”) = 1

Example 3.26 (The map z 7→ z 2 ). Let U := z ∈ C : Arg(z) ∈ 0, π2 , so



f :U →H
z 7→ z 2

maps the first quadrant to H

In that regard, the map

g:H→U
 
1 1
z 7→ z = exp
2 Log(z)
2

is its inverse.
158 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

• (f is injective) Let z12 = z22 , then z1 = ±z2 and only one of z2 , −z2 can be in U .
Since z1 , z2 ∈ U we have that z1 = z2

• (f is surjective) Let w = reiθ with θ ∈ (0, π), so w ∈ H, then z 2 = w has 2

solutions, namely
1 1 iθ
z1,2 = ±w 2 = ±r 2 e 2
1 iθ
and z = r 2 e 2 is in U

In general, let n ∈ N∗ and let the sector Sn = {z ∈ C : Arg(z) ∈ 0, πn }, then the map

f : Sn → H
z 7→ z n

with inverse

f −1 : H → Sn
 
1 1
w 7→ w = exp
n Log(w)
n

Example 3.27. Any horizontal strip of length 2π is conformally equivalent to a cut

plane (slit plane). The map

f : H → C−
z 7→ −z 2

maps H to C− := C \ (−∞, 0] and the map

f˜ : H → C+
z 7→ z 2

maps H to the slit plane cut of the positive reals C+ := C \ [0, +∞).
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 159

In particular, for V = C+

with a fixed branch of logarithm on C+ , such that logC+ (−1) = iπ

Remark 3.23. It is to notice that conformal inequalities do not preserve the bounded-
ness property of sets.
Example 3.28 (Important non-example). Let U := C and V := D, then there is no
biholomorphic map between U and V , since if there were such a map
f :C→D
which is holomorphic. Then f would be bounded, since f (z) < 1. Hence by Liouville’s
Theorem 2.8, it is constant, hence is not injective. Hence
C ̸∼c D

Riemann’s Theorem 3.21 says that any simply connected domain U , which is a proper
subset of C, i.e. U ̸= ∅ and U ̸= C, is conformally equivalent to D.

This leads to the following important Theorem:

160 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Theorem 3.21 (Riemann mapping Theorem). [SS10, Theorem VIII.3.1] Suppose

Ω ⊂ C is proper (i.e. ∅ =
̸ Ω=
̸ C) and simply connected. Then

Ω ∼c C

and if z0 ∈ Ω, then there is a unique conformal equivalence F ∈ H (Ω, D), such

that F (z0 ) = 0 and F ′ (z0 ) ∈ (0, +∞) ⊂ C

From which follows that

Corollary 3.5. Any two proper simply connected open subsets of C are conformally
equivalent.

Remark 3.24. Riemann’s mapping Theorem is remarkable: it classifies all simply

connected open subsets Ω ⊆ C, up to conformal equivalence. There are three of them,
namely ∅, C and D

The proof is though not constructive, as we will see. In general, it is not easy to find
an explicit map. During the rest of the course we will prove this Theorem. The strategy
of he proof is as follows:

1. (Uniqueness) This is going to be easy. It boils down to finding all automorphisms

of the unit disc, since if we have two conformal equivalences

f1 : Ω → D
f2 : Ω → D

then
f2 ◦ f1−1 : D → D
is an automorphism of D

2. If ∅ =
̸ Ω ̸= C, we will show that there is a conformal map f : Ω → D with
f (z0 ) = 0. Hence Ω is conformally equivalent to an open subset of D, hence

Ω ∼c f (Ω) ⊆ D

3. The second step shows that the set F = f ∈ DΩ : f is conformal and f (z0 ) = 0 ̸=
∅. We will see that s := supf ∈F f ′ (z0 ) exists and we will show that ∃f ∈ F
such that f ′ (z0 ) is maximal, i.e. the supremum s is taken. This f has “maximal
expansion speed”.
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 161

4. The f we found in the third step is surjective.

If this is the case, writing f ′ (z0 ) = seiθ and g(z) = e−iθ f gives the map we are
looking for, namely g ∈ DΩ with g (z0 ) = e−iθ f (z0 ) = 0 and g ′ (z0 ) = s > 0

Step 1: Automorphism and uniqueness

For the (conformal) automorphisms of D we have

Theorem 3.22. [SS10, Theorem VIII.2.2] If f ∈ AutH (D) is a (conformal) auto-

morphism of D, then
α−z
∃θ ∈ R∃α ∈ D : f (z) = eiθ
1 − ᾱz
satisfying

f (0) = eiθ α
f ′ (0) = eiθ |α|2 − 1

Conversely, every map of this form is a (conformal) automorphisms of D

Remark 3.25. 1. Note that an immediate Corollary of Theorem 3.22 is that the
only automorphisms of D that fix 0 are rotations, since

f (0) = eiθ α = 0 =⇒ α = 0

=⇒ f (z) = −eiθ z = eiθ̃ z

for some θ̃ ∈ R

2. This Theorem 3.22 is enough to prove the uniqueness of conformal equivalence

f ∈ DΩ

Proof of the uniqueness of F in Theorem 3.21. If f1 , f2 are two such maps with f1 (z0 ) =
f2 (z0 ) = 0 and f1′ (z0 ) , f2′ (z0 ) > 0, then g := f2 ◦ f1−1 : D → D is an automorphism of
D. Hence, by Theorem 3.22
α−z
g(z) = eiθ
1 − ᾱz
for some θ ∈ R and α ∈ D. Since f1 (z0 ) = 0, f2 (z0 ) = 0 we have g(0) = f2 ◦ f1−1 (0) =

0, so α = 0 and g(z) = −eiθ z for z ∈ D and g ′ (z) = −eiθ . Then

′ 1
−eiθ = g ′ (0) = f2′ f1−1 (0) · f1−1 (0) = f2′ (z0 )


f1′ (z0 )
162 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Hence
f2′ (z0 )
= −eiθ > 0
f1′ (z0 )
since fk′ (z0 ) > 0 for k ∈ {1, 2}. It follows that −eiθ ∈ R>0 and hence that θ = π + 2πk
as eiθ = −1. Also, α = 0, so we can conclude that
g(z) = z ⇒ f1 = f2

The proof of Theorem 3.21 uses a simple, but important Lemma:

Lemma 3.4 (Schwarz). [SS10, Lemma VIII.2.1] Let f ∈ H (D, D) with f (0) = 0.
Then

(i) ∀z ∈ D : |f (z)| ≤ |z|

(ii) If for some z0 ̸= 0, we have f (z0 ) = |z0 |, then f is a rotation.

