Further Analysis

Prof. W.T. Gowers

Lent 1997

These notes are maintained by Paul Metcalfe.

Comments and corrections to pdm23@cam.ac.uk.
Revision: 2.9
Date: 2004/07/26 07:28:53

The following people have maintained these notes.

– date Paul Metcalfe

Contents

Introduction v

1 Topological Spaces 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Building New Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Compactness 5
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Some compact sets . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Consequences of compactness . . . . . . . . . . . . . . . . . . . . . 7
2.4 Other forms of compactness . . . . . . . . . . . . . . . . . . . . . . 7

3 Connectedness 9
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2 Connectedness in R . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Path connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Preliminaries to complex analysis 13

4.1 Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 Complex Integration . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.3 Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.4 Path Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

5 Cauchy’s theorem and its consequences 19

5.1 Cauchy’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.2 Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.3 Consequences of Cauchy’s Theorem . . . . . . . . . . . . . . . . . . 23

6 Power Series 27
6.1 Analyticity and Holomorphy . . . . . . . . . . . . . . . . . . . . . . 27
6.2 Classification of Isolated Singularities . . . . . . . . . . . . . . . . . 31

7 Winding Numbers 35
7.1 Introduction and Definition . . . . . . . . . . . . . . . . . . . . . . . 35
7.2 Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

8 Cauchy’s Theorem (homology version) 41

iii
iv CONTENTS
Introduction

These notes are based on the course “Further Analysis” given by Prof. W.T. Gowers1
in Cambridge in the Lent Term 1997. These typeset notes are totally unconnected with
Prof. Gowers.
Other sets of notes are available for different courses. At the time of typing these
courses were:
Probability Discrete Mathematics
Analysis Further Analysis
Methods Quantum Mechanics
Fluid Dynamics 1 Quadratic Mathematics
Geometry Dynamics of D.E.’s
Foundations of QM Electrodynamics
Methods of Math. Phys Fluid Dynamics 2
Waves (etc.) Statistical Physics
General Relativity Dynamical Systems
Combinatorics Bifurcations in Nonlinear Convection

They may be downloaded from

http://www.istari.ucam.org/maths/.

1 Yes, that Prof. Gowers.

v
vi INTRODUCTION
Chapter 1

Topological Spaces

1.1 Introduction
Definition 1.1. A topological space is a set X together with a collection τ of subsets
of X satisfying the following axioms.

1. ∅, X ∈ τ ;
2. If U1 , . . . , Un ∈ τ , then U1 ∩ · · · ∩ Un ∈ τ ; (that is, τ is closed under finite
intersections)
3. Any union of sets in τ is in τ (or τ is closed under any unions).

Definition 1.2. τ is called a topology on X. The sets in τ are called open sets. A
subset of X is closed if its complement is open.
Examples. 1. If (X, d) is a metric space and τ the collection of open sets (in a
metric space sense) then (X, τ ) is a topological space.
2. If X is any set and τ is the power set of X, (X, τ ) is a topological space. τ is
called the discrete topology on X.
3. If X is any set and τ = {∅, X}, (X, τ ) is a topological space. τ is called the
indiscrete topology on X.
4. If X is any infinite set, and τ = {Y ⊂ X : X \ Y is finite} ∪ {∅} then (X, τ ) is
a topological space. τ is called the cofinite topology on X.
5. If X is any uncountable set, and τ = {Y ⊂ X : X \ Y is countable} ∪ {∅} then
(X, τ ) is a topological space. τ is called the cocountable topology on X.
Definition 1.3. Let A be a subset of a topological space. The closure of A, denoted A,
is the intersection of all closed sets containing A. Note that A is closed and any closed
set containing A contains A.
Definition 1.4. Let A be a subset of a topological space. The interior of A, denoted
int A or A0 is the union of all open sets in A. Note that int A is open and any open set
in A is in int A.
Definition 1.5. The boundary ∂A of a set A is A \ int A.

1
2 CHAPTER 1. TOPOLOGICAL SPACES

Definition 1.6. Let (xn )∞

1 be a sequence in a topological space (X, τ ). We say that
xn converges to x (xn → x) if for every open set U such that x ∈ U , ∃ N such that
n ≥ N ⇒ xn ∈ U . This agrees with the usual definition for metric spaces.

Definition 1.7. Let (X, τ ) and (Y, σ) be topological spaces and let f : X 7→ Y . We
say that f is continuous if for every U ∈ σ, f −1 (U ) ∈ τ . (Or inverse image of an open
set is open.)

It follows from results in Analysis that this definition agrees with the usual  − δ
definition if X and Y are metric spaces.

Definition 1.8. Let (X, τ ) be a topological space and let x ∈ X. A neighbourhood of

x is a set N that contains an open set containing x.

Proposition 1.9. Let (X, τ ) and (Y, σ) be topological spaces and let f : X 7→ Y .
Then the following are equivalent :-

1. f is continuous.

2. For every x ∈ X and every neighbourhood M of f (x) there exists a neighbour-

hood N of x such that f (N ) ⊂ M .

Proof. Firstly do 1 ⇒ 2. M contains an open set U containing f (x). Then N =

f −1 (U ) is a neighbourhood of x such that f (N ) ⊂ U .
Now do the case 2 ⇒ 1. Let U be an open subset of Y . For every x ∈ f −1 (U ), U is
a neighbourhood of f (x). We can find a neighbourhood Nx of x such that f (Nx ) ⊂ U .
Now Nx contains an open set Vx containing x. Let
[
V = Vx .
x∈f −1 (U )

Then f (Vx ) ⊂ U ∀x, so f (V ) ⊂ U and V ⊂ f −1 (U ). But V ⊃ f −1 (U ) so V =

f −1 (U ). V is open, so f −1 (U ) is open and f is continuous.

Definition 1.10. Let (X, τ ) and (Y, σ) be topological spaces and f : X 7→ Y . f is a

homeomorphism if it is continuous with a continuous inverse.

Example. R and (0, 1) with usual metrics are homeomorphic.

1.2 Building New Spaces

Definition 1.11. Let (X, τ ) be a topological space and Y ⊂ X. The subspace topology
on Y is
{U ∩ Y : U ∈ τ }.

Proposition 1.12. Let (X, d) be a metric space, Y ⊂ X and U ⊂ Y . Then the

following are equivalent.

1. U is open in the subspace topology on Y .

2. For every u ∈ U, ∃ δ > 0 such that if v ∈ Y, d(u, v) < δ then v ∈ U .

1.2. BUILDING NEW SPACES 3

Proof. Do 1 ⇒ 2. Let u ∈ U . Since U open in Y , ∃ V ⊂ X, V open such that

U = V ∩ Y . Then ∃ δ > 0 such that d(u, v) < δ ⇒ v ∈ V . Now d(u, v) < δ and
v ∈ Y ⇒ v ∈ V ∩ Y = U.
Now do 2 ⇒ 1. Suppose U satisfies 2. For every u ∈ U , pick δ = δ(u) > 0 such
that d(u, v) < δSand v ∈ Y ⇒ v ∈ U . Let Nu = {v ∈ X : d(u, v) < δ}. Now Nu is
open and V = u∈U Nu is open. Now V ∩ Y = U .
Definition 1.13. Let (X, τ ) be a topological space and ∼ be an equivalence relation
on X. Denote the set X/∼ by Y and let q : X 7→ Y be the equivalence map. (ie if
x ∈ X, q(x) is the equivalence class of x). The quotient topology on Y is {U ∈ Y :
q −1 (U ) ∈ τ }.
N.B. 1. q −1 (U ) is the union of the equivalence classes in U.
2. Notice q is continuous and that the quotient topology is the largest topology
making it so.
Examples. 1. Let τ be the usual topology on R. Then the subspace topology on
Z ⊂ R coincides with the discrete topology on Z.
2. Let τ be as in 1. Then the interval ( 21 , 1] is open in the subspace topology on
[0, 1].
3. Let ∼ be the equivalence relation on R defined by x ∼ y ⇔ x − y ∈ Z. Let [x]
denote the equivalence class of x. Then the map

φ : R/∼ 7→ T ≡ {z ∈ C : |z| = 1}

is both well defined and a homeomorphism.

Definition 1.14. Let (X, τ ) and (Y, σ) be topological spaces. The product topology
on X × Y is the collection of all possible unions of sets of the form U × V with U ∈ τ
and V ∈ σ. Similarly, if (Xi , τi )ni=1 are n topological spaces, the product topology on
Q n Qn
i=1 X i is the collection of all unions of sets i=1 U i with Ui ∈ τi .
Example. If R has its usual topology, then the product topology on R × R is the same
as the usual topology on R × R ≡ R2 .
Proof. Let us write π for the product topology on R × R and σ for the usual (Eu-
clidean) topology. Let U ∈ σ. Then given (x1 , x2 ) ∈ U , ∃ δ > 0 such that
d((x1 , x2 ), (y1 , y2 )) < δ ⇒ (y1 , y2 ) ∈ U . Then the π-open set (x − 2δ , x + 2δ ) ×
(x − 2δ , x + 2δ ) ⊂ U and contains (x1 , x2 ), so U is π-open, given σ ⊂ π. Conversely,
every set of the form A × B with A, B open in R is open in R2 . The union of such sets
is σ-open, giving π ⊂ σ and π = σ.
Exercise to show that X × (Y × Z) = X × Y × Z as topological spaces.
Definition 1.15. Let (X, τ ) be a topological space. A basis for τ (or a basis of open
sets) is a subset β ⊂ τ such that every U ∈ τ is a union of sets in β. The sets in β are
called basic open sets. If x ∈ X, then a basis of neighbourhoods of x is a collection
N of neighbourhoods of x such that every neighbourhood of x contains N ∈ N .
Examples.
The sets U × V with U ∈ τ and V ∈ σ are a basis for the product topology on
(X, τ ) × (Y, σ).
4 CHAPTER 1. TOPOLOGICAL SPACES

1
The sets {y : d(x, y) ≤ n} are a basis of neighbourhoods for a point x in a metric
space.

Proposition 1.16. The quotient topology and product topology are topologies.
Proof. The result for the quotient topology follows easily from the fact that q −1 pre-
serves unions and intersections. For products let (Xi , τu )ni=1 be topological spaces.
Everything is simple except closure under finite intersections. By induction it is suffi-
cient to prove for two sets.
If Ui ∈ τi , call U1 × · · · × Un an open box1 . Firstly, observe that

(U1 × · · · × Un ) ∩ (V1 × · · · × Vn ) = (U1 ∩ V1 ) × · · · × (Un ∩ Vn )

and thus the intersection of two open boxes is an open box. Now take
[ [
Bγ and Cδ ,
γ∈Γ δ∈∆

with all Bγ ’s and Cδ ’s open boxes.

