0% found this document useful (0 votes)
92 views75 pages

COMPLEX ANALYSIS - Semester 2016-1: Robert Oeckl - CA NOTES - 01/12/2015

This document provides notes on complex analysis. It begins with an introduction to holomorphic functions, defining the complex derivative and discussing elementary properties like linearity and the Cauchy-Riemann equations. It then covers topics like the exponential function, power series, complex integration along paths, the Cauchy integral formula, and conformal mappings. The notes provide a comprehensive overview of the key concepts and results in complex analysis.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
92 views75 pages

COMPLEX ANALYSIS - Semester 2016-1: Robert Oeckl - CA NOTES - 01/12/2015

This document provides notes on complex analysis. It begins with an introduction to holomorphic functions, defining the complex derivative and discussing elementary properties like linearity and the Cauchy-Riemann equations. It then covers topics like the exponential function, power series, complex integration along paths, the Cauchy integral formula, and conformal mappings. The notes provide a comprehensive overview of the key concepts and results in complex analysis.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 75

Robert Oeckl – CA NOTES – 01/12/2015 1

COMPLEX ANALYSIS – Semester 2016-1

Contents
1 Holomorphic functions 3
1.1 The complex derivative . . . . . . . . . . . . . . . . . . . . . 3
1.2 Elementary properties of holomorphic functions . . . . . . . . 5
1.3 The exponential function . . . . . . . . . . . . . . . . . . . . 6
1.4 Power series and analytic functions . . . . . . . . . . . . . . . 9

2 Complex integration 15
2.1 Integration along paths . . . . . . . . . . . . . . . . . . . . . 15
2.2 Closed paths and winding . . . . . . . . . . . . . . . . . . . . 18
2.3 Integrable functions . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 The Cauchy Integral Formula . . . . . . . . . . . . . . . . . . 23
2.5 General Cauchy Theory . . . . . . . . . . . . . . . . . . . . . 27

3 Basic properties 31
3.1 From local to global structure . . . . . . . . . . . . . . . . . . 31
3.2 Zeros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Holomorphic logarithms and roots . . . . . . . . . . . . . . . 36

4 Singularities 39
4.1 Types of singularities . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Meromorphic functions . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Laurent Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5 Conformal mappings 49
5.1 Conformal mappings as holomorphic functions . . . . . . . . 49
5.2 The Open Mapping Theorem . . . . . . . . . . . . . . . . . . 50
5.3 Biholomorphic mappings . . . . . . . . . . . . . . . . . . . . . 51
5.4 Conformal automorphisms of C and C× . . . . . . . . . . . . 52
5.5 Conformal automorphisms of D . . . . . . . . . . . . . . . . . 55
5.6 Möbius Transformations . . . . . . . . . . . . . . . . . . . . . 56
5.7 Montel’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . 59
5.8 The Riemann Mapping Theorem . . . . . . . . . . . . . . . . 61
2 Robert Oeckl – CA NOTES – 01/12/2015

6 Harmonic functions 65
6.1 Mean value and maximum . . . . . . . . . . . . . . . . . . . . 65
6.2 The Dirichlet Problem . . . . . . . . . . . . . . . . . . . . . . 68

7 The Riemann Sphere 73


7.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.2 Functions on Ĉ . . . . . . . . . . . . . . . . . . . . . . . . . . 74
7.3 Functions onto Ĉ and Aut(Ĉ) . . . . . . . . . . . . . . . . . . 74
Robert Oeckl – CA NOTES – 01/12/2015 3

1 Holomorphic functions
1.1 The complex derivative
The basic objects of complex analysis are the holomorphic functions. These
are functions that posses a complex derivative. As we will see this is quite a
strong requirement and will allow us to make far reaching statements about
this type of functions. To properly understand the concept of a complex
derivative, let us recall first the concept of derivative in Rn .
Definition 1.1. Let U be an open set in Rn and A : U → Rm a function.
Given x ∈ U we say that A is (totally) differentiable at x iff there exists an
m × n-matrix A′ such that,
A(x + ξ) = A(x) + A′ ξ + o(∥ξ∥)
for ξ ∈ Rn sufficiently small. Then, A′ is called the derivative of A at x.
Recall that the matrix elements of A′ are the partial derivatives
∂Ai
A′ij = .
∂xj
Going from the real to the complex numbers, we can simply use the de-
composition z = x + iy of a complex number z into a pair of real numbers
(x, y) to define a concept of derivative. Thus, let U be an open set in C and
consider a function f : U → C. We view U as an open set in R2 with coordi-
nates (x, y) and f = u + iv as a function with values in R2 with coordinates
(u, v). The total derivative of f , if it exists, is then a 2 × 2-matrix f ′ given
by ( )
∂u ∂u
f′ = ∂x
∂v
∂y
∂v .
∂x ∂y
So far we have only recited concepts from real analysis and not made use
of the fact that the complex numbers do not merely form a 2-dimensional
real vector space, but a field. Indeed, this implies that there are special
2 × 2-matrices, namely those that correspond to multiplication by a complex
number. As is easy to see, multiplication by a+ib corresponds to the matrix,
( )
a −b
.
b a
The crucial step that leads us from real to complex analysis is now the
additional requirement that the derivative f ′ take this form. It is then more
useful to think of f ′ as the complex number a + ib, rather than this 2 × 2-
matrix.
4 Robert Oeckl – CA NOTES – 01/12/2015

Definition 1.2. Let U be an open set in C and f : U → C a function.


Given z ∈ U we say that f is complex differentiable at z iff there exists
f ′ (z) ∈ C such that,
f (z + ζ) = f (z) + f ′ (z)ζ + o(|ζ|)
for ζ ∈ C sufficiently small. Then, f ′ (z) is called the complex derivative of f
at z. f is called holomorphic at z iff f is complex differentiable in an open
neighborhood of z.
Proposition 1.3. Let U be an open set in C and f : U → C a function. f
is complex differentiable at z ∈ U iff f is totally differentiable at z and its
partial derivatives at z satisfy the Cauchy-Riemann equations,
∂u ∂v ∂u ∂v
= , =− .
∂x ∂y ∂y ∂x
If U ⊆ C is open we say that f : U → C is holomorphic on U if it
is holomorphic at all z ∈ U . We denote the space of functions that are
holomorphic on U by O(U ). In the following, non-empty connected open
subsets of the complex plane will be of particular importance. We will refer
to such open sets as regions. Since any non-empty open set in the complex
plane is a disjoint union of regions it is sufficient to consider the spaces of
holomorphic functions of the type O(D), where D ⊆ C is a region. The
elements of O(C) are called entire functions.
Exercise 1. Let U be an open set in C and f : U → C a function. Given
z ∈ U we say that f is complex conjugate differentiable at z iff there exists
fz (z) ∈ C such that,
f (z + ζ) = f (z) + fz (z)ζ + o(|ζ|)
for ζ ∈ C sufficiently small. Then, fz (z) is called the complex conjugate
derivative of f at z. f is called anti-holomorphic at z iff f is complex
conjugate differentiable in an open neighborhood of z.
1. Show that the total derivative of f as a real 2 × 2-matrix takes the
form ( )
a b
, for a, b ∈ R,
b −a
where f is complex conjugate differentiable.
2. Deduce the corresponding modified Cauchy-Riemann equations.
3. Show that a function is anti-holomorphic iff it is the complex conjugate
of a holomorphic function.
Robert Oeckl – CA NOTES – 01/12/2015 5

1.2 Elementary properties of holomorphic functions


Proposition 1.4. Let D ⊆ C be a region and f ∈ O(D). Then, f is
constant iff f ′ (z) = 0 for all z ∈ D.

Proof. If f is constant it follows immediately that f ′ = 0. Conversely,


suppose that f ′ = 0. Then, viewing f as a function from an open set D in
R2 to R2 we know that its total derivative is zero. By results of real analysis
it follows that f is constant along any path in D. But since D is connected
it is also path connected and f must be constant on D.

Proposition 1.5. Let D ⊆ C be a region.

1. If f ∈ O(D) and λ ∈ C, then λf ∈ O(D) and (λf )′ (z) = λf ′ (z).

2. If f, g ∈ O(D), then f + g ∈ O(D) and (f + g)′ (z) = f ′ (z) + g ′ (z).

3. If f, g ∈ O(D), then f g ∈ O(D) with (f g)′ (z) = f ′ (z)g(z) + f (z)g ′ (z).

4. If f, g ∈ O(D) and g(z) ̸= 0 for all z ∈ D, then f /g ∈ O(D) and

f ′ (z)g(z) − f (z)g ′ (z)


(f /g)′ (z) = .
(g(z))2

Proof. The proofs are completely analogous to those for real functions on
open subsets of the real line with the ordinary real differential. Alternatively,
1.-3. follow from statements in real analysis by viewing C as R2 .

Note that items 1.-3. imply that O(D) is an algebra over the complex
numbers.

Proposition 1.6. Let D1 , D2 ⊆ C be regions. Let f ∈ O(D1 ) such that


f (D1 ) ⊆ D2 and let g ∈ O(D2 ). Then g ◦ f ∈ O(D1 ) and moreover the
chain rule applies,

(g ◦ f )′ (z) = g ′ (f (z))f ′ (z) ∀z ∈ D1 .

Proof. This is again a result of real analysis, obtained by viewing C as R2 .


(Note that g ′ and f ′ are then 2 × 2-matrices whose multiplication translates
to multiplication of complex numbers here.)

Proposition 1.7. Let D1 , D2 ⊆ C be regions. Let f : D1 → C be continuous


and such that f (D1 ) ⊆ D2 . Let g ∈ O(D2 ) be such that g ◦ f (z) = z for all
6 Robert Oeckl – CA NOTES – 01/12/2015

z ∈ D1 . Let z ∈ D1 . Suppose that g ′ (f (z)) ̸= 0 and that g ′ is continuous at


f (z). Then, f is complex differentiable at z and
1
f ′ (z) = .
g ′ (f (z))

Proof. Again, this is a statement imported from real analysis on R2 . (There,


the condition g ′ (f (z)) ̸= 0 is the condition that the determinant of the 2 × 2-
matrix g ′ (f (z)) does not vanish.)

A few elementary examples together with the properties of holomorphic


functions we have identified so far already allow us to generate considerable
families of holomorphic functions.
Example 1.8. The following are elementary entire functions.
• The constant functions: They have vanishing complex derivative.

• The identity function: f (z) = z has complex derivative f ′ (z) = 1.


Example 1.9. The following are (classes of) holomorphic functions pro-
duced from the elementary entire functions of Example 1.8 by addition,
multiplication and division.

• Polynomials: Any polynomial p(z) = n λn z
n, where λn ∈ C, is entire

with p′ (z) = n̸=0 λn nz n−1 .

• Rational functions: Let p(z) and q(z) be polynomials with q ̸= 0 and


suppose that p and q have no common zeros. Let D = C \ N , where
N is the set of zeros of q. Then, f (z) = p(z)/q(z) ∈ O(D).

1.3 The exponential function


The most important example of a transcendental entire function is the com-
plex exponential function.
Definition 1.10. We define the complex exponential function exp : C → C
as follows. For all a, b ∈ R define

exp(a + ib) := exp(a) (cos(b) + i sin(b)) ,

where exp, cos and sin are the functions known from real analysis.
Proposition 1.11. The complex exponential function has the following
properties:
Robert Oeckl – CA NOTES – 01/12/2015 7

1. exp is entire.

2. exp′ (z) = exp(z) for all z ∈ C.

3. exp(−z) = 1/ exp(z) for all z ∈ C.

4. exp(z + 2πin) = exp(z) for all z ∈ C and all n ∈ Z.

5. exp(C) = C \ {0}.

6. For each z ∈ C \ {0} there is a unique angle θ ∈ [0, 2π), called the
argument or phase, so that z = |z| exp(iθ).

Proof. Exercise.[Suggestion: Use properties of the functions exp, cos and


sin defined on the real numbers.]

Proposition 1.12. Let D ⊆ C be a region and f ∈ O(D). Then, the


following statements are equivalent:

1. f (z) = a exp(bz) for all z ∈ D, where a, b ∈ C are constants.

2. f ′ (z) = bf (z) for all z ∈ D, where b ∈ C is a constant.

Proof. The implication 1. =⇒ 2. is straightforward using elementary prop-


erties of the derivative (Propositions 1.5 and 1.6) together with Proposi-
tion 1.11.2. For the implication 2. =⇒ 1. consider the holomorphic function
g : D → C given by g(z) := f (z) exp(−bz) for all z ∈ D. Then, g ′ = 0,
so by Proposition 1.4 there exists a constant a ∈ C such that g(z) = a for
all z ∈ D. But since exp(−bz) = 1/ exp(bz) due to Proposition 1.11.3, we
obtain 1. as desired.

Remark 1.13. This Proposition shows in particular that the complex ex-
ponential function is uniquely determined by the properties exp′ = exp and
exp(0) = 1.

Proposition 1.14 (Addition Theorem).

exp(z + ζ) = exp(z) exp(ζ) ∀z, ζ ∈ C.

Proof. Fix ζ ∈ C. Then, f (z) := exp(z + ζ) is holomorphic for z ∈ C and


f ′ (z) = f (z). So f (z) = a exp(z) for some a ∈ C by Proposition 1.12. Since
f (0) = exp(ζ) = a we obtain the stated result.

Proposition 1.14 motivates us to use the notation ez := exp(z) as in the


real case.
8 Robert Oeckl – CA NOTES – 01/12/2015

Remark 1.15. It might seem somewhat unsatisfactory to define the com-


plex exponential function by recurrence to transcendental functions from
real analysis. Indeed, one could instead start from a definition in terms of
a power series. One can then derive properties 1.,2.,3. of Proposition 1.11
and consequently Propositions 1.12 and 1.14 from properties of this power
series. We come back to the power series of the exponential function in
Proposition 1.21.
Example 1.16. The following are transcendental entire functions produced
using the exponential function.
• Hyperbolic functions:
exp(z) + exp(−z) exp(z) − exp(−z)
cosh(z) := and sinh(z) := .
2 2
• Trigonometric functions:
exp(iz) + exp(−iz) exp(iz) − exp(−iz)
cos(z) := and sin(z) := .
2 2i

Example 1.17 (The logarithm). Since exp(z + 2πi) = exp(z) we have to


restrict the domain of exp in order to find a unique inverse. It is customary
to make the following choice: Consider the region D2 := R + i(−π, π). Then
exp is a bijective function D2 → D1 , where D1 = C \ R− 0 . We define log as
the unique function such that exp(log(z)) = z for all z ∈ D1 and such that
the image of log lies in D2 ⊆ C. Then, log ∈ O(D1 ) and log′ (z) = 1/z for
all z ∈ D1 . This version of the logarithm is also called the principal branch.
Exercise 2. Suppose f is a holomorphic function on a region D ⊆ C.
Suppose that the real or the imaginary part of f is constant. Show that f
must be constant on D.
Exercise 3. At which points in the complex plane are the following func-
tions complex differentiable and at which points are they holomorphic?
1. f (x + iy) = x4 y 5 + ixy 3

2. f (x + iy) = sin2 (x + y) + i cos2 (x + y)


Exercise 4. Define another version (“branch”) of the logarithm function
that is holomorphic in the region D = C \ R+
0.
sinz
Exercise 5. Define tan z := cos z . Where is this function defined and where
is it holomorphic?
Robert Oeckl – CA NOTES – 01/12/2015 9


Exercise 6. Define a function z 7→ z on C or on a subset of C. Is this
function holomorphic and if yes, where? Comment on possible choices in
the construction.

1.4 Power series and analytic functions


With each sequence {cn }n∈N of complex numbers and each point z0 ∈ C we
can associate a power series


f (z) = cn (z − z0 )n ,
n=0

around z0 . Recall the following result from real analysis.


Lemma 1.18. The radius of convergence r of the power series is given by
1
= lim sup |cn |1/n .
r n→∞

That is, the power series converges absolutely in the open disk Br (z0 ) to a
complex function f : Br (z0 ) → C. For any 0 < ρ < r the convergence is
uniform in the open disk Bρ (z0 ). It diverges for z outside of the closed disk
Br (z0 ).
Proof. Exercise.

Definition 1.19. Let D ⊆ C be a region and f : D → C. We say that f is


analytic in D iff for every point z ∈ D and any r > 0 such that Br (z) ⊆ D
the function f can be expressed as a power series around z with radius of
convergence greater or equal to r.
Theorem 1.20. Let D ⊆ C be a region. Suppose that f is analytic in D.
Then f ∈ O(D) and f (k) is also analytic in D. Moreover, if


f (z) = cn (z − z0 )n , (1)
n=0

converges in Br (z0 ), then



∑ (n + k)!
f (k) (z) = cn+k (z − z0 )n , (2)
n=0
n!

converges in Br (z0 ). Moreover, the coefficients cn of the power series around


a given point are unique.
10 Robert Oeckl – CA NOTES – 01/12/2015

Proof. (Adapted from Rudin.) Fix z0 ∈ D and r > 0 such that Br (z0 ) ⊆ D.
Suppose f is given by the power series (1) and converges in Br (z0 ). Consider
the power series


g(z) := (n + 1)cn+1 (z − z0 )n .
n=0

It is then enough to show that g(z) converges in Br (z0 ) and that g(z) is the
complex derivative of f for all z ∈ Br (z0 ). The statement (2) about the k-th
derivative follows then by iteration.
Firstly, it is clear by Lemma 1.18 that g(z) has the same radius of con-
vergence as f (z). In particular, g(z) converges in Br (z0 ). Fix z ∈ Br (z0 )
and define ξ := z − z0 . Then, set ρ arbitrarily such that |ξ| < ρ < r. Let
ζ ∈ Bs (0) \ {0} where s := ρ − |ξ| and set

f (z + ζ) − f (z)
h(ζ) := − g(z).
ζ

We have to show that h(ζ) → 0 when |ζ| → 0. h(ζ) can be written as




h(ζ) = cn an (ζ),
n=0

where
(ξ + ζ)n − ξ n
an (ζ) := − nξ n−1 .
ζ
Note that a0 (ζ) = 0 and a1 (ζ) = 0. By explicit computation we find for
n ≥ 2,

n−1
an (ζ) = ζ kξ k−1 (ξ + ζ)n−k−1 .
k=1

Now, |ξ| < ρ and |ξ + ζ| < ρ so that we get the estimate,

1
|an (ζ)| < |ζ| n(n − 1)ρn−2 .
2
This implies

1∑
|h(ζ)| < |ζ| |cn |n(n − 1)ρn−2 .
2 n=2
However, since ρ < r, the sum converges by Lemma 1.18 showing that there
is a constant M such that
|h(ζ)| < |ζ|M.
Robert Oeckl – CA NOTES – 01/12/2015 11

This completes the proof of (2). Finally , the uniqueness of the coefficients
cn follows from the special case of (2) given by
f (k) (z0 ) = k!ck .

