INTRODUCTION TO COMPLEX ANALYSIS

W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 1
COMPLEX NUMBERS

1.1. Arithmetic and Conjugates

The purpose of this chapter is to give a review of various properties of the complex numbers that we shall
need in the discussion of complex analysis. As the reader is expected to be familiar with the material,
all proofs have been omitted.

The equation x2 + 1 = 0 has no solution x ∈ R. To “solve” this equation, we have to introduce extra
numbers into our number system. To do this, we define the number i by i2 + 1 = 0, and then extend the
field of all real numbers by adjoining the number i, which is then combined with the real numbers by the
operations addition and multiplication in accordance with the Field axioms of the real number system.
The numbers a + ib, where a, b ∈ R, of the extended field are then added and multiplied in accordance
with the Field axioms, suitably extended, and the restriction i2 + 1 = 0. Note that the number a + 0i,
where a ∈ R, behaves like the real number a.

What we have said in the last paragraph basically amounts to the following. Consider two complex
numbers a + ib and c + id, where a, b, c, d ∈ R. We have the addition and multiplication rules

(a + ib) + (c + id) = (a + c) + i(b + d) and (a + ib)(c + id) = (ac − bd) + i(ad + bc).

These lead to the subtraction rule

(a + ib) − (c + id) = (a − c) + i(b − d),

and the division rule, that if c + id = 0, then

a + ib ac + bd bc − ad
= 2 +i 2 .
c + id c + d2 c + d2
1–2 W W L Chen : Introduction to Complex Analysis

Note the special case a = 1 and b = 0.

Suppose that z = x + iy, where x, y ∈ R. The real number x is called the real part of z, and denoted
by x = Rez. The real number y is called the imaginary part of z, and denoted by y = Imz. The set
C = {z = x + iy : x, y ∈ R} is called the set of all complex numbers. The complex number z = x − iy is
called the conjugate of z.

It is easy to see that for every z ∈ C, we have

z+z z−z
Rez = and Imz = .
2 2i
Furthermore, if w ∈ C, then

z+w =z+w and zw = z w.

1.2. Polar Coordinates

Suppose that z = x + iy, where x, y ∈ R. The real number

r = x2 + y 2

is called the modulus of z, and denoted by |z|. On the other hand, if z = 0, then any number θ ∈ R
satisfying the equations

(1) x = r cos θ and y = r sin θ

is called an argument of z, and denoted by arg z. Hence we can write z in polar form

z = r(cos θ + i sin θ).

Note, however, that for a given z ∈ C, arg z is not unique. Clearly we can add any integer multiple of
2π to θ without affecting (1). We sometimes call a real number θ ∈ R the principal argument of z if θ
satisfies the equations (1) and −π < θ ≤ π. The principal argument of z is usually denoted by Arg z.

It is easy to see that for every z ∈ C, we have |z|2 = zz. Also, if w ∈ C, then

|zw| = |z||w| and |z + w| ≤ |z| + |w|.

Furthermore, if

z = r(cos θ + i sin θ) and w = s(cos φ + i sin φ),

where r, s, θ, φ ∈ R and r, s > 0, then

z r
zw = rs(cos(θ + φ) + i sin(θ + φ)) and = (cos(θ − φ) + i sin(θ − φ)).
w s

1.3. Rational Powers

De Moivre’s theorem, that

(2) cos nθ + i sin nθ = (cos θ + i sin θ)n for every n ∈ N and θ ∈ R,

 Chapter 1 : Complex Numbers 1–3

is useful in finding n-th roots of complex numbers.

Suppose that c = R(cos α + i sin α), where R, α ∈ R and R > 0. Then the solutions of the equation
z n = c are given by
 
√ α + 2kπ α + 2kπ
where k = 0, 1, . . . , n − 1.
n
z= R cos + i sin ,
n n

Finally, we can define cb for any b ∈ Q and non-zero c ∈ C as follows. The rational number b can
be written uniquely in the form b = p/q, where p ∈ Z and q ∈ N have no prime factors in common.
Then there are exactly q distinct numbers z satisfying z q = c. We now define cb = z p , noting that the
expression (2) can easily be extended to all n ∈ Z. It is not too difficult to show that there are q distinct
values for the rational power cb .

Problems for Chapter 1

1. Suppose that z0 ∈ C is fixed. A polynomial P (z) is said to be divisible by z − z0 if there is another

polynomial Q(z) such that P (z) = (z − z0 )Q(z).
a) Show that for every c ∈ C and k ∈ N, the polynomial c(z k − z0k ) is divisible by z − z0 .
b) Consider the polynomial P (z) = a0 + a1 z + a2 z 2 + . . . + an z n , where a0 , a1 , a2 , . . . , an ∈ C are
arbitrary. Show that the polynomial P (z) − P (z0 ) is divisible by z − z0 .
c) Deduce that P (z) is divisible by z − z0 if P (z0 ) = 0.
d) Suppose that a polynomial P (z) of degree n vanishes at n distinct values z1 , z2 , . . . , zn ∈ C, so
that P (z1 ) = P (z2 ) = . . . = P (zn ) = 0. Show that P (z) = c(z − z1 )(z − z2 ) . . . (z − zn ), where
c ∈ C is a constant.
e) Suppose that a polynomial P (z) of degree n vanishes at more than n distinct values. Show
that P (z) = 0 identically.
 
 z−α 

2. Suppose that α ∈ C is fixed and |α| < 1. Show that |z| ≤ 1 if and only if   ≤ 1.
1 − αz 

3. Suppose that z = x + iy, where x, y ∈ R. Express each of the following in terms of x and y:
   
z + 1  z+i 
a) |z − 1|3 b)   c)  
z − 1  1 − iz 

4. Suppose that c ∈ R and α ∈ C with α = 0.

a) Show that αz + αz + c = 0 is the equation of a straight line on the plane.
b) What does the equation zz + αz + αz + c = 0 represent if |α|2 ≥ c?

5. Suppose that z, w ∈ C. Show that |z + w|2 + |z − w|2 = 2(|z|2 + |w|2 ).

6. Find all the roots of the equation (z 8 − 1)(z 3 + 8) = 0.

7. For each of the following, compute all the values and plot them on the plane:
a) (1 + i)−1/2 b) (−4)3/4 c) (1 − i)3/8
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 2
FOUNDATIONS OF COMPLEX ANALYSIS

2.1. Three Approaches

We start by remarking that analysis is sometimes known as the study of the four C’s: convergence,
continuity, compactness and connectedness. In real analysis, we have studied convergence and continuity
to some depth, but the other two concepts have been somewhat disguised. In this course, we shall try
to illustrate these two latter concepts a little bit more, particularly connectedness.

Complex analysis is the study of complex valued functions of complex variables. Here we shall
restrict the number of variables to one, and study complex valued functions of one complex variable.
Unless otherwise stated, all functions in these notes are of the form f : S → C, where S is a set in C.

We shall study the behaviour of such functions using three different approaches. The first of these,
discussed in Chapter 3 and usually attributed to Riemann, is based on differentiation and involves pairs
of partial differential equations called the Cauchy-Riemann equations. The second approach, discussed in
Chapters 4–11 and usually attributed to Cauchy, is based on integration and depends on a fundamental
theorem known nowadays as Cauchy’s integral theorem. The third approach, discussed in Chapter 16
and usually attributed to Weierstrass, is based on the theory of power series.

2.2. Point Sets in the Complex Plane

We shall study functions of the form f : S → C, where S is a set in C. In most situations, various
properties of the point sets S play a crucial role in our study. We therefore begin by discussing various
types of point sets in the complex plane.

Before making any definitions, let us consider a few examples of sets which frequently occur in our
subsequent discussion.
2–2 W W L Chen : Introduction to Complex Analysis

Example 2.2.1. Suppose that z0 ∈ C, r, R ∈ R and 0 < r < R. The set {z ∈ C : |z − z0 | < R}
represents a disc, with centre z0 and radius R, and the set {z ∈ C : r < |z − z0 | < R} represents an
annulus, with centre z0 , inner radius r and outer radius R.

R R
r
z0 z0

Example 2.2.2. Suppose that A, B ∈ R and A < B. The set {z = x + iy ∈ C : x, y ∈ R and x > A}
represents a half-plane, and the set {z = x + iy ∈ C : x, y ∈ R and A < x < B} represents a strip.

A A B

Example 2.2.3. Suppose that α, β ∈ R and 0 ≤ α < β < 2π. The set

{z = r(cos θ + i sin θ) ∈ C : r, θ ∈ R and r > 0 and α < θ < β}

represents a sector.

β
α

We now make a number of important definitions. The reader may subsequently need to return to
these definitions.
 Chapter 2 : Foundations of Complex Analysis 2–3

Definition. Suppose that z0 ∈ C and  ∈ R, with  > 0. By an -neighbourhood of z0 , we mean a

disc of the form {z ∈ C : |z − z0 | < }, with centre z0 and radius  > 0.

Definition. Suppose that S is a point set in C. A point z0 ∈ S is said to be an interior point of S

if there exists an -neighbourhood of z0 which is contained in S. The set S is said to be open if every
point of S is an interior point of S.

z0

Example 2.2.4. The sets in Examples 2.2.1–2.2.3 are open.

Example 2.2.5. The punctured disc {z ∈ C : 0 < |z − z0 | < R} is open.

Example 2.2.6. The disc {z ∈ C : |z − z0 | ≤ R} is not open.

Example 2.2.7. The empty set ∅ is open. Why?

Definition. An open set S is said to be connected if every two points z1 , z2 ∈ S can be joined by the
union of a finite number of line segments lying in S. An open connected set is called a domain.

z1 z2

Remarks. (1) Sometimes, we say that an open set S is connected if there do not exist non-empty
open sets S1 and S2 such that S1 ∪ S2 = S and S1 ∩ S2 = ∅. In other words, an open connected set
cannot be the disjoint union of two non-empty open sets.

(2) In fact, it can be shown that the two definitions are equivalent.
2–4 W W L Chen : Introduction to Complex Analysis

(3) Note that we have not made any definition of connectedness for sets that are not open. In
fact, the definition of connectedness for an open set given by (1) here is a special case of a much more
complicated definition of connectedness which applies to all point sets.

Example 2.2.8. The sets in Examples 2.2.1–2.2.3 are domains.

Example 2.2.9. The punctured disc {z ∈ C : 0 < |z − z0 | < R} is a domain.

Definition. A point z0 ∈ C is said to be a boundary point of a set S if every -neighbourhood of z0

contains a point in S as well as a point not in S. The set of all boundary points of a set S is called the
boundary of S.

z0

Example 2.2.10. The annulus {z ∈ C : r < |z − z0 | < R}, where 0 < r < R, has boundary C1 ∪ C2 ,
where

C1 = {z ∈ C : |z − z0 | = r} and C2 = {z ∈ C : |z − z0 | = R}

are circles, with centre z0 and radius r and R respectively. Note that the annulus is connected and hence
a domain. However, note that its boundary is made up of two separate pieces.

Definition. A region is a domain together with all, some or none of its boundary points. A region
which contains all its boundary points is said to be closed. For any region S, we denote by S the closed
region containing S and all its boundary points, and call S the closure of S.

Remark. Note that we have not made any definition of closedness for sets that are not regions. In
fact, our definition of closedness for a region here is a special case of a much more complicated definition
of closedness which applies to all point sets.

Definition. A region S is said to be bounded or finite if there exists a real number M such that
|z| ≤ M for every z ∈ S. A region that is closed and bounded is said to be compact.

Example 2.2.11. The region {z ∈ C : |z − z0 | ≤ R} is closed and bounded, hence compact. It is called
the closed disc with centre z0 and radius R.

Example 2.2.12. The region {z = x + iy ∈ C : x, y ∈ R and 0 ≤ x ≤ 1} is closed but not bounded.

Example 2.2.13. The square {z = x + iy ∈ C : x, y ∈ R and 0 ≤ x ≤ 1 and 0 < y < 1} is bounded

but not closed.
 Chapter 2 : Foundations of Complex Analysis 2–5

2.3. Complex Functions

In these lectures, we study complex valued functions of one complex variable. In other words, we study
functions of the form f : S → C, where S is a set in C. Occasionally, we will abuse notation and simply
refer to a function by its formula, without explicitly defining the domain S. For instance, when we
discuss the function f (z) = 1/z, we implicitly choose a set S which will not include the point z = 0
where the function is not defined. Also, we may occasionally wish to include the point z = ∞ in the
domain or codomain.

We may separate the independent variable z as well as the dependent variable w = f (z) into real
and imaginary parts. Our usual notation will be to write z = x + iy and w = f (z) = u + iv, where
x, y, u, v ∈ R. It follows that u = u(x, y) and v = v(x, y) can be interpreted as real valued functions of
the two real variables x and y.

Example 2.3.1. Consider the function f : S → C, given by f (z) = z 2 and where S = {z ∈ C : |z| < 2}
is the open disc with radius 2 and centre 0. Using polar coordinates, it is easy to see that the range of
the function is the open disc f (S) = {w ∈ C : |w| < 4} with radius 4 and centre 0.

Example 2.3.2. Consider the function f : H → C, where H = {z = x + iy ∈ C : y > 0} is the upper

half-plane and f (z) = z 2 . Using polar coordinates, it is easy to see that the range of the function is the
complex plane minus the non-negative real axis.

Example 2.3.3. Consider the function f : T → C, where T = {z = x + iy ∈ C : 1 < x < 2} is a strip

and f (z) = z 2 . Let x0 ∈ (1, 2) be fixed, and consider the image of a point (x0 , y) on the vertical line
x = x0 . Here we have

u = x20 − y 2 and v = 2x0 y.

Eliminating y, we obtain the equation of a parabola

v2
u = x20 −
4x20

in the w-plane. It follows that the image of the vertical line x = x0 under the function w = z 2 is this
parabola. Now the boundary of the strip are the two lines x = 1 and x = 2. Their images under the
mapping w = z 2 are respectively the parabolas

v2 v2
u=1− and u=4− .
4 16

It is easy to see that the range of the function is the part of the w-plane between these two parabolas.

z w
8

1 2 1 4
2–6 W W L Chen : Introduction to Complex Analysis

Example 2.3.4. Consider again the function w = z 2 . We would like to find all z = x + iy ∈ C for
which 1 < Rew < 2. In other words, we have the restriction 1 < u < 2, but no rectriction on v. Let
u0 ∈ (1, 2) be fixed, and consider points (x, y) in the z-plane with images on the vertical line u = u0 .
Here we have the hyperbola

x2 − y 2 = u0 .

The boundaries u = 1 and u = 2 are represented by the hyperbolas

x2 − y 2 = 1 and x2 − y 2 = 2.

It is easy to see that the points in question are precisely those between the two hyperbolas.

w z

1 2 1

2.4. Extended Complex Plane

It is sometimes useful to extend the complex plane C by the introduction of the point ∞ at infinity. Its
connection with finite complex numbers can be established by setting z + ∞ = ∞ + z = ∞ for all z ∈ C,
and setting z · ∞ = ∞ · z = ∞ for all non-zero z ∈ C. We can also write ∞ · ∞ = ∞.

Note that it is not possible to define ∞ + ∞ and 0 · ∞ without violating the laws of arithmetic.
However, by special convention, we shall write z/0 = ∞ for z = 0 and z/∞ = 0 for z = ∞.

In the complex plane C, there is no room for a point corresponding to ∞. We can, of course,
introduce an “ideal” point which we call the point at infinity. The points in C, together with the point
at infinity, form the extended complex plane. We decree that every straight line on the complex plane
shall pass through the point at infinity, and that no half-plane shall contain the ideal point.

The main purpose of this section is to introduce a geometric model in which each point of the
extended complex plane has a concrete representative. To do this, we shall use the idea of stereographic
projection.

Consider a sphere of radius 1 in R3 . A typical point on this sphere will be denoted by P (x1 , x2 , x3 ).
Note that x21 + x22 + x23 = 1. Let us call the point N (0, 0, 1) the north pole. The equator of this sphere is
the set of all points of the form (x1 , x2 , 0), where x21 + x22 = 1. Consider next the complex plane C. This
can be viewed as a plane in R3 . Let us position this plane in such a way that the equator of the sphere
lies on this plane; in other words, our copy of the complex plane is “horizontal” and passes through the
origin. We can further insist that the x-direction on our complex plane is the same as the x1 -direction
in R3 , and that the y-direction on our complex plane is the same as the x2 -direction in R3 . Clearly a
typical point z = x + iy on our complex plane C can be identified with the point Z(x, y, 0) in R3 .
 Chapter 2 : Foundations of Complex Analysis 2–7

Suppose that Z(x, y, 0) is on the plane. Consider the straight line that passes through Z and the
north pole N . It is not too difficult to see that this straight line intersects the surface of the sphere at
precisely one other point P (x1 , x2 , x3 ). In fact, if Z is on the equator of the sphere, then P = Z. If Z is
on the part of the plane outside the sphere, then P is on the northern hemisphere, but is not the north
pole N . If Z is on the part of the plane inside the sphere, then P is on the southern hemisphere. Check
that for Z(0, 0, 0), the point P (0, 0, −1) is the south pole.

N y
P

On the other hand, if P is any point on the sphere different from the north pole N , then a straight
line passing through P and N intersects the plane at precisely one point Z. It follows that there is a
pairing of all the points P on the sphere different from the north pole N and all the points on the plane.
This pairing is governed by the requirement that the straight line through any pair must pass through
the north pole N .

We can now visualize the north pole N as the point on the sphere corresponding to the point at
infinity of the plane. The sphere is called the Riemann sphere.

2.5. Limits and Continuity

The concept of a limit in complex analysis is exactly the same as in real analysis. So, for example, we
say that f (z) → L as z → z0 , or

lim f (z) = L,
z→z0

if, given any  > 0, there exists δ > 0 such that |f (z) − L| <  whenever 0 < |z − z0 | < δ.

This definition will be perfectly in order if the function f is defined in some open set containing
z0 , with the possible exception of z0 itself. It follows that if z0 is an interior point of the region S of
definition of the function, our definition is in order. However, if z0 is a boundary point of the region S
of definition of the function, then we agree that the conclusion |f (z) − L| <  need only hold for those
z ∈ S satisfying 0 < |z − z0 | < δ.

Similarly, we say that a function f (z) is continuous at z0 if f (z) → f (z0 ) as z → z0 . A similar

qualification on z applies if z0 is a boundary point of the region S of definition of the function. We also
say that a function is continuous in a region if it is continuous at every point of the region.
2–8 W W L Chen : Introduction to Complex Analysis

Note that for a function to be continuous in a region, it is enough to have continuity at every point of
the region. Hence the choice of δ may depend on a point z0 in question. If δ can be chosen independently
of z0 , then we have some uniformity as well. To be precise, we make the following definition.

Definition. A function f (z) is said to be uniformly continuous in a region S if, given any  > 0, there
exists δ > 0 such that |f (z1 ) − f (z2 )| <  for every z1 , z2 ∈ S satisfying |z1 − z2 | < δ.

Remark. Note that if we fix z2 to be a point z0 and write z for z1 , then we require |f (z) − f (z0 )| < 
for every z ∈ S satisfying |z − z0 | < δ. In other words, δ cannot depend on z0 .

Example 2.5.1. Consider the punctured disc S = {z ∈ C : 0 < |z| < 1}. The function f (z) = 1/z is
continuous in S but not uniformly continuous in S. To see this, note first of all that continuity follows
from the simple observation that the function z is continuous and non-zero in S. To show that the
function is not uniformly continuous in S, it suffices to show that there exists  > 0 such that for every
δ > 0, there exist z1 , z2 ∈ S such that




1

|z1 − z2 | < δ and
− 1
≥ .

z1 z2

Let  = 1. For every δ > 0, choose n ∈ N such that n > δ −1/2 , and let

1 1
z1 = and z2 = .
n n+1

Clearly z1 , z2 ∈ S. It is easy to see that

1 1

1
1



|z1 − z2 | =
− = <δ and
− 1
= 1.
n n + 1
n(n + 1)
z1 z2

Problems for Chapter 2

1. For each of the following functions, find f (z + 3), f (1/z) and f (f (z)):
1−z
a) f (z) = z − 1 b) f (z) = z 2 c) f (z) = 1/z d) f (z) =
3+z

2. Which of the sets below are domains?

a) {z : 0 < |z| < 1} b) {z : Imz < 3|z|} c) {z : |z − 1| ≤ |z + 1|}
d) {z : |z 2 − 1| < 1} e) {z : 0 < Rez ≤ 1}

3. Find the image of the strip {z : |Rez| < 1} and of the disc {z : |z| < 1} under each of the following
mappings:
z+1
a) w = (1 + i)z + 1 b) w = 2z 2 c) w = z −1 d) w =
z−1

4. A function f (z) is said to be an isometry if |f (z1 ) − f (z2 )| = |z1 − z2 | for every z1 , z2 ∈ C; in other
words, if it preserves distance.
a) Suppose that f (z) is an isometry. Show that for every a, b ∈ C with |a| = 1, the function
g(z) = af (z) + b is also an isometry.
b) Show that the function

f (z) − f (0)
h(z) =
f (1) − f (0)

is an isometry with h(0) = 0 and h(1) = 1.

 Chapter 2 : Foundations of Complex Analysis 2–9

c) Suppose that k(z) is an isometry with k(0) = 0 and k(1) = 1. Show that Rek(z) = Rez, and
that k(i) = ±i.
[Hint: Explain first of all why |k(z)| = |z| and |1 − k(z)| = |1 − z|.]
d) Suppose that in (c), we have k(i) = i. Show that Imk(z) = Imz and that k(z) = z for all
z ∈ C.
e) Suppose that in (c), we have k(i) = −i. Show that Imk(z) = −Imz and that k(z) = z for all
z ∈ C.
f) Deduce that every isometry has the form f (z) = az + b or f (z) = az + b, where a, b ∈ C with
|a| = 1.

5. In the notation of Section 2.4, let the point z = x + iy on the complex plane C correspond to the
point (x1 , x2 , x3 ) of the sphere under stereographic projection, so that the three points (0, 0, 1),
(x1 , x2 , x3 ) and (x, y, 0) are collinear. Note that (x1 , x2 , x3 − 1) = λ(x, y, −1) for some λ ∈ R, and
that x21 + x22 + x23 = 1.
 
2x 2y |z|2 − 1
a) Show that (x1 , x2 , x3 ) = , , .
|z|2 + 1 |z|2 + 1 |z|2 + 1
b) Note that a circle on the sphere is the intersection of the sphere with a plane ax1 +bx2 +cx3 = d.
By expressing this equation of the plane in terms of x and y, show that a circle on the sphere
not containing the pole (0, 0, 1) corresponds to a circle in the complex plane. Show also that a
circle on the sphere containing the pole (0, 0, 1) corresponds to a line in the complex plane.
c) Suppose that (x1 , x2 , x3 ) and (x1 , x2 , x3 ) are two points on the sphere corresponding to the com-
plex numbers z and z  respectively. Show that the distance between (x1 , x2 , x3 ) and (x1 , x2 , x3 )
is given by

2|z − z  |
d(z, z  ) =   .
1 + |z|2 1 + |z  |2

[Remark: The number d(z, z  ) is known as the chordal distance.]

6. Each of the following functions is not defined at z = z0 . What value must f (z0 ) take to ensure
continuity at z = z0 ?
z − z0 z 3 − z03
a) f (z) = b) f (z) =
z − z0 z − z0
   
1 1 1 1 1 1
c) f (z) = − d) f (z) = −
z − z0 z z0 z − z0 z 3 z03

7. Suppose that

a0 + a1 z + a2 z 2
f (z) = ,
b 0 + b 1 z + b2 z 2

where a0 , a1 , a2 , b0 , b1 , b2 ∈ C. Examine the behaviour of f (z) at z = 0 and at z = ∞.

 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 3
COMPLEX DIFFERENTIATION

3.1. Introduction

Suppose that D ⊆ C is a domain. A function f : D → C is said to be differentiable at z0 ∈ D if the limit

f (z) − f (z0 )
lim
z→z0 z − z0

exists. In this case, we write

f (z) − f (z0 )
(1) f  (z0 ) = lim ,
z→z0 z − z0

and call f  (z0 ) the derivative of f at z0 .

If z = z0 , then


f (z) − f (z0 )
f (z) = (z − z0 ) + f (z0 ).
z − z0

It follows from (1) and the arithmetic of limits that if f  (z0 ) exists, then f (z) → f (z0 ) as z → z0 , so
that f is continuous at z0 . In other words, differentiability at z0 implies continuity at z0 .

Note that the argument here is the same as in the case of a real valued function of a real variable. In
fact, the similarity in argument extends to the arithmetic of limits. Indeed, if the functions f : D → C
and g : D → C are both differentiable at z0 ∈ D, then both f + g and f g are differentiable at z0 , and

(f + g) (z0 ) = f  (z0 ) + g  (z0 ) and (f g) (z0 ) = f (z0 )g  (z0 ) + f  (z0 )g(z0 ).
3–2 W W L Chen : Introduction to Complex Analysis

If the extra condition g  (z0 ) = 0 holds, then f /g is differentiable at z0 , and


f g(z0 )f  (z0 ) − f (z0 )g  (z0 )
(z0 ) = .
g g 2 (z0 )

One can also establish the Chain rule for differentiation as in real analysis. More precisely, suppose
that the function f is differentiable at z0 and the function g is differentiable at w0 = f (z0 ). Then the
function g ◦ f is differentiable at z = z0 , and

(g ◦ f ) (z0 ) = g  (w0 )f  (z0 ).

Example 3.1.1. Consider the function f (z) = z, where for every z ∈ C, z denotes the complex
conjugate of z. Suppose that z0 ∈ C. Then

f (z) − f (z0 ) z − z0 z − z0
(2) = = .
z − z0 z − z0 z − z0

If z − z0 = h is real and non-zero, then (2) takes the value 1. On the other hand, if z − z0 = ik is purely
imaginary, then (2) takes the value −1. It follows that this function is not differentiable anywhere in C,
although its real and imaginary parts are rather well behaved.

3.2. The Cauchy-Riemann Equations

If we use the notation

f (z + h) − f (z)
f  (z) = lim ,
h→0 h
then in Example 3.1.1, we have examined the behaviour of the ratio

f (z + h) − f (z)
h
first as h → 0 through real values and then through imaginary values. Indeed, for the derivative
to exist, it is essential that these two limiting processes produce the same limit f  (z). Suppose that
f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real valued functions. If h is real, then the
two limiting processes above correspond to

f (z + h) − f (z) u(x + h, y) − u(x, y) v(x + h, y) − v(x, y) ∂u ∂v

lim = lim + i lim = +i
h→0 h h→0 h h→0 h ∂x ∂x
and
f (z + ih) − f (z) u(x, y + h) − u(x, y) v(x, y + h) − v(x, y) ∂v ∂u
lim = lim + i lim = −i
h→0 ih h→0 ih h→0 ih ∂y ∂y

respectively. Equating real and imaginary parts, we obtain

∂u ∂v ∂u ∂v
(3) = and =− .
∂x ∂y ∂y ∂x

Note that while the existence of the derivative in real analysis is a mild smoothness condition, the
existence of the derivative in complex analysis leads to a pair of partial differential equations.
 Chapter 3 : Complex Differentiation 3–3

Definition. The partial differential equations (3) are called the Cauchy-Riemann equations.

We have proved the following result.

THEOREM 3A. Suppose that f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real
valued functions. Suppose further that f  (z) exists. Then the four partial derivatives in (3) exist, and
the Cauchy-Riemann equations (3) hold. Furthermore, we have

∂u ∂v ∂v ∂u
(4) f  (z) = +i and f  (z) = −i .
∂x ∂x ∂y ∂y

A natural question to ask is whether the Cauchy-Riemann equations are sufficient to guarantee
the existence of the derivative. We shall show next that we require also the continuity of the partial
derivatives in (3).

THEOREM 3B. Suppose that f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real
valued functions. Suppose further that the four partial derivatives in (3) are continuous and satisfy the
Cauchy-Riemann equations (3) at z0 . Then f is differentiable at z0 , and the derivative f  (z0 ) is given
by the equations (4) evaluated at z0 .

Proof. Write z0 = x0 + iy0 . Then

f (z) − f (z0 ) (u(x, y) − u(x0 , y0 )) + i(v(x, y) − v(x0 , y0 ))

= .
z − z0 z − z0
We can write



∂u ∂u
u(x, y) − u(x0 , y0 ) = (x − x0 ) + (y − y0 ) + |z − z0 |1 (z)
∂x z0 ∂y z0

and



∂v ∂v
v(x, y) − v(x0 , y0 ) = (x − x0 ) + (y − y0 ) + |z − z0 |2 (z).
∂x z0 ∂y z0

If the four partial derivatives in (3) are continuous at z0 , then

lim 1 (z) = 0 and lim 2 (z) = 0.

z→z0 z→z0

In view of the Cauchy-Riemann equations (3), we have

(u(x, y) − u(x0 , y0 )) + i(v(x, y) − v(x0 , y0 ))



∂u ∂v ∂u ∂v
= (x − x0 ) +i + (y − y0 ) +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂y ∂y z0



∂u ∂v ∂v ∂u
= (x − x0 ) +i + (y − y0 ) − +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂x ∂x z0



∂u ∂v ∂u ∂v
= (x − x0 ) +i + i(y − y0 ) +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂x ∂x z0


∂u ∂v
= (z − z0 ) +i + |z − z0 |(1 (z) + i2 (z)).
∂x ∂x z0

Hence




f (z) − f (z0 ) ∂u ∂v |z − z0 | ∂u ∂v
= +i + (1 (z) + i2 (z)) → +i
z − z0 ∂x ∂x z0 z − z0 ∂x ∂x z0
3–4 W W L Chen : Introduction to Complex Analysis

as z → z0 , giving the desired results.

3.3. Analytic Functions

In the previous section, we have shown that differentiability in complex analysis leads to a pair of partial
differential equations. Now partial differential equations are seldom of interest at a single point, but
rather in a region. It therefore seems reasonable to make the following definition.

Definition. A function f is said to be analytic at a point z0 ∈ C if it is differentiable at every z in

some -neighbourhood of the point z0 . The function f is said to be analytic in a region if it is analytic
at every point in the region. The function f is said to be entire if it is analytic in C.

Example 3.3.1. Consider the function f (z) = |z|2 . In our usual notation, we clearly have

u = x2 + y 2 and v = 0.

The Cauchy-Riemann equations

2x = 0 and 2y = 0

can only be satisfied at z = 0. It follows that the function is differentiable only at the point z = 0, and
is therefore analytic nowhere.

Example 3.3.2. The function f (z) = z 2 is entire.

Example 3.3.3. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant real part u. Then clearly

∂u ∂u
=0 and = 0.
∂x ∂y

Since f is analytic in D, it is differentiable at every point in D, and so the Cauchy-Riemann equations

hold in D. It follows that
∂v ∂v
=0 and = 0.
∂x ∂y

Hence f must have constant imaginary part v, and so f must be constant in D.

Example 3.3.4. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant imaginary part v. A similar argument shows that f must have constant real part u. Hence f
must be constant in D.

Example 3.3.5. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant modulus. In other words, u2 + v 2 = C for some non-negative real number C. Differentiating
this with respect to x and to y, we obtain respectively

∂u ∂v ∂u ∂v
2u + 2v =0 and 2u + 2v = 0.
∂x ∂x ∂y ∂y

In view of the Cauchy-Riemann equations, these can be written as

∂u ∂u ∂u ∂u
2u − 2v =0 and 2v + 2u = 0.
∂x ∂y ∂x ∂y
 Chapter 3 : Complex Differentiation 3–5

In matrix notation, these become

 
∂u



u −v  ∂x 
 = 0 .
v u  ∂u  0
∂y
Note now that


u −v
det = u2 + v 2 = C.
v u

If C > 0, then we must have the unique solution

∂u ∂u
=0 and = 0,
∂x ∂y

so that the real part u is constant. It then follows from Example 3.3.3 that f is constant in D. On the
other hand, if C = 0, then clearly u = v = 0, so that f = 0 in D.

3.4. Introduction to Special Functions

In this section, we shall generalize various functions that we have studied in real analysis to the complex
domain. Consider first of all the exponential function. It seems reasonable to extend the property
ex1 +x2 = ex1 ex2 for real variables to complex values of the variables to obtain

ez = ex+iy = ex eiy , where x, y ∈ R.

This suggests the following definition.

Definition. Suppose that z = x + iy, where x, y ∈ R. Then the exponential function ez is defined for
every z ∈ C by

(5) ez = ex (cos y + i sin y).

If we write ez = u(x, y) + iv(x, y), then

u(x, y) = ex cos y and v(x, y) = ex sin y.

It is easy to check that the Cauchy-Riemann equations are satisfied for every z ∈ C, so that ez is an
entire function. Furthermore, it follows from (4) that

d z ∂u ∂v
e = +i = ex cos y + iex sin y = ex (cos y + i sin y) = ez ,
dz ∂x ∂x
so that ez is its own derivative. On the other hand, note that for every y1 , y2 ∈ R, we have

ei(y1 +y2 ) = cos(y1 + y2 ) + i sin(y1 + y2 ) = (cos y1 + i sin y1 )(cos y2 + i sin y2 ) = eiy1 eiy2 .

Furthermore, if x1 , x2 ∈ R, then

ex1 +x2 ei(y1 +y2 ) = (ex1 ex2 )(eiy1 eiy2 ) = (ex1 eiy1 )(ex2 eiy2 ).

Writing z1 = x1 + iy1 and z2 = x2 + iy2 , we deduce the addition formula

ez1 +z2 = ez1 ez2 .

3–6 W W L Chen : Introduction to Complex Analysis

Finally, note that

|ez | = |ex (cos y + i sin y)| = ex | cos y + i sin y| = ex .

Since ex is never zero, it follows that the exponential function ez is non-zero for every z ∈ C.

Next, we turn our attention to the trigonometric functions. Note first of all that if z = x + iy, where
x, y ∈ R, then iz = −y + ix. Replacing z in (5) by iz and by −iz gives respectively

eiz = e−y (cos x + i sin x) and e−iz = ey (cos x − i sin x).

The special case y = 0 gives respectively

eix = cos x + i sin x and e−ix = cos x − i sin x.

It follows that

eix + e−ix eix − e−ix

cos x = and sin x = .
2 2i
This suggests the following definition.

Definition. Suppose that z ∈ C. Then the trigonometric functions cos z and sin z are defined in terms
of the exponential function by

eiz + e−iz eiz − e−iz

(6) cos z = and sin z = .
2 2i

Since the exponential function is an entire function, it follows easily from (6) that both cos z and
sin z are entire functions. Furthermore, it can easily be deduced from (6) that

d d
cos z = − sin z and sin z = cos z.
dz dz
We can define the functions tan z, cot z, sec z and cosec z in terms of the functions cos z and sin z as in
real variables. However, note that these four functions are not entire. Also, we can deduce from (6) the
formulas

cos(z1 + z2 ) = cos z1 cos z2 − sin z1 sin z2 and sin(z1 + z2 ) = sin z1 cos z2 + cos z1 sin z2 ,

and a host of other trigonometric identities that we know hold for real variables.

Finally, we turn our attention to the hyperbolic functions. These are defined as in real analysis.

Definition. Suppose that z ∈ C. Then the hyperbolic functions cosh z and sinh z are defined in terms
of the exponential function by

ez + e−z ez − e−z
(7) cosh z = and sinh z = .
2 2

Since the exponential function is an entire function, it follows easily from (7) that both cosh z and
sinh z are entire functions. Furthermore, it can easily be deduced from (7) that

d d
cosh z = sinh z and sinh z = cosh z.
dz dz
 Chapter 3 : Complex Differentiation 3–7

We can define the functions tanh z, coth z, sech z and cosech z in terms of the functions cosh z and sinh z
as in real variables. However, note that these four functions are not entire. Also, we can deduce from
(7) a host of hyperbolic identities that we know hold for real variables. Note also that comparing (6)
and (7), we obtain

cosh z = cos iz and sinh z = −i sin iz.

3.5. Periodicity and its Consequences

One of the fundamental differences between real and complex analysis is that the exponential function
is periodic in C.

Definition. A function f is periodic in C if there is some fixed non-zero ω ∈ C such that the identity
f (z + ω) = f (z) holds for every z ∈ C. Any constant ω ∈ C with this property is called a period of f .

THEOREM 3C. The exponential function ez is periodic in C with period 2πi. Furthermore, any
period ω ∈ C of ez is of the form ω = 2πki, where k ∈ Z is non-zero.

Proof. The first assertion follows easily from the observation

e2πi = cos 2π + i sin 2π = 1.
Suppose now that ω ∈ C. Clearly ez+ω = ez implies eω = 1. Write ω = α + iβ, where α, β ∈ R. Then
eα (cos β + i sin β) = 1.
Taking modulus, we conclude that eα = 1, so that α = 0. It then follows that cos β + i sin β = 1.
Equating real and imaginary parts, we conclude that cos β = 1 and sin β = 0, so that β = 2πk, where
k ∈ Z. The second assertion follows.

Consider now the mapping w = ez . By (5), we have w = ex (cos y + i sin y), where x, y ∈ R. It
follows that

|w| = ex and arg w = y + 2πk,

where k ∈ Z. Usually we make the choice arg w = y, with the restriction that −π < y ≤ π. This
restriction means that z lies on the horizontal strip
(8) R0 = {z ∈ C : −∞ < x < ∞, −π < y ≤ π}.
The restriction −π < arg w ≤ π can also be indicated on the complex w-plane by a cut along the negative
real axis. The upper edge of the cut, corresponding to arg w = π, is regarded as part of the cut w-plane.
The lower edge of the cut, corresponding to arg w = −π, is not regarded as part of the cut w-plane.

y v
z w

x u

-π
3–8 W W L Chen : Introduction to Complex Analysis

It is easy to check that the function exp : R0 → C \ {0}, defined for every z ∈ R0 by exp(z) = ez ,
is one-to-one and onto.

Remark. The region R0 is usually known as a fundamental region of the exponential function. In
fact, it is easy to see that every set of the type

(9) Rk = {z ∈ C : −∞ < x < ∞, (2k − 1)π < y ≤ (2k + 1)π},

where k ∈ Z, has this same property as R0 .

Let us return to the function exp : R0 → C \ {0}. Since it is one-to-one and onto, there is an inverse
function.

Definition. The function Log : C \ {0} → R0 , defined by Log(w) = z ∈ R0 , where exp(z) = w, is

called the principal logarithmic function.

Suppose that z = x + iy and w = u + iv, where x, y, u, v ∈ R. Suppose further that we impose the
restriction −π < y ≤ π. If w = exp(z), then it follows from (5) that u = ex cos y and v = ex sin y, and so

|w| = (u2 + v 2 )1/2 = ex and y = Arg(w),

where Arg(w) denotes the principal argument of w. It follows that

x = log |w| and y = Arg(w).

Hence

(10) Log(w) = log |w| + i Arg(w).

In many practical situations, we usually try to define

log w = log |w| + i arg w,

where the argument is chosen in order to make the logarithmic function continuous in its domain of
definition, if this is at all possible. The following three examples show that great care needs to be taken
in the study of such “many valued functions”.

Example 3.5.1. Consider the logarithmic function in the disc {w : |w+2| < 1}, an open disc of radius 1
and centred at the point w = −2. Note that this disc crosses the cut on the w-plane along the negative real
axis discussed earlier. In this case, we may restrict the argument to satisfy, for example, 0 ≤ arg w < 2π.
The logarithmic function defined in this way is then continuous in the disc {w : |w + 2| < 1}.

u
-3 -1
 Chapter 3 : Complex Differentiation 3–9

Example 3.5.2. Consider the region P obtained from the w-plane by removing both the line segment
{u + iv : 0 ≤ u ≤ 1, v = 0} and the half-line {u + iv : u = 1, v > 0}, as shown below.

u
1

Suppose that we wish to define the logarithmic function to be continuous in this region P . One way to
do this is to restrict the argument to the range π < arg w ≤ 3π for any w ∈ P satisfying u ≥ 1, and to
the range 0 < arg w ≤ 2π for any w ∈ P satisfying u < 1.

Example 3.5.3. Consider the annulus {w : 1 < |w| < 2}. It is impossible to define the logarithmic
function to be continuous in this annulus. Heuristically, if one goes round the annulus once, the argument
has to change by 2π if it varies continuously. If we return to the original starting point after going round
once, the argument cannot therefore be the same.
v

u
1 2

It should now be quite clear that we cannot expect to have

Log(w1 w2 ) = Log(w1 ) + Log(w2 ),

or even

log w1 w2 = log w1 + log w2 .

Instead, we have

log w1 w2 = log w1 + log w2 + 2πik for some k ∈ Z.

Let us return to the principal logarithmic function Log : C \ {0} → R0 . Recall (10). We have

Log(z) = log |z| + i Arg(z).

Recall from real analysis that for any t ∈ R, the equation tan θ = t has a unique solution θ satisfying
−π/2 < θ < π/2. This solution is denoted by tan−1 t and satisfies

d 1
tan−1 t = .
dt 1 + t2
3–10 W W L Chen : Introduction to Complex Analysis

It is not difficult to show that if we write




 −1 x π

 − tan − if y < 0,

 y 2


 y
(11) v(x, y) = − tan−1 if x > 0,

 x






 −1 x π
 − tan + if y > 0,
y 2

then Arg(z) = v(x, y). Hence Log(z) = u(x, y) + iv(x, y), where

1
(12) u(x, y) = log(x2 + y 2 ).
2

It now follows from (12) that

∂u x ∂u y
= 2 and = 2 ,
∂x x + y2 ∂y x + y2

and from (11) that

∂v y ∂v x
=− 2 and = 2 .
∂x x + y2 ∂y x + y2

Clearly the Cauchy-Riemann equations are satisfied, and so

d ∂u ∂v x − iy 1 1
Log(z) = +i = 2 = = .
dz ∂x ∂x x + y2 x + iy z

Power functions are defined in terms of the exponential and logarithmic functions. Given z, a ∈ C,
we write z a = ea log z . Naturally, the precise value depends on the logarithmic function that is chosen,
and care again must be exercised for these “many valued functions”.

3.6. Laplace’s Equation and Harmonic Conjugates

We have shown that for any function f = u + iv, the existence of the derivative f  leads to the Cauchy-
Riemann equations. More precisely, we have

∂u ∂v ∂u ∂v
(13) = and =− .
∂x ∂y ∂y ∂x

Furthermore,

∂u ∂v
(14) f  (z) = +i .
∂x ∂x

Suppose now that the second derivative f  also exists. Then f  satisfies the Cauchy-Riemann
equations. The Cauchy-Riemann equations corresponding to the expression (14) are





∂ ∂u ∂ ∂v ∂ ∂u ∂ ∂v
(15) = and =− .
∂x ∂x ∂y ∂x ∂y ∂x ∂x ∂x
 Chapter 3 : Complex Differentiation 3–11

Substituting (13) into (15), we obtain

∂2u ∂2u ∂2v ∂2v

(16) + 2 =0 and + = 0.
∂x2 ∂y ∂x2 ∂y 2
We also obtain

∂2v ∂2v ∂2u ∂2u

= and = .
∂y∂x ∂x∂y ∂y∂x ∂x∂y

Definition. A continuous function φ(x, y) that satisfies Laplace’s equation

∂2φ ∂2φ
+ 2 =0
∂x2 ∂y
in a domain D ⊆ C is said to be harmonic in D.

We have in fact proved the following result.

THEOREM 3D. Suppose that f = u + iv, where u and v are real valued. Suppose further that f  (z)
exists in a domain D ⊆ C. Then u and v both satisfy Laplace’s equation and are harmonic in D.

Definition. Two harmonic functions u and v in a domain D ⊆ C are said to be harmonic conjugates
in D if they satisfy the Cauchy-Riemann equations.

The remainder of this chapter is devoted to a discussion on finding harmonic conjugates. We shall
illustrate the following theorem by discussing the special case when D = C.

THEOREM 3E. Suppose that a function u is real valued and harmonic in a domain D ⊆ C. Then
there exists a real valued function v which satisfies the following conditions:
(a) The functions u and v satisfy the Cauchy-Riemann equations in D.
(b) The function f = u + iv is analytic in D.
(c) The function v is harmonic in D.

Clearly, parts (b) and (c) follow from part (a). We shall now indicate a proof of part (a) in the
special case D = C, and shall omit reference to this domain.

Suppose that u is real valued and harmonic. Then we need to find a real valued function v such
that
∂u ∂v ∂u ∂v
= and =− .
∂x ∂y ∂y ∂x
Let X0 + iY0 ∈ D be chosen and fixed. Integrating the second of these with respect to x, we obtain
 X
∂u
(17) v(X, y) = − (x, y)dx + c(y),
X0 ∂y

where c(y) is some function depending at most on y. Differentiating with respect to y, we obtain
 X
∂v ∂ ∂u
(X, y) = − (x, y)dx + c (y).
∂y ∂y X0 ∂y

Clearly the first of the Cauchy-Riemann equations requires

 X
∂u ∂ ∂u
(X, y) = − (x, y)dx + c (y).
∂x ∂y X0 ∂y
3–12 W W L Chen : Introduction to Complex Analysis

Changing the order of differentiation and integration, we obtain

 X
  X
∂u ∂ ∂u  ∂2u
(X, y) = − (x, y)dx + c (y) = − (x, y)dx + c (y).
∂x X0 ∂y ∂y X0 ∂y 2

Since u is harmonic, we obtain

 X
∂u ∂2u ∂u ∂u
(X, y) = (x, y)dx + c (y) = (X, y) − (X0 , y) + c (y),
∂x X0 ∂x2 ∂x ∂x

so that
∂u
c (y) = (X0 , y).
∂x
Integrating with respect to y, we obtain
 Y
∂u
(18) c(Y ) = (X0 , y)dy + c,
Y0 ∂x

where c is an absolute constant. On the other hand, (17) can be rewritten in the form
 X
∂u
(19) v(X, Y ) = − (x, Y )dx + c(Y ).
X0 ∂y

Combining (18) and (19), we obtain

 X  Y
∂u ∂u
(20) v(X, Y ) = − (x, Y )dx + (X0 , y)dy + c.
X0 ∂y Y0 ∂x

It is easy to check that this function v satisfies the Cauchy-Riemann equations. Indeed, we have
 X  Y
∂ ∂ ∂u ∂ ∂u ∂u
v(X, Y ) = − (x, Y )dx + (X0 , y)dy = − (X, Y ).
∂X ∂X X0 ∂y ∂X Y0 ∂x ∂y

On the other hand, we have

 X  Y  X 2
∂ ∂ ∂u ∂ ∂u ∂ u ∂u
v(X, Y ) = − (x, Y )dx + (X0 , y)dy = − 2
(x, Y )dx + (X0 , Y )
∂Y ∂Y X0 ∂y ∂Y Y0 ∂x X0 ∂y ∂x
 X 2
∂ u ∂u ∂u ∂u ∂u ∂u
= 2
(x, Y )dx + (X0 , Y ) = (X, Y ) − (X0 , Y ) + (X0 , Y ) = (X, Y ).
X0 ∂x ∂x ∂x ∂x ∂x ∂x

This completes our sketched proof.