(iii) f ′ (0) ≤ 1 and equality holds if and only if f is a rotation, i.e. ∃θ ∈ R : f (z) =
eiθ z

Remark 3.26. 1. (ii) and (iii) give conditions on f , so that up to a rotation, f is

the identity map. Since we assume that f (0) = 0, the condition in (ii) about the
f (z0 ) is true ( f (0) = |0|) for z0 = 0, but we cannot conclude from it that f
is a rotation. (iii) is the necessary condition at 0 to conclude that f is a rotation
(i.e. f ′ (0) = 1).
2. This Lemma 3.4 is once again a statement for holomorphic functions. One cannot
conclude for a real differentiable function f : D → D with f (0) = 0 any of
(i),(ii),(iii).
Proof of the Lemma 3.4. It is a consequence of the Maximum Modulus Principle (The-
orem 3.13).
(i) The assumption f (0) = 0 implies that ord0 (f ) ≥ 1, so we can define
f (z)
g : C → C, z 7→ g(z) :=
z
for z ∈ D. Since ord0 (f ) ≥ 1 and ord0 (z) = 1, in fact g has a removable singular-
ity at z = 0, so g ∈ H (D, D)


Fix z ∈ D and let r ∈ |z|, 1 . For w ∈ Cr (0) we have, since f (w) < 1 on D,
that
1 1
|g(z)| ≤ max g(w) = max f (w) ≤
w∈Cr (0) r w∈Cr (0) r
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 163

This holds for all z ∈ Dr (0). By the Maximum Modulus Principle 3.13, we have
1
∀z ∈ Dr (0) : g(z) ≤
r
(the holomorphic function g cannot attain a maximum in Dr (0))

This is true for all z ∈ D such that |z| < r < 1, then by letting r → 1 it follows
that
|g(z)| ≤ 1
and hence
∀z ∈ D : f (z) ≤ |z|
(ii) We proved that (i) gives supz∈D g(z) ≤ 1, but the assumption f (z0 ) = |z0 |
for some z0 ∈ D \ {0} implies that g has a local maximum at z0 ∈ D. By the
Maximum Modulus Principle 3.13 this can only happen if g is constant, hence
∃c ∈ C∀z ∈ D : f (z) = zg(z) = cz
Since f (z0 ) = |z0 | for that z0 ∈ D \ {0}, it follows that |c| = 1. Hence, c = ei θ
for some θ ∈ R and f (z) = eiθ z
(iii)
f (z) f (z) − f (0)
g(0) = lim = lim = f ′ (0)
z→0 z z→0 z−0
so f ′ (0) = g(0) ≤ 1. If f ′ (0) = 1, then again 0 is a local maximum of g and
we conclude as in (ii) that f (z) = eiθ z for some θ ∈ R, namely a rotation.

We can now give the proof of classification of AutH (D)

Proof of Theorem 3.22. First note that any function φα : D → D of the form
α−z
z 7→ φα (z) =
1 − ᾱz
for α ∈ C with |α| < 1, is an automorphism of D. This since:
1. Since |α| < 1, then 1 − āz ̸= 0 for z ∈ D (as |z| < 1), so φα ∈ H (D, D)
2. φα is injective:
φα (z) = φα (w)
α−z α−w
=
1 − ᾱz 1 − ᾱw
α − |α| w − z + ᾱzw = α − |α|2 z − w + ᾱzw
2

1 − |α|2 z = 1 − |α|2 w

z=w
Hence φα is a conformal map φα : D → D
164 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

3. φα (D) ⊆ D: This point might look superfluous, but checking this condition guar-
antees that the map is well-defined on its domain of definition, as any argument
will have an image in the codomain. So if |z| = 1, then z = eiθ and

α − eiθ α − eiθ
 
−iθ w
iθ
= e−iθ


φα e = iθ −iθ =e −iθ
e (e − ᾱ) e − ᾱ −w̄

with w := α − eiθ . Hence

w
φα eiθ = e−iθ


=1
−w̄

By the Maximum Modulus Principle 3.13, we have that ∀z ∈ D : φα (z) < 1 (not
being φα (z) a constant map, it cannot have a local maximum inside of D).

4. We have that
α−z
α− 1−ᾱz α − |α|2 z − α + z (1 − |α|2 ) z
(φα ◦ φα ) (z) = α−z

= = =z
1 − ᾱ 1−ᾱz
1 − ᾱz − |α|2 + ᾱz 1 − |α|2

Hence, φα is its own inverse.

Clearly any rotation R : D → D, z 7→ R(z) = eiθ z is also an automorphism of D

Hence
α−z
(R ◦ φα ) (z) = eiθ
1 − ᾱz
is on automorphism of D

Now, let f be any (conformal) automorphism of D, then ∃!α ∈ D : f (α) = 0.

Consider g = f ◦ φα with g ∈ DD , then g(0) = f (α) = 0

Schwarz Lemma 3.4 (a) applied to g gives

∀z ∈ D : g(z) ≤ |z|

Since g −1 (0) = 0, we can also apply the Schwarz Lemma 3.4 to g −1 and get

∀w ∈ D : g −1 (w) ≤ |w|

Using this for w = g(z) gives

∀z ∈ D : |z| = g −1 g(z) ≤ g(z)

3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 165

Combined with g(z) ≤ |z| we get that |g(z)| = |z|. Once again by Schwarz
Lemma 3.4 (b), g(z) = eiθ z is a rotation with some θ ∈ R. Hence eiθ z =
(f ◦ φα )(z) = g(z)

Replacing z with φα (z) now gives

eiθ φα (z) = g φα (z) = (f ◦ φα ) φα (z) =





= (f ◦ φα ◦ φα ) (z) = f (φα ◦ φα ) (z) = f (z)
using the fact that φα ◦ φα = id

Remark 3.27. Combining automorphisms of D together with the Cayley map

F :H→D
z−i
z 7→
z+i
allows one to find all automorphisms of H in the following manner.

Theorem 3.23. [SS10, Theorem VIII.2.4] Every automorphism g ∈ AutH (H) is of

the form
az + b
g(z) =
cz + d
 
a b
for ∈ GL2 (R) such that ad − bc > 0
c d

Remark 3.28. It is of interest that one can see that AutH (H) can be described via the
action of SL2 (R) ↷ H via fractional linear transformations.

Remark 3.29. Being these maps invariant under re-scaling by a real factor, we dis-
cover that these (conformal) automorphism are in fact represented in GL2 (R) quotient
by R∗ , namely the projective general linear group PGL2 (R), but only considering
those elements with positive determinant, hence a subgroup of it. These elements form
PSL2 (R), the projective special linear group, this being the quotient of SL2 (R) and
{±1}, and can be seen as a subgroup of the former (of the orientation-preserving trans-
formations).
Over R these two groups are different, in particular the second one being a strict sub-
group of the first one, but for instance over C surprisingly
PGL2 (C) ∼
= PSL2 (C)
This is going to have a greater relevance later in geometry, in particular in the context
of Möbius Transformations.
166 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Proof. Exercise: read in the book.