Now [ [ [
Bγ ∩ Cδ = Bγ ∩ Cδ ,
γ∈Γ δ∈∆ γ∈Γ
δ∈∆

which is a union of open boxes, and so open.

1 This is not standard terminology.

Chapter 2

Compactness

2.1 Introduction
Definition 2.1. Let (X, τ ) be a topologicalSspace. An open cover of X is a collection
{Uγ : γ ∈ Γ} of open sets such that X = γ∈Γ S Uγ . If Y ⊂ X then an open cover of
Y is a collection {Uγ : γ ∈ Γ} such that Y ⊂ γ∈Γ Uγ .
A subcover of a cover U = {Uγ : γ ∈ Γ} is a subset V ⊂ U which is still an open
cover.
Examples. 1. {In = (−n, n), n = 1, 2, . . . } is an open cover for R. {In2 } is a
subcover.
2. The intervals In = (n − 1, n + 1) with n ∈ Z form a cover of the reals with no
proper subcover.
Definition 2.2. A topological space (X, τ ) is compact if every open cover has a finite
subcover.
Examples. The open covers mentioned above show that R is not compact. Any finite
topological space is compact, as is any set with the indiscrete topology.

2.2 Some compact sets

Lemma 2.3. Let (X, τ ) be a topological space with Y ⊂ X. Then the following are
equivalent :-
1. Y is compact in the subspace topology.
2. Every cover of Y by Uγ ∈ τ has a finite subcover.
S S
Proof. 1 ⇒ 2. Let Y ⊂ γ∈Γ Uγ with Uγ ∈ τ . Then Y = γ∈Γ (Uγ ∩ Y ) with Uγ
Sn
SnY is compact ∃ γ1 , . . . , γn such that Y = i=1 (Uγi ∩ Y ), which
open in Y . Since
gives that Y ⊂ i=1 USγi .
2 ⇒ 1. Let Y = γ∈Γ Vγ with Vγ open in Y . Since Vγ = Uγ ∩ Y for some Uγ
S Sn
open in X, Y ⊂ γ∈Γ Uγ . By assumption, ∃ γ1 , . . . , γn such that Y ⊂ i=1 Uγi . This
Sn Sn
implies that Y = i=1 (Uγi ∩ Y ) = i=1 Vγi .
Theorem 2.4 (The Heine-Borel Theorem). The closed interval [a, b] ⊂ R is compact.

5
6 CHAPTER 2. COMPACTNESS

S
Proof. Let [a, b] ⊂ γ∈Γ Uγ with Uγ open in R and let

K = {x ∈ [a, b] : [a, x] is contained in a finite union of open sets}.

Let r = sup K. Then r ∈ [a, b], so r ∈ Uγ for some γ. But Uγ is open, so ∃ δ > 0
such that [r − δ, r + δ] ⊂ Uγ . By the definition of r, [a, r − δ] has a finite open cover,
thus [a, r + δ] has a finite open cover. This is a contradiction, unless r = b.

Theorem 2.5. A continuous image of a compact set is compact.

S
Proof. Let X be compact and let f : X 7→ Y be continuous. Now f (X) ⊂ γ∈Γ Uγ
with Uγ open in YS. Since f is continuous, f −1 (Uγ ) is open in X ∀ γ ∈ Γ and
furthermore X = γ∈Γ f −1 (Uγ ). Since X is compact, ∃ γ1 , . . . , γn such that X =
Sn −1
Sn
i=1 f (Uγi ). Thus f (X) ⊂ i=1 Uγi .

Theorem 2.6. A closed subset of a compact set is compact.

S
Proof. Let X be compact and K ⊂ X be closed. Let K ⊂ γ∈Γ Uγ with Uγ open in
S 
X. Now X = (X \ K) ∪ γ∈Γ Uγ . Since X is compact, ∃ a finite subcover and
Sn Sn
X = (X \ K) ∪ ( i=1 Uγi ). Then K ⊂ i=1 Uγi .

Definition 2.7. A topological space (X, τ ) is called Hausdorff, if given any two distinct
x, y ∈ X, there exist disjoint open sets U and V with x ∈ U and y ∈ V .

Examples. 1. Any metric space is Hausdorff. Given x, y distinct, let U = {z :

d(x, z) < d(x, y)/2} and V = {z : d(y, z) < d(x, y)/2}.

2. Any set with more than one element and the indiscrete topology is not Hausdorff.

Theorem 2.8. Every compact subset of a Hausdorff topological space is closed.

Proof. Let X be Hausdorff and K ⊂ X be compact. If x ∈ / K and y ∈ K then one can

find disjoint open sets Uxy and Vxy with x ∈ Uxy and y ∈ Vxy . For fixed S x, the sets
n
Vxy , y ∈ KTform an open coverS of K. Hence ∃ y1 , . . . , yn such that K ⊂ i=1 Vxyi
n n
Let Ux = i=1 Uxyi and S Vx = i=1 Vxyi . Note that Ux ∩ Vx = ∅, x ∈ Ux , K ⊂ Vx
and Ux ∩ K = ∅. But x∈K / Ux = X \ K is open.

Theorem 2.9. A product of finitely many compact sets is compact.

Proof. It is enough to do two sets. The general result follows by induction (and X ×
(Y × Z) = X × Y × Z). S
Let X and Y be compact and let X × Y = γ∈Γ Uγ , Uγ open in X × Y . Uγ is
S of the form V × W , with V open in X and W open in Y . It follows
the union of sets
that X × Y = δ∈∆ Vδ × Wδ with Vδ open in X and Wδ open in Y , Vδ × Wδ ⊂ Uγ .
Let x ∈ X. Now [
{x} × Y ⊂ Vδ × W δ .
δ∈∆
x∈Vδ
Sn
Since Y is compact, one can find δ1 , . . . , δn such that Y = i=1 Wδi . Let Vx =
T n
Si=1
m Sman open cover of X, thus ∃ x1 , . . . , xn such than X =
Vδi . The sets Vx form
V
j=1 xj . Now X ×Y = j=1 Vxj ×Y . But Vxj ×Y has a finite cover of (Vδ ×Wδ )’s.
For each Vδ × Wδ , choose Uγ with Vδ × Wδ ⊂ Uγ to obtain a finite open cover of
X ×Y.
2.3. CONSEQUENCES OF COMPACTNESS 7

Theorem 2.10. A subset of Rn is compact iff it is closed and bounded.

Proof. Let X ⊂ Rn be closed and bounded. Then ∃ M such that d(x, 0) ≤ M ∀ x ∈
X. In particular, X ⊂ [−M, M ]n . But [−M, M ] is compact, so [−M, M ]n is com-
pact. But X is closed, so X is compact.
Conversely, if X is compact then X is closed, since Rn is Hausdorff. If X is not
bounded, then the sets Um = {x ∈ X : d(x, 0) < m} form an open cover with no
finite subcover.

2.3 Consequences of compactness

Theorem 2.11. A continuous real function on a compact metric space is bounded and
attains its bounds.
Proof. The image of such a function is compact, and so closed and bounded. Closed
⇒ the function attains its bounds.
Theorem 2.12. Let X be a compact metric space, let Y be a metric space and let
f : X 7→ Y be continuous. Then f is uniformly continuous.
Proof. Let  > 0 be arbitrary. We must show ∃ δ > 0 ∀ x, y ∈ Xd(x, y) < δ ⇒
d(f (x), f (y)) < . Since f is continuous, ∀ x ∈ X ∃ δx > 0 ∀ y ∈ Xd(x, y) <
2δx ⇒ d(f (x), f (y)) < /2.
Now let Ux = {y : d(x,Sy) < δx }. Then the Ux form an open cover of X, so
n
∃ x1 , . . . , xn such that X = i=1 Uxi . Let δ = min δxi .
Let d(y, z) < δ. Since the Uxi form a cover we can find i such that d(xi , y) < δxi
and d(xi , z) < δxi . By the definition of the Uxi , d(f (y), f (z)) <  by the triangle
inequality.
Lemma 2.13. Let X be a metric space and let K ⊂ X be compact and Y ⊂ X be
closed. Then ∃ x ∈ K such that d(x, Y ) = inf{d(x, w) : x ∈ K, w ∈ Y }.
Proof. Define f : K 7→ R by f (x) = dist(x, Y ). This is continuous, and so attains its
lower bound.
In particular, if K and Y are disjoint, then d(x, Y ) > 0 for every x as Y is closed.
Hence ∃ δ > 0 such that d(x, Y ) ≥ δ ∀ x ∈ X.

2.4 Other forms of compactness

Definition 2.14. X is sequentially compact if every sequence in X has a convergent
subsequence.
Bolzano-Weierstrass is the statement that a closed bounded subset of Rn is sequen-
tially compact.
Theorem 2.15. A compact metric space is sequentially compact.
Proof. Let X be a metric space and let (xn )∞
1 be a sequence with no convergent sub-
sequence.
Now, claim that ∀ x ∈ X ∃ δ > 0 such that d(x, xn ) < δ for at most finitely
many n. Otherwise ∃ x such that d(x, xn ) < m−1 infinitely many times. Now easy to
construct a convergent subsequence.
8 CHAPTER 2. COMPACTNESS

For each x, pick such a δ and let Ux = {y : d(x, y) < δ}. The Ux thus form an
open cover with no finite subcover.
Theorem 2.16. Let X ⊂ Rn . Then the following are equivalent.
1. X is compact.
2. X is sequentially compact.
3. X is closed and bounded.

Proof. We know 3 ⇔ 1 ⇒ 2 from previous theorems. Thus enough to show that

2 ⇒ 3.
/ X such that ∀ m ∃ ym ∈ X such that d(ym , x) < n−1 .
If X is not closed, ∃ x ∈
Then every subsequence of yn converges in Rn to x and does not converge in X. If X
is not bounded then ∀ n ∃ xn such that d(0, xn ) ≥ n, a sequence with no convergent
subsequence.
Chapter 3

Connectedness

3.1 Introduction
Definition 3.1. Let X be a topological space. Suppose U , V ⊂ X such that
1. U , V open,
2. U ∩ V = ∅,
3. X = U ∪ V ,
4. both U and V are non-empty.
Then U , V are said to disconnect X. X is connected if no two subsets disconnect X.
If Y is a subspace of a topological space X, then Y is disconnected in the subspace
topology iff ∃ U, V open in X such that U ∩ Y , V ∩ Y 6= ∅ and Y ⊂ U ∪ V and
Y ∩ U ∩ V = ∅. In this case we shall say that U , V disconnect Y .
Proposition 3.2. Let X be a topological space. Then the following are equivalent.
1. X is connected.
2. Every continuous f : X 7→ Z is constant.
3. The only subsets of X both open and closed are ∅ and X.
Proof. 1 ⇒ 2. Suppose ∃ non-constant f : X 7→ Z. Then we can find m < n such
that both m and n are in f (X). Then f −1 ({k : k ≤ m}) and f −1 ({k : k > m})
disconnect X.
2 ⇒ 1. Suppose U and V disconnect X. Then consider
(
0 x∈U
f (x) =
1 x ∈ V.