In practice, we may use the following technique. Suppose that u is a real valued harmonic function
in a domain D. Write
∂u ∂u
(21) g(z) = −i .
∂x ∂y

Then the Cauchy-Riemann equations for g are





∂ ∂u ∂ ∂u ∂ ∂u ∂ ∂u
=− and = ,
∂x ∂x ∂y ∂y ∂y ∂x ∂x ∂y

which clearly hold. It follows that g is analytic in D. Suppose now that u is the real part of an analytic
function f in D. Then f  (z) agrees with the right hand side of (21) in view of (3) and (4). Hence f  = g
 Chapter 3 : Complex Differentiation 3–13

in D. The question here, of course, is to find this function f . If we are successful, then the imaginary
part v of f is a harmonic conjugate of the harmonic function u.

Example 3.6.1. Consider the function u(x, y) = x3 − 3xy 2 . It is easily checked that

∂2u ∂2u
+ 2 = 0,
∂x2 ∂y

so that u is harmonic in C. Using X0 = Y0 = 0 in (20), we obtain

 X  Y  X  Y
∂u ∂u
v(X, Y ) = − (x, Y )dx + (0, y)dy + c = 6 xY dx − 3 y 2 dy + c = 3X 2 Y − Y 3 + c,
0 ∂y 0 ∂x 0 0

where c is any arbitrary constant. On the other hand, we can write

∂u ∂u
g(z) = −i = 3(x2 − y 2 ) + 6ixy = 3(x2 + 2ixy − y 2 ) = 3(x + iy)2 = 3z 2 .
∂x ∂y

It follows that u is the real part of an analytic function f in C such that f  (z) = g(z) for every z ∈ C.
The function f (z) = z 3 + C satisfies this requirement for any arbitrary constant C. Note that the
imaginary part of f is 3x2 y − y 3 + c, where c is the imaginary part of C.

Example 3.6.2. Consider the function u(x, y) = ex sin y. It is easily checked that

∂2u ∂2u
+ 2 = 0,
∂x2 ∂y

so that u is harmonic in C. Using X0 = Y0 = 0 in (20), we obtain

 X  Y  X  Y
∂u ∂u
v(X, Y ) = − (x, Y )dx + (0, y)dy + c = − x
e cos Y dx + sin ydy + c
0 ∂y 0 ∂x 0 0
= cos Y − eX cos Y − cos Y + 1 + c = c − eX cos Y,

where c is any arbitrary constant. On the other hand, we can write

∂u ∂u
g(z) = −i = ex sin y − iex cos y = −iex (cos y + i sin y) = −iez .
∂x ∂y

It follows that u is the real part of an analytic function f in C such that f  (z) = g(z) for every z ∈ C.
The function f (z) = C − iez satisfies this requirement for any arbitrary constant C. Note that the
imaginary part of f is c − ex cos y, where c is the imaginary part of C.

Problems for Chapter 3

1. a) Suppose that P (z) = (z − z1 )(z − z2 ) . . . (z − zk ), where z1 , z2 , . . . , zk ∈ C. Show that

P  (z) 1 1 1
= + + ... + for every z ∈ C \ {z1 , z2 , . . . , zk }.
P (z) z − z1 z − z2 z − zk

b) Suppose further that Rezj < 0 for every j = 1, . . . , k, and that Rez ≥ 0. Show in this case that
Re(z − zj )−1 > 0 for every j = 1, . . . , k, and deduce that P  (z) = 0.
[Remark: Polynomials all of whose roots have negative real parts are called Hurwitz polynomials.
We have shown here that the derivative of a non-constant Hurwitz polynomial is also a Hurwitz
polynomial.]
3–14 W W L Chen : Introduction to Complex Analysis

2. For each of the following functions f (z), determine whether the Cauchy-Riemann equations are
satisfied:
a) f (z) = x2 − y 2 − 2ixy b) f (z) = log(x2 + y 2 ) + 2i cot−1 (x/y)
c) f (z) = x3 − 3y 2 + 2x + i(3x2 y − y 3 + 2y) d) f (z) = log(x2 − y 2 ) + 2i tan−1 (y/x)

3. Show that a real valued analytic function is constant.

4. We are required to define an analytic function f (z) such that f (x + iy) = ex f (iy) for every x, y ∈ R
and f (0) = 1. Suppose that for every y ∈ R, we write f (iy) = c(y) + is(y), where c(y), s(y) ∈ R for
every y ∈ R.
a) Show by the Cauchy-Riemann equations that c (y) = −s(y) and s (y) = c(y) for every y ∈ R.
b) For every y ∈ R, write g(y) = (c(y) − cos y)2 + (s(y) − sin y)2 . Show that g  (y) = 0 for every
y ∈ R. Deduce that g(y) = 0 for every y ∈ R.
c) Comment on the above.

5. a) Suppose that P (z) = a0 + a1 z + a2 z 2 + . . . + an z n , where a0 , a1 , a2 , . . . , an ∈ C are constants.

Show that for every k = 0, 1, . . . , n, we have

P (k) (0)
ak = .
k!
b) Apply the result to the polynomial (1 + z)n = c0 + c1 z + c2 z 2 + . . . + cn z n and show that for
every k = 0, 1, . . . , n, we have
n!
ck = .
k!(n − k)!

6. a) Show that for every z ∈ C, we have eiz = cos z + i sin z.

b) Show that for every z, w ∈ C, we have

cos(z + w) + i sin(z + w) = (cos z + i sin z)(cos w + i sin w)

and

cos(z + w) − i sin(z + w) = (cos z − i sin z)(cos w − i sin w).

c) Express sin(z + w) and cos(z + w) in terms of sin z, sin w, cos z and cos w.

7. Suppose that a1 , a2 , . . . , an ∈ C are distinct, and consider the polynomial

Q(z) = (z − z1 )(z − z2 ) . . . (z − zn ).

Suppose further that P (z) is a polynomial of degree less than n. Follow the steps below to show
that there exist a1 , a2 , . . . , an ∈ C such that
P (z) a1 a2 an
= + + ... + .
Q(z) z − z1 z − z2 z − zn

a) We shall first of all show that the expression above is possible by multiplying it by Q(z) and
then determining a1 , a2 , . . . , an so that the resulting equation between polynomials of degree
less than n holds when z = z1 , z2 , . . . , zn .
[Hint: Recall Problem 1 in Chapter 1.]
b) Show that for every k = 1, . . . , n, we have

P (z) P (zk )
ak = lim (z − zk ) =  .
z→zk Q(z) Q (zk )

[Hint: Note that Q(zk ) = 0 for every k = 1, . . . , n.]

 Chapter 3 : Complex Differentiation 3–15

8. Suppose that a ∈ C is non-zero. Show that for any fixed choice of value for log a, the function
f (z) = az = ez log a satisfies f  (z) = f (z) log a.

9. For each expression below, compute all possible values and plot their positions in the complex plane:
a) log(−i) b) log(1 + i)
c) (−i)−i d) i2
e) 2πi f) (1 + i)i (1 + i)−i

10. For each of the following equations, find all solutions:

a) Log(z) = πi/3 b) ez = 2i
c) sin z = i d) sin z = − cos z
e) tan2 z = −1

11. For each of the functions below, determine whether the function is harmonic. If so, find also its
harmonic conjugate:
a) x2 − y 2 + y b) ex sin y
c) x − y
3 3
d) xex cos y − yex sin y
e) 3x y − y + xy
2 3
f) x4 − 6x2 y 2 + y 4 + x3 y − xy 3
g) ex −y sin 2xy
2 2

12. a) Suppose that the functions f (z) and g(z) both satisfy the Cauchy-Riemann equations at a
particular point z ∈ C. Show that the functions f (z) + g(z) and f (z)g(z) also satisfy the
Cauchy-Riemann equations at the point z.
b) Show that the constant function and the function f (z) = z both satisfy the Cauchy-Riemann
equations everywhere in C.
c) Deduce that every polynomial P (z) with complex coefficients satisfies the Cauchy-Riemann
equations everywhere in C.

13. A real valued function u(x, y) which is continuous and satisfies the inequality uxx + uyy ≥ 0 in a
region D is said to be subharmonic in D. Show that u = |f (z)|2 is subharmonic in any region where
f (z) is analytic.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 1
COMPLEX NUMBERS

1.1. Arithmetic and Conjugates

The purpose of this chapter is to give a review of various properties of the complex numbers that we shall
need in the discussion of complex analysis. As the reader is expected to be familiar with the material,
all proofs have been omitted.

The equation x2 + 1 = 0 has no solution x ∈ R. To “solve” this equation, we have to introduce extra
numbers into our number system. To do this, we define the number i by i2 + 1 = 0, and then extend the
field of all real numbers by adjoining the number i, which is then combined with the real numbers by the
operations addition and multiplication in accordance with the Field axioms of the real number system.
The numbers a + ib, where a, b ∈ R, of the extended field are then added and multiplied in accordance
with the Field axioms, suitably extended, and the restriction i2 + 1 = 0. Note that the number a + 0i,
where a ∈ R, behaves like the real number a.

What we have said in the last paragraph basically amounts to the following. Consider two complex
numbers a + ib and c + id, where a, b, c, d ∈ R. We have the addition and multiplication rules

(a + ib) + (c + id) = (a + c) + i(b + d) and (a + ib)(c + id) = (ac − bd) + i(ad + bc).

These lead to the subtraction rule

(a + ib) − (c + id) = (a − c) + i(b − d),

and the division rule, that if c + id = 0, then

a + ib ac + bd bc − ad
= 2 +i 2 .
c + id c + d2 c + d2
1–2 W W L Chen : Introduction to Complex Analysis

Note the special case a = 1 and b = 0.

Suppose that z = x + iy, where x, y ∈ R. The real number x is called the real part of z, and denoted
by x = Rez. The real number y is called the imaginary part of z, and denoted by y = Imz. The set
C = {z = x + iy : x, y ∈ R} is called the set of all complex numbers. The complex number z = x − iy is
called the conjugate of z.

It is easy to see that for every z ∈ C, we have

z+z z−z
Rez = and Imz = .
2 2i
Furthermore, if w ∈ C, then

z+w =z+w and zw = z w.

1.2. Polar Coordinates

Suppose that z = x + iy, where x, y ∈ R. The real number

r = x2 + y 2

is called the modulus of z, and denoted by |z|. On the other hand, if z = 0, then any number θ ∈ R
satisfying the equations

(1) x = r cos θ and y = r sin θ

is called an argument of z, and denoted by arg z. Hence we can write z in polar form

z = r(cos θ + i sin θ).

Note, however, that for a given z ∈ C, arg z is not unique. Clearly we can add any integer multiple of
2π to θ without affecting (1). We sometimes call a real number θ ∈ R the principal argument of z if θ
satisfies the equations (1) and −π < θ ≤ π. The principal argument of z is usually denoted by Arg z.

It is easy to see that for every z ∈ C, we have |z|2 = zz. Also, if w ∈ C, then

|zw| = |z||w| and |z + w| ≤ |z| + |w|.

Furthermore, if

z = r(cos θ + i sin θ) and w = s(cos φ + i sin φ),

where r, s, θ, φ ∈ R and r, s > 0, then

z r
zw = rs(cos(θ + φ) + i sin(θ + φ)) and = (cos(θ − φ) + i sin(θ − φ)).
w s

1.3. Rational Powers

De Moivre’s theorem, that

(2) cos nθ + i sin nθ = (cos θ + i sin θ)n for every n ∈ N and θ ∈ R,

 Chapter 1 : Complex Numbers 1–3

is useful in finding n-th roots of complex numbers.

Suppose that c = R(cos α + i sin α), where R, α ∈ R and R > 0. Then the solutions of the equation
z n = c are given by
 
√ α + 2kπ α + 2kπ
where k = 0, 1, . . . , n − 1.
n
z= R cos + i sin ,
n n

Finally, we can define cb for any b ∈ Q and non-zero c ∈ C as follows. The rational number b can
be written uniquely in the form b = p/q, where p ∈ Z and q ∈ N have no prime factors in common.
Then there are exactly q distinct numbers z satisfying z q = c. We now define cb = z p , noting that the
expression (2) can easily be extended to all n ∈ Z. It is not too difficult to show that there are q distinct
values for the rational power cb .

Problems for Chapter 1

1. Suppose that z0 ∈ C is fixed. A polynomial P (z) is said to be divisible by z − z0 if there is another

polynomial Q(z) such that P (z) = (z − z0 )Q(z).
a) Show that for every c ∈ C and k ∈ N, the polynomial c(z k − z0k ) is divisible by z − z0 .
b) Consider the polynomial P (z) = a0 + a1 z + a2 z 2 + . . . + an z n , where a0 , a1 , a2 , . . . , an ∈ C are
arbitrary. Show that the polynomial P (z) − P (z0 ) is divisible by z − z0 .
c) Deduce that P (z) is divisible by z − z0 if P (z0 ) = 0.
d) Suppose that a polynomial P (z) of degree n vanishes at n distinct values z1 , z2 , . . . , zn ∈ C, so
that P (z1 ) = P (z2 ) = . . . = P (zn ) = 0. Show that P (z) = c(z − z1 )(z − z2 ) . . . (z − zn ), where
c ∈ C is a constant.
e) Suppose that a polynomial P (z) of degree n vanishes at more than n distinct values. Show
that P (z) = 0 identically.
 
 z−α 

2. Suppose that α ∈ C is fixed and |α| < 1. Show that |z| ≤ 1 if and only if   ≤ 1.
1 − αz 

3. Suppose that z = x + iy, where x, y ∈ R. Express each of the following in terms of x and y:
   
z + 1  z+i 
a) |z − 1|3 b)   c)  
z − 1  1 − iz 

4. Suppose that c ∈ R and α ∈ C with α = 0.

a) Show that αz + αz + c = 0 is the equation of a straight line on the plane.
b) What does the equation zz + αz + αz + c = 0 represent if |α|2 ≥ c?

5. Suppose that z, w ∈ C. Show that |z + w|2 + |z − w|2 = 2(|z|2 + |w|2 ).

6. Find all the roots of the equation (z 8 − 1)(z 3 + 8) = 0.

7. For each of the following, compute all the values and plot them on the plane:
a) (1 + i)−1/2 b) (−4)3/4 c) (1 − i)3/8
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 2
FOUNDATIONS OF COMPLEX ANALYSIS

2.1. Three Approaches

We start by remarking that analysis is sometimes known as the study of the four C’s: convergence,
continuity, compactness and connectedness. In real analysis, we have studied convergence and continuity
to some depth, but the other two concepts have been somewhat disguised. In this course, we shall try
to illustrate these two latter concepts a little bit more, particularly connectedness.

Complex analysis is the study of complex valued functions of complex variables. Here we shall
restrict the number of variables to one, and study complex valued functions of one complex variable.
Unless otherwise stated, all functions in these notes are of the form f : S → C, where S is a set in C.

We shall study the behaviour of such functions using three different approaches. The first of these,
discussed in Chapter 3 and usually attributed to Riemann, is based on differentiation and involves pairs
of partial differential equations called the Cauchy-Riemann equations. The second approach, discussed in
Chapters 4–11 and usually attributed to Cauchy, is based on integration and depends on a fundamental
theorem known nowadays as Cauchy’s integral theorem. The third approach, discussed in Chapter 16
and usually attributed to Weierstrass, is based on the theory of power series.

2.2. Point Sets in the Complex Plane

We shall study functions of the form f : S → C, where S is a set in C. In most situations, various
properties of the point sets S play a crucial role in our study. We therefore begin by discussing various
types of point sets in the complex plane.

Before making any definitions, let us consider a few examples of sets which frequently occur in our
subsequent discussion.
2–2 W W L Chen : Introduction to Complex Analysis

Example 2.2.1. Suppose that z0 ∈ C, r, R ∈ R and 0 < r < R. The set {z ∈ C : |z − z0 | < R}
represents a disc, with centre z0 and radius R, and the set {z ∈ C : r < |z − z0 | < R} represents an
annulus, with centre z0 , inner radius r and outer radius R.

R R
r
z0 z0

Example 2.2.2. Suppose that A, B ∈ R and A < B. The set {z = x + iy ∈ C : x, y ∈ R and x > A}
represents a half-plane, and the set {z = x + iy ∈ C : x, y ∈ R and A < x < B} represents a strip.

A A B

Example 2.2.3. Suppose that α, β ∈ R and 0 ≤ α < β < 2π. The set

{z = r(cos θ + i sin θ) ∈ C : r, θ ∈ R and r > 0 and α < θ < β}

represents a sector.

β
α

We now make a number of important definitions. The reader may subsequently need to return to
these definitions.
 Chapter 2 : Foundations of Complex Analysis 2–3

Definition. Suppose that z0 ∈ C and  ∈ R, with  > 0. By an -neighbourhood of z0 , we mean a

disc of the form {z ∈ C : |z − z0 | < }, with centre z0 and radius  > 0.

Definition. Suppose that S is a point set in C. A point z0 ∈ S is said to be an interior point of S

if there exists an -neighbourhood of z0 which is contained in S. The set S is said to be open if every
point of S is an interior point of S.

z0

Example 2.2.4. The sets in Examples 2.2.1–2.2.3 are open.

Example 2.2.5. The punctured disc {z ∈ C : 0 < |z − z0 | < R} is open.

Example 2.2.6. The disc {z ∈ C : |z − z0 | ≤ R} is not open.

Example 2.2.7. The empty set ∅ is open. Why?

Definition. An open set S is said to be connected if every two points z1 , z2 ∈ S can be joined by the
union of a finite number of line segments lying in S. An open connected set is called a domain.

z1 z2

Remarks. (1) Sometimes, we say that an open set S is connected if there do not exist non-empty
open sets S1 and S2 such that S1 ∪ S2 = S and S1 ∩ S2 = ∅. In other words, an open connected set
cannot be the disjoint union of two non-empty open sets.

(2) In fact, it can be shown that the two definitions are equivalent.
2–4 W W L Chen : Introduction to Complex Analysis

(3) Note that we have not made any definition of connectedness for sets that are not open. In
fact, the definition of connectedness for an open set given by (1) here is a special case of a much more
complicated definition of connectedness which applies to all point sets.

Example 2.2.8. The sets in Examples 2.2.1–2.2.3 are domains.

Example 2.2.9. The punctured disc {z ∈ C : 0 < |z − z0 | < R} is a domain.

Definition. A point z0 ∈ C is said to be a boundary point of a set S if every -neighbourhood of z0

contains a point in S as well as a point not in S. The set of all boundary points of a set S is called the
boundary of S.

z0

Example 2.2.10. The annulus {z ∈ C : r < |z − z0 | < R}, where 0 < r < R, has boundary C1 ∪ C2 ,
where

C1 = {z ∈ C : |z − z0 | = r} and C2 = {z ∈ C : |z − z0 | = R}

are circles, with centre z0 and radius r and R respectively. Note that the annulus is connected and hence
a domain. However, note that its boundary is made up of two separate pieces.

Definition. A region is a domain together with all, some or none of its boundary points. A region
which contains all its boundary points is said to be closed. For any region S, we denote by S the closed
region containing S and all its boundary points, and call S the closure of S.

Remark. Note that we have not made any definition of closedness for sets that are not regions. In
fact, our definition of closedness for a region here is a special case of a much more complicated definition
of closedness which applies to all point sets.

Definition. A region S is said to be bounded or finite if there exists a real number M such that
|z| ≤ M for every z ∈ S. A region that is closed and bounded is said to be compact.

Example 2.2.11. The region {z ∈ C : |z − z0 | ≤ R} is closed and bounded, hence compact. It is called
the closed disc with centre z0 and radius R.

Example 2.2.12. The region {z = x + iy ∈ C : x, y ∈ R and 0 ≤ x ≤ 1} is closed but not bounded.

Example 2.2.13. The square {z = x + iy ∈ C : x, y ∈ R and 0 ≤ x ≤ 1 and 0 < y < 1} is bounded

but not closed.
 Chapter 2 : Foundations of Complex Analysis 2–5

2.3. Complex Functions

In these lectures, we study complex valued functions of one complex variable. In other words, we study
functions of the form f : S → C, where S is a set in C. Occasionally, we will abuse notation and simply
refer to a function by its formula, without explicitly defining the domain S. For instance, when we
discuss the function f (z) = 1/z, we implicitly choose a set S which will not include the point z = 0
where the function is not defined. Also, we may occasionally wish to include the point z = ∞ in the
domain or codomain.

We may separate the independent variable z as well as the dependent variable w = f (z) into real
and imaginary parts. Our usual notation will be to write z = x + iy and w = f (z) = u + iv, where
x, y, u, v ∈ R. It follows that u = u(x, y) and v = v(x, y) can be interpreted as real valued functions of
the two real variables x and y.

Example 2.3.1. Consider the function f : S → C, given by f (z) = z 2 and where S = {z ∈ C : |z| < 2}
is the open disc with radius 2 and centre 0. Using polar coordinates, it is easy to see that the range of
the function is the open disc f (S) = {w ∈ C : |w| < 4} with radius 4 and centre 0.

Example 2.3.2. Consider the function f : H → C, where H = {z = x + iy ∈ C : y > 0} is the upper

half-plane and f (z) = z 2 . Using polar coordinates, it is easy to see that the range of the function is the
complex plane minus the non-negative real axis.

Example 2.3.3. Consider the function f : T → C, where T = {z = x + iy ∈ C : 1 < x < 2} is a strip

and f (z) = z 2 . Let x0 ∈ (1, 2) be fixed, and consider the image of a point (x0 , y) on the vertical line
x = x0 . Here we have

u = x20 − y 2 and v = 2x0 y.

Eliminating y, we obtain the equation of a parabola

v2
u = x20 −
4x20

in the w-plane. It follows that the image of the vertical line x = x0 under the function w = z 2 is this
parabola. Now the boundary of the strip are the two lines x = 1 and x = 2. Their images under the
mapping w = z 2 are respectively the parabolas

v2 v2
u=1− and u=4− .
4 16

It is easy to see that the range of the function is the part of the w-plane between these two parabolas.

z w
8

1 2 1 4
2–6 W W L Chen : Introduction to Complex Analysis

Example 2.3.4. Consider again the function w = z 2 . We would like to find all z = x + iy ∈ C for
which 1 < Rew < 2. In other words, we have the restriction 1 < u < 2, but no rectriction on v. Let
u0 ∈ (1, 2) be fixed, and consider points (x, y) in the z-plane with images on the vertical line u = u0 .
Here we have the hyperbola

x2 − y 2 = u0 .

The boundaries u = 1 and u = 2 are represented by the hyperbolas

x2 − y 2 = 1 and x2 − y 2 = 2.

It is easy to see that the points in question are precisely those between the two hyperbolas.

w z

1 2 1

2.4. Extended Complex Plane

It is sometimes useful to extend the complex plane C by the introduction of the point ∞ at infinity. Its
connection with finite complex numbers can be established by setting z + ∞ = ∞ + z = ∞ for all z ∈ C,
and setting z · ∞ = ∞ · z = ∞ for all non-zero z ∈ C. We can also write ∞ · ∞ = ∞.

Note that it is not possible to define ∞ + ∞ and 0 · ∞ without violating the laws of arithmetic.
However, by special convention, we shall write z/0 = ∞ for z = 0 and z/∞ = 0 for z = ∞.

In the complex plane C, there is no room for a point corresponding to ∞. We can, of course,
introduce an “ideal” point which we call the point at infinity. The points in C, together with the point
at infinity, form the extended complex plane. We decree that every straight line on the complex plane
shall pass through the point at infinity, and that no half-plane shall contain the ideal point.

The main purpose of this section is to introduce a geometric model in which each point of the
extended complex plane has a concrete representative. To do this, we shall use the idea of stereographic
projection.

Consider a sphere of radius 1 in R3 . A typical point on this sphere will be denoted by P (x1 , x2 , x3 ).
Note that x21 + x22 + x23 = 1. Let us call the point N (0, 0, 1) the north pole. The equator of this sphere is
the set of all points of the form (x1 , x2 , 0), where x21 + x22 = 1. Consider next the complex plane C. This
can be viewed as a plane in R3 . Let us position this plane in such a way that the equator of the sphere
lies on this plane; in other words, our copy of the complex plane is “horizontal” and passes through the
origin. We can further insist that the x-direction on our complex plane is the same as the x1 -direction
in R3 , and that the y-direction on our complex plane is the same as the x2 -direction in R3 . Clearly a
typical point z = x + iy on our complex plane C can be identified with the point Z(x, y, 0) in R3 .
 Chapter 2 : Foundations of Complex Analysis 2–7

Suppose that Z(x, y, 0) is on the plane. Consider the straight line that passes through Z and the
north pole N . It is not too difficult to see that this straight line intersects the surface of the sphere at
precisely one other point P (x1 , x2 , x3 ). In fact, if Z is on the equator of the sphere, then P = Z. If Z is
on the part of the plane outside the sphere, then P is on the northern hemisphere, but is not the north
pole N . If Z is on the part of the plane inside the sphere, then P is on the southern hemisphere. Check
that for Z(0, 0, 0), the point P (0, 0, −1) is the south pole.

N y
P

On the other hand, if P is any point on the sphere different from the north pole N , then a straight
line passing through P and N intersects the plane at precisely one point Z. It follows that there is a
pairing of all the points P on the sphere different from the north pole N and all the points on the plane.
This pairing is governed by the requirement that the straight line through any pair must pass through
the north pole N .

We can now visualize the north pole N as the point on the sphere corresponding to the point at
infinity of the plane. The sphere is called the Riemann sphere.

2.5. Limits and Continuity

The concept of a limit in complex analysis is exactly the same as in real analysis. So, for example, we
say that f (z) → L as z → z0 , or

lim f (z) = L,
z→z0

if, given any  > 0, there exists δ > 0 such that |f (z) − L| <  whenever 0 < |z − z0 | < δ.

This definition will be perfectly in order if the function f is defined in some open set containing
z0 , with the possible exception of z0 itself. It follows that if z0 is an interior point of the region S of
definition of the function, our definition is in order. However, if z0 is a boundary point of the region S
of definition of the function, then we agree that the conclusion |f (z) − L| <  need only hold for those
z ∈ S satisfying 0 < |z − z0 | < δ.

Similarly, we say that a function f (z) is continuous at z0 if f (z) → f (z0 ) as z → z0 . A similar

qualification on z applies if z0 is a boundary point of the region S of definition of the function. We also
say that a function is continuous in a region if it is continuous at every point of the region.
2–8 W W L Chen : Introduction to Complex Analysis

Note that for a function to be continuous in a region, it is enough to have continuity at every point of
the region. Hence the choice of δ may depend on a point z0 in question. If δ can be chosen independently
of z0 , then we have some uniformity as well. To be precise, we make the following definition.

Definition. A function f (z) is said to be uniformly continuous in a region S if, given any  > 0, there
exists δ > 0 such that |f (z1 ) − f (z2 )| <  for every z1 , z2 ∈ S satisfying |z1 − z2 | < δ.

Remark. Note that if we fix z2 to be a point z0 and write z for z1 , then we require |f (z) − f (z0 )| < 
for every z ∈ S satisfying |z − z0 | < δ. In other words, δ cannot depend on z0 .

Example 2.5.1. Consider the punctured disc S = {z ∈ C : 0 < |z| < 1}. The function f (z) = 1/z is
continuous in S but not uniformly continuous in S. To see this, note first of all that continuity follows
from the simple observation that the function z is continuous and non-zero in S. To show that the
function is not uniformly continuous in S, it suffices to show that there exists  > 0 such that for every
δ > 0, there exist z1 , z2 ∈ S such that




1

|z1 − z2 | < δ and
− 1
≥ .

z1 z2

Let  = 1. For every δ > 0, choose n ∈ N such that n > δ −1/2 , and let

1 1
z1 = and z2 = .
n n+1

Clearly z1 , z2 ∈ S. It is easy to see that

1 1

1
1



|z1 − z2 | =
− = <δ and
− 1
= 1.
n n + 1
n(n + 1)
z1 z2

Problems for Chapter 2

1. For each of the following functions, find f (z + 3), f (1/z) and f (f (z)):
1−z
a) f (z) = z − 1 b) f (z) = z 2 c) f (z) = 1/z d) f (z) =
3+z

2. Which of the sets below are domains?

a) {z : 0 < |z| < 1} b) {z : Imz < 3|z|} c) {z : |z − 1| ≤ |z + 1|}
d) {z : |z 2 − 1| < 1} e) {z : 0 < Rez ≤ 1}

3. Find the image of the strip {z : |Rez| < 1} and of the disc {z : |z| < 1} under each of the following
mappings:
z+1
a) w = (1 + i)z + 1 b) w = 2z 2 c) w = z −1 d) w =
z−1

4. A function f (z) is said to be an isometry if |f (z1 ) − f (z2 )| = |z1 − z2 | for every z1 , z2 ∈ C; in other
words, if it preserves distance.
a) Suppose that f (z) is an isometry. Show that for every a, b ∈ C with |a| = 1, the function
g(z) = af (z) + b is also an isometry.
b) Show that the function

f (z) − f (0)
h(z) =
f (1) − f (0)

is an isometry with h(0) = 0 and h(1) = 1.

 Chapter 2 : Foundations of Complex Analysis 2–9

c) Suppose that k(z) is an isometry with k(0) = 0 and k(1) = 1. Show that Rek(z) = Rez, and
that k(i) = ±i.
[Hint: Explain first of all why |k(z)| = |z| and |1 − k(z)| = |1 − z|.]
d) Suppose that in (c), we have k(i) = i. Show that Imk(z) = Imz and that k(z) = z for all
z ∈ C.
e) Suppose that in (c), we have k(i) = −i. Show that Imk(z) = −Imz and that k(z) = z for all
z ∈ C.
f) Deduce that every isometry has the form f (z) = az + b or f (z) = az + b, where a, b ∈ C with
|a| = 1.

5. In the notation of Section 2.4, let the point z = x + iy on the complex plane C correspond to the
point (x1 , x2 , x3 ) of the sphere under stereographic projection, so that the three points (0, 0, 1),
(x1 , x2 , x3 ) and (x, y, 0) are collinear. Note that (x1 , x2 , x3 − 1) = λ(x, y, −1) for some λ ∈ R, and
that x21 + x22 + x23 = 1.
 
2x 2y |z|2 − 1
a) Show that (x1 , x2 , x3 ) = , , .
|z|2 + 1 |z|2 + 1 |z|2 + 1
b) Note that a circle on the sphere is the intersection of the sphere with a plane ax1 +bx2 +cx3 = d.
By expressing this equation of the plane in terms of x and y, show that a circle on the sphere
not containing the pole (0, 0, 1) corresponds to a circle in the complex plane. Show also that a
circle on the sphere containing the pole (0, 0, 1) corresponds to a line in the complex plane.
c) Suppose that (x1 , x2 , x3 ) and (x1 , x2 , x3 ) are two points on the sphere corresponding to the com-
plex numbers z and z  respectively. Show that the distance between (x1 , x2 , x3 ) and (x1 , x2 , x3 )
is given by

2|z − z  |
d(z, z  ) =   .
1 + |z|2 1 + |z  |2

[Remark: The number d(z, z  ) is known as the chordal distance.]

6. Each of the following functions is not defined at z = z0 . What value must f (z0 ) take to ensure
continuity at z = z0 ?
z − z0 z 3 − z03
a) f (z) = b) f (z) =
z − z0 z − z0
   
1 1 1 1 1 1
c) f (z) = − d) f (z) = −
z − z0 z z0 z − z0 z 3 z03

7. Suppose that

a0 + a1 z + a2 z 2
f (z) = ,
b 0 + b 1 z + b2 z 2

where a0 , a1 , a2 , b0 , b1 , b2 ∈ C. Examine the behaviour of f (z) at z = 0 and at z = ∞.

 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 3
COMPLEX DIFFERENTIATION

3.1. Introduction

Suppose that D ⊆ C is a domain. A function f : D → C is said to be differentiable at z0 ∈ D if the limit

f (z) − f (z0 )
lim
z→z0 z − z0

exists. In this case, we write

f (z) − f (z0 )
(1) f  (z0 ) = lim ,
z→z0 z − z0

and call f  (z0 ) the derivative of f at z0 .

If z = z0 , then


f (z) − f (z0 )
f (z) = (z − z0 ) + f (z0 ).
z − z0

It follows from (1) and the arithmetic of limits that if f  (z0 ) exists, then f (z) → f (z0 ) as z → z0 , so
that f is continuous at z0 . In other words, differentiability at z0 implies continuity at z0 .

Note that the argument here is the same as in the case of a real valued function of a real variable. In
fact, the similarity in argument extends to the arithmetic of limits. Indeed, if the functions f : D → C
and g : D → C are both differentiable at z0 ∈ D, then both f + g and f g are differentiable at z0 , and

(f + g) (z0 ) = f  (z0 ) + g  (z0 ) and (f g) (z0 ) = f (z0 )g  (z0 ) + f  (z0 )g(z0 ).
3–2 W W L Chen : Introduction to Complex Analysis

If the extra condition g  (z0 ) = 0 holds, then f /g is differentiable at z0 , and


f g(z0 )f  (z0 ) − f (z0 )g  (z0 )
(z0 ) = .
g g 2 (z0 )

One can also establish the Chain rule for differentiation as in real analysis. More precisely, suppose
that the function f is differentiable at z0 and the function g is differentiable at w0 = f (z0 ). Then the
function g ◦ f is differentiable at z = z0 , and

(g ◦ f ) (z0 ) = g  (w0 )f  (z0 ).

Example 3.1.1. Consider the function f (z) = z, where for every z ∈ C, z denotes the complex
conjugate of z. Suppose that z0 ∈ C. Then

f (z) − f (z0 ) z − z0 z − z0
(2) = = .
z − z0 z − z0 z − z0

If z − z0 = h is real and non-zero, then (2) takes the value 1. On the other hand, if z − z0 = ik is purely
imaginary, then (2) takes the value −1. It follows that this function is not differentiable anywhere in C,
although its real and imaginary parts are rather well behaved.

3.2. The Cauchy-Riemann Equations

If we use the notation

f (z + h) − f (z)
f  (z) = lim ,
h→0 h
then in Example 3.1.1, we have examined the behaviour of the ratio

f (z + h) − f (z)
h
first as h → 0 through real values and then through imaginary values. Indeed, for the derivative
to exist, it is essential that these two limiting processes produce the same limit f  (z). Suppose that
f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real valued functions. If h is real, then the
two limiting processes above correspond to

f (z + h) − f (z) u(x + h, y) − u(x, y) v(x + h, y) − v(x, y) ∂u ∂v

lim = lim + i lim = +i
h→0 h h→0 h h→0 h ∂x ∂x
and
f (z + ih) − f (z) u(x, y + h) − u(x, y) v(x, y + h) − v(x, y) ∂v ∂u
lim = lim + i lim = −i
h→0 ih h→0 ih h→0 ih ∂y ∂y

respectively. Equating real and imaginary parts, we obtain

∂u ∂v ∂u ∂v
(3) = and =− .
∂x ∂y ∂y ∂x

Note that while the existence of the derivative in real analysis is a mild smoothness condition, the
existence of the derivative in complex analysis leads to a pair of partial differential equations.
 Chapter 3 : Complex Differentiation 3–3

Definition. The partial differential equations (3) are called the Cauchy-Riemann equations.

We have proved the following result.

THEOREM 3A. Suppose that f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real
valued functions. Suppose further that f  (z) exists. Then the four partial derivatives in (3) exist, and
the Cauchy-Riemann equations (3) hold. Furthermore, we have

∂u ∂v ∂v ∂u
(4) f  (z) = +i and f  (z) = −i .
∂x ∂x ∂y ∂y

A natural question to ask is whether the Cauchy-Riemann equations are sufficient to guarantee
the existence of the derivative. We shall show next that we require also the continuity of the partial
derivatives in (3).

THEOREM 3B. Suppose that f (z) = u(x, y) + iv(x, y), where z = x + iy, and u and v are real
valued functions. Suppose further that the four partial derivatives in (3) are continuous and satisfy the
Cauchy-Riemann equations (3) at z0 . Then f is differentiable at z0 , and the derivative f  (z0 ) is given
by the equations (4) evaluated at z0 .

Proof. Write z0 = x0 + iy0 . Then

f (z) − f (z0 ) (u(x, y) − u(x0 , y0 )) + i(v(x, y) − v(x0 , y0 ))

= .
z − z0 z − z0
We can write



∂u ∂u
u(x, y) − u(x0 , y0 ) = (x − x0 ) + (y − y0 ) + |z − z0 |1 (z)
∂x z0 ∂y z0

and



∂v ∂v
v(x, y) − v(x0 , y0 ) = (x − x0 ) + (y − y0 ) + |z − z0 |2 (z).
∂x z0 ∂y z0

If the four partial derivatives in (3) are continuous at z0 , then

lim 1 (z) = 0 and lim 2 (z) = 0.

z→z0 z→z0

In view of the Cauchy-Riemann equations (3), we have

(u(x, y) − u(x0 , y0 )) + i(v(x, y) − v(x0 , y0 ))



∂u ∂v ∂u ∂v
= (x − x0 ) +i + (y − y0 ) +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂y ∂y z0



∂u ∂v ∂v ∂u
= (x − x0 ) +i + (y − y0 ) − +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂x ∂x z0



∂u ∂v ∂u ∂v
= (x − x0 ) +i + i(y − y0 ) +i + |z − z0 |(1 (z) + i2 (z))
∂x ∂x z0 ∂x ∂x z0


∂u ∂v
= (z − z0 ) +i + |z − z0 |(1 (z) + i2 (z)).
∂x ∂x z0

Hence




f (z) − f (z0 ) ∂u ∂v |z − z0 | ∂u ∂v
= +i + (1 (z) + i2 (z)) → +i
z − z0 ∂x ∂x z0 z − z0 ∂x ∂x z0
3–4 W W L Chen : Introduction to Complex Analysis

as z → z0 , giving the desired results.

3.3. Analytic Functions

In the previous section, we have shown that differentiability in complex analysis leads to a pair of partial
differential equations. Now partial differential equations are seldom of interest at a single point, but
rather in a region. It therefore seems reasonable to make the following definition.

Definition. A function f is said to be analytic at a point z0 ∈ C if it is differentiable at every z in

some -neighbourhood of the point z0 . The function f is said to be analytic in a region if it is analytic
at every point in the region. The function f is said to be entire if it is analytic in C.

Example 3.3.1. Consider the function f (z) = |z|2 . In our usual notation, we clearly have

u = x2 + y 2 and v = 0.

The Cauchy-Riemann equations

2x = 0 and 2y = 0

can only be satisfied at z = 0. It follows that the function is differentiable only at the point z = 0, and
is therefore analytic nowhere.

Example 3.3.2. The function f (z) = z 2 is entire.

Example 3.3.3. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant real part u. Then clearly

∂u ∂u
=0 and = 0.
∂x ∂y

Since f is analytic in D, it is differentiable at every point in D, and so the Cauchy-Riemann equations

hold in D. It follows that
∂v ∂v
=0 and = 0.
∂x ∂y

Hence f must have constant imaginary part v, and so f must be constant in D.

Example 3.3.4. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant imaginary part v. A similar argument shows that f must have constant real part u. Hence f
must be constant in D.

Example 3.3.5. Suppose that the function f is analytic in a domain D. Suppose further that f has
constant modulus. In other words, u2 + v 2 = C for some non-negative real number C. Differentiating
this with respect to x and to y, we obtain respectively

∂u ∂v ∂u ∂v
2u + 2v =0 and 2u + 2v = 0.
∂x ∂x ∂y ∂y

In view of the Cauchy-Riemann equations, these can be written as

∂u ∂u ∂u ∂u
2u − 2v =0 and 2v + 2u = 0.
∂x ∂y ∂x ∂y
 Chapter 3 : Complex Differentiation 3–5

In matrix notation, these become

 
∂u



u −v  ∂x 
 = 0 .
v u  ∂u  0
∂y
Note now that


u −v
det = u2 + v 2 = C.
v u

If C > 0, then we must have the unique solution

∂u ∂u
=0 and = 0,
∂x ∂y

so that the real part u is constant. It then follows from Example 3.3.3 that f is constant in D. On the
other hand, if C = 0, then clearly u = v = 0, so that f = 0 in D.

3.4. Introduction to Special Functions

In this section, we shall generalize various functions that we have studied in real analysis to the complex
domain. Consider first of all the exponential function. It seems reasonable to extend the property
ex1 +x2 = ex1 ex2 for real variables to complex values of the variables to obtain

ez = ex+iy = ex eiy , where x, y ∈ R.

This suggests the following definition.

Definition. Suppose that z = x + iy, where x, y ∈ R. Then the exponential function ez is defined for
every z ∈ C by

(5) ez = ex (cos y + i sin y).

If we write ez = u(x, y) + iv(x, y), then

u(x, y) = ex cos y and v(x, y) = ex sin y.

It is easy to check that the Cauchy-Riemann equations are satisfied for every z ∈ C, so that ez is an
entire function. Furthermore, it follows from (4) that

d z ∂u ∂v
e = +i = ex cos y + iex sin y = ex (cos y + i sin y) = ez ,
dz ∂x ∂x
so that ez is its own derivative. On the other hand, note that for every y1 , y2 ∈ R, we have

ei(y1 +y2 ) = cos(y1 + y2 ) + i sin(y1 + y2 ) = (cos y1 + i sin y1 )(cos y2 + i sin y2 ) = eiy1 eiy2 .

Furthermore, if x1 , x2 ∈ R, then

ex1 +x2 ei(y1 +y2 ) = (ex1 ex2 )(eiy1 eiy2 ) = (ex1 eiy1 )(ex2 eiy2 ).

Writing z1 = x1 + iy1 and z2 = x2 + iy2 , we deduce the addition formula

ez1 +z2 = ez1 ez2 .

3–6 W W L Chen : Introduction to Complex Analysis

Finally, note that

|ez | = |ex (cos y + i sin y)| = ex | cos y + i sin y| = ex .

Since ex is never zero, it follows that the exponential function ez is non-zero for every z ∈ C.

Next, we turn our attention to the trigonometric functions. Note first of all that if z = x + iy, where
x, y ∈ R, then iz = −y + ix. Replacing z in (5) by iz and by −iz gives respectively

eiz = e−y (cos x + i sin x) and e−iz = ey (cos x − i sin x).

The special case y = 0 gives respectively

eix = cos x + i sin x and e−ix = cos x − i sin x.

It follows that

eix + e−ix eix − e−ix

cos x = and sin x = .
2 2i
This suggests the following definition.

Definition. Suppose that z ∈ C. Then the trigonometric functions cos z and sin z are defined in terms
of the exponential function by

eiz + e−iz eiz − e−iz

(6) cos z = and sin z = .
2 2i

Since the exponential function is an entire function, it follows easily from (6) that both cos z and
sin z are entire functions. Furthermore, it can easily be deduced from (6) that

d d
cos z = − sin z and sin z = cos z.
dz dz
We can define the functions tan z, cot z, sec z and cosec z in terms of the functions cos z and sin z as in
real variables. However, note that these four functions are not entire. Also, we can deduce from (6) the
formulas

cos(z1 + z2 ) = cos z1 cos z2 − sin z1 sin z2 and sin(z1 + z2 ) = sin z1 cos z2 + cos z1 sin z2 ,

and a host of other trigonometric identities that we know hold for real variables.

Finally, we turn our attention to the hyperbolic functions. These are defined as in real analysis.

Definition. Suppose that z ∈ C. Then the hyperbolic functions cosh z and sinh z are defined in terms
of the exponential function by

ez + e−z ez − e−z
(7) cosh z = and sinh z = .
2 2

Since the exponential function is an entire function, it follows easily from (7) that both cosh z and
sinh z are entire functions. Furthermore, it can easily be deduced from (7) that

d d
cosh z = sinh z and sinh z = cosh z.
dz dz
 Chapter 3 : Complex Differentiation 3–7

We can define the functions tanh z, coth z, sech z and cosech z in terms of the functions cosh z and sinh z
as in real variables. However, note that these four functions are not entire. Also, we can deduce from
(7) a host of hyperbolic identities that we know hold for real variables. Note also that comparing (6)
and (7), we obtain

cosh z = cos iz and sinh z = −i sin iz.

3.5. Periodicity and its Consequences

One of the fundamental differences between real and complex analysis is that the exponential function
is periodic in C.

Definition. A function f is periodic in C if there is some fixed non-zero ω ∈ C such that the identity
f (z + ω) = f (z) holds for every z ∈ C. Any constant ω ∈ C with this property is called a period of f .

THEOREM 3C. The exponential function ez is periodic in C with period 2πi. Furthermore, any
period ω ∈ C of ez is of the form ω = 2πki, where k ∈ Z is non-zero.

Proof. The first assertion follows easily from the observation

e2πi = cos 2π + i sin 2π = 1.
Suppose now that ω ∈ C. Clearly ez+ω = ez implies eω = 1. Write ω = α + iβ, where α, β ∈ R. Then
eα (cos β + i sin β) = 1.
Taking modulus, we conclude that eα = 1, so that α = 0. It then follows that cos β + i sin β = 1.
Equating real and imaginary parts, we conclude that cos β = 1 and sin β = 0, so that β = 2πk, where
k ∈ Z. The second assertion follows.

Consider now the mapping w = ez . By (5), we have w = ex (cos y + i sin y), where x, y ∈ R. It
follows that

|w| = ex and arg w = y + 2πk,

where k ∈ Z. Usually we make the choice arg w = y, with the restriction that −π < y ≤ π. This
restriction means that z lies on the horizontal strip
(8) R0 = {z ∈ C : −∞ < x < ∞, −π < y ≤ π}.
The restriction −π < arg w ≤ π can also be indicated on the complex w-plane by a cut along the negative
real axis. The upper edge of the cut, corresponding to arg w = π, is regarded as part of the cut w-plane.
The lower edge of the cut, corresponding to arg w = −π, is not regarded as part of the cut w-plane.

y v
z w

x u

-π
3–8 W W L Chen : Introduction to Complex Analysis

It is easy to check that the function exp : R0 → C \ {0}, defined for every z ∈ R0 by exp(z) = ez ,
is one-to-one and onto.

Remark. The region R0 is usually known as a fundamental region of the exponential function. In
fact, it is easy to see that every set of the type

(9) Rk = {z ∈ C : −∞ < x < ∞, (2k − 1)π < y ≤ (2k + 1)π},

where k ∈ Z, has this same property as R0 .

