MATH 3304

HOLOMORPHIC FUNCTIONS

Lecture notes of Prof. Youssef Helou and Prof. Hicham Gebran

hicham.gebran@gmail.com

Lebanese University, Fanar, Spring 2025

https://hichamgebran.wordpress..com
2

This is a second course on complex variables. First, we shall review what you learned in
Math 2209 but at a deeper and more rigorous level. Second, we shall prove the theorems that
we accepted in Math 2209 as well as the fundamental theorem of algebra. And third, we shall
learn new topics such as the principle of analytic continuation, the maximum principle, the notion
of simple connectedness and the homotopic version of Cauchy’s theorem. We shall also give a
general solution to the problem of existence of antiderivatives.

References

Lars Ahlfors, Complex analysis (McGraw-Hill, 1979).

Henri Cartan, Théorie élémentaire des fonctions analytiques d’une ou plusieurs variables com-
plexes. (Hermann, 1995). Also available in English (Dover, 1995).
Walter Rudin, Real and complex analysis (McGraw-Hill, 1977).
Elias Stein and Rami Shakarchi, Complex analysis (Princeton University Press, 2003).
Contents

1 Functions of a complex variable 5

2 Complex integration 13
2.1 Line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Analycity of holomorphic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Homotopic version of Cauchy’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Laurent series and the residue theorem 39

3.1 Laurent series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 The residue theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3
4 CONTENTS
Chapter 1

Functions of a complex variable

Notational review.

• C denotes the set of complex numbers.

• It is a field (with the usual addition and multiplication).

• A complex number z = x + iy can be thought of as a couple (x, y) ∈ R2 .

• C is a vector space; as a vector space over R it has dimension 2 and as a vector space over
itself it has dimension 1.

• As a vector space over R, C is isomorphic with R2 .

• C is equipped with the usual norm |z| = x2 + y2 (for z = x + iy).

p

• With this norm, C is a Banach space (over R and over itself).

• As a topological space, C is homeomorphic to R2 .

• Therefore C and R2 can be identified algebraically and topologically.

• Throughout this course, Ω will usually denote an open nonempty subset of C.

Definition. Let Ω be an open subset of C and let f : Ω → C. We say that f is holomorphic or

C−differentiable at the point z0 ∈ Ω if the limit
f(z) − f(z0 )
lim
z→z0 z − z0
df
exists in C. In this case, the limit is denoted by f ′ (z0 ) or (z0 ).
dz
If f is holomorphic at every point of Ω, we say that f is holomorphic on Ω and we have a new
function f ′ : Ω → C called the derivative of f.
Remark. Equivalently, f is holomorphic at z0 if there exists c ∈ C such that

f(z) − f(z0 )
∀ ε > 0, ∃ α > 0 such that |z − z0 | < α ⇒ −c ≤ε
z − z0
i.e.
∀ ε > 0, ∃ α > 0 such that |z − z0 | < α ⇒ |f(z) − f(z0 ) − c(z − z0 )| ≤ ε|z − z0 |.
This condition is usually written as

|f(z0 + h) − f(z0 ) − ch| = o(|h|) as h → 0.

5
6 CHAPTER 1. FUNCTIONS OF A COMPLEX VARIABLE

A function f : Ω → C induces a function of two variables f̃ : Ω ⊂ R2 → R2 when we identify C

with R2 . Indeed, let u and v be the real and imaginary parts of f, then we set

f̃(x, y) = (u(x + iy), v(x + iy)).

The function f̃ is called a vector field. We usually still denote this function by the same symbol f.

Example. Let f(z) = z2 . Then

f(z) = (x + iy)2 = x2 − y2 + 2ixy.

So
f̃(x, y) = (x2 − y2 , 2xy).

Figure 1.1: The vector field (x2 − y2 , 2xy) near the origin

Definition. A function f : Ω → C is called R−differentiable at z0 = (x0 , y0 ) ∈ Ω if there exists a

2 × 2−matrix L such that as h = (h1 , h2 ) → 0, we have
 
h1
f̃(x0 + h1 , y0 + h2 ) − f̃(x0 , y0 ) − L = o(∥h∥).
h2

Let u and v be the components of f̃. Recall from differential calculus that if f̃ is R−differentiable,
then u and v have partial derivatives at (x0 , y0 ) and
!
∂u ∂u
∂x ∂y
L = Df̃(x0 , y0 ) = ∂v ∂v .
∂x ∂y

The relation between C−differentiability and R−differentiability is given in the following proposition.

Proposition 1.1. Let Ω be an open subset of C, let f : Ω → C and let z0 = (x0 , y0 ) ∈ Ω. Then the
following conditions are equivalent.

(i) f is C−differentiable (holomorphic) at z0 .

 
α −β
(ii) f is R−differentiable at (x0 , y0 ) and Df(x0 , y0 ) is of the form .
β α

(iii) f is R−differentiable and the Cauchy-Riemann equations are satisfied.

 7

Proof. (i) ⇒ (ii). Suppose first that f is holomorphic at z0 and let c = f ′ (z0 ) = α + iβ. Then, as
h → 0,
|f(z0 + h) − f(z0 ) − ch| = o(|h|).
 
h1
Now observe that for h = = h1 + ih2 , we have
h2

f(z0 + h) − f(z0 ) − ch = f(x0 + h1 , y0 + h2 ) − f(x0 , y0 ) − ch1 − ich2

= f(x0 + h1 , y0 + h2 ) − f(x0 , y0 ) − (α + iβ)h1 − i(α + iβ)h2
= f(x0 + h1 , y0 + h2 ) − f(x0 , y0 ) − (αh1 − βh2 ) − i(βh1 + αh2 )
 
u(x0 + h1 , y0 + h2 ) − u(x0 , y0 ) − (αh1 − βh2 )
=
v(x0 + h1 , y0 + h2 ) − v(x0 , y0 ) − (βh1 + αh2 )
  
α −β h1
= f(x0 + h1 , y0 + h2 ) − f(x0 , y0 ) − .
β α h2
It follows that
  
α −β h1
f(x0 + h1 , y0 + h2 ) − f(x0 , y0 ) − = o(∥h∥).
β α h2
 
α −β
This means that f is R−differentiable at (x0 , y0 ) with Df(x0 , y0 ) = .
β α
(ii) ⇒ (i). Just go back through the steps in (i) ⇒ (ii).
 
α −β
(ii) ⇔ (iii). If f = u + iv and Df(x0 , y0 ) = , then
β α
! 
∂u ∂u 
∂x ∂y α −β
∂v ∂v = .
∂x ∂y
β α

This is equivalent to
∂u ∂v ∂u ∂v
= and =− .
∂x ∂y ∂y ∂x
These relations are the well known Cauchy-Riemann equations. □
In summary,
C-differentiability ⇔ R-differentiability + Cauchy-Riemann equations.
Thus, C−differentiability is a very restrictive condition.

Remark 1. It follows from the proof of the previous proposition that if f is holomorphic at
z0 = (x0 , y0 ) and f = u + iv then
∂u ∂v ∂u ∂u ∂v ∂v
f ′ (z0 ) = +i = −i = +i .
∂x ∂x ∂x ∂y ∂y ∂x
where the partial derivatives are evaluated at (x0 , y0 ).

Remark 2. Let ||Df̃(x0 , y0 )|| = sup∥h∥≤1 ||Df̃(x0 , y0 )h|| be the Euclidean matrix norm of Df̃(x0 , y0 ).
   
h1 α −β
Then, in fact, ||Df̃(x0 , y0 )|| = |f (z0 )|. Indeed let h =
′ and Df̃(x0 , y0 ) = . Then,
h2 β α
 
αh1 − βh2
Df̃(x0 , y0 )h = . Therefore,
βh1 + αh2
h i1/2 h i1/2
||Df̃(x0 , y0 )h|| = (αh1 − βh2 )2 + (βh1 + αh2 )2 = (α2 + β2 )(h21 + h22 ) .
8 CHAPTER 1. FUNCTIONS OF A COMPLEX VARIABLE

Taking the sup over ||h|| = (h21 + h22 )1/2 ≤ 1, we get

||Df̃(x0 , y0 )|| = (α2 + β2 )1/2 = |f ′ (z0 )|.

Examples and counterexamples.

• For n ∈ Z, the function z 7→ zn is holomorphic on C and it derivative is nzn−1 .

• Every polynomial p(z) = a0 + a1 z + · · · + an zn is holomorphic on C and

p ′ (z) = a1 + 2a2 z + · · · + nan zn−1 .

• Finite sums and finite products of holomorphic functions are holomorphic.

• The inverse of a holomorphic function that does not vanish at a point is holomorphic near
that point.

• The composition of two holomorphic functions is holomorphic.

• The function z 7→ z is R−differentiable on C but nowhere holomorphic.

• The function z 7→ |z|2 is R−differentiable on C and holomorphic only at z = 0.

P
Theorem 1.1. (Differentiation term by term). Let n≥0 an (z − a)n be a power series with radius
of convergence R > 0. Then,
P
(1) The series n≥1 nan (z − a)n−1 has the same radius of convergence R.
P
(2) The function S given by S(z) = n≥0 an (z − a)n is holomorphic on B(a, R) and
X
S ′ (z) = nan (z − a)n−1 .
n≥1

S(n) (a)
(3) S has derivatives of all orders and an = for all n ∈ IN.
n!

Definition. Let Ω be an open subset of C and f : Ω → C. We say that f is analyticPat a point

a ∈ Ω, if f has a power series expansion about a, that is, if there exists a power series n≥0 an zn
with radius of convergence r > 0 such that
X
f(z) = an (z − a)n for all z ∈ B(a, r).
n≥0

f is said to be analytic on Ω if it is analytic at every point of Ω.

Remarks.

• By the theorem on differentiation term by term, if a function is analytic at a point, it is

holomorphic in a neighborhood of that point.

• We shall prove in chapter 2, that a function which is holomorphic on an open set is analytic
there.

• However, a function which is holomorphic at only one point is not be analytic at that point.
For example, the function z 7→ |z|2 is holomorphic at 0 but not analytic at 0.
 9

Exercise 1. Show directly that the function f : C∗ → C defined by f(z) = 1

z is analytic on C∗ .

Definition. Let f : Ω → C be analytic and let a ∈ Ω. The germ of f at a is the set

Germ(f, a) = f(a), f ′ (a), . . . , f(n) (a), . . . .

Notation. If f(k) (a) = 0 ∀ k ∈ IN, we write Germ(f, a) = 0.

Proposition 1.2. (Principle of isolated zeros). Let f : Ω → C be analytic and let a ∈ Ω be such that
f(a) = 0 and Germ(f, a) ̸= 0. Then there exists r > 0 such that f(z) ̸= 0 for all z ∈ B(a, r)\{a}.
Otherwise stated, a is an isolated zero of f.
Proof. Since Germ(f, a) ̸= 0, there exists k ∈ IN∗ such that f(k) (a) ̸= 0, that is, the set
{k ≥ 1 | f (a) ̸= 0} is not empty. Let then p = min{k ≥ 1 | f(k) (a) ̸= 0} (every nonempty subset of
(k)

IN has a smallest element). It follows that fk (a) = 0 for all k < p. Now since f is analytic, there
P
exists a disk B(a, r1 ) such that f(z) = ∞ k=0
f(k) (a) k
k! (z − a) for all z ∈ B(a, r1 ). Therefore, f can
be written in the form
∞ (k)
X ∞
X
!
f (a)
f(z) = (z − a)k = (z − a)p bk (z − a)k = (z − a)p f1 (z)
k!
k=p k=0

(p)
with b0 = f p!(a) ̸= 0. Since f1 (a) = b0 ̸= 0, and f1 is continuous, it follows that f1 does not vanish
on a neighborhood say B(a, r) of a. □

Proposition 1.3. Let f : Ω → C be analytic and let a ∈ Ω. Then, the following conditions are
equivalent.

(i) There exists a subset E ⊂ Ω having a as a limit point in Ω and such that f = 0 on E.

(ii) Germ(f, a) = 0.

(iii) f ≡ 0 in a neighborhood of a.