Note that using the map

γ : Aut(D) → Aut(H)
φ 7→ γ(φ) = F −1 ◦ φ ◦ F

any automorphism of D is lead to an automorphism of H. Moreover, γ is an isomorphism

with inverse

γ −1 : Aut(H) → Aut(D)
β 7→ γ −1 (β) = F ◦ β ◦ F −1

Using γ we can pull the automorphisms of D to automorphisms of H and show that

they are of the above form.
Now, we move to step two in the proof of the Riemann mapping Theorem 3.21.

Step 2: There is a conformal map f ∈ DΩ

(i.e. if Ω is a proper, simply connected of C, then it is conformally equivalent to a subset
of D)

We have the following

Proposition 3.9. [SS10, Step 1 in Section VIII.3.3, p.228] Let Ω ⊂ C such that
∅≠ Ω ̸= C, open and simply connected. Then there exists a conformal map f ∈ DΩ
such that 0 ∈ f (Ω), i.e. Ω is conformally equivalent to a subset of D, which contains
the origin.

Proof. Without loss of generality we assume that Ω ⊂ C∗ , hence that 0 ∈/ Ω. In fact,

being Ω proper, ∃α ∈ C : α ∈ / Ω, by replacing Ω with Ω − α := {z − α : z ∈ Ω}, we
can assume that α = 0 ∈ / Ω. Hence Ω ⊂ C∗ ; furthermore, since Ω is simply connected,
there exists a branch of logarithm logΩ ∈ H (Ω)

Note that logΩ is also injective, since if

logΩ (z) = logΩ (w)

then exponentiating both sides, we get that

z = exp logΩ (z) = exp logΩ (w) = w

and hence logΩ is a conformal map.

3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 167

Now let w ∈ Ω, then note that for any z ∈ Ω

logΩ (z) ̸= logΩ (w) + 2πi
otherwise, exponentiating we would get



z = exp logΩ (z) = exp logΩ (w) exp(2πi) = w
Hence z = w, but then logΩ (z) = logΩ (w) results in a contradiction. In fact, logΩ (z)
stays away from logΩ (w) + 2πi in the sense that


∃δ > 0 : D2δ log(w) + 2πi ∩ logΩ (Ω) = ∅
∗
Indeed otherwise, if for all n ∈ N∗ , say with δn = n1 , we got a sequence (zn )n∈N∗ ∈ ΩN
such that

1
logΩ (zn ) − logΩ (w) + 2πi <
n
Hence
n→∞
logΩ (zn ) −−−→ logΩ (w) + 2πi
and exponentiating and using the fact that exp is continuous, we would get that
n→∞ n→∞
zn −−−→ w and hence logΩ (zn ) −−−→ logΩ (w), using the continuity of logΩ , which
n→∞
is a contradiction to logΩ (zn ) −−−→ logΩ (w) + 2πi

Now, we can consider the map (for the same w as before)

F :Ω→C
1
z 7→

logΩ (z) − logΩ (w) + 2πi
Note that F ∈ H (Ω), since
∀z ∈ Ω : logΩ (z) ̸= logΩ (w) + 2πi
Since logΩ is injective, so is F and hence F is a conformal map. Furthermore, the above
estimate gives


∀z ∈ Ω : logΩ (z) − logΩ (w) + 2πi ≥ 2δ
Hence
1 1 1
∀z ∈ Ω : F (z) − 0 = ≤ <
logΩ (z) − (logΩ (w) + 2πi) 2δ δ
and so F (Ω) ⊂ D 1 (0). We can now translate and rescale F to obtain a function f ∈ DΩ
δ
which contains the origin in
its image.
Let f (z) := 4δ F (z) − F (w) , then f ∈ CΩ is conformal, as we have f (w) = 0 and
 
δ 1 1 1
∀z ∈ Ω : |f (z)| ≤ + ≤
4 δ δ 2
Therefore f (Ω) ⊂ D for all z ∈ Ω and since f (w) = 0, i.e. 0 ∈ f (Ω)
168 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Step 3: An extremal problem

Let Ω be a proper, non-empty, simply connected subset of C and z0 ∈ Ω. By Step 2 we
have at least one f ∈ DΩ such that f (z0 ) = 0. Let

F := f ∈ DΩ : f is conformal and f (z0 ) = 0



Then F ̸= ∅. We start by the following (see Step 2 in Section 3.3 in Chapter 8 in

[SS10]).

Lemma 3.5. The set of values f ′ (z0 ) ∈ R≥0 : f ∈ F


is bounded in [0, +∞).
Therefore, it exists
s := sup f ′ (z0 ) < +∞
f ∈F

Proof. Let δ > 0, such that D2δ (z0 ) ⊆ Ω and let f ∈ F. The Cauchy Integral Formula
2.6 gives Z
′ 1 f (z)
f (z0 ) = dz
2πi Cδ (z0 ) (z − z0 )2
Hence, using the standard estimate one finds that

1 f (z) 1
f ′ (z0 ) ≤ 2πδ max 2
≤
2π z∈Cδ δ δ
since f (z) ≤ 1 for all z ∈ Ω. Hence f ′ (z0 ) is bounded by 1δ for an arbitrary f ∈ F,
thus there exists an upper bound and by consequence of it also the supremum.
The next Proposition is key and states that the supremum

s = sup f ′ (z0 )
f ∈F

is taken.

Proposition 3.10. ∃f ∈ F : f ′ (z0 ) = s, hence s = maxf ∈F f ′ (z0 )

The proof of this Proposition 3.10 uses a compactness argument which we will come back
to, but we first see why this is key, in the sense that it gives the conformal equivalence
that we are looking for, between Ω and D (see Step 3 in Section 3.3 in [SS10], p.231).
Remark 3.30. The step 2 shows that Ω is conformally equivalent to an open subset of
D, which contains 0
Why are we looking for an extremal function that realises the extremal value s =
supf ∈F f ′ (z0 ) ?
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 169

We can assure without loss of generality that Ω is an open subset of D that contains 0,
so we can assume that z0 = 0

We want to conformally stretch Ω to fill D

F = f ∈ DΩ : f ∈ H (Ω) is injective and f (0) = 0



We want to choose a function in F with “maximal expansion”, but what does “expand-
ing” mean? Consider
f (0) = 0 =⇒ f (z) ∼ f ′ (0)z
for z near 0, so if f ′ (0) > 1, we say that f is expanding, since the distances between
nearby points are expanding

f (z1 ) − f (z2 ) ≈ f ′ (0) z1 − z2 > |z1 − z2 |

Step 4: f from the key Proposition in Step 3 is surjective

Proposition 3.11. Let f ∈ F be such that f ′ (z0 ) = s, then f ∈ DΩ is a conformal

equivalence (i.e. f is also onto D).

Proof. We want to show that f is a surjection: we assume that it is not, then ∃α ∈ D.

which is not in f (Ω). We are going to construct g ∈ F with g ′ (z0 ) > f ′ (z0 ) which
is going to be a contradiction to

f ′ (z0 ) = s = sup g ′ (z0 )

g∈F
170 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

To do this we are going to use φα and the square root map. Let

φ = φα : D → D
α−z
z 7→
1 − ᾱz
be the automorphism of D, with φα (0) = α and φα (α) = 0

Then φα ◦ f
: Ω → D is conformal and 0 ∈ / (φα ◦ f ) (Ω), since if for some z ∈ Ω held

φα ◦ (f (z)) = 0, then we would have f (z) = α, which we assured is not the case.

