This page intentionally left blank

Applied Complex Variables for Scientists and Engineers

Second Edition

Yue Kuen Kwok

Applied Complex Variables for
Scientists and Engineers
Second Edition

Yue Kuen Kwok

Hong Kong University of Science and Technology
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
São Paulo, Delhi, Dubai, Tokyo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521701389

© Y. K. Kwok 2010

This publication is in copyright. Subject to statutory exception and to the

provision of relevant collective licensing agreements, no reproduction of any part
may take place without the written permission of Cambridge University Press.
First published in print format 2010

ISBN-13 978-0-511-77500-0 eBook (EBL)

ISBN-13 978-0-521-70138-9 Paperback

Cambridge University Press has no responsibility for the persistence or accuracy

of urls for external or third-party internet websites referred to in this publication,
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
 Contents

Preface page ix
1 Complex Numbers 1
1.1 Complex numbers and their representations 1
1.2 Algebraic properties of complex numbers 4
1.2.1 De Moivre’s theorem 7
1.3 Geometric properties of complex numbers 13
1.3.1 nth roots of unity 16
1.3.2 Symmetry with respect to a circle 17
1.4 Some topological definitions 23
1.5 Complex infinity and the Riemann sphere 29
1.5.1 The Riemann sphere and stereographic projection 30
1.6 Applications to electrical circuits 33
1.7 Problems 36
2 Analytic Functions 46
2.1 Functions of a complex variable 46
2.1.1 Velocity of fluid flow emanating from a source 48
2.1.2 Mapping properties of complex functions 50
2.1.3 Definitions of the exponential and trigonometric
functions 53
2.2 Limit and continuity of complex functions 54
2.2.1 Limit of a complex function 54
2.2.2 Continuity of a complex function 58
2.3 Differentiation of complex functions 61
2.3.1 Complex velocity and acceleration 63
2.4 Cauchy–Riemann relations 64
2.4.1 Conjugate complex variables 69

v
vi Contents

2.5 Analyticity 70
2.6 Harmonic functions 74
2.6.1 Harmonic conjugate 75
2.6.2 Steady state temperature distribution 80
2.6.3 Poisson’s equation 84
2.7 Problems 85
3 Exponential, Logarithmic and Trigonometric Functions 93
3.1 Exponential functions 93
3.1.1 Definition from the first principles 94
3.1.2 Mapping properties of the complex exponential
function 97
3.2 Trigonometric and hyperbolic functions 97
3.2.1 Mapping properties of the complex sine function 102
3.3 Logarithmic functions 104
3.3.1 Heat source 106
3.3.2 Temperature distribution in the upper half-plane 108
3.4 Inverse trigonometric and hyperbolic functions 111
3.5 Generalized exponential, logarithmic, and power
functions 115
3.6 Branch points, branch cuts and Riemann surfaces 118
3.6.1 Joukowski mapping 123
3.7 Problems 126
4 Complex Integration 133
4.1 Formulations of complex integration 133
4.1.1 Definite integral of a complex-valued function of a
real variable 134
4.1.2 Complex integrals as line integrals 135
4.2 Cauchy integral theorem 142
4.3 Cauchy integral formula and its consequences 151
4.3.1 Derivatives of contour integrals 153
4.3.2 Morera’s theorem 157
4.3.3 Consequences of the Cauchy integral formula 158
4.4 Potential functions of conservative fields 162
4.4.1 Velocity potential and stream function of fluid
flows 162
4.4.2 Electrostatic fields 175
4.4.3 Gravitational fields 179
4.5 Problems 183
 Contents vii

5 Taylor and Laurent Series 194

5.1 Complex sequences and series 194
5.1.1 Convergence of complex sequences 194
5.1.2 Infinite series of complex numbers 196
5.1.3 Convergence tests of complex series 197
5.2 Sequences and series of complex functions 200
5.2.1 Convergence of series of complex functions 201
5.2.2 Power series 206
5.3 Taylor series 215
5.4 Laurent series 221
5.4.1 Potential flow past an obstacle 230
5.5 Analytic continuation 233
5.5.1 Reflection principle 236
5.6 Problems 238
6 Singularities and Calculus of Residues 248
6.1 Classification of singular points 248
6.2 Residues and the Residue Theorem 255
6.2.1 Computational formulas for evaluating residues 257
6.3 Evaluation of real integrals by residue calculus 268
6.3.1 Integrals of trigonometric functions over [0, 2π ] 268
6.3.2 Integrals of rational functions 269
6.3.3 Integrals involving multi-valued functions 271
6.3.4 Miscellaneous types of integral 275
6.4 Fourier transforms 278
6.4.1 Fourier inversion formula 279
6.4.2 Evaluation of Fourier integrals 285
6.5 Cauchy principal value of an improper integral 288
6.6 Hydrodynamics in potential fluid flows 295
6.6.1 Blasius laws of hydrodynamic force and moment 295
6.6.2 Kutta–Joukowski’s lifting force theorem 299
6.7 Problems 300
7 Boundary Value Problems and Initial-Boundary
Value Problems 311
7.1 Integral formulas of harmonic functions 312
7.1.1 Poisson integral formula 312
7.1.2 Schwarz integral formula 319
7.1.3 Neumann problems 324
7.2 The Laplace transform and its inversion 326
7.2.1 Bromwich integrals 330
viii Contents

7.3 Initial-boundary value problems 336

7.3.1 Heat conduction 337
7.3.2 Longitudinal oscillations of an elastic thin rod 341
7.4 Problems 346
8 Conformal Mappings and Applications 358
8.1 Conformal mappings 358
8.1.1 Invariance of the Laplace equation 364
8.1.2 Hodograph transformations 372
8.2 Bilinear transformations 375
8.2.1 Circle-preserving property 378
8.2.2 Symmetry-preserving property 381
8.2.3 Some special bilinear transformations 390
8.3 Schwarz–Christoffel transformations 399
8.4 Problems 409
Answers to Problems 419
Index 434
 Preface

This textbook is intended to be an introduction to complex variables for mathe-

matics, science and engineering undergraduate students. The prerequisites are
some knowledge of calculus (up to line integrals and Green’s Theorem), though
basic familiarity with differential equations would also be useful.
Complex function theory is an elegant mathematical structure on its own. On
the other hand, many of its theoretical results provide powerful and versatile
tools for solving problems in physical sciences and other branches of math-
ematics. The book presents the important analytical concepts and techniques
in deriving most of the standard theoretical results in introductory complex
function theory. I have included the proofs of most of the important theorems,
except for a few that are highly technical. This book distinguishes itself from
other texts in complex variables by emphasizing how to use complex vari-
able methods. Throughout the text, many of the important theoretical results
in complex function theory are followed by relevant and vivid examples in
physical sciences. These examples serve to illustrate the uses and implications
of complex function theory. They are drawn from a wide range of physical
and engineering applications, like potential theory, steady state temperature
problems, hydrodynamics, seepage flows, electrostatics and gravitation. For
example, after discussing the mathematical foundations of the Laplace trans-
form and Fourier transform, I show how to use the transform methods to solve
initial-boundary problems arising from heat conduction and wave propagation
problems. The materials covered in the book equip students with the analytical
concepts of complex function theory together with the technical skills to apply
complex variable methods to physical problems.
Throughout the whole textbook, both algebraic and geometric tools are
employed to provide the greatest understanding, with many diagrams illustrat-
ing the concepts introduced. The book contains some 340 stimulating exercises,
with solutions given to most of them. They are intended to aid students to grasp

ix
x Preface

the concepts covered in the text and foster the skills in applying complex vari-
able techniques to solve physical problems. Students are strongly advised to
work through as many exercises as possible since mathematical knowledge can
only be gained through active participation in the thinking and learning process.
The book begins by carefully exploring the algebraic, geometric and topo-
logical structures of the complex number field. In order to visualize the complex
infinity, the Riemann sphere and the corresponding stereographic projection are
introduced. Applications of complex numbers in electrical circuits are included.
Analytic functions are introduced in Chapter 2. The highlights of the chapter
are the Cauchy–Riemann relations and harmonicity. The uses of complex func-
tions in describing fluid flows and steady state heat distributions are illustrated.
In Chapter 3, the complex exponential function is introduced as an entire
function which is equal to its derivative. The description of steady state temper-
ature distributions by complex logarithm functions is illustrated. The mapping
properties of complex trigonometric functions are examined. The notion of
Riemann surfaces is introduced to help visualize multi-valued complex func-
tions.
Complex integration forms the cornerstone of complex variable theory. The
key results in Chapter 4 are the Cauchy–Goursat theorem and the Cauchy
integral formulas. Other interesting results include Gauss’ mean value theorem,
Liouville’s theorem and the maximum modulus theorem. The link of analytic
functions and complex integration with the study of conservative fields is
considered. Complex variable methods are seen to be effective analytical tools
to solve conservation field models in potential flows, gravitational potentials
and electrostatics.
Complex power series are the main themes in Chapter 5. We introduce
different types of convergence of series of complex functions. The various
tests that examine the convergence of complex series are discussed. The Taylor
series theorem and Laurent series theorem show that a convergent power series
is an analytic function within its disk or annulus of convergence, respectively.
The notion of analytic continuation of a complex function is discussed. As an
application, the solution to the potential flow over a perturbed circle is obtained
as a power series in a perturbation parameter.
In Chapter 6, we start with the discussion of the classification of isolated
singularities by examining the Laurent series expansion in a deleted neighbor-
hood of the singularity. We then examine the theory of residues and illustrate the
applications of the calculus of residues in the evaluation of complex integrals.
The concept of the Cauchy principal value of an improper integral is intro-
duced. Fourier transforms and Fourier integrals are considered. The residue
 Preface xi

calculus method is applied to compute the hydrodynamic lift and moment of

an immersed obstacle.
The solutions of boundary value problems and initial-boundary value prob-
lems are considered in Chapter 7. The Poisson integral formula and the Schwarz
integral formula for Dirichlet problems are derived. The inversion of the Laplace
transform via the Bromwich contour integral is discussed. The Laplace trans-
form techniques are applied to obtain the solutions of initial-boundary value
problems arising from heat conduction and wave propagation models.
In the last chapter, we explore the rich geometric structure of complex vari-
able theory. The geometric properties associated with mappings represented by
complex functions are examined. The link between analyticity and conformal-
ity is derived. Various types of transformations that perform the mappings of
regions are introduced. The bilinear and Schwarz–Christoffel transformations
are discussed in full context. A wide range of physical examples are included
to illustrate how to use these transformations to transform conservative field
problems with complicated configurations into those with simple geometries.
We also show how to use the hodograph transformations to solve seepage flow
problems.
I would like to thank Ms Odissa Wong for her careful typing and editing of
the manuscript, and her patience in entertaining the seemingly endless changes
in the process. Also, I would like to thank the staff of Cambridge University
Press for their editorial assistance in the production of this book. Last but not
least, special thanks go to my wife Oi Chun and our two daughters, Grace and
Joyce, for their forbearance while this book was written. Their love and care
have been my main source of support in everyday life and work.
 1
Complex Numbers

In this chapter, we explore the algebraic, geometric and topological structures

of the complex number field C. Readers are assumed to have some basic
knowledge of the real number field R. The construction of C serves as a
resolution of the failure to find a real number x that satisfies the simple quadratic
equation x 2 + 1 = 0.
There are similarities and differences between C and R. For example, R is
an ordered field while C is not. Also, the notions of infinity in R and C are
different. Indeed, in order to visualize the complex infinity, the Riemann sphere
and the corresponding stereographic projection are introduced. In addition, we
introduce various topological properties of point sets in the complex number
field. The chapter finishes with a discussion of the applications of complex
numbers in alternating current circuit analysis.

1.1 Complex numbers and their representations

A complex number z is defined as an ordered pair

z = (x, y) , (1.1.1)

where x and y are both real numbers. We commonly write

z = x + iy ,

where i is a symbol yet to be defined. The operations of addition and multipli-

cation of complex numbers will be defined in a meaningful manner, which will
be seen to observe i 2 = −1 [see eq. (1.2.2)]. The set of all complex numbers is
denoted by C. The real numbers x and y in eq. (1.1.1) are called the real part
and imaginary part of z, respectively. Symbolically, we write

Re z = x, Im z = y .

1
2 Complex Numbers

If y = 0 in eq. (1.1.1), then (x, 0) = x reduces to a real number. This shows that
the set of real numbers R is a proper subset of C. When x = 0 in eq. (1.1.1),
(0, y) = iy is called a pure imaginary number. Putting x = 0 and y = 1, we
obtain the special number i = (0, 1).
Since complex numbers are defined as ordered pairs, two complex numbers
(x1 , y1 ) and (x2 , y2 ) are equal if and only if both their real parts and imaginary
parts are equal, that is,

(x1 , y1 ) = (x2 , y2 ) if and only if x1 = x2 and y1 = y2 .

A complex number z = (x, y) is defined by the pair of real numbers x and y,

so it is natural to assume a one-to-one correspondence between the complex
number z = x + iy and the point (x, y) in the x-y plane. We refer to that plane
as the complex plane or the z-plane.
Sometimes, it may be convenient to use polar coordinates

x = r cos θ , y = r sin θ . (1.1.2)

Accordingly, we define the modulus of z to be

|z| = r = x 2 + y 2 . (1.1.3)

The modulus |z| is the distance between the origin and the point (x, y), which
represents z = x + iy in the complex plane. Obviously, Re z ≤ |z| and Im
z ≤ |z|. The polar representation of the complex number z is written as

z = x + iy = r(cos θ + i sin θ ) . (1.1.4)

We call θ, denoted by arg z, the argument of z. Geometrically, arg z represents

the angle between the positive x-axis and the line segment joining the origin and
the point (x, y) (see Figure 1.1). Because of the periodicity of trigonometric
functions, there are infinitely many values for arg z, each differing from the
others by multiples of 2π. The principal value of arg z, denoted by Arg z,
is the particular value of arg z chosen within the principal interval (−π, π].
Summarizing, we have

arg z = Arg z + 2kπ, k is any integer and Arg z ∈ (−π, π]. (1.1.5)

For example: Arg 1 = 0, Arg(−1) = π , arg(−1 + i) = 3π

4
+ 2kπ, where k is
any integer. Note that arg 0 is not defined.
From eq. (1.1.2), we deduce a simple computational formula for finding
Arg z:
y
tan(Arg z) = . (1.1.6)
x
 1.1 Complex numbers and their representations 3

z = (x, y)

|z|

arg z

Figure 1.1. Vectorial representation of a complex number in the complex plane.

 
As a note of caution, tan−1 yx returns a value in the interval  
(− π2 , π2 ]. Therefore, we adjust the value for Arg z by adding π to tan−1 yx
if (x, y) lies in the second quadrant or subtracting π from tan−1 yx if (x, y) lies
in the third quadrant.
  For example:
 Arg(1 − i) = − π4 while Arg(−1 + i) = 3π 4
,
−1
though tan−1 1 = tan −1 = − π4 .
1

In the complex plane, any point that lies on the x-axis represents a real
number. Therefore, the x-axis is termed the real axis. Similarly, any point on
the y-axis represents an imaginary number, so the y-axis is called the imaginary
axis.
The complex number z = x + iy may also be regarded as the vector xi + yj
in the complex plane, where i and j are the respective unit vectors along the
x-axis and y-axis. In this representation, |z| is visualized as the length of the
vector and arg z as the angle included between the vector and the positive real
axis (see Figure 1.1).
The complex conjugate z of a given complex number z = x + iy is defined
by

z = x − iy . (1.1.7)

In the complex plane, the conjugate z = (x, −y) is the reflection of the point
z = (x, y) with respect to the real axis. From eqs. (1.1.3) and (1.1.6), we observe

|z| = |z| and Arg z = −Arg z .

The polar representation of z is seen to be

z = r(cos θ − i sin θ ) .

Also, |z|2 = zz and z is real if and only if z = z.

4 Complex Numbers

Example 1.1.1 Suppose z = x + iy. Prove the following inequalities:

1
√ (|x| + |y|) ≤ |z| ≤ |x| + |y| .
2

Solution First, we consider

2|z|2 − (|x| + |y|)2 = 2(|x|2 + |y|2 ) − (|x|2 + |y|2 + 2|x||y|)
= (|x| − |y|)2 ≥ 0 ,
so that
1
(|x| + |y|)2 ≤ |z|2 .
2
Since moduli are non-negative, we take the square root on both sides of the
above inequality and obtain
1
√ (|x| + |y|) ≤ |z| .
2
Note that equality holds only if |x| = |y|. In an analogous manner, we consider
|z|2 = |x|2 + |y|2 ≤ |x|2 + |y|2 + 2|x||y| = (|x| + |y|)2 ,
so that upon taking the square root of both sides, we obtain
|z| ≤ |x| + |y| .
When either x = 0 or y = 0, equality holds in the above inequality. Combining
the two inequalities, we obtain
1
√ (|x| + |y|) ≤ |z| ≤ |x| + |y| .
2

1.2 Algebraic properties of complex numbers

Given two complex numbers z1 = (x1 , y1 ) and z2 = (x2 , y2 ), we define the
operations of addition z1 + z2 and multiplication z1 z2 to be
(x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ), (1.2.1a)
(x1 , y1 )(x2 , y2 ) = (x1 x2 − y1 y2 , x1 y2 + x2 y1 ). (1.2.1b)
When the complex numbers are restricted to real numbers, that is, y1 = 0 and
y2 = 0, the above formulas reduce to
(x1 , 0) + (x2 , 0) = (x1 + x2 , 0),
(x1 , 0)(x2 , 0) = (x1 x2 , 0),
 1.2 Algebraic properties of complex numbers 5

which agree with the usual operations of addition and multiplication in the real
number field. These properties must be observed since the complex number
field is an extension of the real number field.
Consider the square of the complex number i = (0, 1). By performing mul-
tiplication according to the multiplication rule (1.2.1b), we have

i 2 = (0, 1)(0, 1) = (−1, 0) = −1 . (1.2.2)

If we write (x1 , y1 ) = x1 + iy1 and (x2 , y2 ) = x2 + iy2 , the addition and mul-
tiplication formulas can be formally written as

(x1 + iy1 ) + (x2 + iy2 ) = (x1 + x2 ) + i(y1 + y2 ) ,

(x1 + iy1 )(x2 + iy2 ) = (x1 x2 − y1 y2 ) + i(x1 y2 + x2 y1 ) .

One observes that the formulas can be obtained by formally treating addi-
tion and multiplication of complex numbers as operations on real numbers,
enforcing i 2 = −1, and collecting the real and imaginary parts separately.
The addition and multiplication of complex numbers obey the familiar com-
mutative, associative and distributive rules like those for real numbers.

1. Addition is commutative:

z 1 + z2 = z 2 + z1 . (1.2.3a)

2. Addition is associative:

z1 + (z2 + z3 ) = (z1 + z2 ) + z3 . (1.2.3b)

3. Multiplication is commutative:

z 1 z 2 = z2 z 1 . (1.2.3c)

4. Multiplication is associative:

z1 (z2 z3 ) = (z1 z2 )z3 . (1.2.3d)

5. Multiplication is distributive over addition:

z1 (z2 + z3 ) = z1 z2 + z1 z3 . (1.2.3e)

The proofs of these rules are straightforward and they are left as exercises.
According to rules (1.2.1a) and (1.2.1b), we have 0 + z = z and 1 · z = z for
any complex number z. Therefore, the numbers 0 and 1 retain their ‘identity’
properties in the complex number field. The additive and multiplicative inverses
6 Complex Numbers

of z, denoted by −z and 1
z
respectively, are defined by

z + (−z) = 0 , (1.2.4a)
 
1
z = 1. (1.2.4b)
z
Suppose z = x + iy. Then its additive inverse −z = −x − iy. For the multi-
plicative inverse 1z , it is seen that

1 z z x − iy
= = 2 = 2 .
z zz |z| x + y2
The subtraction of two complex numbers z1 and z2 can be defined via the
additive inverse by

z1 − z2 = z1 + (−z2 ) = (x1 − x2 ) + i(y1 − y2 ) .

z1
Similarly, the division operation (provided z2 = 0) can be defined via the
z2
multiplicative inverse as
 
z1 1 (x1 x2 + y1 y2 ) + i(x2 y1 − x1 y2 )
= z1 = .
z2 z2 x22 + y22
As a rule of thumb, while addition and subtraction are mostly performed
using the cartesian forms, multiplication and division are more conveniently
performed in polar representations. Suppose

z1 = r1 (cos θ1 + i sin θ1 ) and z2 = r2 (cos θ2 + i sin θ2 );

it can be shown by direct computation that

z1 z2 = r1 r2 [cos(θ1 + θ2 ) + i sin(θ1 + θ2 )], (1.2.5a)

z1 r1
= [cos(θ1 − θ2 ) + i sin(θ1 − θ2 )]. (1.2.5b)
z2 r2
One infers directly from the above equations that

arg(z1 z2 ) = arg z1 + arg z2 , (1.2.6a)

 
z1
arg = arg z1 − arg z2 . (1.2.6b)
z2
Since arg z is multi-valued, eq. (1.2.6a) is interpreted as: arg(z1 z2 ) is equal
to the sum of some value of arg z1 and some value of arg z2 . It is rela-
tively straightforward to verify the following properties of conjugates and
 1.2 Algebraic properties of complex numbers 7

moduli:
z1 z1
(i) z1 + z2 = z1 + z2 , (ii) z1 z2 = z1 z2 , (iii) = ,
  z2 z2
 z1  |z1 |
(iv) |z1 z2 | = |z1 ||z2 |, (v)   = .
z2 |z2 |

In view of the satisfaction of the commutative, associative and distributive

rules and the existence of additive and multiplicative inverses, the set of complex
numbers C constitutes a field. Recall that the set of real numbers R is a
proper subset of C. However, the ordering property of R cannot be extended
to the field C. For any pair of real numbers a and b, the ordering property
means a < b, a = b or a > b. It is meaningless to write z1 < z2 when z1
and z2 are complex numbers unless both z1 and z2 happen to be real. To be
an ordered field, the following property must be observed. Taking any non-
zero complex number z in C, we must have either z or −z being positive,
and z2 = (−z)2 being positive. However, we have i 2 = (−i)2 = −1, which is
negative. A contradiction is encountered, so C cannot be an ordered field.

1.2.1 De Moivre’s theorem

Any complex number with unit modulus can be expressed as

cos θ + i sin θ.

Interestingly, this complex number is related to a complex exponential quantity

by the following formula:

eiθ = cos θ + i sin θ. (1.2.7)

This is the renowned Euler formula, the rigorous justification of which is

presented when the formal definition of the exponential complex function is
introduced in Section 3.1. From the Euler formula, we may deduce that

(cos θ + i sin θ)n = (eiθ )n = einθ = cos nθ + i sin nθ, (1.2.8)

where n can be any integer. This result is the statement of de Moivre’s theorem.
To justify the theorem, we consider the following cases:

(i) The theorem is trivial when n = 0.

(ii) When n is a positive integer, the theorem can be proved easily by
mathematical induction.
8 Complex Numbers

(iii) When n is a negative integer, let m = −n so that m is a positive integer.

We then have
1 1
(cos θ + i sin θ)n = =
(cos θ + i sin θ) m cos mθ + i sin mθ
cos mθ − i sin mθ
=
(cos mθ + i sin mθ )(cos mθ − i sin mθ )
= cos mθ − i sin mθ = cos nθ + i sin nθ.

How do we generalize the formula to (cos θ + i sin θ )s , where s is a rational

number? Let s = p/q, where p and q are irreducible integers. It is easily seen
that the modulus of (cos θ + i sin θ )s remains unity. Let the polar representation
of (cos θ + i sin θ )s take the form cos φ + i sin φ for some φ. Now, we write

cos φ + i sin φ = (cos θ + i sin θ)s = (cos θ + i sin θ )p/q .

Taking the power of q of both sides of the above equation, we obtain

cos qφ + i sin qφ = cos pθ + i sin pθ,

which implies
pθ + 2kπ
qφ = pθ + 2kπ or φ = , k = 0, 1, . . ., q − 1.
q
It suffices to limit the set of integers to be {0, 1, · · · , q − 1} since the value
of φ corresponding to k beyond this set of integers equals one of those values
defined in the above equation plus some multiple of 2π . Therefore, integers
beyond the above set do not generate new distinct values of cos φ + i sin φ. As
a result, there are q distinct roots of (cos θ + i sin θ )p/q , namely,
   
pθ + 2kπ pθ + 2kπ
cos + i sin , k = 0, 1, . . ., q − 1.
q q

Example 1.2.1 Find the three possible values of the cube root of the complex
number
1−i
.
1+i

Solution It is more convenient to use the polar representation. The polar forms
of 1 − i and 1 + i are
 
√ −π −π √  π π
1 − i = 2 cos + i sin and 1 + i = 2 cos + i sin .
4 4 4 4
 1.2 Algebraic properties of complex numbers 9

Using eq. (1.2.5b), we obtain

√    

1−i 2 −π π −π π −π −π
= √ cos − + i sin − = cos + i sin .
1+i 2 4 4 4 4 2 2
1−i
By de Moivre’s theorem, the three possible values of the cube root of are
1+i
−π −π
cos + i sin ,
  6 6
−π 2π −π 2π π π
cos + + i sin + = cos + i sin = i,
6 3 6 3 2 2
and
   
−π 4π −π 4π 7π 7π
cos + + i sin + = cos + i sin .
6 3 6 3 6 6

Example 1.2.2 For any two complex numbers z1 and z2 , show that
|z1 − z2 |2 + |z1 + z2 |2 = 2|z1 |2 + 2|z2 |2 .
Hence, deduce that



|α + α 2 − β 2 | + |α − α 2 − β 2 | = |α + β| + |α − β|,
for any real numbers α and β.

Solution For any two complex numbers z1 and z2 ,

|z1 − z2 |2 + |z1 + z2 |2 = (z1 − z2 )(z1 − z2 ) + (z1 + z2 )(z1 + z2 )
= (z1 z1 − z2 z1 − z1 z2 + z1 z2 )
+ (z1 z1 + z2 z1 + z1 z2 + z2 z2 )
= 2(z1 z1 + z2 z2 ) = 2|z1 |2 + 2|z2 |2 .



We let z1 = α + α 2 − β 2 and z2 = α − α 2 − β 2 , where α and β are real,
and apply the above relation to obtain


1
2 
|α + α 2 − β 2 |2 + |α − α 2 − β 2 |2 = |2 α − β 2 |2 + |2α|2
2
= 2|α 2 − β 2 | + 2α 2 .
Next, we consider


2
|α + α 2 − β 2 | + |α − α 2 − β 2 | − (|α + β| + |α − β|)2



= |α + α 2 − β 2 |2 + |α − α 2 − β 2 |2 + 2β 2 − (2α 2 + 2β 2 ) − 2|α 2 − β 2 |



= |α + α 2 − β 2 |2 + |α − α 2 − β 2 |2 − 2|α 2 − β 2 | − 2α 2 = 0 ,
10 Complex Numbers

so that either



|α + α 2 − β 2 | + |α − α 2 − β 2 | = |α + β| + |α − β|

or



|α + α 2 + β 2 | + |α − α 2 − β 2 | + |α + β| + |α − β| = 0.

Since modulus quantities are always non-negative, the latter identity holds only
in the trivial case where α = β = 0. Hence, we obtain



|α + α 2 − β 2 | + |α − α 2 − β 2 | = |α + β| + |α − β|,

for any real numbers α and β.

Example 1.2.3 Find the square roots of a + ib, where a and b are real
constants.

Solution Let u + iv be a square root of a + ib, that is, (u + iv)2 = a + ib.

Equating the corresponding real and imaginary parts, we have

u2 − v 2 = a and 2uv = b .

By eliminating v from the above equations, we obtain a quadratic equation

for u2 :

4u4 − 4au2 − b2 = 0 .

The two real roots for u are found to be

 √
a + a 2 + b2
u=± .
2
From the relation v 2 = u2 − a, we obtain
√
a 2 + b2 − a
v=± .
2
Since u and v each have two possible values, apparently there are four possible
values for u + iv. However, there can be only two values of the square root of
a + ib. By observing the relation 2uv = b, one must choose u and v such that
their product has the same sign as b. This leads to
 √ √ 
a+ a +b 2 2 b a + b − a
2 2
u + iv = ±  +i ,
2 |b| 2
 1.2 Algebraic properties of complex numbers 11

provided that b = 0. The special case where b = 0 is trivial. As a numerical

example, suppose a = 3 and b = −4. We then have
  
3+5 5−3
(3 − 4i) = ±
1/2
−i = ±(2 − i).
2 2

Example 1.2.4 A complex number of the form α + iβ, where α and β are
real integers, is called a Gaussian integer. A Gaussian integer a is said to be
composite if and only if it can be factored into the form a = bc, where b and c
are both Gaussian integers (excluding ±1 and ±i); otherwise it is prime. Show
that, as Gaussian integers, 2 is composite but 3 is prime.

Solution The fact that 2 is composite is obvious since

2 = (1 + i)(1 − i) .

It is easy to check that 3 cannot be expressed as the product of a real number and
a complex number, or as the product of an imaginary number and a complex
number. Assume 3 to be composite, then

3 = (α1 + iβ1 )(α2 + iβ2 ) = (α1 α2 − β1 β2 ) + i(α1 β2 + α2 β1 ) , (i)

where α1 , α2 , β1 , β2 are non-zero integers. Without loss of generality, we

assume α1 ≥ α2 . Since the imaginary part of 3 is zero, we have

α1 β2 + α2 β1 = 0 ,

or equivalently,
α1 β1
= − = k, k ≥ 1.
α2 β2
Putting the above relations into eq. (i), we obtain
 
k α22 + β22 = 3 .

Since α2 and β2 are non-zero integers, and k ≥ 1, the only possible solutions to
the above equation are α2 = β2 = 1 and k = 3/2. This gives α1 = −β1 = 3/2,
which is a non-integer. This is a contradiction, hence 3 cannot be composite.

Example 1.2.5 A complex number lies on or inside the unit circle in the
complex plane if and only if its modulus is less than or equal to 1. A polynomial
P (z) is called a simple von Neumann polynomial if its roots all lie on or inside
the unit circle and any root on the unit circle is simple; that is, has multiplicity 1.
12 Complex Numbers

Consider a quadratic polynomial of the form

P2 (z) = z2 + a1 z + a0 , where a0 and a1 are complex numbers.
Show that
(a) when |a0 | > 1, P2 (z) cannot be a simple von Neumann polynomial;
(b) when |a0 | < 1, P2 (z) is a simple von Neumann polynomial if and only if
|a1 − a1 a0 | ≤ 1 − |a0 |2 ;
(c) when |a0 | = 1, P2 (z) is a simple von Neumann polynomial if and only if
a1 = a1 a0 and |a1 | < 2.

Solution
(a) Let r1 and r2 denote the two roots of P2 (z). Assume P2 (z) to be a simple
von Neumann polynomial. Then
|r1 r2 | = |a0 | ≤ 1.
Hence, when |a0 | > 1, P2 (z) cannot be a simple von Neumann polyno-
mial.
(b) First, note that a0 = r1 r2 and −a1 = r1 + r2 , so
a1 − a1 a0 = r1 (|r2 |2 − 1) + r2 (|r1 |2 − 1). (i)
Assume P2 (z) to be a simple von Neumann polynomial. One then
observes that
0 ≤ (1 − |r1 r2 |)(1 − |r1 |)(1 − |r2 |)
= 1 − |r1 |2 |r2 |2 − |r1 |(1 − |r2 |2 ) − |r2 |(1 − |r1 |2 )
≤ 1 − |a0 |2 − |a1 − a1 a0 |. (ii)
When |a0 | < 1, the above inequality can be rearranged as
|a1 − a1 a0 | ≤ 1 − |a0 |2 . (iii)
Conversely, suppose inequality (iii) holds and |a0 | < 1. One can estab-
lish the following inequality
(1 − |r1 |)(1 − |r2 |) ≥ 0 (iv)
by direct computation (see Problem 1.16). Since |r1 | |r2 | = |a0 | < 1,
together with the result in inequality (iv), these lead to either
|r1 | < 1 and |r2 | ≤ 1, or |r1 | ≤ 1 and |r2 | < 1.
 1.3 Geometric properties of complex numbers 13

Therefore, the two roots must be on or inside the unit circle and any root
on the unit circle must be simple. Hence, P2 (z) is a simple von Neumann
polynomial.
(c) When P2 (z) is a simple von Neumann polynomial and |a0 | = 1, the two
roots must satisfy |r1 | = |r2 | = 1 and r1 = r2 . Equation (i) now becomes

a1 = a1 a0 .

In addition, since |r1 | = |r2 | = 1 and r1 = r2 , we have

|a1 | = |r1 + r2 | < |r1 | + |r2 | = 2.

Conversely, given the condition |a0 | = 1, there are only two possible
cases of distribution of the roots. In one case, one root is inside and the
other root is outside the unit circle. In the other case, both roots are on
the unit circle.
The first case can be shown to be impossible, given that a1 − a1 a0 = 0
and |a1 | < 2. To prove the claim, we note from eq. (i) that

0 = a1 − a1 a0 = r1 (|r2 |2 − 1) + r2 (|r1 |2 − 1),

so Arg r1 = Arg r2 since |r1 |2 − 1 and |r2 |2 − 1 are real and of opposite
signs. Now, consider

|a1 | = |r1 + r2 | = |r1 | + |r2 | (since Arg r1 = Arg r2 )

1
= |r1 | + ≥2 (since |r1 | |r2 | = 1),
|r1 |

contradicting the fact that |a1 | < 2.

Consider the second case, where both roots are on the unit circle.
Given that |a1 | < 2, one can show that the two roots cannot be repeated.
If otherwise, the condition of repeated roots requires a12 = 4a0 . One then
obtains |a1 |2 = 4|a0 | = 4, giving |a1 | = 2, and this contradicts |a1 | < 2.
Combining the results, the two roots must lie on the unit circle and
cannot be repeated, so P2 (z) is a simple von Neumann polynomial.

1.3 Geometric properties of complex numbers

We have seen that a complex number z = x + iy can be associated with a point
(x, y) or a vector xi + yj in the complex plane. In this section, we explore more
properties of these geometric representations of complex numbers.
14 Complex Numbers

z1+z2
z2

z1
x

Figure 1.2. In the vectorial representation, the sum of two complex numbers can be
constructed using the parallelogram law.

The distance between the two points representing z1 and z2 in the complex
plane is given by

|z1 − z2 | = |(x1 − x2 ) + i(y1 − y2 )|

= (x1 − x2 )2 + (y1 − y2 )2 . (1.3.1)

For example, the locus of points z in the complex plane defined by the relation

|z − z0 | = r, z0 is complex and r is real, (1.3.2)

represents a circle centered at z0 and with radius r.

What is the geometric interpretation of the addition of two complex numbers
z1 and z2 in the complex plane? The sum z1 + z2 is represented by the point
(x1 + x2 , y1 + y2 ) or the vector (x1 + x2 )i + (y1 + y2 )j. If we treat z1 as the
vector x1 i + y1 j and z2 as the vector x2 i + y2 j, then z1 + z2 can be found
by adding the two vectors using the parallelogram law (see Figure 1.2). The
difference z1 − z2 can be found similarly by treating z1 − z2 as z1 + (−z2 ).
In Figure 1.2, the lengths of the sides of the two triangles are given by the
moduli of the corresponding complex numbers representing the sides. Since
the sum of any two sides of a triangle must be greater than or equal to the
third side, we deduce immediately the renowned triangle inequalities. For any
complex numbers z1 and z2 , we have

|z1 + z2 | ≤ |z1 | + |z2 |, (1.3.3a)

 
 
|z1 − z2 | ≥ |z1 | − |z2 | . (1.3.3b)

The triangle inequalities can also be proved by an algebraic argument (see

Example 1.3.1). By setting z2 to be −z2 in (1.3.3b), the above two inequalities
 1.3 Geometric properties of complex numbers 15

can be combined into the form

 
 
|z1 | − |z2 | ≤ |z1 + z2 | ≤ |z1 | + |z2 | .

Replacing z2 by z2 + z3 in inequality (1.3.3a), we obtain

|z1 + z2 + z3 | ≤ |z1 + z2 | + |z3 | ≤ |z1 | + |z2 | + |z3 | . (1.3.4)

The result can be extended to any finite number of complex numbers by induc-
tion. Accordingly, the generalized triangle inequality is given as
 n 
   n
 
 z k ≤ |zk | , n = 2, 3, . . . (1.3.5)
 
k=1 k=1

In Example 1.2.2, we establish the following relation

|z1 − z2 |2 + |z1 + z2 |2 = 2(|z1 |2 + |z2 |2 ).

What would be the geometric interpretation of the relation? Considering the
vectorial representation of complex numbers, suppose we take the complex
numbers z1 and z2 as the adjacent sides of a parallelogram. The modulus
quantities |z1 + z2 | and |z1 − z2 | can be visualized as the lengths of the two
diagonals of the parallelogram. The above relation shows that the sum of the
squares of the lengths of the diagonals of a parallelogram is equal to the sum
of the squares of the lengths of the four sides of the parallelogram.
In the complex plane, the various points P1 , P2 , . . . in a geometric figure
can be represented by the complex numbers z1 , z2 , . . ., respectively. A given
geometric property of the figure, such as ‘the triangle P1 P2 P3 is equilateral’ or
‘the four points P1 , P2 , P3 , P4 are concyclic’, corresponds to some algebraic
relation of the complex numbers. It is this type of correspondence between
geometry and algebra that explains why complex numbers can be used as an
effective tool in solving geometric problems.
For example, suppose we want to find the equation of the perpendicular
bisector of the line segment joining the two points z1 and z2 . Since any point z
on the bisector is equidistant from z1 and z2 , the equation of the bisector can
be represented by

|z − z1 | = |z − z2 | . (1.3.6)
For a given equation f (x, y) = 0 of a geometric curve, if we set x =
(z + z)/2 and y = (z − z)/2i, the equation can be expressed in terms of the
pair of conjugate complex variables z and z as
 
z+z z−z
f (x, y) = f , = F (z, z) = 0 . (1.3.7)
2 2i
16 Complex Numbers

For example, the unit circle centered at the origin as represented by the equation
x 2 + y 2 = 1 can be expressed as zz = 1.
The notion of the moduli of complex numbers is a useful tool for describing
a region in the complex plane. Some examples are:
(i) The set {z : |z − a| < r}, a ∈ C, r ∈ R, represents the set of points
inside the circle centered at a and with radius r but excluding the
boundary.
(ii) The set {z : r1 ≤ |z − a| ≤ r2 }, a ∈ C, r1 and r2 ∈ R, represents the
annular region centered at a and bounded by circles of radii r1 and r2 .
Here, the boundary circles are included.
(iii) The set of points z such that |z − α| + |z − β| ≤ 2d, α and β ∈ C and
d ∈ R, is the set of all points on or inside the ellipse with foci α and
β and with the length of the semi-major axis equal to d. What is the
length of the semi-minor axis?

1.3.1 nth roots of unity

By definition, the nth roots of unity satisfy the equation
zn = 1 .
By de Moivre’s theorem, the n distinct roots of unity are
2kπ 2kπ
z = e2kπi/n = cos + i sin , k = 0, 1, . . ., n − 1 . (1.3.8)
n n
If we write ωn = e2πi/n , then the nth roots are 1, ωn , ωn2 , . . ., ωnn−1 . Alterna-
tively, if we pick any one of the roots and call it α, then the other roots are
given by αωn , αωn2 , . . ., αωnn−1 .
In the complex plane, the nth roots of unity correspond to the n vertices
of a regular n-sided polygon inscribed inside the unit circle centered at the
origin, with one vertex at the point z = 1. The vertices are equally spaced on
the circumference of the circle. Figure 1.3 shows the regular octagon inscribed
inside the unit circle when n = 8.
The above argument can be extended easily to the nth roots of an arbi-
trary complex number. Suppose the given complex number in polar form is
represented by r(cos φ + i sin φ). Its nth roots are given by
 
φ + 2kπ φ + 2kπ
r 1/n
cos + i sin , k = 0, 1, 2, . . ., n − 1,
n n
where r 1/n is the positive nth root of the positive real number r. The roots
now lie on the circle |z| = r 1/n . These roots are equally spaced along the
 1.3 Geometric properties of complex numbers 17

Figure 1.3. The vertices of the regular octagon inscribed inside the unit circle centered
at the origin are the eighth roots of unity. One of the vertices of the octagon must be at
z = 1.

circumference but one vertex must be at r 1/n (cos( φn ) + i sin( φn )). For example,
consider the cube roots of 1 + i. In polar form, we have
√  π   π 
1 + i = 2 cos + i sin
4 4
√ √
π
so that the cube roots are 6 2(cos( 12
√ ) + i sin( 12
π
)), 6 2(cos( 3π
4
) + i sin( 3π
4
)) and
6
2(cos( 12 ) + i sin( 12 )). They√form the vertices of an equilateral triangle
17π 17π

inscribed inside the circle |z| = 6 2.

1.3.2 Symmetry with respect to a circle

Given a point α in the complex plane, we would like to construct the symmetry
point of α with respect to the circle CR : |z| = R. The symmetry point of α is
defined to be β = R 2 /α. Conversely, since we may write α = R 2 /β, we can
also consider α to be the symmetry point of β. The two points α and β are
said to be symmetric with respect to the circle CR (more details can be found
in Subsection 8.2.2). We explore some of the geometric properties associated
with a pair of symmetric points.

(i) We first assume that |α| < R so that the symmetry point β lies outside
the circle CR . By observing that
R2 1
Arg β = Arg = Arg = −Arg α = Arg α,
α α
one concludes that α and β both lie on the same ray emanating from
the origin. The symmetry point β can be constructed as follows: draw
18 Complex Numbers

Figure 1.4. Construction of a pair of symmetric points with respect to the circle CR .

the circle CR and a ray L from the origin through α. We then draw a
perpendicular to L through α which intersects the circle CR at P . The
point of intersection of the tangent line to the circle CR at P and the ray
L then gives β (see Figure 1.4).
The proof of the construction is simple. It is seen that α and β lie
on the same ray through the origin so that Arg α = Arg β. From the
construction, it is observed that |β|/R = R/|α| = sec φ, where φ is
the angle between OP and the ray L, so |α| |β| = R 2 . Combining the
results, α and β are related by β = R 2 /α.
(ii) When |α| = R, the symmetry point is just α itself.
(iii) Lastly, when |α| > R, the symmetry point β will be inside the circle
CR . To reverse the method of construction in (i), we find a tangent to
the circle which passes through α and call the point of tangency P . A
ray L is then drawn from the origin through α and a perpendicular is
dropped from P to L. The point of intersection of the perpendicular
with the ray L then gives β.

Example 1.3.1 Show that for any two complex numbers z1 and z2 ,

|z1 + z2 |2 = |z1 |2 + |z2 |2 + 2 Re(z1 z2 ) ,

then deduce the combined triangle inequalities

 
 
|z1 | − |z2 | ≤ |z1 + z2 | ≤ |z1 | + |z2 | .
 1.3 Geometric properties of complex numbers 19

Solution Consider
|z1 + z2 |2 = (z1 + z2 )(z1 + z2 )
= z1 z1 + z2 z2 + z1 z2 + z2 z1
= |z1 |2 + |z2 |2 + 2 Re(z1 z2 ).
By observing that Re(z1 z2 ) ≤ |z1 z2 |, we have
|z1 + z2 |2 ≤ |z1 |2 + |z2 |2 + 2|z1 z2 |
= |z1 |2 + |z2 |2 + 2|z1 ||z2 | = (|z1 | + |z2 |)2 .
Since moduli are non-negative, we take the square root of both sides and obtain
|z1 + z2 | ≤ |z1 | + |z2 | . (i)
To prove the other half of the triangle inequality, we write
|z1 | = |(z1 + z2 ) + (−z2 )| ≤ |z1 + z2 | + | − z2 |,
giving
|z1 | − |z2 | ≤ |z1 + z2 | . (ii)
By interchanging z1 and z2 in the above inequality, we have
|z2 | − |z1 | ≤ |z1 + z2 | . (iii)
Combining the results in eqs. (i), (ii) and (iii), we obtain
 
 
|z1 | − |z2 | ≤ |z1 + z2 | ≤ |z1 | + |z2 | .

Example 1.3.2 Find the locus of the points z satisfying

|z − α| + |z + α| = 2r ,
where α is complex and r is real. Discuss various possibilities and find the
maximum and minimum values of |z|.

Solution The two moduli |z − α| and |z + α| represent the distances from z

to α and −α, respectively. By the triangle inequality, |z − α| + |z + α| ≥ 2|α|
and so |α| ≤ r. We consider the following separate cases:
(a) |α| < r. This represents an ellipse with foci at α and −α. The length
of

the semi-major axis is r and the length of the semi-minor axis is
r 2 − |α|2 . The angle of inclination of the major axis is Arg α. When
α = 0, the two foci coalesce and the ellipse becomes a circle centered at
the origin and with radius r.
20 Complex Numbers

(b) |α| = r. The ellipse collapses into a line segment joining α and −α.

Further, from the geometric properties of an ellipse, we deduce that

(i) max{|z|} = length of the semi-major axis =

r,
(ii) min{|z|} = length of the semi-minor axis = r 2 − |α|2 .

Example 1.3.3 Find the curve or region in the complex plane represented by
each of the following equations or inequalities:
1
(a) Re= 2, (b) |z + 1||z − 1| = 1 , (c) |z| + Re z ≤ 1 ,
z  
z−i π z − 1
(d) 0 < Arg < , (e)   ≤1.
z+i 2 z + 1

Solution
(a) Suppose we write z = reiθ , and so Re 1
z
= cos θ
r
. The equation of the
curve in polar form becomes
cos θ
r= .
2

Note that r 2 = zz and r cos θ = z+z

2
, so the polar equation can be
rewritten as
    2  
z+z 1 1 1  1  1
zz = or z− z− = 
or z −  = .
4 4 4 4 4 4
The locus represents the circle centered at (1/4, 0) and with radius 1/4,
but with the origin deleted since Re 1z is not defined at z = 0.
(b) The curve is symmetrical with respect to both the x-axis and y-axis since
the equation is invariant if we change z to z and −z, respectively. It also
passes through the origin. If we write the equation as |z2 − 1| = 1 and
let z = reiθ , we have

|r 2 cos 2θ − 1 + ir 2 sin 2θ | = 1 ,

which can be simplified to the form

r 2 = 2 cos 2θ .

This is the standard form of a lemniscate. The polar representation reveals

the symmetry of the curve with √ respect to both axes. Also, the curve is
bounded within the circle |z| = 2 (see Figure 1.5).
 1.3 Geometric properties of complex numbers 21

Figure 1.5. The shape of a lemniscate represented by the polar equation r 2 = 2 cos 2θ.

(c) Suppose z = x + iy. The inequality can be written as

x 2 + y 2 ≤ 1 − x ⇔ y 2 ≤ 1 − 2x .
The above inequality represents the region on and inside the parabola
y 2 = 1 − 2x.
(d) Let z = x + iy, then
z−i (x 2 + y 2 − 1) − 2ix
= .
z+i x 2 + (y + 1)2
 
One observes that Arg z−iz+i
∈ 0, π2 if and only if z−i
z+i
has both positive
real and imaginary parts, that is, x < 0 and x 2 + y 2 > 1. Therefore,
0 < Arg z−i
z+i
< π2 represents the region exterior to the unit circle |z| = 1
and on the left half-plane.
(e) Consider
 
z − 1
 
 z + 1  ≤ 1 ⇔ |z − 1| ≤ |z + 1|
2 2

⇔ (z − 1)(z − 1) ≤ (z + 1)(z + 1)
⇔ z + z ≥ 0 ⇔ Re z ≥ 0 .
The region represented by the inequality is the right half-plane, including
the y-axis.

Example 1.3.4 Suppose the four points z1 , z2 , z3 and z4 lie on a circle. Show
that
(z1 − z3 )(z2 − z4 )
(z1 − z4 )(z2 − z3 )
is real.
22 Complex Numbers

y
z
× 1

)
z 2× θ1
θ2

)
×
z4
×
z3
x
Figure 1.6. Angles of turning of the line segments at z1 and z2 are equal, that is, θ1 = θ2 .

Solution Suppose the four points z1 , z2 , z3 and z4 are distributed on a

circle as shown in Figure 1.6. Draw the two line segments, one join-
ing z1 with z3 and the other joining z1 with z4 . The angle of turning of
these two line segments at z1 (denoted by θ1 in Figure 1.6) is given by Arg zz11 −z
−z3
4
.
Similarly, we draw another pair of line segments joining z2 with z3 and z2 with
z4 . The angle of turning of the line segments at z2 (denoted by θ2 in Figure 1.6)
is given by Arg zz22 −z
−z3
4
. Since the four points are concyclic, the two angles of
turning are equal. Therefore, we have
z 1 − z4 z2 − z4
Arg = θ1 = θ2 = Arg ,
z 1 − z3 z2 − z3
and so
(z1 − z3 )(z2 − z4 )
(z1 − z4 )(z2 − z3 )
is real.

Remarks
(i) There are 3! = 6 circular permutations of four distinct points on a circle.
The above argument can be applied analogously to the other 5 circular
permutations.
(ii) One can deduce the equation for the circle passing through three non-
collinear points z1 , z2 , z3 to be given by

(z1 − z3 )(z2 − z)
Im = 0.
(z1 − z)(z2 − z3 )
 1.4 Some topological definitions 23

1.4 Some topological definitions

In this section, some topological definitions commonly encountered in complex
variable theory are introduced. These definitions are useful in our discussion
of various topics related to analyticity, residue calculus, etc. in later chapters.
The set of points z such that |z − z0 | < , where z0 ∈ C,  ∈ R, contains
points that are inside the circle centered at z0 and with radius . We call it a
neighborhood of z0 and denote it by N (z0 ; ). A deleted neighborhood of z0 is
(z0 ; ).
the point set N(z0 ; ) \ {z0 }. We write it as N
A point z0 is called a limit point or an accumulation point of a point set
S if every neighborhood of z0 contains a point of S other than z0 . That is,
 0 ; ) ∩ S = φ for any  > 0. Since this is true for any neighborhood of
N(z
z0 , S must contain infinitely many points. For example, consider the point set
{z : |z| < 1}. The limit points are points on and inside the circle |z| = 1. Be
aware that a limit point z0 of S may or may not belong to the point set S.

