CA Lectures

Lectures on Complex Anal-
ysis
These are lecture notes based on courses on complex analysis offered at
UC Berkeley in Fall 2016 and IISc in the winter semester of 2020.
Date: 01 August 2021
Ved V. Datar
Department of Mathematics, IISc
vvdatar@iisc.ac.in
Indian Institute of Science

www.math.iisc.ac.in
Contents
I Complex differentiable functions 1
1 Introduction 3
1.1 Roadmap for the course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Conjugate, Absolute value, Argument. . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Lecture-2: Limits, continuity and holomorphicity 9

2.1 Topology of the complex plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Functions on the complex plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Path connected sets in C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Convergence of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Holomorphic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Power series 15
3.1 Infinite series of complex numbers: A recap . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Convergence of power series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3 Holomorphicity of power series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 The exponential functions 23

4.1 The exponential functional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Trigonometric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Periodicity and the definition of π. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.4 The logarithm function and complex powers . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4.1 Logarithm as power series. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.4.2 Complex powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5 Cauchy Riemann equations 29

5.1 A review of multivariable calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Cauchy-Riemann equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2.1 New notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6 Harmonic functions 35
6.1 Harmonic functions in the plane and holomorphic functions . . . . . . . . . . . . . . . . . 35
6.2 Mean value property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.3 The Poisson kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
i
ii CONTENTS
II Integration theory 37
7 Complex Integration 39
7.1 Curves in the complex plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7.1.1 Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7.1.2 Notation and conventions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.2 Complex line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.3 A fundamental computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
7.4 Primitives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.4.1 A differential forms interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
8 Cauchy’s theorem : Local versions 47

8.1 Proof of Cauchy’s theorem assuming Goursat’s theorem . . . . . . . . . . . . . . . . . . . 48
8.2 Proof of Goursat’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
8.3 Extensions to punctured domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
9 Cauchy’s integral formula 53

9.1 Cauchy integral formula and Analyticity of holomorphic functions . . . . . . . . . . . . . 53
9.2 Cauchy integral formula for derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
9.3 Liouville Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
10 Further applications of Cauchy integral formula 57

10.1 The fundamental Theorem of algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
10.2 Zeroes of a holomorphic function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
10.3 Sequences of holomorphic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
11 Cauchy’s theorem : Homology version 63

11.1 Index of a curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
11.2 The homology version of Cauchy’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 64
11.3 Simply connected domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
12 Cauchy’s theorem : Multiply connected domains 69

12.1 Characterisations of simply connected sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
12.2 Cauchy’s theorem for multiply connected domain . . . . . . . . . . . . . . . . . . . . . . . 71
12.3 A real variable integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
13 Logarithm and roots revisited 75

13.1 Criteria for existence of holomorphic logarithms . . . . . . . . . . . . . . . . . . . . . . . . 75
13.2 Some examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
13.2.1 Principal branch of the logarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
13.2.2 Branch cut for log (z2 − 1).
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
z+1
13.2.3 Branch cuts for log z−1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
th
13.3 n -roots of holomorphic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
13.3.1 Principal, and other
√ branches of the square root . . . . . . . . . . . . . . . . . . . . 80
13.3.2 Branch cuts for z 2 − 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
14 Isolated Singularities 83
14.1 Removable singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
14.2 Poles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
14.3 Essential singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
15 Laurent series 89
CONTENTS iii
15.1 Application to study of isolated singularities . . . . . . . . . . . . . . . . . . . . . . . . . . 92
16 Meromorphic functions 95
16.1 Definition and basic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
16.2 Partial fraction decomposition of meromorphic functions on C. . . . . . . . . . . . . . . . 96
16.3 Holomorphic self maps on the extended complex plane . . . . . . . . . . . . . . . . . . . . 98
17 The theorems of Mittag Leffler and Weierstrass 101

17.1 Proof of the Mittag-Leffler theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
17.2 Infinite products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
17.3 Weierstrass factorisation theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
18 The Residue theorem 107

18.1 The argument principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
18.1.1 Argument principle as an index calculation . . . . . . . . . . . . . . . . . . . . . . 110
19 Applications of the argument principle 111

19.1 Local mapping properties of holomorphic functions . . . . . . . . . . . . . . . . . . . . . . 111
19.2 The maximum modulus principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
19.3 Rouche’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
19.4 Appendix : details left out in the proof of Corollary 19.1.5 . . . . . . . . . . . . . . . . . . . 115
20 Contour Integration - I 117

20.1 Type-I: Integrals of rational functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
20.1.1 A variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
20.2 Type-II: Rational functions of sine and cosine . . . . . . . . . . . . . . . . . . . . . . . . . . 119
21 Contour Integration - II 123

21.1 Type-III(b): Principal values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
21.2 Type-IV: Products of rational functions and powers of x. . . . . . . . . . . . . . . . . . . . 125
21.3 A bonus integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
III Geometry of holomorphic maps 129

22 Conformality 131
22.1 Conformal maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
22.2 A survey of elementary mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
22.2.1 Rotations and dilations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
22.2.2 Complex powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
22.2.3 The logarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
22.2.4 The exponential map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
23 Mobius transformations 137

23.1 Mobius transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
23.1.1 The group P SL(2, C). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
23.1.2 The cross ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
23.1.3 The Cayley transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
24 Some automorphism groups 143

24.1 Biholomorphisms and automorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
24.2 Automorphism group of the disc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
24.3 Automorphism groups of C and C∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
iv CONTENTS
24.4 Automorphism group of the extended complex plane . . . . . . . . . . . . . . . . . . . . . 147
25 The Riemann mapping theorem 149

25.1 Montel’s and Hurwitz’s theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
25.2 Proof of Riemann mapping theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
25.2.1 Some further preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
25.2.2 Completion of the proof of Riemann mapping . . . . . . . . . . . . . . . . . . . . . 153
25.3 Green’s functions and a generalization of the Riemann mapping theorem . . . . . . . . . . 153
IV Special functions 157

26 Gamma and Zeta functions 159
26.1 Revisiting the principle of analytic continuation . . . . . . . . . . . . . . . . . . . . . . . . 159
26.2 Analytic continuation of the Mellin transform . . . . . . . . . . . . . . . . . . . . . . . . . 160
26.3 The Gamma function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
26.4 The Riemann Zeta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
27 Prime numbers and the Riemann hypothesis 169

27.1 The prime number theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
28 Elliptic functions 171

28.1 Doubly periodic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
28.2 The canonical basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
28.3 Poles and zeroes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
28.4 Weierstrass ℘ function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
29 Modular forms 173

29.1 The modular group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
V Analytic continuation 175

30 Analytic continuation along curves 177
30.1 An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
30.2 Formalizing the main ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
30.3 The monodromy principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
30.4 A (very) surprising application - Picard’s little theorem . . . . . . . . . . . . . . . . . . . . 178
31 An exposition on Riemann surfaces 179
A Problems 181
Part I
Complex differentiable functions
1
Lecture 1
Introduction
1.1 Roadmap for the course

Complex analysis is one of the most beautiful branches of mathematics; a subject that lies at the heart of
several other subjects, such as topology, algebraic geometry, differential geometry, harmonic analysis, and
number theory.
The main objects in calculus are real (scalar or vector) valued functions defined on domains in Rn . The
starting point in complex analysis is to consider complex valued functions
f :Ω→C
defined on subsets Ω ⊂ C of complex numbers. Recall that complex numbers can be added, subtracted,
multiplied and divided (if non-zero) just like real numbers. Every complex number can be written in the
form
z = x + iy
where x and y are real numbers. So complex numbers can be identified as a set with Euclidean plane R2 .
The addition of complex numbers is also equivalent to addition of vectors in R2 . So it might appear as if
we are not adding much, and that nothing is lost by simply treating the complex valued function as a two
variable vector field. In fact this is true, as we will see later, when talking about limits and continuity.
But there is one key difference between R2 and C, that of multiplication and division. Indeed things
change dramatically when we restrict our attention to complex differentiable or holomorphic functions, that
is, functions for which
f (z + h) − f (z)
lim
h→0 h
exists and is finite. The important point being that h could be a complex number. Formally this definition is
identical to that of a differentiable function in one-variable calculus. But quite surprisingly the mere change
of perspective, the fact that h is allowed to take complex values as it goes to zero, produces beautiful new
phenomenon that have no counterparts in one-variable calculus, or indeed even multivariable calculus.
We now summarize some of these remarkable consequences of holomorphicity.
• Analyticity. As we remarked above, complex valued functions can be thought as mapping between
sets in R2 . We will prove later in the course that for a holomorphic function, partial derivatives
of all orders exist. And moreover, one also has that the Taylor series at ever point converges to the
function value, that is holomorphic functions are analytic. Recall that this is not true for one-variable
3
4 LECTURE 1. INTRODUCTION
functions. For instance if ( 2

e−1/x , x > 0
f (x) =
0, x ≤ 0,
then it is easy to see that at x = 0, derivative of any order is zero. So the Taylor series of the function
at x = 0 is zero, but the function is clearly not zero.
• Analytic continuation. Two holomorphic functions defined on an open connected domain are
equal in a small open neighbourhood of a point, no matter how small the neighbourhood is, have to
be identically equal.
• Good convergence properties. If a sequence of holomorphic functions converges uniformly, the
limit function is again holomorphic. This is not true for differentiable one-variable functions. For
instance, if fn : [−1, 1] → R is defined by
r
1
fn (x) = + x2 ,
n
then one can show that fn → |x| uniformly, but |x| is not differentiable.
• Liouville property. A bounded holomorphic function defined on all of C is forced to be a constant.
As a consequence, one can prove the fundamental theorem of algebra.
Part of the richness of the theory of holomorphic functions comes from the variety in the methods used to
study the subject. We next summarize the approaches that we will touch upon in this course.
• Partial differential equations. It turns out that real and imaginary parts of holomorphic functions,
thought of as real valued two-variable functions, satisfy a system of first-order partial differential
functions, called the Cauchy-Riemann equations. As a consequence of this, the real and imaginary
parts are harmonic functions. The theory of harmonic functions is rather well developed, and could be
potentially exploited to study holomorphic functions. We will only touch upon the Cauchy-Riemann
equations, but will not pursue this approach further. We will instead focus on integral methods.
• Integral methods. The viewpoint that we will adopt is centered on a remarkable formula called
the Cauchy’s integral formula. We will develop a notion of integration of complex valued functions
along curves, a generalization of the notion of line integrals in multi-variable calculus. The funda-
mental fact, which will be the theoretical basis for the rest of the course, is that the complex integral
of a holomorphic function around a closed curve is zero. If the real and imaginary parts of the
holomorphic function are assumed to have continuous partial derivatives, this result follows from
Green’s theorem. We will give an independent proof, not because we wish to be clever, but because
remarkably this theorem will imply that the real and imaginary parts of the holomorphic function
indeed have not only continuous partial derivatives but have partial derivatives of all orders, and are
in fact analytic.
• Power series methods. As remarked above, every holomorphic function is represented by a power
series. Since power series are algebraic objects, for the most part they can also be manipulated as if
they were polynomials. Thus algebraic methods can be used to study holomorphic functions.
• Geometric mathods. An elementary but beautiful fact is that holomorphic functions, thought of
as mappings (or transformations) between sets in R2 are conformal maps. That is, holomorphic
mappings preserve angles between curves, and stretch the distances. We will study some standard
examples of conformal maps. Towards the end of the course we will prove the following deep fact,
first discovered by Riemann - Any domain in the complex plain which does not have a ‘hole’ and
which is not the entire complex plane, can be mapped conformally to a disc centered at the origin of
radius one. Our proof will be due to Koebe.
1.2. COMPLEX NUMBERS 5
1.2 Complex numbers

It is known that certain polynomial equations with real coefficients need not have real roots. Complex
numbers are obtained from the reals by formally adjoining a number i that solves the equation
i2 = −1.
More formally, we define the set of complex numbers by
C := R[i] = R[x]/(1 + x2 ).
So a general complex number takes the form z = x + iy, where x and y are real numbers, and are called
the real and imaginary part of z respectively. We use the notations
x = ℜ(z) and y = ℑ(z).
Clearly the real numbers can be identified as a subset of the complex numbers in a natural way as numbers
with ℑ(z) = 0. We define addition and subtraction to be component-wise i.e. if z1 = x1 + iy1 and
z2 = x2 + iy2 , then we define
z1 ± z2 = (x1 + x2 ) ± i(y1 + y2 ).
Using this, we can identify C with R2 as vector spaces. With this interpretation, a complex number rep-
resents a point in the xy-plane; with the x-coordinate given by ℜ(z) and the y-coordinate given by ℑ(z).
This more geometric interpretation will be very useful to us.
But the complex numbers are much more than just 2-dimensional vectors. They also have a multiplicative
structure, induced from the multiplicative structure of R[x]. That is we can multiply two complex numbers
to obtain another complex number. Indeed, if z1 and z2 are as above, we define
z1 · z2 = (x1 x2 − y1 y2 ) + i(x1 y2 + x2 y1 ).
More simply, we define i2 = −1, and then extend the product to satisfy the distributive property. It is not
hard to verify that addition and multiplication satisfy the following properties:
P1 (Additive and Multiplicative identity.) For any complex number z,
z + 0 = z, z · 1 = z.
P2 (Commutativity.) For any z1 , z2 ∈ C,
z1 + z2 = z2 + z1 , z1 · z2 = z2 · z1 .
P3 (Associativity.) For any complex numbers z1 , z2 , z3 ,
(z1 + z2 ) + z3 = z1 + (z2 + z3 ), (z1 · z2 ) · z3 = z1 · (z2 · z3 ).
P4 (Distribution) For any z1 , z2 , z3 ∈ C,
z1 · (z2 + z3 ) = z1 · z2 + z1 · z3 .
P5 (Additive inverse.) For any z ∈ C, −z = (−1) · z satisfies
z + (−z) = 0.
For notational convenience, we sometimes drop the dot when multiplying complex numbers. As remarked
above, geometrically, addition of complex numbers corresponds to addition of vectors. What is the inter-
pretation for multiplication? This is clearer if we use polar coordinates. Recall that any point (x, y) in the
plane that is not the origin, can be represented uniquely by a pair (r, θ), where r > 0 and θ ∈ (−π, π] via
the following transformation law:
x = r cos θ, y = r sin θ.
Then r is the geometric distance from the origin, and θ is the angle made by the line joining (x, y) to
the origin with the positive x-axis. For instance the complex number i corresponds to (1, π/2) in polar
coordinates. So, any complex number can be represented as
z = r(cos θ + i sin θ).
If w = ρ(cos α+i sin α) is another complex number, then it follows from the definition of the multiplication
formula that
zw = rρ[(cos θ cos α − sin θ sin α) + i(cos θ sin α + sin θ cos α)]

= rρ(cos(θ + α) + i sin(θ + α)),
where we used the sum-angle formulas in the last equation. So geometrically multiplication simply corre-
sponds to a dilation (i.e. scaling) and a rotation. For instance multiplication by i corresponds to rotating
the vector representing the complex number by π/2. To form a good number system we will also need to
be be able to divide by complex numbers. For any z2 ̸= 0, we say that w = z1 /z2 if z1 = wz2 . We call
w, the quotient obtained by dividing z1 by z2 . Clearly, if z2 = 0, then by property P1, the quotient cannot
be well defined. We next see that the quotient is in fact well defined when dividing by non-zero complex
numbers.
1.2.1 Conjugate, Absolute value, Argument.

To prove that a quotient always exists on dividing by a non-zero complex number, it is enough to have a
formula for 1/z, where z ̸= 0. Let z = x + iy. Multiplying the numerator and denominator of 1/z by
x − iy (this is similar to rationalizing irrational denominators) we obtain
1 1 x − iy x − iy
= = = 2 .
z x + iy (x + iy)(x − iy) x + y2
Note that in the last equation, the denominator is now a real number, and we already know how to divide
by real non-zero numbers. The numerator, is called the conjugate of z and is denoted by
z̄ = x − iy.
Geometrically this amounts to reflection of the point representing z about the x-axis. Readers will notice
that the denominator is the square of the distance of the point (x, y) from the origin. So we define the
absolute value or the length of the complex number, denote by |z| as
p
|z| = x2 + y 2 .
This is of course the ‘r’ in the polar coordinate representation. Some basic properties of these operations
are the following.
• z = z.
• |z| = |z|.
• |z| = 0 =⇒ z = 0.
1.2. COMPLEX NUMBERS 7
• zw = z̄ w̄, |zw| = |z||w|.

• z + w = z̄ + w̄.
Note that contrary to the conjugate function, the modulus function is not additive. Instead we have an
inequality; see Theorem 1.2.1. With these notations in place, we can re-write the above statement as
1 z̄
= 2,
z |z|
for any z ̸= 0. So we in summary we shown that multiplication has another property, that every no-zero
number has multiplicative inverse. That is we have
P6 (multiplicative inverse.) For every z ∈ C, z ̸= 0 there exists a complex number 1/z = z̄/|z|2 such
that
1
z · = 1.
z
With the two operations of addition and multiplication satisfying these six axioms, the set of complex
numbers become what is called as field by algebraists. In fact, the complex numbers form an algebraically
closed field, which means that any polynomial with complex coefficients can be completely factorized using
complex roots. Later in the course, somewhat remarkably, we will prove this statement in algebra using
our complex analysis techniques. We will in fact give multiple proofs, not just one!
The ‘θ’ in the polar coordinates also has a name, and is called the argument of z, and denote by arg(z).
Using the new notation, it is also easy to see that
z + z̄ z − z̄
ℜ(z) = , ℑ(z) = .
2 2i
We can now define division by
z z w̄
= .
w |w|2
Integer Powers. Given any natural number n ∈ N, we define z n to be z multiplied to itself ‘n’ times. We
also define z 0 := 1. For negative integers −n, we then define z −n to be 1/z n or the multiplicative inverse
to z n .
We end with an important inequality that will be crucial in most of the estimates.
Theorem 1.2.1 (Triangle inequality). Let z, w ∈ C. Then we have the following inequalities
1.
|z + w| ≤ |z| + |w|,
with equality if and only if z = aw where a ∈ R i.e. z and w lie on the same line through the origin.
2.
|z − w| ≥ ||w| − |z||,
with equality again if z and w lie on the same line through the origin.
Proof. 1. We first note that if x, y, u, v ∈ R, then

(xu + yv)2 ≤ (x2 + y 2 )(u2 + v 2 ),
with equality if and only there exists some a ∈ R such that x = au and y = av. This follows
from the elementary observation that the difference of the two sides in the above inequality equals
(xv − yu)2 . Next, we have
|z + w|2 = (z + w)(z + w) = |z|2 + |w|2 + 2ℜ(z w̄).
So if we let z = x + iy and w = u + iv, then

p p
2ℜ(z w̄) = (x + iy)(u − iv) + (x − iy)(u + iv) = 2(xu + yv) ≤ 2 x2 + y 2 u2 + v 2 = 2|z||w|,
which completes the proof of the inequality. We have equality if and only if there exists a a ∈ R
such that x = au and y = av, or equivalently, z = aw.
2. For the second inequality, without loss off generality we may assume that |z| ≤ |w|. Then by the
first part we have
|w| = |z + w − z| ≤ |z| + |w − z|.
Again we have equality if and only if there exists a real number b such that w − z = bz, or w =
(1 + b)z.
Lecture 2
Lecture-2: Limits, continuity and

holomorphicity
2.1 Topology of the complex plane

A consequence of the triangle inequality discussed in the previous lecture is that the function
d(z, w) := |z − w|
defines a distance function on C, and for simplicity, we denote the corresponding metric space by (C, | · |).
Given a z0 ∈ C, the open disc of radius R around z0 is given by
DR (z0 ) = {z ∈ C | |z − z0 | < R}.
We now review a few standard definitions from topology. The complement of a set S, denoted by S c is the
set of all complex numbers NOT in S. Given any set S ⊂ C, a point p ∈ C is a limit or an accumulation
point if for any r > 0, the disc Dr (p) has at least one point in common with S other than possibly p itself.
A point p ∈ S is said to be isolated if p is not a limit point of S. The closure of a set S, denoted by S̄ is the
union of S with all it’s accumulation points. The interior of S, denoted by S̊, is the set of all points p ∈ S
such that Dr (p) ⊂ S for some r > 0. The boundary of a set S is the set of points p ∈ C such that for all
r > 0, the disc Dr (p) contains at least one point from S and S c . For instance the boundary of the open
disc Dr (p) is the circle of radius r centered at p.
A set S is called open is for any point p ∈ S, there exists a disc Dr (p) ⊂ S. That is each point has a
neighborhood that is completely contained in the set. A set is called closed is it’s complement is open. An
equivalent definition (why are they equivalent?) is that a set is closed if and only if it completely contains
it’s boundary. So for any set S, the interior S̊ is the largest open set contained in S and the closure S̄ is
the smallest closed set containing S. A basic property of open and closed sets is the following.
Proposition 2.1.1. • Arbitrary union (possibly infinite) of open sets is again open. Finite intersection of
open sets is open.
• Arbitrary intersection of closed sets is close. Finite union of closed sets is closed.
Given a sequence {zn } we say that it converges to p ∈ C if for all ε > 0, there exists an N such that
|zn − p| < ε.
Proposition 2.1.2. zn → p if and only if ℜ(zn ) → ℜ(p) and ℑ(zn ) → ℑ(p).
9
10 LECTURE 2. LECTURE-2: LIMITS, CONTINUITY AND HOLOMORPHICITY
This is a consequence of the fact that for any z ∈ C,

√
max(|ℜ(z)|, |ℑ(z)|) ≤ |z| ≤ 2 max(|ℜ(z)|, |ℑ(z)|).
A disadvantage of the above definition of convergence is that one needs to know the limit a priori, to even
decide if a sequence is converging. A convenient alternative is of course the notion of a Cauchy sequence.
Recall that a sequence zn is said to be Cauchy if for all ε > 0, there exists an N > 0 such that for all
n, m > N we have
|zn − zm | < ε.
It is easy to see (prove it!) that every convergent sequence is Cauchy. Conversely, we have the following
fundamental fact.
Theorem 2.1.3. Every Cauchy sequence in C converges. That is, (C, | · |) is a complete metric space.
. The theorem follows from the proposition above and the fact that real numbers from a complete metric
space. Recall that a set is called compact if every open cover has a finite sub-cover. A consequence of
completeness is the following useful characterization of compact sets in C.
Theorem 2.1.4. The following are equivalent for a subset K ⊂ C.
1. K is compact.
2. K is closed and bounded.
3. K is sequentially compact. That is, any infinite sequence {zn } ⊂ K has an accumulation point p ∈ K.
The last notion we need is that of a connected set. A subset S ⊂ C is called connected, if
S = (U ∩ S) ∪ (V ∩ S)
for some disjoint open sets U and V . If S itself is open, this reduces to saying that S cannot be written as
the union of two disjoint open sets. An open, connected subset is called a region. We have the following
elementary characterization of regions.
2.2 Functions on the complex plane

Let S ⊂ C be a subset. A function f : S → C is a rule that assigns unique complex number, denoted by
f (z) to every number z ∈ S. The set S is called the domain of the function, and
f (S) := {f (z) | z ∈ S},
is called the range. The pre-image of a set T ⊂ C, denoted by f −1 (T ) is the subset of S defined by
f −1 (T ) = {z ∈ S | f (z) ∈ T }.
A function is called injective or one-one if the pre-image of every point in the range consists of exactly one
point, i.e
f (z) = f (w) =⇒ z = w.
It is said to surjective or onto if the range is all of C.
We say that the limit of f (z) as z tends towards p is L, and denote it by
lim f (z) = L,
z→p
if the following holds - For any ε > 0, there exists a δ > 0 such that
|z − p| < δ, z ∈ S, =⇒ |f (z) − L| < ε.

2.2. FUNCTIONS ON THE COMPLEX PLANE 11
We say that f is continuous at p ∈ S if

lim f (z) = f (p).
z→p
f is simply called continuous if it is continuous at all points in its domain. We then have the basic fact.
Theorem 2.2.1. f : S → C is continuous if and only if Re(f ) and Im(f ) a continuous as real valued
functions of two variables.
So as far as topology, which is the study of continuous functions, is concerned, there is no difference
between C and R2 . With this remark, the following properties follow easily from what is already known
about multivariable functions.
Theorem 2.2.2. Consider a function f : Ω → C, where Ω is open.
1. It is continuous if and only if f −1 (U ) is open for any open set U ⊂ C.
2. It is continuous if and only if f −1 (K) is closed for every closed set K ⊂ C.
3. It is continuous at p ∈ Ω if and only if for any sequence {zn } such that zn → p, we have
lim f (zn ) = f (p).

zn →p
4. If f is continuous, then for any compact subset K ⊂ Ω, f (K) is compact.

5. If f and g : Ω → C are continuous at p then so are f ± g and f g. If g(p) ̸= 0, then f /g is also
continuous at p.
6. if f is continuous at p, and g : f (Ω) → C is continuous at f (p), then the composition g ◦ f is also
continuous at p.
Example 2.2.3. The function f (z) = z n , where n is an integer, is continuous. To see this, note that
z n − pn = (z − p)(z n−1 + z n−2 p · · · + pn−1 )
A polynomial is a function p : C → C of the form
p(z) = an z n + an−1 z n−1 + · · · + a1 z + a0 ,
where ak ∈ C for k = 0, 1, · · · , n. Then by the fact that sums of contiuous functions are continuous, it follows
that polynomials are continuous at all points. A rational function is a quotient of two polynomials
p(z)
R(z) = ,
q(z)
wherever q is non-zero. At all such points, by the quotient rule above, a rational function is also continuous.
Example 2.2.4. The function f (z) = z̄ is continuous. Similarly, the function g(z) = |z| is also continuous.
Example 2.2.5. Arg(z) in not continuous on C. Recall that if z = x + iy, then arg(z) is defined as the
unique angle between (−π, π] that the line joining the origin to (x, y) makes with the positive x-axis. Now
consider any point on the the negative x-axis, say z = −1. Then the function is not continuous. The reason
being that if zn → −1 with ℑ(zn ) > 0, then arg(zn ) → π, while if ℑ(zn ) < 0, then arg(zn ) → −π.
2.2.1 Path connected sets in C.

As a special case, we could take S to be an interval in R thought of as a subset of the x-axis in C. A path is
defined to be a continuous function γ : I → R, where I is an interval. We then have the following useful
characterization of regions in C.
Proposition 2.2.6. Let Ω ⊂ C be an open subset. Then Ω ⊂ C is a region if and only if Ω is path connected,
ie. for any z0 , z1 ∈ Ω, there exists a continuous map γ : [0, 1] → Ω such that γ(0) = z0 and γ(1) = z1 .
Proof. Suppose Ω is a region. Fix z0 ∈ Ω, and let
A = {z ∈ Ω | there exists a path in Ω connecting z to z0 }.
Note that z0 ∈ A, and hence A is non-empty. Since Ω is open, and clearly every disc is path connected,
A is open. Now we claim that Ac := Ω \ A is also open. To see this, if Ac is non-empty we have some
w ∈ Ac . Then since Ω is open, there is disc Dr (w) ⊂ Ω. Clearly, Dr (w) ∩ A = Φ, for if, z1 ∈ Dr (w) ∩ A,
then one could simply connect z to z0 , by connecting z to z1 , and z1 to z0 , contradicting our assumption
that w ∈ Ac . This shows that Dr (w) ⊂ Ac , and hence Ac is open. But since Ω is connected and A is
non-empty, this forces Ac = Φ. Conversely, suppose Ω is path connected, but is disconnected. Then we
can write Ω = A ∪ Ac , where both A and Ac are open and non-empty.
2.2.2 Convergence of functions

There are two notions of convergence, that of point-wise, and uniform convergence. We say that
• the sequence of functions fn : Ω → C converges point-wise to f : Ω → C, if for every z ∈ Ω, the
sequence fn (z) → f (z). Or equivalently, given any ε > 0, and any z ∈ Ω, there exists an N > 0,
possibly depending both on ε and z, such that
n > N =⇒ |fn (z) − f (z)| < ε.
• the sequence of functions is said to converge uniformly if given any ε > 0, there exists an N > 0
depending only on ε such that for all z ∈ Ω and n > N we have that
|fn (z) − f (z)| < ε.
Theorem 2.2.7. If fn : Ω → C is a sequence of continuous functions which converge uniformly to f : Ω → C,

then f itself is continuous.
2.3 Holomorphic functions

Let Ω ⊂ C be an open subset. We say that a function f : Ω → C is complex differentiable at z = p ∈ Ω, if
the limit
d f (z) − f (p) f (a + h) − f (a)
f ′ (p) = f (z) = lim = lim ,
dz z=p z→p z−p h→0 h
exists and is finite. The limit, denoted by f ′ (p), is then called the derivative (or the complex derivative if the
context is not clear) of f at p. The function is called holomorphic, if it is complex differentiable at all points
in the domain. We will denote the set of holomorphic functions on Ω by O(Ω). Note that formally this
definition is identical to the one for real valued functions of one variable. We have also seen that functions
of a complex variable can be thought of as vector fields in two variables. In multivariable calculus, there
is already a notion of derivatives of such functions. A natural question is to ask for the relation between
these two notions. We will return to this question shortly. But first we collect some basic properties of
holomorphic functions, the proofs of which are identical to those in for differentiable functions of one real
variable.
Theorem 2.3.1. Let f, g : Ω → C be differentiable at z, and a, b ∈ C
1. If f is constant in a neighbourhood of z, then f ′ (z) = 0.

2.3. HOLOMORPHIC FUNCTIONS 13
2. (Linearity) af + bg is complex differentiable at z, and
[af + bg]′ (z) = af ′ (z) + bg ′ (z)
3. (Product rule) f g is complex differentiable at z and
(f g)′ (z) = f ′ (z)g(z) + f (z)g ′ (z)
4. (Quotient rule) If g ′ (z) ̸= 0, then f /g is complex differentiable at z and

f ′ f ′ (z)g(z) − f (z)g ′ (z)
(z) = .
g g 2 (z)
5. (Chain rule) If h is complex differentiable at f (z), then h ◦ f is complex differentiable at z and
[h ◦ f ]′ (z) = h′ (f (z)) · f ′ (z).
The main theme of the course is that holomorphicity imposes severe restrictions on the functions under
consideration. Just to get our feet wet, we start with the following elementary observation.
Proposition 2.3.2. If a function f is complex differentiable at z = a then it is automatically continuous at
z = a.
Proof. We proceed by contradiction. So suppose f is not continuous at a. Then there exists an ε > 0 and
a sequence zn → a such that |f (zn ) − f (a)| > ε. By holomorphicity, there exists an N such that
f (zn ) − f (a)
− f ′ (a) < 1,
zn − a
whenever n > N . Or equivalently, that
|f (zn ) − f (a) − (zn − a)f ′ (a)| < |zn − a|.
By the triangle inequality
ε < |f (zn ) − f (a)| = |f (zn ) − f (a) − (zn − a)f ′ (a) + (zn − a)f ′ (a)|
< |f (zn ) − f (a) − (zn − a)f ′ (a)| + |(zn − a)f ′ (a)|
< |zn − a|(1 + |f ′ (a)|)
Suppose now N is chosen large enough so that

ε
|zn − a| < ,
2(1 + |f ′ (a)|)
then the above chain of inequality yields
ε
ε < |f (zn ) − f (a)| < ,
2
which is absurd.
Example 2.3.3. As a first example, we compute the derivative of f (z) = z n , where n is an integer. We first
assume that n ≥ 1. Then it is easy to see (try to prove it!) that
z n − an = (z − a)(z n−1 + z n−2 a + · · · + zan−2 + an−1 ).

So then
z n − an
f ′ (a) = lim = lim z n−1 + z n−2 a + · · · + zan−2 + an−1 = nan−1 .
z→a z − a z→a
For negative integers n we can apply quotient rule to again obtain the same formula. So for integers n we have
that
d
z n = nan−1
dz z=a
Example 2.3.4. Polynomials and Rational functions. Recall that polynomials are functions of the type
p(z) = an z n + an−1 z n−1 + · · · + a0 .
Then by the above theorem, such functions are holomorphic. Moreover, by the above calculation the derivative
of a degree n polynomial is again a polynomial, but of degree n − 1. Recall also that rational functions are
quotients of two polynomials. By the quotient rule, these are holomorphic at all points where the denominator
does not vanish. That is, if
p(z)
R(z) = ,
q(z)
then R(z) is holomorphic at z = a if and only if q(a) ̸= 0.
Example 2.3.5. The function f (z) = z̄ is not holomorphic. To see this, consider the difference quotient
f (z + h) − f (z) h̄
= .
h h
Then if h → 0 along the real axis i.e. h ∈ R, this difference quotient is 1. On the other hand if h → 0 along
the imaginary axis, i.e. h = ik where k ∈ R, then this quotient is always -1. So the limit cannot exist.
Example 2.3.6. The function f (z) = |z| is not holomorphic. By the product rule, it is enough to show that
g(z) = |z|2 is not holomorphic. The difference quotient is
|z + h|2 − |z|2 (z + h)(z + h) − z z̄ z h̄ + z̄h + |h|2 h̄

= = = z + z + h̄.
h h h h
The limit of the last two terms as h → 0 is z̄, but, as we saw in the previous example, the limit of the first term
does not exist. So |z|2 , and hence |z|, is not differentiable.
Lecture 3
Power series
A power series centered at z0 ∈ C is an expansion of the form

∞
X
an (z − z0 )n ,
n=0
where an , z ∈ C. If an and z are restricted to be real numbers, this is the usual power series that you are
already familiar with. A priori it is only a formal expression. But for certain values of z, lying in the so
called disc of convergence, this series actually converges, and the power series represents a function of z.
Before we discuss this fundamental theorem of power series, let us review some basic facts about complex
series, and series of complex valued functions.
3.1 Infinite series of complex numbers: A recap

A series is an infinite sum of the form
∞
X
an ,
n=0
where an ∈ C for all n. We say that the series converges to S, and write
∞
X
an = S,
n=0
if the sequence of partial sums

N
X
SN = an
n=0
converges to S. We need the following basic fact.
P∞ P∞
Proposition 3.1.1. If n=0 |an | converges then n=0 an converges.
P∞
Proof. We will show that SN forms a Cauchy sequence if n=0 |an | converges. Then the theorem will
PN
follow from the completeness of C. If we let TN = n=0 |an |, then {TN } is a Cauchy sequence by the
hypothesis. Then by triangle inequality
M
X M
X
|SN − SM | = an ≤ |an | = |TM − TN |.
n=N +1 n=N +1
15
16 LECTURE 3. POWER SERIES
Now, given ε > 0, there exists a K > 0 such that for all N, M > K, |TM − TN | < ε. But then
|SM − SN | ≤ ε, and so {SN } is also Cauchy.
P P
We say that P∞ absolutely if
an converges |an | converges. Next suppose fn : Ω → C are complex
functions, we say that n=0 fn (z) converges uniformly if the corresponding sequence of partial sums
N
X
SN (z) = fn (z)
n=0
converges uniformly.
Proposition 3.1.2 (Weierstrass’ M-Test). Suppose fn : Ω → C is a sequence of complex functions, and
{Mn } is a sequence of positive real numbers such that
• |fn (z)| ≤ Mn for all n and all z ∈ Ω.
P∞
• n=0 Mn converges.
P∞
Then n=0 fn (z) converges uniformly.
P {SN (z)} is uniformly Cauchy. But

Proof. Like before, this time we show that the sequence of partial sums
again by triangle inequality if TN denotes the N th partial sum of Mn , then
M
X M
X
|SN (z) − SM (z)| ≤ |fn (z)| ≤ Mn = |TM − TN |,
n=N +1 n=N +1
for all z ∈ Ω. Since the right side does no depend on z, given ε > 0, one can make |SN (z) − SM (z)| < ε
by choosing N, M > K where K can be chosen independent of z.
3.2 Convergence of power series

By convergence of the power series, we mean the following. Consider the truncations of the power series
at the N th term, also called the N th partial sum -
N
X
sN (z) = an (z − z0 )n .
n=0
We say that the power series converges (unifomrly) if the sequence of functions {sN (z)} converges (uni-
formly). We say that the series converges absolutely if the sequence of functions
N
X
|an ||z|n
n=0
converges. It is well known, and not difficult to see, that absolute convergence implies convergence. The
fundamental fact is the following.
Theorem 3.2.1 (Fundamental theorem of power series). There exists a 0 ≤ R ≤ ∞ such that
• If |z − z0 | < R, the series
∞
X
an (z − z0 )n
n=0
converges absolutely.
3.2. CONVERGENCE OF POWER SERIES 17
• For any compact set K ⊂ DR (z0 ), the absolute convergence is actually uniform.
• If |z − z0 | > R, then the series diverges.
Moreover, R can be computed using the Cauchy-Hadamard formula:
1
R= .
lim sup |an |1/n
The number R is called the radius of convergence, and the domain DR = {z | |z − z0 | < R} is called the
disc of convergence. Recall that for a sequence {bn } of real numbers,
L = lim sup bn
if the following two conditions hold

• For all ε > 0, and all N > 0, there exists n > N such that
bn > L − ε.
• For all ε > 0, there exists an N such that for all n > N ,
bn < L + ε.
Remark 3.2.2. It is not difficult to show that if

an+1
lim
n→∞ an
exists, then it is equal to 1/R. On many occasions the limiting ratio is easier to calculate.
Remark 3.2.3. Suppose 0 < R < ∞ is the radius of convergence of the above power series. The by
the theorem, the series converges on the open disc |z − z0 | < R. The behavior of the series at points on
boundary however is subtle as the examples below indicate.
Example 3.2.4. The power series
∞
X
zn
n=0
has radius of convergence 1. In fact it easy to see that on |z| < 1,

∞
1 X
= zn.
1 − z n=0
Observe that on the unit disc |z n | = |z|n = 1, and so by the divergence test, the series cannot converge at any
boundary point. On the other hand, the left hand side is defined and holomorphic at all points z ̸= 1 even
though the power series is only defined inside the unit disc. We then say that the holomorphic function 1/1 − z
is an analytic continuation of the power series to the domain C \ {z = 1}. We will say more about analytic
continuation towards the end of the course. We remark that the following misleading formula often appears
in popular culture (most notably in a video on the youtube channel - Numberphile, an otherwise decent math
channel), many times accompanied with a quote with the effect that “Oh look - math is magical!":
1
1 − 1 + 1 − 1..... = .
2
The formula as stated is of course junk since the left hand side is clearly a divergent series. But there are
ways of interpreting the left hand side. For instance, the left hand side is in fact Cesaro summable, which is a
generalization of usual infinite summation in that a convergent series is also Cesaro summable and the Cesaro
sum equals the sum of the series. In this case, the Cesaro sum does turn out to be 1/2. A more fundamental
P∞ way
(at least in my opinion) of interpreting the left hand side, as precisely the analytic continuation of n=0 z n to
z = −1. Then as remarked above, this analytic continuation is given by 1/(1 − z) which of course equals 1/2
at z = −1. Another example of such misleading propogation of math, especially in India, is that Ramanujan
proved the “miraculous" identity that
−1
1 + 2 + 3 + ··· = .
12
We’ll see later
P∞ in the course that the left hand side should in fact be replaced by the analytic continuation of
the series n=1 n−s to s = −1. The infinite seres is a priori only defined on the region Re(s) > 1, but can
be analytically continued to C \ {1}, and this if of course the famous ζ(s) of Riemann.. We’ll then compute
that ζ(−1) = −1/12!
Example 3.2.5. Consider the power series
∞
X zn
.
n=0
n
Again, it is easy to see that the radius of convergence is 1. At z = 1 this is the usual harmonic series, and is
divergent. It turns out in fact, that this series converges for all other points on |z| = 1, z ̸= 1. This follows
from the following test due to Abel, which we state without proof.
Lemma 3.2.6 (Abel’s test). Consider the power series
∞
X
an z n .
n=0
Suppose
• an ∈ R, an ≥ 0.
• {an } is a decreasing sequence such that limn→0 an = 0.
Then the power series converges on |z| = 1 except possibly at z = 1.
Clearly the series in the example above satisfies all the hypothesis, and hence is convergent at all points on
|z| = 1 except at z = 1.
Example 3.2.7. Next, consider the power series
∞
X zn
.
n=0
n2
Again, the radius of convergence is 1, and again by Abel’s test the power series is convergent on |z| = 1 except
possibly at z = 1. But at z = 1, the series is clearly convergent, for instance by the integral test. So in this
example the power series is convergent on the entire boundary.
Example 3.2.8. Finally consider the power series
∞
X zn
.
n=0
n!
To find the radius of convergence, we use the ratio test. Denoting an = 1/n!, we see that
an+1 n! 1
lim = lim = lim = 0.
n→∞ an n→∞ (n + 1)! n→∞ n + 1
3.3. HOLOMORPHICITY OF POWER SERIES 19
So R = ∞. Notice that if z is a real number then this is the usual Taylor expansion of the exponential function.
Inspired by this, we define the complex exponential function, exp(z) : C → C by
∞
X zn
exp(z) = ez = .
n=0
n!
We will study this function in more detail in the next lecture.
3.3 Holomorphicity of power series

From the previous theorem, since the convergence is uniform on comapct subsets of the disc of conver-
gence, it is clear that a power series represents a continuous function. In fact, much more is true.
Theorem 3.3.1. Consider the function defined by
∞
X
f (z) = an (z − z0 )n ,
n=0
on the disc of convergence DR = {z | |z − z0 | < R} where 0 < R ≤ ∞. Then f (z) is holomorphic on DR

with
∞
X
f ′ (z) = nan (z − z0 )n .
n=1
1/n
Proof. Without loss
P∞of generality wen can assume z0 = 0. Firstly, observe that since limn→∞ n = 1,
the power series n=1 nan (z − z0 ) also has radius of convergence R. To prove the theorem, we need to
show that for any p ∈ DR (0),
∞
f (p + h) − f (p) X
lim = nan pn .
h→0 h n=1
Or equivalently, given any ε > 0, we need to find a δ > 0 such that

∞
f (p + h) − f (p) X
|h| < δ =⇒ − nan pn < ε (3.1)
h n=1
Let us denote by
N
X ∞
X
SN (z) = an z n , EN (z) = an z n ,
n=0 n=N +1
th th
the N partial sum, and the N error term respectively, so that
f (z) = SN (z) + EN (z).
Then since the partial sums are polynomials, they are holomorphic, and in fact
∞
X
′
lim SN (z) = nan z n , (3.2)
N →∞
n=1
where the convergence is uniform on compact subsets of DR (0) Now suppose |p| < r < R, then for any
N , we can break the difference that we need to estimate into three parts -
∞
f (p + h) − f (p) X SN (p + h) − SN (p) ′
− nan pn = − SN (p)
h n=1
h
∞
′
X EN (p + h) − EN (p)
+ |SN (p) − nan pn | +
n=1
h
Since SN , being a polynomial, is holomorphic, there exists a δ > 0 so that for |h| < δ, the first term is
smaller than ε/3. Similarly, by equation (3.2), the second term can be made smaller than ε/3 by choosing
N big enough. So all that remains is to control the error term. Using the factorization an − bn = (a −
b)(an−1 + an−2 b + · · · + bn−1 ), we have
∞
EN (p + h) − EN (p) X (p + h)n − pn
≤ |an |
h h
n=N +1
X∞
≤ |an ||((p + h)n−1 + · · · + pn−1 )|
n=N +1
But if |h| < δ for sufficiently small δ (in particular if δ < r − |p|), then |p + h| ≤ |p| + |h| < r, and so
∞
EN (p + h) − EN (p) X
≤ |an |nrn−1 .
h
n=N +1
n|an |rn−1 which converges for r < R, so we can also make this term
P
But this is the tail of the series
smaller than ε/3 by choosing N big enough. This shows that we can find δ small enough so that (3.1) is
satisfied.
Notice that the derivative is again a power series with the same radius of convergence. So applying the
above theorem inductively we obtain -
P∞
Corollary 3.3.2. A power series f (z) = n=0 an (z − z0 )n is infinitely complex differentiable in it’s disc of
convergence. Moreover, the derivatives can be computed by successive term-wise differentiation:
∞
X
f (k) (z) = n(n − 1) · · · (n − k + 1)an (z − z0 )n−k .
n=k
In particular, the coefficients of the power series are given by

f (n) (z0 )
an = .
n!
We say that a function f : Ω → C is analytic if for every p ∈ Ω, there exists an r = r(p) > 0 and a
sequence of numbers {an = an (p)} such that
∞
X
f (z) = an (z − p)n
n=0
for every z ∈ Dr (p). A priori, it is not quite clear that if a function is represented by a power series
expansion on a disc of convergence DR (p) then it is automatically analytic. For instance, it is not clear
if there should be a power series expansion around any other point q ∈ DR (p). The next proposition
answers this question in the afirmative.
P∞
Proposition 3.3.3. Let f (z) = n=0 an (z − z0 )n be a power series with radius of convergence R. Then f
is analytic on DR (z0 ). In fact, for every p ∈ DR (z0 ), and every z ∈ Dr (p) where r := R − |p − z0 |, we have
∞
X f (n) (p)
f (z) = (z − p)n .
n=0
n!
3.3. HOLOMORPHICITY OF POWER SERIES 21
Proof. We use the binomial expansion. As before, without loss of generality, we assume that z0 = 0.
Writing z = z − p + p, and applying the binomial theorem, we see that on |z − p| < R − |p| (since
|z| < R), we have
∞
X
f (z) = an (z − p + p)n
n=0
∞ Xn
X n n−k
= an p (z − p)k
n=0 k=0
k
X∞
= bk (z − p)k ,
k=0
where
∞
X n f (k) (p)
bk = an pn−k =
k k!
n=k
by Corollary 3.3.2.
Remark 3.3.4. In some cases the power series on the right in the conclusion of Proposition 3.3.3 might
have a larger radius of convergence than R − |q − p|. In such cases, the new power will define an analytic
continuation of f . For instance, let us consider the power series
∞
1 X
= zn
1 − z n=0
on |z| < 1 with p = 0. Let q = −1/2. Then if |z + 1/2| < 1 − 1/2 = 1/2, we have
∞
1 1 2 1 2 X 2n 1 n
= = · = z + .
1−z 3/2 − (z + 1/2) 3 1 − 2(z + 1/2)/3 3 n=0 3n 2
On the other hand, it can be easily seen that the power series on the right has a radius of convergence
R = 3/2, and hence defines an extension of the original power series in the new region |z + 1/2| < 3/2.
This was Weierstrass’ method of analytically continuing holomorphic functions.
Corollary 3.3.5. [Principle of analytic continuation for power series] Let f be an analytic function on a
connected open set Ω. If there exists a point p ∈ Ω such that f (n) (p) = 0 for all n ∈ N, f ≡ 0 on all of Ω. In
particular, the conclusion is true if there is an open set U ⊂ Ω such that f ≡ 0.
U
Proof. Let S = {z ∈ Ω | f (n) (z) = 0 for all n = 0, 1, 2, · · · }. Then by the continuity of f , S is closed in
Ω (ie. all limit points of S in Ω, are also contained in S). Also, S is non-empty by the hypothesis. Now
suppose q ∈ S. Since the function is analytic, there is an open disc Dr (q) on which
∞
X f (n) (q)
f (z) = (z − q)n = 0,
n=0
n!
since q ∈ S. But then Dr (q) ⊂ S, and so S is open in Ω. But then since Ω is connected, this forces S = Ω,
and in particular, f ≡ 0 on Ω.
Appendix: Multiplication and composition of power series

Given two power series
∞
X X
an z n , bn z n ,
n=0 n=0
their product can be defined, at least formally, in the following way-

∞
X X
an z n · bn z n = (a0 + a1 z + a2 z 2 + · · · )(b0 + b1 z + b2 z 2 + · · · )
n=0 n=0
= a0 b0 + (a0 b1 + a1 b0 )z + (a0 b2 + a1 b1 + a2 b0 )z 2 + · · ·
X∞
= cn z n ,
n=0
where cn is given by
n
X
cn = ak bn−k .
k=0
This product goes by the name of Cauchy product. The main theorem, which we state without proof, is the
following.
Theorem 3.3.6. If the radius of convergence of two series centered at z0 is R1 and R2 respectively, then their
product power series has a radius of convergence that is at least min (R1 , R2 ).
Proof of a slightly general version, applicable for any infinite series, can be found on the wiki article
https://en.wikipedia.org/wiki/Cauchy_product#Convergence_and_Mertens.
27_theorem.
Lecture 4
The exponential functions
4.1 The exponential functional

Last lecture, we defined the complex exponential function by the power series
∞
X zn
exp(z) := .
n=0
n!
We saw that the power series has infinite radius of convergence, and hence defines a function on the entire
complex plain. In fact by the theorem from last lecture, we now know that the exponential function is
holomorphic on the entire plane. Such functions, that are holomorphic on the entire complex plane, are
called entire functions. As we saw in the previous lecture, to find the complex derivative, it is enough to
differentiate term-wise:
∞ ∞ ∞
d X 1 d n X z n−1 X zn
exp(z) = z = = .
dz n=0
n! dz n=1
(n − 1)! n=0 n!
To see the last equality just replace n − 1 by n in the penultimate term. So we see that
d
exp(z) = exp(z).
dz
In fact we’ll see later that this property characterizes the exponential function. But first, we collect some
important properties of the exponential function.
Theorem 4.1.1. 1. exp(0) = 1.
2. For any complex numbers z, w we have that
exp(z + w) = exp(z) exp(w),
and in particular exp(−z) = [exp(z)]−1 .