Since 0 ∈
/ (φ ◦ f )(Ω), and Ω is simply connected, a logarithm and a square root of of
φ ◦ f exist, i.e.
∃f˜ ∈ H (Ω)∀z ∈ Ω : f˜2 (z) = (φ ◦ f )(z)
′
α ◦f )
One can simply take g̃ as primitive of (φ


(φα ◦f )
, so that exp g̃(z) = (φα ◦ f )(z)

Note that f˜ is also injective: if f˜(z) = f˜(w), then (φ ◦ f )(z) = (φ ◦ f )(w). Since φ ◦ f
is conformal and f, φ are injective, we have that z = w

Now, f˜ is not yet the function we want, since f˜ (z0 ) ̸= 0, as φ f (z0 ) =

̸ 0, due to the
fact that 0 ∈
/ (φ ◦ f )(Ω)

Let f˜ (z0 ) = β and consider the (conformal) automorphism of D

φβ : D → D
β−z
z 7→
1 − β̄z

with φβ (β) = 0. Finally, let g(z) := φβ ◦ f˜ ∈ DΩ . Then g (z0 ) = 0 and g ∈ H (Ω, D),
since f˜ ∈ H (Ω, D) and φβ ∈ AutH (D)
Moreover, g is injective, since φβ and f˜ are injective, also g ∈ F by definition.
Claim. g ′ (z0 ) > f ′ (z0 )
This will give the contradiction that we are looking for.
Proof of the Claim. Recall: we first looked at φα ◦ f as
f φα
φα ◦ f : Ω →
− D −→ D
√
Then we took the · function, call it h

h : (φα ◦ f ) (Ω) → D
 
1
w 7→ exp w
2
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 171

and composed it with φα ◦ f as follows

f˜ := h ◦ φα ◦ f : Ω → D

so that f˜2 = φα ◦ f
Then we composed it with φβ , to get g ∈ DΩ such that

g := φβ ◦ f˜ = φβ ◦ h ◦ φα ◦ f
| {z }
f˜

Being these conformal equivalences, we have that φ−1 ˜

β ◦g = f

2
⇒ φ−1
β ◦g = φα ◦ f
−1

2
⇒ φ−1
α ◦ φβ ◦ g =f

Let s : D → D, z 7→ s(z) = z 2 be the squaring map. Then define Φ such that

−1
f = φ−1
α ◦ s ◦ φβ ◦g = Φ ◦ g
| {z }
=:Φ

Note that Φ is not injective. Now Φ ∈ H (D, D) by composition and since φ−1
β (0) = β

−1
Φ(0) = φ−1 (0) = φ−1 β2



α ◦ s ◦ φβ α

 2
˜ 2 ˜
But recall that f (z0 ) = β, it follows that β = f (z0 ) = (φα ◦ f ) (z0 ). Hence

Φ(0) = φ−1


α ◦ φα ◦ f (z0 ) = f (z0 ) = 0

Hence, we can apply Schwarz Lemma 3.4 (iii) to get Φ′ (0) < 1 (note that |Φ′ (0)| = ̸ 1,
iθ
since if it were so, then Φ(z) = e z for some θ ∈ R and it would mean that Φ is
injective, but Φ cannot be such, since the squaring function is not injective). Using the
chain rule applied to f = Φ ◦ g we have

f ′ (z0 ) = Φ′ g (z0 ) · g ′ (z0 ) = Φ′ (0) · g ′ (z0 )

f ′ (z0 ) = Φ′ g (z0 ) · g ′ (z0 ) = Φ′ (0) · g ′ (z0 ) < g ′ (z0 )

Hence f ′ (z0 ) < g ′ (z0 ) , which is a contradiction.

This concludes the proof.

172 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

Proof of Proposition 3.10 in Step 3

1. Existence of the maximum: We want to prove the existence of f ∈ F, such
that f ′ (z0 ) = s = sup {g ′ (z0 ) : g ∈ F }

Recall: For any bounded subset u ⊆ R there is a non-decreasing sequence

∗
(an )n∈N∗ ∈ U N such that limn→∞ an = sup(U )

∗
Recalling the definition of supremum, we take a sequence (fn )n∈N∗ ∈ (F)N with
n→∞
fn′ (z0 ) −−−→ s. We want to show that this sequence has a limit f in F

Note that the proof will not be constructive and will only guarantee the existence
of a limit f ∈ F

2. Recall: We have seen that a sequence of holomorphic functions that converge

uniformly on compact sets has a holomorphic limit (Theorem 2.14), but we can-
not expect that an arbitrary sequence (fn )n∈N∗ to be uniformly convergent on
compact sets. May it be that a subsequence has this property?

Recall: In the finite dimensional vector space Rn with n ∈ N, every bounded

sequence has a convergent subsequence.

So we are looking for an analogue of this for F. This is provided in the following
by Montel’s Theorem.

Theorem 3.24 (Montel’s Theorem). [SS10, Theorem VIII.3.3] Let Ω ⊆ C

N∗
open and (fn )n∈N∗ ∈ H (Ω) . Suppose that

∀K ⊆ Ω compact ∃MK ≥ 0∀n ∈ N∗ ∀z ∈ K : fn (z) ≤ MK

N∗
Then ∃(fnk )k∈N∗ ∈ H (Ω) , a subsequence of (fn )n∈N∗ , which converges
uniformly on compact subsets of Ω.

∗
3. In application to Riemann’s Theorem 3.21 we have a sequence (fn )n∈N∗ ∈ F N ,
so that ∀z ∈ Ω∀n ∈ N∗ : |fn (z)| ≤ 1 (not only compact sets, as D is bounded).

∗
Hence we can apply Montel’s Theorem 3.24 to find a sequence (fnk )k∈N∗ ∈ F N
which converges uniformly on compact sets and this will give the limk→∞ fnk = f
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 173

(holomorphic) that we are looking for, provided that we can show that f ∈ F

We now need to argue for the injectivity of such f , to this purpose we use the
following result:
∗
Proposition 3.12. Let (fn )n∈N∗ ∈ F N be a sequence in F and suppose that
fn → f uniformly on any compact set K ⊆ Ω. Then, either f is constant or
f ∈ F. Moreover, for any z ∈ Ω

lim fn′ (z) = f ′ (z)

n→∞

Proof. Clearly, if fn → f uniformly on compact sets, then f ∈ H (Ω) and

limn→∞ fn′ (z) = f ′ (z) for z ∈ Ω, by Theorem 2.15.

We need to show that f (Ω) ⊆ D and that f is injective or constant.