Proof. (i) ⇒ (ii). The continuity of f implies that f(a) = 0. Suppose that Germ(f, a) ̸= 0, then,
by the previous proposition, there exists r > 0 such that f(z) ̸= 0 for all z ∈ B(a, r)\{a}. Since a
is a limit point of E, there exists z0 ∈ E ∩ B(a, r)\{a}. Then, on the one hand, f(z0 ) ̸= 0 because
z0 ∈ B(a, r)\{a}; and on the other hand f(z0 ) = 0 because z0 ∈ E; a contradiction.
P (n)
(ii) ⇒ (iii). Since f is analytic, then f(z) = n≥0 f n!(a) (z − a)n for all z in a neighborhood of a.
But f(n) (a) = 0 for all n ∈ IN, hence f(z) = 0 for all z in a neighborhood of a.
(iii) ⇒ (i). Let B(a, r) ⊂ Ω be a disk on which f vanishes. Note that a is a limit point of the set
E := B(a, r)\{a}. □

Theorem 1.2. (Principle of analytic continuation). Let Ω be an open and connected subset of C
and let f : Ω → C be analytic. If there exists a ∈ Ω such that Germ(f, a) = 0, then f ≡ 0 in Ω.
Proof. Let S = {z ∈ Ω ; Germ(f, z) = 0}. Then

1) S ̸= ∅ because a ∈ S.

2) S is open, for if z0 ∈ S, then f = 0 in a neighborhood U of z0 . But then Germ(f, z) = 0 for all

z ∈ U.
10 CHAPTER 1. FUNCTIONS OF A COMPLEX VARIABLE

Ω
3) S is closed in Ω. Indeed, let z ∈ S = S∩Ω. Then, there exists a sequence (zn ) of S converging
/ S then z is a limit point of the set E = {zn ; n ∈ IN}1 . But Germ(f, zn ) = 0 and so
to z. If z ∈
in particular f(zn ) = 0 so that f = 0 on E. By the previous proposition, Germ(f, z) = 0 and so
z ∈ S, contradicting our assumption. Therefore z ∈ S. This means that S ∩ Ω ⊂ S.

Since Ω is connected, it follows that S = Ω and so f(z) = 0 ∀ z ∈ Ω. □

Remarks.

• Said differently, if an analytic function vanishes on an open set, then it vanishes on the
connected component containing that set.

• Suppose that f : Ω → C is analytic and let Ω ′ be an open and connected set containing
Ω. If g : Ω ′ → C and h : Ω ′ → C are two analytic extensions (continuations) of f then
g = h. Otherwise stated, there is at most one analytic continuation of an analytic function
to a connected set. This is why the previous theorem is called the principle of analytic
continuation.

• By the previous proposition, if an analytic function vanishes on a set E having a limit point
a, then it vanishes on the connected component containing a.

• The principle of analytic continuation is a remarkable property which does not hold for
arbitrary smooth functions f : R → R. For example consider the function f defined by

0 if t ≤ 0
f(t) = − 1/t
e if t > 0.

t
1

It is not difficult to see that f is differentiable and even C∞ . However f vanishes on the open set
] − ∞, 0[ without being identically zero.

Corollary 1.1. Let Ω be open and connected and f : Ω → C be analytic and not identically zero.
Then the set of zeros of f is closed in Ω, discrete and therefore countable.
Recall. Let X be a topological space and let A ⊂ X. Then A is discrete (in the subspace topology)
if and only if no limit point of A (in X) belongs to A.
Proof. Let Z(f) = f−1 (0). Z(f) is closed in Ω because f is continuous. If Z(f) were not discrete
then it would have a limit point a ∈ Z(f). Since f vanishes on Z(f), it follows from the previous
theorem that f vanishes on Ω, contrary to our assumption.
Now for every n ∈ IN∗ , set Kn = {z ∈ C ; |z| ≤ n and dist (z, Ωc ) ≥ n1 }. Then Kn is a closed
and bounded set contained in Ω; therefore it is compact. Note also that Ω = ∪n Kn . Now, for
every n,Sthe set Z(f) ∩ Kn is discrete and compact; therefore it is finite. But since we can write
Z(f) = (Kn ∩ Z(f)) we see that Z(f) is countable as a countable union of finite sets. □
1
If U is neighborhood of z, then, zn ∈ E ∩ U for all n large enough and zn ̸= z because zn ∈ S and z ∈
/ S.
 11

Remark. Z(f) as a subset of C may possess limit points outside Ω. For example let f : C∗ → C
be defined by f(z) = sin( z1 ). Then Z(f) = { nπ
1
; n ∈ Z∗ }. Z(f) is indeed discrete, closed in C∗ (but
not in C) and countable. It has however 0 as a limit point (which is not in C∗ ).

Exercise 2. Let Ω be an open and connected subset of C and let f, g : Ω → C be analytic. Show
that fg ≡ 0 ⇒ f ≡ 0 or g ≡ 0. Show by a counterexample that continuity alone does not imply
the conclusion.

Exercise 3. a) Find all analytic functions f : B(0, 1) → C that satisfy f( n1 ) = f(− n1 ) = 1

n2
for all
integers n ≥ 2.
b) Same question for f( n1 ) = f(− n1 ) = 1
2n+1 for all n ≥ 2.
c) Same question for f( n1 ) = sin( nπ ) for all n ≥ 2.
12 CHAPTER 1. FUNCTIONS OF A COMPLEX VARIABLE
Chapter 2

Complex integration

2.1 Line integrals

Definitions. Let Ω ⊂ C. A path in Ω is a continuous function γ : [α, β] → Ω.

γ
γ(β)

γ∗

γ(α)

α β
Ω

The points γ(α) and γ(β) are called the endpoints of the path; γ(α) is called the initial point
and γ(β) is called the final point. The image of [α, β] under a path γ will be denoted by γ∗ and
we will call it a curve.1

A piece-wise smooth path is a path γ such that there exists a subdivision of [α, β], t0 = α <
t1 < · · · < tn−1 < tn = β for which γ|[tk ,tk+1 ] is C1 .

γ
γ(β)

γ(t1 )

γ(α)

α t1 t2 β
Ω

This means that γ ′ exists (and is continuous) except at finitely many points tk but γ ′ has left
and right limits at tk .

A path γ is called a loop or a closed path if γ(α) = γ(β).

Here are some piecewise smooth loops on which we can integrate. Note that corners and
cusps are perfectly acceptable in this course.
1
Therefore, a curve is a subset of C, whereas a path is a function. A path is also called a parametrization of the
curve and it gives an orientation to it.

13
14 CHAPTER 2. COMPLEX INTEGRATION

Remark 1. A curve can be the image of different paths, i.e., it can be described by different
parametrizations. For example let
1 1
C= (x, y); x2 + y2 = 1 and − √ ≤ x ≤ √ .
2 2

e3iπ/4 eiπ/4

Let γ : [ π4 , 34π ] → C be given by γ(t) = ei(π−t) = −e−it . Then C = γ∗ . Also if δ : [− √12 , √12 ] → C
√
is given by δ(t) = t + i 1 − t2 , then C = δ∗ .
Remark 2. Let α and β be two real numbers such that α < β. Set h(t) = (β − α)t + α. Then h
is an increasing homeomorphism from [0, 1] to [α, β].
If γ : [α, β] → C is a path, then γ ◦ h : [0, 1] → C is also a path having the same underlying
curve i.e., γ∗ = (γ ◦ h)∗ and giving the same orientation. Therefore, we can always assume that
the paths are defined on the unit interval I = [0, 1].

Remark 3. All the paths we shall consider in this course are piece-wise smooth.

Definition. Let γ be a piece-wise smooth path and f : γ∗ → C be continuous. The line integral of
f over γ is
Z n−1 Z tk+1
X
f(z) dz = f(γ(t))γ ′ (t) dt
γ k=0 tk

where t0 = α < t1 < · · · < tn−1 < tn = β and γ|[tk ,tk+1 ] is C1 . Actually, the integral depends only
on the oriented curve and not on the parametrization.

Proposition 2.1. (Invariance under reparametrization). Let φ : [0, 1] → [0, 1] be an increasing

diffeomorphism. Let γ : [0, 1] → C be piece-wise smooth path and γ̃ = γ ◦ φ. Let f : γ∗ → C be
continuous. Then Z Z
f(z) dz = f(z) dz.
γ γ̃

Proof. We have
Z n−1 Z tk+1
X n−1 Z tk+1
X
′
f(z) dz = f(γ̃(t))γ̃ (t) dt = f(γ(φ(t))γ ′ (φ(t))φ ′ (t) dt
γ̃ k=0 tk k=0 tk
n−1 Z φ(tk+1 )
X Z
′
= f(γ(s))γ (s) ds = f(z) dz. □
k=0 φ(tk ) γ
2.1. LINE INTEGRALS 15

Z Z
Remark 1. If φ is decreasing, then f(z) dz = − f(z) dz.
γ γ◦φ
Z Z
Remark 2. If orientation is clear we write, f(z) dz instead of f(z) dz.
γ∗ γ

Definition. The length of a path γ : [α, β] → C is the integral

Zβ n−1 Z ti+1
X
′
L(γ) = |γ (t)| dt = |γ ′ (t)| dt.
α i=0 ti

See the appendix to this section for a justification of this formula. The next proposition was proved
in Math 2209.
Z
Proposition 2.2. f(z) dz ≤ L(γ) sup |f(z)|.
γ z∈γ∗

Corollary 2.1. Let γ be a path in C.

R
(a) (Interchange of lim and ). Let (fn ) be a sequence of continuous functions that converges
uniformly to f on γ∗ . Then Z Z
lim fn (z) dz = f(z) dz.
n→∞ γ γ
P
(b) (Integration term by term). If fn is a series of continuous functions that converges uniformly
on γ∗ , then
Z ∞
X
! ∞ Z
X
fn (z) dz = fn (z) dz.
γ n=1 n=1 γ

Proof. (a) follows from

Z Z
fn (z) dz − f(z) dz ≤ sup |fn (z) − f(z)|L(γ).
γ γ z∈γ∗

(b) follows from (a) and the linearity of the integral. □

Let γ1 : [0, 1] → C and γ2 : [0, 1] → C be two paths such that γ1 (1) = γ2 (0) (we say that γ1
and γ2 are juxtaposable). The product of γ1 and γ2 is the path

γ1 (2t) if 0 ≤ t ≤ 1/2
(γ1 ∗ γ2 )(t) =
γ2 (2t − 1) if 1/2 ≤ t ≤ 1.

t 7→ γ1 (2t) γ2 (1)

γ1 (1) = γ2 (0)

γ1 (0)

0 1 1
2
t 7→ γ2 (2t − 1)

Proposition 2.3. Let γ1 and γ2 be two juxtaposable piecewise smooth paths and let f : γ∗1 ∪γ∗2 → C
be continuous. Then Z Z Z
f(z) dz = f(z) dz + f(z) dz.
γ1 ∗γ2 γ1 γ2
16 CHAPTER 2. COMPLEX INTEGRATION

Proof. See the appendix to this section. □

Corollary 2.1. If γ = γ1 ∗ γ2 ∗ · · · ∗ γm and f : γ∗ → C is continuous then

Z Z Z Z
f(z) dz = f(z) dz + f(z) dz + · · · + f(z) dz. (2.1)
γ γ1 γ2 γm

If γ : [α, β] → C is a path, the opposite path of γ is the path γ− defined by γ− (t) = γ(α+β−t).

Proposition 2.4. Let γ be a piecewise smooth path and f : γ∗ → C be continuous. Then

Z Z
f(z) dz = − f(z) dz.
γ− γ

Proof. Straightforward. □

Definition. An antiderivative of a function f : Ω → C is a holomorphic function F : Ω → C such

that F ′ (z) = f(z) for all z ∈ Ω. If f has an antiderivative, we say that the differential form f(z) dz
is exact.

Remark. We shall see shortly that a continuous or even analytic function need not have an
antiderivative. One of our targets in this chapter is to give conditions under which a function has
an antiderivative.

Theorem 2.1. (Barrow’s rule). If f : Ω → C has an antiderivative F and γ : [α, β] → Ω is a

piecewise smooth path, then
Z
f(z) dz = F(γ(β)) − F(γ(α)).
γ

Proof. Suppose first that γ is C1 . Then

Z Z Zβ Zβ
′
f(z) dz = F (z) dz = F (γ(t))γ (t) dt = (F ◦ γ) ′ (t) dt = F(γ(β)) − F(γ(α)).
′ ′
γ γ α α

More generally, let t0 = α < t1 < · · · < tn−1 < tn = β be a subdivision of [α, β] such that
γ|[tk ,tk+1 ] is C1 . Then
Z Z n−1 Z tk+1
X n−1 Z tk+1
X
′ ′ ′
f(z) dz = F (z) dz = F (γ(t))γ (t) dt = (F ◦ γ) ′ (t) dt
γ γ k=0 tk k=0 tk
X
n−1
 
= (F(γ(tk+1 )) − F(γ(tk )) = F(γ(β)) − F(γ(α)). □
k=0

Z
Exercise. Compute z2 dz where γ(t) = et cos3 t + iet sin3 t for 0 ≤ t ≤ 2π.
γ

Solution. A direct computation of the integral leads to the computable but horrible integral
Z Z 2π  2  
2
z dz = e3t cos3 t + i sin3 t cos3 t − 3 cos2 t sin t + i(sin3 t + 3 sin2 t cos t) dt.
γ 0

The path γ looks like this.