3. exp(z) ̸= 0 for all z ∈ C.
4. The restriction of exp(z) to R is a positive strictly increasing function. For x ∈ R, ex = 1 if and only if
x = 0. Moreover, limx→∞ exp(x) = ∞ and limx→−∞ exp(x) = 0. In particular, ex : R → R+ is a
bijective function.
Proof. 1. This is trivial from the definition.
23
24 LECTURE 4. THE EXPONENTIAL FUNCTIONS
2. This follows from the product formula for power series and the binomial theorem. The left hand side
of the equation is
∞
X (z + w)n
exp(z + w) = .
n=0
n!
But then
n n
(z + w)n 1 X n! X zk wn−k
= z k wn−k = · .
n! n! k!(n − k)! k! (n − k)!
k=0 k=0
th
This is exactly the n coefficient of the Cauchy product, and the result follows.
3. Suppose exp(p) = 0 for some p ∈ C. Since f (z) = exp(z) is an analytic function with an infinite
radius of convergence at 0, we also have the following representation formula on all of C:
∞
X f (n) (p)
exp(z) = (z − p)n .
n=0
n!
But since f (n) (p) = exp(p) = 0 for all n, this would mean that exp(z) = 0 for all z, clearly
contradicting (1).
4. Since exp(z) ̸= 0 and exp(0) = 1, by the intermediate value theorem the function f (x) = exp(x)
defined on R is strictly positive. Moreover, since f (0) = 1 and f ′ (x) = ex > 0, the function
f (x) is strictly increasing, and hence f (x) > 1 for all x > 0. Then by the mean value theorem
f (x) − 1 > x for all x > 0. From this it follows that limx→∞ f (x) = ∞. Then property (2) implies
that limx→−∞ f (x) = 0. By the intermediate value theorem ex : R → R+ is then a surjective map,
and hence a bijection.
Remark 4.1.2. Due to property (2) and our familiarity with working with exponents from high-school,
from now one we adopt the more suggestive notation exp(z) = ez .
Theorem 4.1.3. There is a unique holomorphic function f : C → C satisfying
(
f ′ (z) = f (z)
f (0) = 1.
Proof. The proof uses the fact that if a holomorphic function has complex derivative identically zero, then
the function has to be a constant. We will prove this fact later in the course. Assuming this, consider the
function
g(z) = e−z f (z).
Then by the Chain rule, since f ′ (z) = f (z) we see that
g ′ (z) = e−z (−f (z) + f ′ (z)) = 0.
Hence g(z) is a constant. But by the initial condition we see that g(0) = 1. On the other hand by the first
property in Theorem 4.1.1, e−z = 1/ exp(z), and so f (z) = exp(z).
4.2 Trigonometric functions

We can now analogously define the functions sine and cosine using power series:
∞
X z 2m
cos z = (−1)m
m=0
(2m)!
4.2. TRIGONOMETRIC FUNCTIONS 25
∞
X z 2m+1
sin z = (−1)m
m=0
(2m + 1)!
It is easy to check that the radius of convergence of both the power series is infinity, and hence they define
entire functions, just like the exponential function. In fact an easy computation also shows that
d d
cos z = − sin z, sin z = cos z.
dz dz
The same computation then gives the following generalized Euler identity.
Proposition 4.2.1 (Generalized Euler identity). For any z ∈ C,
eiz = cos z + i sin z.
Proof. This follows trivially from the following observations:

(
n (−1)m , n = 2m
i =
(−1)m i, n = 2m + 1.
So then by definition
∞ n n ∞ ∞
X i z X z 2m X z 2m+1
eiz = = (−1)m +i (−1)m .
n=0
n! m=0
(2m)! m=0
(2m + 1)!
Euler’s identity then follows from the observation that the two series on the right are simply the Maclaurin
series for sine and cosine respectively.
Remark 4.2.2. Polar coordinates. The Euler identity can be used to give a third representation of com-
plex numbers in terms of the exponential function. Namely, for z ∈ C, let r = |z| and θ = arg z. Then we
have seen that
z = r cos θ + ir sin θ.
So by the Euler identity, we have the representation
z = reiθ .
This is sometimes very useful in computations.

Next, we collect some properties of the sine and cosine functions. These can be proved using the generalized
Euler identity, and the analogous properties of the exponential function.
Theorem 4.2.3. The sine and cosine function satisfy the following.
1. sin(0) = 0, cos(0) = 1, and for all z ∈ C we have
sin(−z) = − sin z and cos(−z) = cos(z).
2. For z, w ∈ C,
sin (z ± w) = sin z cos w ± cos z sin w

cos (z ± w) = cos z cos w ∓ sin z sin w.
3. For all z ∈ C,
sin2 z + cos2 z = 1.
Proof. 1. This follows easily from the definitions.

2. It follows from the definition and property (1) above that
eiz − e−iz eiz + e−iz

sin z = , and cos z = .
2i 2
The sum-angle fomrulae then follow from this and property (2) in Theorem 4.1.1.
3. This follows from the sum angle properties and property (1) above. An independent proof can be
given by observing that the derivative of f (z) = sin2 z + cos2 z is identically zero, hence f (z) is a
constant function, and hence equal to f (0) = 1.
One can also define the other trigonmetric functions tan z, cot z, sec z and csc z in the usual way.
4.3 Periodicity and the definition of π.

Theorem 4.3.1. There exists a smallest positive real number which we denote by π such that e2πi = 1.
Moreover, eiz = 1 if and only if z = 2nπ for some n ∈ Z.
The above theorem can be taken as the definition of the number π. As an immediate consequence of the
above theorem and Euler’s identity, we have the following.
Corollary 4.3.2. For all z ∈ C and all n ∈ Z,
ez+2nπi = exp(z), sin(z + 2nπ) = sin(z), and cos(z + 2nπ) = cos z.
In particular,
2n + 1
sin(2nπ) = cos π =0
2
for all n.
Proof of Theorem 4.3.1. We first prove that there exists a real number τ such that eiτ = 1. To see this, note
that by property (3) above, −1 ≤ sin(x), cos(x) ≤ 1 for all x ∈ R. Then by the mean value theorem, it is
easy to see that for all x > 0,
x2
sin x < x, and cos x > 1 − .
2
Once again by an application of the mean value theorem, since (sin x)′ = cos(x) > 1 − x2 /2, we see that
x3
sin(x) > x − .
6
Finally, by yet another application of the mean value theorem we obtain
x2 x4
cos x < 1 − + ,
2 24
√ √
√ x = 3, we see that cos 3 < 0, and hence by the intermediate value theorem, there
for all x > 0. Putting
exists a x0 ∈ (0, 3) such that cos(x0 ) = 0. Let τ = 4x0 . Then by the sum, angle formulae, cos(τ ) = 1
and sin(τ ) = 0, and hence by the Euler identity, eiτ = 1.
Next, we argue that any period of eiz has to be a real number. For, note that if p ∈ C is a period, that is if
eiz+ip = eiz for all z ∈ C, then eip = 1. But then, if p = a + ib, then 1 = |eip | = e−b , and so by part(4) in
Theorem 4.1.1, we see that b = 0, or that p has to be real.
4.4. THE LOGARITHM FUNCTION AND COMPLEX POWERS 27
Finally, we show that any period is an integral√multiple of τ . We first show that τ is the smallest positive
period. To see this, note that if 0 < x < x0 (< 3), then
x2 x
sin x > x 1 − > > 0.
6 2
Since (cos x)′ = − sin x, this shows that cos x is strictly decreasing in [0, x0 ]. Then from the identity
sin2 x + cos2 x = 1, since sin x > 0, we see that sin x is strictly increasing in [0, x0 ]. In particular,
0 < sin x < 1, and eix ̸= ±1 or ±i. Hence e4ix ̸= 1, and τ is indeed the smallest positive period. Now, if p
is any other period of eiz , then we can write p/τ = n+c, where n ∈ Z and c ∈ [0, 1). Then 1 = eip = eiτ c .
By our discussion above, if c > 0, then τ c ≥ τ , which is a contradiction. Hence c = 0, and p is an integral
multiple of τ . The proof of the theorem is complete with π := τ /2.
4.4 The logarithm function and complex powers

For functions of one real variable, the logarithm is the inverse function of the exponential function. Indeed
by Theorem 4.1.1, ex : R → R+ is a bijective map, and so we can define
ln x : R+ → R
to be the inverse function. We would like to generalize this to the complex plane. In particular, we would
like to have a definition for logarithm that makes it a holomorphic function. An immediate difficulty is
that while on real line, the exponential function is strictly increasing, and hence one-one, on the complex
plane, we have already seen that the exponential function is not one-one. For instance e0 = e2πin = 1 for
all n ∈ Z. So to define an inverse function, one has to make a choice of the pre-image. For instance, we
can choose to log 1 = 0 or indeed any of 2πin, for n = 1, 2, 3, · · · .
In fact, writing in polar coordinates z = reiθ , f (z) satisfies ef (z) = f (exp(z)) = z on any open connected
set if and only if
f (z) = ln r + iθ + 2πin, n = 0, 1, −1, 2, −2, · · · ,
where ln r is the logarithm on positive real numbers defined above. That is, we can at best define logarithm
as multivalued function. A choice of n corresponds to defining a single valued logarithm, the corresponding
function is called a branch of the logarithm. For instance, choosing n = 0, and defining
log z = log r + iθ,
picks out what is called as the principal branch of the logarithm. This might seem like a good definition
until we realize that the logarithm so defined is not even continuous. To see this, suppose z → −1 from
the 2nd quadrant. Then log z will tend to iπ. But on the other hand, as z → −1 from the third quadrant,
log z will tend to −π. This is not only a minor irritant that can be fixed by some trick, but as we will see
later in the course, is a fundamental issue. In fact, we will see that there is actually no way to define a
holomorphic logarithm on which is defined on all of C \ {0}. The best we can do is to define it outside of
a ray. In fact we have the following.
Theorem 4.4.1. The function
log z := log |z| + i arg z
defines a holomorphic function on C \ Re(z) ≤ 0 satisfying elog z = z. Moreover, applying chain rule, we see
that
d 1
log z = .
dz z
We will see a proof of this later in the course. The logarithm function then has the property that log 1 = 0
and
log zw = log z + log w,
assuming it is defined at all those points.
4.4.1 Logarithm as power series.

0
Since e = 1, and definition of logarithm must satisfy log 1 = 0. Let’s see if we can have a definition of
logarithm using power series centered at z0 = 1. By the chain rule, if have a holomorphic function log z
near z0 = 1, then
d 1
log z = .
dz z
Iteratively, we must have
dn
log z = (−1)n−1 (n − 1)!.
dz n z=1
If log z has a power series expansion around z0 = 1, then the coefficients must be given by
1 dn 1
an = log z = (−1)n−1 .
n! dz n z=1 n
Turning this around, we consider the power series
∞
X (−1)n−1
L(z) = (z − 1)n .
n=1
n
We then have the following

Proposition 4.4.2. The holomorphic function L(z) : D1 (1) → C satisfies
L(z) = log z,
where log z is the principal branch of the logarithm on C \ Re(z) ≤ 0 defined above.
Proof. We already know that d log z/dz = 1/z for the principal branch of the logarithm. Also, L(1) =
log 1 = 0. So, similar to the above proof, all we need to show (modulo the theorem on identically zero
derivatives to be covered later) is that L′ (z) = 1/z. Since F (z) is a power series, by term-wise differenti-
ation,
X∞ X∞
L′ (z) = (−1)n−1 (z − 1)n−1 = (−1)n (z − 1)n .
n=1 n=0
From the geometric series expansion, we know that for |w| < 1,
∞
X 1
wn = .
n=0
1−w
Putting w = 1 − z (we can do this since |z − 1| < 1) in the above expansion

∞ ∞
1 X X
= (1 − z)n = (−1)n (z − 1)n = L′ (z).
z n=0 n=0
4.4.2 Complex powers

Once logarithm is defined, one can also define the powers of complex to other complex numbers by the
simple formula
z w = ew log z .
Example 4.4.3. The nth roots. If n is an integer, then en log z = (elog z )n = z n , and hence the new
definition of complex powers agrees with our usual definition of integer powers of complex numbers.
Lecture 5
Cauchy Riemann equations
5.1 A review of multivariable calculus

Let Ω ⊂ R2 be an open set, and p = (a, b) ∈ Ω. Then a function f : Ω → R2 is said to be (totally)
differentiable at p, if there exists a linear map Dfp : R2 → R2 such that
f (p + h) − f (p) − Dfp (h)

lim = 0.
|h|→0 |h|
The linear map Dfp is called the (total) derviative of f at p. If f is differentiable at p, then f is also
continuous, and moreover the partial derivatives ∂f /∂x and ∂f /∂y exist at p. In fact, if

1 0
e⃗1 = , e⃗2 =
0 1
are the standard basis vectors for R2 , then

∂f ∂f
(p) = Dfp (e⃗1 ), and (p) = Dfp (e⃗2 ).
∂x ∂y
In particular if f = (u, v), the matrix for the linear map Dfp in terms of the standard basis, called the
Jacobian matrix, is given by !
∂u ∂u
∂x (p) ∂y (p)
Jf (p) := ∂v ∂v .
∂x (p) ∂y (p)
The determinant of the Jacobian matrix is called simply the Jacobian, and we will denote it by Jf (p).
Remark 5.1.1. Note that the mere existence of partial derivatives is not sufficient for the function to be
differentiable. On the other hand, if the partial derivatives exist and are continuous, then the function is
indeed differentiable.
5.2 Cauchy-Riemann equations

Recall that a function f : Ω → C is holomorphic if
d f (z + h) − f (z)
f ′ (z) = f (z) = lim
dz h→0 h
29
30 LECTURE 5. CAUCHY RIEMANN EQUATIONS
exists and is finite at all points z ∈ Ω. Putting z = x + iy and decomposing f into it’s real and imaginary
parts, we can write
f (z) = f (x, y) = u(x, y) + iv(x, y),
where u, v : Ω → R are real valued functions of two real variables.
Question 5.2.1. What restrictions does holomorphicity of f put on the functions u and v?
So suppose f is holomorphic. In the limit above, suppose h goes to zero along the the real axis ie. h = h+i0
is a real number. Then one can re-write the difference quotient as
f (z + h) − f (z) u(x + h, y) − u(x, y) v(x + h, y) − v(x, y)

lim = lim + i lim
h→0 h h→0 h h→0 h
∂u ∂v
= (x, y) + i (x, y).
∂x ∂x
Notice that since the limit on the let exists, by definition of the limits of complex values functions, the
individual limits of the real and imaginary parts also exist, and hence the two limits on the right exist. In
other words, if f is holomorphic, then the partials of u and v with respect to x exist. On the other hand, if
h goes to zero along the imaginary axis ie. h = ik, where k ∈ R goes to zero, then
f (z + h) − f (z) u(x, y + k) − u(x, y) v(x, y + k) − v(x, y)

lim = lim + i lim
h→0 h k→0 ik k→0 ik
v(x, y + k) − v(x, y) u(x, y + k) − u(x, y)
= lim − i lim
k→0 k k→0 k
∂v ∂u
= (x, y) − i (x, y).
∂y ∂y
Again, the fact that f is holomorphic implies that the partials of u and v with respect to y also exist. But of
course, if f is holomorphic, then these two limits must coincide. Setting the real and imaginary parts equal
to each other we obtain the so called Cauchy-Riemann equations. In fact we have the following fundamental
characterisation of holomorphicity. To state it, we first define “complex" multiplication of vectors in R2 as
the map J : R2 → R2 given by
a −b
J· = .
b a
This is of course, simply multiplication by i, once we identify R2 with C. It is easy to check that this is a
linear isomorphism satisfying J 2 = −id, and that the matrix with respect to the standard basis is

0 −1
J= .
1 0
Theorem 5.2.2. Let Ω ⊂ C be an open set and f = u + iv : Ω → C a function. Then the following are
equivalent.
1. f = u + iv is complex differentiable at a point z ∈ Ω.
2. Dfp exists and u and v satisfy the Cauchy-Riemann (CR) equations:
(
∂u ∂v
∂x (x, y) = ∂y (x, y)
∂v ∂u
(CR)
∂x (x, y) = − ∂y (x, y)
3. Dfp exists and is C-linear in the sense that for any vector ⃗v ∈ R2 ,
Dfp (J · ⃗v ) = J · Dfp (⃗v ).

5.2. CAUCHY-RIEMANN EQUATIONS 31
Moreover, if f is complex differentiable at p = a + ib, then

∂f ∂f
f ′ (p) = (a, b) = i (a, b). (5.1)
∂x ∂y
Proof. • Proof of (1) =⇒ (2). Firstly, the remarks above prove that if f is complex differentiable
at p = a + ib, then the partial derivatives of u and v exist at (a, b) and satisfy the Cauchy-Riemann
equations. Moreover, f ′ (p) can be computed using the formula (5.1) above. All it is remains to
be shown is that f is in fact differentiable as a vector valued function of two variables. From the
definition of differentiability, given ε > 0, there exists a δ > 0 such that
f (p + h) − f (p) − f ′ (p)h
< ε,
h
whenever |h| < δ. By the Cauchy Riemann equations, the Jacobian matrix at p is given by

ux (a, b) uy (a, b)
Jf (p) = .
−uy (a, b) ux (a, b)
If Dfp is the linear transformation associated to the Jacobian, then an easy computation shows that
for any complex number h = λ + iµ,
Dfp (h) = (λux (a, b) + µuy (a, b)) + i(−λuy (a, b) + µux (a, b)) = f ′ (p)h.
Here the left hand side is a vector in R2 , while the right hand side is a complex number, and the
identification is the usual one. So give ε > 0, for the δ > 0 chosen above, we have that
f (p + h) − f (p) − Dfp (h) f (p + h) − f (p) − f ′ (p)h
= < ε,
|h| h
whenever |h| < δ. Hence f is differentiable at p with derivative given by Dfp .
• Proof of (2) ⇐⇒ (3). Let e⃗1 and e⃗2 be the standard basis vectors for R2 as above. Note that
J(e⃗1 ) = e⃗2 and J(e⃗2 ) = −e⃗1 . Then
" #
∂u
∂y (p)
Dfp (J(e⃗1 )) = Dfp (e⃗2 ) = ∂v
∂y (p)
∂u (p) − ∂v (p)

J(Dfp (e⃗1 )) = J ∂x = ∂u ∂x .
∂v
∂x (p) ∂x (p)
So the Cauchy-Riemann equations are satisfied if and only if Dfp (J(⃗v )) = J(Dfp (⃗v )).
• Proof of (2) =⇒ (1). Let p = a + ib ∈ Ω be a point, and let
(
A = ∂u ∂v
∂x (a, b) = ∂y (a, b)
B = ∂x (a, b) = − ∂u
∂v
∂y (a, b).
By the definition of differentiability,
u(a + h, b + k) = u(a, b) + hA − kB + ε1 (h, k)

v(a + h, b + k) = v(a, b) + hB + kA + ε2 (h, k),
√ √
where ε1 (h, k)/ h2 + k 2 → 0 and ε2 (h, k)/ h2 + k 2 → 0 as (h, k) → 0. So
f (a + ib + (h + ik)) − f (a + ib) h(A + iB) − k(B − iA) ε1 (h, k) + ε2 (h, k)
= + .
h + ik h + ik h + ik
For the first term, multiplying and dividing by the conjugate h − ik and simplifying, we see that
(hA − kB) + i(hB + kA)

= A + iB.
h + ik
For the second term, using triangle inequality, we see that
ε1 (h, k) + ε2 (h, k) |ε1 (h, k) |ε2 (h, k)

≤ √ + √ →0
h + ik 2
h +k 2 h2 + k 2
as (h, k) → 0. And so we see that
f (a + ib + (h + ik)) − f (a + ib)
lim
h+ik→0 h + ik
exists, and is in fact equal to A + iB. This proves that f is holomorphic, with
∂u ∂v ∂u ∂v
f ′ (p) = (a, b) + i (a, b) = i (a, b) − (a, b).
∂x ∂x ∂y ∂y
As a consequence, we have the following useful observation.

Corollary 5.2.3. Let f = u + iv : Ω → R be holomorphic at p ∈ Ω. Then, Jf (p) = |f ′ (z)|2 . In particular,
the Jacobian of is always positive, and hence any holomorphic map is orientation preserving.
Proof. By the Cauchy Riemann equations and (5.1),

∂u ∂v ∂u ∂v ∂u 2 ∂v 2
Jf = − = + = |f ′ (z)|2 .
∂x ∂y ∂y ∂x ∂x ∂x
The Cauchy-Riemann equations help us compute complex derivatives, or rule out the possibilities of some
functions being holomorphic.
Example 5.2.4. The function f (z) = z̄ cannot be holomorphic. To see this, note that f (z) = x − iy.
Applying the above theorem with u(x, y) = x and v(x, y) = −y, we see that ux = 1 while vy = −1, and
so the above equations are not satisfied. For this function f¯(z) = z is holomorphic. Functions such as these
whose conjugates are holomorphic, are called anti-holomorphic functions. It is easy to prove that the only
holomorphic and anti-holomorphic functions are the constants.
Example 5.2.5. We have already seen that f (z) = |z| is not holomorphic. Using the above theorem it is
easy to see that in fact, any function, defined on an open connected set, that takes only real values cannot be
holomorphic, unless it is a constant function. This is because from the above theorem (since v=0) we get that
ux = uy = 0. That is, the gradient of u is zero in an open connected set in R2 . But then, by a standard fact
from multivariable calculus, u (and hence f ) has to be a constant.
p
Example 5.2.6. The function f (x + iy) = |x||y| satisfies the Cauchy-Riemann equations at the origin.
On the other hand, one can show that the function is not holomorphic at z = 0. The problem is of course that
Df0 does not exist.
We now discuss some important consequences of the Cauchy-Riemann equations.
Corollary 5.2.7. Let Ω ⊂ C be a connected, open subset, and f : Ω → C be a holomorphic function. If
f ′ (z) ≡ 0, then f is a constant.
5.2. CAUCHY-RIEMANN EQUATIONS 33
Proof. If f ′ (z) ≡ 0, then by the Cauchy Riemann equations, and (5.1), Dfp ≡ 0. By the mean value
theorem (applied to line segments in R2 ), we see that f has to be a constant.
Corollary 5.2.8. Let Ω be an open subset of C containing the origin. There does not exist a holomorphic
function f (z) on Ω such that exp(f (z)) = z for all z ∈ Ω.
Proof. Since Ω is an open set containing the origin, there exists an δ > 0 such that Dδ (0) ⊂ Ω. We proceed
by contradiction. Let f = u + iv be such a function. Then necessarily
1 x y
f ′ (z) = = 2 2
−i 2 .
z x +y x + y2
By the Cauchy Riemann equations we then have that
∂v y ∂v x
=− 2 := P, = 2 := Q,
∂x x + y2 ∂y x + y2
that is ∇u = F⃗ = (P, Q). Then by the fundamental theorem of line integrals

ˆ
∇u · d⃗r = 0.
∂Dδ (0)
On the other hand

ˆ ˆ
F⃗ · d⃗r = P dx + Q dy
∂Dδ (0) ∂Dδ (0)
= 2π,
which is a constradiction.
5.2.1 New notation

A more compact way to write down the Cauchy-Riemann equations is to introduce the
∂ ∂
,
∂z ∂ z̄
operators. These are the analogs of the partial derivative operators ∂/∂x, ∂/∂y in the complex setting. To
see how to define these operators, note that any point in the plane (x, y) can be described using the (z, z̄)
variables via the formulas
z + z̄ z − z̄
x= , y= .
2 2i
Formally, using chain rule from multivariable calculus (any sensible definition of these operators should
satisfy chain rule after all!), and treating z and z̄ as independent variables, we see that
∂f ∂f ∂x ∂f ∂y
= · + ·
∂z ∂x ∂z ∂y ∂z
1 ∂f 1 ∂f
= +
2 ∂x i ∂y
Similarly we see that

∂f 1 ∂f 1 ∂f
= − .
∂ z̄ 2 ∂x i ∂y
Motivated by these formulae, we define the two operators as
∂ 1 ∂ 1 ∂
:= +
∂z 2 ∂x i ∂y
∂ 1 ∂ 1 ∂
:= −
∂ z̄ 2 ∂x i ∂y
Then for a holomorphic function f (z) the Cauchy-Riemann equations can be re-written as
∂f ∂f
= 0, = f ′ (z).
∂ z̄ ∂z
So holomorphicity implies that the function is independent of the z̄ variable. Note that for any complex
valued two variable function, we then have
∂f ∂f
= .
∂ z̄ ∂z
Remark 5.2.9. (Chain rule in the new notation). Let f and g be function from domains in R2 to R2 .
Suppose f is differentiable at p = (a, b), g is defined in a neighborhood of f (p) and is differentiable at
f (p), then g ◦ f is differentiable at p. Then the chain rule from multivariable calculus is equivalent to
∂g ◦ f ∂g ∂f ∂g ∂ f¯
(p) = (f (p)) · (p) + (f (p)) · (p)
∂z ∂w ∂z ∂ w̄ ∂z
∂g ◦ f ∂g ∂f ∂g ∂ f¯
(p) = (f (p)) · (p) + (f (p)) · (p).
∂ z̄ ∂w ∂ z̄ ∂ w̄ ∂ z̄
Heuristically, this would be the “obvious" chain rule one might write down, if we consider (z, z̄) as two
independent variables of f (and resp. (w, w̄) two independent variables of g). The proof of course requires
an interpretation of the chain rule in terms of multiplication of Jacobian matrices, and the expression above
of the Jacobian matrix in terms of the holomorphic and anti-holomorphic derivatives. In particular, if f
and g are both holomorphic, then
∂g ◦ f ∂g ◦ f
(p) = g ′ (f (p)) · f ′ (p), (p) = 0.
∂z ∂ z̄
Lecture 6
Harmonic functions
6.1 Harmonic functions in the plane and holomorphic functions

For a function of two variables u(x, y) whose first two partials exist and are continuous, the Laplacian is
defined by
∂2u ∂2u
∆u = + 2.
∂x2 ∂y
The function is said to be harmonic if ∆u = 0 at all points. Harmonic functions show up almost everywhere
in physics, and most prominently in electrostatics. For instance the electric potential in a charge free region
is harmonic function! In two dimensions, the study of harmonic functions is equivalent to the study of
holomorphic functions via the following theorem.
Proposition 6.1.1. Let f = u + iv : Ω → C be a holomorphic function, Suppose u and v have continuous
second partial derivatives, then u and v are both harmonic functions.
Proof. The proof is an easy consequence of the Cauchy-Riemann equations. By CR equations, ux = vy ,

and uy = −vx . differentiating the first with respect to ‘x’ and the second equation with respect to ‘y’, and
recalling that since v has continuous second partials, the mixed partials commute, we see that
∆u = uxx + uyy = vyx − vxy = 0.
Conversely we have the following.

Proposition 6.1.2. Let u be a harmonic function on the unit disc DR (z0 ). Then there exists a harmonic
function v on DR (z0 ) such that f := u + iv is a holomorphic function on DR (z0 ). Moreover, if v ′ is another
such function, then v − ṽ is constant on the disc.
Proof. Without loss of generality we may assume that z0 = 0. The uniqueness follows from the Cauchy-
Riemann equations since
∂v ∂v ∂ṽ ∂ṽ
f ′ (z) = +i = +i .
∂y ∂x ∂y ∂x
In particular, ∇(v − ṽ) ≡ 0, and hence v − ṽ is a constant. Let F⃗ = (P, Q) be the vector field obtained by
∂u ∂u
P := − , Q= .
∂y ∂x
35
36 LECTURE 6. HARMONIC FUNCTIONS
Since u is harmonic, we automatically have that curlF⃗ = 0.

Claim. There exists a function v on DR (0) with continuous second partials such that ∇v = F⃗ .
Assuming this the proposition follows since
∂P ∂Q
∆v = + = 0,
∂x ∂y
by Clairaut’s theorem on commuting mixed partials.
Proof of the Claim. We let

ˆ ˆ 1
v(x, y) = F⃗ · d⃗r = P (tx, ty)x + Q(tx, ty)y dt.
l(x,y) 0
Then
ˆ 1
∂v ∂P ∂Q
= (tx, ty) + P (tx, ty) + ty
tx dt
∂x 0 ∂x ∂x
ˆ 1 ˆ 1
d
= t P (tx, ty) dt + P (tx, ty) dt
0 dt 0
1
ˆ 1 ˆ 1
= tP (tx, ty) − P (tx, ty) dt + P (tx, ty) dt
t=0 0 0
= P (x, y).
Similarly we can also prove that

∂v
= Q(x, y).
∂y
6.2 Mean value property

6.3 The Poisson kernel
Part II
Integration theory
37
Lecture 7
Complex Integration
7.1 Curves in the complex plane

A parametrized curve (or simply a curve) in a domain Ω ⊂ C is a continuous function z(t) : [a, b] → Ω.
Writing
z(t) = x(t) + iy(t),
we say that z(t) is differentiable with derivative z ′ (t), if x′ (t) and y ′ (t) exist for all t ∈ (a, b), and then we
set
z ′ (t) = x′ (t) + iy ′ (t).
We call it regular if z ′ (t) exists and z ′ (t) ̸= 0 for all t ∈ (a, b). Geometrically, the vector
x′ (t)î + y ′ (t)ĵ
gives the tangent vector to the curve at the point (x(t), y(t)), and so the complex number z ′ (t) encodes
the information of the tangent vector. Moreover
p
|z ′ (t)| = x′ (t)2 + y ′ (t)2 ,
measures the speed at which the curve is traversed.

For a parametrised curve as above, the points z(a) and z(b) are called the initial and final points of the
curve respectively, and together they are referred to as the end points of the curve. The curve is said to be
closed if z(a) = z(b), and is called simple if z(t) is injective on the interval (a, b).
7.1.1 Orientation
The choice of a parametrization fixes the orientation of the curve. Given a parametrization z(t) : [a, b] → C
of the curve C, we say that w(s) : [c, d] → C is an orientation preserving re-parametrization or that z(t)
and w(s) are equivalent parametrizations, if there exists a strictly increasing function α : [c, d] → [a, b]
with α(c) = a and α(d) = b, such that
w(s) = z(α(s)).
Example 7.1.1. Consider the curve defined by z : [0, 2π] → C where
z(t) = R(cos t + i sin t).
39
40 LECTURE 7. COMPLEX INTEGRATION
The image of the curve is of course a circle of radius R. The tangent vector is given by
z ′ (t) = R(− sin t + i cos t),
and so the speed is |z ′ (t)| = 1. A parametrization is given by w : [0, π] → C,
w(t) = R(cos 2t + i sin 2t),
which traverses the same circle, but with double the speed. On the other hand the parametrization
z(t) = R(cos t − i sin t)
for t ∈ (0, 2π) also describes the same circle with the same speed, but traversed in a clock-wise direction. A
circle is said to be positively oriented if the parametrization traverses the circle in the anti-clockwise direction,
while negatively oriented otherwise.
We will need to consider slightly more general curves. A curve z(t) : [a, b] is said to be piecewise regular if
there is a partition
a = t0 < t1 < · · · < tn = b
such that z(t) restricted to each (ti−1 , ti ) is a regular curve.
7.1.2 Notation and conventions.

If the image of restriction of the curve z(t) to the interval (ti−1 , ti ) is denoted by Ci , and the image of the
full curve is denoted by C, we then write
Xn
C= Cn .
i=1
We denote by −C, the curve C traced in the opposite direction. For instance, if z(t) : [a, b] → C is a
parametrization for C, a parametrization for −C is given by z − (s) : [a, b] → C where
z − (s) = z(a + b − s).
For a positive integer a > 0, we denote by aC to be the curve C traversed ‘a’ times. A circle CR (p) or
|z − p| = R, unless otherwise specified, will always mean a circle of radius R centred at p traversed once
in the anti-clockwise direction.
7.2 Complex line integrals

For a continuous function f = u + iv : [a, b] → C of one real variable, we can extend the definition of
integration by defining
ˆ b ˆ b ˆ b
f (t) dt := u(t) dt + i v(t) dt.
a a a
Now, suppose we are given a smooth curve as above, and a function f : Ω → C, we then define the complex
integral along the curve by
ˆ ˆ b
f (z) dz := f (z(t))z ′ (t) dt,
C a
where C denotes the image of the curve. Note that the multiplication above is the complex multiplication.
It is convenient to think of dz as a complex differential, representing an infinitesimal complex change, and
given by
dz = dx + idy.
7.2. COMPLEX LINE INTEGRALS 41
So if f = u + iv, then
ˆ ˆ ˆ
f (z) dz = (u dx − v dy) + i (u dy + v dx),
C C C
where the integrals on the right are now the usual line integrals from multivariable calculus.
Remark 7.2.1. Recall that the differential forms dx and dy are defined to be duals of the vector fields
∂/∂x and ∂/∂y in the xy-plane. That is,
∂ ∂ ∂ ∂
dx = dy = 1, dx = dy = 0.
∂x ∂x ∂y ∂x
Then it is easy to compute that dz = dx + idy and dz̄ = dx − idy are dual complex valued differential
forms to the complex valued vector fields ∂/∂z and ∂/∂ z̄ vector fields defined in the previous lecture.
To make sure the integral is well defined, we need to show that it is independent of orientation preserving
parametrizations.
Lemma 7.2.2. Let C be a curve with parametrization z(t). Let w(s) = z(α(s)) be another orientation
preserving parametrization, where α : [a, b] → [c, d]. Then
ˆ ˆ
f (w) dw = f (z) dz.
C C
Proof. By the chain rule, since w′ (s) = z ′ (α(s))α′ (s), we see that
ˆ ˆ b ˆ b
f (w) dw = f (w(s))w′ (s) ds = f (z(α(s))z ′ (α(s))α′ (s) ds.
C a a
Putting t = α(s), we see that α′ (s)ds = dt, and hence

ˆ ˆ b ˆ
′
f (w) dw = f (z(t))z (t) dt = f (z) dz.
C a C
This shows that the definition of integration is independent of the parametrization chosen.
We can now extend the definition of complex integrals to piecewise smooth curves by linearity. That is, if
C = C1 + · · · + Cn is a piecewise smooth curve with Cj smooth curve for all j = 1, · · · , n, then we define
ˆ Xn ˆ
f (z) dz := f (z) dz.
C j=1 Cj
´
Similar to our definition of C
f (z) dz we can define the integral with respect to dz̄ by
ˆ ˆ
f (z, z̄) dz̄ := f (z) dz.
C C
We can also define the integral with respect to arc-length. For a complex or real valued function f (z) and
a curve z = z(t) : [a, b] → C we also define
ˆ ˆ b
f (z) |dz| := f (z)|z ′ (t)|dt.
C a
We then define the length of the curve by
ˆ
len(C) = |dz|.
C
We next state, without proof, some basic properties of the complex line integral. The proofs follow from
the definition of the complex integral and corresponding properties of the Riemann integral.
Proposition 7.2.3. Let C be a parametric piecewise regular curve in an open set Ω ⊂ C.

1. For any complex numbers a, b and any complex valued functions f and g we have that
ˆ ˆ ˆ
[af + bg](z) dz = a f (z) dz + b g(z) dz.
C C C
2. For integers aj ∈ Z and piecewise smooth curves Cj , if we denote by C = a1 C1 + · · · + an Cn , we have

ˆ Xn ˆ
f (z) dz = aj f (z) dz.
C j=1 Cj
In particular, ˆ ˆ
f (z) dz = − f (z) dz.
−C C
3. (Triangle inequality)
ˆ ˆ
f (z) dz ≤ |f (z)| |dz| ≤ sup |f (z)| · length(C),
C C z∈C
with all inequalities replaced with equality if and only if f is a constant real number.
u.c
4. If fn −−→ f , then ˆ ˆ
lim fn (z) dz = f (z) dz.
n→∞ C C
7.3 A fundamental computation

We now compute the integrals ˆ
z n dz,
CR
where CR = {z | |z| = R} is the circle of radius R centered at the origin positively oriented and traversed
once. It is not an overstatement to claim that this integral, although elementary, plays a fundamental role in
complex analysis. A parametrization of the circle is given by z(t) = Reiθ , θ ∈ [0, 2π]. Then dz = iReiθ dθ,
and so
ˆ ˆ 2π
z n , dz = Rn einθ (iReiθ ) dθ
CR 0
ˆ 2π
= iRn+1 ei(n+1)θ dθ
0
Now if n + 1 ̸= 0, then ˆ 2π θ=2π

1
ei(n+1)θ dθ = ei(n+1)θ = 0,
0 (n + 1)i θ=0
since eiθ is periodic with period 2π. On the other hand, if n = −1, then
ˆ ˆ 2π ˆ 2π
z n , dz = iRn+1 ei(n+1)θ dθ = i dθ = 2πi.
CR 0 0
So summarizing, we have the following:

ˆ (
1 n 0, n ̸= −1
z dz =
2πi |z|=R 1, n = −1.
In particular, the integral is independent of the radius R. More generally, we have the following.
7.4. PRIMITIVES 43
Proposition 7.3.1. Let D be any disc in C, and let p be a point not lying on the boundary circle C = ∂D. If
C is traversed only once with positive orientation, then
ˆ (
1 1 1, p ∈ D
n(C; p) := dz = .
2πi C z − p 0, p ∈/ D.
Moreover, if n ̸= −1 and n ∈ Z, then ˆ

(z − p)n dz = 0.
C
The number n(C, p) is called the index or the winding number of the circle C around p.
Proof. Without loss of generality, we can assume that z0 = 0. Now suppose p ∈ D. Then for any z ∈ C,
|z| > |p|. Then by the geometric series expansion, for z ∈ C we have
∞
1 1 1 1 X pn−1
= · = + .
z−p z 1 − p/z z n=2 z n
Integrating both sides (this can be done since convergence is uniform), and using the above computation,
we see that n(C, p) = 1. On the other hand, if p ∈/ D, then |z| < |p| for all z ∈ C, and hence
∞
1 1 1 1 X zn
=− · =− .
z−p p 1 − z/p p n=0 pn
Again integrating both sides, we see that n(C, p) = 0, since there are only positive powers of z on the
right. For the second part, if n > 0, then the integrand is a polynomial and hence the integral is zero by
the above computations. If n = −m < 0, then we can write
1 m
(z − p)n = .
z−p
Once again using the geometric series expansions above, in both cases, there will no terms with exponent
−1. And hence by the computation above, the integral will be zero.
Remark 7.3.2. Later in the course, we will define the index n(γ, p) of a general curve γ around a point
p by a similar formula, and we shall prove (rather indirectly) that the index of any closed curve is always
an integer. Assuming this, we can provide a more conceptual explanation of the above result. It is clear
that the n(C, p), as a function of p defined on the open set C \ C is a continuous function. But then being
integer valued, it must be locally constant. From our elementary observation, n(C, 0) = 1, and hence
n(C, p) = 1 for all p ∈ D. On the other hand, clearly as |p| → ∞, n(C, p) approaches 0. Again by virtue
of being locally constant, this implies that n(C, p) = 0 for all p ∈ C \ D.
7.4 Primitives
We then have the following theorem, which is a generalization of the fundamental theorem for line inte-
grals from multivariable calculus.
Proposition 7.4.1 (Fundamental theorem for complex integrals). If C is any curve joining the point p to q,
then ˆ
F ′ (z) dz = F (q) − F (p).
C
Proof. Let z(t) : [0, 1] → C be a parametrization for C such that p = z(0) and q = z(1). Then
ˆ ˆ 1
F ′ (z) = F ′ (z(t))z ′ (t) dt.
C 0
But by Chain rule, if we let g(t) = F (z(t)), then g ′ (t) = F ′ (z(t))z ′ (t), and so
ˆ ˆ 1
′ dg
F (z) dz = dt = g(1) − g(0),
C 0 dt
where we use the usual one variable fundamental theorem of calculus. But g(0) = F (z(0)) = F (p) and
g(1) = F (q), and this completes the proof.
Recall that an open set is called connected is any two points can be joined by a continuous curve lying
completely inside the open set. An important and immediate consequence of the fundamental theorem is
the following.
Corollary 7.4.2. Let Ω ⊂ C be an open connected subset, and f : Ω → C be holomorphic. Then f ′ (z) = 0
for all z ∈ Ω if and only if f (z) is a constant.
For a domain Ω ⊂ C, and F, f : Ω → C complex valued functions, we say that F (z) is a primitive of f (z)
if
F ′ (z) = f (z)
for all z ∈ Ω. Then another direct corollary of the above theorem is the following.
Corollary 7.4.3. Suppose f : Ω → C has a primitive on F , then
ˆ
f (z) dz = 0
C
for every closed curve C ⊂ Ω.

Using the above corollary, we can explain the results of the calculation above. Recall that for any integer
n ̸= −1, we have that
1 d
z n+1 = z n .
n + 1 dz
Or in other words, for n ̸= −1, z n has a primitive at least on C\{0}. Hence the integral on any closed loop
not passing through z = 0, is always zero. In particular, the integrals around |z| = R are zero. That leaves
the case when n = −1. We have already seen (as a consequence of the chain rule) that if a holomorphic
logarithm log z can be defined, then it is a natural primitive for 1/z. But we saw some lectures back, we
saw that going around a circle centered at the origin, makes the logarithm function discontinuous, leave
alone non-holomorphic. In fact, combining the corollary with the calculations above, we have managed to
prove the following.
Proposition 7.4.4. Let Ω ⊂ C \ {0} be an open set containing at least one circle |z| = r. Then there is no
holomorphic function F : Ω → C such that
eF (z) = z.
In particular, there cannot be a holomorphic logarithm defined on all of C∗ , or indeed on any punctured
neighbourhood Dr (0) \ {0} of 0.
In fact, the theorem can also be used to explain the fact that n(∂D, p) = 0 when p ∈/ D. Again without
loss of generality, we assume that D = DR (0). Recall that log w can indeed be defined via a power series
in the region |w − 1| < 1, namely
∞
X (−1)n−1
log w = (w − 1)n ,
n=1
n
7.4. PRIMITIVES 45
and it satisfies (log w)′ = w−1 . Now putting w = z/p − 1, f (z) = log(z/p − 1) is a holomorphic function
in the region |z| < |p| with
1
f ′ (z) = .
z−p
Since C is completely contained in the region |z| < |p|, by Corollary 7.4.3, n(C, p) = 0.
7.4.1 A differential forms interpretation

Recall that (real) differential one forms on an open subset Ω ⊂ R2 are formal C ∞ (Ω, R)-linear combina-
tions of the symbols dx and dy. The space of one forms is usually denoted by A1 (Ω). But one can just
as well consider the set A1C (Ω) of complex valued differential forms where instead of smooth real valued
functions, one considers C ∞ (Ω, C)-linear combinations. Instead of using dx and dy as bases, it is more
convenient to use
dz = dx + idy, dz̄ = dx − idy
as bases elements, so that a general element of A1C (Ω) can then be written as
α = f (z, z̄)dz + g(z, z̄)dz̄.
The form α is real if and only if f¯ = g. A form is said to be of type (1, 0) if b = 0. Recall that one can
integrate (real) differential one-forms on curves. Indeed if α = adx + bdy, and γ = z(t) = x(t) + iy(t) :
[0, 1] → C is a C 1 -curve, then one defines
ˆ ˆ 1
α := a(z(t))x′ (t) + b(z(t))y ′ (t) dt.
γ 0
One can extend this definition to integrating complex one forms by simply using linearity. That is, if
η = α + iβ is a complex valued one form then we simply define
ˆ ˆ ˆ
η= α + i β.
γ γ γ
One can now easily check that if η = f (z) dz is a form on type (1, 0), then
ˆ ˆ 1
f (z) dz = f (γ(t))γ ′ (t) dt,
γ 0
and so our definition of complex integration is simply integration of complex valued forms of type (1, 0).
Lecture 8
Cauchy’s theorem : Local versions
In the previous lecture, we saw that if f has a primitive in an open set, then
ˆ
f dz = 0
γ
for all closed curves γ in the domain. This was a simple application of the fundamental theorem of calculus.
It is somewhat remarkable, that in many situations the converse also holds true. In the next few lectures
we will explore this theme, and prove theorems that will form the basis of all that we will accomplish in
the rest of the course. The simplest version is the following:
Theorem 8.0.1 (Cauchy’s theorem on a disc). Let D be a disc in the complex plane. If f : D → C is
holomorphic, then
ˆ
f dz = 0
γ
for all closed curves γ contained in D.