The first remarkable result of complex analysis is that the converse of


this theorem is also valid: Every holomorphic function is analytic. However,
in order to show this we will have to introduce the integral calculus in the
complex plane. We will do this in the next chapter.
Proposition 1.21. The exponential function is analytic in C and has a
power series representation given as follows:

∑ 1
exp(z) = zn.
n=0
n!
Proof. Exercise.
Lemma 1.22. Let n ∈ Z. Then,
∫ {
2π 2π if n=0
inθ
e dθ =
0 0 if n ̸= 0
Proof. Exercise.
Lemma 1.23. Let z0 ∈ C and r > 0, and suppose the power series


f (z) = cn (z − z0 )n
n=0
has radius of convergence greater than r. Then,
∫ 2π
1
cn = f (z0 + reiθ )e−inθ dθ.
2πrn 0

Proof. By Lemma 1.18 the power series converges uniformly in Br (z0 ). We


can thus in the following interchange integration and summation,
∫ 2π
f (z0 + reiθ )e−inθ dθ
0
∫ ∞
2π ∑
= ck rk ei(k−n)θ dθ
0 k=0

∑ ∫ 2π
= ck rk ei(k−n)θ dθ
k=0 0

= 2πcn rn .
12 Robert Oeckl – CA NOTES – 01/12/2015

Here we have used Lemma 1.22.

Lemma 1.24 (Gutzmer Formula). Let z0 ∈ C and r > 0, and suppose the
power series


f (z) = cn (z − z0 )n
n=0

has radius of convergence greater than r. Set M := supz∈∂Br (z0 ) |f (z)|.


Then,

∑∞
1 2π 2

|cn | r =
2 2n
f (z0 + reiθ ) dθ ≤ M 2 .
n=0
2π 0

Proof. Since


f (z0 + reiθ ) = cn rn e−inθ
n=0

we have,
2 ∞


f (z0 + reiθ ) = cn rn f (z0 + reiθ )e−inθ ,
n=0

where the series converges uniformly as a series of functions on the interval


θ ∈ [0, 2π]. Thus, we can interchange integration and summation in the
following and use Lemma 1.23 to obtain,
∫ 2π 2 ∞
∑ ∫ 2π ∞


f (z0 + re ) dθ = cn r n
f (z0 + reiθ )e−inθ dθ = 2π |cn |2 r2n .
0 n=0 0 n=0

This shows the claimed equality. The stated inequality is obtained by esti-
mating the integral through the maximum of its integrand.

Proposition 1.25 (Cauchy’s Estimates). Let D ⊆ C be a region, f :


D → C analytic, z0 ∈ D and r > 0 such that Br (z0 ) ⊂ D. Set M :=
supz∈∂Br (z0 ) |f (z)|. Then,

n!M
|f (n) (z0 )| ≤ .
rn
Proof. Exercise.

Theorem 1.26 (Liouville Theorem). Every bounded function analytic in C


is constant.
Robert Oeckl – CA NOTES – 01/12/2015 13

Proof. Let f : C → C be analytic in C and be bounded by N , i.e., |f (z)| ≤ N



for all z ∈ C. Since f is analytic in C and its power series f (z) = ∞n=0 cn z
n

around 0 has infinite radius of convergence. Thus, for a radius r > 0 we


have from Lemma 1.24 the estimate,


|cn |2 r2n ≤ M 2 ≤ N 2 .
n=0

Since r can be arbitrarily large, this implies ck = 0 for all k ∈ N.

Exercise 7. Let a, b, c, d ∈ C such that c ̸= 0 and ad − bc ̸= 0. Show that


cz+d is analytic in D := C \ {− c }.
f (z) := az+b d

Exercise 8. Suppose f : C → C is analytic in C and satisfies

|f (z)| ≤ a + b|z|c ∀z ∈ C,

where a, b, c are positive constants. Show that f is a polynomial of degree


less than or equal to c.

Exercise 9. Let f : C → C be analytic in C. Show that the power series of


f at 0 converges uniformly in all of C if an only if f is a polynomial.
14 Robert Oeckl – CA NOTES – 01/12/2015
Robert Oeckl – CA NOTES – 01/12/2015 15

2 Complex integration
2.1 Integration along paths
Definition 2.1. Let I = [a, b] ⊂ R be a compact interval. A continuous
map γ : I → C is called a curve in C. We denote the image of the curve
γ by |γ|. If γ(a) = γ(b), the curve is called closed. A curve γ : I → C
is called a path iff it is piecewise continuously differentiable. That is, there
exist points a = x0 < x1 < · · · < xn = b such that γ restricted to [xk−1 , xk ]
is continuously differentiable for all k ∈ {1, . . . , n}.

Recall that continuous differentiability in a closed interval [a, b] means


differentiability in (a, b) such that the differential is continuous and has a
continuous extension to [a, b].
For the theory of integration along paths what is important in a path is
its image in and in which direction this is retraced. In contrast, the concrete
parametrization of a path via an interval I ⊂ R is not important. To make
this more precise we define the concept of reparametrization of a path.

Definition 2.2. Let γ : [a, b] → C and γ̃ : [ã, b̃] → C be paths. We say that
γ̃ is a reparametrization of γ iff there exists a monotonous, continuous and
piecewise continuously differentiable map ϕ : [ã, b̃] → [a, b] with ϕ(ã) = a
and ϕ(b̃) = b and such that γ̃ = γ ◦ ϕ.

We will be interested only in properties and usages of paths that are


invariant under reparametrization. The first such property we consider is
the length of a path. Intuitively it is quite clear what we mean by this. If a
path γ : [a, b] → C is a straight line

(b − t)x1 + (t − a)x2
γ(t) :=
b−a

with end points x1 and x2 , then its length should be |x2 − x1 | where we use
the standard Euclidean inner product on C. In general, we can approximate
a path by subdividing the interval on which it is defined and replacing the
pieces of paths in subdivisions by straight lines. The length of the path
should then be the limit of the sum of the lengths of these straight lines
when we make the subdivisions arbitrarily fine. That this limit exists is
due to the piecewise continuous differentiability property we have imposed.
(The limit does not necessarily exist for arbitrary curves, even if their image
is bounded.) The result is the following, which we state as a definition.
16 Robert Oeckl – CA NOTES – 01/12/2015

Definition 2.3. Let γ : [a, b] → C be a path. The length of γ, denoted l(γ)


is defined by,
∫ b
l(γ) := |γ ′ (t)| dt.
a

Exercise 10. (a) Show that the definition indeed agrees with the result of
the procedure described above. (b) Give an example of a curve that has
bounded image, but no well defined length.
Exercise 11. Show that the length of a path is invariant under reparametriza-
tion. That is, show that if γ is a path and γ̃ is a reparametrization of γ,
then l(γ) = l(γ̃).
Definition 2.4. Let U ⊆ C be open and f : U → C be a continuous map.
Let γ : I → C be a path such that |γ| ⊂ U . We define the complex integral
of f along the path γ as follows,
∫ ∫ b
f (z) dz := f (γ(t))γ ′ (t) dt. (3)
γ a

To make sense of this definition we note that t 7→ f (γ(t))γ ′ (t) is a piece-


wise continuous function I → C and is therefore bounded and integrable.
Proposition 2.5. The complex integral is invariant under reparametriza-
tions: Given an open set U ⊆ C, a continuous function f : U → C, a path
γ with |γ| ⊂ U and a reparametrization γ̃ of γ. Then,
∫ ∫
f (z) dz = f (z) dz.
γ̃ γ

Proof. Exercise.

Similarly to what we have seen in the context of the concept of derivative,


the concept of integration introduced is quite similar to what we are familiar
with in the case of R or Rn . Nevertheless, again, there is an important
difference that makes crucial use of the fact that the complex numbers form
a field. If we were to discuss integration along paths in R2 weighted by
path length, the formula to use would be almost identical to (3), with one
important difference: γ ′ would be a 2 × 1-matrix and we would insert |γ ′ (t)|
instead of γ ′ (t) on the right hand side. Decomposing γ ′ = reiθ the difference
is that in the real case we would only put the absolute value r. We might
think of the complex case as letting the direction of the curve (encoded in
θ) enter the integrand. As we shall see, this leads to a remarkable interplay
between complex integral and derivative.
Robert Oeckl – CA NOTES – 01/12/2015 17

Suppose γ1 : [a, b] → C and γ2 : [b, c] → C are paths such that γ1 (b) =


γ2 (b). Then, we can form the composite path γ1 · γ2 : [a, c] → C in the
obvious way. We have then,
∫ ∫ ∫
f (z) dz = f (z) dz + f (z) dz.
γ1 ·γ2 γ1 γ2

Because of Proposition 2.5 we are usually interested in paths only up to


reparametrization. That is, we consider two paths as equivalent if one is a
reparametrization of the other. We may then talk about the composition of
paths whenever the endpoint of the first coincides with the initial point of
the second.
Given a path γ : [0, 1] → C we may form the opposite path γ −1 : [0, 1] →
C given by γ −1 (t) = γ(1 − t). Then clearly, (γ −1 )−1 = γ. As is easy to see,
∫ ∫
f (z) dz = − f (z) dz.
γ −1 γ

We also find that the integral of any function along γ · γ −1 vanishes. γ · γ −1


is called a retracing. Because the integral along a retracing vanishes, we
consider a retracing as equivalent to the trivial path.

Exercise 12. The concept of reparametrization can be generalized to in-


clude some form of retracing. To this end remove the monotonicity condition
from Definition 2.2. (a) Is the length of a path invariant under generalized
reparametrization? (b) Is the complex integral along a path invariant under
generalized reparametrization?

Exercise 13 (Transformation rule). Prove the following Proposition: Let


D ⊆ C be a region, g ∈ O(D) such that g ′ : D → C is continuous and γ a
path with |γ| ⊂ D. Then, g ◦ γ is a path and for any continuous function
f : U → C where U ⊆ C is open and |g ◦ γ| ⊂ U we have,
∫ ∫
f (z) dz = f (g(z))g ′ (z) dz.
g◦γ γ

Proposition 2.6. Let U ⊆ C be open, f : U → C continuous, γ be a path


with |γ| ⊂ U . Set ∥f ∥γ := supz∈|γ| |f (z)|. Then,


f (z) dz ≤ ∥f ∥γ l(γ).

γ

Proof. Exercise.
18 Robert Oeckl – CA NOTES – 01/12/2015

Proposition 2.7. Let U ⊆ C be open and {fn }n∈N a sequence of continuous


functions fn : U → C converging uniformly. Let γ be a path in U . Then,
∫ ∫
lim fn (z) dz = lim fn (z) dz.
n→∞ γ γ n→∞

Proof. Exercise.[Hint: Use Proposition 2.6.]

2.2 Closed paths and winding


Definition 2.8. Let γ be a closed path. Let z ∈ C \ |γ| and define the index
of z with respect to γ as,

1 1
Indγ (z) = dζ.
2πi γ ζ −z
Theorem 2.9. Let γ be a closed path and U := C \ |γ|. Then, Indγ (z) ∈ Z
for all z ∈ U . Moreover, Indγ (z) = Indγ (z ′ ) if z and z ′ are in the same
connected component of U . Also, Indγ (z) = 0 if |z| is sufficiently large.
Proof. Parametrizing γ : [a, b] → C we have,

1 b γ ′ (t)
Indγ (z) = dt.
2πi a γ(t) − z
In order to show that Indγ (z) ∈ Z we define ϕ : [a, b] → C via,
(∫ )
t γ ′ (s)
ϕ(t) := exp ds .
a γ(s) − z
It is then sufficient to show that ϕ(b) = 1, which we proceed to do. Observe
that ϕ is continuous and piecewise continuously differentiable with piecewise
differential
ϕ(t)γ ′ (t)
ϕ′ (t) = .
γ(t) − z
The quotient function t 7→ ϕ(t)/(γ(t) − z) is also continuous and piecewise
continuously differentiable with piecewise differential given by,
( )′
ϕ(t)
= 0.
γ(t) − z
Thus, this function is piecewise constant and continuous. So it must be
constant on the connected set [a, b]. Equating its value at a with its value
at b yields,
γ(b) − z
ϕ(b) = ϕ(a) = 1,
γ(a) − z
Robert Oeckl – CA NOTES – 01/12/2015 19

since ϕ(a) = exp(0) = 1 and γ is closed.


Exercise.Show that Indγ (z) = Indγ (z ′ ) if z and z ′ are in the same con-
nected component of U . [Hint: Show first that Indγ : U → C is continuous.]
It remains to show that Indγ (z) = 0 if z is sufficiently large. Let M :=
supt∈[a,b] |γ(t)|. Then, if |z| > M + l(γ) we have, using Proposition 2.6,

1 l(γ)
|Indγ (z)| ≤ < 1.
2π |z| − M

On the other hand Indγ (z) ∈ Z, so we must have in this case Indγ (z) = 0.
This completes the proof.

Exercise 14. Let γ : [0, 1] → C be the path γ(t) := z0 + re2πikt with z0 ∈ C


and r > 0 and k ∈ Z. Show that Indγ (z) = k if z ∈ Br (z0 ) and Indγ (z) = 0
if z ∈ C \ Br (z0 ).

Let B be an open disk in C. We denote by ∂B its boundary, i.e., ∂B =


B \ B. We also denote by ∂B a closed path that traces the boundary ∂B
once with positive (anti-clockwise) direction. If B has center z0 and radius r,
the path ∂B can be represented by the corresponding path γ of Exercise 14
with k = 1.

Lemma 2.10. Let z ∈ C, r > 0 and k ∈ Z. Then,


∫ {
1 1 if k = −1
(ζ − z) dζ =
k
.
2πi ∂Br (z) 0 if k ∈ Z \ {−1}

Proof. Exercise.

2.3 Integrable functions


Definition 2.11. Let D ⊆ C be a region and f : D → C. If F ∈ O(D)
such that F ′ = f , then F is called a primitive of f . f is called integrable in
D if there exists such a primitive.

Theorem 2.12. Let D ⊆ C be a region, f : D → C be continuous and


F : D → C. Then, F is a primitive of f iff for every path γ : [a, b] → D

f (z) dz = F (γ(b)) − F (γ(a)).
γ
20 Robert Oeckl – CA NOTES – 01/12/2015

Proof. Suppose F is a primitive of f . Assume without loss of generality that


γ is continuously differentiable everywhere. Then, using the chain rule,
∫ ∫ b ∫ b
′ ′
f (z) dz = F (γ(t))γ (t) dt = (F ◦ γ)′ (t) dt = F (γ(b)) − F (γ(a)).
γ a a

Conversely, suppose that F satisfies the stated formula for every path γ
in D. Let z ∈ D and choose r > 0 such that Br (z) ⊆ D. For ξ ∈ Br (0) let
γξ : [0, 1] → C be the path γξ (t) := z + tξ. By assumption,
∫ ∫ 1
F (z + ξ) − F (z) = f (ζ) dζ = f (z + tξ)ξ dt
γξ 0

For ξ ̸= 0 we get,

F (z + ξ) − F (z) 1
= f (z + tξ) dt.
ξ 0

The right hand side of this expression converges to f (z) when |ξ| → 0 since,
(∫ ) ∫
1 1
f (z + tξ) dt − f (z) ≤ |f (z + tξ) − f (z)| dt

0 0
≤ sup |f (z + ζ) − f (z)|,
ζ∈B|ξ| (0)

where the right hand side expression converges to zero for |ξ| → 0 by conti-
nuity of f . Thus, F is complex differentiable at z with the differential being
F ′ (z) = f (z). This completes the proof.

Proposition 2.13. Let D ⊆ C be a region and f : D → C be continuous.


Then, f is integrable in D iff for every closed path γ in D we have:

f (z) dz = 0.
γ

Proof. If f is integrable, then by Theorem 2.12 the integral along any close
path must be zero. Conversely, suppose the integral of f along any closed
path is zero. Choose z0 ∈ D arbitrarily. Define

F (z) := f (ζ) dζ,
γz

where γz : [a, b] → D is a path such that γz (a) = z0 and γz (b) = z. Such


a path always exists by the path-connectedness of D. Also, the definition
of F (z) is well, since any other path with the same end points must yield
the same value by assumption. F : D → C defined in this way satisfies the
assumption of Theorem 2.12 and is thus a primitive of f .
Robert Oeckl – CA NOTES – 01/12/2015 21

Definition 2.14. Let D ⊂ C be a region. We call D star-shaped with


center z0 ∈ D iff for every element z ∈ D the path γ : [0, 1] → C given by
γ(t) := z0 + t(z − z0 ) lies entirely in D.

A triangle ∆ is a closed subset of C with the shape of a triangle. Its


boundary ∂∆ is the union of three straight line segments. We also denote by
∂∆ a closed path that traces the boundary of the triangle once in positive
(i.e., counter-clockwise) direction.

Lemma 2.15. Let D ⊆ C be a star-shaped region with center z0 . Let


f : D → C be continuous. Then, f is integrable in D iff for every triangle
∆ in D with z0 a corner,

f (z) dz = 0.
∂∆

Proof. If f is integrable, we obtain the required implication as a special case


of Proposition 2.13. Conversely, we show that f is integrable if the integral
along all triangles in D with one vertex in z0 vanishes. We define a function
F : D → C as follows. Let z ∈ D and define the path γz : [0, 1] → C by
γz (t) := z0 + t(z − z0 ). Since D is star-shaped with center z0 , the path γz
lies entirely in D. Then set,

F (z) := f (z) dz.
γz

Fix z ∈ D. By star-shapedness of D there exist r > 0 such that Br (z) ⊆ D


and for all ζ ∈ Br (z) the path γζ lies entirely in D. For all ξ ∈ Br (0) set
γ̃ξ : [0, 1] → C to be the path γ̃ξ (t) = z + tξ. Then, by assumption,
∫ ∫ ∫
F (z + ξ) − F (z) = f (z) dz − f (z) dz = f (z) dz,
γz+ξ γz γ̃ξ

and we may proceed as in the proof of Theorem 2.12 to show that F is a


primitive of f at z. This completes the proof.