Since for z ∈ Ω we have fn (z) < 1, we deduce that f (z) ≤ 1; if |f (z)| = 1

for some z ∈ Ω, then z would be a local maximum of |f |, which is impossible
by the Maximum Modulus Principle 3.13, unless f is constant. Otherwise indeed
f (Ω) ⊆ D

What is left to show is that f is injective or constant. For this we have

Lemma 3.6. [SS10, Lemma VIII.3.5] Let Ω ⊆ C open and connected. (fn )n∈N∗ ∈

N∗
DΩ conformal. If fn → f ∈ CΩ uniformly on compact sets, then f is either
injective or constant.

Proof. We will suppose that f is not injective and show that then f is constant.
Suppose that for z1 ̸= z2 ∈ Ω : f (z1 ) = f (z2 ). If f is not constant, since the
zeroes of holomorphic functions are isolated, we can find a disc Dδ (z2 ) ⊆ Ω, so
that f (z) − f (z2 ) ̸= 0 in Ḋδ (z2 ). Hence, in particular
∀z ∈ C δ (z2 ) : f (z) − f (z2 ) ̸= 0
2

Note that this also says that z1 ∈

/ C δ (z2 ), since we assumed that f (z1 ) = f (z2 )
2

We apply the Argument Principle 3.10 to the function f (z) − f (z1 ) which has a
zero, namely z2 , in D δ (z2 ) to get
2

f ′ (z)
Z
1
dz ≥ 1
2πi C δ (z2 ) f (z) − f (z1 )
2
174 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

We have fn → f uniformly on compact sets, hence also on C δ (z2 ), in particular

2
fn (z) ̸= fn (z1 ) for all n ∈ N∗ and z ∈ C δ (z2 ), since the fn ’s are injective and
2
z1 ∈/ C δ (z2 ). Hence
2

fn′ (z) n→∞ f ′ (z)

−−−→
fn (z) − fn (z1 ) f (z) − f (z1 )

uniformly on C δ (z2 ). Therefore we get

2

f ′ (z) fn′ (z)

Z Z
1 1
dz = lim dz
2πi C δ (z2 ) f (z) − f (z1 ) n→∞ 2πi C δ (z2 ) fn (z) − fn (z1 )
2 2
| {z }
=0 , for all n∈N∗

since the integrals on the right counts the zeroes of the holomorphic function
fn (z) − fn (z1 ) in C δ (z2 ), so none by the injectivity of fn , but the integer on the
2
left is ≥ 1, which is a contradiction. Hence, f must be a constant.

∗
Remark 3.31. Finally note that in the case (fn )n∈N∗ ∈ F N with limn→∞ fn′ (z0 ) = s
and limn→∞ fn′ (z0 ) = f (z0 ) we have that

1. ∀n ∈ N∗ : fn′ (z0 ) ̸= 0, using Proposition 3.8 and that fn ’s are conformal (hence
∀z ∈ Ω : fn′ (z) ̸= 0).

2. By the definition of supremum,there exists a non-decreasing sequence

0 < f1′ (z0 ) ≤ ... ≤ fn′ (z0 ) ≤ ...

such that limn→∞ fn′ (z0 ) = s > 0. Hence f ′ (z0 ) ̸= 0 and f is not constant.

Proof of Montel’s Theorem 3.24

Montel’s Theorem 3.24 actually consists of 2 parts:

1. The first part is about the complex behaviour of sequences of holomorphic func-
tions, which says

(a) A sequence of holomorphic functions which is uniformly bounded on compact

sets K ⊆ Ω, i.e.

∀K ⊆ Ω compact ∃Mk > 0∀z ∈ K∀n ∈ N∗ : |fn (z)| ≤ Mk

is equicontinuous on compact sets.

3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 175

N∗
(b) A sequence (fn )n∈N∗ ∈ CΩ is equicontinuous on a compact set K ⊆
Ω, if

∀ε > 0∃δ > 0∀z, w ∈ K : |z − w| < δ =⇒ ∀n ∈ N∗ : fn (z) − fn (w) < ε

This is a complex behaviour in the sense that it is not true for sequences of real
functions. For example, fn (x) = sin(nx) on (0, 1) is uniformly bounded on com-
pact sets, but not equicontinuous.

Equicontinuity is a very strong condition and requires uniform continuity uni-

formly in the family.

The family (fn )n∈N∗ ∈ R[0,1] on [0, 1], given by fn (x) = xn , is not equicontinuous,
even though each

fn : [0, 1] → R
x 7→ xn

is uniformly continuous on [0, 1]. The family (fn )n∈N∗ is not equicontinuous. For
n→∞
example take any w ∈ (0, 1), then fn (1) − fn (w) −−−→ 1
2. The second part is known as Arzelà-Ascoli Theorem [EW22], which says that any
family F of functions, which is uniformly bounded and equicontinuous on com-
pact subsets of Ω, has a subsequence which converges uniformly on every compact
subset of Ω (the limit need not be in F).

This part belongs to Topology/Functional Analysis, hence will be assumed with-

out proof.

Theorem 3.25 (Arzelà-Ascoli). [EW22] Let K ⊆ Rn be compact and (fn )n∈N∗ ∈

N∗
C 0 (K; Rm ) a sequence of continuous functions on K. Suppose that

1. ∃x0 ∈ K∃M > 0∀n ∈ N∗ : fn (x0 ) ≤ M (i.e. fn (x0 ) n∈N∗ is bounded in

Rm ).

2. (fn )n∈N∗ is equicontinuous.

N∗
Then ∃(fnk )k∈N∗ ∈ C 0 (K; Rm ) , which converges uniformly on K to some con-
tinuous function f ∈ C 0 (K; Rm )

Assuming the Arzelà-Ascoli Theorem 3.25, the proof of Montel’s Theorem 3.24 reduces
to proving that every sequence of holomorphic functions, which is uniformly bounded
176 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA

on compact sets, is equicontinuous on compact sets.

This uses Cauchy’s Integral Formula 2.6 together with the following Lemma:

Lemma 3.7. [SS10, Lemma VIII.3.4] Let Ω ⊆ C be open. Then ∃ compact sets
{Kl }l∈N∗ such that

1. ∀l ∈ N∗ : Kl ⊆ int(Kl+1 )

2. Any compact set K ⊆ Ω is contained in Kl for some l ∈ N∗ . In particular

∞
[
Ω= Kl
l=1

(Such a sequence {Ke }∞

l=1 of compact subsets of Ω is called an exhaustion)

Proof. Exercise.

Now, assuming the Arzelà-Ascoli Theorem 3.25 and Lemma 3.7 we can give the proof
of Montel’s Theorem 3.24.

Proof of Theorem 3.24. Let (fn )n∈N∗ be a sequence of holomorphic functions, which are
uniformly bounded on compact sets. We want to show that (fn )n∈N∗ is equicontinuous.

1. Let K ⊆ Ω be compact, let r > 0 such that D3r (z) ⊆ Ω for z ∈ K (we can
choose r so that 3r is less than the distance from K to the boundary of Ω, i.e.
3r < d(K, ∂Ω)).