We will prove Cauchy’s theorem on a disc by showing that all holomorphic functions in the disc have a
primitive. The key technical result we need is Goursat’s theorem.
Theorem 8.0.2 (Goursat). Let Ω ⊂ C be an open subset, and T ⊂ Ω be a triangle contained inside Ω. If
f : Ω → C is holomorphic, then
ˆ
f (z) dz = 0.
T
From this we immediately obtain a Goursat theorem for rectangles. We leave the proof as an exercise.
Corollary 8.0.3. Let Ω ⊂ C be an open subset, and R ⊂ Ω be a triangle contained inside Ω. If f : Ω → C
is holomorphic, then
ˆ
f (z) dz = 0.
R
Remark 8.0.4 (Relation to Green’s theorem.). If F = P i + Qj is a vector field, such that P and Q have
continuous first partials, then for any close curve
ˆ ˆ
(Qx − Py ) dx dy = P dx + Q dy.
R ∂R
47
48 LECTURE 8. CAUCHY’S THEOREM : LOCAL VERSIONS
Now, suppose that f = u + iv and that u and v have continuous partials. Then
ˆ ˆ ˆ
f dz = (u dx − v dy) + i (u dy + v dx).
γ γ γ
Consider the vector fields (P1 , Q1 ) = (u, −v) and (P2 , Q2 ) = (v, u). Then for each j = 1, 2, by the
Cauchy Riemann equations
∂Qj ∂Pj
= .
∂x ∂y
Then by Green’s theorem, the line integral is zero and we recover Cauchy’s theorem (even for general
domains). The important point is our assumption that u and v have continuous partials, while in Cauchy’s
theorem we only assume holomorphicity which a priori only guarantees the existence of the partial deriva-
tives. Later in the course, we will in fact show that holomorphicity does imply that the first partials are
continuous partials (actually we will prove that partial derivatives of all orders exists and are continu-
ous), but that will be a consequence of Cauchy’s theorem. And hence we need a proof that avoids Green’s
theorem. In fact the methods we use also yield a proof of Green’s theorem.
8.1 Proof of Cauchy’s theorem assuming Goursat’s theorem

Cauchy’s theorem follows immediately from the theorem below, and the fundamental theorem for complex
integrals (cf. Proposition 7.4.1).
Theorem 8.1.1. Let f : D → C be a holomorphic function. Then f (z) has a primitive on D.
Proof. We first observe that By translation, we can assume without loss of generality that the disc D is
centered at the origin. For any points z, w ∈ C, we denote by lz,w the straight line segment from z to w.
For any z = (x, y) ∈ D, let γz denote the path from the origin to z consisting of a horizontal segment
from 0 to (x, 0) followed by a vertical segment from (x, 0) to (x, y). We then define
ˆ
F (z) = f (w) dw,
γz
and claim that F (z) is holomorphic with F ′ (z) = f (z). To see this, first note that if z1 = (x, 0), then for
any small h ∈ R,
ˆ ˆ ˆ ˆ ˆ
f (w) dw − f (w) dw = f (w) dw + f (w) dw − f (w) dw.
γz+h γz lz1 ,z1 +h lz1 +h,z+h lz1 ,z
On the other hand, by Theorem 8.0.2,

ˆ ˆ ˆ ˆ
f (w) dw + f (w) dw + f (w) dw + f (w) dw = 0,
lz1 ,z1 +h lz1 +h,z+h lz+h,z lz,z1
and so we have the key identity:

ˆ
F (z + h) − F (z) = f (w) dw. (8.1)
lz,z+h
Intuitively, if |h| << 1, then f (w) ≈ f (z) on lz,z+h , and so the integral is approximately f (z)h. To make
this rigorous, we write
ˆ ˆ ˆ
f (w) dw = f (z) dw + (f (w) − f (z)) dw
lz,z+h lz,z+h lz,z+h
8.2. PROOF OF GOURSAT’S THEOREM 49
ˆ
= f (z)h + (f (w) − f (z)) dw,
lz,z+h
and so ˆ
F (z + h) − F (z) 1
= f (z) + (f (w) − f (z)) dw.
h h lz,z+h
Dividing by h and using triangle inequality,

ˆ ˆ
F (z + h) − F (z) 1 1
− f (z) = (f (w) − f (z)) dw ≤ |f (w) − f (z)| |dw|.
h h lz,z+h |h| lz,z+h
By the continuity of f , given any ε > 0 there exists a δ such that if |h| < δ, then for all w ∈ lz,z+h
|f (w) − f (z)| ≤ ε.
Using this in the above estimate, and remembering that length(lz,z+h ) = |h| we see that
F (z + h) − F (z)
− f (z) ≤ ε,
h
so long as |h| ≤ δ. This shows that
∂F F (z + h) − F (z)
(z) := lim = f (z),
∂x h→0 h
Next, consider a path σz consisting of a vertical line segment from 0 to iy followed by a horizontal segment
from iy to z. By Theorem 8.0.2, ˆ
F (z) = f (w) dw.
σz
By an argument similar to the one above, we can prove that ∂F/∂y exists, and that
∂F f (z + ik) − f (z)
(z) = lim = if (z).
∂y k→0 k
The analog of the key identity is that
ˆ
F (z + ik) − F (z) = f (w) dw,
lz,z+ik
which is approximately if (z)k. (integrating in the vertical direction incurs an i). In any case, this shows
that the partials of F exist and are continuous, and hence F is a (totally) differentiable map from D to R2 .
On the other hand, since
∂F ∂F
(z) = i (z),
∂y ∂x
the partials also satisfy the Cauchy-Riemann equations. Hence F is complex differentiable at z, and more-
over, F ′ (z) = f (z).
8.2 Proof of Goursat’s theorem

The main idea of the proof is to localize the integral. We start by choosing a sequence of “nested" triangles
Tn
int(Tn+1 ) ⊂ int(Tn ),
with T0 = T . Note that when we say triangle we mean the one-dimensional object, and we denote by
int(T ) the region bounded by the triangle. Suppose we have already constructed the triangle Tn−1 . The
first step in the construction of Tn is to bisect each side of Tn−1 . This results in four new triangles which
(1) (2) (3) (4)
we label Tn−1 , Tn−1 , Tn−1 and Tn−1 . We also give them the an orientation consistent with the original
triangle (see figure) so that the integrals over the common boundaries cancel, and we have that
ˆ 4 ˆ
X
f (z) dz = f (z) dz.
(j)
Tn−1 j=1 Tn−1
By triangle inequality,
ˆ 4
X ˆ
f (z) dz ≤ f (z) dz ,
(j)
Tn−1 j=1 Tn−1
(j)
and so for at least one triangle Tn−1 ,
ˆ ˆ
f (z) dz ≤ 4 f (z) dz .
(j)
Tn−1 Tn−1
Choose any one such triangle, and label it Tn . Inductively, we have

ˆ ˆ
f (z) dz ≤ 4n f (z) dz . (8.2)
T Tn
Recall that the diameter of a subset of C is the maximum distance between any two points in that subset.
We then have the following elementary observation.
Lemma 8.2.1. If dn and pn denote the diameter and perimeter of the triangle T (n) respectively, then
dn = 2−n d0 , and pn = 2−n p0 .
Moreover, there exists a unique p ∈ ∩∞

n=0 int(Tn ).
(1) (2) (3) (4)

Proof. The diameter of a triangle is the length of the longest side. Since Tn−1 , Tn−1 , Tn−1 and Tn−1
are all congruent triangles with side lengths that are half of the respective side lengths of T (n−1) , clearly
dn = 2−1 dn−1 and pn = 2−1 pn−1 , and we obtain the result by induction. Now, let xn ∈ Tn be any point.
Then since the sequence {xn } is contained in T0 , a compact subset, there exists a limit point p ∈ T = T0 ,
which will trivially lie inside all the triangles. So the intersection is non-empty. On the other hand since
limn→∞ dn = 0 the intersection can contain at most one point.
Continuing with the proof of Goursat’s theorem, since f is holomorphic at z = p, we can write
f (z) = f (p) + f ′ (p)(z − p) + ψ(z)(z − p),
where ψ(z) → 0 as z → p. Now the constant function f (p) and the linear function f ′ (p)(z − p) both have
primitives; f (p)z and f ′ (p)(z − p)2 /2 respectively. So by the fundamental theorem, their line integrals on
Tn are zero. So ˆ ˆ
f (z) dz = ψ(z)(z − p) dz.
Tn Tn
Now since dn → 0 as n → ∞, given any ε > 0 there exists a n such that
|ψ(z)| < ε
8.3. EXTENSIONS TO PUNCTURED DOMAINS 51
on Tn . But then ˆ
f (z) dz ≤ ε sup |z − p|pn = εdn pn ≤ 4−n εd0 p0 .
Tn Tn
Then by the inequality in (8.2) ˆ

f (z) dz ≤ εd0 p0 ,
T
which can be made arbitrarily small. Hence

ˆ
f (z) dz = 0,
T
and this proves the theorem in the case that f is holomorphic everywhere.
Remark 8.2.2. The reader should attempt to understand why a similar argument would fail if f is merely
assumed to be smooth (as opposed to holomorphic) as a function of two variables.
8.3 Extensions to punctured domains

For applications, we need a slightly stronger version of Cauchy’s theorem, which in turn relies on a slightly
stronger version of Goursat’s theorem. For any open set Ω ⊂ C, and any p ∈ Ω we denote Ω∗p := Ω \ {p}.
Theorem 8.3.1. Let D be a disc, and p ∈ D. Let f : Dp∗ → C be a holomorphic function such that
limz→p (z − p)f (z) = 0. Then for any closed curve γ ⊂ Dp∗ ,
ˆ
f (z) dz = 0.
γ
As before, the kye technical input is a version of Goursat’s theorem. This time we will state the version
only for rectangles.
Theorem 8.3.2. We only need to prove the theorem in the case that p ∈ int(R). Let Ω be any open subset of
C. For some p ∈ Ω, let f : Ω∗p → C be a holomorphic function satisfying limz→p (z − p)f (z) = 0. Then for
any rectangle R ⊂ Ω with p ∈ /R ˆ
f (z) dz = 0.
R
Proof. Without loss of generality, we may assume that p lies in the interior of the region bounded by R.
Let ε > 0, and let δ > 0 such that
ε
|f (z)| ≤ ,
|z − p|
whenever |z −p| < δ. Let R0 be a small square of side length δ with p at it’s centre. Note that |z −p| > δ/2
for all z ∈ ∂R. By extending the sides of R0 , divide R into nine rectangles R0 , · · · , R8 . Clearly
ˆ
f (z) dz = 0
Rj
for j = 1, 2, · · · , 8 by Theorem 8.0.2. Since the integrals over the common boundaries cancel out if we
choose the correct (ie. anti-clockwise) orientations
ˆ ˆ
R R0
Next, we estimate the integral over R0 ,

ˆ ˆ
|dz| 2ε
f (z) dz ≤ ε < len(∂R0 ) = 8ε.
∂R0 ∂R0 |z − p| δ
´
Since ε is arbitrary, this shows that ∂R
f (z) dz = 0.
Proof of Theorem 8.3.1. The idea of the proof is to again show that f (z) has a primitive on Dp∗ , and we
proceed as in the proof of Theorem 8.1.1. Let p = (a, b). Pick a point z0 = (x0 , y0 ) such that x0 ̸= a
and y0 ̸= b. Let z + (x, y) ∈ Dp∗ . If x ̸= a, we let γz0 ,z be the path consisting of the line segment
(x0 , y0 ) to (x, y0 ), followed by a vertical line from (x, y0 ) to (x, y). On the other hand, if x = a, then we
let γz0 ,z consist of three segments - A vertical segment from (x0 , y0 ) to (x0 , yη ) followed by a horizontal
segment from (x0 , yη ) to (x, yη ) followed by another vertical segment from (x, yη ) to (x, y). We then
define F : Dp∗ → C by ˆ
F (z) = f (w) dw.
γz0 ,z
First, we claim that ∂F/∂x = f (z) for every z ∈ Dp∗ . Note that the key step in the proof of Theorem
8.1.1 is to obtain the identity (8.1). We obtain the same identity in this new situation. Let h ∈ R be a small
number. If x ̸= a, then the same argument as before will imply that ∂F/∂x exists at (x, y) and equals
f (z). So suppose x = a. Let R be the rectangle with vertices z0 , (a + h, y0 ), (a + h, yη ) and (x0 , yη ) and
let R′ be the rectangle with vertices (a, yη ), (a + h, yη ), z + h = (a + h, y) and z = (a, y). Note that p lies
either in the interior or the exterior of R (but not on the boundary), and so by Theorem 8.0.2 or Theorem
8.3.2, ˆ
f (w) dw = 0.
∂R
On the other hand, p lies in the exterior of R′ , and so by Theorem (8.0.2),

ˆ
f (w) dw = 0.
∂R′
It is easy to check that ˆ

F (z + h) − F (z) = f (w) dw.
lz,z+h
Now the argument proceeds exactly as in the proof of Theorem 8.1.1. Next, as before, one proves that
∂F/∂y exists everywhere on Dp∗ and equals if (z) using a suitable modification of the path σ.
Lecture 9
Cauchy’s integral formula
9.1 Cauchy integral formula and Analyticity of holomorphic func-

tions
The main consequence of Cauchy’s theorem for the punctured disc is the following beautiful formula - a
fundamental mean value property or a localization property. It is not an exaggeration to say that all of
complex analysis is contained in this formula, if only a little bit of cleverness is added to the mix.
Theorem 9.1.1 (Cauchy integral formula (CIF)). Let f : Ω → C be a holomorphic function. If Dr (z0 ) ⊂ Ω,
then for any z ∈ Dr (z0 ), ˆ
1 f (ζ)
f (z) = dζ.
2πi |ζ−z0 |=r ζ − z
Proof. For a fixed z ∈ DR (z0 ), we define hz : D∗ := DR (z0 ) \ {z} → C by

f (ζ) − f (z)
hz (ζ) = .
ζ −z
Clearly hz is holomorphic in D∗ . Also limζ→z (ζ − z)hz (ζ) = 0, since f is holomorphic, and hence
continuous at z. Then by Cauchy’s theorem
ˆ ˆ 1 ˆ
1 1 f (ζ) 1
0= hz (ζ) = dζ − f (z) dζ
2πi |ζ−z0 |=r 2πi |ζ−z0 |=r ζ − z 2πi |ζ−z0 |=r ζ − z
ˆ
1 f (ζ)
= dζ − f (z),
2πi |ζ−z0 |=r ζ − z
since the quantity in the bracket is simply n(C, z), where C = ∂Dr (z0 ), which is equal to 1 because
z ∈ Dr (z0 ).
An immediate consequence of the Cauchy integral formula is that holomorphic functions are analytic
Theorem 9.1.2. Let Ω ⊂ C open, and f : Ω → C a holomorphic function. Then f is analytic.P∞ Moreover, if
DR (z0 ) is any disc whose closure is contained in Ω, then for all z ∈ DR (z0 ), f (z) = n=0 an (z − z0 )n ,
where ˆ
1 f (ζ)
an = dζ. (9.1)
2πi |ζ−z0 |=R (ζ − z0 )n+1
In particular, holomorphic functions are infinitely complex differentiable.
53
54 LECTURE 9. CAUCHY’S INTEGRAL FORMULA
Proof. By CIF, if D is a of radius R centered at a with boundary circle CR , then for any z ∈ D,
ˆ
1 f (ζ)
f (z) = dz.
2πi |ζ−a|=R ζ −z
Writing ζ − z = (ζ − z0 ) − (z − z0 ), we see that
1 1 z − z0 −1
= 1− .
ζ −z ζ − z0 ζ − z0
For z ∈ D and ζ ∈ CR , |z − z0 | < R = |ζ − z0 |, or equivalently |(z − z0 )/(ζ − z0 )| < 1, and hence using
the geometric series
∞
1 1 X z − z0 n
= .
ζ −z ζ − z0 n=0 ζ − z0
Since power series converge uniformly, we can also integrate term-wise (see Appendix), we get that
ˆ ∞
f (ζ) X z − z0 n
2πif (z) = dζ
CR ζ − z0 n=0 ζ − z0
∞ ˆ
X f (ζ)
= n+1
dζ · (z − z0 )n
n=0 CR
(ζ − z0 )
∞
X
= 2πi an (z − z0 )n ,
n=0
where an is given by the formula (9.1), and this completes the proof of the theorem. Infinite complex
differentiability follows from analyticity by Corollary 1.1 from Lecture-3.
As an immediate corollary, we have the following versions of the principle of analytic continuation for
holomorphic functions.
Corollary 9.1.3 (Principle of analytic continuation). Let f : Ω → C be a holomorphic function, and suppose
Ω is connected.
1. If there exists a p ∈ Ω such that f (n) (p) = 0 for n = 1, 2, · · · , then f is a constant function.
2. If there exists a n open subset U such that f ≡ 0, then f ≡ 0.

U Ω
This follows immediately from Theorem 9.1.2 above and Corollary 3.3.5 from Lecture-4. Another important
consequence of the analyticity is the following criteria for holomorphicity.
Corollary 9.1.4 (Morera). Any continuous function on an open set Ω that satisfies
ˆ
f (z) dz = 0
∂R
for all rectangular regions R ⊂ Ω, is holomorphic.
Proof. Let p ∈ Ω and r > 0 such that Dr (p) ⊂ Ω. The given condition is equivalent to f having a primitive
Fp (z) on Dr (p), essentially by the proof of Theorem 2.1 in Lecture-7. Note that continuity of f is crucial
for this. But then by Theorem 9.1.2, Fp′ (z) = f (z) is also holomorphic on Dr (p). In particular, f is complex
differentiable at p. Since this is true for all p ∈ Ω, f ∈ O(Ω).
9.2. CAUCHY INTEGRAL FORMULA FOR DERIVATIVES 55
9.2 Cauchy integral formula for derivatives

Let f : Ω → C be holomorphic. Then by Theorem 9.1.2, the nth complex derivative f (n) (z) exists for all
z ∈ Ω. Moreover, by equation (9.1), for any z ∈ Ω,
ˆ
(n) n! f (ζ)
f (z) = dζ,
2πi |ζ−z0 |=r (ζ − z0 )n+1
where r > 0 such that Dr (z0 ) ⊂ Ω. More generally, just as for the Cauchy integral formula, one can obtain
a similar formula for f n (z0 ) where the integral is over a circle centred at possibly a point other than z0 .
Theorem 9.2.1 (Cauchy integral formula for derivatives). If D is a disc with boundary C whose closure is
contained in Ω, then for any z ∈ D, we have
ˆ
(n) n! f (ζ)
f (z) = dζ. (9.2)
2πi C (ζ − z)n+1
Remark 9.2.2. Essentially what this theorem says is that one can differentiate the Cauchy integral for-
mula, and take the derivative inside the integral.
Proof. Since f (z) is analytic in the neighborhood of any point z ∈ Ω, it is automatically holomorphic in
Ω. To prove the formula (9.2) we use induction. For n = 0, this is simply the Cauchy integral formula.
ˆ
f (n−1) (z) − f (n−1) (z0 ) (n − 1)! 1 1
= f (ζ) − dζ
z − z0 2πi(z − z0 ) C (ζ − z)n (ζ − z0 )n
ˆ
(n − 1)! (ζ − z0 )n−1 + · · · + (ζ − z)n−1
= f (ζ) · dζ
2πi C (ζ − z)n (ζ − z0 )n
ˆ n−1
(n − 1)! f (ζ) X (ζ − z0 )k
= n
dζ.
2πi C (ζ − z0 ) (ζ − z)k+1
k=0
Suppose D = DR (p), and suppose |f (ζ)| < M on C. By choosing z sufficiently close to z0 , we can ensure
that for all ζ ∈ C,
|ζ − z| > (R − |z0 − p|)/2 := λ(z0 ).
Also, trivially, |ζ − z|, |ζ − z0 | < 2R and |ζ − z0 | > λ(z0 ). Then for any k < n,
(ζ − z0 )k 1 |z − z0 |(|ζ − z|k + · · · |ζ − z0 |k ) 2R n+2

− = < n|z − z0 | .
(ζ − z)k+1 ζ − z0 |ζ − z|k+1 |ζ − z0 | λ(z0 )
Hence, given any ε > 0, there exists a δ = δ(n, R, M, z0 ) > 0 such that for any |z − z0 | < δ, and any
ζ ∈ C, we have
f (ζ) (ζ − z0 )k f (ζ)
n k+1
− < ε.
(ζ − z0 ) (ζ − z) (ζ − z0 )n+1
uniformly in the sense that the rate of convergence is independent of ζ. So if |z − z0 | < δ, then for each k,
ˆ ˆ
(n − 1)! f (ζ) (ζ − z0 )k (n − 1)! f (ζ)
n k+1
dζ − dζ ≤ (n − 1)!Rε,
2πi C (ζ − z0 ) (ζ − z) 2πi C (ζ − z0 )n+1
and hence
ˆ ˆ
(n − 1)! f (ζ) (ζ − z0 )k (n − 1)! f (ζ)
lim dζ = dζ.
z→z0 2πi C (ζ − z0 ) (ζ − z)k+1
n 2πi C (ζ − z0 )n+1
56 LECTURE 9. CAUCHY’S INTEGRAL FORMULA
Summing up over k, we see that

ˆ
(n) n! f (ζ)
f (z0 ) = dζ.
2πi C (ζ − z0 )n+1
An extremely useful consequence is the following estimate on the derivatives of a holomorphic function.
Corollary 9.2.3. [Cauchy’s estimate] Let f be a holomorphic function in an open set containing the closure
of a disc DR (z0 ). If we denote the boundary of the disc by C, then for any z ∈ DR (z0 ),
n!R
|f (n) (z)| ≤ ||f ||C ,
(R − |z − z0 |)n+1
where ||f ||C := supζ∈C |f (ζ)|. In particular,
n!
f (n) (z0 ) ≤ ||f ||C ,
Rn
Proof. By CIF for derivatives, we have that
ˆ
n! f (ζ)
f (n) (z) = dζ.
2πi C (ζ − z)n+1
Applying the triangle inequality, and remembering that on C, |ζ − z| ≥ R − |z − z0 |,
ˆ
n! |f (ζ)|
|f (n) (z)| ≤ |dζ|
2π C |ζ − z|n+1
n!
≤ sup |f (ζ)| · len(C)
2π(R − |z − z0 |)n+1 C
n!R
= ||f ||C .
(R − |z − z0 |)n+1
9.3 Liouville Theorem

Recall that an entire function is a function that is holomorphic on the entire complex plane C. We then
have the following surprising fact.
Theorem 9.3.1 (Liouville). There are no bounded non-constant entire functions.
Proof of Theorem 24.3.4. Let f (z) be a bounded entire function. We show that it then has to be a constant.
Suppose |f (ζ)| < M for all ζ ∈ C, and let z ∈ C be an arbitrary point. Since f is entire, it is holomorphic
on any disc DR (z). Denoting the boundary by CR , by the estimate above,
||f ||CR M
|f ′ (z)| ≤ < .
R R
Letting R → ∞ the right hand side approaches zero, and hence f ′ (z) = 0 for all z ∈ C. Since C is
connected, this forces f (z) to be a constant, completing the proof of Liouville’s theorem.
Remark 9.3.2. More generally, we can show that an entire function f (z) satisfying
|f (z)| ≤ M (1 + |z|α ),
for some constants M, α > 0 and all z ∈ C, has to be a polynomial of degree at most ⌊α⌋, where ⌊·⌋ is the
usual floor function. We leave this as an exercise.
Lecture 10
Further applications of Cauchy

integral formula
In this lecture, we give three further applications of the Cauchy integral formula.
10.1 The fundamental Theorem of algebra

Recall that in the previous lecture we proved Liouville’s theorem, namely that there are no bounded, non-
constant, entire functions. As a simple and beautiful consequence of this, we can prove the fundamental
theorem of algebra, namely that any polynomial can be completely factored into linear factors over complex
numbers.
Theorem 10.1.1. Any non-constant polynomial p(z) has a complex root, that is there exists an α ∈ C such
that p(α) = 0. As a consequence, any polynomial is completely factorable, that is we can find α1 , · · · , αn
(some of them might be equal to each other), and c ∈ C such that
p(z) = c(z − α1 ) · · · (z − αn ),
where n = deg(p(z)).
Proof. For concreteness, let

p(z) = an z n + · · · + a1 z + a0 ,
where an ̸= 0. Then we claim that there exists an R such that
|an | n
|z| ≤ |p(z)| ≤ 2|an ||z|n ,
2
whenever |z| > R. To see this, by the triangle inequality,
|an−1 | |a0 | 3
|p(z)| ≤ |an ||z|n + · · · |a1 ||z| = |a0 | = |an ||z|n 1 + + · · · + n ≤ |an ||z|n
|z| |z| 2
if |z| > R and R is sufficiently big. For the other inequality, we use the other side of the triangle inequality.
That is,
an−1 a0
|p(z)| ≥ |an ||z|n 1 − + ··· + n .
z z
57
58 LECTURE 10. FURTHER APPLICATIONS OF CAUCHY INTEGRAL FORMULA
But
an−1 a0 |an−1 | |a0 | 1
+ ··· + n ≤ + ··· + n < ,
z z |z| |z| 2
if |z| > R when R is as above. So when |z| > R,
|an ||z|n
|p(z)| ≥ .
2
Now suppose, that p(z) has no root in C. Then
1
f (z) =
p(z)
will be an entire function. Moreover, on |z| > R, by the lower bound above,
|p(z)| > |an ||z|n /2 > |an |Rn /2,
that is |f (z)| ≤ M for some M on |z| > R. On the other hand, we claim that there exists an ε > 0 such
that |p(z)| > ε on |z| ≤ R. If not, then there is a sequence of points zk ∈ DR (0) such that |p(zk )| → 0.
Since DR (0) is compact, there exists a subsequence, which we continue to denote by zk , such that zk →
a ∈ DR (0). But then by continuity, p(a) = 0, contradicting our assumption that there is no root. This
proves the claim. The upshot is that on |z| ≤ R, |f (z)| ≤ 1/ε. This shows that f (z) is a bounded, entire
function, and hence by Liouville, must be a constant, which in turn implies that p(z) must be a constant.
This proves the first part of the theorem.
The second part follows from induction. If p(z) is a non-constant polynomial, let α1 be a root, which is
guaranteed to exist by the first part. Then by the remainder theorem, p(z) is divisible by (z − α1 ). For a
proof of the remainder theorem, see remark below. Then we define a new polynomial
p(z)
p1 (z) = .
z − α1
The crucial observation is that deg(p1 ) is strictly smaller than deg (p). In finitely many steps we should
reach a linear polynomial which is obviously factorable.
Remark 10.1.2. We used the remainder theorem for the proof of the second part of the theorem which
basically says that for a polynomial p(z),
p(z) − p(a) = (z − a)q(z)
for some polynomial q(z) of a strictly smaller degree. To see this, let
p(z) = an z n + · · · + a1 z + a0 ,
and so
p(z) − p(a) = an (z n − an ) + · · · + a1 (z − a).
Each term is of the form
ak (z k − ak ) = ak (z − a)(z k−1 + z k−2 a + · · · + zak−2 + ak−1 ),
and this completes the proof. It is also easy to see that highest degree term in q(z) is an z n−1 , and hence
in particular, the degree of q(z) is strictly smaller.
10.2. ZEROES OF A HOLOMORPHIC FUNCTION 59
10.2 Zeroes of a holomorphic function

A complex number a is called a zero of a holomorphic function f : Ω → C if f (a) = 0. A basic fact is that
zeroes of holomorphic functions are isolated. This follows from the following theorem.
Theorem 10.2.1. Let f : Ω → C be a holomorphic function that is not identically zero, and let a ∈ Ω be
a zero of f . Then there exists a disc D around a, a non vanishing holomorphic function g : D → C (that is,
g(z) ̸= 0 for all z ∈ D), and a unique positive integer n such that
f (z) = (z − a)n g(z).
Moreover, we have that

n = min{ν ∈ N | f (ν) (a) ̸= 0}.
The positive integer n is called the order or multiplicity of the zero at a.
Proof. By the principle of analytic continuation, if f is not identically zero, there exists an n such that
f (k) (a) = 0 for k = 0, 1, · · · , n − 1 but f (n) (a) ̸= 0. Let Dε (a) be a disc such that Dε (a) ⊂ Ω. Then f
has a power series expansion in the disc centered at z = a with the first n terms vanishing. So we write
f (z) = an (z − a)n + an+1 (z − a)n+1 + · · ·
with an ̸= 0. The the theorem is proved with
g(z) = an + an+1 (z − a) + · · · .
As an immediate corollary to the theorem we have the following.

Corollary 10.2.2. Let f : Ω → C holomorphic.
1. The set of zeroes of f is isolated. That is, for every zero a, there exists a small disc Dε (a) centered at a
such that f (z) ̸= 0 for all z ∈ Dε (a) \ {a}.
2. (strong principle of analytic continuation) If Ω is connected, and f, g : Ω → C are holomorphic functions
such that for some sequence of points pn ∈ Ω, f (pn ) = g(pn ). Then either f ≡ g or {pn } does not
have a limit point in Ω.
Proof. By the theorem, if a ∈ Ω is a root, then there exists a disc D around a, and integer m, and a
holomorphic function g : D → C such that g(a) ̸= 0 and
f (z) = (z − a)m g(z).
Since g(a) ̸= 0, by continuity, there is a small disc Dε (a) ⊂ D such that g(z) ̸= 0 for any z ∈ Dε (a).
But then on this disc (z − a) is also not zero anywhere except at a, and hence for any z ∈ Dε (a) \ {a},
f (z) ̸= 0 exactly what we wished to prove. Part (2) is a trivial consequence of the Lemma.
Example 10.2.3. Even though the zeroes are isolated, they could converge to the boundary. For instance,
consider the holomorphic function π
f (z) = sin
z
on C∗ . Clearly z = 1/n is a sequence of zeroes. They are isolated, but converge to z = 0 which is not in the
domain.
Let f : Ω → C be a holomorphic function. Then by Corollary 10.2.2 any disc D such that D ⊂ Ω contains
only finitely many zeroes of the function in the interior. The next proposition allows one to calculate the
number of zeroes counted with multiplicity.
Corollary 10.2.4. Let f : Ω → C be holomorphic, and let D ⊂ Ω be a disc such that D ⊂ Ω, and C := ∂D
contains no zeroes of f . Let p1 . · · · , pk are the zeroes of f in D with multiplicity n1 . · · · , nk .
1. There exists a no-where vanishing holomorphic function g : D → C such that
f (z) = (z − p1 )n1 · · · (z − pk )nk g(z).
2. The total number of roots (counted with multiplicity) is given by

ˆ k
f ′ (z) X
dz = nj .
C f (z) j=1
Proof. 1. For each j ∈ {1, · · · , k}, there exist a radius rj > 0 and a nowhere vanishing holomorphic
function gj : Drj (pj ) → C such that for all z ∈ Drj (pj ),
f (z) = (z − pj )nj gj (z).
can choose rjsmall enough so that the discs have mutually disjoint closures. Let U := D \
We
∪kj=1 Drj /2 (pj ) . Define the function
gj (z)
(
Πi̸=j (z−pi )ni , z ∈ Drj (pj ) where j = 1, · · · , k
g(z) := f (z)
Πj (z−pj )nj
, z ∈ U.
Clearly g(z) satisfies all the required properties.

2. With g(z) as above, for any z ̸= pj , a simple computation shows that
k
f ′ (z) X nj g ′ (z)
= + .
f (z) j=1
z − pj g(z)
Since g(z) is nowhere vanishing, g ′ (z)/g(z) is holomorphic on D, and hence by Cauchy’s theorem
it’s integral on C vanishes. It then follows that
ˆ k ˆ
f ′ (z) X dz X
dz = nj = nj .
C f (z) j=1 C z − pj j
10.3 Sequences of holomorphic functions

Recall that if a sequence of differentiable functions fn : I → R converges uniformly to f : I → R, for
some interval I ⊂ R, it does not necessarily imply that f is also differentiable. For instance, consider
fn : [−1, 1] → R defined by r
1
fn (x) = x2 + .
n
Then it is not difficult to see that fn → |x| uniformly, but of course |x| is not differentiable at x = 0. It turns
out that holomorphic function behave much better under uniform convergence. We say that a sequence of
10.3. SEQUENCES OF HOLOMORPHIC FUNCTIONS 61
holomorphic functions fn : Ω → C converges to f : Ω → C compactly if the convergence is uniform over

compact subsets K ⊂ Ω. That is, for all ε > 0 and K ⊂ Ω compact, there exists a N = N (ε, K) such that
|fn (x) − f (x)| < ε,
for all n > N and all x ∈ K.

Theorem 10.3.1. If {fn }∞ n=1 is a sequence of holomorphic functions on Ω that converge compactly to f :
Ω → C, then f (z) is holomorphic. Moreover
fn(k) → f (k)
compactly on Ω for all k ∈ N.
Proof. Fix a a ∈ Ω, and let D be a disc around a such that it’s closure is also in Ω. Then for any triangle
T ⊂ D, by Goursat’s theorem ˆ
fn (z) dz = 0.
T
Since D is compact, fn → f converges uniformly on D, and hence

ˆ ˆ
f (z) dz = lim fn (z) dz = 0.
T n→∞ T
But this is true for all triangles in D, and hence by Morera’s theorem f is holomorphic in D and in particular
at a. Since a is arbitrary, this shows that f is holomorphic on Ω.
We prove that fn′ → f ′ compactly. For k > 1, the result will follow from induction. There is no loss of
generality in assuming that Ω ⊂ C is compact. First we define
Ωr = {z ∈ Ω | Dr (z) ⊂ Ω}.
Geometrically, this is the set of all points in Ω that are at least a distance r away from the boundary of
Ω. Given any compact set K, there exists a r > 0 such that K ⊂ Ωr , and hence it suffices to show that
fn′ → f ′ uniformly on Ωr . The key point is the following estimate.
Claim.Let F : Ω → C be holomorphic. Then for any r > 0,
2
sup |F ′ (ζ)| ≤ sup |F (ζ)|.
z∈Ωr r ζ∈Ωr/2
First notice that if z ∈ Ωr then Dr/2 (z) ⊂ Ωr . To see this, let w ∈ Dr/2 (z) i.e. |z − w| ≤ r/2, and we need
to show that w ∈ Ωr/2 or equivalently that Dr/2 (w) ⊂ Ω. But for any w′ ∈ Dr/2 (w), |w′ − w| ≤ r/2 and
hence by the triangle inequality |w′ − z| ≤ r, that is w′ ∈ Dr (z) which is contained in Ω by definition,
since z ∈ Ωr . This shows that Dr/2 (w) ⊂ Ω and hence that Dr/2 (z) ⊂ Ωr/2 . Now by Cauchy’s integral
formula, if we denote the boundary of Dr (z) by Cr (z), then for any z ∈ Ω2r ,
ˆ
1 F (ζ)
F ′ (z) = dζ.
2πi Cr/2 (z) (ζ − z)2
By the above observation, if z ∈ Ωr then Cr/2 (z) ⊂ Ωr/2 , and hence by triangle inequality, for all z ∈ Ωr ,
since |ζ − z| = r/2 for ζ ∈ Cr/2 (z) we have
1
|F ′ (z)| ≤ sup |F (ζ)|len(Cr/2 (z))
2πr2 ζ∈Cr/2 (z)
2
≤ sup |F (ζ)|
r ζ∈Ωr
This proves the claim. Now given any ε > 0, since Ωr/2 ⊂ Ω is compact (remember we are assuming
without any loss of generality that Ω is bounded), there exists an N = N (r, ε) such that for all n > N and
all ζ ∈ Ωr/2 ,
εr
|fn (ζ) − f (ζ)| < .
2
But then by the claim, for n > N and for all z ∈ Ωr , we have the estimate
2 εr
|f (z) − fn (z)| ≤ · = ε,
r 2
proving that fn → f uniformly on Ωr , and this completes the proof of the theorem.
Recall that Weiestrass’ theorem states that any continuous function on a compact interval is the uniform
limit of polynomials. On the other, by the above theorem, a continuous non-holomorphic function cannot
be the uniform limit of polynomials. Instead we have the following, which we state without a proof.
Theorem 10.3.2 (Runge’s thoerem). Let K ⊂ C and let f be a function that is holomorphic in a neighbour-
hood of K.
u.c
1. There exists a sequence of rational functions Rn (z) such that Rn −−→ f on K, and such that the
singularities of the rational functions all lie in K c .
u.c
2. If K c is connected, then one there exists a sequence of polynomials pn (z) such that pn −−→ f .
Lecture 11
Cauchy’s theorem : Homology

version
11.1 Index of a curve

For a piecewise smooth (not necessarily close) curve, we defined the index or winding number around a
point p ∈
/ γ by ˆ
1 1
n(γ, p) := dz.
2πi γ z − p
We have already seen that if γ is a circle traversed n number of times, then
(
n, p is inside the disc bounded by γ
n(γ, p) =
0, otherwise
We now argue that this number is a measure of the change in the argument along the curve. For simplicity,
lets suppose that p = 0, and that the curve γ joins u = reiθ to v = reiφ (note that γ need not be a circular
arc), where θ, φ ∈ (−π, π). In particular, the curve lies in C\{z < 0}. On this domain, 1/z has a primitive,
which we take to be the principal branch of the logarithm. Then by the fundamental theorem
ˆ
1 1 φ−θ
n(γ, 0) = dz = log v − log u = ,
2πi γ z 2π
and so up to a factor of 2π, the index measures the change in the argument. The following theorem contains
the basic properties of the index.
Theorem 11.1.1. Let γ be any closed curve. Then
1. n(γ, p) is an integer for any p ∈ C \ γ.
2. n(γ, p) is a continuous function on C \ γ, and hence is locally constant.
3. If γ is any curve lying in the interior of a disc D, then n(γ, p) = 0 for all p ∈ C \ D.
Proof. 1. If we could take holomorphic logarithms freely, and argument as above would suffice. Instead
we will give a computational proof. Let γ : [0, 1] → C be parametrization. Then
ˆ 1
1 γ ′ (t)
n(γ, p) = dt.
2πi 0 γ(t) − p
63
64 LECTURE 11. CAUCHY’S THEOREM : HOMOLOGY VERSION
Consider a function ˆ s
1 γ ′ (t)
g(s) = dt.
2πi 0 γ(t) − p
This is a continuous function on [0, 1], and is differentiable wherever γ ′ (t) is continuous (and hence
at all but finitely many points) with derivative
1 γ ′ (t)
g ′ (s) = .
2πi γ(t) − p
Then, letting
G(s) = exp(−2πig(s))(γ(s) − p),
at all but finitely many points, G′ (s) = 0. This shows that G(s) is locally constant, which together
with continuity, forces it to be a constant. In particular, G(1) = G(0), from which it follows (since
g(0) = 1 and γ(0) = γ(1)) that
γ(0) − p
e2πig(1) = = 1.
γ(1) − p
So g(1) = n(γ, p) has to be an integer.
2. Continuity is easy to check since p ∈

/ γ. Since the index is integer valued it has to be then locally
constant.
3. If |p| >> 1, then clearly n(γ, p) can be made really small. But then since the index is locally constant,
and C \ D is connected, it ought to be zero for all p ∈ C \ D.
Remark 11.1.2. There is a deeper reason that the index is always an integer, and a full explanation requires
some knowledge of covering space theory. If a ∈ C does not lie on γ : [0, 1] → C, we can think of γ as a
curve in C∗a := C \ {a}. Then it follows from standard covering space theory that γ has a “lift" to a curve
γ̃ : [0, 1] → C such that e2πiγ̃(t) = γ(t) − a. The relevant jargon is that exp : C → C∗ is a covering space
map. Now additionally if γ is closed, then γ(1) = γ(0), and hence γ̃(1) − γ̃(0) is an integer. On the other
hand, by chain rule, γ ′ (t) = 2πie2πiγ̃(t) γ̃ ′ (t), and hence
1 γ ′ (t)
γ̃ ′ (t) = · .
2πi γ(t) − a
Integrating both sides we see that
ˆ 1 ˆ
1 γ ′ (t) dz
Z ∋ γ̃(1) − γ̃(0) = dt = := n(γ, a),
2πi 0 γ(t) − a γ z−a
and hence n(γ, a) is an integer.
11.2 The homology version of Cauchy’s theorem

A chain is a formal linear combination of curves γ = a1 γ1 + · · · + an γn , where ai ∈ Z and each γi is a
regular curve. We interpret aγ as γ traversed a times if a > 0 and γ traversed in the the reverse direction
−a times if a < 0. Chains can be then added in an obvious way. The union of the set theoretic images of
γj is called the support of γ and denoted by Supp(γ). A chain is called a cycle if each of it’s components
γj is a closed curve. We can extend the definition of the index n(γ, p) for a cycle γ and a point p that does
not lie on any of the components of γ. Then it has the same properties as in Theorem 11.1.1. Additionally,
we have the linearity property that
n(γ1 + γ2 , p) = n(γ1 , p) + n(γ2 , p).
11.2. THE HOMOLOGY VERSION OF CAUCHY’S THEOREM 65
We say that a chain γ is homologous to zero in Ω, and write γ ∼Ω 0 (or γ ∼ 0(modΩ)), if for any point
a ∈ Ωc , n(γ, p) = 0. We also say that γ1 is homologous to γ2 in Ω and write γ1 ∼Ω γ2 if γ1 − γ2 ∼Ω 0, or
equivalently if n(γ1 , p) = n(γ2 , p) for all p ∈ Ωc . Note that if Ω ⊂ Ω′ then γ ∼Ω 0 implies that γ ∼Ω′ 0,
but the converse need not be true as can be seen in the example below.
Example 11.2.1. Consider the disc D3 (0) and the annulus A1,2 (0) := {z ∈ C | 1 < |z| < 3}. By Cauchy’s
theorem for discs, any curve γ in D3 (0) is homologous to zero. On the other hand the curve γ(t) = 2e2πit , t ∈
[0, 1] is NOT homologous to zero in A1,2 (0). This is because n(γ, 0) = 1 ̸= 0.
Now we are ready to state the most general form of Cauchy’s theorem.
Theorem 11.2.2 (Generalised Cauchy’s theorem). If f : Ω → C is holomorphic and γ ∼Ω 0, then
ˆ
f (z) dz = 0.
γ
More generally, if γ1 , γ2 are curves in Ω such that γ1 ∼Ω γ2 , then

ˆ ˆ
γ1 γ2
In other words, Cauchy’s theorem says that if the integrals of the holomorphic functions 1/(z − a) is zero
on a closed curve in Ω, then the integral of any holomorphic function on γ is zero.
Proof. This beautiful proof is taken from [1]. We first assume that Ω is bounded. For a small δ > 0, we
cover the plane with a net consisting of squares with sides parallel to the axes, and length δ. Since Ω is
bounded, and if δ > 0 is chosen sufficiently small, there exists a finite set of number of cubes Q1 , · · · , QN
such that
1. The collection of squares {Q1 , · · · , Qj } are all the squares in the net which lie completely inside Ω.
That is, if Q is a square in the net, then Q = Qj for some j if and only if Q ⊂ Ω.
◦
2. γ ⊂ Ωδ := ∪N j=1 Qj .
Consider the cycle

X
Γδ = ∂Qj ,
j
where the boundary of each Qj is oriented in the anti-clockwise direction. Then Γδ is equivalent to ∂Ωδ
in the sense that for any function ˆ ˆ
f (z) dz = f (z) dz,
Γδ ∂Ωδ
since the integrals over the common boundaries cancel.