Let us return to the function exp : R0 → C \ {0}. Since it is one-to-one and onto, there is an inverse
function.

Definition. The function Log : C \ {0} → R0 , defined by Log(w) = z ∈ R0 , where exp(z) = w, is

called the principal logarithmic function.

Suppose that z = x + iy and w = u + iv, where x, y, u, v ∈ R. Suppose further that we impose the
restriction −π < y ≤ π. If w = exp(z), then it follows from (5) that u = ex cos y and v = ex sin y, and so

|w| = (u2 + v 2 )1/2 = ex and y = Arg(w),

where Arg(w) denotes the principal argument of w. It follows that

x = log |w| and y = Arg(w).

Hence

(10) Log(w) = log |w| + i Arg(w).

In many practical situations, we usually try to define

log w = log |w| + i arg w,

where the argument is chosen in order to make the logarithmic function continuous in its domain of
definition, if this is at all possible. The following three examples show that great care needs to be taken
in the study of such “many valued functions”.

Example 3.5.1. Consider the logarithmic function in the disc {w : |w+2| < 1}, an open disc of radius 1
and centred at the point w = −2. Note that this disc crosses the cut on the w-plane along the negative real
axis discussed earlier. In this case, we may restrict the argument to satisfy, for example, 0 ≤ arg w < 2π.
The logarithmic function defined in this way is then continuous in the disc {w : |w + 2| < 1}.

u
-3 -1
 Chapter 3 : Complex Differentiation 3–9

Example 3.5.2. Consider the region P obtained from the w-plane by removing both the line segment
{u + iv : 0 ≤ u ≤ 1, v = 0} and the half-line {u + iv : u = 1, v > 0}, as shown below.

u
1

Suppose that we wish to define the logarithmic function to be continuous in this region P . One way to
do this is to restrict the argument to the range π < arg w ≤ 3π for any w ∈ P satisfying u ≥ 1, and to
the range 0 < arg w ≤ 2π for any w ∈ P satisfying u < 1.

Example 3.5.3. Consider the annulus {w : 1 < |w| < 2}. It is impossible to define the logarithmic
function to be continuous in this annulus. Heuristically, if one goes round the annulus once, the argument
has to change by 2π if it varies continuously. If we return to the original starting point after going round
once, the argument cannot therefore be the same.
v

u
1 2

It should now be quite clear that we cannot expect to have

Log(w1 w2 ) = Log(w1 ) + Log(w2 ),

or even

log w1 w2 = log w1 + log w2 .

Instead, we have

log w1 w2 = log w1 + log w2 + 2πik for some k ∈ Z.

Let us return to the principal logarithmic function Log : C \ {0} → R0 . Recall (10). We have

Log(z) = log |z| + i Arg(z).

Recall from real analysis that for any t ∈ R, the equation tan θ = t has a unique solution θ satisfying
−π/2 < θ < π/2. This solution is denoted by tan−1 t and satisfies

d 1
tan−1 t = .
dt 1 + t2
3–10 W W L Chen : Introduction to Complex Analysis

It is not difficult to show that if we write




 −1 x π

 − tan − if y < 0,

 y 2


 y
(11) v(x, y) = − tan−1 if x > 0,

 x






 −1 x π
 − tan + if y > 0,
y 2

then Arg(z) = v(x, y). Hence Log(z) = u(x, y) + iv(x, y), where

1
(12) u(x, y) = log(x2 + y 2 ).
2

It now follows from (12) that

∂u x ∂u y
= 2 and = 2 ,
∂x x + y2 ∂y x + y2

and from (11) that

∂v y ∂v x
=− 2 and = 2 .
∂x x + y2 ∂y x + y2

Clearly the Cauchy-Riemann equations are satisfied, and so

d ∂u ∂v x − iy 1 1
Log(z) = +i = 2 = = .
dz ∂x ∂x x + y2 x + iy z

Power functions are defined in terms of the exponential and logarithmic functions. Given z, a ∈ C,
we write z a = ea log z . Naturally, the precise value depends on the logarithmic function that is chosen,
and care again must be exercised for these “many valued functions”.

3.6. Laplace’s Equation and Harmonic Conjugates

We have shown that for any function f = u + iv, the existence of the derivative f  leads to the Cauchy-
Riemann equations. More precisely, we have

∂u ∂v ∂u ∂v
(13) = and =− .
∂x ∂y ∂y ∂x

Furthermore,

∂u ∂v
(14) f  (z) = +i .
∂x ∂x

Suppose now that the second derivative f  also exists. Then f  satisfies the Cauchy-Riemann
equations. The Cauchy-Riemann equations corresponding to the expression (14) are





∂ ∂u ∂ ∂v ∂ ∂u ∂ ∂v
(15) = and =− .
∂x ∂x ∂y ∂x ∂y ∂x ∂x ∂x
 Chapter 3 : Complex Differentiation 3–11

Substituting (13) into (15), we obtain

∂2u ∂2u ∂2v ∂2v

(16) + 2 =0 and + = 0.
∂x2 ∂y ∂x2 ∂y 2
We also obtain

∂2v ∂2v ∂2u ∂2u

= and = .
∂y∂x ∂x∂y ∂y∂x ∂x∂y

Definition. A continuous function φ(x, y) that satisfies Laplace’s equation

∂2φ ∂2φ
+ 2 =0
∂x2 ∂y
in a domain D ⊆ C is said to be harmonic in D.

We have in fact proved the following result.

THEOREM 3D. Suppose that f = u + iv, where u and v are real valued. Suppose further that f  (z)
exists in a domain D ⊆ C. Then u and v both satisfy Laplace’s equation and are harmonic in D.

Definition. Two harmonic functions u and v in a domain D ⊆ C are said to be harmonic conjugates
in D if they satisfy the Cauchy-Riemann equations.

The remainder of this chapter is devoted to a discussion on finding harmonic conjugates. We shall
illustrate the following theorem by discussing the special case when D = C.

THEOREM 3E. Suppose that a function u is real valued and harmonic in a domain D ⊆ C. Then
there exists a real valued function v which satisfies the following conditions:
(a) The functions u and v satisfy the Cauchy-Riemann equations in D.
(b) The function f = u + iv is analytic in D.
(c) The function v is harmonic in D.

Clearly, parts (b) and (c) follow from part (a). We shall now indicate a proof of part (a) in the
special case D = C, and shall omit reference to this domain.

Suppose that u is real valued and harmonic. Then we need to find a real valued function v such
that
∂u ∂v ∂u ∂v
= and =− .
∂x ∂y ∂y ∂x
Let X0 + iY0 ∈ D be chosen and fixed. Integrating the second of these with respect to x, we obtain
 X
∂u
(17) v(X, y) = − (x, y)dx + c(y),
X0 ∂y

where c(y) is some function depending at most on y. Differentiating with respect to y, we obtain
 X
∂v ∂ ∂u
(X, y) = − (x, y)dx + c (y).
∂y ∂y X0 ∂y

Clearly the first of the Cauchy-Riemann equations requires

 X
∂u ∂ ∂u
(X, y) = − (x, y)dx + c (y).
∂x ∂y X0 ∂y
3–12 W W L Chen : Introduction to Complex Analysis

Changing the order of differentiation and integration, we obtain

 X
  X
∂u ∂ ∂u  ∂2u
(X, y) = − (x, y)dx + c (y) = − (x, y)dx + c (y).
∂x X0 ∂y ∂y X0 ∂y 2

Since u is harmonic, we obtain

 X
∂u ∂2u ∂u ∂u
(X, y) = (x, y)dx + c (y) = (X, y) − (X0 , y) + c (y),
∂x X0 ∂x2 ∂x ∂x

so that
∂u
c (y) = (X0 , y).
∂x
Integrating with respect to y, we obtain
 Y
∂u
(18) c(Y ) = (X0 , y)dy + c,
Y0 ∂x

where c is an absolute constant. On the other hand, (17) can be rewritten in the form
 X
∂u
(19) v(X, Y ) = − (x, Y )dx + c(Y ).
X0 ∂y

Combining (18) and (19), we obtain

 X  Y
∂u ∂u
(20) v(X, Y ) = − (x, Y )dx + (X0 , y)dy + c.
X0 ∂y Y0 ∂x

It is easy to check that this function v satisfies the Cauchy-Riemann equations. Indeed, we have
 X  Y
∂ ∂ ∂u ∂ ∂u ∂u
v(X, Y ) = − (x, Y )dx + (X0 , y)dy = − (X, Y ).
∂X ∂X X0 ∂y ∂X Y0 ∂x ∂y

On the other hand, we have

 X  Y  X 2
∂ ∂ ∂u ∂ ∂u ∂ u ∂u
v(X, Y ) = − (x, Y )dx + (X0 , y)dy = − 2
(x, Y )dx + (X0 , Y )
∂Y ∂Y X0 ∂y ∂Y Y0 ∂x X0 ∂y ∂x
 X 2
∂ u ∂u ∂u ∂u ∂u ∂u
= 2
(x, Y )dx + (X0 , Y ) = (X, Y ) − (X0 , Y ) + (X0 , Y ) = (X, Y ).
X0 ∂x ∂x ∂x ∂x ∂x ∂x

This completes our sketched proof.

In practice, we may use the following technique. Suppose that u is a real valued harmonic function
in a domain D. Write
∂u ∂u
(21) g(z) = −i .
∂x ∂y

Then the Cauchy-Riemann equations for g are





∂ ∂u ∂ ∂u ∂ ∂u ∂ ∂u
=− and = ,
∂x ∂x ∂y ∂y ∂y ∂x ∂x ∂y

which clearly hold. It follows that g is analytic in D. Suppose now that u is the real part of an analytic
function f in D. Then f  (z) agrees with the right hand side of (21) in view of (3) and (4). Hence f  = g
 Chapter 3 : Complex Differentiation 3–13

in D. The question here, of course, is to find this function f . If we are successful, then the imaginary
part v of f is a harmonic conjugate of the harmonic function u.

Example 3.6.1. Consider the function u(x, y) = x3 − 3xy 2 . It is easily checked that

∂2u ∂2u
+ 2 = 0,
∂x2 ∂y

so that u is harmonic in C. Using X0 = Y0 = 0 in (20), we obtain

 X  Y  X  Y
∂u ∂u
v(X, Y ) = − (x, Y )dx + (0, y)dy + c = 6 xY dx − 3 y 2 dy + c = 3X 2 Y − Y 3 + c,
0 ∂y 0 ∂x 0 0

where c is any arbitrary constant. On the other hand, we can write

∂u ∂u
g(z) = −i = 3(x2 − y 2 ) + 6ixy = 3(x2 + 2ixy − y 2 ) = 3(x + iy)2 = 3z 2 .
∂x ∂y

It follows that u is the real part of an analytic function f in C such that f  (z) = g(z) for every z ∈ C.
The function f (z) = z 3 + C satisfies this requirement for any arbitrary constant C. Note that the
imaginary part of f is 3x2 y − y 3 + c, where c is the imaginary part of C.

Example 3.6.2. Consider the function u(x, y) = ex sin y. It is easily checked that

∂2u ∂2u
+ 2 = 0,
∂x2 ∂y

so that u is harmonic in C. Using X0 = Y0 = 0 in (20), we obtain

 X  Y  X  Y
∂u ∂u
v(X, Y ) = − (x, Y )dx + (0, y)dy + c = − x
e cos Y dx + sin ydy + c
0 ∂y 0 ∂x 0 0
= cos Y − eX cos Y − cos Y + 1 + c = c − eX cos Y,

where c is any arbitrary constant. On the other hand, we can write

∂u ∂u
g(z) = −i = ex sin y − iex cos y = −iex (cos y + i sin y) = −iez .
∂x ∂y

It follows that u is the real part of an analytic function f in C such that f  (z) = g(z) for every z ∈ C.
The function f (z) = C − iez satisfies this requirement for any arbitrary constant C. Note that the
imaginary part of f is c − ex cos y, where c is the imaginary part of C.

Problems for Chapter 3

1. a) Suppose that P (z) = (z − z1 )(z − z2 ) . . . (z − zk ), where z1 , z2 , . . . , zk ∈ C. Show that

P  (z) 1 1 1
= + + ... + for every z ∈ C \ {z1 , z2 , . . . , zk }.
P (z) z − z1 z − z2 z − zk

b) Suppose further that Rezj < 0 for every j = 1, . . . , k, and that Rez ≥ 0. Show in this case that
Re(z − zj )−1 > 0 for every j = 1, . . . , k, and deduce that P  (z) = 0.
[Remark: Polynomials all of whose roots have negative real parts are called Hurwitz polynomials.
We have shown here that the derivative of a non-constant Hurwitz polynomial is also a Hurwitz
polynomial.]
3–14 W W L Chen : Introduction to Complex Analysis

2. For each of the following functions f (z), determine whether the Cauchy-Riemann equations are
satisfied:
a) f (z) = x2 − y 2 − 2ixy b) f (z) = log(x2 + y 2 ) + 2i cot−1 (x/y)
c) f (z) = x3 − 3y 2 + 2x + i(3x2 y − y 3 + 2y) d) f (z) = log(x2 − y 2 ) + 2i tan−1 (y/x)

3. Show that a real valued analytic function is constant.

4. We are required to define an analytic function f (z) such that f (x + iy) = ex f (iy) for every x, y ∈ R
and f (0) = 1. Suppose that for every y ∈ R, we write f (iy) = c(y) + is(y), where c(y), s(y) ∈ R for
every y ∈ R.
a) Show by the Cauchy-Riemann equations that c (y) = −s(y) and s (y) = c(y) for every y ∈ R.
b) For every y ∈ R, write g(y) = (c(y) − cos y)2 + (s(y) − sin y)2 . Show that g  (y) = 0 for every
y ∈ R. Deduce that g(y) = 0 for every y ∈ R.
c) Comment on the above.

5. a) Suppose that P (z) = a0 + a1 z + a2 z 2 + . . . + an z n , where a0 , a1 , a2 , . . . , an ∈ C are constants.

Show that for every k = 0, 1, . . . , n, we have

P (k) (0)
ak = .
k!
b) Apply the result to the polynomial (1 + z)n = c0 + c1 z + c2 z 2 + . . . + cn z n and show that for
every k = 0, 1, . . . , n, we have
n!
ck = .
k!(n − k)!

6. a) Show that for every z ∈ C, we have eiz = cos z + i sin z.

b) Show that for every z, w ∈ C, we have

cos(z + w) + i sin(z + w) = (cos z + i sin z)(cos w + i sin w)

and

cos(z + w) − i sin(z + w) = (cos z − i sin z)(cos w − i sin w).

c) Express sin(z + w) and cos(z + w) in terms of sin z, sin w, cos z and cos w.

7. Suppose that a1 , a2 , . . . , an ∈ C are distinct, and consider the polynomial

Q(z) = (z − z1 )(z − z2 ) . . . (z − zn ).

Suppose further that P (z) is a polynomial of degree less than n. Follow the steps below to show
that there exist a1 , a2 , . . . , an ∈ C such that
P (z) a1 a2 an
= + + ... + .
Q(z) z − z1 z − z2 z − zn

a) We shall first of all show that the expression above is possible by multiplying it by Q(z) and
then determining a1 , a2 , . . . , an so that the resulting equation between polynomials of degree
less than n holds when z = z1 , z2 , . . . , zn .
[Hint: Recall Problem 1 in Chapter 1.]
b) Show that for every k = 1, . . . , n, we have

P (z) P (zk )
ak = lim (z − zk ) =  .
z→zk Q(z) Q (zk )

[Hint: Note that Q(zk ) = 0 for every k = 1, . . . , n.]

 Chapter 3 : Complex Differentiation 3–15

8. Suppose that a ∈ C is non-zero. Show that for any fixed choice of value for log a, the function
f (z) = az = ez log a satisfies f  (z) = f (z) log a.

9. For each expression below, compute all possible values and plot their positions in the complex plane:
a) log(−i) b) log(1 + i)
c) (−i)−i d) i2
e) 2πi f) (1 + i)i (1 + i)−i

10. For each of the following equations, find all solutions:

a) Log(z) = πi/3 b) ez = 2i
c) sin z = i d) sin z = − cos z
e) tan2 z = −1

11. For each of the functions below, determine whether the function is harmonic. If so, find also its
harmonic conjugate:
a) x2 − y 2 + y b) ex sin y
c) x − y
3 3
d) xex cos y − yex sin y
e) 3x y − y + xy
2 3
f) x4 − 6x2 y 2 + y 4 + x3 y − xy 3
g) ex −y sin 2xy
2 2

12. a) Suppose that the functions f (z) and g(z) both satisfy the Cauchy-Riemann equations at a
particular point z ∈ C. Show that the functions f (z) + g(z) and f (z)g(z) also satisfy the
Cauchy-Riemann equations at the point z.
b) Show that the constant function and the function f (z) = z both satisfy the Cauchy-Riemann
equations everywhere in C.
c) Deduce that every polynomial P (z) with complex coefficients satisfies the Cauchy-Riemann
equations everywhere in C.

13. A real valued function u(x, y) which is continuous and satisfies the inequality uxx + uyy ≥ 0 in a
region D is said to be subharmonic in D. Show that u = |f (z)|2 is subharmonic in any region where
f (z) is analytic.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 4
COMPLEX INTEGRALS

4.1. Curves in the Complex Plane

Integration of functions of a complex variable is carried out over curves in C and leads to many important
results useful in pure and applied mathematics. In this section, we give a brief introduction to curves in
C. We are interested in complex valued functions of the form

φ(t) = φ1 (t) + iφ2 (t),

where the functions φ1 and φ2 are real valued and defined on some closed interval [A, B] in R. The
functions φ1 and φ2 are called the real and imaginary parts of the function φ respectively.

Example 4.1.1. The function φ(t) = eit , defined on the interval [0, π], represents the upper half of a
circle centred at the origin and of radius 1. As t varies from 0 to π, φ(t) follows this half-circle in an
anticlockwise direction. Also φ1 (t) = cos t and φ2 (t) = sin t.

φ 2(t)
1

t
φ 1 (t)
4–2 W W L Chen : Introduction to Complex Analysis

Example 4.1.2. The function φ(t) = (4t + 1) + i(t + 1), defined on the interval [0, 1], represents a line
segment from the point 1 + i to the point 5 + 2i.

2i φ(1)

i
φ(0)

1 2 3 4 5

The functions φ1 and φ2 are real valued functions of a real variable, and we have already studied
continuity, differentiability and integrability of such functions. We can now extend these definitions to
the function φ.

We say that φ is continuous at t0 if both φ1 and φ2 are continuous at t0 . We also say that φ is
continuous in an interval if both φ1 and φ2 are continuous in the interval. It is simple to show that the
arithmetic of limits, as applied to continuity, holds.

We say that φ is differentiable at t0 if both φ1 and φ2 are differentiable at t0 , and write

φ
(t0 ) = φ
1 (t0 ) + iφ
2 (t0 ).

It is simple to show that the arithmetic of derivatives holds.

We also have the Chain rule: Suppose that f is analytic at the point z0 = φ(t0 ), and that φ is
differentiable at t0 . Then the complex valued function ψ(t) = f (φ(t)) is differentiable at t0 , and

ψ
(t0 ) = f
(z0 )φ
(t0 ).

We say that φ is integrable over the interval [A, B] if both φ1 and φ2 are integrable over [A, B], and
write

B
B
B
φ(t) dt = φ1 (t) dt + i φ2 (t) dt.
A A A

Many rules of integration for real valued functions can be carried over to this case. For example, if φ is
continuous in [A, B], then there exists a function Φ satisfying Φ
= φ, and the Fundamental theorem of
integral calculus can be generalized to

B
t
d
φ(t) dt = Φ(B) − Φ(A) and φ(τ ) dτ = φ(t).
A dt A

Definition. A complex valued function ζ : [A, B] → C is called a curve. The curve is said to be
continuous if ζ is continuous in [A, B], and differentiable if ζ is also differentiable in [A, B]. The set
ζ([A, B]) is called the trace of the curve. The point ζ(A) is called the initial point of the curve, and the
point ζ(B) is called the terminal point of the curve.

Remarks. (1) Of course, we can only have continuity and differentiability from the right at A and
from the left at B.
 Chapter 4 : Complex Integrals 4–3

(2) Usually, we do not distinguish between the curve and the function ζ, and simply refer to the
curve ζ.

Definition. A curve ζ : [A, B] → C is said to be simple if ζ(t1 ) = ζ(t2 ) whenever t1 = t2 , with the
possible exception that ζ(A) = ζ(B). A curve ζ : [A, B] → C is said to be closed if ζ(A) = ζ(B).

simple and not closed simple and closed

not simple and not closed closed and not simple

4.2. Contour Integrals

Definition. A curve ζ : [A, B] → C is said to be an arc if ζ is differentiable in [A, B] and ζ
is

continuous in [A, B].

Example 4.2.1. The unit circle is a simple closed arc, since we can described it by ζ : [0, 2π] → C,
given by ζ(t) = eit = cos t + i sin t. It is easy to check that ζ(t1 ) = ζ(t2 ) whenever t1 = t2 , the only
exception being ζ(0) = ζ(2π). Furthermore, ζ
(t) = − sin t + i cos t is continuous in [0, 2π].

Definition. Suppose that C is an arc given by the function ζ : [A, B] → C. A complex valued function
f is said to be continuous on the arc C if the function ψ(t) = f (ζ(t)) is continuous in [A, B]. In this
case, the integral of f on C is defined to be


B
(1) f (z) dz = f (ζ(t))ζ
(t) dt.
C A

Remarks. (1) Note that (1) can be obtained by the formal substitution z = ζ(t) and dz = ζ
(t) dt.

(2) If we describe the arc C in the opposite direction from t = B to t = A, this opposite arc can
be designated by −C. Since

A
B
f (ζ(t))ζ
(t) dt = − f (ζ(t))ζ
(t) dt,
B A

we have



f (z) dz = − f (z) dz.
−C C
4–4 W W L Chen : Introduction to Complex Analysis

(3) Integration of functions on arcs is a linear operation. More precisely, suppose that f and g are
continuous on the arc C. Then for any α, β ∈ C, we have




(αf (z) + βg(z)) dz = α f (z) dz + β g(z) dz.
C C C

(4) Suppose that F is analytic in a domain D and has a continuous derivative f = F
in D.

Suppose further that C, defined by ζ : [A, B] → C, is an arc lying in D, with initial point z1 and
terminal point z2 .

D z2

z1

Define ψ : [A, B] → C by ψ(t) = F (ζ(t)). Then by the Fundamental theorem of integral calculus applied
to the function ψ
(t) = F
(ζ(t))ζ
(t) = f (ζ(t))ζ
(t), we have

B
f (z) dz = f (ζ(t))ζ
(t) dt = F (ζ(B)) − F (ζ(A)) = F (z2 ) − F (z1 ).
C A

We can extend the case of one arc to the case of a finite number of arcs joined together.

Definition. Suppose that A1 < B1 = A2 < B2 = . . . = Ak < Bk are real numbers, and that
for every j = 1, . . . , k, Cj is an arc given by the function ζj : [Aj , Bj ] → C. Suppose further that
ζj (Bj ) = ζj+1 (Aj+1 ) for every j = 1, . . . , k − 1. Then

C = C1 ∪ C2 ∪ . . . ∪ Ck

is called a contour. The point ζ1 (A1 ) is called the initial point of the contour C, and the point ζk (Bk )
is called the terminal point of the contour C. A complex valued function f is said to be continuous on
the contour C if it is continuous on the arc Cj for every j = 1, . . . , k. In this case, the integral of f on
C is defined to be





f (z) dz = f (z) dz + f (z) dz + . . . + f (z) dz.
C C1 C2 Ck

The following result follows immediately from this definition and Remark (4) above.
 Chapter 4 : Complex Integrals 4–5

THEOREM 4A. Suppose that F is analytic in a domain D and has a continuous derivative f = F

in D. Suppose further that C is a contour lying in D, with initial point z1 and terminal point z2 .

D z2

z1

Then


(2) f (z) dz = F (z2 ) − F (z1 ).
C

Remarks. (1) Note that the right hand side of (2) is independent of the contour C. It follows that
under the hypotheses of Theorem 4A, we have


(3) f (z) dz = 0
C

for any closed contour C in D.

(2) Naturally, we would like to extend (3) to all analytic functions f in D and all closed contours
C in D. Note, however, the restrictive nature of the hypotheses of Theorem 4A in this case. In many
situations, no analytic functions F satisfying F
= f may be at hand. Consider, for example, the function
f (z) = cos z 2 .

Example 4.2.2. Consider the contour C given by ζ : [A, B] → C, where ζ(t) = eit . Then




1 dz 1 B
ζ
(t) 1 B
ieit B−A
= dt = dt = .
2πi C z 2πi A ζ(t) 2πi A e it 2π

Suppose that A = 0 and B = 2kπ, so that the contour “winds” round the origin k times in the
anticlockwise direction. In this case, we have


1 dz
= k.
2πi C z

Note, however, that in this case, the initial point and terminal point of the contour are the same. Yet,
(3) does not hold. Clearly the function 1/z is analytic in the domain D = {z : 1/2 < |z| < 3/2} and the
contour C lies in D. However, we cannot find an analytic function F in D such that F
(z) = 1/z in D.
The logarithmic function log z appears to be a candidate; however, it is not possible to define log z to
be continuous in this annulus D. See Example 3.5.3.

Example 4.2.3. Suppose that C is any contour in C with initial point z1 and terminal point z2 . Then



ez dz = ez2 − ez1 and cos z dz = sin z2 − sin z1 .
C C
4–6 W W L Chen : Introduction to Complex Analysis

These follow from Theorem 4A since the entire functions ez and sin z satisfy
d z d
e = ez and sin z = cos z.
dz dz

Example 4.2.4. Suppose that f is a polynomial in z with coefficients in C. Then

f (z) dz = 0
C

for any closed contour C in C. This is a special case of (3). For any such polynomial f , it is easy to find
a polynomial F in z with coefficients in C and such that F
(z) = f (z) in C.

4.3. Inequalities for Contour Integrals

Suppose that φ : [A, B] → C is continuous in [A, B]. Let



 B 
 
I= φ(t) dt .
 A 

If I = 0, then clearly



 B  B
 
(4)  φ(t) dt ≤ |φ(t)| dt.
 A  A

If I > 0, then there exists a real number θ such that

B
φ(t) dt = Ieiθ ,
A

so that

B
B
B
−iθ −iθ
(5) I= e φ(t) dt = Re[e φ(t)] dt + i Im[e−iθ φ(t)] dt.
A A A

Since I is real, the last integral on the right hand side of (5) must be 0. It follows that

B
I= Re[e−iθ φ(t)] dt.
A

On the other hand, clearly

Re[e−iθ φ(t)] ≤ |e−iθ φ(t)| = |φ(t)|

for every t ∈ [A, B], and so it follows from the theory of real integration that

B
I≤ |φ(t)| dt.
A

Hence the inequality (4) always holds.

Consider now an arc C given by the function ζ : [A, B] → C. Suppose that the function f is
continuous on C. Then it follows from (4) that

 
B 


   
B
(6)  f (z) dz  =  f (ζ(t))ζ

(t) dt ≤ |f (ζ(t))||ζ
(t)| dt.
   
C A A
 Chapter 4 : Complex Integrals 4–7

Suppose that |f (z)| ≤ M on C, where M is a real constant, then we have




  B
(7)  f (z) dz  ≤ M |ζ
(t)| dt.
 
C A

Let us investigate the integral on the right hand side of (7) more closely. If ζ(t) = x(t) + iy(t), then
    2   2
 dx dy  dx dy ds
|ζ
(t)| =  + i  = + = ,
dt dt dt dt dt

where s(t) is the length of the arc C between ζ(A) and ζ(t). It follows that the integral

B
|ζ
(t)| dt
A

is the length of the arc C. We have proved the following result.

THEOREM 4B. Suppose that a function f is continuous on a contour C. Then



 
 f (z) dz  ≤ M L,
 
C

where L is the length of the contour C and where M is a real constant such that |f (z)| ≤ M on C.

Remark. We usually write

B
f (z)| dz| = f (ζ(t))|ζ
(t)| dt.
C A

In this notation, we have

L= |dz|,
C

and (6) can be represented by




 
 f (z) dz  ≤ |f (z)| |dz|.
 
C C

4.4. Equivalent Curves

Example 4.4.1. Consider the arc C
given by the function ζ : [0, 1] → C where ζ(t) = (1 + i)t2 .
Consider also the arc C

given by the function ξ : [0, π/2] → C where ξ(τ ) = (1 + i) sin τ . Clearly
ζ(0) = ξ(0), so that the two arcs have the same initial point. Also, ζ(1) = ξ(π/2), so that the two arcs
have the same terminal point. Furthermore, ζ([0, 1]) = ξ([0, π/2]), so that the two arcs have the same
trace. Suppose that the function f is continuous on C
and C

. On the one hand, we have


1
1


f (z) dz = f (ζ(t))ζ (t) dt = f ((1 + i)t2 )2(1 + i)t dt.
C
0 0

On the other hand, we have

π/2
π/2
f (z) dz = f (ξ(τ ))ξ
(τ ) dτ = f ((1 + i) sin τ )(1 + i) cos τ dτ.
C

0 0
4–8 W W L Chen : Introduction to Complex Analysis

If we perform a formal change of variables

t2 = sin τ and 2t dt = cos τ dτ,

then we see in fact that

f (z) dz = f (z) dz.
C
C

This is not surprising, considering that basically the two arcs “are the same”.

Definition. Two curves C
and C

are said to be equivalent if they have the same trace and if



f (z) dz = f (z) dz
C
C

holds for all functions f which are continuous in a region containing this trace.

Remarks. (1) It can be shown that two simple arcs are equivalent if they have the same initial points,
the same terminal points and the same trace, with the convention that in the case of closed arcs, the
arcs must be followed in the same direction.

(2) In fact, the definition

f (z) dz = f (z) dz + f (z) dz + . . . + f (z) dz
C C1 C2 Ck

of a contour integral in terms of integrals over arcs as discussed earlier is made in the same spirit.

(3) The practical importance of these considerations is that when we consider integrals over a
simple contour, we may choose the most convenient parameterization of the given contour.

Example 4.4.2. Consider the integral

dz
,
C z
where C is a contour that avoids the origin. Suppose first of all that C is an arc ζ : [A, B] → C. Then


B

dz ζ (t)
= dt = var(log z, C).
C z A ζ(t)

Here the variation function var(log z, C) is interpreted in the following way: We choose a branch of the
logarithmic function at the initial point z1 = ζ(A) of the arc C and then let log z vary continuously as
z follows C to the terminal point z2 = ζ(B) of the arc C. In other words, the function log ζ(t) must be
continuous in [A, B]. Then we calculate log ζ(B) − log ζ(A). This result is then extended to contours
by addition. Let us interpret this geometrically. Note that log z = log |z| + i arg z. Since log |z| is single
valued, its variation around a closed contour C is 0. In this case, we have

var(log z, C) = var(i arg z, C),

and this gives a value 2πi every time the closed contour winds round the origin.

4.5. Riemann Sums

The following brief discussion of complex integrals in terms of Riemann sums will further demonstrate
the independence of the integral from the parameterization of the arcs in question. The discussion is
heuristic, as we only want to illustrate ideas.
 Chapter 4 : Complex Integrals 4–9

Suppose that C is an arc, with initial point z0 and terminal point zk . Suppose further that we
divide the arc C into subarcs C1 , . . . , Ck in the following way. The points z0 , z1 , . . . , zk are points on the
arc C, and they occur in the given order as we follow the arc C from z0 to zk . For every j = 1, . . . , k,
the subarc Cj is then the part of C between zj−1 and zj , with initial point zj−1 and terminal point zj .

For every j = 1, . . . , k, we write ∆zj = zj − zj−1 , and we let z̃j denote a point on the subarc Cj .

z3 z4
z2
z1 z k-2

z0 zk-1 zj -1
z~j
zk zj

As in real variables, we can then construct the Riemann sum


k
S= f (z̃j )∆zj .
j=1

We now consider subdivisions of the arc C which are made finer and finer by subdivision into more and
more subarcs. The precise requirement will be

k→∞ and max |∆zj | → 0.

1≤j≤k

When the subdivision becomes arbitrarily fine, the Riemann sum S has a unique limit, independent of
the manner of subdivision. This limit is the integral


f (z) dz.
C

Problems for Chapter 4

1. Consider the integral z n dz, where n ∈ Z and C is a closed contour on the complex plane.
C
a) Suppose that n ≥ 0. Use Theorem 4A to explain why the integral is equal to zero.
b) Suppose that n < −1, and that the contour C does not pass through the origin z = 0. Use
Theorem 4A to explain why the integral is equal to zero.
c) What is the value of the integral if n = −1 and C is the unit circle {z : |z| = 1}, followed in
the positive (anticlockwise) direction?

2. a) Sketch each of the arcs z = 2 + it, z = e−πit , z = e4πit and z = 1 + it + t2 for t ∈ [0, 1].
b) Using Theorem 4A if appropriate, integrate each of the functions f (z) = 4z 3 , f (z) = z and
f (z) = 1/z over each of the arcs in part (a).
4–10 W W L Chen : Introduction to Complex Analysis

3. Suppose that a function f (z) satisfies f
(z) = 0 throughout a domain D ⊆ C. Use Theorem 4A to
prove that f (z) is constant in D.

4. Suppose that C1 is the semicircle from 1 to −1 through i, followed in the positive (anticlockwise)
direction. Suppose also that C2 is the semicircle from 1 to −1 through −i, followed in the negative
(clockwise) direction. Show that





z 3 dz = z 3 dz and z dz = z dz.
C1 C2 C1 C2

Use Theorem 4A to comment on the two results.

5. Suppose that α = a + ib, where a, b ∈ R are fixed. By integrating the function eαt over an interval
[0, T ] and equating real parts, show that

T
(a2 + b2 ) eat cos bt dt = eaT (a cos bT + b sin bT ) − a.
0

6. Suppose that f (t) = f1 (t)+if2 (t) and g(t) = g1 (t)+ig2 (t) are differentiable complex valued functions
of a real variable t. Deduce the formulas

(f + g)
(t) = f
(t) + g
(t) and (f g)
(t) = f (t)g
(t) + f
(t)g(t)

from known results of these types for real valued functions f1 (t), f2 (t), g1 (t) and g2 (t).

7. Consider an arc z(t) = x(t) + iy(t), where t ∈ [A, B]. Use

dz dx dy
= +i
dt dt dt

to interpret z
(t) as the complex representation of a vector tangent to the arc at any point where
z
(t) is non-zero.

8. Suppose that ζ(t) = t2 for t ∈ [−1, 0] and ζ(t) = it2 for t ∈ [0, 1]. Show that the curve ζ : [−1, 1] → C
is an arc (although its trace has a corner).

9. Consider the curve ζ : [−1, 1] → C, given by ζ(t) = −t for t ∈ [−1, 0] and ζ(t) = t + it3 sin(1/t) for
t ∈ (0, 1].
a) Show that ζ : [−1, 1] → C is an arc.
b) Determine all points of self-intersection of this arc; in other words, find all points z ∈ C such
that there exist t1 , t2 ∈ [−1, 1] satisfying t1 = t2 and z = ζ(t1 ) = ζ(t2 ).

10. Suppose that C = {z : |z| = 1} is the unit circle, followed in the positive (anticlockwise) direction.
Evaluate each of the following integrals:




dz dz |dz|
a) b) c)
C z C |z| C z




dz dz |dz|
d) e) f)
C |z |
2 2 2
C z C z
 Chapter 4 : Complex Integrals 4–11

11. Suppose that C = {z : |z| = 1} is the unit circle, followed in the positive (anticlockwise) direction.
a) Use Theorem 4B to show that


 dz 

  ≤ 2π.
C 4 + 3z

b) By dividing the circle C into its left half and its right half and applying Theorem 4B to each
half, establish the better bound


 dz  6π

 ≤ 5 .
C 4 + 3z

12. Consider the circle C = {z : |z − 1| = 1/2}, followed in the positive (anticlockwise) direction with
initial point z = 1/2. Evaluate the integral


dz
,
C (z 2 − 1)1/2

given that the integrand is equal to the derivative of the function log(z + (z 2 − 1)1/2 ).

13. Writing f = u + iv and z = x + iy, computation suggests the identity

f (z) dz = (u(x, y) dx − v(x, y) dy) + i (u(x, y) dy + v(x, y) dx).
C C C

Suppose now that the arc C is given by ζ(t) = ξ(t) + iη(t) for t ∈ [A, B]. Show from first definition
that the identity holds.

14. Suppose that the arc C1 is given by z = ζ1 (t) for t ∈ [A, B], and that the arc C2 is given by z = ζ2 (τ )
for τ ∈ [α, β]. Suppose further that there is a differentiable function φ : [A, B] → [α, β] such that
φ(A) = α, φ(B) = β. Show that the two arcs C1 and C2 are equivalent.

15. Suppose that C denotes the ellipse x = a cos t and y = b sin t, where a, b ∈ R are positive and fixed,
and t ∈ [0, 2π], so that C is followed in the positive (anticlockwise) direction.


1 dz
a) By referring to Example 4.4.2 if necessary, explain why = 1.
2πi C z

2π
1 dt 1
b) Hence show that 2 = .
2π 0 a cos t + b sin t
2 2 2 ab
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 5
CAUCHY’S INTEGRAL THEOREM

5.1. A Restricted Case

Cauchy’s integral theorem states that in a simply connected domain, the integral of an analytic function
over a closed contour is zero. The proof of this general result is rather involved. Here we first study a
special case of the theorem in order to develop the basic properties of analytic functions.

THEOREM 5A. Suppose that a function f is analytic in a domain D. Suppose further that the
closed triangular region T lies in D, and that C denotes the boundary of T in the positive (anticlockwise)
direction.

D
C

Then


f (z) dz = 0.
C
5–2 W W L Chen : Introduction to Complex Analysis

We shall give two proofs of this result, usually known as Cauchy’s integral theorem for a triangular
path. The first of these proofs, given next, is based on an additional assumption that the derivative
f
(z) is continuous in D.

Proof of Theorem 5A. Write f (z) = u(x, y) + iv(x, y), where u and v are real valued. Since f

exists and is continuous, the Cauchy-Riemann equations

∂u ∂v ∂u ∂v
= and =−
∂x ∂y ∂y ∂x

hold, and the four partial derivatives are continuous. On the other hand, we can write





f (z) dz = (u + iv)(dx + idy) = (u dx − v dy) + i (v dx + u dy).
C C C C

Suppose that C = C1 ∪ C2 ∪ C3 , a union of the three straight directed edges.

C3

C1
C2

X1 X2

Consider the integral

u dx.
C

We can write





u dx = u dx + u dx + u dx.
C C1 C2 C3

For each of the three integrals on the right hand side, y can be represented as a linear function of x,
unless the edge is vertical, in which case the integral vanishes. Suppose that the projection of the triangle
T on the x-axis is the line segment X1 ≤ x ≤ X2 . Suppose also that the vertical line with abscissa x
intersects the triangle in h1 (x) and h2 (x), where h1 (x) ≤ h2 (x) (in the diagram, h1 (x) describes C1 and
C2 , while h2 (x) describes C3 ). Then


X2
X1
X2
u dx = u(x, h1 (x)) dx + u(x, h2 (x)) dx = − (u(x, h2 (x)) − u(x, h1 (x))) dx
C X1 X2 X1


X2
h2 (x) 

∂u ∂u
=− (x, y) dy dx = − dxdy.
X1 h1 (x) ∂y T ∂y

Note that the third equality above follows from the continuity of ∂u/∂y. Similarly



∂v
v dy = dxdy.
C T ∂x
 Chapter 5 : Cauchy’s Integral Theorem 5–3

Hence


 
∂u ∂v
(u dx − v dy) = − + dxdy = 0.
C T ∂y ∂x

We can also show that

∂v ∂u
v dx = − dxdy and u dy = dxdy,
C T ∂y C T ∂x

so that


 
∂u ∂v
(v dx + u dy) = − dxdy = 0.
C T ∂x ∂y

The result follows.

5.2. Analytic Functions in a Star Domain

In this section, we shall use Theorem 5A to establish the existence of an indefinite integral and the
Cauchy integral theorem for analytic functions in a certain class of domains.

Definition. A domain D ⊆ C is called a star domain if there exists a point z0 ∈ D such that for every
point z ∈ D, the line segment joining z and z0 also lies in D. In this case, the point z0 is called a star
centre of the domain D.

Example 5.2.1. The disc {z : |z| < 1} is a star domain. Every point in this domain is a star centre.

Example 5.2.2. The complex plane C is a star domain. Again, every point in this domain is a star
centre.

Example 5.2.3. The complex plane C with the non-negative real axis {x + iy : x ≥ 0, y = 0} deleted
is a star domain. Every point on the remaining part of the real axis is a star centre.

Example 5.2.4. The set {x + iy : |xy| < 1} is a star domain. The point 0 is the only star centre.

Example 5.2.5. The interior of the set shown below is a star domain, with a star centre z0 as shown.

z0

THEOREM 5B. Suppose that a function f is analytic in a star domain D. Then there exists a
function F , analytic in D and such that F
(z) = f (z) for every z ∈ D.
5–4 W W L Chen : Introduction to Complex Analysis

Proof. Suppose that z0 ∈ D is a star centre. For every z ∈ D, define

(1) F (z) = f (ζ) dζ,
[z0 ,z]

where, for every z1 , z2 ∈ D, [z1 , z2 ] denotes the directed line segment from z1 to z2 . Since z ∈ D, there
exists an -neighbourhood of z which is contained in D. Furthermore, for every h ∈ C satisfying |h| < ,
the point z + h lies in this -neighbourhood of z. It follows that the closed triangular region with vertices
z0 , z and z + h lies in D.

z z+h
D

z0

By Theorem 5A, we have

f (ζ) dζ + f (ζ) dζ + f (ζ) dζ = 0.
[z0 ,z] [z,z+h] [z+h,z0 ]

In other words,




f (ζ) dζ − f (ζ) dζ = f (ζ) dζ.
[z0 ,z+h] [z0 ,z] [z,z+h]

It follows from (1) that

F (z + h) − F (z) = f (ζ) dζ.
[z,z+h]

If h = 0, then


F (z + h) − F (z) 1
(2) − f (z) = (f (ζ) − f (z)) dζ.
h h [z,z+h]

Since the function f is continuous at z, it follows that given any  > 0, there exists δ > 0 such that
|f (ζ) − f (z)| <  whenever |ζ − z| < δ. This means that if |h| < δ, then |f (ζ) − f (z)| <  holds for every
ζ ∈ [z, z + h]. Theorem 4B now gives


 
 
(3)  (f (ζ) − f (z)) dζ  ≤ |h|.
 [z,z+h] 

Combining (2) and (3), we have

 
 F (z + h) − F (z) 
 − f (z) ≤ .
 h
 Chapter 5 : Cauchy’s Integral Theorem 5–5

This gives

F (z + h) − F (z)
lim = f (z),
h→0 h
and completes the proof of the theorem.

If we examine our proof carefully, then it is not difficult to see that we have in fact established the
following result.

THEOREM 5C. Suppose that a function f is continuous in a star domain D. Suppose further that


f (z) dz = 0
C

for every closed triangular contour C lying in D. Then there exists a function F , analytic in D and
such that F
(z) = f (z) for every z ∈ D.

We can also deduce the Cauchy integral theorem for a star domain.

THEOREM 5D. Suppose that a function f is analytic in a star domain D. Suppose further that C
is a closed contour lying in D. Then


f (z) dz = 0.
C

Proof. By Theorem 5B, there exists a function F , analytic in D and such that F
(z) = f (z) for every
z ∈ D. The result now follows from Remark (1) immediately after Theorem 4A.

Example 5.2.6. Consider the contour integral

ez + sin z
dz,
|z|=3 z − 16
2

where the contour of integration is the circle centred at 0 and with radius 3, followed in the positive
(anticlockwise) direction. Note that the function in question is analytic in the disc D = {z : |z| < 4},
clearly a star domain. It follows from Theorem 5D that the integral is 0.

Example 5.2.7. Suppose that 0 < r < R. Consider the contour integral


R+z
dz,
|z|=r (R − z)z

where the contour of integration is the circle centred at 0 and with radius r, followed in the positive
(anticlockwise) direction. For every z ∈ C, note that using partial fractions, we have

R+z 1 2
= + .
(R − z)z z R−z

It follows that




R+z 1 2
dz = dz + dz.
|z|=r (R − z)z |z|=r z |z|=r R−z

Next, note that the function

2
R−z
5–6 W W L Chen : Introduction to Complex Analysis

is analytic in the star domain D = {z : |z| < R}. It follows from Theorem 5D that the last integral is 0,
so that



R+z 1
(4) dz = dz = 2πi,
|z|=r (R − z)z |z|=r z

in view of Example 4.4.2. On the other hand, the contour can be described by z = reiθ , where θ ∈ [0, 2π].
This formal substitution leads to the expression dz = ireiθ dθ = iz dθ and


2π
R+z R + reiθ
dz = i dθ.
|z|=r (R − z)z 0 R − reiθ

Next, note that

R + reiθ (R + reiθ )(R − re−iθ ) R2 − r2 + 2iRr sin θ

= = ,
R − reiθ (R − reiθ )(R − re−iθ ) R2 − 2Rr cos θ + r2

so that



R+z 2π
R2 − r2 + 2iRr sin θ
(5) dz = i dθ.
|z|=r (R − z)z 0 R2 − 2Rr cos θ + r2

Combining (4) and (5) and equating imaginary parts, we obtain

1 2π
R2 − r 2
dθ = 1.
2π 0 R2 − 2Rr cos θ + r2

5.3. Nested Triangles

In this section, we shall give a second proof of Theorem 5A, without the additional assumption that
the derivative f
(z) is continuous in D. This proof is based on the following well-known result in real
analysis: Suppose that

a1 ≤ a2 ≤ a3 ≤ . . . and b 1 ≥ b2 ≥ b3 ≥ . . . .

Suppose further that ak ≤ bk for every k ∈ N, and that bk − ak → 0 as k → ∞. Then there exists a
unique number  ∈ R such that ak →  and bk →  as k → ∞. This is a special case of the Cantor
intersection theorem. In other words, if the intervals

[a1 , b1 ] ⊇ [a2 , b2 ] ⊇ [a3 , b3 ] ⊇ . . .

are nested, so that each contains all subsequent ones, and if their lengths decrease to 0, then the intervals
collapse to a unique point.

We shall now prove Theorem 5A by the method of bisection.