Let z, w ∈ K with |z − w| < r. The Cauchy Integral Formula 2.6 gives

Z  
1 1 1
fn (z) − fn (w) = fn (ξ) − dξ
2πi γ (ξ − z) (g − w)

where im(γ) = C2r (w). On C2r (w) we have

1 1 |z − w| |z − w|
− = ≤
ξ−z ξ−w |ξ − w||ξ − z| 2r · r

since |ξ − z| ≥ |ξ − w| − |w − z| ≥ 2r − r = r and |ξ − w| = 2r. Hence using the

standard estimate for the integral

1 1 |z − w|
fn (z) − fn (w) ≤ 2π(2r)|z − w| 2 Mk ≤ M
2π 2r r
3.9. CONFORMAL MAPS AND THE RIEMANN MAPPING THEOREM 177

since ∀ξ ∈ C2r (w)∀n ∈ N∗ : fn (ξ) ≤ M , where M in the uniform bound for all
fn ∈ F in D2δ (w)
So, for any ε > 0 we get

∀n ∈ N∗ ∀z, w ∈ K : fn (z) − fn (w) < ε

as soon as n εr o
|z − w| ≤ min r,
M
2. To extract a subsequence which converges uniformly on all compact sets we use
a standard trick, called the “diagonal argument”.

Let {Kl }l∈N∗ be the sequence of compact sets given by the last Lemma 3.7.
By the first step and the Arzelà-Ascoli Theorem 3.25, there is a subsequence of
(fn )n∈N∗ converging uniformly on K1 , say

(fn )n∈L1

Then there exists a subsequence of (fn )n∈L1 , with L1 ⊂ N infinite, converging

uniformly on K2 , hence on K1 and K2 , say (fn )n∈L2 L2 ⊂ L1 ⊂ N infinite. Induc-
tively, we get a subsequence (fn )n∈Lk converging uniformly on K1 , K2 , ..., Kk with
Lk ⊂ Lk−1 ⊂ ... ⊂ L1 ⊂ N

Now let

n1 := min {n ∈ N∗ : n ∈ L1 }
n2 := min {n ∈ N∗ : n ∈ L2 \ {n1 }} ∈ L2 ⊂ L1
n3 := min {n ∈ N∗ : n ∈ L3 \ {n1 , n2 }} ∈ L3 ⊂ L2 ⊂ L1
..
.

We get n1 < n2 < ... with nk ∈ Lk ⊂ Lk−1 ⊂ ... ⊂ L1

Note that L := {n1 , n2 , ...} has the property that L \ Lk is finite for each k

For each k ∈ N∗ , (fn )n∈L = (fnj )j∈N is up to finitely many terms (which has no
effect on convergence) a subsequence of (fn )n∈Lk

Hence (fnj )j∈N converges uniformly on every Kk , since any compact set K is
contained in Kk for some k ∈ N∗
178 CHAPTER 3. MEROMORPHIC FUNCTIONS AND RESIDUE FORMULA
Appendix A

The analytic continuation of the

Riemann Zeta Function

So far we have seen the following results:

1. We have seen
P∞ 1
Proposition A.1. [SS10, Proposition VI.2.1] The series ζ(s) = n=1 ns con-
verges absolutely and uniformly on every half plane

∀δ > 0 : Uδ := {s ∈ C : Re(s) ≥ 1 + δ}

and is holomorphic in {s ∈ C : Re(S) > 1}

2. We have also seen

P∞ iπn2 z
Proposition A.2. ∀z ∈ H : θ(z) := n=0 e converges and defines a
holomorphic function in H = {z ∈ C : Im(z) > 0}

We have seen that

3. RFor a function f ∈ CR , which is Riemann integrable on every [a, b] and for which
∞
−∞
f (t) dt converges, its Fourier transform is defined as
Z ∞
fˆ(ξ) := f (x)e−2πiξx dx
−∞

We also have shown that f (x) = e−πx has fˆ(ξ) = e−πξ . What we have not seen, but
2 2

can be proved, are the following results about Fourier Transform, which can be found

179
180APPENDIX A. THE ANALYTIC CONTINUATION OF THE RIEMANN ZETA FUNCTION

in Chapter IV in [SS10].

For a > 0, denote by Fa the class of all functions f that satisfy the following two
conditions:

1. f is holomorphic in the horizontal strip Sa := z ∈ C : Im(z) < a

2. It exists a constant A > 0 such that

A
∀x ∈ R∀y ∈ (−a, a) : f (x + iy) ≤
1 + x2
i.e. f is of moderate decay on each horizontal line Im(z) = y uniformly in y ∈
(−a, a)
2
Example A.1. f (z) = e−πz ∈ Fa for all a > 0. Let F := f ∈ CC : ∃a > 0 : f ∈ Fa


The Fourier Inversion says

Theorem A.1 (Fourier Inversion). [SS10, Proposition VI.2.2] If f ∈ F, then

Z ∞
∀x ∈ R : f (x) = fˆ(ξ)e2πixξ dξ
−∞

and (fˆ)(x) = f (−x)

c

The Poisson Summation says

Theorem A.2 (Poisson Summation Formula). [SS10, Proposition VI.2.4] If f ∈ F,

then X X
f (n) = fˆ(n)
n∈Z n∈Z

2
Corollary A.1. The Poisson Summation Formula applied to ft (x) = e−πtx , for
t ∈ R>0 gives
∞ ∞
2 −1 −πn2
X X
e−πtn = t2e t
n=−∞ n=−∞

Hence  
X
−πtn2 −1
X −πn2 1 1
ϑ(t) := e =t 2 e t =√ ϑ
n∈Z n∈Z
t t
 181

So  
1 1
ϑ(t) = √ ϑ
t t
and note that ϑ(t) = θ(it)

We can now use this transformation of the θ-function to give analytic continuation and
functional equation of ζ(s), namely

Theorem A.3. [SS10, Theorem VI.2.3] Let for Re(s) > 1 and
s
−s/2
Λ(s) := π Γ g(s)
2
Then Λ(s) has a meromorphic continuation to all of s-plane, with simple poles at
s = 0, 1 and satisfies the functional equation

Λ(1 − s) = Λ(s)

Recall: For Re(s) > 0 we have

Z ∞
dt
Γ(s) := e−t ts
0 t

Theorem A.4. Γ has analytic continuation to a meromorphic function on C, with

simple poles at s = 0, −1, −2, ... and residue at s = −n equal to

(−1)n
Res−n (Γ) =
n!