Now, let γ be a cycle homologous to zero in Ω. Let ζ ∈ Ω \ Ωδ , and let Q be a square in the net such that
ζ ∈ Q. By definition of Ωδ , Q ̸= Qj for any j. Again, by our choice of the squares that make up Ωδ , there
exists a point ζ0 ∈ Q ∩ Ωc . Since ζ0 ∈
/ Ω and γ ∼Ω 0, we have that n(γ, ζ0 ) = 0. But then by continuity,
since ζ0 and ζ can be joined by a straight in Q and hence not intersecting γ, we have n(γ, ζ) = 0. In
particular, n(γ, ζ) = 0 for all ζ ∈ ∂Ωδ .
Suppose now that f (z) is holomorphic on Ω. For any z ∈ Q˚j0 , we have
ˆ (
1 f (ζ) f (z), j = j0 ,
dζ =
2πi ∂Qj ζ − z 0, otherwise,
and hence ˆ
1 f (ζ)
f (z) = dζ. (11.1)
2πi Γδ ζ −z
Since j0 was arbitrary, (11.1) holds for all z ∈ ∪j Q◦j . But since both sides are continuous in z, clearly (11.1)
must hold on all of Ωδ . As a consequence,
ˆ ˆ ˆ
1 f (ζ)
f (z) dz = dζ dz.
γ γ 2πi Γδ ζ − z
By Fubini’s theorem (which can be applied since the integrand is continuous in both z and ζ),
ˆ ˆ 1 ˆ dz ˆ
f (z) dz = f (ζ) dζ = f (ζ)n(γ, ζ) dζ = 0.
γ Γδ 2πi γ ζ − z Γδ
Finally, if Ω is not bounded, consider a large disc DR (0) which contains supp(γ) in the interior, and let
Ω′ = Ω ∩ DR (0). Then since γ ∼DR (0) 0, one can easily see that γ ∼Ω′ 0, and the previous argument
then applies to Ω′ completing the proof.
Using the generalised Cauchy theorem, we can prove the following generalisation of the CIF.
Theorem 11.2.3 (Generalised Cauchy integral formula (GCIF)). Let f ∈ O(Ω), and γ ⊂ Ω a cycle. If
γ ∼Ω 0, then for any z ∈ Ω \ Supp(γ),
ˆ
1 f (ζ)
n(γ, z)f (z) = dζ.
2πi γ ζ − z
Proof. Fix a z ∈ Ω \ Supp(γ), and let ε0 > 0 such that Dε0 (z) ∩ Supp(γ) = ϕ. For ε ∈ (0, ε0 ), let Cε be
the circle centred at z with radius ε.
Claim. For every ε ∈ (0, ε0 ), γ ∼Ω∗z n(γ, z)Cε , where as usual Ω∗z = Ω \ {z}.
11.3. SIMPLY CONNECTED DOMAINS 67
Proof of the Claim. We need to prove that for any ζ ∈ C \ Ω∗z ,

n(γ, ζ) = n(γ, z)n(Cε , ζ). (11.2)
First suppose ζ ∈ C \ Ω. Then in particular, ζ lies outside Dε (z) and hence n(Cε , ζ) = 0. On the other
hand since γ ∼Ω 0, we also have n(γ, ζ) = 0, and hence (11.2) is verified. The only other possibility is that
ζ = z. But then n(Cε , ζ) = 1, and hence again (11.2) is verified.
Now, applying Cauchy’s theorem to the holomorphic function f (ζ)/(ζ − z) on Ω∗z , we see that for all
ε ∈ (0, ε0 ), ˆ ˆ
1 f (ζ) n(γ, z) f (ζ)
dζ = dζ. (11.3)
2πi γ ζ − z 2πi Cε ζ − z
The integral on the right is n(γ, z)f (z) by the CIF for discs, and we are done. But it is in fact possible to
avoid the CIF altogether, and in the process provide a second proof for the CIF on discs. The argument is
as follows. Given any η > 0, by choosing ε << 1, we can make sure that
|f (ζ) − f (z)| < η
for all ζ ∈ Cε . Then
ˆ ˆ
1 f (ζ) 1 f (ζ) − f (z)
dζ − f (z) = dζ < η.
2πi Cε ζ −z 2πi Cε ζ −z
Hence from (11.3),
ˆ ˆ
1 f (ζ) 1 f (ζ)
dζ = n(γ, z) lim dζ = n(γ, z)f (z).
2πi γ ζ −z ε→0 2πi Cε ζ −z
11.3 Simply connected domains

There are many equivalent ways of defining simply connected domains. Following Ahlfors, we take a
slightly non-standard route. Consider the sphere
S2 = {(x, y, z) ∈ R3 | x2 + y 2 + z 2 = 1}.
We denote it’s north pole by N = (0, 0, 1), and south pole by S = (0, 0, −1). Then consider the stereo-
graphic projection ΦN : S2 \ {N } → C defined by
x + iy
Φ(x, y, z) = .
1−z
Then ΦN is a bijection, and is in fact a homeomorphism. We can thus identify S2 as a one-point compact-
ification of C, and call it the extended complex plane, and think of N as the “point at infinity".
We then say that a domain Ω ⊂ C is simply connected if it is connected, and S2 \ Ω is also connected. We
note that this is not a commonly used definition since it does not work in higher dimensions. In the next
lecture, we will provide equivalent characterisations, one of which will be what is the modern “textbook"
definition.
Example 11.3.1. A disc, C itself, and half planes are simply connected. A parallel strip, say {z | a ≤ Im(ζ) ≤
b} is also simply connected. This shows the importance of taking the complement of Ω in the extended plane,
rather than C itself. Similarly C \ {ζ ≤ 0} is also simply connected. In this case the complement of the set in
the extended plane is the complement of half a great circle in S2 . On the other hand the complement of a line
L passing through the origin in C in the extended plane is an entire great circle, and hence C \ L is not simply
connected. Similarly, C∗ is not simply connected, since S2 \ C∗ = {N, S}
Lecture 12
Cauchy’s theorem : Multiply

connected domains
12.1 Characterisations of simply connected sets

Recall that a connected subset Ω ⊂ C is called simply connected if the complement in the extended complex
plane is also connected. We now provide several equivalent characterisations of being simply connected.
First we need to introduce two important notions.
Recall that a curve γ : [a, b] → Ω is a simple, closed curve if γ is injective on (a, b) and γ(a) = γ(b). Such
curves are called Jordan curves, and their name stems from the following historically significant theorem.
Theorem 12.1.1 (Jordan curve theorem). Let γ be a Jordan curve and C be it’s image. Then it’s complement
C \ C consists of exactly two open connected subsets. One of these components is bounded while the other is
unbounded.
The bounded component is called the interior and the unbounded component is called the exterior, denoted
respectively by int(γ) and ext(γ). While intuitively obvious, the proof is extremely non-trivial. So much
so that the the theorem is notorious for numerous incorrect proofs from well known mathematicians. In
fact it will not be an exaggeration to say that attempts to prove this theorem led to the modern development
of algebraic topology.
We also need to introduce the notion of homotopy. We say that a closed piecewise regular curve γ :
[0, 1] → Ω is contractible, or null homotopic in Ω if there exists a point p ∈ Ω and a continuous function
H : [0, 1] × [0, 1] → Ω such that
(
H(0, t) = γ(t), H(1, t) = p, ∀t ∈ [0, 1]
H(s, 0) = H(s, 1), ∀s ∈ [0, 1].
We use the notation γs (t) := H(s, t).

Example 12.1.2. 1. Let D be a disc centred at p. Then for every curve γ, consider the homotopy
H(s, t) = (1 − s)(γ(t) − p) + p.
Then H(1, t) = p and hence γ is null homotopic. More generally any convex domain has the property
that every closed curve is null homotopic. To see this, let γ be an aribitrary closed curve. Then
2. On the other hand, consider the curve γ(t) = eit in the annulus A0,2 (0) := {z | 0 < |z| < 2}. Then γ
is not null-homotopic, as can for instance be seen using the theorem below.
69
70 LECTURE 12. CAUCHY’S THEOREM : MULTIPLY CONNECTED DOMAINS
We then have the following fundamental result.

Theorem 12.1.3. Let Ω ⊂ C be a connected set. Then the following are equivalent.
1. Ω is simply connected.
2. For every Jordan curve γ in Ω, int(γ) ⊂ Ω.
3. For all cycles γ in Ω, γ ∼Ω 0.
4. (Cauchy theorem for simply connected domains) For all holomorphic functions f ∈ O and all cycles γ
in Ω,
ˆ
f (z) dz = 0.
γ
5. Every closed piecewise regular curve is null homotopic.

We need to use the following crucial lemma which we state without proof.
Lemma 12.1.4. Let γ be a piecewise regular curve which is null homotopic. Then one can choose the homotopy
H such that each γs is piecewise regular.
Proof. The implication (2) ⇐⇒ (3) is a consequence of the generalised Cauchy theorem. W
• (1) =⇒ (2). We can assume that Ω ̸= C for else the implication is trivial. Suppose a ∈ int(γ) ∩ Ωc .
Then since Ω is simply connected, S2 \ Ω is connected, and hence there exists a p ∈ Ωc ∩ ext(γ),
and a path σ lying in Ωc and connecting a to p. But since int(γ) is connected this is a contradiction.
• (1) =⇒ (3). Let γ be a cycle in Ω and p ∈ / Ω. Since S2 \ Ω is connected, there is a sequence of
points pn such that |pn | → ∞ and there is a path σn from p to pn . Since limn→∞ n(γ, pn ) = 0 and
index is locally constant, this implies that n(γ, p) = 0.
• (3) =⇒ (1). Suppose Ω is not simply connected. Then S2 \ Ω = A ∪ B, where B is the component
at infinity, and A is a compact (possibly disconnected) set. Let
δ := inf{|z − w| | z, ∈ A, w ∈ B}.
Then δ > 0. Now we cover the entire √ plane with a net N of squares of a fixed side length δ/4 (any
side length strictly smaller than δ/ 2 will do). We choose the net so that a certain square, say Q1
has the point a ∈ A at it’s centre. Let Q1 , · · · , QN be the squares whose interiors have a non-empty
intersection with A, and let

Γ = ∂ ∪N j=1 Q j
oriented in an anticlockwise direction. Note that

(
1, j = 1
n(∂Qj , a) =
0, j > 1,
and hence n(∂Γ, a) = 1, since the integrals over the common boundaries vanish. But then since Γ
clearly does not meet B, we have found a cycle in Ω such that n(Γ, a) ̸= 0 but a ∈ Ωc . This is a
contradiction.
• (2) =⇒ (1). Suppose Ω is not simply connected, then the Γ produced above gives a Jordan curve
whose interior is not completely contained inside Ω.
12.2. CAUCHY’S THEOREM FOR MULTIPLY CONNECTED DOMAIN 71
• (4) =⇒ (2). It is enough to prove that every closed smooth curve that index zero. Let γ be one
such curve, and let p ∈ Ωc . There exists a homotopy H : [0, 1] × [0, 1] → Ω contracting γ to a point
a ∈ Ω. For s ∈ [0, 1], consider the function
ˆ 1
1 γs′ (t)
f (s) := dt.
2πi 0 γs (t) − p
Note that the definition makes sense because of the Lemma above. Now, since p ∈ Ωc , clearly f (s)
is a continuous function. Moreover, f (1) = 0, and hence f (0) = n(γ, p) = 0.
• (2) =⇒ (4). Since this implication will not play any further role in the course, we simply direct
the reader to the argument on page 252 of Complex Analysis by Theodore Gamelin.
An important consequence of this is the following.

Theorem 12.1.5. Let Ω be a simply connected domain and f ∈ O. Then f has a primitive on Ω.
Proof. The proof is along the lines of the proof for existence of primitives on disc that was used in the proof
of Cauchy’s theorem. So we fix a p ∈ Ω, and define
ˆ
F (z) = f (w) dw,
γz
where the integral is along some path γz joining p to z. If we choose another path γ̃z joining the two
points, then γz − γ̃z will form a cycle. Since the domain is simply connected, γz ∼Ω γ̃z , and hence the
integral of f along both would be the same. Hence our definition is actually independent of the path. By
openness of Ω, for any h small, the straight line joining z to z + h will lie entirely in Ω, and we call this
path as l. Then γz+h − l and γz are both piecewise smooth paths joining p to z, so once again by simple
connectedness of Ω and Theorem 12.1.3
ˆ ˆ ˆ
f (w) dw − f (w) dw = f (w) dw,
γz+h l γz
or equivalently ˆ
F (z + h) − F (z) = f (w) dw.
l
Then the same argument as in the proof of Theorem 2.1 in Lecture-7 implies that F (z) is holomorphic
with F ′ (z) = f (z).
12.2 Cauchy’s theorem for multiply connected domain

A connected domain Ω ⊂ C is said to be n-connected if it’s complement in the extended complex plane has
n-connected components. So for instance, a simply connected set is 1-connected, and an n-connected set
has n − 1 number of “holes". Our convention will be to label the components as A1 , · · · , An , where An
is component containing the north pole (or the point at “infinity"). Using the argument in the proof of the
implication (2) =⇒ (1) in Theorem 12.1.3 we obtain the following.
Theorem 12.2.1. For every i = 1, · · · , n − 1, there exists a cycle γi such that
(
1, p ∈ Ai
n(γi , p) = (12.1)
0, p ∈ Ωc \ Ai .
Moreover we have the following observations.
Pn
1. The set {γ1 , · · · , γn } is a linearly independent set, in the sense that i=1 ci γi ∼Ω 0 if and only if
ci = 0 for all i.
2. The set {γ1 , · · · , γn } is a spanning set, in the sense that if γ is any other cycle in Ω, then
γ ∼Ω c1 γ1 + · · · cn−1 γn−1 ,
where ci = n(γ, pi ) for any pi ∈ Ai .

3. For any f ∈ O(Ω), we have
ˆ n−1
X ˆ
f (z) dz = n(γ, pi ) f (z) dz,
γ i=1 γi
where (p1 , p2 , · · · , pn−1 ) is any collection of points in A1 × · · · An−1 .

4. If {σ1 , · · · , σm } is a linearly independent spanning set, then m = n.
We call the collection {γ1 . · · · , γn } the homology basis for Ω. In general there will be multiple choices of
homology bases, but by an elementary theorem in linear algebra, all of these must be n in number. In fact,
if
Example 12.2.2. Consider the domain Ω := D2 (0) \ {−1, i, 1}. Then clearly the domain is 4-connected. In
fact if we label the points by p1 = −1, p2 = i, p3 = 1, then Ai = {pi } for i = 1, 2, 3 and A4 = S2 \ D2 (0).
Let γi be given by
1
γi (t) = pi + eit .
2
Clearly n(γi , pi ) = 1. On the other hand if p ∈ Ωc \{pi }, then p lies outside D1/2 (pi ) and hence n(γi , p) = 0.
Hence by the above theorem, {γ1 , γ2 , γ3 } forms a homology basis for Ω.
As a consequence of the above result, we have the following. We say that a Jordan curve γ is positively
oriented if while the curve is traversed, int(γ) remains to the left.
Corollary 12.2.3. Let Γ, Γ1 , · · · , Γn be positively oriented, pairwise non intersecting,
piecewise
smooth Jor-
n
dan curves such that Γj ⊂ int(Γ) for all j = 1, · · · , n. Let Ω = int(Γ) ∩ ∩j=1 ext(Γj ) . Let f be a
function that is holomorphic in a neighbourhood of Ω. Then
ˆ Xˆ
Γ j Γj
Proof. Firstly, one can construct Jordan curves C, C1 , · · · , Cn such that

int(Γ) ⊂ int(C)
and int(C) ⊂
′ n
int(Γ) for all i, and such that f is holomorphic in Ω := int(C) ∩ ∩j=1 ext(Cj ) which of course
contains Ω. Let pi ∈ int(Ci ). It is then easy to check that Γ1 , · · · , Γn forms a homology basis for Ω′ and
the result then follows from Theorem 12.2.1. Note that constructing the Jordan curves Ci is non-trivial. If
Γi is smooth, then one can construct Ci by perturbing slightly in the direction of the inner normal (and
similarly C be perturbing Γ a little bit along the outer normal). But since our curves are only piecewise
smooth, extra care must be take to “round off"the “corners".
12.3 A real variable integral

We will now apply Cauchy’s theorem to compute a real variable integral. Later in the course, once we
prove a further generalization of Cauchy’s theorem, namely the residue theorem, we will conduct a more
12.3. A REAL VARIABLE INTEGRAL 73
systematic study of the applications of complex integration to real variable integration. For now, let us
compute ˆ ∞
cos x
2
dx.
−∞ 1 + x
This is an improper integral which is convergent, so by definition

ˆ ∞ ˆ R
cos x 1 − cos x
dx = lim dx,
−∞ x2 R→∞ −R x2
Consider now the function

eiz
f (z) = .
1 + z2
This is a holomorphic function on C \ {i}. Moreover, on the real line the real part of this function is
precisely the function that we are looking to integrate. Now consider a contour ΓR := {z ∈ C | |z| =
R, Im(z) > 0} ∪ {(x, 0) | x ∈ (−R, R)} oriented in the anti-clockwise direction. Let Cε ; = {|z − i| = ε}
be a small circle around i. Then for R >> ε, we clearly have that ΓR ∼C∗i Cε , and hence by Theorem
12.2.1, ˆ ˆ
ΓR Cε
Now note that

g(z)
f (z) = ,
z−i
where g(z) = eiz /(z + i) is holomorphic in a neighbourhood of the disc Dε (i), and hence by the Cauchy
integral formula (applied to g(z)),
ˆ ˆ
g(z) π
f (z) dz = dz = 2πig(i) = .
Cε Cε z − i e
Letting SR be the semi-circle {z ∈ C | |z| = R, Im(z) > 0}, we see that

ˆ ˆ ˆ R ˆ ˆ R
eix cos x
f (z) dz = f (z) dz + = f (z) dz + ,
ΓR γR −R 1 + x2 γR −R 1 + x2
where we have used the fact that sin x is an odd function. On γR we claim that
1 − eiz 2
≤ 2.
z2 R
To see this, for z ∈ γR , we can write z = x + iy with y > 0. So |eiz | = e−y < 1, and hence by triangle
inequality |1 − eiz | < 2 which proves the claim since |z| = R on γR . Using this we can estimate that
ˆ
1 − eiz 2 2π
2
dz ≤ len(γR ) = →0
γR z R R
as R → ∞. So the contribution on γR goes to zero as R goes to infinity, and hence

ˆ ∞
cos x π
2
= .
−∞ 1 + x e
Lecture 13
Logarithm and roots revisited
13.1 Criteria for existence of holomorphic logarithms

The purpose of this lecture is twofold - first, to characterize domains on which a holomorphic logarithm can
be defined, and second, to show that the only obstruction to defining a holomorphic logarithm is in defining
a continuous logarithm. From henceforth, we let O∗ (Ω) be the set of nowhere vanishing holomorphic
functions on Ω. For instance, ez ∈ O∗ (C).
To set the stage, let us revisit the difficulties we had in defining a holomorphic logarithm. For z = reiθ ,
consider the function
log z = log |z| + iθ, (13.1)
where, for instance, we can let θ ∈ [θ0 , θ0 + 2π). If arg p = θ0 , and we traverse a circle of radius |p| centred
at 0 and return to the point p, the argument goes from θ0 to θ0 + 2π, and hence the log z does not return
to the original value. In other words log z as defined is not continuous. On the other hand, if we return
to p along a small circle not containing 0 in the interior, then the argument does return to θ0 , and log z
does not jump in value. The difference between the situations if of course that the first curve goes around
0 while the second does not. Thus logarithm is an example of a multivalued function, and zero in this case
is called a branch point.
In general, we can consider any holomorphic function f : Ω → C∗ . Then, a holomorphic function g :
Ω → C (if it exists) is called a branch of the logarithm of f , and denoted by log f (z), if
eg(z) = f (z)
for all z ∈ Ω. A natural question to ask is the following.
Question 13.1.1. Given a holomorphic function f : Ω → C∗ , when can we define a holomorphic branch of
log f (z).
From the point of view of the Cauchy theory, the multivalued behaviour of the logarithm function is esen-
tially because 1/z, which would be the derivative of a holomorphic logarithm function, does not integrate
out to zero around curves that contain the origin in the interior. Keeping this in mind, we have the follow-
ing basic theorem.
Theorem 13.1.1 (Fundamental theorem on existence of holomorphic logarithms). Let Ω be a connected
domain, and f ∈ O∗ (Ω) such that ˆ ′
f (z)
dz = 0
γ f (z)
for all closed loops γ ⊂ Ω.
75
76 LECTURE 13. LOGARITHM AND ROOTS REVISITED
• Then there exists a holomorphic function g : Ω → C, denoted by g(z) = log f (z), such that
eg(z) = f (z).
• g ′ = f ′ /f , and hence for any fixed p ∈ Ω,

ˆ z
f ′ (w)
g(z) = g(p) + dw.
p f (w)
Moreover, if g̃ is another function satisfying eg̃(z) = f (z), then there exists an n ∈ Z such that
g̃(z) − g(z) = 2πn, ∀z ∈ Ω.
Remark 13.1.2. Note that different choices of g(p) corresponding to the countable number of solutions to
ez = f (p) give different formulae for g(z), all of which differ by integral multiples of 2πi. Conversely, if g1
and g2 are two logarithms, then they have to differ by a multiple of 2πi. The various logarithm functions
are called branches.
Theorem 13.1.1 above combined with Cauchy’s theorem for simply connected domains gives the following.
Corollary 13.1.3. Let Ω be a simply connected domain.
1. Then for any f ∈ O∗ (Ω), there exists a holomorphic log f (z) with (log f )′ = f ′ /f , and hence
ˆ z ′
f (w)
log f (z) = log f (p) + dw,
p f (w)
w here log f (p) is any solution to ez = f (p).

/ Ω, then there is a holomorphic branch of log z on Ω with (log z)′ = 1/z. Moreover,
2. In particular, if 0 ∈
for any p ∈ Ω, ˆ z
1
log z = log p + dz,
p z
where we integrate along any path from p to z, and log p is any solution to ez = p.
Proof of Theorem 13.1.1. Fix a point p ∈ Ω, and let g(p) be a solution to eg(p) = f (p). Since f (p) ̸= 0, such
a solution always exists. We then define g(z) by
ˆ z ′
f (w)
g(z) = g(p) + dw,
p f (w)
where we integrate over any curve joining p and z. By the hypothesis, this is independent of the path
chosen. Then, by the same argument used before (as in the proof of Theorem ), g(z) is holomorphic with
f ′ (z)
g ′ (z) = .
f (z)
Now, consider the function F (z) = e−g(z) f (z). Then
F ′ (z) = −eg(z) g ′ (z)f (z) + e−g(z) f ′ (z) = 0.
Since Ω is connected, this implies that F (z) is a constant, and hence F (z) = F (p) = 1. This completes
the proof.
13.2. SOME EXAMPLES 77
Recall that the branch of log z defined by (13.1) is not even a continuous function over C∗ . This is not a
coincidence. Our next theorem, says that continuity in fact, is the only obstruction to define a holomorphic
logarithm.
Theorem 13.1.2. Let Ω ⊂ C and g : Ω → C be continuous. If eg(z) is holomorphic, then so is g(z).
In other words if f (z) is holomorphic, and we can define a continuous log f (z), then log f (z) is automat-
ically holomorphic.
Proof. Let f (z) = eg(z) , which by the hypothesis, is holomorphic, and fix a p ∈ Ω. There is a δ > 0
such that Dδ (p) ⊂ Ω and a holomorphic function gp (z) on Dδ (p) such that egp (z) = f (z). Then for all
z ∈ Dδ (p),
g(z) − gp (z)
∈ Z.
2πi
g(z)−g (z)
p
Now, since g(z) is continuous, 2πi is a continuous function on Dδ (p) which only takes integer
values, and hence has to be a constant. That is,
g(z) = gp (z) + 2πin,
for some fixed n ∈ Z. But then g(z) has to be holomorphic in Dδ (p), since gp (z) is holomorphic, and
hence is in complex differentiable at p. Since p was arbitrary, this completes the proof of the theorem.
13.2 Some examples

Somewhat vaguely, for a (multi-valued) function g(z), the point z = a is defined to be a branch point
if g(z) is discontinuous while traversing an arbitrarily small circle around the point. We define infinity
to be a branch point if z = 0 is the branch point of g(1/z). An alternate way is to consider a curve
“enclosing" infinity. This is a large curve that contains all the other branch points in it’s interior. Then
infinity is a branch point if along this large curve, the function g(z) is discontinuous. A branch cut is a
union of curves such that g(z) defines a single valued holomorphic function on the complement. Branch
cuts should usually connect branch points to prevent the possibility of going around branch points and
making the function value jump.
13.2.1 Principal branch of the logarithm

Since C\{Re(z) ≤ 0} is simply connected, this immediately implies that there is a holomorphic logarithm
on that domain. The principal branch of the logarithm is defined to be the one with
log 1 = 0.
By Theorem 13.1.1, for any z ∈ C \ {Re(z) ≤ 0}, we then have

ˆ z
1
log z = dw,
1 w
where we integrate over any piecewise smooth path from 1 to z. Suppose z = reiθ , then one such path is
C = C1 + C2 where C1 is parametrized by z1 (t) : [0, 1] → C with z1 (t) = tr + (1 − t), and C2 is given
it
by z2 (t) : [0, θ] → C where z2 (t) = ree So C is simply the path going first from 1 to r along the x-axis,
and then the circular arc to z. Then
ˆ ˆ 1
1 r−1 t=1
dw = dt = log(t(r − 1) + 1) = log r,
C1 w 0 t(r − 1) + 1 t=0
where the log is the usual logarithm defined on real numbers. On the other hand,
ˆ ˆ θ
1
dw = i dt = iθ.
C2 w 0
So the principal branch of the logarithm is given by
log z = log r + iθ,
where θ ∈ (−π, π). We end with the following remark.

Remark 13.2.1. Unlike the real logarithm, in the complex case, in general
log z1 z2 ̸= log z1 + log z2 .
For example, let z1 = e3πi/4 ,z2 = eπi/2 and log z be the principal branch. Then log z1 = 3πi/4 and
log z2 = πi/2. But z1 z2 = e5πi/4 = e−3πi/4 (remember the range of arg is (−π, π], and so log z1 z2 =
−3πi/4 ̸= log z1 + log z2 . Similarly, even though elog z = z for all z, log ez ̸= z generally, again due to
the periods of ez .
13.2.2 Branch cut for log (z 2 − 1).

The points where z 2 − 1 = 0, namely z = ±1 are certainly branch points. Any branch cuts should include
these two points. To see that infinity is also a branch point, note that the logarithm should be defined as a
primitive of
2z 1 1
= + .
z2 − 1 z+1 z−1
It is clear that as we integrate along a curve of radius R > 1 around the origin, both terms will make a
contribution with the same sign, and hence the integral will not be zero. In other words if we define log
as a primitive of z22z−1 , it will have a jump if we traverse a large curve. Hence ∞ is a also branch point.
Possible branch cuts, that will prevent going around z = ±1 or z = ∞ are
(−∞, 1] or (−∞, 1] ∪ [1, ∞) or [−1, ∞).
Of course there are infintely many choices of branch cuts. Each of the above branch cuts renders the
domain simply connected, and hence a holomorphic branch does exist.
A convenient way to write down a formula for the branch is by using “double polar coordinates". That
is, we let
z = −1 + r1 eiθ1 = 1 + r2 eiθ2 .
If we restrict the “phases" θ1 , θ2 to be in the usual range (−π, π), then using the principal branch of the
log we obtain
log (z + 1) = log r1 + iθ1

log (z − 1) = log r2 + iθ2 .
Adding up these two we see that
g(z) = log r1 + log r2 + i(θ1 + θ2 ) (13.2)
does define a branch of log (z 2 − 1) since it is easy to see that eg(z) = z 2 − 1.
Claim. g(z) defines a holomorphic branch of log (z 2 − 1) on C \ (−∞, 1].

13.3. N T H -ROOTS OF HOLOMORPHIC FUNCTIONS 79
By Theorem 13.1.2 it is enough to check that it defines a continuous branch. But this is obvious since θ1
is continuous everywhere except z ≤ −1 and θ2 is continuous everywhere except z ≤ 1, and since these
points are removed in the branch cut, g(z) is continuous everywhere else.
If we instead, restrict θ1 ∈ (−π, π] and θ2 ∈ (0, 2π), then formula (13.2) defines a holomorphic branch on
the complement of the branch cut (−∞, 1] ∪ [1, ∞). The reader should work these out carefully.

z+1
13.2.3 Branch cuts for log z−1
.
A holomorphic definition would have primitive
d z + 1 1 1
log = − .
dz z−1 z+1 z−1
Clearly z = ±1 are branch points. To analyze branching at infinity, consider a large disc DR (0) with
R > 2. Then both the terms contribute an integral of 2πi but with opposite signs, and hence the integral
vanishes. In other words the argument does not change as we traverse this big circle. Hence infinity is
NOT a branch point. Hence we can then choose the branch cut to be [−1, 1], even though C \ [−1, 1] is
not simply connected.
Again, lets analyze this using the double polar coordinates z = −1 + r1 eiθ1 = 1 + r2 eiθ2 , This time let
θ1 , θ2 ∈ [−π, π), and define
g(z) = log r1 − log r2 + i(θ1 − θ2 ).

z+1
Clearly this defines a branch of log z−1 , and we need to check if it is a continuous branch. The function
so defined is surely continuous (and hence holomorphic by Theorem 13.1.2) everywhere on C \ [−1, 1]
except possibly the real axis to the left of z = −1. As you approach this part of the real axis from the
top, both θ1 , θ2 → π. On the other hand when you approach from the bottom, both θ1 , θ2 → −π, and
so their
difference
cancels out. So the resulting function defines a continuous, and hence a holomorphic,
z+1
log z−1 . Again if we change the domain for either θ1 or θ2 to (0, 2π), we are forced ton consider other
branch cuts. Once again, the reader should work both these cases out carefully.
13.3 nth -roots of holomorphic functions

Given a logarithm function, and a w ∈ C, one can define a holomorphic complex power, by
z w = ew log z . (13.3)
When w = 1/n is the reciprocal of a natural number, we call z 1/n the nth root of z.
Example 13.3.1. Roots of unity. Consider the polynomial z n − 1. Clearly ζn = e2πi/n , is a root. Moreover,
ζnk , for k = 0, 1, · · · , n − 1 is also a root, and since the degree of the polynomial is n, these are all the possible
roots. We call ζn the primitive nth root of unity.
We then have the following analogs of Corollary 13.1.3 and Theorem 13.1.2.
Theorem 13.3.1. 1. If Ω is simply connected and f (z) is holomorphic and zero free, then there exists a
holomorphic function g(z) such that g(z)n = f (z). Moreover, if g1 (z) is any other such function, then
g1 (z) = ζnk g(z) where ζn = e2πi/n is the primitive nth root of unity and k = 0, 1, · · · , n − 1.
2. If g : Ω → C is continuous such that g(z)n is holomorphic for some positive integer n, then g(z) itself
is holomorphic.
Proof. 1. For the first part, by Corollary 13.1.3, there exists a holomorphic log f (z). We then simply
take log f (z)
g(z) = e n .
It is also clear that if g1 (z) is another such function, then (g1 (z)/g(z))n = 1, and hence there exists
2. We proceed as in the proof of Theorem 13.1.2. Let f (z) = g(z)n . Then for any p ∈ Ω if r > 0
such that Dr (p) ⊂ Ω, by the first part, there exists a holomorphic function gp (z) on Dr (p) such
that gp (z)n = f (z). But then on the disc, (g/gp )n = 1 and hence by continuity, there exists a fixed
integer 0 ≤ k ≤ n − 1 (independent of z) such that g(z) = gp (z)e2πik/n , which in turn implies that
g(z) is holomorphic.
13.3.1 Principal, and other branches of the square root

√
We can define the principal branch of the square root so that 1 = 1. Doing a similar computation as
above, we can then see that if z = reiθ with θ ∈ (−π, π), then
√
z = reiθ/2 .
√
On the other hand, if we want 1 = −1, then
√
z = reiπ+iθ/2 .
√
13.3.2 Branch cuts for z2 − 1
√
Any of the branch cuts for log (z 2 − 1) will allow us to define z 2 − 1 on their complement by equation
(13.3).
√ Each of those branch cuts extend out to infinity. But it turns out we can define a holomorphic branch
of z 2 − 1 on the complement of finite cut. This is possible because ∞ is not a branch point (even though
it is a branch point of log (z 2 − 1)).
To see this, we again make use of double polar coordinates. Let z = −1 + r1 eiθ1 = 1 + r2 eiθ2 as before,
where we let θj ∈ (−π, π), and we define

θ1 +θ2
√ i 2
g(z) = r1 r2 e .
√
Clearly this defines a branch of z 2 − 1 since
g(z)2 = r1 r2 ei(θ1 +θ2 ) = (z + 1)(z − 1) = z 2 − 1.
All we need to now do, is to find a branch cut such that g(z) is continuous in the complement.
Clearly θ1 is continuous everywhere except (−∞, −1] and θ2 is continuous everywhere except (−∞, 1].
So g(z) is continuous everywhere except possibly for (−∞, 1]. Let us analyze the two intervals (−∞, −1)
and [−1, 1]. If z approaches [−1, 1] from above, θ1 → 0 but θ2 → π, and so
√ √
g(z) → r1 r2 eiπ/2 = i r1 r2 .
But if it approaches from below, then θ1 → 0 while θ2 → −π. So

√ √
g(z) → r1 r2 e−πi/2 = −i r1 r2 ,
and so g(z) is discontinuous on [−1, 1]. Next we analyse continuity along (−∞, −1). As z approaches
(−∞, −1) from above, θ1 → π and θ2 → π, hence
√ √
g(z) → e2iπ/2 r1 r2 = − r1 r2 .
13.3. N T H -ROOTS OF HOLOMORPHIC FUNCTIONS 81
On the other hand when z approaches (−∞, −1) from below θ1 , θ2 → −π, and so
√ √
g(z) → r1 r2 e−iπ = − r1 r2 ,
and hence g(z) defines a continuous√branch on (−∞, −1). The upshot is that g(z) defines a continuous
(and hence holomorphic) branch of z 2 − 1 on C \ [−1, 1].
√
2
√ we can also compute the value of z − 1. For instance we demonstrate how
Note that with above formula, √
2
to calculate the value of i − 1 for our particular branch. Of course the answer has to be either ±i 2,
but the question is which one of these values? We can write (draw a picture to see what is happening
geometrically) √ √
i = −1 + 2eiπ/4 = 1 + 2e3πi/4 ,
√
so that r1 = r2 = 2 and θ1 = π/4 and θ2 = 3π/4. Since θ1 + θ2 = π by the formula above
√ √ iπ/2 √
q
g(i) = 2 2e = i 2.
Lecture 14
Isolated Singularities
A punctured domain is an open set with a point removed. For p ∈ Ω, we use the notation
Ω∗p = Ω \ {p},
or simply Ω∗ for Ω \ {0}, if 0 ∈ Ω or when there is no confusion about the point removed. The aim of this
lecture is to study functions that are holomorphic on punctured domains. The puncture, that is the point
p in the above case, is called an isolated singularity. These come in three types -
• Removable singularities
• Poles
• Essential singularities
14.1 Removable singularities

A holomorphic function f ∈ O(Ω∗p ) is said to have a removable singularity at p if there exists a holomorphic
function f˜ ∈ O(Ω) such that
f˜ = f.
Ω∗
p
Theorem 14.1.1. Let f ∈ O(Ω∗p ). Then the following are equivalent.

1. f has a removable singularity at p.
2. f can be extended to a continuous function on Ω.
3. f is bounded in a neighborhood of p.
4. limz→p (z − p)f (z) = 0.
Proof. The implications (1) =⇒ (2) =⇒ (3) =⇒ (4) are trivial. To complete the proof, we need to
show that (4) =⇒ (1).
For convenience, suppose p = 0. So suppose f (z) satisfies
lim zf (z) = 0,
z→0
83
84 LECTURE 14. ISOLATED SINGULARITIES
and define a new function (

z 2 f (z), z ̸= 0
g(z) =
0, z = 0
Claim. g(z) is holomorphic on Ω and moreover, g ′ (0) = 0.

Clearly g(z) is holomorphic on Ω∗ . So we only need to prove holomorphicity at z = 0. Let us compute
the difference quotient. Since g(0) = 0, g ′ (0) if it exists is equal to
g(h)
lim = lim hf (h) = 0
h→0 h h→0
by hypothesis. This proves the claim. By analyticity, g(z) has power series expansion in a neighborhood
of z = 0. That is, there is a small disc Dε (0) such that for all z ∈ Dε (0),
∞
X g (n) (0) n
g(z) = z .
n=0
n!
Now, since g(0) = g ′ (0) = 0,

∞
X g (n) (0) n−2
g(z) = z 2 z .
n=2
n!
By comparing to the definition of g(z) we see that for z ∈ Dε (0) \ {0},

∞
X g (n) (0) n−2
f (z) = z .
n=2
n!
and so we define (P
∞ g (n) (0) n−2
n! z , z ∈ Dε (0)
f˜(z) = n=2
∗
f (z), z ∈ Ω .
This is a well defined function, since on the intersection Ω∗ ∩ Dε (0), f (z) is equal to the infinite series. f˜
is clearly holomorphic on Ω∗ since it equals f (z) in this region. Moreover, since it is a power series in a
neighborhood of z = 0, it is also holomorphic at z = 0. Hence f˜ satisfies all the properties in (1), and this
completes the proof.
Remark 14.1.1. Recall that in lecture-7 we proved that Goursat’s theorem was valid for functions that are
holomorphic at all but one point in a domain, so long as they are bounded near that point. In view of the
above theorem, such a result is not surprising, since the function does extend to a holomorphic function
on the entire domain, to which Goursat’s theorem applies.
Example 14.1.2. Consider the holomorphic function Si : C∗ → C defined by
sin z
Si(z) = .
z
Then clearly
lim z · Si(z) = lim sin z = 0.
z→0 z→0
Hence by the theorem, Si(z) has a removable singularity at z = 0 and hence can be extended to an entire
function. It is instructive to look at the power series of sin z. Recall that
z3 z5
sin z = z − + + ...,
3! 5!
14.2. POLES 85
and so dividing by z, we see that for z ̸= 0,

∞
z2 z4 X z 2n
Si(z) = 1 − + = (−1)n .
3! 5! n=0
(2n + 1)!
The power series on the right clearly defines an entire function (an is in particular also defined at z = 0), and
hence Si(z) defines an entire function.
14.2 Poles
Let f ∈ O(Ω∗p ). We say that p is a pole if
lim |f (z)| = ∞.
z→p
Theorem 14.2.1. Let f ∈ O(Ω∗p ). Then the following are equivalent.

1. f has a pole at p.
2. There exists a small disc Dε (p) and a holomorphic function h : Dε (p) → C such that h(p) = 0 and
h(z) ̸= 0 for any other z ∈ Dε (p), and
1
f (z) =
h(z)
for all z ∈ Dε (p) \ {p}.
3. There exists a holomorphic function g : Ω → C such that g(p) ̸= 0, and an integer m ≥ 1 such that
for all z ∈ Ω∗p ,
g(z)
f (z) = .
(z − p)m
4. There exists a M > 1 and integer m ≥ 1 such that on some disc Dε (p) around p, we have the estimates
1 M
m
≤ |f (z)| ≤ .
M |z − p| |z − p|m
Note that the integer m in (3) and (4) above has to be the same, and is called the order of the pole at p, and
written as νf (p).
Proof. Again for convenience, lets assume p = 0, and we denote Ω∗p = Ω∗ . Suppose |f (z)| → ∞ as
z → 0. Then clearly there is a small disc Dε (0) on which f does not have a zero. Then h(z) = 1/f (z) is
holomorphic in the punctured disc Dε (0)∗ . Moreover,
1
lim |h(z)| = = 0,
z→0 limz→0 |f (z)|
and hence in particular is bounded near z = 0. By Theorem 14.1.1, h(z) actually extends to a holomorphic
function to the entire disc Dε (0), which we continue to call h(z), and from the limit it is clear that h(p) = 0.
So h(z) atisfies all the conditions in (2), and this proves that (1) =⇒ (2).
To show that (2) =⇒ (3), note that by the theorem on zeroes, since h is not identically zero, there exists
an integer m such that for all z ∈ Dε (0),
h(z) = z m g1 (z),
where g1 (p) ̸= 0. Moreover, since h(p) = 0, we must necessarily have m ≥ 1. Now consider the function
g(z) = z m f (z)
holomorphic on Ω∗ . Then on Dε (0) \ {0}, g(z) = 1/g1 (z). Since g1 (p) ̸= 0 and g1 is holomorphic
on Dε (0), we see that 1/g1 (z) is bounded on Dε (0). Hence by the removable singularity theorem, g(z)
extends to a holomorphic function on all of Ω, and satisfies all the conditions in (3).
To show that (3) =⇒ (4), note that since g is holomorphic near z = 0, it will in particular be bounded
in a neighborhood. So there exists M > 0 such that for all z ∈ Dε (0),
|g(z)| ≤ M.
On the other hand, since g(p) ̸= 0, by continuity, for the ε > 0 above, there exists a δ such that
|g(z)| ≥ δ
for all z ∈ Dε (0). Take M large enough so that 1/M < δ, then we see that on Dε (0),
1
≤ |g(z)| ≤ M,
M
and this proves (4).
(4) =⇒ (1) also holds trivially, thus completing the proof of the Theorem.
Example 14.2.1. The function

cos z
cot z = .
sin z
has poles at all the zeroes of sin z (since cos z and sin z do not share any zeroes, there is no “cancellation" of
the poles). Let us find the order of the zero at z = 0. Near z = 0, sin z ≈ z. More precisely,
z cos z cos z
z cot z = = ,
sin z Si(z)
where Si(z) is the function from the last section. Then we saw from the power series expansion, that Si(0) = 1
and hence cos z/Si(z) → 1 as z → 0. In particular, for a small ε > 0, 1/2 < | cos z/Si(z)| < 2, and hence
1 cos z 2
≤ ≤ ,
2z sin z z
and so z = 0 is a pole of order m = 1. It is once again instructive to look at an expansion near z = 0. For
z ̸= 0,
cos z 1 − z 2 /2 + · · ·
=
sin z z − z 3 /6 + · · ·
1 1 − z 2 /2 + · · ·
= ·
z 1 − z 2 /6 + · · ·
1 z2 z2
= 1− + ··· 1 + + ···
z 2 6
1 z
= − + ···
z 3
From this it is clear that cot z has a pole of order z = 0.
14.3. ESSENTIAL SINGULARITIES 87
Remark 14.2.2. The idea of an expansion for a singular function near it’s pole can be generalized. Let p
be a pole for f : Ω∗p → C. Then from the theorem, we can write
g(z)
f (z) = ,
(z − p)m
for some holomorphic g : Ω → C with g(p) ̸= 0. By analyticity, in a neighborhood of p we can write
∞
X
g(z) = an (z − p)n ,
n=0
with a0 ̸= 0. Hence for z ̸= p, we have the expansion

a0 a1
f (z) = m
+ + · · · + am + am+1 (z − p) + am+2 (z − p)2 + · · · .
(z − p) (z − p)m−1
Such an expansion is called a Laurent series expansion, which we will study in greater detail in the next
lecture. The part with the negative powers is called the principal part of f near p. In fact, if we denote
by
Qp (w) := a0 wm + · · · + am−1 w,
then we can write 1
f (z) = Qp + hp (z),
z−p
where hp extends to a holomorphic function across p.
14.3 Essential singularities

If f : Ω∗p → C is holomorphic, then p is called an essential singularity if it is neither a removable singularity
nor a pole. Unlike in the case of removable singularities and poles, the function behaves rather erratically
in any neighborhood around an essential singularity.
Theorem 14.3.1 (Casorati-Weierstrass). The following are equivalent.
1. f has an essential singularity at p.
2. For any disc Dε (p), f (Dε (p)) is dense in C, that is for any disc Dε (p) and any a ∈ C, there exists a
sequence {zn } ∈ Dε (p) such that
lim f (zn ) = a.
n→∞
Proof. We first show that (2) =⇒ (1). If p is a removable singularity, then for some disc Dε (p), f (Dε (0))
is a bounded set in C, and so cannot be dense. On the other hand if p is a pole, then |f (z)| → ∞ as z → p.
In particular, there is a disc Dε (p) such that for all z ∈ Dε (p),
|f (z)| > 1,
and hence once again Dε (p) cannot be dense in C. This forces p to be an essential singularity.
Conversely, suppose p is an essential singularity. We then have to show that (2) holds. If not, then there
is a disc Dε0 (p) such that f (Dε0 (p) \ {p}) is not dense in C. Hence there exists an a ∈ C and an r > 0
such that
|f (z) − a| > r
for all z ∈ Dε (p) \ {p}. Then define g : Dε (p) \ {p} → C by
1
g(z) = .
f (z) − a
Since f (z) ̸= a on that punctured disc, g(z) is holomorphic. Moreover |g(z)| ≤ 1/r in Dε (p) \ {p}, and
hence by the removable singularity Theorem 14.1.1, there exists an extension g̃ holomorphic on Dε (p).
There are now two cases.
Case-1. g̃(p) ̸= 0. Then by continuity, there is a smaller r < ε and a δ > 0 such that |g̃(z)| > δ on Dr (p).
But away from p,
1
f (z) = + a,
g(z)
and so on Dr (p) \ {p},
1 1
|f (z)| ≤ + |a| < + |a|,
|g̃(z)| δ
and so |f (z)| is bounded in a neighborhood of p. By the removable singularity theorem, f must have a
removable singularity at z = p which is a contradiction.
Case-2. g̃(p) = 0. Then for any ε > 0, there exists a r > 0 such that on Dr ,
|g̃(z)| ≤ ε.
So by triangle inequality, if ε small enough so that |a| < 1/2ε, then on Dr (p) \ {p} we have
1 1 1 1
|f (z)| = +a ≥ − |a| ≥ − |a| >
g(z) |g(z)| ε 2ε
for all z ∈ Dr (p). This shows that limz→p |f (z)| = ∞, which is a contradiction, completing the proof of
the theorem.
Remark 14.3.1. It is a theorem of Picard’s that in any neighbourhood of an essential singularity, the image
under f is not only dense in C but misses at most one point of C!
Example 14.3.2. The function f (z) = e1/z , which is holomorphic on C∗ , has an essential singularity at
z = 0. To see this, we need to rule out the possibilities of f having a removable singularity or a pole at z = 0.
Since
n→∞
f (1/n) = en −−−−→ ∞,
f (z) is not bounded in any neighborhood of z = 0, and hence cannot have a removable singularity. On the
other hand, 1
f = e2πin = 1.
2πni
Hence the limit limz→0 f (z) cannot be infinity, and hence f cannot have a pole at z = 0. This shows that
f (z) has to have an essential singularity at z = 0. Again looking at an expansion, we see that for z ̸= 0,
1 1
e1/z = 1 + + + ··· .
z 2!z 2
So the expansion has infinitely many terms with negative powers of z. As we will see when we discuss Laurent
series, this in fact characterizes essential singularities.
Remark 14.3.3. We finally remark that non-isolated singularities can exist. For instance the function
1
f (z) = tan
z
has singularities at 0 and points pn = 2/nπ which converge to 0. The analysis in the present lecture does
not apply to such singularities.
Lecture 15
Laurent series
A Laurent series centered at z = a is an infinite series of the form

∞ ∞
X bn X
+ cn (z − a)n (15.1)
n=1
(z − a)n n=0
We can combine this into one infinite sum

∞
X a−1
an (z − a)n = · · · + + a0 + a1 (z − a) + a2 (z − a)2 + · · · ,
n=−∞
z−a
by setting (
b−n , n ≤ −1
an = (15.2)
cn , n ≥ 0.
We say that the Laurent series in (15.1) is convergent at z if both the infinite series are convergent. The first
term above is an infinite series of the form
b1 (z − a)−1 + · · · . (15.3)
Changing the variable to w = (z − a)−1 , we can re-write this as a usual power series -
b1 w + b2 w2 + · · · .
Then by the fundamental theorem for power series, there exists an R1 such that the series converges on
the disc |w| < R1−1 (or equivalently the annulus |z| > R1 ), where
R1 = lim sup |bn |1/n

n→∞
Or equivalently, the series (15.3) converges for |z − a| > R1 . On the other hand the second series in (15.1)
is a regular power series, and hence setting
R2 = (lim sup |cn |1/n )−1 ,
n→∞
the second series is convergent for |z − a| < R2 . Combining this, we have the following theorem.
Theorem 15.0.1. If R1 , R2 given by the formulae above satisfy R1 < R2 , then the Laurent series 15.1
converges for all z ∈ C such that R1 < |z − a| < R2 . Moreover, the convergence is uniform and absolute in
the region r1 ≤ |z − a| ≤ r2 for any r1 , r2 satisfying R1 < r1 < r2 < R2 . As a consequence, the limiting
function is holomorphic in the annulus R1 < |z − a| < R2 .
89
90 LECTURE 15. LAURENT SERIES
Henceforth if R1 < R2 we will denote the annulus of inner radius R1 and outer radius R2 by
AR1 ,R2 (a) = {z ∈ C | R1 < |z − a| < R2 }.
Our main result in this chapter is a converse.