2.1. LINE INTEGRALS 17

z3
However, an antiderivative of the function z2 is given F(z) = . Therefore by Barrow’s rule
3
Z
1  6π 
z2 dz = F(γ(2π)) − F(γ(0)) = F(e2π ) − F(1) = e −1 .
γ 3

Theorem 2.2. Let Ω be a connected open set of C and let f : Ω → C be continuous. Then the
following conditions are equivalent.

(i) f has an antiderivative.

Z
(ii) f(z) dz = 0 for every piecewise smooth loop γ.
γ
Z
(iii) f(z) dz = 0 for every polygonal loop γ.
γ

If moreover Ω is convex, then the above conditions are equivalent to

Z
(iv) f(z) dz = 0 for every triangular loop γ.
γ

Proof. (i) ⇒ (ii). Let F be an antiderivative of f and let γ : [0, 1] → Ω be a piecewise smooth loop.
By the previous theorem, Z
f(z) dz = F(γ(1)) − F(γ(0)) = 0
γ
because γ is closed.
(ii) ⇒ (iii). A polygonal path is piecewise smooth.
(iii) ⇒ (i). Let a ∈ Ω be given and let z vary in Ω. Let if γ and µ be two polygonal paths joining
a to z.
γ z

a µ

Ω
We claim that Z Z
f(z) dz = f(z) dz.
γ µ

Indeed, the path γ ∗ µ− is a closed polygonal path and so

Z
f(z) dz = 0
γ∗µ−
18 CHAPTER 2. COMPLEX INTEGRATION

by assumption. But it follows from Proposition 2.3 and Proposition 2.4 that
Z Z Z Z Z
f(z) dz = f(z) dz + f(z) dz = f(z) dz − f(z) dz.
γ∗µ− γ µ− γ µ

Hence the claim.

This claim permits to define the function F by
Z
F(z) = f(u) du
γz

where γz is a polygonal path joining a to z. We will show that F is holomorphic on Ω and that
F ′ = f.
Consider now a point z ∈ Ω. Since Ω is open, there exists a disc B(z, r) ⊂ Ω. Then
[z, z + h] ⊂ Ω whenever |h| < r.

γz z

Ω
Now take γz+h = γz ∗ [z, z + h]. Then, by Proposition 2.3.,
Z Z Z
f(u) du = f(u) du + f(u) du
γz+h γz [z,z+h]

It follows that
Z Z
F(z + h) − F(z) − hf(z) = f(u) du − f(u) du − hf(z)
γz+h γz
Z
= f(u) du − hf(z)
[z,z+h]
Z Z1
= f(u) du − hf(z) du
[z,z+h] 0
Z1
=h [f(z + th) − f(z)] dt.
0

Therefore, Z1
|F(z + h) − F(z) − hf(z)| ≤ |h| |f(z + th) − f(z)| dt.
0
Let now ε > 0 be given. Since f is continuous at z, there exits α > 0 that we can take < r such
that |f(z + k) − f(z)| < ε whenever |k| < α. Let h satisfy |h| < α. Then for all t ∈ [0, 1], |th| < α.
Therefore |f(z + th) − f(z)| < ε for all t ∈ [0, 1] and so
Z1
|f(z + th) − f(z)| dt ≤ ε.
0

Thus,
|F(z + h) − F(z) − hf(z)| ≤ ε|h|
2.1. LINE INTEGRALS 19

whenever |h| < α. This means that F is holomorphic at z and F ′ (z) = f(z). Since z ∈ Ω was
arbitrary, F is an antiderivative of f on Ω.

Suppose now that Ω is convex. Then (iii)⇒ (iv) because a triangular path is a closed polygonal
path.

(iv) ⇒ (i) Let a ∈ Ω be fixed. For z ∈ Ω, set

Z
G(z) = f(u) du.
[a,z]

This makes sense because now [a, z] ⊂ Ω.

z z+h

Since the integral of f over a triangular loop is zero, it follows that

Z
f(u) du = 0.
∆[a,z,z+h]

By Proposition 2.3,
Z Z Z Z
f(u) du = f(u) du + f(u) du + f(u) du.
∆[a,z,z+h] [a,z] [z,z+h] [z+h,a]

It follows that Z Z Z
f(u) du = f(u) du + f(u) du.
[a,z+h] [a,z] [z,z+h]

Therefore, Z Z Z
G(z + h) − G(z) = f(u) du − f(u) du = f(u) du.
[a,z+h] [a,z] [z,z+h]

Continuing as in the proof of (iii)⇒(i), we deduce that G is holomorphic and that G ′ = f. □

1
Consequence. The function z 7→ has no antiderivative on C∗ . Indeed, take γ(t) = eit , t ∈ [0, 2π].
z
Then
Z Z 2π Z 2π it
1 1 ′ ie
dz = γ (t) dt = it
dt = 2πi ̸= 0.
γ z 0 γ(t) 0 e

Exercise. Show that the function z 7→ z has no antiderivative on C.

Definition. By a triangular region, we mean a triangle together with its interior, that is, the smallest
convex set containing the three vertices of a triangle. If T is a triangular region, its boundary
(which is a triangle) is denoted by ∂T .
20 CHAPTER 2. COMPLEX INTEGRATION

∂T

Remark. Let γ be a triangular loop in an open set Ω ⊂ C. If Ω contains holes the interior of γ
need not be contained in Ω. In this case γ is not the boundary of a triangular region in Ω.

If Ω is convex and γ is a triangular loop in Ω, then the interior of γ is contained in Ω and so

γ = ∂T for a triangular region in T ⊂ Ω.

Lemma. Let z1 , z2 , z3 , z4 be complex numbers and let z = z1 + z2 + z3 + z4 . Then there is

k ∈ {1, 2, 3, 4} such that 41 |z| ≤ |zk |.

Proof. If not, then 41 |z| > |zk | for all k and so |z| > |z1 | + |z2 | + |z3 | + |z4 | ≥ |z| by the triangle
inequality, a contradiction. Alternatively, one can take |zk | = max(|z1 |, |z2 |, |z3 |, |z4 |). Then, by the
triangle inequality, |z| ≤ 4|zk |. □

Theorem 2.3. (Goursat-Cauchy version 1). Let f : Ω → C be holomorphic and let T be a

triangular region in Ω. Then Z
f(z) dz = 0.
∂T
2.1. LINE INTEGRALS 21

R
Proof. Let I = ∂T f(z) dz. By means of the midpoints of the edges T , divide T into four equal
triangular regions T 1 , T 2 , T 3 and T 4 .

T4 ∂T

T2
T1 T3
T

Then,
Z 4 Z
X
I= f(z) dz = f(z) dz.
∂T i=1 ∂T i

By the lemma, there exists among the T i , a triangular region T1 such that
Z
1
|I| ≤ f(z) dz .
4 ∂T1

T1
T2

By induction, we construct a decreasing sequence (Tn ) of triangular regions such that

Z
1
|I| ≤ f(z) dz = |In |. (2.2)
4n ∂Tn
22 CHAPTER 2. COMPLEX INTEGRATION

The sequence (Tn ) is such that diam(Tn ) = diamT

2n → 0 as n → ∞. By a theorem of topology,
there exits a ∈ Ω such that
\∞
Tn = {a}.
n=1

At the point a, f(z) = f(a) + (z − a)f ′ (a) + (z − a)ε(z) where limz→a ε(z) = 0. Then
Z Z Z
′
In = f(z) dz = [f(a) + (z − a)f (a)] dz + (z − a)ε(z) dz.
∂Tn ∂Tn ∂Tn

The first integral on the right hand side vanishes because the function z 7→ f(a) + (z − a)f ′ (a)
′
has an antiderivative (which is z 7→ f(a)z + f (a) 2
2 (z − a) ) and ∂Tn is a loop. Therefore,
Z Z
f(z) dz = (z − a)ε(z) dz
∂Tn ∂Tn
≤ L(∂Tn ) sup |z − a| sup |ε(z)|
z∈∂Tn z∈∂Tn
L(∂T ) diam(T )
≤ sup |ε(z)|
2n 2n z∈Tn

By inequality (1), we get

|I| L(∂T ) diam(T )
n
≤ sup |ε(z)|.
4 2n 2n z∈Tn

Consequently,
|I| ≤ L(∂T )diam(∂T ) sup |ε(z)|.
z∈Tn

Letting n → ∞, we get I = 0. □

Now we have a first answer to the problem of existence of antiderivatives.

Corollary 2.2. Let Ω be open and convex and f : Ω → C be holomorphic. Then f has an
antiderivative on Ω.
Proof. Let γ be a triangular loop ∗
Z in Ω. Since Ω is convex, γ = ∂T where T is a triangular region
in Ω. By the previous theorem f(z) dz = 0. By Theorem 2.2., f has an antiderivative because Ω
γ
is convex. □

Corollary 2.3. Let f : Ω → C be holomorphic. Then every z ∈ Ω has a neighborhood on which f

has an antiderivative. Otherwise stated, a holomorphic function has locally an antiderivative.
Proof. A ball B(z, r) is convex. □
In the next section, we need a slightly different version of the above theorem. □

Theorem 2.4. (Goursat-Cauchy version 2). Let Ω ⊂ C be open and a ∈ Ω. Let f : Ω → C be

continuous and holomorphic on Ω\{a}. If T is a triangular region in Ω, then
Z
f(z) dz = 0.
∂T

Proof. If a ∈ / T , then T ⊂ Ω\{a} and the result follows from version 1 of this theorem. So we
assume that a ∈ T . Let v1 , v2 and v3 be the vertices of T . We will distinguish between three cases:
a is a vertex of T , a is interior to T and a is one side of T .
2.1. LINE INTEGRALS 23

Case 1. After relabeling the vertices of T , we can assume a = v1 . Let ε > 0 be given. Choose
a2 ∈ [v1 , v2 ] and a3 ∈ [v1 , v3 ] such that |v1 − a2 | + |a2 − a3 | + |a3 − v1 | < ε (that is, a2 and a3 are
sufficiently close to v1 so that the perimeter of the triangle v1 a2 a3 is small).

v1 = a

a2 a3

∂T

v2 v3
T

Then,
Z Z Z Z
f(z) dz = f(z) dz + f(z) dz + f(z) dz.
∂T ∆v2 a3 a2 ∆v3 a3 v2 ∆v1 a2 a3

The first two integrals are zero by version 1 of this theorem. Let M be a bound of |f(z)| on T .
Then,
Z Z
f(z) dz = f(z) dz ≤ L(v1 a2 a3 )M ≤ εM.
∂T ∆v1 a2 a3
Z
Since ε was arbitrary, we conclude that f(z) dz = 0.
∂T

Case 2. a is interior to T .

v1

a
∂T

v2 v3
T

Ω
24 CHAPTER 2. COMPLEX INTEGRATION

Then, Z Z Z Z
f(z) dz = f(z) dz + f(z) dz + f(z) dz.
∂T ∆av1 v2 ∆av2 v3 ∆av3 v1

By step 1, each of these integrals is zero.

Case 3. a is on a side of T . This case is similar to Case 2. □

Remark. The assumptions of this theorem imply that actually f is holomorphic at a (see the next
section or chapter 3).
We shall deal later with integrals depending on a complex parameter. The following result is
fundamental and will be useful.

Theorem 2.5. (Differentiation under the integral sign). Let φ : [a, b] × Ω → C be continuous,
holomorphic with respect to the second variable and such that ∂2 φ is continuous on [a, b] × Ω.
We set Z b
f(z) = φ(t, z) dt.
a

Then f is holomorphic and

Zb
′
f (z) = ∂2 φ(t, z) dt.
a

Proof. Let z ∈ Ω. Then [z, z + h] ⊂ Ω for all h whose modulus is small enough. Let g(z) =
Zb
∂2 φ(t, z) dt. Then g is continuous (prove this). Now,
a
Z z+h Z z+h Z b
g(u) du = ∂2 φ(t, u) dt du
z z a
Z b Z z+h
= ∂2 φ(t, u) du dt by Fubini’s theorem
a z
Zb
= [φ(t, z + h) − φ(t, z)] dt by Barrow’s rule
a
= f(z + h) − f(z).