Example 1.4.1 Find all the limit points of each of the following point sets
and determine whether the point set contains all of its limit points.
 
n
(a) E = z : z = (−1)n , n is an integer ;
n+1
 
1 i
(b) F = z : z = + , m and n are integers .
m n

Solution
(a) The point z = 1 is a limit point of E, the proof of which is as follows.
Given any  > 0, we want to show that there exists a point in E other
than z = 1 such that
 
 
(−1)n n − 1 <  .
 n+1 

If we choose n to be even and n + 1 > 1 , then the above inequality holds.

Therefore, every neighborhood N(1; ) contains at least one point in E
other than z = 1. This completes the proof that z = 1 is a limit point. The
proof that z = −1 is another limit point can be done similarly. The point
set E has only two limit points. It is seen that E does not contain the two
limit points.
(b) One may deduce judiciously that the set of limit points of F is the union
of the two sets:
   
1 i
z : z = , m is any integer ∪ z : z = , n is any integer .
m n
24 Complex Numbers

It can be shown that every neighborhood of any limit point chosen from
the above set contains a point in F other than the limit point itself. It is
apparent that all limit points of F are contained in F .

Interior points, boundary points, and exterior points

A point z0 is called an interior point of a point set S if there exists a neighborhood
of z0 , all points of which belong to S. If every neighborhood of z0 contains
points of S and also points not belonging to S, then z0 is called a boundary
point. The set of all boundary points of the point set S is called the boundary of
S. If a point is neither an interior nor a boundary point of S, then it is called an
exterior point of S. Indeed, if z0 is not a boundary point of S, then there exists
a neighborhood of z0 such that it is completely inside S or completely outside
S. In the former case, it is an interior point; otherwise, it is an exterior point.
For example, consider the point set S = {z : |z| ≤ 1}. Any point inside the
unit circle is an interior point of S and any point on the circumference of the unit
circle is a boundary point. The point z = 1 + i is an exterior point. Actually,
given any point that lies outside the circle |z| = 1, obviously it cannot be an
interior point. Also, one can always find a neighborhood of this outside point
that does not intersect the circle, so it cannot be a boundary point. These facts
illustrate why any point lying outside the circle is an exterior point.

Open sets and closed sets

A point set which consists only of interior points is called an open set. Another
way of looking at an open set is to observe that every point of this set has a
neighborhood contained completely in the set. Intuitively, we may think of any
two-dimensional set without boundary as an open set. A point set is said to be
closed if it contains all its boundary points. For example, the set {z : |z| < 1} is
an open set while the set {z : |z| ≤ 1} is a closed set. The closure of a point set
S is the closed set that contains all points in S together with the whole boundary
of S. According to this definition, the closed set {z : |z| ≤ 1} is the closure of
the open set {z : |z| < 1}.

Example 1.4.2 In this example, we examine whether a point set is open or

closed or neither.

(a) Consider the point set

A = {z : Re z > Im z}.

Show that it is an open set. Find the closure of A. Is its complement Ac

a closed set?
 1.4 Some topological definitions 25
x y
y

x0 y0
2

( x0 , y0 )

Figure 1.7. A neighborhood of (x0 , y0 ) can always be found that lies completely inside
set A, where A = {z : Re z > Im z}.

(b) Consider another point set

B = {z : Re z > Im z if Re z ≥ 0 and Re z ≥ Im z if Re z < 0}.

Is this set open or closed or neither?

Solution
(a) Consider an arbitrary point z0 = (x0 , y0 ) that lies in A. We would like to
show that there exists a neighborhood of z0 which lies completely inside
A (see Figure 1.7). Note that the shortest distance from the point (x0 , y0 )
to the line x = y is x0√−y
2
0
. We choose  < x0√−y
2
0
so that the neighborhood
N(z0 ; ) lies completely inside A. Since A consists of interior points only,
it is an open set.
The boundary points of A are points that lie along the line Re z = Im z,
so the closure of A is given by

A = {z : Re z ≥ Im z}.

The complement of A is

Ac = {z : Re z ≤ Im z},

which is seen to contain all its boundary points. Hence, Ac is closed.

(b) The set B is not open since a point (x0 , y0 ) that satisfies x0 = y0
and x0 < 0 is not an interior point. This is because any neighborhood
of such a point contains points inside B and points outside B. Actually,
it is a boundary point. On the other hand, B is not closed since it does
not include all its boundary points that satisfy x0 = y0 and x0 ≥ 0.
26 Complex Numbers

Can we devise some effective techniques to determine whether a given point

set is open or closed? More properties of open and closed sets are presented in
the following theorems.

Theorem 1.4.1 A set is open if and only if it contains none of its boundary
points.

Proof
“if” part
Suppose that D is an open set, and let p be a boundary point of D. Suppose
p is in D. Then by virtue of the property of an open set, there is an open
disc centered at p that lies completely inside D. This contradicts that p is a
boundary point.
“only if” part
Suppose D is a set that contains none of its boundary points. For any z0 ∈ D,
z0 cannot be a boundary point of D. Hence, there is some disc centered at z0
that is either a subset of D or a subset of the complement of D. The latter is
impossible since z0 itself is in D. Hence, each point of D is an interior point
so D is open.

Corollary A set C is closed if and only if its complement D = {z : z ∈ C} is

open.
To show the claim, by virtue of the definition of the boundary point, we
observe that the boundary of a point set coincides exactly with the boundary
of the complement of that set. Recall that a closed set contains all its boundary
points. Its complement shares the same boundary, but these boundary points
are not contained in the complement. By Theorem 1.4.1, the complement is
open.

Recall that a closed set contains all its boundary points. But what is the
relation between a closed set and the set of its limit points?

Theorem 1.4.2 A set S is closed if and only if S contains all its limit
points.

Proof
 ) as the deleted -neighborhood of z, and S c as the
We write N(z;
(z; ) ∩ S. We then
complement of S. Note that for z ∈ S, N(z; ) ∩ S = N
 1.4 Some topological definitions 27

have

S is closed ⇔ S c is open
⇔ given z ∈ S, there exists  > 0 such that N (z; ) ⊂ S c
(z; ) ∩ S = φ
⇔ given z ∈ S, there exists  > 0 such that N
⇔ no point of S c is a limit point of S.

Consider the two point sets in Example 1.4.1. By virtue of Theorem 1.4.2,
we can deduce immediately that set E is not closed while set F is closed. Is E
an open set? Check whether the point z = 23 lying in E is an interior point. If
the answer is “no”, then E cannot be open.

Compact sets
A bounded set is one that can be contained in a large enough circle centered at
the origin, that is, |z| < M for all points z in S where M is some sufficiently
large constant. An unbounded set is one that is not bounded. A set which is
both closed and bounded is called compact. For example, the set {z : Re z ≥ 1}
is closed but not bounded while the set {z : |z + 1| + |z − 1| ≤ 3} is compact
since it is both closed and bounded.

Connectedness and domain

A set S is said to be connected if any two points of S can be joined by a
continuous curve lying entirely inside S. For example, a neighborhood N (z0 ; )
is connected. An open connected set is called an open region or domain. For
example, the set {z : Re z ≥ z} is not a domain since it is not open. The set
{z : 0 < Re z < 1 or 2 < Re z < 3} is also not a domain since it is open but not
connected. To the point set S, suppose we add all of its boundary points. The
new set S is the closure of S and the closure is a closed set. The closure of an
open region is called a closed region. However, to an open region we may add
none, some, or all of its boundary points. We simply call the resulting point set
a region.
To illustrate the importance of connectedness in calculus, let us consider
an example from real calculus. It is well known that if f  (x) = 0 for all x in
an open interval (a, b), then f is constant throughout that interval. However,
suppose the domain of definition of the function is not connected [say, f (x) is
defined over (−1, 1) ∪ (2, 3)]. Then one can easily construct a function that is
not constant in the domain of definition but where f  (x) = 0 throughout the
domain of definition.
28 Complex Numbers

Jordan arc
Let x(t) and y(t) be real continuous functions of the real parameter t, α ≤ t ≤
β. The set of points z(t) = x(t) + iy(t) defines a continuous arc in the complex
plane beginning at z(α) and ending at z(β). A point z0 is called a multiple
point of the arc if the equation z0 = x(t) + iy(t) is satisfied by more than one
value of t in α ≤ t ≤ β. A continuous arc without multiple points is called a
Jordan arc. If the arc has only one double point, corresponding to the initial
and terminal values α and β of t, that is, z(α) = z(β), then it is called a simple
closed Jordan arc.

Simply connected domains

A domain S is said to be simply connected if every simple closed Jordan arc in
S can be shrunk to a point in S without passing through points not belonging
to S. That is, S is a simply connected domain if for any closed Jordan arc lying
in the domain S, the points inside the closed curve also belong to S. A domain
that is not simply connected is said to be multiply connected. According to the
definition, it is then always possible to construct some closed curve inside a
multiply connected domain in such a manner that one or more points inside the
curve do not belong to the domain. Intuitively, there are holes contained inside
some Jordan arc lying completely in the domain. We have one hole in a doubly
connected domain and two holes in a triply connected domain (see Figure 1.8).
For example, the domain {z : 1 < |z| < 2} is doubly connected.

Example 1.4.3 Let S be the set of points defined by z = p

m
+ qn i, where
p, q, m, n are positive integers, and p < m, q < n.

(a) What are the limit points of S, if any?

(b) Is S a closed set, or an open set, or neither?
(c) Is S a compact set?
(d) What is the closure of S?
(e) Is the closure of S compact?
(f) Is S a domain?

Solution
(a) Every point of the form α + βi, where α and β are real numbers inside
the interval [0, 1], is a limit point.
(b) The set S does not contain the points z = 1, z = i, and z = 1 + i,
and these points are limit points of S. Since it does not include
 1.5 Complex infinity and the Riemann sphere 29

Jordan arc lies

inside the domain

(a) doubly connected domain

y

Jordan arc lies

inside the domain

(b) triply connected domain

Figure 1.8. The holes inside the Jordan arc correspond to points that do not belong to
the domain. A doubly connected domain contains one hole while a triply connected
domain contains two holes.

all its limit points, S is not closed. None of the points in S is

an interior point since the points in S are discrete, so S is not
open.
(c) Since S is not closed, S cannot be compact.
(d) We obtain the closure of S by adding all its boundary points to S. It is
seen that the closure of S is simply the set of its limit points. √
(e) The closure of S is closed, and it is bounded by the circle |z| = 2.
Hence, the closure is compact.
(f) Since S is not open, it is not a domain.

1.5 Complex infinity and the Riemann sphere

It is convenient to augment the complex plane with the point at infinity, denoted
by ∞. The set C ∪ {∞} is called the extended complex plane. The algebra
30 Complex Numbers

involving ∞ is defined as follows:

z + ∞ = ∞ + z = ∞, for all z ∈ C,
z · ∞ = ∞ · z = ∞, for all z ∈ C/{0}.

We allow ∞ + ∞ = ∞ and ∞ · ∞ = ∞. In particular, we have −1 · ∞ = ∞.

We adopt the conventions:

z
= ∞ for all z ∈ C/{0},
0
z
= 0 for all z ∈ C.
∞
∞
= ∞ for all z ∈ C.
z

However, expressions like ∞ − ∞, 0 · ∞, 00 , ∞ 0

, and ∞
∞
are not defined.
Topologically, any set of the form {z : |z| > R}, where R ≥ 0, is called
a neighborhood of ∞. To approach the point at infinity, we let |z|
increase without bound while Arg z can assume any value. Note that
the open upper half-plane Im z > 0 does not contain the point at infinity
since Arg z is restricted to take value in the interval (0, π ). The modu-
lus value of the point of infinity is infinite while the argument value is
indeterminate.

1.5.1 The Riemann sphere and stereographic projection

In order to visualize the point at infinity, we consider the Riemann sphere that
has radius 12 and is tangent to the complex plane at the origin (see Figure 1.9).
We call the point of contact the south pole (denoted by S) and the point
diametrically opposite S the north pole (denoted by N ). Let z be an arbitrary
complex number in the complex plane, represented by the point P . We draw
the straight line P N that intersects the Riemann sphere at a unique point P  ,
distinct from N. Conversely, to each point P  on the sphere, other than the north
pole N , we draw the line P  N which cuts uniquely one point P in the complex
plane. Clearly there exists a one-to-one correspondence between points on the
Riemann sphere, except N , and all the finite points in the complex plane. We
assign the north pole N as the point at infinity. With such an assignment, we then
establish a one-to-one correspondence between all the points on the Riemann
sphere and all the points in the extended complex plane. This correspondence
is known as the stereographic projection.
 1.5 Complex infinity and the Riemann sphere 31

Figure 1.9. The Riemann sphere sitting on the complex plane.

Under the stereographic projection, meridians of the Riemann sphere are

projected onto straight lines through the origin in the complex plane and lati-
tudes are mapped onto concentric circles around the origin. The equator on the
Riemann sphere is mapped onto the unit circle in the complex plane. Therefore,
the southern hemisphere goes to the interior of the unit circle while the north-
ern hemisphere goes to the exterior of the unit circle in the extended complex
plane.
Consider an arbitrary point P (x, y), z = x + iy, in the z-plane and let the
corresponding point on the Riemann sphere be P  (ξ, η, ζ ) (see Figure 1.9).
From the geometry of the figure, the coordinates are related by
ξ η
x= and y= . (1.5.1)
1−ζ 1−ζ
Further, the equation of the Riemann sphere is given by
   2
1 2 1
ξ +η + ζ −
2 2
= or ξ 2 + η2 + ζ 2 = ζ. (1.5.2)
2 2
The above three equations are now used to solve for ξ, η and ζ in terms of x
and y. First, we substitute ξ = x(1 − ζ ) and η = y(1 − ζ ) into eq. (1.5.2) to
obtain
x2 + y2 zz
ζ = = 2 . (1.5.3a)
x2 + y2 + 1 |z| + 1
Once ζ is available, we use eq. (1.5.1) to obtain
x 1 z+z
ξ= = , (1.5.3b)
x2 + y2 + 1 2 |z|2 + 1
y 1 z−z
η= 2 = . (1.5.3c)
x +y +12 2i |z|2 + 1
32 Complex Numbers

Example 1.5.1 Derive a formula for the length of the chord joining the images
on the Riemann sphere corresponding to the two points z1 and z2 in the complex
plane. Examine the case when one of the points is the point at infinity.

Solution Let the corresponding image points of z1 and z2 on the Riemann

sphere be P1 (ξ1 , η1 , ζ1 ) and P2 (ξ2 , η2 , ζ2 ), respectively. Using eqs. (1.5.2) and
(1.5.3a,b,c), we have

d(z1 , z2 )2 = (ξ1 − ξ2 )2 + (η1 − η2 )2 + (ζ1 − ζ2 )2

= (ξ12 + η12 + ζ12 ) + (ξ22 + η22 + ζ22 ) − 2(ξ1 ξ2 + η1 η2 + ζ1 ζ2 )
= (ζ1 + ζ2 ) − 2(ξ1 ξ2 + η1 η2 + ζ1 ζ2 )
|z1 |2 |z2 |2 x1 x2 + y1 y2 + |z1 |2 |z2 |2
= + − 2
1 + |z1 |2 1 + |z2 |2 (1 + |z1 |2 )(1 + |z2 |2 )
|z1 − z2 | 2
= .
(1 + |z1 |2 )(1 + |z2 |2 )
Taking the positive square root of the expressions on both sides, we have
|z1 − z2 |
d(z1 , z2 ) =

.
1 + |z1 |2 1 + |z2 |2
Suppose one of the points, say z2 , goes to infinity. We obtain
1
d(z1 , ∞) =
.
1 + |z1 |2

Example 1.5.2 Show that any small circle whose circumference contains the
north pole N on the Riemann sphere corresponds to a straight line not passing
through the origin in the complex plane.

Solution A small circle whose circumference contains the north pole N on

the Riemann sphere is determined by the intersection of the Riemann sphere
ξ 2 + η2 + ζ 2 − ζ = 0 and a plane through N . In terms of ξ, η and ζ , a plane
passing through the north pole N (0, 0, 1) has the general form

Aξ + Bη + C(ζ − 1) = 0, C = 0.

To find the image curve of this circle in the complex plane, we use the following
transformations [see eqs. (1.5.3a,b,c)]:

x y |z|2
ξ= , η= and ζ = .
1 + |z|2 1 + |z|2 1 + |z|2
 1.6 Applications to electrical circuits 33

Substituting the above relations into the equation of the plane through N , the
equation of the image curve is found to be

Ax + By − C = 0, C = 0.

Thus, the image curve in the complex plane is shown to be a straight line not
passing through the origin. The converse statement can be proved similarly by
reversing the above argument.

1.6 Applications to electrical circuits

In this section, we discuss the application of complex numbers to alternating
current circuit analysis. The effectiveness of the formulation using complex
numbers lies in the close link between the algebraic and geometric properties
of complex numbers.
An alternating current I with magnitude Iˆ and angular frequency ω is rep-
resented by

I = Iˆ cos ωt = Re(Iˆeiωt ). (1.6.1)

As an alternative representation, the current may be conveniently represented

by the phasor

I = Iˆeiωt (1.6.2)

in the complex plane. The phasor is a complex number which revolves around
the circle of radius Iˆ with angular frequency ω. Its projection onto the real axis
gives the magnitude of the current (see Figure 1.10).
Suppose we connect in series two alternating currents of the same angular
frequency ω but with different phases as represented by

I1 = Iˆ1 cos(ωt + φ1 ) and I2 = Iˆ2 cos(ωt + φ2 ).

Their resultant is then given by

I = Iˆ1 cos(ωt + φ1 ) + Iˆ2 cos(ωt + φ2 ).

Suppose we set I formally as Iˆ cos(ωt + φ). What would be the relation

between Iˆ, φ, and Iˆ1 , Iˆ2 , φ1 , φ2 ?
Using the phasor notation, the above equation can be rewritten as

I = Re(Iˆei(ωt+φ) ) = Re(Iˆ1 ei(ωt+φ1 ) + Iˆ2 ei(ωt+φ2 ) ).

If we treat the phasors Iˆ1 ei(ωt+φ1 ) and Iˆ2 ei(ωt+φ2 ) as vectors, then Iˆei(ωt+φ) is
the vector sum of the two phasors. The magnitude and phase of the resultant
34 Complex Numbers

Figure 1.10. A phasor Iˆeiωt revolves around the circle of radius Iˆ with angular
frequency ω.

phasor are given by

Iˆ2 = (Iˆ1 cos φ1 + Iˆ2 cos φ2 )2 + (Iˆ1 sin φ1 + Iˆ2 sin φ2 )2
2 2
= Iˆ1 + Iˆ2 + 2Iˆ1 Iˆ2 cos(φ1 − φ2 )
and
Iˆ1 sin φ1 + Iˆ2 sin φ2
tan φ = .
Iˆ1 cos φ1 + Iˆ2 cos φ2
The resultant phasor can be found graphically by the parallelogram law of
vector addition. Since all the three phasors revolve with the same angular
frequency ω, the parallelogram remains in the same shape at all times.
The general formula for finding the magnitude and phase of the resultant
phasor of the sum of three or more current phasors with the same angular
frequency can be derived analogously.
Next, we consider a closed alternating current circuit with an applied voltage
V and the three basic circuit elements: resistor, inductor, and capacitor. This is
usually called an R-L-C circuit. The voltage V (excitation) and the current I
(response) are related by the formula
V = I Z, (1.6.3)
where Z is called the impedance of the circuit. If V and I are represented by
phasors, then Z becomes a complex quantity.
Suppose the alternating current I = Iˆ cos ωt is flowing through a pure resis-
tor R. The voltage drop Vr across the resistor is given by
Vr = R Iˆ cos ωt.
 1.6 Applications to electrical circuits 35

In this case, the voltage Vr is in phase with the current. When the current is
flowing through a pure inductor L, the voltage drop Vi across the inductor is
given by
dI  π
Vi = L = −ωLIˆ sin ωt = ωLIˆ cos ωt + .
dt 2
That is, the voltage Vi leads the current by a phase angle of π2 . Suppose the
same current is flowing through a pure capacitor C. The corresponding voltage
drop Vc across the capacitor is given by
 
1 Iˆ Iˆ π
Vc = I dt = sin ωt = cos ωt − .
C ωC ωC 2
In this case, the voltage Vc lags the current by a phase angle of π2 .
Suppose we connect the three elements in series and assume that steady state
responses have been attained. The total voltage drop V is the sum of drops
across the three circuit elements. We then have


π 1  π ˆ
V = R cos ωt + ωL cos ωt + + cos ωt − I,
2 ωC 2
and in phasor notation, it takes the form
 

1
V = V̂ e i(ωt+φ)
= R + i ωL − Iˆ eiωt .
ωC
According to eq. (1.6.3), the impedance of the R-L-C circuit is found to be
 
1
Z = R + i ωL − .
ωC

Example 1.6.1 In elementary circuit theory, the time-dependent response in

an R-L-C series circuit is modeled by the second-order ordinary differential
equation
d 2I dI 1
L +R + I = 0,
dt 2 dt C
where I is the instantaneous alternating current flowing through the circuit.
Suppose the solution is assumed to be of the form

I = Iˆeαt .

Find an expression for α in terms of R, L, and C. Show that, as a complex

quantity, α is never in the right half complex plane if R, L and C are non-
negative.
36 Complex Numbers

Solution Suppose the governing equation admits a solution of the form

I = Iˆeαt .
Its first and second derivatives are
dI d 2I
= α Iˆeαt and = α 2 Iˆeαt .
dt dt 2
Substituting into the differential equation, we obtain the following algebraic
equation for α:
1
Lα 2 + Rα + = 0.
C
Solving the quadratic equation for α, we have

−R ± R 2 − 4L C
α= .
2L
Consider the following cases:

(i) When R = 0, α = ± LC 1
i. The two roots are on the imaginary axis.
√ 4L 2
−R
(ii) When 0 < 2L < LC , α = − 2L ± C2L i. The two roots are in the
R √ 1 R

left half complex plane.

(iii) When 2L R
= √LC1
, α = − 2L R
(a double root). The root lies on the negative
real axis.
R 1
(iv) When 2L > √LC , the two distinct roots both lie on the negative real
axis.
Physically, the magnitude of the alternating current decays to zero as time
tends to infinity whenever α has a negative real part (corresponding to R > 0).

1.7 Problems
1.1. Express the following complex numbers in the form x + iy:
(a) (1 + 2i)3 ; (b) i 17 ;
(c) (1 + i)n + (1 − i)n , n is any positive integer;
5 i 1+i
(d) ; (e) + .
−3 + 4i 1+i i
1.2. Find the modulus of each of the following complex numbers:
(3 + i) (2 − i)
(a) −i (2 + i) (1 + 2i) (1 + i); (b) ;
(3 − i) (2 + i)
(3 + 4i) (1 + i)6
(c) .
(i)5 (2 + 4i)2
 1.7 Problems 37

1.3. Let z1 and z2 be two complex numbers. Supposing z1 + z2 and z1 z2 are

both real, show that either both z1 and z2 are real, or that one is the
complex conjugate of the other.
1.4. Prove the commutative, associative and distributive rules for the addition
and multiplication of complex numbers.
1.5. Show that if the product of two complex numbers is zero, then at least
one of them is zero.
1.6. Show that

 
θ nθ nθ
(1 + cos θ + i sin θ) = 2 cos
n n n
cos + i sin .
2 2 2

1.7. Use de Moivre’s theorem to show that

sin(2m + 1)θ = sin2m+1 θ Pm (cot2 θ ), for 0 < θ < π

2
,

where

m
n!
Pm (x) = 2m+1 m−k
(−1)k C2k+1 x , Crn = .
k=0
(n − r)!r!

Hence, deduce that


m
kπ m(2m − 1)
cot2 = .
k=1
2m + 1 3

1.8. A complex number, represented by x + iy, may also be visualized as a

2 × 2 matrix
 
x y
.
−y x

Verify that addition and multiplication of complex numbers defined via

matrix operations are consistent with the usual addition and multiplica-
tion rules. What is the matrix representation corresponding to (x + iy)−1 ?
1.9. For any pair of complex numbers z1 and z2 , show that

(a) Re(z1 z2 ) ≤ |z1 | |z2 |;

(b) |1 − z1 z2 |2 − |z1 − z2 |2 = (1 − |z1 |2 ) (1
 − |z2 | );
2

1  z1 z 
(c) |z1 + z2 | ≥ (|z1 | + |z2 |) 
2 
+ .
2 |z1 | |z2 | 
38 Complex Numbers

1.10. Let a, b, u, v be complex numbers; prove that

(a) |au + bv|2 + |bu − av|2 = (|a|2 + |b|2 ) (|u|2 + |v|2 );
(b) |au + bv|2 − |bu + av|2 = (|a|2 − |b|2 ) (|u|2 − |v|2 ).
1.11. Let z1 and z2 be complex numbers, and α1 and α2 be real numbers
satisfying α12 + α22 = 0. Prove the following inequalities:
|α1 z1 + α2 z2 |2
|z1 |2 + |z2 |2 − |z12 + z22 | ≤ 2 ≤ |z1 |2 + |z2 |2 + |z12 + z22 |.
α12 + α22
1.12. Consider two complex numbers z1 and z2 such that |z1 | = 1 and |z2 | = 1.
Show that
 
 z1 − z2 
 
 1 − z z  = 1.
2 1

1.13. If we write z = re and w = Re , where 0 ≤ r < R, show that

iθ iφ

 
w+z R2 − r 2
Re = 2 .
w−z R − 2Rr cos(θ − φ) + r 2
This is called the Poisson kernel (see Subsection 7.1.1).
1.14. Show that if r1 eiθ1 + r2 eiθ2 = reiθ , then
r 2 = r12 + 2r1 r2 cos(θ1 − θ2 ) + r22
 
−1 r1 sin θ1 + r2 sin θ2
θ = tan .
r1 cos θ1 + r2 cos θ2
Generalize the result to the sum of n complex numbers.
1.15. One may find the square roots of a complex number using polar repre-
sentation. First, we write formally
z = r(cos θ + i sin θ ).
Recalling the identities
θ 1 − cos θ θ 1 + cos θ
sin2 = and cos2 = ,
2 2 2 2
show that
  
√ 1 + cos θ 1 − cos θ
z 1/2
=± r +i , 0 ≤ θ ≤ π.
2 2
Explain why the above formula becomes invalid for −π < θ < 0. Find
the corresponding formula under this case. Use the derived formula to
compute (3 − 4i)1/2 . Compare the results with those obtained using the
formula derived in Example 1.2.3.
 1.7 Problems 39

1.16. This problem is related to the proof of inequality (iv) in Example 1.2.5.
Using the same set of notation as in Example 1.2.5, suppose we write

α1 = |r2 |2 − 1 and α2 = |r1 |2 − 1

so that

a1 − a 1 a0 = α1 r1 + α2 r2 .

Show that

|a1 − a 1 a0 |2 = α12 (1 + α2 ) + α22 (1 + α1 ) + α1 α2 (r1 r 2 + r2 r 1 )

(1 − |a0 |2 )2 = (α1 + α2 + α1 α2 )2

so that

(1 − |a0 |2 )2 − |a1 − a 1 a0 |2
= (1 − |r1 |)(1 − |r2 |)(1 + |r1 |)(1 + |r2 |)(1 + r1 r 2 )(1 + r2 r 1 ).

Given that |a1 − a 1 a0 | ≤ 1 − |a0 |2 , use the above relation to deduce the
following inequality

(1 − |r1 |)(1 − |r2 |) ≥ 0.

1.17. Let z be a root of the following equation:

zn + (z + 1)n = 0,

where n is any positive integer. Show that

1
Re z = − .
2
1.18. The nth order Chebyshev polynomial is defined by

Tn (x) = cos(n cos−1 x), n is a positive integer; − 1 ≤ x ≤ 1.

Using de Moivre’s theorem, show that Tn (x) has the formal polynomial
representation
1  n  n 
Tn (x) = x + (x 2 − 1)1/2 + x − (x 2 − 1)1/2 .
2

Hint: One needs not be concerned with the occurrence of fractional

powers of x 2 − 1. Why?
40 Complex Numbers

1.19. Show that the distance of the point c from the line

az + az = b, b is real,

|ac + ac − 2b|
is given by .
2|a|

Hint: Use the mapping w = az so that the line

az + az = b

in the z-plane is mapped onto the line

b
Re w =
2

in the w-plane. The point c is mapped to ac. Find the distance

between the image point ac and the mapped line in the w-plane.

1.20. The triangle inequality can be generalized to n complex numbers. Given

n complex numbers z1 , z2 , . . ., zn , show that

|z1 + z2 + · · · + zn | ≤ |z1 | + |z2 | + · · · + |zn |.

When does equality hold?

1.21. Let z1 , z2 , z3 , and z4 be four arbitrary points in the complex plane. Show
that

|z1 − z3 | |z2 − z4 | ≤ |z1 − z2 | |z3 − z4 | + |z1 − z4 | |z2 − z3 |.

1.22. Find the necessary and sufficient condition for any three points z1 , z2 and
z3 to be collinear in the complex plane.
1.23. Suppose z1 + z2 + z3 = 0 and |z1 | = |z2 | = |z3 | = 1. Show that z2 =
ωz1 and z3 = ω2 z1 , where ω is a root of the quadratic equation z2 + z +
1 = 0. Hence, show that z1 , z2 and z3 are the vertices of an equilateral
triangle inscribed inside the unit circle |z| = 1.
1.24. For z = 0 and −π < Arg z ≤ π , show that
 
 
|z − 1| ≤  |z| − 1 + |z||Arg z|.
 1.7 Problems 41

Hint: For |z| > 1, consider the following figure:

y
z

z 1
z |z|

|z| 1 x
1 |z|

The circular arc is centered at z = 0 and passes through the

two points z and |z|. Perform similar calculations for |z| < 1 and
|z| = 1.

1.25. When |α| < 1 and |β| < 1, show that

 
 α−β 
 
 1 − αβ  < 1.

1.26. Let z1/2 denote the square root of z that has a positive imaginary part,
where z is a non-real complex number. For any two non-real complex
numbers z1 and z2 with the same modulus value, find a real number p such
that
1/2 1/2
z1 − z2 = ipz1 z2 .

1.27. Show that the necessary and sufficient condition for the existence of z
satisfying the following equation

|z − α| + |z + α| = 2|β|

is given by |α| ≤ |β|. Find the maximum and minimum values of |z|.
1.28. Show that the equation

|z| = ε(Re z − k), ε and k are real,

represents a conic with its focus at the origin, eccentricity ε, and directrix
along the line Re z = k.
42 Complex Numbers

1.29. Determine the maximum and minimum distance from the origin to the
curve that is defined by
 
 
z + b  = a,
 z
where a and b are real.
1.30. Find the region in the complex plane that is represented by each of the
following inequalities:
 
π 1
(a) Re(z ) ≤ 1; (b) |Arg z| < ; (c)   < 3;
2
3 z
z−1 π
(d) |z + 1| − |z − 1| < 1; (e) 0 < Arg < .
z+1 4
1.31. Determine the family of curves represented by each of the following
equations:
 
 z − z1 

(a)   = λ; (b) Arg z − z1 = α, −π < α ≤ π .
z − z2  z − z2
1.32. Let z1 , z2 , z3 and w1 , w2 , w3 be the vertices of two triangles in the com-
plex plane. Show that the necessary and sufficient condition for the two
triangles to be similar is given by
 
 1 1 1 

 z1 z2 z3  = 0.
 
w w w 
1 2 3

z 2 − z1 w 2 − w1
Hint: The given condition is equivalent to = .
z 3 − z1 w 3 − w1

1.33. If z1 , z2 and z3 represent the vertices of an equilateral triangle, show that

z12 + z22 + z32 = z1 z2 + z2 z3 + z3 z1 .
1.34. Show that the area of the triangle whose vertices are z1 , z2 , z3 is given
by the absolute value of
 
1 z 3 − z1
|z3 − z2 | Im
2
.
2 z 3 − z2
1.35. Consider any triangle ABC whose sides are of length α, β, γ and for
which the distances from the centroid to the vertices are λ, µ, ν. Show
that
α2 + β 2 + γ 2
= 3.
λ2 + µ2 + ν 2
 1.7 Problems 43

Hint: Let the centroid of the triangle be at the origin and let z1 , z2 , and
z3 be the vertices; then z1 + z2 + z3 = 0.

1.36. Let ω (ω = 1) be any one of the nth roots of unity. Show that
1 + ω + ω2 + · · · + ωn−1 = 0.
Hence, deduce the value of
1 + ωk + ω2k + · · · + ω(n−1)k , k is any non-negative integer.
Also, find the value of
1 + 2ω + 3ω2 + 4ω3 + · · · + nωn−1 .
iπ
1.37. Let ω = e n and z be any complex number. Show that
(z − ω) (z − ω2 ) · · · (z − ωn−1 ) = 1 + z + z2 + · · · + zn−1 .
1.38. Let ω be a cube root of unity and define
f (a, b, c) = a 2 + b2 + c2 − bc − ca − ab.
(a) Show that
f (a, b, c) = (a + ωb + ω2 c) (a + ω2 b + ωc).
(b) Prove that if n is not a multiple of 3, then
(b − c)n + (c − a)n + (a − b)n
contains f (a, b, c) as a factor.
(c) Suppose α, β and γ are complex numbers representing the vertices
of an equilateral triangle in the complex plane. Explain why
γ − α = ω(β − γ ) or γ − α = ω2 (β − γ ).
Hence, show that α, β and γ satisfy
f (α, β, γ ) = 0.
1.39. Suppose a, b, c and d are real numbers. Find the conditions under which
the quadratic equation
x 2 + (a + ib)x + (c + id) = 0
has at least one real root.
1.40. Assume |αk | < 1 and λk ≥ 0, k = 1, 2, . . ., n, and
λ1 + λ2 + · · · + λn = 1.
44 Complex Numbers

Here, α1 , α2 , . . ., αn are complex numbers. Show that

|λ1 α1 + λ2 α2 + · · · + λn αn | < 1.

1.41. Let w1 and w2 be two complex numbers. Denote w = w2 − w1 . Sup-

pose w2 = rw1 , where r is real and non-zero. Show that

1 + cos(Arg w2 − Arg w) 1
when 0 < r < 1
= r
2
.
1 + cos(Arg w1 − Arg w) 1 when r > 1

When r < 0, show that it becomes undefined.

1.42. For each sequence {zn }, find the corresponding limit points (if they exist):
1 nπi
(a) zn = (−1)n ; (b) zn = (−1)n ; (c) zn = e 4 ;
n
n nπi
(d) zn = e 4 .
n+1
1.43. Show that the whole complex plane and the empty set are both open. Are
they both closed?
1.44. For each of the following point sets, find the interior points, exterior
points, boundary points and limit points:

(a) 0 ≤ Re(iz) ≤ 3; (b) 0 ≤ Arg z < π

4
and |z| > 2.

1.45. Classify the following sets according to the properties: open, closed,
bounded, unbounded, compact.

(a) S1 = {z : a ≤ Re z ≤ b}; (b) S2 = {z : 0 < Arg z < π

2
};
2kπi
(c) S3 = {z : z = e 5 , k = 0, 1, 2, 3, 4}.

1.46. Consider the set of points (x, y) which are solutions to

π
y=0 and sin =0
x
lying inside the punctured disk: 0 < |z| < 1, z = x + iy. Does there exist
any limit point of the point set? If so, find the limit point.
1.47. Find the image point of each of the following complex numbers on the
Riemann sphere:

(a) 1; (b) i; (c) −1; (d) −i.

1.48. What is the relation between z1 and z2 if their images on the Riemann
sphere are diametrically opposite each other?
1.49. Prove that the angle between two curves in the complex plane is equal to
the angle between their image curves on the Riemann sphere.
 1.7 Problems 45

1.50. Show that any circle or straight line in the complex plane corresponds to
a circle on the Riemann sphere.
1.51. To find a particular solution xp (t) of the constant coefficient differential
equation
d 2x dx
2
+c + kx = F cos ωt,
dt dt
one may consider the associated problem
d 2v dv
+c + kv = F eiωt ,
dt 2 dt
and find its particular solution vp (t). Show that
xp (t) = Re vp (t),
and use the result to find xp (t).
1.52. Show that the modulus and argument of the impedance of the R-L-C
circuit are given by

 
1 2 ωL − ωC
1
|Z| = R + ωL −
2 and Arg Z = tan−1 ,
ωC R
respectively. Find the condition under which the voltage is in phase with
the current.
 2
Analytic Functions

In this chapter, we introduce functions of a complex variable and examine how

some of their mathematical properties may differ from those of real-valued
functions. The mapping properties associated with complex functions are illus-
trated. The theory of differentiation of complex functions and the concept of
differentiability are developed. The highlights of the chapter are the Cauchy–
Riemann relations and the definition of an analytic function. Analyticity plays
a central role in complex variable theory. The relations between harmonic func-
tions and analytic functions are established. We show how to solve the Poisson
equation effectively using the formulation of complex conjugate variables. In
addition, the application of complex differentiation in dynamics problems and
the use of complex functions in describing fluid flows and steady state heat
distribution are illustrated.

2.1 Functions of a complex variable

Let S be a set of complex numbers in the complex plane. For every point
z = x + iy ∈ S, we specify the rule for assigning a corresponding complex
number w = u + iv. This defines a function of the complex variable z, and the
function is denoted by

w = f (z). (2.1.1)

The set S is called the domain of definition of the function f and the collection
of all values of w is called the range of f . Below are some examples of complex
functions:

(1) f1 (z) = z2 ; (2) f2 (z) = Im z; (3) f3 (z) = Arg z;

z+3
(4) f4 (z) = 2 .
z +1

46
 2.1 Functions of a complex variable 47

Note that f3 (z) = Arg z is defined everywhere except at z = 0, and this function
can assume all possible real values in the interval (−π, π ]. The domain of
definition of f4 (z) = zz+3
2 +1 is seen to be C\{i, −i}. What is the range of this

function?
A complex function of the complex variable z may be visualized as a pair of
real functions of the two real variables x and y, where z = x + iy. Let u(x, y)
and v(x, y) be the real and imaginary parts of f (z), respectively. We may write

f (z) = u(x, y) + iv(x, y), z = x + iy. (2.1.2)

For example, consider the function

f1 (z) = z2 = (x + iy)2 = x 2 − y 2 + 2ixy;

its real and imaginary parts are the real functions

u(x, y) = x 2 − y 2 and v(x, y) = 2xy,

respectively.
In a single-valued complex function, only one value of w is assigned to
1
each value of z. However, functions like f (z) = z 2 and f (z) = arg z are multi-
valued. In complex variable theory, we may treat a multi-valued function as a
collection of single-valued functions. Each member is called a branch of the
1
function. In the above examples, f (z) = z 2 has two branches and f (z) = arg z
has infinitely many branches. We usually choose one of the branches as the
principal branch of the multi-valued function. For example, Arg z is chosen
as the principal branch of f (z) = arg z. A more detailed discussion on the
characterization of the branches of multi-valued complex functions will be
presented in Section 3.6.

Example 2.1.1 For each of the following functions, determine whether it is

a many-to-one, one-to-one or one-to-many function:
1 iz + 4 1
(a) w(z) = z + ; (b) w(z) = ; (c) w(z) = z 2 .
z 2z + 3i

Solution
(a) Consider w = z + 1z . It is obvious that
 
1 1 1
w = + 1 = w(z);
z z z
48 Analytic Functions

that is, both z and 1z are mapped to the same point. In fact, the function
may be expressed as a quadratic polynomial in z, where

z2 − zw + 1 = 0.

For a given value of w, there are in general two values of z which satisfy
the above relation. Therefore, the function w(z) = z + 1z is many-to-one.
(b) For any two complex numbers z1 and z2 , we have

w(z1 ) = w(z2 )
iz1 + 4 iz2 + 4
⇐⇒ =
2z1 + 3i 2z2 + 3i
⇐⇒ 2iz1 z2 + 8z2 − 3z1 + 12i = 2iz1 z2 + 8z1 − 3z2 + 12i
⇐⇒ z1 = z2 .

Therefore, the function w(z) = 2z+3i

iz+4
is one-to-one.
(c) Consider the following relation:

z = reiθ = rei(2π+θ) .

We have two possible values, namely,

1 θ 1 2π+θ 1 θ
w(z) = r 2 ei 2 or w(z) = r 2 ei 2 = −r 2 ei 2 .

Therefore, the function w(z) = z1/2 is one-to-many.

2.1.1 Velocity of fluid flow emanating from a source

The following example illustrates how a complex function arises from the
description of a physical phenomenon. Consider the steady state flow of fluid
in the z-plane emanating from a fluid source placed at the origin. We would
like to find a complex function v(z) which gives the velocity of the flow at any
point z. Physically, the direction of the flow is radially outward from the fluid
source and the speed is inversely proportional to the distance from the source
(see Figure 2.1).
These physical properties can be described mathematically as
k
Arg v = Arg z and |v| = , (2.1.3)
|z|
where k is some positive real constant. If we write z = reiθ , the velocity
function of the fluid source is given by
k iθ k
v(z) = |v|eiArg v = e = . (2.1.4)
|z| z
 2.1 Functions of a complex variable 49

Figure 2.1. Fluid emanates from the source at the origin. The flow direction is radially
outward from the source and the speed is inversely proportional to the distance from the
source.

The constant k may be called the strength of the source. Physically, it

is related to the amount of fluid flowing out from the source per unit
time.

Example 2.1.2 Suppose we have a fluid source (i.e. fluid flowing out) at
z = α and a fluid sink (i.e. fluid flowing in) at z = β in the complex plane,
both of the same strength k. Find the resulting velocity at an arbitrary point z.
What happens if the source and the sink are approaching each other such that
(α − β) → 0 and k → ∞ but µ = k(α − β) is kept finite?

Solution From eq. (2.1.4), the velocities at z due to the source and the sink are
given by
k k
and − ,
z−α z−β
respectively. Assuming that the superposition principle of velocities is applica-
ble, the combined velocity at z is given by the sum of the two velocity functions,
so
 
1 1 k(α − β)
v(z) = k − = .
z−α z−β (z − α) (z − β)
Consider the limits (α − β) → 0 and k → ∞, while µ = k(α − β) is kept
finite; such a configuration is called a doublet. The velocity of the flow fluid at
z due to the doublet is found to be
µ
v(z) = .
(z − α)2
50 Analytic Functions

2.1.2 Mapping properties of complex functions

A complex function w = f (z) may be considered as the assignment of a point
(x, y) in the x-y plane to another point (u, v) in the u-v plane, where z = x + iy
and w = u + iv. It is not possible to superimpose these two planes to visualize
the graph of f (z). To examine how f (z) works, we put the two planes side by
side, select some convenient curves or sets in the x-y plane and then plot their
corresponding image curves or image sets in the u-v plane. In this manner, we
may consider w = f (z) as a mapping from the x-y plane onto the u-v plane. The
realization of a complex function as a mapping is an extremely useful concept
in complex variable theory. A full discussion on mappings represented by
complex functions will be presented in Chapter 8. The two examples presented
below illustrate some interesting mapping properties of complex functions.

Example 2.1.3 The complex numbers z = x + iy and w = u + iv are rep-

resented by points P and Q in the z-plane and w-plane, respectively. Suppose
P traverses along the vertical line x = − 12 , and z and w are related by
w
z= .
1−w
Find the trajectory traced out by Q.

Solution It is convenient to express the mapping relation as

z
w= .
1+z
Since P moves along x = − 12 , we may represent P by z = − 12 + iy, where
y can take any real value. The modulus of the image point Q in the w-plane
satisfies

 1    2
 − 2 + iy   − 12 + y 2
 
 1 + iy  =  1 2 = 1,
2 2
+ y 2

indicating that the locus of Q is the unit circle centered at the origin in the
w-plane (see Figure 2.2). Furthermore, we observe that
   
− 12 + iy y 2 − 14 − 12 + iy y
Re 1 = 2 1 and Im 1 = 2 1.
2
+ iy y + 4 2
+ iy y +4

Therefore, z = − 12 (corresponding to y = 0) is mapped onto w = −1, and

z = ∞ (corresponding to y → ±∞) is mapped onto w = 1. Since Im w and
y have the same sign, the portion of the line segment y < 0 is mapped onto the
lower unit circle and the portion y > 0 is mapped onto the upper unit circle.
 2.1 Functions of a complex variable 51

y v

|w| = 1

<

>
x u
1  1 w( ∞ ) = 1
z = − + iy w  −  = −1
2  2
>>

>>
Figure 2.2. The mapping behavior of z = w/(1 − w).

Correspondingly, the region on the right hand side of the vertical line x = − 12
is mapped onto the interior of the unit circle |w| = 1. The relation between the
two loci traced by P and Q is shown in Figure 2.2.

Example 2.1.4 Consider the function f (z) = z2 and write it as

f (z) = u(x, y) + iv(x, y), where z = x + iy.

Find the curves in the x-y plane such that u(x, y) = α and v(x, y) = β. Also,
find the curves in the u-v plane whose preimages in the x-y plane are x = a
and y = b. What is the image curve in the u-v plane of the closed curve
r = 2(1 + cos θ) in the x-y plane, where z = reiθ ?

Solution For z = x + iy,

f (z) = (x + iy)2 = x 2 − y 2 + 2ixy,

so that

u(x, y) = x 2 − y 2 and v(x, y) = 2xy.

For all points on the hyperbola x 2 − y 2 = α in the x-y plane, the corresponding
image points in the w-plane are on the coordinate curve u = α. Similarly, the
points on the hyperbola 2xy = β are mapped onto the coordinate curve v = β.
To find the curves in the u-v plane whose preimage curves are the coordinate
curves x = a and y = b, we use the result obtained for the square roots of
52 Analytic Functions

a complex number in Example 1.2.3. The image curve in the w-plane corre-
sponding to x = a is given by
√
u + u2 + v 2
= a 2 ⇐⇒ 4a 2 (a 2 − u) = v 2 .
2
Similarly, the image curve in the w-plane corresponding to y = b is given by
√
u2 + v 2 − u
= b2 ⇐⇒ 4b2 (b2 + u) = v 2 .
2
Both image curves are parabolas in the w-plane.

Remark Mapping of regions As deduced from the above relations, the com-
plex function w = z2 is seen to have the following mapping properties:

(i) the line x = 0, y > 0 is mapped onto the line u < 0, v = 0;

(ii) the line 0 ≤ x ≤ a, y = 0 is mapped onto the line 0 ≤ u ≤ a 2 , v = 0;
and
(iii) the line x = a, y > 0 is mapped onto the upper portion of the parabola
4a 2 (a 2 − u) = v 2 , v > 0.
Hence, the semi-infinite strip
{(x, y) : 0 ≤ x ≤ a, y ≥ 0}
in the z-plane is mapped onto the semi-infinite parabolic wedge
{(u, v) : 4a 2 (a 2 − u) ≥ v 2 , v ≥ 0}
in the w-plane (see Figure 2.3).
The preimage curve r = 2(1 + cos θ ) is expressed in the polar form, so it is
convenient to seek the polar representation of the image curve. Let Reiφ be the
polar representation of w. In polar form, the complex function w = z2 becomes
Reiφ = r 2 e2iθ . Consider a point on the curve
θ
r = 2(1 + cos θ) = 4 cos2 ;
2
we see that
θ
r 2 = 16 cos4 .
2
The corresponding image point in the w-plane can be represented as
16 cos4 θ2 e2iθ . By comparing like terms in Reiφ and 16 cos4 θ2 e2iθ , we deduce
that
θ φ
φ = 2θ and R = 16 cos4 = 16 cos4 .
2 4
 2.1 Functions of a complex variable 53

Figure 2.3. The complex function w = z2 maps a semi-infinite strip in the z-plane onto
a semi-infinite parabolic wedge in the w-plane.

Therefore, the polar form of the image curve in the w-plane is found to be
φ
R = 16 cos4 .
4

2.1.3 Definitions of the exponential and trigonometric functions

We illustrate how the complex exponential and trigonometric functions can be
defined as natural extensions of their real counterparts. These functions are
introduced earlier so that they can be used to serve as examples to illustrate the
concepts of limit, continuity and differentiability of complex functions in later
sections. In the next chapter, we show how to define the complex exponential
and trigonometric functions using the first principles.
Let z = x + iy. The complex exponential function is defined by
ez = ex (cos y + i sin y), for all z in C. (2.1.5)
How would we justify the above definition? First, we require that the complex
exponential function reduces to its real counterpart when z is real. In fact, if
y = 0, ez reduces to ex , which is the real exponential function. By setting
x = 0, eq. (2.1.5) becomes
eiy = cos y + i sin y, (2.1.6)
which is consistent with the Euler formula (see Subsection 1.2.1). The real and
imaginary parts of the complex exponential functions are, respectively,
u(x, y) = ex cos y and v(x, y) = ex sin y.
54 Analytic Functions

From eq. (2.1.6), one can deduce that the real sine and cosine functions are
related to eiy and e−iy by
eiy − e−iy eiy + e−iy
sin y = and cos y = .
2i 2
The complex extension of the sine and cosine functions are then deduced to be
eiz − e−iz eiz + e−iz
sin z = and cos z = , for all z in C. (2.1.7)
2i 2
Correspondingly, other common complex trigonometric functions are defined
by
sin z 1 1 1
tan z = , cot z = , sec z = , cosec z = . (2.1.8)
cos z tan z cos z sin z

2.2 Limit and continuity of complex functions

Similar to the calculus of real variables, the differentiability of a complex
function is defined using the concept of limit. Here, we start with the precise
definition of the limit of a complex function.

2.2.1 Limit of a complex function

Let w = f (z) be defined in the point set S and z0 be a limit point of S. The
mathematical statement
lim f (z) = L, z ∈ S, (2.2.1)
z→z0

means that the value w = f (z) can be made arbitrarily close to L if we choose
z to be close enough, but not equal, to z0 . The formal definition of the limit of
a function is stated as:
For any  > 0, there exists δ > 0 (usually dependent on ) such that

|f (z) − L| <  if 0 < |z − z0 | < δ.