Theorem 15.0.2. Let R1 < R2 , and f be holomorphic on a domain containing the closure of the annulus
AR1 ,R2 (a). Then for all z ∈ AR1 ,R2 (a),
∞
X
f (z) = an (z − a)n ,
n=−∞
where ˆ
1 f (ζ)
an = dζ,
2πi Cr (ζ − a)n+1
for any r ∈ [R1 , R2 ]. Moreover, the series converges uniformly and absolutely on any compact subset of
AR1 ,R2 (a).
First we need the following elementary observations.
Lemma 15.0.1. Let F be holomorphic on any domain containing the closure of the annulus AR1 ,R2 (a). Then
ˆ
F (z) dz
Cr (a)
is independent of r ∈ [R1 , R2 ].
Proof. Let R1 < r1 < r2 < R2 . For simplicity let us denote Cri (a) = Ci . We claim that C1 ∼AR1 ,R2 (a) C2 .
The lemma then follows from the generalized Cauchy theorem. To prove the claim, we need to compute
indices. Let w ∈
/ AR1 ,R2 (a). Then either |w| > R2 or |w| < R1 . If it is the former, then w ∈/ Int(C1 ) and
w∈ / Int(C2 ). Hence n(C2 , w) = n(C1 , w) = 0. On the other hand, if it is the latter, then w ∈ Int(C1 ) ⊂
Int(C2 ), and so n(C1 , w) = n(C2 , w) = 1. In either case, for all w ∈/ AR1 ,R2 (a), n(C1 , w) = n(C2 , w),
and hence by definition C1 ∼AR1 ,R2 (a) C2 .
Lemma 15.0.2 (CIF for annuli). Let f be holomorphic on a domain containing the closure of the annulus
AR1 ,R2 (a). Then for all z ∈ C such that R1 < |z − a| < R2 we have
ˆ ˆ
1 f (ζ) 1 f (ζ)
f (z) = dζ − dζ.
2πi CR2 ζ −z 2πi CR1 ζ −z
Proof. For convenience, we use the notation A = AR1 ,R2 (a). Fix z ∈ A, and consider the function
(
f (ζ)−f (z)
ζ−z ζ ̸= z
,
g(ζ) = ′
f (z), ζ = z.
Clearly g(ζ) is holomorphic on the punctured annulus A \ {z}. But it is continuous on the whole of the
annulus since f is holomorphic at z. Hence by the theorem on removable singularities, g(ζ) is holomorphic
on all of A. Then by the above lemma
ˆ ˆ
g(ζ) dζ = g(ζ) dζ.
CR2 CR1
91
Since z ∈
/ CR2 or CR1 , the above is equivalent to
ˆ ˆ ˆ ˆ
f (ζ) f (ζ) f (z) f (z)
dζ − dζ = dζ − dζ
CR2 ζ − z CR1 ζ − z CR2 ζ − z CR1 ζ −z
ˆ ˆ
dζ dζ
= f (z) − f (z)
CR2 ζ − z CR1 ζ −z
= 2πif (z)(n(CR2 , z) + n(CR1 , z))
Since z ∈ Int(CR2 ) but lies in Ext(CR1 ), n(CR2 , z) = 1 and n(CR1 , z) = 0, and this completes the proof
of the Lemma.
Proof of theorem 15.0.2

This is similar to the proof of analyticity, and the key tool as before is the geometric series expansion
∞
1 X
= wn ,
1 − w n=0
which is valid in the region |w| < 1. By the Lemma above

ˆ ˆ
1 f (ζ) 1 f (ζ)
f (z) = dζ − dζ := I2 − I1 .
2πi CR2 ζ − z 2πi CR1 ζ − z
To evaluate I2 , we write
1 1 1 1
= =
ζ −z ζ − a − (z − a) ζ − a 1 − (z − a)/(ζ − a)
Since ‘a’ is the center of the annulus, if ζ ∈ CR2 , and z ∈ Int(CR2 ), then
|z − a| |z − a|
= < 1.
|ζ − a| R2
Applying the geometric series expansion with w = (z − a)/(ζ − a) we see that
ˆ ∞
1 f (ζ) X (z − a)n
I2 = n
dζ
2πiCR2 ζ − a n=0 (ζ − a)
∞ ˆ
X 1 f (ζ)
= n+1
dζ · (z − a)n
n=0
2πi CR2 (ζ − a)
∞
X
= an (z − a)n ,
n=0
where ˆ
1 f (ζ)
an = dζ.
2πi CR2 (ζ − a)n+1
To analyze I1 , we write
1 1 1
=− .
ζ −z z − a 1 − (ζ − a)/(z − a)
But now, if ζ ∈ CR1 , then for z ∈ AR1 ,R2 (a) we have that
|ζ − a| R1
= < 1,
|z − a| |z − a|
and so again from the geometric series expansion it follows that

ˆ ∞
1 f (ζ) X (ζ − a)k
I1 = − k
dζ
CR1 z − a k=0 (z − a)
2πi
∞ ˆ
X 1
=− f (ζ)(ζ − a)k dζ (z − a)−k−1 .
2πi CR1
k=0
Putting k + 1 = −n, we can write

−∞
X
I1 = − an (z − a)n ,
n=−1
where ˆ
1 f (ζ)
an = dζ.
2πi CR1 (ζ − a)n+1
This completes the proof of the theorem.
15.1 Application to study of isolated singularities

Corollary 15.1.1. Let f : Ω∗p → C holomorphic. The for any disc DR (p) such that DR (p) ⊂ Ω,
∞
X
f (z) = an (z − p)n ,
n=−∞
where ˆ
1 f (ζ)
an = dζ.
2πi ∂DR (p) (ζ − p)n+1
Proof. Apply the theorem to the annulus Ar,R (p) and let r → 0.
We then have the following characterization of isolated singularities based on the Laurent series expansion.
Theorem 15.1.1. Let f : Ω∗p holomorphic with Laurent series expansion
∞
X
f (z) = an (z − p)n
n=−∞
around p. Then
1. p is a removable singularity if and only if an = 0 for all n < 0.
2. p is a pole of order m if and only if an = 0 for all n < −m.
3. p is an essential singularity if and only if for any N > 0, there exists an n < −N such that an ̸= 0.
That is, there are infinitely many non-zero negative exponent terms in the Laurent series expansion.
Proof. Note that if DR (p) ⊂ Ω, then the coefficients are given by the formula
ˆ
1 f (ζ)
an = dζ.
2πi ∂DR (p) (ζ − p)n+1
So if p is a removable singularity, then for integers n < 0, (ζ − p)−n−1 f (ζ) is holomorphic on the entire
disc DR (p), and hence by Cauchy’s theorem for discs, an = 0 for all n < 0. Conversely, if an = 0 for
15.1. APPLICATION TO STUDY OF ISOLATED SINGULARITIES 93
n < 0, the Laurent series reduces to a power series, and we know that power series are holomorphic on
the entire disc of convergence.
To prove the characterization of poles, apply the same argument to the function (ζ − p)m f (ζ). The char-
acterization of essential singularities then follows from the definition and the first two parts.
Remark 15.1.2. Note that in the event the function has only poles, the Laurent series of the function
centered at some other points might have infinitely many negative exponent terms. The theorem only
states that the Laurent series centered at the isolated singularity can have only finitely many negative
exponent terms. For an illustration of this, see Example 15.1.4 below.
Example 15.1.3. Consider the function
1 1 1
= − .
z 2 − 3z + 2 z−2 z−1
It has two singularities at z = 1 and z = 2 which are clearly poles. We can expand the function as a Laurent
series centered at either of the poles. To illustrate this, let us find the Laurent series expansion centered at z = 1.
One approach is to use the formula for the coefficients in Theorem 15.0.2 and compute out all the integrals. An
easier approach is to use the geometric series expansion, namely that
∞
1 X
= wn
1 − w n=0
whenever |w| < 1. Note that the function is holomorphic on the annulus 0 < |z − 1| < 1, and so we can hope
to have a Laurent series expansion on that domain. Writing
1 1 1
= =− .
z−2 z−1−1 1 − (z − 1)
Since |z − 1| < 1, using the geometric series expansion (with w = z − 1) we see that
∞
1 X
=− (z − 1)n ,
z−2 n=0
and so
∞
1 1 X
= − − (z − 1)n .
z 2 − 3z + 2 z − 1 n=0
Example 15.1.4. Sticking with the function from the previous example, one can also try to find a Laurent series
expansion on other annuli. For instance the function is holomorphic on the annulus A1,2 (0) = 1 < |z| < 2.
We consider each of the terms in the partial fraction decomposition separately. For z ∈ A1,2 (0), |z| > 1 and
so applying the geometric series expansion above to w = 1/z, we see that
∞ ∞
1 1 1 1 X 1 X 1
= = = .
z−1 z 1 − 1/z z n=0 z n n=1
z n
On the other hand, for z ∈ A1,2 (0), |z| < 2 and hence once again applying the geometric series expansion to
w = z/2,
∞ ∞
1 1 1 1 X zn X zn
=− =− n
= − n+1
.
z−2 2 1 − z/2 2 n=0 2 n=0
2
Putting it all together, we see that on 1 < |z| < 2,
∞ ∞
1 X 1 X zn
= − − .
z 2 − 3z + 2 n=1
z n n=0 2n+1
So even though the function only has poles, it Laurent series centred around z = 0 has infinitely many negative
exponent terms.
Lecture 16
Meromorphic functions
16.1 Definition and basic properties

A function on a domain Ω is called meromorphic, if there exists a sequence of points p1 , p2 , · · · with no
limit point in Ω such that if we denote Ω∗ = Ω \ {p1 , · · · }
• f : Ω∗ → C is holomorphic.
• f has poles at p1 , p2 · · · .
We denote the collection of meromorphic functions on Ω by M(Ω). We have the following observation,
whose proof we leave as an exercise.
Proposition 16.1.1. The class of meromorphic function forms a field over C. That is, given any meromorphic
functions f, g, h ∈ M(Ω), we have that
1. f ± g ∈ M(Ω),
2. f g ∈ M(Ω),
3. f (g + h) = f g + f h.
4. f ± 0 = f, f · 1 = f ,
5. 1/f ∈ M.
Recall that if a holomorphic function has finitely many roots, then it can be “factored" as a product of a
polynomial and a no-where vanishing holomorphic function. Something similar holds true for meromor-
phic functions.
Proposition 16.1.2. Let f ∈ M(Ω) such that f has only finitely many poles {p1 , · · · , pn } with orders
{m1 , · · · , mn }. Then there exist holomorphic functions g, h ∈ O(Ω) such that for all z ∈ Ω \ {p1 , · · · , pn },
g(z)
f (z) = .
h(z)
Moreover, we can choose g and h such that f (z) and g(z) have the exact same roots with same multiplicities,
while h(z) has zeroes precisely at p1 , · · · , pn with multiplicities exactly m1 , · · · , mn .
Proof. We define g : Ω \ {p1 , · · · , pm } by

g(z) = Πnk=1 (z − pk )mk f (z).
95
96 LECTURE 16. MEROMORPHIC FUNCTIONS
This is clearly a holomorphic function. Moreover, since f (z) has a pole of order mk at pk , g(z) is bounded
in a neighbourhood of pk . Thus, by the theorem on removable singularities, g(z) can be extended as a
holomorphic function on Ω. The theorem is then proved with
h(z) = (z − p1 )m1 · · · (z − pn )mn .
Remark 16.1.3. The same is true even if the meromorphic function has infinite number of poles. This is
a consequence of Weierstrass’ factorization theorem. We will prove this theorem for the special case when
Ω = C. For a general open set the proof requires the use of Runge’s approximation theorem.
16.2 Partial fraction decomposition of meromorphic functions on

C.
Recall that if f has a pole of order m at p, then the Laurent series expansion can be written as
1
f (z) = Qp f + Hp f (z),
z−p
where Hp f is holomorphic near p, and Qp f (w) is a polynomial
Qp f (w) = a−m wm + · · · a−1 w,
where for each n = 1, 2, · · · , m and each ε << 1, we have
ˆ
1
a−n = f (z)(z − p)n−1 .
2πi |z−p|=ε
The difference f (z) − Hp f (z) is called the principal part of f (z) at p. We then have the following funda-
mental theorem.
Theorem 16.2.1 (Mittag-Leffler). Let {pk } be a discrete set of points in Ω, and for each k, let Qk (w) be
a polynomial without a constant term. There there exists a f ∈ M(Ω) with poles at pn and holomorphic
everywhere else, with principle part at pk given by Qk (1/(z − pk )). Moreover, all such meromorphic functions
are of the form
X 1
f (z) = Qk − qk (z) + H(z),
z − pk
k
where each qk (z) and H(z) are holomorphic functions on Ω. Furthermore:
1. If {pk } is a finite sequence, then one could take qk ≡ 0.
2. If Ω = C, and |pk | → ∞, then one could take each qk to be a polynomial.
We will prove parts (1) and (2) in the next lecture. For a general Ω and infinitely many poles, the proof
require’s Runge’s theorem, and an outline will be provided in the appendix to the next lecture.
P
Remark 16.2.2. Note that if k gk is a compactly convergent series on Ω (for instance, take a power
P
series), then q̃k = qk + gk , and H̃ = H + k gk will give another representation for the function f (z),
and hence qk and H are by no means unique.
Remark 16.2.3. Note that given a meromorphic function, the theorem does not say whether that particular
function has a partial fraction decomposition (unlike say for rational functions, as we will see in the next
section, or more generally meromorphic functions with only a finite number of poles). In other words, we
are not claiming that the converse holds (even when Ω = C). It does turn out that the converse holds under
some additional conditions on the distribution of poles. But in particular examples, one can get away by a
more hands-on approach. We will see a beautiful illustration of this below.
16.2. PARTIAL FRACTION DECOMPOSITION OF MEROMORPHIC FUNCTIONS ON C. 97
Example 16.2.4. Consider the meromorphic function f (z) = π 2 / sin2 πz which is a meromorphic function
with poles at integers. Near zero,
π2 π2 1 1 z2 2
2 = 2 = 2 = 2 1 + + ··· ,
sin πz z 6
πz + O(z 3 ) z 2 1 − z 2 /6 + · · ·
and so the principal part of f (z) is given by 1/z 2 . Using the identity sin2 (π(z − n)) = sin2 πz, it is easy to
see that the principal parts at z = n are given by (z − n)−2 . Now consider the series
∞
X 1
.
n=−∞
(z − n)2
This converges uniformly on all compact subsets of C \ Z, and hence represents a meromorphic function
on C with poles of order two at all integers. Moreover the principle parts at each pole z = n is given by
Qn ((z − n)−1 ), where Qp (w) = w2 . It is then easy to see that the difference
∞
π2 X 1
H(z) := −
sin2 πz n=−∞ (z − n)2
extends to an entire function.
Claim. H ≡ 0.
Proof. Note that the series and the function on the left are both periodic with period 1, and hence so is
H(z). That is, H(z + 1) = H(z) for all z ∈ C. Also by Euler’s identity, if z = x + iy, then
eiπz − e−iπz
sin πz = = sin(πx) cosh(πy) + sinh(πy) cos(πx),
2i
and so
| sin πz|2 = cosh2 (πy) − cos2 (πx) ≥ cosh2 (πy) − 1 → ∞
uniformly as |y| → ∞. As a consequence π 2 / sin2 πz converges uniformly to zero as |y| → ∞. But the
infinite series also shares this property. Indeed, since the series converges uniformly on |y| ≥ 1, we can take
pointwise limit, and clearly each (z − n)−2 → 0 uniformly as |y| → ∞. The upshot is that H(z) converges
uniformly to zero as |y| → 0. In particular, H(z) is bounded on the strip {z ∈ C | 0 ≤ Re(z) ≤ 1}. But
then since H is periodic with period one, this means that H is a bounded entire function, and hence a
constant by Liouville. But since limy→0 H(iy) = 0, we can conclude that H ≡ 0.
Assuming this, we get the identity

∞
π2 X 1
=
sin2 πz n=−∞ (z − n)2
for all z ∈
/ Z. Plugging in z = 1/2, we obtain the identity
∞ ∞
π2 X 1 X 1
= 2
= 2 .
4 n=−∞
(2n − 1) n=1
(2n − 1)2
P∞
Now let, S = m=0 m−2 . Then we have
∞ ∞
X 1 X 1
S= +
n=1
(2m − 1)2 n=1 (2n)2
π2 S
= + .
8 4
Solving for S, we get the beautiful identity
∞
X 1 π2
2
= .
m=1
m 6
Proving this identity was the so-called Basel problem, first “solved" by Euler. But his “proof" would not pass our
modern day standards of rigour. Euler used a “facotrization" for sine, but a rigorous development of the theory
of infinite product factorizations of entire functions had to wait till Weierstrass came along many decades later.
Nevertheless, Euler’s insights were of course crucial in all subsequent developments.
16.3 Holomorphic self maps on the extended complex plane

It is often useful to think of z = ∞ on the same footing as other points in the complex plane, and to define
the extended complex plane
C = C ∪ ∞.
We can then think of meromorphic functions f : Ω \ {p1 , · · · , pj , · · · } → C, as functions f : Ω → Ĉ, by
defining
f (pj ) = ∞
for all the poles pj s.
Definition 16.3.1. Let Ω ⊂ C. We say that a map f : Ω → C is holomorphic if for any p ∈ Ω, either
f (p) ̸= ∞ and f is holomorphic near p, or f (p) = ∞ and 1/f (z) is holomorphic near p.
Proposition 16.3.2. A function f : Ω → C is holomorphic if and only if f ∈ M(Ω) with poles precisely on
the polar set f −1 (∞).
Similarly, in studying meromorphic functions on C, it is also useful to consider the extension of the function
themselves to Ĉ. We say that z = ∞ is a pole of order m (resp. removable or essential singularity) if z = 0
is a pole of order m (resp. removable or essential singularity) for the function
fˆ(z) = f (1/z).
Similarly we can also define a zero of order m at infinity. We then say that a meromorphic function on C
is meromorphic on the extended plane, if it does not have an essential singularity at z = ∞. It turns out
that meromorphic functions on Ĉ can be classified. Recall that a rational function on C is a function of the
form
P (z)
R(z) =
Q(z)
where both P (z) and Q(z) are polynomials.
Example 16.3.3. 1. A polynomial P (z) = an z n + · · · + a0 with an ̸= 0 has a pole of order n at infinity.
In fact, conversely, ever entire function p(z) with a pole of order n at infinity is a polynomial of degree
n. This follows from the Cacuhy estimates.
2. The function ez has an essential singularity at infinity.
3. A rational function has a pole or removable singularity at infinity. In fact a rational function R(z) =
P (z)/Q(z) as above has
• a pole of order deg P − deg Q at infinity if deg(P ) > deg(Q),
• a removable singularity at infinity if deg(P ) ≤ deg(Q),
16.3. HOLOMORPHIC SELF MAPS ON THE EXTENDED COMPLEX PLANE 99
• a zero of order deg(Q) − deg(P ) if deg(P ) < deg(Q).

Theorem 16.3.4. The only meromorphic functions on Ĉ are rational functions.
Proof. Let F : Ĉ → Ĉ be a meromorphic function.
Claim-1. F has only finitely many poles {p1 , · · · , pn } in the complex plane C.
To see this, note that F (1/z) has either a pole or zero at z = 0. In either case there is a small neighborhood
|z| < ε which has no other pole. Which is the same as saying that F has no finite pole in |z| > 1/ε. But
|z| ≤ 1/ε is compact, and since all poles are isolated, this shows that there are only finitely many poles.
Now, corresponding to each of the poles pk ∈ C there exists a polynomial Qk (see Remark 0.2 in Lecture-20)
such that 1
F (z) = Qk + Hk (z),
z − pk
where Gk is holomorphic on a whole neighborhood around pk (including at the point pk ). Similarly if
|z| > R, we can write
1
F (z) = Q∞ (z) + H∞ ,
z
where as before, H∞ (z) is holomorphic in a neighborhood of z = 0.
Claim-2. The function

n
X 1
G(z) = F (z) − Q∞ (z) − Qk
z − pk
k=1
is an entire and bounded function.

Assuming the claim, by Liouville’s theorem, G(z) is a constant, and hence F (z) must be rational, and the
theorem is proved. To prove the claim, first note that clearly, G(z) is holomorphic away from {p1 , · · · , pn }.
At some z = pk , Qj (1/z − pj ) is holomorphic for all j ̸= k. On the other hand, near pk ,
1
F (z) − Qk = Hk (z)
z − pk
which is holomorphic. This shows that G(z) is entire. As a consequence, to show boundedness, we only
need to show boundedness on |z| > R for some large R. To see, first observe that since Qk are polynomials,
1
lim Qk = 0.
z→∞ z − pk
Hence it is enough to show that F (z) − Q∞ (z) is bounded near infinity. But this follows immediately
from noting that
1 1
H∞ (z) = F − Q∞
z z
is holomorphic near z = 0 and hence is bounded on |z| < ε for some ε > 0. In particular F (z) − Q∞ (z)
is bounded on |z| > 1/ε. This proves the claim, and hence completes the proof of the theorem.
Remark 16.3.5. A meromorphic function f ∈ M(C) gives rise to a holomorphic map F : P! → P1 .

Conversely, given any map F : P1 → P1 , one gets a meromorphic map from C → C with poles at
F −1 ([0, 1]). So the theorem can be reformulated in the following way - all holomorphic maps from F :
P1 → P1 are given by rational functions of two variables, where the numerator and denominator are
homogenous polynomials.
A simple consequence of the proof is the following theorem on partial fraction decomposition that we take
for granted as an important tool in integration theory, but never see the proof of.
Corollary 16.3.6. For any rational function R(z) = P (z)/Q(z) has a partial fraction decomposition of the
form
Xn 1
R(z) = Q∞ (z) + Qk ,
z − pk
k=1
where pk is a root of Q(z) of order mk , Qk is a polynomial of degree mk , and deg Q∞ = deg P − deg Q if
this number is non-negative. Else we have that Q∞ ≡ 0.
Lecture 17
The theorems of Mittag Leffler and

Weierstrass
17.1 Proof of the Mittag-Leffler theorem.

Recall that the Mittag-Leffler theorem was as follows.
Theorem 17.1.1 (Mittag-Leffler). Let {pk } be a discrete set of points in Ω, and for each k, let Qk (z) be
a polynomial without a constant term. There there exists a f ∈ M(Ω) with poles at pk and holomorphic
everywhere else, with principle part at pk given by Qk (1/(z − pk )). Moreover, all such meromorphic functions
are of the form
X 1
f (z) = Qk − qk (z) + H(z),
z − pk
k
where each qk (z) and H(z) are holomorphic functions on Ω, and qk depends only on Qk . Furthermore:
1. If {pk } is a finite sequence, then one could take qk ≡ 0.
2. If Ω = C, and |pk | → ∞, then one could each qk to be a polynomial.
Proof. The proof of the theorem in case (1) is trivial, and we leave it as an exercise. We prove the theorem
only in the case (2) above. For a general Ω with possibly infinite sequence {pk }, the proof relies on Runge’s
theorem and is out of the scope of the present course.
So from now, suppose Ω = C. Without loss of generality, we can assume that no pk is equal to 0. Suppose
we order them such that 0 < |p1 | ≤ · · · . By the first part, we can assume that the number of poles is
infinite, and hence that limk→∞ |pk | = ∞. Since each Qk (1/(z − pk )) is holomorphic on |z| < |pk |, it
can be expanded as a Taylor series around z = 0. Let qk be the partial sum of degree dk of this Taylor
expansion. Let
1
Mk = sup Qk .
|z|≤|pk |/2 z − pk
Claim. For all z such that |z| ≤ |pk |/4, we have the estimate
1 2|z| dk +1
Qk − qk (z) ≤ 2Mk ≤ Mk 2−dk .
z − pk |pk |
101
102 LECTURE 17. THE THEOREMS OF MITTAG LEFFLER AND WEIERSTRASS
We assume the claim for the moment. Now pick dk >> 1 such that 2dk ≥ Mk 2k , and consider the series
X 1
f (z) = Qk − qk (z) .
z − pk
k
For any compact set K ⊂⊂ C \ {p1 , · · · , }, there exists a N such that for k > N , K ⊂ D|pk |/4 (0). By the
claim and our choice of dk , for all k > N , each term of the infinite series
∞
X 1
Qk − qk (z)
z − pk
k=N +1
−k
is dominated by 2 , and hence by Weierstarss test the tail, represents a holomorphic function. On the
other hand
N
X 1
Qk − qk (z)
z − pk
k=1
is a meromorphic function on K with poles at p1 , · · · , pN with prescribed principal parts. In particular,
f (z) is meromorphic on C with poles at pk with principal part Qk ((z − pk )−1 ). To finish the proof, if f˜ is
another such function, then clearly f˜ − f extends to an entire function.
Proof of the claim. Suppose
∞
1 X
Qk − qk (z) = aj z j
z − pk
j=dk +1
2j Mk
in the disc |z| < |pk |/2. By the Cauchy estimate, we have that |aj | ≤ |pk |j , and so
∞
1 X 2j |z|j
Qk − qk (z) ≤ Mk
z − pk |pk |j
j=dk +1
2|z| dk +1 X∞
1
≤ Mk j
|pk | j=0
2
2|z| dk +1
≤ 2Mk ,
|pk |
where we used the fact that |z| ≤ |pk |/4 in the penultimate line.
Example 17.1.2. In the previous lecture, we illustrated the theorem by obtaining an expansion for π 2 / sin2 πz.
We now obtain an expansion for π cot πz which is meromorphic on C with Ponly simple poles at z = n ∈ Z. In
fact, the principal part is precisely (z − n)−1 . Unfortunately, the series (z − n)−1 is divergent, and hence
one has to subtract off a polynomial, which in this case turns out to be a constant. Consider the series
X 1 1 X 1
+ = ,
z−n n (z − n)n
n̸=0 n̸=0
n−2 . Hence the series

P
which is compactly convergent on C \ Z as can be seen by comparing with the series
represents a meromorphic function on C with simple poles at z = n. Then clearly
1 X 1 1
H(z) = π cot πz − − +
z z−n n
n̸=0
is an entire function. Moreover, by direct calculation, one can see that for z ∈
/ Z (and hence everywhere),
π2 X 1
H ′ (z) = 2 − =0
sin πz n∈Z (z − n)2
17.2. INFINITE PRODUCTS 103
by our expansion from previous lecture. Hence H(z) is a constant. Now, rewriting the function as
1 m ∞
X 1 1 1 X 2z
H(z) = π cot πz − lim + + = π cot πz − − .
m→∞ z n=−m
z−n n z n=1 z 2 − n2
The right hand side is an odd function, and hence H(z) must be zero. Thus we obtain the identity
1 X 1 1
π cot πz = + + . (17.1)
z z−n n
n̸=0
17.2 Infinite products

A infinite product of non-zero complex numbers Π∞
n=1 pn is said to converge if
P := lim Πnk=1 pk
n→∞
exists. If some of the terms are allowed to be zero, then we say that the infinite product converges if the
following two conditions hold
1. At most a finite number of terms are zero.
2. If N > 0 such that pn ̸= 0 for all k > N , then Π∞
k=N +1 pn converges in the above sense.
. If we denote the nth partial product by Pn = Πnk=1 pk , then it is clear that it any convergent product
pn = Pn /Pn−1 converges to 1. Denoting pn = 1 + bn , we say that the product converges absolutely if
Π(1 + |bn |) converges. We then have the following basic fact.
Proposition 17.2.1. Let {bn } be a sequence of complex functions,
P none of which is zero. Then the infinite
product Π(1 + bn ) (absolutely) converges if and only the series log(1 + bn ) (absolutely) converges, where
log is the principal branch of the logarithm.
One can similarly talk about uniform convergence and compact convergence of infinite products. We then
have the following counterpart of the above theorem.
Proposition 17.2.2. Let {fn } be a sequence of entire. Suppose that for every compact set K, all but finitely
∞
many
P fn s are zero free in K. Then Πn=1 (1 + fn (z)) converges compactly (resp. absolutely) on C if and only
if log fn (z) converges compactly (resp. absolutely) on C. In such a case the infinite product converges to an
entire function.
Note that absolutely convergent products also satisfy the “rearrangement property".
17.3 Weierstrass factorisation theorem

Theorem 17.3.1. Let {an } be an arbitrary sequence of non-zero complex numbers ordered such that limn→∞ |an | =
∞, if the sequence is infinite. Then there exists an entire function with zeroes at precisely the points an . More-
over, every entire function with these and no other zeroes (except possibly at z = 0) is given by
z qn (z)
f (z) = z m eg(z) Π∞n=1 1 − e , (17.2)
an
where qn is a polynomial given by
z 1 z 2 1 z m n
qn (z) = + + ··· + .
an 2 an mn an
The convergence here is absolute, and uniform on compact sets. Furthermore, if there exists some integer h > 0,
such that
X 1
1+h
< ∞, (17.3)
n
|an|
then we can take mn = h for all n.

The expression (17.2) above is called the canonical product associated with {an }, and the smallest integer h
satisfying (17.3) (if it exists) is called the genus of the canonical product. Else the genus is said to be infinite.
If g(z) above reduces to a polynomial, then we say that f (z) is of finite genus, and the genus of f (z) is
defined to be the maximum of the degree of g(z) and h.
Proof. We only prove the second part, and leave the more general statement as an exercise. So from now
on assume that our sequence satisfies (17.3). We need to prove the existence of polynomials qn (z) such
that the infinite product z qn (z)
Π∞n=1 1 − e
an
converges to an entire function. By Proposition 17.2.2, this holds if and only if the series with the general
term z
rn (z) = log 1 − + qn (z)
an
converges compactly. Let R > 1 be arbitrary. We only consider those an such that |an | > 2R. Then on
|z| < R, since |z/an | < 1, the first term in rn (z) has a power series expansion:
z z 1 z 2
log 1 − =− − − ··· .
an an 2 an
For the h as in the hypothesis, we let
z 1 z 2 1 z h
qn (z) = + + ··· + ,
an 2 an h an
so that
1 z h+1 1 z h+2
rn (z) = − − − ··· .
1 + h an 2 + h an
By comparison with a geometric series, we obtain the estimate that
1 R 1+h R −1 2 R 1+h

|rn (z)| ≤ 1− < .
1 + h |an | |an | 1 + h |an |
P
By hypothesis, rn (z) converges absolutely and compactly on |z| ≤ R, and hence so must the product.
Note that even though the proof required us to work with |an | > 2R, just multiplying in the terms on
the product with |an | ≤ 2R does not affect convergence of the product, since such an ’s are only finite
in number. This shows that the product with our choice of qn converges to an entire function with the
required properties. Now suppose f (z) is any other such entire function with a zero of order m at z = 0.
Then consider the entire function
f (z)
F (z) = .
z m Π∞
n=1 1 − azn eqn (z)
Since the zeroes of the numerator and denominator match up, F (z) clearly extends as an entire function
which is no-where zero. But then by simple connectivity of C, this implies that one can take a holomorphic
branch g(z) = log F (z). Then clearly f (z) has the expression above.
17.3. WEIERSTRASS FACTORISATION THEOREM 105
Example 17.3.2. Consider the function sin πz. This is an entire function with zeroes at z = ±n. Since
X 1
n
n1+h
converges for h = 2 (and 2 is the smallest such integer), we can apply the above proposition to obtain
z z/n
sin πz = zeg(z) Πn∈Z\{0} 1 − e .
n
Claim. eg(z) ≡ π.
To see this, we take a logarithmic derivative. Then

1 X 1 1
π cot πz = + g ′ (z) + − .
z z−n n
n̸=0
Comparing with the expansion of π cot πz, we see that g ′ (z) has to vanish, and hence g(z) is a constant. On
the other hand,
sin πz
π = lim = eg(0) ,
z→0 z
and so eg(z) ≡ π. It follows that f (z) is an entire function of genus 1. Because of absolute convergence, we
can rewrite
sin πz z2
= Π∞ n=1 1 − .
πz n2
Expanding, and comparing the coefficients of z 2 , we once again see that
∞
π2 X 1
= .
6 n=1
n2
This was the original approach of Euler.

We finish with the following elementary corollary, whose proof we leave as an exercise.
Corollary 17.3.3. Any meromorphic function in C is the quotient of two entire functions.
Lecture 18
The Residue theorem
Let f be a holomorphic function in Dε (p) \ {p} with a Laurent series expansion

∞
X
f (z) = an (z − p)n .
n=−∞
The residue of f (z) at z = p is then defined by
Resz=p f (z) = a−1 .
Then by Theorem 2 in Lecture 16, for any r < ε,

ˆ
1
Resz=p f (z) = f (z) dz.
2πi |z−p|=r
More generally, we have the following fundamental result.

Theorem 18.0.1 (Residue Theorem). Let Ω be open, {pk } ∈ Ω a sequence of isolated points, and f ∈ O(Ω∗ ),
where Ω∗ := Ω \ {p1 , · · · }. Then for any cycle γ ∼Ω 0 in Ω such that no pk lies on Supp(γ), we have
ˆ X
f (z) dz = 2πi n(γ, pk )Respk f (z).
γ k
Moreover, for any given γ as above, n(γ, pk ) = 0 for all but finitely many k, and hence the summation above
has only a finite number of non-zero terms.
Proof. First let us assume that there are only a finite number of singular points. Let Ck be a small circle
around pk enclosing a disc Dk , such that Dk ⊂ Ω and such that Ck does not intersect Supp(γ). We now
claim that X
γ ∼Ω∗ n(γ, pk )Ck . (18.1)
k
Assuming this, we are then done by the generalised Cauchy theorem, since
ˆ X ˆ
f (z) dz = n(γ, pk ) f (z) dz
γ k Ck
X
= 2πi n(γ, pk )Resz=pk f (z).
k
107
108 LECTURE 18. THE RESIDUE THEOREM
/ Ω∗ . We need to show that

To prove (18.1) let a ∈
X
n(γ, a) = n(γ, pk )n(Ck , a). (18.2)
k
If a ∈
/ Ω, then by construction a ∈/ Dk , and since a disc is simply connected, n(Ck , a) = 0. On the other
hand, since γ ∼Ω 0, we also have that n(γ, a) = 0, and hence (18.2) is trivially satisfied. If a ∈ Ω, then
a = pj for some j. Once again, as above, n(Ck , a) = 0 for all k ̸= j. On the other hand, n(Cj , a) = 1.
This verifies (18.2) and completes the proof of (18.1).
Finally, suppose the number of singularities is infinite in number. It is enough to prove that n(γ, pk ) = 0
for all but finitely many k. Note that
U0 := {a ∈ C | n(γ, a) = 0}
is an open set (since the index is locally constant). Moreover it contains the annulus AR.∞ (0) for R large
enough. As a consequence, the set U c = C \ U0 is a compact set and must contain only finitely many pk
(since the singularities are isolated).
An important corollary is the following.

Corollary 18.0.2. Let f be holomorphic in Ω except possibly at isolate points {p1 , p2 , · · · , } in Ω, and let γ
be a positively oriented, simple, closed curve in Ω not passing through any of the singularities. Then
ˆ X
f (z) dz = 2πi Resz=pk f (z).
γ pk ∈Int(γ)
Proof. This follows from the residue theorem and the fact that
(
1, z ∈ Int(γ)
n(γ, pk ) =
0, z ∈ Ext(γ).
Our next result helps in computing the residue at poles.

Proposition 18.0.3. Let f have a pole of order m at p. Then
1 dm−1
Resz=p f (z) = (z − p)m f (z).
(m − 1)! dz m−1 z=p
Proof. If f has a pole of order m, then (z − p)m f (z) has a removable singularity at z = p. Moreover, if the
Laurent series expansion for f at p is given by
∞
X
f (z) = an (z − p)n ,
n=−m
then
∞
X
m
(z − p) f (z) = ak−m (z − p)k ,
k=0
and hence
1 dm−1
(z − p)m f (z) = am−1−m = a−1 = Resz=p f (z).
(m − 1)! dz m−1 z=p
18.1. THE ARGUMENT PRINCIPLE 109
Example 18.0.4. Let us evaluate

ˆ
eπ/2z
dz,
SR 1 + z2
where SR is a square of side length 2R centred at the origin and oriented in an anti-clockwise direction. Let
f (z) be the integrand. Then it has three isolated singularities, namely an essential one at 0 and poles of order
one at ±i. Let us compute the residue at each of the singularities.
• Residue at z = 0. The Laurent series expansion is given by
∞ ∞
e(π/2z) X π n −n X 2m 2m

= z (−1) z ,
z2 + 1 n=0
2n n! m=0
hence the residue, which is the coefficient of z −1 is given by

∞
X (−1)m π 2m+1 π
Resz=0 f (z) = = sin = 1.
m=0
(2m + 1)! 2 2
• Residue at z = i. By Proposition 18.0.3, the residue is given by
eπ/2z eπ/2i 1
Resz=i f (z) = lim (z − i) 2
= =− .
z→i z +i 2i 2
• Residue at z = −i. Once again by Proposition 18.0.3, the residue is given by
eπ/2z e−π/2i 1
Resz=−i f (z) = lim (z + i) = =− .
z→−i z2 + i −2i 2
Then by the residue theorem, we have

ˆ (
eπ/2z 1, R < 1,
dz =
SR 1 + z2 0, R > 1.
18.1 The argument principle

Theorem 18.1.1 (The argument principle). Let Ω be a domain and f ∈ M(Ω) zeroes at {aj } or orders
{mj } and poles at {bk } of orders {nk }. Then for every cycle γ ∼Ω 0 which does not pass through any zeroes
or poles, we have that ˆ ′
1 f (z) X X
dz = n(γ, aj )mj − n(γ, bk )nk .
2πi γ f (z) j k
Furthermore the two summations are finite summations.

As a simple corollary we have the following.
Corollary 18.1.2. Let Ω be a simply connected domain and f ∈ O(Ω) with zeroes at {aj } or orders {mj }.
Then for any simple, closed, positively oriented curve γ no passing through any of the roots, we have
ˆ ′
1 f (z) X
dz = mj .
2πi γ f (z)
aj ∈Int(γ)
Proof of the argument principle. Note that f ′ /f has poles precisely at the zeroes and poles of f (z), and is
holomorphic everywhere else. So the integral can be computed by using the residue theorem. To do so,
we need to compute the residues of f ′ /f . There are two cases.
110 LECTURE 18. THE RESIDUE THEOREM
1. Residue of f ′ /f at z = aj . Near aj , say on Dεj (aj ), we can write f (z) = (z − aj )mj gj (z), where
gj (z) is holomorphic and zero free on Dεj (aj ). Then
f ′ (z) mj gj′ (z)

= + .
f (z) z − aj gj (z)
Since gj′ /gj is holomorphic, we have that
f ′ (z)
Resz=aj = mj .
f (z)
2. Residue of f ′ /f at z = bk . Near bk , say on Dεk (bk ), we can write f (z) = (z − bk )−nk g(z), where
gk (z) is holomorphic and zero free on Dεk (bk ). Then
f ′ (z) nk g ′ (z)
=− + k .
f (z) z − aj gk (z)
Once again, since gk′ /gk is holomorphic, we have that
f ′ (z)
Resz=bk = −nk .
f (z)
The theorem then follows by an application of the residue theorem.
Remark 18.1.3. More generally, if f is holomorphic, and we take f (z) − w, then for any simple closed
curve γ such that w ∈
/ f (Supp(γ)),
ˆ
1 f ′ (z) X
dz = mj ,
2πi γ f (z) − w
aj ∈Int(γ),f (aj )=w
where mj is the order of the zero of f (z) − w at aj .
18.1.1 Argument principle as an index calculation

If γ is a short curve such that |γ(0)| = |γ(1)|, and not passing through the origin, then the index n(γ, 0)
computes the change in the argument of γ(t) (upto a factor of 2π). To see this, note that is γ is short
enough, then one can define a holomorphic branch of the logarithm log z in a neighborhood of Supp(γ).
Then ˆ
1 dz log γ(1) − log γ(0) arg γ(1) − arg(0)
n(γ, 0) := = = .
2πi γ z 2πi 2π
Now, let γ : [0, 1] → Ω be a curve and f ∈ O(Ω). Then Γ(t) = f (γ(t)) defines a curve in C with
Γ′ (t) = f ′ (γ(t))γ ′ (t). Then
ˆ ′ ˆ 1 ′ ˆ 1 ′ ˆ
1 f (z) 1 f (γ(t)) ′ 1 Γ (t) 1 dw
dz = γ (t) dt = dt = ,
2πi γ f (z) 2πi 0 f (γ(t)) 2πi 0 Γ(t) 2πi Γ w
and hence we conclude that ˆ
1 f ′ (z)
n(Γ, 0) = dz.
2πi γ f (z)
So the integral of f ′ /f along γ essentially measures the change in argument of f (z). More generally,
together with the remark above, we see that for any w ∈ / f (Supp(γ)),
ˆ
1 f ′ (z)
n(Γ, w) = dz.
2πi γ f (z) − w
Lecture 19
Applications of the argument

principle
19.1 Local mapping properties of holomorphic functions

Theorem 19.1.1. Let Ω be connected, and f ∈ O(Ω) be a non-constant holomorphic function. Suppose
f (z0 ) = w0 , and that f (z) − w0 has a zero of order m at z0 . Then there exists an ε0 > 0 such that for any
ε < ε0 , there exists a δ = δ(ε) > 0 such that whenever w with 0 < |w − w0 | < δ, the equation f (z) = w
has exactly m distinct solutions in Bε (z0 ) each of multiplicity one.
Proof. First we choose ε0 > 0 such that

1. f (z) − w0 has no other root in Dε0 (z0 ), and
2. For all z ∈ Dε0 (z0 ), f ′ (z) ̸= 0.
The first condition can be achieved since f (z) − w0 is not constant, and zeroes are isolated. For the second
condition, if m = 1, then f ′ (z0 ) ̸= 0, and hence an ε0 > 0 as above can be picked by continuity of f ′ (z).
If m > 1, then f ′ (z0 ) = 0. But f ′ (z) is also holomorphic, and hence it’s zeroes must also be isolated. Let γ
be the circle |z − z0 | = ε oriented in the positive sense, and let Γ = f ◦ γ. Now w0 ∈ / Supp(Γ) by propert
(1) above, and hence there exists a δ > 0 such that Dδ (w0 ) ⊂ C\Supp(Γ). For any w ∈ Dδ (w0 ), since the
index is locally constant, n(Γ, w) = n(Γ, w0 ). By our discussion in the previous lecture, n(Γ, w0 ) counts
the number of zeroes of f (z) − w0 (with multiplicity) in the interior of γ, which in this case is m. Hence
n(Γ, w) = m, and so f (z)−w also has exactly m solutions in Dε (z0 ) counted with multiplicity. Now, look
at g(z) = f (z)−w. Since g ′ (z) ̸= 0 for all z ∈ Dε (z0 ), none of the roots of g(z) can have multiplicity more
than one. Hence f (z) = w has exactly m distinct solutions in Dε (z0 ), each with multiplicity one.
Remark 19.1.2. The theorem essentially says that locally, holomorphic functions are “branched" or “rami-
fied" covers. That is if f (z0 ) = w0 with multiplicity m, and with ε, δ as above, the map f : Dε (z0 )\{z0 } →
Dδ (w0 ) \ {w0 } is m : 1 covering map, and the m branches come together at z0 . If m > 1, we say that
z0 is a branch point, and that m is the branching order. The prototypical example that one should keep
in mind is f (z) = z m . Then in any small neighbourhood of z = 0 (excluding at zero), then function is
m : 1. Namely, for any w ̸= 0, then if w = reiθ , f (r1/m ζm k iθ/m
e ) = w for k = 0, 1, · · · , m − 1, where
2πi/m th
ζm = e is the primitive m root of unity.
Corollary 19.1.3 (Open mapping theorem). Let U be and open set, and f : U → C be any non-constant
holomorphic function. Then f (U ) is an open subset of C.
111
112 LECTURE 19. APPLICATIONS OF THE ARGUMENT PRINCIPLE
Proof. Let w0 ∈ f (U ). Then there exists a z0 ∈ U such that f (z0 ) = w0 . By the above theorem, there
exists a ε > 0 and δ > 0 such that Dε (z0 ) ⊂ U and f (z) = w has at least one solution in Dε (z) for each
w ∈ Dδ (w0 ). In particular, Dδ (w0 ) ⊂ U , and since w0 was arbitrary, f (U ) is open.
Remark 19.1.4. This is of course not true in the real setting, even for polynomials, much less real analytic
functions. For instance, consider f (x) = x2 on (−1, 1). Then f ((−1, 1)) = [0, 1) which is not open.
Given two open sets U and V , we say that f : U → V is a biholomorphism if f is bijective, holomorphic,
and it’s inverse f −1 : V → U is also holomorphic.
Corollary 19.1.5 (Inverse function theorem). Let f ∈ O(Ω), and z0 ∈ Ω such that f ′ (z0 ) ̸= 0, and put
w0 = f (z0 ). Then there exist ε, δ > 0 such that for every w ∈ Dδ (w0 ) there exists a unique zw ∈ Dε (z0 )
such that f (zw ) = w. Moreover we have the following explicit formula for zw :
ˆ
1 f ′ (z)
zw = z dz, (19.1)
2πi |z−z0 |=r f (z) − w
where |zw − z0 | < r < ε. In particular, if we set U = f −1 (Dδ (w0 )) ∩ Dε (z0 ), then f : U → Dδ (w0 ) is a
biholomorphism with f −1 (w) = zw and (f −1 )′ (w) = 1/f ′ (zw ).
Proof. Since f ′ (z0 ) ̸= 0, the multiplicity of f (z) = w0 is exactly one at z = z0 . By Theorem 19.1.1, there
exists ε, δ > 0 such that for all w ∈ Dδ (w0 ), there is a unique zw such that f (zw ) = w in the disc Dε (z0 ).
Also note that f ′ (z) ̸= 0 for all z ∈ Dε (z0 ). locally To prove the formula for zw , we use the residue
theorem. Consider the function
zf ′ (z)
Hw (z) = .
f (z) − w
Then since f (z) = w has a unique solution in |z − z0 | < ε, Hw (z) has a pole exactly order one at z = zw ,
and is holomorphic everywhere else. Also note that f ′ (zw ) ̸= 0. This follows since f (z) − w has a zero of
multiplicity one at zw . We then compute the residue
zf ′ (z)
Resz=zw Hw (z) = lim (z − zw )
z→zw f (z) − w
z − zw
= zw f ′ (zw ) lim = zw .
z→zw f (z) − f (zw )
Then (19.1) is proved by an application of the residue theorem. In particular, as in the statement of the
theorem, if we set U = f −1 (Dδ (w0 )) ∩ Dε (z0 ), then f : U → Dδ (w0 ) is an injective function with a
well defined inverse function f −1 : Dδ (w0 ) → U . By the open mapping theorem this inverse function is
continuous. In fact since in the formula for f −1 , the integrand depends holomorphically on w, an argument
similar to the proof of the CIF for derivative, shows that f −1 is holomorphic. By the chain rule then
(f −1 )′ (w) = 1/f ′ (zw ).
Remark 19.1.6. Another proof can be obtained by using the inverse function theorem from multivariable
calculus. Recall that if Jf (z0 ) is the Jacobian (determinant) of f when thought of as a map from subset of
R2 to R2 , then Jf (z0 ) = |f ′ (z0 )|2 ̸= 0. Hence from the inverse function theorem in calculus, there exists
a local inverse f −1 on an open neighbourhood V of w0 with continuous partial derivatives. Possibly by
shrinking V one can assume that that f ′ (z) ̸= 0 on U . All one needs to do is to show that f −1 : V → U
is holomorphic. It is enough to prove that f −1 satisfies CR equations. By chain rule,
∂ ∂f ∂f −1 ∂f ∂f −1 ∂f −1
0= f ◦ f −1 = + = f ′ (z) ,
∂ w̄ ∂z ∂ w̄ ∂ z̄ ∂ w̄ ∂ w̄
since f is holomorphic. But then since f ′ (z) ̸= 0, we see that ∂f −1 /∂ w̄ = 0 at each point.
19.2. THE MAXIMUM MODULUS PRINCIPLE 113
An elementary but important consequence of the proof is the following.