Therefore,
Z z+h
f(z + h) − f(z) 1
= g(u) du.
h h z
Z
1 z+h
Now we claim that limh→0 g(u) du = g(z). Let ε be given. Since g is continuous at z, there
h z
is δ > 0 such that |g(u) − g(z)| < ε for all u such that |u − z| < δ. Then for |h| < δ, we have
Z z+h Z z+h Z z+h
1 1 1
g(u) du − g(z) = g(u) du − g(z) du
h z h z h z
Z z+h
1
= [g(u) − g(z)] du
h z
1
≤ |h|ε = ε.
|h|

It follows from the claim that f is holomorphic at z and f ′ (z) = g(z). □

2.1. LINE INTEGRALS 25

Appendix
In this optional appendix, we will

• give a justification of the formula for the length of a path and

• prove Proposition 2.3.

Definition. The length of a path γ : [α, β] → C is the integral

Zβ n−1 Z ti+1
X
′
L(γ) = |γ (t)| dt = |γ ′ (t)| dt.
α i=0 ti

Justification. It is enough to justify the formula when γ is C1 . The idea is to approximate the path
by polygonal paths (or broken line paths).

Recall that a partition of [α, β] is a finite sequence P = {t0 , t1 , . . . , tn } such that t0 = α < t1 <
· · · < tn = β. Let ∆tk = tk+1 − tk . The norm of the partition is ||P|| = max ∆tk . If P is a partition
0≤k≤n
of [α, β] the length of the polygonal line joining the points γ(tk ) is

X
n−1
SP = |γ(tk+1 − γ(tk )| .
k=0

γ(tk+1 )

γ(tk )

We will show that Zβ

SP → |γ ′ (t)| dt as ||P|| → 0.
α
Now Z tk+1 Z tk+1 Z tk+1
γ ′ (t) dt = γ ′ (tk ) dt + γ ′ (t) − γ ′ (tk ) dt


γ(tk+1 ) − γ(tk ) =
tk tk tk
Z tk+1
= γ ′ (tk )∆tk + γ ′ (t) − γ ′ (tk ) dt.


tk

By the second form of the triangle inequality |x| − |y| ≤ |x − y|

|γ(tk+1 ) − γ(tk )| − |γ ′ (tk )|∆tk ≤ γ(tk+1 ) − γ(tk ) − γ ′ (tk )∆tk

Z tk+1 Z tk+1
′ ′
γ ′ (t) − γ ′ (tk ) dt.


≤ γ (t) − γ (tk ) dt ≤
tk tk
26 CHAPTER 2. COMPLEX INTEGRATION

It follows that

X
n−1 X
n−1 X
n−1
′
SP − γ (tk ) ∆tk = |γ(tk+1 ) − γ(tk )| − γ ′ (tk ) ∆tk
k=0 k=0 k=0
X
n−1
|γ(tk+1 ) − γ(tk )| − γ ′ (tk ) ∆tk


=
k=0
X1
n−
≤ |γ(tk+1 ) − γ(tk )| − γ ′ (tk ) ∆tk
k=0
n−1 Z tk+1
X
≤ γ ′ (t) − γ ′ (tk ) dt.
k=0 tk

Let now ε > 0 be given. Since γ ′ is uniformly continuous on [α, β], there exists δ > 0 such that
|γ(t) − γ(s)| < ε whenever |t − s| < δ. Let ||P|| < δ, then ∆tk < δ for all k and |t − tk | < δ for all
t ∈ [tk , tk+1 ]. It follows that |γ ′ (t) − γ ′ (tk )| < ε for all t ∈ [tk , tk+1 ]. Therefore,
Z tk+1 Z tk+1
′ ′
γ (t) − γ (tk ) dt ≤ ε dt = ε∆tk
tk tk

and so
n−1 Z tk+1
X X
n−1
γ ′ (t) − γ ′ (tk ) dt ≤ ε∆tk = (β − α)ε.
k=0 tk k=0

Therefore, we have proved the following. For all ε > 0 there exists δ > 0 such that whenever
||P|| < δ, we have
X
n−1
SP − γ ′ (tk ) ∆tk ≤ (β − α)ε.
k=0

This means precisely that

X
n−1
SP − γ ′ (tk ) ∆tk → 0 as ||P|| → 0.
k=0

The Riemann sum

X
n−1 Zβ
γ ′ (tk ) ∆tk → γ ′ (t) dt as ||P|| → 0.
k=0 α

Thus,
Zβ
SP → γ ′ (t) dt as ||P|| → 0.
α

□
Proposition 2.3. Let γ1 and γ2 be two juxtaposable piecewise smooth paths and let f : γ∗1 ∪γ∗2 → C
be continuous. Then Z Z Z
f(z) dz = f(z) dz + f(z) dz.
γ1 ∗γ2 γ1 γ2

Proof. It’s better to start with the case where γ1 : [0, 1] → C and γ2 : [0, 1] → C are C1 . In this
case, γ := γ1 ∗ γ2 is piecewise smooth with respect to the partition {0, 21 , 1}, that is γ is C1 on [0, 21 ]
2.2. ANALYCITY OF HOLOMORPHIC FUNCTIONS 27

and on [ 21 , 1] (but need not be differentiable at 21 . Then, by definition,

Z Z 1/2 Z1
′
f(z) dz = f(γ(t))γ (t) dt + f(γ(t))γ ′ (t) dt
γ 0 1/2
Z 1/2 Z1
= f(γ1 (2t))2γ1′ (2t) dt + f(γ2 (2t − 1))2γ2′ (2t − 1) dt
0 1/2
Z1 Z1
= f(γ1 (s))γ1′ (s) ds + f(γ2 (s))γ2′ (s) ds
Z0 Z 0

= f(z) dz + f(z) dz.

γ1 γ2

Now more generally, let γ1 be piecewise smooth on the partition {t0 , t1 , . . . , tn } and γ2 be piecewise
smooth on the partition {s0 , s1 , . . . , sm }. Then γ is piecewise smooth on the partition
t0 t1 tn s0 + 1 1 sm + 1
= 0, , . . . , = = ,..., = {α0 , . . . , αn+m } .
2 2 2 2 2 2
Then, by definition,
Z X 1 Z αk+1
n+m−
f(z) dz = f(γ(t))γ ′ (t) dt
γ k=0 αk

X Z
n−1 αk+1 X 1 Z αk+1
n+m−
′
= f(γ(t))γ (t) dt + f(γ(t))γ ′ (t) dt
k=0 αk k=n αk

Z n−1 Z tk+1 /2
X X1 Z (sk+1 +1)/2
m−
f(z) dz = f(γ1 (2t))2γ1′ (2t) dt + f(γ2 (2t − 1))2γ2′ (2t − 1) dt
γ k=0 tk /2 k=0 (sk +1)/2
n−1 Z tk+1
X X1 Z sk+1
m−
= f(γ1 (s))γ1′ (s) dt + f(γ2 (s))γ2′ (s) dt
tk k=0 (sk
Zk=0 Z
= f(z) dz + f(z) dz.
γ1 γ2

2.2 Analycity of holomorphic functions

Given a piecewise smooth loop γ and a point a ∈/ γ∗ , we will define a integer I(γ, a) that counts
the oriented number of times the loop winds around that point.

γ
a1

a3
a2

Figure 2.1: I(γ, a1 ) = 0; I(γ, a2 ) = 1; I(γ, a3 ) = 2

28 CHAPTER 2. COMPLEX INTEGRATION

γ
a1

a3
a2

Figure 2.2: I(γ, a1 ) = 0; I(γ, a2 ) = 2; I(γ, a3 ) = 4

γ
a1

a3
a2

Figure 2.3: I(γ, a1 ) = 0; I(γ, a2 ) = −1; I(γ, a3 ) = −2

Definition. Let γ : [α, β] → C be a piecewise smooth loop and let a ∈ / γ∗ . The index or the
winding number of γ with respect to a is the number
Z Z
1 dz 1 β γ ′ (t)
I(γ, a) = = dt.
2πi γ z − a 2πi α γ(t) − a
Z
1 2π rieit
Example. Let γ(t) = a + reit , 0 ≤ t ≤ 2π. Then I(γ, a) = = 1.
2πi 0 reit

Recall. 1) ez = 1 ⇐⇒ z = 2nπi for some n ∈ Z.

2) exp : C → C∗ is surjective (eα+iθ = z if α = ln |z| and θ = arg(z)).

Proposition 2.5. I(γ, a) ∈ Z.

Proof. Let γ : [0, 1] → C be a piecewise smooth loop not passing through a. Set
Z t
γ ′ (s)

g(t) = exp ds , 0 ≤ t ≤ 1.
0 γ(s) − a

Then, except at a finite number of points where γ is not differentiable, we have by the funda-
mental theorem of calculus
Z t
γ ′ (s) γ ′ (t) γ ′ (t)

′
g (t) = exp ds = g(t) .
0 γ(s) − a γ(t) − a γ(t) − a
2.2. ANALYCITY OF HOLOMORPHIC FUNCTIONS 29

g(t) g ′ (t)[γ(t) − a] − γ ′ (t)g(t)

Let φ(t) = . Then φ ′ (t) = = 0. It follows that φ is piecewise
γ(t) − a (γ(t) − a)2
constant. But since it is continuous, it is constant and so φ(1) = φ(0). This means that

exp(2πiI(γ, a)) 1
= .
γ(1) − a γ(0) − a

Consequently, (since γ(0) = γ(1)), we get exp(2πiI(γ, a)) = 1 and therefore I(γ, a) ∈ Z. □
Remark. I(γ, a) is the number of times γ winds around a in the counterclockwise direction. Let
us see why. Set
Zt
γ ′ (s)
f(t) = ds.
0 γ(s) − a
γ (t) ′
By the fundamental theorem of calculus, f is piecewise smooth and f ′ (t) = γ(t)−a except at finitely
many points. Keeping the same notation as in the above proof, we have φ(t) = φ(0) and g(0) = 1.
Therefore,
γ(t) − a
ef(t) = g(t) = .
γ(0) − a
1
Now, there exists a complex number C such that eC = γ(0)−a . Therefore ef(t)−C = γ(t) − a.
Setting h(t) = f(t) − C, we get
γ ′ (t)
(1) h ′ (t) = γ(t)−a and

(2) eh(t) = γ(t) − a.

It follows from the first point that

Z1
1 1
I(γ, a) = h ′ (t) dt = (h(1) − h(0)).
2πi 0 2πi

Next, let α(t) and θ(t) denote respectively the real and imaginary part of h(t). It follows from the
second point that α(t) = ln |γ(t) − a| and so α(1) = α(0). Therefore h(1) − h(0) = i(θ(1) − θ(0))
and so
1
I(γ, a) = (θ(1) − θ(0)).
2π
But θ(t) is the angle that γ(t) − a makes with the positive x-axis (it follows from point 2 that
eα(t) eiθ(t) = γ(t) − a). Therefore θ(1) − θ(0) is the variation of this angle as γ goes from 0 to 1.
Therefore I(γ, a) is the oriented number of times γ winds around a (I(γ, a) is positive if γ winds
counterclockwise, negative it it winds clockwise, and 0 if a is outside γ).

Exercise. a) Check that I(γ− , a) = −I(γ, a).

b) Let γ(t) = eint for t ∈ [0, 2π] and n ∈ Z. Find I(γ, 0).
c) More generally, let γ : [0, 1] → C be a loop. Extend γ by periodicity to R. Let µ(t) = γ(nt) for
t ∈ [0, 1]. Show that I(µ, a) = nI(γ, a).

Proposition 2.6. Let γ be a piecewise smooth loop in C. Then the function ζ 7→ I(γ, ζ) from
C\γ∗ → Z is constant on the connected component of C\γ∗ .
Proof. First method. Let f(ζ) = I(γ, ζ). The function f is continuous and takes values in Z.
Therefore the image of every connected component is a connected subset of Z. But the only
connected subsets of Z are singletons.
30 CHAPTER 2. COMPLEX INTEGRATION

Z Z
1 dz 1 1 γ ′ (t)
Second method. We can write f(ζ) = = dt. It follows from Theorem
2πi γ z − ζ 2πi 0 γ(t) − ζ
2.5, that f is holomorphic on C\γ∗ and that
Z
′ 1 dz
f (ζ) = .
2πi γ (z − ζ)2

Now the function z 7→ (z−ζ) 1

2 has an antiderivative on C\{ζ} which is z 7→ z−ζ . Consequently,
−1
Z
dz
2
= 0. It follows that f ′ (ζ) = 0 for all ζ ∈ C\γ∗ . This implies that f is constant on the
γ (z − ζ)
connected components of C\γ∗ (draw a figure). □

Proposition 2.7. I(γ, a) = 0 whenever a belongs to the unbounded connected component of

C\γ∗ .