Proposition 2.16 (Integral Lemma of Goursat). Let D ⊆ C be a region,


f ∈ O(D) and ∆ ⊂ D a triangle. Then,

f (z) dz = 0.
∂∆

Proof. We produce a sequence of triangles {∆n }n∈N with ∆n ⊂ D by itera-


tion. Set ∆1 := ∆. To produce ∆n+1 from ∆n proceed as follows. Subdivide
22 Robert Oeckl – CA NOTES – 01/12/2015

∆n into four triangles ∆n,1 , . . . , ∆n,4 by subdividing each of its sides into
two pieces of equal length. Now choose k ∈ {1, 2, 3, 4} such that the absolute
value ∫


f (z) dz
∆n,k
is maximized and set ∆n+1 := ∆n,k . This defines a sequence of triangles.

Note that the intersection n∈N ∆n is a single point z0 ∈ D.
By the addition property of the integral along paths we have for every
n ∈ N the identity
∫ ∫ ∫ ∫ ∫
f= f+ f+ f+ f.
∂∆n ∂∆n,1 ∂∆n,2 ∂∆n,3 ∂∆n,4

By the maximality condition of our construction, this implies, for all n ∈ N,


∫ ∫


f ≤ 4 f , (4)
∂∆n ∂∆n+1

and thus, ∫ ∫


f ≤ 4 n−1
f .

∂∆ ∂∆n
For the circumference of the triangles we obtain the relation,
1
l(∂∆n ) = n−1 l(∂∆). (5)
2
Now set ϵ > 0 arbitrarily and choose r > 0 such that Br (z0 ) ⊆ D and
|g(z)| ≤ ϵ|z − z0 |, where g(z) := f (z) − f (z0 ) − f ′ (z0 )(z − z0 )
for all z ∈ Br (z0 ). (This is possible since f is complex differentiable at z0 .)
Now fix n ∈ N such that ∆n ⊂ Br (z0 ). Note that the constant function and
the identity function are integrable so that with Proposition 2.13 we have,
∫ ∫ ∫
( )
f (z) dz = f (z0 ) + f ′ (z0 )(z − z0 ) + g(z) dz = g(z) dz.
∂∆n ∂∆n ∂∆n

Using the estimate of Proposition 2.6, and (5),



ϵ ϵ
f ≤ ∥g∥∂∆n l(∂∆n ) ≤ l(∂∆n )2 = 2n−1 l(∂∆)2 .
2 2
∂∆n

On the other hand, combining this with (4) we get,



ϵ
f ≤ l(∂∆)2 .
2
∂∆
Since ϵ was arbitrary, we conclude that the integral of f along ∂∆ vanishes.
Robert Oeckl – CA NOTES – 01/12/2015 23

Proposition 2.17 (Cauchy Integral Theorem for star-shaped regions). Let


D ⊆ C be a star-shaped region and f ∈ O(D). Then, f is integrable in D.
Proof. This is obtained by combining Lemma 2.15 with Proposition 2.16.

We arrive at the important conclusion that a holomorphic function is


integrable (in star-shaped regions). Soon we will see that the converse is
also true: An integrable function is holomorphic.
Exercise 15. Let D := C \ [0, 1]. Show that f (z) := z(z−1)
1
is integrable in
D. [Hint: Observe that f (z) = z−1 − z and use primitives for the summands.
1 1

Be careful about the domain of definition.]


Exercise 16. Let D ⊆ C be a region and {fn }n∈N a sequence of continuous
integrable functions converging uniformly to a function f : D → C. Show
that f is integrable in D.
Exercise 17. Let D1 , D2 ⊆ C be regions such that D1 ∩ D2 is connected.
Let f : D1 ∪ D2 → C be continuous. (a) Show that if f is integrable in
D1 and also integrable in D2 , then f is integrable in D1 ∪ D2 . (b) Give a
counter example in the case when the connectedness condition is removed.

2.4 The Cauchy Integral Formula


In order to obtain Cauchy’s integral formula, a key result of complex anal-
ysis, we need a sharpened version of Proposition 2.16.
Proposition 2.18. Let D ⊆ C be a region and p ∈ D. Let f : D → C be
continuous function which is moreover holomorphic in D \ {p}. Let ∆ be a
triangle in D such that p is one of the corners of ∆. Then,

f = 0.
∂∆

Proof. Fix ϵ > 0. Denote the corner points of ∆ by (p, x, y). Define the
triangle ∆t for t ∈ [0, 1] as the triangle with corner points (p, xt , yt ), where
xt := p + t(x − p) and yt := p + t(y − p). Then, l(∂∆t ) → 0 as t → 0. By
continuity of f on the compact set ∆, Proposition 2.6 implies that there
exists t > 0 such that ∫
f < ϵ.
∂∆t
Now, subdivide the triangle ∆ into the triangle ∆t and the triangles with
corners given by (xt , x, y) and (xt , y, yt ). The integral over boundary paths
24 Robert Oeckl – CA NOTES – 01/12/2015

of the latter two triangles vanishes by Proposition 2.16. On the other hand,
the sum of the integrals over the boundary paths of the three triangles equals
the integral over the boundary path of ∆. Thus,
∫ ∫
f= f < ϵ.
∂∆ ∂∆t

Since ϵ was arbitrary the statement follows.

Exercise 18. The above Proposition can be strengthened considerably.


Show the following: Let ∆ ⊂ C be a triangle and let f : ∆ → C be con-
tinuous. Furthermore, assume that f is holomorphic in the interior of ∆.
Then, ∫
f = 0.
∂∆

The above proposition implies a corresponding stronger version of Propo-


sition 2.17.

Proposition 2.19. Let D ⊆ C be a star-shaped region with center z0 ∈ D


and f : D → C continuous. Furthermore assume that f is holomorphic in
D \ {z0 }. Then, f is integrable in D.

Proof. Combine Lemma 2.15 with Proposition 2.18.

Theorem 2.20 (Cauchy Integral Formula). Let D ⊆ C be a star-shaped


region with center z, f ∈ O(D), γ a closed path in D \ {z}. Then,

1 f (ζ)
f (z)Indγ (z) = dζ.
2πi γ ζ −z

Proof. Define the function g : D → C as follows,


{ f (ζ)−f (z)
ζ−z if ζ ̸= z
g(ζ) := .
f ′ (z) if ζ = z

By the property of the complex derivative of f at z, g is continuous in all


of D. Moreover, g is holomorphic in D \ {z}. So, by Proposition 2.19, g is
integrable in D. By Proposition 2.13 this implies,
∫ ∫
f (ζ)
0= g= dζ − f (z) 2πi Indγ (z).
γ γ ζ −z
Robert Oeckl – CA NOTES – 01/12/2015 25

The Cauchy Integral Formula is often used in the special case where the
path is the boundary of a disk: Let D ⊆ C be a region, f ∈ O(D), z ∈ D
and r > 0 such that Br (z) ⊂ D. Then,

1 f (ζ)
f (z) = dζ.
2πi ∂Br (z) ζ −z
Lemma 2.21. Let U ⊆ C be open, f : U → C continuous and γ a closed
path in U . Define the function F : C \ |γ| → C via

f (ζ)
F (z) := dζ.
γ ζ −z
Then, F is analytic in C \ |γ|. Moreover, for all n ∈ N0 ,

(n) f (ζ)
F (z) = n! dζ.
γ (ζ − z)n+1
Proof. Fix z0 ∈ C \ |γ| and define for all n ∈ N0 ,

f (ζ)
cn := dζ.
γ (ζ − z0 )n+1
Set r := inf t∈[a,b] |γ(t) − z0 |. We proceed to show that the power series


G(z) := cn (z − z0 )n
n=0

converges in Br (z0 ) and agrees there with F (z). Fix z ∈ Br (z0 ). Define the
partial sums gn : |γ| → C for n ∈ N0 via,

n
f (ζ)(z − z0 )k
gn (ζ) := .
k=0
(ζ − z0 )k+1

Since |ζ −z0 | ≥ r > |z−z0 | and f is bounded on |γ|, the sequence of functions
{gn }n∈N0 converges uniformly on |γ|. Thus, by Proposition 2.7,
∫ ∫
G(z) = lim gn (ζ) dζ = lim gn (ζ) dζ.
n→∞ γ γ n→∞

In particular, G(z) is well defined and its radius of convergence is at least


r. Consider now the identity


1
= xk ,
1 − x k=0
26 Robert Oeckl – CA NOTES – 01/12/2015

for x ∈ B1 (0) ⊂ C. Inserting x = (z − z0 )/(ζ − z0 ) and dividing by (ζ − z0 )


we get,


1 (z − z0 )k
= .
ζ − z k=0 (ζ − z0 )k+1
This implies,
f (ζ)
lim gn (ζ) = ,
n→∞ ζ −z
and hence G(z) = F (z).
Finally, Theorem 1.20 tells us that F is holomorphic and its complex
derivatives are again analytic in the same region. In particular, formula (2)
of that Theorem yields,
F (n) (z) = n! cn ,
and thus the stated formula.

Theorem 2.22 (Cauchy-Taylor Representation Theorem). Let D ⊆ C be a


region, f ∈ O(D). Then, f is analytic in D. Moreover, for any z0 ∈ D and
r > 0 such that Br (z0 ) ⊆ D we have,

(n) n! f (ζ)
f (z) = dζ.
2πi ∂Br (z0 ) (ζ − z)n+1
for all z ∈ Br (z0 ).
Proof. Fix z0 ∈ D and ρ > 0 such that Bρ (z0 ) ⊆ D. Then choose r such that
0 < r < ρ. This implies, Br (z0 ) ⊂ D and by Theorem 2.20 and Exercise 14
we have, ∫
1 f (ζ)
f (z) = dζ.
2πi ∂Br (z0 ) ζ − z
for z ∈ Br (z0 ). Lemma 2.21 then tells us that f is analytic in Br (z0 ) and
that ∫
(n) n! f (ζ)
f (z) = dζ.
2πi ∂Br (z0 ) (ζ − z)n+1
for z ∈ Br (z0 ) and n ∈ N0 . But since r can be chosen arbitrarily close to ρ,
the radius of convergence of the power series for f around z0 is actually at
least ρ. Thus, f is analytic in D. This completes the proof.

This Theorem finally yields the remarkable result that holomorphic func-
tions are analytic. Together with Theorem 1.20 this means that the proper-
ties of holomorphicity and analiticity are really equivalent. Furthermore, it
implies that the derivative of a holomorphic function is again a holomorphic
function.
Robert Oeckl – CA NOTES – 01/12/2015 27

Definition 2.23. Let D ⊆ C be a region. We call f : D → C locally analytic


iff for every point z ∈ D there is r > 0 so that f can be represented by a
power series around z with radius of convergence r.
Definition 2.24. Let D ⊆ C be a region. We call f : D → C locally
integrable iff for every point z ∈ D there is a neighborhood U ⊆ D of z such
that f is integrable in U .
Theorem 2.25. Let D ⊆ C be a region. For a function f : D → C the
following statements are equivalent:
1. f is holomorphic in D.

2. f is analytic in D.

3. f is locally analytic in D.

4. f is locally integrable in D.
Proof. Exercise.

Exercise 19. Calculate the following integrals. [Hint: Use the Cauchy
Integral formula]
1. ∫
ez
dz
∂B2 (0) (z + 1)(z − 3)2
2. ∫
1
dz
∂B2 (−2i) z2 +1

Exercise 20. Determine all entire functions f ∈ O(C) which satisfy the
differential equation f ′′ + f = 0.

2.5 General Cauchy Theory


We have already seen that the index Indγ (z) of a point z with respect to
a path γ is zero, if z lies in the connected component of C \ |γ| which is
unbounded. This motivates the following definition.
Definition 2.26. Let γ be a closed path in C. We define the interior of γ
as the subset Intγ := {z ∈ C \ |γ| : Indγ (z) ̸= 0}. Similarly, we define the
exterior of γ as the subset Extγ := {z ∈ C \ |γ| : Indγ (z) = 0}.
Obviously, we have the disjoint union C = Intγ ∪ |γ| ∪ Extγ .
28 Robert Oeckl – CA NOTES – 01/12/2015

Lemma 2.27. Let D ⊆ C be a region and γ a path in D. Suppose g :


|γ| × D → C is a continuous function such that z 7→ g(ζ, z) is holomorphic
for all ζ ∈ |γ|. Then, the function h : D → C given by

h(z) := g(ζ, z) dζ
γ

is holomorphic.

Proof. Fix z ∈ D. Let U ⊆ D be a star-shaped neighborhood of z with


center z (e.g. a disc centered at z). Then, for all ζ ∈ |γ| and all closed paths
γ̃ in U we have, ∫
g(ζ, z) dz = 0,
γ̃

by Proposition 2.13 since z 7→ g(ζ, z) is holomorphic and thus integrable in


U by Proposition 2.17. But we can interchange the order of integration by
Fubini’s Theorem to get
∫ ∫ (∫ ) ∫ (∫ )
h(z) dz = g(ζ, z) dζ dz = g(ζ, z) dz dζ = 0.
γ̃ γ̃ γ γ γ̃

Thus, h is integrable in U by Proposition 2.13 and therefore holomorphic in


U by Theorem 2.25. Since z was arbitrary, h is holomorphic in D.

Theorem 2.28. Let D ⊆ C be a region and γ a closed path in D. Then,


the following conditions are equivalent:

1. All f ∈ O(D) satisfy ∫


f = 0.
γ

2. For every f ∈ O(D) and every z ∈ D \ |γ| we have,



1 f (ζ)
f (z)Indγ (z) = dζ.
2πi γ ζ −z

3. Intγ ⊂ D.

Proof. To show 2.⇒1. for a given f ∈ O(D), choose z ∈ D \ |γ| arbitrarily


and define h ∈ O(D) via h(ζ) := (ζ − z)f (ζ). Applying the formula of 2. to
h yields 1. since h(z) = 0 by construction.
Robert Oeckl – CA NOTES – 01/12/2015 29

We proceed to show 1.⇒3. If D = C there is nothing to demonstrate.


So assume the contrary and let z0 ∈ C \ D. We have to demonstrate that
Indγ (z0 ) = 0. Define f ∈ O(D) via f (z) := (z − z0 )−1 . By 1.,

0= f = 2πi Indγ (z0 ).
γ

It remains to demonstrate 3.⇒2. Define the function g : D × D → C as


follows, { f (ζ)−f (z)
ζ−z if ζ ̸= z
g(ζ, z) := ′
.
f (z) if ζ = z
We proceed to show that g is continuous. For (ζ, z) ∈ D ×D such that ζ ̸= z
this is immediate. Thus, consider the case ζ = z and fix z ∈ D. Let r > 0
such that Br (z) ⊂ D. Consider the power series expansion of f around z,


f (ζ) = cn (ζ − z)n ,
n=0

for all ζ ∈ Br (z). Then, for ζ, ξ ∈ Br (z),



∑ ∑
n
g(ζ, ξ) = f ′ (z) + cn (ζ − z)n−k (ξ − z)k−1 .
n=2 k=1

For ζ, ξ ∈ Bρ (z) with 0 < ρ < r we have thus the estimate,



∑ ∑
n ∞

k−1
|g(ζ, ξ) − g(z, z)| ≤ cn (ζ − z)n−k
(ξ − z) ≤ n|cn |ρn−1 .

n=2 k=1 n=2

However, the series on the right hand side converges for all ρ < r to a
continuous function which goes to 0 when ρ → 0. Thus, g is continuous
at (z, z). Since z was arbitrary in D this completes the proof that g is
continuous in D × D. Exercise.Show that g is holomorphic in the second
argument. Now we apply Lemma 2.27 to conclude that the function h :
D → C defined by ∫
h(z) := g(ζ, z) dζ
γ

is holomorphic in D.
Now observe that for z ∈ D \ |γ| we have
∫ ∫
f (ζ)
h(z) = g(ζ, z) dζ = −2πi f (z)Indγ (z) + dζ.
γ γ ζ −z
30 Robert Oeckl – CA NOTES – 01/12/2015

Thus, to show 2. we need to show that h = 0. But if z ∈ D ∩ Extγ, then


Indγ (z) = 0 and we get

f (ζ)
h(z) = dζ.
γ ζ −z

However, this formula actually defines a holomorphic function in all of Extγ


by Lemma 2.21. Thus, we use it to extend h to a holomorphic function on
the open set D ∪ Extγ. Now recall that the assumption is that Intγ ⊂ D.
But this means D ∪Extγ = C, i.e. h is entire. Exercise.Complete the proof.
[Hint: Use Liouville’s Theorem 1.26].
Robert Oeckl – CA NOTES – 01/12/2015 31

3 Basic properties
3.1 From local to global structure
Definition 3.1. Let T be a topological space and A a subset. We say that
p ∈ A is an isolated point of A in T iff there exists a neighborhood U ⊆ T
of p such that U ∩ A = {p}. We say that A is discrete in T iff all its points
are isolated.
Theorem 3.2 (Riemann Continuation Theorem). Let D ⊆ C be a region
and A ⊂ D a discrete and relatively closed subset. Suppose that f ∈ O(D \
A). Then, the following assertions are equivalent.
1. f extends to a holomorphic function on D.

2. f extends to a continuous function on D.

3. f is bounded in some neighborhood of any point of A.

4. limz→z0 (z − z0 )f (z) = 0 for each point z0 ∈ A.