Suppose that a function f is analytic in a domain D. Suppose further that the closed triangular
region T lies in D, and that C denotes the boundary of T in the positive (anticlockwise) direction. Write


I(T ) = f (z) dz.
C
 Chapter 5 : Cauchy’s Integral Theorem 5–7

We now divide T into four triangular regions by joining the midpoints of the three sides of T as shown
in the diagram.

T(3)
T(1)
T(2)
T
T(4)

Suppose that the four triangular regions so obtained are denoted by T (j) , where j = 1, 2, 3, 4, with
boundaries C (j) in the positive (anticlockwise) direction. Then since integrals over the common sides
cancel each other, we have

I(T ) = I(T (1) ) + I(T (2) ) + I(T (3) ) + I(T (4) ),

where for j = 1, 2, 3, 4,


I(T (j) ) = f (z) dz.
C (j)

Since the maximum is never less than the average, at least one of these four triangular regions T (j) must
satisfy

1
(6) |I(T (j) )| ≥ |I(T )|.
4
We denote this triangular region by T1 , with the convention that if more than one of the four triangular
regions T (j) satisfies (6), then we choose one under some fixed rule. This process can now be repeated
indefinitely, so that we obtain a sequence of nested triangles

T = T0 ⊇ T1 ⊇ T2 ⊇ T3 ⊇ . . . ⊇ Tk ⊇ . . .

with the property

1
|I(Tk )| ≥ |I(Tk−1 )|,
4
so that

(7) |I(Tk )| ≥ 4−k |I(T )|.

Note now that the sequence of nested triangular regions must collapse to a point z ∗ ∈ D. Suppose now
that  > 0 is chosen. Since D is open and the function f is analytic at z ∗ , there exists a δ-neighbourhood
{z : |z − z ∗ | < δ} of z ∗ , contained in D and such that
 
 f (z) − f (z ∗ ) 
(8)  − f
∗ 
(z )
 z − z∗ <

whenever |z − z ∗ | < δ. Furthermore, we can choose k so large that

(9) Tk ⊂ {z : |z − z ∗ | < δ}.

5–8 W W L Chen : Introduction to Complex Analysis

Note that since

dz = 0 and z dz = 0,
Ck Ck

we have



I(Tk ) = f (z) dz = (f (z) − f (z ∗ ) − (z − z ∗ )f
(z ∗ )) dz.
Ck Ck

In view of (8) and (9), we have

|f (z) − f (z ∗ ) − (z − z ∗ )f
(z ∗ )| ≤ |z − z ∗ | ≤ dk ,

where dk denotes the diameter of Tk . It follows from Theorem 4B that

(10) |I(Tk )| ≤ dk Lk ,

where Lk denotes the perimeter of Tk . Observe now that

(11) dk = 2−k d and Lk = 2−k L,

where d and L denote respectively the diameter and perimeter of T . Combining (7), (10) and (11), we
obtain

|I(T )| ≤ dL.

Since  > 0 is arbitrary, we must have I(T ) = 0. This completes the proof of Theorem 5A.

5.4. Further Examples

Example 5.4.1. Suppose that C is any contour. For any z ∈ C not lying on C, consider the integral


dζ
I(z) = .
C ζ −z

We shall show that the function I(z) is continuous at z. Since z ∈ C, there exists  > 0 such that the
-neighbourhood of z does not meet C. Suppose that h ∈ C satisfies |h| < /2.

ε C
z z+h

Then

 

1 1 dζ
I(z + h) − I(z) = − dζ = h .
C ζ −z−h ζ −z C (ζ − z − h)(ζ − z)
 Chapter 5 : Cauchy’s Integral Theorem 5–9

Note next that for any ζ ∈ C, we have


|ζ − z| >  and |ζ − z − h| > ,
2

and so it follows from Theorem 4B that

2L|h|
|I(z + h) − I(z)| ≤ ,
2

where L is the length of C. This clearly tends to 0 as h → 0.

The final example in this chapter exhibits the possibility of defining a continuous logarithm.

Example 5.4.2. Consider the domain obtained by deleting from C the origin 0 and a half-line starting
from 0. This is a star domain in which the function 1/z has a continuous derivative. Suppose that C is
a closed contour that does not meet this half-line.

Then


dζ
= 0.
C ζ

Furthermore, the integral

z
dζ
z0 ζ

is independent of the path joining z0 to z in this domain, and can therefore be used to define a continuous
logarithm.

Problems for Chapter 5

1. Give an example to show that the conclusion of Theorem 5D may not hold if D is not a star domain.

2. Suppose that R > 0 is fixed. By integrating the function (R − z)−1 over the circle C = {z : |z| = r},
where 0 < r < R, and referring to Example 5.2.7, show that

2π
1 R cos θ r
dθ = 2 .
2π 0 R2 − 2Rr cos θ + r 2 R − r2
5–10 W W L Chen : Introduction to Complex Analysis

3. a) Suppose that C is the rectangle with vertices at ±b and ±b + ia, where a, b > 0. Explain why


e−z dz = 0.
2

b) Let C = C1 ∪ C2 ∪ C3 ∪ C4 , where C1 , C2 , C3 , C4 represent the four edges of C followed in the

positive (anticlockwise) direction, with initial point z = −b. Show that


a 

a
   
 e −z 2
dz  ≤ e−b2 e y2
dy and  e −z 2
dz  ≤ e−b2 2
ey dy.
   
C2 0 C4 0

c) Explain why

b
b
−(x+ia)2
e−x dx → 0
2
e dx − as b → ∞.
−b −b

Deduce that the integral

∞
e−(x+ia) dx
2

−∞

is independent of the choice of a > 0.

4. Suppose that a function f (z) is analytic in {z : |z| < R} and continuous in {z : |z| ≤ R}, where
R > 0 is fixed. Suppose further that C denotes the circle {z : |z| = R}.
a) Suppose that r < R. Explain why


2π
2π
f (z) dz = f (Re )Re i dθ −
iθ iθ
f (reiθ )reiθ i dθ.
C 0 0

b) The function f (z)z is continuous in {z : |z| ≤ R}, and so uniformly continuous in {z : |z| ≤ R}.
This implies that given any  > 0, there exists δ > 0 such that |f (Reiθ )Reiθ − f (reiθ )reiθ | < 
whenever R − δ < r < R. Use this to show that


 
 f (z) dz  < 2π.
 
C

c) Explain why it follows that

f (z) dz = 0.
C

d) Explain also why this result does not follow directly from Theorem 5D.

5. Suppose that a function f (z) is continuous on a closed contour C. Suppose further that f (z) can be
uniformly approximated with arbitrary precision by a polynomial; in other words, given any  > 0,
there exists a polynomial P (z) such that |f (z) − P (z)| <  for every z ∈ C. Prove that


f (z) dz = 0.
C

6. Suppose that a function f (z) is analytic in {z : |z| ≤ 1}. By considering a suitable integral over the
unit circle {z : |z| = 1}, show that
 
1 
max  − f (z) ≥ 1.

|z|=1 z
 Chapter 5 : Cauchy’s Integral Theorem 5–11

7. Suppose that C is a closed contour, and that D is a domain not containing any point of C. By
noting Examples 4.4.2 and 5.4.1, show that the integral


1 dz
n(C, z0 ) =
2πi C z − z0

is independent of the choice of z0 ∈ D.

[Remark: The value n(C, z0 ) is called the winding number of the contour C round the point z0 ,
and measures the number of times the contour winds round the point z0 .]

8. Suppose that C is a contour z = r(θ)eiθ for θ ∈ [0, 2π], where r(θ) > 0 for every θ ∈ [0, 2π]. Suppose
further that r(0) = r(2π), so that C is a closed contour. Let D be the domain containing the origin
z = 0 and with boundary C.
a) Show that D is a star domain with the origin z = 0 as a star centre.
b) Suppose that z0 ∈ D ∪ C. Explain why the half line L = {λz0 : λ ∈ [1, ∞)} satisfies L ∩ C = ∅.
Show also that C \ L is a star domain with star centre z = 0.
c) Explain why


1 dz 0 if z0 ∈ D ∪ C,
=
2πi C z − z0 1 if z0 ∈ D.

[Hint: For the case z0 ∈ D, refer to Problem 7 if necessary.]

d) Suppose that P (z) is a polynomial with no roots on the contour C. By referring to Problem 1
in Chapter 3 if necessary, show that the number of roots of P (z) in D is given by


1 P
(z)
dz.
2πi C P (z)

9. Suppose that P (z) is a polynomial of degree k and with distinct roots z1 , . . . , zk . Suppose further
that C is a closed contour which does not contain any of these roots. By referring to Problem 7 if
necessary, show that


1 P
(z)
dz = n(C, z1 ) + . . . + n(C, zk ).
2πi C P (z)

10. Suppose that two star domains D1 and D2 both have the point z0 as star centre. Show that D1 ∩ D2
and D1 ∪ D2 are both star domains with star centre z0 .
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 6
CAUCHY’S INTEGRAL FORMULA

6.1. Introduction

In this chapter, we study a remarkable formula due to Cauchy and which shows that the values of an
analytic function at the interior points of a disc are determined by the values of the function on the
boundary of the disc.

THEOREM 6A. Suppose that a function f is analytic in a domain D. Suppose further that the
closed disc {z : |z − α| ≤ r} is contained in D, and that C denotes the circle {z : |z − α| = r} followed
in the positive (anticlockwise) direction.

D C
r
α

Then for every z ∈ D satisfying |z − α| < r, we have

1 f (ζ)
(1) f (z) = dζ.
2πi C ζ − z
6–2 W W L Chen : Introduction to Complex Analysis

Remark. Theorem 6A is a special case of Cauchy’s integral formula in a simply connected domain
which we shall study in Chapter 9.

Proof of Theorem 6A. Suppose that γ is a circle of radius ρ and centred at z, followed in the
positive direction. Suppose further that ρ is sufficiently small so that γ lies in the interior of C. Note
that a horizontal line through the point z intersects C at two points and intersects γ at two points and
gives rise to two line segments inside C and outside γ.

C+
C
ρ
z z
γ
α α
r

C−

The part of the two circles above this line and the two line segments give rise to a simple closed contour
C + followed in the positive direction and which can be shown to lie in a star domain lying in D but not
containing z, so that


f (ζ)
dζ = 0,
C+ ζ − z

in view of Theorem 5D. Similarly, the part of the two circles below this line and the two line segments
give rise to a simple closed contour C − followed in the positive direction and which again can be shown
to lie in a star domain lying in D but not containing z, so that


f (ζ)
dζ = 0.
C− ζ −z

It is easily seen that

f (ζ) f (ζ) f (ζ) f (ζ)
dζ − dζ = dζ + dζ,
C ζ −z γ ζ −z C+ ζ −z C− ζ −z

so that



f (ζ) f (ζ)
dζ = dζ.
C ζ −z γ ζ −z

We can write




f (ζ) dζ f (ζ) − f (z)
(2) dζ = f (z) + dζ.
C ζ −z γ ζ −z γ ζ −z
 Chapter 6 : Cauchy’s Integral Formula 6–3

The first integral on the right hand side of (2) is studied in a similar way as in Example 4.2.2, and we
have


dζ
(3) = 2πi.
γ ζ −z

On the other hand, note that f is continuous at z. It follows that given any  > 0, there exists δ > 0
such that |f (ζ) − f (z)| <  whenever |ζ − z| < δ. If we choose ρ so that ρ < δ, then
 
 f (ζ) − f (z) 
 ≤ 
 ζ −z  ρ

for every ζ ∈ γ, and so it follows from Theorem 4B that



 f (ζ) − f (z) 
(4)  dζ  ≤ 2π.
 ζ −z 
γ

Combining (2)–(4), we obtain



 f (ζ) 
 dζ − 2πif (z) ≤ 2π.
 
C ζ −z

The result follows immediately, since  is arbitrary.

6.2. Derivatives

An important consequence of Cauchy’s integral formula is that we can show that an analytic function
possesses derivatives of all orders. This is a rather remarkable result, and much nicer than in real
analysis. We shall establish this result by a number of steps.

THEOREM 6B. Suppose that a function f is analytic in a domain D. Then the derivative f
is
analytic in D.

Remark. Recall that in our first proof of Theorem 5A, we use the additional assumption that f
is
continuous in D. In view of Theorem 6B, it appears that this extra assumption is superfluous. However,
our proof of Theorem 6B below will depend on Theorem 6A, whose proof uses Theorem 5D, which
follows somewhat from Theorem 5A. Hence we cannot reasonably use our first proof of Theorem 5A
without running into the danger of a “circular argument” of deducing two results from each other and
possibly establishing neither. Note, however, that our second proof of Theorem 5A in Section 5.3 saves
us from this dubious distinction.

THEOREM 6C. Suppose that a function f is analytic in a domain D. Then the derivative f (n)
exists for every n ∈ N and is analytic in D.

THEOREM 6D. Suppose that a function f is analytic in a domain D. Then, in the notation of
Theorem 6A, we have


(n) n! f (ζ)
(5) f (z) = dζ
2πi C (ζ − z)n+1

for every n ∈ N.
6–4 W W L Chen : Introduction to Complex Analysis

Proof of Theorem 6B. Note that every z ∈ D is contained inside some circle C with α as centre,
and so (1) is valid. Suppose that h ∈ C has sufficiently small modulus so that z + h, as well as z, lies
inside the circle C.

D z+h δ
z
α
C

Then by (1), we have

 

f (z + h) − f (z) 1 f (ζ) f (ζ) 1 f (ζ)
= − dζ = dζ
h 2πih C ζ − z − h ζ − z 2πi C (ζ − z − h)(ζ − z)



1 f (ζ) h f (ζ)
= dζ + dζ.
2πi C (ζ − z)2 2πi C (ζ − z)2 (ζ − z − h)

Since z is inside the circle C, the number

δ = min |ζ − z| > 0.
ζ∈C

If |h| < δ/2, then for every ζ ∈ C, we have

δ δ
|ζ − z − h| ≥ |ζ − z| − |h| > δ − = .
2 2
On the other hand, the circle C is a closed and bounded set. It follows that there exists some real
constant M such that |f (ζ)| ≤ M for every ζ ∈ C. Recall that the circle C has radius r. It now follows
from Theorem 4B that


 h f (ζ)  |h| 2M 2M r|h|
 dζ  ≤ 2πr = →0
 2πi (ζ − z)2 (ζ − z − h) 2π δ 3 δ3
C

as h → 0. This establishes the existence of f
in D and (5) for n = 1. We now repeat the argument,
starting with (5) with n = 1. This establishes the existence of f

in D, and so the analyticity of f
in
D.

Proof of Theorem 6C. Suppose that f (n) is analytic in D. Applying Theorem 6B to the function
f (n) , we conclude that f (n+1) is analytic in D. The conclusion now follows by induction.

Proof of Theorem 6D. It follows from Theorem 6C that f (n) is analytic in D. Applying Theorem
6A to the function f (n) , we have


1 f (n) (ζ)
f (n) (z) = dζ.
2πi C ζ −z

Integrating this by parts n times gives (5).

 Chapter 6 : Cauchy’s Integral Formula 6–5

6.3. Further Consequences

THEOREM 6E. (CAUCHY’S ESTIMATE) Suppose that a function f is analytic in a domain D.

Suppose further that the closed disc {z : |z − α| ≤ r} is contained in D, and that there exists a positive
constant M such that |f (z)| ≤ M in this disc. Then

n!M
|f (n) (α)| ≤ .
rn

Proof. It follows from Theorem 6D that

n! f (ζ)
f (n) (α) = dζ.
2πi C (ζ − α)n+1

Note now that for every ζ ∈ C, we have |ζ − α| = r, so that

 
 f (ζ)  M
 
 (ζ − α)n+1  ≤ rn+1 .

It now follows from Theorem 4B that

n! M n!M
|f (n) (α)| ≤ 2πr = n
2π rn+1 r

as required.

THEOREM 6F. (LIOUVILLE’S THEOREM) An entire and non-constant function f cannot be

bounded in C.

Proof. Suppose on the contrary that f is bounded. Then there exists a positive constant M such
that |f (z)| ≤ M for every z ∈ C. For every α ∈ C, Cauchy’s estimate for n = 1 gives

M
|f
(α)| ≤ .
r

This inequality is valid for every r > 0 since the closed disc {z : |z − α| ≤ r} is clearly contained in C.
It follows that we must have f
(α) = 0 for every α ∈ C. Hence it follows from Theorem 4A that

z
f (z) − f (0) = f
(ζ) dζ = 0,
0

so that f is constant, a contradiction.

THEOREM 6G. (MORERA’S THEOREM) Suppose that f is continuous in a domain D. Suppose

further that


f (z) dz = 0
C

holds for every closed triangular contour C which together with its interior lies in D. Then f is analytic
in D.

Proof. Suppose that z ∈ D. Then there exists an -neighbourhood Dz of z lying entirely in D.

Clearly Dz is a star domain. It follows from Theorem 5C that there exists a function F , analytic in Dz
and such that F
= f in Dz . By Theorem 6B, f is analytic in Dz , and so analytic at z. Since z ∈ D is
arbitrary, the result follows immediately.

6–6 W W L Chen : Introduction to Complex Analysis

Example 6.3.1. We can now prove the Fundamental theorem of algebra, that every non-constant
polynomial P (z) has at least one root. It is easily checked that every such non-constant polynomial
satisfies |P (z)| → ∞ as |z| → ∞. Hence the function 1/P (z) is bounded outside some circle {z : |z| = r}.
Suppose that P (z) does not vanish. Then 1/P (z) is an entire function. Hence it is continuous and so
bounded in the closed set {z : |z| ≤ r}. It follows that it is bounded in C. By Liouville’s theorem, it
must be constant, a contradiction.

Problems for Chapter 6

1. Suppose that a function f = u + iv is analytic in a region D. Show that all partial derivatives of u
and v are continuous in D. Show also that uv is harmonic in D.

2. Suppose that f (z) is an entire function. Suppose further that there exist M ∈ R and m ∈ N such
that |f (z)| ≤ M |z|m whenever |z| is large.
a) Use Cauchy’s estimate to show that f (n) (0) = 0 for every integer n > m.
b) Deduce that f (z) is a polynomial of degree at most m.

3. Suppose that a function f (z) is analytic in the closed disc {z : |z| ≤ R}, where R > 0 is fixed.
a) Prove Gauss’s mean value theorem, that

2π
1
f (0) = f (Reiθ ) dθ.
2π 0
b) Prove also that for every n ∈ N ∪ {0}, we have

2π
n!
|f (n) (0)| ≤ |f (Reiθ )| dθ.
2πRn 0

c) Suppose that L is the length of the image of the circle {z : |z| = R} under f , so that


2π
L= |f
(z)||dz| = R |f
(Reiθ )| dθ.
C 0

Show that L ≥ 2πR|f
(0)|.

d) By first expressing the integral in polar coordinates, show that


f (x + iy) dxdy = πR2 f (0).
{z:|z|≤R}

4. Suppose that f (z) is an entire function. Suppose further that there exists M ∈ R such that
Ref (z) ≤ M for every z ∈ C. By applying Liouville’s theorem to the function ef (z) , show that f (z)
is constant.

5. Suppose that f (z) is an entire function, and that g(R) → 0 as R → ∞. Suppose further that for
all large R, the inequality |f (z)| ≤ Rg(R) holds whenever |z| = R. By proceeding along the lines of
the proof of Liouville’s theorem, show that f (z) is constant.

6. Suppose that a function f (ζ) is continuous on a contour C. Show that the function



f (ζ)
f (ζ)
F (z) = dζ satisfies F (z) = dζ
C ζ −z C (ζ − z)
2

for every z ∈ C, so that F (z) is analytic off the contour C.

7. Suppose that a function f (ζ) is continuous on a contour C. Consider again the function F (z) in
Problem 6. Use Morera’s theorem to show that F (z) is analytic off the contour C.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 7
TAYLOR SERIES, UNIQUENESS
AND THE MAXIMUM PRINCIPLE

7.1. Remarks on Series

The purpose of this chapter is to show that every analytic function can be represented by a Taylor series,
and to use the Taylor series to study further properties of such functions.

Here we do not propose to have a systematic study of series. Such a study is postponed until
Chapter 16. In this section, we shall make a brief review of standard terminology.

We are concerned with power series of the form

(1) a0 + a1 (z − z0 ) + a2 (z − z0 )2 + . . . ,

where a0 , a1 , a2 , . . . ∈ C and z0 ∈ C are fixed, and where z belongs to some region in the complex plane
C. For every N ∈ N, the N -th partial sum of the series (1) is defined by

(2) sN (z) = a0 + a1 (z − z0 ) + a2 (z − z0 )2 + . . . + aN (z − z0 )N .

We are interested in the sequence sN (z) of partial sums.

Suppose that a function f (z) and a sequence of functions sN (z) are defined in a region G in the
complex plane C. The sequence sN (z) is said to converge uniformly to f (z) in G if, given any
> 0,
there exists N0 = N0 (
) such that

|sN (z) − f (z)| <

7–2 W W L Chen : Introduction to Complex Analysis

for every N > N0 and every z ∈ G. Note here that the notion of uniformity implies the independence of
N0 from the choice of z in G. We also write

f (z) = lim sN (z)

N →∞

uniformly in G.

7.2. Taylor Series

In particular, if sN (z) is given by (2), then we write

∞


f (z) = an (z − z0 )n
n=0

uniformly in G.

THEOREM 7A. (TAYLOR’S THEOREM) Suppose that a function f is analytic in the domain
{z : |z − z0 | < R}, where z0 ∈ C and R > 0 are fixed. Suppose further that 0 ≤ r < R. Then

∞
f (n) (z0 )
(3) f (z) = (z − z0 )n
n=0
n!

uniformly in {z : |z − z0 | ≤ r}.

Definition. The series (3) is called the Taylor series of the function f at z0 .

Theorem 7A follows easily from the following special case.

THEOREM 7B. Suppose that a function g is analytic in the domain {z : |z| < R}, where R > 0 is
fixed. Suppose further that 0 ≤ r < R. Then

∞
g (n) (0) n
(4) g(z) = z
n=0
n!

uniformly in {z : |z| ≤ r}.

We shall show that Theorem 7A follows from Theorem 7B. Suppose that |z − z0 | < R. If we write
ζ = z − z0 , then |ζ| < R. We now use the substitution f (z) = g(ζ). Suppose that f is analytic in the
region {z : |z − z0 | < R}. Then clearly g is analytic in the region {ζ : |ζ| < R}. It then follows from
Theorem 7B that

∞ ∞
g (n) (0) n
f (n) (z0 )
f (z) = g(ζ) = ζ = (z − z0 )n
n=0
n! n=0
n!

uniformly in {ζ : |ζ| ≤ r}, and so uniformly in {z : |z − z0 | ≤ r}. It remains to prove Theorem 7B.

Proof of Theorem 7B. Let the real number ρ be chosen to satisfy r < ρ < R, and let C denote the
circle {ζ : |ζ| = ρ}, followed in the positive (anticlockwise) direction. By Theorem 6A, we have

1 g(ζ)
g(z) = dζ
2πi C ζ −z
 Chapter 7 : Taylor Series, Uniqueness and the Maximum Principle 7–3

for every z satisfying |z| ≤ r. Since (see Remark (1) below)

1 1 z z2 z n−1 zn 1
(5) = + 2 + 3 + ... + n + n ,
ζ −z ζ ζ ζ ζ ζ ζ −z

we have
 
1 g(ζ) z n−1 g(ζ)
(6) g(z) = dζ + . . . + dζ + z n gn (z),
2πi C ζ 2πi C ζn

where

1 g(ζ)
(7) gn (z) = dζ.
2πi C − z)
ζ n (ζ

By Theorems 6A and 6D, we have


1 g(ζ) g (k) (0)
(8) dζ =
2πi C ζ k+1 k!

for every k = 0, . . . , n − 1, so that (6) becomes

g (n−1) (0) n−1

(9) g(z) = g(0) + g  (0)z + . . . + z + z n gn (z).
(n − 1)!

To complete the proof of Theorem 7B, it suffices to show that

|z n gn (z)| → 0

uniformly in {z : |z| ≤ r} as n → ∞. Clearly, the circle C = {ζ : |ζ| = ρ} is closed and bounded,

and the function g is continuous on C. It follows that there exists a positive real constant M such that
|g(ζ)| ≤ M for every ζ ∈ C. Hence for every ζ ∈ C and every |z| ≤ r, we have
 
 g(ζ)  M
 
 ζ n (ζ − z)  ≤ ρn (ρ − r) .

It follows from Theorem 4B that

 
 g(ζ)  M
 dζ  ≤ n 2πρ,
 ζ n (ζ − z) ρ (ρ − r)
C

and so
 n
rn M Mρ r
|z gn (z)| ≤
n
2πρ = .
2π ρ (ρ − r)
n ρ−r ρ

Since r < ρ, the right hand side clearly converges to 0 as n → ∞ independently of the choice of z in the
set {z : |z| ≤ r}.

Remarks. (1) To derive the identity (5), note that if w = 1, then the identity

1 wn
= 1 + w + w2 + w3 + . . . + wn−1 +
1−w 1−w

is easily verified. Now substitute w = z/ζ and then divide by ζ to obtain (5).

(2) It is easily seen that the function gn (z) in (7) is analytic in the domain {z : |z| < R}. If z = 0,
then the analyticity follows immediately from (9). To show that gn (z) is analytic at 0, note that the
7–4 W W L Chen : Introduction to Complex Analysis

function g(ζ)/ζ n is continuous on the circle C, and the result follows from Problem 6 in Chapter 6. Note
also that (7) and (8) for k = n give

g (n) (0)
gn (0) = .
n!
These observations, combined with (9), immediately lead to the following finite version of Taylor’s
theorem.

THEOREM 7C. Under the hypotheses of Theorem 7A, we have

f (n−1) (z0 )
f (z) = f (z0 ) + f  (z0 )(z − z0 ) + . . . + (z − z0 )n−1 + fn (z)(z − z0 )n ,
(n − 1)!

where fn (z) is analytic in the domain {z : |z − z0 | < R} and

f (n) (z0 )
fn (z0 ) = .
n!

A very good reason for studying Taylor series is their polynomial-like behaviour, although great
care needs to be exercised. An example is the following result.

THEOREM 7D. Suppose that a function f is analytic in the domain {z : |z − z0 | < R}, where z0 ∈ C
and R > 0 are fixed. Then the series obtained through term-by-term differentiation of the Taylor series
(3) of f (z) converges uniformly to f  (z) in any closed disc {z : |z − z0 | ≤ r}, where r < R. Furthermore,
the differentiated series is the Taylor series of f  (z).

Proof. Since f is analytic in the domain {z : |z − z0 | < R}, it follows from Theorem 6B that f  is
also analytic in {z : |z − z0 | < R}. By Theorem 7A, the function f  has its Taylor series

∞
∞
f (n+1) (z0 ) f (n) (z0 )
f  (z) = (z − z0 )n = n(z − z0 )n−1 ,
n=0
n! n=1
n!

the same series as obtained through term-by-term differentiation of the Taylor series of f . The uniform
convergence of this series to f  (z) in {z : |z − z0 | ≤ r} follows from Theorem 7A.

Example 7.2.1. The function g(z) = ez is entire. Also, for every n ∈ N, we have g (n) (z) = ez , so that
g (n) (0) = 1. It now follows from Theorem 7B that

∞
z zn z2 z3 z4
e = =1+z+ + + + ...
n=0
n! 2! 3! 4!

uniformly in {z : |z| ≤ r} for every r > 0.

Example 7.2.2. The function g(z) = sin z is entire. Also, for every n ∈ N, we have

 cos z if n = 1, 5, 9, . . .,

− sin z if n = 2, 6, 10, . . .,
g (n) (z) =

 − cos z if n = 3, 7, 11, . . .,
sin z if n = 4, 8, 12, . . .,
so that

 1 if n = 1, 5, 9, . . .,

(n) 0 if n = 2, 6, 10, . . .,
g (0) =
 −1
 if n = 3, 7, 11, . . .,
0 if n = 4, 8, 12, . . ..
 Chapter 7 : Taylor Series, Uniqueness and the Maximum Principle 7–5

It now follows from Theorem 7B that

z3 z5 z7
sin z = z − + − + ...
3! 5! 7!

uniformly in {z : |z| ≤ r} for every r > 0. Applying Theorem 7D and differentiating term-by-term, we
obtain

z2 z4 z6
cos z = 1 − + − + ...
2! 4! 6!

uniformly in {z : |z| ≤ r} for every r > 0.

Example 7.2.3. Consider the function

sin z
h(z) = .
z
From Example 7.2.2 and Theorem 7C, we have

sin z = z + z 2 g2 (z),

where the function g2 (z) is entire. It follows that for z = 0, we have

sin z
= 1 + zg2 (z),
z

and so h(z) → 1 as z → 0. Note that the function h(z) is analytic at any z = 0. If we define h(0) = 1,
then the function h is continuous at 0. We say that h has a removable singularity at 0.

Example 7.2.4. Consider the function

1 − cos z
k(z) = .
z2
From Example 7.2.2 and Theorem 7C, we have

z2
cos z = 1 − + z 3 g3 (z),
2

where the function g3 (z) is entire. It follows that for z = 0, we have

1 − cos z 1
= − zg3 (z),
z2 2

and so k(z) → 1/2 as z → 0. Note that the function k(z) is analytic at any z = 0. If we define
k(0) = 1/2, then the function k is continuous at 0.

7.3. Uniqueness

Recall the Cauchy integral formula as given by Theorem 6A. To determine the value of an analytic
function at interior points of a disc, we need the values of the function on the boundary of the disc.

On the other hand, if we know the values of f (z0 ), f  (z0 ), f  (z0 ), . . . of an analytic function f at a
point z0 in a domain D, then the Taylor series determines f (z) in some disc {z : |z − z0 | < R} centred at
z0 . It follows that if f (z) is known in some infinitely differentiable short arc in the disc {z : |z − z0 | < R},
7–6 W W L Chen : Introduction to Complex Analysis

then f (z) is uniquely determined in the disc {z : |z − z0 | < R}, since the derivatives of f (z) can be
calculated by differentiation of f (z) on this arc.

The purpose of this section is to extend this second observation to show that an analytic function
f is uniquely determined in a domain D, and not just a disc centred at z0 , by the values of f (n) (z0 ) at
some z0 in D. We shall also show that an analytic function f is determined in a domain D by the values
on a short continuous curve C in D.

THEOREM 7E. Suppose that two functions f and g are analytic in a domain D. Suppose further
that z0 ∈ D, and that f (z) = g(z) in some
-neighbourhood of z0 . Then f (z) = g(z) for every z ∈ D.

Remark. The proof is based on the following argument: A domain D is open and connected, and
therefore cannot be written as a disjoint union of two non-empty open sets.

Proof of Theorem 7E. For every z ∈ D, write h(z) = f (z) − g(z). Then the function h is analytic
in D. Let

S1 = {z1 ∈ D : h(z) = 0 in some neighbourhood of z1 } and S 2 = D \ S1 .

To prove the theorem, it clearly suffices to show that S1 = D.

Suppose that z1 ∈ S1 . Then there exists
1 > 0 such that h(z) = 0 in {z : |z − z1 | <
1 }. Suppose
now that |z  − z1 | <
1 .

z’
z ’-z1
ε1
z1

Then clearly

{z : |z − z  | <
1 − |z  − z1 |} ⊆ {z : |z − z1 | <
1 },

and so it follows that h(z) = 0 in the neighbourhood {z : |z − z  | <
1 − |z  − z1 |} of z  . This shows that
z  ∈ S1 whenever |z  − z| <
1 . It follows that {z : |z − z1 | <
1 } ⊆ S1 , so that S1 is open.

Suppose next that z2 ∈ S2 . Since z2 ∈ D, there exists R > 0 such that the disc {z : |z − z2 | < R}
is contained in D and so it follows from Theorem 7A that the Taylor series expansion

∞
h(n) (z2 )
h(z) = (z − z2 )n
n=0
n!

is valid in the disc {z : |z − z2 | ≤ r} for every r < R. Since h is not identically zero in this latter disc,
there exists a smallest n such that h(n) (z2 ) = 0. By Theorem 7C, we can then write

h(z) = hn (z)(z − z2 )n ,
 Chapter 7 : Taylor Series, Uniqueness and the Maximum Principle 7–7

where hn (z) is analytic in the disc {z : |z − z2 | < R} and

h(n) (z2 )
hn (z) → = 0
n!

as z → z2 . It follows from continuity of hn that there exists
2 > 0 such that h(z) = 0 in the punctured
disc {z : 0 < |z − z2 | <
2 }, and so {z : |z − z2 | <
2 } ⊆ S2 . Hence S2 is open.

Clearly S1 = ∅, since z0 ∈ S1 . Suppose, on the contrary, that S1 = D. Then the two open sets S1
and S2 are both non-empty. Clearly

S1 ∪ S2 = D and S1 ∩ S2 = ∅.

In view of our earlier remark, this is absurd. Hence we must have S2 = ∅, and so S1 = D.

In fact, a slight elaboration of the ideas in part of the above proof gives the following two results.

THEOREM 7F. Suppose that a function f is analytic in the domain D. Suppose further that z0 ∈ D,
and that f (z0 ) = 0. Then either f (z) is identically zero in D or else there exists n ∈ N such that

f (z) = (z − z0 )n g(z),

where the function g is analytic in D, and

f (n) (z0 )
g(z0 ) = = 0.
n!

Definition. If the latter conclusion of Theorem 7F holds, then we say that the function has a zero of
order n at z0 . Furthermore, if n = 1, then we say that the function f has a simple zero at z0 .

THEOREM 7G. Suppose that a function f is analytic in the domain D. Suppose further that f (z)
is not identically zero in D. Then for every z0 ∈ D such that f (z0 ) = 0, there exists
> 0 such that
f (z) = 0 for every 0 < |z − z0 | <
. In other words, the zeros of f are isolated.

Proof of Theorem 7F. Clearly there exists R > 0 such that the disc {z : |z − z0 | < R} is contained
in D. Suppose that f (n) (z0 ) = 0 for every n ∈ N. Then it follows from Theorem 7A that f (z) is
identically zero in this disc. Let the function g be identically zero in D. Then f (z) = g(z) in some
neighbourhood of z0 . It now follows from Theorem 7E that f (z) = 0 for every z ∈ D. Suppose next
that f (k) (z0 ) = 0 for some k ∈ N. Then there exists a smallest k ∈ N such that f (k) (z0 ) = 0. Denote
this value of k by n. By Theorem 7C, we can then write

f (z) = fn (z)(z − z0 )n ,

where fn (z) is analytic in the disc {z : |z − z0 | < R} and

f (n) (z0 )
fn (z) → = 0
n!

as z → z0 . We now define g(z) = fn (z) in this disc, and by g(z) = f (z)(z − z0 )−n in the remainder of
D to complete the proof.

Proof of Theorem 7G. It follows from Theorem 7F that there exists n ∈ N and an analytic function
g in D such that f (z) = (z − z0 )n g(z), where g(z0 ) = 0. It follows from the continuity of g that there
exists
> 0 such that g(z) = 0 if |z − z0 | <
. The result follows immediately.

7–8 W W L Chen : Introduction to Complex Analysis

Example 7.3.1. Suppose that two functions f and g are analytic in a domain D and not identically
zero in D. Suppose further that z0 ∈ D, and that f (z0 ) = g(z0 ) = 0. Then by Theorem 7F, there exist
m, n ∈ N and functions F and G analytic in D and satisfying F (z0 ) = 0 and G(z0 ) = 0 such that

f (z) = (z − z0 )m F (z) and g(z) = (z − z0 )n G(z).

Then

f (z) F (z) f  (z) 

k mF (z) + (z − z0 )F (z)
= (z − z0 )k and = (z − z 0 ) ,
g(z) G(z) g  (z) nG(z) + (z − z0 )G (z)

where k = m − n. Consider now the special case m = n = 1, so that k = 0. We have

f  (z) F (z) + (z − z0 )F  (z) F (z) f (z)

lim 
= lim 
= lim = lim .
z→z0 g (z) z→z0 G(z) + (z − z0 )G (z) z→z0 G(z) z→z0 g(z)

This is l’Hopital’s rule.

We complete this section by proving the following result, which shows that an analytic function is
determined in a domain D by the values on a short continuous curve in D.

THEOREM 7H. Suppose that a function f is analytic in the domain D. Suppose further that zn
is a sequence of distinct points having a limit z0 ∈ D, and that f (zn ) = g(zn ) for every n ∈ N. Then
f (z) = g(z) for every z ∈ D.

Proof. For every z ∈ D, write h(z) = f (z) − g(z). Then the function h is analytic in D. Furthermore,
it follows from continuity that h(z0 ) = 0. Since h(zn ) = 0 for every n ∈ N and zn → z0 as n → ∞, the
zero z0 of the function h is not isolated. It follows from Theorem 7G that h(z) is identically zero in D.

7.4. The Maximum Principle

Let us return to Cauchy’s integral formula, as given by Theorem 6A. If we take z to be the centre α of
the circle C, then (1) in Chapter 6 gives

1 f (ζ)
(10) f (α) = dζ,
2πi C ζ −α

so that

(11) |f (α)| ≤ max |f (ζ)|,

ζ∈C

in view of Theorem 4B. In other words, the modulus of an analytic function at a point in a domain never
exceeds the maximum modulus of the function on the boundary of any disc centred at that point and
contained in the domain.

In this section, we shall establish the following stronger result.

THEOREM 7J. (MAXIMUM PRINCIPLE) Suppose that a function f is analytic in a domain D.

Then |f (z)| cannot have a maximum anywhere in D unless f (z) is constant in D.

Proof. Suppose on the contrary that there exists α ∈ D such that

(12) |f (α)| ≥ |f (z)|

 Chapter 7 : Taylor Series, Uniqueness and the Maximum Principle 7–9

for every z ∈ D. Since D is open, there exists an
-neighbourhood S of α which is contained in D. If

|f (z)| = |f (α)| for every z ∈ S, then it follows from Example 3.3.5 that f (z) is constant in S, and so
constant in D by Theorem 7E. We may therefore assume that there exists z1 ∈ S such that

(13) |f (z1 )| < |f (α)|.

D
ε z1
r
S α
C

Let |z1 − α| = r. Clearly r <
. If we denote by C the circle in the positive (anticlockwise) direction
centred at α and with radius r, then (10) holds. Furthermore, writing ζ = α + reit , we have
 2π
1
f (α) = f (α + reit ) dt,
2π 0

so that
 2π
1
(14) |f (α)| ≤ |f (α + reit )| dt.
2π 0

On the other hand, it clearly follows from (12) that

 2π
1
(15) |f (α)| ≥ |f (α + reit )| dt.
2π 0

However, note that z1 = α + reit1 for some t1 ∈ [0, 2π]. It follows from continuity that there exists an
interval I ⊆ [0, 2π] for which |f (α + reit )| < |f (α)| for every t ∈ I. It follows that equality cannot hold
in (15), so that
 2π
1
(16) |f (α)| > |f (α + reit )| dt.
2π 0

Note now that (14) and (16) contradict each other, and this concludes the proof of the theorem.

The following alternative form of the Maximum principle is perhaps more useful.

THEOREM 7K. Suppose that a function f is analytic in a bounded domain D, and continuous in
the closed region D. Then |f (z)| attains its maximum on the boundary of D.

Proof. It is well known from real analysis that |f (z)| assumes its maximum somewhere in the closed
bounded region D. By Theorem 7J, this maximum cannot be attained in D, and so must be attained
on the boundary of D.

7–10 W W L Chen : Introduction to Complex Analysis

Problems for Chapter 7

1. Obtain the Taylor series for the function (1 − z)−1 at z = 0. Deduce from this the Taylor series for
the function (1 − z)−2 at z = 0. In what open discs centred at z = 0 are these series valid?

2. Suppose that a function f (z) is analytic in the disc {z : |z| < R}, where R > 0 is fixed, and has
Taylor series
∞


f (z) = an z n .
n=0

a) Show that
 z
∞
an n+1
f (ζ) dζ = z
0 n=0
n+1

uniformly in {z : |z| ≤ r} for every r < R.

b) Denote the integral by F (z). Explain why the series in (a) is the Taylor series for F (z).

3. Deduce from the Taylor series for (1 − z)−1 at z = 0 in Problem 1 the Taylor series for the function
log(1 − z) at z = 0, where log 1 = 0. In what open discs centred at z = 0 is this series valid?

4. Suppose that α ∈ C is fixed. By interpreting the function (1 + z)α as eα log(1+z) , with log 1 = 0,
show that

α(α − 1) 2 α(α − 1)(α − 2) 3

(1 + z)α = 1 + αz + z + z + ...
2! 3!

uniformly in {z : |z| ≤ r} for every r < 1.

5. a) Show that the function f (z), defined by f (0) = 1 and f (z) = z −1 sin z when z = 0, is entire.
b) Obtain the Taylor series for f (z) at z = 0.
 z
c) Obtain the Taylor series for the integral f (ζ) dζ at z = 0.
0

6. Suppose that P (z) is a polynomial of degree at most 3. Using partial fractions if necessary, find a0 ,
a1 and a2 such that

P (z)
= a0 + a1 z + a2 z 2 + . . .
(z 2 + 1)(z − 1)(z − 2)

valid whenever |z| < 1.

7. Suppose that a power series

∞


an z n
n=0

converges uniformly to an analytic function f (z) in the disc D = {z : |z| ≤ R}, where R > 0 is
fixed. For every N ∈ N and z ∈ D, let

N
sN (z) = an z n .
n=0

Then given any
> 0, there exists N0 such that |f (z) − sN (z)| <
for every N > N0 and z ∈ D.
 Chapter 7 : Taylor Series, Uniqueness and the Maximum Principle 7–11

a) Show that for every k ∈ N ∪ {0} and every N > N0 ,

 
 
 (f (z) − SN (z))z −k−1 dz  ≤ 2π
,
  Rk
C

where C = {z : |z| = R}, followed in the positive (anticlockwise) direction.

b) Now let N > k. Show that
  
 1 f (z) 

 dz − ak  ≤ k .
 2πi k+1
C z R

c) Deduce that ak = f (k) (0)/k! for every k ∈ N ∪ {0}.

[Remark: This shows that if a power series converges uniformly to an analytic function f (z) in D,
then it is the Taylor series for f (z).]

8. Suppose that two functions f (z) and g(z) are analytic in the disc D = {z : |z| < R}, where R > 0
is fixed, with Taylor series

f (z) = a0 + a1 z + a2 z 2 + . . . and g(z) = b0 + b1 z + b2 z 2 + . . .

respectively.
a) Without worrying about convergence problems, multiply the two series together to obtain
another power series c0 + c1 z + c2 z 2 + . . . . Check that cn = a0 bn + a1 bn−1 + . . . + an b0 for every
n ∈ N ∪ {0}.
b) Suppose that f (z)g(z) has a Taylor series. Show that the coefficient of the term z n in the
Taylor series is given by the value at z = 0 of the function
1 dn
(f (z)g(z)).
n! dz n
c) Explain why the power series in (a) is the Taylor series for f (z)g(z).

9. Suppose that two functions f (z) and g(z) are analytic in a bounded region D and continuous in
D. Suppose further that f (z) = g(z) for every z on the boundary of D. Show from the Maximum
principle that f (z) = g(z) for every z ∈ D.

10. Suppose that a function f (z) is analytic in a closed bounded region D. Suppose further that f (z) = 0
for any z ∈ D. Show that |f (z)| assumes its minimum value on the boundary of D.

11. Using l’Hopital’s rule, or otherwise, evaluate each of the following limits:
sin z eπz + 1
a) lim b) lim 2
z→π π − z z→i z + 1

12. Suppose that f (z) is analytic in the disc D = {z : |z| ≤ R}, where R > 0 is fixed. Suppose further
that f (0) = 0 and |f (z)| ≤ M whenever |z| = R.
a) Explain why f (z) = zg(z) for some function g(z) analytic in D.
b) By applying the Maximum principle on the function g(z), prove Schwarz’s lemma, that
M
|f (z)| < |z|
R
whenever 0 < |z| < R, unless f (z) = cz for some constant c ∈ C.

13. Suppose that C is a contour of length L. Suppose further that a function f (z) is continuous on C,
and |f (z)| ≤ M for every z ∈ C. Show that, unless |f (z)| = M for every z ∈ C, we have
 
 
 f (z) dz  < M L.
 
C
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 8
ISOLATED SINGULARITIES
AND LAURENT SERIES

8.1. Removable Singularities

Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}. Observe that it is not
necessary for f to be defined at the point z0 . We say that the function f has an isolated singularity at z0 .
Our purpose is to show that there are only three possible ways in which f (z) can behave in a punctured
neighbourhood of z0 . To illustrate the first of these, let us first consider the following examples.

Example 8.1.1. The function

sin z
f (z) =
z

is analytic in the punctured disc {z : 0 < |z| < R}. However, the quotient is not defined at z = 0.
However, note that the function sin z is entire. By Theorem 7C, we can write

sin z = z + z 3 g(z),

where g is an entire function. It follows that for z = 0, we have

sin z
f (z) = = 1 + z 2 g(z).
z

Note that the function 1 + z 2 g(z) is entire. It follows that if we make the further definition f (0) = 1,
then f is now analytic at z = 0, and we have removed the isolated singularity.
8–2 W W L Chen : Introduction to Complex Analysis

Example 8.1.2. Suppose that a function f is analytic in a domain D, and that z0 ∈ D. We define
the function g in D by writing

(1) g(z0 ) = f
(z0 ),

and writing

f (z) − f (z0 )
(2) g(z) =
z − z0
if z = z0 . It is easily seen from Theorem 7C that g is analytic in D. However, note that the function
g, defined by (2), is analytic in the domain D \ {z0 }. It also has an isolated singularity at z0 , which is
removed by the definition (1).

Definition. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}.
Suppose further that by assigning a suitable value for f (z0 ), the function f can be made to be analytic
in the disc {z : |z − z0 | < R}. Then we say that f has a removable singularity at z0 .

THEOREM 8A. (RIEMANN’S THEOREM ON REMOVABLE SINGULARITIES) Suppose that

a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}. Suppose further that

(3) lim (z − z0 )f (z) = 0.

z→z0

Then f has a removable singularity at z0 .

Proof. Suppose that z is a point in the punctured disc {z : 0 < |z − z0 | < R}. Then 0 < |z − z0 | < R.
Let r1 and r2 satisfy 0 < r1 < |z − z0 | < r2 < R, and let C1 and C2 denote two circles in the positive
(anticlockwise) direction, centred at z0 , and of radius r1 and r2 respectively.