Proof. Since γ∗ is bounded, it is contained in some disk B(0, r). Take a ∈ C such that |a| > r
(draw a figure). Then a belongs to the unbounded component of C\γ∗ . Then, for z ∈ γ∗ ,
1 1 1
z−a ≤ |a|−|z| ≤ |a|−r
Z
1 dz L(γ)
|I(γ, a)| = ≤ .
2π γ z − a 2π(|a| − r)
This estimate is true for any a outside C, and so letting |a| → ∞ we get I(γ, a) = 0. □

Riddle. Someone asks you to compute the derivative of the function f : R∗ → R given by
1
f(t) = arctan t + arctan .
t
When you compute it, you find f ′ (t) = 0. You conclude that f is constant. However, f(1) = π
2 and
f(−1) = − π2 . What’s wrong?

Proposition 2.8. (Cauchy’s integral formula in a special case). Let f : B(z0 , r) → C be holomor-
phic. Let γ(t) = z0 + r0 eit where 0 < r0 < r and t ∈ [0, 2π]. Then
Z
1 f(u)
f(z) = du
2πi γ u − z
for all z ∈ B(z0 , r0 ).
Proof. Fix z ∈ B(z0 , r0 ) and define g : B(z0 , r) → C by

f(u)−f(z)
u−z if u ̸= z
g(u) = ′
f (z) if u = z.

Then g is continuous
Z on B(z0 , r) and holomorphic on B(z0 , r)\{z}. Now by version 2 of Goursat-
Cauchy theorem, g(u) du = 0 for every triangular loop µ in B(z0 , r). But it follows from Theorem
Z µ Z
f(u) − f(z)
2.2 that g(u) du = 0. This means that / γ∗ . Therefore,
du = 0 since z ∈
γ γ u − z
Z Z Z
1 f(u) 1 f(z) 1 du
du = du = f(z) = f(z)I(γ, z).
2πi γ u − z 2πi γ u − z 2πi γ u − z

Now, since z and z0 are in the same connected component of C\γ∗ (both are inside γ), it follows
that
Z Z Z
1 du 1 2π γ ′ (t) 1 2π r0 ieit
I(γ, z) = I(γ, z0 ) = = dt = dt = 1. □
2πi γ u − z0 2πi 0 γ(t) − z0 2πi 0 r0 eit
2.2. ANALYCITY OF HOLOMORPHIC FUNCTIONS 31

Z u
e
Exercise. a) Compute du where γ is a parametrization of the unit circle (positively oriented).
γ u
Z
1
b) Compute 2
dz where γ is a parametrization of the circle |z| = 2 (positively oriented).
γ z +1

Theorem 2.6. (Analycity of holomorphic functions). Let Ω ⊂ C be an open set and let f : Ω → C
be holomorphic. Then f is analytic on Ω.
Proof. Let z0 ∈ Ω. Then, there exists r > 0 such that B(z0 , r) ⊂ Ω. Let now r0 satisfy 0 < r0 < r.
Let γ(t) = z0 + r0 eit for 0 ≤ t ≤ 2π. Let z ∈ B(z0 , r0 ). By Cauchy’s integral formula,
Z
1 f(u)
f(z) = du.
2πi γ u − z

Now observe that

1 1 1 1
= = z−z0
u−z (u − z0 ) − (z − z0 ) u − z0 1 − u−z
0

z−z0 |z−z0 |
and that u−z0 = r0 < 1 for u ∈ γ∗ . Therefore

∞
X (z − z0 )n
1
= .
u−z (u − z0 )n+1
n=0

Consequently,
X ∞
f(u) f(u)
= (z − z0 )n .
u−z (u − z0 )n+1
n=0

Therefore,
Z Z X
∞ 
1 f(u) 1 f(u) n
f(z) = du = (z − z0 ) du.
2πi γ u−z 2πi γ (u − z0 )n+1
n=0

Now let M be a bound of |f| on γ∗ . Then,

n
|z − z0 |

f(u) M
(z − z0 )n ≤
(u − z0 )n+1 r0 r0
P
for all u ∈ γ∗ . This means that the series ∞ f(u) n
n=0 (u−z0 )n+1 (z − z0 ) is normally (and therefore
uniformly) convergent. Thus, we can interchange the operations of integration and summation.
Therefore,
∞ 
X Z  ∞
X
1 f(u) n
f(z) = du (z − z 0 ) = an (z − z0 )n ∀z ∈ B(z0 , r0 )
2πi γ (u − z0 )n+1
n=0 n=0

where Z
1 f(u)
an = du
2πi γ (u − z0 )n+1
for all n ∈ IN. □

Remark 1. If f is analytic at a point z0 , then it is analytic in a neighborhood of z0 . Indeed,

we know that f is holomorphic on a neighborhood U of z0 . Let z1 ∈ U, then there exists a disk
B(z1 , r1 ) ⊂ U. The function f is holomorphic on B(z1 , r1 ), and according to the previous theorem,
it is analytic at z1 .
32 CHAPTER 2. COMPLEX INTEGRATION

Remark 2. Suppose that f cannot be extended to a holomorphic function outside Ω (i.e., Ω is the
biggest open set on which f is holomorphic). Let z0 ∈ Ω. We know that
∞
X
f(z) = an (z − z0 )n .
n=0

We claim that the radius of convergence of this series is R = dist (z0 , Ωc ) (if Ω = C, R = ∞).
Points in Ω are called regular points of f and points in Ωc are called singular points. So in words,
the claim says that the radius of convergence of the above series is the distance from z0 to the
set of singular points of f.
Indeed, let us go back to the proof of the previous theorem. We know that there exists r > 0
such that B(z0 , r) ⊂ Ω. Take r as large as possible, that is, take r = dist (z0 , Ωc ) with the convention
that dist (z0 , ∅) = inf ∅ = +∞.

z0

Ω
P
Let 0 < r0 < r. Our proof shows that the series ∞ n
n=0 an (z − z0 ) converges absolutely in
B(z0 , r0 ) so R ≥ r0 . Since r0 was an arbitrary number less than r, it follows that R ≥ r. Next,
if R > r, then by a property of the infimum, B(z0 , R) would contain a point in Ωc ; however f is
holomorphic on B(z0 , R). Thus R = r.
z
Let us illustrate the
Puse of this fact in practice. Let f(z) = 1e−z for z ̸= 1. Here Ω = C\{1}.
∞
We know that f(z) = n=0 an zn for z ∈ B(0, R). Even without computing the an , we know that
R = dist (0, 1) = 1.

Exercise. a) Compute the an in the previous example and check directly that R = 1.
b) Show in the context of the previous remark, that dist (z0 , Ωc ) = dist (z0 , ∂Ω).
c) Show that S(z0 , R) ∩ Ωc ̸= ∅. In words, there is at least one singular point on the circle of
convergence. Therefore, we can say: the radius of convergence is the distance to the nearest
singularity.
Solution. c) Suppose that S(z0 , R) ∩ Ωc = ∅. Since S(z0 , R) is compact and Ωc is closed, we
get δ := dist (S(z0 , R), Ωc ) > 0. Since R = dist (z0 , Ωc ), then by a property of the infimum, there
exists z ′ ∈ Ωc such that |z ′ − z0 | < R + δ. Now z ′ is outside B(z0 , R) and z0 is inside B(z0 , R). By
connectedness, the straight line segment [z0 , z ′ ] meets the boundary S(z0 , R) at some point z1 .
Therefore
R + δ > |z ′ − z0 | = |z ′ − z1 | + |z1 − z0 | ≥ δ + R,
a contradiction.
2.2. ANALYCITY OF HOLOMORPHIC FUNCTIONS 33

Corollary 2.4. Let f : Ω → C be holomorphic. Then f has derivatives of all orders given by
Z
(n) n! f(u)
f (z0 ) = du
2πi γ (u − z0 )n+1

where γ is as in the proof of the above theorem.

Remark. In particular, the derivative of a holomorphic function is holomorphic. Therefore if a
function has an antiderivative, it is necessarily holomorphic. Thus for example z 7→ z and z 7→ |z|2
have no antiderivatives.

Theorem
R 2.7. (Morera; converse to Goursat-Cauchy). Let f : Ω → C be continuous and satisfy
∂T f(z) dz = 0 for every triangular region T ⊂ Ω. Then f is holomorphic on Ω.
Proof. Let z ∈ Ω be given and let U a disk containing z and contained in Ω. Note that U is convex.
By Theorem 2.2 (iv), f has an antiderivative F on U. By the previous remark, f is holomorphic on
U and in particular at z. Since z was arbitrary, it follows that f is holomorphic on Ω. □

Corollary 2.5 Let f : Ω → C be continuous and holomorphic on Ω\{a} for some a ∈ Ω. Then f
is holomorphic on Ω.
R
Proof. By Goursat-Cauchy theorem version 2, ∂T f(z) dz = 0 for every triangular region T ⊂ Ω.
By Morera’s theorem, f is holomorphic on Ω. □
P
Proposition 2.9. (Cauchy’s inequality). Let f(z) = n≥0 an (z − z0 )n be holomorphic on a disk
B(z0 , R). Let 0 < r < R and γ(t) = z0 +reit for 0 ≤ t ≤ 2π. Finally, let M(r) = sup|u−z0 |=r |f(u)| =
supu∈γ∗ |f(u)|. Then
M(r)
|an | ≤ n
r
for all n ∈ IN.
Z
1 f(u)
Proof. We know from Theorem 2.6 that an = du. Therefore,
2πi γ (u − z0 )n+1

1 |f(u)| 1 M(r) M(r)

|an | ≤ L(γ) sup = 2πr n+1 = n . □
u∈γ∗ |u − z0 |
2π n+ 1 2π r r

Definition. An entire function is a holomorphic function f : C → C.

P
It follows that an entire function is the sum of a power series n≥0 an zn with infinite radius
of convergence.

Corollary 2.6. (Liouville’s theorem). A bounded entire function is constant.

P
Proof. Let f(z) = an zn for all z ∈ C. Let r > 0 and let A be a bound for f. By Cauchy’s
M(r)
inequality, |an | ≤ rn ≤ rAn . Letting r → ∞, we get an = 0 for all n ≥ 1. Consequently, f(z) = a0
for all z ∈ C. □

Corollary 2.7. (Fundamental theorem of Algebra). Every nonconstant polynomial with complex
coefficients has a complex root.
Proof. Let P(z) = a0 + a1 z + · · · + an zn with an ̸= 0. By writing P(z) = zn [an + an− a0
z + · · · + zn ],
1

we see that lim|z|→∞ |P(z)| = ∞. Suppose that P has no roots in C. Then z 7→ P(z) 1
is holomorphic
on C. Since lim|z|→∞ | P(z)
1
| = 0, there is a number R > 0 such that 1
P(z) ≤ 1 for all z such that
|z| > R. On the other hand, the map z 7→ 1
P(z) is continuous on the compact set B ′ (0, R) and
34 CHAPTER 2. COMPLEX INTEGRATION

therefore it is bounded there. Consequently, P1 is bounded on C. It follows from Liouville theorem

that P1 is constant and therefore P is constant, contradicting our assumption. □

Definition. Let g : X → R be a real valued function on a topological space and let a ∈ X. We say
that g has a local maximum at a if g(x) ≤ g(a) for every x in a neighborhood of a. If g(x) ≤ g(a)
for all x ∈ X, we say that g has a global maximum at a. We have a similar definition for local and
global minima.