Corollary 19.1.7. A holomorphic function is locally injective on an open set U if and only if for all z ∈ U ,
f ′ (z) ̸= 0.
Proof. Suppose f ′ (z) is never zero, then the inverse function theorem implies that the function is locally
injective. Conversely, suppose the function is injective on some Dr (z0 ), but f ′ (z0 ) = 0. Then by Theorem
19.1.1 there exists a δ > 0, and w ∈ Dδ (f (z0 )) such that f (z) = w has at least two distinct solutions in
Dr (z0 ) contradicting injectivity.
Once again, the counterpart in real variable theory is false, as can be seen by considering the function
f (x) = x3 . This is globablly (and hence locally) injective, but f ′ (0) = 0.
19.2 The maximum modulus principle

The next theorem says that for non-constant holomorphic functions f (z), |f | cannot have local maximums.
Theorem 19.2.1 (Max modulus principle). Let Ω be connected and f ∈ O(Ω). If there exists a z0 ∈ Ω and
a neighbourhood U such that |f (z)| ≤ |f (z0 )| for all z ∈ U , then f (z) is a constant.
Proof. By assumption |f (z0 )| = supz∈U |f (z)|. If f (z) is non-constant on U , then by the open mapping
theorem f (U ) is an open set. In particular, there exists a δ > 0 such that for any w ∈ Dδ (f (z0 )), there
exists z ∈ U such that f (z) = w. Now simply pick w1 such that |w1 | > |f (z0 )|. Then there exists a z1 ∈ U
such that |f (z0 )| < |f (z1 )| which is a contradiction. Hence f (z) must be a constant on U . But then by
analytic continuation, f (z) must be a constant on all of Ω.
As a consequence we have the following estimate.

Corollary 19.2.2. Let Ω be a bounded set and f ∈ O(Ω) such that f extends continuously to the boundary
∂Ω. Then
sup |f (z)| ≤ sup |f (z)|.
z∈Ω z∈∂Ω
Proof. It is enough to assume that Ω is connected (or else one could work with each connected component).
Since Ω is compact, there exists a z0 ∈ Ω such that |f (z0 )| = supz∈Ω |f (z)|. If z0 ∈ ∂Ω, there is nothing
to prove. If not, then by the above theorem, z0 is an interior local maximum for |f | and hence f (z) must
be a constant. But in that case the above inequality is trivial.
Note that a minimum principle does not hold, as can be seen easily by considering the function f (z) = z
on any neighbourhood of the origin. It turns out that this is only way in which a minimum principle can
fail. Recall that O∗ (Ω) stands for holomorphic functions that are nowhere vanishing on Ω.
Corollary 19.2.3 (Minimum principle). Let Ω be connected and f ∈ O∗ (Ω). If there exists a z0 ∈ Ω and a
neighbourhood U such that |f (z)| ≥ |f (z0 )| for all z ∈ U , then f (z) is a constant.
Proof. Simply apply the maximum modulus principle to the holomorphic function g(z) = 1/f (z).
Remark 19.2.4. A function u is said to be subharmonic if ∆u ≥ 0 and superharmonic if ∆u ≤ 0. It is

a general fact that subharmonic functions satisfy a maximum principle while super harmonic functions
satisfy a minimum principle. In particular, harmonic functions satisfy both a minimum and a maximum
principle. If f (z) is holomorphic, we can compute that
1 ∂2 1 ∂
∆|f |2 = |f (z)|2 = f (z)f ′ (z) = |f ′ (z)|2 ≥ 0.
4 ∂z∂ z̄ 4 ∂z
Hence |f (z)|2 is subharmonic, and must satisfy a maximum principle. Hence |f (z)| satisfies a maximum
principle. On the other hand if |f (z)| is nowhere vanishing, then log |f (z)|2 is smooth function, and in
fact is harmonic as can be seen from the following computation
1 ∂2 1 ∂2
∆ log |f |2 = log |f (z)|2 = (log f (z) + log f (z)) = 0.
4 ∂z∂ z̄ 4 ∂z∂ z̄
Note that since f (z) is no-where vanishing at least locally near z one can define a holomorphic branch of
log. The upshot is that log |f (z)|2 must satisfy a minimum principle, and hence must |f (z)| (since log is
increasing).
19.3 Rouche’s theorem

Theorem 19.3.1. Let γ be a simple closed curve in Ω and f, g ∈ O(Ω) such that for all z ∈ Supp(γ),
|f (z) − g(z)| < |g(z)|.
Then f (z) and g(z) have the same number of zeroes in Int(γ).
Proof. Firstly, note that f (z) and g(z) have no zero on γ (the strictness of the inequality above is crucial
precisely for this purpose). Moreover, for all z ∈ Supp(γ), we have
f (z)
− 1 < 1.
g(z)
Put F (z) = f (z)/g(z). Then F (z) ∈ M(Ω). Moreover, at the points where f (z) and g(z) are non-zero
(in particular on Supp(γ)), one can easily see that
F ′ (z) f ′ (z) g ′ (z)
= − .
F (z) f (z) g(z)
A quick way to see this is that in the nieghbourhood of such points, log F (z) is well defined and holomor-
phic, and moreover, log F (z) = log f (z) − log g(z). Now consider Γ := F ◦ γ, then Γ is a close curve in
D1 (1). Since D1 (1) is simply connected, and 0 ∈ / D1 (1), n(Γ, 0) = 0. By the argument principle,
ˆ ˆ ′ ˆ ′
1 F ′ (z) 1 f (z) 1 g (z)
0 = n(Γ, 0) = dz = dz − dz.
2πi γ F (z) 2πi γ f (z) 2πi γ g(z)
Once again by argument principle, we see that f (z) and g(z) must have the same number of zeroes in
Int(γ).
Typically, as can be seen in the example below, the theorem is applied to count the number of zeroes of
f (z). The heart of the matter is to come up with a suitable g(z), whose zeroes can be counted easily, and
such that the above (strict) inequality can be satisfied.
Example 19.3.2. Consider the polynomial p(z) = z 4 − 6z + 3. We claim that all it’s roots are contained in
the disc D2 (0), and three of them are contained in the annulus A1,2 (0). We divide the proof into the following
two cases.
• The disc |z| < 2. On the circle |z| = 2 we have the following estimate
|p(z) − z 4 | = |6z − 3|
≤ 6|z| + 3 = 12 + 3 = 15 < |z|4 .
By Rouche’s theorem, p(z) has the same number of roots as z 4 in |z| < 2, and hence has four roots in
that disc. But p(z) is also a polynomial of degree four, and hence these four roots must include all the
possible roots of p(z).
19.4. APPENDIX : DETAILS LEFT OUT IN THE PROOF OF COROLLARY ?? 115
• The disc |z| < 1 + ε for ε << 1. Lets take ε < 1/10. On the circle |z| = 1 + ε we have the following
estimate
|p(z) − (−6z)| = |z 4 + 3|
≤ |z|4 + 3 = (1 + ε)4 + 3 < 4.5 < 6(1 + ε) = | − 6z|.
Once again by Rouche’s theorem, p(z) has exactly one root in |z| < 1 + ε, and hence has exactly three
roots in 1 + ε ≤ |z| < 2. Since this is true for all ε << 1, in particular, it has exactly three roots in
A1,2 (0).
Remark 19.3.3. Rouche’s theorem can be used to give another proof of the fundamental theorem of
algebra. Let p(z) = an z n + an−1 z n−1 + · · · + a0 be a general degree n polynomial (so an ̸= 0). It is easy
to see that for R >> 1, if |z| = R, then
|p(z) − an z n | < |an ||z n |.
This is essentially because p(z)−an z n is a polynomial of a strictly lower degree. Now by Rouche’s theorem
p(z) and an z n have the same number of roots on |z| < R. In particular, p(z) must have exactly n roots on
|z| < R. In fact it can be shown easily (by induction for instance) that it cannot have any further zeroes.
19.4 Appendix : details left out in the proof of Corollary 19.1.5

To spell out the details on the holomorphicity of f −1 and that the derivative is 1/f ′ (zw ), we first note that
ˆ
f −1 (w + h) − f −1 (w) 1 1 1
= zf ′ (z) − dz
h 2πih |z−z0 |=r f (z) − w − h f (z) − w
ˆ
1 zf ′ (z)
= dz
2πi |z−z0 |=r (f (z) − w − h)(f (z) − w)
Now the integrand is continuous and bounded for |h| << 1, and hence we can take compute the limit by
swapping the integral and the limit. That is,
ˆ
f −1 (w + h) − f −1 (w) 1 zf ′ (z)
lim = dz.
h→0 h 2πi |z−z0 |=r (f (z) − w)2
Another application of the residue theorem shows that the second integral is precisely 1/f ′ (zw ). To see
this, we observe that
zf ′ (z) (z − zw )f ′ (z) f ′ (z)
2
= 2
+ zw .
(f (z) − w) (f (z) − w) (f (z) − w)2
From the geometric series expansion, one can see that the second term is of the form
f ′ (z) zw
zw = ′ (z − zw )−2 + g(z),
(f (z) − w)2 f (zw )2
where g(z) is holomorphic near zw . This relies on the fact that f ′ (zw ) ̸= 0, and hence the numerator has
a non-zero constant term in it’s Taylor expansion. The upshot is that the second term does not contribute
to the residue. The advantage now is that the first term has a simple pole at z = zw , and hence we can
compute the residue as
zf ′ (z) (z − zw )2 f ′ (z) 1
Resz=zw 2
= lim = ′ .
(f (z) − w) z→zw (f (z) − w)2 f (zw )
Lecture 20
Contour Integration - I
In the next two lectures, we’ll systematically study applications of the residue theorem to the computation
of real variables integrals. We will follow the exposition in Ahlfor’s book [1] very closely. For instance the
classification of the integrals into Types is taken directly from [1].
20.1 Type-I: Integrals of rational functions

In this section we study integrals of the form
ˆ ∞
R(x) dx,
−∞
where R(x) is a rational function.

Assumption. R(x) = P (x)/Q(x) with deg(Q) ≥ deg(P ) + 2 and Q(x) has no real root. Recall that by
definition,
ˆ ∞ ˆ R2
R(x) dx = lim R(x) dx.
−∞ R1 ,R2 →∞ −R1
By the hypothesis on degree and no real root of Q(x), the integral is absolutely convergent and it follows
that in fact,
ˆ ∞ ˆ R
R(x) dx = lim R(x) dx.
−∞ R→∞ −R
The method. We consider the contour γR consisting of a semi-circle of radius R centred at the origin
and traversed in the anti-clockwise direction. We can decompose γR = lR + CR , where lR is the interval
(−R, R) and CR is the semi-circular part, and hence
ˆ R ˆ ˆ
R(x) dx = R(z) dz − R(z) dz.
−R γR CR
If R >> 1, then all roots of Q(x) in the upper half plane lie in the interior of γR . By the residue theorem,
ˆ X
R(z) dz = 2πi Resz=α R(z).
γR Q(α)=0
Im(α)>0
117
118 LECTURE 20. CONTOUR INTEGRATION - I
On the other hand, since deg Q ≥ deg P + 2, when R >> 1,

M
|R(z)| ≤
R2
for some M > 0. Hence, ˆ
2πM R→∞
R(z) dz ≤ −−−−→ 0.
CR R
Putting it all together we have
ˆ ∞ X
R(x) dx = 2πi Resz=α R(z).
−∞ Q(α)=0
Im(α)>0
Example 20.1.1. Consider the integral

ˆ ∞ ˆ
x2 1 ∞ x2 X
dx = dx = πi Resz=α R(z),
0 1 + x6 2 −∞ 1 + x6
1+α6 =0
Im(α)>0
where R(z) = z 2 /(1+z 6 ). The roots of the denominator are given by αk = eπi/6+2πik/6 for k = 0, 1, · · · , 5.
Of these only α0 = eπi/6 , α1 = eπi/2 = i and α3 = e5πi/6 are in the upper half region. Each of these is a
simple pole, and hence we can compute that

z 2
z − αk 1 −i/6, k = 0

2
Resz=αk R(z) = lim (z − αk ) = αk lim = = i/6, k = 1
z→αk 1 + z6 z→αk 1 + z 6 6αk3 
−i/6, k = 2.

Finally we get that ˆ ∞

x2 i π
dx = πi − = .
0 1 + x6 6 6
20.1.1 A variation
One also often encounters integrals (as in the above example) of the form
ˆ ∞
R(x) dx.
0
If R(x) is an even function as in the example above, then one can simply convert it into an integral on all
of R at the cost of an additional factor. But this trick will not work in general, so one might have to get
more creative.
20.2. TYPE-II: RATIONAL FUNCTIONS OF SINE AND COSINE 119

ˆ ∞ ˆ R
x x
I= dx = lim dx.
0 1 + x4 R→∞ 0 1 + x4
As usual, we let IR denote the integral on the right (inside the limit), and we let f (z) = z/(1 + z 4 ). We take
the contour γR = γ1,R + CR + γ2,R , where γ1,R consists of the straight line from (0.0) to (R, 0), CR is the
quadrant of the circle from (R, 0) to (0, R), and γ2,R is the straight line from (0, R) to (0, 0). By the residue
theorem, since the only pole of f (z) in the interior of γR is a simple pole at z = eiπ/4 , we have
ˆ
f (z) dz = 2πiResz=eiπ/4 f (z)
γR
(z − eiπ/4 )z
= 2πi lim
z→eiπ/4 1 + z4
2πieiπ/4 π
= = .
4e3iπ/4 2
Next, we observe that ˆ
f (z) dz = IR ,
γ1,R
and by the discussion above, ˆ

lim f (z) dz = 0.
R→∞ CR
′
To compute the integral over γ2,R we parametrize γ2,R (t) = −it, where t ∈ (−R, 0). Then γ2,R (t) = −i,
and so ˆ ˆ 0
−it
f (z) dz = (−idt) = IR .
γ2,R −R 1 + t4
Hence if R > 1, by the residue computation above,
π
2IR + CR = .
2
Taking the limit as R → ∞ we see that
π
I= .
4
Remark 20.1.3. This example was merely for illustration, since the integral can of course be computed
in a more elementary way by a change of variables x2 = u. The reader can challenge himself/herself with
the following: ˆ ∞
x3
.
0 1 + x5
20.2 Type-II: Rational functions of sine and cosine

The type 2 integrals are of the following kind:
ˆ 2π
R(cos θ, sinθ) dθ,
0
where R(x) is again a rational function.

Assumption. No pole for θ ∈ [0, 2π).
The method. Consider z = z(θ) = eiθ . Then

(
z+1/z
cos θ = z+z̄
2 = 2
z−1/z
sin θ = z−z̄
2i = 2i .
Moreover, dz = z ′ (θ)θ = ieiθ θ, and hence

dz
dθ = −i .
z
One can then transform the integral to
ˆ 2π ˆ z + 1/z z − 1/z dz
R(cos θ, sinθ) dθ = −i R , .
0 |z|=1 2 2i z

ˆ π ˆ
dθ 1 2π dθ
I := = , a > 1.
0 a + cos θ 2 0 a + cos θ
Using the above transformations we can re-write the second integral as
ˆ 2π ˆ ˆ
dθ 1 dz dz
= −i z+1/z z
= −2i 2 + 1 + 2az
,
0 a + cos θ |z|=1 a + |z|=1 z
2
and so our required integral is ˆ

dz
I = −i .
|z|=1 z2 + 2az + 1
√ √
Now the denominator has two real roots, α = −a + a2 − 1 and β = −a − a2 − 1. Since a > 1, it is easy
to see that |α| < 1 while |β| > 1. So
1 z−α 2π π
I = (−i)(2πi)Resz=α = 2π lim 2 = =√ .
z 2 + 1 + 2az z→α z + 2az + 1 α−β a2 − 1
Type-III: Products of rational functions and triganometric func-

tions
These are integrals of the kind
ˆ ∞ ˆ ∞
R(x) cos x dx, R(x) sin x, dx.
−∞ −∞
These can be combined into the analysis of one single integral

ˆ ∞
R(x)eix dx.
−∞
Assumptions. These come in two sub-types where the analysis substantially differs. As before, we let
R(x) = P (x)/Q(x).
• Type-III(a). We assume that deg Q ≥ deg P + 2, and no root of Q is real.
• Type-III(b). Here we assume that deg Q = deg P + 1, and no root of Q is real.
20.2. TYPE-II: RATIONAL FUNCTIONS OF SINE AND COSINE 121
The method. For Type-III(a), one proceeds exactly as in Type-I, and we do not spend additional time on
this. For Type-III(b), it is not clear at all if the integral converges. It certainly will not absolutely converge,
and hence we cannot use the semi-circle contour. But in many cases, the oscillating factor eix might make
the integral converge conditionally. By definition, if the integral converges, then
ˆ ∞ ˆ R2
R(x)eix dx = lim R(x)eix dx.
−∞ R1 ,R2 →∞ −R1
Now let Γ be the rectangle with vertices (−R1 , 0), (R2 , 0), (R2 , H) and (−R1 , H). with sides given by
straight-line curves γ1 , γ2 , γ3 and γ4 (see figure below). If R1 , R2 and H are big enough, the rectangle will
contain all the roots of Q with positive imaginary part. Hence

4 ˆ
X X
R(z) dz = 2πi Resz=α R(z)eiz .
i=1 γi Q(α)=0
Im(α)>0
The idea is to first fix R1 , R2 and let h → ∞. And then let R1 , R2 → ∞. We illustrate with an example.
ˆ ∞ ˆ ∞
x sin x 1 xeix
I= = dx,
−∞ 1 + x2 i −∞ 1 + x2
since x cos x/(1 + x2 ) is an odd function and hence the integral will be zero. We let f (z) = zeiz /(1 + z 2 ).
If H > 1, then the rectangle will contain the only pole of f (z) with positive imaginary part, namely z = i.
Hence ˆ
zeiz iei2 πi
f (z) dz = 2πiResz=i f (z) = 2πi lim (z − i) = 2πi = .
Γ z→i 1 + z2 i+i e
Now let us analyze each smooth curve in Γ.
• The curve γ1 . This is the integral we are interested in
ˆ ˆ R2
xeix
f (z) dz = dx.
γ1 −R1 1 + x2
• The curve γ2 . On γ2 , |z| > R2 , and hence if R2 >> 1, then
z 2
≤ .
1 + z2 R2
Then parametrizing the curve by γ2 (t) = R2 + it, and noting that |eiz | = e−t , we see that
ˆ ˆ H
2 2
f (z) dz ≤ e−t dt ≤ .
γ2 R2 0 R2
• The curve γ3 . Again if H >> 1, we will have the estimate

z 2
≤ .
1 + z2 H
Also |eiz | = e−H on γ3 . Hence |f (z)| ≤ 2e−H /H, and we have

ˆ
2e−H (R2 + R1 )
f (z) dz ≤ .
γ2 H
• The curve γ4 . As for the second curve, we once again have the bound
ˆ
2
f (z) dz ≤ .
γ4 R 1
Together, if we first let H → ∞, and then let R1 , R2 → ∞ the integrals on the last three curves converge to
0, and hence ˆ ∞ ˆ
xeix πi
2
dx = f (z) dz = .
−∞ 1 + x Γ e
Our original integral is then ˆ ∞
x sin x π
= .
−∞ 1 + x2 e
Lecture 21
Contour Integration - II
This is a continuation of our previous lecture on computing real variable integrals using the residue theo-
rem. In the present lecture, we’ll focus on integrations involving branch cuts. But first we refine a method
introduced in the previous lecture.
21.1 Type-III(b): Principal values

Now suppose that Q(x) has a simple zero on the real line. In this case even though the integral of R(x)eix
will not converge, the integrals of R(x) sin x or R(x) cos x might converge if the simple zero happens to
coincide with a zero of sin x and cos x respectively. To illustrate the work-around in this case, suppose we
wish to evaluate the integral
ˆ ∞
R(x) sin x.
−∞
Assumption. deg Q(x) = deg P (x) + 1 and Q(x) has a simple zero at x = 0.
The method. Consider the contour Γ in the figure below. Using the analysis from the final section of the
previous lecture, we can see that

ˆ −ε ˆ ∞ ˆ X
R(x)eix dx + R(x)eix dx − R(z)eiz dz = 2πi Resz=α R(z)eiz .
−∞ ε Cε Q(α)=0
Im(α)>0
123
124 LECTURE 21. CONTOUR INTEGRATION - II
To compute the third integral, note that near z = 0,
A
R(z)eiz = + R0 (z),
z
where R0 (z) is holomorphic near z = 0 (and hence bounded) and A = Resz=0 R(z)eiz . Then by an explicit
calculation it is easy to see that ˆ
lim R(z)eiz dz = πiA,
ε→0+ Cε
and hence
ˆ −ε ˆ ∞ 1 X
lim R(x)eix dx + R(x)eix dx = 2πi Resz=0 R(z)eiz + Resz=α R(z)eiz .
ε→0+ −∞ ε 2
Q(α)=0
Im(α)>0
(21.1)
The principal value of the integral of R(x)eix on R is defined to be
ˆ ∞ ˆ −ε ˆ ∞
ix ix
p.v. R(x)e dx := lim R(x)e dx + R(x)eix dx ,
−∞ ε→0+ −∞ ε
if the limit exists. Note that if Q(x) has a simple pole at x = 0, then the above limit will exist, as the
analysis above shows. Moreover, under these assumptions, clearly R(z) sin z is integrable near zero and
infinity, and so ˆ ∞ ˆ ∞
R(x) sin x dx = Im p.v. R(x)eix dx ,
−∞ −∞
and the latter principal value can be computed by the analysis above. Let us illustrate this via a famous
integral.
Example 21.1.1. Consider the integral of
ˆ ∞
sin x
I= dx.
−∞ x
Then by the above analysis,

eiz
I = Im πiResz=0 ,
z
and so ˆ ∞
sin x
dx = π.
−∞ x
Remark 21.1.2. This is the standard example of a function whose improper integral is finite, and yet
the Lebesgue integral does not converge. I cannot resist the temptation to include another method of
computing the integral, using the so called “Feynman technique". This is essentially just differentiation
under the integral sign, but was popularised by Feynman as his way of handling integrals that others
needed residue calculus for! Let I be the integral above. We introduce a parameter ‘a > 0’ and consider
the the family of integrals ˆ ∞
sin(x)e−ax
I(a) = dx.
0 x
One can check that I(a) is a differentiable function of a, and that differentiation under the integral works.
Then, ˆ ∞
I ′ (a) = − e−ax sin x dx.
0
21.2. TYPE-IV: PRODUCTS OF RATIONAL FUNCTIONS AND POWERS OF X. 125
We can compute the integral on the right, by applying integration by parts twice. Indeed
ˆ ∞ ˆ
1 ∞
J(a) := e−ax sin x dx = − sin x de−ax
a 0
0
ˆ
1 ∞ −ax
= e d sin x
a 0
ˆ ∞
1
= e−ax cos x dx
a 0
ˆ ∞
1
=− 2 cos x de−ax
a 0
ˆ ∞
1
= 2+ e−ax d cos x
a 0
1 J(a)
= 2− 2 .
a a
Solving for J(a) we get that
1
I ′ (a) = − ,
1 + a2
and hence
I(a) = − arctan(a) + C.
Since lima→∞ I(a) = 0, clearly C = π/2. But then taking a → 0+ we see that
π
I = lim+ I(a) = C = .
a→0 2
21.2 Type-IV: Products of rational functions and powers of x.

In this section we study integrals of the form
ˆ ∞
xα R(x) dx,
−∞
where R(x) = P (x)/Q(x) is a rational function and α ∈ (0, 1).

Assumption. deg Q ≥ deg P + 2, and Q(x) has a simple zero at x = 0 and no other real zero.
The method. Note that the assumption implies that the integral is absolutely convergent, and so
ˆ ∞ ˆ R
α
x R(x) dx = lim xα R(x) dx.
−∞ R→∞ −R
Consider the contour in the figure below. We integrate the function z α R(z) on the contour. Since we are
dealing with fractional powers, we have to make a choice of a branch cut and a a corresponding branch of
the power. Given the geometry of the contour, it is clear that we have to use the branch cut (0, ∞). Recall
that z α = eα log z , where log z = log |z| + i arg(z), and arg(z) ∈ (0, 2π).
We first have the following observations:
Lemma 21.2.1. With the orientations as in the figure,
ˆ ˆ R
lim z α R(z) dz = xα R(x) dx,
δ→0 γ+,δ ε
ˆ ˆ R
lim z α R(z) dz = e2πiα xα R(x) dx.
δ→0 γ−,δ ε
The proof relies on the fact that continuous functions on compact sets are uniformly continuous. For a
fixed ε and R, and for small δ > 0, the function z α R(z) is continuous, and hence uniformly continuous,
on the rectangle with vertices (−δ, ε), (−δ, R), (δ, R) and (δ, ε). The e2πiα factor in the second integral is
due to the fact that there is a jump of e2πiα in the value of z α across the branch cut z > 0. We leave the
details to the reader.
By the residue theorem,
ˆ ˆ ˆ ˆ X
α α α
z R(z) dz + z R(z) dz − z R(z) dz − z α R(z) dz = 2πi Resz=β z α R(z).
γ+,δ CR γ−,δ Cε Q(β)=0
β̸=0
Letting δ → 0, by the Lemma above,

ˆ R X ˆ ˆ
(1 − e2πiα ) xα R(x) dx = 2πi Resz=β z α R(z) − z α R(z) dz + z α R(z) dz.
ε Q(β)=0 CR Cε
β̸=0
Finally, letting ε → 0+ and R → ∞ one proves that the integrals on the right converge to zero under the
given assumptions. We illustrate with an example.
ˆ ∞
x−a
I := dx, a ∈ (0, 1).
0 1+x
To get it into the above form, we re-write this as
ˆ ∞
x1−a
I= dx,
0 x(1 + x)
and let Iε,R be the corresponding integral from ε to R. Let f (z) = z 1−a /(z(1 + z)). Recall that we are using
the branch z 1−a = e(1−a) log z , where log(reiθ ) = log r + iθ and θ ∈ (0, 2π). Now f (z) has simple poles at
z = 0 and z = −1. By the above discussion,
ˆ ˆ
z −a z 1−a z 1−a
(1 − e2πi(1−a) )Iε,R = 2πiResz=−1 − dz + dz (21.2)
1+z CR z(1 + z) Cε z(1 + z)
21.3. A BONUS INTEGRAL 127
We first estimate the two remaining integrals on the right.

• The integral on CR . When |z| = R >> 1, we have |1 + z| ≥ |z| − 1 > R/2, and so
z −a R−a 2R−a 2
= ≤ = 1+a .
1+z |1 + z| R R
The integral then satisfies
ˆ
z 1−a 4πR 4π R→∞
dz ≤ 1+a = a −−−−→ 0.
CR z(1 + z) R R
• The integral on Cε . When |z| = ε << 1, we have |1 + z| ≥ 1 − |z| > 1/2, and so
z −a ε−a
= ≤ 2ε−a .
1+z |1 + z|
The integral then satisfies ˆ
z 1−a ε→0+
dz ≤ 4πε1−a −−−−→ 0.
Cε z(1 + z)
Finally we compute the residue at z = −1,

Resz=−1 f (z) = lim z −a = e−a log(−1) = e−iπa .
z→−1
Putting all of this together, taking ε → 0+ and R → ∞ in (21.2) we see that

2πie−iπa 2πie−iπa 2πi π
I= 2πi(1−a)
= −2πia)
= πia −πia)
= .
(1 − e ) (1 − e ) (e − e ) sin πa
21.3 A bonus integral

As a final integral, let us compute ˆ ∞
(log x)2
I := dx.
0 1 + x2
We denote the corresponding integral over (ε, R) by Iε,R . As an exercise, the reader should attempt to
use the contour above too evaluate this integral. We will instead use a semicircular contour Γ in the
figure below. We use the branch cut {iy | y ∈ (−∞, 0)}, and the branch of the logarithm defined by
log(reiθ ) = log r + iθ, with θ ∈ (−π/2, 3π/2). Let

(log z)2
f (z) = .
1 + z2
Then clearly,
ˆ ˆ R
(log z)2 (log x)2
= dx = Iε,R ,
γ1 1 + z2 ε 1 + x2
ˆ 2 ˆ −ε
(log z) (log |t| + iπ)2
= dt.
γ2 1 + z2 −R 1 + t2
The limits follow from the fact that the integral is absolutely convergent. For the second integral above we
parametrize γ2 (x) = t = |t|eiπ with t ∈ (−R, −ε). Putting x = −t int he second integral, we see that
ˆ ˆ −ε
(log z)2 (log |t| + iπ)2
2
= dx
γ2 1 + z −R 1 + t2
ˆ R
(log x + iπ)2
= dx
ε 1 + x2
ˆ R ˆ R
log x 2 dx
= Iε,R + 2πi 2
dx − π 2
.
ε 1+x ε 1+x
Notice that the third integral is arctan(R/ε), and so letting ε → 0+ and R → ∞,

ˆ (log z)2 ˆ (log z)2 π3
ˆ R
log x
lim 2
+ 2
= 2I − + 2πi dx.
ε→0+ γ1 1 + z γ1 1 + z 2 ε 1 + x2
R→∞
By the exact analysis as the previous section, one can prove that
ˆ ˆ
(log z)2 (log z)2
lim dz = lim dz = 0,
R→∞ C
R
1 + z2 ε→0+ Cε 1 + z 2
and so by the residue theorem,

ˆ R
π3 log x (log z)2
2I − + 2πi 2
dx = 2πiResz=i (21.3)
2 ε 1+x 1 + z2
(log z)2
= 2πi lim (z − i)
z→i 1 + z2
2
(log i)
= 2πi
2i
π3
=− .
4
Note that in the penultimate line, we used the fact that for our chosen branch of the logarithm, we have
log i = iπ/2. Equating the real parts, and solving for I, we get that
ˆ ∞
(log x)2 π3
dx = .
0 1 + x2 8
Remark 21.3.1. Notice that since 2πi times the residue above was completely real, the imaginary part in
equation (21.3) above has to be zero. This yields a curious integral identity
ˆ ∞
log x
= 0.
0 1 + x2
A simple change of variables yields an explanation. Namely, denoting the integral by J, if we let t = 1/x,
we have ˆ 0 2
−t log t −dt
J= 2
= −J,
∞ 1+t t2
and hence J = 0.
Part III
Geometry of holomorphic maps
129
Lecture 22
Conformality
In this lecture, we will begin our study of some geometric properties of holomorphic maps.
22.1 Conformal maps

Loosely speaking, a conformal map is a map that preserves angles. We now try to make this concept more
rigorous, and give a precise definition of conformal maps. Throughout this introductory part, we’ll work
with either R2 or C, as is convenient, always remembering the natural identification. Let f : Ω → R2 be
a C 1 map ie. the partial derivatives ∂f /∂x, ∂f /∂y exist and are continuous. Given two curves γ1 (t) =
(x1 (t), y1 (t)) : (−ε, ε) → C and γ2 (t) = (x2 (t), y2 (t)) : (−ε, ε) → C with γ1 (0) = γ2 (0) = p, we define
the angle between them as
∡γ1 (0), γ2 (0)
Recall that if v⃗1 and v⃗2 are non-zero vectors in R2 , then the angle between them is defined to be
⟨v⃗ , v⃗ ⟩
1 2
∡v⃗1 , v⃗2 = arccos ,
|v⃗1 ||v⃗2 |
where ⟨·, ·⟩ is the usual dot product on R2 and | · | is the usual norm (given by the square-root of the dot-
product of the vector with itself). Motivate by this, we say that a linear map T : R2 → R2 preserves angles
⃗ ∈ R2 \ {⃗0},
or is conformal if det(T ) > 0 1 , and for any pair of non-zero vectors ⃗v , w
⟨⃗v , w⟩
⃗ ⟨T ⃗v , T w⟩
⃗
= .
|⃗v ||w|
⃗ |T ⃗v ||T w|
⃗
More generally, we sat that a C 1 mapping f : Ω ⊂ R2 → R2 is conformal at (x0 , y0 ) ∈ Ω if the total

derivative D(x0 ,y0 ) f : R2 → R2 is conformal. We say that f is conformal if it is conformal at all points in
Ω. Recall that in the standard basis the matrix representing D(x0 ,y0 ) f is given by the Jacobian matrix
!
∂u ∂u
∂x (x0 , y0 ) ∂y (x0 , y0 )
D(x0 ,y0 ) f = ∂v ∂v ,
∂x (x0 , y0 ) ∂y (x0 , y0 )
where (u, v) are the components of f ie. in complex notation f = u + iv.

1 This condition is equivalent to the map being orientation preserving. Some authors choose to not impose this extra condition
on conformal maps
131
132 LECTURE 22. CONFORMALITY
A slightly more geometric insight is obtained by looking at curves. Consider a pair of curves γ1 (t), γ2 (t) :
(−ε, ε) → C intersecting at γ1 (0) = γ2 (0) = z0 . We say that they intersect at an angle θ at z0 if the
angle between the tangent vectors γ1′ (0) and γ2′ (0) is θ. The a mapping is conformal at z0 if and only if
for any two curves γ1 and γ2 as above, the angle between them is equal to the angle between their images
f (γ1 (t)) and f (γ2 (t)) at f (z0 ). To see this, it is enough to note that if γ(t) : (−ε, ε) → Ω is a curve with
γ(0) = z0 = (x0 , y0 ) and γ ′ (0) = ⃗v , then
d
D(x0 .y0 ) f (⃗v ) = f (γ(t)).
dt t=0
We are now ready to prove our main observation.

Proposition 22.1.1. Let f : Ω → C be a C 1 map. Then f is conformal at z0 if and only if f is complex
differentiable at z0 and f ′ (z0 ) ̸= 0. In particular, f is conformal if and only if it is holomorphic with nowhere
vanishing complex derivative. 2 .
We first need two elementary lemmas from linear algebra.
Lemma 22.1.2. A linear map C : R2 → R2 is conformal if and only if C = λQ, where Q is an orientation
preserving orthogonal transformation and λ > 0.
Recall that Q is orthogonal if and only if |Q⃗v | = |⃗v | for all ⃗v or equivalently QT Q = I, where QT is the
transpose of Q, and I is the identity matrix.
Proof. Let e⃗1 and e⃗2 denote the standard basis vectors in the x and y directions, and let C T C = (aij ) be
the matrix representation of C T C in this basis. By conformality, Then
0 = ⟨e⃗1 , e⃗2 ⟩ =⇒ 0 = ⟨C e⃗1 , C e⃗2 ⟩ = e⃗1 T (C T C)e⃗2 = a12 .
By the symmetry of C T C, we also have that a21 = 0, and hence C T C is diagonal. On the other hand,
0 = ⟨e⃗1 − e⃗2 , e⃗1 + e⃗2 ⟩ =⇒ 0 = ⟨C(e⃗1 − e⃗2 ), C(e⃗1 + e⃗2 )⟩

= ⟨C e⃗1 , C e⃗1 ⟩ − ⟨C e⃗2 , C e⃗2 ⟩
= a11 − a22 .
That is, the diagonal terms in C T C are equal, and hence C T C = µI for some µ ∈ R. Since det C > 0 and
the diagonal terms of C T C have to be non-negative, we have that µ > 0 Now, let Q = µ−1/2 C. Then
C T C = µI =⇒ QT Q = I,
√
and this proves the claim with λ = µ. The converse, that a matrix C = λQ is conformal if Q is orthogonal
is trivial.
Lemma 22.1.3. Any orientation preserving orthogonal 2 × 2 matrix Q is a rotation matrix, that is, it is given
by
cos θ − sin θ
Q = Rθ := ,
sin θ cos θ
for some θ ∈ [0, 2π).
Proof. Since QT Q = I and det Q > 0, we have that det Q = 1. Suppose

a b
Q= .
c d
2 If we drop orientation preserving from the definition of conformality, then a conformal map is either a holomorphic or anti-
holomorphic map
22.2. A SURVEY OF ELEMENTARY MAPPINGS 133
Then the combination of QT Q = I and det Q = 1 gives us the three equations


2 2
a + c = 1

b2 + d 2 = 1

ad − bc = 1.

In particular, (a − d)2 + (b + c)2 = 0, and hence a = d and c = −b. Since a2 + c2 = 1, we can set a = cos θ
and c = sin θ. The result then follows.
Proof. • Suppose f is conformal at z0 . Then C := Dz0 f is a conformal linear map, and by the above
two lemmas, C = λRθ for some θ. In particular, the partial derivatives of f satisfy the Cauchy-
Riemann equations, and since f is a C 1 map (in particular the total derivative exists), this implies
that f is complex differentiable at z0 . Moreover, 0 < det C = |f ′ (z0 )|2 , and hence f ′ (z0 ) ̸= 0.
• Conversely, suppose f is complex differentiable at z0 and f ′ (z0 ) ̸= 0. Then C := Dz0 f satisfies
det C = |f ′ (z0 )|2 > 0. Moreover, it can be checked by direct calculation (using the fact that the
Cauchy-Riemann equations are satisfied) that Q = |f ′ (z0 )|−1 C is orthogonal. Then by Lemma
22.1.2, C is a conformal linear map, and hence by definition, f is a conformal map.
Corollary 22.1.4. Any holomorphic, injective function is conformal.
Proof. Let f ∈ O(Ω) and injective, but not conformal. By Proposition 22.1.1 there exists a point z0 ∈ Ω
such that f ′ (z0 ) ̸= 0. If w0 = f (z0 ), then the equation f (z) = w0 has a root of multiplicity m > 1 at
z = z0 . By our fundamental theorem on the local mapping properties of holomorphic functions (Theorem
1 in Lecture-19), there exists ε, δ > 0 such that for each w such that 0 < |w−w0 | < δ, there exists at least m
distinct points in Bε (z0 ) such that f (z) = w. In particular f is not injective, which is a contradiction.
Remark 22.1.5. Of course, there are plenty of conformal maps which are not injective. For example
f (z) = z n on C∗ is conformal, but not injective if n > 1. Other examples include ez , sin z, cos z etc.
22.2 A survey of elementary mappings

Given two regions Ω1 and Ω2 , we wish to construct a conformal map one to the other. A reasonable
strategy is to first try to map Ω1 into the unit disc, and then to map the unit disc into Ω2 . A priori, t his
strategy might seem limiting. After all why would there exist such a conformal map from Ω1 into the unit
disc. In a couple of lectures time, we’ll prove a deep fact - The Riemann mapping theorem - that for a large
class of regions, namely simply connected strict subsets of C, such a mapping always exists. In view of
this it is important to build up a toolkit of familiar mappings, so that more complicated mapping can be
constructed by taking products, compositions etc. So in this section we’ll familiarise ourselves with the
mapping properties of complex powers, exponentials and logarithms. A good strategy in finding the image
of a certain region under a conformal mapping is to find the image of the boundary. Our convention is to
use z = x + iy as the complex coordinate in the domain, and w = u + iv as the complex coordinate on
the image.
22.2.1 Rotations and dilations

Clearly rotations Rθ (z) = eiθ z, dilations and Tλ (z) = λz are examples of conformal maps. In fact these
form a subgroup of the full group of Mobius transformations.
22.2.2 Complex powers

Complex powers are useful in mapping sectors and half planes to each each other. We illustrate this using
two examples.
• Case-1: f (z) = z n , n ∈ N. An example of such a mapping is in Figure 22.1
Figure 22.1: integer powers turn a sector into a half plane
• Case-1: f (z) = z n , n ∈ N. An example of such a mapping is in Figure 22.2
Figure 22.2: fractional powers can turn a half plane into a sector
22.2.3 The logarithm

If L(z) = log z is a branch of the logarithm, then L′ (z) = 1/z ̸= 0 on it’s domain of definition. Hence
it defines a conformal map. As an illustration, in the Figures 22.3 and 22.4, we consider the branch of the
logarithm
log z = log |z| + i arg(z),
where arg(z) ∈ (−π/2, 3π/2). That is, our branch cut is the line {z = iy | y < 0} or the −ve y axis.
In the Figure 22.3, the boundary of the domain H contained in the domain of definition of log z consists
of two components, namely the negative and positive x (or real) axis. The negative axis consists of points
with arg(z) = π, and hence is mapped via the log to the line v = π. Similarly the positive x-axis has
arg(z) = 0, and hence is mapped to v = 0. Hence H is mapped by the above branch of logarithm to the
infinite strip {0 < v < π}. The reader should similarly work out the mapping in Figure 22.4.
22.2.4 The exponential map

The map f (z) = ez is clearly a conformal map by Proposition 22.1.1 since it’s derivative never vanishes.
Figures 22.5 and 22.6 illustrate some of the mapping properties of the exponential. The reader should try
to work out why the images of the two mappings below are given by the figures on the left. As in the
discussion above, the trick is to work out the images of the various boundary components.
22.2. A SURVEY OF ELEMENTARY MAPPINGS 135
Figure 22.3: The logarithm mapping the H into an infinite horizontal strip
Figure 22.4: The logarithm mapping a semi-circular region into a half infinite strip
Figure 22.5: The exponential mapping a half infinite strip to a semi-circular region.
Figure 22.6: eiz mapping a half infinite strip to a semi-circle.

Lecture 23
Mobius transformations
23.1 Mobius transformations

A fractional linear transformation or a Mobius transformation is a map of the form
az + b
w = T (z) =
cz + d
with ad−bc ̸= 0. Clearly if each of a, b, c and d are scaled by the same complex number, then T (z) remains
invariant. Hence it is often convenient to normalize so that ad−bc = 1. The set of Mobius transformations
is denoted by Mob(C). Note that the map is defined and holomorphic for all z except z = −d/c. Moreover,
a
lim T (z) = .
z→∞ c
In view of this, it is sometimes more convenient to define T (−d/c) = ∞ and T (∞) = a/c and think of
T (z) as a map from the extended complex plane Ĉ := C ∪ ∞ to itself. We say that a map T : C → Ĉ is
holomorphic, and write T ∈ O(C) if the following three conditions hold:
1. If R is a meromorphic function and

C
2. F (z) = T (1/z) is holomorphic in a neighbourhood of z = 0.

Lemma 23.1.1. O(Ĉ) can be identified with the set of rational functions on C.
This is essentially Theorem 0.2 in Lecture 16.
Theorem 23.1.2. 1. A Mobius transformation T is a holomorphic, bijective map from Ĉ onto Ĉ, and it’s
inverse is also a Mobius transformation.
2. In particular if T (z0 ) = ∞ for some z0 ∈ C, then T → C \ C is a conformal map.

C\{z0 }
3. Moreover, if T and S are Mobius transformations, then so is S ◦ T . In other words, Mob(C) forms a
group under the law of composition with the identity element given by the transformation I(z) = z.
Proof. Since T is a rational function, it is clearly in O(C). The second point follows from the first easily.
Let T (z) be as above.
• T is injective. Suppose T (z1 ) = T (z2 ) and neither of z1 or z2 is infinity or −d/c. Then rearranging,
it is easy to see that (ad − bc)z1 = (ad − bc)z2 , and since ad − bc ̸= 0, we have z1 = z2 . Now,
137
138 LECTURE 23. MOBIUS TRANSFORMATIONS
suppose z1 = ∞ then T (z1 ) = ac = T (z2 ). This forces z2 to be infinity and hence z1 = z2 . On the
other hand, if z1 = −d/c, then T (z1 ) = ∞ = T (z2 ). Again, this means that z2 = −d/c = z1 .
• T is surjective. To prove this, we can simply solve the equation w = T (z). That is, w = T (z), if
and only
dw − b
z= .
a − wc
Combining this with the injectivity, we then have a well defined map T −1 : Ĉ → Ĉ defined by
dw − b
T −1 (w) =
a − wc
which is again a Mobius transformation.
• S ◦ T is a Mobius transformation. Suppose T is as above, and S is another Mobius transformation
pw + q
S(w) = .
rw + s
Then a simply computation gives
(pa + cq)z + pb + qd
S ◦ T (z) = .
(ra + cs)z + rb + ds
Moreover,
(pa + cq)(rb + ds) − (ra + cs)(pb + qd) = (ad − bc)(pr − qs) ̸= 0,
and so S ◦ T is again a Mobius transformation.
23.1.1 The group P SL(2, C).