R z

r1
r2 z0
C1 C2

The function g, defined by g(z) = f
(z) and for ζ = z by

f (ζ) − f (z)
(4) g(ζ) = ,
ζ −z

is clearly analytic in the punctured disk {ζ : 0 < |ζ − z0 | < R}. Then it can be shown, as in the proof
of Theorem 6A, that



g(ζ) dζ = g(ζ) dζ.
C1 C2
 Chapter 8 : Isolated Singularities and Laurent Series 8–3

Combining this with (4), we have

f (ζ) dζ f (ζ) dζ
(5) dζ − f (z) = dζ − f (z) .
C1 ζ −z C1 ζ −z C2 ζ −z C2 ζ −z

Note now that the function

1
ζ −z

is analytic in the star domain {ζ : |ζ − z0 | < |z − z0 |} which contains the contour C1 . It follows that


dζ
(6) = 0.
C1 ζ −z

On the other hand, by Cauchy’s integral formula as given by Theorem 6A, we have


dζ
(7) = 2πi.
C2 ζ −z

Furthermore, in view of the condition (3), we have, given any  > 0, there exists δ > 0 such that
|(ζ − z0 )f (ζ)| <  whenever |ζ − z0 | < δ. Without loss of generality, we may assume that

1
(8) δ< |z − z0 |.
2
If we now take r1 = δ, then

 

 f (ζ)   (ζ − z0 )f (ζ)   2π 4π
  
dζ  =  dζ  ≤ 2πδ = ≤ ,
 ζ − z (ζ − z 0 )(ζ − z) δ(|z − z 0 | − δ) |z − z 0 | − δ |z − z0 |
C1 C1

in view of Theorem 4B and (8). Since  > 0 is arbitrary, we conclude that

f (ζ)
(9) dζ = 0.
C1 ζ −z

Combining (5)–(7) and (9), we obtain

1 f (ζ)
(10) f (z) = dζ.
2πi C2 ζ −z

Note now that (10) holds for every z in the punctured disc {z : 0 < |z − z0 | < r2 }. Note also that the
integral on the right hand side of (10) represents an analytic function in the disc {z : |z − z0 | < r2 } (see
the proof of Theorem 6B). It follows that if we define


1 f (ζ)
f (z0 ) = dζ,
2πi C2 ζ − z0

then the function f is analytic in the disc {z : |z − z0 | < r2 }.

Remarks. (1) Note that condition (3) will be satisfied if f (z) is continuous at z0 , or if |f (z)| is
bounded.

(2) Since an analytic function is continuous, it follows that removable singularities at z0 can be
overcome by defining

f (z0 ) = lim f (z).

z→z0
8–4 W W L Chen : Introduction to Complex Analysis

8.2. Poles

Definition. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}.
Suppose further that

g(z)
(11) f (z) = ,
(z − z0 )n

where n ∈ N and the function g is analytic in some neighbourhood of z0 , with g(z0 ) = 0. Then we say
that f has a pole of order n at z0 . Furthermore, if n = 1, then we say that f has a simple pole at z0 .

THEOREM 8B. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}.
Then f has a pole at z0 if and only if

(12) lim |f (z)| = ∞;

z→z0

in other words, given any E > 0, there exists δ > 0 such that |f (z)| > E whenever 0 < |z − z0 | < δ.

Proof. Note first of all that (12) follows immediately from (11), since g(z0 ) = 0. Suppose now that
(12) holds. Then f (z) = 0 in some punctured disc {z : 0 < |z − z0 | < r}, where r ≤ R. It follows that
the function
1
F (z) =
f (z)

is analytic in {z : 0 < |z − z0 | < r}, and has an isolated singularity at z0 . On the other hand, it follows
from (12) that F (z) → 0 as z → z0 . Hence by Theorem 8A, F has a removable singularity at z0 . If we
define F (z0 ) = 0, then F is now analytic in the disc {z : |z − z0 | < r}. Clearly F (z) is not identically
zero in {z : |z − z0 | < r}. It follows from Theorem 7F that there exists n ∈ N such that

F (z) = (z − z0 )n h(z),

where the function h is analytic in {z : |z − z0 | < r}, with h(z0 ) = 0. Hence

1
g(z) =
h(z)

is analytic in some neighbourhood of z0 , and (11) holds. Clearly g(z0 ) = 0.

Remark. Note that a function f has a pole of order n at z0 if and only if the function 1/f has a zero
of order n at z0 .

8.3. Essential Singularities

Definition. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R}.
Suppose further that the isolated singularity at z0 is neither removable nor a pole. Then we say that f
has an essential singularity at z0 .

Example 8.3.1. The function e1/z is analytic at every z = 0. It has an isolated singularity at z = 0.
Let us restrict z to be real numbers, and consider e1/x , where x > 0. Clearly

lim e1/x = lim ey = ∞,

x→0+ y→+∞
 Chapter 8 : Isolated Singularities and Laurent Series 8–5

so that the singularity is not removable. On the other hand, for every n ∈ N,

ey
lim xn e1/x = lim = ∞,
x→0+ y→+∞ y n

so that the singularity is not a pole of order n. Hence e1/z has an essential singularity at z = 0.

To illustrate the wild behaviour of an analytic function near an essential singularity, we mention
Picard’s theorem that such a function assumes all values except possibly one in any neighbourhood of
an essential singularity. The following result is somewhat weaker, and shows that such a function comes
arbitrarily close to any given complex number in any neighbourhood of an essential singularity.

THEOREM 8C. (CASORATI-WEIERSTRASS) Suppose that a function f is analytic in the punc-

tured disc {z : 0 < |z − z0 | < R}, with an essential singularity at z0 . Then given any w ∈ C and any
real numbers  > 0 and δ > 0, there exists z in the punctured disc satisfying

0 < |z − z0 | < δ and |f (z) − w| < .

Proof. Suppose on the contrary that the conclusion does not hold. Then there exist w ∈ C and real
numbers  > 0 and δ > 0 such that |f (z) − w| ≥  whenever 0 < |z − z0 | < δ. It follows that the function

1
g(z) =
f (z) − w

is analytic and bounded in the punctured disc {z : 0 < |z − z0 | < δ}, with an isolated singularity at z0
which is removable, in view of Theorem 8A. It follows that by defining g(z0 ) appropriately, the function
g is analytic in the disc {z : |z − z0 | < δ}. On the other hand, the function g is clearly not identically
zero in {z : |z − z0 | < δ}. Furthermore, note that

1
f (z) = w + .
g(z)

If g(z0 ) = 0, then f is analytic at z0 . If g(z0 ) = 0, then f has a pole at z0 . In either case, the conclusion
contradicts the assumption that f has an essential singularity at z0 , and this completes the proof.

8.4. Isolated Singularities at Infinity

The behaviour of a function f (z) at z = ∞ can be studied via the behaviour of the function f (1/ζ) at
ζ = 0. A punctured neighbourhood {ζ : 0 < |ζ| < R−1 } of 0 then plays the same role as the “punctured”
neighbourhood {z : R < |z| < ∞} of ∞.

Suppose now that a function f (z) is analytic in the domain {z : R < |z| < ∞}. Then by using
z = 1/ζ and considering ζ = 0, we see that the function f (z) has an isolated singularity at z = ∞. This
may be a removable singularity, a pole or an essential singularity.

Corresponding to Theorem 8A, suppose that |f (z)/z| → 0 as |z| → ∞. Then the singularity is
removable by defining f (∞) suitably to make f (z) continuous at z = ∞. In other words, we need to
define
 
1
f (∞) = lim f .
ζ→0 ζ
8–6 W W L Chen : Introduction to Complex Analysis

In the special case that f (∞) = 0, then we say that f has a zero at z = ∞. Furthermore, if f is not
identically zero, then, corresponding to Theorem 7F, there exists n ∈ N such that

h(z)
f (z) = ,
zn

where h(z) is analytic in {z : R < |z| < ∞}, and h(∞) = 0. In this case, we say that f has a zero of
order n at z = ∞.

Corresponding to Theorem 8B, suppose that |f (z)| → ∞ as |z| → ∞. Then f has a pole at z = ∞,
and there exists n ∈ N such that

f (z) = z n h(z),

where h(z) is analytic in {z : R < |z| < ∞}, and h(∞) = 0. In this case, we say that f has a pole of
order n at z = ∞.

Corresponding to Theorem 8C, suppose that the isolated singularity at z = ∞ is neither removable
nor a pole. Then it is an essential singularity. In this case, given any w ∈ C and any real numbers  > 0
and N > 0, there exists z in the domain {z : R < |z| < ∞} satisfying

|z| > N and |f (z) − w| < .

In other words, the function f (z) comes arbitrarily close to any given complex number in any neigh-
bourhood of z = ∞.

8.5. Further Examples

Example 8.5.1. The function

ez − 1
f (z) =
z(z − 1)

is analytic at every z ∈ C except for isolated singularities at z = 0, 1. At z = 1, it has a simple pole;

note that we can write

g(z) ez − 1
f (z) = with g(z) = ,
z−1 z

and g(1) = 0. At z = 0, it has a removable singularity, since

ez − 1 ez
lim = lim = −1
z→0 z(z − 1) z→0 2z − 1

by l’Hopital’s rule. It follows that if we define f (0) = −1, then f is analytic at z = 0. The function f (z)
also has an isolated singularity at z = ∞. To study the isolated singularity at z = ∞, note first of all
that
ez − 1
lim
|z|→∞ z(z − 1)

does not exist. To see this, note that

ex − 1 ex − 1
lim = +∞ and lim = 0.
x→+∞ x(x − 1) x→−∞ x(x − 1)
 Chapter 8 : Isolated Singularities and Laurent Series 8–7

Hence the singularity is not removable. Suppose next that n ∈ N is given and fixed. Then

f (z) ez − 1
h(z) = =
zn z n+1 (z − 1)

is not analytic at z = ∞, since

ez − 1
lim
|z|→∞ z n+1 (z − 1)

does not exist. To see this, note that

ex − 1 ex − 1
lim = +∞ and lim = 0.
x→+∞ xn+1 (x − 1) x→−∞ xn+1 (x − 1)

Hence the singularity is not a pole. It follows that f (z) has an essential singularity at z = ∞.

Example 8.5.2. The function

(z 2 − 4)(z − 1)4
f (z) =
(sin πz)4

is analytic at every z ∈ C except for isolated singularities at z = 0, ±1, ±2, . . . , where the denominator
vanishes. Note also that the numerator vanishes at z = 1, ±2. Note that the function sin πz has simple
zeros at z = 0, ±1, ±2, . . . . It follows that f has poles of order 4 at z = 0, −1, ±3, ±4, ±5, . . . . Next, note
that the function (z 2 − 4)(z − 1)4 has simple zeros at z = ±2. It follows that f has poles of order 3 at
z = ±2. To study the isolated singularity at z = 1, note that by Theorem 7C, we have

sin πz = −π(z − 1) + g(z)(z − 1)2 ,

where g is entire. It follows that

(z 2 − 4)(z − 1)4 z2 − 4 3
lim = lim = − 4,
z→1 (sin πz)4 z→1 (π − g(z)(z − 1))4 π

and so f has a removable singularity at z = 1. Finally, the singularity at z = ∞ is not isolated, since
there does not exist any R > 0 such that the function f (z) is analytic in the domain {z : R < |z| < ∞}.

8.6. Laurent Series

Example 8.6.1. Suppose that the function f is analytic in the punctured disc {z : 0 < |z − z0 | < R},
with a pole of order m at z0 . Then

g(z)
f (z) = ,
(z − z0 )m

where the function g is analytic in {z : |z − z0 | < R}, with g(z0 ) = 0. By Theorem 7C, we have

g

(z0 ) g (m−1) (z0 )

g(z) = g(z0 ) + g
(z0 )(z − z0 ) + (z − z0 )2 + . . . + (z − z0 )m−1 + gm (z)(z − z0 )m ,
2! (m − 1)!

where gm (z) is analytic in the disc {z : |z − z0 | < R}. It follows that

g(z0 ) g
(z0 ) g

(z0 ) g (m−1) (z0 )

f (z) = + + + ... + + gm (z).
(z − z0 ) m (z − z0 ) m−1 2!(z − z0 ) m−2 (m − 1)!(z − z0 )
8–8 W W L Chen : Introduction to Complex Analysis

The expression

g(z0 ) g
(z0 ) g

(z0 ) g (m−1) (z0 )

+ + + . . . +
(z − z0 )m (z − z0 )m−1 2!(z − z0 )m−2 (m − 1)!(z − z0 )

is called the principal part of f at z0 . If we use Theorem 7A instead, then we can show that
∞

f (z) = an (z − z0 )n
n=−m

for suitable choices of the coefficients an .

Our main task in this section is to generalize this example. The first step in this direction can be
summarized by the following result.

THEOREM 8D. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R},
with an isolated singularity at z0 . Then there exist unique functions f1 and f2 such that
(a) f (z) = f1 (z) + f2 (z) in {z : 0 < |z − z0 | < R},
(b) f1 is analytic in C except possibly at z0 ,
(c) f1 (z) → 0 as |z| → ∞, and
(d) f2 is analytic in the disc {z : |z − z0 | < R}.

Proof. We begin the proof in the same way as for Theorem 8A. Suppose that z is a point in the
punctured disc {z : 0 < |z − z0 | < R}. Let r1 and r2 satisfy 0 < r1 < |z − z0 | < r2 < R, and let C1
and C2 denote two circles in the positive (anticlockwise) direction, centred at z0 , and of radius r1 and
r2 respectively. On combining (5)–(7), we obtain



1 f (ζ) 1 f (ζ)
(13) f (z) = dζ − dζ.
2πi C2 ζ − z 2πi C1 ζ − z

Write



1 f (ζ) 1 f (ζ)
(14) f1 (z) = − dζ and f2 (z) = dζ.
2πi C1 ζ −z 2πi C2 ζ −z

Part (a) follows immediately. For part (d), note that the second integral in (14) represents an analytic
function in the disc {z : |z − z0 | < r2 } (as in the proof of Theorems 6B and 8A). For part (b), note that
the first integral in (14) represents an analytic function in the annulus {z : |z − z0 | > r1 } (similar to
the proof of Theorem 6B). Note next that f2 (z) and f (z) are independent of the choice of r1 , so that it
follows from (a) that f1 (z) is also independent of the choice of r1 . Similarly, f2 (z) is independent of the
choice of r2 . It is easy to see that


f (ζ)
lim dζ = 0.
|z|→∞ C1 ζ − z

Part (c) follows immediately. To show that the functions f1 and f2 are unique, suppose that g1 and g2
are functions having the same properties as f1 and f2 respectively. Then

f1 (z) − g1 (z) = g2 (z) − f2 (z)

in the punctured disc {z : 0 < |z − z0 | < R}. Let


g2 (z) − f2 (z) if |z − z0 | < R,
F (z) =
f1 (z) − g1 (z) if |z − z0 | > 0.

Then F is entire. On the other hand, it follows from part (c) that F (z) → 0 as |z| → ∞. Hence F is
bounded. It follows from Liouville’s theorem that F is constant in C, and so we must have F (z) = 0 for
every z ∈ C. This completes the proof.

 Chapter 8 : Isolated Singularities and Laurent Series 8–9

We can now state our generalization of Example 8.6.1.

THEOREM 8E. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R},
with an isolated singularity at z0 . For every n ∈ Z, let


1 f (z)
(15) an = dz,
2πi C (z − z0 )n+1

where C is a circle in the positive (anticlockwise) direction centred at z0 and of radius r, where 0 < r < R.
Then the series
∞

(16) f (z) = an (z − z0 )n
n=−∞

is convergent in the punctured disc {z : 0 < |z − z0 | < R}. Furthermore, this convergence is uniform in
any annulus {z : r1 < |z − z0 | < r2 }, where 0 < r1 < r2 < R.

Remark. To say that the series converges uniformly to f (z) in the annulus {z : r1 < |z − z0 | < r2 },
we mean given any  > 0, there exists N0 = N0 (, r1 , r2 ), independent of the choice of z, such that
 
 
N2 
 n
f (z) − an (z − z0 )  < 
 
n=−N1

for every z in the annulus {z : r1 < |z − z0 | < r2 } whenever N1 > N0 and N2 > N0 .

Definition. The series (16) is called the Laurent series for the function f at z0 .

Proof of Theorem 8E. The first step in our proof is to show that if the series in (16) converges to
f (z) uniformly on the circle C centred at z0 and of radius r, where 0 < r < R, then the coefficients an
are given by (15). Suppose that n ∈ Z is chosen and fixed. For any  > 0, we can choose N1 and N2 so
large that −N1 ≤ n ≤ N2 and
 
 
 N2

f (z) − a (z − z j
0 <
)
 j
 j=−N1 

for every z ∈ C. Then it follows from Theorem 4B that

   


 1 N2
dz 
(17)  f (z) − aj (z − z0 ) 
j ≤  .
 2πi (z − z0 )n+1  rn
 C j=−N1 

Since


1 1 if k = −1,
(z − z0 )k dz =
2πi C 0  −1,
if k =

we have
 


N2
1  dz
aj (z − z0 )j  = an ,
2πi C (z − z0 )n+1
j=−N1

so that (17) can be simplified to



 1 f (z)  
 dz − an  ≤ n .
 2πi
C (z − z0 )
n+1 r
8–10 W W L Chen : Introduction to Complex Analysis

Since  > 0 is arbitrary, (15) follows immediately. It now remains to show that f (z) can be represented
in the form (16) in the punctured disc {z : 0 < |z − z0 | < R}, and that the convergence is uniform in any
annulus {z : r1 < |z − z0 | < r2 }, where 0 < r1 < r2 < R. Suppose that 0 < r1 < r < r2 < R. Following
Theorem 8D, we can write

(18) f (z) = f1 (z) + f2 (z),

where f1 (z) and f2 (z) are uniquely determined and satisfy conditions (b)–(d) of Theorem 8D. Since f2
is analytic in the disc {z : |z − z0 | < R}, it follows from Theorem 7A that the Taylor series
∞

(19) f2 (z) = An (z − z0 )n
n=0

converges in the disc {z : |z − z0 | < R}, uniformly in the closed disc {z : |z − z0 | ≤ r2 }. To study f1 (z),
write
1 1
w= or z= + z0 .
z − z0 w
Then
 
1
f1 (z) = f1 + z0
w
is an entire function of w, and so it follows from Theorem 7A that the Taylor series
   ∞
1
(20) f1 + z0 = Bm wm
w m=1

converges in C, uniformly in the closed disc {w : |w| ≤ 1/r1 }. Note that the constant term B0 in the
Taylor series is missing, since B0 corresponds to the value of the function at w = 0, or z = ∞, and this
is 0 in view of condition (c) in Theorem 8D. However, (20) is equivalent to saying that the series
∞

(21) f1 (z) = Bm (z − z0 )−m
m=1

converges in C \ {0}, uniformly in {z : |z − z0 | ≥ r1 }. The result now follows on combining (18), (19)
and (21).

Definition. The series

−1

f1 (z) = an (z − z0 )n ,
n=−∞

where an is given by (15), is called the principal part of the function f at z0 .

The next result highlights the relationship between the principal part of a function and the nature
of the isolated singularity.

THEOREM 8F. Suppose that a function f is analytic in the punctured disc {z : 0 < |z − z0 | < R},
with an isolated singularity at z0 . Suppose further that the Laurent coefficients an are given by (15).
(a) The function f either is analytic or has a removable singularity at z0 if and only if an = 0 for every
n < 0.
(b) The function f has a pole at z0 if and only if a positive but finite number of coefficients an with
n < 0 are non-zero.
(c) The function f has an essential singularity at z0 if and only if an infinite number of coefficients an
with n < 0 are non-zero.
 Chapter 8 : Isolated Singularities and Laurent Series 8–11

Proof. Note first of all that if f has a removable singularity at z0 , then f can be made analytic at z0
by a suitable choice of f (z0 ). Part (a) now follows on observing that an analytic function has a Taylor
series, and that a Laurent series with no principal part is a Taylor series. To prove part (b), note first
of all that if a positive but finite number of coefficients an with n < 0 are non-zero, then there exists
m > 0 such that a−m = 0 but an = 0 for every n < −m. In this case, we have
∞

f (z) = an (z − z0 )n ,
n=−m

so that

g(z)
f (z) = ,
(z − z0 )m

where m ∈ N and the function g is analytic in some neighbourhood of z0 , with g(z0 ) = a−m = 0. This
shows that f has a pole of order m at z0 . The converse is given in Example 8.6.1. Part (b) follows. Part
(c) follows immediately from (a) and (b).

Example 8.6.2. The observation that a Laurent series is unique enables us to use different methods
to find the coefficients apart from the formula (15). Consider, for example, the function e1/z . Using the
substitution z = 1/w on the Taylor series

∞
wn
ew = ,
n=0
n!

we obtain the Laurent series

∞
 1 1 1 1
e1/z = n
= ... + 3
+ 2
+ + 1.
n=0
n!z 3!z 2!z z

We conclude this chapter by making a remark on various equivalent definitions of analyticity in a

domain D. The reader is advised to check the following theorem very carefully.

THEOREM 8G. For any function f and any domain D, the following statements are equivalent:
(a) f (z) is analytic in D.
(b) f (z) has continuous derivatives of all orders in D.
(c) f
(z) exists and is continuous in D.
(d) f
(z) exists in D.
(e) f
(z) exists in D except possibly at a finite number of points in D, and f (z) is continuous at these
exceptional points.
(f ) f (z) can be represented uniformly by its Taylor series in the neighbourhood of every point in D.

Problems for Chapter 8

1. For each of the functions below, classify all the singular points in C:
cos z z2 + 1
a) f (z) = ez b) f (z) = c) f (z) = 2
z z −1
z4 z
d) f (z) = 3 e) f (z) =
z +z cos z

2. Show that the principal parts of the function f (z) = 8z 3 (z + 1)−1 (z − 1)−2 at z = −1 and z = 1
are respectively −2(z + 1)−1 and 4(z − 1)−2 + 10(z − 1)−1 .
8–12 W W L Chen : Introduction to Complex Analysis

3. For each of the functions below, find the principal part at the given points:
ez z6
a) f (z) = 4 at the point z = 0 b) f (z) = at the point z = 1
z (1 − z)3
sin z
c) f (z) = at the point z = 2π
(z − 2π)2

4. Expand the function (z − 1)/(z + 1) in powers of 1/z.

5. For each of the functions below, use partial fractions if appropriate and find the principal part at
each of its singular points in C:
12 z4 + 1
a) f (z) = 2 2 b) f (z) =
z (z + 4) z(z 2 + 1)2
6
48z z9 + 1
c) f (z) = d) f (z) =
(z − 1)2 (z − 2) (z − 1)3 (z 2 + 4)2

6. Suppose that f (z) = b−m z −m + b−m+1 z −m+1 + . . . + b0 + b1 z + . . . + bk z k , where m, k ∈ N. Suppose

further that f (z) has Laurent series
∞

an z n
n=−∞

at the point z = 0. Show by direct calculation that an = bn whenever −m ≤ n ≤ k and an = 0

otherwise.

7. a) Consider the function f (z) = e1/z . Note that for every k ∈ Z, the coefficient for the term z k in
the Laurent series of f (z) at z = 0 is given by


1 e1/ζ
ak = dζ,
2πi C ζ k+1

where C is the circle {z : |z| = 1} followed in the positive (anticlockwise) direction. Show that

π
1
ak = ecos θ cos(sin θ + kθ) dθ.
2π −π

b) Find the Laurent series for the function f (z) = e1/z at z = 0 without using part (a).
c) Deduce that for every n ∈ N ∪ {0},

π
1 1
ecos θ cos(sin θ − nθ) dθ = .
π 0 n!
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 9
CAUCHY’S INTEGRAL THEOREM REVISITED

9.1. Simply Connected Domains

Cauchy’s integral theorem states that in a simply connected domain, the integral of an analytic function
over a closed contour is zero. In Chapter 5, we have studied a special case of the theorem in order to
develop the basic properties of analytic functions. More precisely, we have studied the special case when
the domain in question is a star domain. For most purposes in pure and applied mathematics, this
special case is adequate. However, by introducing further line segments, one may be able to extend the
results to curves which may not lie in star domains.

Example 9.1.1. It is easy to see that the domain D = {z : 2 < |z| < 6 and Imz > 0} is the part of
the annulus {z : 2 < |z| < 6} in the upper half plane. It is not difficult to check that D is not a star
domain.

C
-6 -2 2 6

Let

D1 = {z : 2 < |z| < 6 and Imz > 0 and Rez < 1}

and

D2 = {z : 2 < |z| < 6 and Imz > 0 and Rez > −1}.
9–2 W W L Chen : Introduction to Complex Analysis

It is not difficult to see that D = D1 ∪ D2 , and that D1 and D2 are star domains with star centres
−2 + 5i and 2 + 5i respectively.

D1 D2

C1 C2
-6 -2 2 6

If C is a simple closed contour lying in D, then by introducing line segments along the imaginary axis,
it is not difficult to see that there is a simple closed contour C1 in D1 and a simple closed contour C2 in
D2 such that for any function f analytic in D, we have




f (z) dz = f (z) dz + f (z) dz.
C C1 C2

Note now that we can apply Theorem 5D to the two integrals on the right hand side.

However, it is of great theoretical interest to formulate results that are less restricting. Here we
introduce the idea of a simply connected domain. This can be done in a number of ways. Here we use
the Jordan curve theorem for simple closed polygons.

Definition. By a polygonal curve, we mean a curve ζ : [A, B] → C which is continuous and piecewise
linear. In other words, there exists a dissection

A = A1 < B1 = A2 < B2 = . . . = Ak < Bk = B

such that for every j = 1, . . . , k, the edge ζ : [Aj , Bj ] → C is of the form ζ(t) = αj t+βj , where αj , βj ∈ C
and we assume further that αj = 0.

Definition. By a simple closed polygon, we mean a polygonal curve that is closed and does not
intersect itself; in other words, if ζ(t1 ) = ζ(t2 ) whenever t1 = t2 , with the one exception that ζ(A) = ζ(B).
 Chapter 9 : Cauchy’s Integral Theorem Revisited 9–3

It is not hard to prove the Jordan curve theorem for a simple closed polygon, that such a polygon
divides the plane C into two domains, the bounded one interior to the polygon and the unbounded one
exterior to the polygon. Here we shall not give the proof.

Definition. A domain D is said to be simply connected if the interior of every simple closed polygon
in D is contained in D.

Remark. Recall that a domain is an open connected set. A simply connected domain is one which is
free of holes or cuts in its interior.

D D

simply connected not simply connected

9.2. Cauchy’s Integral Theorem

In this section, we indicate the proof of the following generalization of Theorem 5B.

THEOREM 9A. Suppose that a function f is analytic in a simply connected domain D. Then there
exists a function F , analytic in D and such that F
(z) = f (z) for every z ∈ D.

In view of Remark (1) immediately after Theorem 4A, Theorem 9A immediately leads to the
following generalization of Theorem 5D.

THEOREM 9B. Suppose that a function f is analytic in a simply connected domain D. Suppose
further that C is a closed contour lying in D. Then


f (z) dz = 0.
C

Example 9.2.1. The punctured plane D = {z : z = 0} is not simply connected. Although the function
f (z) = 1/z is analytic in D,


dz
= 2πi = 0
C z

if C is the positive (anticlockwise) oriented unit circle centred at 0. This also shows that there is no single
valued branch of log z in D, and confirms that the condition that D is simply connected in Theorems
9A and 9B is essential.
9–4 W W L Chen : Introduction to Complex Analysis

The proof of Theorem 9A depends on a process known sometimes as “triangulation”. It can be

shown by induction that a simple closed polygon of k + 2 sides can be decomposed by diagonals into a
set of k triangles.

We shall first prove the following generalization of Theorem 5A.

THEOREM 9C. Suppose that a function f is analytic in a simply connected domain D. Suppose
further that C is a simple closed polygon in D. Then


f (z) dz = 0.
C

Proof. We may assume, without loss of generality, that C is in the positive (anticlockwise) direction.
Suppose that the triangulation process gives rise to triangles T1 , . . . , Tk , with boundaries C1 , . . . , Ck in
the positive (anticlockwise) direction. Then it is easy to see that




f (z) dz = f (z) dz + . . . + f (z) dz.
C C1 Ck

The result follows on applying Theorem 5A to each of the integrals on the right hand side.

We now sketch a proof of Theorem 9A. Suppose that z0 ∈ D is fixed. For every z ∈ D, let C1 and
C2 denote polygonal curves from z0 to z that lie entirely in D.

z
D
C2

z0

C1

We shall first of all indicate that

(1) f (z) dz = f (z) dz.
C1 C2

Let C be the closed polygonal curve obtained by C1 followed by −C2 . To show (1), it suffices to show
that


(2) f (z) dz = 0.
C
 Chapter 9 : Cauchy’s Integral Theorem Revisited 9–5

It can be shown by induction that the closed polygonal curve C consists of a number of line segments
followed in opposite directions and a number of simple closed polygonal curves. (2) now follows in view
of Theorem 9C. Note that (1) shows that the integral is independent of the polygonal curve chosen. We
can therefore define

z
F (z) = f (ζ) dζ,
z0

where the integral is taken along any polygonal curve from z0 to z. It follows that if |h| is sufficiently
small, then the segment [z, z + h] lies entirely in D, and that


F (z + h) − F (z) = f (ζ) dζ.
[z,z+h]

The proof of Theorem 9A is now completed in the same way as in the proof of Theorem 5B.

9.3. Cauchy’s Integral Formula

Suppose that f is analytic in a simply connected domain D. Suppose further that C is a closed contour
in D, and that the point z does not lie on C. If z ∈ D, then the function
f (ζ) − f (z)
g(ζ) =
ζ −z
is analytic in D, apart from a removable singularity at ζ = z. Furthermore, this singularity is removed
by defining g(z) = f
(z). It now follows from Theorem 9B that




f (ζ) − f (z) f (ζ) dz
dz = 0, and so dz = f (z) .
C ζ − z C ζ − z C ζ −z
Here


1 dz var(i arg(ζ − z), C)
n(C, z) = =
2πi C ζ −z 2πi
is the winding number, and counts the number of times the contour C winds round the point z in the
positive (anticlockwise) direction. Hence


1 f (ζ)
(3) dz = n(C, z)f (z).
2πi C ζ − z
Note also that if z ∈ D, then both sides of (3) are 0. It follows that (3) holds whenever z ∈ C.

9.4. Analytic Logarithm

Suppose that f is analytic and non-zero in a simply connected domain D. Let z0 ∈ D be fixed. For
every z ∈ D, we can define

z

f (ζ)
(4) g(z) = dζ + Log f (z0 ),
z0 f (ζ)

where the integral is over any contour in D from z0 to z, and is independent of the choice of the contour,
in view of Theorem 9B. Differentiating (4) with respect to z, we obtain g
(z)f (z) = f
(z). It follows that
d −g(z)
(e f (z)) = e−g(z) f
(z) − g
(z)e−g(z) f (z) = 0,
dz
9–6 W W L Chen : Introduction to Complex Analysis

so that e−g(z) f (z) is contant in D. Since e−g(z0 ) f (z0 ) = 1, it follows that e−g(z) f (z) = 1 for every z ∈ D.
In other words, f (z) = eg(z) for every z ∈ D, so that f has an analytic logarithm in D.

Problems for Chapter 9

1. Suppose that T is a triangle, followed in the positive (anticlockwise) direction. Suppose further that
for any a ∈ C \ T , we write


1 dz
n(T, a) = .
2πi T z − a
a) Show that n(T, a) = 0 for every a ∈ C outside T .
b) Suppose now that a ∈ C is inside T . By relating T to a suitably small circular path centred at
the point a, show that n(T, a) = 1.

2. Suppose that C is simple closed polygon, followed in the positive (anticlockwise) direction. Suppose
further that for any a ∈ C \ C, we write


1 dz
n(C, a) = .
2πi C z − a
a) Show that n(C, a) = 0 for every a ∈ C outside C.
b) Suppose now that a ∈ C is inside C. Apply the “triangulation” process to C and show that
n(C, a) = 1 if the point a does not lie on the boundary of any of the triangles that arise from
the process.
c) Suppose now that a ∈ C is inside C and lies on the boundary of some of the triangles that arise
from the “triangulation” process. Explain why we also have n(C, a) = 1.

3. Suppose that D ⊆ C is a domain. For every contour C lying in D and for every a ∈ D, write


1 dz
n(C, a) = .
2πi C z − a
a) Suppose that D is simply connected. Explain why n(C, a) = 0.
b) Suppose that D is not simply connected. Show that there exists a contour C lying in D and a
point a ∈ D such that n(C, a) = 0.

4. Deduce from the conclusion of Problem 3 that every star domain is simply connected.

5. Suppose that D ⊆ C is a domain. Suppose further that every function which is analytic and non-
zero in D has an analytic logarithm in D, so that in particular, for every a ∈ D, there exists a
function g(z) analytic in D and such that z − a = eg(z) for every z ∈ D. Use the conclusion of
Problem 3 to show that D is simply connected.

6. Suppose D is a bounded domain. For any  > 0, let D
denote the larger domain containing D
and every point in C whose distance from D is less than . Give an example of a simply connected
domain D such that D
is not simply connected for any  > 0.

7. Suppose that f (z) is an entire function with a finite number of zeros. Show that there exist a
polynomial P (z) and an entire function g(z) such that f (z) = P (z)eg(z) .

8. Suppose that f (z) is analytic and non-zero in the disc {z : |z| ≤ R}, where R > 0 is fixed. Prove
the following special case of Jensen’s formula, that

2π
1
log |f (0)| = log |f (Reiθ | dθ.
2π 0
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 10
RESIDUE THEORY

10.1. Cauchy’s Residue Theorem

If we extend Cauchy’s integral theorem to functions having isolated singularities, then the integral is in
general not equal to zero. Instead, each singularity contributes a term called the residue. Our principal
aim in this section is to show that this residue depends only on the coefficient of (z −z0 )−1 in the Laurent
expansion of the function near the singularity z0 , since all the other powers of z − z0 has single valued
integrals and so integrate to zero.

Definition. By a simple closed contour or Jordan contour, we mean a contour ζ : [A, B] → C such
that ζ(t1 ) = ζ(t2 ) whenever t1 = t2 , with the one exception ζ(A) = ζ(B).

THEOREM 10A. Suppose that a function f is analytic in a simply connected domain D, except for
an isolated singularity at z0 , and that

−1


f1 (z) = an (z − z0 )n
n=−∞

is the principal part of f at z0 . Suppose further that C is a Jordan contour in D followed in the positive
(anticlockwise) direction and not passing through z0 . Then

 
1 a−1 if z0 lies inside C,
f1 (z) dz =
2πi C 0 if z0 lies outside C.
10–2 W W L Chen : Introduction to Complex Analysis

Proof. Suppose first of all that z0 is outside C. Then z0 is in the exterior domain of C which also
contains the point at ∞. It follows that z0 can be joined to the point at ∞ by a simple polygonal curve
L, as shown in the picture below.

z0

The Jordan contour C is clearly contained in the simply connected domain obtained when L is deleted
from the complex plane. In fact, it is contained in a simply connected domain which is a subset of D \ L,
as shown by the shaded part in the picture above. Clearly f is analytic in this simply connected domain,
so it follows from Theorem 9B that

f (z) dz = 0.
C

Suppose next that z0 is inside C. Then there exists r > 0 such that the closed disc {z : |z − z0 | ≤ r} is
inside C. Let γ denote the boundary of this disc, followed in the positive (anticlockwise) direction.

z0

We now draw a horizontal line through the point z0 . Following this line to the left from z0 , it first
intersects γ and then C (for the first time). Draw a line segment joining these two intersection points.
Similarly, following this line to the right from z0 , it first intersects γ and then C (for the first time).
Again draw a line segment joining these two intersection points. Note that these two line segments are
inside C and outside γ. We now divide C into two parts by cutting it at the two intersection points
 Chapter 10 : Residue Theory 10–3

mentioned. It can be shown that one part of this, together with the part of γ above the horizontal line
and the two line segments, gives rise to a simple closed contour C + followed in the positive direction
and which can be shown to lie in a simply connected domain lying in D but not containing z0 . Clearly
f is analytic in this simply connected domain, so that

f (z) dz = 0,
C+

in view of Theorem 9B.

C+
z0

Similarly, the other part of C, together with the part of γ below the horizontal line and the two line
segments, gives rise to a simple closed contour C − followed in the positive direction and which again can
be shown to lie in a simply connected domain lying in D but not containing z0 . Clearly f is analytic in
this simply connected domain, so that

f (z) dz = 0.
C−

It is easily seen that

   
f (z) dz − f (z) dz = f (z) dz + f (z) dz,
C γ C+ C−

so that
 
f (z) dz = f (z) dz.
C γ

By Theorem 8E, we have


f (z) dz = 2πia−1 .
γ

It follows that

f (z) dz = 2πia−1 .
C

Finally, note that f2 (z) = f (z) − f1 (z) is analytic in D, so that


f2 (z) dz = 0,
C
10–4 W W L Chen : Introduction to Complex Analysis

whence
 
f (z) dz = f1 (z) dz.
C C

The result follows.

Definition. The value a−1 in Theorem 10A is called the residue of the function f at z0 , and denoted
by res(f, z0 ).

We are now in a position to state and prove a simple version of Cauchy’s residue theorem.

THEOREM 10B. Suppose that the function f is analytic in a simply connected domain D, except
for isolated singularities at z1 , . . . , zk . Suppose further that C is a Jordan contour in D followed in the
positive (anticlockwise) direction and not passing through z1 , . . . , zk . Then


k
1
f (z) dz = res(f, zj ).
2πi C j=1
zj inside C

Proof. For every j = 1, . . . , k, let fj (z) denote the principal part of f (z) at zj . By Theorem 8D, fj
is analytic in C except at zj . It follows that the function

k
g(z) = f (z) − fj (z)
j=1

is analytic in D, so that

g(z) dz = 0
C

by Theorem 9B, and so

 k 


f (z) dz = fj (z) dz.
C j=1 C

The result now follows from Theorem 10A.

10.2. Finding the Residue

In order to use Theorem 10B to evaluate the integral


f (z) dz,
C

we need a technique to evaluate the residues at the isolated singularities.

Suppose that f (z) has a removable singularity at z0 . Then f (z) has a Taylor series expansion which
is valid in a neighbourhood of z0 . The residue is clearly 0.

Suppose that f (z) has a simple pole at z0 . Then we can write

a−1
f (z) = + g(z),
z − z0
 Chapter 10 : Residue Theory 10–5

where g(z) is analytic at z0 , so that (z − z0 )g(z) → 0 as z → z0 . It follows that the residue is given by

a−1 = lim (z − z0 )f (z).

z→z0

Suppose that f (z) has a pole of order m at z0 . Then we can write

a−m a−m+1 a−1
f (z) = + + ... + + g(z),
(z − z0 )m (z − z0 )m−1 z − z0

where g(z) is analytic at z0 , so that

(z − z0 )m f (z) = a−m + a−m+1 (z − z0 ) + . . . + a−1 (z − z0 )m−1 + (z − z0 )m g(z)

is analytic at z0 . Differentiating m − 1 times gives

dm−1 dm−1
((z − z 0 )m
f (z)) = a−1 (m − 1)! + ((z − z0 )m g(z)).
dz m−1 dz m−1
Since g(z) is analytic at z0 , we have

dm−1
lim ((z − z0 )m g(z)) = 0.
z→z0 dz m−1

It follows that the residue is given by

1 dm−1
a−1 = lim ((z − z0 )m f (z)).
(m − 1)! z→z0 dz m−1

Definition. A function is said to be meromorphic in a domain D if it is analytic in D except for poles.

Example 10.2.1. The function

e2iz
f (z) =
1 + 4z 2

has simple poles at z = ±i/2, with residues

   
i i e2iz e−1
res f, = lim z− f (z) = lim =
2 z→i/2 2 z→i/2 4(z + i/2) 4i

and
   
i i e2iz e
res f, − = lim z+ f (z) = lim =− .
2 z→−i/2 2 z→−i/2 4(z − i/2) 4i

i/2

1
-i/2
C
10–6 W W L Chen : Introduction to Complex Analysis

It follows from Cauchy’s residue theorem that if C = {z : |z| = 1} is the circle with centre 0 and radius
1, followed in the positive (anticlockwise) direction, then
    
e2iz e−1 e π 1
dz = 2πi − = −e .
C 1 + 4z 2 4i 4i 2 e

Example 10.2.2. The function

ez
f (z) =
z4
has a pole of order 4 at z = 0, with residue

1 d3 1 d3 1
res(f, 0) = lim 3 (z 4 f (z)) = lim 3 ez = .
3! z→0 dz 3! z→0 dz 6

It follows from Cauchy’s residue theorem that if C is any Jordan contour with 0 inside and followed in
the positive (anticlockwise) direction, then
  
ez 1 πi
dz = 2πi = .
C z4 6 3

Example 10.2.3. Suppose that a function f is analytic in a simply connected domain D, and that
z0 ∈ D. Suppose further that C is a Jordan contour in D, followed in the positive (anticlockwise)
direction and with z0 inside. If f (z0 ) = 0, then the function

f (z)
F (z) =
z − z0

has a simple pole at z0 , with residue

lim (z − z0 )F (z) = f (z0 ).

z→z0

Applying Cauchy’s residue theorem, we obtain Cauchy’s integral formula


1 f (z)
dz = f (z0 ).
2πi C z − z0

If f (z0 ) = 0, then F (z) has a removable singularity at z0 . The same result follows instead from Cauchy’s
integral theorem.

10.3. Principle of the Argument

In this section, we shall show that the residue theorem, when applied suitably, can be used to find the
number of zeros of an analytic function, as well as the number of zeros minus the number of poles of a
meromorphic function.

The main idea underpinning our discussion can be summarized by the following two results.

THEOREM 10C. Suppose that a function f is analytic in a neighbourhood of z0 . Suppose further

that f has a zero of order m at z0 . Then the function f  /f is analytic in a punctured neighbourhood of
z0 , with a simple pole at z0 with residue m.
 Chapter 10 : Residue Theory 10–7

THEOREM 10D. Suppose that a function f is analytic in a punctured neighbourhood of z0 . Sup-

pose further that f has a pole of order m at z0 . Then the function f  /f is analytic in a punctured
neighbourhood of z0 , with a simple pole at z0 with residue −m.

Proof of Theorem 10C. We can write f (z) = (z − z0 )m g(z), where g(z) is analytic in a neighbour-
hood of z0 and g(z0 ) = 0. Then

f  (z) m(z − z0 )m−1 g(z) + (z − z0 )m g  (z) m g  (z)

= = + .
f (z) (z − z0 ) g(z)
m z − z0 g(z)

Since g(z) is analytic in a neighbourhood of z0 and g(z0 ) = 0, the function g  (z)/g(z) is analytic in a
neighbourhood of z0 . The result follows.

Proof of Theorem 10D. We can write f (z) = (z − z0 )−m g(z), where g(z) is analytic in a neigh-
bourhood of z0 and g(z0 ) = 0. Then

f  (z) −m(z − z0 )−m−1 g(z) + (z − z0 )−m g  (z) −m g  (z)

= = + .
f (z) (z − z0 )−m g(z) z − z0 g(z)

Since g(z) is analytic in a neighbourhood of z0 and g(z0 ) = 0, the function g  (z)/g(z) is analytic in a
neighbourhood of z0 . The result follows.

The main result in this section is the Principle of the argument, as stated below.

THEOREM 10E. Suppose that a function f is meromorphic in a simply connected domain D. Sup-
pose further that C is a Jordan curve in D, followed in the positive (anticlockwise) direction, and that
f has no zeros or poles on C. If N denotes the number of zeros of f in the interior of C, counted with
multiplicities, and if P denotes the number of poles of f in the interior of C, counted with multiplicities,
then

1 f  (z)
dz = N − P.
2πi C f (z)

Proof. Note that by Theorems 10C and 10D, the poles of the function f  /f are precisely at the zeros
and poles of f . Furthermore, a zero of f of order m gives rise to a residue m for f  /f , so that the residues
of f  /f arising from the zeros of f are equal to the number of zeros of f counted with multiplicities, and
this number is N . On the other hand, a pole of f of order m gives rise to a residue −m for f  /f , so
that the residues of f  /f arising from the poles of f are equal to minus the number of poles of f counted
with multiplicities, and this number is P . It follows that the sum of the residues is equal to N − P . The
result now follows from Theorem 10B applied to the function f  /f .

Remarks. (1) Note that


1 f  (z) 1 1 1
dz = var(log f (z), C) = var(i arg f (z), C) = var(arg f (z), C).
2πi C f (z) 2πi 2πi 2π

It follows that the conclusion of Theorem 10E can be expressed in the form
1
N −P = var(arg f (z), C),
2π
in terms of the variation of the argument of f (z) along the Jordan curve C.

(2) Note also that

 
1 f  (z) 1 dw
dz = = n(f (C), 0).
2πi C f (z) 2πi f (C) w
10–8 W W L Chen : Introduction to Complex Analysis

(3) Theorem 10E can be generalized in the following way. Suppose that a function f is meromor-
phic in a simply connected domain D, and that all its zeros and poles in D are simple. Suppose further
that C is a Jordan curve in D, followed in the positive (anticlockwise) direction, and that f has no zeros
or poles on C. If a1 , . . . , aN denote the zeros of f in the interior of C, and if b1 , . . . , bP denote the poles
of f in the interior of C, then


N
P
1 f  (z)
(1) g(z) dz = g(aj ) − g(bk )
2πi C f (z) j=1 k=1

for every function g analytic in D. To see this, simply note that any simple zero or simple pole z0 of f ,
where g(z0 ) = 0, gives rise to a simple pole of (f  /f )g with residue

f  (z) f  (z) g(z0 ) if z0 is a simple zero of f ,
lim (z − z0 ) g(z) = g(z0 ) lim (z − z0 ) =
z→z0 f (z) z→z0 f (z) −g(z0 ) if z0 is a simple pole of f ;

on the other hand, if g(z0 ) = 0, then (f  /f )g has a removable singularity at z0 . In fact, (1) remains
valid if the zeros and poles of f are of higher order, provided that all zeros and poles are counted with
multiplicities. Note also that the choice g(z) = 1 in D gives Theorem 10E again. A particular useful
choice of f is given by the entire function f (z) = sin πz, with simple zeros at every n ∈ Z. Since

f  (z) π cos πz
= = π cot πz,
f (z) sin πz

it follows from (1) that



1
(2) g(z)π cot πz dz = g(n)
2πi C n inside C

for every function g analytic in D. This may be used to obtain a variety of infinite series expansions.
See Chapter 16.

Example 10.3.1. To find the number of zeros of the function f (z) = z 4 + z 3 − 2z 2 + 2z + 4 in the first
quadrant of the complex plane, we use the Jordan curve C = C1 ∪ C2 ∪ C3 , where C1 = [0, R] is the
straight line segment along the real axis from 0 to R, C2 is the circular path ζ : [0, π/2] → C, given by
ζ(t) = Reit , and C3 = [iR, 0] is the straight line segment along the imaginary axis from iR to 0. Here R
is taken to be a large positive real number.