1 ⇔ 3. Note that if U ⊂ X, U ∪ (X \ U ) = X.
Proposition 3.3. A continuous image of a connected space is connected.
Proof. Let f : X 7→ Y be continuous with X a connected topological space. Then if
U and V disconnect f (X), f −1 (U ) and f −1 (V ) disconnect X.

9
10 CHAPTER 3. CONNECTEDNESS

3.2 Connectedness in R
Definition 3.4. A subset I of the reals is an interval if, whenever x ≤ y ≤ z, x, z ∈
I ⇒ y ∈ I.
Every interval is of one of these nine forms: [a, b], [a, b), [a, ∞), (a, b], (a, b),
(a, ∞), (−∞, b), (−∞, b], (−∞, ∞).
To see this note that at the upper end of the interval there are three possibilities -
bounded and achieves bound, bounded and does not achieve bound and unbounded.
Similarly for the lower end of the interval.
Theorem 3.5. A subset of R is connected iff it is an interval.
Proof. If X ⊂ R is not an interval we can find x ≤ y ≤ z such that x, z ∈ X and
y∈/ X. Then (−∞, y) and (y, ∞) disconnect X.
Now let I ⊂ R be an interval and suppose U and V disconnect I. We can find
u ∈ U ∩ I and v ∈ V ∩ I and without loss of generality take u < v. Since I is an
interval, [u, v] ⊂ I. Let s = sup{[u, v] ∩ U }. If s ∈ U then s < v. Since U is open
∃ δ > 0 such that (s − δ, s + δ) ⊂ U .
If s ∈ V then (s − δ, s + δ) ⊂ V ⇒ (s − δ, s + δ) ∩ U = ∅.
Corollary 3.6 (The Intermediate Value Theorem). Let a < b and f : [a, b] 7→ R be
continuous. If f (a) < y < f (b) then ∃ x ∈ [a, b] such that f (x) = y.
Proof. If not, f −1 ((−∞, y)) and f −1 ((y, ∞)) disconnect [a, b].

3.3 Path connectedness

Definition 3.7. Let X be a topological space and let x, y ∈ X. A (continuous) path
from x to y is a continuous function φ : [a, b] 7→ X such that φ(a) = x and φ(b) = y.
Definition 3.8. X is path connected if ∀ x, y ∈ X, ∃ a path from x to y.
Proposition 3.9. Path-connectedness implies connectedness.
Proof. Suppose X is a topological space and U and V disconnect X. Let u ∈ U and
v ∈ V . If X is path connected then ∃ continuous φ : [a, b] 7→ X with φ(a) = u and
φ(b) = v. Then φ−1 (U ) and φ−1 (V ) disconnect [a, b].
Definition 3.10. Let X be a topological space and φ : [a, b] 7→ X and ψ : [c, d] 7→ X
be continuous paths with φ(b) = ψ(c). Then the join of φ and ψ, written φ ∨ ψ is
defined as φ ∨ ψ : [a, b + d − c] 7→ X
(
φ(t) a≤t<b
φ ∨ ψ : t 7→
ψ(t + c − b) b ≤ t ≤ b + d − c

Definition 3.11. The reverse of φ, written −φ is the path −φ : [−b, −a] 7→ X with
−φ : t 7→ φ(−t).
Let us write x → y if ∃ a continuous path in X from x to y. This is an equivalence
φ
relation. Writing x −
→ y for “φ is a path from x to y”, we have that :-
φ −φ
1. x −
→ y ⇒ y −−→ x
3.3. PATH CONNECTEDNESS 11

φ ψ φ∨ψ
2. x −
→ y, y −
→ z ⇒ x −−−→ z
χ
3. the path χ : [0, 1] 7→ X, χ(t) = x satisfies x −
→ x.

Thus → is an equivalence relation. The equivalence classes of → are called path-

components.
Definition 3.12. Let X ⊂ Rn . Then a polygonal path is a path φ : [a, b] 7→ X which
is piecewise linear, that is ∃ a = x0 < x1 < · · · < xn = b such that φ((1 − t)xi−1 +
txi ) = (1 − t)φ(xi−1 ) + tφ(xi ) for t ∈ [0, 1] and 1 ≤ i ≤ n. X is said to be
polygonally connected if any two points of X can be joined by polygonal paths.
Write x  y 1 for “∃ a polygonal path in X from x to y”. It is easy to see that 
is an equivalence relation.

Theorem 3.13. Let X ⊂ Rn be open. Then the following are equivalent :-

1. X is connected,
2. X is path-connected,
3. X is polygonally connected.
Proof. 3 ⇒ 2 is trivial and 2 ⇒ 1 has been done before. Thus it suffices to show that
1 ⇒ 3.
Suppose X is connected and for x ∈ X let U = {y ∈ X : x  y}. It remains to
show that U is both open and closed in X, thus since X is connected, U = X.
Now, let y ∈ U . Since X is open ∃ δ > 0 such that Bδ (y) ⊂ X. Then for
z ∈ Bδ (y), y  z, thus x  z by transitivity and U is open.
Let y ∈ X \ U . Since X is open, ∃ δ > 0 such that Bδ (y) ⊂ X. As before,
z ∈ Bδ (y) ⇒ y  z. If x  z then x  y. Thus X \ U is open and U is closed.
N.B. The above argument also shows that if X ⊂ Rn is open then all path components
of X are open.
Lemma 3.14. Let n ≥ 2 and X ⊂ Rn be a compact topological space. Then X c has
only open path-components and precisely one of these is unbounded.
Proof. X c is an open subset of Rn , so its path components are open. X is bounded
so ∃ M such that X ⊂ {y : kyk ≤ M }. It is easy to check that {y : kyk > M } is
path-connected, so it is contained in an unbounded path-component of X c . The other
path-components lie in {y : kyk ≤ M } and are therefore bounded.

1 Not 100% standard notation.

12 CHAPTER 3. CONNECTEDNESS
Chapter 4

Preliminaries to complex
analysis

4.1 Paths
Definition 4.1. A path φ : [a, b] 7→ C is smooth if the function φ is continuously differ-
entiable, with the appropriate one-sided limits at the endpoints a and b. In particular,
φ and φ0 are bounded.
Two paths φ : [a, b] 7→ C and ψ : [c, d] 7→ C are equivalent if ∃ a continuously
differentiable function1 γ : [a, b] 7→ [c, d] with γ 0 (t) > 0 ∀ t ∈ [a, b], γ(a) = c,
γ(b) = d and φ(t) = ψ(γ(t)) ∀ t ∈ [a, b]. Note ∃ δ > 0 such that φ0 (t) ≥ δ ∀ t ∈ [a, b].
This is easily shown to be an equivalence relation.
Definition 4.2. If φ is a path, φ : [a, b] 7→ C, the track φ∗ of φ is defined by φ∗ =
{φ(t) ∈ C : a ≤ t ≤ b}.
N.B. Equivalent paths have the same track.
Definition 4.3. A path φ : [a, b] 7→ C is piecewise smooth
 if ∃ a = x0 < x1 < · · · <
xn = b such that the restriction of φ to [xi−1 , xi ], φ[x ,x ] is smooth. Equivalently,
i−1 i
φ is piecewise smooth if it can be written as the join of finitely many smooth paths.
Two piecewise smooth paths φ : [a, b] 7→ C and ψ : [c, d] 7→ C are equivalent if we
can find a = x0 < x1 < · · · < xn = b and c = y0 < y1 < · · · < yn = d such that for
all 1 ≤ i ≤ n, φ[xi−1 ,xi ] and ψ [yi−1 ,yi ] are equivalent.
Definition 4.4. A piecewise smooth path φ : [a, b] 7→ C is closed if φ(a) = φ(b). It is
simple if φ(x) = φ(y) ⇒ {x, y} ⊂ {a, b} or x = y.

4.2 Complex Integration

A function f : [a, b] 7→ C is said to be Riemann integrable if both its real and imaginary
parts are Riemann integrable. The integral is defined to be
Z b Z b Z b
f (t)dt = <(f (t))dt + ı =(f (t))dt.
a a a
1 Including appropriate one-sided limits at the endpoints.

13
14 CHAPTER 4. PRELIMINARIES TO COMPLEX ANALYSIS

Simple Simple, Closed

Closed

Figure 4.1: Examples of paths

Lemma 4.5. Let f : [a, b] 7→ C be continuous. Then

Z  Z
 b  b
f (t)dt ≤ |f (t)| dt.
 

 a  a

Proof. For suitable θ ∈ R,

Z 
 b  Z b Z b
f (t)dt = eıθ f (t)dt = eıθ f (t)dt
 

 a  a a
Z b Z b
=< eıθ f (t)dt = <(eıθ f (t))dt
a a
Z b
≤ |f (t)| dt using the real result.
a

4.3 Domains
Definition 4.6. A domain is a connected open subset of C.

Examples. Some examples of domains are

1. C,

2. ∆ = {z : |z| < 1},

3. C \ {0},

4. {z : a < |z| < b}.

Definition 4.7. Given x, y ∈ C, let [x → y] be the path φ : [0, 1] 7→ C, φ(t) =

(1 − t)x + ty.

Definition 4.8. A domain D is convex if x, y ∈ D ⇒ [x → y]∗ ⊂ D.

4.4. PATH INTEGRALS 15

Convex Star-shaped, not convex Neither

Figure 4.2: Examples of domains

Definition 4.9. A domain D is star-shaped if ∃ z0 ∈ D such that [z0 → z]∗ ⊂ D ∀ z ∈

D.

A star-shaped domain is sometimes called a star-domain. Note that every (non-

empty) convex domain is star-shaped.

4.4 Path Integrals

Definition 4.10. Let D be a domain, f : D 7→ C be continuous and φ : [a, b] 7→ D be
a smooth path. Then the integral of f along φ is defined as
Z Z b
f (z)dz = f (φ(t))φ0 (t)dt.
φ a


If φ is piecewise smooth, then pick a = x0 < · · · < xn = b such that φ[xi−1 ,xi ] is
smooth and then
Z Xn Z xi
f (z)dz = f (φ(t))φ0 (t)dt.
φ i=1 xi−1

Lemma 4.11. Equivalent paths give the same integral.