The limit L, if exists, must be unique. The value of L is independent of the
direction along which z → z0 (see Figure 2.4).
The function f (z) needs not be defined at z0 in order for the function to
have a limit at z0 . However, we do require z0 to be a limit point of S so that
it would never occur that f (z) is not defined in some deleted neighborhood of
z0 . For example, let us consider limz→0 sinz z . The domain of definition of sinz z is
C\{0}. Though sinz z is not defined at z = 0, this is a limit point of the domain
of definition. Hence, the above limit is well defined.
 2.2 Limit and continuity of complex functions 55

Figure 2.4. The region 0 < |z − z0 | < δ in the z-plane is mapped onto the region
enclosed by the curve  in the w-plane. The curve  lies completely inside the annulus
0 < |w − L| < .

If L = α + iβ, f (z) = u(x, y) + iv(x, y), z = x + iy and z0 = x0 + iy0 ,

then

|u(x, y) − α| ≤ |f (z) − L| ≤ |u(x, y) − α| + |v(x, y) − β|,

|v(x, y) − β| ≤ |f (z) − L| ≤ |u(x, y) − α| + |v(x, y) − β|.

From the above inequalities, it is obvious that eq. (2.2.1) is equivalent to the
following pair of limits

lim u(x, y) = α (2.2.2a)

(x,y)→(x0 ,y0 )

lim v(x, y) = β. (2.2.2b)

(x,y)→(x0 ,y0 )

Therefore, the study of the limiting behavior of f (z) is equivalent to that of a

pair of real functions u(x, y) and v(x, y). Consequently, theorems concerning
the limit of the sum, difference, product and quotient of complex functions hold
as to those for real functions. Suppose that

lim f1 (z) = L1 and lim f2 (z) = L2

z→z0 z→z0

Then

lim (f1 (z) ± f2 (z)) = L1 ± L2 , (2.2.3a)

z→z0

lim f1 (z)f2 (z) = L1 L2 , (2.2.3b)

z→z0
f1 (z) L1
lim = , L2 = 0. (2.2.3c)
z→z0 f2 (z) L2
56 Analytic Functions

Example 2.2.1 Prove that

lim z2 = α 2 , α is a fixed complex number,

z→α

using the -δ criterion.

Solution It suffices to establish that for any given  > 0, there exists a positive
number δ such that

|z2 − α 2 | <  whenever 0 < |z − α| < δ.

Observing

z2 − α 2 = (z − α)(z + α) = (z − α)(z − α + 2α)

and applying the triangle inequality, we obtain

|z2 − α 2 | = |z − α||z − α + 2α| ≤ |z − α|(|z − α| + 2|α|)

provided that z liesinside the deleted δ-neighborhood of α. Here, δ is chosen


to be less than min 1, 1+2|α| , so

|z2 − α 2 | ≤ |z − α|(|z − α| + 2|α|) < (1 + 2|α|) = .
1 + 2|α|
Note that the choice of δ depends on  and α.

Limit at infinity
The definition of limit holds even when z0 or L is the point at infinity. We can
simply replace the corresponding neighborhood of z0 or L by the neighborhood
of infinity. The mathematical statement

lim f (z) = L (2.2.4)

z→∞

can be understood as:

For any  > 0, there exists δ() > 0 such that

|f (z) − L| <  whenever |z| > 1δ .

Here, z refers to a point in the finite complex plane and |z| > 1δ is visualized
as a deleted neighborhood of ∞. Also, we must be cautious that the results in
eqs. (2.2.3a,b,c) hold for z0 , L1 and L2 in the finite complex plane only.
Suppose we define w = 1z . Then z → ∞ is equivalent to w → 0. It is then
not surprising to have the following properties on limit at infinity.
 2.2 Limit and continuity of complex functions 57

Theorem 2.2.1 If z0 and w0 are points in the z-plane and the w-plane respec-
tively, then

(a) limz→z0 f (z) = ∞ if and only if limz→z0 1

f (z)
= 0;
1
(b) limz→∞ f (z) = w0 if and only if limz→0 f z = w0 .

Proof
(a) limz→z0 f (z) = ∞ implies that for any  > 0, there exists a positive
number δ such that
1
|f (z)| > whenever 0 < |z − z0 | < δ.

The above result may be rewritten as
 
 1 
 
 f (z) − 0 <  whenever 0 < |z − z0 | < δ,

so we obtain
1
lim = 0.
z→z0 f (z)
(b) limz→∞ f (z) = w0 implies that for any  > 0, there exists δ > 0 such
that
1
|f (z) − w0 | <  whenever |z| > .
δ
Replacing z by 1z , we obtain
   
 
f 1 − w0  <  whenever 0 < |z − 0| < δ,
 z 

so we obtain
 
1
lim f = w0 .
z→0 z

The above results provide the convenient tools to evaluate limits on infinity.
For example,
z−1 z2 + 1
lim =0 so lim = ∞.
z→1 z2 + 1 z→1 z − 1

Also,
1 + 4z2
1
z2
+4 1
lim = lim =
z→0 5 + iz2 z→0 5
+i 5
z2
58 Analytic Functions

would imply
z2 + 4 1
lim = .
z→∞ 5z2 + i 5

2.2.2 Continuity of a complex function

Continuity of a complex function is defined in the same manner as that of a
real function. The complex function f (z) is said to be continuous at z0 if
lim f (z) = f (z0 ). (2.2.5)
z→z0

The above statement implicitly implies the existence of both lim f (z) and
z→z0
f (z0 ). Alternatively, the statement can be understood as:
For any  > 0, there exists δ > 0 (usually dependent on ) such that
|f (z) − f (z0 )| <  whenever |z − z0 | < δ.

Since z0 lies in the domain of definition of f (z), the deleted

δ-neighborhood of z0 in the former definition of the limit of a function is now
replaced by the δ-neighborhood of z0 in this new definition.
A complex function is said to be continuous in a region R if it is continuous
at every point in R.
Since continuity of a complex function is defined using the concept of limits,
it can be shown similarly that eq. (2.2.5) is equivalent to
lim u(x, y) = u(x0 , y0 ), (2.2.6a)
(x,y)→(x0 ,y0 )

lim v(x, y) = v(x0 , y0 ). (2.2.6b)

(x,y)→(x0 ,y0 )

For example, consider f (z) = ez ; its real and imaginary parts are, respectively,
u(x, y) = ex cos y and v(x, y) = ex sin y. Since both u(x, y) and v(x, y) are
continuous at any point (x0 , y0 ) in the finite x-y plane, we conclude that ez is
continuous at any point z0 = x0 + iy0 in C.
Theorems on real continuous functions can be extended to complex contin-
uous functions. If two complex functions are continuous at a point, then their
sum, difference and product are also continuous at that point; and their quotient
is continuous at any point where the denominator is non-zero. For example,
since g(z) = z2 is continuous everywhere, we conclude by the above remark
that both z2 ± ez and z2 /ez are continuous in C. Examples of complex contin-
uous functions in C are polynomials, exponential functions and trigonometric
functions.
Another useful result is that a composition of continuous functions is contin-
uous. If f (z) is continuous at z0 and g(z) is continuous at ξ , and if ξ = f (z0 ),
 2.2 Limit and continuity of complex functions 59

then the composite function g(f (z)) is continuous at z = z0 . Thus, functions

like sin(z2 ) and cos(z2 ) are continuous functions in C.
Since continuity of f (z) implies continuity of its real and imaginary parts,
the real function


|f (z)| = u(x, y)2 + v(x, y)2 , f = u + iv and z = x + iy,

is also continuous. By applying the well-known result on boundedness of a

continuous real function in a closed and bounded region, we can deduce a
related property on boundedness of the modulus of a continuous complex
function. We state without proof the following theorem.

Theorem 2.2.2 If f (z) is continuous in a closed and bounded region R, then

|f (z)| is bounded in the region, that is,

|f (z)| < M, for all z ∈ R, (2.2.7)

for some constant M. Also, |f (z)| attains its maximum value at some point z0
in R.

Example 2.2.2 Discuss the continuity of the following complex functions at

z = 0:

0 z=0
(a) f (z) = Re z ;
 z = 0
|z|
Im z
(b) f (z) = .
1 + |z|

Solution
(a) Let z = x + iy, z = 0. Then
Re z x
=
.
|z| x + y2
2

Suppose z approaches 0 along the half straight line y = mx (x > 0).

Then
Re z x x 1
lim = lim+ √ = lim+ √ =√ .
z→0 |z| x→0 x +m x
2 2 2 x→0 x 1+m2 1 + m2
y = mx, x>0

Since the limit depends on m, lim f (z) does not exist. Therefore, f (z)
z→0
cannot be continuous at z = 0.
60 Analytic Functions

(b) Let z = x + iy. Then

Im z y
=
.
1 + |z| 1 + x2 + y2
Now, consider the limit
y
lim f (z) = lim
= 0 = f (0).
z→0 (x,y)→(0,0) 1 + x2 + y2
Therefore, f (z) is continuous at z = 0.

Uniform continuity
Suppose f (z) is continuous in a region R. Then by definition, at each
point z0 inside R and for any  > 0, we can find δ > 0 such that |f (z) −
f (z0 )| <  whenever |z − z0 | < δ. Usually δ depends on  and z0 together.
However, if we can find a single value of δ for each , independent of z0 chosen
in R, we say that f (z) is uniformly continuous in the region R.

Example 2.2.3 Show that

(a) f1 (z) = z2 is uniformly continuous in the region |z| < R, where 0 <
R < ∞.
(b) f2 (z) = 1z is not uniformly continuous in the region 0 < |z| < 1.

Solution
(a) It suffices to show that given any  > 0, we can find δ > 0 such that
|z2 − z02 | <  when |z − z0 | < δ, where δ depends on  but not on the
particular point z0 of the region. If z and z0 are any two points inside
|z| < R, then

|z2 − z02 | = |z + z0 | |z − z0 | ≤ {|z| + |z0 |}|z − z0 | < 2R|z − z0 |. (i)

This relation between |f1 (z) − f1 (z0 )| and |z − z0 | dictates the choice
of δ = 2R

, where δ depends on  but not on z0 . Now, given any  > 0,
suppose |z − z0 | < δ. Then by inequality (i), we have

|f1 (z) − f1 (z0 )| = |z2 − z02 | < 2R|z − z0 | < 2Rδ = .

Hence, f1 (z) = z2 is uniformly continuous in |z| < R.

(b) For z and z0 inside 0 < |z| < 1, we observe that
 
1 1  |z − z0 |

|f2 (z) − f2 (z0 )| =  −  = ,
z z0 |z0 | |z|
 2.3 Differentiation of complex functions 61

and |f2 (z) − f2 (z0 )| can be made to be larger than any positive number
when z0 becomes sufficiently close to 0. It is not possible to find δ that
depends on  but not z0 such that for any given , we have

|f2 (z) − f2 (z0 )| < 

for |z − z0 | < δ. Hence, f2 (z) = 1

z
is not uniformly continuous in 0 <
|z| < 1.

Most of the theorems related to the properties of continuity for real functions
can be extended to complex functions. However, this is not quite so when we
consider differentiation.

2.3 Differentiation of complex functions

Let the complex function f (z) be single-valued in a neighborhood of a point
z0 . The derivative of f (z) at z0 is defined by
df f (z) − f (z0 )
(z0 ) = lim
dz z→z0 z − z0
f (z0 + z) − f (z0 )
= lim , z = z − z0 , (2.3.1)
z→0 z
provided that the above limit exists. The value of the limit must be independent
of the path of z approaching z0 .
Since the derivative of a complex function is defined in a similar manner
to that of a real function, many formulas for the computation of derivatives of
complex functions are the same as those for the real counterparts. For example,
the derivative of a complex polynomial is the same as the real case:
d n
z = nzn−1 .
dz
Standard theorems on differentiation in real calculus also hold in the complex
counterpart. If the derivatives of two functions f and g exist at a point z, then
d
[cf (z)] = cf  (z), where c is a constant,
dz
d
[f (z) ± g(z)] = f  (z) ± g  (z),
dz
d
[f (z)g(z)] = f  (z)g(z) + f (z)g  (z),
dz

d f (z) f  (z)g(z) − g  (z)f (z)

= .
dz g(z) [g(z)]2
62 Analytic Functions

The chain rule for differentiation of composite functions also holds. Suppose
f has a derivative at z0 and g has a derivative at f (z0 ). Then the derivative of
g(f (z)) at z0 is given by

d
g(f (z0 )) = g  (f (z0 )) f  (z0 ).
dz
Since we cannot graph a complex function in the usual sense as a real
function, it is meaningless to visualize f  (z0 ) as the ‘slope’ of some curve as
we do in the real case.
Like real calculus, existence of the derivative of a complex function at a point
implies continuity of the function at the same point. Supposing f  (z0 ) exists,
we consider
f (z) − f (z0 )
lim [f (z) − f (z0 )] = lim lim (z − z0 ) = 0
z→z0 z→z0 z − z0 z→z0

so that

lim f (z) = f (z0 ).

z→z0

This shows that f (z) is continuous at z0 . However, continuity of f (z) may not
imply the differentiability of f (z) at the same point.
It may occur that a complex function can be differentiable at a given point
but not so in any neighborhood of that point (see Example 2.3.1).

Example 2.3.1 Show that the functions z and Re z are nowhere differentiable,
while |z|2 is differentiable only at z = 0.

Solution According to definition (2.3.1), the derivative of z is given by

d z + z − z z
z = lim = lim = lim e−2iArg z .
dz z→0 z z→0 z z→0

The value of the limit depends on the path approaching z. Therefore, z is

nowhere differentiable. Similarly,

d d 1
Re z = (z + z)
dz dz 2
1 (z + z + z + z) − (z + z)
= lim
2 z→0 z
1 z + z 1 1 z
= lim = + lim .
2 z→0 z 2 2 z→0 z
 2.3 Differentiation of complex functions 63

Again, Re z is shown to be nowhere differentiable. Lastly, the derivative of |z|2

is given by

d 2 |z + z|2 − |z|2 z
|z| = lim = lim z + z + z .
dz z→0 z z→0 z

The above limit exists only when z = 0, that is, |z|2 is differentiable only at
z = 0.

2.3.1 Complex velocity and acceleration

A complex number z can be visualized geometrically as a position vector in
the complex plane. Suppose z(t) is considered as a position vector with the
running parameter t. The terminal point of the position vector traverses a curve
C in the complex plane. Similar to the differentiation of a vector function, we
define the derivative of z(t) with respect to t to be
dz z(t + t) − z(t)
= lim .
dt t→0 t
Suppose we separate z(t) into its real and imaginary parts and write z(t) =
x(t) + iy(t). Then the derivative of z(t) can be expressed as
dz dx dy
= +i . (2.3.2)
dt dt dt
The derivative gives the direction of the tangent vector to the curve at t. If the
parameter t is considered as the time variable, then dz dt
represents the velocity
with which the terminal point moves along the curve. Also, the second-order
2
derivative ddt 2z gives the acceleration of the motion along the curve.

Example 2.3.2 Suppose the motion of a particle is described using the polar
coordinates (r, θ ) and its position in the complex plane is represented by

z(t) = r(t)eiθ(t) .

By differentiating z(t) with respect to the time variable t, find the velocity
and acceleration of the particle, separating them into their radial and tangential
components.

Solution Starting with z = reiθ , where z, r and θ are all functions of t, we

differentiate z with respect to the time variable t and obtain

u = ż = ṙeiθ + ireiθ θ̇ .
64 Analytic Functions

Here, u is called the complex velocity and the dot over a variable denotes
differentiation of the variable with respect to t. Also, eiθ and ieiθ represent the
unit vector in the radial direction and tangential direction, respectively. The
radial component of velocity ur and the tangential component of velocity uθ
are then given by

ur = ṙ and uθ = r θ̇.

The complex velocity may be written as

u = (ur + iuθ )eiθ .

The complex acceleration can be found by differentiating u again with respect

to t. We obtain
du
a= = (r̈ − r θ̇ 2 )eiθ + (2ṙ θ̇ + r θ̈)ieiθ .
dt
The radial component of acceleration ar and the tangential component of
acceleration aθ are then given by

ar = r̈ − r θ̇ 2 and aθ = 2ṙ θ̇ + r θ̈.

2.4 Cauchy–Riemann relations

This section discusses the necessary and sufficient conditions for the existence
of the derivative of a complex function f (z) = u(x, y) + iv(x, y), z = x + iy.
The necessary conditions are given by the Cauchy–Riemann relations. The
sufficient conditions require, additionally, the continuity of all first-order partial
derivatives of u and v.
Let f (z) be defined in a neighborhood of the point z0 = x0 + iy0 , and suppose
it is differentiable at z0 , that is, the limit
f (z0 + z) − f (z0 )
f  (z0 ) = lim
z→0 z
exists. This limit is independent of the direction along which z approaches 0.

(i) First, we take z → 0 in the direction parallel to the x-axis, that is,
z = x. We then have

f (z0 + z) − f (z0 ) = u(x0 + x, y0 ) + iv(x0 + x, y0 )

− u(x0 , y0 ) − iv(x0 , y0 ),
 2.4 Cauchy–Riemann relations 65

so that
u(x0 + x, y0 ) − u(x0 , y0 )
f  (z0 ) = lim
x→0 x
v(x0 + x, y0 ) − v(x0 , y0 )
+ i lim
x→0 x
∂u ∂v
= (x0 , y0 ) + i (x0 , y0 ).
∂x ∂x
(ii) Next, we let z → 0 in the direction parallel to the y-axis, that is,
z = iy. Now, we have

f (z0 + z) − f (z0 ) = u(x0 , y0 + y) + iv(x0 , y0 + y)

− u(x0 , y0 ) − iv(x0 , y0 ),

so that
u(x0 , y0 + y) − u(x0 , y0 )
f  (z0 ) = lim
y→0 iy
v(x0 , y0 + y) − v(x0 , y0 )
+ i lim
y→0 iy
1 ∂u ∂v
= (x0 , y0 ) + (x0 , y0 ).
i ∂y ∂y
Combining the above two equations, we obtain
∂u ∂v ∂v ∂u
f = +i = −i . (2.4.1)
∂x ∂x ∂y ∂y
Equating the respective real and imaginary parts gives
∂u ∂v ∂v ∂u
= and =− . (2.4.2)
∂x ∂y ∂x ∂y
The results in eq. (2.4.2) are called the Cauchy–Riemann relations. They give
the necessary conditions for the existence of the derivative of a complex func-
tion. The above results are summarized in the following theorem.

Theorem 2.4.1 Suppose f (z) = u(x, y) + iv(x, y) is differentiable at a point

z0 = x0 + iy0 . Then at that point

f  (z0 ) = ux (x0 , y0 ) + ivx (x0 , y0 ) = vy (x0 , y0 ) − iuy (x0 , y0 ).

Accordingly, we have

ux (x0 , y0 ) = vy (x0 , y0 ) and vx (x0 , y0 ) = −uy (x0 , y0 ).

66 Analytic Functions

Readers should be wary that the satisfaction of the Cauchy–Riemann

relations may not guarantee the existence of f  at the point (see
Example 2.4.2). However, a partial converse can be salvaged by adding some
extra assumptions.

Theorem 2.4.2 Given f (z) = u(x, y) + iv(x, y), z = x + iy, and assume
that
(i) the Cauchy–Riemann relations hold at a point z0 = x0 + iy0 ,
(ii) ux , uy , vx , vy are all continuous at the point (x0 , y0 ).
The derivative f  (z0 ) then exists and it is given by
f  (z0 ) = ux (x0 , y0 ) + ivx (x0 , y0 ) = vy (x0 , y0 ) − iuy (x0 , y0 ).

Proof Since u(x, y) and v(x, y) have continuous first-order partial derivatives
at (x0 , y0 ) and satisfy the Cauchy–Riemann relations at the same point, we have
u(x, y) − u(x0 , y0 )
= ux (x0 , y0 )(x − x0 ) + uy (x0 , y0 )(y − y0 ) + 1 (|z|)
= ux (x0 , y0 )(x − x0 ) − vx (x0 , y0 )(y − y0 ) + 1 (|z|), (i)
and
v(x, y) − v(x0 , y0 )
= vx (x0 , y0 )(x − x0 ) + vy (x0 , y0 )(y − y0 ) + 2 (|z|)
= vx (x0 , y0 )(x − x0 ) + ux (x0 , y0 )(y − y0 ) + 2 (|z|), (ii)
where 1 and 2 satisfy
1 (|z|) 2 (|z|)

lim = lim = 0, |z| = (x − x0 )2 + (y − y0 )2 .
|z|→0 |z| |z|→0 |z|
(iii)
Adding eq. (i) and i times eq. (ii) together, we obtain
f (z) − f (z0 ) = [ux (x0 , y0 ) + ivx (x0 , y0 )](z − z0 ) + 1 (|z|) + i2 (|z|),
and subsequently,
f (z) − f (z0 ) 1 (|z|) + i2 (|z|)
− [ux (x0 , y0 ) + ivx (x0 , y0 )] = .
z − z0 z − z0
Note that
 
 1 (|z|) + i2 (|z|)  1 (|z|) 2 (|z|)
 ≤ + ;
 z − z0  |z| |z|
 2.4 Cauchy–Riemann relations 67

and as z → 0, so does |z| → 0. It then follows from the results in eq. (iii)
that
1 (|z|) + i2 (|z|)
lim = 0.
z→0 z − z0
Hence, we obtain
f (z) − f (z0 )
f  (z0 ) = lim
z→z0 z − z0
= ux (x0 , y0 ) + ivx (x0 , y0 ) = vy (x0 , y0 ) − iuy (x0 , y0 ).

Example 2.4.1 Express the Cauchy–Riemann relations in polar coordinates.

Solution Consider the polar coordinates r 2 = x 2 + y 2 and θ = tan−1 yx .

Differentiating r and θ with respect to both x and y, we obtain
∂r x ∂r y
= = cos θ, = = sin θ,
∂x r ∂y r
and
∂θ y 1 ∂θ x 1
=− 2 = − sin θ, = 2 = cos θ.
∂x x + y2 r ∂y x + y2 r
Using the chain rule, the first-order partial derivatives of u are given by
∂u ∂u ∂r ∂u ∂θ ∂u 1 ∂u
= + = cos θ − sin θ, (i)
∂x ∂r ∂x ∂θ ∂x ∂r r ∂θ
∂u ∂u ∂r ∂u ∂θ ∂u 1 ∂u
= + = sin θ + cos θ. (ii)
∂y ∂r ∂y ∂θ ∂y ∂r r ∂θ
Similarly, the first-order partial derivatives of v are given by
∂v ∂v 1 ∂v
= cos θ − sin θ, (iii)
∂x ∂r r ∂θ
∂v ∂v 1 ∂v
= sin θ + cos θ. (iv)
∂y ∂r r ∂θ
Using one of the Cauchy–Riemann relations, we combine eqs. (i) and (iv) to
give
∂u ∂v
−
∂x ∂y
   
∂u 1 ∂v ∂v 1 ∂u
= − cos θ − + sin θ = 0. (v)
∂r r ∂θ ∂r r ∂θ
68 Analytic Functions

Similarly, using eqs. (ii) and (iii) and applying the other Cauchy–Riemann
relation, we have
∂u ∂v
+
∂y ∂x
   
∂u 1 ∂v ∂v 1 ∂u
= − sin θ + + cos θ = 0. (vi)
∂r r ∂θ ∂r r ∂θ
In order that eqs. (v) and (vi) are satisfied for all θ , we must have
∂u 1 ∂v ∂v 1 ∂u
= and =− .
∂r r ∂θ ∂r r ∂θ
These are the Cauchy–Riemann relations expressed in polar coordinates.

Example 2.4.2 Discuss the differentiability of the function

f (z) = f (x + iy) = |xy|

at z = 0.

Solution Write f (x + iy) = u(x, y) + iv(x, y) so that

u(x, y) = |xy| and v(x, y) = 0.

Since u(x, 0) and u(0, y) are identically equal to zero, we have

ux (0, 0) = uy (0, 0) = 0.

Also, since v(x, y) is identically zero, it is obvious that

vx (0, 0) = vy (0, 0) = 0.

Hence, the Cauchy–Riemann relations are satisfied at the point (0, 0).
However, suppose z approaches the origin along the ray x = αt, y = βt, t >
0, assuming that α and β cannot be zero simultaneously. For z = αt + iβt, we
then have
√
f (z) − f (0) f (z) |αβ|
= = .
z−0 z α + iβ
The limit of the above quantity as z → 0 depends on the values of α and β, so
the limit is non-unique. Therefore, f (z) is not differentiable at z = 0, though
the Cauchy–Riemann relations are satisfied at z = 0.
Let us check the continuity of ux at (0, 0). Since
∂u
d

= |y| |x|,
∂x dx
 2.4 Cauchy–Riemann relations 69

ux fails to be continuous at (0, 0). By virtue of Theorem 2.4.2, it is not surprising

√
that f (z) = |xy| can fail to be differentiable at z = 0 since the Cauchy–
Riemann relations are necessary but not sufficient for differentiability.

2.4.1 Conjugate complex variables

Let z = x + iy; its complex conjugate is z = x − iy. Formally, we may treat
the pair of conjugate complex variables, z and z, as two independent variables.
They are related to the real variables x and y by
z+z z−z
x= , y= . (2.4.3)
2 2i
Take the example: f (x, y) = x + ixy, the function can be expressed in terms
of z and z as
z + z z2 − z2
f (x, y) ≡ f (z, z) = + .
2 4
Applying the transformation rules in calculus, we have
 
∂ ∂ ∂ ∂ ∂ ∂
= + and =i − . (2.4.4)
∂x ∂z ∂z ∂y ∂z ∂z
∂ ∂
Here, ∂z and ∂z are visualized as the symbolic derivatives with respect to z and
z, respectively. Suppose we write
f (x, y) = u(x, y) + iv(x, y);
the Cauchy–Riemann relations can be expressed as
∂f ∂u ∂v ∂v ∂u ∂f
= +i = −i = −i .
∂x ∂x ∂x ∂y ∂y ∂y
Using the relations between the differential operators given in eq. (2.4.4), the
above equation can be reduced to
∂f
= 0. (2.4.5)
∂z
This is an elegant representation of the Cauchy–Riemann relations.
As an application of the result, one can conclude by a simple calculation that
the complex function
z+z
f (z) = Re z =
2
is nowhere differentiable since
∂ 1
Re z = = 0.
∂z 2
70 Analytic Functions

As another example, recall that the complex function

f (z) = |z|2 = zz

is differentiable only at z = 0. This is because it satisfies eq. (2.4.5) only at

z = 0 as
∂
zz = z
∂z
equals zero only at z = 0.

2.5 Analyticity
Analyticity is one of the central concepts in the calculus of complex variables.
It may be visualized as some form of an extended notion of differentiability.
Analyticity is a property defined over an open set while differentiability is
confined to one single point. Let us start with the definition of a function that
is analytic at a point.

Definition 2.5.1 A function f (z) is said to be analytic at some point z0 if

it is differentiable at every point inside a certain neighborhood of z0 . In other
words, f (z) is analytic at z0 if and only if there exists a neighborhood N (z0 ; ),
 > 0, such that f  (z) exists for all z ∈ N (z0 ; ).

Remark Some authors use the terms holomorphic or regular as synonyms for
analytic.

Since z0 ∈ N (z0 ; ), analyticity at z0 implies differentiability at z0 . The

converse statement is not true, that is, differentiability of f (z) at z0 does not
guarantee the analyticity of f (z) at z0 . For example, the function f (z) = |z|2
is nowhere differentiable except at the origin, hence f (z) = |z|2 is not analytic
at z = 0. A point at which f is not analytic is called a singular point (or
singularity) of f .

Analyticity in a domain
We say that a function is analytic in a region R if it is analytic at every point in
R. We would like to show that the region of analyticity must be open. Recall
that an open region is called a domain, so analyticity of a function is defined in
a domain. To prove the claim, consider a function that is analytic at some point
z0 inside R. According to the requirement of analyticity at a point, there exists
a neighborhood around z0 that lies completely inside the region. This would
 2.5 Analyticity 71

imply implicitly that z0 must be an interior point of the region. The region of
analyticity contains only interior points, so analytic functions are defined in
domains only.
Suppose a function f is said to be analytic in a closed region, say |z| ≤ R.
Then it is implicit that the function is analytic in some domain D containing
the closed region. If a complex function is analytic in the entire complex plane,
then the function is called an entire function. Examples of entire functions are
polynomials, exponential functions and trigonometric functions.
It is relatively straightforward to show that the existence of f  for all points
inside a domain implies analyticity of f in the domain. Since any point in the
domain of analyticity of a complex function is an interior point, there exists
a neighborhood of that point that is contained completely inside the domain
of analyticity. This implies implicitly that f  exists for all points inside the
neighborhood, so f is analytic at that point. In other words, in order to show
that f (z) is analytic in a domain D, it suffices to show either

(i) f  (z) exists for all z in D, or

(ii) the real and imaginary parts of f (z) have continuous first-order partial
derivatives, and these derivatives satisfy the Cauchy–Riemann relations
at every point inside D.

Suppose two complex functions are analytic in some domain D. Then their
sum, difference and product are all analytic in D. Also, their quotient is analytic
in D given that the denominator function is non-zero at all points in D. The
composition of two analytic functions is also analytic. Suppose f1 (z) is analytic
at z0 while f2 (z) is non-analytic at z0 . Is the sum f1 (z) + f2 (z) analytic at z0 ?
Let us assume

S(z) = f1 (z) + f2 (z)

to be analytic at z0 . Then S(z) − f1 (z) is analytic at z0 since the difference

of two functions analytic at z0 remains analytic at that point. This leads to a
contradiction since

S(z) − f1 (z) = f2 (z)

is non-analytic at z0 . Hence, f1 (z) + f2 (z) fails to be analytic at z0 . However,

given two functions that are non-analytic at z0 , it may be possible that their sum
becomes analytic at z0 . For example, f1 (z) = sin z + 1z and f2 (z) = ez − 1z are
both non-analytic at z = 0. However, their sum

f1 (z) + f2 (z) = sin z + ez

is an entire function.
72 Analytic Functions

Can we find a non-constant analytic function f (z) such that both f (z) and
f (z) are analytic in a domain D? This is not possible since we cannot find
a non-constant f (z) such that the Cauchy–Riemann relations are satisfied for
both f (z) and f (z) at every point in D. To prove the claim, we write

f (z) = u(x, y) + iv(x, y), z = x + iy,

f (z) = u(x, y) − iv(x, y).

Their respective pairs of Cauchy-Riemann relations are given by

ux = vy , uy = −vx ,

and

ux = −vy , uy = vx .

Combining the above relations, we obtain

ux = uy = vx = vy = 0

so that

f  (z) = ux (x, y) + ivx (x, y) = 0

in D. Observing the connectedness property of the domain D, we deduce that

f (z) is constant throughout D.

Example 2.5.1 Find the domains in which the function

f (z) = |x 2 − y 2 | + 2i|xy|, z = x + iy,

is analytic.

Solution The functional values of f (z) depend on the signs of x 2 − y 2 and xy.
When x 2 − y 2 > 0 and xy > 0, f (z) = z2 ; also when x 2 − y 2 < 0 and xy <
0, f (z) = −z2 . Both functions are known to be analytic. However, when x 2 −
y 2 > 0 and xy < 0, f (z) = x 2 − y 2 − 2ixy, the Cauchy–Riemann relations
are not satisfied, and so f (z) fails to be analytic. Also, for x 2 − y 2 < 0 and xy >
0, the function becomes f (z) = −(x 2 − y 2 ) + 2ixy, which is non-analytic.
Thus, the domains of analyticity of f (z) occupy alternative sectors, each
subtending an angle of π4 in the complex plane. Also, any neighborhood drawn
around any point on the boundary rays: x = ±y, x = 0 and y = 0 must overlap
with some region where f (z) is not differentiable. Hence, f (z) is not analytic
along these rays.
 2.5 Analyticity 73

Figure 2.5. Domains of analyticity (shown in shaded areas) of f (z) = |x 2 − y 2 | +

2i|xy|.

Summary
The function is analytic within the following domains (see Figure 2.5):
π π 3π
0 < Arg z < 4
, 2
< Arg z < 4
,
−π < Arg z < − 3π
4
and − π
2
< Arg z < − π4 .

Example 2.5.2 Given an analytic function

w = f (z) = u(x, y) + iv(x, y), z = x + iy,

the equations

u(x, y) = α and v(x, y) = β, α and β are constants,

define two families of curves in the complex plane. Show that the two families
are mutually orthogonal to each other.

Solution Consider a particular member from the first family

u(x, y) = α1 , for some constant α1 .
dy
The slope of the curve, given by dx , can be found by differentiating the above
equation with respect to x. This gives
∂u ∂u dy
+ = 0,
∂x ∂y dx
74 Analytic Functions

and upon rearranging, we obtain

$
dy ∂u ∂u
=− .
dx ∂x ∂y
Similarly, the slope of any member from the other family is given by
$
dy ∂v ∂v
=− .
dx ∂x ∂y
The product of the slopes of the two curves, one from each family, at their point
of intersection is found to be
 $  $ 
∂u ∂u ∂v ∂v
− − = −1,
∂x ∂y ∂x ∂y
by virtue of the Cauchy–Riemann relations. Therefore, the two families of
curves are mutually orthogonal to each other.

Remark The image curves of these two families of curves are mutually orthog-
onal to each other in the w-plane since they are simply the horizontal lines and
vertical lines in the w-plane. Orthogonal families of curves are mapped onto
orthogonal families of image curves. This preservation of the orthogonality
property under mapping stems from analyticity of the complex function f (z).
This interesting geometric property is closely related to the conformal property
of an analytic function (see Section 8.1).

2.6 Harmonic functions

A real-valued function φ(x, y) of two real variables x and y is said to be
harmonic in a given domain D in the x-y plane if φ has continuous partial
derivatives up to the second order in D and satisfies the Laplace equation

φxx (x, y) + φyy (x, y) = 0. (2.6.1)

Interestingly, analytic functions are closely related to harmonic functions. Sup-

pose

f (z) = u(x, y) + iv(x, y), z = x + iy,

is analytic in D; we will show that both the component functions u(x, y) and
v(x, y) are harmonic in D. We need one result to prove the above claim. It will
be shown in Section 4.3 that if a complex function is analytic at a point, then
its real and imaginary parts have continuous partial derivatives of all orders at
that point (see Remark (ii) in Theorem 4.3.2).
 2.6 Harmonic functions 75

Suppose f is analytic in D. Then the Cauchy–Riemann relations hold

throughout D; that is

ux = vy and vx = −uy .

The analyticity of f (z) in D dictates the continuity of partial derivatives of

u(x, y) and v(x, y) of all orders in D. Differentiating both sides of the Cauchy–
Riemann relations with respect to x, we obtain

uxx = vyx and vxx = −uyx .

Similarly, differentiation of the Cauchy–Riemann relations with respect to y

gives

uxy = vyy and vxy = −uyy .

Since the above partial derivatives are all continuous, we have

uxy = uyx and vxy = vyx .

Combining the above relations, we obtain

vyy = uxy = uyx = −vxx so vxx + vyy = 0; (2.6.2a)

and similarly,

−uyy = vxy = vyx = uxx so uxx + uyy = 0. (2.6.2b)

Therefore, both u(x, y) and v(x, y) are harmonic in D.

2.6.1 Harmonic conjugate

Given two harmonic functions φ(x, y) and ψ(x, y) that satisfy the Cauchy–
Riemann relations throughout a domain D, with

φx = ψy and φy = −ψx , (2.6.3)

we call ψ a harmonic conjugate of φ in D.

There is a close link between analyticity and harmonic conjugacy. Suppose
f = u + iv is analytic in D. We have shown that u and v are harmonic in
D. Also, from Theorem 2.4.1, u and v satisfy the Cauchy–Riemann relations.
Therefore, v is a harmonic conjugate of u in D. How about the validity of
the converse statement? If v is a harmonic conjugate of u in D, is the complex
function f = u + iv analytic in D? The proof of the converse statement follows
directly from Theorem 2.4.2. These results can be succinctly summarized by
the following theorem.
76 Analytic Functions

Theorem 2.6.1 A complex function f (z) = u(x, y) + iv(x, y), z = x + iy,

is analytic in a domain D if and only if v is a harmonic conjugate of u in D.

Note that harmonic conjugacy is not a symmetric relation because of the

minus sign in the second Cauchy–Riemann relation. In fact, while ψ is a
harmonic conjugate of φ, −φ is a harmonic conjugate of ψ.
Given that φ(x, y) is harmonic in a simply connected domain D, it can be
shown that it is always possible to obtain its harmonic conjugate ψ(x, y) by
integration. Starting from the differential form
dψ = ψx dx + ψy dy,
and using the Cauchy–Riemann relations, we have
dψ = −φy dx + φx dy.
The right-hand side of the above equation contains known terms φx and φy
since φ(x, y) has been given. To obtain ψ, we integrate along some path 
joining a fixed point (x0 , y0 ) to (x, y), that is,

ψ(x, y) = − φy dx + φx dy. (2.6.4)


The above line integral is an exact differential provided that

− (−φy )y + (φx )x = 0. (2.6.5)
The required condition is satisfied since it is simply the harmonicity property
of φ. Further, since D is a simply connected domain, the value of the above line
integral is independent of the path  chosen. To simplify the computation, we
choose the path that consists of horizontal and vertical line segments as shown
in Figure 2.6. This gives
 x  y
ψ(x, y) = − φy (x, y0 ) dx + φx (x, y) dy. (2.6.6)
x0 y0

The choice of a different starting point (x0 , y0 ) of the integration path simply
leads to a different additive constant in ψ(x, y). Indeed, a harmonic conjugate
is unique up to an additive constant since the governing equations for ψ(x, y)
involve derivatives of ψ(x, y) only [see eq. (2.6.3)].

Example 2.6.1 Let w = φ + iψ be an analytic function so that both φ and

ψ are harmonic. Suppose
∂ 2φ ∂ 2φ
∇ 2φ = + = 0 in D, and φ = g on ∂D,
∂x 2 ∂y 2
 2.6 Harmonic functions 77

y
(x, y)

(x0, y0) (x, y0)

x

Figure 2.6. To ease the computation in performing the line integration, the path of
integration is chosen to be composed of horizontal and vertical line segments.

where ∂D denotes the boundary of D. Show that the boundary condition for ψ
is given by

∂ψ ∂φ
=− on ∂D,
∂n ∂s

where ∂s∂ and ∂n

∂
denote the operators that perform differentiation with respect
to arc length s and outward normal n, respectively. Verify the result for the
harmonic function

sin x sinh y
φ(x, y) =
sin a sinh b

in the rectangle 0 ≤ x ≤ a and 0 ≤ y ≤ b.

Solution From the Cauchy–Riemann relations, we have

∂φ ∂ψ ∂φ ∂ψ
= and =− .
∂x ∂y ∂y ∂x

On the boundary ∂D, the tangential derivative of φ and the normal derivative
of ψ are related by
 
∂φ ∂φ ∂φ ∂ψ ∂ψ ∂ψ
− =−  +m =m − = ,
∂s ∂x ∂y ∂x ∂y ∂n

where  and m are the direction cosines of the unit tangent vector t.
78 Analytic Functions

Table 2.1. Verification of ∂ψ

∂n
= − ∂φ
∂s
along the four sides of the rectangular
domain.

∂g ∂ψ
φ = g on ∂D −
∂s ∂n
∂g ∂ψ
x = 0, 0 ≤ y ≤ b, (0, y) = 0 − (0, y) = 0
∂y ∂x
g(0, y) = 0
∂g ∂ψ
y = 0, 0 ≤ x ≤ a, − (x, 0) = 0 − (x, 0) = 0
∂x ∂y
g(x, 0) = 0
∂g cosh y ∂ψ cosh y
x = a, 0 ≤ y ≤ b, − (a, y) = − (a, y) = −
∂y sinh b ∂x sinh b
sinh y
g(a, y) =
sinh b
∂g cos x ∂ψ cos x
y = b, 0 ≤ x ≤ a, (x, b) = (x, b) =
∂x sin a ∂y sin a
sin x
g(x, b) =
sin a

The harmonic conjugate to φ(x, y) is given by (unique up to an additive

constant)
 (x,y)
ψ(x, y) = − φy dx + φx dy
(0,0)
 x  y
= − φy (x, 0) dx + φx (x, y) dy
0
 x 0
 y

1
= − sin x dx + cos x sinh y dy
sin a sinh b 0 0
cos x cosh y − 1
= .
sin a sinh b
The verification of ∂ψ
∂n
= − ∂φ
∂s
on the four sides of the rectangle is revealed in
Table 2.1.

Example 2.6.2 Given the harmonic function

u(x, y) = e−x cos y + xy,
find the family of curves that is orthogonal to the family
u(x, y) = α, α is constant.
 2.6 Harmonic functions 79

Solution First, it is observed that u(x, y) is harmonic in the whole complex

plane. Let v(x, y) denote the harmonic conjugate of u(x, y) (unique up to an
additive constant). From the result in Example 2.5.2, the required orthogonal
family of curves is given by

v(x, y) = β, β is constant.

Besides the line integration method given in eq. (2.6.6), we would like to
illustrate three other methods of finding the harmonic conjugate.

Method One
From the first Cauchy–Riemann relation, we have
∂v ∂u
= = −e−x cos y + y.
∂y ∂x
Integrating with respect to y, we obtain

y2
v(x, y) = −e−x sin y + + η(x),
2
where η(x) is an arbitrary function arising from integration. Using the second
Cauchy–Riemann relation, we have
∂v ∂u
= e−x sin y + η (x) = − = e−x sin y − x.
∂x ∂y
Comparing like terms, we obtain

η (x) = −x,

and subsequently,

x2
η(x) = − + C, where C is an arbitrary constant.
2
Hence, a harmonic conjugate is found to be (taking C to be zero for convenience)

y2 − x2
v(x, y) = −e−x sin y + .
2
The corresponding analytic function, f = u + iv, is seen to be

iz2
f (z) = e−z − , z = x + iy,
2
which is an entire function.
80 Analytic Functions

Method Two
It is readily seen that
1
e−x cos y = Re e−z and xy = Im z2 .
2
A harmonic conjugate of Re e−z is Im e−z , while that of 12 Im z2 is − 12 Re z2 .
Therefore, a harmonic conjugate of u(x, y) can be taken to be
1 y2 − x2
v(x, y) = Im e−z − Re z2 = −e−x sin y + .
2 2

Method Three
It is known that
∂u ∂u
f  (z) = (x, y) − i (x, y), z = x + iy.
∂x ∂y
Putting y = 0, we obtain
∂u ∂u
f  (x) = (x, 0) − i (x, 0).
∂x ∂y
By replacing x by z formally, this gives
∂u ∂u
f  (z) = (z, 0) − i (z, 0).
∂x ∂y
In the present problem, we have
∂u ∂u
(z, 0) = −e−z and (z, 0) = z
∂x ∂y
so that
f  (z) = −e−z − iz.
Integrating with respect to z, we obtain
iz2
f (z) = e−z − .
2

2.6.2 Steady state temperature distribution

One of the most interesting examples of applying the theory of harmonic
functions to physical modeling is the study of two-dimensional steady state
temperature fields. When steady state prevails, the temperature function inside
a two-dimensional body T (x, y) can be shown to be harmonic. To verify the
claim, we need to examine the fundamental physical process in heat conduction.
 2.6 Harmonic functions 81

∂T  ∆y 
−K  x, y +  ∆x
∂y  2 

∂T  ∆x  ∂T  ∆x 
−K x− , y  ∆y ∆y −K x+ , y  ∆y
∂x  2  ∂x  2 
∆x

∂T  ∆y 
−K  x, y −  ∆x
∂y  2 

Figure 2.7. An infinitesimal control volume of widths x and y is contained inside a

two-dimensional body. The heat fluxes across the four sides of the rectangular control
volume are shown.

The heat flux Q across a surface inside a solid body at a point is defined as the
amount of heat flowing normal to the surface per unit area per unit time at that
point. The empirical law of heat conduction states that the flux across a surface
is proportional to the normal temperature gradient at the point on the surface.
We consider a two-dimensional solid body bounded by a simple closed curve 
in the complex plane. Let T (x, y) denote the steady state temperature at a point
(x, y) inside . The mathematical statement of the law of heat conduction can
be written as
∂T
Q = −K , K > 0, (2.6.7)
∂n
where the constant K is called the thermal conductivity of the material of
the solid body. For a homogeneous solid, K may be assumed to be constant
throughout the body. The negative sign indicates that the heat flux flows from
the high temperature to the low temperature region.
Physical argument dictates that steady state temperature distribution prevails
if there is no heat source or sink present inside the body and there is no net heat
flux across the bounding surface. It is reasonable to assume that the temperature
function T (x, y) and all its partial derivatives up to second order are continuous
at all points inside the body.
We consider an infinitesimal control volume of rectangular shape inside a
two-dimensional body as shown in Figure 2.7. Let x and y be the widths
82 Analytic Functions

of the control volume along the x- and y-directions, respectively, and (x, y) be
the point at the center of the control volume.
Within a unit time interval, the amount of heat flowing across the
left vertical side into the rectangular control volume is −K ∂T ∂x
(x − x
2
, y)y.
Likewise, the amount of heat flowing across the right vertical side out of the
control volume is −K ∂T ∂x
(x + x
2
, y)y. Similarly, the amount of heat flowing
into the control volume across the bottom side is −K ∂T ∂y
(x, y − y
2
)x, and
the amount of heat flowing out across the top side is −K ∂T ∂y
(x, y + y2
)x.
The net accumulation of heat per unit time inside the control volume is
% &
∂T
(x + x , y) − ∂T (x − x , y)
∂T
∂y
(x, y + y
2
) − ∂T
∂y
(x, y − y
2
)
Kxy ∂x 2 ∂x 2
+ .
x y

When steady state prevails, there should be no heat accumulation in any

infinitesimal control volume inside the body. In the limits x → 0 and
y → 0, we then obtain
∂ 2T ∂ 2T
+ = 0. (2.6.8)
∂x 2 ∂y 2
Hence, the steady state temperature function T (x, y) is a harmonic function.

Example 2.6.3 Isothermal curves in a temperature field are curves that con-
nect points with the same temperature value. Supposing the isothermal curves
of a steady state temperature field are given by the family of parabolas

y 2 = α 2 + 2αx, α is real positive,

in the x-y plane, find the general solution of the temperature function T (x, y).

Solution First, we solve for the parameter α in the equation of the isothermal
curves. This gives


α = −x + x 2 + y 2 ,

where the positive sign is chosen since α > 0. A naive guess may suggest that
the temperature function T (x, y) is given by


T (x, y) = −x + x 2 + y 2 .

However, since T (x, y) has to be harmonic, the above function cannot be a

feasible solution.
The systematic approach to finding the temperature function is to set

T (x, y) = f (t),
 2.6 Harmonic functions 83


where t = x 2 + y 2 − x, and f is some function to be determined such that
T (x, y) is harmonic. To solve for f (t), we first compute
 2
∂ 2T  x  y2
= f (t)
− 1 + f (t) ,
∂x 2 x2 + y2 (x 2 + y 2 )3/2
∂ 2T  y2  x2
= f (t) + f (t) .
∂y 2 x2 + y2 (x 2 + y 2 )3/2
The requirement that T (x, y) satisfies the Laplace equation leads to
 
x 1 f  (t) 1
2 1−
f  (t) +
f  (t) = 0 or  (t)
=− .
x +y
2 2 x +y
2 2 f 2t

The above equation can be integrated twice to give

√
f (t) = C1 t + C2 ,

where C1 and C2 are arbitrary constants. The temperature function is then given
by


T (x, y) = f (t) = C1 x 2 + y 2 − x + C2 .

When expressed in polar coordinates, the temperature function becomes

√ θ
T (r, θ ) = C1 r(1 − cos θ ) + C2 = C1 2r sin + C2 .
2
√
Remark Since T (r, θ ) can be expressed as 2C1 Im z1/2 + C2 , the harmonic
conjugate of T (r, θ ) is easily seen to be
√ 

θ
F (r, θ ) = −C1 2r cos + C3 = −C1 x + x 2 + y 2 + C3 ,
2

where C3 is another arbitrary constant. Using the result in Example 2.5.2, the
family of curves defined by


x + x 2 + y 2 = β or y 2 = β 2 − 2βx, β > 0,

are orthogonal to the isothermal curves y 2 = α 2 + 2αx, α > 0.

Physically, the direction of heat flux is normal to the isothermal lines. There-
fore, the family of curves orthogonal to the isothermal lines are called the flux
lines. These flux lines indicate the flow directions of heat in the steady state
temperature field.
84 Analytic Functions

2.6.3 Poisson’s equation

A partial differential equation of the form
∇ 2 φ(x, y) = h(x, y) (2.6.9)
is called the Poisson equation. The use of conjugate complex variables may
facilitate the solution of this type of equation. Recall from eq. (2.4.4) that
 
∂ ∂ ∂ ∂ ∂ ∂
= + and =i − ,
∂x ∂z ∂z ∂y ∂z ∂z
so that
 2 
∂2 ∂2 ∂2 ∂2 ∂2 ∂ ∂2 ∂2
= + 2 + , = − − 2 + .
∂x 2 ∂z2 ∂z∂z ∂z2 ∂y 2 ∂z2 ∂z∂z ∂z2
In terms of the pair of conjugate complex variables, the Poisson equation can
be expressed as
 
∂ 2φ z+z z−z
4 =h , . (2.6.10)
∂z∂z 2 2i
The general solution to eq. (2.6.10) can be obtained by integrating the equation
with respect to z and z successively.

Example 2.6.4 Find the general solution to the Poisson equation

∇ 2 φ(x, y) = 8(x 2 − y 2 ).