C2

C3

C1 R

On C1 , we have z = x > 0, so that


x4 + x3 + 4 if 0 ≤ x ≤ 1
f (z) = f (x) = x4 + x3 − 2x2 + 2x + 4 ≥
2x + 4 if x ≥ 1

is clearly positive, so that var(arg f (z), C1 ) = 0. Next, note that

 
z 3 − 2z 2 + 2z + 4
f (z) = z 4 1 + .
z4
 Chapter 10 : Residue Theory 10–9

On C2 , we have |z| = R, so that

 3 
 z − 2z 2 + 2z + 4  R3 + 2R2 + 2R + 4 2R3 2
 ≤ < 4 =
 z 4  R 4 R R

whenever R > 8, say. It follows that on C2 when R is large enough, we have f (z) = R4 e4it (1 + w), where
|w| < 2/R, so that var(arg f (z), C2 ) = 2π + 1 , where 1 → 0 as R → ∞. Finally, on C3 , we have z = iy,
where y > 0, so that

f (z) = f (iy) = (y 4 + 2y 2 + 4) + i(2y − y 3 ) = (y 2 + 1)2 + 3 + i(2y − y 3 ).

Note that Ref (iy) > 0, so that f (iy) is in the first or fourth quadrant of the complex plane. In fact,
when R > 0 is large, f (iR) is much nearer the real axis than the imaginary axis, while f (0) = 4 is on the
positive real axis. It follows that var(arg f (z), C3 ) = 2 , where 2 → 0 as R → ∞. We now conclude that
var(arg f (z), C) = 2π + 1 + 2 , where 1 , 2 → 0 as R → ∞. On the other hand, C is a closed contour,
so that var(arg f (z), C) must be an integer multiple of 2π. It follows that var(arg f (z), C) = 2π. Note
now that the function f has no poles in the first quadrant. It follows from the Argument principle that
f has exactly one zero inside the contour C for all large R. Hence f has exactly one zero in the first
quadrant of the complex plane.

To find the number of zeros in a region, the following result provides an opportunity to either bypass
the Argument principle or at least enable one to apply the Argument principle to a simpler function.
Needless to say, the proof is based on an application of the Argument principle.

THEOREM 10F. (ROUCHÉ’S THEOREM) Suppose that functions f and g are analytic in a sim-
ply connected domain D, and that C is a Jordan contour in D. Suppose further that |f (z)| > |g(z)| on
C. Then f and f + g have the same number of zeros inside C.

We shall give two proofs of this result. The first is the one given in most texts.

First Proof of Theorem 10F. Consider the function

f (z) + g(z)
F (z) = .
f (z)

The condition |f (z)| > |g(z)| on C ensures that both f and f + g have no zeros on C. On the other
hand, note that
 
 g(z) 
(3) 
|F (z) − 1| =  <1
f (z) 

for every z ∈ C. By Remark (2) after Theorem 10E, we have

 
1 F  (z) 1 dw
dz = = n(F (C), 0).
2πi C F (z) 2πi F (C) w

In view of (3), the closed contour F (C) is contained in the open disc {w : |w − 1| < 1} with centre 1
and radius 1. This disc does not contain the point 0, so that n(F (C), 0) = 0. Hence

1 F  (z)
dz = 0.
2πi C F (z)

It follows from the Argument principle that the function F has the same number of zeros and poles
inside C. Note now that the poles of F are precisely the zeros of f , and the zeros of F are precisely the
zeros of f + g.

10–10 W W L Chen : Introduction to Complex Analysis

Second Proof of Theorem 10F. For every τ ∈ [0, 1], let


1 f  (z) + τ g  (z)
N (τ ) = dz.
2πi C f (z) + τ g(z)

The condition |f (z)| > |g(z)| on C ensures that

|f (z) + τ g(z)| ≥ |f (z)| − τ |g(z)| ≥ |f (z)| − |g(z)| > 0

on C, so that f + τ g does not have any zeros (or poles) on C. In fact, there is a positive lower bound
for |f (z) + τ g(z)| on C independent of τ . It follows easily from this that N (τ ) is continuous in [0, 1]. By
the Argument principle, N (τ ) is an integer for every τ ∈ [0, 1]. Hence N (τ ) must be constant in [0, 1].
In particular, we must have N (0) = N (1). Clearly, N (0) is the number of zeros of f inside C, and N (1)
is the number of zeros of f + g inside C.

Example 10.3.2. To determine the number of solutions of ez = 2z + 1 with |z| < 1, we write

f (z) = −2z and g(z) = ez − 1,

so that f (z) + g(z) = ez − 2z − 1. We therefore need to find the number of zeros of f + g inside the unit
circle C = {z : |z| = 1}. Clearly, f has precisely one zero, at z = 0, inside C. On the other hand, note
that
  1
e −1=
z ζ
e dζ = ezt z dt.
[0,z] 0

If z ∈ C, then |ezt | ≤ et , and so

 1  1
|g(z)| = |e − 1| ≤
z
|e z| dt ≤
zt
et dt = e − 1.
0 0

Since |f (z)| = 2 whenever z ∈ C, it follows that |f (z)| > |g(z)| on C. By Rouché’s theorem, f + g has
precisely one zero inside C.

Problems for Chapter 10

1. a) Write down the Taylor series for ew about the origin w = 0.

2
b) Using the substitution w = 1/z 2 in (a), find the Laurent series for the function e1/z about the
origin z = 0.
2
c) Find the residue of the function e1/z at the origin z = 0.
2
d) What type of singularity does the function e1/z have at the origin z = 0?

2. Suppose that f (z) = g(z)/h(z), where the functions g(z) and h(z) are analytic at z0 . Suppose
further that g(z0 ) = 0 and h(z) has a simple zero at z0 . Use l’Hopital’s rule to show that

g(z) g(z0 )
res(f, z0 ) = lim =  .
z→z0 h (z) h (z0 )
 Chapter 10 : Residue Theory 10–11

3. For each of the functions f (z) given below, find all the singularities in C, find the residues at these
singularities, and evaluate the integrals
 
f (z) dz and f (z) dz,
C
C

where C  and C  are circular paths centred at the origin z = 0, of radius 1/2 and 2 respectively,
followed in the positive (anticlockwise) direction:
1 z z3 + 2
a) f (z) = b) f (z) = 4 c) f (z) = 4
z(z − 1) z +1 (z − 1)(z + 1)

4. Suppose that C is a circular path centred at the origin z = 0, of radius 1, followed in the positive
(anticlockwise) direction. Show each of the following:
  z
eπz e
a) 2+1
dz = πi; b) 3
dz = πi.
C 4z C z

5. Find the number of zeros of f (z) = z 4 + z 3 + 5z 2 + 2z + 4 in the first quadrant of the complex plane.
Find also the number of zeros of the function in the fourth quadrant.

6. Consider the equation 2z 5 + 8z − 1 = 0.

a) Writing f (z) = 2z 5 and g(z) = 8z − 1, use Rouché’s theorem to show that all the roots of this
equation lie in the open disc {z : |z| < 2}.
b) Writing f (z) = 8z − 1 and g(z) = 2z 5 , use Rouché’s theorem to show that this equation has
exactly one root in the open disc {z : |z| < 1}.
c) How many roots does this equation have in the open annulus {z : 1 < |z| < 2}? Justify your
assertion.

7. Show that the equation z 6 + 4z 2 = 1 has exactly two roots in the open disc {z : |z| < 1}.
[Hint: Use Rouché’s theorem. You will need to make a good choice for f (z) and g(z). Do not give
up if your first guess does not work.]
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 11
EVALUATION OF DEFINITE INTEGRALS

11.1. Introduction

The calculus of residues often provides an efficient method for evaluating certain real and complex
integrals. This is particularly important when it is not possible to find indefinite integrals explicitly.
Even in cases when ordinary methods of calculus can be applied, the use of residues often proves to be
a labour saving device.

Naturally, the calculus of residues gives rise to complex integrals, and this suggests that we may
be at a disadvantage if we want to evaluate real integrals. In practice, this is seldom the case, since a
complex integral is equivalent to two real integrals.

However, there are limitations to this approach. The integrand must be closely associated with some
analytic function. We usually want to integrate some elementary functions, and these can be extended
to the complex domain. Also, the techniques of complex integration applies to closed curves while a real
integral is over an interval. It follows that we need a device to reduce our problem to one which concerns
integration over closed curves. There are a number of ways to achieve this, depending on circumstances.
The technique is best learned by studying typical examples, and complete mastery does not guarantee
success.

11.2. Rational Functions over the Unit Circle

We shall be concerned mainly with integrals of the type

2π
f (cos θ, sin θ) dθ,
0
11–2 W W L Chen : Introduction to Complex Analysis

where f (x, y) is a real valued rational function in the real variables x and y.

If we use the substitution

z = eiθ = cos θ + i sin θ and dz = ieiθ dθ,

then
1 dz
= cos θ − i sin θ and dθ = −i .
z z
We can therefore write
   
1 1 1 1
cos θ = z+ and sin θ = z− ,
2 z 2i z

so that

2π
    
1 1 1 1 dz
f (cos θ, sin θ) dθ = −i f z+ , z− ,
0 C 2 z 2i z z

where C is the unit circle {z : |z| = 1}, followed in the positive (anticlockwise) direction.

Example 11.2.1. Suppose that the real number a > 1 is fixed. Consider the integral

π

dθ 1 2π dθ
= .
0 a + cos θ 2 0 a + cos θ

Using the substitution z = eiθ = cos θ + i sin θ, we have

2π


dθ dz 2dz
= −i 1 1 = −i 2 + 2az + 1
,
0 a + cos θ C z(a + 2 (z + z )) C z

where C is the unit circle {z : |z| = 1}, followed in the positive (anticlockwise) direction. It follows that

π

dθ dz
= −i 2 + 2az + 1
.
0 a + cos θ C z

If we factorize the denominator z 2 + 2az + 1, we obtain roots

 
α = −a + a2 − 1 and β = −a − a2 − 1.

Clearly |β| > 1. Since αβ = 1, it follows that |α| < 1. Hence the function
1
z 2 + 2az + 1
 Chapter 11 : Evaluation of Definite Integrals 11–3

is analytic in some simply connected domain containing the unit circle C, except for a simple pole at
z = α inside C, with residue
 
1 1 1 1 1
res ,α = lim (z − α) = lim = = √ .
2
z + 2az + 1 z→α z2 + 2az + 1 z→α z−β α−β 2 a2 − 1

It follows from Cauchy’s residue theorem that

 
dz 1 2πi
= 2πi res ,α = √ ,
C z 2 + 2az + 1 z 2 + 2az + 1 2 a2 − 1

and so

π
dθ π
=√ .
0 a + cos θ a −1
2

Example 11.2.2. Now let w ∈ C \ [−1, 1], and consider the integral

π
dθ
F (w) = .
0 w + cos θ

Note that we have excluded the closed interval [−1, 1] to ensure that the denominator of the integrand
does not vanish. One can show that F  (w) exists in the domain C \ [−1, 1], so that F (w) is analytic
there. We know from Example 11.2.1 that
π
(1) F (w) =
(w2 − 1)1/2

on the real axis to the right of the point w = 1. It follows from Theorem 7H that (1) holds for every
w ∈ C \ [−1, 1]. Note, however, that the branch of the square root must be chosen so that it is positive
for w > 1.

11.3. Rational Functions over the Real Line

We shall be concerned mainly with integrals of the type

∞
(2) f (x) dx,
−∞

where f (x) is a real valued rational function in the real variable x. Here we shall assume that the degree
of the denominator of f exceeds the degree of the numerator of f by at least 2, and that f has no poles
on the real line, so that the integral (2) is convergent.

Consider first of all the integral

R
f (x) dx,
−R

where R > 0. We then extend the definition of the rational function f to the complex domain, and
consider also the integral


f (z) dz,
CR
11–4 W W L Chen : Introduction to Complex Analysis

where CR is the semicircular arc given by z = Reit , where t ∈ [0, π]. Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where [−R, R] denotes the line segment from −R to R.

CR

-R R

By Cauchy’s residue theorem, we have

R

f (x) dx + f (z) dz = 2πi res(f, zi ),
−R CR zi inside C

where the summation is taken over all the poles of f inside the Jordan contour C. It is easily shown
that


f (z) dz → 0
CR

as R → ∞, so that

∞ 
f (x) dx = 2πi res(f, zi ),
−∞ Imzi >0

where the summation is taken over all the poles of f in the upper half plane.

Example 11.3.1. Suppose that the real number a > 0 is fixed. Consider the integral

∞
x2
2 2 3
dx.
−∞ (x + a )

To evaluate this integral, note that the rational function

z2
f (z) =
(z 2 + a2 )3
has poles of order 3 at z = ±ia. Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where R > a.

CR

ia

-R -ia R
 Chapter 11 : Evaluation of Definite Integrals 11–5

By Cauchy’s residue theorem, we have

R

f (x) dx + f (z) dz = 2πi res(f, ia).
−R CR

Since
   
1 d2 z2 1 2 12z 12z 2
res(f, ia) = lim (z − ia)3
= lim − +
2 z→ia dz 2 (z 2 + a2 )3 2 z→ia (z + ia)3 (z + ia)4 (z + ia)5
 
1 2 12ia 12a2 i
= 3
− 4
− 5
=− ,
2 (2ia) (2ia) (2ia) 16a3

it follows that

R

π
f (x) dx + f (z) dz = .
−R CR 8a3

Note now that



  R2
 f (z) dz  ≤ πR → 0
 (R2 − a2 )3
CR

as R → ∞. Hence

∞
∞
x2 π
dx = f (x) dx = .
−∞ (x2 + a2 )3 −∞ 8a3

Example 11.3.2. Consider the integral

∞
x2 + 3
dx.
−∞ x4 + 5x2 + 4

To evaluate this integral, note that the rational function

z2 + 3 z2 + 3
f (z) = = 2
z4 2
+ 5z + 4 (z + 1)(z 2 + 4)

has simple poles at z = ±i and z = ±2i. Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where R > 2.

CR
2i
i

-R R
-i
-2i

By Cauchy’s residue theorem, we have

R

f (x) dx + f (z) dz = 2πi(res(f, i) + res(f, 2i)).
−R CR
11–6 W W L Chen : Introduction to Complex Analysis

By Problem 2 in Chapter 10, we have

z2 + 3 1 z2 + 3 1
res(f, i) = lim 3
= and res(f, 2i) = lim 3
= .
z→i 4z + 10z 3i z→2i 4z + 10z 12i
It follows that

R

5π
f (x) dx + f (z) dz = .
−R CR 6

Note now that



  R2 + 3
 f (z) dz  ≤ 4 πR → 0
 R − 5R2 − 4
CR

as R → ∞. Hence

∞
∞
x2 + 3 5π
dx = f (x) dx = .
−∞ x4 + 5x2 + 4 −∞ 6

11.4. Rational and Trigonometric Functions over the Real Line

We shall be concerned mainly with integrals of the type

∞
(3) f (x)eix dx,
−∞

where f (x) is a real valued rational function in the real variable x. Here we shall assume that the degree
of the denominator of f exceeds the degree of the numerator of f by at least 2, and that f has no poles
on the real line, so that the integral (3) is convergent. Note that the real and imaginary parts of the
integral (3) are respectively

∞
∞
f (x) cos x dx and f (x) sin x dx.
−∞ −∞

Consider first of all the integral

R
f (x)eix dx,
−R

where R > 0. We consider also the integral

f (z)eiz dz,
CR

where CR is the semicircular arc given by z = Reit , where t ∈ [0, π].

CR

-R R
 Chapter 11 : Evaluation of Definite Integrals 11–7

Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where [−R, R] denotes the line segment from −R to R. By Cauchy’s residue theorem, we have

R

ix
(4) f (x)e dx + f (z)eiz dz = 2πi res(f (z)eiz , zi ),
−R CR zi inside C

where the summation is taken over all the poles of f (z)eiz inside the Jordan contour C.

To study the second integral in (4), we prove the following estimate.

THEOREM 11A. (JORDAN’S LEMMA) Suppose that R > 0. Suppose further that CR is the
semicircular arc given by z = Reit , where t ∈ [0, π]. Then


(5) |eiz || dz| < π.
CR

Proof. Note that

π
π
iReit
(6) |e ||dz| =
iz
|e ||iRe | dt = R
it
|eiR(cos t+i sin t) | dt
CR 0 0

π
π/2
=R e−R sin t dt = 2R e−R sin t dt.
0 0

Since
2 π
sin t ≥ t whenever 0 ≤ t ≤ ,
π 2
it follows that

π/2
π/2
π π
(7) e−R sin t dt ≤ e−2Rt/π dt = (1 − e−R ) < .
0 0 2R 2R

The inequality (5) follows on combining (6) and (7).

It follows easily from Theorem 11A that

f (z)eiz dz → 0
CR

as R → ∞, so that

∞ 
f (x)eix dx = 2πi res(f (z)eiz , zi ),
−∞ Imzi >0

where the summation is taken over all the poles of f (z)eiz in the upper half plane.

Remark. In view of Jordan’s lemma, we may consider integrals of the form (3) where the degree of
the denominator of the rational function f exceeds the degree of the numerator of f by only 1. Note,
however, that the argument in this case only establishes the existence of the integral (3) as

R
lim f (x)eix dx.
R→∞ −R
11–8 W W L Chen : Introduction to Complex Analysis

Example 11.4.1. Suppose that the real number a > 0 is fixed. Consider the integral

∞
cos x
2 + a2
dx.
−∞ x
To evaluate this integral, note that the function
eiz
F (z) =
z2 + a2
has simple poles at z = ±ia. Consider now the Jordan contour
C = [−R, R] ∪ CR ,
where R > a.

CR

ia

-R -ia R

By Cauchy’s residue theorem, we have

R

F (x) dx + F (z) dz = 2πi res(F, ia).
−R CR

Since
 
eiz eiz e−a
res(F, ia) = lim (z − ia) 2 = lim = ,
z→ia z + a2 z→ia z + ia 2ia
it follows that



R
πe−a
F (x) dx + F (z) dz = .
−R CR a
Note now that



  1 π
 
F (z) dz  ≤ 2 |eiz ||dz| < 2 →0
 R − a2 R − a2
CR CR

as R → ∞. Hence

∞
πe−a
F (x) dx = ,
−∞ a
so that

∞
∞
cos x πe−a
dx = Re F (x) dx = .
−∞ x2 + a2 −∞ a

Example 11.4.2. Suppose that the real numbers a > 0 and b > 0 are fixed and different. Consider
the integral

∞
x3 sin x
dx.
−∞ (x2 + a2 )(x2 + b2 )
 Chapter 11 : Evaluation of Definite Integrals 11–9

To evaluate this integral, note that the function

z 3 eiz
F (z) =
(z 2 + a2 )(z 2 + b2 )

has simple poles at z = ±ia and z = ±ib. Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where R > max{a, b}.

CR
ia
ib

-R R
-ib
-ia

By Cauchy’s residue theorem, we have

R

F (x) dx + F (z) dz = 2πi(res(F, ia) + res(F, ib)).
−R CR

Since
 
z 3 eiz z 3 eiz a2 e−a
res(F, ia) = lim (z − ia) 2 = lim =
z→ia (z + a2 )(z 2 + b2 ) z→ia (z + ia)(z 2 + b2 ) 2(a2 − b2 )

and
 
z 3 eiz z 3 eiz b2 e−b
res(F, ib) = lim (z − ib) = lim = ,
z→ib (z + a2 )(z 2 + b2 )
2 z→ib (z 2 2
+ a )(z + ib) 2(b2 − a2 )

it follows that


 
R
a2 e−a b2 e−b
F (x) dx + F (z) dz = πi + .
−R CR a2 − b2 b2 − a2

Note now that




  R3 πR3
 F (z) dz  ≤ |e iz
||dz| < →0
 (R2 − a2 )(R2 − b2 ) CR (R2 − a2 )(R2 − b2 )
CR

as R → ∞. Hence

∞
πi(a2 e−a − b2 e−b )
F (x) dx = ,
−∞ a2 − b2

so that

∞
∞
x3 sin x π(a2 e−a − b2 e−b )
dx = Im F (x) dx = .
−∞ (x2 + a2 )(x2 + b2 ) −∞ a2 − b2
11–10 W W L Chen : Introduction to Complex Analysis

Example 11.4.3. Consider the integral

∞
sin x
dx.
−∞ x
To evaluate this integral, note that the function

eiz
F (z) =
z
has a simple pole at z = 0. We consider instead the function

eiz − 1
G(z) =
z
which has a removable singularity at z = 0. Consider now the Jordan contour

C = [−R, R] ∪ CR ,

where R > 0.

CR

-R R

By Cauchy’s integral theorem, we have

R

G(x) dx + G(z) dz = 0,
−R CR

so that

R


dz eiz
G(x) dx = − dz.
−R CR z CR z

Note that




dz  eiz  1 π
= πi and  dz  ≤ |eiz ||dz| < .
z  z R CR R
CR CR

Hence


 R 
  π
 G(x) dx − πi < .
 −R  R

Since

R
R
sin x
dx = Im G(x) dx,
−R x −R

it follows that


 R sin x 
  π
 dx − π  < ,
 −R x  R
 Chapter 11 : Evaluation of Definite Integrals 11–11

so that letting R → ∞, we obtain

∞
sin x
dx = π.
−∞ x

We shall return to this example later.

Note that the previous two examples do not fit the discussion at the beginning of this section, since
the degrees of the denominators of the rational functions in question do not exceed the degrees of the
numerators by at least 2. In fact, we have a non-trivial convergence problem for the integral

∞
f (x) sin x dx.
−∞

The argument formally establishes the existence of this integral as

R
lim f (x) sin x dx,
R→∞ −R

and not as

X2
lim f (x) sin x dx.
X1 →∞ −X1
X2 →∞

However, it turns out that this does not cause any difficulties, since the functions f (x) sin x in question
turn out to be even functions of x, so that for X1 , X2 > 0, we have

X2
X1
X2
f (x) sin x dx = f (x) sin x dx + f (x) sin x dx.
−X1 0 0

Let us examine the problem more carefully. Consider the integral

∞
f (x)eix dx,
−∞

where f (x) is a real valued rational function in the real variable x. Suppose now that the degree of the
denominator of f exceeds the degree of the numerator of f by exactly 1, and that f has no poles on the
real line. To establish the existence of the integral, we need to study the integral

X2
f (x)eix dx,
−X1

where −X1 < 0 < X2 , and consider the limit as X1 → ∞ and X2 → ∞. Clearly we cannot use the
semicircular arc. We shall use instead a rectangular contour

C = [−X1 , X2 ] ∪ [X2 , X2 + iY ] ∪ [X2 + iY, −X1 + iY ] ∪ [−X1 + iY, −X1 ],

where Y > 0. Here [Z1 , Z2 ], where Z1 , Z2 ∈ C, denotes the line segment from Z1 to Z2 .

-X 1 +iY X2+ iY

-X 1 X2
11–12 W W L Chen : Introduction to Complex Analysis

By Cauchy’s residue theorem, we have

X2



ix iz iz
(8) f (x)e dx + f (z)e dz + f (z)e dz + f (z)eiz dz
−X1 [X2 ,X2 +iY ] [X2 +iY,−X1 +iY ] [−X1 +iY,−X1 ]

= 2πi res(f (z)eiz , zi ),
zi inside C

where the summation is taken over all the poles of f (z)eiz inside the rectangular contour C. When
X1 , X2 and Y are large, then all the poles of the function f (z)eiz in the upper half plane are inside the
contour C.

Under our hypotheses, the function zf (z) is bounded. Suppose that |zf (z)| ≤ M for every z ∈ C.

Note first of all that

Y
f (z)eiz dz = i f (X2 + iy)ei(X2 +iy) dy.
[X2 ,X2 +iY ] 0

Since
M M
|f (X2 + iy)| ≤ ≤ ,
|X2 + iy| X2

we have



  M Y −y
  M M
(9)  f (z)e dz  ≤
iz
e dy = (1 − e−Y ) < .
 [X2 ,X2 +iY ]  X2 0 X2 X2

Similarly,


 
 iz  M
(10)  f (z)e dz  < .
 [−X1 +iY,−X1 ]  X1

Next, note that

X2
f (z)e dz = −
iz
f (x + iY )ei(x+iY ) dx.
[X2 +iY,−X1 +iY ] −X1

Since
M M
|f (x + iY )| ≤ ≤ ,
|x + iY | Y

we have


  M
X2 M e−Y
 
(11)  f (z)e dz  ≤
iz
e−Y dx = (X1 + X2 ).
 [X2 +iY,−X1 +iY ]  Y −X1 Y

Combining (8)–(11), we conclude that for sufficiently large X1 , X2 and Y , we have



 X2   M e−Y
  M M
(12)  f (x)eix dx − 2πi res(f (z)eiz , zi ) < + + (X1 + X2 ).
 −X1  X1 X2 Y
Imzi >0

Note that the left hand side of (12) is independent of Y . For fixed X1 and X2 , we have

M e−Y
(X1 + X2 ) → 0
Y
 Chapter 11 : Evaluation of Definite Integrals 11–13

as Y → ∞. It follows that


 X2  
  M M
 f (x)e dx − 2πi
ix
res(f (z)e , zi ) ≤
iz
+ .
 −X1  X1 X2
Imzi >0

Letting X1 → ∞ and X2 → ∞, we conclude that

∞ 
f (x)eix dx = 2πi res(f (z)eiz , zi ).
−∞ Imzi >0

11.5. Bending Round a Singularity

We shall first indicate the ideas by two examples.

Example 11.5.1. Recall Example 11.4.3, and consider again the integral

∞
sin x
dx.
−∞ x

If we use the function

eiz
F (z) =
z

to evaluate this integral, then the Jordan contour C = [−R, R] ∪ CR discussed earlier is unsuitable, since
the singular point z = 0 is on the contour. Let us consider instead the Jordan contour

C = [−R, −δ] ∪ K(δ) ∪ [δ, R] ∪ CR ,

where R > δ > 0, and where K(δ) denotes the semicircular arc z = δeit , where t ∈ [π, 2π].

CR

-R -δ δ R
K(δ)

By Cauchy’s residue theorem, we have

−δ

R

F (x) dx + F (z) dz + F (x) dx + F (z) dz = 2πi res(F, 0).
−R K(δ) δ CR

Note that the function F (z) in analytic in C apart from a simple pole at z = 0 with residue 1, so that
res(F, 0) = 1. It follows that
1
F (z) = + G(z),
z
11–14 W W L Chen : Introduction to Complex Analysis

where G(z) is entire. Furthermore, it is easy to show that

dz
= πi.
K(δ) z

Hence

−δ
R


F (x) dx + F (x) dx + G(z) dz + F (z) dz = πi.
−R δ K(δ) CR

Since G(z) is entire, there exists M > 0 such that |G(z)| < M whenever |z| ≤ 1, so that for every δ < 1,
we have


 
 
 G(z) dz  ≤ M πδ.
 K(δ) 

On the other hand, a simple application of Jordan’s lemma gives



  π
 F (z) dz  < .
 R
CR

It follows that if δ < 1, then



R 
 −δ 
  π
 F (x) dx + F (x) dx − πi < M πδ + .
 −R δ  R

Letting δ → 0 and R → ∞, we obtain

∞
F (x) dx = πi.
−∞

Taking imaginary parts gives

∞
sin x
dx = π.
−∞ x

Example 11.5.2. Suppose that the real number a > 0 is fixed. Consider the integral

∞
cos x
dx.
−∞ a − x
2 2

To evaluate this integral, note that the function

eiz
F (z) =
a2 − z 2
has simple poles at z = ±a. Consider now the Jordan contour

C = [−R, −a − δ1 ] ∪ J1 (δ1 ) ∪ [−a + δ1 , a − δ2 ] ∪ J2 (δ2 ) ∪ [a + δ2 , R] ∪ CR ,

where R > 2a and 0 < δ1 , δ2 < a.

CR
J2(δ2)
J1(δ1)

-R -a a R
 Chapter 11 : Evaluation of Definite Integrals 11–15

Here −J1 (δ1 ) denotes the semicircular arc z = −a + δ1 eit , where t ∈ [0, π], and −J2 (δ2 ) denotes the
semicircular arc z = a + δ2 eit , where t ∈ [0, π]. By Cauchy’s integral theorem, we have

−a−δ1

a−δ2
F (x) dx − F (z) dz + F (x) dx
−R −J1 (δ1 ) −a+δ1


R

− F (z) dz + F (x) dx + F (z) dz = 0.
−J2 (δ2 ) a+δ2 CR

Note that the function F (z) in analytic in C apart from simple pole at z = ±a with residues

e−ia eia
res(F, −a) = lim (z + a)F (z) = and res(F, a) = lim (z − a)F (z) = − .
z→−a 2a z→a 2a
It follows that

e−ia eia
F (z) = + G1 (z) = − + G2 (z),
2a(z + a) 2a(z − a)

where G1 (z) is analytic in {z : |z + a| ≤ a} and G2 (z) is analytic in {z : |z − a| ≤ a}. Furthermore, it is

easy to show that



e−ia πie−ia eia πieia
dz = and dz = .
−J1 (δ1 ) 2a(z + a) 2a −J2 (δ2 ) 2a(z − a) 2a

Hence

−a−δ1
a−δ2
R
F (x) dx + F (x) dx + F (x) dx
−R −a+δ1 a+δ2




− G1 (z) dz − G2 (z) dz + F (z) dz
−J1 (δ1 ) −J2 (δ2 ) CR
−ia ia
πie πie πi −ia
= − = (e − eia ).
2a 2a 2a
Since G1 (z) is analytic in {z : |z + a| ≤ a}, there exists M1 > 0 such that |G1 (z)| < M1 whenever
|z + a| ≤ a, so that for every δ1 < a, we have


 
 
 G1 (z) dz  ≤ M1 πδ1 .
 −J1 (δ1 ) 

Since G2 (z) is analytic in {z : |z − a| ≤ a}, there exists M2 > 0 such that |G2 (z)| < M2 whenever
|z − a| ≤ a, so that for every δ2 < a, we have


 
 
 G2 (z) dz  ≤ M2 πδ2 .
 −J2 (δ2 ) 

On the other hand, a simple application of Jordan’s lemma gives



  π
 F (z) dz <
  R2 − a2 .
CR

It follows that if 0 < δ1 , δ2 < a and R > 2a, then



a−δ2
R 
 −a−δ1 
 πi −ia 
 F (x) dx + F (x) dx + F (x) dx − (e − eia )
 −R −a+δ1 a+δ2 2a 
π
< M1 πδ1 + M2 πδ2 + 2 .
R − a2
11–16 W W L Chen : Introduction to Complex Analysis

Letting δ1 , δ2 → 0 and R → ∞, we obtain

∞
πi −ia
F (x) dx = (e − eia ).
−∞ 2a
Taking real parts gives

∞
cos x π sin a
dx = .
−∞ a2 − x2 a

Remark. Note that in Example 11.5.1, we have bent round the singularity in question by using a con-
tour with the singularity in the interior, whereas in Example 11.5.2, we have bent round the singularities
in question by using a contour with the singularities in the exterior.

We have used the following general result.

THEOREM 11B. Suppose that a function F (z) is analytic in an -neighbourhood of z0 , apart from
a simple pole at z0 with residue a−1 . Suppose further that 0 ≤ t1 < t2 ≤ 2π. For every positive δ < ,
let J(δ) denote a circular arc of the form z = z0 + δeit , where t ∈ [t1 , t2 ].

J(δ)
δ
z0 ε

Then


lim F (z) dz = ia−1 (t2 − t1 ).
δ→0 J(δ)

Proof. We can write

a−1
F (z) = + G(z),
z − z0
where G(z) is analytic in the closed disc {z : |z − z0 | ≤ δ}. Then




a−1
F (z) dz = dz + G(z) dz.
J(δ) J(δ) z − z0 J(δ)

It is easy to check that

a−1
dz = ia−1 (t2 − t1 ).
J(δ) z − z0
On the other hand, since G(z) is analytic in the closed disc {z : |z − z0 | ≤ δ}, it is bounded in this disc,
and so there exists M > 0 such that |G(z)| ≤ M whenever |z − z0 | ≤ δ, whence



 
 
 G(z) dz  ≤ M |dz| ≤ 2πM δ → 0
 J(δ)  J(δ)

as δ → 0. The result follows.

 Chapter 11 : Evaluation of Definite Integrals 11–17

11.6. Integrands with Branch Points

Consider an integral of the type

∞
(13) xα f (x) dx,
0

where f (x) is a real valued rational function in the real variable x. Here we shall assume that the degree
of the denominator of f exceeds the degree of the numerator of f by at least 2, and that f has no poles
on the positive real axis and at most a simple pole at the origin. We shall also assume that 0 < α < 1.

The problem here is that the function z α f (z) is not single valued. However, this is precisely the
circumstance that makes it possible to find the integral. The simplest technique is to first of all make
the substitution x = u2 , and note that

∞
∞
(14) xα f (x) dx = 2 u2α+1 f (u2 ) du.
0 0

We now consider the function

F (z) = z 2α+1 f (z 2 ).

For the function z 2α , by choosing the branch so that the argument of z 2α lies between −πα and 3πα, it
is easy to see that this is well defined and analytic in the region obtained from C by deleting the origin
and the negative imaginary axis. It follows that as long as a Jordan contour avoids this cut, then we
can use Cauchy’s residue theorem on F (z).

We shall consider the Jordan contour

C = [−R, −δ] ∪ J(δ) ∪ [δ, R] ∪ CR ,

where R > δ > 0, −J(δ) denotes the semicircular arc of the form z = δeit , where t ∈ [0, π], and CR
denotes the semicircular arc of the form z = Reit , where t ∈ [0, π].

CR

J(δ)

-R -δ δ R

By Cauchy’s residue theorem, we have

−δ

R

F (z) dz + F (z) dz + F (z) dz + F (z) dz = 2πi res(F, zi ),
−R J(δ) δ CR zi inside C

where the summation is taken over all the poles of F inside the Jordan contour C. It is easily shown
that



F (z) dz → 0 and F (z) dz → 0
J(δ) CR

as δ → 0 and R → ∞ respectively, so that

∞ 
F (z) dz = 2πi res(F, zi ),
−∞ Imzi >0
11–18 W W L Chen : Introduction to Complex Analysis

where the summation is taken over all the poles of F in the upper half plane. On the other hand, note
that (−z)2α = e2πiα z 2α , and so

∞
∞
∞
z 2α+1 f (z 2 ) dz = (z 2α+1 + (−z)2α+1 )f (z 2 ) dz = (1 − e2πiα ) u2α+1 f (u2 ) du.
−∞ −∞ 0

Since e2πiα = 1, this gives us a way of calculating the integral on the right hand side of (14).

Example 11.6.1. Suppose that the real number α ∈ (0, 1) is fixed. Consider the integral

∞
xα−1
dx.
0 1+x

The substitution x = u2 gives

∞
∞
xα−1 u2α+1
(15) dx = 2 du.
0 1+x 0 u2 + u4

To evalaute this integral, note that the function

z 2α+1
F (z) =
z2 + z4
has a singularity at z = 0 and simple poles at z = ±i. Consider now the Jordan contour

C = [−R, −δ] ∪ J(δ) ∪ [δ, R] ∪ CR ,

where R > 1 > δ > 0.

CR
i
J(δ)

-R -δ δ R
-i

By Cauchy’s residue theorem, we have

−δ

R

F (z) dz + F (z) dz + F (z) dz + F (z) dz = 2πi res(F, i).
−R J(δ) δ CR

Since
1
res(F, i) = lim(z − i)F (z) = − eπiα ,
z→i 2
it follows that

−δ

R

F (z) dz + F (z) dz + F (z) dz + F (z) dz = −πieπiα .
−R J(δ) δ CR

Note now that


 

  δ 2α+1   R2α+1
  
 F (z) dz  ≤ 2 πδ → 0 and  F (z) dz  ≤ 4 πR → 0
 J(δ)  δ − δ4 CR R − R2
 Chapter 11 : Evaluation of Definite Integrals 11–19

as δ → 0 and R → ∞ respectively. Hence

∞
∞
u2α+1 z 2α+1
(1 − e2πiα ) du = dz = −πieπiα ,
0 u2 + u4 −∞ z2 + z4

so that

∞
u2α+1 πieπiα πi π
du = − = − −πiα = .
0
2
u +u 4 1 − e2πiα e − eπiα 2 sin πα

It now follows from (15) that

∞
xα−1 π
dx = .
0 1+x sin πα

In fact, the integral (13) can be studied without the substitution x = u2 . However, the contour that
we use will not be a Jordan contour. We shall consider the function

F (z) = z α f (z),

and choose a branch of z α so that the argument lies between 0 and 2πα.

We now consider the contour

C = [δ, R] ∪ S(R) ∪ [R, δ] ∪ L(δ),

where R > δ > 0, S(R) denotes the circle of the form z = Reit , where t ∈ [0, 2π], and −L(δ) denotes the
circle of the form z = δeit , where t ∈ [0, 2π].

S(R)

L(δ)

δ R

Clearly C is not a Jordan contour. However, there clearly exists t0 ∈ (0, 2π) such that the line segment
joining δeit0 and Reit0 does not pass through any singularities of f (z) in {z : |z| ≤ R}. Now let

C1 = [δ, R] ∪ S1 (R) ∪ [Reit0 , δeit0 ] ∪ L1 (δ),

where S1 (R) denotes the circular arc of the form z = Reit , where t ∈ [0, t0 ], and −L1 (δ) denotes the
circular arc of the form z = δeit , where t ∈ [0, t0 ]. Also let

C2 = [R, δ] ∪ L2 (δ) ∪ [δeit0 , Reit0 ] ∪ S2 (R),

where S2 (R) denotes the circular arc of the form z = Reit , where t ∈ [t0 , 2π], and −L2 (δ) denotes the
circular arc of the form z = δeit , where t ∈ [t0 , 2π].
11–20 W W L Chen : Introduction to Complex Analysis

It is easy to see that

(16) F (z) dz = F (z) dz + F (z) dz.
C C1 C2

Clearly there exists 0 > 0 such that C1 is a Jordan contour in the simply connected domain

D1 = {z = 0 : −0 < arg z < t0 + 0 }

and C2 is a Jordan contour in the simply connected domain

D2 = {z = 0 : t0 − 0 < arg z < 2π + 0 }.

C1

C2

Applying Cauchy’s residue theorem, we obtain



F (z) dz = 2πi res(F, zi ) and F (z) dz = 2πi res(F, zi ),
C1 zi inside C1 C2 zi inside C2

so that


(17) F (z) dz = 2πi res(F, zi ).
C zi inside C

Note that






(18) F (z) dz = F (z) dz + F (z) dz + F (z) dz + F (z) dz,
C1 [δ,R] S1 (R) [Reit0 ,δeit0 ] L1 (δ)

and






(19) F (z) dz = F (z) dz + F (z) dz + F (z) dz + F (z) dz.
C2 [R,δ] L2 (δ) [δeit0 ,Reit0 ] S2 (R)

When we evaluate the integrals in (18), we need 0 ≤ arg z ≤ t0 . Hence

R
(20) F (z) dz = F (x) dx.
[δ,R] δ

When we evaluate the integrals in (19), we need t0 ≤ arg z ≤ 2π. Hence

R
R
(21) F (z) dz = z f (z) dz = −
α α 2πiα
x e f (x) dx = −e 2πiα
F (x) dx.
[R,δ] [R,δ] δ δ
 Chapter 11 : Evaluation of Definite Integrals 11–21

Also



(22) F (z) dz + F (z) dz = 0.
[Reit0 ,δeit0 ] [δeit0 ,Reit0 ]

It follows from (16), (18)–(22) that

(23) F (z) dz = F (z) dz + F (z) dz + F (z) dz + F (z) dz
C [δ,R] [R,δ] S1 (R) S2 (R)



+ F (z) dz + F (z) dz
L1 (δ) L2 (δ)

R


= (1 − e2πiα ) F (x) dx + F (z) dz + F (z) dz.
δ S(R) L(δ)

It is easily shown that

F (z) dz → 0 and F (z) dz → 0
L(δ) S(R)

as δ → 0 and R → ∞ respectively, so it follows from (17) and (23) that

∞ 
(1 − e 2πiα
) F (x) dx = 2πi res(f, zi ),
0 zi

where the summation is taken over all the poles of F in C \ {0}.

Example 11.6.2. Suppose that the real number α ∈ (0, 1) is fixed. Consider again the integral

∞ α−1
x
dx.
0 1 +x
To evaluate this integral, note that the function
z α−1 zα
F (z) = = ,
1+z z(1 + z)
has a simple pole at z = −1, with residue

res(F, −1) = lim (z + 1)F (z) = lim z α−1 = eπi(α−1) = −eπiα .

z→−1 z→−1

Consider now the contour

C = [δ, R] ∪ S(R) ∪ [R, δ] ∪ L(δ),

where R > 1 > δ > 0.

S(R)

L(δ)

δ R
-1
11–22 W W L Chen : Introduction to Complex Analysis

By Cauchy’s residue theorem and our earlier observation, we have

R


(24) F (z) dz = (1 − e2πiα ) F (x) dx + F (z) dz + F (z) dz = −2πieπiα .
C δ S(R) L(δ)

Note now that


 

  δ α−1   Rα−1
   
 F (z) dz  ≤ 2πδ → 0 and  F (z) dz  ≤ 2πR → 0
 L(δ)  1−δ  S(R)  R−1

as δ → 0 and R → ∞ respectively, so it follows from (24) that

∞
xα−1 2πieπiα π
dx = − = .
0 1+x 1 − e2πiα sin πα

We next turn our attention to integrals of the types

∞
∞
(25) f (x) log x dx and f (x) log2 x dx,
0 0

where f (x) is a real valued rational function in the real variable x. Here we shall assume that the degree
of the denominator of f exceeds the degree of the numerator of f by at least 2, and that f has no poles
on the non-negative real axis, so that the integrals (25) are convergent. We shall also assume that f is
an even function; in other words, f (−x) = f (x) for every x ∈ R \ {0}.

We shall consider the function

F (z) = f (z) log2 z

and the Jordan contour

C = [δ, R] ∪ CR ∪ [−R, −δ] ∪ J(δ),

where R > δ > 0. Here CR denotes the semicircular arc z = Reit , where t ∈ [0, π], and −J(δ) denotes
the semicircular arc z = δeit , where t ∈ [0, π].

CR

J(δ)

-R -δ δ R

By Cauchy’s residue theorem, we have


F (z) dz + F (z) dz + F (z) dz + F (z) dz = 2πi res(F, zi ).
[δ,R] CR [−R,−δ] J(δ) zi inside C

If we impose the restriction −π/2 ≤ arg z < 3π/2, then

R
F (z) dz = f (x) log2 x dx,
[δ,R] δ
 Chapter 11 : Evaluation of Definite Integrals 11–23

and


R
R
2
F (z) dz = (log x + iπ) f (−x) dx = (log x + iπ)2 f (x) dx
[−R,−δ] δ δ

R
R
R
= f (x) log2 x dx + 2πi f (x) log x dx − π 2 f (x) dx.
δ δ δ

It is easily shown that

F (z) dz → 0 and F (z) dz → 0
J(δ) CR

as δ → 0 and R → ∞ respectively. It follows that

∞
∞
∞ 
(26) 2 2
f (x) log x dx + 2πi f (x) log x dx − π 2
f (x) dx = 2πi res(F, zi ),
0 0 0 Imzi >0

where the summation is taken over all the poles of F in the upper half plane. The integrals (25) can
then be found on equating real and imaginary parts.

Example 11.6.3. Consider the integrals

∞
∞
log x log2 x
dx and dx.
0 1 + x2 0 1 + x2

To evaluate these integrals, note that the function

log2 z
F (z) =
1 + z2

has simple poles at z = ±i. In particular,

log2 z π2
res(F, i) = lim(z − i)F (z) = lim =− .
z→i z→i z + i 8i

Consider now the Jordan contour

C = [δ, R] ∪ CR ∪ [−R, −δ] ∪ J(δ),

where R > 1 > δ > 0.

CR
i
J(δ)

-R -δ δ R
-i

By Cauchy’s residue theorem, we have

π3
F (z) dz + F (z) dz + F (z) dz + F (z) dz = 2πi res(F, i) = − .
[δ,R] CR [−R,−δ] J(δ) 4
11–24 W W L Chen : Introduction to Complex Analysis

On J(δ), we have z = δeit , so that

 
 log2 z  | log δ + it|2 log2 δ + π 2
 ≤ ≤ .
 1 + z2  1 − δ2 1 − δ2

Hence


  log2 δ + π 2
 
 F (z) dz  ≤ πδ → 0
 J(δ)  1 − δ2

as δ → 0. On CR , we have z = Reit , so that

 
 log2 z  | log R + it|2 log2 R + π 2
 ≤ ≤ .
 1 + z2  R2 − 1 R2 − 1

Hence


  log2 R + π 2
 F (z) dz  ≤ πR → 0
 R2 − 1
CR

as R → ∞. It follows from (26) that

∞
∞
∞
log2 x log x dx π3
2 dx + 2πi dx − π 2 = − .
0 1 + x2 0 1 + x2 0 1 + x2 4

It is well known that

∞
dx π
= .
0 1 + x2 2

Hence

∞
∞
log2 x log x π3
2 dx + 2πi dx = .
0 1 + x2 0 1 + x2 4

Equating real and imaginary parts, we obtain

∞
∞
log2 x π3 log x
2
dx = and dx = 0.
0 1+x 8 0 1 + x2

We conclude this chapter by considering an example which does not easily fall into any general
discussion. Some of the calculation is unpleasant, and we shall omit some details.

Example 11.6.4. We wish to evaluate the integral

π
log sin x dx.
0

To do this, consider the function

(27) 1 − e2iz = eiz (e−iz − eiz ) = −2ieiz sin z.

Note that if z = x + iy, where x, y ∈ R, then

1 − e2iz = 1 − e−2y (cos 2x + i sin 2x),

 Chapter 11 : Evaluation of Definite Integrals 11–25

so that the function is real and non-positive if and only if x = nπ and y ≤ 0, where n ∈ Z. Consider the
simply connected domain D obtained from C by deleting all half lines of the form {z = nπ + iy : y ≤ 0},
where n ∈ Z.