Proof. Let D be a domain and φ : [a, b] 7→ D and ψ : [c, d] 7→ D be equivalent smooth

paths. Let f : D 7→ C be continuous and γ : [a, b] 7→ [c, d] give the equivalence. Then
Z Z d Z b
f (z)dz = f (ψ(s))ψ 0 (s)ds = f (ψ(γ(t)))ψ 0 (γ(t))γ 0 (t)dt
ψ c a
Z b Z
= f (φ(t))φ0 (t)dt = f (z)dz.
a φ

If φ and ψ are piecewise smooth write φ = φ1 ∨ · · · ∨ φn and ψ = ψ1 ∨ · · · ∨ ψn ,

with φi and ψi smooth, φi equivalent to ψi . Then
Z n Z
X n Z
X Z
f (z)dz = f (z)dz = f (z)dz = f (z)dz.
φ i=1 φi i=1 ψi ψ
16 CHAPTER 4. PRELIMINARIES TO COMPLEX ANALYSIS

Example. Let D be C \ {0} and f (z) = z n for some n ∈ Z, φ : [0, 2π] 7→ D be

t 7→ eıt . Then (
2πı if n = −1
Z
f (z)dz =
φ 0 otherwise.

Proof.
Z Z 2π Z 2π
f (z)dz = enıt ıeıt dt = ı e(n+1)ıt dt
φ 0 0
h
 e(n+1)ıt 2π
i
n+1 = 0 if n 6= −1
= 0
2πı if n = −1.

Definition 4.12. Let D be a domain and φ : [a, b] 7→ D be a smooth path. Then the
length of φ, L(φ) is defined as
Z b
L(φ) = |φ0 (t)| dt.
a

Lemma 4.13. Let D be a domain, f : D 7→ C be continuous and φ : [a, b] 7→ D be a

smooth path. Then Z 
 
 f (z)dz  ≤ sup |f (z)| L(φ).
  z∈φ∗
φ

Proof.
Z  Z b


f (φ(t))φ0 (t)dt
 
 f (z)dz  =  
  
φ a 
Z b
≤ |f (φ(t))| |φ0 (t)| dt
a
≤ sup |f (z)| L(φ) by real result.
z∈φ∗

Remark. The above generalises easily to piecewise smooth paths.

Henceforth, all paths are piecewise smooth unless otherwise stated.
Proposition 4.14 (Fundamental Theorem of Calculus). Let D be a domain and let
f : D 7→ C be continuous. Suppose f has an antiderivative F (i.e. a function F (z)
such that F 0 (z) = f (z) ∀ z ∈ D). Let φ : [a, b] 7→ D be a path. Then
Z
f (z)dz = F (φ(b)) − F (φ(a)).
φ

Proof. If φ is smooth, then

Z Z b Z b
f (z)dz = f (φ(t))φ0 (t)dt = (F ◦ φ)0 (t)dt = F ◦ φ(b) − F ◦ φ(a).
φ a a
4.4. PATH INTEGRALS 17

In general if a = x0 < x1 < · · · < xn = b and φ[xi−1 ,xi ] is smooth, then the above
argument gives that
Z n
X
f (z)dz = (F (φ(xi )) − F (φ(xi−1 ))) = F (φ(b)) − F (φ(a)).
φ i=1

Corollary 4.15. If D is a domain, f : D 7→ C is continuous with antiderivative F and

φ is a closed path, then Z
f (z)dz = 0.
φ

Proof. Immediate.
Lemma 4.16. Let D be a star-domain and f : D 7→ C be continuous. Then the fol-
lowing are equivalent.
1. f has an antiderivative F on D.
R
2. φ f (z)dz = 0 for all closed paths φ in D.
R
3. ∂T f (z)dz = 0 for the boundary ∂T of any triangle T such that T ⊂ D 2 .
Proof. It is enough to do 3 ⇒ 1. Take z0 ∈ D such that [z0 → z]∗ ⊂ D ∀ z ∈ D and
define Z
F (z) = f (w)dw.
[z0 →z]∗

Then take T to be the triangle with vertices z0 , z and z + h. Since D is open,

[z → z + h]∗ ⊂ D for |h| sufficiently small which gives that T ⊂ D. Now
Z
F (z + h) − F (z) = f (w)dw, so that
[z→z+h]∗

Z 
 
|F (z + h) − F (z) − hf (z)| =  (f (w) − f (z))dw
 
 [z→z+h]∗ 
≤ sup |f (w) − f (z)| |h| .
w∈[z→z+h]

Choose δ > 0 such that |h| < δ ⇒ |f (z + h) − f (z)| < . Then

|F (z + h) − F (z) − hf (z)| ≤  |h| .

2 Including the boundary and interior.

18 CHAPTER 4. PRELIMINARIES TO COMPLEX ANALYSIS
Chapter 5

Cauchy’s theorem and its

consequences

5.1 Cauchy’s theorem

Definition 5.1. Let D be a domain and f : D 7→ C be continuous. f is analytic (or
holomorphic)1 if f is differentiable at z ∀z ∈ D.
Theorem 5.2 (Cauchy’s theorem for triangles). Let D be a domain and T be a
triangle lying entirely in D. If f : D 7→ C is analytic, then
Z
f (z)dz = 0.
∂T
R 
Proof. Let η =  ∂T f (z)dz  and let l = L(∂T ). Now let T0 = T . We can split T into
4 equally sized triangles T 1 , T 2 , T 3 , T 4 as shown, with all boundaries oriented in the
same direction as that of T .

Figure 5.1: Splitting up the triangle

Since the contributions from internal edges cancel,

Z X4 Z
f (z)dz = f (z)dz,
∂T i=1 ∂T i

and ∃ i ≤ 4 such that Z 

  η
 f (z)dz  ≥ .

∂T i 4
1 Outside Cambridge, an analytic function is one which has a power series expansion and a holomorphic

function is C differentiable on a domain.

19
20 CHAPTER 5. CAUCHY’S THEOREM AND ITS CONSEQUENCES

Put T1 = T i for this i and repeat the process. We produce a sequence T0 , T1 , T2 , . . .

such that Z 
  η
 f (z)dz  ≥ n and

∂Tn 4
l
L(∂Tn ) = .
2n
T∞
Since the Tn are closed, we can find z0 ∈ i=1 Tn . As f is differentiable at z0 ,
∀  > 0, ∃ δ > 0 such that

|w − z0 | < δ ⇒ |f (w) − f (z0 ) − (w − z0 )f 0 (z0 )| <  |w − z0 | .

Pick n such that Tn ⊂ Bδ (z0 ). Then

Z  Z 
0
   

 f (z)dz  = 
  (f (z) − f (z0 ) − (z − z0 )f (z0 ))dz 
∂Tn ∂Tn
≤ L(∂Tn ) sup |z − z0 |
z∈∂Tn

≤ L(∂Tn )2 .
R 
But  ∂Tn f (w)dw ≥ 4−n ηl2 . This gives that η < . But  > 0 is arbitrary, so
 
η = 0.
Corollary 5.3 (Cauchy’s Theorem for a star-domain). Let D be a star-domain and
f : D 7→ C be analytic. Then
Z
f (z)dz = 0 for all closed paths φ in D.
φ

Proof. Result true for triangles. Thus f has an anti-derivative and thus
Z
f (z)dz = 0.
φ

5.2 Homotopy
Definition 5.4. Let φ : [0, 1] 7→ D and ψ : [0, 1] 7→ D be piecewise smooth closed
paths in a domain D. A homotopy from φ to ψ is a function γ : [0, 1]2 7→ D such that
1. γ is continuous,
2. γ(0, t) = φ(t) ∀ t ∈ [0, 1],
3. γ(1, t) = ψ(t) ∀ t ∈ [0, 1],
4. ∀ s ∈ [0, 1], the path γs (t) defined by γs (t) = γ(s, t) is closed and piecewise
smooth.

Definition 5.5. ψ is said to be an elementary deformation of φ if ∃ 0 = x0 < x1 <

· · · < xn = 1 and convex open subsets C1 , . . . , Cn ⊂ D such that xi−1 ≤ t ≤ xi ⇒
φ(t) ∈ Ci , ψ(t) ∈ Ci .
5.2. HOMOTOPY 21

Figure 5.2: Two non-homotopic paths.

D

φ (xi )
φ (xi-1 )
ψ (xi )

ψ (xi-1 ) Ci

Figure 5.3: Elementary deformation

Lemma 5.6. Let D be a domain, f : D 7→ C be analytic, φ : [0, 1] 7→ D be a closed

path and ψ be an elementary deformation of φ. Then
Z Z
f (z)dz = f (z)dz.
φ ψ

Proof. Let φi and ψi be the restrictions to [xi−1 , xi ] of φ and ψ respectively. Let

γi = [φ(xi ) → ψ(xi )]. By Cauchy’s theorem for a star-domain,
Z Z Z Z
f (z)dz + f (z)dz − f (z)dz − f (z)dz = 0.
φi γi ψi γi−1

Now summing from i = 1 . . . n gives that

Z Z
f (z)dz = f (z)dz.
φ ψ

Proposition 5.7. Let D be a domain and let φ : [0, 1] 7→ D and ψ : [0, 1] 7→ D be

homotopic. Then ∃ φ = φ0 , φ1 , . . . , φn = ψ such that φi is an elementary deformation
of φi−1 .
22 CHAPTER 5. CAUCHY’S THEOREM AND ITS CONSEQUENCES

Proof. Let γ : [0, 1]2 7→ D be a homotopy from φ to ψ. [0, 1]2 is compact, so γ([0, 1]2 )
is a compact subset of D. Now C \ D is closed and disjoint from γ([0, 1]2 ) so
∃  > 0 such that ∀ z ∈ γ([0, 1]2 ) and w ∈/ D, |z − w| ≥ . Thus ∀(s, t) ∈ [0, 1]2 ,
B (γ(s, t)) ⊂ D. Also, γ is uniformly continuous on [0, 1]2 , so ∃ δ > 0 such that

1/2
(s − s0 )2 + (t − t0 )2 < δ ⇒ |γ(s, t) − γ(s0 , t0 )| < .

j
Now pick n ∈ N such that n2 < δ and let φi = γ ni , i = 0, . . . , n. Let xj = n and
Cij = B (γ(xi , xj )).
j−1 j
But if i−1 i
n ≤ s ≤ n and n ≤ t ≤ n then

1/2 2
(s − s0 )2 + (t − t0 )2 < < δ ⇒ |γ(s, t) − γ(i/n, j/n)| <  ⇒ γ(s, t) ∈ Cij .
n

Thus φi is an elementary deformation of φi−1 .