Solution In terms of the pair of conjugate complex variables, the Poisson

equation can be expressed as
∂ 2φ
4 = 4(z2 + z2 ).
∂z∂z
Integrating the above equation with respect to z and z successively, we obtain
z3 z zz3
φ(z, z) = + + F (z) + G(z),
3 3
where F (z) and G(z) are arbitrary functions in z and z, respectively. In terms
of x and y, the solution takes the form
2 4
φ(x, y) = (x − y 4 ) + F (x + iy) + G(x − iy).
3
Note that F (x + iy) + G(x − iy) is the general solution to the homogeneous
equation ∇ 2 φ = 0, while 23 (x 4 − y 4 ) is a particular solution corresponding to
the non-homogeneous term 8(x 2 − y 2 ).
 2.7 Problems 85

2.7 Problems
2.1. Letting z = x + iy, find the real and imaginary parts of the following
functions:
1 i+z
(a) 2z3 − 3z; (b) ; (c) .
z i−z
2.2. Represent each of the following functions in terms of z and z:
(a) w = x 2 − y 2 + 2ixy;
(b) w = x(x 2 − 3y 2 ) − y(3x 2 − y 2 )i;
2(x 2 + y 2 ) − (x + iy)
(c) w = .
4(x 2 + y 2 ) − 4x + 1
2.3. For the function
1
w = f (z) = z + , z = x + iy and w = u + iv,
z
find the image curve corresponding to the circle |z| = r0 . Also, find the
preimage curves in the x-y plane corresponding to the coordinates curves
u = α and v = β in the u-v plane.
2.4. An isometry is a function f : C → C such that
|f (z1 ) − f (z2 )| = |z1 − z2 |,
for all z1 and z2 in C. Define
f (z) − f (0)
g(z) = .
f (1) − f (0)
(a) Show that g(z) is also an isometry if f is an isometry.
(b) By observing g(1) = 1 and g(0) = 0, show that the real parts of g(z)
and z are equal, for all z in C; and g(i) = i or −i.
(c) Show that
i. if g(i) = i, then g(z) = z;
ii. if g(i) = −i, then g(z) = z.
(d) Prove that any isometry f must be of the form
f (z) = αz + β or f (z) = αz + β,
where α and β are constants, and |α| = 1.
2.5. This problem finds the velocity of fluid flow emanating from a vortex.
Physically, the direction of the velocity of fluid flow due to a vortex is
tangential to the concentric circles drawn with the vortex as the center,
and the magnitude is inversely proportional to the distance from the
86 Analytic Functions

vortex. Suppose the vortex is placed at the point α in the complex plane.
Find the velocity function that describes the velocity of fluid flow at the
point z due to the vortex. How does the strength of the vortex enter into
the velocity function?
2.6. Suppose
lim f (z) = A and lim g(z) = B.
z→z0 z→z0

Show that
f (z) A
(i) lim f (z)g(z) = AB, (ii) lim = , B = 0.
z→z0 z→z0 g(z) B
Given that limz→z0 f (z) exists, show
lim f (z) = lim f (z)
z→z0 z→z0

directly from the definition of the limit of a complex function.

2.7 Show that
lim f (z) = ∞
z→∞

if and only if
1
lim  1  = 0.
z→0 f z

2.8. For each of the following functions, examine whether the function is
continuous at z = 0:
 
 0 z=0  0 z=0
(a) f (z) = Re z ; (b) f (z) = (Re z)2 .
 z = 0  z = 0
|z| |z|
2.9. Suppose a complex function f (z) is continuous in a region R; show that
its modulus |f (z)| is also continuous within the same region.
2.10. Show that f (z) = 1z is continuous in 0 < |z| < 1 (see Example 2.2.3).
2.11. Suppose a function f (z) is continuous and nonzero at a point z0 . Show
that f (z) is nonzero throughout some neighborhood of that point.
2.12. The motion of a particle in the complex plane is given by
z(t) = z1 cos2 t + z2 sin2 t,
where t is the time variable and z1 and z2 are some fixed complex
numbers. Describe the path traversed by the particle.
2.13. If a particle moves with the instantaneous speed v along any plane curve
C, show that the normal component of the acceleration at any point on
 2.7 Problems 87

C is given by v 2 /r, where r is the radius of curvature of the curve C at

that point.
2.14. Suppose the trajectory of a particle in the complex plane is described by
z(t) = a cos ωt + ib sin ωt.
Show that the acceleration at any point is always directed toward the
origin. Find the equation of the periodic trajectory.
2.15. For each of the following functions, determine the region in the complex
plane where the corresponding Cauchy–Riemann relations are satisifed:
1
(a) w = 3 − z + 2z2 ; (b) w = ; (c) w = |z|2 z; (d) w = z5 .
z
2.16. Consider the function f (z) = xy 2 + ix 2 y, z = x + iy. Find the region
where
(a) the Cauchy–Riemann relations are satisfied;
(b) the function is differentiable;
(c) the function is analytic.
2.17. Show that the function f (z) = z Re z is nowhere differentiable except at
the origin; hence find f  (0).
2.18. Find a complex function that is continuous for |z| < 1 but differentiable
only at the origin. Find another complex function that is continuous
in a region but differentiable only along certain lines in the region of
continuity.
2.19. Discuss the analyticity of each of the following functions. If the function
is analytic, find its derivative.
(a) f (z) = z|z|;
(b) f (z) = x 2 − y 2 − 3x + 2 + i(2x − 3)y, z = x + iy;
x+y x−y
(c) f (z) = 2 +i 2 , z = x + iy.
x + y2 x + y2
2.20. Let f (z) = (x − y)2 + 2i(x + y).
(a) Show that the Cauchy–Riemann relations are satisfied only along the
curve x − y = 1.
(b) Deduce that f (z) has a derivative along that curve and find the deriva-
tive value. Then explain why f (z) is nowhere analytic.
2.21. (a) Show that the function

f (z) = |Im(z2 )|
88 Analytic Functions

satisfies the Cauchy–Riemann relations at z = 0, but is not differen-

tiable at z = 0.
(b) Consider the function
 1
−
e z4 z = 0
f (z) = .
0 z=0

Show that the Cauchy–Riemann relations are satisfied at all points in

the complex plane, but f (z) is not analytic (and not continuous) at
z = 0.
2.22. Consider the following function

 Im(z2 )
(1 + i) z = 0
f (z) = |z|2 .

0 z=0
Show that the Cauchy–Riemann relations are satisfied at z = 0. Is f (z)
differentiable at z = 0?
2.23. Determine the set on which

z if |z| ≤ 1
f (z) = 2
z if |z| > 1
is analytic and compute its derivative.
2.24. Suppose the complex function

f (z) = u(x, y) + iv(x, y), z = x + iy,

satisfies the following conditions at z0 = x0 + iy0 :

(a) u(x, y) and v(x, y) are differentiable at the point (x0 , y0 ), and
 
 f (z) − f (z0 ) 
(b) lim   exists.
z→z0 z−z 
0

Show that either f (z) or f (z) is differentiable at z0 .

2.25. Let f (z) = u(x, y) + iv(x, y), z = x + iy. Determine the validity of the
following statements.

(a) Suppose u(x, y) and v(x, y) are differentiable at (x0 , y0 ); then f (z)
is also differentiable at z0 , z0 = x0 + iy0 .
(b) Suppose f (z) is analytic in a domain D, and u is a real constant; then
f (z) is constant throughout D.

2.26. Suppose f (z) is analytic inside the domain D. Show that f (z) is constant
inside D if it satisfies any one of the following conditions:
 2.7 Problems 89

(a) |f (z)| is constant inside D;

(b) Re f (z) = [Im f (z)]2 inside D.
2.27. Let f (z) be analytic in a domain D, and let α and β be two points inside
D. Assuming that all points on the line segment L joining α and β lie
inside D, show that there exist z1 and z2 on L such that
f (β) − f (α)
= Re f  (z1 ) + i Im f  (z2 ).
β −α
This result somewhat resembles the mean value theorem in real variable
calculus.

Hint: Parametrize the line joining z1 and z2 by

z(t) = α + (β − α)t, 0 ≤ t ≤ 1.
Write f = u + iv and consider the function value of f (z) along
L as a function of t. Set
f (z(t)) = u(α + (β − α)t) + iv(α + (β − α)t), 0 ≤ t ≤ 1.
Next, consider
f (β) − f (α) u(β) − u(α) v(β) − v(α)
= +i ,
β −α β −α β −α
then apply the mean value theorem in real variable calculus to
u and v by treating them as real variable functions of t.

2.28. Let  be a two-dimensional curve defined by the differentiable functions

x = x(t) and y = y(t), t1 ≤ t ≤ t2 ,
in the z-plane, where z = x + iy. Let w = f (z) be an analytic function
in a domain containing . Show that
dw dz
= f  (z) .
dt dt
2.29. Suppose f (z) is differentiable at z = z0 . Let z0 + z approach z0 along
a line making an angle α with the x-axis so that z = s eiα , where s is
the arc length variable. Show that
 
 −iα du dv
f (z0 ) = e +i ,
ds ds
where
du dv
= ∇α u and = ∇α v
ds ds
90 Analytic Functions

are the directional derivatives of u and v in the direction chosen. Similarly,

show that

f  (z0 ) = e−iα (∇α+ π2 v − i∇α+ π2 u).

Use the above results to deduce the Cauchy–Riemann relations.

2.30. Find an analytic function f (z) whose real part u(x, y) is

(a) u(x, y) = y 3 − 3x 2 y, f (i) = 1 + i;

y
(b) u(x, y) = 2 , f (1) = 0;
x + y2
(c) u(x, y) = (x − y)(x 2 + 4xy + y 2 ).

2.31. Find the orthogonal trajectories of each of the following families of

curves:

(a) x 3 y − xy 3 = α;
(b) 2e−x sin y + x 2 − y 2 = α;
(c) (r 2 + 1) cos θ = αr.

2.32. Determine the values of the parameters appearing in the following func-
tions such that the functions become analytic:

(a) f (z) = (x 2 + axy + by 2 ) + i(cx 2 + dxy + y 2 ), z ∈ C;

x + k − iy
(b) f (z) = 2 , z ∈ C\{−1};
x + y 2 + 2x + 1
(c) f (z) = ex (cos y + i sin y), z ∈ C.

2.33. The pair of real functions

y
u(x, y) = x 2 − y 2 and v(x, y) =
x2 + y2
are harmonic; however, the function f (z) = u(x, y) + iv(x, y),
z = x + iy, is non-analytic. Explain why.
2.34. If u(x, y) is a harmonic function, determine whether u2 is harmonic.
2.35. If f (z) is analytic and does not vanish in a domain D, determine whether
the modulus function |f (z)| is harmonic inside D.
2.36. If v is a harmonic conjugate of u in a domain D, is uv harmonic in D?
Give an explanation to your answer.
2.37. If u and v are harmonic in a domain D, prove that
   
∂u ∂v ∂u ∂v
− +i +
∂y ∂x ∂x ∂y
is analytic in D.
 2.7 Problems 91

2.38. Let θ = AP B, where A and B are the fixed points (−a, 0) and (a, 0),
respectively, and P is the variable point z = x + iy. Show that θ (x, y) is
a harmonic function. Find the corresponding harmonic conjugate v such
that θ + iv is an analytic function.

P (x, y)

(
q

x
A (–a, 0) B (a, 0)

2.39. From the polar form of the Cauchy–Riemann relations (see Exam-
ple 2.4.1), derive the Laplace equation in polar form.
2.40. Find a particular solution to each of the following Poisson equations:

(a) ∇ 2 u(x, y) = xyey ; (b) ∇ 2 u(r, θ ) = r 2 cos θ .

2.41. Consider
 
z+z z−z
w(z) = f (x, y) = f , = φ(z, z), z = x + iy,
2 2i
where the pair of conjugate complex variables z and z are considered to
be independent variables. Show that
   
∂ 1 ∂ ∂ ∂ 1 ∂ ∂
= −i and = +i .
∂z 2 ∂x ∂y ∂z 2 ∂x ∂y
Let u(x, y) and v(x, y) denote the real and imaginary parts of w(z),
respectively. Prove that
∂φ 1
(a) = [(ux + vy ) + i(−uy + vx )];
∂z 2
∂φ 1
(b) = [(ux − vy ) + i(uy + vx )];
∂z 2
dw ∂φ ∂φ −2iα
(c) = + e , α = Arg dz.
dz ∂z ∂z
92 Analytic Functions

2.42. Suppose f (z) = u(x, y) + iv(x, y) is analytic in a domain D, where

z = x + iy. Show that the Jacobian of the transformation from (x, y) to
(u, v) is given by
∂(u, v)
= |f  (z)|2 .
∂(x, y)
Also, show that
 
∂2 ∂2
2
+ 2 |f (z)|2 = 4|f  (z)|2 .
∂x ∂y
Show that, in general,
 2 
∂ ∂2
+ |f (z)|k = k 2 |f (z)|k−2 |f  (z)|2 ,
∂x 2 ∂y 2
where k is any positive integer.
2.43. Suppose a complex function f (z) is analytic in a domain D, and that it
assumes the form
f (z) = u(x, y) + iv(x, y) = XeiY , z = x + iy,
where X and Y are real, and Y is independent of x. Show that X must be
independent of y.
2.44. Suppose the isothermal lines of a steady state temperature field are the
family of curves
x 2 + y 2 = α, α > 0.
Find the general solution of the temperature function, and the equation
of the family of flux lines.
 3
Exponential, Logarithmic and
Trigonometric Functions

The full details of the properties of the complex exponential, logarithmic,

trigonometric and hyperbolic functions are discussed in this chapter. The
mapping properties of some of these complex functions are also examined.
These complex functions are defined as a natural extension of their real coun-
terparts (actually, they can be reduced to their corresponding real functions
by setting y = 0 in the independent complex variable z = x + iy). How-
ever, some of the properties of these complex functions may not be shared
by their real counterparts. For example, the complex exponential function is
periodic but the real one is not; the equation sin z = 2 has solutions only if z is
complex.
The logarithmic function, the inverse trigonometric and hyperbolic functions,
and the generalized power functions are multi-valued functions. The notion
of the Riemann surface for a multi-valued function is introduced in the final
section. The Riemann surface can be considered as an extension of the complex
plane to a surface which has more than one ‘sheet’. By virtue of the construction
of the Riemann surface, the multi-valued function assumes only one value
corresponding to each point on the Riemann surface.

3.1 Exponential functions

Recall that the complex exponential function is defined by (see Section 2.1.3)

ez = ex (cos y + i sin y), z = x + iy. (3.1.1)

Its real and imaginary parts are, respectively,

u(x, y) = ex cos y and v(x, y) = ex sin y. (3.1.2)

93
94 Exponential, Logarithmic and Trigonometric Functions

The derivative of ez is found to be

d z ∂ ∂
e = u(x, y) + i v(x, y)
dz ∂x ∂x
= ex cos y + iex sin y
= ez , for all z in C, (3.1.3)

that is, the derivative of ez is equal to itself. This is one of the fundamental
properties of the exponential function. Since the derivative of ez exists for all
z in the whole z-plane, the exponential function is an entire function. Also, it
can be verified that

ez1 +z2 = ez1 ez2 , (3.1.4)

another basic property of the exponential function. The modulus of ez is non-

zero since

|ez | = ex = 0, for all z in C, (3.1.5)

and so ez = 0 for all z in the complex z-plane. The range of the complex
exponential function is the entire complex plane except the zero value. Further,
it can be shown that

ez+2kπi = ez , for any z and integer k, (3.1.6)

that is, ez is periodic with the fundamental period 2π i. Interestingly, the com-
plex exponential function is periodic while its real counterpart is not.

3.1.1 Definition from the first principles

From the first principles, it seems natural to define the complex exponential
function as a complex function f (z) that satisfies the following defining prop-
erties:

(1) f (z) is entire,

(2) f  (z) = f (z),
(3) f (x) = ex , x is real.

We would like to show how the definition of the complex exponential functon
can be derived from these properties. Let f (z) = u(x, y) + iv(x, y), z = x +
iy. From property (1), u and v are seen to satisfy the Cauchy–Riemann relations.
Combining properties (1) and (2), we obtain the following relations:

ux + ivx = vy − iuy = u + iv.

 3.1 Exponential functions 95

First, we observe that

ux = u and vx = v.

From these two relations, we deduce that

u = ex g(y) and v = ex h(y),

where g(y) and h(y) are arbitrary functions of y. In addition, we also have the
relations

vy = u and uy = −v,

from which we deduce that the arbitrary functions are related by

h (y) = g(y) and − g  (y) = h(y).

By eliminating g(y) in the above relations, we obtain

h (y) = −h(y).

The general solution of the above equation is given by

h(y) = A cos y + B sin y,

where A and B are arbitrary constants. Furthermore, using g(y) = h (y), we

have

g(y) = −A sin y + B cos y.

To determine the arbitrary constants A and B, we use property (3) where

ex = u(x, 0) + iv(x, 0) = g(0)ex + ih(0)ex = Bex + iAex .

We then obtain B = 1 and A = 0. Putting all the results together, the complex
exponential function is found to be

f (z) = ez = ex cos y + iex sin y,

which agrees with the earlier definition given in eq. (3.1.1).

Example 3.1.1 Find all roots of the equation

ez = i.

Solution Equating the real and imaginary parts of ez and i leads to the
following pair of equations:

ex cos y = 0 and ex sin y = 1.

96 Exponential, Logarithmic and Trigonometric Functions

Since ex = 0 for all x, we deduce from the first equation that

π
y= + kπ, k is any integer.
2
From the second equation, since ex > 0 so that sin y > 0, the value of y is
restricted to
π
y= + 2kπ, k is any integer.
2
Substituting y into the second equation, we have ex = 1 and so x = 0. The
roots of the given equation are then found to be
 
1
z = 2k + π i, k is any integer.
2

Example 3.1.2 Consider the following function:

α
f (z) = e z , α is real.

Show that |f (z)| is constant on the circle x 2 + y 2 − ax = 0, where a is a real

constant.

Solution The equation of the given circle can be written in the standard form
 a 2  a 2
x− + y2 = ,
2 2
which reveals that the circle is centered at (a/2, 0) and has radius a/2. A
possible parametric representation of the circle is
a a
x= (1 + cos θ ) and y= sin θ, −π < θ ≤ π.
2 2
Geometrically, the parameter θ is the angle between the positive x-axis and
the line joining the center (a/2, 0) to the point (x, y). Correspondingly, the
complex representation of the circle can be expressed as
a
z= (1 + eiθ ), −π < θ ≤ π.
2
The modulus of f (z) when z lies on the circle is found to be
 2α   2α 1+cos θ −i sin θ 
    α
|f (z)| = e a(1+eiθ )  = e a 2(1+cos θ)  = e a .
 3.2 Trigonometric and hyperbolic functions 97
y v
w = ez

Arg w =

y=

x u

|w| = e

x=

Figure 3.1. Mapping properties of w = ez . Vertical lines are mapped onto concentric
circles centered at the origin and horizontal lines are mapped onto rays through the
origin.

The modulus value is equal to a constant with no dependence on θ , that is,

independent of the choice of the point on the circle.

3.1.2 Mapping properties of the complex exponential function

Since the complex exponential function is periodic with fundamental period
2π i, it is a many-to-one function. Suppose the domain of interest is chosen
to be the infinite strip −π < Im z ≤ π. Then the mapping w = ez becomes
one-to-one. The vertical line x = α is mapped onto the circle |w| = eα , while
the horizontal line y = β is mapped onto the ray Arg w = β (see Figure 3.1).
When the vertical line x = α moves further to the left, the mapped cir-
cle |w| = eα shrinks to a smaller radius. When the horizontal line in the
z-plane moves vertically from y = −π to y = π , the image ray in the w-
plane traverses in the anticlockwise sense from Arg w = −π to Arg w = π .
In particular, the whole x-axis is mapped onto the whole positive u-axis,
and the portion of the y-axis, −π < y ≤ π , is mapped onto the unit circle
|w| = 1.

3.2 Trigonometric and hyperbolic functions

Using the Euler formula

eiy = cos y + i sin y,

98 Exponential, Logarithmic and Trigonometric Functions

the real sine and cosine functions can be expressed in terms of eiy and e−iy as
follows:
eiy − e−iy eiy + e−iy
sin y = and cos y = .
2i 2
It is natural to define the complex sine and cosine functions in terms of the
complex exponential functions eiz and e−iz in the same manner as for the real
functions, that is,
eiz − e−iz eiz + e−iz
sin z = and cos z = . (3.2.1)
2i 2
The other complex trigonometric functions are defined in terms of the complex
sine and cosine functions by the usual formulas:
sin z cos z 1 1
tan z = , cot z = , sec z = , cosec z = . (3.2.2)
cos z sin z cos z sin z
The complex sine and cosine functions are entire since they are formed by the
linear combination of the entire functions eiz and e−iz . The functions tan z and
sec z are analytic in any domain that does not include points where cos z = 0.
Similarly, the functions cot z and cosec z are analytic in any domain excluding
those points z such that sin z = 0.
Let z = x + iy. Then

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

The real and imaginary parts of the complex sine and cosine functions are seen
to be

sin z = sin x cosh y + i cos x sinh y,

cos z = cos x cosh y − i sin x sinh y.

Moreover, their moduli are found to be

 
| sin z| = sin x + sinh y, | cos z| = cos2 x + sinh2 y.
2 2

Since sinh y is unbounded at large values of y, the above modulus values can
increase (as y does) without bound. While the real sine and cosine functions are
always bounded between −1 and 1, their complex counterparts are unbounded.
Next, we define the complex hyperbolic functions in the same manner as
their real counterparts. They are defined as
ez − e−z ez + e−z sinh z
sinh z = , cosh z = , tanh z = . (3.2.3)
2 2 cosh z
 3.2 Trigonometric and hyperbolic functions 99

The other hyperbolic functions cosech z, sech z and coth z are defined as the
reciprocal of sinh z, cosh z and tanh z, respectively.
In fact, the hyperbolic functions are closely related to the trigonometric
functions. Supposing z is replaced by iz in eq. (3.2.3), we obtain

sinh iz = i sin z.

Similarly, one can show that

sin iz = i sinh z, cosh iz = cos z, cos iz = cosh z.

The real and imaginary parts of sinh z and cosh z are found to be

sinh z = sinh x cos y + i cosh x sin y,

cosh z = cosh x cos y + i sinh x sin y,

and their moduli are given by


| sinh z| = sinh2 x + sin2 y,

| cosh z| = cosh2 x − sin2 y.

The complex hyperbolic functions sinh z and cosh z are periodic with
fundamental period 2π i, and tanh z is periodic with fundamental period
π i. Thus the complex hyperbolic functions are periodic, unlike their real
counterparts.
A zero α of a function f (z) satisfies f (α) = 0. Like their real counterparts,
the zeros of sin z are kπ and the zeros of cos z are kπ + π2 , k is any integer.
While the real cosh has no zero and the real sinh has only one zero at z = 0,
the complex cosh and sinh have infinitely many zeros.
To find the zeros of sinh z, we observe that

sinh z = 0 ⇔ | sinh z| = 0 ⇔ sinh2 x + sin2 y = 0.

Hence, x and y must satisfy sinh x = 0 and sin y = 0, thus giving x = 0 and
y = kπ, k is any integer. The zeros of sinh z are z = kπ i, k is any integer.
Similarly, the zeros of cosh z are z = (k + 12 )π i, k is any integer.
Knowing the derivative of ez , the derivatives of the trigonometric and hyper-
bolic functions can be found easily. Indeed, the derivative formulas for the
complex trigonometric and hyperbolic functions are exactly the same as those
for their real counterparts. In addition, the compound angle formulas for real
100 Exponential, Logarithmic and Trigonometric Functions

trigonometric and hyperbolic functions also hold for their complex counter-
parts. For example,

cos(z1 ± z2 ) = cos z1 cos z2 ∓ sin z1 sin z2 ,

sinh(z1 ± z2 ) = sinh z1 cosh z2 ± cosh z1 sinh z2 ,
tan z1 ± tan z2
tan(z1 ± z2 ) = .
1 ∓ tan z1 tan z2

Example 3.2.1 Show that cos z = cos z.

Solution Let z = x + iy, and consider

eiz = ei(x+iy) = e−y eix = e−y e−ix = e−i(x−iy) = e−iz .

Similarly, we have

e−iz = eiz .

Now, consider

1 iz 1
cos z = (e + e−iz ) = (e−iz + eiz ) = cos z.
2 2

Remark The above result is a manifestation of the reflection principle, which

states:

Suppose a function f is analytic in some domain D which includes

part of the real axis and D is symmetric about the real axis. Further,
f (x) is real whenever x is a point on that part of the real axis. We
then have

f (z) = f (z) for any z in D.

The conditions required in the reflection principle are seen to be satisfied by

f (z) = cos z, since the domain of analyticity is the whole complex plane and the
complex cosine function reduces to the real cosine function when the argument
is real.
By virtue of the reflection principle, a similar property about con-
jugation holds for all complex trigonometric and hyperbolic functions.
More detailed discussion of the reflection principle can be found in
Section 5.5.1.
 3.2 Trigonometric and hyperbolic functions 101

Example 3.2.2 Let z = x + iy. Derive the following lower bounds:

|ey − e−y | |ey − e−y |
(a) | sin z| ≥ ; (b) | tan z| ≥ y .
2 e + e−y

Solution

(a) Consider
 iz 
 e − e−iz  |eiz − e−iz |
| sin z| =  =
 .
2i 2
Use the triangle inequality to give
 
 iz 
 −iz 
|e − e | ≥ |e | − |e | = |ey − e−y |,
iz −iz

and finally
|ey − e−y |
| sin z| ≥ .
2
(b) From the relation
|eiz − e−iz |
| tan z| = ,
|eiz + e−iz |
and using the triangle inequality, we obtain

|eiz + e−iz | ≤ |eiz | + |e−iz | = ey + e−y .

Combining with the result in (a), we then have

|ey − e−y |
| tan z| ≥ .
ey + e−y

Example 3.2.3 Suppose z moves to infinity along a ray through the origin.
Discuss the possible values for

lim tan z.
z→∞

Solution Let z = reiθ , where θ is fixed for a given ray through the origin.
Consider the following cases:

(i) 0 < θ < π

We have
e2iz − 1
tan z = −i
e2iz + 1
102 Exponential, Logarithmic and Trigonometric Functions

and observe that

|e2iz | = e−2r sin θ and sin θ > 0.

As r → ∞, e−2r sin θ → 0 and so lim tan z = i.

r→∞
z=reiθ

(ii) −π < θ < 0

Now, we write
1 − e−2iz
tan z = −i
1 + e−2iz
and observe that

|e−2iz | = e2r sin θ and sin θ < 0.

As r → ∞, e2r sin θ → 0 and so lim tan z = −i.

r→∞
z=reiθ

(iii) θ = 0 or θ = π.
In these cases,

tan r when θ = 0
tan z = .
−tan r when θ = π
As r → ∞, the limit of tan r does not exist and so the same holds for
tan z.

3.2.1 Mapping properties of the complex sine function

Consider the complex sine function

w = sin z = sin x cosh y + i cos x sinh y, z = x + iy.

Suppose we write w = u + iv. Then

u = sin x cosh y and v = cos x sinh y.

To comprehend the mapping properties of w = sin z, we find the images of

the coordinates lines x = α and y = β. When x = α, u = sin α cosh y and
v = cos α sinh y. By eliminating y in the above equations, we obtain
u2 v2
2
− = 1, (3.2.4)
sin α cos2 α
which represents a hyperbola in the w-plane (see Figure 3.2a). When
0 < α < π/2, u = sin α cosh y > 0 for all values of y, so the line x = α
is mapped onto the right-hand branch of the hyperbola. Likewise, the line
 3.2 Trigonometric and hyperbolic functions 103

w = sin z
v
y D' A'
D A

E B E' B'
x u

F C
x= x= x= x= F'
2 2 C'

Figure 3.2.a Vertical lines are mapped onto hyperbolas under the mapping w = sin z.

y v
w = sin z

y= B'
C B A

C' A'
x u
D' F'
D E F y=
E'

x= x=
2 2

Figure 3.2.b Horizontal lines are mapped onto ellipses under the mapping w = sin z.

x = −α is mapped onto the left-hand branch of the same hyperbola. In par-

ticular, when α = π/2, the line x = π/2 is mapped onto the line segment
v = 0, u ≥ 1, which is a degenerated hyperbola (see Example 3.4.2). Also,
the y-axis is mapped onto the v-axis. We conclude that the infinite strip
{0 ≤ Re z ≤ π/2} is mapped to the right half-plane {u ≥ 0}; and by symme-
try, the other infinite strip {−π/2 ≤ Re z ≤ 0} is mapped to the left half-plane
{u ≤ 0}.
104 Exponential, Logarithmic and Trigonometric Functions

Next, we consider the image of a horizontal line y = β (β > 0), −π/2 ≤

x ≤ π/2 under the mapping w = sin z. When y = β, u = sin x cosh β and
v = cos x sinh β. By eliminating x in the above equations, we obtain
u2 v2
+ = 1, (3.2.5)
cosh2 β sinh2 β
which represents an ellipse in the w-plane (see Figure 3.2b). The upper line
y = β (the lower line y = −β) is mapped onto the upper (lower) portion of the
ellipse. In particular, when β = 0, the line segment y = 0, −π/2 ≤ x ≤ π/2
is mapped onto the line segment v = 0, −1 ≤ u ≤ 1, which is a degenerated
ellipse.
Since w = sin z is an entire function, by virtue of the orthogonality result
obtained in Example 2.5.2, the family of hyperbolas defined in eq. (3.2.4) are
orthogonal to the family of ellipses defined in eq. (3.2.5).

3.3 Logarithmic functions

The complex logarithmic function is defined as the inverse of the complex
exponential function. Suppose the logarithmic function is represented by
w = log z. (3.3.1)
Since it is the inverse of the exponential function, we then have
z = ew . (3.3.2)
Let u(x, y) and v(x, y) denote the real and imaginary parts of w = log z. From
eq. (3.3.2), we have
z = x + iy = eu+iv = eu cos v + ieu sin v.
Equating the real and imaginary parts on both sides gives
x = eu cos v and y = eu sin v.
It then follows that
y
e2u = x 2 + y 2 = |z|2 = r 2 and v = tan−1 . (3.3.3)
x
Using the polar form z = reiθ , we deduce from eq. (3.3.3) that
u = ln r = ln |z|, r = 0 and v = θ = arg z.
Putting the results together, we have
w = log z = ln |z| + i arg z, z = 0. (3.3.4)
 3.3 Logarithmic functions 105

Remark To avoid confusion, we follow the convention that ‘ln’ refers to real
logarithm while ‘log’ refers to complex logarithm.

Recall that arg z is multi-valued; so then is log z. For a fixed z, there are
infinitely many possible values of log z, each differing by a multiple of 2π i.
Among the possible values of argz, we then choose some value θ0 and restrict
arg z to θ0 < θ ≤ θ0 + 2π. In this way, we obtain a branch of arg z, and cor-
respondingly a branch of log z. Within a branch, the function arg z is single-
valued. In Section 1.1, we chose the principal value of arg z, denoted by Arg z,
as the branch where −π < Arg z ≤ π . This particular branch of log z corre-
sponding to the principal value of arg z is called the principal branch of log z.
From now on, the principal branch of the complex logarithmic function is
denoted by Log z, that is,

Log z = ln |z| + i Arg z, −π < Arg z ≤ π. (3.3.5)

One may write

log z = Log z + 2kπ i, k is any integer. (3.3.6)

For example, Log i = π2 i and log i = π2 i + 2kπ i, k is any integer.

By definition, z = elog z , however it would be incorrect to write z = log ez
since the logarithmic function is multi-valued.
Given two non-zero complex numbers z1 and z2 , we have

ln |z1 z2 | = ln |z1 | + ln |z2 |,

arg(z1 z2 ) = arg z1 + arg z2 ,

from which it can be deduced that

log(z1 z2 ) = log z1 + log z2 . (3.3.7)

The equality sign in eq. (3.3.7) actually means that any value of log(z1 z2 ) equals
some value of log z1 plus some value of log z2 .
From eq. (3.3.5), the real and imaginary parts of Log z are seen to be ln |z|
and Arg z, respectively. One then deduces that ln |z| is harmonic everywhere
except at z = 0, and Arg z is harmonic inside the domain −π < Arg z < π .

Example 3.3.1 Show from the first principles that

d 1
log z = , z = 0, ∞.
dz z
106 Exponential, Logarithmic and Trigonometric Functions

Solution Supposing we take differentiation along the x-axis, the derivative

of log z becomes
d ∂ ∂
log z = ln r + i (θ + 2kπ), z = 0, ∞,
dz ∂x ∂x


where r = x 2 + y 2 , θ = tan−1 y
x
and k is any integer. The above expression
can be simplified to
d 1 ∂r ∂θ
log z = +i
dz r ∂x ∂x
1 x −y
=

+i 2
x +y
2 2 x +y
2 2 x + y2
x − iy z 1
= 2 = = , z = 0, ∞.
x +y 2 zz z

3.3.1 Heat source

Consider the complex function
1
T (z) = −λ Re(Log z) + A = λ ln + A, λ > 0, z = reiθ , (3.3.8)
r
where A is a constant and z = 0. One argues that this function is a feasible
steady state temperature function since it is harmonic everywhere except at
z = 0. Since the temperature at the origin becomes infinite, T (z) is not analytic
at z = 0. The temperature field has radial symmetry since T (z) is independent
of θ.
In polar coordinates, the local heat flux across a curve  is given by (see
Section 2.6)

Q = −K∇T · n,

where n is the local normal vector to the curve , and K is the thermal
conductivity of the material. When the steady state temperature field has radial
symmetry, the normal gradient ∇T · n is simply dT /dr. For the temperature
field defined in eq. (3.3.8), the net rate of heat energy flowing across the circle
|z| = R is then given by
  2π   2π
 dT  1

Q = −K
dr  R dθ = K λ R dθ = 2π λK, (3.3.9)
R
r=R 0 r=R 0

which is constant and independent of R. Since the net heat flux is positive,
this indicates that heat is flowing radially outward from the origin. The inde-
pendence of R in the above expression for the heat flux reveals that the same
 3.3 Logarithmic functions 107

amount of heat energy flows across every circle centered at the origin when the
steady state is attained. Let m denote the amount of heat energy flowing out
from the origin per unit time. Then λ and m are related by
m
m = 2π λK or λ= .
2π K
Combining the above results, we deduce that T (z) as defined in eq. (3.3.8)
refers to the steady state temperature distribution due to a heat source of
intensity 2π λK placed at the origin. In particular, the temperature along the
circumference of the unit circle |z| = 1 is equal to A. The steady state condition
can only be attained when there is no net heat energy accumulated within any
pair of concentric circles centered at the source. The amount of heat flux flowing
across any circle centered at the heat source is exactly equal to the heat energy
generated at the heat source per unit time. In terms of source intensity m,
thermal conductivity K and temperature value along the unit circle T (1), the
temperature function due to a heat source at the origin can be expressed as
m 1
T (r) = ln + T (1). (3.3.10)
2π K r

Remark In the above derivation, we start with a feasible temperature function

and attempt to derive a physical interpretation of the function. Conversely, one
may start from the laws of physics and derive the required form of the steady
state temperature distribution. This approach is illustrated in the following
example.

Example 3.3.2 Suppose a heat source of intensity m is placed at the origin

so that heat flows radially outward uniformly in all directions. When the steady
state condition prevails, the net rate of heat energy flowing across any circle
|z| = r should be the same, independent of r. Otherwise, there will be net heat
accumulation. Show that the temperature function must be of the form

T (z) = α ln |z| + β, α < 0 and β is real.

Give a physical interpretation to the parameters α and β.

Solution In order that the net rate of heat energy flowing across any circle
|z| = r is independent of r, we deduce from eq. (3.3.9) that the temperature
gradient must be of the form
dT α
= , α is some real constant, (i)
dr r
108 Exponential, Logarithmic and Trigonometric Functions

so that the source intensity

 2π  2π
dT
m = −K r dθ = −K α dθ = −2π Kα (ii)
0 dr 0

is a constant. The temperature function can be obtained by integrating eq. (i)

to give

T (r) = α ln r + β, β is an arbitrary constant.

The constant β is seen to be equal to T (1), which is a real quantity [see

eq. (3.3.10)]. From eq. (ii), α is equal to − 2πK
m
, which is a negative quantity.
The resulting formula agrees with the temperature function given in eq. (3.3.10).
The sign of α is reversed when we deal with a heat sink (negative source). In
terms of the complex variable z, where z = reiθ , the temperature function takes
the form

T (z) = α ln |z| + β, α < 0 and β is real.

3.3.2 Temperature distribution in the upper half-plane

We would like to illustrate that the solution to the steady state temperature
distribution in the upper half-plane, Im z > 0, can be found readily if the
boundary temperature values along the x-axis assume some discrete constant
values. First, let T (x, 0) be given by

0 x>0
T (x, 0) = (3.3.11)
U x<0
which has a jump of discontinuity at x = 0. Consider the function Arg z =
Im(Log z) which is harmonic in the upper half-plane since it is the imaginary
part of the analytic function Log z in the same domain. Furthermore,

0 x>0
Arg x = .
π x<0
Hence, one deduces that the function
U U
T (z) = Arg z = Im(Log z), Im z > 0, (3.3.12)
π π
is a solution to this temperature distribution problem since T (z) is harmonic in
the domain Im z > 0 and satisfies the prescribed boundary temperature values.
Indeed, by the uniqueness property of solutions to the Laplace equation that
governs T (x, y) (see Subsection 7.1.1), this is the unique solution to the present
temperature distribution problem.
 3.3 Logarithmic functions 109

Figure 3.3. The angle θ is facing the segment of the x-axis with boundary temperature
value U .

We note that Arg x is related to the well-known Heaviside function H (x)

defined by

1 x>0
H (x) = . (3.3.13)
0 x<0

In terms of H (x), we observe that

T (x, 0) = U [1 − H (x)] and Arg x = π [1 − H (x)].

The above solution method can be easily generalized to the following form
of boundary temperature distribution along the x-axis:

U x 1 < x < x2
T (x, 0) =
0 x < x1 or x > x2
= U [H (x − x1 ) − H (x − x2 )]. (3.3.14)

Since the Laplace equation is linear, the superposition of solutions is also a

solution. The solution to this new problem is seen to be
U
T (z) = [Arg(z − x2 ) − Arg(z − x1 )], Im z > 0. (3.3.15)
π
A simple geometrical visualization may be helpful to recognize the form
of the above solution. In Figure 3.3, we set θ to be the angle subtended at z
and bounded by the rays joining z to the points z = x1 and z = x2 on the x-
axis. Geometrically, the subtended angle faces the segment of the x-axis with
boundary temperature value U . The value of θ is seen to be Arg(z − x2 ) −
Arg(z − x1 ), and so
U
T (z) = θ, Im z > 0. (3.3.16)
π
110 Exponential, Logarithmic and Trigonometric Functions

w = z2
y v

y
0

[U ]

[U ] [U ]
x u
x0 y 02 x2
0
Figure 3.4. The mapping w = z2 maps the first quadrant in the z-plane onto the upper
half
 w-plane.
  temperature value along the u-axis becomes T (u, 0) =
 The boundary
U H u + y02 − H u − x02 .

For points at close proximity to the two points of discontinuity z = x1 and

z = x2 , the temperature function can assume any value ranging from 0 to
U . Consider the case where z → x2 . It is seen that Arg(z − x1 ) ≈ 0 and so
T (z) ≈ (U/π ) Arg(z − x2 ). For Im z > 0, Arg(z − x2 ) takes values from 0 to
π , depending on the direction of approach of the point z to x2 . For example,
suppose z approaches x2 along the vertical direction such that Arg(z − x2 ) =
π/2. Then
U
lim
z→x2
T (z) = .
Arg(z−x2 ) = π/2
2

Readers are invited to seek the generalization of the present problem to

a boundary temperature distribution which has a finite number of discrete
constant values along the x-axis (see Problem 3.25).

Example 3.3.3 Find the steady state temperature distribution inside the first
quadrant x > 0, y > 0, where the boundary temperature values are given by
 
U 0 ≤ x < x0 U 0 ≤ y < y0
T (x, 0) = and T (0, y) = .
0 x > x0 0 y > y0

Solution The mapping w = z2 maps the first quadrant in the z-plane, z =

x + iy, onto the upper half w-plane, w = u + iv (see Figure 3.4).
In the w-plane,
 the boundary temperature  value along the u-axis is given
by T (u, 0) = U H (u + y02 ) − H (u − x02 ) . Using eq. (3.3.15), the temperature
 3.4 Inverse trigonometric and hyperbolic functions 111

function in the upper half w-plane is found to be

U
T (w) = [Arg(w − x02 ) − Arg(w + y02 )], Im w > 0.
π
Transforming back to the z-plane, since w = z2 , the temperature function in
the first quadrant in the z-plane is given by
U 
T (z) = Arg(z2 − x02 ) − Arg(z2 + y02 ) , Re z > 0 and Im z > 0.
π
This is the solution to the given problem since each term in the above solution
is harmonic, and T (z) satisfies the prescribed boundary conditions along the
boundary of the first quadrant.

Remark The above solution technique requires the following invariance prop-
erty of the Laplace equation: if φ(x, y) is harmonic in a certain domain D in the
z-plane, with z = x + iy, and if w = f (z) is an analytic function which maps D
onto a domain D  in the w-plane, where w = u + iv, then φ(u, v) is harmonic
in D . This invariance property is discussed in detail in Subsection 8.1.1.

3.4 Inverse trigonometric and hyperbolic functions

Since the complex trigonometric and hyperbolic functions are defined in terms
of the complex exponentials, we would expect that the inverses of these func-
tions will be expressible in terms of the complex logarithms. Similarly to their
real counterparts, the complex inverse trigonometric and hyperbolic functions
are multi-valued.
First, we consider the inverse sine function and write

w = sin−1 z, (3.4.1)

or equivalently,

eiw − e−iw
z = sin w = . (3.4.2)
2i
Equation (3.4.2) can be considered as a quadratic equation in eiw , that is,

e2iw − 2izeiw − 1 = 0.

Solving this gives

1
eiw = iz + (1 − z2 ) 2 .
112 Exponential, Logarithmic and Trigonometric Functions

Taking the logarithm of both sides of the above equation, it then follows that
1  
w = sin−1 z = log iz + (1 − z2 ) 2 .
1
(3.4.3)
i
1
When z = ±1, the quantity (1 − z2 ) 2 has two possible values. For each
value, the logarithm generates infinitely many values. Therefore, sin−1 z has
two sets of an infinite number of values. For example, consider
 √ 
−1 1 1 i 3
sin = log ±
2 i 2 2
'  (  

1 π 1 5π
= ln 1 + i + 2kπ or ln 1 + i + 2kπ
i 6 i 6
π 5π
= + 2kπ or + 2kπ, k is any integer.
6 6
In a similar manner, we can derive the following formulas for the other
inverse trigonometric and hyperbolic functions:
1
cos−1 z =
1
log(z + (z2 − 1) 2 ), (3.4.4a)
i
−1 1 1 + iz −1 1 z+i
tan z = log cot
, z= log , (3.4.4b)
2i 1 − iz 2i z−i
sinh−1 z = log(z + (1 + z2 ) 2 ), cosh−1 z = log(z + (z2 − 1) 2 ),
1 1
(3.4.4c)
1 1+z 1 z+1
tanh−1 z = log , coth−1 z = log , etc. (3.4.4d)
2 1−z 2 z−1
The derivative formulas for the inverse trigonometric functions are
d 1 d 1
sin−1 z = , cos−1 z = − 1 ,
2 21
dz (1 − z ) dz (1 − z2 ) 2
d 1
tan−1 z = , and so forth. (3.4.5)
dz 1 + z2

Example 3.4.1 Find explicitly all values of tan−1 (i − 2).

Solution Applying formula (3.4.4b), we have

1 1 + i(i − 2)
tan−1 (i − 2) = log
2i 1 − i(i − 2)
1 −2i
= log
2i 2(1 + i)
3π i
= nπ − + ln 2, n = 0, ±1, ±2, . . . .
8 4
 3.4 Inverse trigonometric and hyperbolic functions 113

Example 3.4.2 Show that if θ is real and sin θ sin φ = 1, then

   
1  θ
φ = n+ π ± i ln tan ,
2 2
where n is an integer, even or odd, according to whether sin θ > 0 or sin θ < 0.

Solution Let φ = α + iβ. The given equation then takes the form

sin(α + iβ) = cosec θ.

By equating the real and imaginary parts of sin(α + iβ) to those of cosec θ , we
obtain

sin α cosh β = cosec θ and cos α sinh β = 0.

The second equation implies either cos α = 0 or sinh β = 0. However, when

sinh β = 0, we have cosh β = 1. Substituting back into the first equation, we
get

sin α = cosec θ.
 
There will be no solution for real
 α except
 when θ = n + 1
2
π , where n is any
integer. In this case, φ = θ = n + 2 π, which is just trivial.
1
 Excluding
 this
trivial case, we are then left with cos α = 0. This gives α = n + 12 π , where
n is any integer, and β is determined from

cosec θ cosec θ if n is even
cosh β =   = .
1 −cosec θ if n is odd
sin n + π
2
Since cosh β > 0 for all β, we have to choose n to be even when sin θ > 0 or n
to be odd when sin θ < 0. The above results can be represented in the following
succinct form:
eβ + e−β
cosh β = = cosec(θ + nπ ).
2
The above equation can be considered as a quadratic equation in eβ . On solving
for eβ , we obtain
1 ± cos(θ + nπ )
eβ =
sin(θ + nπ )
θ + nπ θ + nπ
= tan or cot .
2 2
Combining these results, we observe that
114 Exponential, Logarithmic and Trigonometric Functions

(i) when sin θ > 0, n has to be even (write n = 2k, k is an integer), so

   
θ + nπ θ  θ 
tan = tan 
+ kπ =  tan ;
2 2 2
(ii) when sin θ < 0, n has to be odd (write n = 2k + 1, k is an integer), so
   
θ + nπ θ +π θ  θ 
tan = tan + kπ = −cot =  cot .
2 2 2 2
By following a similar argument, we obtain
 
θ + nπ  cot θ  when sin θ > 0
cot =  2
 .
2  tan θ  when sin θ < 0
2

Therefore, the possible solutions to β are given by

     
 θ  θ  θ
β = ln tan  or β = ln cot  = − ln tan .
2 2 2
Summing up the results, the solutions for φ can be represented by
   
1  θ
φ = α + iβ = n + π ± i ln tan ,
2 2
where n is an integer, even or odd according to sin θ > 0 or sin θ < 0, respec-
tively.
As a numerical example, consider the solution to
1
sin φ = 2 = .
sin π6
Here θ = π
6
and sin θ > 0, so we obtain
  
1 π
φ = 2k + π ± i ln tan , k is any integer.
2 12
Similarly, the solution to
sin φ = −2
is given by
  
1 π
φ = 2k − π ± i ln tan , k is any integer.
2 12
Referring to
 theπ mapping
 of w = sin z (see Figure 3.2a), the pair of points
z = 2 ± i ln tan 12 in the z-plane are mapped to the same point w = 2. This
π

is consistent with the mapping property that the vertical line Re z = π2 in the
z-plane is mapped onto the semi-infinite line Re w ≥ 1 and Im w = 0 in the
w-plane.
 3.5 Generalized exponential, logarithmic and power functions 115

In general, given w = α0 , where α0 > 1, we find θ0 , 0 < θ0 < π2 , such that

sin θ0 = α10 . The pair of preimages of w = α0 within the strip − π2 ≤ Re z ≤ π2
under the mapping w = sin z are
 
π θ0
z = ± i ln tan .
2 2
Similarly, the pair of preimages of w = −α0 within the same strip under the
same mapping are
 
π θ0
z = − ± i ln tan .
2 2

3.5 Generalized exponential, logarithmic and power functions

In this section, we would like to give definitions of the following quantities:
i z , zi , logi z, etc. First, consider the generalized exponential function
f (z) = a z , (3.5.1)
where a is complex in general and z = x + iy is a complex variable. Suppose
we write
log a = ln |a| + i(Arg a + 2kπ), k = 0, ±1, ±2, . . . ;
since a = elog a , we then have
a z = e[ln |a| + i(Arg a + 2kπ)] (x+iy)
= e[x ln |a| − y(Arg a + 2kπ)] ei[y ln |a| + x(Arg a + 2kπ)]
= |a|x e−y(Arg a + 2kπ) [cos(y ln |a| + x(Arg a + 2kπ))
+ i sin(y ln |a| + x(Arg a + 2kπ))], k = 0, ±1, ±2, . . . . (3.5.2)
The function is not multi-valued, though apparently we have infinitely many
choices for k. The value k is related to the choice of the value for the argument
of a and it has nothing to do with the complex variable z. Each choice of k
corresponds to a separate function but not a particular branch of a multi-valued
function. We usually take k = 0 as convention so that when a is real, the
expression for a z in eq. (3.5.2) reduces to
a z = e(ln a)z . (3.5.3)
The inverse of the generalized exponential function a z is the logarithm
function to the complex base a, denoted by loga z. Suppose we write w =
loga z; then z = a w . Similarly, we choose a branch of a w as above, that is, fix
the value of k in log a, where log a = Log a + 2kπ i. Once the branch is fixed
116 Exponential, Logarithmic and Trigonometric Functions

(say k = k0 for some chosen k0 ), then z = eλw , where λ = Log a + 2k0 π i.

This leads to
log z log z
w = loga z = = . (3.5.4)
λ Log a + 2k0 π i
In the denominator, λ is one of the infinitely many possible values of log a.
Note that loga z remains a multi-valued function, like other logarithm functions
with a real base. For example, suppose we choose k0 to be 2. Then

Log(1 − i) 1
ln 2 − π4 i
Log1+i (1 − i) =  = 2
.
1
2
ln 2 + i 4π + π4 1
2
ln 2 + 17π
4
i
Lastly, we consider the generalized power function

f (z) = za , (3.5.5)

where a is complex in general and

z = x + iy = reiθ = |z|ei(Arg z + 2kπ)

is a complex variable. Consider the following cases:

(i) When a = n, n is an integer,

zn = |z|n einArg z .

(ii) When a is rational, a = m/n where m, n are irreducible integers, we

have
m m
z n = e n log z
m m m
= |z| n ei n Arg z
e2k n πi , k = 0, 1, . . . , n − 1.
2k mn πi
The factor e takes on n different values for k = 0, 1, . . . , n − 1,
then repeats itself with period n if k continues to increase through
the integers. This is precisely the same result as that of De Moivre’s
theorem (see Subsection 1.2.1). The power function has n different
branches, corresponding to the different values of k.
(iii) When a = α + iβ,

za = e(α + iβ) [ln |z| + i(Arg z + 2kπ)]

= eα ln |z|−β(Arg z + 2kπ) ei[β ln |z| + α(Arg z + 2kπ)]
= |z|α e−β(Arg z + 2kπ) [cos(β ln |z| + α(Arg z + 2kπ))
+ i sin(β ln |z| + α(Arg z + 2kπ))], k = 0, ±1, ±2, . . . .

In this case, za has infinitely many branches.

 3.5 Generalized exponential, logarithmic and power functions 117

Example 3.5.1 Find the principal value of each of the following complex
quantities:

(a) (1 − i)1+i ; (b) 33−i ; (c) 22i .