Proof. The implications 1.⇒2.⇒3.⇒4. are clear. It remains to show 4.⇒1.
It is sufficient to consider a single point z0 ∈ A. Moreover, without loss of
generality we may assume z0 = 0. Since 0 is isolated, there exists an open
neighborhood U ⊆ D of 0 such that U ∩ A = {0}. Define g : U → C as
follows, {
zf (z) if z ̸= 0
g(z) := .
0 if z = 0
By assumption, g is continuous in U . Define h : U → C by h(z) := zg(z).
Since g is holomorphic in U \ {0} so is h. Moreover, h(z) = h(0) + zg(z) =
h(0) + o(|z|), so h is complex differentiable at 0 with differential h′ (0) = 0.
Thus, h is actually holomorphic in U . By Theorem 2.22 it can be represented

for some radius of convergence r > 0 as a power series h(z) = ∞ n=0 cn z
n

around 0. But since h(0) = 0 and h (0) = 0 we actually have c0 = 0 and

c1 = 0 and thus h(z) = z 2 ∞ n
n=0 cn+2 z , where the series still converges
pointwise in Br (0). But since h(z) = z 2 f (z) in U \ {0}, this implies that the

power series ∞ n=0 cn+2 z coincides with f in U ∩ Br (0) \ {0}. This yields an
n

analytic (and therefore holomorphic) extension of f to (D \ A) ∪ {zo }.

Theorem 3.3 (Identity Theorem). Let D be a region and f, g ∈ O(D). The


following statements are equivalent:
1. f = g
32 Robert Oeckl – CA NOTES – 01/12/2015

2. The coincidence set {z ∈ D|f (z) = g(z)} is not empty and not discrete.

3. There exists a point z0 ∈ D such that f (n) (z0 ) = g (n) (z0 ) for all n ∈ N.
Proof. The implication 1.⇒2. is trivial. We show 2.⇒3. Let h := f − g.
Suppose z0 ∈ {z ∈ D|h(z) = 0} is not an isolated point. Suppose there
exists m ∈ N0 such that h(m) (z0 ) ̸= 0 and choose the smallest such m. Since
h is holomorphic in D it is also analytic by Theorem 2.22 and has a power
series expansion around z0 for some radius r > 0, given by

∑ h(n) (z0 )
h(z) = (z − z0 )n = (z − z0 )m k(z),
n=m n!

where k : Br (0) → C is the analytic function given by the power series,



∑ h(n+m) (z0 )
k(z) = (z − z0 )n .
n=0
(n + m)!

In particular, k(z0 ) = h(m) (z0 )/m! ̸= 0. But continuity of k at z0 implies


that there must be a neighborhood U ⊆ D of z0 such that k(z) ̸= 0 for
z ∈ U . But this implies h(z) ̸= 0 for z ∈ U \ {z0 }, a contradiction to the
assumption that z0 is not an isolated point of the coincidence set.
We proceed to show the implication 3.⇒1. Set Sn := {z ∈ D|h(n) (z) =
0} for all n ∈ N0 . Then, each Sn is closed in D and so is the intersection

S := ∞ n=0 Sn . On the other hand, S is open since given z1 ∈ S the power
series expansion of h around z1 has non-zero radius r of convergence by
Theorem 2.22, but is identical to zero. So every point z ∈ Br (z1 ) is element
of S. Thus S is both open and closed in D. Connectedness of D implies
that S is either empty or S = D. The first possibility is excluded by the
assumption that z0 ∈ S. So the power series of h is zero around any point
of D, hence h = 0, implying f = g in D.

Corollary 3.4. Let I ⊆ R be an interval and f : I → C some function.


For any region D ⊆ C such that I ⊂ D there is at most one holomorphic
function g : D → C such that f (z) = g(z) for all z ∈ I.
This is relevant when we are interested in extending functions on R or
some interval I ⊂ R to holomorphic functions on the complex plane.
Theorem 3.5 (Maximum Modulus Principle). Let D ⊆ C be a region and
f ∈ O(D). Suppose that |f | has a local maximum at some point z ∈ D,
i.e., that |f (z)| = ∥f ∥U := supζ∈U |f (ζ)| for some neighborhood U ⊆ D of
z, then f is constant.
Robert Oeckl – CA NOTES – 01/12/2015 33

Proof. Given a point z ∈ D and a neighborhood U of z as described, consider



the power series expansion f (ζ) = ∞n=0 cn (ζ − z) of f around z. Let ρ > 0
n

such that Bρ (z) ⊆ U . Then, the power series converges with radius at least
ρ and for 0 < r < ρ we have, by Lemma 1.24,


|cn |2 r2n ≤ ∥f ∥2∂Br (z) ≤ ∥f ∥2U = |f (z)|2 = |c0 |2 .
n=0

This implies ck = 0 for all k ∈ N, i.e., f is constant in Bρ (z). But then the
Identity Theorem (Theorem 3.3) ensures that f is constant in all of D.

Proposition 3.6. Let D ⊆ C be a bounded region and K its closure. Sup-


pose f : K → C is continuous and its restriction to D is holomorphic.
Then,
|f (z)| ≤ ∥f ∥∂D ∀z ∈ D.
In case of equality for some z ∈ D, f is constant.
Proof. If f is constant the inequality is an equality and is valid trivially.
Thus, suppose that f is not constant. Since K is compact and f is continuous
on K there exists a point z ∈ K such that |f (z)| = ∥f ∥K . We have to show
that necessarily z ∈ ∂D = K \ D. Assume to the contrary that z ∈ D. Since
|f (z)| = ∥f ∥K = ∥f ∥D we can apply Theorem 3.5 with U = D, concluding
that f is constant, a contradiction.

Proposition 3.7 (Minimum Principle). Let D ⊆ C be a region and f ∈


O(D). Suppose that |f | has a local minimum at some point z ∈ D, i.e., that
|f (z)| = inf ζ∈U |f (ζ)| for some neighborhood U ⊆ D of z. Then, f (z) = 0
or f is constant in D.
Proof. Let z ∈ D be a local minimum and U a neighborhood of z as de-
scribed. Without loss of generality we may assume that U is connected, i.e.
a region. If f (z) = 0 we are done. Thus, suppose f (z) ̸= 0. Since z is local
minimum of |f | in U , f (ζ) ̸= 0 for all ζ ∈ U . So, 1/f ∈ O(U ). But |1/f | has
a local maximum at z and we may apply Theorem 3.5 to conclude that 1/f
is constant in U . But then f is constant in U and by Theorem 3.3 constant
in D.

Proposition 3.8. Let D ⊆ C be a bounded region and K its closure. Sup-


pose f : K → C is continuous and its restriction to D is holomorphic. Then,
either f has zeros in D or

|f (z)| ≥ inf |f (ζ)| ∀z ∈ D.


ζ∈∂D
34 Robert Oeckl – CA NOTES – 01/12/2015

Proof. Exercise.

Exercise 21. Let D ⊆ C a region, a ∈ D. Suppose that f ∈ O(D \ {a}).


Show that f has a holomorphic extension to D if f ′ has.

Exercise 22. Let f, g be entire functions satisfying |f (z)| ≤ |g(z)| for all
z ∈ C. Show that there is a ∈ C such that f = ag.

Exercise 23. Let D ⊆ C be a region and L ⊂ C be a straight line. Let


f : D → C be continuous and f holomorphic in D \ L. Show that f is
actually holomorphic in all of D.

Exercise 24. Let D ⊆ C be a region and f ∈ O(D). Suppose that there


exists z ∈ D such that f (n) (z) = 0 for almost all n ∈ N. Show that f is a
polynomial.

Exercise 25. Let D ⊆ C be a region such that if z ∈ D then z ∈ D. Show


that for f ∈ O(D) the following statements are equivalent:

1. f (D ∩ R) ⊆ R.

2. f (z) = f (z) for all z ∈ D.

Exercise 26. For each of the following properties give an example for a
holomorphic function defined in some disk around 0 with that property or
show that there can be no such function.

1. f (1/n) = (−1)n /n for almost all n ∈ N.

2. f (1/n) = 1/(n2 − 1) for almost all n ∈ N \ {1}.

3. |f (n) (0)| ≥ (n!)2 for almost all n ∈ N0 .

4. |f (1/n)| ≤ e−n for almost all n ∈ N and f ̸= 0.

3.2 Zeros
Definition 3.9. Let D ⊆ C be a region, z0 ∈ D and f ∈ O(D) such that
f (z0 ) = 0. We say that f has a zero of order n at z0 iff there exists g ∈ O(D)
such that g(z0 ) ̸= 0 and f (z) = (z − z0 )n g(z) for all z ∈ D.

Proposition 3.10. Let D ⊆ C be a region, z0 ∈ D and f ∈ O(D) such that


f (z0 ) = 0. If f is not constant, then there exists a unique n ∈ N such that
f has a zero of order n at z0 . Moreover, n = inf{k ∈ N : f (k) (z0 ) ̸= 0}.
Robert Oeckl – CA NOTES – 01/12/2015 35

Proof. Exercise.

Proposition 3.11 (Fundamental Theorem of Algebra). Let n ∈ N and



p(z) = nk=0 ck z k be a polynomial of degree n (i.e., cn ̸= 0). Then, there
are constants a1 , . . . , an ∈ C such that p factorizes as

p(z) = cn (z − a1 ) · · · (z − an ).

Proof. Exercise.[Hint: First show the existence of one zero and factorize
it, then proceed recursively.]

Theorem 3.12. Let D ⊆ C be a region, f ∈ O(D) such that it has distinct


zeros a1 , . . . , am ∈ D with orders n1 , . . . , nm . Suppose γ is a closed path in
D \ {a1 , . . . , an } such that Intγ ⊂ D. Then,

∑ ∫
m
1 f ′ (z)
nk Indγ (ak ) = dz.
k=1
2πi γ f (z)

Proof. Knowing the zeros, we can factorize f as

f (z) = (z − a1 )n1 · · · (z − am )nm g(z),

where g ∈ O(D) has no zeros in D. Using the product rule for the derivative
we find for z ∈ D \ {a1 , . . . , an },

f ′ (z) g ′ (z) ∑m
nk
= + .
f (z) g(z) k=1 z − ak

The term g ′ /g on the right hand side is a holomorphic function in D. So,


by Theorem 2.28 its integral along γ vanishes. The second term yields the
desired sum over the indices of the ak .

Exercise 27. Let D ⊆ C be a region and a ∈ D. For a function f ∈ O(D)


we denote by na (f ) the order of its zero at a. (If f (a) ̸= 0 then na (f ) = 0.)
For all f, g ∈ O(D) show the following:

1. na (f g) = na (f ) + na (g).

2. na (f + g) ≥ min{na (f ), na (g)} and equality if na (f ) ̸= na (g).


36 Robert Oeckl – CA NOTES – 01/12/2015

3.3 Holomorphic logarithms and roots


Definition 3.13. A region D ⊆ C is called homologically simply connected
iff all holomorphic functions in D are integrable.
Remark 3.14. Theorem 2.28 together with Proposition 2.13 imply that
all holomorphic functions are integrable in a region D ⊆ C iff every closed
path γ in D satisfies Intγ ⊂ D. So this provides an alternative definition
of homologically simple connectedness. In fact it turns out that the ad-
jective “homologically” is superfluous as the notion is equivalent to simple
connectedness. However, we will not prove this here.
Definition 3.15. Let D ⊆ C be a region and f ∈ O(D). Then, g ∈ O(D)
is called a holomorphic logarithm of f iff f = exp g.
Theorem 3.16. Let D ⊆ C be a homologically simply connected region and
f ∈ O(D) zero-free. Then, there exists a holomorphic logarithm of f in D.
Proof. By the assumptions f ′ /f ∈ O(D) and integrable. Let h ∈ O(D) be
a primitive. Define k := f exp(−h) ∈ O(D). As is easy to check, k ′ = 0 so
k = c for all z ∈ D for some constant c ∈ C. This implies f = c exp h and
c ̸= 0 since f is zero-free. Since exp takes all complex values except zero,
there is b ∈ C with c = exp(b). Then, g := h + b ∈ O(D) is the looked for
holomorphic logarithm with f = exp g.

Definition 3.17. Let D ⊆ C be a region, f ∈ O(D) and n ∈ N. Then, a


(holomorphic) nth root of f is a function g ∈ O(D) such that f = g n .
Theorem 3.18. Let D ⊆ C be a homologically simply connected region and
f ∈ O(D) zero-free. Then, there exists an nth root of f for every n ∈ N.
Proof. According to Theorem 3.16 there is a holomorphic logarithm g ∈
O(D) of f . An nth root of f is given by
( )
1
z→
7 exp g ∀z ∈ D.
n

Exercise 28. Let D, D′ ⊆ C be homologically simply connected regions.


Suppose that D′′ := D ∩ D′ is connected and non-empty. Show that D′′ is
homologically simply connected.
Exercise 29. Let D ⊆ C be a region, f ∈ O(D) such that f is not constant.
Let a ∈ D. Show the equivalence of the following statements:
Robert Oeckl – CA NOTES – 01/12/2015 37

1. There exists a neighborhood U ⊆ D of a such that f has a holomorphic


square-root in U .

2. f (a) ̸= 0 or f (a) = 0 and the order of the zero is even.


38 Robert Oeckl – CA NOTES – 01/12/2015
Robert Oeckl – CA NOTES – 01/12/2015 39

4 Singularities
4.1 Types of singularities
Definition 4.1. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}).
Then, we say that f has an isolated singularity at a. Moreover, a is called
a removable singularity iff f can be extended to a holomorphic function on
all of D.

We have already seen criteria for identifying removable singularities in


the Riemann Continuation Theorem 3.2.

Definition 4.2. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}). We


say that a is a pole of f iff f diverges at a, i.e. if for any M > 0 there exists
r > 0 such that |f (z)| > M for all z ∈ Br (a) \ {a}. We say that a is an
essential singularity of f iff a is not removable and is not a pole.

We now consider poles.

Proposition 4.3. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}).


Suppose that a is a pole of f . Then, there exists a unique m ∈ N such that
there is a g ∈ O(D) with g(a) ̸= 0 and

g(z)
f (z) = ∀z ∈ D \ {a}.
(z − a)m

Proof. Since f has a pole at a there exist r > 0 such that Br (a) ⊆ D
and f (z) ̸= 0 for all z ∈ Br (a) \ {a}. Thus we can define h ∈ O(Br (a) \
{a}) by h(z) := 1/f (z). But limz→a h(z) = 0, so by Theorem 3.2, a is a
removable singularity of h and h can be extended to a holomorphic function
on all of Br (a). By Proposition 3.10 there exists a unique m ∈ N such
that h(z) = (z − a)m k(z), where k ∈ O(Br (a)) and k(a) ̸= 0. Moreover,
k(z) ̸= 0 for all z ∈ Br (a) so we can invert it, defining g ∈ O(Br (a)) by
g(z) = 1/k(z). But notice that g(z) = (z − a)m f (z) for all z ∈ Br (a) \ {a},
which obviously extends to a holomorphic function on D \ {a}. So g really
extends to a holomorphic function on all of D. Observe also that g(a) ̸= 0.
This completes the proof.

Definition 4.4. Let D ⊆ C be a region, a ∈ D, f ∈ O(D \ {a}) such that


a is a pole of f . Then, the integer m ∈ N such that g(z) := (z − a)m f (z)
extends to a holomorphic function in D with g(a) ̸= 0 is called the order of
the pole. If m = 1 we also say that the pole is simple.
40 Robert Oeckl – CA NOTES – 01/12/2015

Proposition 4.5. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}) with


a pole at a of order m. Then, there is a function g ∈ O(D) and there are
constants b1 , . . . , bm ∈ C with bm ̸= 0 such that


m
bn
f (z) = g(z) + ∀z ∈ D \ {a}.
n=1
(z − a)n

Proof. Exercise.

The second term on the right hand side of the equation above is also
called the singular part of f at a.
We now turn to essential singularities. In some sense they are more
“wild” than poles, as shows the following Theorem.

Theorem 4.6 (Casorati, Weiserstrass). Let D ⊆ C be a region, a ∈ D and


f ∈ O(D \ {a}). The following statements are equivalent:

1. The point a is an essential singularity of f .

2. For every neighborhood U ⊆ D of a the set f (U \ {a}) is dense in C.

3. There exists a sequence {zn }n∈N in D \ {a} such that limn→∞ zn = a,


but {f (zn )}n∈N has no limit in C ∪ {∞}.

Proof. We start with the implication 1.⇒2. Assume the contrary of 2. Let
U ⊆ D be a neighborhood of a such that f (U \ {a}) is not dense in C.
Thus, there exists p ∈ C and r > 0 such that f (U \ {a}) ∩ Br (p) = ∅.
This implies |f (z) − p| ≥ r for all z ∈ U \ {a}. Define g ∈ O(U \ {a}) by
g(z) := 1/(f (z)−p). Then, |g(z)| ≤ 1/r for all z ∈ U \{a} so by Theorem 3.2,
g has a removable singularity at a. Thus, c := limz→a g(z) exists. If c ̸= 0,
f (z) = p + 1/g(z) is bounded near a and thus has a removable singularity
at a. If c = 0, then limz→a |f (z)| = ∞ and f has a pole at a. In both cases,
a is not an essential singularity, contradicting 1. Exercise.Complete the
proof.

Exercise 30. Find and classify the isolated singularities of the following
functions and specify the order in case of a pole:

z4 1 − cos(z)
1. 2.
(z 4 + 16)2 sin z
1
3. exp(1/z) 4.
cos(1/z)
Robert Oeckl – CA NOTES – 01/12/2015 41

Exercise 31. Let f be a function that is holomorphic in C except for poles.


Show that the set of poles is discrete and closed.

Exercise 32. Investigate how the different types of singularities behave with
respect to addition, multiplication, quotienting and composition (whenever
the corresponding operations make sense)!

Exercise 33. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}). Show


that if a is a non-removable singularity of f , then exp ◦f ∈ O(D \ {a}) has
an essential singularity at a.

Exercise 34. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}). Let


P ∈ O(C) be a non-constant polynomial. Show that f and P ◦ f have the
same type of singularity at a.

4.2 Meromorphic functions


Definition 4.7. Let D ⊆ C be a region and A ⊂ D a discrete and rela-
tively closed subset. Then, f ∈ O(D \ A) is called meromorphic in D if all
points a ∈ A are either removable singularities or poles of f . The set of
meromorphic functions in D is denoted by M(D).

Proposition 4.8. Let D ⊆ C be a region. Then, the set M(D) forms a


vector space over C and moreover forms a field. That is, sums, scalar mul-
tiples, products and quotients of meromorphic functions are meromorphic.
(Except the quotient by the zero function.)

Proof. Exercise.

Exercise 35. Show that the set of rational functions forms a proper subfield
of M(C).