Corollary 5.8. Let D be a domain, f : D 7→ C be analytic and φ, ψ be homotopic

closed paths in D. Then
Z Z
f (z)dz = f (z)dz.
φ ψ

Proof. Immediate from above.

Definition 5.9. Let D be a domain. A closed path φ is contractible if it is homotopic

to a constant path.

Definition 5.10. A domain D is simply connected if every closed path is contractible.

Corollary 5.11 (Cauchy’s theorem for a simply connected domain).

Let D be a domain and f : D 7→ C be analytic. If the closed path φ is contractible,
then Z
f (z)dz = 0.
φ

If D is simply connected then

Z
f (z)dz = 0 for all closed paths φ.
φ

Proof. Immediate.

Notation.

Br (z0 ) ≡ B(z0 , r) ≡ {z ∈ C : |z − z0 | < r}

thus Br (z0 ) ≡ B(z0 , r) ≡ {z ∈ C : |z − z0 | ≤ r}.

C(z0 , r) ≡ Cr (z0 ) is the path t 7→ z0 + re2πıt for t ∈ [0, 1].

5.3. CONSEQUENCES OF CAUCHY’S THEOREM 23

z0
z
δ r

5.3 Consequences of Cauchy’s Theorem

Theorem 5.12 (Cauchy’s Integral Formula). Let D be a domain and let f : D 7→ C
be analytic. Let z0 , r be such that Br (z0 ) ⊂ D. Then ∀ z ∈ Br (z0 ),
Z
1 f (w)
f (z) = dw.
2πı Cr (z0 ) w−z

Proof. Take  > 0. Let δ > 0 be such that Bδ (z) ⊂ Br (z0 ) and |w − z| = δ ⇒
|f (w) − f (z)| ≤ . Then
 Z   Z 
 1 f (w)   1 f (w) 
f (z) − dw = f (z) − dw
   
 2πı Cr (z0 ) w − z   2πı Cδ (z) w − z 

since γ(s, t) = (1 − s)(z0 + re2πıt ) + s(z + δe2πıt ) is a homotopy from Cr (z0 ) to

Cδ (z) in D \ {z} in which fw−z
(w)
is analytic. Thus

 
 1 Z f (z) − f (w) 
= dw

 2πı Cδ (z) w−z 
1 2πδ
≤ = .
2π δ

But  > 0 is arbitrary, so result follows.

Remark. Note that the proof of Cauchy’s integral formula given does not need the full
strength of homotopy invariance, since Cδ (z) is clearly an elementary deformation of
Cr (z0 )

Theorem 5.13 (Liouville’s Theorem). Every bounded entire function is constant.

Proof. Let f : C 7→ C be analytic and |f (z)| ≤ M ∀ z ∈ C. Take z1 , z2 ∈ C and let

R ≥ 2 max{|z1 | , |z2 |}. Then
24 CHAPTER 5. CAUCHY’S THEOREM AND ITS CONSEQUENCES

   
 1 Z f (w) f (w)
|f (z1 ) − f (z2 )| =  − dw
 
 2πı CR (0) w − z1 w − z2 
 
f (w)(z1 − z2 )
 1 Z 
= dw
 
 2πı CR (0) (w − z1 )(w − z2 ) 
1 2πRM |z1 − z2 |
≤
2π R 2


2
4M |z1 − z2 |
= .
R

But R can be arbitrarily large, so result follows.

Theorem 5.14 (The Fundamental Theorem of Algebra). Every non-constant poly-

nomial has at least one root in C.

Proof. Let p be a non-constant polynomial and suppose that p has no roots. Then the
1
function p(z) is analytic on C. Suppose that

p(z) = an z n + · · · + a0 , with an 6= 0.

Then if
 
|an−1 | + · · · + |a0 |
|z| ≥ max 1, 2 ,
|an |

n n−1
|p(z)| ≥ |an | |z| − (|an−1 | + · · · + |a0 |) |z|
1 n 1
≥ |an | |z| ≥ |an | .
2 2
 
 1 
So ∃ M such that |z| ≥ M ⇒  p(z)  ≤ a2n . Now p(z)1
is continuous and BM (0) is
compact, so p1 is bounded in BM (0). Hence 1
p bounded on all of C and thus constant.
This is a contradiction.

Proposition 5.15. Let g be a continuous function from {z ∈ C : |z − z0 | = r} to C.

Then
Z
g(w)
f (z) = n dw
Cr (z0 ) (w − z)

is analytic on Br (z0 ) and

Z
g(w)
f 0 (z) = n n+1 dw
Cr (z0 ) (w − z)

Proof. Let 2 = r − |z − z0 | such that |w − z0 | = r ⇒ |w − z| ≥ 2. Now

5.3. CONSEQUENCES OF CAUCHY’S THEOREM 25

 
 1 1 
 (w − z − h)n − (w − z)n  =
 
  n−1 
 1 1 X 1 
−
 
 w−z−h w−z
 k n−1−k 
k=0 (w − z − h) (w − z)

 n−1

 h X 1 
=
 
 (w − z − h) (w − z) k n−1−k 
k=0 (w − z − h) (w − z)
hn
≤
n+1
→ 0 as |h| → 0 independently of w.

When |h| ≤ ,
 Z 
 g(w) 
f (z + h) − f (z) − hn dw =
 
n+1
 Cr (z0 ) (w − z) 
Z ! 
n−1
 1 X 1 
hg(w) dw
 

(w − z − h) (w − z) k n−1−k
 Cr (z0 ) (w − z − h) (w − z)
k=0


Using the same estimate as above, the bit in brackets converges to 0 as h → 0.

Since g(w) is bounded on Cr (z0 )∗ for |h| sufficiently small the whole integral is at
most  |h|.
Corollary 5.16. Let D be a domain and f : D 7→ C be analytic. Then f is infinitely
differentiable inside Br (z0 ), Br (z0 ) ⊂ D and ∀ z ∈ Br (z0 ),
Z
n! f (w)
f (n) (z) = dw.
2πı Cr (z0 ) (w − z)n+1

Proof. The case n = 0 is Cauchy’s integral formula. If we have it for n, then the
proposition gives it for n + 1.
Theorem 5.17 (Morera’s Theorem). Let D be a star-shaped domain and
f : D 7→ C be continuous. If Z
f (z)dz = 0
∂T
for all triangles T ⊂ D then f is analytic.
Proof. The condition implies that f has an antiderivative F , which is analytic. F is
therefore infinitely differentiable, so f is analytic.

Remark. Now let D be an arbitrary domain and let z ∈ D. Since ∃  > 0 such that
B (z) ⊂ D and B (z) is star-shaped, one can easily extend Morera’s Theorem to any
domain.
26 CHAPTER 5. CAUCHY’S THEOREM AND ITS CONSEQUENCES
Chapter 6

Power Series

6.1 Analyticity and Holomorphy

P∞
Lemma 6.1. Consider a power series n=0 an (z − z0 )n . If this sum converges for
some z with |z − z0 | = ρ, then it converges for all w with |w − z0 | < ρ and for any
r < ρ, the convergence is uniform in Br (z0 ).
P∞ n
Proof. Since n=0 an (z − z0 )n converges, then ∃ M such that |an | |z − z0 | =
|an | ρn ≤ M ∀ n. If |w − z0 | < ρ, then
 ∞  ∞ 
 w − z0  n
X  
X
n
an (w − z0 )  ≤ M
  

   ρ 
n=N n=N
∞  n
X r
≤
ρ
n=N
 n
r 1
=M
ρ 1 − ρr
→ 0 independently of w.

P∞
Definition 6.2. The radius of convergence of a power series n=0 an (z − z0 )n is
∞
X
R = sup{r : ∃ z such that |z − z0 | < r and an (z − z0 )n converges.}.
n=0

Lemma 6.3. Let D be a domain, φ : [a, b] 7→ D be a path and fn : D 7→ C be

continuous. Suppose fn → f uniformly on φ∗ . Then
Z Z
fn (z)dz → f (z)dz.
φ φ

Proof. Let  > 0. Then ∃ N such that ∀ n ≥ N ∀ z ∈ φ∗ , |fn (z) − f (z)| ≤ 

L(φ) .
Then
Z Z  Z 
   
 fn (z)dz − f (z)dz  =  fn − f dz 
   
φ φ φ
≤ .

27
28 CHAPTER 6. POWER SERIES

But  > 0 is arbitrary, so result follows.

P∞ P∞
Lemma 6.4. Let f (z) = n=0 an (z − z0 )n and g(z) = n=0 bn (z − z0 )n . Suppose
there exists a sequence (zk )∞k=1 → z0 , zk 6= z0 and f (zk ) = g(zk ). Then an = bn for
all n.
Proof. Suppose otherwise. Then let N be minimal such that aN 6= bN . Now let
cn = an − bn . Then
∞
X
f (z) − g(z) = cn (z − z0 )n
n=N
∞
!
X
n n−N −1
= (z − z0 ) cN + (z − z0 ) cn (z − z0 ) .
n=N +1

For |z0 − zk | sufficiently small, then

 ∞ 
 X  1
n−N −1 
|zk − z0 |  cn (z − z0 )  < |cN |

  2
n=N +1

and thus f (zk ) − g(zk ) 6= 0.

P∞ n
Lemma 6.5. Let f (z) = n=0 aP n (z − z0 ) with radius of convergence R. Then f is
0 ∞
analytic in BR (z0 ) and f (z) = n=1 nan (z − z0 )n−1 .
Proof. Let z ∈ BR (z0 ). Pick r such that |z − z0 | < r < R. Let
N
X
fN (z) = an (z − z0 )n .
n=0

Then fN → f uniformly on Br (z0 ). Since |w − z0 | = r ⇒ |w − z| ≥ r − |z − z0 |,

we have
fN (w) f (w) fN (w) f (w)
→ and 2
→ uniformly for |w − z0 | = r.
w−z w−z (w − z) (w − z)2
But Z Z
1 fN (w) 1 f (w)
fN (z) = dw → dw.
2πı Cr (z0 ) w−z 2πı Cr (z0 ) w−z
Therefore Z
1 f (w)
f (z) = dw.
2πı Cr (z0 ) w−z
Hence f is differentiable at z and
Z
1 f (w)
f 0 (z) = dw.
2πı Cr (z0 ) (w − z)2

Also,
Z Z
0 1 fN (w) 1 f (w)
fN (z)= dw → dw = f 0 (z).
2πı Cr (z0 ) (w − z)2 2πı Cr (z0 ) (w − z)2
P∞
Hence f 0 (z) = n=1 nan (z − z0 )n as claimed.
6.1. ANALYTICITY AND HOLOMORPHY 29

Theorem 6.6 (Taylor’s Theorem). Let D be a domain and then f : D 7→ C be an-

alytic. Let z0 ∈ D and let R be such that BR (z0 ) ⊂ D. Then there exist unique
coefficients (an )∞
n=0 such that

∞
X
f (z) = an (z − z0 )n ∀ z ∈ BR (z0 ).
n=0

Proof. Let z ∈ BR (z0 ) and let r be such that |z − z0 | < r < R. Then by Cauchy’s
integral formula,
Z
1 f (w)
f (z) = dw
2πı Cr (z0 ) w−z
Z
1 f (w)
= dw
2πı Cr (z0 ) (w − z0 ) − (z − z0 )
Z
1 f (w) 1
= z−z0 dw
2πı Cr (z0 ) w − z0 1 − w−z 0
∞  n
z − z0
Z
1 f (w) X
= dw.
2πı Cr (z0 ) w − z0 n=0 w − z0
 
 z−z0 
As  w−z 0
 < 1, the convergence is uniform, so can exchange the sum and integral to
get
∞ n Z
X (z − z0 ) f (w)
f (z) = n+1 dw.
n=0
2πı Cr (z0 ) (w − z0 )

To get uniqueness use above lemma.