For a Mobius transformation, if we write the coefficients as a matrix, we get what we call the coefficient
matrix
a b
M (T ) = .
c d
Since ad − bc ̸= 0, the matrix M (T ) is an invertible matrix, that is M (T ) ∈ GL(2, C), the group of all
invertible 2 × 2 complex valued matrices. If we normalize so that ad − bc = 1, then M (T ) ∈ SL(2, C), the
so-called special linear group of 2 × 2 complex valued matrices. Even with this normalization, a particular
Mobius transformation actually corresponds to 2 matrices, namely M (T ) and −M (T ), and hence a Mobius
transformation actually corresponds to an equivalence class of matrices. To make this more precise we
define the projective special linear group as
P SL(2, C) := SL(2, C)/A ∼ ±A.
We denote any element of P SL(2, C) as [A] where A is a matrix in SL(2, C). One can prove that
P SL(2, C) forms a group with the multiplication
[A] · [B] := [AB],
where AB is the usual matrix multiplication. One has to of course check that if you pick different represen-
tatives in [A] (ie. -A instead of A) and/or [B], then the multiplication gives the same element in P SL(2, C).
With this definition in place, we then obtain a map Φ : M ob(C) → P SL(2, C), given by
Φ(T ) = [M (T )].
23.1. MOBIUS TRANSFORMATIONS 139
By Theorem 23.1.2, part(3), it is clear that
[M (S ◦ T )] = [M (S)M (T )].
where the multiplication on the right is simply the usual matrix multiplication. In fancier language, this
says that the map Φ is a group homomorphism. In fact we have the following.
Theorem 23.1.3. The map Φ is an isomorphism between the groups M ob(C) and P SL(2, C).
Remark 23.1.4. One can also define P GL(2, C) as GL(2, C)/ ∼ where A ∼ λA where λ ∈ C∗ . It is then
easy to see that P GL(2, C) is isomorphic as a group to P SL(2, C).
Remark 23.1.5. A convenient way to represent Mobius transformations is using homogenous coordinates,
and this makes the role of P SL(2, C) much more transparent. The extended complex plane C can be
identified with the set P1 of complex lines passing through the origin in C2 . The identification is given by
the complex slope. A line L in C2 passing through the origin is determined by a point (ξ1 , ξ2 ) ̸= (0, 0), and
any other point on the line is given by (tξ1 , tξ2 ) for t ∈ C. Hence we represent points in P1 as equivalence
classes of these points [ξ1 : ξ2 ]. The complex slope is then given by z = ξ1 /ξ2 , and is a well defined number
in C. For instance the points [0 : 1] and [1 : 0] correspond to the points 0 and ∞ respectively in Ĉ. We say
that [ξ1 : ξ2 ] are the homogenous coordinates of z. Note that homogenous coordinates are unique, only
up to scaling ie. both (ξ1 , ξ2 ) and tξ1 , tξ2 ) for t ̸= 0, represent the same point z in C.
With this identification, if w = ζ1 /ζ2 and z = ξ1 /ξ2 , we can rewrite w = T z as

ζ1 a b ξ1
= .
ζ2 c d ξ2
So the action of the Mobius transformation on Ĉ is exactly the linear action of a 2 × 2 matrix on C2 \
{(0, 0)} = {set of homogenous coordinates}.
23.1.2 The cross ratio

Given any four numbers z1 , z2 , z3 , z4 in the extended complex plane Ĉ, the cross ratio is defined to be
z3 − z1 z4 − z2
(z1 , z2 , z3 , z4 ) = · .
z3 − z2 z4 − z1
Note that if one of the points is infinity, then the cross ratio is defined by taking a limit. For instance, if
z1 = ∞, then
z4 − z2
(∞, z2 , z3 , z4 ) = .
z3 − z2
The importance of the cross ratio comes form the following theorem.
Theorem 23.1.6. 1. Given any three points z2 , z3 , z4 , there exists a unique Mobius transformation map-
ping these points to 1, 0 and ∞ respectively. In fact we can take
S(z) = (z, z2 , z3 , z4 ).
2. If T is any Mobius transformation, then
(T z1 , T z2 , T z3 , T z4 ) = (z1 , z2 , z3 , z4 ).
Proof. 1. Clearly, S(z) = (z, z2 , z3 , z4 ) is a Mobius transformation that takes (z2 , z3 , z4 ) to (1, 0, ∞).
Let T be another such Mobius transformation. Then S ◦ T −1 takes (1, 0, ∞) to itself. Then it is not
hard to prove that S ◦ T −1 (z) = z, and hence S(z) = T (z).
2. Let S(z) = (z, z2 , z3 , z4 ). Then ST −1 carries (T z2 , T z3 , T z4 ) to (1, 0, ∞). Then by part (1),
ST −1 (z) = (z, T z2 , T z3 , T z4 ). Applying this to T z1 , we see that
(z1 , z2 , z3 , z4 ) = Sz1 = ST −1 T z1 = (T z1 , T z2 , T z3 , T z4 ).
As a consequence we have the following.

Corollary 23.1.7. Given any pair of three points (z1 , z2 , z3 ) and (w1 , w2 , w3 ), there exists a unique Mobius
transformation that takes the first triple to the second.
Proof. Let S and T be the Mobius transformations that take (z1 , z2 , z3 ) and (w1 , w2 , w3 ) to (1, 0, ∞) re-
spectively. Then T ◦ S −1 is the Mobius transformation that we need. Uniqueness follows from the fact
that there is unique transofrmation (namely the identity) that takes (1, 0, ∞) to itself.
There is another, more geometric, application of the cross ratio. A generalised circle in C is either a circle
(given by an equation |z − z0 | = r or a straight line (given by the equation āz + az̄ + b = 0, where b ∈ R).
In principle, a straight line is being thought of as a circle with infinite radius. An additional justification for
this terminology is that both circles and straight lines in C correspond to circles on the Riemann sphere via
the stereographic projection (or rather via it’s inverse). A key observation is that three points determine a
unique generalised circle (in the case of a line, one of the points will be at infinity).
Theorem 23.1.8. The cross ratio of of (z1 , z2 , z3 , z4 ) is real if and only if the four points lie on a generalised
circle. Consequently, a Mobius transformation maps generalised circles to generalised circles.
Proof. The second part follows from the first part, and the fact that the cross ratio is invariant under Mobius
transformations. So we focus on proving the first part. The key is the following claim.
Claim. If T is a Mobius transformation, then T −1 maps the (extended) real axis R̂ to a generalised circle.
Assuming this we complex the proof. There are two directions.
• =⇒ . Suppose the cross ratio (z1 , z2 , z3 , z4 ) is real. Consider T z = (z, z2 , z3 , z4 ). Then T z1 ∈ R.
Moreover, (T z2 , T z3 , T z4 ) = (1, 0, ∞) and hence they already lie on the real line. Since T −1 maps
R to a generalised circle, z1 , z2 , z3 and z4 lie on a generalised circle.
• ⇐= . Suppose z1 , z2 , z3 , z4 lie on a generalised circle. Once again consider T z = (z, z2 , z3 , z4 ).
Once again, (T z2 , T z3 , T z4 ) = (1, 0, ∞) and hence they already lie on the real line. So z2 , z3 , z4 lie
on the generalised circle T −1 (R̂). Hence z1 also lies on T −1 (R̂), and so T z1 ∈ R. In particular the
cross ratio is real.
Proof of the claim: Let z = T −1 (w). If w is real, then T z = T z. If T = az+b
cz+d , then this condition
translates to
az + b āz̄ + b̄
= .
cz + d c̄z̄ + d¯
Cross multiplying and simplifying, we get
(ac̄ − cā)|z|2 + (ad¯ − cb̄)z + (bc̄ − dā)z̄ + bd¯ − db̄ = 0.
There are two cases:

• Case-1: ac̄ − cā = 0. Then since ad − bc ̸= 0 we have that ad¯ − cb̄ ̸= 0. Hence the equation above
represents a straight line.
23.1. MOBIUS TRANSFORMATIONS 141
• Case-2: ac̄ − cā ≠= 0. Completing the square, we can rewrite the equation as
ād − c̄b ad − bc
z+ = ,
āc − c̄a āc − c̄a
which clearly defines a circle.
23.1.3 The Cayley transform

Perhaps the most important Mobius transformation is the the Cayley transform:
z−i
β(z) = .
z+i
To compute it’s inverse, we set w = β(z). and solve for z. It is easy to see that
1+w
β −1 (w) = i .
1−w
Figure 23.1: The Cayley transform
Lemma 23.1.9. The Cayley transform is a biholomorphism from the upper half plane H onto the unit disc
D. Moreover, it maps the boundary ∂H bijectively to ∂D.
Proof. • β(z) maps H into D and ∂H to ∂D. To see this, we compute
1 + |z|2 + i(z − z̄) 1 + |z|2 − 2Im(z)

|β(z)|2 = = <1
1 + |z|2 − i(z − z̄) 1 + |z|2 + 2Im(z)
if Im(z) > 0 i.e. if z ∈ H. On the other hand it is also clear from the computation that |β(z)| = 1 if
and only if Im(z) = 0.
• β(z) is surjective from H onto D. It is easy to see by direct computation that β −1 (w) given by
the above formula is an inverse. It is a nice exercise to check that indeed β −1 (w) is in the upper half
plane if w ∈ D.
Lecture 24
Some automorphism groups
In this lecture, we compute the automorphism groups of the disc and the complex plane.
24.1 Biholomorphisms and automorphisms

A holomorphic function f : Ω → Ω′ is said to be a bi-holomorphism if it is a bijective function, and f −1 is
also holomorphic.
Lemma 24.1.1. A holomorphic function f : Ω → Ω′ is a bi-holomorphism if and only if it is bijective.
Proof. If f is a bi-holomorphism, then it is automatically bijective from the definition. Conversely suppose
f : Ω → Ω′ is a bijective, holomorphic function. Injectivity implies that f ′ (z) ̸= 0 for all z ∈ Ω. Then the
holomorphic inverse function theorem implies that the inverse is holomorphic.
For any domain Ω ⊂ C, a bi-holomorphic function f : Ω → Ω is called an automorphism of Ω.

Lemma 24.1.2. The set of automorphisms of a domain Ω
Aut(Ω) := {f : Ω → Ω | f is holomorphic and bijective},
forms a group under the law of composition.

We call Aut(Ω) the automorphism group of Ω. Note that if Ω and Ω′ are bi-holomorphic, then Aut(Ω)
and Aut(Ω′ ) are isomorphic as groups. In fact, if φ : Ω → Ω′ is a biholomorphism, then Φ : Aut(Ω) →
Aut(Ω′ ) defined by
Φ(f ) = φ ◦ f ◦ φ−1 ,
is the required group isomorphism.
The aim of this note is to compute the automorphism groups of the complex plane, the punctured plane
and the disc (and hence the upper half plane).
24.2 Automorphism group of the disc

In this section, we compute Aut(D).Recall that in Problem-5 from Assignment-1, you were asked to prove
that for any |α| < 1,
α−z
ψα (z) :=
1 − ᾱz
143
144 LECTURE 24. SOME AUTOMORPHISM GROUPS
is a biholomorphism from the disc to itself. Our main theorem is that upto rotation these are the only au-
tomorphisms. En route, we’ll also provide a short proof that ψα is indeed an element in the automorphism
group, using some of the tools we have developed over the past few months, but which were of course
unavailable when you were asked to solve the problem!
Theorem 24.2.1. The automorphism group of the disc is given by
α−z
Aut(D) = {ψα,θ (z) = eiθ | α ∈ D, θ ∈ [0, 2π)}.
1 − z ᾱ
Moreover, the automorphism ψα,θ is precisely the automorphism that takes z = 0 to z = α. In particular, the
automorphisms of the disc fixing the origin are all given by z → eiθ z for some fixed θ.
The key tool in the proof is the Schwarz lemma, whose utility extends well beyond the computation of the
automorphism groups of the disc.
Lemma 24.2.2 (Schwarz lemma). Let f : D → D be a holomorphic map such that f (0) = 0. Then
• |f (z)| ≤ |z| for all z ∈ D.
• |f ′ (0)| ≤ 1.
Moreover, if |f (z)| = |z| for some non-zero z or |f ′ (0)| = 1, then f (z) = az for some a ∈ C with |a| = 1
Proof. Since f (0) = 0, there exists a holomorphic function
g:D→D
such that
f (z) = zg(z).
For any fixed z ∈ D, let 1 > r > |z|. Then by the maximum modulus principle,
|f (w)| 1
|g(z)| ≤ max < .
|w|=r r r
Letting r → 1− we see that |g(z)| < 1 for all z ∈ D, and hence
|f (z)| < |z|
for all z ∈ D. This directly implies that |f ′ (0)| ≤ 1. Now, suppose that |f (z0 )| = |z0 | for some z0 ∈ D\{0}.
Then |g(z0 )| = 1, and hence by the maximum modulus principle, g(z) must be a constant, and hence
f (z) = az for some a with |a| = 1.
Finally, suppose |f ′ (0)| = 1. Then there exists a sequence zn → 0, zn ̸= 0 such that
1 |f (zn )|
1− ≤ ≤ 1.
n |zn |
By the definition of g, we then have that 1 − 1/n ≤ |g(zn )| ≤ 1, and hence by continuity, |g(0)| = 1.
Then 0 is an interior maximum point for |g(z)|, and hence again by maximum modulus principle, g(z) is
a constant. That is, g(z) = a, where a = g(0) satisfies |a| = 1, and once again f (z) = az.
Proof of the theorem. There are two steps in the proof. We let ψα := ψα,0 as above (ie. ψα is the map
with rotation by zero angle). Note that ψα exchanges 0 and α. That is, ψα (0) = α and ψα (α) = 0.
24.3. AUTOMORPHISM GROUPS OF C AND C∗ 145
• We’ll first prove that each ψα,θ is actually an automorphism. To see this, first we observe that
|α|2 + |z|2 − z ᾱ − z̄α

|ψα,θ (z)|2 = ψα,θ (z)ψα,θ (z) = ,
1 + |α|2 − z ᾱ − z̄α
and so |ψα,θ (z)| = 1 if |z| = 1. But then by the maximum principle, since ψα,z is clearly non-
constant, |ψα,θ (z)| < 1 for all z ∈ D. Hence ψα,θ maps D into itself. Next, consider φα := ψα ◦ ψα .
Clearly, φα,θ (0) = 0 and φα (α) = α, and φα maps D into itself. By equality in Schwarz lemma,
φα (z) = z for all z ∈ D. In particular, ψα is surjective and injective, and hence a biholomorphism,
with inverse function given by itself ie. ψα−1 = ψα . But then ψα,θ = eiθ ψα is also clearly a biholo-
morphism.
• It remains to show that these are the only automorphisms. Let F ∈ Aut(D) such that F (0) = 0. By
the Schwarz lemma, |F (z)| ≤ |z|. But the same also holds true for F −1 (z), and so
|z| = |F −1 (F (z))| ≤ |F (z)| ≤ |z|,
and hence all inequalities must be equalities. That is, |F (z)| = |z| for all z ∈ D. By the equality part
of Schwarz lemma, we have that F (z) = eiθ z for some θ ∈ [0, 2π). Now suppose F ∈ Aut(D) such
that F (α) = 0. Consider the automorphism Fα (z) := F (ψα (z)). Then Fα (0) = 0, and hence by
the above argument, Fα (z) = eiθ z for some θ. But then, since ψα−1 = ψα , we have that
F (z) = Fα (ψα−1 (z)) = eiθ ψα (z) = ψα,θ (z).
This completes the proof of the theorem.
24.3 Automorphism groups of C and C∗

Theorem 24.3.1. Let f : C → C be an injective holomorphic map. Then
f (z) = az + b
for some a, b ∈ C with a ̸= 0.Since any such map is automatically surjective, we have that
Aut(C) = {az + b | a, b ∈ C, a ̸= 0}.
Theorem 24.3.2. Any injective holomorphic map f : C∗ → C∗ is given by either

a
f (z) = az or f (z) = ,
z
for some a ∈ C and a ̸= 0. Since any such map is automatically surjective, we have that
a
Aut(C∗ ) = {az, | a ∈ C, a ̸= 0}.
z
The key lemma needed to compute both the automorphism groups is the following.
Lemma 24.3.3. Let g : C∗ → C be holomorphic and injective. Then g cannot have an essential singularity
at z = 0.
Proof. Suppose, for the sake of contradiction, g has an essential singularity at z = 0. Consider the disc
D1 (2) of radius 1 around the point z = 2, and the unit disc D. Note that both these discs are disjoint
from each other. By the open mapping theorem, V = g(D1 (2)) is an open neighborhood of g(2). By
the Casorati-Weierstrass theorem, g(D) is dense in C. That means there is some w ∈ g(D1 (2)) and some
z1 ∈ D such that
g(z1 ) = w.
But since w ∈ g(D1 (2)), there is already a z2 ∈ D1 (2) such that
g(z2 ) = w.
Since the open discs D1 (2) and D are disjoint sets, z1 ̸= z2 , but g(z1 ) = w = g(z2 ). This contradicts the
injectivity of g. Hence z = 0 cannot be an essential singularity.
We also need an elementary generalization of Liouville’s theorem, which was a homework problem some-
time back. We provide a proof for the sake of completeness.
Lemma 24.3.4. Let f : C → C be an entire function such that
|f (z)| ≤ M (1 + |z|n )
for some M > 0 and all z ∈ C. Then f is a polynomial of degree less than or equal to n.
Proof. By the Cauchy estimates (Corollary 3 in Lecture-8), if R > 1, we have that,
k!M (1 + Rn )
|f (k) (0)| ≤ . ≤ 2M k!Rn−k .
Rk
But this holds no matter what R is chosen. So letting R → ∞, if k > n, the right hand side goes to zero.
Hence
f (k) (0) = 0
for all k = n + 1, n + 2, · · · . But since f is entire, it has a power series expansion whose coefficients are
given by
f (k) (0)
ak = = 0,
k!
for k > n. Hence the power series terminates, and f is a polynomial of degree less than or equal to n.
We are now ready to compute the automorphism groups of C and C∗ .
Proof of Theorem 24.3.1. Let f : C → C be an injective, holomorphic map. The key idea is to study the
function at infinity. That is, define a holomorphic function g : C∗ → C by
1
g(z) := f .
z
Then it is easy to see that g is also injective. Applying lemma 24.3.3, g either has a zero or a pole at z = 0.
1
In any case, this means that there exists a constant M > 0 and integer n > 0 such that |z|n |g(z)| < M on
|z| < 1. Transferring the estimates to f , we see that
|f (z)| < M |z|n ,
for |z| > 1. On the other hand, since |z| ≤ 1 is compact, we actually get that (possibly by choosing a
bigger M ), that
|f (z)| ≤ M (1 + |z|n ),
for all z ∈ C. Then by lemma 24.3.4, f (z) is a polynomial. By the fundamental theorem of algebra, f (z)
has at most n roots. We claim that n = 1. To see this, note that by injectivity, all the roots have to be
identical, or equivalently, f (z) = a(z − α)n for some a, α ∈ C with c ̸= 0. If n > 1, then f ′ (α) = 0, and
24.4. AUTOMORPHISM GROUP OF THE EXTENDED COMPLEX PLANE 147
hence f (z) cannot even be locally injective (see Theorem 1 or Corollary 0.3 in Lecture 19), and hence we
must have n = 1. But then clearly f (z) = az + b, where we put b = −aα. This proves the first part of the
theorem. For the second part, notice that any linear polynomial is surjective, and hence the f above will
automatically be surjective, and hence give an automorphism.
Proof of Theorem 24.3.2. Let f : C∗ → C∗ be an injective map. Then by lemma 24.3.3, z = 0 is either a
removable singularity or a pole.
Case 1. Suppose z = 0 is a removable singularity. Then f extends to f˜ : C → C. We then claim that f˜ is

also injective. Suppose not, then since f is injective, the only possibility is that there are z0 , w0 ∈ C with
z0 ̸= 0 such that
f˜(z0 ) = f˜(0) = w0 .
By the argument principle there is a neighborhood U of w0 and disjoint discs Dr (z0 ), Dρ (0) around z0
and 0 respectively such that for any w ∈ U , w ̸= w0 there are solutions z1 ∈ Dr (z0 ) and z2 ∈ Dρ (0) to
f˜(z) = w.
. But z2 ̸= 0 since w ̸= w0 . So the two distinct solutions are actually solutions to
f (z) = w,
contradicting the injectivity of f . This proves that the extension f˜ : C → C is an injective holomorphic
map. But then by Theorem 24.3.1, f˜(z) = az + b for some a, b ∈ C and a ̸= 0. All that is needed now is
to show that b = 0. If not, then 0 = f˜(−b/a) = f (−b/a) which is a contradiction since f takes non-zero
values being a map from C∗ into C∗ . To sum up, in this case f (z) = az.
Case-2: Suppose z = 0 is a pole. Then if we define
1
h(z) = ,
f (z)
then h extends to an holomorphic function h̃ : C → C. Then by the proof in the first case, we can see that
h̃(z) = cz,
for some c ̸= 0. But then this shows that f (z) = z/c, and proves the theorem with a = 1/c.
24.4 Automorphism group of the extended complex plane

. Recall that a map F : C → Ĉ is called holomorphic if
1. If F is a meromorphic function and

C
2. G(z) = F (1/z) is holomorphic in a neighbourhood of z = 0.

Theorem 24.4.1. The automorphism group of the extended complex plane is given by
n az + b o
Aut(C) = T (z) = | ad − bc = 1 ,
cz + d
and hence Aut(C) ∼

= P SL(2, C) = SL(2, C)/ ± I.
Proof. From our discussion in the previous lecture, it is clear that any T (z) defined as above is an auto-
morphism of Ĉ, and hence the set on the right is contained in Aut(C). To show the reverse containment,
let T ∈ Aut(Ĉ). Then there is a unique point z0 ∈ C such that T (z0 ) = ∞, and a unique point w0 such
that T (∞) = w0 .
• Case-1: z0 = ∞. In this case, w0 = ∞, and so F (z) = T is a bi-holomorphism of C, and

C
by Theorem 24.3.1, F (z) = az + b for some a, b and a ̸= 0. But then extending this to infinity,
T (z) = az + b and hence we are done.
• Case-2: z0 ̸= ∞. In this case, since T is one-one, w0 ̸= ∞ (else it will map both z0 and ∞ to ∞,
contradicting the one-one property). Now consider the function F : C∗ → C defined by
F (z) = T (z + z0 ) − w0 .
It is easy to check that F maps C∗ to C∗ , and is infact an automorphism of C∗ is (this follows

essentially from the fact that F has a zero at infinity). Moreover, since z = 0 is a pole for F , by
Theorem 24.3.2, there exists an a ∈ C∗ such that F (z) = a/z. But then solving for z,
a
T (z) = + w0 ,
z − z0
and is a Mobius transformation.
Lecture 25
The Riemann mapping theorem
Recall that two domains are called conformally equivalent or biholomorphic if there exists a holomorphic
bijection from one to the other. This automatically implies that there is an inverse holomorphic function.
The aim of this lecture is to prove the following deep theorem due to Riemann. Denote by D the unit disc
centered at the origin.
Theorem 25.0.1. Let Ω ⊂ C be a simply connected set that is not all of C. Then for any z0 ∈ Ω, there exists
a unique biholomorphism F : Ω → D such that
F (z0 ) = 0, and F ′ (z0 ) > 0.
Here F ′ (z0 ) > 0 stands for F ′ (z0 ) being real and positive, and can be thought of as a normalization, to
ensure that the above map is unique. The precise normalization by itself is not very important. The reader
should try to test her/his understanding of the proof by coming up with other normalizations that work,
and also some that do not work (for instance, you might not be able to impose that F ′ (z0 ) = 1). Note that
by Liouville’s theorem, such a statement is patently false if Ω = C, and so the hypothesis that Ω is a proper
subset is a necessary condition. As a consequence of the Theorem, we have the following corollary.
Corollary 25.0.1. Any two proper, simply connected subsets for C are conformally equivalent.
Proof of uniqueness in Theorem 25.0.1. Let F1 : Ω → D and F2 : Ω → D be two such mappings. Then
f = F2 ◦ F1−1 satisfies the following properties
• f : D → D is injective and onto.
• f (0) = 0.
• f ′ (0) > 0.
• f −1 also satisfies both these properties.
By Schwarz lemma, |f (z)| ≤ |z| for all z ∈ D and |f −1 (w)| ≤ |w| for all w ∈ D. Let w = f (z), then
second inequality gives |z| ≤ |f (z)|, and hence |z| = |f (z)|. But then by the equality part of Schwarz
lemma, we see that f (z) = az for some a ∈ C with |a| = 1. But then f ′ (0) = a, which forces a = 1 (since
f ′ (0) > 0). Hence f (z) = z for all z ∈ D or equivalently F2 (w) = F1 (w) for all w ∈ Ω.
25.1 Montel’s and Hurwitz’s theorems

The proof relies on two theorems on sequences of holomorphic functions. Recall that we say that a se-
quence of functions fn converges compactly on Ω to f if it converges uniformly on any compact subset
149
150 LECTURE 25. THE RIEMANN MAPPING THEOREM
K ⊂ Ω. More precisely, for every compact set K ⊂ Ω and ε > 0, there exists an N = N (ε, K) such that
sup |fn (z) − f (z)| < ε

z∈K
whenever n > N . In Lecture 9 we proved the following theorem:

Theorem 25.1.1. If {fn }∞ n=1 is a sequence of holomorphic functions on Ω that converge compactly to f :
Ω → C, then f (z) is holomorphic. Moreover
fn(k) → f (k)
compactly on Ω for all k ∈ N.

We say that family of continuous functions F on an open set Ω is normal if every sequence of functions in
F has a subsequence that converges compactly on Ω. Note that the definition does not require the limiting
function to be contained in F. On the other hand, by Theorem 25.1.1 above, the limiting function will
certainly be holomorphic. The family is said to be locally uniformly bounded if for any K ⊂ Ω compact,
there exists a constant MK such that
sup |f (z)| < MK
z∈K
for all f ∈ F.
Theorem 25.1.2 (Montel’s theorem). A family F of holomorphic functions on Ω is normal if and only if it
is locally uniformly bounded.
To prove this, we first recall the Arzela-Ascoli theorem. Recall that family F of continuous functions on Ω
is said to locally equicontinuous if for all a ∈ Ω and all ε > 0 there exists a δ = δ(a, ε) such that
z, w ∈ Dδ (a) =⇒ |f (z) − f (w)| < ε,
for all f ∈ F. Then we have the following basic theorem, which we state without proof.
Theorem 25.1.3 (Arzela-Ascoli). If a family of functions is locally equicontinuous and locally uniformly
bounded, then for every sequence of functions {fn } ∈ F, there exists a continuous function f and a subse-
quence {fnk } which converges to f compactly on Ω.
Remark 25.1.1. Generally, Arzela-Ascoli is stated for compact sets assuming equicontinuity. One can
prove the above theorem by taking an exhaustion of Ω by compact sets ie. K1 ⊂ K2 · · · ⊂ Kn · · · such
that Ω = ∪Kn . Local equicontinuity will then imply that the family is genuinely equicontinuous on
each compact set Kn . Then apply the standard Arzela-Ascoli to F restricted to each Kn , and use Cantor
diagonalization argument.
Proof of Montel’s theorem. First, suppose that F is a locally uniformly bounded family of holomorphic
functions. By the Arzela-Ascoli theorem, if we show that F is automatically locally equicontinuous, then
for every sequence, there will exist a subsequence which converges compactly on Ω to a continuous func-
tion f . By Theorem 25.1.1, the limit function will then be holomorphic, and hence F would be a normal
family.
Hence it is enough to show that the family F is locally equicontinuous. To do this, we use the Cauchy
integral formula. Fix an a ∈ Ω and ε > 0. We need to choose a delta that works. Let r > 0 be such that
D2r (a) ⊂ Ω, and let Mr such that
|f (ζ)| ≤ Mr ,
for all ζ ∈ D2r (a) and all f ∈ F. By the Cauchy estimates (see Corollary 3 from Lecture 8), we have that
for any ζ ∈ Dr (a),
2Mr
|f ′ (ζ)| ≤ .
r
25.2. PROOF OF RIEMANN MAPPING THEOREM 151
Then by the fundamental theorem for complex integrals, for any z, w ∈ Dr (a),
ˆ
2Mr
|f (z) − f (w)| = f ′ (ζ) dζ ≤ |z − w|.
lw,z r
Given an ε > 0, let us pick δ < 2Mr ε/r. Then whenever |z − w| < δ, we have |f (z) − f (w)| < ε. This
proves local equicontinuity.
Conversely, suppose F is a normal family, but not locally uniformly bounded. Then there exists a compact
set K ⊂ Ω, and a sequence of functions fn ∈ F such that
sup |fn (z)| ≥ n.

z∈K
Since the family is normal, there exists a subsequence fnk which converges uniformly on K. But then
supz∈K |fnk | would be a bounded sequence which is a contradiction.
We also need the following theorem due to Hurwitz on the limit of injective holomorphic functions.
Theorem 25.1.4 (Hurwitz). Let fn : Ω → C be a sequence of holomorphic, injective functions on an open
connected subset, which converge uniformly on compact subsets to F : Ω → C. Then either F is injective, or
is a constant.
Proof. We argue by contradiction. So suppose F is non constant and not injective. Then for some w ∈ C,
there exists a, b ∈ Ω such that F (a) = F (b) = w. Let fn (a) = wn , then wn → w. Choose an r > 0 small
enough so that there does not exist any z ∈ Dr (b) such that F (z) = w. This is possible by the principle
of analytic continuation since we are assuming that F is non-constant. In particular a ∈
/ Dr (b). Since fn
is injective for any n, there exists no solution to
fn (z) = wn
in the closure of the disc Dr (b), and so by the argument principle applied to fn (z) − wn , we see that
ˆ
1 fn′ (ζ)
dζ = 0.
2πi |ζ−b|=r fn (ζ) − wn
But since fn → F uniformly on compact sets, in particular, on the compact set Dr (a) we have fn′ (ζ) →
F ′ (ζ) and fn (ζ) − wn → F (ζ) − w uniformly. Hence the integral also converges uniformly, and from this
we conclude that ˆ
1 F ′ (ζ)
dζ = 0.
2πi |ζ−b|=r F (ζ) − w
This integral calculates the number of zeroes of F (ζ) − w = 0 in Dr (b) which we know is at least one
(counting multiplicity) since F (b) = w. This is a contradiction, and hence if F is non-constant, it has to
be injective.
25.2 Proof of Riemann mapping theorem

25.2.1 Some further preparation
For a fixed z0 ∈ Ω, we define a family F of holomorphic function functions by
F = {f : Ω → D | f holomorphic and injective, f (z0 ) = 0}.
The required biholomorphic map will be obtained by maximizing the modulus of the derivative at z0 ,
amongst all functions in this family. We first show that this family is non-empty.
Lemma 25.2.1. There is an injective holomorphic function f : Ω → D such f (z0 ) = 0. That is, F ̸= ϕ.
Proof. Since Ω ̸= C, there is an a ∈ C \ Ω. Then z − a is never zero on Ω. Since Ω is simply connected,

we can choose a holomorphic branch of log (z − a), or in other words, there is a holomorphic function
l : Ω → C such that
el(z) = z − a
for all z ∈ Ω. Clearly l(z) is injective. Moreover, if z1 , z2 ∈ Ω and z1 ̸= z2 then l(z2 ) − l(z1 ) ∈
/ 2πiZ, i.e
their difference cannot be an integral multiple of 2πi. In particular, l(z) ̸= l(z0 ) + 2πi. We in fact claim
that |f (z) − (f (z0 ) + 2πi)| is bounded strictly away from zero. That is,
Claim. There exists an ε > 0 such that |l(z) − (l(z0 ) + 2πi)| > ε for all z ∈ Ω.
To see this, assume the claim is false. Then there is a sequence {zn } ∈ Ω such that l(zn ) → l(z0 )+2πi. But
then exponentiating, since the exponential function is continuous, we see that zn → z0 . But then, since
l(z) is continuous, this implies that l(zn ) → l(z0 ) contradicting the assumption that l(zn ) → l(z0 ) + 2πi.
This proves the claim.
Now consider the function
1
f˜(z) = .
l(z) − l(z0 ) − 2πi
By the claim, this is a bounded, injective and holomorphic function on Ω, and hence f˜ : Ω → DR (0),
where R can be taken to be R = 1/ε where ε is from the claim above. Suppose f˜(z0 ) = a, then
f˜(z) − a
f (z) =
R + |a|
is the required function.
Next, let
λ = sup |f ′ (z0 )|.
f ∈F
We claim that λ > 0. To see this, consider the f ∈ F constructed above. Since f (z) is injective, by
Corollary 0.3 from Lecture 19, |f ′ (z0 )| > 0 and hence λ > 0.
Lemma 25.2.2. There is a function F ∈ F such that |F ′ (z0 )| = λ. In particular, λ is also finite.
Proof. Let fn ∈ F be a sequence of functions that maximize |f ′ (z0 )|; that is
lim |fn′ (z0 )| = λ.

n→∞
Since |fn (z)| < 1, by Montel’s theorem there is a subsequence that converges uniformly on compact sets
to a holomorphic function F satisfying |F (z)| ≤ 1, and F (z0 ) = 0. Moreover since the derivatives also
converge, we must have |F ′ (z0 )| = λ ̸= 0. In particular, F cannot be a constant. Then by the maximum
modulus principle, |F (z)| < 1 for all z ∈ Ω, since otherwise, there will be a point z ∈ Ω with |F (z)| = 1,
and hence will be an interior maximum point for |F |. Finally to show that F ∈ F, we need to show
that F is injective. But this follows from Hurwitz’s theorem since F is non-constant and all fn ∈ F are
injective.
25.3. GREEN’S FUNCTIONS AND A GENERALIZATION OF THE RIEMANN MAPPING THEOREM 153
25.2.2 Completion of the proof of Riemann mapping

Since F (z0 ) ̸= 0, by composing with a suitable rotation, we can assume that F ′ (z0 ) is real and positive.
′
We claim that this F is the required bi-holomorphism. We already know that F : Ω → D, F (z0 ) = 0 and
F is injective. To complete the proof, we need to show that F is surjective. If not, then there exists a α ∈ D
such that F (z) = α has no solution in Ω. We then exhibit a G ∈ F with |G′ (z0 )| > |F ′ (z0 )| contradicting
the choice of F . To do this, consider ψα : D → D defined by
α−z
ψα (w) = ,
1 − ᾱz
and let p
g(z) = ψα ◦ F (z).
Since ψα (z) = 0 if and only if z = α, we see that ψα ◦ F is always zero free, and so a holomorphic branch
of log ψα ◦ F can be defined since Ω is simply connected. We can then choose a holomorphic branch for
g(z) by letting
1
g(z) = e 2 log ψα ◦F (z) .
√
Note that g(z0 ) = α. To construct a member of the family, we need to bring this back to the origin, and
hence we define
G(z) = ψ√α ◦ g(z).
Then G(z0 ) = 0. Moreover, G(z) is also injective since ψ√α and g(z) are injective, and so G ∈ F.
Claim. |G′ (z0 )| > |F ′ (z0 )|.
To see this, observe that

−1
F (z) = ψα−1 ◦ s ◦ ψ√α
◦ G(z) = Φ ◦ G(z),
−1
where s(w) = w2 is the squaring function and Φ = ψα−1 ◦ s ◦ ψ√ α
: D → D. We compute that Φ(0) =
−1 −1 −1
√ −1
ψα (s(ψ√α (0)) = ψα (s( α)) = ψα (α) = 0. By Schwarz lemma, |Φ(z)| ≤ |z|, and so
|Φ′ (0)| ≤ 1.
We claim that |Φ′ (0)| < 1. Suppose, |Φ′ (0)| = 1, then by the second part of Schwarz lemma, Φ(z) = az
for some unit complex number a. In particular, Φ(z) is injective. But Φ cannot be injective since s(z) is
a 2 − 1 function ie. sends two points to a single point, and ψα and ψ√α are injective. This shows that
|Φ′ (0)| < 1. But then F ′ (z0 ) = Φ′ (G(z0 )) · G′ (z0 ) = Φ′ (0) · G′ (z0 ), and hence |F ′ (z0 )| < |G′ (z0 )| which
proves the claim, and completes the proof of the theorem.
25.3 Green’s functions and a generalization of the Riemann map-

ping theorem
For the purposes of this section, we assume that Ω is bounded and has a sufficiently nice boundary ∂Ω (for
concreteness, assume that ∂Ω is a union of piecewise regular curves). Often one is interested in solving
the Dirichlet problem on Ω: Namely, given any smooth (real valued) function f on Ω and a continuous
function u0 on ∂Ω, to find a smooth function u such that
(
∆u = f
u = u0 .
∂Ω
This problem arises in many (seemingly) different areas in mathematics and physics. For instance, one
interpretation of solutions of the above problem, is that u represents the voltage distribution on a conductor
Ω with charge distribution given by f (z) and the boundary being held at voltage u0 .
A Green’s function for Ω based at z0 ∈ Ω is a function Gz0 : Ω → R such that
1. Gz0 (z) is a harmonic function on Ω \ {z0 }.
2. Gz0 ≡ 0.
∂Ω
1
3. Gz0 (z) − 2π log |z − z0 | is bounded in a neighbourhood of z0 .
Note that log |z − z0 | is a harmonic function in a punctured neighbourhood of z0 . By an analog of the
Riemann removable singularity theorem for harmonic functions, condition (3) is equivalent to Gz0 (z) −
1
2π log |z − z0 | extending as a harmonic function in a disc Dr (z0 ). It is often useful to think of the Green’s
function as a function of two variables G : Ω × Ω → R, where we set
G(z, w) := Gw (z).
It turns out that the function G is actually symmetric in the two variables. The reason why Green’s function
is important is that it is a fundamental solution to the Dirichlet problem on Ω. That is, with f and u0 as
above, a solution to the Dirichlet poblem is given by
ˆ ˆ
dG
u(z) = G(z, w)f (w) dwdw̄ − (z, w)g(w) dσ,
Ω ∂Ω dν
where dG/dν is the outward normal derivative of G on the boundary (where differentiation is with respect
to the variable w), dwdw̄ is the usual Lebesgue (or Euclidean) measure on Ω, and dσ is the surface measure
on ∂Ω.
Conversely, if one can Dirichlet problems with continuous data, then one can construct a Green’s function.
The idea is to simply find a harmonic function Hz0 (z) with boundary data u0 = − log |z −z0 |. The Green’s
function Gz0 (z), will then be
1 1
Gz0 (z) = log |z − z0 | + Hz (z).
2π 2π 0
Now suppose that Ω is simply connected. Fix a z0 ∈ Ω, and as above, let Hz0 (z) := 2πGz0 (z)−log |z −z0 |,
which is harmonic by property (3) above. Since Ω is simply connected, Hz0 (z) has a harmonic conjugate,
that is a function Hz∗0 (z) : Ω → R which is harmonic, and such that f (z) = Hz0 ( z) + iHz∗0 (z) is a
holomorphic function on Ω (The proof is essentially the same as that for a disc, and this was an exam
problem on the midterm). We let F (z) = (z − z0 )ef (z) . For any z ∈ ∂Ω, Gz0 (z) = 0, Hz0 (z) =
− log |z − z0 |, and so |F (z)| = 1. By the maximum principle, |F (z)| < 1 for all z ∈ Ω. Hence F is a map
from Ω to the unit disc D. Furthermore, F (z) has only one zero in Ω, and that too a simple one, namely
at z = z0 . Next, let w0 ∈ D such that |w0 | < 1 − ε < 1, and let γ be the curve given by |F (z)| = 1 − ε
and Γ = F ◦ γ. Then Γ is of course the circle |w| = 1 − ε (but possibly traversed multiple times). Since
F (z) = 0 has only one solution in Ω, by the argument principle (rather the index version of it), we see
that ˆ ˆ
F ′ (z) 1 F ′ (z)
dz = n(Γ, w0 ) = n(Γ, 0) = = 1.
γ F (z) − w0 2πi γ F (z)
This shows that F (z) = w0 has a unique solution for any w0 ∈ D, and hence shows both surjectivity and
injectivity of F (z). Finally, we can compose F with a rotation to ensure that F ′ (z0 ) is real and positive.
The reader is encouraged to read a detailed and complete account of a proof of the Riemann mapping theo-
rem along the lines of Riemann’s original “proof" from https://link.springer.com/article/
10.1186/s40627-016-0009-7.
25.3. GREEN’S FUNCTIONS AND A GENERALIZATION OF THE RIEMANN MAPPING THEOREM 155
We end with a vastly more general Riemann mapping theorem. Recall that a domain Ω is said to be n-
connected, if C \ Ω has n connected components. For instance, Ω is 1-connected if and only if it is simply
connected.
Theorem 25.3.1. Let Ω be an n-connected domain in C such that no component of C \ Ω consists of a single
point. Then there exists a biholomorphism F : Ω → D, where
1. D = D if n = 1,
2. D is an annulus Ar,R (0) = {z ∈ C | r < |z| < R} if n = 2,
3. D is Ar,R (0) \ ∪n−2
i=1 Supp(γi ) if n > 2, where γi are concentric arcs lying on circles |ζ| = ri , with
r < ri < R.
A historical note. One of the first pushes towards making Dirichlet’s problem and harmonic functions a
part of mainstream mathematics arose out of Riemann’s (faulty) proof of his theorem on conformal map-
pings into the disc. In fact the first systematic and rigorous study of the Dirichlet problem was to fix
the error in Riemann’s original proof. By the turn of the twentieth century, the vastly more general uni-
formization theorem had also been proved using similar methods, and elliptic partial differential equations
and calculus of variations (of which the above problem is the simplest example) had become a part of
mainstream mathematics. So much so that, they were the subject of two of Hilbert’s problems in his 1900
address to the Congress of mathematicians. Finally, this whole circle of ideas of using solutions of partial
differential equations to say something about the topology continues to be a fruitful area of mathematical
research. Some of the spectacular successes include Hodge theory (characterizing cohomology groups via
harmonic forms) and Donaldson theory (characterising smooth structures on four manifolds via solving
Yang-Mills equations, which are a non-linear generalization of Dirichlet’s problem).
Part IV
Special functions
157
Lecture 26
Gamma and Zeta functions
In this lecture, we study two important functions, namely the Gamma function and the Zeta function. Each
function is initially defined in a certain region in the complex plain; the Gamma function by an integral
and the Zeta function by an infinite series. Both the functions are then extended to obtain meromorphic
functions on the entire complex plain. The key techinical tool is the principal of analytic continuation.
26.1 Revisiting the principle of analytic continuation

Recall that the principal of analytic continuation says that if two functions agree on some open set, then
they must agree on the entire connected component containing the open set. Given a holomorphic function
f : Ω → C it is natural to ask for the biggest possible open set Ω′ containing Ω on which f has a
holomorphic extension. It is in fact much more natural to ask for meromorphic extensions. So we pose
the following question.
Question 26.1.1. Given f : Ω → C holomorphic, what is the biggest Ω′ containing Ω such that there exists
a meromorphic function F : Ω′ → C such that
F|Ω = f.
Is the extension unique.