-2π -π π 2π

In this domain, the principal branch of log(1 − e2iz ), with imaginary part between −π and π, is single
valued and analytic. Consider the Jordan contour

C = [δ, π − δ] ∪ T1 (δ) ∪ [π + iδ, π + iY ] ∪ [π + iY, iY ] ∪ [iY, iδ] ∪ T2 (δ),

where δ > 0 is small and Y > 0 is large. Here −T1 (δ) denotes the circular arc z = π + δeit , where
t ∈ [π/2, π], and −T2 (δ) denotes the circular arc z = δeit , where t ∈ [0, π/2].

π +iY
iY

iδ
π

By Cauchy’s integral theorem, we have

log(1 − e2iz ) dz + log(1 − e2iz ) dz + log(1 − e2iz ) dz
[δ,π−δ] T1 (δ) [π+iδ,π+iY ]




+ log(1 − e ) dz +
2iz
log(1 − e ) dz +
2iz
log(1 − e2iz ) dz = 0.
[π+iY,iY ] [iY,iδ] T2 (δ)

Using the periodicity of the integrand, we have

log(1 − e2iz ) dz + log(1 − e2iz ) dz = 0.
[π+iδ,π+iY ] [iY,iδ]

Furthermore, it can be shown that

log(1 − e2iz ) dz → 0 and log(1 − e2iz ) dz → 0
T1 (δ) T2 (δ)

as δ → 0, and that


log(1 − e2iz ) dz → 0
[π+iY,iY ]

as Y → ∞. It follows that

π
(28) log(1 − e2ix ) dx = 0.
0
11–26 W W L Chen : Introduction to Complex Analysis

Next, consider the function log(−2ieix sin x). If we choose log eix = ix, then the imaginary part lies
between 0 and π. To obtain the principal branch of the logarithm with imaginary part between −π and
π, we must choose log(−i) = −iπ/2. Hence

iπ
log(−2ieix sin x) = log 2 − + ix + log sin x,
2
so that

π
π
iπ 2 iπ 2
(29) log(−2ieix sin x) dx = π log 2 − + + log sin x dx.
0 2 2 0

Combining (27)–(29), we conclude that

π
log sin x dx = −π log 2.
0

Problems for Chapter 11

1. Show each of the following:

2π
2π
dθ 2π dθ 2π
a) =√ b) = , where a ∈ C and |a| < 1
0 2 + sin θ 3 0 1 − 2a cos θ + a2 1 − a2

∞
∞
∞
dx x2 π dx π
c) 2
=π d) 2 )2
dx = e) 6
=
−∞ 1 + x 0 (1 + x 4 0 1 + x 3

∞
∞
cos x π cos x π
f) 2 )2
dx = g) 2 + a2
dx = e−a , where a ∈ R and a > 0
−∞ (1 + x e −∞ x a

∞  −b −a

cos x π e e
h) dx = − , where a, b ∈ R and a > b > 0
2 2 2
−∞ (x + a )(x + b )
2 a −b
2 2 b a

2. Suppose that a ∈ R and 0 < a < 1. By integrating the function

eaz
ez+1

around a rectangle with vertices at ±R and ±R + 2πi, show that

∞
eax π
dx = .
−∞ ex + 1 sin πa

3. Show each of the following:

∞
∞
1 − cos x sin πx
a) dx = π b) dx = π
−∞ x2 0 x(1 − x2 )

∞ √
∞
log x π2 2 log x π2
c) dx = − d) dx =
0 x4 + 1 16 0 x −1
2 4
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 12
HARMONIC FUNCTIONS
AND CONFORMAL MAPPINGS

12.1. A Local Property of Analytic Functions

Consider an arc C given by z(t), where t ∈ [A, B]. For every t ∈ [A, B], we can write

z(t) = x(t) + iy(t),

where x(t), y(t) ∈ R. Then the vector

z
(t) = x
(t) + iy
(t)

has slope dy/dx, which is also the slope of the arc C. It follows that if z
(t0 ) = 0, then the vector z
(t0 )
is tangent to the arc at the point z0 = z(t0 ), and arg z
(t0 ) is the angle this directed tangent makes with
the positive x-axis.

z
(t0 )

z(t0 ) arg z(t0 )

C
12–2 W W L Chen : Introduction to Complex Analysis

Suppose now that C lies in a domain D, and that a function f (z) is analytic in D. Consider the
arc f (C) given by w(t) = f (z(t)), where t ∈ [A, B]. By the Chain rule,

w
(t) = f
(z(t))z
(t).

Suppose now that z
(t0 ) = 0 and f
(z0 ) = 0, where z0 = z(t0 ). Then w
(t0 ) = 0, and

arg w
(t0 ) = arg f
(z0 ) + arg z
(t0 ).

We can interpret this geometrically in the following way: The angle between the directed tangent to C
at z0 = z(t0 ) and the directed tangent to f (C) at f (z0 ) is arg f
(z0 ).

In other words, under the mapping f , the directed tangent to any arc through z0 is rotated by an
angle arg f
(z0 ), independent of the choice of the arc through z0 . This also means that if two arcs C1
and C2 intersect at z0 at an angle, then the two arcs f (C1 ) and f (C2 ) intersect at f (z0 ) at the same
angle.

Definition. An analytic function f is said to be conformal at z0 if the following condition is satisfied:

If two arcs C1 and C2 meet at z0 , then the angle from f (C1 ) to f (C2 ) at f (z0 ) is the same as the angle
from C1 to C2 at z0 .

z w

same
angle f(C2 )

C1 C2
f(C1 )

We have in fact proved the following result.

THEOREM 12A. Suppose that a function f is analytic in a domain D, and that z0 ∈ D. Suppose
further that f
(z0 ) = 0. Then f is conformal at z0 .

Remark. Conformality is considered a local property of analytic functions. Note also that

|f (z) − f (z0 )|
lim = |f
(z0 )|.
z→z0 |z − z0 |

This shows that |f
(z0 )| is a local scaling factor of the function f at z0 , and is independent of the
direction of z from z0 .

To say that an analytic function is conformal usually means that it is locally one-to-one. In partic-
ular, we have the following result.

THEOREM 12B. Suppose that a function f is analytic and one-to-one in a domain D. Then f is
conformal at every point in D.

This follows immediately from the result below and Theorem 12A.

THEOREM 12C. Suppose that a non-constant function f is analytic in a domain D, and that z0 ∈ D.
Suppose further that f
(z0 ) = 0. Then f cannot be one-to-one in any disc containing z0 .
 Chapter 12 : Harmonic Functions and Conformal Mappings 12–3

Proof. Write w0 = f (z0 ). Since f (z) is not identically constant, the function g(z) = f (z) − w0 is not
identically zero. If f
(z0 ) = 0, then g(z) has a zero of finite order at least 2 at z0 . Since the zeros of an
analytic function are isolated, we can choose r > 0 so small that both g(z) and f
(z) have no zeros in
the punctured disc {z : 0 < |z − z0 | ≤ r}. Then

m = min |g(z)| > 0,

z∈C

where C = {z : |z − z0 | = r} denotes the boundary of the disc. Let w ∈ C satisfy 0 < |w − w0 | < m.
Then

|w0 − w| < |g(z)|

on C. It follows from Rouché’s theorem that the functions g(z) and g(z) + (w0 − w) have the same
number of zeros inside C. Hence

g(z) + (w0 − w) = f (z) − w

has at least two zeros inside C. Clearly none of these zeros can be z0 . Since f
(z) = 0 inside the
punctured disc, it follows that these zeros must be simple, and so distinct.

Example 12.1.1. The exponential function f (z) = ez has non-zero derivative at every z ∈ C, and is
therefore conformal at every z ∈ C. Note, however, that f : C → C is not one-to-one. On the other
hand, the function is one-to-one if we restrict its domain of definition to any strip of the form

{z = x + iy : a ≤ y ≤ b},

where 0 < b − a < 2π. We therefore say that the exponential function is locally one-to-one.

12.2. Laplace’s Equation

Recall that a continuous function φ(x, y) that satisfies Laplace’s equation

∂2φ ∂2φ
+ 2 =0
∂x2 ∂y

in a domain D ⊆ C is said to be harmonic in D. In Theorem 3E, we have shown that such a function
can be written as the real part of an analytic function f in D.

Our main task in this section is to show that a harmonic function can be carried from one domain
to another by analytic functions. More precisely, we prove the following result.

THEOREM 12D. Suppose that D, D
⊆ C are domains, and that f : D → D
is a one-to-one and
onto analytic function. Suppose further that for every z ∈ D, we write w = f (z), where z = x + iy and
w = u + iv, with x, y, u, v ∈ R. Then for every function φ(x, y) harmonic in D, the function ψ(u, v),
defined by

(1) ψ(u, v) = φ(x(u, v), y(u, v)),

is harmonic in D
.

Theorem 12D is particularly useful in applications which involve the solution of the Dirichlet problem
concerning the question of finding a harmonic function in a domain D which takes specified values on
the boundary of D. Once we solve this problem for a particular domain, we can use Theorem 12D to find
solutions on all domains which can be obtained from D by a one-to-one and onto analytic function, so
12–4 W W L Chen : Introduction to Complex Analysis

long as the boundary values correspond. We can therefore select the domain which makes the problem
simplest.

Note that (1) can be written in the form

ψ(w) = φ(f −1 (w)).

Our first task is therefore to establish the analyticity of the inverse function f −1 : D
→ D.

THEOREM 12E. Suppose that D, D
⊆ C are domains, and that f : D → D
is a one-to-one and
onto analytic function. Then the inverse function f −1 : D
→ D is analytic in D
. Suppose further that
for every z ∈ D, we write w = f (z), where z = x + iy and w = u + iv, with x, y, u, v ∈ R. Then

∂x ∂y ∂x ∂y
(2) = and =− .
∂u ∂v ∂v ∂u

Proof. To prove the first assertion, it suffices to prove that for every w0 = f (z0 ) ∈ D
, the limit
z − z0
lim
w→w0 w − w0
exists. To establish this, note first of all that by Theorem 12C, we have
w − w0
lim = 0.
z→z0 z − z0
Since f is continuous in D, we clearly have w → w0 as z → z0 . Since f is one-to-one in D, we clearly
have w = w0 when z = z0 . It follows that

−1
z − z0 w − w0
lim = lim .
w→w0 w − w0 z→z0 z − z0

To complete the proof of the theorem, note that (2) are the Cauchy-Riemann equations of the analytic
function f −1 .

Proof of Theorem 12D. Note that

∂ψ ∂φ ∂x ∂φ ∂y ∂x ∂φ ∂y ∂φ
= + = + ,
∂u ∂x ∂u ∂y ∂u ∂u ∂x ∂u ∂y
so that




∂2ψ ∂ ∂x ∂φ ∂y ∂φ ∂ 2 x ∂φ ∂x ∂ ∂φ ∂ 2 y ∂φ ∂y ∂ ∂φ
(3) 2
= + = + + + .
∂u ∂u ∂u ∂x ∂u ∂y ∂u2 ∂x ∂u ∂u ∂x ∂u2 ∂y ∂u ∂u ∂y

On the other hand,




∂ ∂φ ∂x ∂ ∂φ ∂y ∂ ∂φ ∂x ∂ 2 φ ∂y ∂ 2 φ
(4) = + = + .
∂u ∂x ∂u ∂x ∂x ∂u ∂y ∂x ∂u ∂x2 ∂u ∂x∂y

Similarly,




∂ ∂φ ∂x ∂ ∂φ ∂y ∂ ∂φ ∂x ∂ 2 φ ∂y ∂ 2 φ
(5) = + = + .
∂u ∂y ∂u ∂x ∂y ∂u ∂y ∂y ∂u ∂x∂y ∂u ∂y 2

Combining (3)–(5), we obtain

2
2
∂2ψ ∂ 2 x ∂φ ∂ 2 y ∂φ ∂x ∂2φ ∂x ∂y ∂ 2 φ ∂y ∂2φ
(6) 2
= + + +2 + .
∂u ∂u2 ∂x ∂u2 ∂y ∂u ∂x2 ∂u ∂u ∂x∂y ∂u ∂y 2
 Chapter 12 : Harmonic Functions and Conformal Mappings 12–5

A similar argument gives

2
2
∂2ψ ∂ 2 x ∂φ ∂ 2 y ∂φ ∂x ∂2φ ∂x ∂y ∂ 2 φ ∂y ∂2φ
(7) 2
= + + +2 + .
∂v ∂v 2 ∂x ∂v 2 ∂y ∂v ∂x2 ∂v ∂v ∂x∂y ∂v ∂y 2
Adding (6) and (7), we have

2 
2 

∂2ψ ∂2ψ ∂ x ∂ 2 x ∂φ ∂ y ∂ 2 y ∂φ ∂x ∂y ∂x ∂y ∂ 2 φ
(8) + = + + + + 2 +
∂u2 ∂v 2 ∂u2 ∂v 2 ∂x ∂u2 ∂v 2 ∂y ∂u ∂u ∂v ∂v ∂x∂y


  

2  2
2 2 2 2
∂x ∂x ∂ φ ∂y ∂y ∂ φ
+ + 2
+ + .
∂u ∂v ∂x ∂u ∂v ∂y 2

Suppose now that φ(x, y) harmonic in D. Then

∂2φ ∂2φ
(9) + 2 =0
∂x2 ∂y
in D. On the other hand, the Cauchy-Riemann equations (2) give
∂2x ∂2x ∂2y ∂2y
(10) + 2 =0 and + 2 = 0,
∂u2 ∂v ∂u 2 ∂v
as well as

2
2
2
2
∂x ∂y ∂x ∂y ∂x ∂x ∂y ∂y
(11) + =0 and + = +
∂u ∂u ∂v ∂v ∂u ∂v ∂u ∂v
in D
. Combining (8)–(11), it is easily seen that
∂2ψ ∂2ψ
+ =0
∂u2 ∂v 2
in D
, so that ψ(u, v) is harmonic in D
.

12.3. Global Properties of Analytic Functions

We begin by studying the following result which can be considered both local and global. It can be
proved by means of Rouché’s theorem in the same spirit as the proof of Theorem 12C.

THEOREM 12F. (OPEN MAPPING THEOREM) Suppose that a non-constant function f is ana-
lytic in a domain D. Then f (S) is open for any open set S ⊆ D. More specifically, suppose that z0 ∈ D,
and that w0 = f (z0 ). Then for all sufficiently small  > 0, there exists δ > 0 such that
{w : |w − w0 | < δ} ⊆ f ({z : |z − z0 | < }).
In other words, w0 is an interior point of f (D).

z w

ε δ
z0 w0
12–6 W W L Chen : Introduction to Complex Analysis

Proof. Since f (z) is not identically constant, the function g(z) = f (z) − w0 is not identically zero,
and has a zero of finite order at z0 . Since the zeros of an analytic function are isolated, we can choose
r <  so small that g(z) has no zeros in the punctured disc {z : 0 < |z − z0 | ≤ r}. Then

δ = min |g(z)| > 0,

z∈C

where C = {z : |z − z0 | = r} denotes the boundary of the disc. Let w ∈ C satisfy |w − w0 | < δ. Then

|w0 − w| < |g(z)|

on C. It follows from Rouché’s theorem that the functions g(z) and g(z) + (w0 − w) have the same
number of zeros inside C. Hence

g(z) + (w0 − w) = f (z) − w

has a solution inside C, so that w ∈ f ({z : |z − z0 | < r}). It follows that

{w : |w − w0 | < δ} ⊆ f ({z : |z − z0 | < r}).

The result follows.

Let us examine Theorem 12D again. We have assumed that both D and D
are domains. The
following result allows us to somewhat relax our assumptions.

THEOREM 12G. Suppose that a non-constant function f is analytic in a domain D. Then f (D) is
a domain.

Remark. Recall that an open set S is connected if every two points in S can be joined by the union
of a finite number of line segments lying in S. An easy theorem in real analysis states that any contour
can be approximated arbitrarily well by the union of a finite number of line segments. It follows that S
is connected if every two points in S can be joined by a contour lying in S.

Proof of Theorem 12G. To show that f (D) is a domain, we need to show that it is open and
connected. In view of Theorem 12F, it remains to show that f (D) is connected. Suppose that w1 , w2 ∈
f (D). Then there exist z1 , z2 ∈ D such that f (z1 ) = w1 and f (z2 ) = w2 . Since D is connected, z1 and
z2 can be joined by the union of a finite number of line segments lying in S. The image of each line
segment under f is an arc in f (D), since f is differentiable in D.

z w

w1

f(D)
D
z1 z2 w2

It follows that w1 and w2 can be joined by a contour lying in f (D).

We conclude this section by stating the following result.

THEOREM 12H. (RIEMANN MAPPING THEOREM) Suppose that D is a simply connected do-
main in C which is different from C. Then there exists a one-to-one and onto analytic function of the
type f : D → U, where U = {w : |w| < 1} denotes the unit open disc.
 Chapter 12 : Harmonic Functions and Conformal Mappings 12–7

Remarks. (1) If we prescribe a point in D and a direction through this point, then there is a unique
function of the type described which maps this point and direction to the origin and the positive x-axis
respectively.

(2) The proof can be split into three steps. Let z0 ∈ D be fixed. One begins by showing that
the collection S of one-to-one analytic functions of the type f : D → U and satisfying the conditions
f (z0 ) = 0 and f
(z0 ) > 0 is non-empty. One then shows that there is an extremal member in S with
greatest f
(z0 ). Finally, one shows that if a member in S is not an onto function, then it cannot be this
extremal member. It follows that the extremal member satisfies the requirements of the theorem.

(3) Unfortunately, Theorem 12H is a purely existence theorem, and so cannot be used in conjunc-
tion with Theorem 12D. In Chapters 13–14, we shall study some techniques which may enable us to
construct such a function.

Problems for Chapter 12

1. Let C1 and C2 be two straight lines that meet at the origin at an angle φ. Consider the function
f (z) = z 3 .
a) At what angle do the two lines f (C1 ) and f (C2 ) meet?
b) Comment on the solution in (a).

2. Discuss angles at the origin under the mapping f (z) = z α , where 0 < α < 1.

3. Use the Open mapping theorem to prove the Maximum principle.

4. Explain why the conclusion of the Riemann mapping theorem cannot hold when D = C.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 13
MÖBIUS TRANSFORMATIONS

13.1. Linear Functions

Example 13.1.1. Consider the square {z = x + iy ∈ C : 0 ≤ x ≤ 1 and 0 ≤ y ≤ 1}. The pictures

below show the images of this square under the functions f (z) = z + 1 + i, f (z) = eiφ z and f (z) = 2z.
Note that the image of the square in each case is also a square.

1 2 1 2
original square image under f (z) = z + 1 + i

φ
1 2 1 2
image under f (z) = eiφ z image under f (z) = 2z

The function f (z) = z + 1 + i is an example of a function of the type f (z) = z + c, where c ∈ C is fixed.
This function describes a translation on the complex plane C, where every point is shifted by a vector
corresponding to the complex number c. The function f (z) = eiφ z, where φ ∈ R is fixed, describes a
rotation on the complex plane C, where every point is rotated in the anticlockwise direction by an angle
φ about the origin. The function f (z) = 2z is an example of a function of the type f (z) = ρz, where
13–2 W W L Chen : Introduction to Complex Analysis

ρ ∈ R is positive and fixed. This function describes a magnification on the complex plane C, where
the distance between points is magnified by a factor ρ, noting that |ρz1 − ρz2 | = ρ|z1 − z2 | for every
z1 , z2 ∈ C.

It is easily seen that if we take the domain and codomain of each of the above functions to be the
complex plane C, then f : C → C is both one-to-one and onto. Furthermore, any geometric object in C
has an image under f which is similar to itself.

Definition. A linear function is a function f : C → C of the form f (z) = az + b, where a, b ∈ C are

fixed, and a = 0.

Example 13.1.2. Let us return to the three examples earlier. For the function f (z) = z + c, we have
a = 1 and b = c. For the function f (z) = eiφ z, we have a = eiφ and b = 0. For the function f (z) = ρz,
we have a = ρ and b = 0.

THEOREM 13A. Any linear function f : C → C is the composition of a rotation, a magnification

and a translation. Furthermore, it is one-to-one and onto.

Proof. Suppose that f (z) = az + b for every z ∈ C. Write a = ρeiφ , where ρ, φ ∈ R and ρ > 0. Then
f = f3 ◦ f2 ◦ f1 , where

f1 (z) = eiφ z and f2 (z) = ρz and f3 (z) = z + b.

We have the picture below:

rotation translation

magnifica tion

The last assertion follows from the observation that composition of functions preserves the one-to-one
and onto properties.

THEOREM 13B. The composition of any two linear functions is also a linear function.

Proof. Suppose that f1 (z) = a1 z + b1 and f2 (z) = a2 z + b2 , where a1 , b1 , a2 , b2 ∈ C and a1 , a2 =  0.

Then (f2 ◦ f1 )(z) = a2 (a1 z + b1 ) + b2 = a1 a2 z + (a2 b1 + b2 ). Clearly a1 a2 , a2 b1 + b2 ∈ C and a1 a2 = 0.

Example 13.1.3. Suppose that z0 ∈ C is fixed. Consider the linear function f : C → C which rotates
the complex plane C in the anticlockwise direction by an angle θ about the point z0 . We may adopt the
following strategy: Translate the point z0 to the origin, then rotate in the anticlockwise direction by an
angle θ about the origin, and then translate the origin back to the point z0 . Then f = f3 ◦ f2 ◦ f1 , where

f1 (z) = z − z0 and f2 (z) = eiθ z and f3 (z) = z + z0 .

 Chapter 13 : Möbius Transformations 13–3

We have the picture below:

z0 z0

f1 (z) = z − z0 f3 (z) = z + z0

f2 (z) = eiθ z

Hence

f (z) = eiθ (z − z0 ) + z0 = eiθ z + z0 (1 − eiθ ).

Alternatively, we may adopt the following strategy: Rotate in the anticlockwise direction by an angle θ
about the origin, and then translate the image of z0 under this rotation back to z0 . Then f = g2 ◦ g1 ,
where

g1 (z) = eiθ z and g2 (z) = z + (z0 − eiθ z0 ).

Hence

f (z) = eiθ z + (z0 − eiθ z0 ) = eiθ z + z0 (1 − eiθ ).

Example 13.1.4. Consider the linear function f : C → C which maps the horizontal arrow shown to
the other arrow shown.

2i

-2 -1 1 2 3

We may adopt the following

√ strategy: Translate the tip of the arrow from 3 + i to the origin, magnify
the arrow by a factor 1/ 2, rotate it about its tip (now at the origin) in the anticlockwise direction by
an angle 3π/4, and finally translate its tip from the origin to the point −2 + 2i. Then f = f4 ◦ f3 ◦ f2 ◦ f1 ,
where
1
f1 (z) = z − (3 + i) and f2 (z) = √ z and f3 (z) = e3πi/4 z and f4 (z) = z + (−2 + 2i).
2

Hence



e3πi/4 1 i 1 i
f (z) = √ (z − 3 − i) − 2 + 2i = − + (z − 3 − i) − 2 + 2i = − + z + i.
2 2 2 2 2
13–4 W W L Chen : Introduction to Complex Analysis

13.2. The Inversion Function

Consider the inversion function

1
w = f (z) = .
z

This function can be considered a function of the type f : C → C, where C denotes the extended complex
plane, so that C = C ∪ {∞}. We write formally f (0) = ∞ and f (∞) = 0.

Let us first study some geometric properties of this function. For our purposes, the point at ∞ is
considered to belong to every line on the extended complex plane.

Remarks. (1) A line passing through the origin contains all points of the form z = reiθ , where θ ∈ R
is fixed and r ∈ R. The images of these points under the inversion function are of the form

1 1
w= = e−iθ .
z r

They form a line through the origin. Note that the point at ∞ and the origin change roles under the
inversion function.

(2) A line not passing through the origin consists of all points of the form z = x + iy, where
x, y ∈ R and Ax + By = C, where A, B, C ∈ R are fixed and C = 0. The images of these points under
the inversion function are of the form w = u + iv, where u, v ∈ R and w = 1/z. It is easy to see that

1 1 u − iv
z= = = 2 ,
w u + iv u + v2

so that
u v
(1) x= and y=− .
u2 + v 2 u2 + v 2

It follows that
Au Bv
− 2 = C.
u2 +v 2 u + v2

This can be rewritten in the form

A B
u2 + v 2 − u + v = 0,
C C

the equation of a circle passing through the origin.

(3) Note now that the inverse of the inversion function is the inversion function itself. It follows
from the previous observation that a circle passing through the origin becomes a line not passing through
the origin under the inversion function.

(4) A circle not passing through the origin consists of all points of the form z = x + iy, where
x, y ∈ R and x2 + y 2 + Ax + By = C, where A, B, C ∈ R are fixed and C = 0. The images of these
points under the inversion function are of the form w = u + iv, where u, v ∈ R and w = 1/z. In view of
(1), we have

u2 v2 Au Bv
+ 2 + 2 − 2 = C.
(u2 2
+v ) 2 2
(u + v )2 u +v 2 u + v2
 Chapter 13 : Möbius Transformations 13–5

This can be rewritten in the form

A B 1
u2 + v 2 − u+ v = ,
C C C

the equation of a circle not passing through the origin.

We now state a result which includes these four remarks.

THEOREM 13C. The inversion function f : C → C, given by f (z) = 1/z for every non-zero z ∈ C,
and f (0) = ∞ and f (∞) = 0, is one-to-one and onto. On the other hand, its inverse function is itself.
Furthermore, the image under this function of a line or a circle in C is also a line or a circle in C.

Remarks. (1) We have in fact shown the following: Under the inversion function, the image of a
line through the origin is a line through the origin, the image of a line not through the origin is a circle
through the origin, the image of a circle through the origin is a line not through the origin, and the
image of a circle not through the origin is a circle not through the origin.

(2) If we think of a line as a circle of infinite radius, then we can think of circles and lines as
belonging to the “same” class. The inversion function therefore maps members of this class to members
of this class.

13.3. A Generalization

If we extend any linear function discussed in §13.1 to a function of the type f : C → C by writing
f (∞) = ∞, then it is easy to see that this extended function f : C → C is also one-to-one and onto, and
that its inverse function f −1 : C → C is also a linear function.

Note also that the class of all circles and lines in C is carried to itself by all linear functions as well
as the inversion function. We now try to generalize these two types of functions.

Definition. A Möbius transformation, or a bilinear transformation, is a rational function T : C → C

of the form

az + b
(2) T (z) = ,
cz + d

where a, b, c, d ∈ C are fixed and ad − bc = 0. We write formally


d a
(3) T − =∞ and T (∞) = .
c c

Remarks. (1) Note that if ad − bc = 0, then

ad − bc
T
(z) = = 0,
(cz + d)2

so that T (z) is constant. Hence the requirement ad − bc = 0 is essential.

(2) To justify (3), note that the function T (z) has a simple pole at z = −d/c, and that

a
lim T (z) = .
|z|→∞ c
13–6 W W L Chen : Introduction to Complex Analysis

(3) Since
ad − bc
T
(z) = = 0
(cz + d)2

for every z ∈ C satisfying z = −d/c, it follows that a Möbius transformation is conformal at every point
in C where it is analytic.

(4) The case c = 0 and d = 1 reduces to T (z) = az + b, a linear function.

(5) The case a = d = 0 and b = c = 1 reduces to T (z) = 1/z, the inversion function.

(6) If c = 0, then it is easy to check that


az + b a ad 1
(4) = + b− .
cz + d c c cz + d

(7) Writing w = T (z), then (2) can be written in the form cwz − az + dw − b = 0, and this is
linear in both z and w. This is the reason for calling such a function a bilinear transformation.

The following result is a generalization of Theorems 13A and 13C.

THEOREM 13D. Suppose that T : C → C is a Möbius transformation. Then

(a) T is the composition of a sequence of translations, magnifications, rotations and inversions;
(b) T : C → C is one-to-one and onto;
(c) the inverse function T −1 : C → C is also a Möbius transformation;
(d) T maps the class of circles and lines in C to itself; and
(e) for every Möbius transformation S : C → C, S ◦ T : C → C is also a Möbius transformation.

Proof. (a) Suppose that

az + b
T (z) = ,
cz + d

where ad − bc = 0. If c = 0, then we must have ad = 0, so that

az + b a b
T (z) = = z+ .
d d d

In this case, T is a linear function, and the result follows from Theorem 13A. On the other hand, if
c = 0, then we use the identity (4). We can write T = T3 ◦ T2 ◦ T1 , where


1 ad a
T1 (z) = cz + d and T2 (z) = and T3 (z) = b− z+ .
z c c

It is easy to check that T1 and T3 are linear functions, while T2 is the inversion function. The result now
follows from Theorem 13A.

(b) and (d) follow from (a) on noting that translations, magnifications, rotations and inversions all
have the properties in question, and that composition of functions preserves these properties.

(c) and (e) are left as exercises.

 Chapter 13 : Möbius Transformations 13–7

Example 13.3.1. Suppose that a ∈ C is fixed and |a| < 1. Consider the Möbius transformation
T : C → C, given by
a−z
T (z) = ,
1 − az

where a ∈ C denotes the complex conjugate of a. Note that

|a − z|2 |a|2 − 2Re(az) + |z|2

|T (z)|2 = = .
|1 − az|2 1 − 2Re(az) + |a|2 |z|2

It is easy to see that if |z| = 1, then |T (z)| = 1. It follows from Theorem 13D(d) that the image under
T of the unit circle {z : |z| = 1} is the unit circle itself. On the other hand, the inequality |T (z)| < 1 is
equivalent to the inequality

|a|2 + |z|2 < 1 + |a|2 |z|2 ,

which is equivalent to the inequality

(1 − |a|2 )(1 − |z|2 ) > 0,

which is equivalent to the inequality |z| < 1, in view of the assumption |a| < 1. It now follows from this
observation and Theorem 13D(b) that the interior {z : |z| < 1} of the unit circle must be mapped onto
itself by T .

Definition. A fixed point z ∈ C of a Möbius transformation T : C → C is a solution of the equation

T (z) = z.

THEOREM 13E. A Möbius transformation T : C → C has at most two distinct fixed points in C
unless T (z) = z identically.

Proof. Suppose that

az + b
T (z) = ,
cz + d

where ad − bc = 0. If c = 0, then ad = 0, so that T is a linear function. In this case, the equation

T (z) = z becomes az + b = dz. If T (z) is not identically equal to z, then a = d or b = 0, so that this
equation has at most one solution in C. Suppose next that c = 0. Then clearly ∞ is not a fixed point.
The equation T (z) = z is now a quadratic equation, and so has at most two distinct roots in C.

It follows from Theorem 13E that a Möbius transformation must be the identity function if it has
three fixed points. Suppose now that S and T are Möbius transformations, and that there exist distinct
z1 , z2 , z3 ∈ C such that S(zj ) = T (zj ) for j = 1, 2, 3. By Theorem 13D(c)(e), the composition S −1 ◦ T is
also a Möbius transformation. Clearly (S −1 ◦ T )(zj ) = zj for j = 1, 2, 3, so that S −1 ◦ T has three fixed
points, and so must be the identity function. In other words, (S −1 ◦ T )(z) = z, and so S(z) = T (z), for
every z ∈ C. We summarize this observation below.

THEOREM 13F. Suppose that two Möbius transformations S : C → C and T : C → C are equal at
three distinct points in C. Then S(z) = T (z) for every z ∈ C.

Next, we show that three points determine uniquely a Möbius transfomation.

THEOREM 13G. Suppose that z1 , z2 , z3 ∈ C are distinct, and that w1 , w2 , w3 ∈ C are also distinct.
Then there exists a unique Möbius transformation T : C → C such that T (zj ) = wj for j = 1, 2, 3.
13–8 W W L Chen : Introduction to Complex Analysis

Proof. To establish the existence of such a function, note that T1 : C → C, given by

(z − z1 )(z2 − z3 )
T1 (z) = ,
(z − z3 )(z2 − z1 )

is a Möbius transformation, with T1 (z1 ) = 0, T1 (z2 ) = 1 and T1 (z3 ) = ∞. Similarly, T2 : C → C, given

by

(w − w1 )(w2 − w3 )
T2 (w) = ,
(w − w3 )(w2 − w1 )

is a Möbius transformation, with T2 (w1 ) = 0, T2 (w2 ) = 1 and T2 (w3 ) = ∞. Clearly T = T2−1 ◦ T1 is a

Möbius transformation such that T (zj ) = wj for j = 1, 2, 3. The uniqueness follows from Theorem 13F.

Example 13.3.2. To find a Möbius transformation T : C → C such that T (0) = 2, T (1) = 3 and
T (6) = 4, note that T1 : C → C, given by

(z − 0)(1 − 6) −5z
T1 (z) = = ,
(z − 6)(1 − 0) z−6

is a Möbius transformation, with T1 (0) = 0, T1 (1) = 1 and T1 (6) = ∞. Similarly, T2 : C → C, given by

(w − 2)(3 − 4) −w + 2
T2 (w) = = ,
(w − 4)(3 − 2) w−4

is a Möbius transformation, with T2 (2) = 0, T2 (3) = 1 and T2 (4) = ∞. We now have to calculate
T = T2−1 ◦ T1 . Note that
4z + 2
T2−1 (z) = ,
z+1
so that


−5z
4 +2
z−6 −20z + 2(z − 6) −18z − 12 9z + 6
T (z) =
 = = = .
−5z −5z + (z − 6) −4z − 6 2z + 3
+1
z−6

13.4. Finding Particular Möbius Transformations

Recall that a Möbius transformation T : C → C, given by

az + b
T (z) = ,
cz + d

where ad − bc = 0, maps the class of circles and lines in C to itself. Suppose that a circle or line contains
the pole z = −d/c of T , then its image under T is unbounded, and is therefore a line rather than a
circle. Suppose, on the other hand, that a circle or line does not contain the pole z = −d/c of T , then
its image under T cannot contain the point at ∞, and is therefore a circle rather than a line.

Note next that a circle or line splits the extended complex plane into two domains. Here we adopt
the convention that a line contains the point at ∞, whereas a half plane not including its boundary line
does not contain the point at ∞. Since T : C → C is one-to-one and onto, and since an analytic function
 Chapter 13 : Möbius Transformations 13–9

maps domains to domains, it follows that the image of any domain arising from a circle or line must be
mapped onto a domain arising from the image of this circle or line under T .

Remark. Strictly speaking, the function

az + b
T (z) =
cz + d

is one-to-one, onto and analytic if we take the domain of T to be C \ {−d/c} and the codomain of T to
be C \ {a/c}.

Example 13.4.1. Suppose that z0 , w0 ∈ C and r1 , r2 > 0 are fixed, and that we are required to
find a Möbius transformation T : C → C which maps the disc {z : |z − z0 | < r1 } to the annulus
{w : |w − w0 | > r2 }. This can be achieved by taking T = T4 ◦ T3 ◦ T2 ◦ T1 , where

1
T1 (z) = z − z0 and T2 (z) = and T3 (z) = r1 r2 z and T4 (z) = z + w0 .
z
We have the picture below:

r2
r1 z0
w0

T1 (z)=z-z 0 r1-1 T4 (z)=z +w0

r2
r1
T2 (z)= 1z T3(z)=r1 r2 z

Note that T1 is a translation which takes the centre of the disc {z : |z − z0 | < r1 } to the origin. Then the
inversion T2 turns a disc into an annulus. We now apply a magnification T3 and then use the translation
T4 to position the disc so that its centre is at w0 . It is easy to see that

r1 r2 w0 z + (r1 r2 − w0 z0 )
T (z) = + w0 = .
z − z0 z − z0

Example 13.4.2. Suppose that we are required to find a Möbius transformation T which maps the
unit disc {z : |z| < 1} to the right half plane {w : Rew > 0}.

T
1
13–10 W W L Chen : Introduction to Complex Analysis

Our first step is to find a Möbius transformation S : C → C which maps the unit circle {z : |z| = 1} to
the imaginary axis {w : Rew = 0}. For
az + b
S(z) =
cz + d

to map the unit circle to a line, the unit circle must contain the pole z = −d/c of S. Suppose that we
choose the point z = 1 to be this pole. In this case, we may take, for example, c = 1 and d = −1. Next,
some point on the unit circle must have image 0 under S. Suppose that we choose z = −1 to be this
point. In this case, we may take, for example, a = 1 and b = 1. Note that ad − bc = 0, and these choices
give
z+1
S(z) = .
z−1

Note that S(1) = ∞ and S(−1) = 0. We also know that the image of the unit circle under S is a line
through the origin, but at this point, we do not know whether this line is the imaginary axis. To check
what this line is, we use a third point on the unit circle, the point z = i, say. It is easy to check that
S(i) = −i, on the imaginary axis. We therefore conclude that S maps the unit circle to the imaginary
axis. It follows that S maps the unit disc {z : |z| < 1} to one of the half planes arising from the imaginary
axis, but at this point, we do not know whether it is {w : Rew < 0} or {w : Rew > 0}. To check which
half plane this is, we can use the point z = 0. It is easy to check that S(0) = −1. Unfortunately, this
is in {w : Rew < 0} instead of {w : Rew > 0}. This little problem can be eradicated by rotating S(z)
about the origin by an angle π. In other words, the Möbius transformation

−z − 1
T (z) = eiπ S(z) = −S(z) =
z−1

satisfies our requirements.

Example 13.4.3. Suppose that we are required to find a Möbius transformation S which maps the
half plane {z = x + iy : y < x} to the annulus {w : |w − 3| > 5}.

y=x
4i
S
-2 3 8

Our first step is to find a Möbius transformation S : C → C which maps the line {z = x + iy : y = x} to
the circle {w : |w − 3| = 5}. Consider first of all the transformation
√
S1 (z) = 2e−iπ/4 z = (1 − i)z
√
(the magnification here by 2 serves only to simplify the arithmetic), where we attempt to map the line
{z = x + iy : y = x} to the real axis {z : Imz = 0}. Next, we shall find a Möbius transformation S2
which maps the real axis {z : Imz = 0} to the circle {w : |w − 3| = 5}. To do this, we shall use the
Möbius transformation S2 which maps the points 0, 1, ∞, say, on the real axis to the points 8, 4i, −2,
say, on the circle. From the proof of Theorem 13G, the inverse Möbius transformation S2−1 is given by

(w − 8)(4i + 2) w−8
z = S2−1 (w) = = .
(w + 2)(4i − 8) 2i(w + 2)
 Chapter 13 : Möbius Transformations 13–11

Simple calculation gives

4z − 8i
w = S2 (z) = ,
−2z − i

and so the Möbius transformation S = S2 ◦ S1 , given by

4(1 − i)z − 8i
S(z) = (S2 ◦ S1 )(z) = ,
−2(1 − i)z − i

maps the line {z = x + iy : y = x} to the circle {w : |w − 3| = 5}. It follows that S maps the half plane
{z = x + iy : y < x} to the disc {w : |w − 3| < 5} or the annulus {w : |w − 3| > 5}. To check which
this is, we can use the point z = 1. It is easy to check that S(1) = −4 + 4i. This is in the annulus
{w : |w − 3| > 5}. It follows that

4(1 − i)z − 8i
S(z) =
−2(1 − i)z − i

satisfies our requirements.

13.5. Symmetry and Möbius Transformations

Definition. We say that two points z1 , z2 ∈ C are symmetric with respect to a line L if L is the
perpendicular bisector of the line segment joining z1 and z2 .

Suppose that z1 , z2 ∈ C are symmetric with respect to a line L. Then it is easy to see that every
circle or line passing through both z1 and z2 intersects L at right angles.

z1
L

z2

Using this observation, we make the following definition.

Definition. We say that two points z1 , z2 ∈ C are symmetric with respect to a circle C if every circle
or line passing through both z1 and z2 intersects C at right angles.

C
z1 z2
13–12 W W L Chen : Introduction to Complex Analysis

Remarks. (1) Consider the circle C = {z : |z − z0 | = r} with centre z0 and radius r. Then it can be
shown that two points z1 inside C and z2 outside C are symmetric with respect to C if and only if there
exist ρ, θ ∈ R satisfying 0 < ρ < r and such that

r2 iθ
z1 = z0 + ρeiθ and z 2 = z0 + e ;
ρ

in other words, if and only if (z1 − z0 )(z2 − z0 ) = r2 .

(2) We also say that the centre of a circle C and the point at ∞ are symmetric with respect to
the circle C.

(3) Note that a line can be interpreted as a circle of infinite radius. It follows that our definition
covers symmetry with respect to both lines and circles.

THEOREM 13H. (SYMMETRY PRINCIPLE) Suppose that T : C → C is a Möbius transforma-

tion. Suppose further that C is a circle or line in C. Then two points z1 , z2 ∈ C are symmetric with
respect to C if and only if T (z1 ) and T (z2 ) are symmetric with respect to T (C).

Proof. Note that T maps the class of lines and circles in C to itself. Note also that T is conformal at
all points where it is analytic, and so preserves orthogonality.

Example 13.5.1. Let us return to Example 13.3.1. Suppose that a ∈ C is fixed and |a| < 1. Suppose
also that λ ∈ C is fixed and |λ| = 1. Then the Möbius transformation T : C → C, given by

z−a
T (z) = λ ,
az − 1

maps the unit disc D = {z : |z| < 1} onto itself. Note here that we have introduced an extra rotation
λ about the origin. We shall now attempt to show that any Möbius transformation T : C → C which
maps the unit disc D onto itself must be of this form. Clearly T maps the unit circle C = {z : |z| = 1}
onto itself. Next, let a ∈ C be the unique point satisfying T (a) = 0. Then |a| < 1. Suppose now that
a and a∗ are symmetric with respect to the unit circle C. Then by the Symmetry principle, T (a) and
T (a∗ ) are symmetric with respect to the circle T (C) = C. Since T (a) = 0, we must have T (a∗ ) = ∞.
It follows that T (z) must have a zero at z = a and a pole at z = a∗ . Note now that aa∗ = 1, so that
a∗ = 1/a. Hence
z−a
T (z) = λ
az − 1

for some λ ∈ C. Recall now that |T (z)| = 1 whenever |z| = 1. In particular, we require
 
 1 − a

|T (1)| = λ  = 1.
a − 1

It follows that |λ| = 1.

Problems for Chapter 13

1. Find a Möbius transformation that takes the points 0, 2, −2 to the points −2, 0, 2 respectively.

z−i
2. Show that the Möbius transformation w = maps the upper half plane {z : Imz > 0} onto the
z+i
disc {w : |w| < 1}.
 Chapter 13 : Möbius Transformations 13–13

3. Suppose that C is a given circle or line, and that C
is also a given circle or line. Does there exist a
Möbius transformation that maps C onto C
? If so, is this Möbius transformation unique? Justify
your assertions.

4. The cross ratio of four distinct points z1 , z2 , z3 , z4 ∈ C is defined by

(z1 − z2 )(z3 − z4 )
X(z1 , z2 , z3 , z4 ) =
(z1 − z4 )(z3 − z2 )

and by the obvious limit if one of the points is ∞. Show that the cross ratio is invariant under
Möbius transformation; in other words, for every Möbius transformation T : C → C, we have

X(T (z1 ), T (z2 ), T (z3 ), T (z4 )) = X(z1 , z2 , z3 , z4 ).

[Hint: Note that every Möbius transformation is a composition of translations, rotations, magnifi-
cations and inversions.]

5. Use the invariance of the cross ratio to find a Möbius transformation that takes the points 0, 1, ∞
to the points −i, 1, i respectively.
[Hint: Suppose that the Möbius transformation takes z to w.]

6. Show that a Möbius transformation w = f (z) maps the upper half plane {z : Imz > 0} onto the
disc {w : |w| < 1} if and only if it is of the form

z−a
w=λ ,
z−a

where a, λ ∈ C satisfy |λ| = 1 and Ima > 0.

7. a) Construct a one-parameter family of Möbius transformations that map the real axis onto the
unit circle by mapping the points 0, λ, ∞ to the points −i, 1, i respectively, where λ is a non-zero
real parameter.
b) What point of the upper half plane gets mapped to the centre of the circle?
c) For what values of λ is the upper half plane {z : Imz > 0} mapped onto the disc {w : |w| < 1}?
Onto the annulus {w : |w| > 1}?
d) Taking note of Problem 6, comment whether the family includes all Möbius transformations
that map the real axis onto the unit circle.

8. Find all Möbius transformations that map the disc {z : |z − 1| < 2} onto the upper half plane
{w : Imw > 0} and takes z = 1 to w = i.

bz bz
9. Show that if either of the transformations w = a + and w = c + maps the unit disc
1 − cz 1 − az
onto the unit disc, then they both do.

10. Find a transformation that maps A = {z = x + iy : |z| < 1 and y > 0}, the upper half of the unit
disc, onto the first quadrant.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 14
SCHWARZ-CHRISTOFFEL TRANSFORMATIONS

14.1. Introduction

Recall that a function f (z) is conformal at every point where it is analytic and has non-zero derivative.
In this chapter, we shall study the situation at points where f (z) is not conformal.

Suppose that x0 ∈ R is fixed. Consider a function f (z) with derivative

f
(z) = (z − x0 )α ,

where −1 < α < 1. Here we have chosen the branch of the argument so that
π 3π
− < arg(z − x0 ) ≤ ,
2 2
introducing a branch cut along the axis {x0 + iy : y ≤ 0}. We shall study the image of the real axis
under this mapping f .

Suppose first of all that z lies on the real axis and z > x0 . Then f (z) is conformal at such a point
z, since f
(z) = 0. Note also that

arg f
(z) = α arg(z − x0 ) = 0

for all such points z, ignoring multiples of 2π. Since the tangent at every point of the half line (x0 , ∞) has
slope 0, it follows that the tangent at every point of the image curve f ((x0 , ∞)) has slope arg f
(z) = 0.
Hence f ((x0 , ∞)) is a half line parallel to the real axis and has left hand end point f (x0 ).

Suppose next that z lies on the real axis and z < x0 . Again f (z) is conformal at such a point z,
since f
(z) = 0. Note also that

arg f
(z) = α arg(z − x0 ) = απ

14–2 W W L Chen : Introduction to Complex Analysis

for all such points z, again ignoring multiples of 2π. It follows easily that f ((−∞, x0 )) is a half line
making an angle απ with the horizontal axis.

Summarizing the above, we have the following diagram which describes the image of the real axis
under f .
;
;;;;;
; ; ; ; ;;;;;
;;
;;;;;
f (x0 ); ; ; ; ; ; ; απ
oo
•
o
oo
ooo
oo
oo
ooo
o
oo
ooo
o
oo

14.2. A Generalization

Again, suppose that x0 ∈ R is fixed. Consider a function f (z) with derivative

f
(z) = λ(z − x0 )α ,

where λ ∈ C is non-zero and −1 < α < 1. Then

arg f
(z) = arg λ + α arg(z − x0 ).

In other words, there is an extra rotation by arg λ from the case in the previous section. This leads to
the following diagram which describes the image of the real axis under f .