Solution
√
2− π4 i)
(a) Principal value of (1 − i)1+i = e(1+i)Log(1−i) = e(1+i)(ln
 √   √ 
ln 2+ π4 +i ln 2− π4
=e
√ π  √ 
= 2e 4 cos ln 2 − π4
 √ 
+ i sin ln 2 − π4 .

(b) Principal value of 33−i = e(3−i)Log 3 = e3 ln 3−i ln 3

= 27 [cos(ln 3) − i sin(ln 3)].

(c) Principal value of 22i = e2i ln 2 = cos(ln 4) + i sin(ln 4).

Example 3.5.2 The power function

w = f (z) = [z(z − 1)(z − 2)]1/2

√ √
be either − 6i or 6i. Suppose the
has two branches. Show that f (−1) can√
branch that corresponds to f (−1) = − 6i is chosen; find the value of the
function at z = i.

Solution The given power function can be expressed as

w = f (z) = |z(z − 1)(z − 2)|1/2 ei[Arg z + Arg(z−1) + Arg(z−2)]/2 eikπ , k = 0, 1,

where the two possible values of k correspond to the two branches of the
double-valued power function. Note that at z = −1,
√
Argz = Arg(z − 1) = Arg(z − 2) = π and |z(z − 1)(z − 2)|1/2 = 6,
√ i3π/2 √ √ √
so f (−1) can be either
√ 6e = − 6i or 6ei3π/2 eiπ = 6i. The branch
that gives f (−1) = − 6i corresponds to k = 0. With the choice of that branch,
118 Exponential, Logarithmic and Trigonometric Functions

we have
f (i) = |i(i − 1)(i − 2)|1/2 ei[Arg i + Arg(i−1) + Arg(i−2)]/2
√ √ −1 1
= ( 2 5)1/2 ei ( 2 + 4 +π −tan 2 )/2
π 3π

−1 1
= (10)1/4 eiπ e 2 ( 4 −tan 2 )
i π

i −1 −1 1
= −(10)1/4 e 2 (tan 1−tan 2 )
 
1− 21
i
tan−1
= −(10)1/4 e 1+ 21
2

i −1 1
= −(10)1/4 e 2 tan 3 .

3.6 Branch points, branch cuts and Riemann surfaces

We have discussed the multi-valuedness of power functions and logarithmic
functions in previous sections. The mapping property of a complex function
would be much easier to visualize if the mapping were one-to-one. A function
f (z) is said to be univalent in a domain D if it is one-to-one and analytic in D.
Correspondingly, D is called a domain of univalence for f (z). For example,
the function w = zn , n is a positive integer, is univalent in the sectoral domain
{z : θ0 < arg z < θ0 + 2π n
, θ0 is any value}.
The inverse of the above function is w = z1/n , known to be multi-valued.
Given any value of z, there are n distinct nth roots of z. Each root corresponds
to a specific branch of the multi-valued function. Every branch of a multi-
valued function has the same domain. How do we separate the branches so that
every branch has its own copy of the domain? Each branch of the function now
becomes single-valued in its own domain.
In this section, we would like to discuss an ingenious construction, known
as the Riemann surfaces, in order to achieve the above objective. A Riemann
surface consists of overlapping sheets (the number of sheets can be finite or
infinite) and these sheets are connected by branch cuts. The end points of
a branch cut are called the branch points. The concepts of branch points,
branch cuts and Riemann surfaces for multi-valued functions are exemplified
below.
First, consider the simple example w = z1/2 . For a given z0 = r0 eiθ0 (0 ≤
√ √ √
θ0 < 2π), the two roots are r0 eiθ0 /2 and r0 ei(θ0 +2π )/2 = − r0 eiθ0 /2 . Sup-
pose we start from r0 eiθ0 in the z-plane and follow the path of z which traverses
one complete loop around the origin and back to the starting point; the argu-
ment of z increases from θ0 to θ0 + 2π. The corresponding image point in
√ √ √
the w-plane changes from r0 eiθ0 /2 to r0 ei(θ0 +2π )/2 = − r0 eiθ0 /2 . Note that
the two image points lie in different branches of the double-valued function
w = z1/2 . Suppose the path of z traverses once more around z = 0 in the
 3.6 Branch points, branch cuts and Riemann surfaces 119

(top) (bottom)
z1 - plane z2 - plane

y y

arg z increases arg z increases

from 0 to 2 from 2 to 4

x x

Figure 3.5. The Riemann surface of w = z1/2 consists of two sheets: z1 -plane (top) and
z2 -plane (bottom). The branch points are z = 0 and z = ∞, and the branch cut is taken
to be along the positive real axis on each sheet. The path of z traversing a closed circuit
around the origin moves from one sheet to the other sheet.

anticlockwise direction and back to the same starting point; the argument of z
now becomes θ0 + 4π and leads w = z1/2 to return to the value r0 eiθ0 /2 . Note
that the value w = z1/2 remains unchanged if the complete circuit in the z-plane
does not include the point z = 0.
Suppose any closed loop around a point always carries every branch of a
given multi-valued function into another branch. Then the enclosed point is
called a branch point of the function. The branch point is said to be of order
n − 1 (a finite positive integer) if n complete circuits in the same direction
around the point carry every branch of the function back to itself. For example,
z = 0 is a branch point of order one of the function w = z1/2 .
Is the point at infinity a branch point of w = z1/2 ? To answer this question,
we need to examine the case of a complete loop traversing around the complex
infinity. Recall that the complex infinity in the extended z-plane corresponds
to the north pole on the Riemann sphere, and so a small closed curve around
the north pole on the Riemann sphere corresponds to a large closed curve in
the complex plane. When the path of z moves around a large closed loop in the
complex plane, w = z1/2 is seen to move to a new branch. Hence, z = ∞ is
also a branch point of w = z1/2 .
A multi-valued function may be regarded as single-valued if we suitably gen-
eralize its domain of definition. For the function w = z1/2 , suppose we take two
copies of the z-plane superimposed upon each other. The argument of z on the
top sheet (called it the z1 -plane) ranges from 0 to 2π , while that on the bottom
sheet (called it the z2 -plane) ranges from 2π to 4π (see Figure 3.5). In order
that the path of z that traverses a complete circuit around the origin moves
120 Exponential, Logarithmic and Trigonometric Functions

0
y

S0

S1
S2

x
Figure 3.6. Sketch of the three-sheeted structure of the Riemann surface for the multi-
valued function w = z1/3 . The bottom sheet S2 is joined back to the top sheet S0 . The
branch cuts are chosen to be along the positive real axis.

from one sheet to the other sheet, we place a cut along the positive real
axis of each sheet. This cut is called the branch cut, and it links the two
branches of the multi-valued function. Naturally, the two ends of this branch cut
are the two branch points of the function. The lower edge along the branch cut
of the top sheet (the ray defined by arg z = 2π − ) is joined to the upper edge
along the branch cut of the bottom sheet (the ray defined by arg z = 2π + ).
Likewise, the lower edge of the cut on the bottom sheet (the ray defined by arg
z = 4π − ) is joined to the upper edge of the cut on the top sheet (the ray defined
by arg z = 0+ ). These two sheets together with the branch cuts constitute the
Riemann surface of the double-valued function w = z1/2 .
The construction of the Riemann surface can be easily generalized to w =
1/n
z , where n is a positive integer. There will be n sheets in the corresponding
Riemann surface. The branch points are z = 0 and z = ∞, and the order is
equal to n − 1. A sketch of the three-sheeted structure of the Riemann surface
for w = z1/3 is illustrated in Figure 3.6.

Logarithmic function
The logarithmic function w = log z has infinitely many values for each z, so
we expect its Riemann surface to consist of infinitely many sheets. The sheets
are joined together in a similar manner to those of w = z1/2 , that is, the lower
 3.6 Branch points, branch cuts and Riemann surfaces 121

edge of the cut on the zk -sheet is joined to the upper edge of the cut on the
zk+1 -sheet. However, it is not necessary to join the first and the last sheet, unlike
those of w = z1/n . Indeed it is impossible to define which sheet is the first and
which is the last when the number of sheets is infinite. The branch points can
be deduced to be z = 0 and z = ∞, using the same technique of observing the
change of branch when z moves around a closed loop containing the branch
point.
Where do we place the branch cut? Supposing the principal branch of the
logarithmic function is chosen such that Arg z is lying between −π and π ,
the choice of the branch cut is then taken along the negative real axis of each
Riemann sheet.

Example 3.6.1 Consider the multi-valued function

w = f (z) = (z2 − 1)1/2 .

Find the two branch points of the function. Describe the possible branch cut
and the Riemann surface of the function.

Solution We let

z − 1 = r1 eiθ1 and z + 1 = r2 eiθ2

so that
√ iθ1 iθ2
w = [r1 r2 ei(θ1 +θ2 ) ] 2 =
1
r1 r2 e 2 e 2 .

Suppose we start with a particular point z0 such that θ1 = α1 and θ2 = α2 .

When z moves in the anticlockwise sense once around z = 1 but not z = −1,
the value of θ1 increases from α1 to α1 + 2π but θ2 remains the same value.
The new value of w becomes
√ i(α1 +2π ) iα2 √ iα1 iα2
w = r1 r2 e 2 e 2 = − r1 r2 e 2 e 2 ,

which is different from the original value. This signifies a change in branch
and therefore z = 1 is a branch point of f (z). Similarly, if we consider a closed
path around z = −1 but not z = 1, θ1 remains the same value but θ2 increases
from α2 to α2 + 2π . This causes a change in the value of w. Therefore, z = −1
is the other branch point of f (z). If z moves in the anticlockwise sense around
a sufficiently large circuit that includes both branch points z = ±1, then θ1
increases from α1 to α1 + 2π and θ2 also increases from α2 to α2 + 2π . Now,
the value of w becomes
√ i(α1 +2π ) i(α2 +2π ) √ iα1 iα2
w = r1 r2 e 2 e 2 = r1 r2 e 2 e 2 ,
122 Exponential, Logarithmic and Trigonometric Functions

Figure 3.7. One choice of the branch cut of the multi-valued function f (z) = (z2 − 1)1/2
is along the line segment joining z = −1 and z = 1. A new branch of the function is
encountered when the path of z traverses across the branch cut. Another possible choice
of the branch cut is the union of the two line segments: z = −1 to z = ∞ along the
negative real axis and z = 1 to z = ∞ along the positive real axis. In this case, one may
visualize the two branch points as being joined by a line segment going through the
complex infinity.

which is the same as the original value. There is no change in branch, so z = ∞

is not a branch point.
The branch cut can be taken to be either (i) a cut between the branch points
z = 1 and z = −1, or (ii) two cuts along the real axis, one from z = 1 to z = ∞
and the other from z = −1 to z = ∞ (see Figure 3.7). The single-valuedness of
the function is ensured when the path of z does not cross these branch cuts. The
Riemann surface of the function consists of two sheets superimposed on each
other and they are joined along the branch cuts. The end points of the cuts are
z = −1 and z = 1. On making one complete loop around either z = −1 or z =
1 but not both, we start on one sheet and wind up on another sheet of the Riemann
surface.

Example 3.6.2 The inverse cosine function is related to the logarithm function
by [see eq. (3.4.4a)]
1
w = cos−1 z = log(z + (z2 − 1)1/2 ).
i
Find the branch points and branch cuts of this multi-valued function.

Solution Since z = −1 and z = 1 are known to be the branch points of

1
the double-valued function (z2 − 1) /2 , the same is true for the inverse cosine
 3.6 Branch points, branch cuts and Riemann surfaces 123

function. Though z = ∞ is not a branch point of (z2 − 1)1/2 , the argument

of z + (z2 − 1)1/2 increases by 2π when z traverses in a simple closed curve
encircling both branch points z = −1 and z = 1. Therefore, z = ∞ is seen to
be a branch point of 1i log(z + (z2 − 1)1/2 ). Since z = ∞ is a branch point of
cos−1 z, the branch cuts are chosen to be semi-infinite line segments emanating
from the two branch points z = −1 and z = 1 and extending to infinity.
A convenient choice of the branch cuts would be the two line segments
along the real axis: −∞ < x < −1 and 1 < x < ∞. Let D be the domain of
definition of the branch with respect to these two branch cuts. Each branch of
the inverse cosine function maps D in the z-plane onto a vertical infinite strip
of width 2π in the w-plane. The single-valued branch function takes the upper
half-plane of D onto a vertical infinite strip of width π and the lower half-plane
of D onto a neighboring vertical infinite strip of the same width.

3.6.1 Joukowski mapping

The complex function
 
1 1
w = f (z) = z+ , z = 0, (3.6.1)
2 z

is called the Joukowski function.  1 This function is analytic everywhere except

at z = 0. Since w = f (z) = f z , the pair of points z1 and z2 in the z-plane
which satisfy z1 z2 = 1 are mapped onto the same image point in the w-plane.
Hence, the Joukowski mapping is univalent in a domain D if and only if
there are no two points z1 and z2 in D which satisfy z1 z2 = 1. For example,
the Joukowski mapping is univalent in the following domains: (i) |z| < 1,
(ii) |z| > 1, (iii) Im z > 0, (iv) Im z < 0.
The Joukowski mapping is a two-to-one mapping. To examine its mapping
properties, we write z = reiθ and w = u + iv, and consider
 
1 1
w = u + iv = reiθ + e−iθ
2 r
   
1 1 i 1
= r+ cos θ + r− sin θ,
2 r 2 r

so that
   
1 1 1 1
u= r+ cos θ and v = r− sin θ.
2 r 2 r
124 Exponential, Logarithmic and Trigonometric Functions

1
r0 r0

(0, 21 ( r1
0
)
− r0 )

u
−1 1

( 21 ( r 0 + 1),0
r0 )

Figure 3.8. The Joukowski mapping takes concentric circles centered at the origin in
the z-plane onto cofocal ellipses in the w-plane. Any two circles with the product of their
radii equal to one are mapped onto the same ellipse. In particular, the unit circle |z| = 1
is mapped onto the line segment [−1, 1] on the u-axis (considered as a degenerated
ellipse).

The circle |z| = r0 < 1 is mapped onto the ellipse defined by (see Figure 3.8)
   
1 1 1 1
u= r0 + cos θ and v = r0 − sin θ.
2 r0 2 r0
 
The semi-major and semi-minor axes of the ellipse are seen to be 12 r0 + r10
 
and 12 r10 − r0 , respectively. The ellipse degenerates into the line segment
[−1, 1] on the u-axis when r0 → 1.
 3.6 Branch points, branch cuts and Riemann surfaces 125

The domain {z : |z| < 1} in the z-plane is mapped onto the whole w-plane
minus the line segment
 1  [−1, 1] along the real axis. By virtue of the reciprocity
property f (z) = f z , the domain exterior to the unit circle in the z-plane is
also mapped onto the whole w-plane (see Figure 3.8).
The inverse of the Joukowski function is found to be

w = z + (z2 − 1)1/2 , (3.6.2)

which is a double-valued function. Following similar techniques used in Exam-

ple 3.6.1, one can check that z = 1 and z = −1 are branch points of order one,
but z = ∞ is not a branch point of the Riemann surface of the inverse Joukowski
function.
Suppose we choose the domain D to be the whole z-plane minus the line
segment [−1, 1] on the x-axis. Then the two branches of the inverse Joukowski
function are given by

√
w1 = z + z2 − 1 with w1 (i) = (1 + 2)i, (3.6.3a)

√
w2 = z − z2 − 1 with w2 (i) = (1 − 2)i. (3.6.3b)

They map the domain D onto the exterior and interior of the unit circle |w| = 1
in the w-plane, respectively (see also Problem 3.37).

Example 3.6.3 Find an analytic function w = f (z) which maps the interior
of the unit circle |z| < 1 minus the segments (−1, −1 + h] and [1 − h, 1),
where 0 < h < 1, on the real axis onto the interior of the unit circle
|w| < 1.

Solution Recall that the Joukowski mapping

 
1 1
w1 = z+
2 z
takes the domain {z : |z| < 1} to the whole w1 -plane minus the segment [−1, 1]
along the real axis. Also, the segments (−1, −1 + h] and [1 − h, 1) along the
 −1) and (1, δ] along
real axis in the z-plane are mapped onto the segments [−δ,
the real axis in the w1 -plane, where δ = 12 1 − h + 1−h1
. Note that [−δ, δ] is
the union of the three segments: [−δ, −1), [−1, 1] and (1, δ]. Hence, the above
Joukowski mapping takes the given domain in the z-plane onto the whole
w1 -plane minus the segment [−δ, δ] along the real axis. Suppose we take
w1
w2 = ;
δ
126 Exponential, Logarithmic and Trigonometric Functions

then the segment [−δ, δ] on the real axis in the w1 -plane becomes [−1, 1]
on the real axis in the w2 -plane. By virtue of eq. (3.6.3b), the inverse of the
Joukowski mapping

w = w2 − w22 − 1

takes the whole w2 -plane minus the segment [−1, 1] on the real axis onto the
interior of the unit circle |w| < 1. Combining the three transformations, the
analytic function which effects the mapping is found to be

 2
z + 1z  z + 1z
w= − − 1.
1 − h + 1−h1
1 − h + 1−h1

3.7 Problems
3.1. Verify the following identities:

(a) cosh2 z − sinh2 z = 1; (b) cosh2 z + sinh2 z = cosh 2z;

(c) sinh(z1 + z2 ) = sinh z1 cosh z2 + cosh z1 sinh z2 .

3.2. Show that

(a) tan iz = i tanh z; (b) cot iz = −i coth z;

sinh 2x − i sin 2y
(c) coth z = , z = x + iy.
cosh 2x − cos 2y
3.3. Show that if cos(z + w) = cos z for any z in the complex plane, then
w = 2kπ, where k is any integer.
3.4. Find the general solution for each of the following equations:

(a) ez = 2i; (b) sin z = cosh 3.

3.5. Show that

(a) |ez | is bounded when Re z ≤ α;

(b) | cos(x + iy)| is unbounded as y → ∞.

3.6. For each of the functions

(a) sin z, (b) tan z, (c) coth z,

find the set of points z such that the function assumes
(i) real value,
(ii) purely imaginary value.
 3.7 Problems 127

3.7. Let z = x + iy; express the following modulus quantities in terms of x

and y:

(a) | tan z|; (b) | tanh z|.

3.8. For 0 < |z| < 1, show that

|z| 7|z|
< |ez − 1| < .
4 4

Hint: Consider
 
|z| |z|2
|ez − 1| ≤ |z| 1 + + + ···
2! 3!
 

1 1 1
≤ |z| 1 + 1 + + 2 + ··· since 0 < |z| < 1.
2 3 3
In a similar manner, we obtain
 

1 1
|ez − 1| ≥ |z| 1 − + + ··· .
2! 3!

3.9. Prove that

(a) |ez − 1| ≤ e|z| − 1 ≤ |z| e|z| ;

(b) |Im z| ≤ | sin z| ≤ e|Im z| .

Hint: To prove the left-hand side inequality of part (a), consider

 
−z/2 z/2 
 z 
|e − 1| = |e ||e − e
z z/2 z/2
| = 2|e | sinh 
2
|z|
e|z| − 1 = e|z|/2 (e|z|/2 − e−|z|/2 ) = 2e|z|/2 sinh .
2
In addition, try to establish
 
 
 sinh z  ≤ sinh |z| .
 2 2

3.10. Find a complex number z that satisfies

| sin z| > 1 and | cos z| > 1.

3.11. Show that

1
| log z| = | log(1 − z)| when Re z = .
2
128 Exponential, Logarithmic and Trigonometric Functions

3.12. Suppose |z| ≤ R; show that

| sin z| ≤ cosh R and | cos z| ≤ cosh R.

3.13. Show that

sin 2α + sinh 2β
(1 + i) cot(α + iβ) + (1 − i) cot(α − iβ) = 2 .
cosh 2β − cos 2α
3.14. Let z = x + iy; show that
 z n   z n

lim  1 +  = ex and lim Arg 1 + = y.
n→∞ n n→∞ n

tan−1 y
Hint: lim y
n
=1
n→∞
n

3.15. Find all roots of each of the following equations:

(a) sin z + cos z = 2; (b) sin z − cos z = 3; (c) sin z − cos z = i;

(d) cosh z − sinh z = 1; (e) sinh z − cosh z = 2i;
(f) 2 cosh z + sinh z = i; (g) cos z = cosh z;
(h) sin z = i sinh z; (i) cos z = i sinh 2z.

3.16. (a) Suppose z moves along the parabola y = x 2 ; find

lim ez .
z→∞

(b) Suppose z moves along the imaginary axis; describe the behavior of
the following functions: (i) sin z, (ii) cosh z.
(c) Show that | sin z| is bounded when Im z = α.
3.17. Evaluate the following quantities:

(a) Log(2 − 3i); (b) log(−2 + 3i);

(c) cos(2+ i); (d) coth(2 + i);
πi
(e) tanh ln 3 + ; (f) cos−1 i; (g) sin−1 i; (h) tan−1 i.
4
3.18. For each of the following functions, find the domain of analyticity:
ez ez 1
(a) ; (b) ; (c) .
z cos z sin z + cos z cosh z + ez
3.19. Show that
 
ia − 1 ib
= exp(−2b cot−1 a), where a and b are real.
ia + 1
 3.7 Problems 129

3.20. Show that

z 1 z+a
coth−1 = log
a 2 z−a
 
1 (x + a)2 + y 2 i −1 2ay
= ln + tan + 2kπ ,
4 (x − a)2 + y 2 2 a2 − x 2 − y 2
where a is positive real and k is any integer.
3.21. Find the image of the infinite horizontal strip {z: −1 < Im z ≤ 1} under
the mapping w = f (z) = e2πz . Find the corresponding inverse function
for f (z) whose image is the above horizontal strip.
3.22. Show that the mapping function
w = cosh z
maps the semi-infinite strip {z = x + iy : x ≥ 0 and 0 ≤ y ≤ π2 } in the
z-plane onto the first quadrant of the w-plane.
3.23. Detect any fault in the following argument: using the fact that (−z)2 = z2 ,
we obtain
2 Log(−z) = 2 Log z,
and so
Log(−z) = Log z, for all z in C\{0}.
3.24. Consider the temperature function
T (z) = ln |z − z1 | − ln |z − z2 |,
where z1 and z2 are distinct points in the complex plane. Explain why
T (z) is harmonic everywhere except at z1 and z2 . Where are the locations
of the heat source and heat sink in the temperature field defined by this
temperature function? Find the set of points at which the temperature
value equals zero.
3.25. Find the steady state temperature distribution in the upper half-plane,
where the boundary temperature values along the x-axis are given by



U1 x < x1



 2U x1 < x < x2
T (x, 0) = .
.. .. .
 .



 U xn−1 < x < xn
 n
Un+1 xn < x
3.26. Find the steady state temperature distribution T (z) in the semi-infinite
vertical strip {z : − π2 < Re z < π2 and Im z > 0}, where the boundary
130 Exponential, Logarithmic and Trigonometric Functions

temperature values are

 π   
T 2 + iy = T − π2 + iy = 0, y>0
.
T (x) = U, − π2 < x < π
2
, U is real positive

Hint: The mapping w = sin z maps the given semi-infinite strip in the
z-plane onto the upper half w-plane.

3.27. Consider the multi-valued function

w = f (z) = z1/3 , z ∈ C\[0, ∞).

5
Supposing we choose the branch such that f (i) = e 6 πi , compute f (−2).
3.28. Consider the multi-valued function

f (z) = [(1 − z)3 z]1/4 .

Show that z = 0 and z = 1 are branch points of the function, and find
their order. Is z = ∞ a branch point? Describe the Riemann surface of
the function.
3.29. Consider the logarithmic function

w = log(z − α), α is complex.

Show that z = α and z = ∞ are branch points of the function. Explain

why any simple curve that starts at z = α and ends at ∞ can be a branch
cut of the Riemann surface of the function.
3.30. The inverse tangent function can be expressed as
1 i−z
w = tan−1 z = log .
2i i+z
Examine whether the points z = i, z = −i and z = ∞ are branch points
of the function. Compute tan−1 i, tan−1 (−i) and tan−1 ∞.
3.31. Recall that
sinh z ez − e−z
tanh z = = z .
cosh z e + e−z
If we choose the principal branch of tanh−1 z to be that for which
tanh−1 0 = 0, prove that
 
−1 1 1+z
tanh z = Log .
2 1−z
Find the derivative of tanh−1 z.
 3.7 Problems 131

3.32. Consider the multi-valued function

f (z) = log(1 − z2 )

where the domain D is defined to be the whole complex plane minus the
following three branch cuts (see the figure):
(i) line segment (including the end points) joining −1 and i,
(ii) line segment (including the end points) joining 1 and i,
(iii) semi-infinite line segment: x = 0, y ≥ 1.
(a) Show that the function f can be separated into single-valued
branches in the above domain D.
(b) Supposing we choose the branch where f (0) = 0, find the cor-
responding value of this branch of the function at z = 2.
(c) Is the choice of the point z = i (starting point of the semi-infinite
branch cut) unique?

x
1 1

3.33. Is [z(z + 1)]1/2 the same function as z1/2 (z + 1)1/2 ?

3.34. Consider the many-to-one mapping

f (z) = (z − α)(z − β),

and let  be any line through the point (α + β)/2 which divides the
z-plane into two half-planes. Show that each of the open half-planes
determined by  is a domain of univalence for f (z).
3.35. Show that the necessary and sufficient condition for the function

f (z) = eαz , α = a + ib = 0,
132 Exponential, Logarithmic and Trigonometric Functions
* +
to be single-valued inside the infinite strip z : − π2 < Im z < π2 is given
by
a 2 + b2 ≤ 2|a|.
3.36. Consider the multi-valued function
w = f (z) = [z(1 − z)3 ]1/4 .
Supposing we choose the branch where
f (−1) = 81/4 eiπ/4 eiπ/2 = 21/4 (−1 + i),
show that the value of w corresponding to this branch along the segment
(0, 1) on the x-axis is given by


f (x) = i 4 x(1 − x)3 , x ∈ (0, 1).
3.37. Consider the mapping represented by the inverse Joukowski function
w = f (z) = z + (z2 − 1)1/2 .
Find the preimage in the z-plane of the ray Arg w = θ0 in the w-plane.
Suppose we choose the domain D to be the whole z-plane minus the two
line segments (−∞, −1) and (1, ∞) on the x-axis. Show that the two
branches of the inverse Joukowski function



w1 = z + z2 − 1 and w2 = z − z2 − 1
map the domain D onto the upper and lower half w-plane, respectively.
 4
Complex Integration

The methods of integration of complex functions and their underlying theories

are discussed in this chapter. The cornerstones in complex integration are
the Cauchy–Goursat theorem and the Cauchy integral formula. A fascinating
result deduced from the Cauchy integral formula is that if a complex function
is analytic at a point, then its derivatives of all orders exist and these derivatives
are analytic at that point. Other important theorems include Gauss’ mean value
theorem, Liouville’s theorem, and the maximum modulus theorem.
Many properties of the complex integrals are very similar to those of the
real line integrals. For example, when the integrand satisfies certain conditions,
the integral can be computed by finding the primitive function of the integrand
and evaluating the primitive function at the two end points of the integration
path. However, there are other properties that are unique to integration in the
complex plane.
In the last section, we link the study of conservative fields in physics with the
mathematical theory of analytic functions and complex integration. The pro-
totype conservative fields considered include the gravitational potential fields,
electrostatic fields and potential fluid flow fields. The potential functions in
these physical models are governed by the Laplace equation, and so their solu-
tions are harmonic functions. Complex variables techniques are seen to be
effective analytical tools for solving these physical models.

4.1 Formulations of complex integration

The integration of complex functions in the complex plane is seen to resemble
closely the integration of real functions in the two-dimensional plane. In this
section, we first consider the formulation of the integration of an arbitrary
complex function following a similar approach to that for a real line integral. The

133
134 Complex Integration

celebrated Cauchy integral theory for dealing with complex integrals involving
analytic integrand functions will be discussed in the next section.

4.1.1 Definite integral of a complex-valued function of a real variable

Consider a complex-valued function f (t) of a real variable t:
f (t) = u(t) + iv(t), (4.1.1)
which is assumed to be a piecewise continuous function defined in the closed
interval a ≤ t ≤ b. To integrate f (t) from t = a to t = b, we use the natural
definition
 b  b  b
f (t) dt = u(t) dt + i v(t) dt. (4.1.2)
a a a

Properties of a complex integral with real variable of integration

 b  b  b
1. Re f (t) dt = Re f (t) dt = u(t) dt. (4.1.3a)
a a a

 b  b  b
2. Im f (t) dt = Im f (t) dt = v(t) dt. (4.1.3b)
a a a

 b  b  b
3. [γ1 f1 (t) + γ2 f2 (t)] dt = γ1 f1 (t) dt + γ2 f2 (t) dt, (4.1.3c)
a a a

where γ1 and γ2 are any complex constants.

 b   b
 
4.  f (t) dt  ≤ |f (t)| dt. (4.1.3d)
 
a a

The proofs of the first three properties are obvious. The last property can be
shown using the following argument. We consider
 b   b  b
 
 f (t) dt  = e−iφ f (t) dt = e−iφ f (t) dt,
 
a a a
  b  
b  
where φ = Arg 
f (t) dt . Since  f (t) dt  is real, we deduce that
a a
 b   b  b
 
 f (t) dt  = Re e−iφ f (t) dt = Re [e−iφ f (t)] dt
 
a a a
 b  b
≤ |e−iφ f (t)| dt = |f (t)| dt.
a a
 4.1 Formulations of complex integration 135

Example 4.1.1 Suppose α is real. Show that

|e2απi − 1| ≤ 2π|α|.

Solution Let f (t) = eiαt , where α and t are real. Substituting the function
into eq. (4.1.3d), we obtain
 2π   2π
 
 e dt  ≤
iαt
|eiαt | dt = 2π.

0 0

The left-hand side of the above inequality is equal to

 2π   iαt 2π 
   e   |e2απi − 1|
 =  
  iα   =
iαt
 e dt .
0 0
|α|

Combining the results, we obtain

|e2απi − 1| ≤ 2π|α|, α is real.

4.1.2 Complex integrals as line integrals

Consider a curve C which is a set of points z = (x, y) in the complex plane
defined by

x = x(t), y = y(t), a ≤ t ≤ b,

where x(t) and y(t) are continuous functions of the real parameter t. One may
write

z(t) = x(t) + iy(t) and a ≤ t ≤ b.


Recall that |zz (t)
(t)|
gives the unit tangent vector to the curve at the point t, which is
well defined provided that z (t) = 0 in the open interval a < t < b. The curve
is said to be smooth if z(t) has a continuous derivative in the closed interval
a ≤ t ≤ b and nonzero in the open interval a < t < b. A contour is defined as
a curve consisting of a finite number of smooth curves joined end to end. A
contour is said to be a simple closed contour if only the initial and final values
of z(t) are the same.
Let f (z) be any complex function defined in a domain D in the complex
plane and let C be any contour contained in D with initial point z0 and termi-
nal point z. We divide the contour C into n subarcs by the discrete points
z0 , z1 , z2 , . . . , zn−1 , zn = z arranged consecutively along the direction of
increasing t. Let ζk be an arbitrary point on the subarc zk zk+1 (see Figure 4.1)
136 Complex Integration

Figure 4.1. Subdivision of the contour into n subarcs by the discrete points
z0 , z1 , . . . , zn−1 , zn = z.

and form the sum


n−1
f (ζk )(zk+1 − zk ).
k=0

We write zk = zk+1 − zk . Let λ = max |zk | and take the limit
k


n−1
lim f (ζk ) zk .
λ→0
n→∞ k=0

The above limit is defined to be the contour integral of f (z) along the contour
C. Symbolically, we write
 
n−1
f (z) dz = lim f (ζk ) zk . (4.1.4)
C λ→0
n→∞ k=0

If the above limit exists, then the function f (z) is said to be integrable along
the contour C.
The contour integral defined in eq. (4.1.4) can be related to the integral of a
complex function of a real variable. If we write

dz(t) dx(t) dy(t)

= +i , a ≤ t ≤ b,
dt dt dt
then
  b
dz(t)
f (z) dz = f (z(t)) dt.
C a dt
 4.1 Formulations of complex integration 137

Writing f (z) = u(x, y) + iv(x, y) and dz = dx + i dy, we have

  
f (z) dz = u dx − v dy + i u dy + v dx
C C C
 b

dx(t) dy(t)
= u(x(t), y(t)) − v(x(t), y(t)) dt (4.1.5)
a dt dt
 b

dy(t) dx(t)
+i u(x(t), y(t)) + v(x(t), y(t)) dt.
a dt dt
Since a contour integral can be defined in terms of real line integrals, the usual
properties of real line integrals are carried over to their complex counterparts.
Some of these properties are included below.

(i) f (z) dz is independent of the parametrization of C.
C 
(ii) f (z) dz = − f (z) dz, where −C is the opposite curve of C.
−C C
(iii) The integral of f (z) along a string of contours is equal to the sum of the
integrals of f (z) along each of these contours.

Estimation of the modulus value of a complex integral The upper bound of

the modulus value of a complex integral can be related to the arc length of the
contour C and the maximum value of |f (z)| along C. In fact,
 
 
 f (z) dz ≤ ML, (4.1.6)
 
C

where M is the upper bound of |f (z)| along C and L is the arc length of the
contour C. To show the above modulus inequality, we consider
   b 
   
 f (z) dz =  f (z(t)) dz(t) dt 
   dt 
C a
 b  
 dz(t) 
≤ |f (z(t))|   dt
a dt 
 b  
 dz(t) 
≤ M   dt
a dt 

 b    
dx(t) 2 dy(t) 2
=M + dt = ML.
a dt dt

Example 4.1.2 Evaluate the integral

,
1
dz,
C z − z0
138 Complex Integration

where C is a circle centered at z0 and of any radius. The path is traced out once
in the anticlockwise direction.

Solution The circle C can be parametrized by

z(t) = z0 + reit , 0 ≤ t ≤ 2π,

where r is any positive real number. The contour integral becomes

,  2π  2π
1 1 dz(t) ireit
dz = dt = dt = 2π i.
C z − z0 0 z(t) − z0 dt 0 reit
Interestingly, the value of the integral is independent of the radius of the circle.

Example 4.1.3 Evaluate the integral


1
dz
C z2
along the following two paths:

(a) the straight line segment joining 1 and 2 + i;

(b) the horizontal line from 1 to 2, then the vertical line from 2 to 2 + i.

Solution
(a) The parametric form of the straight line segment joining 1 and 2 + i
is given by z(t) = 1 + (1 + i)t, 0 ≤ t ≤ 1. The contour integral can be
expressed as
 1
1 1 3+i
(1 + i) dt = 1 − = .
0 [1 + (1 + i)t]
2 2+i 5
(b) For the second path, the contour integral can be expressed as
 2  1    
1 1 1 1 1 3+i
dx + i dy = 1 − + − = .
1 x 2
0 (2 + iy) 2 2 2 2 + i 5

Remark Apparently, the value of the integral is independent of the path of

integration and can be found directly from
 2+i 2+i
1 1  3+i
2
dz = −  = ,
1 z z 1 5

where − 1z is a primitive function of 1

z2
.
 4.1 Formulations of complex integration 139

Example 4.1.4 Evaluate the integral


|z|2 dz,
C

where the contour C is

(a) the line segment with initial point −1 and final point i;
(b) the arc of the unit circle |z| = 1 traversed in the clockwise direction with
initial point −1 and final point i.

Do the two results agree?

Solution
(a) Parametrize the line segment by

z = −1 + (1 + i)t, 0 ≤ t ≤ 1,

so that

|z|2 = (−1 + t)2 + t 2 and dz = (1 + i) dt.

The value of the contour integral becomes

  1
2
|z|2 dz = (2t 2 − 2t + 1)(1 + i) dt = (1 + i).
C 0 3

(b) Along the unit circle |z| = 1, we have z = eiθ and dz = ieiθ dθ . The
initial and final points of the path correspond to θ = π and θ = π/2,
respectively. The contour integral can be evaluated as
   π2

π

ieiθ dθ = eiθ  = 1 + i.
2
|z|2 dz =
C π π

The results in (a) and (b) do not agree. Hence, the value of this contour
integral does depend on the path of integration.

Example 4.1.5 Let C be the closed contour consisting of four straight line
segments, Re z = ±a and Im z = ±a (a > 0), oriented in the anticlockwise
direction (C represents a square of sides 2a). Evaluate the integral
,
Re z dz.
C
140 Complex Integration

Figure 4.2. The contour is a closed square of sides 2a.

Solution As shown in Figure 4.2, the closed contour C consists of the four
line segments

C1 = {z : z = a + it, −a ≤ t ≤ a}, dz = i dt,

C2 = {z : z = −t + ia, −a ≤ t ≤ a}, dz = −dt,
C3 = {z : z = −a − it, −a ≤ t ≤ a}, dz = −i dt,
C4 = {z : z = t − ia, −a ≤ t ≤ a}, dz = dt,

where C = C1 ∪ C2 ∪ C3 ∪ C4 . The respective contour integrals along the line

segments are found to be
  a
Re z dz = ai dt = 2a 2 i
C1 −a
  a
Re z dz = (−t)(−dt) = 0
C2 −a
  a
Re z dz = (−a)(−i dt) = 2a 2 i
C3 −a
  a
Re z dz = t dt = 0.
C4 −a

The contour integral around the closed square C is given by the sum of the
contour integrals along C1 , C2 , C3 and C4 . Adding the four contour integrals
together, we obtain
,    
Re z dz = Re z dz + Re z dz + Re z dz + Re z dz
C C1 C2 C3 C4
= 4a 2 i.
 4.1 Formulations of complex integration 141

Example 4.1.6 Show that

 
 

(a)  (x + iy ) dz ≤ 2, where C is the line segment joining −i to i;
2 2
C
 
 
(b)  (x + iy ) dz ≤ π, where C is the right half-circle |z| = 1 and
 2 2
C
Re z ≥ 0;
 
 1 

(c)  dz ≤ 2, where C is the line segment joining −1 + i and 1 + i.
z 2
C

Solution
(a) Using modulus inequality (4.1.6), we obtain
  
 
 (x 2 + iy 2 ) dz ≤ |x 2 + iy 2 | |dz|, z = x + iy.
 
C C

By the triangle inequality, we deduce that

|x 2 + iy 2 | ≤ |x 2 | + |iy 2 | = |x|2 + |y|2 .

The line segment joining −i to i can be defined as z = iy, −1 ≤ y ≤ 1.

Along the line segment, we have |x|2 + |y|2 ≤ 1. Combining the results,
we obtain
    1
 
 (x 2 + iy 2 ) dz ≤ |dz| = |i dy| = 2.
 
C C −1

(b) Similarly, we use modulus inequality (4.1.6) to establish

   
 
 (x 2 + iy 2 ) dz ≤ (|x|2 + |y|2 ) |dz| = |z|2 |dz|.
 
C C C

The contour C is the right half unit circle, |z| = 1 and Re z ≥ 0. We then
have
   π/2
|z|2 |dz| = |dz| = |ieiθ | dθ = π,
C C −π/2

and so
 
 
 (x 2 + iy 2 ) dz ≤ π.
 
C

(c) Along
√ the contour C, we have z = x + i, −1 ≤ x ≤ 1, so that 1 ≤ |z| ≤
2 and, correspondingly, 12 ≤ |z|1 2 ≤ 1. Here, M = maxz∈C |z|1 2 = 1 and
142 Complex Integration

the arc length L = 2. Using modulus inequality (4.1.6), we have

 
 1 
 
 z2 dz ≤ ML = 2.
C

4.2 Cauchy integral theorem

One is tempted to ask under what conditions does
 
f (z) dz = f (z) dz, (4.2.1)
C1 C2

where C1 and C2 are two contours in a domain D with the same initial and final
points and f (z) is piecewise continuous inside D. We observe that the property
of path independence is valid for f (z) = z12 in Example 4.1.3, but it fails when
f (z) = |z|2 in Example 4.1.4. The above query is equivalent to the question:
when does
,
f (z) dz = 0 (4.2.2)
C

hold, where C is any closed contour lying completely inside D? The equivalence
of eqs. (4.2.1) and (4.2.2) is revealed if we treat C as C1 ∪ −C2 . In the above
examples, we observe that f (z) = z12 is analytic everywhere except at z = 0 but
f (z) = |z|2 is nowhere analytic. The observation suggests that analyticity of the
integrand may play an important role in establishing the validity of eq. (4.2.2).
The conjecture is confirmed by the renowned Cauchy integral theorem.

Theorem 4.2.1 (Cauchy integral theorem) Let f (z) be analytic on and inside
a simple closed contour C and let f  (z) be continuous on and inside C. Then
,
f (z) dz = 0.
C

Proof The proof of the Cauchy integral theorem requires Green’s theorem
from real calculus. Green’s theorem can be stated as: given a positively oriented
closed contour C, if the two real functions P (x, y) and Q(x, y) have continuous
first-order partial derivatives throughout the closed region consisting of all
points on and inside C, then
, 
P dx + Q dy = (Qx − Py ) dxdy, (4.2.3)
C D

where D is the simply connected domain bounded by C.

 4.2 Cauchy integral theorem 143

From eq. (4.1.5), suppose we write

f (z) = u(x, y) + iv(x, y), z = x + iy;

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

One can infer from the continuity of f  (z) that u(x, y) and v(x, y) have con-
tinuous derivatives on and inside C. Using Green’s theorem, the two real line
integrals can be transformed into double integrals. This gives
,  
f (z) dz = (−vx − uy ) dxdy + i (ux − vy ) dxdy.
C D D

Both integrands in the double integrals are equal to zero due to the Cauchy–
Riemann relations, hence the Cauchy integral theorem is established.
In 1903, Goursat was able to obtain the same result as in eq. (4.2.2) without
assuming the continuity of f  (z). This stronger version is called the Goursat
theorem. The omission of the continuity assumption is important. Based on
the Goursat theorem, we can show later that the derivative f  is also analytic
without assuming continuity of f  (see Theorem 4.3.2).

Theorem 4.2.2 (Goursat theorem) Given a simple closed contour C, let f (z)
be analytic on and inside C. Then
,
f (z) dz = 0.
C

We choose to omit the proof of the Goursat theorem here since the procedures
are rather technical. Readers interested in the proof may consult some other
texts in complex variables.†

Corollary 1 The integral of a function f (z) that is analytic throughout a

simply connected domain D depends on the end points and not on the particular
contour taken. Suppose α and β are inside D, and C1 and C2 are any contours
inside D joining α to β. We have
 
f (z) dz = f (z) dz. (4.2.4)
C1 C2

† A detailed proof of the Goursat theorem can be found in Complex Variables and Applications
by J.W. Brown and R.V. Churchill, 7th edition, McGraw-Hill, 2003.
144 Complex Integration

The basic essence of this corollary has been discussed at the beginning of
this section. As deduced from this corollary, the Goursat theorem can be stated
in the following alternative form:
If a function f (z) is analytic throughout a simply connected domain D, then
for any simple closed contour C lying completely inside D, we have
,
f (z) dz = 0.
C

Corollary 2 Let f (z) be analytic throughout a simply connected domain D.

Consider a fixed point z0 ∈ D; by virtue of Corollary 1, we deduce that
 z
F (z) = f (ζ ) dζ, for any z ∈ D, (4.2.5)
z0

is a well-defined function in D. Moreover, it can be shown that (see Problem 4.9)

F  (z) = f (z), for any z ∈ D, (4.2.6)
so F (z) is analytic throughout D. Here, F (z) may be visualized as a primitive
function of f (z).
This corollary may be considered as the complex counterpart of the funda-
mental theorem of real calculus. If we integrate f (z) along any contour joining
α and β inside D, then the value of the integral is given by
 β
f (z) dz = F (β) − F (α), α and β ∈ D. (4.2.7)
α

The above formula has been verified in Example 4.1.3.

Corollary 3 Let C, C1 , C2 , . . . , Cn be positively oriented closed contours,

where C1 , C2 , . . . , Cn are all inside C. For C1 , C2 , . . . , Cn , each of these
contours lies outside of the other contours. Let int Ci denote the collection
of all points bounded by Ci , i = 1, 2, . . . , n. Let f (z) be analytic on the set
S : C ∪ int C \ int C1 \ int C2 \ · · · \ int Cn (see the shaded area in Figure 4.3).
Then
, n ,
f (z) dz = f (z) dz. (4.2.8)
C k=1 Ck

The proof for the case when n = 2 is presented below. The extension to the
general case is straightforward.

Proof To each interior closed contour Ci , i = 1, 2, we place one cut which

joins Ci with the exterior contour C. The cut lines provide the passage from
 4.2 Cauchy integral theorem 145

Figure 4.3. The shaded region becomes simply connected with the introduction of the
cuts.

the exterior contour to the interior contours. We now construct the boundary
curve for the multiply connected region: C ∪ int C \ int C1 \ int C2 (see the
shaded area in Figure 4.3). The constructed boundary curve is composed of
C ∪ −C1 ∪ −C2 together with the cut lines. Each cut line travels twice in
opposite directions. To explain the negative signs in front of C1 and C2 , we
note that the interior contours traverse in the clockwise sense as parts of the
positively oriented boundary curve. With the introduction of these cuts, the
shaded region bounded within this constructed boundary curve becomes a
simply connected set.

By the Cauchy–Goursat theorem, the integral of f (z) along the above bound-
ary curve vanishes. Also, since f (z) is integrated along the cuts twice but in
opposite directions, the various contributions to the line integral along the cuts
are canceled off. We then have
,  
f (z) dz + f (z) dz + f (z) dz = 0,
C −C1 −C2

so that
,  
f (z) dz = f (z) dz + f (z) dz.
C C1 C2

Example 4.2.1 Evaluate the integral

 β
cos z esin z dz,
α

where α and β are any complex constants.

146 Complex Integration

Solution The function esin z is a primitive function of cos z esin z , and they are
both entire functions. By Corollary 2, we have
 β β

cos z esin z dz = esin z  = esin β − esin α .
α α

Example 4.2.2 Find the value of the integral

,
(|z| − ez sin z2 + z) dz,
C

where C is the circle |z| = a.

Solution The integral is split into three individual integrals:

, , , ,
(|z| − e sin z + z) dz = |z| dz − e sin z dz + z dz.
z 2 z 2
C C C C
z 2
The second integral equals zero since the integrand e sin z is an entire function.
The first integral is found to be
, ,
|z| dz = a dz = 0,
C C

since C is the circle |z| = a. To compute the third integral, we write z = aeiθ
and substitute z with ae−iθ and dz with aieiθ dθ . This gives
,  2π  2π
z dz = ae−iθ aieiθ dθ = a 2 i dθ = 2π a 2 i.
C 0 0

Combining the results, the value of the integral is found to be 2π a 2 i.

Example 4.2.3 Let D be the domain that contains the whole complex plane
except the origin and the negative real axis. Let  be an arbitrary contour, lying
completely inside D, that starts from 1 and ends at a point α (see Figure 4.4).
Show that

dz
= Log α.
 z

Solution Let 1 be the line segment from 1 to |α| along the real axis, and 2
be the circular arc centered at the origin and of radius |α| which extends from
|α| to α. The union 1 ∪ 2 ∪ − forms a closed contour (see Figure 4.4).
Since the integrand 1z is analytic everywhere inside D, by the Cauchy–Goursat
theorem, we have
 4.2 Cauchy integral theorem 147

Figure 4.4. The contour  starts from z = 1 and ends at z = α. The arc 2 is part of
the circle |z| = |α|.

  
dz dz dz
= + .
 z 1 z 2 z

Since α does not lie on the negative real axis, Arg α cannot assume the
value π. If we write α = |α|eiArg α (−π < Arg α < π ), then
  |α|
dz dt
= = ln |α|
1 z 1 t
and
 
dz Arg α
ireiθ
= dθ = i Arg α.
2 z 0 reiθ
Combining the results, the integral is found to be

dz
= ln |α| + i Arg α = Log α.
 z

The primitive function of 1z is Log z. The above result can be obtained by

applying the computational formula in eq. (4.2.7). Note that the given domain
D is the domain of definition of Log z, the principal branch of the complex
logarithm function. The branch cut is taken to be the negative real axis and the
branch points are z = 0 and z = ∞. By excluding the origin and the negative
real axis in D, we avoid multi-valuedness of the integral.

Example 4.2.4 Consider the integration of the function e−z around the rect-
2

angular contour  with vertices ±a, ±a + ib and oriented positively as shown

148 Complex Integration
y

Γ3
(−a, b) (a, b)

Γ4 Γ2

x
(−a, 0) Γ1 (a, 0)

Figure 4.5. The configuration of the closed rectangular contour .

in Figure 4.5. By letting a → ∞ while keeping b fixed, show that

 ∞  ∞ √
e−x e±2ibx dx = e−x cos 2bx dx = e−b
2 2 2
π.
−∞ −∞

The result is called the Poisson integral.

Since e−z is an entire function, we have

2
Solution
,
e−z dz = 0,
2

by virtue of the Cauchy–Goursat theorem. The closed contour  consists of

four line segments:  = 1 ∪ 2 ∪ 3 ∪ 4 , where (see Figure 4.5)

1 = {x : −a ≤ x ≤ a},
2 = {a + iy : 0 ≤ y ≤ b},
3 = {x + ib : −a ≤ x ≤ a},
4 = {−a + iy : 0 ≤ y ≤ b},

and  is oriented in the anticlockwise direction.