Theorem 4.9 (Argument Principle). Let D ⊆ C be a region, f ∈ M(D).


Suppose Z ⊂ D is the set of zeros of f and P ⊂ D is the set of poles of f .
Suppose γ is a closed path in D \ (Z ∪ P ) such that Intγ ⊂ D. Then,

∑ ∑ ∫
1 f ′ (z)
N (z)Indγ (z) − N (z)Indγ (z) = dz,
z∈Z z∈P
2πi γ f (z)

where N (z) is the order of the zero or pole z.

Proof. Exercise.[Hint: Generalize the proof of Theorem 3.12.]


42 Robert Oeckl – CA NOTES – 01/12/2015

Theorem 4.10 (Rouché’s Theorem). Let D ⊆ C be a region and f, g ∈


M(D). Let Zf , Zg ⊂ D be the sets of zeros of f and g and Pf , Pg ⊂ D the
sets of poles of f and g. Let γ be a closed path such that |γ| ⊂ D \ (Pf ∪ Pg )
and Intγ ⊂ D. Suppose that

|f (ζ) + g(ζ)| < |f (ζ)| + |g(ζ)| ∀ζ ∈ |γ|.

Then,
∑ ∑ ∑ ∑
N (z)Indγ (z)− N (z)Indγ (z) = N (z)Indγ (z)− N (z)Indγ (z),
z∈Zf z∈Pf z∈Zg z∈Pg

where N (z) denotes the order of the zero or pole z.

Proof. First, note that the inequality also implies |γ|∩Zf = ∅ and |γ|∩Zg =
∅. Set U := D \ (Zf ∪ Zg ∪ Pf ∪ Pg ) and h(z) := f (z)/g(z) for all z ∈ U .
Then, h ∈ O(U ). Note that the hypothesis is equivalent to the inequality

|h(z) + 1| < |h(z)| + 1 ∀z ∈ |γ|.

This inequality implies that h(z) cannot be a non-negative real number


(since in that case there would be equality). That is, h(z) ∈ C \ R+ 0 for
all z ∈ |γ|. But since C \ R+ 0 is open, there must a neighborhood V ⊆U
of |γ| such that h(z) ∈ C \ R0 for all z ∈ V . Now, C \ R0 is star-shaped
+ +

so that z 7→ 1/z is integrable there (Proposition 2.17), i.e., has a primitive


l ∈ O(C \ R+ 0 ). (l is in fact a branch of the logarithm.) But l ◦ h ∈ O(V )
is a primitive of h′ /h ∈ O(V ), so the integral of h′ /h along γ vanishes (by
Proposition 2.13). This means,
∫ ∫ ∫
h′ (z) f ′ (z) g ′ (z)
0= dz = dz − dz.
γ h(z) γ f (z) γ g(z)

The result follows then from Theorem 4.9.

Exercise 36. Let D, D′ ⊆ C be regions such that D′ ⊂ D. Consider the


linear map O(D) → O(D′ ) induced by the restriction of functions on D to
D′ . (a) Show either that this map must be injective or that it cannot be
injective. (b) Show either that this map must be surjective or that it cannot
be surjective.

Exercise 37. Let D ⊆ C be a bounded region. Define Õ(D) ⊆ O(D)


to be the set of holomorphic functions f on D such that f extends to a
holomorphic function on some open neighborhood of D. Likewise, define
Robert Oeckl – CA NOTES – 01/12/2015 43

M̃(D) ⊆ M(D) to be the set of meromorphic functions f on D such that f


extends to a meromorphic function on some neighborhood of D. (a) Show
that Õ(D) is a proper subalgebra of O(D). Likewise, show that M̃(D) is a
proper subfield of M(D). (b) Show that M̃(D) is the quotient field of Õ(D).
In other words, show that for every element f ∈ M̃(D) there exist elements
g, h ∈ Õ(D) such that f = g/h. (c) Comment on the possible problems that
would appear if one replaces in this exercise Õ(D) with O(D) and M̃(D)
with M(D).

Exercise 38. Let D ⊆ C be a region such that B1 (0) ⊂ D and f ∈ O(D).


Suppose |f (z)| < 1 for all z ∈ ∂B1 (0). Show that f has precisely one fixed
point in B1 (0).

Exercise 39. Determine the number of zeros (counted with order) of the
following functions in the specified domain:

1. z 5 + 13 z 3 + 14 z 2 + 1
3 in B1 (0) and in B1/2 (0).

2. z 5 + 3z 4 + 9z 3 + 10 in B1 (0) and B2 (0).

3. 9z 5 + 5z − 3 in B5 (0) \ B1/2 (0).

4. z 8 + z 7 + 4z 2 − 1 in B1 (0) and B2 (0).

4.3 Laurent Series


The representation of a holomorphic function with a pole as in Proposi-
tion 4.5 can be written as an “extended” power series that starts not with
the power 0, but with the power −n. Indeed, we will see that even essential
singularities can be captured by such an “extended” power series, if we start
at −∞. Such series are called Laurent series.
Let z ∈ C and 0 < r1 < r2 . In the following we use the notation

Ar1 ,r2 (z) := Br2 (z) \ Br1 (z).

This type of region is called an (open) annulus. Note the special case of the
punctured disk A0,r (z) = Br (z) \ {z}.

Definition 4.11. Let {an }n∈Z be an indexed set of complex numbers. We


∑ ∑ ∑∞
say that n∈Z an converges (absolutely) iff ∞ n=0 an and n=1 a−n both
converge (absolutely). Let S be a set and {fn }n∈Z be an indexed set of
∑ ∑
functions fn : S → C. We say that n∈Z fn converges uniformly iff ∞ n=0 fn
∑∞
and n=1 f−n both converge uniformly.
44 Robert Oeckl – CA NOTES – 01/12/2015

Proposition 4.12. Let {cn }n∈Z be an indexed set of complex numbers. De-
fine r1 , r2 ∈ [0, ∞] via

r1 := lim sup |c−n |1/n and 1/r2 := lim sup |cn |1/n .
n→∞ n→∞

Iff r1 < r2 then the Laurent series



f (z) = cn z n
n∈Z

converges absolutely for all z ∈ Ar1 ,r2 (0) and uniformly on Aρ1 ,ρ2 (0) where
r1 < ρ1 < ρ2 < r2 . Moreover, it diverges for z ∈ C \ Ar1 ,r2 (0).
Proof. Exercise.[Hint: Split the series into the parts with positive and
negative indices and apply Lemma 1.18.]

Proposition 4.13. Let D ⊆ C be a region, z0 ∈ C and 0 ≤ r1 < r2 such


that Ar1 ,r2 (z0 ) ⊂ D. Then, for all f ∈ O(D) we have,
∫ ∫
f= f.
∂Br1 (z0 ) ∂Br2 (z0 )

Moreover, for all z ∈ Ar1 ,r2 (z0 ) we have,


∫ ∫
1 f (ζ) 1 f (ζ)
f (z) = dζ − dζ
2πi ∂Br2 (z0 ) ζ −z 2πi ∂Br1 (z0 ) ζ −z

Proof. Exercise.

Theorem 4.14 (Laurent Decomposition). Let z0 ∈ C and 0 ≤ r1 < r2 ≤ ∞


and f ∈ O(Ar1 ,r2 (z0 )). Then, there exists a unique pair of holomorphic
functions f + ∈ O(Br2 (z0 )) and f − ∈ O(C \ Br1 (z0 )) such that

f (z) = f + (z) + f − (z), ∀z ∈ Ar1 ,r2 (z0 ) and lim f − (z) = 0


|z|→∞

Proof. For any r1 < s < r2 define fs : C \ ∂Bs (z0 ) → C via



1 f (ζ)
fs (z) := dζ,
2πi ∂Bs (z0 ) ζ −z

By Lemma 2.27, fs is holomorphic. Now define f + : Br2 (z0 ) → C as follows.


For a given z choose r1 < s < r2 such that |z| < s and set f + (z) := fs (z).
Proposition 4.13 ensures that this definition does not depend on the choice
of s. Moreover, it is clear that this defines a holomorphic function. Similarly,
Robert Oeckl – CA NOTES – 01/12/2015 45

we define f − : C \ Br1 (z0 ) → C as follows. For a given z choose r1 < s < r2


such that s < |z| and set f − (z) := −fs (z). Again, this definition does not
depend on the choice of s and f − is holomorphic.
Now let z ∈ Ar1 ,r2 (z0 ) and choose s1 , s2 such that r1 < s1 < |z| < s2 <
r2 . Then, by Proposition 4.13 we have,
∫ ∫
1 f (ζ) 1 f (ζ)
f (z) = dζ − dζ = f + (z) + f − (z).
2πi ∂Bs2 (z0 ) ζ −z 2πi ∂Bs1 (z0 ) ζ −z

Fix r1 < s < r2 and choose ϵ > 0. Now if

s∥f ∥∂Bs (z0 )


|z| > + s + |z0 |,
ϵ

then we have |f − (z)| < ϵ by an application of the integral estimate of Propo-


sition 2.6. Thus lim|z|→∞ f − (z) = 0.
To see uniqueness suppose there is another pair of holomorphic functions
g ∈ O(Br2 (z0 )) and g − ∈ O(C \ Br1 (z0 )) with the same properties. Then,
+

h(z) := f + (z) − g + (z) defines a holomorphic function on Br2 (z0 ). Moreover,


for z ∈ Ar1 ,r2 (z0 ) we also have h(z) = g − (z) − f − (z). But the latter are
even defined on C \ Br1 (z0 ). So h extends to an entire function. But,
lim|z|→∞ h(z) = lim|z|→∞ g − (z) − lim|z|→∞ f − (z) = 0. So by Liouville’s
Theorem (Theorem 1.26) h must be constant and therefore can only be
equal to zero.

Definition 4.15. In the above Theorem, f + is called the regular part of f


while f − is called the principal or singular part of f .

Theorem 4.16 (Laurent Series). Let z0 ∈ C and 0 ≤ r1 < r2 and f ∈


O(Ar1 ,r2 (z0 )). Then, there exist a unique set of coefficients {cn }n∈Z such
that

f (z) = cn (z − z0 )n ,
n∈Z

where the series converges absolutely for all z ∈ Ar1 ,r2 (z0 ) and uniformly on
As1 ,s2 (z0 ), when r1 < s1 < s2 < r2 . Also, the coefficients are given by

1 f (ζ)
cn = dζ,
2πi ∂Br (z0 ) (ζ − z0 )n+1

where r1 < r < r2 .


46 Robert Oeckl – CA NOTES – 01/12/2015

Proof. We use the decomposition f = f + + f − of Theorem 4.14. Define


g ∈ O(B1/r1 (0) \ {0}) via
( )
− 1
g(z) := f + z0 .
z

Since lim|z|→∞ f − (z) = 0 it follows that limz→0 g(z) = 0. In particular, g


has a continuous extension to B1/r1 (0) and thus a holomorphic one by the
Riemann Continuation Theorem (Theorem 3.2). Consider its power series
expansion


g(z) = bn z n ,
n=1

which converges pointwise in B1/r1 (0) and uniformly in B1/s1 (0) for any
s1 > r1 . Thus
( ) ∞

− 1
f (z) = g = bn (z − z0 )−n
z − z0 n=1

converges pointwise in C \ Br1 (z0 ) and uniformly on C \ Bs1 (z0 ) for any
s1 > r1 . On the other hand, the power series expansion


f + (z) = cn (z − z0 )n
n=0

converges pointwise in Br2 (z0 ) and uniformly on Bs2 (z0 ) for any 0 < s2 < r2 .
Summing both expansions and setting c−n := bn for all n ∈ N yields the
Laurent series with the desired properties.
Set r1 < r < r2 . Using Lemma 2.10 together with convergence of the
Laurent series and interchangeability of limit and integral (Proposition 2.7)
yields the desired formula for the coefficients cn .

Proposition 4.17. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}). Let


r > 0 such that A0,r (a) ⊂ D. Let

f (z) = cn (z − a)n
n∈Z

be the Laurent series for f in A0,r (a). Then,

1. a is a removable singularity of f iff cn = 0 for all n < 0.

2. a is a pole of order m of f iff c−m ̸= 0 and cn = 0 for all n < −m.


Robert Oeckl – CA NOTES – 01/12/2015 47

3. a is an essential singularity of f iff there exist infinitely many n < 0


such that cn ̸= 0.

Proof. Exercise.

Exercise 40. Let f ∈ O(C \ {0, 1, 2}) be given by


1
f (z) := .
z(z − 1)(z − 2)

Give the Laurent series expansion of f in the following regions: A0,1 (0),
A1,2 (0), A2,∞ (0).

Exercise 41. Give the Laurent series expansion of z 7→ exp(1/z).

4.4 Residues
Definition 4.18. Let a ∈ C and 0 < r, f ∈ O(Br (a) \ {a}) and

f (z) = cn (z − a)n
n∈Z

the Laurent series of f at a. Then, Res(f, a) := c−1 is called the residue of


f at a.

Theorem 4.19 (Residue Theorem). Let D ⊆ C be a region, A ⊂ D a


discrete and relatively closed subset, and f ∈ O(D \ A). Let γ be a closed
path with |γ| ⊂ D \ A and Intγ ⊂ D. Then,
∑ ∫
1
Res(f, a)Indγ (a) = f (z) dz.
a∈A
2πi γ

Proof. Define à := Intγ ∩ A. This is finite since Intγ ∪ |γ| is compact. Thus,
suppose à = {a1 , . . . , an }. Observe that the sum in the statement really
only runs over Ã, since the index of the other elements of A vanishes. Now,
decompose f into a sum

f (z) = f1 (z) + · · · + fn (z) + g(z) ∀z ∈ D \ A,

where fk ∈ O(C \ {ak }) and g ∈ O((D \ A) ∪ Ã) as follows. Let f1 be


the singular part f − of f at a1 (according to Theorem 4.14). In particular
Res(f, a1 ) = Res(f1 , a1 ). Note that f −f1 has one singularity less than f (the
one at a1 ) and moreover Res(f, ak ) = Res(f −f1 , ak ) for all k > 1. Now, take
f2 to be the singular part of f −f1 at a2 etc. Finally, let g := f −f1 −· · ·−fn
48 Robert Oeckl – CA NOTES – 01/12/2015

and notice that g has no singularities in Intγ left. Note that the integral
over g along γ vanishes by Theorem 2.28. Thus, the Theorem reduces to
proving the identity,

1
Res(h, a)Indγ (a) = h(z) dz
2πi γ

for functions h ∈ O(C \ {a}) such that lim|z|→∞ h(z) = 0. Consider the
Laurent series of h around a,
−1

h(z) = cn (z − a)n .
n=−∞

Since this converges uniformly on the compact set |γ|, we can interchange
integration and summation,
∫ −1
∑ ∫
h(z) dz = cn (z − a)n dz.
γ n=−∞ γ

Now note that (z − a)n has a primitive if n ≤ −2, i.e., is then integrable in
C \ {a}. Thus, by Proposition 2.13 its integral vanishes. Hence,
∫ ∫
h(z) dz = c−1 (z − a)−1 dz = Res(h, a)2πiIndγ (a).
γ γ

This completes the proof.

Exercise 42. Let D ⊆ C be a region and a ∈ D. Let g, h ∈ O(D) such that


g(a) ̸= 0 and h(a) = 0, but h′ (a) ̸= 0. Show that f := g/h ∈ M(D) has a
simple pole at a and,
g(a)
Res(f, a) = ′ .
h (a)
Exercise 43. Calculate the following integrals:
∫ ∞ ∫ ∞
x2 cos(x) − 1
1. dx 2. dx
0 x4 + x2 + 1 0 x2
∫ π ∫ π
cos(2θ) 1
3. dθ, a2 < 1 4. , a>1
0 1 − 2a cos(θ) + a2 0 (a + cos(θ))2
Exercise 44. Show that the following identities hold:
∫ ∞ ∫ ∞
1 π 1 π
1. 2
dx = 2. dx = 3 , a>0
0 1+x 2 0 (x2 2
+a )2 4a
Robert Oeckl – CA NOTES – 01/12/2015 49

5 Conformal mappings
5.1 Conformal mappings as holomorphic functions
Recall that we have the standard Euclidean scalar product on the complex
plane, by viewing C as a two-dimensional real vector space. That is, we
have
⟨z, z ′ ⟩ := aa′ + bb′ = ℜ(zz ′ ),

where z = a+ib and z ′ = a′ +ib′ . Recall also that |z| = ⟨z, z⟩. In geometric
terms we have,
⟨z, z ′ ⟩ = |z||z ′ | cos θ,
where θ is the angle between z and z ′ , viewed as vectors in the complex
plane.
We shall now be interested in mappings A : C → C that preserve angles
between intersecting curves. First, we consider R-linear mappings. Then,
for A to be angle-preserving clearly is equivalent to the identity,

|z||z ′ |⟨A(z), A(z ′ )⟩ = |A(z)||A(z ′ )|⟨z, z ′ ⟩ ∀z, z ′ ∈ C.

(We also require of course that A not be zero.)


We write A as a real 2 × 2 matrix
( )
r s
A= , acting as a + ib 7→ ra + sb + i(ta + ub).
t u

Lemma 5.1. Let A : C → C be an R-linear mapping. Then, A preserves


angles iff ( ) ( )
a −b a b
A= , or A =
b a b −a
where a, b ∈ R and a and b are not both equal to zero.
Proof. Exercise.

More generally, to make sense of the concept of angle-preservation for


a map f : D → C, where D is a region, it is necessary that f possesses a
continuous total differential. Then, f preserves angles iff its total differential
f ′ preserves angles at every point of D.
Proposition 5.2. Let D ⊆ C be a region and f : D → C a function
possessing a continuous total differential in D. Then, f is angle-preserving
iff f is holomorphic in D or anti-holomorphic in D and its derivative never
vanishes.
50 Robert Oeckl – CA NOTES – 01/12/2015

Proof. Exercise.

A conformal mapping is a mapping that preserves both angles and orien-


tation. Recall that a linear map is orientation preserving iff its determinant
is positive. More generally, a mapping is orientation preserving iff its total
derivative has positive determinant everywhere.

Proposition 5.3. Let D ⊆ C be a region and f : D → C a function


possessing a continuous total differential in D. Then, f is conformal iff f
is holomorphic in D and its derivative never vanishes.