The uniqueness part is answered in the affirmative by the following extension of the principle of analytic
continuation to meromorphic functions.
Lemma 26.1.2. Let Ω be a connected open set, and f, g : Ω → C be meromorphic functions with poles at
isolated sets Sf and Sg respectively. Let S = Sf ∪ Sg . Suppose there is a sequence of pairwise distinct points
zn ∈ Ω \ S such that zn → z0 ∈ Ω and f (zn ) = g(zn ) for all n, then Sf = Sg = S, the poles of f and g are
of the same order and the Laurent series expansions match up, and f = g as meromorphic functions.
Proof. Since f, g ∈ O(Ω \ S), by the usual principle of analytic continuation,
f =g .
Ω\S Ω\S
To complete the proof, we need to show that Sf = Sg = S. Let p ∈ Sf . Then there is an ε > 0 such that
Dε (p) does not contain any point of S apart from p. That is, Dε (p) \ {p} ⊂ Ω \ S, and hence f (z) = g(z)
for all z ∈ Dε (p) \ {p}. Since p is a pole of f , f (z) → ∞ as z → p, and hence g(z) → ∞ as z → p. This
shows that Sf ⊂ Sg . By symmetry we get the reverse inclusion and this proves that Sf = Sg . Since f
159
160 LECTURE 26. GAMMA AND ZETA FUNCTIONS
and g are equal in the complement, it is also clear that the poles will be of the same order, and the Laurent
series expansions match up. Hence f = g as meromorphic functions.
Example 26.1.3. Consider the function defined by the power series

∞
X
f (z) = zn.
n=0
This defines a holomorphic function on the unit disc D. Moreover, by the summation formula for geometric
series, the function is precisely f (z) = (1 − z)−1 . On the other hand, the function F (z) = (1 − z)−1
is meromorphic on the entire complex plane with a single pole of order one at z = 1. So F (z) defines a
meromorphic extension of f (z) to C and an analytic continuation to C \ {1}. Note that F (−1) = 1/2, since
F (z) is an analytic continuation of f (z), naively (and rather thoughtlessly) one might be tempted to plug in
z = −1 in the power series and write
1
1 − 1 + 1 − 1 + ···“ = ” . (26.1)
2
Of course as stated the above equality is meaningless since the series on the left is a divergent series. The correct
way to make sense out of this is by analytic continuation. There are other ways to make sense out of the series,
for instance by using Cesaro summability. G.H Hardy wrote an entire book on divergent series, and this was
quite a hot topic for research in Britain in the early 20th century.
26.2 Analytic continuation of the Mellin transform

In this section we answer the above question completely in the following model case, for the (truncated)
Mellin transform. Recall that a function φ : [−1, 1] → R is called smooth, if derivatives of all orders exist
on (−1, 1), and are continuous on [−1, 1].
Theorem 26.2.1. Let φ(t) be a smooth function on the unit interval [−1, 1], and consider the function
ˆ 1
f (z) = xz φ(x) dx,
0
where we use the principal branch, namely xz = ez ln x . Then

1. f (z) defines a holomorphic function on Re(z) > −1.
2. f (z) admits a unique extension as a meromorphic function on C with at most simple poles at the negative
integers with
φ(n−1) (0)
Resz=−n f (z) = ,
(n − 1)!
where φ(k) (0) as usual denotes the k th derivative.
Note that the usual Mellin transform involves the integral over all of (0, ∞), and hence we call the above
a truncated Mellin transform (possibly non standard terminology).
Proof. To prove this, we first show that the integral is absolutely convergent for Re(z) > −1. We only need
to worry about convergence near x = 0. It is easy to see that |xz | = |ez ln x | = xRe(z) , and so if |φ(x)| < M
on [0, 1], then |xz φ(x)| < M xRe(z) , which, by the p-test, is integrable near x = 0 if Re(z) > −1. So by
the comparison theorem, f (z) is well defined for Re(z) > −1. To show that it is holomorphic, we look at
the difference quotient. Note that xz is holomorphic for all x ∈ (0, 1) with derivative
dxz
= xz ln x.
dz
26.2. ANALYTIC CONTINUATION OF THE MELLIN TRANSFORM 161
We claim
ˆ 1
f ′ (z) = xz φ(x) ln x dx.
0
To prove this, it is enough to show the following.
Claim. For all ε > 0, there exists a δ > 0 such that

ˆ 1
f (z + h) − f (z)
− xz φ(x) ln x dx < ε,
h 0
whenever |h| < δ.
Note that
ˆ 1 ˆ 1
f (z + h) − f (z) xz+h − xz
− xz φ(x) ln x dx = ( − xz ln x)φ(x) dx
h 0 0 h
ˆ 1
eh ln x − 1
≤ |xz φ(x)| − ln x dx. (26.2)
0 h
By the power series expansion of ez , we have that

∞
eh ln x − 1 X hn (ln x)n
− ln x = h(ln x)2 .
h n=0
(n + 2)!
Now the infinite series on the right is convergent. In fact we have that
∞ ∞
X hn (ln x)n X hn (ln x)n
≤ < e|h|| ln x| = x−|h| ,
n=0
(n + 2)! n=0
n!
where we used the fact that | ln x| = ln(1/x) since x ∈ (0, 1). From the power series expansion above, we
then have the estimate
eh ln x − 1
− ln x ≤ h(ln x)2 x−|h| ≤ Cη hx−|h|−η ,
h
for any η > 0 and |h| small. Here Cη is a constant that possibly depends on η but is independent of h. This
estimate holds because limx→0 xη (log x)2 for any η > 0.
Now suppose M = supx∈[0,1] |φ(x)|, then going back to the integral estimate in (26.2) we see that if
|h| < Re(z), then
ˆ 1 ˆ 1
z eh ln x − 1
|x φ(x)| − ln x dx ≤ hCη M xRe(z)−|h|−η dx.
0 h 0
We choose η > 0 small enough Re(z) − 2η > −1. Suppose |h| < η, then
ˆ 1 ˆ 1
Re(z)−|h|−η
x dx ≤ xRe(z)−2η dx := Aη .
0 0
Note that Aη of course depends on z, but z is fixed throughout this argument, and hence we hide the
dependence of Aη on z. So putting all of this together with (26.2)
ˆ 1
f (z + h) − f (z)
− xz φ(x) ln x dx ≤ M Aη Cη |h| < ε,
h 0
if |h| < ε/M Cη Aη . So the claim is proved by choosing

ε
δ = min ,η .
M C η Aη
Next, we show that a meromorphic extension exists on all of C. For any given integer N > 0, we can write
the Taylor expansion of φ around x = 0 as
N −1
X φj (0) j
φ(x) = x + EN (x),
j=0
j!
where EN (x) is a smooth function on [−1, 1] such that |EN (x)| ≤ C|x|N for some constant C > 0. So
for Re(z) > −1,
N −1 ˆ 1 ˆ 1
X φk (0) k+z
f (z) = x + EN (x)xz dx
0 k! 0
k=0
N −1 ˆ 1
X φk (0) 1
= · + EN (x)xz dx
k! z+k+1 0
k=0
´1
Since |EN (x)| ≤ C|x|N , 0 EN (x)xz dx is convergent for Re(z + N ) > −1, and hence defines a holo-
morphic function on Re(z) > −(N + 1) by the first part. On the other hand, putting k + 1 = j,
the first term on the right defines a meromorphic function on all of C with simple poles at z = −j,
j = 1, 2, · · · , N with residue φj−1 (0)/(j − 1)!. So the right hand side defines a meromorphic function
fN (z) on Re(z) > −(N + 1), which restricts to f (z) on Re(z) > −1. By uniqueness of meromorphic
extensions, for M > N , the restriction of fN and fM to Re(z) > −(N + 1) are equal, and hence letting
N → ∞, fN converges to a meromorphic function on all of C with simple poles at z = −n with residue
φ(n−1) (0)/(n − 1)!.
Remark 26.2.1. Note that if φ(n−1) (0) = 0 for some n, then the residue of f (z) at z = −n would be zero.
And since z = −n can at most be a simple pole, this would imply that z = −n is in fact not a pole at all,
but is a removable singularity.
26.3 The Gamma function

The Gamma function Γ(s) is defined on Re(s) > 0 as the Mellin transform of e−x . That is,
ˆ ∞
Γ(s) = e−x xs−1 dx.
0
It is easy to see (exercise!) that the integral is convergent on Re(s) > 0 and hence is well defined and finite
on this region. Note also that Γ(1) = 1.
Theorem 26.3.1. With Γ(s) defined as above for Re(s) > 0, we have the following.
1. There exists a meromorphic extension of Γ(s) on C with simple poles at s = 0, −1, −2, · · · with residue
(−1)n
Ress=−n Γ(s) = .
n!
26.3. THE GAMMA FUNCTION 163
2. (Functional equation) For s ̸= −n, n = 0, 1, 2, · · · ,
Γ(s + 1) = sΓ(s),
and hence for integers n, Γ(n + 1) = n!.
Proof. For Re(s) > 0, we can write

ˆ 1 ˆ ∞
−x s−1
Γ(s) = e x dx + e−x xs−1 dx.
0 1
The second integral is convergent for all s, and hence defines an entire function by similar arguments as
in the proof of the first part of Theorem 26.2.1. The first integral, by Theorem 26.2.1 (applied to s − 1 = z),
can be extended to a meromorphic function with simple poles at s = −n, for n = 0, 1, 2, · · · . The residue
also comes from the first term. Applying theorem 26.2.1 with φ = e−x , and s − 1 = z, we see that the
residue at s = −n (or z = −(n + 1)) is given by
1 dn (−1)n
Ress=−n Γ(s) = · e−x = .
n! dxn x=0 n!
We first prove part (2) when Re(s) > 0. In this range we can use the integral formula,
ˆ ∞
Γ(s + 1) = e−x xs dx
0
ˆ ∞
=− xs−1 de−x
0
x=∞
ˆ ∞
−x s
=e x + e−x dxs
x=0
ˆ ∞ 0
=s e−x xs−1 dx = sΓ(s).

0
To prove the equality over the entire complex plane, consider F (s) = Γ(s + 1) − sΓ(s).
Claim. F (s) extends to an entire function.
Assuming this, since F (s) ≡ 0 on Re(s) > 0, by the principle of analytic continuation, F (s) is identically
zero, and we are done.
Proof of the Claim. Clearly F (s) is holomorphic everywhere except possibly at the negative integers
and s = 0. Moreover, F can only have simple poles at these points. At s = 0, Γ(s + 1) is holomorphic,
and so is sΓ(s) since Γ has a simple pole at s = 0. So F can have pole at only negative integers. To rule
this out, let us calculate the residue. For n ∈ N,
(−1)n−1
Ress=−n Γ(s + 1) = Resz=−(n−1) Γ(z) = .
(n − 1)!
On the other hand,
(−1)n
Ress=−n sΓ(s) = lim s(s + n)Γ(s) = −nRess=−n Γ(s) = − ,
s→−n (n − 1)!
and hence Ress=−n F (s) = 0. Since s = −n is a simple pole, this implies that s = −n is a removable
singularity.
Theorem 26.3.2 (Euler reflection formula). The Gamma function satisfies the identity
π
Γ(s)Γ(1 − s) = .
sin πs
Proof. By analytic continuation, it enough to prove the identity for s ∈ R ∩ (0, 1). Recall that in Example-2
in Lecture-21, we proved the following identity: For 0 < a < 1,
ˆ ∞ −s
v π
dv = .
0 1+v sin πs
We now rewrite ˆ ∞ ˆ ∞
−x −s
Γ(1 − s) = e x dx = t e−vt (vt)−s dv,
0 0
where we made the change of variables x = vt. Note that the above formula for Γ(1 − s) is valid for all
t ≥ 0. Now we compute
ˆ ∞
Γ(s)Γ(1 − s) = e−t ts−1 Γ(1 − s) dt
ˆ0 ∞ ˆ ∞
= e−t ts−1 t1−s e−vt v −s dv dt
ˆ0 ∞ ˆ ∞ 0
−t(1+v) −s
= e v dt dv
0 0
ˆ ∞ −s
v
= dv
0 1+v
π
= .
sin πs
26.4 The Riemann Zeta function

For Re(s) > 1, we define the zeta function by the infinite series
∞
X 1
ζ(s) = s
,
n=1
n
where as before, ns = es ln (n) . This time by comparison test for series, since |ns | = nRe(s) , this is
a convergent series for Re(s) > 1, and by the Weierstrass M -test defines a holomorphic function on
Re(s) > 1.
Theorem 26.4.1. The zeta function above satisfies the following properties
1. For Re(s) > 1, we have the identity,
ˆ ∞
1 1
ζ(s) = ts−1 dt.
Γ(s) 0 et −1
That is, the zeta function (upto a factor of Γ(s)) is the Mellin transform of (et − 1)−1 .
2. ζ(s) can be extended to a meromorphic function on C with a simple pole at z = 1 and holomorphic on
C \ {1}. Moreover, we have
Resz=1 ζ(s) = 1.
26.4. THE RIEMANN ZETA FUNCTION 165
3. ζ(s) = 0 whenever s = −2n for some n ∈ N. These are the so called trivial zeroes of the zeta function.
The Riemann hypothesis conjectures that in fact all other zeroes (the so called non-trivial zeroes) lie on
the line Re(s) = 1/2.
Proof. For the first identity, we observe that

ˆ ˆ ∞
Γ(s) 1 ∞ −x
= e (x/n)s−1
dx = e−nt ts−1 dt.
ns n 0 0
where we changed variables x = nt in the second equality. Summing up we obtain

ˆ ∞ ∞ ˆ ∞ ˆ ∞
X 1 1
ζ(s)Γ(s) = e−nt ts−1 dt = − 1 ts−1 dt = ts−1 dt.
0 n=1 0 1 − e−t 0 et − 1
´∞ 1 s−1
Now let f (s) = 0 et −1 t dt. We can write
ˆ ∞
f (s) = φ(t)ts−2 ,
0
where φ(t) = t/(et − 1). From the Tayor expansion of et we can see that φ(t) is smooth on [−1, 1], and so
by Theorem 26.2.1 with z = s − 2, f (s) is holomorphic on Re(s − 2) > −1 or equivalently on Re(s) > 1,
and admits a meromorphic extension with simple poles at Re(s) = 1, 0, −1, −2, · · · . But Γ(s) itself has
simple poles at s = 0, −1, −2, · · · , and hence the meromorphic extension ζ(s) = f (s)/Γ(s) will have a
pole only at s = 1. When s = 1, s − 2 = −1, and so from Theorem 26.2.1, Ress=1 f (s) = φ(0). But
t t 1
= = ,
et −1 2
t + t /2 + · · · 1 + t/2 + · · ·
and so φ(0) = 1, and hence Ress=1 f (s) = 1. But since Γ(1) = 1, we then have that Ress=1 ζ(s) = 1. It
follows from Problem-7 in Assignment-4 that
∞
t t X Bn 2n
φ(t) = t
=1− + (−1)n−1 z ,
e −1 2 n=1 (2n)!
where Bn is the nth Bernoulli number. In particular, φ(2n+1) (0) = 0, for all n = 1, 2, · · · , and hence
φ(2n+1)
Ress=−2n f (s) = Resz=−2n−2 f (z) = = 0.
(2n + 1)!
So f (s) has a removable singularity at s = −2n. But since Γ(s) has a simple pole at s = −2n, it follows
that ζ(−2n) = 0 for all n ∈ N.
Example 26.4.1. Let us calculate ζ(0). First, note that if two functions f (z) and g(z) have a simple pole at
z = 0, then h(z) = f (z)/g(z) has a removable singularity at z = 0. Moreover the extension, which we also
denote by h(z), satisfies h(0) = Resz=0 f (z)/Resz=0 g(z). We apply this to the meromorphic extension of
ˆ ∞
1
f (s) = t−1
ts−1 dt,
0 e
and g(s) = Γ(s). For Re(s) > 1, we re-write

ˆ 1 ˆ ∞
t 1
f (s) = ts−2 dt + ts−1 dt.
0 et − 1 1 et − 1
The second part, by the argument above is an entire function. So the residue comes from the meromorphic
extension of the first integral. We apply Theorem 26.2.1 with φ(t) = t/(et − 1) and z = s − 2. To find the
residue at s = 0 we apply the second part of Theorem 26.2.1 (with n = 2, since z = −2 is the same as s = 0)
φ′ (0)
Ress=0 f (s) = .
1!
But we can write down the Taylor expansion of
t t 1 t
φ(t) = = = = 1 − + higher order terms,
et − 1 t + t2 /2 + · · · 1 + t/2 + · · · 2
and so φ′ (0) = −1/2. On the other hand the residue of the Gamma function is given by Theorem 26.3.1, and
we see that Ress=0 Γ(s) = 1, and putting everything together, we obtain that
1
ζ(0) = − .
2
Example 26.4.2. By working a bit harder, we can compute ζ(−1). Once again applying Theorem 26.2.1, this
time with n = 3 (since s = −1 is z = −3) we have
φ(2) (0)
Ress=−1 f (s) = .
2!
Computing the next term in the Taylor expansion,
1 t t2 t2 3 t t2
φ(t) = = 1 − − + + O(t ) = 1 − + + O(t3 ),
1 + t/2 + t2 /6 + O(t3 ) 2 6 4 2 12
and so Ress=−1 f (s) = 1/12. On the other hand, Ress=−1 Γ(s) = (−1)1 /1 = −1, and hence
1
ζ(−1) = − .
12
Remark 26.4.3. Recall that for Re(s) > 1,

∞
X
ζ(s) = n−s .
n=0
We can then formally (and formally is the key word here) “plug in" s = −1, and write
1
1 + 2 + · · · “ = ”ζ(−1) = − . (26.3)
12
Of course this does not make any “real" sense since 1 + 2 + 3 · · · is a divergent series and ζ(−1) is a
finite number since ζ(s) is holomorphic at s = −1. The equation (26.3) is found in one of Ramanujan’s
notebooks. Apparently Ramanujan had stumbled upon a way of summing up certain divergent series, and
being unaware of analytic continuation, used the rather crude notation that seems to suggest that the sum
of all natural numbers is not only a finite number but also negative!
There is a particularly misleading video posted by numpherphile, an otherwise decent youtube math chan-
nel, on this “astounding" identity - https://www.youtube.com/watch?v=w-I6XTVZXww,
which gives a “derivation" of the above “identity" using other misleading identities such as (26.1). Fol-
lowing the barrage of criticism that this video received, other channels made better videos. For instance
this video - https://www.youtube.com/watch?v=jcKRGpMiVTw by mathlogger clarifies the
identity using Cesaro summability. There is also a very beautiful video on the analytic continuation of the
zeta function by 3Blue1Brown - https://www.youtube.com/watch?v=sD0NjbwqlYw.
26.4. THE RIEMANN ZETA FUNCTION 167
We end this lecture, with the beautiful functional equation of Riemann’s which we state without proof.
Theorem 26.4.2. [Functional equation]
πs
ζ(s) = 2s π s−1 sin Γ(1 − s)ζ(1 − s).
2
Example 26.4.4. As an application we can calculate ζ(2). Namely,

πs
ζ(2) = 4πζ(−1) lim sin Γ(1 − s).
s→2 2
We can compute the limit directly by using the Residue of Γ(1 − s) at s = 2. Alternately, by Theorem 26.3.2
πs sin πs π
sin Γ(1 − s) = Γ(1 − s) = ,
2 2 cos(πs/2) 2Γ(s) cos πs/2
and so
π2 π2
ζ(2) = − = .
3Γ(2)(−2) 6
This gives a third derivation of the famous Basel-Euler identity:
∞
X 1 π2
= ,
n=1
n2 6
and is a good place to end.

.
Lecture 27
Prime numbers and the Riemann

hypothesis
The Riemann hypothesis is one of the most famous unsolved problems in all of mathematics. In the previous
chapter we saw that the zeta function has zeroes at all the negative even integers s = 0, −2, −4, · · · . Any
zero that is not one of these, is called the a non-trivial zero of the Riemann zeta function. The following is
Riemann’s conjecture:
Conjecture 27.0.1. All the non-trivial zeroes of the zeta function lie on the line
1
s(t) = + it.
2
To explain the significance of the Riemann hypothesis, we need to introduce the prime counting function.
Recall that a positive integer p is called a prime if it’s only positive factors are 1 and p. We then let
π(x) := number of primes less than or equal to x.
For x > 1, we also let

ˆ x ˆ 1−ε ˆ x
dt dt
Li(x) = p.v. := lim+ + .
0 log t ε→0 0 1+ε log t
Theorem 27.0.2. Riemann hypothesis implies the following estimate on π(x):

√
π(x) = Li(x) + O( x log x).
What does the zeta function have to do with the primes? The following observations goes back to Euler.
Indeed, what we call the Riemann zeta function, was first studied in detail by Euler himself.
Proposition 27.0.3. For ℜ(s) > 1,
Y 1
ζ(s) = .
p prime
1 − p−s
The proof is an elementary consequence of the unique factorisation theorem, and is left as an exercise.
169
170 LECTURE 27. PRIME NUMBERS AND THE RIEMANN HYPOTHESIS
27.1 The prime number theorem

Theorem 27.1.1.
π(x)
lim = 1.
x→∞ x log x
Lecture 28
Elliptic functions
28.1 Doubly periodic functions

We have already encountered many holomorphic functions that periodic. A function f : C → C is periodic
if there exists a ω ̸= 0 such that f (z + ω) = f (z) for all √
z ∈ C. Examples include most notably the
exponential function which is periodic with a period of 2π −1. In fact, as we saw earlier, this can be
taken as a definition of the real number π. In this chapter we will be interested in functions f : C → C
that are doubly periodic, that is,
f (z + ω1 ) = f (z) = f (z + ω2 )
for some ω1 , ω2 ∈ C∗ and all z ∈ C. At the end of this lecture, we shall give an application of such functions
to the evaluation of certain real valued integrals that arise in computing lengths of arcs on ellipses - hence
the name elliptic functions. But first, we note that the study of such doubly periodic functions falls naturally
into two cases - either ω2 = λω1 for some λ ∈ R or not. We claim that the first case is not very interesting.
Indeed we have the following:
Proposition 28.1.1. Let f be a meromorphic function on C. Suppose there exists ω ∈ C∗ and λ ∈ R∗ and
λ ̸= 1 such that
f (z + ω) = f (z + λω) = f (z), ∀z ∈ C,
then we have the following dichotomy:
1. If λ ∈ Q, then f is periodic with a single period.
2. If λ is irrational, then f is a constant.
In view of the above we restrict to the second case, that is, ω1 and ω2 are linearly independent as vectors in
R2 . It is then clear that any holomorphic function which is doubly periodic will be forced to be a constant
by Liouville’s theorem. Indeed the values that f takes are determined by the fundamental parallelogram of
f:
P0 = {aω1 + bω2 | 0 ≤ a < 1, 0 ≤ b < 1}.
So f itself is bounded, and hence constant. So the upshot is that to get something interesting, we need to
consider doubly periodic meromorphic functions whose periods are linearly independent. Such functions
are called elliptic functions. Any translate P = P0 + h, h ∈ C is called a period parallelogram of f .
171
172 LECTURE 28. ELLIPTIC FUNCTIONS
28.2 The canonical basis

Suppose now f is an elliptic function with bases ω1 and ω2 . Then clearly for any integers m and n,
mω1 + nω2 is a period for f . Let Λ = Zω1 + Zω2 be the lattice generated by ω1 and ω2 .
28.3 Poles and zeroes

Proposition 28.3.1. Let f be an elliptic function with fundamental period P0 = ⟨1, τ ⟩. Then the number
of poles and zeroes of f in P0 (counted with multiplicity) are equal. We call this number the order νf of the
elliptic function. Consequently, for any c ∈ C, the equation
f (z) = c
has exactly νf solutions when counted with multiplicity.
Proof. Without loss of generality (for instance by perturbing P0 to a neighbouring period parallelogram)
we may assume that there are no zeroes or poles on ∂P0 . If we denote the number of poles and zeroes by
N∞ and N0 , by the residue theorem we have
ˆ
1 f ′ (z)
dz = N0 − N∞ .
2πi ∂P0 f (z)
A natural question is what integers can appear as orders elliptic functions? We have the following elemen-
tary observation.
Proposition 28.3.2. For any elliptic function f , νf ≥ 2.
28.4 Weierstrass ℘ function

A natural question now arises on the existence of such functions. In this section we give an explicit (and
very classical) construction of an elliptic function of order two.
Lecture 29
Modular forms
29.1 The modular group
173
174 LECTURE 29. MODULAR FORMS
Part V
Analytic continuation
175
Lecture 30
Analytic continuation along curves
30.1 An example
Suppose we have an analytic function defined in some disc D, which we assume to be centred at z0 = 0.
Suppose we want to extend it to an analytic function at some z1 ∈ / D. A naive idea would be to join z1
to the origin by a path, say a straight line, and to attempt to extend the analytic function along the the
straight line by using power series expansions centred at points on the straight line. If one is lucky, this
procedure can be carried out until one reaches the desired point z1 . To illustrate this, consider analytic
function f : D → C on the standard unit disc:
∞
X
f (z) = zn.
n=0
Suppose we want to extend this to the point z = −1. We consider a point close to z = −1. In this case
z0 = −1/2 will work. We try to expand the function f (z) as a power series around z = −1/2. To do this,
we recall that whenever |z| < 1,
1
f (z) = .
1−z
In a neighbourhood of z0 = −1/2 we write this as
1
f (z) =
3/2 − (z + 1/2)
2 1
= ,
31−w
where w = 23 (z + 1/2). But then as long as |w| < 1, we can expand the above and obtain
∞ ∞
2 X n X 2 n+1 1 n
f (z) = w = z+ .
3 n=0 n=0
3 2
The right hand side is a power series which converges as long as

1 3
z+ < .
2 2
Since z = −1 satisfies the above inequality we have the required extension of our original function. This
is not surprising, since we already know that f (z) = (1 − z)−1 , and the latter function is clearly analytic
177
178 LECTURE 30. ANALYTIC CONTINUATION ALONG CURVES
at −1 (indeed it is analytic everywhere except z = 1). But one can imagine that this process holds good
for many other analytic functions.
On the other hand, note that there cannot be any such extension at z = 1. For if there were such an
extension, say F (z), then
1
F (1) = lim F (x + 0i) = lim = ∞.
x+0i→1 x→1 1−x
30.2 Formalizing the main ideas

30.3 The monodromy principle
30.4 A (very) surprising application - Picard’s little theorem
Having set-up the beautiful theory of modular forms, and in particular having constructed an elliptic mod-
ular function, we now combine with the monodromy principle to obtain an immediate and stunning con-
sequence. Recall that if f : C → C is an entire function, then f (C) is dense in C. This follows from the
Casorati-Weierstrass theorem by analyzing the singularity at infinity. We have the following vast gener-
alization.
Theorem 30.4.1 (Picard’s little theorem). Let f : C → C be a non-constant entire function. Then f (C) can
miss at most one point in C. That is, if there exist a ̸= b such that f (C) ⊂ C \ {a, b}, then f is a constant.
Of course we do have entire functions that miss exactly one point - the function ez for instance. So this
theorem is sharp.
Proof.
Lecture 31
An exposition on Riemann surfaces
179
180 LECTURE 31. AN EXPOSITION ON RIEMANN SURFACES
Appendix A
Problems
1. Find all possible solutions to the following equations:

(a) z 4 = i + 1.
(b) z n = 1.
2. Determine all values of 2i , ii and (−1)2i .
3. For what values of z is ez equal to 2, i/2?
4. Show that there are complex numbers z satisfying
|z − a| + |z + a| = 2|c|
if and only if |a| ≤ |c|. If this condition is satisfied, what are the smallest and largest values of |z|.
5. (a) Let z, w ∈ C such that z w̄ ̸= 1. Then prove that if |z|, |w| < 1, then
w−z
< 1.
1 − z w̄
Moreover
w−z
=1
1 − z w̄
if and only if either |z| = 1 or |w| = 1.
(b) Now, fix a w ∈ D := {z ∈ C | |z| < 1}, and consider the mapping ϕw : D → C,
w−z
ϕw (z) := .
1 − z w̄
Prove that ϕw has the following properties:
i. ϕw is a holomorphic map of D into itself.
ii. ϕw interchanges 0 and w. That is, ϕw (0) = w and ϕw (w) = 0.
iii. ϕw is a biholomorphism. Hint. Compute ϕw ◦ ϕw .
6. Prove that f (z) is holomorphic if and only if f (z̄) is holomorphic. How are the two complex deriva-
tives related?
7. Find the radius of convergence of the following power series.
181
182 APPENDIX A. PROBLEMS
P∞ 2 n
(a) n=1 (log n) z .
P∞ (n!)3 n
(b) n=1 (3n)! z .
(c) The Bessel function:

∞
z r X (−1)n z 2n
Jr (z) = .
2 n=0
n!(n + r)! 2
8. For what values of z ∈ C is the series

∞
X z n
n=0
1+z
convergent. In the regions that the series does converge, is it absolutely and/or uniformly conver-
gent?
p
9. Prove that the function f (x + iy) = |x||y| satisfies the Cauchy-Riemann equations at the origin,
and yet, using only the definition, prove that the function is not complex differentiable at 0.
10. Suppose f is holomorphic on a region Ω. If any one of the following holds:
(a) Re(f ) is a constant,
(b) Im(z) is a constant,
(c) |f | is a constant,
prove that f is a constant function.
11. Let Ω ⊂ C be open and f = u + iv : Ω → C be a smooth map.
(a) If h is a smooth map in a neighborhood of f (p), then prove that
∂h ◦ f ∂h ∂f ∂h ∂ f¯
(p) = (f (p)) · (p) + f (p) · (p)
∂z ∂w ∂z ∂ w̄ ∂z
∂h ◦ f ∂h ∂f ∂h ∂ f¯
(p) = (f (p)) · (p) + f (p) · (p)
∂ z̄ ∂w ∂ z̄ ∂ w̄ ∂ z̄
(b) If f is holomorphic, prove that det Jf (z) = |f ′ (z)|2 for all z ∈ Ω.

(c) Using the usual inverse function theorem from multivariable calculus, prove the following holo-
morphic inverse function theorem: If f is holomorphic at p with f ′ (p) ̸= 0, then there there
exists open sets U and V around p and f (p) respectively, such that f : U → V has a holomor-
phic inverse f −1 : V → U .
12. Let Ω ⊂ C be a region and f : Ω → C a holomorphic function satisfying |f (z) − 1| < 1 in Ω. Prove
that for any closed regular curve γ in Ω,
ˆ ′
f (z)
dz = 0.
γ f (z)
13. If P (z) is a polynomial and R > 0 prove that

ˆ
P (z)dz̄ = −2πiR2 P ′ (a).
|z−a|=R
14. (a) Let f (z) be an entire function such that Re(f (z)) > −a for all z ∈ C. Prove that f is a
constant.
183
(b) Let f (z) be an entire function such that f (z) is real whenever |z| = 1. Prove that f is a
constant.
(c) Let f (z) be an entire function such that |f (z)| ≤ |z|n . Prove that f (z) = cz n for some constant
c ∈ D.
(d) Let f be an entire function, and α, C > 0 be constants. Suppose
|f (z)| ≤ C(1 + |z|α )
for all z ∈ C. Prove that f (z) is a polynomial of degree at most ⌊α⌋, where ⌊α⌋ denotes the
greatest integer less than or equal to α.
(e) Find all entire functions such that |f (z)| > 1 whenever |z| > 1.
15. Compute the following integrals. The circles are traversed once in the anti-clockwise direction.
´
(a) |z|=1 ez z −n dz.
´ dz
(b) |z|=2 1+z 2.
´ |dz|
(c) |z|=r |z−a|2
, where |a| =
̸ r. Hint. First prove that |dz| = −irdz/z on the circle |z| = r.
16. Is there a holomorphic function f on a domain Ω such that there exists and integer N ∈ N and a
complex number z ∈ Ω such that |f (n) (z)| > n!nn for all n > N ? Can you formulate a sharper
theorem of the same kind?
17. Let Ω ⊂ C be open, and let f : Ω → Ω be a holomorphic function. Suppose there exists a point
z0 ∈ Ω such that
f (z0 ) = z0 , f ′ (z0 ) = 1.
Prove that f is linear, that is, there exists a, b ∈ Ω such that f (z) = az +b. Hint. First prove that one
can assume without loss of generality that z0 = 0. If f is not linear, then f (z) = z +an z n +O(z n+1 )
for some an ̸= 0 and n > 1. If fk denote f composed with itself k-times, prove that fk (z) =
z + kan z n + O(z n+1 ). Now, apply Cauchy inequalities and let k → ∞ to conclude the desired
result.
18. Suppose f : D := D1 (0) → C is holomorphic. Prove that the diameter of the image d :=
supz,w∈D |f (z) − f (w)| satisfies the estimate
d ≥ 2|f ′ (0)|,
and that equality holds if and only if the function is linear.

19. Let γ be a simple closed curve, and a ∈
/ Supp(γ). The prove that
(
±1, a ∈ int(γ)
n(γ, a) =
0, a ∈ ext(γ).
20. Are there holomorphic functions f (z) and g(z) in a neighbourhood of 0 such that for n = 1, 2 · · ·
we have
(a) f (1/n) = f (−1/n) = 1/n2 ,
(b) g(1/n) = g(−1/n) = 1/n3 ?
21. Determine all holomorphic functions on the unit disc D := D1 (0) such that
1 1
f ′′ +f = 0,
n n
for all n = 2, 3, · · · .
22. For any open set Ω ⊂ C and any complex valued function, the L2 -norm on U is defined to be
ˆ 1/2
||f ||L2 (Ω) := |f (x, y)|2 dx dy
Ω
if it is finite. We then define the space H(Ω) by
H(Ω) := {f ∈ O(Ω) | ||f ||L2 (Ω) < ∞}.
We equip it with the norm || · ||L2 (Ω) defined above.
(a) Let z0 ∈ Ω and Dr (z0 ) ⊂ Ω. For any 0 < s < r, prove that
1
sup |f (z)| ≤ √ ||f ||L2 (Dr (z0 )) .
z∈Ds (z0 ) π(r − s)
Hint. Write f (z) in terms of an integral on ∂Dr (z0 ) by the Cauchy integral formula and use
polar coordinates.
(b) Prove that if {fn } is a sequence in H(Ω) that is Cauchy with respect to the L2 (Ω) norm, then
fn → f compactly on Ω.
(c) Hence prove that H(Ω) with the metric
d(f, g) := ||f − g||L2 (Ω)
is a complete metric space.

23. Suppose that f and g are holomorphic in a region containing D1 (0). Suppose f has a simple zero at
0 (ie. the order is one), and has no other zero in D1 (0). Let
fε (z) = f (z) + εg(z).
Show that if ε is sufficiently small, then

(a) fε has a unique zero (counted with multiplicity) in D1 (0).
(b) Moreover, if that unique zero is pε , then ε → pε is a continuous function.
24. Find the branch points (including infinity) for the following functions. Also give a branch cut that
will make the function a single valued holomorphic function on the complement of the cut.
√
(a) z−1
(b) log (z 2 + z + 1)
25. (a) Let f : D → C such that the functions g = f 2 and h = f 3 are holomorphic on D. Prove that
f is holomorphic. Is the statement true if either g is not holomorphic or h is not holomorphic?
If so, give a proof. Else give counterexamples.
(b) Either prove or provide a counter-example to the following statement: If f is a continuous
function on a connected open subset Ω such that f 2 is holomorphic. Then so is f .
185
26. (a) Let QR be the rectangle with vertices (−R, 0), (R, 0), (R, R) and (−R, R). Compute the inte-
gral ˆ
dz
2 n+1
,
∂QR (1 + z )
where QR has the anti-clockwise orientation.

(b) Use this to prove that ˆ ∞
dx (2n)!
= n π.
−∞ (1 + x2 )n+1 4 (n!)2
27. In each of the cases below, classify the isolated singularities, and in case of poles, compute the order.
z 2 −π 2
(a) sin2 z
.
1−cos z
(b) sin z .
1 1
(c) ez −1 − z−2πi .
1
(d) cos(1/z) .
28. If f and g are entire functions such that |f (z)| < |g(z)| for |z| > 1, then show that f (z)/g(z) is a
rational function.
29. Let R(z) be a rational function such that |R(z)| = 1 for |z| = 1.
(a) Show that α is a zero or a pole of order m, if and only if 1/ᾱ is a pole or zero of order m
respectively. Hint. First show that
1
M (z) = R(z)R
z̄
is a rational function such that M (z) = 1 on |z| = 1.
(b) Let {αj }Nj=1 be zeroes and poles of R(z) of order mj in the unit disc |z| < 1. Here mj > 0 is
αj is a zero and mj < 0 is it is a pole. Define
z − α m1 z − α m2 z − α mN
1 2 N
B(z) = ··· .
1 − z ᾱ1 1 − z ᾱ2 1 − z ᾱN
Show that R(z) = λB(z) for some λ ∈ C with |λ| = 1
30. Recall that a function is said to have a removable singularity (resp. pole or essential singularity) at
infinity if the function f (1/z) has a removable singularity (resp. pole or essential singularity) at
z = 0.
(a) Show that an isolated singularity (including at infinity) of f (z) cannot be a pole for exp(f (z)).
(b) In particular, if f is a non-constant entire function, then exp(f (z)) has an essential singularity
at infinity.
31. Let f and g be entire functions such that h(z) = f (g(z)) is a non-constant polynomial. Prove that
both f (z) and g(z) are polynomials.
32. Show that when 0 < |z| < 4,
X zn ∞
1 1
= + .
4z − z 2 4z n=0 4n+2
Using the Laurent series, evaluate ˆ
1
dz,
|z|=2 4z − z 2
where the circle is given positive orientation.

33. Show that the Laurent series for (ez − 1)−1 at the origin takes the form
∞
1 1 X Bk 2k−1
− + (−1)k−1 z ,
z 2 n=1 (2k)!
where the numbers Bk are the called the Bernoulli numbers. Calculate B1 , B2 , B3 .
34. Find a Laurent series that converges in the annulus 1 < |z| < 2 to a branch of the function
z(2 − z)
f (z) = log .
1−z
35. Recall that we proved the identity

1 X 1 1
π cot πz = + + .
z z−n n
n∈Z\{0}
Recall also, the definition of Bernoulli numbers from the previous assignment
∞
z z X Bn 2n
= 1 − + (−1)n−1 z .
ez − 1 2 n=1 (2n)!
P∞
Finally, for any complex number Re(s) > 1, we define ζ(s) = m=1 m−s .
(a) Prove that
∞
X
πz cot πz = 1 − ζ(2n)z 2n .
n=1
(b) Prove that

Bn 2n
ζ(2n) = 22n−1 π ,
(2n)!
In particular, from your answers in the previous assignment, you should be able to calculate
ζ(2), ζ(4) and ζ(6). Hint. First observe that
2πiz
πz cot πz = πiz + .
e2πiz−1
36. Let Ω ⊂ C be an open set containing the closure Dr (0) of the disc of radius r centred at the origin.
Suppose f : Ω → C be is a holomorphic function with zeroes α1 , α2 , · · · , αn in Dr (0) with multi-
plicities m1 , · · · , mn respectively, and no zero on ∂Dr (0). For any entire function φ : C → C, show
that ˆ n
1 f ′ (z) X
φ(z) dz = mj φ(αj ).
2πi |z|=r f (z) j=1
37. Let f be a function that is holomorphic on the annulus AR,∞ (0). The residue of f (z) at infinity is
defined to be ˆ
1
Resz=∞ f (z) = − f (z) dz,
2πi |z|=r
where r > R. Note that by Cauchy’s theorem, the definition is independent of r. The reason for the
negative sign is that morally, one would like to define the residue at infinity, in the same way as for a
point in C, namely via an integral on a small circle around the point with positive orientation. But a
small circle with positive orientation around the point at infinity is a large circle in C with negative
orientation, and hence the negative sign in the above expression.
187
(a) Prove that

1 1
Resz=∞ f (z) = −Resz=0 f .
z2 z
(b) If f is holomorphic in C \ {p1 , · · · , pn }. Then prove that

n
X
Resz=∞ f (z) + Resz=pk f (z) = 0.
k=1
38. Show that if f is an injective entire function, then it must be linear. That is, f (z) = az + b, for some
a, b ∈ C with a ̸= 0. Hint. First show that f (z) cannot have an essential singularity at infinity.
39. Let f be a non-constant holomorphic map defined in an open set Ω containing the unit disc D centred
at the origin.
(a) Suppose |f (z)| = 1 whenever |z| = 1, then show that f (Ω) contains the unit disc.
(b) Show that if |f (z)| ≥ 1 whenever |z| = 1, and there exists a point z0 ∈ D such that |f (z0 )| < 1,
the prove that f (Ω) contains the unit disc.
40. Show that there is no holomorphic function on D that extends continuously to ∂D such that f (z) =
1/z for all z ∈ ∂D.
41. In each of the cases below, calculate the total number of solutions (with multiplicity) in the regions
indicated.
(a) z 7 − 2z 5 + 6z 3 − z + 1 = 0 in |z| < 1.
(b) cz n = ez , |c| > e in |z| < 1.
42. Compute the following real-variable integrals using the residue theorem.
´∞ x2
(a) 0 (x2 +a 2 )2 dx, a ∈ R.
´ ∞ x sin x
(b) 0 x2 +a2 dx, a ∈ R.
´ ∞ 1/3
(c) 0 xx2 +1 dx.
´ ∞ log x
(d) 0 1+x 2 dx.
´∞ dx
(e) 0 log(1 + x2 ) x1+α , 0 < α < 2.
43. (a) Find a fractional linear transformation that takes the points z1 = 2, z2 = i and z3 = −2 to
w1 = 1, w2 = i and w3 = −1.
(b) Find the fractional linear transformation that maps z1 = −i, z2 = 0 and z3 = i to w1 = −1,
w2 = i and w3 = 1. What curve does the x-axis transform into?
44. Find a conformal map from the wedge
π π
W = {z ∈ C | − ≤ arg z ≤ }
6 6
onto the unit disc D = {z | |z| ≤ 1}.
45. (a) Show that f (z) = − 12 (z + 1/z) is a conformal map from the half disc {z = reiθ | r < 1, 0 <
θ < π} to the upper half plane. Hint. The equation w = f (z) reduces to the quadratic
z 2 + 2wz + 1 = 0 which has two distinct roots if w ̸= ±1, which is certainly the case when w
lies in the upper half plane.
(b) Use this to show that g(z) = sin z maps the infinite strip {z = x + iy | − π/2 < x < π/2, y >
0} conformally onto the upper half plane.
46. If a Mobius transformation takes two concentric circles of radii r < R to two other concentric
circles with radii s < S. Prove that r/R = s/S. Hint. First argue that by pre-composing and post-
composing, one can assume that both the pairs of concentric circles have centres at 0 and s = r = 1,
and that the inner circle is sent to the inner circle. After this has been arranged, argue that 0 and ∞
will have to be fixed points.
47. (Schwarz-Pick theorem) If f : D → D is a holomorphic map from the unit disc to itself, prove the
following.
(a) For any z1 ̸= z2 ,
f (z1 ) − f (z2 ) z1 − z2
≤ .
1 − f (z1 )f (z2 ) 1 − z1 z2
(b) For any z ∈ D,

|f ′ (z)| 1
≤ .
1 − |z|2 1 − |z|2
(c) In either of the two inequalities, when (and only when) can equality occur?
48. Let H := {z = x + iy | y > 0} be the upper half plane, and let P SL(2, R) := SL(2, R)/ ± I be the
set of 2 × 2 real matrices modded out by the equivalence relation A ∼ −A. Then prove that Aut(H)
is isomorphic to P SL(2, R)
49. (a) Let fn : Ω → C be a sequence of holomorphic functions which converge compactly to f :
Ω → C which is not identically zero. If z0 is a root of f (z) of multiplicity ‘m’, then show that
for all sufficiently small ε > 0, there exists an N = N (ε) such that fn has exactly ‘m’ roots
(counting multiplicity) in Dε (z0 ) for all n > N .
(b) From this conclude that if fn : Ω → C is injective, and the sequence converges compactly to a
non constant function f : Ω → C, then f is injective.
(c) Let fn : D → D be a sequence of holomorphic functions such that fn (0) → 1. Then show
that fn → 1 compactly on D. Hint. First show that you get convergence after passing to a
subsequence.
50. For Re(s) > 0, consider the function
ˆ 1
ts
f (s) = √ dt,
0 1 − cos t
and let
f (s)
Λ(s) = .
Γ(s)
(a) Show that f (s) is holomorphic on Re(s) > 0 and can be extended to a meromorphic function
on all of C with simple poles at s = 0, −1, −2, · · · .
√
Hint. What is the behavior of 1 − cos t near t = 0. Use this to compare with the model case
discussed in the lecture.
(b) Find
Ress=−1 f (s).
189
(c) Use the identity

π
Γ(s)Γ(1 − s) =
sin πs
to show that the Gamma function has no zeroes in C. Caution: If Γ has a root at s, Γ(1 − s)
might have a pole at 1 − s. So you need to rule out this case.
(d) Show that Λ(s) is an entire function, and evaluate Λ(−1).
Bibliography
[1] Ahlfors, Lars, Complex Analysis, An introduction to the theory of analytic functions of one complex
variable. Third edition. International Series in Pure and Applied Mathematics. McGraw-Hill Book Co.,
New York, 1978. xi+331 pp.
[2] Stein, Elias and Shakarchi, Rami, Complex analysis. Princeton Lectures in Analysis, 2. Princeton
University Press, Princeton, NJ, 2003. xviii+379 pp.
191

CA Lectures

Uploaded by

CA Lectures

Uploaded by

Lectures on Complex Anal-

Indian Institute of Science

I Complex differentiable functions 1

2 Lecture-2: Limits, continuity and holomorphicity 9

4 The exponential functions 23

5 Cauchy Riemann equations 29

8 Cauchy’s theorem : Local versions 47

9 Cauchy’s integral formula 53

10 Further applications of Cauchy integral formula 57

11 Cauchy’s theorem : Homology version 63

12 Cauchy’s theorem : Multiply connected domains 69

13 Logarithm and roots revisited 75

15.1 Application to study of isolated singularities . . . . . . . . . . . . . . . . . . . . . . . . . . 92

17 The theorems of Mittag Leffler and Weierstrass 101

18 The Residue theorem 107

19 Applications of the argument principle 111

20 Contour Integration - I 117

21 Contour Integration - II 123

III Geometry of holomorphic maps 129

23 Mobius transformations 137

24 Some automorphism groups 143

24.4 Automorphism group of the extended complex plane . . . . . . . . . . . . . . . . . . . . . 147

25 The Riemann mapping theorem 149

IV Special functions 157

27 Prime numbers and the Riemann hypothesis 169

28 Elliptic functions 171

29 Modular forms 173

V Analytic continuation 175

31 An exposition on Riemann surfaces 179

Complex differentiable functions

1.1 Roadmap for the course

functions. For instance if ( 2

1.2 Complex numbers

More formally, we define the set of complex numbers by

x = ℜ(z) and y = ℑ(z).

P2 (Commutativity.) For any z1 , z2 ∈ C,

P3 (Associativity.) For any complex numbers z1 , z2 , z3 ,

(z1 + z2 ) + z3 = z1 + (z2 + z3 ), (z1 · z2 ) · z3 = z1 · (z2 · z3 ).

P4 (Distribution) For any z1 , z2 , z3 ∈ C,

P5 (Additive inverse.) For any z ∈ C, −z = (−1) · z satisfies

z = r(cos θ + i sin θ).

zw = rρ[(cos θ cos α − sin θ sin α) + i(cos θ sin α + sin θ cos α)]

1.2.1 Conjugate, Absolute value, Argument.

• zw = z̄ w̄, |zw| = |z||w|.

Proof. 1. We first note that if x, y, u, v ∈ R, then

So if we let z = x + iy and w = u + iv, then

Lecture-2: Limits, continuity and

2.1 Topology of the complex plane

DR (z0 ) = {z ∈ C | |z − z0 | < R}.

Proposition 2.1.2. zn → p if and only if ℜ(zn ) → ℜ(p) and ℑ(zn ) → ℑ(p).

This is a consequence of the fact that for any z ∈ C,

2.2 Functions on the complex plane

f (S) := {f (z) | z ∈ S},

|z − p| < δ, z ∈ S, =⇒ |f (z) − L| < ε.

We say that f is continuous at p ∈ S if

lim f (zn ) = f (p).

4. If f is continuous, then for any compact subset K ⊂ Ω, f (K) is compact.

z n − pn = (z − p)(z n−1 + z n−2 p · · · + pn−1 )

A polynomial is a function p : C → C of the form

p(z) = an z n + an−1 z n−1 + · · · + a1 z + a0 ,

2.2.1 Path connected sets in C.

Proof. Suppose Ω is a region. Fix z0 ∈ Ω, and let

A = {z ∈ Ω | there exists a path in Ω connecting z to z0 }.

2.2.2 Convergence of functions

n > N =⇒ |fn (z) − f (z)| < ε.

|fn (z) − f (z)| < ε.

Theorem 2.2.7. If fn : Ω → C is a sequence of continuous functions which converge uniformly to f : Ω → C,

2.3 Holomorphic functions

1. If f is constant in a neighbourhood of z, then f ′ (z) = 0.

2. (Linearity) af + bg is complex differentiable at z, and

[af + bg]′ (z) = af ′ (z) + bg ′ (z)

3. (Product rule) f g is complex differentiable at z and

(f g)′ (z) = f ′ (z)g(z) + f (z)g ′ (z)

4. (Quotient rule) If g ′ (z) ̸= 0, then f /g is complex differentiable at z and

5. (Chain rule) If h is complex differentiable at f (z), then h ◦ f is complex differentiable at z and

[h ◦ f ]′ (z) = h′ (f (z)) · f ′ (z).