The contour integral can be split into four contour integrals, namely,
,    
−z2 −z2 −z2 −z2
e−z dz.
2
e dz = e dz + e dz + e dz +
 1 2 3 4
 4.2 Cauchy integral theorem 149

The four contour integrals can be expressed as real integrals as follows:

  a
e−z dz = e−x dx,
2 2

1 −a
 b
e−z dz = e−(a+iy) i dy,
2 2

2 0
  −a
e−z dz = e−(x+ib) dx,
2 2

3 a
 a  a

b2 −x 2 −x 2
= −e e cos 2bx dx − i e sin 2bx dx ,
−a −a
  0
e−z dz = e−(−a+iy) i dy.
2 2

4 b

First, we consider the bound on the modulus of the second integral. By the
modulus inequality (4.1.6), we have
   b
 
 e−z2 dz ≤ |e−(a −y +2iay) i| dy
2 2

 
2 0
 b
−a 2 2
=e ey dy
0
 b
≤ e−a
2 2
eb dy (since 0 ≤ y ≤ b)
0
2
beb
= 2 → 0 as a → ∞ and b is fixed.
ea

e−z dz → 0 as a → ∞.
2
Therefore, the value of
2
Using a similar argument, the fourth integral can be shown to be zero as
a → ∞. Thus, by taking the limit a → ∞ but keeping b fixed, the contour
integral around  is found to be equal to the sum of the first and third integrals
,  a  a 
−z2 −x 2 b2 −x 2
lim e dz = lim e dx − e e cos 2bx dx
a→∞  a→∞ −a −a
  a 
b2 −x 2
+ i lim e e sin 2bx dx = 0.
a→∞ −a

Using the known value of the following integral (see Problem 4.13)
 ∞ √
e−x dx =
2
π,
−∞
150 Complex Integration

we obtain
 ∞  ∞  ∞
−x 2 −x 2 −b2
√
e−x dx = e−b
2 2
e cos 2bx dx − i e sin 2bx dx = e π.
−∞ −∞ −∞

By equating the imaginary parts of the above equation, we observe

 ∞
e−x sin 2bx dx = 0.
2

−∞

Finally, we obtain
 ∞  ∞
−x 2 ±2ibx
√
e−x cos 2bx dx = e−b
2 2
e e dx = π.
−∞ −∞

Example 4.2.5 Let W (z, z) be a function with continuous first-order partial

derivatives in a domain D and on its boundary C, where z = x + iy and
z = x − iy. In terms of the pair of conjugate complex variables z and z, show
that the complex variable formulation of Green’s theorem is
, 
∂W
W (z, z) dz = 2i dxdy.
C D ∂z

Supposing a closed contour C encloses a region of area A, show that

, , ,
,
1 1 1
A= z dz = z dz − z dz = Im z dz.
2i C 4i C C 2 C

Use the formula to find the area enclosed by the ellipse with parametric repre-
sentation: x = 5 cos t, y = 4 sin t, 0 ≤ t < 2π .

Solution Let W (z, z) = P (x, y) + iQ(x, y). The contour integral can be
written as
, ,
W (z, z) dz = (P + iQ) (dx + idy)
C
,C ,
= P dx − Q dy + i Q dx + P dy.
C C

The above two real integrals can be transformed into double integrals by Green’s
theorem as follows:
,      
∂Q ∂P ∂P ∂Q
W (z, z) dz = − + dxdy + i − dxdy
C D ∂x ∂y D ∂x ∂y
  
∂ ∂
=i +i (P + iQ) dxdy
D ∂x ∂y

∂W ∂ ∂ ∂
= 2i dxdy, since 2 = +i .
D ∂z ∂z ∂x ∂y
 4.3 Cauchy integral formula and its consequences 151

By setting W (z, z) = z in the above formula and observing that

∂W
= 1,
∂z
we obtain
, 
z dz = 2i dxdy = 2iA.
C D

Taking the complex conjugate on both sides of the above equation, we have
,
z dz = −2iA.
C

The above results can be combined into the form

, , , 
1 1
A= z dz = z dz − z dz .
2i C 4i C C
,
Since A is real, the value of z dz must be purely imaginary. We then have
C
, ,
1 1
A= z dz = Im z dz.
2i C 2 C

Using the complex representation, the ellipse can be parametrized as

z(t) = 5 cos t + 4i sin t, 0 ≤ t < 2π.

The area of the ellipse is given by

 2π
1
A= (5 cos t − 4i sin t) (−5 sin t + 4i cos t) dt
2i 0
 2π  2π 
1
= − 9 cos t sin t dt + i 20 dt = 20π.
2i 0 0

4.3 Cauchy integral formula and its consequences

The Cauchy integral formula is considered to be one of the most important
and useful results in the theory of analytic functions. At the elementary level,
it provides a useful tool for the evaluation of a wide variety of complex inte-
grals. More importantly, the Cauchy integral formula plays a major role in the
development of more advanced topics in analytic function theory.

Theorem 4.3.1 (Cauchy integral formula) Let the function f (z) be analytic
on and inside a positively oriented simple closed contour C and z be any point
152 Complex Integration

Cr

×
z

Figure 4.6. A circle Cr centered at the point z is drawn which lies completely inside
the closed contour C.

inside C. Then
,
1 f (ζ )
f (z) = dζ. (4.3.1)
2π i Cζ −z

Proof We draw a circle Cr , with radius r, centered at the point z, small enough
(ζ )
to be completely inside C (see Figure 4.6). Since fζ −z is analytic in the region
lying between Cr and C, by Corollary 3 of the Cauchy–Goursat theorem, we
have
, ,
1 f (ζ ) 1 f (ζ )
dζ = dζ
2π i C ζ −z 2π i Cr ζ − z
, ,
1 f (ζ ) − f (z) f (z) 1
= dζ + dζ.
2π i Cr ζ −z 2π i Cr ζ − z

By virtue of the result obtained in Example 4.1.2, the last integral is seen to be
equal to f (z). To complete the proof, it suffices to show that the first integral
has the value zero.

Since f is continuous at z, for each  > 0, there exists δ > 0 such that

|f (ζ ) − f (z)| <  whenever |ζ − z| < δ.

Now, suppose we choose r < δ and thus we ensure that Cr lies com-
pletely inside the contour C. The modulus of the first integral is bounded
 4.3 Cauchy integral formula and its consequences 153

by
 ,  ,
 1 f (ζ ) − f (z)  1 |f (ζ ) − f (z)|
 
dζ  ≤ |dζ |
 2π i ζ −z 2π Cr |ζ − z|
Cr
,
1
= |f (ζ ) − f (z)| |dζ |
2π r Cr
,
 
< |dζ | = 2π r = .
2π r Cr 2π r

Since the modulus of the above integral is less than any positive number ,
however small, the value of that integral must be zero.
The Cauchy integral formula is a remarkable result. The value of f (z) at any
point inside the closed contour C is determined by the values of the function
along the bounding contour C.

4.3.1 Derivatives of contour integrals

Suppose we differentiate both sides of the Cauchy integral formula in eq. (4.3.1)
formally with respect to z (holding ζ fixed). Assuming that differentiation under
the integral sign is legitimate, we obtain
, ,
1 d f (ζ ) 1 f (ζ )
f  (z) = dζ = dζ. (4.3.2)
2π i C dz ζ − z 2π i C (ζ − z)2

Can we establish the legitimacy of direct differentiation of the integrand in

the Cauchy integral formula? Assuming that z and z + h both lie inside C, we
consider the expression
,
f (z + h) − f (z) 1 f (ζ )
− dζ
h 2π i C (ζ − z)2
 ,

1 1 f (ζ ) f (ζ ) f (ζ )
= − −h dζ
h 2π i C ζ − z − h ζ − z (ζ − z)2
,
h f (ζ )
= dζ.
2π i C (ζ − z − h) (ζ − z)2

To show the validity of eq. (4.3.2), it suffices to show that the value of the
last integral goes to zero as h → 0. To estimate the value of the last integral,
we draw the circle C2d : |ζ − z| = 2d that lies completely inside the domain
bounded by C and choose h such that 0 < |h| < d. Every point ζ on the curve
C is then outside the circle C2d so that

|ζ − z| > d and |ζ − z − h| > d.

154 Complex Integration

Let M be the upper bound of |f (z)| on C and L be the total arc length of C.
Using the modulus inequality (4.1.6) and together with the above inequalities,
we obtain
 , 
 h f (ζ )  |h| ML
 
 2π i (ζ − z − h) (ζ − z)2 dζ  ≤ 2π d 3 .
C

In the limit h → 0, we observe that

 , 
 h 
lim 
f (ζ )
dζ  ≤ lim |h| ML = 0;
h→0 2π i C (ζ − z − h)(ζ − z)2  h→0 2π d 3

therefore,
,
f (z + h) − f (z) 1 f (ζ )
f  (z) = lim = dζ.
h→0 h 2π i C (ζ − z)2
By induction, the generalized Cauchy integral formula can be established as
follows:
,
k! f (ζ )
f (z) =
(k)
dζ, k = 1, 2, 3, . . . . (4.3.3)
2π i C (ζ − z)k+1
An immediate consequence of the generalized Cauchy integral formula
(4.3.3) is that the mere assumption of analyticity of f at a point is sufficient to
guarantee the existence of the derivatives of f of all orders at the same point.
The precise statement of the result is summarized in the following theorem.

Theorem 4.3.2 If a function f (z) is analytic at a point, then its derivatives

of all orders are also analytic at the same point.

Proof Suppose f is analytic at a point z. Then there exists a neighborhood

|ζ − z| <  around z such that f is analytic throughout the neighborhood. Let
C  denote the positively oriented circle |ζ − z| <   , where   < , so that f
is analytic inside and on C  . According to eq. (4.3.3), the derivative f  exists
at each point inside C  , so f  is analytic at the point z. Repeating the argument
for the function f  that is analytic at z, we can conclude that f  is also analytic
at z, and so forth for all higher-order derivatives of f .

Remarks

(i) The above theorem is limited to complex functions only. In fact, no

similar statement can be made on real differentiable functions. It is easy
to find examples of real-valued functions f (x) such that f  (x) exists at
a point but f  (x) does not exist at that point.
 4.3 Cauchy integral formula and its consequences 155

(ii) Suppose we express an analytic function inside a domain D as f (z) =

u(x, y) + iv(x, y), z = x + iy. Since the derivatives of f of all orders
are analytic functions, it then follows that the partial derivatives of u(x, y)
and v(x, y) of all orders exist and are continuous. This result is consistent
with the earlier claim on continuity of higher-order derivatives when
we discuss the theory of harmonic functions in Section 2.6: the mere
assumption of analyticity of f at a point would guarantee the continuity
of all second-order derivatives of the real part and imaginary part of f .

The Cauchy integral formula can be extended to the case where the simple
closed contour C can be replaced by the oriented boundary of a multiply
connected domain as described in Corollary 3 of the Cauchy–Goursat theorem.
In fact, suppose C, C1 , C2 , . . . , Cn and f (z) obey the same conditions as in
Corollary 3; then for any point z ∈ C ∪ int C \int C1 \ int C2 \ · · · \ int Cn ,
we have
,  n ,
1 f (ζ ) 1 f (ζ )
f (z) = dζ − dζ. (4.3.4)
2π i C ζ − z k=1
2π i Ck ζ − z

The proof follows the same approach of introducing cuts that join the exterior
contour with the interior contours (see Figure 4.3).

Cauchy inequality
Suppose f (z) is analytic on and inside the disk |z − z0 | = r, 0 < r < ∞, and
let

M(r) = max |f (z)|.

|z−z0 |=r

Then
|f (k) (z)| M(r)
≤ k , k = 0, 1, 2, . . . . (4.3.5)
k! r
This inequality follows from the generalized Cauchy integral formula (4.3.3).

Example 4.3.1 Suppose f (z) is defined by the integral

,
3ζ 2 + 7ζ + 1
f (z) = dζ.
|ζ |=3 ζ −z

Find f  (1 + i).
156 Complex Integration

Solution By setting k = 1 in the generalized Cauchy integral formula (4.3.3),

we have
,
3ζ 2 + 7ζ + 1
f  (z) = dζ
|ζ |=3 (ζ − z)2
,
3(ζ − z)2 + (6z + 7)(ζ − z) + 3z2 + 7z + 1
= dζ
|ζ |=3 (ζ − z)2
, ,
1
= 3 dζ + (6z + 7) dζ
|ζ |=3 |ζ |=3 ζ − z
,
1
+ (3z2 + 7z + 1) dζ.
|ζ |=3 (ζ − z)2
The first integral equals zero since the integrand is entire (as it is a constant
function). For the second integral, we observe that
, 
1 0 if |z| > 3
dζ = .
|ζ |=3 ζ − z 2π i if |z| < 3
Furthermore, we deduce that the third integral is zero since
, ,

1 d 1
dζ = dζ = 0.
|ζ |=3 (ζ − z) dz |ζ |=3 ζ − z
2

Combining the results, we have


 (2π i)(6z + 7) if |z| < 3
f (z) = .
0 if |z| > 3
√
We observe that 1 + i is inside |z| < 3 since |1 + i| = 2 < 3. Therefore, we
obtain

f  (1 + i) = 2π i [6(1 + i) + 7] = −12π + 26π i.

Example 4.3.2 Suppose f (z) is analytic inside the unit circle |z| = 1 and
1
|f (z)| ≤ .
1 − |z|
Show that
 n
1
|f (n)
(0)| ≤ (n + 1)! 1 + .
n

(ζ )
Solution We integrate ζfn+1 around the circle |ζ | = n+1
n
, where f (ζ ) is analytic
on and inside the circle. Using the generalized Cauchy integral formula, we
 4.3 Cauchy integral formula and its consequences 157

have
,
n! f (ζ )
f (n)
(0) = dζ
2π i ζ n+1
|ζ |= n+1
n

 
n iθ
 2π f e  
n! n+1 n
=  n+1 eiθ i dθ
2π i 0 n i(n+1)θ
n+1
e
n+1
    
1 n n! 2π n iθ −inθ
= 1+ f e e dθ.
n 2π 0 n+1
The modulus |f (n) (0)| is bounded by
     
1 n n! 2π  n iθ 
|f (0)| ≤ 1 +
(n)
f e  dθ
n 2π 0  n+1
  
1 n n! 2π 1
≤ 1+ dθ
n 2π 0 1 − n
n+1
 
1 n n!
= 1+ [2π(n + 1)]
n 2π
 
1 n
= (n + 1)! 1 + .
n

4.3.2 Morera’s theorem

The converse of the Cauchy–Goursat theorem is Morera’s theorem.

Theorem 4.3.3 (Morera’s theorem) Suppose f (z) is continuous inside a

simply connected domain D and
,
f (z) dz = 0, (4.3.6)
C

for any closed contour C lying inside D. Then f (z) is analytic throughout D.

Proof The continuity property of f (z) and the property defined in eq. (4.3.6)
induce the following primitive function of f (z):
 z
F (z) = f (ζ ) dζ, for all z ∈ D, (4.3.7a)
z0

where z0 is a fixed point inside D. The function F (z) is single-valued; otherwise,

the property in eq. (4.3.6) would be violated. Further, it can be shown that (see
158 Complex Integration

Problem 4.9)

F  (z) = f (z), for all z ∈ D, (4.3.7b)

and so F (z) is analytic throughout D. By virtue of Theorem 4.3.2, F  (z) is also

analytic in the same domain. Hence, f (z) is analytic throughout D.

Remark This proof and the argument presented in Corollary 2 of the Cauchy–
Goursat theorem look quite similar. However, readers are reminded that the
assumption of the analyticity of f (z) is not needed to establish eqs. (4.3.7a,b).

4.3.3 Consequences of the Cauchy integral formula

The Cauchy integral formula leads to a wide variety of important theorems,
many of which find applications in the further development of complex analysis
and in other fields of mathematics. We discuss a few of these theorems that are
considered to be both important and interesting.

Theorem 4.3.4 (Gauss’ mean value theorem) If f (z) is analytic on and

inside the disk Cr : |z − z0 | = r, then

1 2π
f (z0 ) = f (z0 + reiθ ) dθ. (4.3.8)
2π 0

Proof From the Cauchy integral formula, we have

,
1 f (z)
f (z0 ) = dz
2π i Cr z − z0
 2π
1 f (z0 + reiθ )ireiθ
= dθ
2π i 0 reiθ
 2π
1
= f (z0 + reiθ ) dθ.
2π 0

Example 4.3.3 Find the mean value of x 2 − y 2 + x on the circle |z − i| = 2.

Solution First, we observe that x 2 − y 2 + x = Re(z2 + z). The mean value

of x 2 − y 2 + x on |z − i| = 2 is defined by

1 2π
f (i + 2eiθ ) dθ,
2π 0
 4.3 Cauchy integral formula and its consequences 159

where f (z) = Re(z2 + z). By Gauss’ mean value theorem, we have

 
1 2π 
f (i + 2e ) dθ = Re(z + z) = Re(−1 + i) = −1.
iθ 2
2π 0 z=i

Theorem 4.3.5 (Liouville’s theorem) The only bounded entire functions are
constant functions. Equivalently, suppose f (z) is entire and there exists a
constant B ∈ R such that |f (z)| ≤ B, for all z ∈ C; then f (z) = K for some
constant K ∈ C.

Proof It suffices to show that f  (z) = 0, for all z ∈ C. We integrate (ζf−z)

(ζ )
2

around the circle CR : |ζ − z| = R. By virtue of the generalized Cauchy integral

formula, we have
,
1 f (ζ )
f  (z) = dζ.
2π i CR (ζ − z)2
The above result remains valid for any sufficiently large R since f (z) is an entire
function. Using the modulus inequality (4.1.6), the bound on the modulus of
the integral can be estimated as
  , 
 1  f (ζ )  1 B B
 
|f (z)| =    dζ  ≤ 2π R = .
 
CR (ζ − z)
2π i 2 2π R 2 R
By letting R → ∞, we have
|f  (z)| = 0.
This implies that f (z) = K for some K ∈ C.

An interesting deduction from the Liouville theorem is that non-constant

entire functions must be unbounded. Actually we have already seen that entire
functions like sin z and cos z are unbounded, unlike their real counterparts.

Theorem 4.3.6 (Maximum modulus theorem) Let D be a bounded domain

that is enclosed by the closed contour C. If f (z) is analytic on the domain D
and continuous on the bounding contour C, then the maximum value of |f (z)|
occurs on C, unless f (z) is a constant function.

Proof Given that the function f (z) is continuous on and inside C, the modulus
function |f (z)| is also continuous on and inside C and the maximum value
of |f (z)| on or inside C always exists. To prove the theorem, it suffices to
establish the following claim. Supposing |f (z)| attains its maximum value at
some interior point α ∈ D, then f (z) is constant for all z ∈ D.
160 Complex Integration

First, we take a small neighborhood N(α; r) around α that lies completely

inside D. We then apply modulus inequality (4.1.3d) to the formula of Gauss’
mean value theorem [see eq. (4.3.8)] and obtain

 2π
1
|f (α)| ≤ |f (α + reiθ )| dθ. (4.3.9a)
2π 0

Suppose |f (α)| were a maximum; then

|f (α + reiθ )| ≤ |f (α)|, for all θ, (4.3.9b)

which implies that

 2π  2π
1 1
|f (α + re )| dθ ≤ iθ
|f (α)| dθ = |f (α)|. (4.3.9c)
2π 0 2π 0

Combining the results in inequalities (4.3.9a,c), we deduce that

 2π
[|f (α)| − |f (α + reiθ )|] dθ = 0.
0

In addition, inequality (4.3.9b) dictates that the above integrand is non-negative.

We then argue that |f (α)| = |f (α + reiθ )| for all points on the circle, in view
that it is not possible to have |f (α + reiθ )| less than |f (α)| even at some
isolated points. Assume the contrary, suppose |f (α + reiθ )| is less than |f (α)|
at a single point; by continuity of f (z), |f (α + reiθ )| will be less than |f (α)|
for a finite arc on the circle. This leads to

 2π
1
|f (α + reiθ )| dθ < |f (α)|, (4.3.9d)
2π 0

a contradiction to inequality (4.3.9a). Furthermore, since r can be any

value, it then follows that |f (z)| is constant in any neighborhood of α lying
inside D.
Next, we would like to show that |f (z)| is constant not only in every neigh-
borhood of α, but also at every point in D. We take any point z ∈ D, and
join α to z by a curve γ lying completely inside D. A sequence of points
 4.3 Cauchy integral formula and its consequences 161

×
z0

× γ
z1 ×
z2

×
zn−1
zn×= z

Figure 4.7. The point z0 is joined to the point z via a curve γ in D. The points
z0 , z1 , . . . , zn are points lying on γ and each of them is the center of a disk
(plus its boundary) lying completely inside D and containing the preceding point.

z0 = α, z1 , . . . , zn = z on γ are chosen with the property that each of these

points is the center of a disk (plus its boundary) lying inside D and containing the
preceding point (see Figure 4.7). Since z1 is contained in the disk centered at α,
from the above result, we have |f (z1 )| = |f (α)|. Applying the same argument to
|f (z)| in the disk centered at z1 , we conclude that |f (z)| must be constant within
this disc. In particular, |f (z2 )| = |f (z1 )| = |f (α)|. Repeating the argument,
we finally have |f (z)| = |f (zn−1 )| = · · · = |f (z1 )| = |f (α)|, valid for any
point z in D.
Now, we have established that |f (z)| is constant for every point in D if
|f (z)| attains its maximum value at some interior point α ∈ D. Recall that
f (z) is constant in D if |f (z)| is constant in D (see Problem 2.26). Therefore,
for a non-constant f (z), the maximum value of |f (z)| cannot occur inside
D. Equivalently, the maximum value of |f (z)| occurs only on the boundary
curve C.

Corollary Suppose D is a bounded domain and u(x, y) is harmonic inside D

and non-constant; then u(x, y) cannot attain its maximum or minimum value
inside D. In other words, if there exists (x0 , y0 ) ∈ D such that

u(x0 , y0 ) = sup u(x, y) or u(x0 , y0 ) = inf u(x, y),

D D

then u(x, y) is constant inside D. These results are called the maximum and
minimum principles for harmonic functions.
162 Complex Integration

The proof of the above principles for harmonic functions follows immedi-
ately from the maximum modulus theorem for analytic functions. First, we
construct an analytic function f (z) in D such that
Ref (z) = u(z), z = x + iy.
Next, we define the function g(z) = ef (z) , which is also analytic in D, and
|g(z)| = eu(z) . Applying the maximum modulus theorem to g(z), the maximum
value of eu(z) occurs only on the boundary of D, unless eu(z) is a constant
function. Since the exponential function is monotonically increasing, we deduce
that the maximum value of u(x, y) occurs on the boundary, unless u(x, y) is
a constant function. The minimum principle for harmonic functions can be
proved similarly by applying the maximum modulus theorem to −u(z).

Example 4.3.4 At what point inside the region |z − z0 | ≤ 1 does |ez | attain
its maximum value? Find this maximum value.

Solution By virtue of the maximum modulus principle, since ez is an entire

function, |ez | attains its maximum value inside the given region only along the
boundary |z − z0 | = 1. On te circle |z − z0 | = 1, z = z0 + eiθ , the value of
|ez | along the boundary is given by
|ez | = |ez0 +e | = |eRe z0 +cos θ+i(Im z0 +sin θ) | = eRe z0 ecos θ .
iθ

Note that ecos θ attains its maximum value at θ = 0. Therefore, |ez | attains its
maximum value eRe z0 +1 at the boundary point z = z0 + 1.

4.4 Potential functions of conservative fields

In this section, we examine the application of analytic function theory to the
study of several prototype conservative fields in physics, like fluid flow, elec-
trostatics and gravitation. The potential functions derived from these conserva-
tive fields are shown to be harmonic, thus explaining why complex variables
techniques are commonly applied to solve physical problems in fluid flow,
electrostatics and gravitation.

4.4.1 Velocity potential and stream function of fluid flows

The potential theory of fluid flows rests on the assumption of the incompress-
ible, inviscid and irrotational properties of the fluid motion. When the Mach
number (which is the ratio of the fluid velocity to the speed of sound) of the
fluid flow is below 0.3, the assumption of incompressibility is acceptable. For
 4.4 Potential functions of conservative fields 163

Figure 4.8. Volume fluxes across the four surfaces of a differential control volume in a
two-dimensional flow field.

common fluids, like air and water, viscosity effects are negligible in regions
distant from boundaries with solids. First, we derive the continuity equation
based on the property of incompressibility of the fluid. We then examine the
rotational property of fluid motion, characterized by the angular velocity of
a fluid element. A fluid motion is irrotational if the angular velocity of each
element throughout the whole flow field is zero. The condition of irrotationality
allows the definition of a scalar quantity called the velocity potential, whose
gradient gives the velocity vector. An irrotational, inviscid and incompressible
flow is termed a potential flow.

Continuity equation
Let u and v denote the horizontal and vertical components of the fluid veloc-
ity vector. Consider a control volume of infinitesimal widths x and y in a
two-dimensional flow field (see Figure 4.8). Let (x, y) be the coordinates of the
center of the control volume. The horizontal velocities at the left and right faces
are u − ∂x∂u x
2
and u + ∂x∂u x
2
, respectively. Likewise, the vertical velocities at
the bottom and top faces are v − ∂v y
∂y 2
and v + ∂v y
∂y 2
, respectively. The volume
flux of fluid flowing into the control volume (assuming unit depth  of thecon-
 
trol volume normal to the plane) equals u − ∂x 2 y + v − ∂v
∂u x y
∂y 2
x,
while the volume flux of fluid  flowing out of the control volume equals
 
u + ∂x
∂u x
2
y + v + ∂v y
∂y 2
x. The net rate of accumulation of fluid in
the control volume is equal to the volume flux inflow minus the volume flux
outflow. Thus, within the differential control volume,
 
∂u ∂v
net rate of accumulation = − + xy.
∂x ∂y
164 Complex Integration

u x, y y
2
vx x, y
vx x, y
2
2
(x, y)

u x, y y
2
Figure 4.9. The local rotational motion of the horizontal and vertical differential line
segments through the point (x, y).

Since the fluid is incompressible, the net rate of accumulation in any control
volume in the flow field is zero. Therefore, the condition of incompressibility
is given by
∂u ∂v
+ = 0. (4.4.1)
∂x ∂y
The above equation is called the continuity equation.

Irrotationality condition
To derive the condition of irrotationality, let us consider the local rota-
tional motion of the horizontal and vertical differential line segments through
(x, y) as shown in Figure 4.9. The horizontal line segment of width x
through (x, y) is subject to angular velocity (in the anti-clockwise sense) as
given by
   
v x + x , y − v x − x ,y ∂v
2 2
≈ (x, y).
x ∂x
Similarly, the vertical line segment of width y through the center is subject
to angular velocity (in the anti-clockwise sense) as given by
   
u x, y + y 2
− u x, y − y
2 ∂u
− ≈ − (x, y).
y ∂y
The average local angular velocity of the differential fluid elementof widths
x and y centered at the point (x, y) is then given by 12 ∂x ∂v
− ∂u
∂y
.
 4.4 Potential functions of conservative fields 165

Let v(x, y) denote the velocity vector of the fluid motion. We define the
vorticity ω to be the curl of the velocity vector, that is,
ω = ∇ × v. (4.4.2)
In a two-dimensional flow field, the vorticity has only one component that is
normal to the x-y plane. Its magnitude is given by
∂v ∂u
|ω| = − . (4.4.3)
∂x ∂y
The local angular velocity of a fluid element is seen to be |ω|
2
. A fluid motion is
said to be irrotational when the angular velocity at each point throughout the
flow field is zero. Therefore, the condition of irrotationality is given by
∂v ∂u
− = 0. (4.4.4)
∂x ∂y
To understand the physical interpretation of an irrotational fluid motion,
we consider an infinitesimal straw immersed in the flow field. If the fluid
is truly irrotational, the tiny straw always moves parallel to itself since the
fluid has zero angular velocity everywhere. Consider the following interesting
example: the flow field due to a vortex such as a tornado can be irrotational even
though the global motion of the fluid is circulating around in concentric circular
patterns.
In the above discussion, the continuity equation and irrotationality condition
have been formulated using the differential approach based on the physics of the
underlying processes. On the other hand, the integral formulations presented
below exhibit better linkage with the theory of conservative fields, and lead
naturally to the definitions of two scalar functions via a theorem in vector
calculus. These two scalar functions are termed the velocity potential and the
stream function, and they play fundamental roles in the description of potential
fluid flows.

Flux and circulation

Let r(s) = x(s)i + y(s)j be a smooth curve parametrized by its arc length s.
The unit tangent and normal vectors are, respectively, given by
dr dx dy
t= = i+ j,
ds ds ds
dy dx
n= i− j.
ds ds
The flux Fγ across a curve γ is defined by
 
Fγ = v · n ds = u dy − v dx. (4.4.5)
γ γ
166 Complex Integration

When γ is a closed curve, by the Gauss theorem in vector calculus, the above
line integral can be converted into a double integral:
,    
∂u ∂v
Fγ = v · n ds = ∇ · v dxdy = + dxdy, (4.4.6)
γ Aγ Aγ ∂x ∂y

where Aγ is the area bounded by the closed curve γ . Physically, if the fluid is
incompressible, then the flux across any closed curve γ is zero. Mathematically,
we deduce from eq. (4.4.6) that the integrand in the double integral vanishes at
any point in the flow field. This result is precisely the continuity equation given
in eq. (4.4.1).
The circulation Cγ along a curve γ is defined by
 
Cγ = v · dr = u dx + v dy. (4.4.7)
γ γ

When γ is a closed curve, by the Stokes theorem in vector calculus, Cγ can be

expressed as a double integral:
,  
Cγ = v · dr = ∇ × v · dA = |ω| dxdy, (4.4.8)
γ Aγ Aγ

where dA is the differential area vector and ω is the vorticity. We deduce that
the irrotationality condition [see eq. (4.4.4)] is equivalent to the vanishing of
Cγ for any closed curve γ inside the flow field.

Velocity potential and stream function

A continuously differentiable vector function F is said to be conservative in a
domain D if and only if
,
F · dr = 0, (4.4.9)
γ

for any closed curve γ inside D. By a well-known theorem in vector calculus,

the above condition is equivalent to the existence of a scalar potential φ(x, y)
such that

F = ∇φ. (4.4.10)

Supposing the circulation Cγ along any closed curve γ inside the domain D
is zero, by virtue of eq. (4.4.10), we claim that there exists a scalar function
φ(x, y) such that
∂φ ∂φ
v = ∇φ or u = , v= . (4.4.11)
∂x ∂y
 4.4 Potential functions of conservative fields 167

This scalar function φ is called the velocity potential of the flow field in D. The
level curves φ(x, y) = constant are called the equipotential lines of the flow
field.
If we write F = −vi + uj, then the incompressibility condition is given by
, ,
Fγ = v · n ds = F · dr = 0, (4.4.12)
γ γ

for any closed curve γ inside the flow field. Applying eq. (4.4.10) again, we
find that there exists a scalar function ψ(x, y) such that
∂ψ ∂ψ
− vi + uj = ∇ψ or u = , v=− . (4.4.13)
∂y ∂x
This scalar function ψ is called the stream function of the flow field in D. The
level curves ψ(x, y) = constant are called the streamlines of the flow field.
Note that
∇φ · ∇ψ = (ui + vj) · (−vi + uj) = 0, (4.4.14)
so that the streamlines and equipotential lines are orthogonal to each other.

Physical properties of the stream function

(i) A flow particle moves along a streamline. To verify this statement, we
consider
∂ψ ∂ψ
dψ = dx + dy = −v dx + u dy.
∂x ∂y
Along a streamline, we have dψ = 0 so that
 
dy v
0 = −v dx + u dy or = . (4.4.15)
dx ψ u
Therefore, the direction of the velocity vector at a point on a streamline
coincides with the slope of the streamline.
(ii) The difference between the values of ψ along two different stream-
lines gives the volume of fluid flowing between these two stream-
lines. To prove the statement, we consider two streamlines ψ = ψ1
and ψ = ψ2 and place a curve AB joining the two streamlines as shown
in Figure 4.10. The volume flow rate Q between ψ1 and ψ2 (assuming
ψ2 > ψ1 ) is given by
 B  B
∂ψ ∂ψ
Q= u dy − v dx = dx + dy
A A dx ∂y
 B
= dψ = ψ2 − ψ1 . (4.4.16)
A
168 Complex Integration
ψ2

ψ1

Figure 4.10. The two streamlines behave like impenetrable walls. The amount of fluid
flowing within the streamlines ψ1 and ψ2 (assuming ψ2 > ψ1 ) is given by ψ2 − ψ1 .

Complex potential
The incompressibility and irrotationality conditions are expressed by
∂u ∂v ∂v ∂u
+ =0 and − = 0,
∂x ∂y ∂x ∂y
respectively. The scalar functions φ(x, y) and ψ(x, y) are derived based on
irrotationality and incompressibility, respectively. Suppose we substitute u =
∂φ
∂x
and v = ∂φ∂y
into the continuity equation (4.4.1). We obtain

∂ 2φ ∂ 2φ
+ = 0, (4.4.17)
∂x 2 ∂y 2
where φ(x, y) is assumed to be twice differentiable. Hence, φ satisfies the
Laplace equation, so φ is a harmonic function.
Similarly, if we substitute u = ∂ψ
∂y
and v = − ∂ψ
∂x
into the irrotationality con-
dition [see eq. (4.4.4)], we obtain
∂ 2ψ ∂ 2ψ
+ = 0, (4.4.18)
∂x 2 ∂y 2
so ψ is also harmonic.
Note that the velocity potential and stream function are related to the velocity
components by
∂φ ∂ψ ∂φ ∂ψ
u= = and v= =− . (4.4.19)
∂x ∂y ∂y ∂x
The above relations are recognized to be the Cauchy-Riemann relations for the
harmonic functions φ(x, y) and ψ(x, y). If we define the complex potential
f (z) of a potential flow field by

f (z) = φ(x, y) + iψ(x, y), z = x + iy, (4.4.20)

 4.4 Potential functions of conservative fields 169

and assume that velocity components are continuous functions, then f (z) is
an analytic function by virtue of the relations in eq. (4.4.19) and the continuity
properties of the velocity functions. Conversely, for any analytic function, its
real and imaginary parts can be considered as a feasible velocity potential and
stream function of a potential flow field, respectively.
Since the partial derivatives of φ and ψ are related to the velocity components,
the derivative of the complex potential f is expected to have a similar physical
interpretation. Indeed, we observe that
df ∂φ ∂ψ
= +i = u − iv, (4.4.21)
dz ∂x ∂x
and the quantity u − iv is called the complex velocity.
For example, the complex velocity for a uniform flow with speed V0 and
at an angle α to the positive x-axis is given by V0 eiα . Using eq. (4.4.21) and
performing the integration, the complex potential of this uniform flow is found
to be
f (z) = V0 e−iα z.
The complex potential is unique up to an additive constant.
Let ur and uθ denote the radial and tangential components of velocity.
The relations between the velocity components in the rectangular and polar
coordinates are
u = ur cos θ − uθ sin θ and v = ur sin θ + uθ cos θ,
so that the complex velocity in polar coordinates is given by
u − iv = (ur cos θ − uθ sin θ) − i(ur sin θ + uθ cos θ )
= (ur − iuθ )e−iθ . (4.4.22)

Example 4.4.1 Discuss the two-dimensional flow field as represented by the

complex potential
f (z) = cz1/2 , c is real.

Solution It is more convenient to express the complex potential in polar

coordinates z = reiθ . The given complex potential can be expressed as
θ θ
f (r) = cr 1/2 eiθ/2 = cr 1/2 cos + icr 1/2 sin .
2 2
The velocity potential and stream function are, respectively,
θ θ
φ(r, θ ) = cr 1/2 cos and ψ(r, θ ) = cr 1/2 sin , −π < θ ≤ π.
2 2
170 Complex Integration

y =0

Figure 4.11. Pattern of the streamlines (solid) and equipotential lines (dashed) of the
flow field around a flat plate that is placed along the whole positive x-axis.

The line θ = 0 is the streamline corresponding to ψ = 0. The other streamlines

are defined by r 1/2 sin θ2 = constant. Some of the streamlines are drawn in
Figure 4.11. The equipotential lines are given by r 1/2 cos θ2 = constant. The
complex potential represents the potential flow around a flat plate placed along
the whole positive x-axis (the plate corresponds to the streamline ψ = 0).
The complex velocity is given by
 
df c c θ θ
= z−1/2 = 1/2 cos + i sin e−iθ .
dz 2 2r 2 2
Using eq. (4.4.22), the radial and tangential components of velocity are seen
to be
c θ c θ
ur = cos and uθ = − sin .
2r 1/2 2 2r 1/2 2
At the corner of the plate where r = 0, the velocity components become infinite.
The velocity is singular of order r −1/2 .

Complex potentials of basic fluid elements

The fluid source and fluid vortex are considered as the basic fluid elements.
Here, we would like to derive the complex potential of each of them. Also, we
give the physical interpretation of some flow fields obtained by superposition
of these basic elements.
 4.4 Potential functions of conservative fields 171

Source
The velocity function of a fluid source placed at the origin was derived in
Subsection 2.1.1. The complex velocity of the source is given by
k
f  (z) = v(z) = , k is real.
z
Upon integration, the complex potential is found to be
f (z) = k Log z.
Separating the complex potential into its real and imaginary parts, and writing
z = reiθ , the velocity potential and stream function of the flow field due to the
source are
φ(r, θ ) = k ln r and ψ(r, θ ) = kθ.
The streamlines are the radial lines θ = θ0 , −π < θ0 ≤ π . The equipotential
lines are orthogonal to the streamlines, and they are concentric circles centered
at the origin: r = r0 , r0 > 0.
The flow pattern exhibits radial symmetry at the origin. The volume flow
rate m (usually called the source strength) of the flow through the circle r = r0
is found to be
 2π
k
m= r0 dθ = 2π k.
0 r0
As expected, m is independent of r0 since there is no accumulation of fluid
within any circle. This is the volume of fluid per unit time flowing out from
the source. The direction of the flow is radially outward and the velocity at the
source is infinite. In terms of the source strength m, the complex potential of
the flow field due to the source is given by
m
f (z) = Log z. (4.4.23)
2π
The flux across any closed curve γ enclosing the origin in the flow field due
to the above source is given by
,
Fγ = u dy − v dx
γ
,
= Im f  (z) dz, dz = dx + idy,
γ
,
m 1
= Im dz = m, (4.4.24)
γ 2π z

by virtue of the Cauchy integral formula. Again, as there is no accumulation

of fluid inside γ , the flux across any closed curve γ enclosing the source is
172 Complex Integration

equal to the source strength. However, if γ does not encircle the source, the flux
becomes zero. On the other hand, it can be shown that the circulation across
any closed curve γ in the flow field due to a source is always zero.
In general, suppose there are n sources located at z1 , z2 , . . . , zn in the flow
field, whose respective source strengths are m1 , m2 , . . . , mn . The flux across a
closed curve γ that encircles all the sources is given by
,  
m1 1 m2 1 mn 1
Fγ = Im + + ··· + dz
γ 2π z − z1 2π z − z2 2π z − zn
= m1 + m2 + · · · + m n , (4.4.25)

which equals the sum of the source strengths.

Vortex
Another basic fluid element is the vortex, of which the common whirlpool or
tornado is a close approximation. The streamlines of the flow pattern due to
a vortex are concentric circles centered at the vortex. Supposing the vortex is
placed at the origin, the stream function may take the form

ψ(r, θ ) = −k ln r, k is real.

Based on this assumed form of ψ, the derived velocity components agree with
the required physics,
  namely, the radial velocity is zero and the tangential
velocity has O 1r dependence (see Problem 2.5). Since ψ is the harmonic
conjugate of the velocity potential φ, we deduce that

φ(r, θ ) = kθ,

so that the complex potential of the flow field due to the vortex becomes

f (z) = kθ − ik ln r = −ik Log z, z = reiθ . (4.4.26)

The above complex potential resembles that of a source except that the multi-
plicative constant in front of Log z becomes an imaginary number.
The circulation around any closed curve γ enclosing the origin in the flow
field due to the above vortex placed at the origin is given by
,
Cγ = u dx + v dy
γ
,
= Re f  (z) dz, dz = dx + idy
γ
,
−ik
= Re dz = 2π k. (4.4.27)
γ z
 4.4 Potential functions of conservative fields 173

We define the strength of a vortex to be the magnitude of the circulation around

a closed curve enclosing the vortex. Let  denote the strength of the vortex; we
then have  = 2π k, and so the complex potential of the fluid flow due to the
vortex with strength  can be expressed as

f (z) = Log z. (4.4.28)
2π i
From eq. (4.4.27), it is seen that the circulation around any closed contour
not enclosing the vortex is zero. By following a similar procedure to that in
eq. (4.4.24), one can show that the flux around any closed contour in the flow
field due to a vortex is always zero.

Doublet
A doublet is formed by the coalescence of a source and a sink (a negative source)
of equal strength placed closed to each other (see Example 2.1.1). Consider
a source of strength m placed at z = − and a sink of the same strength m
placed at z = , where  is infinitesimal. The complex potential for this flow
configuration is given by
m m
f (z) = Log(z + ) − Log(z − )
2π 2π
   2 

m   
= Log 1 + 1+ +O ,   |z|
2π z z z2
  2 
m  
= Log 1 + 2 + O .
2π z z2
In the limits  → 0 and m → ∞ while keeping m = π µ, where µ is finite,
the complex potential due to a doublet then becomes
µ
f (z) = . (4.4.29)
z
The stream function of the flow field due to the doublet at the origin is found
to be
y
ψ(x, y) = −µ 2 .
x + y2
For the streamline ψ(x, y) = ψ0 , the corresponding equation is
   
µ 2 µ 2
x2 + y + = . (4.4.30)
2ψ0 2ψ0
The pattern of the streamlines forms a coaxial system of circles along the y-axis
(see Figure 4.12).
174 Complex Integration
y

Figure 4.12. Pattern of the streamlines of the flow field due to a doublet at the origin.

The complex velocity of the doublet flow is given by

1 cos θ − i sin θ −iθ
f  (z) = −µ 2
= −µ e . (4.4.31)
z r2
Therefore, the radial and tangential velocity components are, respectively,
cos θ sin θ
ur = −µ and uθ = −µ . (4.4.32)
r2 r2

Example 4.4.2 Discuss the potential flow field formed by the superposition
of a uniform flow of speed U along the positive x-axis and a doublet of strength
µ placed at the origin. Investigate the streamline pattern of the flow field.

Solution By the superposition principle, the complex potential of the resulting

flow field is given by
µ
f (z) = U z + , z = reiθ .
z
The stream function of the flow is found to be
 µ
ψ(r, θ ) = Im f (z) = U r − sin θ.
r
 
We see that ψ becomes zero when r = Uµ , implying that the circle |z| = Uµ
is a streamline in the flow field. Apparently, we may place a solid circle of
the same radius in the flow field where the body surface coincides with the
streamline. The flow field is not disturbed since fluid particles are flowing past
a streamline as if flowing tangentially past a solid surface.
 4.4 Potential functions of conservative fields 175

Figure 4.13. Streamline pattern of uniform potential flow past a circular obstacle of
radius a placed at the origin.

The present flow field

 is exactly the same as that of a uniform flow past
a circle of radius a = Uµ placed at the origin. In other words, the complex
potential of the uniform flow with speed U in the positive x-direction that
streams past the circle |z| = a is given by
 
a2
f (z) = U z + .
z

The stream function can be expressed as

 
a2
ψ(r, θ ) = U r − sin θ.
r

The streamlines ψ = constant of the above flow configuration are the family of
cubic curves y(x 2 + y 2 − a 2 ) = c(x 2 + y 2 ), where c is constant. The pattern
of the streamlines of the flow field is plotted in Figure 4.13.

4.4.2 Electrostatic fields

An electrostatic field is a vector field defined by the forces caused by station-
ary charged bodies. The Coulomb law in electrostatic theory states that the
magnitude of the force F between two point charges q1 and q2 is given by
1 q1 q2
|F| = , (4.4.33)
4π  r 2
176 Complex Integration

where r is the distance between the two charges and  is the permittivity of the
medium of the electric field. The electric field E at the field point (x, y, h) due
to a point charge q at (ξ, η, ζ ) is defined to be the force acting on a particle
of positive unit charge at the point (x, y, h). Therefore, the electric field E at
(x, y, h) produced by a point charge q placed at the origin is given by
q
E= r, (4.4.34)
4π r 3
where r 2 = x 2 + y 2 + h2 and r is the position vector of (x, y, h). It can be
shown by direct differentiation that the divergence of the electric field is zero,
that is, ∇ · E = 0 (see Problem 4.44).
The electric flux FS across a closed surface S is defined to be the integral over
the surface of the component of the electric field normal (outward oriented) to
the surface, that is,

FS = E · dS, (4.4.35)
S

where dS is the differential area vector. We use the Gauss theorem in vector
calculus to transform the above area integral into a volume integral and obtain
 
FS = E · dS = ∇ · E dV ,
S V

where V is the volume bounded by the closed surface S.

We now compute the flux across a closed surface S due to the point charge
q placed at the origin. Suppose the surface does not contain the origin (that
is, not enclosing the point charge); since the divergence of the electric field is
zero, we then have
 
q q
FS = 3
r · dS = ∇· r dV = 0.
S 4π r V 4π r 3
However, when V does contain the origin, the above volume integral over V
has the singularity r = 0. We delete a small ball Bd of radius d around the
origin from the volume V . Inside the remaining volume V \Bd , the electric
field has zero divergence. By applying the Gauss theorem over the deleted
volume V \Bd , we obtain

q
0= ∇· r dV
V \Bd 4π r 3
 
q q
= 3
r · dS − 3
r · dS,
S 4π r Sd 4π r

where Sd is the surface of the ball Bd . The result indicates that the electric
fluxes across S and Sd have the same value. The electric flux across Sd can be
 4.4 Potential functions of conservative fields 177

evaluated in a simple manner. On the spherical surface Sd , we have

r · dS dS
3
= 2
r d
and so
 
q q q
FS = FSd = r · dS = dS = .
Sd 4π r 3 4π d 2 Sd 
In summary, the electric flux across the closed surface S is given by
 q
if V contains the point charge q
FS = E · dS =  . (4.4.36)
V 0 otherwise
In general, if we consider a distributed charge of volume charge density ρ
inside a closed volume V , the electric flux is then given by
 
ρ
FS = E · dS = dV ,
S V 
---
where ρ dV is the total charge enclosed inside V . By transforming the
above surface integral into a volume integral, we obtain
 
ρ
∇ · E dV = dV .
V V 

Since the volume V is arbitrary, the integrand functions on both sides of the
above equation must be equal. We then obtain
ρ
∇ ·E= . (4.4.37)

Equation (4.4.37) is one of the Maxwell equations in electromagnetic theory.
The electric field E is known to be a conservative field, that is,
,
E · dr = 0,
γ

for any closed path γ . Therefore, there exists a scalar electric potential φ such
that

∇φ = −E. (4.4.38)

By convention, the negative sign is included since the electric potential defines
its value as the work required to move a positive unit test charge against
the electric force from a point with a lower potential to another point with a
higher potential. For example, using both eqs. (4.4.34) and (4.4.38), the electric
potential at the field point (x, y, h) due to the point charge q placed at the origin
178 Complex Integration

is found to be
q
φ(x, y, z) = . (4.4.39)
4π r
For q > 0, the electric potential value increases as the field point approaches
the point charge.
Substituting eq. (4.4.38) into eq. (4.4.37), we then obtain the Poisson equation
ρ
∇ 2φ = − , (4.4.40)

which is the governing equation for the electric potential. In particular, when
ρ = 0, the Poisson equation reduces to the Laplace equation
∇ 2 φ = 0. (4.4.41)
Suppose we confine ourselves to two-dimensional electrostatic problems in
free space without generating charges; then the electric potential is a harmonic
function. In this case, the complex variable techniques may provide useful tools
for solving these problems. Given that φ is harmonic, there exists a harmonic
conjugate ψ to φ such that
(z) = φ(x, y) + iψ(x, y), z = x + iy, (4.4.42)
is an analytic function. The harmonic conjugate ψ is called the flux function
and  is called the complex potential. The orthogonal families of curves in the
two-dimensional plane defined by
φ(x, y) = α and ψ(x, y) = β
are called the equipotential lines and flux lines of the electric field, respectively.
The derivative of  is seen to be
d ∂φ ∂ψ ∂φ ∂φ
= +i = −i = −(Ex − iEy ), (4.4.43)
dz ∂x ∂x ∂x ∂y
where Ex and Ey are the respective x- and y-components of the electric field
vector E.

Example 4.4.3 Find the complex potential due to an infinitely long line charge
with uniform linear charge density ρ.

Solution Assume the line charge to be aligned with the vertical ζ -axis. The
electric field vector due to the line charge is expected to have circular symmetry.
By eq. (4.4.34), its value at the field point (x, y, h) is given by
 ∞
ρ R
E= 3
dζ,
−∞ 4π  R
 4.4 Potential functions of conservative fields 179

where ρ dζ gives the amount of charge over the differential segment dζ along
the line charge, R = xi + yj + (h − ζ )k, and R 2 = x 2 + y 2 + (h − ζ )2 . We
see that
 ∞  ∞
1 2 h−ζ
dζ = and dζ = 0,
−∞ R
3 x +y
2 2
−∞ R
3

and so
ρ xi + yj ρ r
E= = ,
2π  x 2 + y 2 2π  r 2
where r = xi + yj and r 2 = x 2 + y 2 . Using eq. (4.4.38), the electric potential
φ can be found by integrating
ρ
∇φ = − r
2π r 2
to give
ρ
φ=− ln r.
2π 
The corresponding complex potential is then found to be
ρ
(z) = − Log z, z = reiθ .
2π 
The equipotential lines are concentric circles around the line charge and the flux
lines are rays emanating from the line charge. The line charge configuration
closely resembles the fluid source in a potential flow field.

4.4.3 Gravitational fields

The gravitational vector field at the field point (x, y, h) due to a particle of mass
m placed at (ξ, η, ζ ) is defined to be the gravitational attractive force acting on
a particle of unit mass at (x, y, h). Let

r = (x − ξ )i + (y − η)j + (h − ζ )k

denote the vector from (ξ, η, ζ ) to (x, y, h). By Newton’s law of universal
gravitation, the gravitational vector field is given by
m
F = −G 3 r, (4.4.44)
r
where r 2 = (x − ξ )2 + (y − η)2 + (h − ζ )2 and G is the universal gravita-
tional constant. The negative sign in the above equation reflects the attractive
nature of the gravitational force. The gravitational vector field F is known to
be conservative, which is reminiscent of the property that the work done in
180 Complex Integration

traversing around a closed loop in the field is zero. Therefore, there exists a
scalar gravitational potential function φ such that

∇φ = F.