Proof. Exercise.

5.2 The Open Mapping Theorem


Definition 5.4. Let X, Y be topological spaces. A map f : X → Y is called
open iff for every open set U ⊆ X the image f (U ) is open in Y .

Lemma 5.5. Let D ⊆ C be a region and f ∈ O(D). Let z ∈ D and


r > 0 such that Br (z) ⊂ D and 2δ := inf ζ∈∂Br (z) |f (ζ) − f (z)| > 0. Then,
Bδ (f (z)) ⊆ f (Br (z)).

Proof. Let a ∈ Bδ (f (z)). Then,

|f (ζ) − a| ≥ |f (ζ) − f (z)| − |a − f (z)| > δ ∀ζ ∈ ∂Br (z).

In particular, inf ζ∈∂Br (z) |f (ζ) − a| > |f (z) − a|. Thus, by Proposition 3.8
f − a must have zeros in the region Br (z). That is, there exists ξ ∈ Br (z)
such that f (ξ) = a.

Theorem 5.6 (Open Mapping Theorem). Let D ⊆ C be a region and f ∈


O(D) such that f is not constant. Then f is an open map D → C.

Proof. Let U ⊆ D be open. Let z ∈ U . It is enough to show that f (U )


contains a disc centered around f (z). Since f is not constant, by the Identity
Theorem (Theorem 3.3) there is a radius r > 0 such that f (z) ∈ / f (∂Br (z))
while Br (z) ⊆ U . Then 2δ := inf ζ∈∂Br (z) |f (ζ) − f (z)| > 0 and Lemma 5.5
can be applied, showing that Bδ (f (z)) ⊆ f (Br (z)) ⊆ f (U ).
Robert Oeckl – CA NOTES – 01/12/2015 51

5.3 Biholomorphic mappings


Definition 5.7. Let X be a topological space and S a set. A function
f : X → S is called locally injective at x ∈ X iff there is a neighborhood
U ⊆ X of x such that f restricted to U is injective. f is called locally
injective iff it is locally injective at each x ∈ X.

Theorem 5.8. Let D ⊆ C be a region, f ∈ O(D), a ∈ D and p := f (a).


Suppose that f − p has a zero of order m at a. Then there exist ϵ > 0 and
δ > 0 with Bδ (a) ⊂ D such that for q ∈ Bϵ (p) \ {p} the function f − q has
exactly m distinct simple zeros for z ∈ Bδ (a) and f − p has no further zeros
in z ∈ Bδ (a).

Proof. Since f is not constant (otherwise f − p could not have a zero of


finite order according to Proposition 3.10), neither f − p nor f ′ are constant
zero. So the zeros of both f − p and f ′ are isolated. This implies that we
can find δ > 0 with Bδ (a) ⊂ D such that f (z) − p ̸= 0 and f ′ (z) ̸= 0 for all
z ∈ Bδ (a) \ {a}. Now set ϵ := minζ∈∂Bδ (a) {|f (ζ) − p|}. Then, if q ∈ Bϵ (p),

|(f (ζ) − p) − (f (ζ) − q)| < ϵ ≤ |f (ζ) − p| ∀ζ ∈ ∂Bδ (a).

So, by Rouché’s Theorem (Theorem 4.10), f − p and f − q must have the


same numbers of zeros, counted with multiplicity, in Bδ (a), namely m. If
q ̸= p these are all simple by Proposition 3.10 because f ′ (z) ̸= 0 for z ∈
Bδ (a) \ {a}.

Proposition 5.9. Let D ⊆ C be a region and f ∈ O(D). Then, f is locally


injective at a ∈ D iff f ′ (a) ̸= 0. Moreover, f is locally injective in D iff f ′
is nowhere zero in D.

Proof. Let a ∈ D and p := f (a). Suppose first that f ′ (a) = 0. Then, either
f is constant or f − p has a zero of order m ≥ 2 at a. In the first case
the lack of local injectivity is trivial. In the second case consider an open
neighborhood U ⊆ D of a. Applying Theorem 5.8, there exists ϵ > 0 such
that for q ∈ Bϵ (p) \ {p} the equation f (z) = q has at least two distinct
solutions for z ∈ U . In particular, f is not injective in U . Since U was
arbitrary, f is not locally injective at a.
Now suppose f ′ (a) ̸= 0. Then, f − p has a simple zero at a. Applying
Theorem 5.8, there exist ϵ > 0 and δ > 0 with Bδ (a) ⊂ D such that for
all q ∈ Bϵ (p) the equation f (z) = q has exactly one solution in Bδ (a). By
continuity of f , U := f −1 (Bϵ (p)) ∩ Bδ (a) is an open neighborhood of a.
Clearly, f is injective in U , showing that f is locally injective at a.
52 Robert Oeckl – CA NOTES – 01/12/2015

Recalling Section 5.1 we see that the concept of conformality is equivalent


to holomorphicity combined with local injectivity.

Definition 5.10. Let D, D′ ⊆ C be regions. A map f : D → C with


f (D) = D′ is called a biholomorphic map from D to D′ iff f is holomorphic
and has a holomorphic inverse f −1 : D′ → C. If such a map exists, D and
D′ are said to be conformally equivalent.

Theorem 5.11. Let D ⊆ C be a region and f ∈ O(D). Then, f is a


biholomorphic mapping from D to f (D) iff f is injective.

Proof. Clearly, biholomorphicity implies injectivity. For the converse as-


sume that f is injective. By continuity, the image D′ := f (D) is connected.
Moreover, by the Open Mapping Theorem 5.6, D′ is open. So D′ is a region
as it cannot be empty. Since f is injective, the inverse map f −1 : D′ → D ex-
ists. Again using the Open Mapping Theorem, f −1 is continuous. Moreover,
by Proposition 5.9 f ′ is nowhere zero. Applying Proposition 1.7 we conclude
that f −1 is everywhere complex differentiable, i.e., it is holomorphic.

In the following H := {z ∈ C : ℑ(z) > 0} denotes the upper half-plane in


C.

Exercise 45. Show that z 7→ −z 2 restricted to H is a biholomorphic map-


ping. Onto which region?

Exercise 46. Let D ⊆ C be a region, f ∈ O(D) such that f is not constant.


Show that for any a ∈ D there exists a neighborhood U ⊆ D of a such that
there is m ∈ N and g ∈ O(U ) biholomorphic with the property f (z) =
f (a) + (g(z))m for all z ∈ U .

5.4 Conformal automorphisms of C and C×


Definition 5.12. Let D ⊆ C be a region. A biholomorphic mapping from
D to D is called a conformal automorphism of D. The group of conformal
automorphisms of D is denoted Aut(D).

As a first example we consider conformal automorphisms of C. The


following ones are obvious:

1. Ta : z 7→ z + a where a ∈ C is the translation by a.

2. Rθ : z 7→ eiθ z where θ ∈ [0, 2π) is the rotation by the angle θ around


the origin in positive direction.
Robert Oeckl – CA NOTES – 01/12/2015 53

3. Sr : z 7→ rz where r ∈ R+ is the scaling by the factor r around the


origin.

Exercise 47. Show that the group generated by translations, rotations and
scalings of C consists precisely of the biholomorphic transformations C → C
of the form
z 7→ az + b with a ∈ C \ {0}, b ∈ C.

As we shall see soon there are in fact no further automorphisms of C.


Another interesting example is the punctured plane C× := C \ {0}. In
addition to the rotations and scalings already seen above, there is another
elementary automorphism of C× given by
1
I : z 7→ , called inversion.
z
We shall see that there are no further automorphisms of C× than those
generated by rotations, scalings and inversions.

Lemma 5.13. Let D ⊆ C be a region, a ∈ D and f ∈ O(D \ {a}) be


injective. Then, either a is a pole of order one or it is a removable singularity
and the continuation of f to D is injective.

Proof. Suppose that a is a removable singularity and denote the continuation


of f by f˜ ∈ O(D). Assume that f˜ is not injective. Since f is injective this
means there exists z ∈ D \ {a} such that f˜(a) = f˜(z). Choose r > 0 such
that r < |z − a|/2 and Br (a) ⊆ D and Br (z) ⊆ D. By the Open Mapping
Theorem (Theorem 5.6) f˜(Br (z)) and f˜(Br (a)) are open and so is their
intersection U := f˜(Br (z)) ∩ f˜(Br (a)). But by assumption U is not empty
as it contains f˜(a). Since U is open there exists p ∈ U with p ̸= f˜(a). Then
there must exist z1 ∈ Br (a)\{a} and z2 ∈ Br (z) such that f (z1 ) = p = f (z2 )
contradicting the injectivity of f . Thus, f˜ must be injective.
Suppose now that a is not a removable singularity. Let r > 0 such that
Br (a) ⊂ D and define D′ := D \ Br (a). By the Open Mapping Theorem
(Theorem 5.6) the sets f (D′ ) and f (Br (a) \ {a}) are both open and non-
empty, but their intersection is empty by injectivity. Thus, f (Br (a) \ {a})
cannot be dense in C. By the Casorati-Weierstrass Theorem (Theorem 4.6)
this implies that a is not an essential singularity. Hence, it must be a pole.
This implies that there is s > 0 such that Bs (a) ⊆ D and f (z) ̸= 0 for all
z ∈ Bs (a) \ {a}. Define g ∈ O(Bs (a) \ {a}) by g(z) := 1/f (z). Note that g
is injective since f is. Also, a is a pole of f , so a is a removable singularity
of g. This implies by the above part of the proof that the continuation
54 Robert Oeckl – CA NOTES – 01/12/2015

g ∈ O(Bs (a)) is still injective. In particular, g is locally injective at a, so


Proposition 5.9 implies that g ′ (a) ̸= 0. On the other hand g(a) = 0, so a is
a zero of order one of g, implying that it is a pole of order one of f .

Theorem 5.14. Every injective holomorphic function f : C → C is an


automorphism of C and can be written in the form

z 7→ az + b for some a ∈ C× , b ∈ C.

Proof. Let


f (z) = cn z n
n=0

be the power series expansion of f . Define the function g ∈ O(C× ) by


g(z) := f (1/z). Then, g is injective and has the Laurent series expansion


g(z) = cn z −n
n=0

in A0,∞ (0). By Lemma 5.13, 0 is either a removable singularity of g or a


pole of order one. This implies cn = 0 for all n ≥ 2 by Proposition 4.17.
By injectivity c1 ̸= 0, so f has the stated form and is an automorphism of
C.

Corollary 5.15. C is not conformally equivalent to any proper subset.


Theorem 5.16. Every injective holomorphic mapping f : C× → C× is an
automorphism of C× and takes either the form
a
z 7→ az or z 7→ for some a ∈ C× .
z
Proof. According to Lemma 5.13, 0 can either be a removable singularity of f
or a pole of order one. In the first case, the continuation f˜ ∈ O(C) is injective
by the same Lemma. Thus, f˜ is automorphism of C and f˜(z) = az + b for
some a ∈ C× and b ∈ C by Theorem 5.14. But must have f˜−1 ({0}) ̸= ∅
while f −1 ({0}) = ∅, implying f˜(0) = 0. Thus, b = 0. In the second case
define the injective holomorphic function g : C× → C× by g(z) := 1/f (z).
Since f has a pole at 0, g has a removable singularity at 0. So we can apply
the first part of the proof to g showing that g(z) = ãz for some ã ∈ C× .
Setting a := 1/ã we find f (z) = a/z, completing the proof.

Exercise 48. Show that C× is conformally equivalent to C \ {p} for any


p ∈ C, but not to any other subset of C.
Robert Oeckl – CA NOTES – 01/12/2015 55

5.5 Conformal automorphisms of D


We now consider the conformal automorphisms of the open unit disk D :=
B1 (0). Among the transformations we have seen so far, the rotation by an
angle θ around the origin is obviously an automorphism of D. A less obvious
automorphism is given by
z−w
Dw : z 7→ , where w ∈ D.
wz − 1
Exercise 49. Verify the following properties of the transformation Dw :
(a) it is an automorphism of D, (b) it is self-inverse, i.e., composing the
transformation with itself yields the identity on D, (c) it interchanges the
points 0 and w.
We shall see that the group generated by rotations Rθ and by transfor-
mations Dw is already the full automorphism group of D.
Lemma 5.17 (Schwarz Lemma). Let f : D → D be a holomorphic function
such that f (0) = 0. Then,

|f (z)| ≤ |z| ∀z ∈ D and |f ′ (0)| ≤ 1.

Moreover, if |f (z)| = |z| for some z ∈ D \ {0} or if |f ′ (0)| = 1, then there is


a ∈ C with |a| = 1 such that f (z) = az for all z ∈ D.
Proof. Since f has a zero at 0, there is g ∈ O(D) such that f (z) = zg(z)
and moreover, f ′ (0) = g(0). Since |f (z)| < 1 for all z ∈ D, we have for any
0 < r < 1,
1
∥g∥∂Br (0) < .
r
On the other hand, applying Proposition 3.6 to Br (0) we have
1
|g(z)| ≤ ∥g∥∂Br (0) < ∀z ∈ Br (0).
r
Since r can be chosen arbitrarily close to 1, we get, for all z ∈ D, |g(z)| ≤ 1.
This translates to the first stated inequality if z ̸= 0 and to the second stated
inequality if z = 0. If either |f (z)| = |z| for some z ∈ D\{0} or if |f ′ (0)| = 1,
then |g(z)| = 1 for some z ∈ D. Then, by Theorem 3.5, g is constant, i.e,
there is a ∈ C such that g(z) = a for all z ∈ D. Consequently, f (z) = az.
Observe also that |a| = 1.

Proposition 5.18. Let f : D → D be biholomorphic and f (0) = 0. Then,


f is a rotation, i.e., there exists θ ∈ [0, 2π) such that f = Rθ .
56 Robert Oeckl – CA NOTES – 01/12/2015

Proof. Applying Lemma 5.17 to both f and f −1 yields,

|f (z)| ≤ |z| and |f −1 (z)| ≤ |z| ∀z ∈ D.

Replacing z by f (z) in the second inequality yields, |z| ≤ |f (z)| for all z ∈ D.
Thus, we actually find |f (z)| = |z| for all z ∈ D. By Lemma 5.17 this implies
that there exists a ∈ C with |a| = 1 and f (z) = az, i.e., f is a rotation.

Theorem 5.19. The group of automorphisms of D is generated by rotations


Rθ and transformations Dw . In particular, any automorphism of D can be
written uniquely as a composition Rθ ◦ Dw for some θ ∈ [0, 2π) and some
w ∈ D.
Proof. Let f ∈ Aut(D). Set w := f −1 (0) and define g := f ◦ Dw . Then
g ∈ Aut(D) with the property that g(0) = 0. Applying Proposition 5.18
to g yields that g is a rotation. That is, there exists θ ∈ [0, 2π) such that
g = Rθ . Then, f = Rθ ◦ Dw , since Dw ◦ Dw = id. To see uniqueness suppose
−1
that also f = Rθ′ ◦ Dw′ . Then f −1 (0) = (Rθ′ ◦ Dw′ )−1 (0) = Dw ′
′ (0) = w ,

so w = w. But composing with Dw yields then Rθ′ = Rθ which implies
θ′ = θ.

Exercise 50. Show that the set of automorphisms of D is identical to the


set of transformations D → D of the form
xz + y
z 7→ with x, y ∈ C and |x| > |y|.
yz + x
Exercise 51. Let f : D → D be holomorphic and a ∈ D such that f (a) = 0.
Show that
|z − a|
|f (z)| ≤ ∀z ∈ D.
|az − 1|
Moreover, in case of equality for some z ∈ D \ {a}, f is automorphism of D.

5.6 Möbius Transformations


It turns out that all the biholomorphic transformations we have considered
so far can be written as rational maps that arise as quotients of polynomials
of degree one. It turns out that maps of this type are always biholomorphic
and permit the understanding of a variety of conformal equivalences and
automorphism groups.
To each complex matrix
( )
a b
A=
c d
Robert Oeckl – CA NOTES – 01/12/2015 57

with c ̸= 0 or d ̸= 0 we associate the rational function MA ∈ M(C) given


by
az + b
MA (z) := .
cz + d
Since
det A
MA′ (z) =
(cz + d)2
we see that MA is constant if det A = 0. In the following we shall restrict to
the case det A ̸= 0. MA is then called a Möbius transformation or fractional
linear transformation. We denote the set of these meromorphic functions
by Möb. Recall that GL2 (C), the group of general linear transformations
in C2 , is the group of complex 2 × 2-matrices with non-zero determinant.
Proposition 5.20. The set of Möbius transformations Möb forms a group
by composition. Moreover, the map GL2 (C) → Möb given by A 7→ MA is a
group homomorphism, i.e., we have

MAB = MA ◦ MB ∀A, B ∈ GL2 (C).

Proof. Exercise.

Exercise 52. Verify that the upper triangular matrices (with non-vanishing
determinant) form a subgroup of GL2 (C). Show that the image of this
subgroup under the map GL2 (C) → Möb is the group Aut(C). Identify
the upper triangular matrices corresponding to translations, rotations and
dilations.
Exercise 53. Verify that the other Möbius transformations also define bi-
holomorphic mappings. Between which regions?
Recall that GL+
2 (R) is the group of orientation-preserving general linear
transformations of R2 , i.e., these are 2 × 2-matrices with real entries and
positive determinant.
Proposition 5.21. The restriction of the map GL2 (C) → Möb to the sub-
group GL+ 2 (R) yields Möbius transformations that are conformal automor-
phisms of H. That is, we obtain a group homomorphism GL+ 2 (R) → Aut(H).

Proof. Exercise.

Proposition 5.22. Let D, D′ ⊆ C be regions such that D and D′ are confor-


mally equivalent. Then Aut(D) and Aut(D′ ) are isomorphic. In particular,
every biholomorphic mapping D → D′ yields such an isomorphism.
58 Robert Oeckl – CA NOTES – 01/12/2015

Proof. Let f : D → D′ be a biholomorphic mapping. Then, an isomorphism


Aut(D) → Aut(D′ ) is given by g 7→ f ◦ g ◦ f −1 .

Exercise 54. Show that the Cayley map MC ∈ M(C) given by


( )
1 −i
C :=
1 i

is a biholomorphic map from H to D.