Proof of Theorem A.3. Idea: to relate Λ(s) and ϑ(t)
= θ(it) via an integral transform
and use the transformation property of ϑ(t) = √1t ϑ 1t inside the integral to analytically
continue Λ(s).
We start by collecting growth and decay property of ϑ(t), for example
−1 t→0
ϑ(t) ≤ Ct 2 −−→ 0
(follows from the functional equation) and v(t) − 1 ≤ Ce−πt for some C > 0 and for
all t ≥ 1. Since for t ≥ 1, we get
2
X X
2 e−πn t ≤ 2 e−πnt ≤ Ce−πt
n∈N∗ n∈N∗
182APPENDIX A. THE ANALYTIC CONTINUATION OF THE RIEMANN ZETA FUNCTION

The relation between Λ and ϑ is now given by the fact that for any s ∈ C with Re(s) > 1
we have Z ∞
−s
s
s dt
Λ(s) = π Γ2 ζ(s) = ϑ(t) − 1 t 2
2 0 t
This is based on the simple observations that
R∞ 2 s −s −s
1. 0 e−πn t t 2 dtt = (πn2 ) 2 Γ(s) = π 2 Γ(s)n−s

2. ϑ(t) − 1 = 2 ∞ −πn2 t
P
n=1 e
R
Pthe estimates on ϑ(t) as t → 0 and as t → ∞, one can justify the change of
Using
and to get for Re(s) > 1
Z ∞ ∞ Z ∞ ∞
1
s dt X −πn2 t s dt −s
X 1
ϑ(t) − 1 t 2 = e t
2 = π Γ(s)
2 = Λ(s)
2 0 t n=1 0 t n=1
ns
Z ∞
1
s dt
Λ(s) = ϑ(t) − 1 t 2
2 0 t
Now, we will see that we can make sense of the right hand side for s ∈ C. Now

∀t ≥ 1 : ϑ(t) − 1 < e−πt

1 ∞
Z
dt
=⇒∀s ∈ C : (ϑ(t) − 1)ts/2 converges
2 1 t
R1
Hence, it defines an analytic function for all s ∈ C. On the other hand, for 21 0 ϑ(t) −

s −1 −1
1 t 2 dtt we use the functional equation ϑ(t) = t 2 ϑ 1t . This gives ϑ(t) − 1 = t 2 ϑ 1t −



−1 −1
1 = t 2 ϑ 1t − 1 + t 2 − 1. Hence

1
1 1 −1
Z Z      
1
s dt 1 −1 s dt
ϑ(t) − 1 t 2 = t2 ϑ − 1 + t 2 − 1 t2 =
2 0 t 2 0 t t
Z 1   Z 1
1 1 s dt
 Z
1 1 s−1 dt 1 s−1 dt
= ϑ −1 t 2 + t 2 − t2 =
2 0 t t 2 0 t 2 0 t
Z 1    " s−1 #1  s 1
1 1 s−1 dt 1 t 2 1 t2
= ϑ −1 t 2 + − =
2 0 t t 2 s−1
2
2 2s 0
0
1 1
Z    
1 s−1 dt 1 1
= ϑ −1 t 2 + −
2 0 t t s−1 s
1 du
Now we make the change of variables u = t
with u
= − dtt we get
Z 1     Z ∞
1 1 s−1 dt
1−s du
ϑ −1 t 2 = ϑ(u) − 1 u 2
2 0 t t 0 u
 183

Hence, overall we have

1 ∞
Z  Z ∞  dt 

dt 1 1 1
 s 1−s
Λ(s) = ϑ(t) − 1 = − + ϑ(t) − 1 t 2 + t 2
2 0 t 2 s−1 s 1 t
Note that the integral on the right defines an entire function for all s ∈ C due to the
exponential decay of
∀t ≥ 1 : ϑ(t) − 1 < Ce−πt
Hence Λ(s) has a continuation to all of the s-plane, which is holomorphic except for
the simple poles at s = 1, 0 with resides 1 and −1 respectively.

1 1
R∞
 s 1−s

dt
Both s−1 − s and 1 ϑ(t) − 1 t + t 2 2
t
are invariant under s 7→ 1 − s, hence

Λ(s) = Λ(1 − s)
−s
Λ(s) = π 2 Γ 2s ζ(s) has analytic continuation to all s ∈ C except for the simple poles

at s = 0 and s = 1

Since Γ 2s has poles at 2s = 0, −1, −2, ...; ζ(s) does not have a pole at s = 0 and must

vanish at s = −2, −4, −6, ...; since Λ(s) does not have poles at s = −2, −4, ...

These are called the trivial zeroes of ζ(s)

P∞ 1
For Re(s) > 1, we have that ζ(s) = n=1 ns also has an infinite product expansion,
called Euler product
Y 1
ζ(s) =
p primes
1 − p1s

which is an analytic statement of the Fundamental Theorem of Arithmetic: every

positive integer is a unique product of prime powers.

Due to the Euler product, we have that ζ(s) ̸= 0 for Re(s) > 1 and by the functional
equation, that ζ(s) ̸= 0 for Re(s) < 0

Definition A.1 (Riemann Hypothesis). If s ̸= −2, −4, ..., s ∈ C with ζ(s) = 0, then
Re(s) = 12

Theorem A.5 (Prime Number Theorem (PNT)). Let

π(x) := #{p prime : p < x}

184APPENDIX A. THE ANALYTIC CONTINUATION OF THE RIEMANN ZETA FUNCTION

Then
π(x)
lim x =1
x→∞
log(x)

The Prime Number Theorem is a consequence of

Theorem A.6. If Re(s) = 1, then ζ(s) ̸= 0

Appendix B

Other results

B.1 Substitution rule in complex line integration

A straightforward fact that is quite different from the case of real analysis is that,
once the reader shall have gained some experience in the practice of the calculation of
complex line integrals, it will be evident that the integration by substitution does not
work as easily as in the real case.

Example B.1. Consider the integral

Z
eiz dz = e2i − 1
ℓ[0,2]

on another hand one tries to approach it by naively setting w = iz and consequently

dw
dz
= i, the result would turn out to be
Z Z
w
e idw = i ew dw = i(e2i − 1)
ℓ[0,2i] ℓ[0,2i]

which leads to a different incorrect result.

Part of the reason for this limit to integration is linkable to the Homotopy Theorem
3.14, as a change of variable, once composed with the path, deforms it in possibly
non-viable ways. For instance, exiting the domain of the integrand function, crossing
a singularity, etc. To name one more, also changes in the “simplicity” of the curve can
have effects.