For example, the gravitational potential due to a particle of mass m placed at

(ξ, η, ζ ) can be found by integrating
m
∇φ = −G r
r3
to give
m
φ=G . (4.4.45)
r
In general, suppose the mass in a body is distributed continuously with density
ρ(ξ, η, ζ ) throughout the volume V . The gravitational potential due to the body
at the field point (x, y, h) is given by

ρ(ξ, η, ζ )
φ(x, y, h) = G dξ dηdζ. (4.4.46)
V r
Note that there is a singularity r = 0 of the integrand in the above integral when
the field point is inside or on the surface of the body.
Following the usual convention, the vertical h-axis and ζ -axis are taken to
be positive downward. The vertical gravity gh due to a three-dimensional body
is defined by

∂φ ζ −h
gh = (x, y, h) = G ρ(ξ, η, ζ ) 3 dξ dηdζ. (4.4.47)
∂h V r
For example, consider a horizontal cylinder with uniform density ρ whose
longitudinal axis is parallel to the η-axis. Suppose the longitudinal extent of
the cylinder ranges from η1 to η2 (assuming η1 < η2 ); the vertical gravity due
to the cylinder can be obtained by computing the integral
   η2 
ζ −h
gh = Gρ dη dξ dζ
S η1 r3
 %
ζ −h η2 − y
= Gρ

S (ξ − x) 2 + (ζ − h)2
(ξ − x) + (η2 − y)2 + (ζ − h)2
2
&
η1 − y
−
dξ dζ,
(ξ − x)2 + (η1 − y)2 + (ζ − h)2
where S is the cross-section of the horizontal cylinder projected onto the ξ -ζ
plane. Suppose we take the limits η1 → −∞ and η2 → ∞ but keep y finite in
 4.4 Potential functions of conservative fields 181

the above integral. The resulting integral for gh of a horizontal cylinder with
an infinite extent becomes

ζ −h
gh = 2Gρ dξ dζ. (4.4.48)
S (ξ − x) + (ζ − h)
2 2

Conjugate complex variables formulation

The evaluation of the above integral for gh can be greatly facilitated by complex
integration. We illustrate how to transform the above double integral into a
contour integral by Green’s theorem using the conjugate complex variables
formulation. Let ω = ξ + iζ and ω = ξ − iζ be a pair of conjugate complex
variables. Given that the function F (ω, ω) is continuous up to its first-order
partial derivatives in a simply connected domain S and its boundary C, we
show in Example 4.2.5 that Green’s theorem can be formulated as
, 
∂F
F (ω, ω) dω = 2i dξ dζ. (4.4.49)
C S ∂ω

In the present gravitational problem, we further define another pair of conjugate

complex variables: s = x + ih and s = x − ih. In terms of the two pairs of
conjugate complex variables, we observe that eq. (4.4.49) can be expressed as

1
gh = −2Gρ Im dξ dζ. (4.4.50)
S ω−s
1
The function ω−s is analytic throughout S if the field point s is outside the
boundary curve C. In this case, Green’s theorem as formulated in eq. (4.4.49)
can be applied. When the point s lies on or inside C, the above form of Green’s
theorem has to be modified accordingly.
We set
∂F 1
(ω, ω) =
∂ω ω−s
so that
ω−s
F (ω, ω) = .
ω−s
By virtue of eq. (4.4.50), the double integral for gh then becomes
 ,
1 ω−s
gh = −2Gρ Im dξ dζ = Gρ Re dω, (4.4.51)
S ω−s C ω−s

where s = x + ih lies outside C.

Example 4.4.4 Find the vertical gravity gh of a horizontal infinite cylin-

der, whose cross-section is an n-sided polygon with vertices Pj (ξj , ζj ),
182 Complex Integration

x
x x

x z

Figure 4.14. The cross-section of the horizontal infinite cylinder is an n-sided polygon
with vertices Pj (ξj , ζj ), j = 1, 2, . . . , n. By convention, the vertical ζ -axis and h-axis
are chosen to be positive downward, so correspondingly, the vertices are looped in the
clockwise sense.

j = 1, 2, . . . , n looped in the positive sense (see Figure 4.14). The field point
is assumed to be outside the bounding curve of the polygonal cross-section.
For convenience of notation, it is taken to be at the origin, that is, s = 0.

Solution We write ωj = ξj + iζj , j = 1, 2, . . . , n, and define ωj =

ωj +1 − ωj (take ωn+1 to be ω1 ). The equation of the polygonal side joining the
two adjacent vertices Pj and Pj +1 can be expressed as

ω = αj ω + βj ,

where
ωj
αj = = e−2iArg ωj ,
ωj
ωj 2i(ξj ζj +1 − ξj +1 ζj )
βj = ωj − ωj = .
ωj ωj

Substituting the above relations into the integral formula for gh [see
eq. (4.4.51)], the vertical gravity of the horizontal polygonal cylinder is given by
n 
 ωj +1
αj ω + βj
gh = Gρ Re dω
j =1 ωj ω
n 
 
ωj +1
= Gρ Re αj ωj + βj Log .
j =1
ωj
 4.5 Problems 183

Note that

n 
n
αj ωj = ωj +1 − ωj = 0
j =1 j =1

and
ωj +1 rj +1
Log = ln + i(θj +1 − θj ),
ωj rj
where

rj2 = ξj2 + ζj2 and θj = Arg ωj .

Finally, we obtain

n
ξj ζj +1 − ξj +1 ζj
gh = 2Gρ
j =1
(ξj +1 − ξj )2 + (ζj +1 − ζj )2

rj +1
× (ζj +1 − ζj ) ln − (ξj +1 − ξj )(θj +1 − θj )
rj
 n
ξj ζj +1 − ξj +1 ζj
= 2Gρ

j +1 − ξj ) + (ζj +1 − ζj )
(ξ 2 2
j =1

rj +1
× sin(Arg ωj ) ln − cos(Arg ωj )(θj +1 − θj ) .
rj

Remark If the field point is placed at (x, h) instead of the origin, then the
corresponding formula for gh can be obtained simply by changing ξj to ξj − x
and ζj to ζj − h, j = 1, 2, . . . , n.

4.5 Problems
4.1. For any curve  joining z1 and z2 in the complex plane, show that

z3 − z13
z2 dz = 2 .
 3
4.2. Evaluate the following integrals:
 π+2i  i
z
(a) cos dz; (b) (2 + iz)2 dz.
0 2 1

In each case, the contour is the line segment joining the lower point to
the upper point.
184 Complex Integration

4.3. Evaluate the contour integral


z dz,
C

where the contour C is

(a) the line segment joining (0, 0) and (1, 1);

(b) a union of two line segments: C1 runs from (0, 0) to (1, 0) and C2
runs from (1, 0) to (1, 1).

Do the two results agree? If not, explain why.

4.4. Show that
  
 
 dz ≤ |dz|
 
C C

and explain its geometric significance.

4.5. Show that
 
f (z) dz = f (z) dz.
C C

4.6. Estimate the upper bound of the modulus of the integral


Log z
I= dz,
C z − 4i

where C is the circle |z| = 3.

   
   
  ln |z| + i Arg z ln |z| + |Arg z|
 
 Log z 
    .
Hint:  z − 4i  = |z − 4i|
≤
 
|z| − |4i|
 

4.7 Find an upper bound for

, 
 
 e z dz,
1


C1 (0)

where C1 (0) is the positively oriented unit circle centered at the origin.

Hint: Along C1 (0), we have

1
|e z | ≤ e.
 4.5 Problems 185

4.8. Let C be the union of three line segments, namely

π   π π
C= + it : 0 ≤ t ≤ 1 ∪ i − t : − ≤ t ≤
2  2 2
π
∪ − − it : −1 ≤ t ≤ 0 .
2
Estimate an upper bound of the modulus of each of the following inte-
grals:
  
(a) |z|2 dz; (b) | sin z| dz; (c) Im e−z dz.
C C C

4.9. Suppose f (z) is analytic throughout a simply connected domain D. Take

a fixed point z0 ∈ D and define the function
 z
F (z) = f (ζ ) dζ, z ∈ D.
z0

Show that
F  (z) = f (z), z ∈ D.

Hint: Consider
 z+z
F (z + z) − F (z) = f (ζ ) dζ.
z

By the Cauchy theorem, the last integral is independent of the

path joining z and z + z as long as the path is completely inside
D. Suppose we take the integration path from z to z + z to be
a line segment; then
 z+z
F (z + z) − F (z) 1
− f (z) = [f (ζ ) − f (z)] dζ.
z z z
Using the continuity property of f (z), show that the above right-
hand integral tends to zero as z → 0.

4.10. Suppose 0 < r < 1; show that

 2π  
1  reiθ 
  dθ = r .
2π 0 (1 − re ) 
 iθ 2 1 − r2

Hint: Consider
,  
1 1 1
1= − dζ,
2π i |ζ |=1 ζ −z ζ − 1
z

where |z| = r < 1 and ζ ζ = 1.

186 Complex Integration

4.11. Evaluate the following integrals:

, , ,
1 3z − 1 sin πz
4
(a) dz; (b) dz; (c) dz;
|z|=10 1 + z |z|=2 z(z − 1) |z+1|=1 z − 1
3 2
, ,
ez 1
(d) dz; (e) dz;
1 + z 2 1 + z4
,|z+i|=1 x +y =2x
2 2
,
1 1
(f) 2 − 1)(z3 − 1)
dz; (g) dz;
(z 3 (z2 + 1)(z2 + 4)
|z|=r<1 |z|= 2
, , ,
sin z ez
(h)  2 dz; (i) 3
dz; (j) |z − 1| |dz|.
|z|=2 z − π |z|=1 z |z|=1
2

4.12. Suppose the contour joining −i and i lies in the complex plane excluding
the origin and the negative real axis. Evaluate

 i
1
z 2 dz
−i

1
where the principal branch of z 2 (−π < Arg z ≤ π ) is taken.
4.13. By considering the contour integral

,
eiπz +2πz
2

dz,
C e2πz + 1

where C = C1 ∪ C2 ∪ C3 ∪ C4 is the closed contour shown in the figure,

show that
 ∞ √
π
e−x dx =
2
.
0 2

The four line segments are

C1 = {x + ix : −R ≤ x ≤ R},
C2 = {R + i(R + y) : 0 ≤ y ≤ 1},
C3 = {x + i(x + i) : −R ≤ x ≤ R},
C4 = {−R + i(y − R) : 0 ≤ y ≤ 1}.

The integrand has a simple pole at z = 2i . Show that the line integral
along the vertical line segment C2 or C4 vanishes as R → ∞.
 4.5 Problems 187

C2

R C3 i C1
4 x
R

C4

4.14. By evaluating the contour integral

,  
1 2n dz
z+ ,
|z|=1 z z
show that
 2π
1 · 3 · 5 · · · (2n − 1)
cos2n θ dθ = 2π .
0 2 · 4 · 6 · · · (2n)
4.15. Apply the Cauchy integral formula to the integral
,
ekz
dz, k is a real constant,
|z|=1 z

to show that
 π
ek cos θ cos(k sin θ ) dθ = π.
0
4.16. Suppose f (z) is analytic on and inside the unit circle |z| = 1, and f (0) =
1. Show that

2 2π θ
f (eiθ ) cos2 dθ = 2 + f  (0)
π 0 2
and
 2π
2 θ
f (eiθ ) sin2 dθ = 2 − f  (0).
π 0 2
,  
1 1 f (z)
Hint: Evaluate 2±z+ dz.
2π i |z|=1 z z

4.17. Consider the function

z−a
f (z) = .
z+a
188 Complex Integration

Use the generalized Cauchy integral formula to show that

 
1 n
f (n) (0) = −2 − n!.
a

Hint: Replace the variable of integration by its reciprocal.

4.18. Let
, πζ
e3
f (z) = dζ.
|ζ |=2 ζ − z

Find the values f (i) and f (−i). Moreover, evaluate f (z) when |z| > 2.
4.19 Suppose C is any contour that has the starting point z = 0 and ending
point z = 1 but does not go through z = i and z = −i. Show that the
contour integral

1
dz
C 1+z
2

has value π
4
+ kπ, where k is some integer.
4.20. Let
,
2ζ 2 − ζ + 1
g(z) = dζ.
|ζ |=2 ζ −z
Compute (a) g(1); (b) g(z0 ), |z0 | > 2. Can we evaluate g(2)?
4.21. If C is any closed contour around the point z = −1, show that
,  
1 zezt t 2 −t
dz = t − e , t > 0.
2π i C (z + 1)3 2

4.22. If f (z) is an nth-degree polynomial with nonzero leading coefficient,

f (z) = a0 zn + a1 zn−1 + · · · + an ,
and C is a simple closed contour enclosing all the zeros of f (z), show
that
,
1 zf  (z) a1
(a) dz = − ;
2π i C f (z) a0
, 2 
1 z f (z) a12 − 2a0 a2
(b) dz = .
2π i C f (z) a02
4.23. If f (z) is analytic on and inside the unit circle |z| = 1, show that
,
1 1 − zζ
(1 − |z| ) f (z) =
2
f (ζ ) dζ, |z| < 1,
2π i |ζ |=1 ζ −z
 4.5 Problems 189

and deduce that

 2π
1
(1 − |z|2 ) |f (z)| ≤ |f (eiθ )| dθ, |z| < 1.
2π i 0

Hint: |1 − zζ | = |ζ − z| when |ζ | = 1.

4.24. Supposing f (z) is analytic on and inside the unit circle |z| = 1, and that
Re f (z) > 0, f (0) = α > 0, show that
 
 f (z) − α 
  
 f (z) + α ≤ |z| and |f (0)| ≤ 2α.

4.25. Suppose f (z) is analytic and |f (z)| ≤ M inside the circle |z| = R, and
f (0) = 0; show that
M M
|f (z)| ≤ |z| and |f  (0)| ≤ ,
R R
and equality holds only if f (z) = M R
eiθ z, where θ is real.
4.26. Suppose f (z) is analytic inside the domain 0 < |z| < 1, and
,
f (z) dz = 0,
|z|=r

for all values of r ∈ (0, 1). Is f (z) analytic at z = 0? If not, give a counter
example.
4.27. Suppose f (z) is analytic and nonzero inside the domain D. Explain why
,
f  (z)
dz = 0,
γ f (z)

where γ is any simple closed contour inside D.

4.28. Let f be entire and suppose Im f (z) ≤ M for all z. Show that f must be
a constant function.
4.29. Let f be analytic on and inside a simple closed curve C and |f (z) − 1| <
1 for all z on C. Use the maximum modulus theorem to show that f is
nonzero inside C.
4.30. Find the maximum value of |z2 + 3z − 1| in the disk |z| ≤ 1.
4.31. Let R denote the rectangular region:

0 ≤ x ≤ π, 0 ≤ y ≤ 1.

By the maximum modulus theorem, the modulus of the entire function

f (z) = sin z
190 Complex Integration

has a maximum value in R that occurs on the boundary. Find the point
on the boundary that gives the maximum value.
4.32. Suppose f (z) is analytic in the domain |z| < R and

M(r) = max |f (z)|, r < R.

|z|=r

Show that M(r) is an increasing function of r.

4.33. Use the mean value theorem to show that
 π
ln(1 − 2r cos θ + r 2 ) dθ = 0, −1 < r < 1.
0

4.34. Suppose u(z) and v(z) are harmonic in a domain that contains the unit
disk |z| ≤ 1; prove the following claims:

(a) If u(z) > v(z) for |z| = 1, then u(z) > v(z) for all z inside |z| = 1;
(b) If u(z) = v(z) for |z| = 1, then u(z) = v(z) for all z inside |z| = 1;
(c) If u(0) ≥ u(z) for |z| = 1, then u(z) is constant for |z| ≤ 1.

If u(z) and v(z) are treated as the temperature functions in a steady state
temperature field, do the above results agree with your physical intuition
about temperature fields?
4.35. Show that the integral of the complex velocity around any simple closed
contour in a flow field is equal to  + im, where  is the net strength of
vortices inside the contour, and m is the net strength of sources and sinks
inside the contour.
4.36. Interpret the flow field with the complex potential
K1 − iK2 z−α
f (z) = Log ,
2π z−β
where K1 and K2 are real constants. Find the velocity potential and
stream function, and sketch the pattern of streamlines and equipotential
lines of the flow field.
4.37. Suppose the velocity potential and stream function of a potential flow are
defined by

x + iy = c cos(φ + iψ).

Show that
x2 y2
+ = 1.
c2 cosh2 ψ c2 sinh2 ψ
Explain why the streamlines are confocal ellipses, and show that the
circulation around any one of these ellipses is 2π .
 4.5 Problems 191

4.38. Suppose the complex potential f = φ + iψ of a potential flow in the

complex plane is defined by
z = cosh f.
Show that the streamlines are confocal hyperbolas and explain why the
pattern might represent the flow through an aperture.
4.39. Suppose a source of strength m is located at z = a, a > 0, and a wall is
placed along the vertical line x = 0. Explain why the complex potential
of this flow configuration in the right half-plane, Re z > 0, is given by
m
f (z) = [Log(z − a) + Log(z + a)].
2π
Note that the vertical wall x = 0 must be a streamline. Sketch some other
streamlines of the flow field. This solution approach is called the method
of images. The image source is placed at z = −a, which is the mirror
image of z = a with respect to the vertical wall x = 0.
4.40. A uniform flow of velocity U∞ > 0 is disturbed by the vertical line
segment joining z = −i and z = i. Find the complex potential and the
equation representing the family of streamlines. Sketch some of the
streamlines of the flow field.
√
Hint: Note that the imaginary part of the function z2 + 1 becomes
zero when z assumes values along the line segment joining
z = −i and z = i. Also, see Example 8.1.5 for a more sophisti-
cated approach to solving this problem.

4.41. Consider an n-sided polygonal vortex patch with uniform vorticity K

placed in a two-dimensional flow field. Let the vertices of the polygon
be denoted by Pj (ξj , ηj ), j = 1, 2, . . . , n, and P (x, y) be an arbitrary
field point outside the vortex patch. Show that the velocity components
at the field point (x, y) induced by the n-sided polygonal vortex patch
are given by

K y−η
u(x, y) = − dξ dη,
2π (x − ξ )2 + (y − η)2
 S
K x−ξ
v(x, y) = dξ dη,
S (x − ξ ) + (y − η)
2π 2 2

where S is the region covered by the vortex patch. Find the explicit
formulas for the velocity components by evaluating the above integrals.
4.42. Suppose a potential flow field is described by the complex potential
w = f (z) which is free from singularities inside the circle |z| = a. A
192 Complex Integration

stationary circular cylinder of radius a is now placed at the origin to

modify the flow. Show that the complex potential of the modified flow
configuration is given by
 2
a
f (z) + f .
z
This is called the circle theorem in potential flow theory.
4.43. Consider the uniform potential flow of speed U in the positive x-direction
which streams past a circle of radius a, whose center coincides with the
origin (see Example 4.4.2). Show that the flow speed V at an arbitrary
point (r, θ ), r ≥ a, in the flow field is given by

a4 2a 2
V = U 1 + 4 + 2 (sin2 θ − cos2 θ).
r r
Find the position in the flow field that has the greatest flow speed. Also,
find the equation representing the family of equipotential lines.
4.44. Show that
df (r)
∇ · rf (r) = 3f (r) + r .
dr
By setting f (r) = r −3 , verify that
r
∇· = 0, r = 0.
r3
Can you give the physical interpretation of the above result with reference
to the electric field generated by a point charge placed at the origin?
4.45. Suppose a line charge with linear charge density ρ is placed at the point
z1 = x1 + iy1 , z1 = 0, in the two-dimensional plane. An infinite plane is
placed along the real axis which is earthed to the potential zero. Find the
complex potential of the resulting electrostatic field. Describe the pattern
of the equipotential lines of this electrostatic field.
4.46. Consider the region bounded by two infinitely long concentric cylindrical
conductors of radii a and b (assuming b > a) which are charged to
uniform electric potentials φa and φb , respectively. Find the electric
potential of the electrostatic field inside the region.

Hint: The solution of the electric potential takes the form

φ(r) = A ln r + B,
where A and B are arbitrary constants.
 4.5 Problems 193

4.47. For the horizontal polygonal cylinder considered in Example 4.4.4, show
that the x- and h-derivatives of the vertical gravity gh are given by
 n

∂gh rj +1
= −Gρ cos(2Arg ωj ) ln + sin(2Arg ωj )(θj +1 − θj ) ,
∂x j =1
rj
 n

∂gh rj +1
= Gρ cos(2Arg ωj )(θj +1 − θj ) − sin(2Arg ωj ) ln .
∂h j =1
rj

4.48. Consider the horizontal infinite cylinder whose cross-section is a circle of

radius r and centered at (ξ0 , ζ0 ). Find the corresponding contour integral
for computing the vertical gravity gh due to the body. Evaluate the integral
to obtain an explicit formula for gh , assuming that the field point (0, 0)
is outside the circular cross-section.

4.49. Suppose the density function of an infinite horizontal cylinder has linear
variation of the form
ρ(ξ, ζ ) = a0 + a1 ξ + a2 ζ .
Show that the integral formula for the vertical gravity gh [see eq. (4.4.51)
with s = 0] becomes
, ,
ω a1 ω2
gh = G a0 Re dω − Re dω
C ω 4 C 2
,

a2 ω2
+ a2 (area of S) − Im dω .
4 C ω
 5
Taylor and Laurent Series

A power series with non-negative power terms is called a Taylor series. In

complex variable theory, it is common to work with power series with both
positive and negative power terms. This type of power series is called a Laurent
series. The primary goal of this chapter is to establish the relation between
convergent power series and analytic functions. More precisely, we try to
understand how the region of convergence of a Taylor series or a Laurent series
is related to the domain of analyticity of an analytic function. The knowledge
of Taylor and Laurent series expansion is linked with more advanced topics,
like the classification of singularities of complex functions, residue calculus,
analytic continuation, etc.
This chapter starts with the definitions of convergence of complex sequences
and series. Many of the definitions and theorems for complex sequences and
series are inferred from their counterparts in real variable calculus.

5.1 Complex sequences and series

An infinite sequence of complex numbers, denoted by {zn }, can be considered
as a function defined on a set of positive integers into the unextended complex
plane. In other words, the sequence of complex numbers z1 , z2 , z3 , . . . is
arranged sequentially and defined by some specific rule.

5.1.1 Convergence of complex sequences

Given a complex sequence {zn }, if for each positive quantity , there exists a
positive integer N such that

|zn − z| <  whenever n > N , (5.1.1)

194
 5.1 Complex sequences and series 195

then the sequence is said to converge to the limit z. In general, the choice of N
depends on . The definition implies that every -neighborhood of z contains
all except a finite number of members of the sequence. Symbolically, we write
lim zn = z . (5.1.2)
n→∞

The limit of a convergent sequence is unique. If the sequence fails to converge,

it is said to be divergent.
Suppose we separate zn , n = 1, 2, . . . , and z into their real and imaginary
parts and write zn = xn + iyn and z = x + iy. Then
|xn − x| ≤ |zn − z| ≤ |xn − x| + |yn − y|,
|yn − y| ≤ |zn − z| ≤ |xn − x| + |yn − y| .
Using the above inequalities, it is easy to show that
lim zn = z ⇐⇒ lim xn = x and lim yn = y . (5.1.3)
n→∞ n→∞ n→∞

Therefore, the study of the convergence of a complex sequence is equivalent

to the consideration of the convergence of two real sequences. Therefore most
theorems concerning the convergence of real series can be generalized and
extended to complex series. For example, the Cauchy criterion for convergence
of a real sequence can also be extended to a complex sequence.

Cauchy criterion
A complex sequence {zn } converges if and only if for each positive , there
exists N() such that
|zm − zn | <  for n, m > N . (5.1.4)
This can be proved easily by referring the convergence of {zn } to the con-
vergence of the real sequences {xn } and {yn } and then applying the Cauchy
criterion for the two real sequences.
As expected, theorems on the sum, difference, product and quotient of two
complex sequences are the same as those for two real sequences. If the two
complex sequences {an } and {bn } converge to their respective limits A and B,
then
lim (an ± bn ) = A ± B,
n→∞
lim an bn = AB,
n→∞
an A
lim = , provided B = 0.
n→∞ bn B
196 Taylor and Laurent Series

Limit superior
The root test discussed in Section 5.1.3 requires the concept of the limit superior
of a real sequence, denoted by the symbol lim. Consider a sequence {xn } of real
numbers, and let S denote the set of all of its limit points. The limit superior of
the sequence {xn } is defined to be the supremum (least upper bound) of S. For
example, consider the real sequence xn = 3 + (−1)n , n = 1, 2, . . . . The limit
points of this real sequence are 2 and 4, so the limit superior is max{2, 4} = 4.

5.1.2 Infinite series of complex numbers

An infinite series of complex numbers z1 , z2 , z3 , . . . is the infinite sum of the
sequence {zn } given by
 n 

z1 + z2 + z3 + · · · = lim zk . (5.1.5)
n→∞
k=1

To study the properties of an infinite series, we define the sequence of partial

sums {Sn } by

n
Sn = zk . (5.1.6)
k=1

If the limit of the sequence {Sn } converges to S, then the series is said to be
convergent and S is its sum; otherwise, the series is divergent. The consideration
of an infinite series is relegated to that of an infinite sequence of partial sums.
Given that an infinite series converges, we define the remainder after n terms
by

Rn = S − Sn , (5.1.7)

and obviously

lim Rn = 0 . (5.1.8)
n→∞

Conversely, it can be shown that if eq. (5.1.8) holds, then the infinite series is
convergent.
A necessary condition for the infinite series in eq. (5.1.5) to converge is that

lim zn = 0 . (5.1.9)
n→∞

However, the above condition is not sufficient for convergence. This can be
revealed by the harmonic sequence defined by zn = n1 . Though n1 → 0 as
.
n → ∞, the harmonic series ∞ 1
n=1 n is known to be divergent.
 5.1 Complex sequences and series 197

Absolute convergence
.
An infinite series zn is said to be absolutely convergent if the associated
.
series of absolute values |zn | converges. It is relatively straightforward to
show that an absolutely convergent series is always convergent (see Problem
. . .
5.1). If zn converges but |zn | does not, the series z is said to be
. n einθ
conditionally convergent. For example, the complex series ∞ n=1 n , θ = 0,
can be shown to be conditionally convergent (see Problem 5.13).

5.1.3 Convergence tests of complex series

The following tests are most commonly used for examining the convergence
of series.

Comparison test
. .
If |zn | ≤ αn and αn converges, then the series zn is absolutely convergent.

Ratio test
 
 zn+1 
Suppose lim   converges to L; then . zn is absolutely convergent if
n→∞ zn 
L < 1 and divergent if L > 1. When L = 1, the ratio test fails.

Root test
1
Suppose the limit superior of the real sequence {|zn | n } equals L. The series
.
zn converges absolutely if L < 1 and diverges if L > 1. The root test fails
when L = 1.

Gauss’ test
When the ratio test fails, this method may be useful. If the ratio has the asymp-
totic expansion
 
 zn+1  k αn
 
 z  = 1 − n + n2 + · · · ,
n

where |αn | is bounded for all n > N for sufficiently large N , then the series
.
zn converges absolutely if k > 1, and diverges or converges conditionally if
k ≤ 1.
198 Taylor and Laurent Series

Remark
(i) A bounded monotonically increasing or decreasing sequence is conver-
gent.
(ii) The removal or addition of a finite number of terms from or to an infinite
series does not change the convergence or divergence of the series.

Example 5.1.1 Check the convergence of the following series:

∞ ∞
1 1 einθ
(a) + in n, (b) .
n=1
n2 n=1
n2

Solution
.
(a) First, we observe that lim n1/n = 1. The series ∞ n=1 n
1/n
is divergent
n→∞ . ∞
since limn→∞ n1/n = 0. Therefore, n=1 n12 + in1/n is also divergent.
(b) The associated series of absolute values of the terms in the given complex
.
series is seen to be ∞ 1
n=1 n2 , which is known to be convergent. By the
.
comparison test, the given series ∞ einθ
n=1 n2 is absolutely convergent, and
so the series is convergent.

Example 5.1.2 Consider the complex sequence {zn } defined by

qn
zn = .
1 + q 2n
Give a judicious guess for the limit

α = lim zn
n→∞

for various values of q, then prove the claims.

Solution We distinguish the following three cases:

(i) |q| > 1
 
 qn  |q|n
 ≤ |q|
n

|zn | =  ∼ → 0 as n → ∞;
1 + q 2n  |q|2n − 1 |q|2n

(ii) |q| < 1

 
 qn 
 ≤ |q|
n

|zn | =  ∼ |q|n → 0 as n → ∞;
1 + q 2n  1 − |q|2n
 5.1 Complex sequences and series 199

(iii) |q| = 1, that is, q = eiθ for some real value of θ

einθ 1
zn = = ,
1 + e2inθ 2 cos nθ
and obviously lim zn does not exist.
n→∞

The proofs of the first two claims are presented below.

(i) |q| > 1
Given any  > 0, we want to find N such that for all n > N ,
 
 qn 
 ≤ |q|
n
|zn | =   < .
1+q 2n |q| − 1
2n

The above inequality is equivalent to either

√ √
1 + 1 + 4 2 1 − 1 + 4 2
|q| >
n
or |q| <n
.
2 2
The second inequality is discarded since we have assumed |q| > 1. The
first inequality can be expressed as
 √ 
1+4 2
ln 1+ 2
n> .
ln |q|
Suppose we choose N to be the largest integer smaller than or equal
to the right-hand term in the above inequality. Then for all n > N , the
above inequality holds. Reversing the above procedure by rearranging
the terms, we can deduce
|zn | <  for any  > 0 whenever n > N .
This establishes the proof that
lim zn = 0.
n→∞

(ii) |q| < 1

Given any  > 0, we consider
  1
 qn  |q|n |q|n
 
 1 + q 2n  ≤ 1 − |q|2n = 1
−1
< ,
|q|2n

which takes a similar form as in case (i) if we replace |q| by |q|

1
. Corre-
spondingly, we choose N to be the largest integer smaller than or equal
to
 √ 
1+4 2
ln 1+ 2
  .
1
ln |q|
200 Taylor and Laurent Series

For any  > 0 and all n > N , we then have

  1
 qn  |q|n

|zn | =  ≤ < ,
1 + q 2n  1
|q|2n
−1
thus establishing
lim zn = 0.
n→∞

In summary, we have
qn
lim = 0 when |q| = 1;
n→∞ 1 + q 2n
 n 
q
and the limit of the sequence 1+q 2n does not exist when |q| = 1.

5.2 Sequences and series of complex functions

In this section, we discuss the convergence of sequences and series of complex
functions. Let f1 (z), f2 (z), . . . , fn (z), . . . , denoted by {fn (z)}, be a sequence
of complex functions of z that are defined and single-valued in a point set R
in the complex plane. For some point z0 ∈ R, {fn (z0 )} becomes a sequence
of complex numbers. Supposing the sequence {fn (z0 )} converges, the limit is
unique. The value of the limit depends on z0 , and we write
f (z0 ) = lim fn (z0 ) . (5.2.1)
n→∞

If this holds for every z ∈ R, the sequence of complex functions {fn (z)} defines
a complex function f (z) in the point set R. Symbolically, we write
f (z) = lim fn (z) . (5.2.2)
n→∞

Definition 5.2.1 A sequence of complex functions {fn (z)} defined in a set R

is said to converge to a complex function f (z) defined in the same set if and only
if, for any given small positive quantity , we can find a positive integer N (; z)
[in general, N(; z) depends on  and z] such that |f (z) − fn (z)| <  for all
n > N (; z). The set R is called the region of convergence† of the sequence of
complex functions.
In general, we may not be able to find a single N () that works for all points
in R. In this case, the convergence is said to be pointwise. Otherwise, {fn (z)}
is said to converge uniformly to f (z) in R when N depends on  only.

† The conventional use of “region” in characterizing “region of convergence” is often but not
always in the same sense of “region” that has been defined in Section 1.4.
 5.2 Sequences and series of complex functions 201

5.2.1 Convergence of series of complex functions

An infinite series of complex functions
∞

f1 (z) + f2 (z) + f3 (z) + · · · = fk (z) (5.2.3)
k=1

is related to the sequence of partial sums {Sn (z)}, the relation being defined by

n
Sn (z) = fk (z) . (5.2.4)
k=1

Similarly, the discussion of the convergence of the infinite series in

eq. (5.2.3) can be relegated to that of the sequence of partial sums defined
in eq. (5.2.4). The infinite series in eq. (5.2.3) is said to be convergent if
lim Sn (z) = S(z) , (5.2.5)
n→∞

where S(z) is called the sum; otherwise the series is divergent.

Many of the properties related to convergence in sequences and series of
complex functions can be extended from their counterparts in sequences and
series of complex numbers. For example, a necessary but not sufficient condi-
tion for the infinite series in eq. (5.2.3) to converge is that
lim fk (z) = 0 . (5.2.6)
k→∞

Example 5.2.1 Consider the following infinite series of complex functions

∞
 sin kz
.
k=1
k2

Show that it is absolutely convergent when z is real, but that when z is non-real,
it becomes divergent.

Solution
(i) When z is real, we have
 
 sin kz  1
 
 k2  ≤ k2 , for all positive integer values of k.
. .∞ sin kz
Since ∞ 1
k=1 k 2 is known to be convergent, k=1 k 2 is absolutely con-
vergent.
(ii) When z is non-real, we let z = x + iy, y = 0. From the relation
sin kz e−ky eikx − eky e−ikx
= ,
k2 2k 2 i
202 Taylor and Laurent Series

we deduce that
 
 sin kz  ek|y| − e−k|y|
 
 k2  ≥ 2k 2
→ ∞ as k → ∞.
  .
 
Since  sink2kz  is unbounded as k → ∞, ∞ sin kz
k=1 k 2 is divergent.

Uniform convergence
Suppose we let

n
Rn (z) = S(z) − fk (z) = S(z) − Sn (z) (5.2.7)
k=1

be the remainder after n terms of the infinite complex series in eq. (5.2.3). The
infinite series is said to converge uniformly to S(z) in some set R if and only if
for any positive quantity , we can find a sufficiently large positive integer N ,
independent of z, such that for all z ∈ R,
|Rn (z)| <  whenever n > N . (5.2.8)
Uniform convergence is a strong property of an infinite series of complex
functions. Properties such as continuity or analyticity of the constituent func-
.
tions fk (z) are carried over to the sum S(z). To be precise, suppose fk (z)
converges uniformly to S(z) in some set R and let fk (z) be continuous in R.
Then the sum S(z) is also continuous in the set R.
Moreover, a uniformly convergent series of continuous functions on a contour
C can be integrated term by term; that is,
   ∞ ∞ 

S(z) dz = fk (z) dz = fk (z) dz , (5.2.9)
C C k=1 k=1 C

where C is any contour on which the constituent functions fk (z) are continuous.
A similar statement can be made for term by term differentiation. Given that
each constituent function fk (z) is analytic in a simply connected domain D
.
and the infinite series ∞k=1 fk (z) converges uniformly to S(z) in any compact
subset of D, then S(z) is also analytic in D and
∞ ∞
d  
S  (z) = fk (z) = fk (z). (5.2.10)
dz k=1 k=1

Weierstrass M test
The Weierstrass M test provides a simple tool for testing the uniform conver-
gence of an infinite series of complex functions. It states that if |fk (z)| ≤ Mk ,
 5.2 Sequences and series of complex functions 203
.∞
where Mk is independent of z in the set R and the infinite series k=1 Mk
.
converges, then the infinite series ∞ k=1 fk (z) is absolutely convergent and
uniformly convergent in R.
We use the comparison test and the property on uniform convergence shown
.
in eq. (5.2.8) to establish the M test. Given that |fk (z)| ≤ Mk in R and ∞ k=1 Mk
.
converges, the absolute convergence of ∞ k=1 kf (z) in R follows from the
comparison test. To prove uniform convergence, we consider the bound on the
remainder after n terms
 ∞  ∞ ∞
    
|Rn (z)| ≤  fk (z) ≤ |fk (z)| ≤ Mk .
k=n+1 k=n+1 k=n+1
.∞ .
Given that k=1 Mk converges, the corresponding remainder ∞ k=n+1 Mk can
be made less than any  by choosing n > N for some N . Clearly, N is inde-
pendent of z. We have

|Rn (z)| <  for n > N,

.
so uniform convergence of ∞ k=1 fk (z) in R follows.

Example 5.2.2 Consider the following infinite series of complex functions

∞

z(1 − z) + z2 (1 − z) + · · · + zk (1 − z) + · · · = zk (1 − z).
k=1

Show that the series converges for |z| < 1 and find its sum. Examine the uniform
convergence of the series for (i) |z| ≤ r0 < 1, a closed disk that lies completely
inside the unit circle; (ii) |z| ≤ 1, the closed unit disk.

Solution Define the partial sum

∞

Sn (z) = zk (1 − z)
k=1

= (z − z2 ) + (z2 − z3 ) + · · · + (zn − zn+1 )

= z − zn+1 .

Supposing |z| < 1, a judicious guess of the limit of the partial sum as n → ∞
is z. For |z| < 1, given any  > 0, we consider

|Sn (z) − z| = |z − zn+1 − z| = |z|n+1 < .

The above inequality is valid provided that we choose n such that

(n + 1) ln |z| < ln  for z = 0;

204 Taylor and Laurent Series

or
ln 
n> − 1 for z = 0.
ln |z|

We set
 
ln 
N(; z) = f l −1 ,
ln |z|

where f l(x) denotes the largest integer less than or equal to x. We then have

|Sn (z) − z| <  whenever n > N (; z), z = 0.

Here, N (; z) has dependence on both  and z. Lastly, when z = 0, we observe

|Sn (0) − 0| = 0 <  for all n.

.
Hence, the series ∞ k=1 z (1 − z) converges to its sum z for |z| < 1.
k

Next, we examine uniform convergence of the series.

 
(i) Given that |z| ≤ r0 < 1, we may choose N to be f l lnln|r0 | − 1 , which
.
has dependence on  only. Hence, the series ∞ k=1 z (1 − z) converges
k

uniformly for |z| ≤ r0 < 1.

(ii) However, the above choice of N cannot be applied to the case when
r0 = 1 since ln |r0 | becomes infinite. Therefore, we cannot establish
uniform convergence of the series when |z| ≤ 1.

Example 5.2.3 Using the Weierstrass M test, establish the uniform conver-
gence of the following series in their respective regions of convergence.
∞
 zn
(a) √ , |z| ≤ 1;
n=1
n n+1
∞
1
(b) , 1 < |z| < 2.
n=1
n2 + z2

Solution
(a) Given that |z| ≤ 1, we observe
 
 n  1
 √z 
 n n + 1  ≤ n3/2 .
 5.2 Sequences and series of complex functions 205
.
Accordingly, we choose Mn = n3/2
1
. Since ∞ 1
n=1 n3/2 is known to be
convergent,
∞
 zn
√ converges uniformly for |z| ≤ 1.
n=1
n n+1
(b) For n ≥ 3 and 1 < |z| < 2, we observe
n2
|n2 + z2 | ≥ |n2 | − |z|2 ≥ n2 − 4 ≥
2
so that
 
 1  2
 
 n2 + z2  ≤ n2 .
. .∞
We take Mn = n22 and note that ∞ 2
n=3 n2 converges. Hence,
1
n=3 n2 +z2
converges uniformly for 1 < |z| < 2. Adding two extra terms does not
.
affect uniform convergence, so ∞ 1
n=1 n2 +z2 shares the same property of
uniform convergence.

Example 5.2.4 Consider the Riemann zeta function defined by

∞

ζ (z) = n−z .
n=1

Show that the function is analytic inside the region {z : Re z > 1}. Also, find
its derivative.

Solution The constituent functions n−z = e−(ln n)z , n = 1, 2, . . . , are seen to

be analytic in Re z > 1. We would like to establish the uniform convergence of
.∞ −z
n=1 n for Re z ≥ γ > 1. Noting that for Re z ≥ γ > 1,
|n−z | = |n−Re z | ≤ n−γ ,
.
we choose Mn = n−γ , γ > 1, in the M test. It is well known that ∞ 1
n=1 nγ con-
.∞ −z
verges for γ > 1. By the Weierstrass M test, n=1 n converges uniformly for
.
Re z ≥ γ > 1. The infinite series ∞ n=1 n
−z
is seen to be uniformly convergent
in any compact subset of Re z > 1. By the termwise differentiation property of
uniformly convergent series and eq. (5.2.10), we deduce that ζ (z) is analytic in
Re z > 1 and its derivative is given by
∞ ∞
d −z 
ζ  (z) = n = (ln n)n−z ,
n=1
dz n=1

valid for Re z > 1.

206 Taylor and Laurent Series

5.2.2 Power series

The choice of the constituent functions

fn (z) = an (z − z0 )n , n = 0, 1, 2, . . . ,

in an infinite complex series leads to a power series expanded at the point

z = z0 . A power series defines a function f (z) for those points z at which it
converges. The properties of absolute convergence of infinite power series are
stated in Theorem 5.2.1.

Theorem 5.2.1 (Absolute convergence of power series) If the infinite power

.
series ∞ n=0 an (z − z0 ) converges at z = z1 , where z1 is some point other
n

than z0 , then it is absolutely convergent at each point z inside the open disk
|z − z0 | < R1 , where R1 = |z1 − z0 |.
.
Proof Since ∞ n=0 an (z1 − z0 ) converges for z1 = z0 , an (z1 − z0 ) tends to 0
n n

as n → ∞. Therefore, there exists N such that when n > N we have |an (z1 −
z0 )n | < 1. Suppose z lies in the open disk |z − z0 | < |z1 − z0 |; we derive the
following bound on the remainder after N terms of the associated series of
absolute terms
∞
 ∞
 ∞
|z − z0 |n
|an (z − z0 )n | = |an | |z − z0 |n ≤ .
n=N+1 n=N+1
|z − z0 |n
n=N+1 1

The ratio of the successive terms in the last infinite series is less than 1 since z
satisfies |z − z0 | < |z1 − z0 |. By the ratio test, the infinite series
∞
|z − z0 |n
,
|z − z0 |
n=N+1 1

converges for |z − z0 | < |z1 − z0 |. By the comparison test, the infinite

.
power series ∞n=0 an (z − z0 ) converges absolutely inside the circle |z − z0 | <
n

|z1 − z0 |.

The above theorem states that the infinite power series converges absolutely
at all points inside some circle centered at z0 provided that it converges at some
point other than z0 . One then deduces that there is a largest circle centered
at z0 such that the series converges absolutely for all points inside that circle.
This circle is termed the circle of convergence of the series. The series fails to
converge at any point outside the circle of convergence. Otherwise, the series
converges at all points inside a new circle that is centered at z0 and passes
through an outside point. This violates the assumption that the earlier circle is
 5.2 Sequences and series of complex functions 207

the defined circle of convergence since there exists a larger circle with absolute
convergence of the series at all points inside it.
.
In summary, given an infinite power series ∞ n=0 an (z − z0 ) , there exists
n

a non-negative real number R, where R can be zero or infinity, such that the
series converges absolutely for |z − z0 | < R and diverges for |z − z0 | > R.
Here, R is called the radius of convergence and the circle |z − z0 | = R is
called the circle of convergence. The power series may or may not converge
at a point on the circle of convergence. Convergence of the power series
must be determined for each point on the circle of convergence (see Exam-
ple 5.2.7).
The radius of convergence of a power series can be found by the following
formulas:
 
 an 
(i) R = lim    if the limit exists.
n→∞ an+1 

(ii) R = lim 1 √n |a | if the limit exists; in general, we always have R =

n→∞ n
1√
limn→∞ n
|an |
.

These formulas are derived directly from the ratio test and root test, respectively.
Applications of the above formulas in finding the radius of convergence of a
given infinite power series are illustrated in Examples 5.2.5 and 5.2.6.
.∞ k
Example 5.2.5 Suppose the radii of convergence of k=0 ak z and
.∞ k
b
k=0 k z are R a and Rb , respectively. Find the corresponding radii of conver-
gence of the following series:
∞
 ∞

(a) (ak + bk )zk , (b) ak bk zk .
k=0 k=0

Solution
(a) Let R = min(Ra , Rb ). Inside the domain |z| < R, the two limits

n 
n
lim ak zk and lim bk zk
n→∞ n→∞
k=0 k=0

exist; so the summed infinite series


n
lim (ak + bk )zk
n→∞
k=0

also exists and the radius of convergence of the summed series is at least
R. The following example shows that the radius of convergence of a
208 Taylor and Laurent Series

summed series can be greater than R. Consider

1 1 1
ak = and bk = − k.
2k 3k 2
. .∞
The series ∞ k
k=0 ak z and
k
k=0 bk z have the same radius of conver-
gence, namely 2. However, their sum
∞ ∞  
∞
1 1 1 zk
(ak + bk )z =
k
+ − z k
=
k=0 k=0
2k 3k 2k k=0
3k

has the radius of convergence 3. This is greater than the minimum of the
two radii of convergence of the constituent series.
.
(b) Let R be the radius of convergence of ∞ k
k=0 ak bk z . Then
  
1


1 1
= lim k |ak bk | ≤ lim k |ak | lim k |bk | = · ,
R k→∞ k→∞ k→∞ Ra Rb
so R ≥ Ra Rb .

Example 5.2.6 Find the circle of convergence of each of the following power
series:
∞
 ∞
 ∞  

1 z k
(a) (z − i)k , (b) k ln k (z − 2)k , (c) ,
k=1
k k=1 k=1
k
∞  k 2 ∞

1 k
(d) 1+ zk , (e) (−1)k z2 .
k=1
k k=1

Solution
(a) By the ratio test, we have
1
R = lim k
= 1;
k→∞ 1
k+1

so the circle of convergence is |z − i| = 1.

(b) Using the root test, we have
1
√
= lim k |ak | = lim k ln k .
k

R k→∞ k→∞

To evaluate the limit, we consider the logarithm

√ √ (ln k)2
k ln k = lim ln k ln k = lim = 0;
k k
ln lim
k→∞ k→∞ k→∞ k
 5.2 Sequences and series of complex functions 209

so
1 √
= lim k ln k = e0 = 1.
k

R k→∞
The circle of convergence is |z − 2| = 1.
(c) By the ratio test, we have
 1 k+1
1
= lim k+1  1 k
R k→∞
k
1 1
= lim lim  
k→∞ k + 1 k→∞ 1 + 1 k
k
1
= 0 · = 0;
e
so the circle of convergence is the whole complex plane.
(d) By the root test, we have

 2  
1 k 1 k 1 k
= lim 1+ = lim 1 + = e,
R k→∞ k k→∞ k
so the circle of convergence is |z| = 1e .
(e) The coefficients are of the form

0 if m = 2k
am = ;
(−1)k if m = 2k
so
1

= lim m |am | = 1.
R m→∞

The circle of convergence is then found to be |z| = 1.

Example 5.2.7 Consider the infinite power series

∞
zn z z2
= 1 + + + ··· , p ≥ 0;
n=0
np 1p 2p

find its circle of convergence. Show that

(a) when p > 1, the series converges for all points on the circle of conver-
gence;
(b) when p = 0, the series diverges for all points on the circle of conver-
gence;
210 Taylor and Laurent Series

(c) when p = 1, the series converges for all points on the circle of conver-
gence except one point.

Solution We apply the root test and consider

 1/n  1/n
1 1
= e−p n ,
ln n
|an |1/n = =
np ep ln n
and as
ln n
lim =0
n→∞ n
we have
lim |an |1/n = 1.
n→∞

The circle of convergence is |z| = 1, for all values of p.

(a) When p > 1, at any point on the circle of convergence, we have
 n
z 
 = 1 .
 np  np

Now, the infinite power series is dominated by the convergent series

. .∞ zn
1+ ∞ 1
n=1 np . Hence, the infinite series n=0 np converges absolutely
at all points on the circle of convergence.
(b) When p = 0, the infinite power series becomes the geometric series
.∞ n
n=0 z . At any point on the circle of convergence, |z| = 1 = 0, so the
n

series diverges.
(c) When p = 1, the infinite power series becomes
∞
 zn
1+ .
n=1
n

At any point z on the circle of convergence, we may write z = eiθ . It can

be shown that (see Problem 5.13)
∞  
cos nθ θ
= − ln 2 sin , θ = 0;
n=1
n 2
∞
 sin nθ π −θ
= , 0 < θ < 2π.
n=1
n 2
.
At z = 1, the infinite series becomes the harmonic series 1 + ∞ 1
n=1 n ,
which is known to be divergent. Therefore, the series converges at all
points on the circle of convergence except at the point z = 1.
 5.2 Sequences and series of complex functions 211

Uniform convergence of infinite power series

We have observed the absolute convergence property of an infinite power series
.∞
n=0 an (z − z0 ) at all points inside its circle of convergence |z − z0 | < R.
n

Does an infinite power series also observe uniform convergence at all points
inside the circle of convergence? It turns out that at best we can only establish
uniform convergence in any closed disk |z − z0 | ≤ r0 < R that lies inside the
circle of convergence (see Theorem 5.2.2).
To motivate the concept of uniform convergence of infinite power series, let
.
us consider uniform convergence of the infinite geometric series ∞ n
n=0 z for
any closed disk |z| ≤ r0 < 1 inside its circle of convergence |z| < 1. To apply
the M test, it suffices to find an appropriate choice of Mn such that |zn | is
bounded by Mn . Since z lies inside the closed disk |z| ≤ r0 < 1, we have

|z|n ≤ r0n < 1;

so we choose Mn = r0n . The associated series

∞
 ∞

Mn = r0n
n=0 n=0

is known to be convergent for r0 < 1. Hence, uniform convergence of the

infinite geometric series is established for |z| ≤ r0 < 1.
Interestingly, though we can establish absolute convergence in the whole
open disk |z| < 1, uniform convergence fails in the same open disk. Let us
consider the remainder after n terms, which is found to be
∞
 zn+1
Rn (z) = zk = .
k=n+1
1−z

To establish absolute convergence of the infinite geometric series in the open

disk |z| < 1, we follow a similar procedure to Example 5.2.2. In order that
 n+1 
z 
|Rn (z)| =   < , z = 0,
1 − z
we have to choose n > N , where
 
ln  ln |1 − z|
N (; z) = f l + −1 .
ln |z| ln |z|
Note that N (; z) has dependence on z and it becomes infinite when |z| → 1,
so uniform convergence fails.
Alternatively, we may demonstrate the failure of uniform convergence in the
open disk |z| < 1 from another perspective. Take any point z = reiθ inside the
212 Taylor and Laurent Series

open disk |z| < 1, where r < 1. By virtue of the relation

|1 − z| ≤ 1 + |z| = 1 + r
we have
1 1
≥ ;
|1 − z| 1+r
we then obtain
|z|n+1 r n+1
|Rn (z)| = ≥ , r < 1.
|1 − z| 1+r
n+1
As r tends to 1 from below, r1+r tends to 12 . Uniform convergence requires that
.
the partial sum Sn (z) = ∞k=0 z converges to the sum S(z) = 1−z for all z in
k 1

the open disk |z| < 1 simultaneously. This is equivalent to requiring

max |S(z) − Sn (z)| = max |Rn (z)| → 0
|z|<1 |z|<1

for all z in |z| < 1. Since this condition fails, uniform convergence of the infinite
geometric series is not observed in the open disk |z| < 1.
.
Theorem 5.2.2 Consider the infinite power series ∞ n=0 an (z − z0 ) , whose
n

circle of convergence is |z − z0 | < R, 0 < R < ∞. The series is uniformly

convergent to its pointwise sum function in the closed disk |z − z0 | ≤ r0 , where
0 < r0 < R.