Theorem 6.7 (Identity Theorem). Let D be a domain and f, g : D 7→ C be analytic.

Suppose zk → z0 (∈ D), zk 6= z0 and f (zk ) = g(zk ) for all k. Then f (z) = g(z) ∀ z ∈
D. In particular, setting g ≡ 0 gives that the zeros of a non-constant analytic function
are isolated.

Proof. Define U = {z ∈ D : f (n) (z) = g (n) (z) ∀n}. Now U 6= ∅ since z0 ∈ U as

the earlier result on uniqueness of power series implies that the Taylor expansions of f
and g at z0 are the same.
Now U is closed, since
∞ 
\ −1
U= f (n) − g (n) ({0}) .
n=0

If z ∈ U , then the Taylor expansions of f and g agree at z, so f ≡ g in some Bδ (z)

and then for y ∈ Bδ (z), f and g must have the same nth derivatives at y. Thus U is
open and since D is connected, U = D.

Proposition 6.8. Let D be a domain, z0 ∈ D and f : D 7→ C be analytic such that

f 6≡ 0. Then there exist a unique k ≥ 0 and a unique analytic function g such that
g(z0 ) 6= 0 and f (z) = (z − z0 )k g(z).
30 CHAPTER 6. POWER SERIES

P∞
Proof. Let the Taylor expansion of f at z0 be n=0 an (z − z0 )n . Now choose N
minimal such that aN 6= 0. (This is do-able since f 6≡ 0.) Thus we can write
∞
X
f (z) = (z − z0 )N an+N (z − z0 )n
n=0

in some Bδ (z0 ). Set

(P
∞
n=0an+N (z − z0 )n if |z − z0 | < δ
g(z) =
(z − z0 )−N f (z) if z 6= z0 .

These two cases agree when 0 < |z − z0 | < δ and f (z) = (z − z0 )N g(z), g(z0 ) =
aN 6= 0. Now if f (z) = (z − z0 )k1 g1 (z) = (z − z0 )k2 g2 (z) ∀ z ∈ D, take k1 ≤ k2
without loss of generality. Then for z 6= z0 we have

g1 (z) = (z − z0 )k2 −k1 g2 (z).

Thus if k1 < k2 , g1 → 0 as z → z0 and so g(z0 ) = 0. Thus g1 (z) = g2 (z) if z 6= z0

and hence g1 ≡ g2 .
Theorem 6.9 (Riemann’s Removable Singularity Theorem). Consider a domain D
with z0 ∈ D. Let f : D \ {z0 } 7→ C be analytic. Then if f is bounded near z0 (i.e.
∃ δ > 0, M such that z ∈ Bδ (z0 ) ⇒ |f (z)| ≤ M ), f → a as z → z0 and the function
(
f (z) z 6= z0
g(z) =
a z = z0

is analytic.
Proof. Define h : D 7→ C by
(
(z − z0 )2 f (z) z 6= z0
h(z) =
0 z = z0 .

This is differentiable at z 6= z0 , and also

 
 h(z) − h(z0 ) 
  ≤ M |z − z0 | when |z − z0 | < δ.
 z − z0 
P∞
Hence h is analytic and so P has Taylor series n=0 an (z−z0 )n . Now h(z0 ) = h0 (z0 ) =
∞
0, so if we define g(z) = n=0 an+2 (z − z0 )n , then g(z) = f (z) for z 6= z0 . Hence
g(z) → a as z → z0 , so f (z) → a as z → z0 .
Proposition 6.10. Let D be a domain, z0 ∈ D and f : D \ {z0 } 7→ C be analytic.
Suppose |f (z)| → ∞ as z → z0 . Then there are a unique integer k ≥ 1 and unique
analytic function g : D 7→ C such that g(z0 ) 6= 0 and f (z) = (z − z0 )−k g(z) when
z 6= z0 .
Proof. Since |f (z)| → ∞ as z → z0 . Then we can find δ > 0 such that z ∈ Bδ (z0 ) ⇒
|f (z)| ≥ 1. Let (
1
0 < |z − z0 | < δ
h(z) = f (z)
0 z = z0 .
6.2. CLASSIFICATION OF ISOLATED SINGULARITIES 31

1
As f (z) → 0 as z → z0 , h is analytic in Bδ (z0 ). Then ∃ k and l : Bδ (z0 ) 7→ C
such that l is analytic and h(z) = (z − z0 )k l(z) if z ∈ Bδ (z0 ), l(z0 ) 6= 0. Since l is
continuous, we can find 0 < δ1 ≤ δ such that l(z) 6= 0 if z ∈ Bδ1 (z0 ). Now let
(
1
0 ≤ |z − z0 | < δ1
g(z) = l(z) k
(z − z0 ) f (z) z 6= z0 .

The definitions agree and g has the required properties. Uniqueness follows as before.

6.2 Classification of Isolated Singularities

Definition 6.11. Let D be a domain, z0 ∈ D and f : D \ {z0 } 7→ C be analytic. Then
z0 is a singularity of f .

• If f is bounded in a neighbourhood of z0 the singularity is called removable

as we can define an analytic function g : D 7→ C such that f and g agree on
D \ {z0 }.

• If |f (z)| → ∞ as z → z0 and if k is the integer from previous proposition the

singularity is a pole of order k.

• All other singularities are called essential.

Theorem 6.12 (Casorati-Weierstrass Theorem). Let D be a domain, z0 ∈ D and

f : D \ {z0 } 7→ C be analytic with an essential singularity at z0 . Then for every
w ∈ C, ∃ zn → z0 such that f (zn ) → w.

Proof. Suppose otherwise. Then we can find w ∈ C such that 0 < |z − z0 | < δ ⇒
|f (z) − w| ≥ . Then g(z) = 1/(f (z) − w) is analytic and bounded in {z : 0 <
|z − z0 | < δ}. Then g has a removable singularity at z0 (in Bδ (z0 )), so we can find an
analytic function h : Bδ (z0 ) 7→ C such that h(z) = g(z) when z 6= z0 . Thus, when
1
z 6= z0 , f (z) = w + h(z) and so f has either a pole or removable singularity at z0 .

Theorem 6.13 (Laurent’s Theorem). 1 Let D be the (non-empty) domain {z : a <

|z − z0 | < b} and let f : D 7→ C be analytic. Then there exist unique coefficients
(an )n∈Z such that
∞
X
f (z) = an (z − z0 )n ∀ z ∈ D.
n=−∞

Proof. Pick r and ρ such that a < r < |z − z0 | < ρ < b. Let γ be the straight line
path from z0 + r to z0 + ρ. 2 It is not hard to see that the closed path Cρ (z0 ) ∨ −γ ∨-
Cr (z0 ) ∨ γ is homotopic in D \ {z} to a path of the form Cδ (z). Hence (by Cauchy’s
Integral Formula and homotopy invariance)
Z Z
1 f (w) 1 f (w)
f (z) = dw − dw.
2πı Cρ (z0 ) w − z 2πı Cr (z0 ) w − z
1 This is not strictly in the schedules, but is covered in Complex Methods. The proof that follows is

slightly sketchy.
2 Unless this goes through z, in which case take a small detour about z. I told you it was sketchy.
32 CHAPTER 6. POWER SERIES

Just as in the proof of Taylor’s theorem, expand in binomial series to get

Z ∞
1 f (w) X
dw = an (z − z0 )n
2πı Cρ (z0 ) w−z n=0
Z
1 f (w)
an = dw
2πı Cρ (z0 ) (w − z0 )n+1

and
Z Z
1 f (w) 1 f (w)
− dw = dw
2πı Cr (z0 ) w−z 2πı Cr (z0 ) z − z0 − (w − z0 )

as before,

−1
X
= an (z − z0 )n
n=−∞
Z
1 f (w)
an = dw
2πı Cr (z0 ) (w − z0 )n+1

or, using homotopy invariance

Z
1 f (w)
an = dw.
2πı Cρ (z0 ) (w − z0 )n+1

For uniqueness, note that

Z ∞ Z
1 f (z) X an n−k−1
dw = (w − z0 ) dw
2πı Cρ (z0 ) (w − z0 )k+1 n=−∞
2πı Cρ (z0 )

= ak .

Let D be a domain, z0 ∈ D, f : D \ {z0 } 7→ C be analytic. P∞ Pick R such nthat

BR (z0 ) ⊂ D. Then f has a Laurent expansion f (z) = n=−∞ an (z − z0 ) in
{z : 0 < |z − z0 | < R}. Let k = inf{n : an 6= 0}. Then if k ≥ 0, f has a removable
singularity at z0 and if k is finite but negative, f has a pole of order −k at z0 . If k is
not finite, then f has an essential singularity at z0 . The converse is also clear.

Theorem 6.14 (Maximum Modulus Theorem). Let D be a domain and f : D 7→ C

be analytic. Suppose |f | has a local maximum. Then f is constant.

Proof. Suppose z0 ∈ D and δ are such that |f (z0 )| ≥ |f (z)| whenever z ∈ Bδ (z0 ).
Then Cauchy’s integral formula implies
Z
1 f (z)
f (z0 ) = dz.
2πı Cδ (z0 ) z − z0
6.2. CLASSIFICATION OF ISOLATED SINGULARITIES 33

Now pick θ such that |f (z0 )| = eıθ f (z0 ).