Theorem B.1. Let Ω ⊆ C be an open subset and let γ ∈ C 1 [a, b]; Ω be a path
in Ω. If ϕ : Ω → ϕ(Ω) is biholomorphic, i.e. is a bijection that is holomorphic and

185
186 APPENDIX B. OTHER RESULTS

with holomorphic inverse, then

Z Z
f ϕ(w) ϕ′ (w)dw


f (z)dz =
γ ϕ−1 ◦γ

A useful result to accompany this Theorem is the following Lemma, which states the
form that biholomorphisms need to assume from C to C, namely Affine maps, denoted
by AutH (C)

Lemma B.1. All biholomorphisms from C to C are affine maps, namely

∀f ∈ AutH (C)∃a, b ∈ C∀z ∈ C : f (z) = az + b

Proof. We want to show this in two steps, first showing that such a function is a
polynomial and second that its degree is 1
Let f ∈ AutH (C). First, the series expansion of f at 0 is
∞
X
f (z) = aℓ z ℓ
n=0

and converges over all of C, since f is entire. Composing f with the inversion η : C∗ →
C, z 7→ z1 it results that for
∞
X
g := f ◦ η : C∗ → C, z 7→ aℓ z −ℓ
n=0

we have that g(C∗ ) = f (C∗ ) = C \ {f (0)}. Now, if 0 were an essential singularity

of g, then g
Ḋ1 (0) = f C \ D1 (0) = C and since f is non-constant we have that
f C \ D1 (0) is open in C and so



f C \ D1 (0) ∩ f D1 (0) ̸= ∅
Hence, a contradiction to the bijectivity of f . So, 0 is either a pole of order N ∈ N∗ or
removable. In the latter case, g(z) = a0 = f (z) for all z ∈ C, otherwise
N
X
f (z) = aℓ z ℓ
n=0

for N as above. To prove that N = 1, assume that N ≥ 2, thus f (z) − a0 either

has multiple zeroes (infraction to the injectivity, as the preimage of 0 has more than
one element) of one with higher multiplicity. In this case, we again conclude that the
injectivity of f produces a contradiction. Therefore N = 1 and the first direction is
shown.
Conversely, every affine function is bijective and biholomorphic on C
B.2. L’HÔPITAL’S RULE IN C 187

There are many more (conformal) automorphism between two proper subset of C with
specific characteristics, these can be obtained using the Riemann mapping Theorem
3.21.

B.2 L’Hôpital’s Rule in C

Theorem B.2 (L’Hôpital’s Rule in C). Let a ∈ C and f, g ∈ H (Ua ) with Ua ∈ OC
and a ∈ Ua , such that f (a) = g(a) = 0, then

f (z) f ′ (z)
lim = lim ′
z→a g(z) z→a g (z)

Proof. Let f ∈ H (Ua ) and f ′ (a) = limz→a f (z)−f

z−a
(a)
= limz→a fz−a
(z)
, since f (a) = 0.
′ ′ ′
Hence, f is continuous and therefore limz→a f (z) = f (a), from which follows that
f ′ (z)
   
f (z) f (z) z − a f (z) z−a
lim = lim · = lim · lim = lim ′
z→a g(z) z→a z − a g(z) z→a z − a z→a g(z) z→a g (z)

B.3 Properties of the Residue

Proposition B.1. Let Ḋr (z0 ) ∈ OC for some r > 0 and f, g ∈ H Ḋr (z0 ) with

z0 ∈ C being a pole of both, then

(i) Resz0 (f + g) = Resz0 (f ) + Resz0 (g)

(ii) Resz0 (f g) has a finitary expression.

(iii) If the extension f˜ ∈ H Dr (z0 ) , then Resz0 (f˜′ ) = 0


N∗
(iv) If a sequence (fn )n∈N∗ ∈ H Ḋr (z0 ) is such that fn ⇒ f , then
n→∞

∞
! ∞
X X
Resz0 fn = Resz0 (fn )
n=0 n=0

Proof. (i) Assume that f and g are analytic in a punctured neighbourhood of z0 .

Then by definition of the residue one has
Z
1

Resz0 (f + g) = f (z) + g(z) dz
2πi γ
188 APPENDIX B. OTHER RESULTS

where γ is a sufficiently small circle with center z0 . By linearity of integrals, it

follows that
Resz0 (f + g) = Resz0 (f ) + Resz0 (g)

(ii) There is no “nice” formula for Resz0 (f g). However, when f and g just have poles
at z0 it is possible to compute Resz0 (f g) in a ”finitary” fashion from the Laurent
expansions of f and g as follows: assuming z0 = 0 for simplicity we have
∞
X
f (z) = ak z k
k=−m
X∞
g(z) = bℓ z ℓ
ℓ=−n

for n, m ∈ N; therefore
n−1
! m−1
!
X X
f (z)g(z) = ak z k bℓ z ℓ + h(z)
k=−m ℓ=−n

where h is holomorphic at 0. The residue of f g at 0 can now be extracted:

n−1
X
Res0 (f g) = ak b−k−1
k=−m

(iii) We have that for z in a punctured neighbourhood of z0

h(z)
f˜(z) =
z
for some holomorphic h in the same neighbourhood. Then

˜′ h′ (z) h(z) zh′ (z) − h(z)

f (z) = − 2 =
z z z2

(iv) Let ε > 0, when the series ∞

P
n=0 fn converges uniformly in annuli A(z0 , ε, 2ε),
then one has
∞
! Z X ∞
! ∞ Z ∞
X 1 X 1 X
Resz0 fn = fn dz = fn (z) dz = Resz0 (fn )
n=0
2πi γ n=0 n=0
2πi γ n=0

for any curve γ with im(γ) ⊆ A(z0 , ε, 2ε)

B.4. A WINDING NUMBER ALGORITHM 189

B.4 A winding number algorithm

To check if a given point lies inside or outside a polygon:

1. Draw a horizontal line to the right of each point and extend it to infinity.

2. Count the number of times the line intersects with polygon edges.

3. A point is inside the polygon if either count of intersections is odd or point lies on
an edge of the polygon. If none of the conditions are true, then point lies outside.

Furthermore, this algorithm can be generalized to an arbitrary direction of the ray,

preserving the sign of every crossing direction, hence left to right or right to left (in
the direction of the ray pointing to infinity). An improved version of this is already far
beyond the scope of this section and is known as “Dan Sunday’s RAY Algorithm”.
190 APPENDIX B. OTHER RESULTS
Change Log

23.12.2024 - Andreas Compagnoni - First version.

191
192 APPENDIX B. OTHER RESULTS
References

[SS10] E.M. Stein and R. Shakarchi. Complex Analysis. Princeton lectures in anal-
ysis. Princeton University Press, 2010. isbn: 9781400831159. url: https :
//books.google.ch/books?id=0ECHh9tjPUAC.
[CE16] R. Cavalieri and Miles E. Riemann Surfaces and Algebraic Curves: A First
Course in Hurwitz Theory. Cambridge University Press, 2016. url: https:
//www.cambridge.org/core/books/riemann-surfaces-and-algebraic-
curves/4D91A22369F62D519F43A8A0713E29D9#.
[EW22] M. Einsiedler and A. Wieser. Analysis I und II. ETH Zürich, 2022. url:
https://metaphor.ethz.ch/x/2023/fs/401- 1262- 07L/sc/Analysis-
Skript.pdf.
[Ima23] Ö. Imamoḡlu. Lecture Notes in Complex Analysis. ETH Zürich, 2023. url:
https://metaphor.ethz.ch/x/2023/hs/401-2303-00L/.
[Da 24] F. Da Lio. Analysis III. ETH Zürich, 2024. url: https://people.math.
ethz.ch/~fdalio/lecture-notes.

193