Proposition 5.23. Consider the group homomorphism GL+ (R) → Aut(D)
given by A 7→ MC ◦ MA ◦ MC−1 induced by the Cayley map MC : H → D.
This group homomorphism is surjective, i.e., every automorphism of D can
be obtained in this way.
( )
1 1
Proof. If C is the matrix of Exercise 54, then C −1 = 1
and MC−1 =
2 i −i
( )
a b
MC −1 . It is easy to verify by matrix multiplication that for A =
c d
the indicated group homomorphism yields the automorphism D → D given
by
xz + y
z 7→ ,
yz + x
where x := a+d+ib−ic and y := a−d−ib−ic. If a, b, c, d were arbitrary real
numbers, x, y would be arbitrary complex numbers. It is easy to verify that
|x|2 −|y|2 = 4 det A. Thus, the condition det A > 0 on (a, b, c, d) corresponds
precisely to the condition |x| > |y| on (x, y). Recalling Exercise 50, we
recognize that we obtain all automorphisms of D.

Exercise 55. Let A, B ∈ GL2 (C). Show that MA = MB iff there exists


λ ∈ C \ {0} such that B = λA.
PGL2 (C) is the group of projective general linear transformations of C2 .
It is the quotient GL2 (C)/C∗ , where C∗ is the subgroup of GL2 (C) given by
non-zero complex multiples of the unit matrix.
Exercise 56. Show that PGL2 (C) is isomorphic to SL2 (C)/Z2 , where Z2 is
the subgroup of SL2 (C) consisting of {1, −1}.
Proposition 5.24. PGL2 (C) ≈ Möb.
PGL+2 (R) is the group of projective orientation-preserving general lin-
ear transformations of R2 . It is the quotient GL+ ∗ ∗
2 (R)/R , where R is the
subgroup of GL+ 2 (R) given by non-zero real multiples of the unit matrix.
Robert Oeckl – CA NOTES – 01/12/2015 59

Exercise 57. Show that PGL+ 2 (R) is isomorphic to SL2 (R)/Z2 , where Z2
is the subgroup of SL2 (R) consisting of {1, −1}.

2 (R) ≈ Aut(H) ≈ Aut(D).


Proposition 5.25. PGL+

5.7 Montel’s Theorem


Let X be a topological space. We denote by C(X) the set of complex valued
continuous functions on X.

Definition 5.26. A topological space is called separable iff it contains a


countable dense subset.

Definition 5.27. Let X be a topological space, F ⊆ C(X). F is called


pointwise bounded iff for each a ∈ X there is a constant M > 0 such that
|f (a)| < M for all f ∈ F . F is called locally bounded iff for each a ∈ X there
is a constant M > 0 and a neighborhood U ⊆ X of a such that |f (x)| < M
for all x ∈ U and for all f ∈ F .

Definition 5.28. Let X be a topological space. A subset F ⊆ C(X) is


called equicontinuous at a ∈ X iff for every ϵ > 0 there exists a neighborhood
U ⊆ X of a such that

|f (x) − f (y)| < ϵ ∀ x, y ∈ U, ∀ f ∈ F.

A subset F ⊆ C(X) is called locally equicontinuous iff F is equicontinuous


at a for all a ∈ X.

Definition 5.29. Let X be a topological space. A subset F ⊆ C(X) is


called normal iff every sequence of elements of F has a subsequence that
converges uniformly on every compact subset of X.

Theorem 5.30 (Arzela-Ascoli). Let X be a separable topological space and


F ⊆ C(X). Suppose that F is pointwise bounded and locally equicontinuous.
Then, F is normal.

Proof. Let {fn }n∈N be a sequence of elements of F . We have to show that


there exists a subsequence that converges uniformly on any compact subset
of X. We encode subsequences of a sequence through infinite subsets of N in
the obvious way. Let {xk }k∈N be a sequence of points which is dense in X.
Set N0 := N and construct iteratively Nk ⊆ Nk−1 as follows. The sequence
{fn (xk )}n∈Nk−1 is bounded by the assumption of pointwise boundedness of
F . Thus there exists a convergent subsequence given by an infinite subset
60 Robert Oeckl – CA NOTES – 01/12/2015

Nk ⊆ Nk−1 . Proceeding in this way we obtain a sequence of decreasing


infinite subsets N0 ⊇ N1 ⊇ N2 ⊇ . . . . Now consider the sequence {nl }l∈N of
strictly increasing natural numbers nl obtained as follows: nl is the lth ele-
ment of the set Nl . It is then clear that the sequence {fnl (xk )}l∈N converges
for every k ∈ N.
Now let K ⊆ X be compact and choose ϵ > 0. Since F is locally
equicontinuous, we find for each a ∈ K an open neighborhood Ua ⊆ X such
that |f (x) − f (y)| < ϵ for all f ∈ F if x, y ∈ Ua . Since K is compact there
are finitely many points a1 , . . . , am ∈ K such that Ua1 , . . . , Uam cover K.
Since {xk }k∈N is dense in X there exists for each j ∈ 1, . . . , m an index kj
such that xkj ∈ Uaj . Now, {fnl (xkj )}l∈N converges and is Cauchy for all
j ∈ {1, . . . , m}. In particular, by taking a maximum if necessary we can
find l0 ∈ N such that |fni (xkj ) − fnl (xkj )| < ϵ for all i, l ≥ l0 and for all
j ∈ {1, . . . , m}.
Now fix p ∈ K. Then, there is j ∈ {1, . . . , m} such that p ∈ Uaj . For
i, l ≥ l0 we thus obtain the estimate

|fni (p) − fnl (p)| ≤ |fni (p) − fni (xkj )|


+ |fni (xkj ) − fnl (xkj )| + |fnl (xkj ) − fnl (p)| < 3ϵ.

In particular, this implies that {fnl }l∈N converges uniformly on K.

Theorem 5.31 (Montel). Let D ⊆ C be a region and F ⊆ O(D). Suppose


that F is locally bounded. Then, F is normal.
Proof. We show that F is locally equicontinuous. The result follows then
from the Arzela-Ascoli Theorem 5.30. Let z0 ∈ D and choose ϵ > 0. Since F
is locally bounded, there exists a constant M > 0 and r > 0 with B2r (z0 ) ⊂
D and such that |f (z)| < M for all z ∈ B2r (z0 ) and all f ∈ F . The Cauchy
Integral Formula (Theorem 2.20) yields for all f ∈ F and z, w ∈ B2r (z0 )
∫ ( )
1 f (ζ) f (ζ)
f (z) − f (w) = − dζ
2πi ∂B2r (z0 ) ζ − z ζ − w

z−w f (ζ)
= dζ.
2πi ∂B2r (z0 ) (ζ − z)(ζ − w)

If we restrict to z, w ∈ Br (z0 ) we have the estimate |(ζ − z)(ζ − w)| > r2


for all ζ ∈ ∂B2r (z0 ). Combining this with the standard integral estimate
(Proposition 2.6) we obtain,
2∥f ∥∂B2r (z0 ) 2M
|f (z) − f (w)| ≤ |z − w| < |z − w| .
r r
Robert Oeckl – CA NOTES – 01/12/2015 61

{rϵ
}
Choosing δ := min r, 4M yields the estimate

|f (z) − f (w)| < ϵ ∀z, w ∈ Bδ (z0 ),

showing local equicontinuity. This completes the proof.

Exercise 58. Let X be a metric space and F ⊆ C(X). Suppose that F is


normal. Show that F is (a) locally bounded and (b) locally equicontinuous.

Exercise 59 (Vitali’s Theorem). Let D ⊆ C be a region and {fn }n∈N a


locally bounded sequence of holomorphic functions on D. Set A := {z ∈ D :
limn→∞ fn (z) exists}. Suppose that A has a limit point in D. Show that
{fn }n∈N converges uniformly on compact subsets of D.

5.8 The Riemann Mapping Theorem


Proposition 5.32. Let D ⊆ C be a region and {fn }n∈N a sequence of
holomorphic functions fn ∈ O(D) that converges uniformly on any compact
(k)
subset of D to f . Then, f ∈ O(D) and the sequence {fn }n∈N converges
uniformly on any compact subset of D to f (k) for all k ∈ N.

Proof. Let z0 ∈ D and set r > 0 such that Br (z) ⊂ D. By Proposition 2.17
fn is integrable in Br (z0 ). For any closed path γ in Br (z0 ) we thus have
∫ ∫ ∫
f= lim fn = lim fn = 0,
γ γ n→∞ n→∞ γ

where we have used Proposition 2.7 to interchange the integral with the
limit. Thus, f is integrable in Br (z0 ) and hence holomorphic there by Theo-
rem 2.25. Since the choice of z0 was arbitrary we find that f is holomorphic
in all of D.
Fix k ∈ N and consider z0 ∈ D. Choose r > 0 such that B2r (z0 ) ⊆ D.
Now for each z ∈ Br (z0 ) we have the Cauchy estimate (Proposition 1.25),

k! k!
|fn(k) (z) − f (k) (z)| ≤ k
∥fn − f ∥∂Br (z) ≤ k ∥fn − f ∥B2r (z0 ) .
r r
For ϵ > 0 there is by uniform convergence of {fn }n∈N an n0 ∈ N such
that |fn (z) − f (z)| < ϵ rk /k! for all n ≥ n0 and all z ∈ B2r (z0 ). Hence,
(k) (k)
|fn (z) − f (k) (z)| < ϵ for all n ≥ n0 and all z ∈ Br (z0 ). That is, {fn }n∈N
converges to f (k) uniformly on some neighborhood of every point of D.
To obtain uniform convergence on a compact subset K ⊂ D it is merely
necessary to cover K with finitely many such neighborhoods.
62 Robert Oeckl – CA NOTES – 01/12/2015

Theorem 5.33 (Hurwitz). Let D ⊆ C be a region and {fn }n∈N a sequence


of functions fn ∈ O(D) converging uniformly in every compact subset of D
to f . Let a ∈ D and r > 0 such that Br (a) ⊂ D. Suppose that f (z) ̸= 0 for
all z ∈ ∂Br (a). Then, there exists n0 ∈ N such that f and fn have the same
number of zeros in Br (a) for all n ≥ n0 .
Proof. Set δ := inf{|f (z)| : z ∈ ∂Br (a)}. By the assumptions δ > 0 and
{fn }n∈N converges uniformly on ∂Br (a). Thus, there exists n0 ∈ N such that
|fn (z) − f (z)| < δ/2 for all n ≥ n0 and all z ∈ ∂Br (a). But this implies,
δ
|f (z) − fn (z)| < < |f (z)| ≤ |f (z)| + |fn (z)| ∀n ≥ n0 , ∀z ∈ ∂Br (a).
2
Applying Rouché’s Theorem 4.10 yields the desired result.

Proposition 5.34. Let D ⊆ C be a region and {fn }n∈N a sequence of


functions fn ∈ O(D) converging uniformly in every compact subset of D to
f . Suppose that for all n ∈ N, fn has no zeros. Then, either f = 0 or f has
no zeros.
Proof. Exercise.

Proposition 5.35. Let D ⊆ C be a region and {fn }n∈N a sequence of


injective functions fn ∈ O(D) converging uniformly in every compact subset
of D to f . Then, either f is constant or f is injective.
Proof. Suppose that f is not constant. Let a in D and set p := f (a) and
pn := fn (a) for all n ∈ N. By injectivity fn − pn never vanishes on D \ {a}.
On the other hand, the sequence {fn − pn }n∈N converges uniformly in any
compact subset of D to f − p. Since f − p ̸= 0, Proposition 5.34 implies
that f − p has no zeros in D \ {a}. In other words, f does not take the value
p at any point of D \ {a}. Since we chose a arbitrarily it follows that f is
injective.

Theorem 5.36 (Riemann Mapping Theorem). Every homologically simply


connected region which is different from C is conformally equivalent to D.
Proof. Let D be the region in question. Fix z0 ∈ D arbitrarily. Let F ⊆
O(D) be the set of holomorphic functions f ∈ O(D) which are injective,
whose image is contained in D and such that f (z0 ) = 0. Our strategy is to
find an element of F which is a biholomorphism D → D.
First we show that F is not empty. By assumption D ̸= C, so we can
choose a ∈ C \ D. The function f (z) := z − a is holomorphic and zero-
free in D, so according to Theorem 3.18 there is a holomorphic square root
Robert Oeckl – CA NOTES – 01/12/2015 63

g ∈ O(D) with g 2 = f . If g(z1 ) = g(z2 ) then (g(z1 ))2 = (g(z2 ))2 and
so z1 = z2 since f is injective. Therefore also g is injective. Moreover, if
g(z1 ) = −g(z2 ) we can draw the same conclusion z1 = z2 , but this time we
get a contradiction, since g is zero-free. Thus, if z ∈ C is in the image of
g, then −z cannot be in the image of g. Now since g is not constant the
Open Mapping Theorem 5.6 ensures that g(D) is open. In particular there
exists w ∈ C and r > 0 such that Br (w) ⊂ g(D). But applying the previous
statement to all elements of Br (w) we obtain Br (−w) ∩ g(D) = ∅. It is now
easy to see that the function h ∈ O(D) defined by h(z) := r/(g(z) + w) is
also injective and satisfies h(D) ⊆ D. Setting v := h(z0 ), we have Dv ◦ h ∈ F
since Dv ∈ Aut(D) and Dv (v) = 0.
Since D is open, there exists r > 0 such that Br (z0 ) ⊂ D. Using the
Cauchy estimate (Proposition 1.25) we find the bound |f ′ (z0 )| < 1/r for all
f ∈ F . This implies that

M := sup{|f ′ (z0 )| : f ∈ F }

is well defined. On the other hand we will show that if f (D) ̸= D for some
f ∈ F , then there exists g ∈ F such that |g ′ (z0 )| > |f ′ (z0 )|. This implies
that h ∈ F is a biholomorphism D → D if |h′ (z0 )| = M . We will then show
that such an h exists.
Consider some f ∈ F such that f (D) ̸= D. Choose p ∈ D \ f (D).
Since Dp ∈ Aut(D), the composition Dp ◦ f is injective and Dp ◦ f (D) ⊂ D.
Furthermore, Dp ◦f is zero-free since Dp−1 (0) = {p}. Since D is homologically
simply connected we can find a holomorphic square root g ∈ O(D) with
g 2 = Dp ◦ f according to Theorem 3.18. In fact, it is clear that g is injective
and g(D) ⊆ D. Set w := g(z0 ). Then h := Dw ◦ g ∈ F . Consider now the
holomorphic map k : D → D given by k(z) = Dp ((Dw (z))2 ). Then, f = k ◦ h
and applying the chain rule for derivatives we obtain

f ′ (z0 ) = k ′ (h(z0 ))h′ (z0 ) = k ′ (0)h′ (z0 ).

Noting that k(0) = 0 we can apply the Schwarz Lemma 5.17. Since k is not
a rotation, this implies |k ′ (0)| < 1. Hence, |f ′ (z0 )| < |h′ (z0 )| since h′ (z0 ) ̸= 0
by injectivity of h.
The image of all functions in F is contained in the bounded set D, so in
particular F is locally bounded. According to Montel’s Theorem 5.31 this
implies that F is normal. Consider now a sequence {fn }n∈N of elements
of F such that |fn′ (z0 )| → M as n → ∞. Since F is normal, there is a
subsequence {fnk }k∈N which converges uniformly on any compact subset of
D to a function f ∈ O(D) by Proposition 5.32. By the same Proposition
64 Robert Oeckl – CA NOTES – 01/12/2015

we have convergence of the derivative and thus |f ′ (z0 )| = M as desired. It


remains to show that f ∈ F . From the limit process it is clear that f (z0 ) = 0
and f (D) ⊆ D. Since f is not constant (in particular, f ′ (z0 ) ̸= 0) the Open
Mapping Theorem 5.6 implies that f (D) must be open and so we must have
f (D) ⊆ D. The injectivity of f follows from Proposition 5.35. Hence f ∈ F .
This completes the proof.

Proposition 5.37. Let D ⊂ C be a homologically simply connected region,


a ∈ D. Then, there exists exactly one biholomorphism f : D → D such that
f (a) = 0 and f ′ (a) > 0.

Proof. Exercise.

Exercise 60. Show that a homologically simply connected region cannot


be conformally equivalent to a region that is not homologically simply con-
nected.
Robert Oeckl – CA NOTES – 01/12/2015 65

6 Harmonic functions
6.1 Mean value and maximum
We coordinatize the complex plane by coordinates (x, y) ∈ R2 with z =
x + iy ∈ C. The Laplace operator on the complex plane is then given by

∂2 ∂2
∆ := + .
∂x2 ∂y 2

Definition 6.1. Let U ⊆ C be an open set and f : U → R be twice contin-


uously partially differentiable. Then, f is called harmonic iff it satisfies the
Laplace equation
∆f = 0.

Proposition 6.2. The real and the imaginary part of a holomorphic func-
tion are harmonic.

Proof. Exercise.

Proposition 6.3. Let U, V ⊆ C be open and f ∈ O(U ) such that f (U ) ⊆ V .


If g : V → R is harmonic, then g ◦ f : U → R is also harmonic.

Proof. Exercise.

Lemma 6.4. Let D = C or D = D and u : D → R a harmonic function.


Then, there exists a harmonic function v : D → R such that u + iv ∈ O(D).

Proof. Define the continuously partially differentiable function v : D → R


given by
∫ y ∫ x
v(x, y) := ux (x, t) dt − uy (s, 0) ds ∀(x, y) ∈ D.
0 0

Differentiating by ∂/∂x and using that u is harmonic we get


∫ y
vx (x, y) = uxx (x, t) dt − uy (x, 0)
0
∫ y
=− uyy (x, t) dt − uy (x, 0)
0
= −uy (x, y) + uy (x, 0) − uy (x, 0)
= −uy (x, y).
66 Robert Oeckl – CA NOTES – 01/12/2015

Note that the interchange of differentiation and integration in the first step
is permitted since the integrand is continuously differentiable and the in-
tegration range compact. On the other hand, differentiating by ∂/∂y we
obtain
vy (x, y) = ux (x, y).

Thus, the pair (u, v) satisfies the Cauchy-Riemann equations so that u + iv


is holomorphic according to Proposition 1.3.