 
 1 Z eıθ f (z)
Z
f (z)  1
dz  = dz
 
 2πı Cδ (z0 ) z − z0  2πı Cδ (z0 ) z − z0

Z 2π ıθ
1 e f (z0 + δeıφ ) ıφ
= ıδe dφ
2πı 0 δeıφ
Z 2π
1
< eıθ f (z0 + δeıφ ) dφ


=
2π 0
Z 2π
1  ıθ
e f (z0 + δeıφ ) dφ

≤
2π 0
≤ |f (z0 )| .

But we know that equality occurs, so eıθ f (z0 ) = eıθ f (z0 + δeıφ ) ∀φ (using a
result of real analysis). So there exist non-isolated z where f (z) = f (z0 ), hence f is
constant.
34 CHAPTER 6. POWER SERIES
Chapter 7

Winding Numbers

7.1 Introduction and Definition

Definition 7.1. Let z ∈ C \ {0}. A value of log z is a complex number w such that
ew = z. If w = a + ıb is a value of log z, then a = log |z|. It is clear that a + ıb is a
value of log z iff a + ı(b + 2nπ) is a value of log z ∀ n ∈ Z.

Definition 7.2. Let z ∈ C \ {0}. A value of arg z is a real number θ such that z =
|z| eıθ . θ is a value of arg z iff θ + 2nπ is a value of arg z ∀ n ∈ Z iff log |z| + ıθ is a
value of log z.

log z log |z| + ıθ

Definition 7.3. The principal value of is such that −π < θ ≤
arg z θ
π.
log z
Definition 7.4. Let D be a domain such that 0 ∈
/ D. A continuous branch of
arg z
log z
is a continuous function f : D 7→ C such that f (z) is a value of ∀ z ∈ D. This
arg z
need not exist (for instance if D = C \ {0}).

Before doing anything with this, a lemma is useful.

Lemma 7.5. Let D be a simply connected domain and f : D 7→ C be analytic. Then

f has an antiderivative.

Proof. Take z0 ∈ D (D tacitly assumed to be non-empty) and define F by setting

Z
F (z) = f (w)dw, where φ is some path from z0 to z.
φ

By Cauchy’s Theorem, this is well-defined and R the proof that F is an antiderivative

of f is more or less identical to the proof that ∂T f (w)dw = 0 for all triangles T
implies that f has an antiderivative in a star-shaped domain.

Lemma 7.6. Let D be a simply connected domain not containing 0. Let z0 be in D

and w0 be a value of log z0 . Then there is a unique continuous branch L of log on D
such that L(z0 ) = w0 .

35
36 CHAPTER 7. WINDING NUMBERS

Proof. f (z) = z1 is analytic on D so by above lemma has an antiderivative L. By

adding a suitable constant we may assume that L(z0 ) = w0 . Now consider g(z) =
ze−L(z) . g 0 (z) = e−L(z) (1 − zL0 (z)) = 0 so g is constant on D. Now g(z0 ) = 1, so L
is a continuous (and even analytic) branch of log z. If L1 is another continuous branch,
then L1 (z)−L(z)
2πı is an integer for all z ∈ D. But L1 and L are continuous, so this must
be constant and as L(z0 ) = L1 (z0 ), L ≡ L1 .

Definition 7.7. Let D be a simply connected domain, take z0 ∈ C \ D and consider a

path φ : [a, b] 7→ D. The change (or variation) in log(z − z0 )is defined as L(φ(b)) −
L(φ(a)) where L is any continuous branch of log(z − z0 ) on D. Note that this is well
defined, and from the way that L was produced, is also equal to
Z
dz
.
φ z − z0

Now let D be any domain, z0 ∈ D and φ : [a, b] 7→ D be a path such that z0 ∈ / φ∗ .

∗
Since φ is compact and (C \ D) ∪ {z0 } is closed, there exists  > 0 such that B (z) ⊂
/ B (z) for all z ∈ φ∗ .
D and z0 ∈
Also φ is uniformly continuous, so ∃ n ∈ N such that
1
|x − y| ≤ ⇒ |φ(x) − φ(y)| < .
n
Now let xi = a + ni (b − a) for i = 0, 1, . . . , n and let φi be the restriction of φ to
[xi−1 , xi ] and Ci = B (φ(xi )). Then φ∗i ⊂ Ci .

Definition 7.8. The change in log(z − z0 ) along φ is defined to be the sum of the
changes in log for each of the φi . This is not circular – the Ci ’s are manifestly simply
connected and do not contain z0 . Thus the change in log is also equal to
Z
dz
.
φ z − z0

When we chose continuous branches Li of log(z − z0 ) in each Ci , we could, by

adding suitable constants, ensure that Li (φ(xi )) = Li+1 (φ(xi )). If we do that, then
the change along φ is
n
X
Li (φ(xi )) − Li (φ(xi−1 )) = Ln (φ(b)) − L1 (φ(a))
i=1

If φ is closed, this must be 2πkı for some k ∈ Z.

Definition 7.9. The winding number of a closed path φ about z0 is defined as this k. It
is denoted as w(φ, z0 ) and is equal to
Z
1 dz
.
2πı φ z − z0

/ φ∗ , w(φ, z0 ) is an analytic function

From the above formula, we see that for z0 ∈
of z0 with derivative Z
1 dz
.
2πı φ (z − z0 )2
7.2. RESIDUES 37

It is therefore continuous, and as it only takes integer values must be constant on

components of C \ φ∗ . Since φ∗ is compact, there is a unique unbounded component
of C \ φ∗ where the winding number is zero.
To see this, let |z0 | > 2 max{|z| : z ∈ φ∗ }. Then
Z 
1  dz  L(φ)
w(φ, z0 ) = ≤ → 0 as |z0 | → ∞.
2π  φ z − z0  π |z0 |

7.2 Residues
Let D be a domain and f be a function analytic on D except at finitely many points
z1 , . . . , zk . Given z ∈ D, we can find δ > 0 such that Bδ (z) contains none of the
zi unless z = zi , in which case Bδ (z) ∩ {z1 , . . . , zk } = {zi }. Inside Bδ (z), f has a
Laurent expansion
X∞
f (w) = an (w − z)n .
n=−∞

Definition 7.10. The residue of f at z is defined as a−1 and is written Res(f, z).

If z ∈
/ {z1 , . . . , zk }, then Res(f, z) = 0. Now, at zi , write
∞
X
f (z) = a(i) n
n (z − zi ) .
n=−∞

Then Z
1
Res(f, zi ) = f (z)dz.
2πı Cδ (zi )

This gives an alternative definition of Res(f, zi ) not involving Laurent expansions.

Definition 7.11. The principal part of f at zi is defined to be the function

−1
X
gi (z) = a(i) n
n (z − zi ) .
n=−∞

g is analytic on D \ {zi } and f − gi has a removable singularity at zi .

Theorem 7.12 (Cauchy’s Residue Theorem). Let D be a simply connected domain

and f and z1 , . . . , zk be as above. Let φ be a closed path in D such that φ∗ ∩
{z1 , . . . , zk } = ∅. Then
Z k
X
f (z)dz = 2πı Res(f, zi )w(φ, zi ).
φ j=1

Proof. f − (g1 + · · · + gk )1 is analytic on D except for removable singularities at

z1 , . . . , zk . Let h : D 7→ C be analytic such that h(z) = f (z) − (g1 (z) + · · · +
1 The gi ’s are the relevant principal parts.
38 CHAPTER 7. WINDING NUMBERS

gk (z)) ∀z ∈ D \ {z1 , . . . , zk }. Then by Cauchy’s theorem,

Z
h(z)dz = 0 and hence
φ
Z k Z
X
f (z)dz = gj (z)dz
φ j=1 φ

k
X
= 2πı Res(f, zi )w(φ, zi ).
j=1

Definition 7.13. Let D be a domain, z0 ∈ D and let f be a function analytic when

0 < |z − z0 | <  for some  > 0. Recall that if f has a removable singularity or
pole at z0 we can write f (z) = (z − z0 )k g(z) where k and g are uniquely determined,
g(z0 ) 6= 0 and g analytic. Then the integer k is called the order of f at z0 and is
written ord(f, z0 ).
Theorem 7.14. Let D be a domain and let f : D 7→ C be analytic except at finitely
many poles. Suppose also that f has finitely many zeros in D, and let the zeros and
poles be z1 , . . . , zk . Let φ be a closed path in D such that φ∗ ∩ {z1 , . . . , zk } = ∅. Then
k
f 0 (z)
Z
1 X
dz = ord(f, zj )w(φ, zj ).
2πı φ f (z) j=1

Proof. By the residue theorem

k
f 0 (z) f0
Z
1 X
dz = Res( , zj )w(φ, zj ).
2πı φ f (z) j=1
f

Near zj , f = (z − zj )r g(z) with g(zj ) 6= 0 and r = ord(f, zj ). Then

f 0 (z) r g 0 (z)
= +
f (z) z − zj g(z)
0
and Res( ff , zj ) = ord(f, zj ). Summing over j gives the result.

1
R f 0 (z)
Notation. Write ZP (f, φ) for 2πı φ f (z)
dz.
N.B.
f 0 (z)
Z Z
1 1 dz
dz = = w(f ◦ φ, 0).
2πı φ f (z) 2πı f ◦φ z
Theorem 7.15 (Rouché’s Theorem). Let D be a domain, φ be a closed path in D and
f and g be functions with the following properties:
1. f and g are analytic on D except for finitely many poles, none of which lie on
φ∗ .
2. f and f + g have finitely many zeros, none of which lie on φ∗ .
3. |g(z)| < |f (z)| ∀ z ∈ φ∗ .
7.2. RESIDUES 39

Then ZP (f + g, φ) = ZP (f, φ).

Proof. It follows from the definition of order at a point that ord(f +g, z) = ord(f, z)+
g(z)
ord( f +g
f , z). Hence, putting h(z) = 1 + f (z) , we have ZP (f + g, φ) = ZP (f, φ) +

∗  g(z) 
ZP (h, φ). But ZP (h, φ) = w(h ◦ φ, 0), and for z ∈ φ , <h(z) ≥ 1 −  f (z)  > 0. But
there is a continuous branch of log on the right half-plane, so w(h ◦ φ, 0) = 0.
Theorem 7.16 (Local Mapping Theorem). Let D be a domain, z0 ∈ D and f : D 7→
C be analytic and non-constant. Then for  > 0 sufficiently small, there exists δ > 0
such that whenever 0 < |w − w0 | < δ, there are exactly k values of z such that
0 < |z − z0 | <  and f (z) = w, where k = ord(f − w0 , z0 ).
Proof. Choose  > 0 small enough such that whenever 0 < |z − z0 | < 2,
1. f (z) 6= w0 ,

2. f 0 (z) 6= 0,
3. z ∈ D.
Note that 1 and 2 are possible by the Identity Theorem. Now C (z0 )∗ is compact,
so put δ = inf{|f (z) − w0 | : z ∈ C (z0 )∗ } > 0.
Then ∀ z ∈ C (z0 )∗ , |w − w0 | < |f (z) − w0 |. Hence by Rouché’s Theorem,

k = number of zeros up to multiplicity of f (z) − w0 in B (z0 )

= number of zeros of f (z) − w0 (= f (z) − w0 − (w − w0 )).