;;
;; ;;;
;; ;;;
;; ;;;
;;;
oo
oo
; ; ; απ ooo
;; o; ; oooarg λ
;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
f (x0 )•o

Suppose now that x1 , . . . , xk ∈ R are fixed, and that x1 < . . . < xk . Consider a function f (z) with
derivative

(1) f
(z) = λ(z − x1 )α1 . . . (z − xk )αk ,

where λ ∈ C is non-zero and −1 < α1 , . . . , αk < 1. Then

arg f
(z) = arg λ + α1 arg(z − x1 ) + . . . + αk arg(z − xk ).

It is easy to see that if z is on the real axis, then


 arg λ if z > xk ,


 arg λ + αk π
 if xk−1 < z < xk ,


arg f (z) = ..
 .



 arg λ + α2 π + . . . + αk π if x1 < z < x2 ,
arg λ + α1 π + . . . + αk π if z < x1 .
 Chapter 14 : Schwarz-Christoffel Transformations 14–3

This leads to the following diagram which describes the image of the real axis under f .

;;; ;;; oo
;; ;;; ;;;
;;; oo
oo

;;; ;;;
o
;;; αk π oo
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
•; ;
f (x1 ) oo arg λ
• ?f (x3 ) ?o
o •
;
; α1 π f (xk ) ??
; ? ?
; ?
; ? ?
; ?
; ?
;  ?
; ?
; ?
; 
? f (xk−1 )
; ?
; 
f (x2 )•−α2 π O _ _ _ •
;; O
;; O

;; O
;; O

; O 

Suppose now that a function f (z) satisfies (1). Then it is analytic on the complex plane C with a
few branch cuts at x1 , . . . , xk . More precisely, it is analytic in the domain

C \ ({x1 + iy : y ≤ 0} ∪ . . . ∪ {xk + iy : y ≤ 0}).

It follows that for any z ∈ H, where H denotes the upper half plane, we can write
 


(2) f (z) = f (ζ) dζ + B = λ (ζ − x1 )α1 . . . (ζ − xk )αk dζ + B.
[z0 ,z] [z0 ,z]

Here, z0 is a suitably chosen point in H or its boundary. Also, for every z ∈ H, [z0 , z] denotes the
straight line segment from z0 to z.

Definition. A function f (z) of the form (2) is called a Schwarz-Christoffel transformation.

14.3. Polygons

Note that the function (2) maps the real axis onto a polygonal path. We now wish to construct a
one-to-one analytic function that maps the upper half plane H onto the interior of a given polygon P .
The idea is to tailor a Schwarz-Christoffel transformation to achieve this.

Suppose that the vertices of the polygon P are given by w1 , . . . , wk in the anticlockwise direction.
Let us follow the edges of the polygon P . At vertex wj , suppose that we make a right turn of angle θj π,
where −1 < θj < 1, with the convention that θj < 0 denotes a left turn.

;;;;;
;;;;; ;;;;;;
;;;; ;;;;;;
;;;
•; ; ;
wj
@
pp @ θj π
ppp @
@
ppp @
pp @
wj−1 p @
pp • wj+1•

Since P is a polygon and its vertices are given in the anticlockwise direction, we must have

θ1 π + . . . + θk π = −2π.

It is an elementary fact in geometry that if we know the vertices w1 , . . . , wk−1 and angles θ1 π, . . . , θk−1 π
of the polygon P , then the last vertex wk and angle θk π are uniquely determined. The idea is therefore
14–4 W W L Chen : Introduction to Complex Analysis

to find real numbers x1 < . . . < xk−1 to act as preimages of the vertices w1 , . . . , wk−1 , and to assume
that x = ∞ is the preimage of the vertex wk .

Suppose that x1 < . . . < xk−1 . Clearly the function


g(z) = (ζ − x1 )θ1 . . . (ζ − xk−1 )θk−1 dζ
[z0 ,z]

maps the real line onto some polygon Q of k sides. However, the polygon Q may not be the polygon
P , but at least it has the required right hand turn angles θ1 , . . . , θk−1 at the vertices g(x1 ), . . . , g(xk−1 ).
We can adjust the lengths of the sides of the polygon Q by choosing x1 , . . . , xk−1 carefully, so that Q is
similar to the polygon P . Once this is achieved, we can then map the polygon Q to the polygon P by a
linear transformation.

We state, without proof, the following important result.

THEOREM 14A. Suppose that P is a polygon with vertices w1 , . . . , wk in the anticlockwise direction,
with corresponding right turns of angles θ1 π, . . . , θk π respectively, where −1 < θ1 , . . . , θk < 1. Then there
exists a function of the form

f (z) = A (ζ − x1 )θ1 . . . (ζ − xk−1 )θk−1 dζ + B,
[z0 ,z]

where A, B ∈ C, that maps the upper half plane H one-to-one and conformally onto the interior of P ,
with

f (x1 ) = w1 , ..., f (xk−1 ) = wk−1 , f (∞) = wk .

Remarks. (1) Note that we do not even need to have very precise information on wk and θk .

(2) Certain infinite regions can sometimes be thought of as infinite polygons. In this case, it is
sometimes convenient to take wk as the point at infinity, as we need no information on the angle θk
when we use Theorem 14A.

(3) It can be shown that a Schwarz-Christoffel transformation can be uniquely determined by three
points, as is the case for Möbius transformations. This can be interpreted as three degrees of freedom
in our construction of the transformation. One of these is used by taking f (∞) = wk . We can therefore
afford to choose x1 and x2 freely, subject to the restriction that −∞ < x1 < x2 < ∞.

(4) Occasionally, we may choose extra points apart from x1 and x2 due to symmetry properties
of the polygon P . We shall illustrate this point in Examples 14.4.3–14.4.5 below.

(5) Note that the integrals involved may be impossible to calculate in practice. Numerical tech-
niques are often used. However, we shall not discuss these here.

14.4. Examples

Example 14.4.1. We wish to find a Schwarz-Christoffel transformation that maps the upper half plane
H to the inside of the triangle with vertices at −1, 0 and i. The boundary of the triangle is described
by the solid edges in the picture below.



 i
;
; 
;

; 
;;
_ _ _ 
_ _ _  _ _ _
−1 0


 Chapter 14 : Schwarz-Christoffel Transformations 14–5

Let us write, in our notation,

w1 = i, w2 = −1, w3 = 0,

so that

θ1 = −3/4, θ2 = −3/4, θ3 = −1/2.

Following Theorem 14A, we consider a function of the form


f (z) = A (ζ − x1 )−3/4 (ζ − x2 )−3/4 dζ + B.
[z0 ,z]

We may choose x1 = −1 and x2 = 1, and obtain, using z0 = 0,

 
f (z) = A (ζ + 1)−3/4 (ζ − 1)−3/4 dζ + B = A (ζ 2 − 1)−3/4 dζ + B.
[0,z] [0,z]

We need f (−1) = i and f (1) = −1. It follows that

 −1  1
A (ζ 2 − 1)−3/4 dζ + B = i and A (ζ 2 − 1)−3/4 dζ + B = −1.
0 0

Writing
 1
κ= (ζ 2 − 1)−3/4 dζ,
0

we have

−Aκ + B = i and Aκ + B = −1,

so that
−1 − i i−1
A= and B= .
2κ 2
Hence

−1 − i i−1
f (z) = (ζ 2 − 1)−3/4 dζ + .
2κ [0,z] 2

Example 14.4.2. We wish to find a Schwarz-Christoffel transformation that maps the upper half plane
H to the set

P = {z = x + iy : x > 0 and y > 0} ∪ {z = x + iy : x ≤ 0 and y > 1}.

The boundary of P is described by the solid edges in the picture below.




 i


_ _ _ _ _ _ _ _ _  _ _ _
0


14–6 W W L Chen : Introduction to Complex Analysis

Let us write, in our notation,

w1 = i, w2 = 0, w3 = ∞,

so that

θ1 = 1/2 and θ2 = −1/2.

Following Theorem 14A, we consider a function of the form


f (z) = A (ζ − x1 )1/2 (ζ − x2 )−1/2 dζ + B
.
[z0 ,z]

We may choose x1 = −1 and x2 = 1, and obtain

   1/2
ζ +1
f (z) = A (ζ + 1)1/2 (ζ − 1)−1/2 dζ + B
= A dζ + B

[z0 ,z] [z0 ,z] ζ −1

= A (z − 1) 2 1/2
+ log(z + (z − 1) 2 1/2
) + B.

We shall omit some of the painful analysis, and claim that we can choose a branch of the function which
is analytic in the upper half plane H. We need f (−1) = i and f (1) = −1. It follows that by choosing a
suitable branch of the logarithm, we have

A log(−1) + B = i and A log 1 + B = 0,

so that A = 1/π and B = 0. Hence

1 2
f (z) = (z − 1)1/2 + log(z + (z 2 − 1)1/2 ) .
π

Example 14.4.3. We wish to find a Schwarz-Christoffel transformation that maps the upper half
plane H to the inside of the rectangle with vertices at ±1 and ±1 + i. The boundary of the rectangle is
described by the solid edges in the picture below.



−1+i i 1+i



_ _ _  _ _ _
−1 1



Let us write, in our notation,

w1 = −1 + i, w2 = −1, w3 = 1, w4 = 1 + i, w5 = i

(here we have used an extra point w5 in order to create some symmetry; see Remark (4) in the previous
section), so that

θ1 = θ2 = θ3 = θ4 = −1/2 and θ5 = 0.

Following Theorem 14A, we consider a function of the form


f (z) = A (ζ − x1 )−1/2 (ζ − x2 )−1/2 (ζ − x3 )−1/2 (ζ − x4 )−1/2 dζ + B.
[z0 ,z]
 Chapter 14 : Schwarz-Christoffel Transformations 14–7

We shall choose

x1 = −α, x2 = −1, x3 = 1, x4 = α,

where α > 1 will be determined later. Note that we are attempting to benefit from the symmetry here.
With such a choice, we obtain, using z0 = 0,

f (z) = A (ζ + α)−1/2 (ζ + 1)−1/2 (ζ − 1)−1/2 (ζ − α)−1/2 dζ + B
[0,z]
 
dζ
=A (ζ 2 − 1)−1/2 (ζ 2 − α2 )−1/2 dζ + B = A
+ B.
[0,z] [0,z] (1 − ζ 2 )(α2 − ζ 2 )

We need

f (−α) = −1 + i, f (−1) = −1, f (1) = 1, f (α) = 1 + i.

It follows that
 −α
dζ
(3) A
+ B = −1 + i,
0 (1 − ζ 2 )(α2 − ζ 2 )
 −1
dζ
(4) A
+ B = −1,
0 (1 − ζ 2 )(α2 − ζ 2)
 1
dζ
(5) A
+ B = 1,
0 (1 − ζ 2 )(α2 − ζ 2)
 α
dζ
(6) A
+ B = 1 + i.
0 (1 − ζ 2 )(α2 − ζ 2 )

Subtracting (4) from (3) and subtracting (5) from (6), we obtain respectively

 −α  α
dζ dζ
A
=i and A
= i,
−1 (1 − ζ 2 )(α2 − ζ 2 ) 1 (1 − ζ 2 )(α2 − ζ 2 )

which are in fact the same equation (note that symmetry is at work here). Multiplying the denominator
by i, we obtain
 α
dζ
(7) A
= 1.
1 (ζ 2 − 1)(α2 − ζ 2 )

On the other hand, if B = 0, then (4) and (5) are the same, and can be represented by

 1
dζ
(8) A
= 1.
0 (1 − ζ 2 )(α2 − ζ 2 )

It follows that our choice of α should be made so that

 1  α
dζ dζ

=
.
0 (1 − ζ 2 )(α2 − ζ 2) 1 (ζ − 1)(α2 − ζ 2 )
2

We can then take A to be the reciprocal of the common value of these two integrals.
14–8 W W L Chen : Introduction to Complex Analysis

Example 14.4.4. We wish to find a Schwarz-Christoffel transformation that maps the upper half plane
H to the domain

P = C \ {z = x ± i : x ≤ 0}.

The boundary of the set P is described by the solid edges in the picture below when the point w2 is
taken to infinity along the negative real axis.



j i
jjjj
jjjj
j
jjjjjj 
jjjj
jjjj
jjjjjjj 
Tj _ _ _ _ _ _ _ _  _ _ _
j
_ _ _ T
j
_
w2 TTTTTTTTT
TTTT
TTTT 
TTTT
TTTT
TTTT 
TTT
 −i



Let us write, in our notation,

w1 = i, w2 = ∞, w3 = −i, w4 = ∞

(note again the symmetry; see Remark (4) in the previous section), so that

θ1 = 1, θ2 = −1, θ3 = 1.

Following Theorem 14A, we consider a function of the form


f (z) = A (ζ − x1 )(ζ − x2 )−1 (ζ − x3 ) dζ + B
.
[z0 ,z]

We shall choose

x1 = −1, x2 = 0, x3 = 1,

and note that we are attempting to benefit from the symmetry here. We obtain
 
−1

f (z) = A (ζ + 1)ζ (ζ − 1) dζ + B = A (ζ 2 − 1)ζ −1 dζ + B

[z0 ,z] [z0 ,z]
    2

1 z
=A ζ− dζ + B
= A − log z + B.
[z0 ,z] ζ 2

We need

f (−1) = i, f (0) = −∞, f (1) = −i.

It follows that by choosing a suitable branch of the logarithm, we have

   
1 1
A − iπ + B = i and A − 0 + B = −i,
2 2

so that A = −2/π and B = 1/π − i. Hence

   
2 z2 1
f (z) = − − log z + −i .
π 2 π

Note that |f (z)| → ∞ as z → 0.

 Chapter 14 : Schwarz-Christoffel Transformations 14–9

Example 14.4.5. We wish to find a Schwarz-Christoffel transformation that maps the upper half plane
H to the domain

P = H \ {z = yi : y ≤ 1}.

The boundary of the set P is described by the solid edges in the picture below.



i


0



Let us write, in our notation,

w1 = 0, w2 = i, w3 = 0, w4 = ∞

(note again the symmetry as well as the use of the point 0 twice), so that

θ1 = −1/2, θ2 = 1, θ3 = −1/2.

Following Theorem 14A, we consider a function of the form


f (z) = A (ζ − x1 )−1/2 (ζ − x2 )(ζ − x3 )−1/2 dζ + B
.
[z0 ,z]

We shall choose

x1 = −1, x2 = 0, x3 = 1,

and note again that we are attempting to benefit from the symmetry here. We obtain
 
−1/2 −1/2

f (z) = A (ζ + 1) ζ(ζ − 1) dζ + B = A (ζ 2 − 1)−1/2 ζ dζ + B
= A(z 2 − 1)1/2 + B.
[z0 ,z] [z0 ,z]

We need

f (−1) = 0, f (0) = i, f (1) = 0.

It follows that by choosing a suitable branch of the function which is positive for large positive z, we
have

Ai + B = i and B = 0,

so that A = 1 and B = 0. Hence

f (z) = (z 2 − 1)1/2 .

Problems for Chapter 14

1. Use these notes and without reproducing proofs, find a transformation that maps the unit disc
D = {z : |z| < 1} onto the domain D
= H \ {z = yi : y ≤ 1}, where H denotes the upper half plane.
14–10 W W L Chen : Introduction to Complex Analysis

2. For each of the sets A below, find a Schwarz-Christoffel transformation that maps the upper half
plane H onto the set A: √
a) A is an open triangular region with vertices ±1 and i 3.
b) A is the region above the polygonal path

{z = x + i : x ≤ 0} ∪ {z = x + (1 − x)i : 0 ≤ x ≤ 1} ∪ {z = x : x ≥ 1}.

c) A = {z = x + iy : y > 0 or |x| < 1}.

 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 15
LAPLACE’S EQUATION REVISITED

15.1. Use of Möbius Transformations

Recall that Laplace’s equation involves finding a harmonic function in a given region and which satisfies
given boundary conditions. In this chapter, we shall illustrate very briefly the use of transformations to
simplify this problem. Note, however, that we are not discussing the general problem of the solution of
Laplace’s equation; that is a topic in partial differential equations. Here we shall satisfy ourselves on
how to use a few simple cases of Laplace’s equation to obtain solutions in more complicated situations.
We first discuss an example which uses Möbius transformations.

Example 15.5.1. Consider the lens region formed by the intersection of the two discs

{z ∈ C : |z + 1|2 < 2} ∩ {z ∈ C : |z − 1|2 < 2}.

√
Here the two discs both have radius 2 and are centred at z = −1 and z = 1 respectively (see the picture
on the next page). Suppose that we are required to find a harmonic function φ in this region with φ = 1
on the right hand boundary and φ = 0 on the left hand boundary. Note that both the right hand and
left hand boundaries are parts of circles and intersect at z = ±i. If we use a Möbius transformation with
pole at z = i, then both boundaries are transformed into straight lines. Let us try the transformation
z+i
w = f (z) = .
z−i
√ √
Then f (i) = ∞ and f (−i) = 0. By considering, for example, f ( 2 − 1) and f (1 − 2), it is not difficult
to show that the right hand and left hand boundaries are transformed into the half lines arg w = 3π/4
and arg w = 5π/4 respectively. Note also that f (0) = −1. It follows that the lens region is transformed
into the region


3π 5π
w∈C: < arg w < .
4 4
15–2 W W L Chen : Introduction to Complex Analysis

We summarize the above discussion in the pictures below:

z w
i

-1 1 -1
-i

It is easy to check that the function

 
2 5π
ψ(w) = − arg w
π 4

is harmonic in this region and satisfies ψ(w) = 1 when arg w = 3π/4 and ψ(w) = 0 when arg w = 5π/4.
It follows that our required harmonic function is given by
  
2 5π z+i
φ(z) = − arg .
π 4 z−i

15.2. Use of Schwarz-Christoffel Transformations

We now discuss three examples which use Schwarz-Christoffel transformations.

Example 15.2.1. We wish to find a non-constant harmonic function in the region above the polygonal
path given in Example 14.4.2, with boundary condition φ = 0 on the polygonal path. Here φ = const can
be interpreted as lines of flow on a river over a step on the river bed. Recall that the Schwarz-Christoffel
transformation
1 2 
f (z) = (z − 1)1/2 + log(z + (z 2 − 1)1/2 )
π
maps the upper half plane onto the region in question. We now need to find a non-constant harmonic
function ψ on the upper half plane with boundary condition ψ = 0 on the real line. For example, the
function

ψ(z) = Imz

satisfies the requirements. We now need to invert the function f (z) to obtain a harmonic function

φ(w) = Im(f −1 (w))

in the original region.

Example 15.2.2. We wish to find a harmonic function in the slit plane given in Example 14.4.4, with
boundary conditions φ = 1 on the upper slit and φ = −1 on the lower slit. Here φ = const can be
interpreted as equipotential lines in a region around two semi-infinite conducting plates with opposite
charges. Recall that the Schwarz-Christoffel transformation
   
2 z2 1
f (z) = − − log z + −i
π 2 π
 Chapter 15 : Laplace’s Equation Revisited 15–3

maps the upper half plane onto the region in question. Furthermore, it maps the negative and positive
real axis onto the upper and lower slits respectively. We now need to find a harmonic function ψ on the
upper half plane with boundary conditions ψ = 1 on the negative real axis and ψ = −1 on the positive
real axis. For example, the function
2
ψ(z) = arg z − 1
π

satisfies the requirements (here we take the principal value of the argument). We now need to invert the
function f (z) to obtain a harmonic function

2
φ(w) = arg(f −1 (w)) − 1
π
in the original region.

Example 15.2.3. We wish to find a non-constant harmonic function in the slit upper half plane given
in Example 14.4.5, with boundary condition φ = 0 on the slit and the real axis. Here φ = const can be
interpreted as lines of flow past an obstacle. Recall that the Schwarz-Christoffel transformation

f (z) = (z 2 − 1)1/2

maps the upper half plane onto the region in question. We now need to find a harmonic function ψ on
the upper half plane with boundary conditions ψ = 0 on the real line. For example, the function

ψ(z) = Imz

satisfies the requirements. Note now that

f −1 (w) = (w2 + 1)1/2 .

We therefore obtain the harmonic function

φ(w) = Im((w2 + 1)1/2 )

in the original region. Here we choose a branch of the square root that is positive for large positive w.
 INTRODUCTION TO COMPLEX ANALYSIS
W W L CHEN

c W W L Chen, 1996, 2003.
This chapter originates from material used by the author at Imperial College, University of London, between 1981 and 1990.
It is available free to all individuals, on the understanding that it is not to be used for financial gains,
and may be downloaded and/or photocopied, with or without permission from the author.
However, this document may not be kept on any information storage and retrieval system without permission
from the author, unless such system is not accessible to any individuals other than its owners.

Chapter 16
UNIFORM CONVERGENCE

16.1. Uniform Convergence of Sequences

Recall that if a sequence an of complex numbers converges to a, then, given any
> 0, there exists
N ∈ R such that |an − a| <
whenever n > N .

We can extend this to pointwise convergence in a region D ⊆ C in a natural way. A sequence of

complex valued functions an (z) defined on D converges pointwise to a function a(z) defined on D if,
given any
> 0 and any z ∈ D, there exists N ∈ R such that |an (z) − a(z)| <
whenever n > N . Here
the value of N may depend on the choice of z ∈ D. Indeed, for any fixed z ∈ D, we simply consider the
convergence of the sequence an (z) of complex numbers to the complex number a(z). The region D does
not play any essential part in the argument apart from providing the complex numbers z in question.

In this chapter, we introduce the idea of uniformity to the question of convergence. Put simply,
uniformity transfers the dependence of N on z to dependence of N only on the region D containing the
complex numbers z in question. More precisely, we have the following definition.

Definition. Suppose that D ⊆ C is a region. We say that a sequence of complex valued functions
an (z) converges uniformly in D to a function a(z), denoted by an (z) → a(z) as n → ∞ uniformly in D,
if, given any
> 0, there exists N ∈ R such that for every z ∈ D, |an (z) − a(z)| <
whenever n > N .

Remark. Note that N no longer depends on the choice of z ∈ D. Note also that a precise definition
can be given by requiring N ∈ R to satisfy

sup |an (z) − a(z)| <

z∈D

whenever n > N .
16–2 W W L Chen : Introduction to Complex Analysis

Example 16.1.1. Consider the sequence

z
an (z) =
n

in the region D = {z : |z| < 1}. Note first of all that for every fixed z ∈ D, we have an (z) → 0 as
n → ∞. On the other hand, given any
> 0, we have, for every z ∈ D, that

|z| 1
|an (z) − 0| = < <

n n

whenever n > 1/
. Hence an (z) → 0 as n → ∞ uniformly in D. Now consider the same sequence in the
region D = C. Note that

|z|
|an (z) − 0| <
if and only if n> .

It is therefore impossible to find a suitable N independent of the choice of z ∈ C. Hence an (z) converges
to 0, but not uniformly, in C.

16.2. Consequences of Uniform Convergence

In this section, we show that uniform convergence carries a number of properties of the sequence over
to the limit function. The following three results concern respectively continuity, integrability and
differentiability.

THEOREM 16A. Suppose that for every n ∈ N, the function an (z) is continuous in a region D ⊆ C.
Suppose further that an (z) → a(z) as n → ∞ uniformly in D. Then a(z) is continuous in D.

Proof. Suppose that z0 ∈ D is fixed. For every z ∈ D, we have

a(z) − a(z0 ) = a(z) − an (z) + an (z) − an (z0 ) + an (z0 ) − a(z0 ),

so that

(1) |a(z) − a(z0 )| ≤ |an (z) − a(z)| + |an (z) − an (z0 )| + |an (z0 ) − a(z0 )|.

Given any
> 0, there exists N (independent of the choice of z ∈ D) such that

(2) |an (z) − a(z)| < and |an (z0 ) − a(z0 )| <
3 3

whenever n > N . We now choose any n > N and consider the function an (z). Clearly this function is
continuous at z0 . Hence given any
> 0, there exists δ > 0 such that


(3) |an (z) − an (z0 )| < whenever |z − z0 | < δ.
3

Combining (1)–(3), we conclude that |a(z) − a(z0 )| <
whenever |z − z0 | < δ, so that a(z) is continuous
at z0 . Since z0 ∈ D is arbitrary, the result follows.

Example 16.2.1. Consider the sequence an (z) = z n on the real interval [0, 1]. Each function an (z) is
clearly continuous in [0, 1]. Also an (z) → 0 as n → ∞ if z ∈ [0, 1) and an (1) → 1 as n → ∞, so that the
limit function is not continuous in [0, 1]. In view of Theorem 16A, it is clear that this discontinuity is
caused by the lack of uniform convergence of an (z) in [0, 1].
 Chapter 16 : Uniform Convergence 16–3

THEOREM 16B. Suppose that for every n ∈ N, the function an (z) is continuous in a region D ⊆ C.
Suppose further that an (z) → a(z) as n → ∞ uniformly in D. Then for any contour C lying in D, we
have



lim an (z) dz = a(z) dz.
n→∞ C C

Proof. Note first of all that the integrals exist, since integrability over C is a consequence of continuity
in D. Suppose now that the contour C has length L. Given any
> 0, there exists N ∈ R such that for
every z ∈ D, |an (z) − a(z)| <
/L whenever n > N . Then



 
 an (z) dz − a(z) dz  ≤ L sup |an (z) − a(z)| ≤

  z∈C
C C

whenever n > N .

THEOREM 16C. Suppose that for every n ∈ N, the function an (z) is analytic in a disc D = {z :
|z − z0 | < R}. Suppose further that an (z) → a(z) as n → ∞ uniformly in Dr = {z : |z − z0 | ≤ r} for
every r ∈ [0, R). Then a(z) is analytic in D, and an (z) → a (z) as n → ∞ uniformly in Dr for every
r ∈ [0, R).

Proof. Suppose that T is any triangular path in D. We now choose r ∈ [0, R) so that T ⊆ Dr . Then



a(z) dz = lim an (z) dz = 0.
T n→∞ T

Here the second equality follows from Cauchy’s integral theorem, while the first equality follows from
Theorem 16B, in view of uniform convergence in Dr . The assertion that a(z) is analytic in D now
follows from Morera’s theorem (Theorem 6G). Suppose next that r ∈ [0, R) is fixed. We now choose
ρ = (r + R)/2, so that r < ρ < R, and let Cρ denote the circle {ζ : |ζ − z0 | = ρ}, followed in the
positive (anticlockwise) direction (the reader is advised to draw a picture). For every z ∈ Dr , we have,
by Cauchy’s integral formula, that


  1 an (ζ) − a(ζ)
an (z) − a (z) = dζ.
2πi Cρ (ζ − z)2

Note that for every ζ ∈ Cρ , we have |ζ − z| ≥ ρ − r. Also, in view of the uniform convergence of the
sequence an (z) in Dρ , we have, given any
> 0, there exists N such that for every z ∈ Dρ ,

(ρ − r)2

|an (z) − a(z)| <
ρ
whenever n > N . It follows that for every z ∈ Dr , we have
 
 an (ζ) − a(ζ) 
|an (z) − a (z)| < ρ sup  ≤

ζ∈Cρ (ζ − z)2 

whenever n > N . Hence an (z) → a (z) as n → ∞ uniformly in Dr .

Note that Theorem 16C is restricted to discs. However, as far as application is concerned, this is
not a serious restriction. For any point z in an arbitrary domain D ⊆ C, we can always find an open
disc D such that z ∈ D ⊆ D, and so we can apply Theorem 16C to the disc D . We immediately have
the following result.

THEOREM 16D. Suppose that for every n ∈ N, the function an (z) is analytic in a domain D ⊆ C.
Suppose further that an (z) → a(z) as n → ∞ uniformly in D. Then a(z) is analytic in D. Furthermore,
(k)
for every z ∈ D and every k ∈ N, we have an (z) → a(k) (z) as n → ∞.
16–4 W W L Chen : Introduction to Complex Analysis

16.3. Cauchy Sequences

Suppose that a sequence of complex numbers an converges to a. Then given any
> 0, there exists
N ∈ R such that |an − a| <
/2 whenever n > N . It follows that

|an − am | ≤ |an − a| + |am − a| <

whenever m, n > N .

Definition. We say that a sequence of complex numbers an is a Cauchy sequence if, given any
> 0,
there exists N ∈ R such that |an − am | <
whenever m, n > N .

In the last section of this chapter, we shall prove the following result.

THEOREM 16E. (GENERAL PRINCIPLE OF CONVERGENCE) A sequence of complex num-

bers an is convergent if and only if it is a Cauchy sequence. In other words, a sequence an of complex
numbers is convergent if and only if, given any
> 0, there exists N ∈ R such that |an − am | <

whenever m, n > N .

Definition. Suppose that D ⊆ C is a region. We say that a sequence of complex valued functions
an (z) is a uniform Cauchy sequence in D, if, given any
> 0, there exists N ∈ R such that for every
z ∈ D, |an (z) − am (z)| <
whenever m, n > N .

We have the following important result.

THEOREM 16F. (GENERAL PRINCIPLE OF UNIFORM CONVERGENCE) Suppose that D ⊆

C is a region. A sequence of complex valued functions an (z) converges uniformly in D if and only if it
is a uniform Cauchy sequence in D.

Proof. It is simple to show that uniform convergence implies uniform Cauchy. To prove the converse,
note that for every fixed z ∈ D, the sequence of complex numbers an (z) is a Cauchy sequence. It follows
from Theorem 16E that an (z) converges to a(z), say. Since an (z) is a uniform Cauchy sequence in D,
it follows that, given any
> 0, there exists N ∈ R such that for every z ∈ D, |an (z) − am (z)| <

whenever m, n > N . Letting m → ∞, we conclude that |an (z) − a(z)| ≤
whenever n > N .

16.4. Uniform Convergence of Series

Recall that the convergence of a series depends on the convergence of the sequence of partial sums.

Definition. Suppose that D ⊆ C is a region. We say that a series of complex valued functions
∞

an (z)
n=1

converges uniformly in D if the sequence of partial sums


N
sN (z) = an (z)
n=1

converges uniformly in D.

We immediately have the following analogues of Theorems 16A, 16B, 16D, 16E and 16F. They can
be established by applying the earlier results to the sequence of partial sums.
 Chapter 16 : Uniform Convergence 16–5

THEOREM 16G. Suppose that for every n ∈ N, the function an (z) is continuous in a region D ⊆ C.
Suppose further that the series
∞

an (z)
n=1

converges uniformly to a function s(z) in D. Then s(z) is continuous in D.

THEOREM 16H. Suppose that for every n ∈ N, the function an (z) is continuous in a region D ⊆ C.
Suppose further that the series
∞

an (z)
n=1

converges uniformly to a function s(z) in D. Then for any contour C lying in D, we have

∞



an (z) dz = s(z) dz.
n=1 C C

In other words, we can interchange the order of summation and integration.

THEOREM 16J. Suppose that for every n ∈ N, the function an (z) is analytic in a domain D ⊆ C.
Suppose further that the series
∞

an (z)
n=1

converges uniformly to a function s(z) in D. Then s(z) is analytic in D. Furthermore, for every z ∈ D
and every k ∈ N, we have
∞

an(k) (z) = s(k) (z).
n=1

In other words, we can interchange the order of summation and differentiation.

THEOREM 16K. (GENERAL PRINCIPLE OF CONVERGENCE) A series

∞

an
n=1

of complex numbers converges if and only if, given any
> 0, there exists N0 ∈ R such that

 N 
 2 
 
 an  <

 
n=N1 +1

whenever N2 > N1 > N0 .

16–6 W W L Chen : Introduction to Complex Analysis

THEOREM 16L. (GENERAL PRINCIPLE OF UNIFORM CONVERGENCE) Suppose that D ⊆

C is a region. A series
∞

an (z)
n=1

of complex valued functions converges uniformly in D if and only if, given any
> 0, there exists N0 ∈ R
such that for every z ∈ D,
 N 
  2 
 
 an (z) <

 
n=N1 +1

whenever N2 > N1 > N0 .

We can also establish the following uniform versions of the Comparison test and the Ratio test.

THEOREM 16M. (WEIERSTRASS M -TEST) Suppose that D ⊆ C is a region. Suppose further

that an (z) is a sequence of complex valued functions such that |an (z)| ≤ Mn for every z ∈ D, where the
real series
∞

Mn
n=1

of non-negative terms is convergent. Then the series

∞

an (z)
n=1

converges uniformly (and absolutely) in D.

Proof. Using the Triangle inequality, we have

 N 
 2  
N2 
N2
 
 an (z) ≤ |an (z)| ≤ Mn .
 
n=N1 +1 n=N1 +1 n=N1 +1

Given any
> 0, it follows from Theorem 16K that there exists N0 such that


N2
Mn <

n=N1 +1

whenever N2 > N1 > N0 . It follows that for every z ∈ D,

 N 
 2 
 
 an (z) <

 
n=N1 +1

whenever N2 > N1 > N0 . The result now follows from Theorem 16L.

 Chapter 16 : Uniform Convergence 16–7

THEOREM 16N. (RATIO TEST) Suppose that D ⊆ C is a region. Suppose further that an (z) is
a sequence of complex valued functions such that a1 (z) is bounded in D, and
 
 an+1 (z) 
 
(4)  an (z)  ≤ R < 1

for every z ∈ D, where R is constant. Then the series

∞

an (z)
n=1

converges uniformly (and absolutely) in D.

Proof. Note that the condition (4) implies |an (z)| ≤ Rn−1 |a1 (z)| for every n ∈ N. On the other hand,
there exists M ∈ R such that |a1 (z)| ≤ M for every z ∈ D. It follows that for every z ∈ D and every
n ∈ N, we have |an (z)| ≤ M Rn−1 . The result now follows from the Weierstrass M -test, noting that the
geometric series
∞

M Rn−1
n=1

converges.

Example 16.4.1. The series

∞
1
(5) ζ(z) =
n=1
nz

converges absolutely for every z satisfying Rez > 1. To see this, note that writing z = x + iy, where
x, y ∈ R, we have

1 1 1 1 1
= x+iy = x n−iy = x e−iy log n = x (cos(y log n) − i sin(y log n)),
nz n n n n
so that
 
 1 
 = 1 .
 nz  n x

Since x > 1, the series

∞
1
n x
n=1

of non-negative terms is convergent. It follows from the Comparison test that the series (5) converges
absolutely. Suppose now that δ > 0 is fixed. Consider the region D = {z : Rez > 1 + δ}. Then for every
z ∈ D, we have
 
 1 
 = 1 < 1 .
 nz  nx n1+δ

The series
∞
 1
n 1+δ
n=1
16–8 W W L Chen : Introduction to Complex Analysis

of non-negative terms is convergent. It follows from the Weierstrass M -test that the series (5) converges
uniformly in D. We comment here that the series (5) is called the Riemann zeta function, and is crucial
in the study of the distribution of prime numbers. Indeed, the study of this function has led to much of
the development in complex analysis.

Example 16.4.2. In Chapter 10, we discussed the function π cot πz, and showed that it has simple
poles at the (real) integers with residue 1. Here we shall make a more detailed study. Consider the
function
∞
1  2z
f (z) = + .
z n=1 z 2 − n2

Let us first of all study this function in the region DR = {z : |z| < R}, where R > 0 is fixed. Let N ∈ N
satisfy N > 2R, and write f (z) = f1 (z) + f2 (z), where
∞
1  2z 
N
2z
f1 (z) = + and f2 (z) = .
z n=1 z 2 − n2 z 2 − n2
n=N +1

Clearly the function f1 (z) is analytic in DR , with the exception of simple poles at the (real) integers in
DR . Consider next the function f2 (z) in DR . For every z ∈ DR and every n > N > 2R, we have
 
 2z  2R 1 2R 8R
 
 z 2 − n2  ≤ n2 − R2 = n2 1 − (R/n)2 < 3n2 .

It follows from the Weierstrass M -test that the series for f2 (z) converges uniformly in DR , and is analytic
in DR in view of Theorem 16J. Hence f (z) is analytic in DR , with the exception of simple poles at the
(real) integers in DR . It follows that f (z) is meromorphic in C, with simple poles at the (real) integers.
It is easy to check that all these simple poles have residue 1. Note also that we can write
 1
f (z) = z .
z2 − n2
n∈Z

We shall show that f (z) = π cot πz. For convenience, we shall change notation, and show that
 1 π cot πa
(6) =
a2 − n2 a
n∈Z

whenever a ∈ Z. Consider the function

π cot πz
g(z) = .
a2 − z 2
Since the function π cot πz has simple poles at every n ∈ Z with residue 1, and since a ∈ Z, it follows
that g(z) has simple poles at every n ∈ Z and at z = ±a, with residues

1 π cot πa
res(g, n) = and res(g, ±a) = − .
a2 − n2 2a
For every N ∈ N, let CN denote the boundary of the rectangular domain
 
1
z = x + iy : |x| < N + and |y| < N ,
2

followed in the positive (anticlockwise) direction. If N > |a|, then we have


1 π cot πz 1 π cot πa
dz = − .
2πi CN a − z
2 2 a −n
2 2 a
−N ≤n≤N
 Chapter 16 : Uniform Convergence 16–9

Clearly (6) will follow if we show that the integral on the left hand side converges to 0 as N → ∞. It
can be shown that | cot πz| ≤ coth π for every z ∈ CN . Hence for every N > |a|, we have

  
 π cot πz   π cot πz  (8N + 2)π coth π
 
dz  ≤ (8N + 2) sup  2 ≤ → 0 as N → ∞.
 a2 − z 2 z∈C a − z2  N 2 − |a|2
CN N

16.5. Application to Power Series

Let z, α ∈ C. In this section, we shall study series of the type

∞

(7) an (z − α)n (a0 , a1 , a2 , . . . ∈ C),
n=0

known commonly as power series.

THEOREM 16P. Suppose that the series given by (7) converges for a particular value z = z0 . Then,
for every r < |z0 − α|, the series converges uniformly (and absolutely) in the disc Dr = {z : |z − α| ≤ r}.

Proof. Suppose that

∞

an (z0 − α)n
n=0

converges. Then an (z0 − α)n → 0 as n → ∞, and so there exists M ∈ R such that |an (z0 − α)n | ≤ M
for every n ∈ N ∪ {0}. For every z ∈ Dr , we have
   n
 z − α n  
|an (z − α)n | ≤ M   ≤M r 
z0 − α   z0 − α 

for every n ∈ N ∪ {0}. The result now follows from the Weierstrass M -test, noting that the geometric
series
∞
  n
 r 
M 
n=0
z0 − α 

converges.

THEOREM 16Q. (CONVERGENCE THEOREM FOR POWER SERIES) For the power series
given by (7), exactly one of the following holds:
(a) The series converges absolutely for every z ∈ C.
(b) There exists a positive real number R such that the series converges absolutely for every z ∈ C
satisfying |z − α| < R and diverges for every z ∈ C satisfying |z − α| > R.
(c) The series diverges for every z = α.

Sketch of Proof. In the notation of Theorem 16P, consider

S = {r ≥ 0 : (7) converges absolutely in Dr }.

Then S contains the number 0. In view of Theorem 16P, S must be an interval with lower end-point 0,
so that S = [0, ∞), S = {0} or there exists some positive number R such that S = [0, R) or S = [0, R].
The first two possibilities correspond to (a) and (c) respectively, while the last possibility corresponds
to (b).

16–10 W W L Chen : Introduction to Complex Analysis

Definition. The number R in Theorem 16Q is called the radius of convergence of the series (7). We
also say that R = 0 if case (c) occurs, and that R = ∞ if case (a) occurs.

We now show that differentiation of a power series can be carried out term by term, and that the
series so obtained converges to the derivative.

THEOREM 16R. Suppose that the power series given by (7) has radius of convergence R > 0. Then
it represents an analytic function f (z) in the open disc D = {z : |z − α| < R}. Furthermore, the
derivatives of f (z) can be obtained by differentiating the series term by term.

Proof. For every r < R, it follows from Theorem 16P that the series converges uniformly in the disc
Dr = {z : |z −α| < r}. It now follows from Theorem 16J that the series converges to an analytic function
f (z) in Dr , and the derivatives of f (z) can be obtained by differentiating the series term by term. Since
the above holds for any r < R, the result follows.

Example 16.5.1. Suppose that f (t) is a complex valued function continuous (and so bounded) on the
closed real interval [0, 1]. Consider the function

1
F (z) = e−zt f (t) dt.
0

For any fixed z ∈ C, we have the power series (here t is the variable)

∞
(−zt)n
(8) e−zt = ,
n=0
n!

with infinite radius of convergence. It follows from Theorem 16P that the series (8) converges uniformly
in [0, 1], and so can be multiplied by the bounded function f (t) and integrated term by term, in view of
Theorem 16H. Hence
∞

 1 ∞

(−zt)n (−z)n 1 n
(9) F (z) = f (t) dt = t f (t) dt.
n=0 0 n! n=0
n! 0

Furthermore, if |f (t)| ≤ M , where M is a fixed positive number, then




 1  1
M
 tn f (t) dt ≤ M tn dt = .
 n+1
0 0

Suppose now that R > 0 is fixed. If |z| < R, then



 (−z)n 1 n  n
 t f (t) dt  ≤ MR .
 n!  (n + 1)!
0

Note that the series

∞
M Rn
n=0
(n + 1)!

converges, so it follows from the Weierstrass M -test that the series in (9) converges uniformly in the disc
{z : |z| < R}. By Theorem 16J, the function F (z) is analytic in {z : |z| < R}. Since R > 0 is arbitrary,
it follows that F (z) is entire.
 Chapter 16 : Uniform Convergence 16–11

16.6. Cauchy Sequences

In this section, we shall prove Theorem 16E. Clearly a convergent sequence of complex numbers is
Cauchy. It remains to show that a Cauchy sequence of complex numbers is convergent.

The proof of this result usually involves the Bolzano-Weierstrass theorem which states that every
bounded sequence of complex numbers has a convergent subsequence. Here, we shall give a proof without
using the Bolzano-Weierstrass theorem.

Assume, first of all, that the sequence an is real. Since an is a Cauchy sequence, it follows that
there exists an increasing sequence of natural numbers

N1 < N2 < . . . < Np < . . .

such that
1
|an − am | <
2p

whenever n, m ≥ Np (we simply take
= 2−p for every p ∈ N). In particular, we have

1
|aNp+1 − aNp | <
2p

for every p ∈ N. For every p ∈ N, let

1
bp = aNp − .
2p−1

Then
1 1
bp+1 − bp = aNp+1 − aNp + ≥ p − |aNp+1 − aNp | > 0,
2p 2

so that the sequence bp is increasing. Note next that

 
 1  1 1 1
|bp | = aNp − ≤ |aNp − aN1 | + |aN1 | + p−1 ≤ + |aN1 | + p−1 ,
2p−1  2 2 2

so that the sequence bp is bounded. Hence the sequence bp converges to L, say, as p → ∞.

We now show that an → L as n → ∞. Given any
> 0, we now choose p ∈ N so large that

1


< and |bp − L| < .
2p 4 4

Suppose that n ≥ Np . Then

1 1

|an − L| ≤ |an − aNp | + |aNp − bp | + |bp − L| < p
+ p−1 + <

2 2 4

as required.

Suppose now that the sequence an is complex valued. Then we can write an = xn + iyn , where
xn , yn ∈ R. If an is a Cauchy sequence, then it is easy to see that the real sequences xn and yn are real
Cauchy sequences. It follows that both xn and yn converge, and so an converges.
16–12 W W L Chen : Introduction to Complex Analysis

Problems for Chapter 16

1. Suppose that an (z) → a(z) and bn (z) → b(z) as n → ∞ uniformly in a region D.

a) Show that an (z) + bn (z) → a(z) + b(z) as n → ∞ uniformly in D.
b) Suppose that f (z) is bounded in D. Show that an (z)f (z) → a(z)f (z) as n → ∞ uniformly in
D.
c) Write f (z) = 1/z and an (z) = 1/n. Find a region D such that an (z) converges uniformly in D
but an (z)f (z) does not converge uniformly in D.

2. For each of the following power series, find a number R such that the series converges for |z| < R
and diverges for |z| > R:
∞ ∞
a) 2n z n b) n2 z n
n=0 n=1
∞ n 2n ∞
2 z 3n z n
c) 2+n
d)
n=1
n n=0
4n+ 5n

3. Show that each of the following represents an entire function:

∞ ∞
zn zn
a) b)
n=1
(n!)1/2 n=1
2n2
∞
 1
c) n nz
n=1
2

4. Show that each of the following functions is meromorphic in C, and find the residues at the poles:
∞ ∞

(−1)n 1
a) b)
n=0
n!(n + z) n=1
(z + n)2

∞
 1  π 2
5. Show that for every z ∈ Z, we have 2
= .
n=−∞
(n + z) sin πz

∞
 z π
6. a) Show that except at the poles, we have = .
n=−∞
n2 +z 2 tanh πz
∞
 1
b) By writing the series as 1/z plus a sum over all natural numbers, evaluate 2 + n2
.
n=1
z
∞
1 π2
c) By letting z → 0, show that 2
= .
n=1
n 6

7. Consider the exponential series

∞
zn
n=0
n!

which converges for every z ∈ C. Suppose further that e(z) is the sum of the series.
a) Show that the series converges uniformly in the disc DR = {z : |z| < R} for every real number
R > 0.
b) Suppose that D is a bounded region in C. Explain why the series converges uniformly in D.
c) Show that for every z ∈ C satisfying |z| = R, we have
   
 
M
z n  RM
 1 1 1 1 1
 ≥ − RM + 2 + . . . + M −N −1 ≥ RM − .
 n!  M! R R R M! R − 1
n=N +1
 Chapter 16 : Uniform Convergence 16–13

d) Use (c) to show that the series does not converge uniformly in C.
e) Explain carefully why e(z) is an entire function in C.
[Remark: In view of the unfavourable conclusion of (d), you should take extra care here.]
f) Show that e (z) = e(z) for every z ∈ C and e(0) = 1.
g) Let g(z) = e(−z)e(z). Show that g  (z) = 0 for every z ∈ C, and deduce that e(−z)e(z) = 1 for
every z ∈ C.
h) Suppose that a ∈ C is fixed. By studying the function ga (z) = e(−z)e(z + a), show that
e(z + a) = e(z)e(a) for every z ∈ C.

8. This question makes use of the function e(z) discussed in Problem 7. Suppose that for every z ∈ C,
we write

e(iz) + e(−iz) e(iz) − e(−iz)

c(z) = and s(z) = .
2 2i

a) By using the Taylor series for e(iz) and e(−iz), find the Taylor series for c(z) and s(z).
b) Show that c (z) = −s(z) and s (z) = c(z) for every z ∈ C.
c) By studying the function h(z) = c2 (z) + s2 (z), show that c2 (z) + s2 (z) = 1 for every z ∈ C.