Theorem 2.8. (Strong maximum principle). Let Ω ⊂ C be open and connected and let f : Ω → C
be holomorphic. If |f| has a local maximum at a point a ∈ Ω, then f is constant.
Proof. By definition, there exists a disk B(a, ρ) ⊂ Ω such that |f(z)| ≤ |f(a)| for all z ∈ B(a, ρ).
Let 0 < r < ρ and let γ(t) = a + reit , t ∈ [0, 2π]. By Cauchy’s integral formula,
Z Z 2π Z 2π
1 f(u) 1 f(a + reit )rieit 1
f(a) = du = it
dt = f(a + reit ) dt.
2πi γ u−a 2πi 0 re 2π 0
Z 2π Z 2π Z 2π
1 1 1
|f(a)| = f(a + reit ) dt ≤ |f(a + reit )| dt ≤ |f(a)| dt = |f(a)|.
2π 0 2π 0 2π 0
It follows that Z 2π h i
|f(a)| − |f(a + reit )| dt = 0.
0

Since the integrand is nonnegative and continuous, it is zero. That is, |f(a + reit )| = |f(a)| for
all t ∈ [0, 2π]. But r was an arbitrary number in ]0, ρ[. This means that |f(z)| = |f(a)| for all
z ∈ B(a, ρ), i.e., |f| is constant on B(a, ρ) (which is open and connected). It follows that f is
constant on B(a, ρ). Since Ω is connected, the principle of analytic continuation implies that f is
constant on Ω. □

Corollary 2.8. (Weak maximum principle). Let Ω ⊂ C be open, bounded and connected. Let
f : Ω → C be continuous and holomorphic on Ω. Then

sup |f(z)| = sup |f(z)| = sup |f(z)|.

z∈Ω z∈Ω z∈∂Ω

Proof. We prove first that supz∈Ω |f(z)| = supz∈Ω |f(z)|. This is true under only the continuity
assumption. Let m = supz∈Ω |f(z)| and M = supz∈Ω |f(z)|. It is clear that m ≤ M since Ω ⊂ Ω.
Conversely, let A = {z ∈ Ω||f(z)| ≤ m}. Then A is closed in Ω and therefore in C. Note that
Ω ⊂ A and so Ω ⊂ A since A is closed. It follows that A = Ω. This means that for all z ∈ Ω,
|f(z)| ≤ m and so M ≤ m.
Next, we prove the second part of the equality. Since Ω is compact and |f| is continuous, it
achieves its maximum value at a point a ∈ Ω = Ω ∪ ∂Ω. If a ∈ Ω, then f is constant by the
previous theorem and so the equality holds. If a ∈ ∂Ω, then supz∈∂Ω |f(z)| ≤ supz∈Ω |f(z)| =
|f(a)| ≤ supz∈∂Ω |f(z)|. Hence the equality. □

2.3 Homotopic version of Cauchy’s theorem

Definition. Let Ω ⊂ C be open. Let γ0 and γ1 be two loops in Ω. We say that γ0 is Ω−homotopic
(or homotopic in Ω) to γ1 if there exists a continuous function H : [0, 1] × [0, 1] → Ω such that

(i) ∀t ∈ [0, 1], H(t, 0) = γ0 (t) and H(t, 1) = γ1 (t).

(ii) ∀s ∈ [0, 1], H(0, s) = H(1, s).

2.3. HOMOTOPIC VERSION OF CAUCHY’S THEOREM 35

Let us explain this definition. Set γs (t) = H(t, s). Then, for each s ∈ [0, 1], γs is a loop in Ω since
γs (0) = γs (1). We can think of a homotopy as a continuous deformation that takes γ0 to γ1 . We
can also think of a homotopy as a path in the space of loops.
Remark. The homotopy is an equivalence relation between loops.
Examples. 1) An two loops γ0 , γ1 : [0, 1] → C in C are C−homotopic. Indeed, let H(t, s) =
(1 − s)γ0 (t) + sγ1 (t). H is called the straight line homotopy. It is clear that H is a homotopy
between γ0 and γ1 .
2) More generally if Ω is convex, then any two loops γ0 , γ1 : [0, 1] → Ω are Ω−homotopic.
3) Let γ0 (t) = eit and γ1 (t) = 2eit for 0 ≤ t ≤ 2π. Then γ0 and γ1 are C∗ −homotopic.
However, if γ2 (t) = e2it for t ∈ [0, 2π], then γ0 and γ2 which parametrize the same circle are not
C∗ −homotopic; γ0 runs one time through the circle, whereas γ2 runs twice.
4) Let γ0 (t) = eit and γ1 (t) = 2 + eit for t ∈ [0, 2π]. We shall see later that γ0 and γ1 are not
C∗ −homotopic. The intuitive idea behind this is that if we want to deform continuously γ0 to γ1 ,
we have at some point to pass by 0. So there is an obstruction in the domain that prevents the
existence of a homotopy.

Theorem 2.9. (Cauchy). Let f : Ω → C be holomorphic and let γ0 and γ1 be two piecewise
smooth loops that are homotopic in Ω. Then
Z Z
f(z) dz = f(z) dz.
γ0 γ1

Corollary 2.9. Let Ω be an open set and a ∈ Ω. If γ and µ are two piecewise smooth loops that
are homotopic in Ω\{a}, then I(γ, a) = I(µ, a).
Z
1 dz
Proof. The function z 7→ is holomorphic Ω\{a}. Therefore by Cauchy’s theorem =
Z z − a γ z − a
dz
. Dividing by 2πi we get the result. □
µ z − a

Examples. 1) Let γ0 (t) = eit and γ1 (t) = e2it for 0 ≤ t ≤ 2π. Then γ0 and γ1 are not
∗
C −homotopic. Indeed, I(γ0 , 0) = 1 whereas I(γ1 , 0) = 2.
2) Let γ0 (t) = eit and γ1 (t) = e−it for 0 ≤ t ≤ 2π. Then γ0 and γ1 are not C∗ −homotopic.
Indeed, I(γ0 , 0) = 1 whereas I(γ1 , 0) = −1.
3) Let γ0 (t) = eit and γ1 (t) = 2 + eit for 0 ≤ t ≤ 2π. Then γ0 and γ1 are not C∗ −homotopic.
Indeed, I(γ0 , 0) = 1 whereas I(γ1 , 0) = 0 (since 0 is in the unbounded component of C\γ∗1 ).

Definition. Let Ω ⊂ C be an open and connected. We say that Ω is simply connected if every
loop is homotopic to a point in Ω. Intuitively, a simply connected domain is a domain without
”holes”. Every convex set is simply connected. On the other hand C∗ is not simply connected. The
annulus 1 < |z| < 2 is not simply connected.
Now we have a very general answer to the problem of existence of antiderivatives.
Corollary 2.10. Let Ω be simply connected and f : Ω → C be holomorphic. Then f has an
antiderivative on Ω.
Proof. LetR γ be a loop
R in Ω. Since Ω is simply connected, γ is Ω−homotopic to a point z0 .
Therefore γ f(z) dz = {z0 } f(z) dz = 0. It follows from Theorem that f has an antiderivative on
Ω. □
Remark. We have therefore a topological condition on the domain of a holomorphic function that
guarantees the existence of a global antiderivative. This condition is optimal: if every holomorphic
36 CHAPTER 2. COMPLEX INTEGRATION

function f : Ω → C has an antiderivative, then Ω is necessarily simply connected; see Theorem

13.11 in the book of Rudin.

Definition. An open set Ω is called star convex with respect to a point a ∈ Ω if [a, z] ⊂ Ω for
any z ∈ Ω.
It is clear that a convex set is star convex (with respect to any of its points). The set C\R− is
star convex (with respect to the point 1 for example) but not convex. More generally if D is a half
line then C\D is star convex.

Proposition 2.10. Every open star convex set is simply connected.

Proof. Let Ω be a star convex set. It should be clear that Ω is path connected and therefore
connected. Next, let γ be a closed path in Ω and let a be a point with respect to which Ω is star
convex. Let H(t, s) = (1 − s)γ(t) + sa. Then it is easy to check that H is a homotopy in Ω joining
γ to a. □

Theorem 2.10. (Cauchy’s integral formula). Let f : Ω → C be holomorphic and a ∈ Ω.

1) If γ and µ are two piecewise smooth loops that are homotopic in Ω and not passing through
a, then Z Z
1 f(u) 1 f(u)
du − du = f(a)[I(γ, a) − I(µ, a)].
2πi γ u − a 2πi µ u − a

2) If γ is a piecewise smooth loop not passing through a and homotopic to a point in Ω, then
Z
1 f(u)
du = I(γ, a)f(a).
2πi γ u − a

Proof. Set 
f(z)−f(a)
z−a if z ̸= a
g(z) =
f ′ (a) if z = a.
It follows Rfrom the corollary
R of Morera’s theorem that g is is holomorphic in Ω. By Cauchy’s
theorem, γ g(u) du = µ g(u) du. This implies that
Z Z Z Z
f(u) du f(u) du
du − f(a) = du − f(a) .
γ u−a γ u−a µ u−a µ u−a
Dividing by 2πi, we get the result.
The proof of 2) is similar. □
By differentiating under the integral, we obtain the following.
Corollary 2.11. (Cauchy’s integral formula of order n). Let f : Ω → C be holomorphic and a ∈ Ω.
If γ is a piecewise smooth loop not passing through a and homotopic to a point in Ω, then
Z
(n) n! f(u)
I(γ, a)f (a) = du.
2πi γ (u − a)n+1

Exercise. Let R > 1 and let CR be the upper half circle of center 0 and radius R (oriented
positively). Set ΓR = [−R, R] ∪ CR .
Z
eiz
(a) Compute 2
dz by using properly Cauchy’s formula.
ΓR 1 + z
Z∞
cos x
(b) Deduce the value of 2
dx.
0 1+x
2.3. HOMOTOPIC VERSION OF CAUCHY’S THEOREM 37

Appendix

In this optional appendix, we prove the homotopic version of Cauchy’s theorem. For this, we need
some preliminary results.

Lemma 1. Let Ω ⊂ C be open and let γ : [0, 1] → Ω be a path. Then there exists a subdivision
t0 = 0 < t1 < · · · < tn−1 < tn = 1 of [0,1] and open disks D0 , . . . , Dn−1 contained in Ω such that
γ([tk , tk+1 ]) ⊂ Dk for all k = 0, . . . , n − 1.

The subdivision {t0 , . . . , tn } and the collection of disks {D0 , . . . , Dn−1 } are called a chain covering
of γ.

Proof. Note that γ is uniformly continuous on the compact space [0,1]; this means that for every
ε > 0, there exists η > 0 such that |γ(t)−γ(t ′ )| < ε whenever |t−t ′ | < η. Since γ∗ is compact and
C\Ω is closed, the distance d(γ∗ , C\Ω) between γ∗ and C\Ω is > 0. Take ε < d(γ∗ , C\Ω) and
η as in the uniform continuity condition. Next let n ∈ IN∗ be such that n1 < η. Define a subdivision
{t0 , . . . , tn } by t0 = 0 and tk+1 = tk + n1 for k = 0, . . . , n − 1. Finally, let Dk = B(γ(tk ), ε) for
k = 0, . . . , n − 1. Now one can easily check that Dk ⊂ Ω and γ([tk , tk+1 ]) ⊂ Dk . □

Definition. Let f : Ω → C be holomorphic. Let γ, {t0 , . . . , tn } and {D0 , . . . , Dn−1 } be as in the

above lemma. Let finally γ(tk ) = zk . Since each Dk is convex, there exists on Dk an antiderivative
Fk of f. The number
X
n−1
[Fk (zk+1 ) − Fk (zk )]
k=0

is called the variation of antiderivatives of f along γ.

Lemma 2.R If γ is a piecewise smooth path, then the variation of antiderivatives of f along γ is
equal to γ f(z) dz.

Proof. Let γk denote the restriction of γ to [tk , tk+1 ]. Then by Barrow’s rule, Fk (zk+1 ) − Fk (zk ) =
R
γk f(z) dz. The result follows by summation over k. □

Lemma 3. The variation of antiderivatives along a path is independent of the choice of antideriva-
tive.

Proof. For each k = 0, . . . , n, let Fk and Gk be two antiderivatives of f on Dk . Then they differ by
a constant ck . But then Fk (zk+1 ) − Fk (zk ) = Gk (zk+1 ) + ck − (Fk (zk ) + ck ) = Gk (zk+1 ) − Gk (zk ). □

Lemma 4. Let γ and µ be two piecewise smooth loops in Ω. If we can associate to them the
same chain covering, then γ and µ have the same variation of antiderivatives.