Theorem 6.5. Let D ⊆ C be a homologically simply connected region and


u : D → R be harmonic. Then, there exists a harmonic function v : D → R
such that u + iv : D → C is holomorphic.

Proof. If D = C then Lemma 6.4 directly applies and we are done. Suppose
therefore that D ̸= C. By the Riemann Mapping Theorem 5.36 there exists
a biholomorphic map f : D → D. By Proposition 6.3, u ◦ f −1 : D → R is
harmonic. Applying Lemma 6.4, there exists a harmonic function w : D → R
such that u ◦ f −1 + iw : D → C is holomorphic. Define v : D → R by
v := w ◦ f . Then, v is harmonic by Proposition 6.3 and u + iv : D → C is
holomorphic.

Proposition 6.6. Harmonic functions are infinitely differentiable.

Proof. Exercise.

Theorem 6.7 (Mean Value Theorem). Let D ⊆ C be a region and u : D →


R harmonic. Suppose a ∈ D and r > 0 such that Br (a) ⊂ D. Then,
∫ 2π ( )
1
u(a) = u a + reiθ dθ.
2π 0

Proof. Choose s > r such that Bs (a) ⊆ D. By Theorem 6.5 there exist a
harmonic function v : Bs (a) → R such that f := u + iv : Bs (a) → C is
holomorphic. Applying the Cauchy Integral Formula (Theorem 2.20) to f
at the point a with path ∂Br (a) we obtain,
∫ ∫ 2π ( )
1 f (ζ) 1
f (a) = dζ = f a + reiθ dθ.
2πi ∂Br (a) ζ −a 2π 0

Taking the real part on both sides yields the desired result.
Robert Oeckl – CA NOTES – 01/12/2015 67

Definition 6.8. Let D ⊆ C be a region and u : D → R continuous. We say


that u has the mean value property iff for all a ∈ D and all r > 0 such that
Br (a) ⊂ D we have
∫ 2π ( )
1
u(a) = u a + reiθ dθ.
2π 0

It turns out that the mean value property implies harmonicity.

Theorem 6.9 (Maximum Principle). Let D ⊆ C be a region and u : D → R


a continuous function with the mean value property. Suppose that u has a
maximum at some point a ∈ D, i.e., that u(z) ≤ u(a) for all z ∈ D. Then
u is constant.

Proof. Define
A := {z ∈ D : u(z) = u(a)}.
Since u is continuous, A must be closed in D. We proceed to show that A is
also open. Let z0 ∈ A and r > 0 such that Br (z0 ) ⊆ D. Choose b ∈ Br (z0 )
and set s := |b − z0 |. By the mean value property
∫ 2π ( )
1
u(z0 ) = u z0 + seiθ dθ.
2π 0

The integrand is continuous and everywhere(smaller or)equal to u(z0 ). Hence,


for the equality to hold, we must have u z0 + seiθ = u(z0 ) for all θ ∈
[0, 2π). In particular, u(b) = u(z0 ) = u(a) and hence b ∈ A. Since b was
chosen arbitrarily we have Br (z0 ) ⊆ A, showing that A is open. Since A is
non-empty, closed in D and open, we must have A = D. Thus, u(z) = u(a)
for all z ∈ D and u is constant.

Proposition 6.10. Let D ⊂ C be a bounded region and u : D → R a


continuous function with the mean value property in D, satisfying u|∂D = 0.
Then, u = 0.

Proof. Exercise.

Exercise 61. Show the following version of the maximum principle, which
is more similar to Theorem 3.5: Let D ⊆ C be a region and f : D → C
a continuous function satisfying the mean value property. Suppose that |f |
has a maximum at some point a ∈ D, i.e., that |f (z)| ≤ |f (a)| for all z ∈ D.
Then f is constant. [Hint: Consider the function g(z) := ℜ(f (z)/f (a)).]
68 Robert Oeckl – CA NOTES – 01/12/2015

6.2 The Dirichlet Problem


Definition 6.11. The function P : D → R given by
( )
1+z
P (z) := ℜ ∀z ∈ D
1−z

is called the Poisson kernel. For 0 ≤ r < 1 and θ ∈ R it is also common to


use the notation ( )
Pr (θ) := P reiθ .

Proposition 6.12. The Poisson kernel P has the following properties:

1. P is harmonic.

2. For all z = reiθ ∈ D,

1 − |z|2 1 − r2
Pr (θ) = P (z) = = .
|1 − z|2 1 − 2r cos θ + r2

3. P (z) > 0 for all z ∈ D.

4. Pr (−θ) = P (z) = P (z) = Pr (θ) for all z = reiθ ∈ D.

5. for all z = reiθ ∈ D,



∑ ∞

Pr (θ) = P (z) = 1 + (z n + z n ) = r|n| einθ .
n=1 n=−∞

6. For all 0 ≤ r < 1 we have


∫ ∫ π
1 P (ζ) 1
dζ = Pr (θ) dθ = 1.
2πi ∂Br (0) ζ 2π −π

7. For all 0 < r < 1 and 0 < |δ| < |θ| ≤ π we have Pr (θ) < Pr (δ).

8. For each 0 < δ < π and ϵ > 0 there exists 0 < ρ < 1 such that for all
ρ < r < 1 and δ < |θ| ≤ π we have |Pr (θ)| < ϵ.

Proof. 1. By definition, the Poisson kernel is the real part of a holomorphic


function. Thus, it is harmonic by 6.2. 2. Elementary calculation. 3. This fol-
lows immediately from 2. 4. This follows immediately from 2. 5. Exercise.
Robert Oeckl – CA NOTES – 01/12/2015 69

6. Note that for 0 ≤ r < 1 the series representation given in 5. converges


uniformly. So, we can exchange summation and integration to get,
∫ π ∞
∑ ∫ π
1 1
Pr (θ) dθ = einθ dθ = 1.
2π −π n=−∞
2π −π

7. This follows easily from 2. 8. Fix 0 < δ < π and ϵ > 0. Then, Pr (δ) → 0
for r → 1− using 2. Thus, there is 0 < ρ < 1 so that |Pr (δ)| < ϵ if ρ < r < 1.
Using 7. completes the proof of 8.

Theorem 6.13. Let b : ∂D → R be continuous. Then, there exists a unique


continuous function u : D → R such that u|∂D = b and u is harmonic in D.
Moreover, for all 0 ≤ r < 1 and θ ∈ R,
( ) ∫ π ( )
1
u reiθ = Pr (θ − ϕ)b eiϕ dϕ.
2π −π

Proof. Define u(z) for z ∈ D by the stated formula and u(z) := b(z) for
z ∈ ∂D. We first show that u is harmonic in D. We note that for z ∈ D,
∫ ( )
1 π 1 + ze−iϕ ( )
u(z) = ℜ b eiϕ dϕ
2π −π 1 − ze−iϕ
( ∫ )
1 π eiϕ + z ( iϕ )
=ℜ b e dϕ .
2π −π eiϕ − z

Note that the integrand in the last expression is continuous as a function


of (ϕ, z) ∈ R × D and holomorphic as a function of z ∈ D for each value
of ϕ ∈ R. That is, we can apply Lemma 2.27 to conclude that the integral
defines a holomorphic function in D. But by Proposition 6.2, the real part
of this function is harmonic.
We proceed to show that u is continuous in D. Since continuity in D
follows from harmonicity it suffices to consider points in the boundary of
D. In particular, it is enough to show the following: Given ψ ∈ [−π, π) and
ϵ > 0, there exist δ > 0 and 0 < ρ < 1 such that
( ) ( )

u reiθ − b eiψ < ϵ ∀ρ < r ≤ 1, ∀θ ∈ (ψ − δ, ψ + δ).

We proceed to find such δ and ρ given ψ and ϵ. By continuity of b, there


exists δ > 0 such that
( ) ( )
ϵ
b eiθ − b eiψ < ∀θ ∈ (ψ − 2δ, ψ + 2δ).
2
70 Robert Oeckl – CA NOTES – 01/12/2015

By Proposition 6.12.8, there exists 0 < ρ < 1 such that


ϵ
Pr (θ) < ∀ρ < r < 1, ∀δ < |θ| ≤ π.
4(∥b∥∂D + 1)

Now let θ ∈ (ψ − δ, ψ + δ) and 0 < r < ρ. Then,


( ) ( ) 1 ∫ π ( ) ( )

u reiθ − b eiψ = Pr (θ − ϕ)b eiϕ dϕ − b eiψ
2π −π
∫ π ( ( ) ( ))
1
= Pr (θ − ϕ) b eiϕ − b eiψ dϕ
2π −π
∫ ( ) ( )
1 π
≤ Pr (θ − ϕ) b eiϕ − b eiψ dϕ
2π −π
∫ ( ) ( )
1
= Pr (θ − ϕ) b eiϕ − b eiψ dϕ
2π |ϕ−ψ|<2δ
∫ ( ) ( )
1
+ Pr (θ − ϕ) b eiϕ − b eiψ dϕ
2π |ϕ−ψ|≥2δ

1 ϵ
≤ Pr (θ − ϕ) dϕ
2π |ϕ−ψ|<2δ 2

1 ϵ
+ 2∥b∥∂D dϕ
2π |ϕ−ψ|≥2δ 4(∥b∥∂D + 1)
∫ ∫
1 π ϵ 1 π ϵ
< Pr (θ − ϕ) dϕ + dϕ
2π −π 2 2π −π 2
= ϵ.

Here, we have used the properties of the Poisson kernel given in Proposi-
tion 6.12 parts 3. and 6.
It remains to show uniqueness of the function u. Suppose there was
another function v : D → R with the required properties. Then, the dif-
ference u − v would be continuous on D and harmonic in D. Furthermore,
(u − v)|∂D = 0, so by Proposition 6.10, u − v = 0, i.e., u = v.

Definition 6.14. We call a region D ⊆ C disk-like iff there exists a confor-


mal equivalence D → D which extends to a homeomorphism D → D.

Remark 6.15. A disk-like region is in particular homologically simply con-


nected and bounded.

Theorem 6.16. Let D ⊂ C be a disk-like region. Let b : ∂D → R be


continuous. Then, there exists a unique continuous function u : D → R
such that u|∂D = b and u is harmonic in D.
Robert Oeckl – CA NOTES – 01/12/2015 71

Proof. Exercise.

Theorem 6.17. Let U ⊆ C be open and u : U → R be continuous with the


mean value property. Then, u is harmonic.

Proof. Let a ∈ U and r > 0 such that Br (a) ⊂ U . It is sufficient to


show that u is harmonic in Br (a). Since Br (a) is disk-like there exists by
Theorem 6.16 a continuous function v : Br (a) → R which is harmonic in
Br (a) and coincides with u in ∂Br (a). But the difference u − v : Br (a) →
R is continuous, has the mean value property in Br (a) and vanishes on
the boundary ∂Br (a). Thus u = v also in Br (a) by Proposition 6.10. In
particular, u is harmonic in Br (a).
72 Robert Oeckl – CA NOTES – 01/12/2015
Robert Oeckl – CA NOTES – 01/12/2015 73

7 The Riemann Sphere


7.1 Definition
Definition 7.1. A topological space is called locally compact iff every point
has a compact neighborhood.
Proposition 7.2 (One-Point Compactification). Let X be a Hausdorff topo-
logical space that is locally compact. Consider the set X̂ := X ∪{∞} equipped
with the following topology: A set U ⊆ X̂ is open iff U ⊆ X and U is open
in X or if U = V ∪ {∞} where V ⊆ X such that X \ V is compact in X.
Then, X̂ is a compact Hausdorff space.
Proof. Exercise.

Proposition 7.3. Consider the topological space Ĉ with the subsets U0 :=


Ĉ \ {∞} and U∞ := Ĉ \ {0}. Consider the maps ϕ0 : U0 → C given by
ϕ0 (z) := z for all z ∈ U0 and ϕ∞ : U∞ → C given by ϕ∞ (z) := 1/z for all
z ∈ U∞ \ {∞} and ϕ∞ (∞) := 0. Then, ϕ0 and ϕ∞ are homeomorphisms.
Moreover, ϕ0 ◦ ϕ−1∞ |C\{0} is the biholomorphism C \ {0} → C \ {0} given by
z 7→ 1/z.
Proof. Exercise.

Remark 7.4. The topological space Ĉ together with the structures intro-
duced in the preceding Proposition is called the Riemann sphere. It is an
example of a complex manifold. The maps ϕ0 , ϕ∞ are called charts.
Exercise 62. Let {zn }n∈N be a sequence of complex numbers. Show that
limn→∞ zn = ∞ in Ĉ if and only if for each M > 0 there exists n0 ∈ N such
that |zn | > M for all n ≥ n0 .
Exercise 63. Consider the symmetric function d : Ĉ × Ĉ → R+
0 given by

2|z − z ′ |
d(z, z ′ ) := √ ∀z, z ′ ∈ C
(1 + |z|2 )(1 + |z ′ |2 )
2
d(∞, z) := √ ∀z ∈ C
1 + |z|2
d(∞, ∞) := 0.
Show that d defines a metric on the Riemann sphere that is compatible with
its topology.
Remark 7.5. The metric introduced above can be obtained from the stere-
ographic projection of Ĉ identified with the unit disk to the complex plane.
74 Robert Oeckl – CA NOTES – 01/12/2015

7.2 Functions on Ĉ
Exercise 64. Let D ⊆ Ĉ be a region and f : D → C be continuous. Let
a ∈ D \ {0, ∞}. Show that f ◦ ϕ−1
0 is holomorphic/conformal at ϕ0 (a) iff
−1
f ◦ ϕ∞ is holomorphic/conformal at ϕ∞ (a).
Definition 7.6. Let D ⊆ Ĉ be a region and f : D → C be continuous.
Let a ∈ D. If a ̸= ∞, we say that f is holomorphic/conformal at a iff
f ◦ ϕ−1
0 is holomorphic/conformal at ϕ0 (a). If a ̸= 0, we say that f is
−1 is holomorphic/conformal at ϕ (a).
holomorphic/conformal at a iff f ◦ ϕ∞ ∞
We say that f is holomorphic/conformal in D iff f is holomorphic/conformal
at each point a ∈ D.
Exercise 65. Let D ⊆ Ĉ be a region and a ∈ D\{0, ∞}. Let f ∈ O(D\{a}).
Show that the type and order of the singularity of f ◦ ϕ−10 at ϕ0 (a) is the
same as the type and order of the singularity of f ◦ ϕ−1
∞ at ϕ∞ (a).
Definition 7.7. Let D ⊆ Ĉ be a region, a ∈ D and f ∈ O(D \ {a}).
If a ̸= ∞, we say that f has a removable singularity/a pole of order n/an
essential singularity at a iff f ◦ϕ−1
0 has a removable singularity/a pole of order
n/an essential singularity at ϕ0 (a). If a ̸= 0, we say that f has a removable
singularity/a pole of order n/an essential singularity at a iff f ◦ ϕ−1 ∞ has a
removable singularity/a pole of order n/an essential singularity at ϕ∞ (a).
Proposition 7.8. Let f ∈ O(Ĉ). Then, f is constant.
Proof. Exercise.

Definition 7.9. Let D ⊆ Ĉ be a region and A ⊂ D be a discrete and


relatively closed subset. A function f ∈ O(D \ A) is called meromorphic iff
each point a ∈ A is either a removable singularity or a pole of f .
Proposition 7.10. Let f ∈ M(Ĉ). Then, f is a rational function.
Proof. Exercise.[Hint: First assume that f has a pole only at ∞ and show
that |f (z)| < M |z|n for some constants M > 0 and n ∈ N. Conclude that f
must be a polynomial. In the general case show and use the fact that f can
only have finitely many poles.]

7.3 Functions onto Ĉ and Aut(Ĉ)


Exercise 66. Let D ⊆ Ĉ be a region and f ∈ M(D). Let P ⊂ D be the
set of poles of f and Z ⊆ D the set of zeros of f . Define fˆ : D → Ĉ by
fˆ(z) := ϕ−1 ˆ ˆ
0 (f (z)) if z ∈ D \ P and f (z) := ∞ if z ∈ P . Show that f is
ˆ ˆ
continuous and that ϕ0 ◦ f |D\P as well as ϕ∞ ◦ f |D\Z are holomorphic.
Robert Oeckl – CA NOTES – 01/12/2015 75

Exercise 67. Let D ⊆ Ĉ be a region and fˆ : Ĉ → Ĉ be continuous. Let


Z := {z ∈ Ĉ : fˆ(z) = 0} and P := {z ∈ Ĉ : fˆ(z) = ∞}. Suppose that
ϕ0 ◦ fˆ|D\P as well as ϕ∞ ◦ fˆ|D\Z are holomorphic. Define f : D \ P → C by
f := ϕ0 ◦ fˆ|D\P . If P ̸= D, then f ∈ M(D).

Definition 7.11. Let D ⊆ Ĉ be a region and f : D → Ĉ be continuous. Let


a ∈ D. If f (a) ̸= ∞, we say that f is conformal at a iff ϕ0 ◦ f is conformal
at a. If f (a) ̸= 0, we say that f is conformal at a iff ϕ∞ ◦ f is conformal at
a. We say that f is conformal in D iff f is conformal at each point a ∈ D.

Definition 7.12. A conformal mapping Ĉ → Ĉ that has a conformal inverse


is called a conformal automorphism of Ĉ.

Proposition 7.13. Möbius transformations are conformal automorphisms


of Ĉ.

Proof. Exercise.

Theorem 7.14. Suppose that f : Ĉ → Ĉ is conformal and injective. Then,


f is a Möbius transformation.

Proof. (Sketch.) As in Exercise 67 we can think of f as a meromorphic


function on Ĉ. Thus, by Proposition 7.10, f is rational, i.e., a quotient p/q
of polynomials. Without loss of generality we may assume p and q not to
have common divisors. Since f is injective, p can only have one zero which
must be simple. Similarly, q can only have one pole which must be simple.
Thus, f is a Möbius transformation.

Corollary 7.15. Aut(Ĉ) = Möb.

Theorem 7.16. Let (a, b, c) and (a′ , b′ , c′ ) be triples of distinct points in Ĉ.
Then, there exists exactly one Möbius transformation f such that f (a) = a′ ,
f (b) = b′ , f (c) = c′ .

Proof. Exercise.

You might also like