But every zero of f (z) − w is simple, since f 0 6= 0.

Corollary 7.17 (Open Mapping Theorem). Let D be a domain and f : D 7→ C be
analytic and non-constant. Then if U ⊂ D is open, f (U ) is open.
Proof. Let w0 ∈ f (U ) and z0 be such that f (z0 ) = w0 . The Local Mapping Theorem
provides δ > 0 such that Bδ (w0 ) ⊂ f (B (z0 )) ⊂ f (U ). Hence f (U ) is open.
Remark. The Maximum Modulus Theorem follows immediately.
40 CHAPTER 7. WINDING NUMBERS
Chapter 8

Cauchy’s Theorem (homology

version)

Let D be the domain and φ the path shown. It can be shown by methods R of algebraic
topology that φ is not contractible in D. However, it is also clear that φ f (z)dz = 0
for all analytic functions f on D. We shall ask which paths have this property. This
section of the course is starred.
A chain in a domain D is a finite sequence (φ1 , . . . , φN ) of paths. Two chains
(φ1 , . . . , φN ) and (ψ1 , . . . , ψN ) are directly equivalent if there is a permutation π
of the set {1, 2, . . . , N } such that ψi = φπ(i) for every i. A subdivision of a chain
(φ1 , . . . , φN ) is a chain
(φ11 , φ12 , . . . , φ1M1 , φ21 , . . . , φ2M2 , . . . , φN 1 , . . . , φN MN )
such that φi = φi1 ∨ φi2 ∨ · · · ∨ φiMi for every i. Two chains are equivalent if they
have directly equivalent subdivisions.

41
42 CHAPTER 8. CAUCHY’S THEOREM (HOMOLOGY VERSION)

A cycle is a chain (φ1 , . . . , φN ) such that each φi is a closed path1 . Two cycles
(φ1 , . . . , φN ) and (ψ1 , . . . , ψN ) are homotopic if φi is homotopic ψi for every i. Two
cycles Φ = (φ1 , . . . , φN ) and Ψ = (ψ1 , . . . , ψN ) are homologous if there is a sequence
Φ = Φ0 , Φ1 , . . . , ΦK = Ψ such that for every i, Φi−1 and Φi are either equivalent or
homotopic.
If Φ = (φ1 , . . . , φN ) is a chain in a domain D and f : D 7→ C is continuous, then
PN R
(z)dz is defined to be i=1 φi f (z)dz. If Φ is also a cycle, and z0 ∈ C \ Φ∗ ,
R
Φ
f
then the winding number of Φ about z0 is defined to be
N Z
X 1 dz
w(Φ, z0 ) = w(φi , z0 ) = .
i=1
2πı Φ z − z0

RTheorem 8.1. Let D be a domain and Φ be a cycle homologous in D to a point. Then

Φ
f (z)dz = 0 for every analytic function f : D 7→ C.
Proof. Trivial consequence of homotopy invariance.
This certainly deals with the path shown above. The converse of this theorem is also
true, but somewhat harder to prove. The main result of this section is a characterization
in terms of winding numbers of those cycles for which the integral of any analytic
function vanishes.
Theorem 8.2. Let D be a domain
R and let Φ be a cycle in D with w(Φ, z0 ) = 0 for every
complex number z0 ∈
/ D. Then Φ f (z)dz = 0 for every analytic function f : D 7→ C.
The converse of this theorem is obvious, using the function f (z) = (z − z0 )−1 . In
order to prove this theorem we need another set of definitions from algebraic topology
and 3 easy lemmas.
Given a real number δ > 0, we define Σ(δ) to be the set of all squares S ⊂ C of
the form
{z ∈ C : mδ ≤ <z ≤ (m + 1)δ, nδ ≤ =z ≤ (n + 1)δ}
where m and n are integers. Given such a square S, we denote by ∂S the boundary
of S, oriented anticlockwise. A square complex of mesh δ is a subset Σ ⊂ Σ(δ). If
Σ = {S1 , . . . , SN } is a square complex, then an edge of one of the Si is called internal
if it is shared by some other Sj , and is otherwise called external. The boundary ∂Σ
of Σ is defined to be the chain of all external edges of Σ (with their directions coming
from the orientations of the relevant ∂Si ). We write Σ∗ for the union of the squares
that make up Σ (so that Σ∗ ⊂ C and Σ ⊂ P [C]).
Lemma 8.3. Let Σ be a square complex. Then ∂Σ is equivalent to a cycle.
Proof. We can form a directed graph, where the vertices are all points of the form
δ(m + nı) and the edges are the external edges of Σ (with their directions). It does not
take long to check that at any vertex the number of edges going in equals the number
of edges coming out. Now start at a vertex v which has at least one vertex which has
at least one edge coming out of it, and move along edges in the forward direction for
as long as possible without repeating an edge. As there are finitely many edges this
process must stop, and because of the condition just mentioned must stop at v. The
result is a closed path. If we remove this path, we obtain a directed graph with fewer
edges satisfying the same condition, so by induction the result is proved.
1 It is more usual to define a cycle to be a chain equivalent to what I have called a cycle.
 43

Lemma 8.4. Let Σ = {S1 , . . . , SN } be a square complex and let D be a domain

PN R
containing Σ∗ . Let f : D 7→ C be continuous. Then ∂Σ f (z)dz = i=1 ∂Si f (z)dz.
R

Proof. The contributions to the right-hand side from integrals along internal edges can-
cel, leaving only the integrals along external edges, which by definition is the left-hand
side.
Lemma 8.5. Let Σ = {S1 , . . . , SN } be a square complex and let z0 ∈ int(Σ∗ ). Let f
f (z)
be a function analytic on a domain that includes Σ∗ . Then f (z0 ) = 2πı
1
R
∂Σ z−z0
dz.
Proof. Firstly, suppose z0 lies in the interior of Si for some i. When j 6= i, Cauchy’s
R f (z)
theorem implies that ∂Sj z−z 0
dz. Also, ∂Si is homotopic to Cδ (z0 ) for some δ > 0,
1
R f (z)
so by Cauchy’s integral formula f (z0 ) = 2πı ∂Si z−z0
dz. The result then follows
from above lemma.
Now if z0 lies on an internal edge the result follows by continuity.
Proof of theorem. Let X ⊂ D be C \ {z : w(Φ, z) = 0} = {z : w(Φ, z) 6= 0} ∪ Φ∗ .
Since w(Φ, z) is continuous on the open set C \ Φ∗ , we see that {z : w(Φ, z) = 0} is
open, so that X is closed. Also, since w(Φ, z) = 0 on the unbounded component of
C \ Φ∗ , X is bounded. Thus X is compact, from which it follows that we can choose
δ > 0 such that B2δ (z) ⊂ D whenever z ∈ X.
Define a square complex Σ of mesh δ by taking every square S ∈ Σ(δ) such that
S ∩ C 6= ∅. Note that S is included even if S intersects X only on the boundary - this
is important. It is clear that X ⊂ Σ∗ , but we also have X ⊂ int Σ∗ as the definition of
Σ does not allow z ∈ X to be on an external edge. Also, by our choice of δ, we have
Σ∗ ⊂ D. R f (w)
Hence, if z ∈ Φ∗ ⊂ X, above lemma gives that f (z) = 2πı 1
∂Σ w−z
dw. It follows
that
Z Z Z
1 f (w)
f (z)dz = dwdz
Φ 2πı Φ ∂Σ w −z
Z Z
1 dz
= f (w) dw
∂Σ 2πı Φ w − z
Z
=− f (w)w(Φ, w)dw.
∂Σ

The above change of integrals will not be justified, but we are talking about con-
tinuous functions on closed, bounded subsets of R, where the justification is relatively
easy. Now since X ∩∂Σ = ∅, we have w(Φ, w) = 0 for every w ∈ ∂Σ. This completes
the proof.
44 CHAPTER 8. CAUCHY’S THEOREM (HOMOLOGY VERSION)
References

◦ G.J.O. Jameson, A First Course on Complex Functions, Chapman and Hall, out of
print.
This is a good book for the second half of the course — but probably not as good as
Priestley.

◦ H.A. Priestley, Introduction to Complex Analysis, OUP, 1990.

This book is highly recommended for the second half of the course.

◦ W.A. Sutherland, Introduction to Metric and Topological Spaces, OUP, 1975.

This book is quite good for the first half of the course.

There are a lot of books on this course. Quality is variable.

45
Index

analytic, 19 interval, 10

basis, 3 Laurent’s Theorem, 31

boundary, 1 Liouville’s Theorem, 23
Local Mapping Theorem, 39
Casorati-Weierstrass Theorem, 31
Cauchy’s Integral Formula, 23 Maximum Modulus Theorem, 32
Cauchy’s Residue Theorem, 37 Morera’s Theorem, 25
Cauchy’s Theorem
for a simply connected domain, 22 neighbourhood, 2
for a star-domain, 20 basis of neighbourhoods, 3
for triangles, 19
chain, 41 open cover, 5
equivalent, 41 Open Mapping Theorem, 39
subdivision, 41 open sets, 1
change in log, 36 order, 38
closed sets, 1
closure, 1 path, 10
continuous branch, 35 closed, 13
cycle, 42 contractible, 22
homology, 42 piecewise smooth, 13
homotopic, 42 polygonal, 11
simple, 13
directly equivalent, 41 smooth, 13
domain, 14 track, 13
convex, 14 principal part, 37
simply connected, 22 principal value, 35
star-shaped, 15 Principle of the Argument, 38

elementary deformation, 20 radius of convergence, 27

Removable Singularity Theorem, 30
Fundamental Theorem of Algebra, 24 Rouché’s Theorem, 38
Fundamental Theorem of Calculus, 16
singularity, 31
Heine-Borel Theorem, 5 essential, 31
holomorphic, 19 pole, 31
homeomorphism, 2 removable, 31
homotopy, 20 square complex, 42

Identity Theorem, 29 Taylor’s Theorem, 29

interior, 1 topological space, 1
Intermediate Value Theorem, 10 compact, 5

46
INDEX 47

connected, 9
Hausdorff, 6
path connected, 10
sequentially compact, 7
topology, 1
cocountable, 1
cofinite, 1
discrete, 1
indiscrete, 1
product, 3
quotient, 3
subspace, 2

winding number, 36