Proof. Let {t0 , . . . , tn } be a subdivision of [0,1] and D0 , . . . , Dn−1 be disks such that γ([tk , tk+1 ]) ⊂
Dk ⊂ Ω and µ([tk , tk+1 ] ⊂ Dk ⊂ Ω. Let F0 , . . . Fn−1 be antiderivatives of f respectively on
D0 , . . . , Dn−1 . Finally let zk = γ(tk ) and wk = µ(tk ). Note that z0 = zn since γ(0) = γ(1) and
similarly w0 = wn . Let

X
n−1 X
n−1
∆= [Fk (zk+1 ) − Fk (zk )] − [Fk (wk+1 ) − Fk (wk )] .
k=0 k=0
38 CHAPTER 2. COMPLEX INTEGRATION

Then,

X
n−1

∆= [Fk (zk+1 ) − Fk (zk )] − [Fk (wk+1 ) − Fk (wk )]
k=0
X
n−1

= [Fk (zk+1 ) − Fk (wk+1 )] − [Fk (zk ) − Fk (wk )]
k=0
= [Fn−1 (zn ) − Fn−1 (wn )] − [Fn−1 (zn−1 ) − Fn−1 (wn−1 )]+
+ [Fn−2 (zn−1 ) − Fn−2 (wn−1 )] − [Fn−2 (zn−2 ) − Fn−2 (wn−2 )] + · · ·
+ [F0 (z1 ) − F0 (w1 )] − [F0 (z0 ) − F0 (w0 )] .

Now observe that zn−1 ∈ Dn−1 ∩ Dn−2 and that wn−1 ∈ Dn−1 ∩ Dn−2 . On the convex set
Dn−1 ∩ Dn−2 , the antiderivatives Fn−1 and Fn−2 differ by a constant. Thus, we see that ∆ = 0
(the underlined terms cancel each other). □

Proof of Cauchy’s theorem. Let H : [0, 1] × [0, 1] → Ω be a homotopy between γ0 and γ1 . Note
that H is uniformly continuous on the compact set [0, 1] × [0, 1]. Therefore for any ε > 0, there
exists α > 0 such that |H(t, s) − H(t ′ , s ′ )| < ε whenever |t − t ′ | < α and |s − s ′ | < α.
Let δ = dist(H([0, 1] × [0, 1]), C\Ω). Let 0 < ε < δ. Take two subdivisions {t0 , t1 , . . . , tn } and
{s0 , s1 , . . . , sn } such that max |ti+1 − ti | < α and max |si+1 − si | < α. Let Di,j = B(H(ti , sj ), ε). It
follows from the uniform continuity condition that H([ti , ti+1 ]×[sj , sj+1 ]) ⊂ Di,j . Set αj (t) = H(t, sj )
for all t ∈ [0, 1]. Then αj is a loop and αj ([ti , ti+1 ]) ⊂ Di,j and αj+1 ([ti , ti+1 ]) ⊂ Di,j . This means
that αj and αj+1 have the same chain covering. By Lemma 4., the variation of antiderivatives of f
along
R αj is equalR to the variation of antiderivatives of f along αj+1 . It follows from Lemma 2. that
αj f(z) dz = αj+1 f(z) dz for every j = 0, . . . , n − 1. Therefore
Z Z Z Z Z Z
f(z) dz = f(z) dz = f(z) dz = f(z) dz = · · · = f(z) dz = f(z) dz.
γ0 α0 α1 α2 αn γ1

□
Chapter 3

Laurent series and the residue theorem

3.1 Laurent series

X
+∞
Definition. A Laurent series is a series of the form an (z − a)n . Such series is convergent
n=−∞
X
+∞ X
−1 X
+∞
a−p
if the two series an (z − a)n and an (z − a) = n
are both convergent. Same
n=−∞
(z − a)p
n=0 p=1
definition for absolutely and uniformly convergent. The first series is called the regular part and
the second series is called the singular part.
P
Proposition 3.1. Let n∈Z an (z − a)n be a Laurent series. Then there exist two values p1 and p2
in [0, ∞] such that the Laurent series is absolutely convergent in the annulus p2 < |z − a| < p1 .
X
+∞
Proof. Let p1 denote the radius of convergence of an (z − a)n . Then this series is absolutely
n=0
convergent for |z − a| < p1 and divergent for |z − a| > p1 .
X
−1 X
+∞
The series an (z − a)n can be written as bp up where bp = a−p and u = 1
z−a . Let R2
n=−∞ p=1
X
+∞
be the radius of convergence of the series bp up . Then, this series is absolutely convergent for
p=1
X
−1
|u| < R2 and divergent for |u| > R2 . Let p2 = 1
R2 . Then the series an (z − a)n is absolutely
n=−∞
convergent for |z − a| > p2 . Hence the conclusion. □

X
+∞ X
−1
Remark 1. Since limu→0 bp up = 0, it follows that lim an (z − a)n = 0.
|z−a|→∞
p=1 n=−∞

P n
Remark 2. A Laurent series need not be convergent. For example the series n∈Z z is nowhere
convergent (it is convergent on the empty annulus 1 < |z| < 1).
P P
Proposition 3.2. If two Laurent series n∈Z an (z − a)n and n∈Z bn (z − a)n coincide on a
nonempty annulus p2 < |z − a| < p1 then an = bn for all n ∈ Z.
Proof. Let S(z) denote the common value of the two series. A complex number P z in the annulus
can be written in the form z = a + re with p2 < r < p1 . Then S(a + re ) = n∈Z an r einθ and
iθ iθ n

this series is normally convergent with respect to θ ∈ [0, 2π]. Therefore, we can multiply by e−ipθ

39
40 CHAPTER 3. LAURENT SERIES AND THE RESIDUE THEOREM

and integrate term by term:

Z 2π Z 2π X X Z 2π
S(a + reiθ )e−ipθ dθ = an rn ei(n−p)θ dθ = an r n ei(n−p)θ dθ = 2πap rp
0 0 n∈Z n∈Z 0

since Z 2π 
0 if n ̸= p
ei(n−p)θ dθ =
0 2π if n = p.
P
Repeating the same reasoning with n∈Z bn rn einθ , we get
Z 2π
S(a + reiθ )e−ipθ dθ = 2πbp rp .
0

Hence, ap = bp for all p ∈ Z. □

Proposition 3.3. Let R > 0 and f be holomorphic

P on {z ∈ C | |z−a| > R} and satisfy lim|z|→∞ f(z) =
a−p
0. Then there exists a Laurent series p≥1 (z−a) p which converges to f(z) on {z ∈ C | |z − a| > R}.

Proof. Set u = z−a1

for |z − a| > R and g(u) = f( u1 + a) = f(z). Then g is holomorphic on the set
{u ∈ C | 0 < |u| < R1 }. Note that limu→0 g(u) = lim|z−a|→∞ f(z) = 0 (by assumption). So letting
g(0) = 0, we see that g is continuous on {u ∈ C | |u| < R1 }. By a corollary of Morera’s theorem, g
P
is holomorphic B(0, R1 ). Therefore, there exists a power series ∞ p
p=1 bp u that converges to g(u)
P∞ bp
on the above disk. Hence f(z) = p=1 (z−a) p for all |z − a| > R. □

Theorem 3.1. LetPf be a holomorphic function on the annulus p2 < |z − a| < p1 . Then there exists
a Laurent series n∈Z an (z − a)n that converges to f(z) on the annulus.
Proof. Let z belong to the annulus A. Then there exists r1 and r2 such that p2 < r2 < |z − a| <
r1 < p1 . Let γ1 (t) = a + r1 eit and γ2 (t) = a + r2 eit . Then γ1 and γ2 are homotopic in the annulus
A (for example by the homotopy H(t, s) = (1 − s)γ2 (t) + sγ1 (t)). Then by Cauchy’s integral
formula, Z Z
1 f(u) 1 f(u)
du − du = f(z)[I(γ1 , z) − I(γ2 , z)].
2πi γ1 u − z 2πi γ2 u − z
Now note that I(γ1 , z) = 1 since z is inside γ1 and I(γ2 , z) = 0 since z is outside γ2 . Therefore,
Z Z
1 f(u) 1 f(u)
f(z) = du − du = f1 (z) + f2 (z).
2πi γ1 u − z 2πi γ2 u − z
It follows from the theorem of differentiation under the integral, that f1 is holomorphic on C\γ∗1
and f2 is holomorphic on C\γ∗2 . In particular
P f1 is holomorphic on the disk B(a, r1 ). Therefore f1
has a power series expansion f1 (z) = n≥0 an (z − a)n for |z − a| < r1 . Also f2 is holomorphic
P disk B (a, r2 )nwith lim|z|→∞ |f2 (z)| = 0. By the previous proposition, f2 (z) =
outside the closed ′
P a−p
p≥1 (z−a)p = n≤−1 an (z − a) for |z − a| > r2 . Consequently,
X X X
f(z) = f1 (z) + f2 (z) = an (z − a)n + an (z − a)n = an (z − a)n . □
n≥0 n≤−1 n∈Z

Remark. This Laurent series depends a priori on r1 and r2 . But according to the uniqueness of
a Laurent series expansion, this series does not depend on r1 and r2 .

Examples. 1) Let us find the Laurent series of f(z) = z−1 1 in the annulus 1 < |z| < +∞. Since
1
z < 1, on this annulus, we can write
∞   ∞
1 1 1X 1 n X 1 1 1
f(z) = · 1
= = n+ 1
= + 2 + ··· .
z 1− z z z z z z
n=0 n=0
3.1. LAURENT SERIES 41

1
2) Let us find the Laurent series of f(z) = e z in the annulus 0 < |z| < ∞. We can write
∞
X 1 1 n
 
1 1 1
e =
z = 1 + + 2 + ··· .
n! z z 2z
n=0

3) Let us find the Laurent series of the function f(z) = 1

z2 −3z+2
in the annulus 1 < |z| < 2. We first
decompose the function in simple fractions. We have

1 1 1 −1/2 1 1
f(z) = = − = − .
(z − 1)(z − 2) z−2 z−1 1 − 2z z 1 − z1

Therefore,
∞ ∞ ∞
− 1 X  z n 1 X X X
 n −1
1 n −1 n
f(z) = − = (−z ) + z .
2 2 z z n=−∞
2n+1
n=0 n=0 n=0

Definition. A point a ∈ C is called an isolated singular point of a function f if f is holomorphic on

the annulus 0 < |z − a| < R. The function f then has a Laurent series expansion on this annulus:
∞
X X ∞
a−n
f(z) = n
+ an (z − a)n
(z − a)
n=1 n=0

for 0 < |z − a| < R.

Let A = {n ∈ IN∗ | a−n ̸= 0}. There are three possibilities.

1. A = ∅.

2. A ̸= ∅ and is finite.

3. A is infinite.

In case 1, f extends to a holomorphic function at a and we say that a is a removable singularity

of f.
Example. Consider the function f(z) = sinz z . A priori, z = 0 is a singularity of f. However, we can
write
z3 z2
 
1
f(z) = z− + ··· = 1 − + ··· .
z 6 6
Since there is no singular part in the Laurent series about z = 0, we conclude that z = 0 is a
removable singularity.
a−p a−1
In case 2, f(z) = (z−a)p + · · · + z−a + a0 + a1 (z − a) + · · · . We say that a is pole of order p for f.
sin z
Example. Consider the function f(z) = z4
. We can write

z3 z 5
 
1 1 1 z
f(z) = 4 z− + + ··· = 3 − + + ··· .
z 6 5! z 6z 5!

Thus, z = 0 is a pole of order 3.

In case 3, we say that a is an essential singularity for f.
Example. z = 0 is an essential singularity of f(z) = e1/z .

Proposition 3.4. Let f : B(a, R)\{a} → C be holomorphic. Then the following conditions are
equivalent.
42 CHAPTER 3. LAURENT SERIES AND THE RESIDUE THEOREM

(i) a is a removable singularity for f.

(ii) f has a limit as z → a.

(iii) |f| has a limit as z → a.

(iv) |f| is bounded on some annulus C = {z ∈ C | 0 < |z − a| < δ}, (δ < R).

Proof. (i) ⇒ (ii). f extends to a holomorphic and therefore continuous function f̃ : Ω → C.

Therefore f has a limit as z → a.
(ii) ⇒ (iii) is clear.
(iii) ⇒ (iv). Since |f| has a limit as z → a, then |f| extends to a continuous function g : B(a, R) → C.
Let 0 < δ < R. The function g is continuous on the compact set B ′ (a, δ), and therefore is bounded
there.
P
(iv) ⇒ (i). We can write f(z) = n∈Z an (z − a)n for 0 < |z − a| < R. Let 0 < r < δ. Then, it
follows from the proof of Proposition 3.2 that
Z 2π
f(a + reiθ )e−ipθ dθ = 2πap rp
0

for all p ∈ Z. Therefore |ap |rp ≤ sup0≤θ≤2π |f(a + reiθ )|.

Since f is bounded on S(a, r) ⊂ B(a, δ)\{a}, there exists M > 0 such that

sup |f(a + reiθ )| ≤ M

0≤θ≤2π

and therefore |ap | ≤ Mr−p . If p < 0, then −p > 0 and limr→0 r−p → 0. Thus letting r → 0, we
get ap = 0. This means that A = ∅ and so a is a removable singularity for f. □

Proposition 3.5. Let f : B(a, R)\{a} → C be holomorphic. Then, the following conditions are
equivalent.

(i) a is a pole of order p for f.

(ii) There exists δ > 0 and a holomorphic function g : B(a, δ) → C such that g(a) ̸= 0 and

g(z)
f(z) =
(z − a)p

for all 0 < |z − a| < δ.

(iii) limz→a |f(z)| = ∞.

Proof. (i) ⇒ (ii). We have

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

for all z ∈ B(a, R)\{a}, with a−p ̸= 0. Reducing to the same denominator, we get

a−p + · · · + a−1 (z − a)p−1 + a0 (z − a)p + · · · g(z)

f(z) = =
(z − a)p (z − a)p

and g(a) = a−p ̸= 0.

3.1. LAURENT SERIES 43

g(z)
(ii) ⇒ (i). Let f(z) = (z−a) p for all z ∈ B(a, δ)\{a}, where g is holomorphic on B(a, δ) and g(a) ̸= 0.
P∞
Then g(z) = n=0 bn (z − a)n for all z ∈ B(a, δ) and therefore

b0 bp−1
f(z) = p
+ ··· + + bp + bp+1 (z − a) + · · · + an (z − a)n + · · · .
(z − a) z−a

This means that a is a pole of order p for f since b0 = g(a) ̸= 0.

(ii) ⇒ (iii). We have
|g(z)| 1
lim |f(z)| = lim = |g(a)| lim = ∞.
z→a z→a |z − a| p z→a |z − a|p

(iii) ⇒ (ii). Since limz→a |f(z)| = ∞, there exists δ ∈]0, R[ such that |f(z)| > 1 for all z ∈ B(a, δ)\{a}.
Let h(z) = 1/f(z). Then h is holomorphic on B(a, δ)\{a} and |h(z)| < 1 there. This means that |h|
is bounded on B(a, δ)\{a}. By the previous proposition, h extends to a holomorphic function (still
denoted by h) h : B(a, δ) → C. Moreover, limz→a h(z) = limz→a 1/f(z) = 0, so that h(a) = 0.
Therefore there exists p ≥ 1 such that the Taylor series expansion of h is
∞
X
h(z) = bk (z − a)k = (z − a)p G(z)
k=p

where bp ̸= 0 and G is holomorphic on B(a, δ). Moreover, G(a) = bp ̸= 0 and G does not vanish
on B(a, δ). Letting g = 1/G, we see that g is holomorphic on B(a, δ) and that

1 1 g(z)
f(z) = = p
=
h(z) (z − a) G(z) (z − a)p

for all z ∈ B(a, δ)\{a}. □

Proposition 3.6. Let f : B(a, R)\{a} → C be holomorphic. Then the following conditions are
equivalent.

(i) a is an essential singularity of f.

(ii) For any ε ∈]0, R], the image of B(a, ε)\{a} under f is dense in C (Cassorati-Weierstrass
theorem).

(iii) |f| has no limit as z → a.

Proof. Note that the following conditions are equivalent to (ii).

1. For all ε ∈]0, R] and for all w ∈ C, there exists a sequence (zn ) ⊂ B(a, ε)\{a} such that
f(zn ) → w as n → ∞. We can take zn → a.

2. For every ε ∈]0, R], every nonempty open set meets f(B(a, ε)\{a}).

3. For every ε ∈]0, R], for every w ∈ C and every δ > 0, the disc B(w, δ) meets f(B(a, ε)\{a}).

To prove that (i) ⇒ (ii), we reason by contradiction. Let ε ∈]0, R] be given. Assume that
there exists w ∈ C, and δ > 0 such that B(w, δ) does not meet f(B(a, ε)\{a}). Then for every
z ∈ B(a, ε)\{a}, |f(z) − w| ≥ δ. Accordingly, let for z ∈ B(a, ε)\{a},

1
g(z) = .
f(z) − w
44 CHAPTER 3. LAURENT SERIES AND THE RESIDUE THEOREM

Then g is holomorphic on B(a, ε)\{a} and bounded there. It follows from Proposition 3.4 that g has
a limit L ∈ C as z → a. Moreover, L ̸= 0 because otherwise we would have lim |f(z) − w| = ∞ so
lim |f(z)| = ∞ and this implies by the previous proposition that a is a pole of f. Setting g(a) = L,
we see that g is holomorphic on B(a, ε) with g(a) ̸= 0. Therefore G := 1/g is holomorphic on
B(a, ε) (because g does not vanish there). In particular, lim G(z) = G(a) = 1/g(a) = L1 . But
lim G(z) = lim[f(z) − w]. Therefore lim f(z) = w + L1 . By Proposition 3.4, this means that a is a
removable singularity for f, contrary to the assumption.
(ii) ⇒ (iii). Let w = 1, then there exists a sequence (zn ) converging to a such that f(zn ) → 1 and
so |f(zn )| → 1. Letting, w = 2, there exists a sequence (tn ) converging to a such that f(tn ) → 2
and so |f(tn )| → 2. Thus |f| has no limit as z → a.
(iii) ⇒ (i). By the previous two propositions, a cannot be neither a removable singularity nor a
pole. Therefore a is an essential singularity. □

3.2 The residue theorem

Definition. Let f be holomorphic on B(a, R)\{a}. We know that f is expandable in a Laurent
series
∞
X ∞
X
a−n
f(z) = n
+ an (z − a)n
(z − a)
n=1 n=0

on B(a, R)\{a}. The residue of f at a denoted by Res(f, a) is the coefficient a−1 in the Laurent
series.

Theorem 3.2. (Residue) Let Ω be a simply connected subset of C and S = {z1 , z2 , . . . , zk } be a

finite subset of Ω. Let f : Ω\S → C be holomorphic. Let γ be a piecewise smooth loop in Ω not
intersecting S. Then,
Z X
k
f(z) dz = 2πi I(γ, zm )Res(f, zm ).
γ m=1

Proof. PFor each s ∈ S, let fs denote the singular part of f. Then fs is holomorphic on C\{s}. Let
g = f− s∈S fs . Then g is holomorphic on the Ω. RSince Ω is simply connected, γ is Ω−homotopic
to a point. It follows from Cauchy’s theorem that γ g(z) dz = 0. Therefore,
Z XZ
f(z) dz = fs (z) dz.
γ s∈S γ
R P∞ b−n
So let us compute γ fs (z) dz. We know that fs (z) = n=1 (z−s)n and the series converges
uniformly on γ (because it converges on the compact subsets of C\{s}). Now Robserve that for
n ̸= 1, the function z 7→ (z − s)−n has an antiderivative on C\{s} and therefore γ (z−s)dz
n = 0 for
n ̸= 1 since γ is a loop. Accordingly,
Z ∞
X Z Z
dz dz
fs (z) dz = b−n n
= b−1 = b−1 2πiI(γ, s) = 2πiRes(f, s)I(γ, s).
γ n=1 γ (z − s) γ z−s

Practical methods for computing residues

1. Suppose that f is holomorphic on a punctured disk 0 < |z − z0 | < R and that z0 is a simple
pole of f. Then,
Res(f, z0 ) = lim (z − z0 )f(z).
z→z0
3.2. THE RESIDUE THEOREM 45

g(z)
Indeed, we know from Proposition 3.5 that f(z) = where g is holomorphic on B(z0 , R).
P z−z0
Therefore, g(z) = ∞ n
n=0 bn (z − z0 ) and so

b0
f(z) = + b1 + b2 (z − z0 ) + · · ·
z − z0

It follows that Res(f, z0 ) = b0 = g(z0 ) = limz→z0 g(z) = limz→z0 (z − z0 )f(z).

P(z)
Consequence. Suppose that f(z) = Q(z) where P and Q are holomorphic in B(z0 , R) and z0 is a
simple zero of Q. Then,

P(z) P(z0 )
Res(f, z0 ) = lim (z − z0 ) = ′ .
z→z0 Q(z) Q (z0 )

2 2
e−z e−z
Example. Let f(z) = z(z−1) .Then, Res(f, 0) = limz→0 z−1 = −1.
2. Suppose that f is holomorphic on a punctured disk 0 < |z − z0 | < R and that z0 is a pole of
g(z)
order p of f. By Proposition 3.5, we know that f(z) = (z−z 0)
p where g is holomorphic on B(z0 , R).
Reasoning as in case 1, we see that

g(p−1) (z0 )
Res(f, z0 ) = .
(p − 1)!
1
Example. Let f(z) = z2 (z−1)
. Then, 0 is a pole of order 2 and so
 
d  2  d 1 −1
Res(f, 0) = z f(z) = = = −1.
dz z=0 dz z−1 z=0 (z − 1)2 z=0

Remark 1. We can also find the residue by expanding f in a Laurent series. We have
∞
1 X n 1 1
f(z) = 2 −z = − 2 − − . . . .
z z z
n=0

The residue is by definition the coefficient of z1 . It is -1.

Remark 2. If z0 is an essential singularity, there is no formula for the residue. We have to expand
1
f in a Laurent series about z0 and find the coefficient of z−z 0
.

Example of the use of the residue theorem

Z
sin z
Let us compute tan z dz where γ(t) = 2eit , 0 ≤ t ≤ 2π. We can write tan z = cos z .
γ
The zeros of cos are π2 + nπ, n ∈ Z. But only π2 and −π 2 are inside the circle. We can take
Ω = {z = x + iy ∈ C, −π < x < π} (a vertical band) and f(z) = tan z. Then,

(1) Ω is an open convex set and therefore simply connected,

(2) γ is a loop in Ω

(3) f is holomorphic on Ω\{− π2 , π2 }.

Then, we can apply the residue theorem and get

Z      
π  π −π −π
f(z) dz = 2πi I γ, Res f, + I γ, Res f, .
γ 2 2 2 2
46 CHAPTER 3. LAURENT SERIES AND THE RESIDUE THEOREM

Now, I γ, π2 =
1 since γ is a simple, positively oriented loop and containing π

2 in its interior.
Similarly, I γ, π2 = 1. Next, by the method explained above
 π sin( π2 )
Res f, = = −1.
2 cos ′ ( π2 )

Similarly,
sin( −π
 
−π 2 )
Res f, = ′ −π = −1.
2 cos ( 2 )
Z
Thus finally, tan z dz = −4πi.
γ

The residue theorem permit the computation of real integrals of the form
Z 2π
(a) R(sin t, cos t) dt where R is a rational function.
0
Z∞ Z∞
(b) R(x) cos x dx and R(x) sin x dx where R is a rational function.
−∞ −∞
Z∞
(c) R(x) dx where R is a rational function.
−∞

Let us explain the method for computing the first class of integrals. The method for computing
the other two classes will be explored in the exercises. We first express the integral as line integral
of some rational function of a complex variable and then compute the integral by using the residue
theorem (or Cauchy’s integral formula). For t ∈ [0, 2π], we set z = eit = γ(t). Then,
   
1 1 1 1
sin t = z− , cos t = z+ , dz = iz dt.
2i z 2 z

It follows that
Z      Z 2π
1 1 1 1 1
R( z− , z+ dz = R(sin t, cos t) dt.
γ iz 2i z 2 z 0

Z 2π Z
dt
Example. Compute I = . According to what we said, I = f(z) dz where
0 2 − cos t γ

1 1 2i 2i
f(z) = 1 1

= 2 = √ √ .
iz 2 − 2 z+ z z − 4z + 1 (z − (2 + 3))(z − (2 − 3))
√ √
This function has two simple poles: 2 + 3 and 2 − 3 but only the second one is inside the unit
circle. Accordingly, let for example Ω = {z = x + iy ∈ C, x < 2}. Then,

(1) Ω is an open convex set and therefore simply connected,

(2) γ is a loop in Ω,
√
(3) f is holomorphic on Ω\{2 − 3}.

Therefore the residue theorem applies and yields

√ √
I = 2πiI(γ, 2 − 3)Res(f, 2 − 3).
3.2. THE RESIDUE THEOREM 47

√ √
Now as usual, I(γ, 2 − 3) = 1, since γ is a simple positively oriented loop containing 2 − 3 in
its interior. Next, by the method explained above
√ 2i −i
Res(f, 2 − 3) = √ =√ .
z − (2 + 3) √ 3
z=2− 3

It follows that
−i 2π
I = 2πi √ = √ .
3 3
Remark. In this example, we can also use Cauchy’s integral formula.