Proof For |z − z0 | ≤ r0 , we have

|an (z − z0 )n | ≤ |an |r0n , for all n.
By Theorem 5.2.1, the infinite power series converges absolutely so
.∞
|an |r0n converges. Applying the M test, we take Mn = |an |r0n . Since
.n=0
∞
n=0 Mn converges, the power series converges uniformly in any closed disk
|z − z0 | ≤ r0 , where 0 < r0 < R.

As a consequence of the uniform convergence property of an infinite power

series in any closed disk centered at z0 that lies inside the circle of convergence,
we can deduce the termwise integration property of an infinite power series
inside its circle of convergence. For any contour C lying inside the circle of
.
convergence, ∞ n=0 an (z − z0 ) is a uniformly convergent series of continuous
n

functions on C, so
  ∞ ∞ 
an (z − z0 )n dz = an (z − z0 )n dz. (5.2.11)
C n=0 n=0 C
 5.2 Sequences and series of complex functions 213

For example, consider the infinite geometric series

 ∞
1
= zn , |z| < 1,
1−z n=0

we consider the termwise integration along a contour C starting from 0 and

ending at z inside the circle of convergence. Since 1−z1
and zn , n = 0, 1, . . . ,
are analytic functions inside |z| < 1, the integrals of these functions along C
are path independent. We then obtain
  ∞ 
1
dζ = ζ n dζ
C 1−ζ n=0 C
 z  ∞  z
1
⇔ dζ = ζ n dζ
0 1−ζ n=0 0

∞
zn+1
⇔ −Log(1 − z) = , |z| < 1, (5.2.12)
n=0
n+1

where Log 1 = 0.
A similar statement can be made for termwise differentiation of an infi-
nite power series inside its circle of convergence. Suppose that the infinite
.
power series ∞ n=0 an (z − z0 ) converges absolutely
n
to S(z) inside its cir-
.
cle of convergence, where |z − z0 | < R; then ∞ n=1 na n (z − z0 )n−1 converges

absolutely to S (z) inside the same circle of convergence. The validity of
termwise differentiation stems from (i) analyticity of the power functions
an (z − z0 )n , n = 0, 1, . . . , inside the circle of convergence, and (ii) uniform
.
convergence of ∞ n=0 an (z − z0 ) .
n
in any closed disk centered at z0 inside the
circle of convergence. Let S(z) = ∞ n=0 an (z − z0 ) ; then S(z) is analytic inside
n

|z − z0 | < R and its derivative is given by [see eq. (5.2.10)]

∞  ∞
 d  
S (z) = an (z − z0 ) =
n
nan (z − z0 )n−1 , |z − z0 | < R.
dz n=0 n=1
(5.2.13)
Hence, an infinite power series defines an analytic function inside its circle of
convergence.
For example, consider the Bessel function of order n as defined by

(−1)k  z 2k+n
∞

Jn (z) = . (5.2.14)
k=0
k!(k + n)! 2
214 Taylor and Laurent Series

The radius of convergence R of the above infinite power series is found to be

 
1  ak+1 
= lim  
R k→∞  ak 
 
 22k+n k!(k + n)! 

= lim  2(k+1)+n 
k→∞ 2 (k + 1)!(k + 1 + n)! 
 
1 1 

= lim  2  = 0.
k→∞ 2 (k + 1)(k + 1 + n) 

Hence, the infinite power series converges absolutely for all z in the whole
(unextended) complex plane. The power series defines an analytic function
Jn (z) for all z, so Jn (z) is entire. The derivative of Jn (z) is given by termwise
differentiation of the infinite power series, where
(−1)k d  z 2k+n
∞
Jn (z) =
k=0
k!(k + n)! dz 2

1 (−1)k (2k + n)  z 2k+n−1

∞

= . (5.2.15)
k=0
2 k!(k + n)! 2

Power series in its Taylor series representation

Suppose a power series represents a function f (z) inside the circle of conver-
gence; that is,
∞

f (z) = an (z − z0 )n . (5.2.16)
n=0

Since it can be differentiated termwise, we have

∞

f  (z) = nan (z − z0 )n−1 ,
n=1
∞

f  (z) = n(n − 1)an (z − z0 )n−2 ,
n=2

etc. By putting z = z0 successively in the original series and in the above

derivative series, we obtain the relations
f (n) (z0 )
an = , n = 0, 1, 2, . . . .
n!
The corresponding series representation
∞
 f (n) (z0 )
f (z) = (z − z0 )n (5.2.17)
n=0
n!
 5.3 Taylor series 215

is called the Taylor series of f (z). The special case z0 = 0 is called the Maclau-
rin series of f (z).
Suppose a function can be represented by a Taylor series; then implicitly
it is differentiable of infinite order. This is not surprising since every Taylor
series with a nonzero radius of convergence defines an analytic function and an
analytic function is differentiable at all orders. If a Taylor series converges at
every point inside the circle of convergence R, then it converges to a function
that is analytic at least in R.

Cauchy product
If the two power series

∞

f (z) = an (z − z0 )n
n=0

and

∞

g(z) = bn (z − z0 )n
n=0

are convergent for |z − z0 | < R, then the sum and difference,

∞

f (z) ± g(z) = (an ± bn )(z − z0 )n , (5.2.18)
n=0

and the product,

∞
 
n
f (z)g(z) = αn (z − z0 )n , where αn = ak bn−k , (5.2.19)
n=0 k=0

are convergent inside the same circle of convergence. The new series
.∞
n=0 αn (z − z0 ) is called the Cauchy product of f (z) and g(z).
n

5.3 Taylor series

We have seen that an infinite power series defines an analytic function inside
its circle of convergence. In this section, we show how an analytic function can
216 Taylor and Laurent Series

Figure 5.1. The circle C1 : |z − z0 | = R1 lies completely inside the domain D and
contains the point z. Also, R is the minimum distance from z0 to the boundary of the
domain.

be expanded in an infinite power series. Indeed, the class of analytic functions

is closely related to the class of convergent power series.

Theorem 5.3.1 (Taylor series theorem) Let f (z) be analytic in a domain D

with boundary ∂D and z0 ∈ D. We determine R such that

R = min {|z − z0 |, z ∈ ∂D} .

Then there exists an infinite power series

∞

ak (z − z0 )k
k=0

which converges to f (z) for |z − z0 | < R. The coefficients ak are given by

,
1 f (ζ ) f (k) (z0 )
ak = dζ = , k = 0, 1, . . . , (5.3.1)
2π i C (ζ − z0 ) k+1 k!
where C is any simple closed contour enclosing z0 and lying completely inside
D. The power series is said to be the Taylor series of f (z) expanded at z0 .

Proof We take a point z inside the circle N (z0 ; R) and denote |z − z0 | by r so

that r < R. A circle C1 is drawn around z0 with radius R1 , where r < R1 < R
(see Figure 5.1). Since the point z also lies inside C1 , the Cauchy integral
formula gives
,
1 f (ζ )
f (z) = dζ .
2π i C1 ζ − z
 5.3 Taylor series 217

By observing the relation |z − z0 | < |ζ − z0 |, where ζ is any point on C1 , we

may perform the following binomial expansion:
1 1 1
=
ζ −z ζ − z0 1 − ζz−z
−z0
0

  n+1 
 n z−z0
1  z − z0 z − z0 ζ −z0 
= 1 + + ··· + + .
ζ − z0 ζ − z0 ζ − z0 1− z−z0
ζ −z0

(ζ )
Next, we multiply both sides of the above equation by f2πi and perform the
contour integration along C1 . Using the property
,
1 f (ζ ) f (k) (z0 )
dζ = (5.3.2)
2π i C1 (ζ − z0 )k+1 k!
and collecting the terms, we obtain

n
f (k) (z0 )
f (z) = (z − z0 )k + Rn ,
k=0
k!

where the remainder Rn can be expressed as

,  
1 f (ζ ) z − z0 n+1
Rn = dζ. (5.3.3)
2π i C1 ζ − z ζ − z0
To complete the proof, it suffices to show that

lim Rn = 0 .
n→∞

We estimate |Rn | using the modulus inequality from Section 4.1. The arc length
along the integration path is given by L = 2π R1 . Since f (z) is continuous
inside D, its modulus is bounded by some constant M on C1 , that is,

|f (ζ )| ≤ M, ζ ∈ C1 .

Note that

|z − z0 | = r and |ζ − z0 | = R1 .

Moreover, we have

|ζ − z| = |(ζ − z0 ) − (z − z0 )| ≥ |ζ − z0 | − |z − z0 | = R1 − r,

so
    n+1
 f (ζ )   z − z0 n+1 M r
   ≤ .
ζ − z ζ − z  R1 − r R1
0
218 Taylor and Laurent Series

These results are combined to give

   n+1  n+1
 1  M r MR1 r
|Rn | ≤   2π R = .
2π i  R1 − r R1
1
R1 − r R1

It is then obvious that |Rn | vanishes as n → ∞ since Rr1 < 1.

From eq. (5.3.2), the Taylor coefficients ak are given by
,
1 f (ζ )
ak = dζ, k = 0, 1, . . . . (5.3.4)
2π i C1 (ζ − z0 )k+1

Since the integrand in the above integral is analytic in the domain D except at
z0 , by virtue of the Cauchy–Goursat integral theorem, the integration path C1
can be replaced by any simple closed contour C enclosing z0 and lying entirely
inside D. This completes the proof.

Remark The Taylor theorem states that a complex function can be expanded
in an infinite power series at a point around which the function is analytic. In
fact, any infinite power series expansion of an analytic function f (z) must be
its Taylor series [see eq. (5.2.17)]. In other words, the expansion of f (z) in a
Taylor power series at a given point is unique.

Example 5.3.1 Let f (z) = 1

(1+z)2
. Use the relation

f (z)(1 + z)2 = 1

to compute the Taylor series of f (z) at z = 0 and find the radius of convergence
of the power series.

Solution If we assume the Taylor series expansion

∞

f (z) = ak zk ,
k=0

then
∞

(1 + z)2 f (z) = a0 + (2a0 + a1 )z + (ak−2 + 2ak−1 + ak )zk = 1.
k=2

By equating coefficients of like power terms on both sides, we deduce that

a0 = 1, a1 = −2, a2 = 3, a3 = −4, . . . .
 5.3 Taylor series 219

The Taylor series of f (z) at z = 0 is then found to be

∞

f (z) = (−1)k (k + 1)zk .
k=0

The radius of convergence R of the Taylor series can be found by the ratio test,
and is shown to be
 
 (−1)k (k + 1) 
R = lim    = 1.
k→∞ (−1)k+1 (k + 2) 

As expected, the radius of convergence equals the distance from z = 0 to the

point z = −1, where the function is not analytic.

Remark The Taylor series expansion of the generalized power function

(1 + z)α , where α is complex, is considered in Problem 5.17.

Example 5.3.2 Find the Maclaurin series up to z3 for each of the following
functions:

(a) sin−1 z (principal value),

z
(b) ee .

Solution
(a) Let f (z) = sin−1 z; then z = sin f (z). Differentiating the equation with
respect to z repeatedly, we obtain

1 = f  (z) cos f (z),

0 = f  (z) cos f (z) − [f  (z)]2 sin f (z),
0 = f  (z) cos f (z) − 3f  (z)f  (z) sin f (z) − [f  (z)]3 cos f (z).

Putting z = 0 in the above relations, we have

f (0) = 0 (principal value), f  (0) = 1, f  (0) = 0, f  (0) = 1.

The first four terms of the Maclaurin series are found to be

f  (0) 2 f  (0) 3
sin−1 z = f (0) + f  (0)z + z + z + ···
2! 3!
z3
=z+ + ··· .
3!
220 Taylor and Laurent Series
z
(b) Writing f (z) = ee , the successive higher-order derivatives of f (z) are
found to be

f  (z) = ez ee ,
z

f  (z) = ez ee + (ez )2 ee ,
z z

f  (z) = ez ee + 3(ez )2 ee + (ez )3 ee , etc.

z z z

Putting z = 0 into the above relations, we obtain

f (0) = e, f  (0) = e, f  (0) = 2e, f  (0) = 5e.

Therefore, the Maclaurin series is found to be

f  (0) 2 f  (0) 3
ee = f (0) + f  (0)z +
z
z + z + ···
2! 3!
5e
= e + ez + ez2 + z3 + · · · .
6

Example 5.3.3 By using an appropriate auxiliary analytic function, expand

each of the following real functions:

(a) y = ex cos x in a Taylor series of the real variable x,

(b) y = ecos x cos(sin x) in a Fourier cosine series.

Solution
(a) Consider the entire function ez and let z = (1 + i)x, x being real. We
take the real part of the function and obtain
∞
 (1 + i)n x n
ex cos x = Re ez = Re e(1+i)x = Re
n=0
n!
∞
 nπ x n
= 2n/2 cos .
n=0
4 n!

(b) First, we consider the expansion

∞
 ∞
 ∞

eiθ einθ cos nθ sin nθ
e = = +i .
n=0
n! n=0
n! n=0
n!

By taking the real parts of both sides of the above equation, we obtain
∞

iθ cos nθ
Re ee = Re ecos x+i sin x = ecos x cos(sin x) = .
n=0
n!

The expanded series is in the form of a Fourier cosine series.

 5.4 Laurent series 221

Example 5.3.4 The generating function of the Bernoulli numbers Bn is given

by
 Bn∞
z 1
= zn where B0 = 1, B1 = − , B2n+1 = 0, n ≥ 1.
e −1
z
n=0
n! 2

Show that the Maclaurin series for tan z is given by

∞
22n (22n − 1)
tan z = (−1)n−1 B2n z2n−1 .
n=1
(2n)!

What is the radius of convergence of the series?

Solution We consider
(e − e−iz )
1 iz
tan z = 2i
(e + e−iz )
1 iz
2
2i 4i
= − −i
e2iz − 1 e4iz − 1
%∞ & %∞ &
 Bn  Bn
= n n−1
(2i) z − n n−1
(4i) z −i
n=0
n! n=0
n!
∞
 Bn
= [(−2i)B1 − i] + [(2i)n − (4i)n ] zn−1 .
n=2
n!

By observing that B1 = − 12 and B2n+1 = 0, n ≥ 1, the above expression can

be simplified to give
∞
22n (22n − 1)
tan z = (−1)n−1 B2n z2n−1 .
n=1
(2n)!

The tangent function is not analytic at the zeros of cos z, namely, z =

± π2 , ± 3π
2
, . . . . The distance from z = 0 to the nearest singular point is π2 ,
so the radius of convergence of the above Maclaurin series is deduced to
be π2 .

5.4 Laurent series

The Taylor series expansion represents an analytic function inside its circle of
convergence. It is common to encounter functions which are analytic in some
punctured domains. In these cases, the Taylor series representation is not the
correct form for describing the infinite power series expansion of these types
of complex function.
222 Taylor and Laurent Series

Let us consider an infinite series with negative powers of the form

∞

b1 (z − z0 )−1 + b2 (z − z0 )−2 + · · · = bn (z − z0 )−n . (5.4.1)
n=1

To find the region of convergence of this series, we set w = z−z

1
0
. The series in
eq. (5.4.1) then becomes a Taylor series in w. By the ratio test, supposing
 
 bn 

R = lim   
n→∞ bn+1 

exists, then the new series converges for all w such that |w| < R  . In terms of
z, the region of convergence is the region outside the circle |z − z0 | = R1 .
More generally, we consider an infinite series with positive and negative
power terms of the form
∞
 ∞

an (z − z0 )n + bn (z − z0 )−n .
n=0 n=1

This is called a Laurent series expanded at z = z0 . The second summation term

with negative power terms is called the principal part of the Laurent series.
Suppose the limits
 a   b 
 n   n 
R = lim   and R  = lim  
n→∞ an+1 n→∞ bn+1
* +
exist, and RR  > 1; then inside the annular domain z : R1 < |z − z0 | < R ,
the Laurent series is convergent. The annulus may degenerate into a hollow
plane if R = ∞, or a punctured disk if R  = ∞. When RR  ≤ 1, such an
annular region of convergence does not exist.
A Laurent series defines a function f (z) in its annular region of conver-
gence. Conversely, the Laurent series theorem states that a function defined
and analytic in an annulus can be expanded in a Laurent series.

Theorem 5.4.1 (Laurent series theorem) Let f (z) be analytic in the annulus
A : R1 < |z − z0 | < R2 ; then f (z) can be represented by the Laurent series,
∞

f (z) = ck (z − z0 )k , (5.4.2)
k=−∞

which converges to f (z) throughout the annulus. The Laurent coefficients are
given by
,
1 f (ζ )
ck = dζ, k = 0, ±1, ±2, . . . , (5.4.3)
2π i C (ζ − z0 )k+1
 5.4 Laurent series 223

Figure 5.2. The function f (z) is analytic inside the annulus A: R1 < |z − z0 | < R2 . We
choose two positively oriented circular paths γ and , centered at z0 and with radii r1
and r2 , respectively. The point z lies within the two concentric circles γ and .

where C is any simple closed contour lying completely inside the annulus and
going around the point z0 .

Proof Take any point z ∈ A and write r = |z − z0 |. Choose two positively

oriented circles  and γ with radii r2 and r1 , respectively, where R1 < r1 <
r < r2 < R2 (see Figure 5.2). Since the annular region bounded by  and γ is
doubly connected, the application of the Cauchy integral formula to this region
gives

, ,
1 f (ζ ) 1 f (ζ )
f (z) = dζ − dζ .
2π i  ζ −z 2π i γ ζ −z

For the first integral, we perform exactly the same computation as in the proof
of the Taylor series theorem to obtain

,  ∞
1 f (ζ )
dζ = ak (z − z0 )k ,
2π i  ζ −z k=0

where
,
1 f (ζ )
ak = dζ, k = 0, 1, . . . . (5.4.4)
2π i  (ζ − z0 )k+1
224 Taylor and Laurent Series

For the second integral, since |ζ − z0 | < |z − z0 | for ζ ∈ γ , we write

1
−
ζ −z
1 1
= −z0
(z − z0 ) 1 − ζz−z
 0
 n 
 n−1 ζ −z0
1  ζ − z0 ζ − z0 z−z0 
= 1 + + ··· + + ζ −z0 .
z − z0 z − z0 z − z0 1− z−z0

(ζ )
Multiplying both sides of the above equation by f2πi and integrating along the
circle γ , we obtain
,
1 f (ζ )
− dζ
2π i γ ζ − z
,

1 1
= f (ζ ) dζ
2π i γ z − z0
,

1 1 n ,
+··· + f (ζ )(ζ − z0 )n−1 dζ +R
2π i γ (z − z0 )n
where
,  n
n = 1 f (ζ ) ζ − z0
R dζ.
2π i γ z−ζ z − z0
n → 0 as n → ∞. Let
It remains to be shown that R
Mγ = max|f (ζ )|,
ζ ∈γ

and note that for ζ lying on γ , we have

|z − ζ | = |(z − z0 ) − (ζ − z0 )| ≥ |z − z0 | − |ζ − z0 | = r − r1 .
The modulus of the remainder R n is bounded by
   r n
 
n | ≤  1  Mγ 1 r 1  r1  n
|R  2π i  r − r 2π r1 = M γ ,
1 r r − r1 r
which tends to zero as n → ∞ since r1
r
< 1. Once the result
n = 0
lim R
n→∞

is established, we then have

, ∞
1 f (ζ )
− dζ = bk (z − z0 )−k ,
2π i γ ζ − z k=1
 5.4 Laurent series 225

where
,
1
bk = f (ζ )(ζ − z0 )k−1 dζ, k = 1, 2, . . . . (5.4.5)
2π i γ

Since the integrand functions in eqs. (5.4.4) and (5.4.5) are analytic inside
the annulus A, we may replace both circles  and γ by any simple closed
contour C lying entirely inside A and enclosing the point z0 . The two results
can be combined to give the Laurent series expansion as defined in eqs. (5.4.2)
and (5.4.3).

Remark
(i) Suppose f (z) is analytic in the full disk |z − z0 | < R2 without the
punctured hole (the same assumption on the domain of analyticity of
f (z) as in the Taylor series theorem); then the integrand in eq. (5.4.5)
becomes analytic inside |z − z0 | < R2 . By the Cauchy–Goursat the-
orem, we then have bk = 0, k = 1, 2, . . . . This is expected since the
Laurent series should reduce to a Taylor series.
(ii) As revealed by later examples, we normally do not find the Laurent
series by computing the Laurent coefficients using eq. (5.4.3). The Lau-
rent series can be found by any method, and by virtue of the uniqueness
property of the Laurent series expansion, the Laurent series obtained
using different methods would all agree.
(iii) Some of the Laurent series coefficients may be related to an integral
whose value is desired. For example, when k = −1 in eq. (5.4.3), we
have
,
2π ic−1 = f (ζ ) dζ . (5.4.6)
C

Example 5.4.1 Find all the possible Laurent series of each of the following
functions at the given point α:

1 1
(a) , α = −1; (b) , α = i.
1 − z2 z3

Solution
2 is not analytic at z = −1 and z = 1. There are only two
1
(a) The function 1−z
possible annular domains centered at z = −1 inside which the function
is analytic throughout. These two annular domains are 0 < |z + 1| < 2
226 Taylor and Laurent Series

and |z + 1| > 2. Recall that the binomial expansion

 ∞
1
= ζk
1−ζ k=0

is  provided that |ζ | < 1. In the first annular1 domain, we have

valid
 z+1
 < 1. This motivates us to express 1 2 as and perform
2 1−z 2(z+1)(1− z+1
2 )
1
the binomial expansion of 1− z+1 accordingly. We then have
2

1 1
=  
1−z 2
2(z + 1) 1 − z+1
2
∞ ∞
1 (z + 1)k (z + 1)k−1
= = .
2(z + 1) k=0 2k k=0
2k+1

Note that there is only one negative power term in the Laurent
expansion.
In the second annular domain |z + 1| > 2, we have |z+1|
2
< 1; so the
Laurent expansion can be found as follows:

1 1
=−  
1−z 2
(z + 1) 1 −
2 2
z+1
∞
 ∞
1 2k 2k
=− = − .
(z + 1) k=0 (z + 1)
2 k
k=0
(z + 1)k+2

Here, the Laurent expansion has an infinite number of negative power

terms but no positive power term.
(b) The function z13 is not analytic at z = 0. We seek annular domains,
centered at z = i, such that the function is analytic throughout. Two
such domains are found to be |z − i| < 1 and |z − i| > 1. To expand the
function z13 in both domains, we make use of the binomial expansion

1 (−3)(−4) 2 (−3)(−4)(−5) 3
= 1 + (−3)ζ + ζ + ζ + ···
(1 + ζ )3 2! 3!
∞
k(k − 1) k−2
= (−1)k ζ ,
k=2
2
 5.4 Laurent series 227

which holds for |ζ | < 1. By setting ζ = z−i

i
in the above binomial expan-
sion, we obtain
1 1
=  
z3 z−i 3
i 1+
3
i
∞
1  (z − i)k−2
= (−1) k
k(k − 1)
2i 3 k=2 i k−2
= i − 3(z − i) − 6i(z − i)2 + 10(z − i)3 + · · · .
 
The series converges in the region z − i < 1, which is precisely the first
domain. Since |z − i| < 1 is a solid disk, the power series expansion in
this domain is actually a Taylor series with no negative
 i power
 term.
In the second domain |z − i| > 1, the property  z−i < 1 holds. We
now set ζ = z−ii
; the corresponding Laurent expansion inside |z − i| > 1
is found to be
1 1
=  
z3 i 3
(z − i) 1 +
3
z−i
∞

1 i k−2
= (−1)k k(k − 1)
2(z − i) k=2
3 (z − i)k−2
∞
1  i k−2
= (−1)k k(k − 1)
2 k=2 (z − i)k+1
1 3i 6 10i
= − − + + ··· .
(z − i)3 (z − i)4 (z − i)5 (z − i)6

Example 5.4.2 Find the Laurent series for

1
f (z) =
ez − e2z
expanded at z = 0 that is convergent in the annulus 0 < |z| < 2π .

Solution The given function is not analytic at the zeros of ez − e2z . These
zeros are given by z = 2kπ i, where k is any integer. To perform the Laurent
expansion of the function at z = 0, we choose an annular domain centered at
z = 0 inside which the function is analytic. One such possible choice is the
annulus 0 < |z| < 2π. First, we decompose f (z) into
 
−z 1 z
f (z) = e − .
z ez − 1
228 Taylor and Laurent Series
z
The Taylor series expansions of e−z and are
ez −1
∞
zn
e−z = (−1)n , valid for |z| < ∞,
n=0
n!
 Bn∞
z
= zn , valid for |z| < 2π,
ez − 1 n=0
n!

where Bn are the Bernoulli numbers (see Example 5.3.4). Combining the
results, the Laurent series is found to be
∞
 ∞

(−1)n Bn
f (z) = zn − zn−1
n=0
n! n=0
n!
∞


1 (−1)n Bn+1
=− + − zn ,
z n=0 n! (n + 1)!

where B0 = 1. The Laurent series has only one negative power term and it is
seen to be convergent in the annulus 0 < |z| < 2π .

Example 5.4.3 Find the Laurent expansion of the function

1
f (z) =
z−k
that is valid inside the domain |z| > |k|, where k is real and |k| < 1. Using the
Laurent expansion, deduce that
∞
 k cos θ − k 2
k n cos nθ = ,
n=1
1 − 2k cos θ + k 2
∞
k sin θ
k n sin nθ = .
n=1
1 − 2k cos θ + k 2

Solution For |z| > |k|, by performing the appropriate binomial expansion,
the Laurent expansion of f (z) at z = 0 is found to be
   
1 1 1 k k2 kn k 
=  = 1 + + 2 + ··· + n + ··· for   < 1.
z−k z 1− z k z z z z z
Since |k| < 1, the unit circle |z| = 1 lies inside the region of convergence of
the above Laurent series. Therefore, the series obtained by substituting z = eiθ
into the above Laurent series is guaranteed to be convergent. We then have
1 1
= iθ (1 + ke−iθ + k 2 e−2iθ + · · · + k n e−inθ + · · · ).
eiθ − k e
 5.4 Laurent series 229

Rearranging the terms, we obtain

eiθ (e−iθ − k) k cos θ − k 2 − ik sin θ
−iθ
−1=
(eiθ− k) (e − k) 1 − 2k cos θ + k 2
∞
 ∞

= k n cos nθ − i k n sin nθ.
n=1 n=1

By equating the real and imaginary parts of the above expressions, we obtain
the required results.

Example 5.4.4 The Bessel functions of integer order are defined by the Lau-
rent series
∞

z 1
e 2 (w− w ) = Jn (z)w n .
n=−∞
z
w− w1
The function e ( 2 ) is called the generating function of the Bessel function.
Show that

(−1)k  z n+2k
2π ∞

1
Jn (z) = cos(nt − z sin t) dt = ,
2π 0 k=0
(n + k)!k! 2
n = 0, ±1, ±2, . . . ,
and deduce the relation
Jn (z) = (−1)n J−n (z).

Solution The generating function is analytic in the annulus 0 < |w| < ∞.
The Bessel functions are the Laurent coefficients of the generating function.
Using eq. (5.4.3), we have
,
1 z 1 dw
Jn (z) = e 2 (w− w ) n+1 , n = 0, ±1, ±2, . . . ,
2π i C w
where C is any simple closed contour lying completely inside the annulus of
z 1
analyticity of e 2 (w− w ) and enclosing the origin. Here, we choose C to be the
unit circle |w| = 1. We parametrize the unit circle by w = eit , 0 ≤ t < 2π .
The above integral for Jn (z) then becomes
 2π
1
Jn (z) = i eiz sin t e−int dt
2π i 0
 2π  2π
1 i
= cos(nt − z sin t) dt − sin(nt − z sin t) dt,
2π 0 2π 0
n = 0, ±1, ±2, . . . .
230 Taylor and Laurent Series

Setting t = 2π − τ in the second integral, we have

 2π  2π
sin(nt − z sin t) dt = − sin(nτ − z sin τ ) dτ ;
0 0

so the second integral is seen to have the value zero. Hence, the integral
representation of Jn (z) reduces to
 2π
1
Jn (z) = cos(nt − z sin t) dt, n = 0, ±1, ±2, . . . .
2π 0
To find the Laurent series expansion of Jn (z), we consider the multiplication of
the series for e 2 and e− 2w using the Cauchy product formula [see eq. (5.2.19)].
zw z

This gives
%∞ &% ∞ &
z 1
 1  z n  (−1)n  z n 1
e2 ( w− w ) = w n

n=0
n! 2 n=0
n! 2 wn
%∞ &
∞  (−1)k  z n+2k
= wn .
n=−∞ k=0
(n + k)!k! 2

By equating like power terms of wn in the Laurent expansion, we obtain

(−1)k  z n+2k
∞

Jn (z) = , n = 0, ±1, ±2, . . . .
k=0
(n + k)!k! 2

Lastly, by swapping w with − w1 in the defining relation for Jn (z), we have

∞
   ∞
z 1 1 n
e 2 (w− w ) = Jn (z) − = (−1)n J−n (z)w n ;
n=−∞
w n=−∞

so

Jn (z) = (−1)n J−n (z).

5.4.1 Potential flow past an obstacle

Consider a potential flow with uniform upstream velocity U∞ past an obstacle
(see Figure 5.3). Assuming that the origin is inside the obstacle, the complex
velocity V (z) of the potential flow can be represented by the following Laurent
series expansion in z:
c−1 c−2 c−n
V (z) = U∞ + + 2 + ··· + n + ··· , (5.4.7)
z z z
 5.4 Laurent series 231

Figure 5.3. Plot of the streamlines that represent the potential flow with uniform
upstream velocity U∞ past an obstacle.

where U∞ , c−1 , c−2 , . . . , c−n are complex numbers. To substantiate the claim,
we observe that V (z) is analytic in the domain outside the obstacle, and at
points far from the obstacle, |z| → ∞, V (z) tends to the uniform velocity U∞ .
The coefficient c−1 has a special physical interpretation. The contour integral
of the complex velocity around the body defined by the closed curve γ is equal
to Cγ + iFγ , where Cγ and Fγ are the circulation and flux around the body,
respectively [see eqs. (4.4.8) and (4.4.12)]. On the other hand, from eq. (5.4.6),
the Laurent coefficient c−1 and the contour integral around γ are related by
,
2π ic−1 = V (z) dz. (5.4.8)
γ

Combining the results, we then have

2π ic−1 = Cγ + iFγ . (5.4.9)

If the circulation and flux around the body are both zero, then c−1 = 0.
The determination of the other coefficients, c−2 , . . . , c−n , . . . , depends on
the configuration of the obstacle. As an illustrative example, we consider the
potential flow past the perturbed circle r = 1 −  sin2 θ , where  is a small
quantity. To find the flow past the perturbed circle, it is more convenient to
formulate the problem in terms of the stream function ψ(r, θ ). Let the uniform
upstream velocity be parallel to the x-axis and have unit magnitude so that the
far-stream conditions are
∂ψ ∂ψ
→1 and → 0 as r → ∞. (5.4.10)
∂y ∂x
232 Taylor and Laurent Series

The surface of the perturbed circle is a streamline and this leads to the following
surface boundary condition:

ψ(1 −  sin2 θ, θ ) = 0. (5.4.11)

The flux across a non-penetrative body should be zero. For simplicity, we

assume the circulation to be zero. Solving the flow problem amounts to finding
a harmonic function ψ(r, θ ) that satisfies the above boundary conditions.
The stream function ψ(r, θ ) takes the form
sin θ sin 2θ sin nθ
ψ(r, θ ) = r sin θ + a1 + a2 2 + · · · + an n + · · · , (5.4.12)
r r r
since each term is harmonic and the far-stream conditions are satisfied auto-
matically. We expect the coefficients a1 , a2 , . . . , an , . . . to depend on the small
parameter  in the problem. When  = 0, the perturbed circle becomes the per-
fect unit circle. The stream function for the flow past the unit circle is known
to be (see Example 4.4.2)
 
1
ψ(r, θ ;  = 0) = r − sin θ. (5.4.13)
r
Alternatively, we may write the stream function as a perturbation series in
powers of :

ψ(r, θ ) = ψ0 (r, θ ) + ψ1 (r, θ ) +  2 ψ2 (r, θ ) + · · · +  n ψn (r, θ ) + · · · .

When  is small, the first few terms of the perturbation series may already give
a very good approximation to the true solution.
We now determine the successive order terms ψ0 , ψ1 , ψ2 , . . . , sequentially.
The zeroth-order solution corresponds to  = 0, and it has been given in
eq. (5.4.13). To find the higher-order solutions, we apply the boundary condition
(5.4.11). When expressed in perturbation expansion, it becomes

ψ0 (1 −  sin2 θ, θ ) + ψ1 (1 −  sin2 θ, θ )

+  2 ψ2 (1 −  sin2 θ, θ ) + · · · = 0. (5.4.14)

The perturbation parameter  appears explicitly in front of the successive order

terms and also implicitly in the arguments of the functions. Next, we expand
the above series in powers of  and equate like power terms of  to generate
the successive boundary conditions for ψ1 , ψ2 , . . . , etc. To determine ψ1 , we
keep only the linear terms in  and obtain
∂ψ0
ψ0 (1, θ ) −  sin2 θ (1, θ ) + ψ1 (1, θ ) + · · · = 0.
∂r
 5.5 Analytic continuation 233

Hence ψ1 (ψ, θ ) has to satisfy the boundary condition

∂ψ0 3 sin θ − sin 3θ
ψ1 (1, θ ) = sin2 θ (1, θ ) = 2 sin2 θ sin θ = .
∂r 2
To satisfy the above boundary condition, the harmonic function ψ1 (r, θ ) is
found to be
3 sin θ 1 sin 3θ
ψ1 (r, θ ) = − . (5.4.15)
2 r 2 r3
By equating like power terms of O( 2 ) in the expansion of eq. (5.4.14), we
obtain the following boundary condition for ψ2 (r, θ ):
sin4 θ ∂ 2 ψ0 ∂ψ1
(1, θ ) − sin2 θ (1, θ ) + ψ2 (1, θ ) = 0.
2 ∂r 2 ∂r
Simplifying the trigonometric terms and rearranging, the boundary condition
can be expressed as
7 13 5
ψ2 (1, θ ) = − sin θ + sin 3θ − sin 5θ. (5.4.16)
8 16 16
The corresponding harmonic function ψ2 (r, θ ) that satisfies the above boundary
condition is found to be
7 sin θ 13 sin 3θ 5 sin 5θ
ψ2 (r, θ ) = − + 3
− . (5.4.17)
8 r 16 r 16 r 5
The solution procedure can be routinely applied to find the higher-order solu-
tions, though the algebraic manipulations become more daunting. Fortunately,
there exists an ingenious algorithm for computing the coefficients of the terms
in the higher-order solutions, the details of which are shown in Problem 5.33.

5.5 Analytic continuation

Suppose we are given an infinite power series or an analytic formula which
defines an analytic function f in a domain D. It is natural to ask whether one
can extend its domain of analyticity. More precisely, can we find a function F
which is analytic in a larger domain and whose values agree with those of f
for points in D? Here, F is called an analytic continuation of f . For example,
the complex exponential function ez is the analytic continuation of the real
exponential function ex defined over the real interval (−∞, ∞). The complex
function ez is analytic in the finite complex plane and ez = ex when z = x, x
being real.
More generally, suppose f1 is analytic in a domain D1 and f2 is analytic in
another domain D2 . If D1 ∩ D2 = ∅ and f1 (z) = f2 (z) in the common inter-
section D1 ∩ D2 , then f2 is said to be the analytic continuation of f1 to D2 and
234 Taylor and Laurent Series

f1 is the analytic continuation of f2 to D1 . Now, the function F defined by


f1 (z) when z ∈ D1
F(z) = (5.5.1)
f2 (z) when z ∈ D2

is analytic in the union D1 ∪ D2 . The function F is the analytic continuation

to D1 ∪ D2 of either f1 or f2 ; and f1 and f2 are called elements of F.

Example 5.5.1 Show that the two functions

∞

f (z) = a k zk = 1 + az + a 2 z2 + · · ·
k=0

and
∞
 (−1)k (1 − a)k zk 1 (1 − a)z (1 − a)2 z2
g(z) = = − + + ···
k=0
(1 − z)k+1 1−z (1 − z)2 (1 − z)3

are analytic continuations of each other.

Solution Consider the function

1
w(z) =
1 − az
which is analytic everywhere except at z = a1 . The Taylor expansion of w(z)
about z = 0 in the domain |z| < |a|
1
is given by

 ∞
1
= 1 + az + a 2 z2 + · · · = a k zk .
1 − az k=0

The above Taylor series is precisely f (z). Suppose we write w(z) in the alter-
native form
1 1 1
= .
1 − az 1 − z 1 + (1−a)z
1−z

Provided that
 
 (1 − a)z 
 
 1 − z  < 1,

w(z) can be expressed in the form

1 1 (1 − a)z (1 − a)2 z2
= − + + ··· ,
1 − az 1−z (1 − z)2 (1 − z)3
 5.5 Analytic continuation 235

which is precisely g(z). Both f (z) and g(z) are elements of the analytic function
w(z). The respective regions of convergence of f (z) and g(z) are given by
1
|z| < and |(1 − a)z| < |1 − z|,
|a|
which overlap with each other. Actually, it can be easily shown that both
domains contain a circle centered around z = 0 with a finite radius. Therefore,
f (z) and g(z) are analytic continuations of each other.

Example 5.5.2 The two power series

z2 z3
z+ + + ···
2 3
and
1 1
iπ − (z − 2) + (z − 2)2 − (z − 2)3 + · · ·
2 3
have no common region of convergence. However, they are analytic continua-
tions of the same function. Find the function and explain why.

Solution Let the function f1 (z) be defined by

z2 z3
f1 (z) = z + + + ··· .
2 3
The infinite power series converges in the domain |z| < 1. In fact, it is the
Maclaurin series expansion of the function

g(z) = −Log(1 − z).

Similarly, the second series

1 1
f2 (z) = iπ − (z − 2) + (z − 2)2 − (z − 2)3 + · · ·
2 3
is seen to be the Taylor series expansion of

h(z) = iπ − Log(z − 1)

at z = 2. The expansion is valid for |z − 2| < 1. Note that the two regions
of convergence, |z| < 1 and |z − 2| < 1, are disjoint. However, when Im(z −
1) > 0, h(z) = g(z); also, both of the above regions overlap with the domain
Im(z − 1) > 0. Hence

−Log(1 − z), {z : Im(z − 1) > 0},

236 Taylor and Laurent Series

is a direct continuation of both f1 (z) and f2 (z). Therefore, the two series are
each an analytic continuation of the other.

Theorem 5.5.1 If a function f is analytic in a domain D and f (z) = 0 at all

points on an arc inside D, then f (z) = 0 throughout D.

Proof We take any point z0 on the arc where f (z) = 0. Inside some circle C
that is centered at z0 and lying completely inside D (the circle can be extended
at least to the boundary of D), f (z) can be expanded in a Taylor expansion in
the form
f  (z0 )
f (z) = f (z0 ) + f  (z0 )(z − z0 ) + (z − z0 )2 + · · · .
2!
Since f (z) = 0 for all points on the arc that lies inside the circle C, this would
imply that

f (z0 ) = f  (z0 ) = f  (z0 ) = · · · = 0 .

Therefore, f (z) = 0 for all points inside the circle C. Continuing the process
with another arc that lies completely inside C and expanding f (z) in a region
bounded by another circle, f (z) can be similarly shown to be zero for all points
inside the new circle. Eventually, we can find a sufficient number of circles to
cover the whole domain D. Since f (z) = 0 inside all these circles, f (z) = 0
throughout D.

Corollary Let the two functions f1 (z) and f2 (z) be analytic in some domain
D. Suppose f1 (z) = f2 (z) on an arc inside D; then f1 (z) = f2 (z) throughout
the domain D. The validity of the corollary can be revealed by writing f (z) =
f1 (z) − f2 (z) and applying the theorem. For an interesting application of this
corollary, see Problem 5.38.

5.5.1 Reflection principle

Some functions like z + 1 and cos z possess the property that z2 + 1 = z2 + 1
2

and cos z = cos z. However, other functions, like z2 + i and i cos z, do not
observe the property f (z) = f (z); that is, reflection of z with respect to the
real axis does not correspond to reflection of f (z) with respect to the real axis.
The following theorem states precisely the condition under which f (z) = f (z)
holds.
 5.5 Analytic continuation 237

Theorem 5.5.2 (Reflection principle) Let the function f be analytic in some

domain D, which contains a segment of the real axis and is symmetric with
respect to the real axis. For any point z in D, the property

f (z) = f (z) (5.5.2)

holds if and only if f (z) assumes real values on the real axis contained in D.

Proof Writing f (z) = u(x, y) + iv(x, y), z = x + iy, we have

f (z) = u(x, −y) − iv(x, −y) .

Note that f (z) is well defined for all z ∈ D since D is symmetric with respect
to the real axis; so for any z ∈ D, we have z ∈ D. Suppose f (z) = f (z); we
then have u(x, −y) − iv(x, −y) = u(x, y) + iv(x, y). When z is real, we put
y = 0 in the above relation and obtain

u(x, 0) − iv(x, 0) = u(x, 0) + iv(x, 0) , (5.5.3)

which implies that v(x, 0) = 0. Hence, f (z) is real when z is real.

Conversely, given that f (z) assumes real values on the segment of the real
axis within D, from eq. (5.5.3) we deduce that f (z) and f (z) have the same
values along the segment of the real axis. To prove f (z) = f (z) for all z ∈ D,
it suffices to show that f (z) is analytic in D. This is because once f (z) and
f (z) are known to be analytic in the same domain D and they share the same
values on a segment in D, then by the corollary to Theorem 5.5.1, we have
f (z) = f (z) throughout D.
The real and imaginary parts of f (z) are u(x, −y) and −v(x, −y), respec-
tively. Since f (z) is analytic in D, both u(x, y) and v(x, y) have continuous
first-order partial derivatives in D; correspondingly, u(x, −y) and −v(x, −y)
share the same continuity properties. The next step is to show that both u(x, −y)
and −v(x, −y) satisfy the Cauchy–Riemann relations. From the analyticity of
f (z), we have
∂u ∂v ∂u ∂v
(x, y) = (x, y) and (x, y) = − (x, y),
∂x ∂y ∂y ∂x
from which we can deduce that
∂u ∂(−v) ∂u ∂(−v)
(x, ỹ) = (x, ỹ) and (x, ỹ) = − (x, ỹ),
∂x ∂ ỹ ∂ ỹ ∂x
where ỹ = −y. These are precisely the Cauchy–Riemann relations for the real
and imaginary parts of f (z). Hence, the analyticity of f (z) in D is established.
238 Taylor and Laurent Series

5.6 Problems
5.1. Let zn = xn + iyn , n = 1, 2, . . . .
. .∞
(a) Given that ∞ n=1 |zn | converges, show that n=1 zn also converges.
That is, an absolutely convergent series is always convergent.
. .∞
(b) Show that ∞ n=1 zn converges absolutely if and only if both n=1 xn
.∞
and n=1 yn converge absolutely.

Hint: Use the comparison test to establish the convergence of the two
real series
∞
 ∞

|xn | and |yn |.
n=1 n=1

5.2. Suppose z1 , z2 , . . . , zn , . . . all lie in the right half-plane Re z ≥ 0, and

. .∞ 2 .∞
both series ∞ n=1 zn and n=1 zn converge; show that n=1 |zn | also
2

converges.
5.3. Suppose z1 , z2 , . . . , zn , all lie within the sector −α ≤ Arg z ≤ α, 0 <
α < π2 ; show that the two series
∞
 ∞

zn and |zn |
n=1 n=1

both either converge or diverge.

Hint: Observe that |zn | ≤ Re zn

cos α
.

5.4. Examine the convergence of the following series:

∞ ∞ ∞

cos in 2n i(2n + i) n
(a) ; (b) √ ; (c) .
n=1
3n n=1
n + in n=1
5n

5.5. Discuss the convergence of

∞

(zn+1 − zn ).
n=0

5.6. Show that the series

∞
 z2
n=0
(1 + z2 )n

does not converge uniformly when z assumes values along the real axis.
 5.6 Problems 239
.∞
5.7. Let R > 0 be the radius of convergence of the series n=0 an zn . Show
.
that the radius of convergence of the series ∞ (Re a )zn is at least R.
.∞ n n
n=0
5.8. Suppose the radius of convergence of the series n=0 an z is R; find the
corresponding radius of convergence of each of the following series:
∞
 ∞
 ∞
 ∞

an
(a) n10 an zn ; (b) (2n − 1)an zn ; (c) zn ; (d) ank zn .
n=0 n=0 n=0
n! n=0

√ the following examples: an = α n!;

n
Hint: For part (c), consider
an = (n!) ; an = n!.
2

5.9. Show that the series

∞
 zn
n=1
1 + z2n

converges in both the domains |z| < 1 and |z| > 1. Discuss the conver-
gence of the series on the unit circle |z| = 1.
5.10. Consider the series of complex functions
∞
 z2
.
k=1
(1 + |z|2 )k

The partial sum Sn (z) is defined by


n
z2
Sn (z) = .
k=0
(1 + |z|2 )k

(a) Show that

z2
S(z) = lim Sn (z) = (1 + |z|2 ),
n→∞ |z|2
valid for |z| > 0.
(b) Show that the convergence of Sn (z) to S(z) inside |z| < 1 as n → ∞
is not uniform convergence.
5.11. Find the radii of convergence of the following series:
∞
 ∞
 ∞

2 n2 n
(a) n q z , |q| < 1; (b) n!
z ; (c) [3 + (−1)n ]n zn ;
n=0 n=0 n=0
∞
n! n
(d) z .
n=1
nn
240 Taylor and Laurent Series

5.12 Suppose
an+1
L = lim
n→∞ an
exists; show that the following three power series have the same radius
of convergence.
∞
 ∞ ∞

an n+1
(a) an zn ; (b) z ; (c) nan zn−1 .
n=0 n=0
n+1 n=0

5.13. Show that

∞
 zn
= −Log(1 − z) for z ∈ D = {z : |z| < 1}.
n=1
n
.
Explain why the series ∞ einθ
n=1 n , θ = 0, is conditionally convergent.
What happens when θ = 0? Use the above series to show that
∞  
cos nθ θ
(a) = − ln 2 sin ;
n=1
n 2
∞
 sin nθ π −θ
(b) = , 0 < θ < 2π.
n=1
n 2

Hint: Note that −Log(1 − z) is analytic inside the domain D and

d 1
[−Log(1 − z)] = .
dz 1−z

5.14. Expand both of the following functions in Maclaurin series up to the z4

term:

(a) e−z sin z; (b) Log(1 + ez ).

5.15. Find the Maclaurin series of the following functions:

1−z sin z
(a) ; (b) ; (c) sinh(z3 );
1 +
 z z + z 2 + z3 1 + z 2
1 − cos ζ
(d) dζ.
0 ζ
5.16. Suppose the Maclaurin series expansion of sec z is given by
∞
 (−1)k E2k E2 2 E4 4
sec z = z2k = E0 − z + z − ··· + ··· .
k=0
(2k)! 2! 4!
 5.6 Problems 241

Find the values for E0 , E2 , E4 , E6 . These numbers are called the Euler
numbers. Find the circle of convergence of the above series.

Hint: Consider
1 1
sec z = = z2 z4
cos z 1− 2!
+ 4!
− ···
so that
  
E2 2 E4 4 z2 z4
1 = E0 − z + z − ··· + ··· 1− + − ··· .
2! 4! 2! 4!

5.17. Consider the principal branch of the generalized power function (1 + z)α
for α complex; show that its Taylor series expansion at z = 0 is given
by
α(α − 1) 2 α(α − 1)(α − 2) 3
(1 + z)α = 1 + αz + z + z
2! 3!
α(α − 1) · · · (α − n + 1) n
+··· + z + ··· .
n!
Find the corresponding circle of convergence of the above series.

Hint: Suppose we write the function as

f (z) = eαLog(1+z) , f (0) = 1;

show that
(1 + z)f  (z) = αf (z).

5.18. Show that the coefficient cn of the Taylor series expansion

 ∞
1
= cn zn
1−z−z 2
n=0

satisfies the recurrence relation

cn = cn−1 + cn−2 , n ≥ 2.
Find the general form for cn and the radius of convergence of the series.
5.19. Find the Taylor series expansion of
 z
1
f (z) = sin−1 z =
dξ.
0 1 − ξ2
Determine the circle of convergence.
242 Taylor and Laurent Series

Hint: Consider
∞
1 (2n)! 2n
√ = z , valid for |z| < 1.
1 − z2 n=0
2 (n!)2
2n

5.20. Suppose the rational polynomial

a0 + a1 z + a2 z2
b0 + b1 z + b2 z2 + b3 z3
.
is expanded in a Taylor series in the form ∞ n
n=0 cn z ; find c0 , c1 , c2 and
the recurrence relation between cn , cn−1 , cn−2 and cn−3 , n ≥ 3.
5.21. Suppose f (z) admits a power series of the form
∞

f (z) = cn zn ;
n=0

prove that
(a) the coefficients of the odd powers of z vanish if f (z) is even;
(b) the coefficients of the even powers of z vanish if f (z) is odd.
1
5.22. The function 1+x 2 is differentiable of all orders for real values of x, but

its Taylor series expansion at x = 0

1
= 1 − x 2 + x 4 − · · · + (−1)n x 2n + · · · ,
1 + x2
converges only for |x| < 1. Explain why.
.
5.23. Supposing f (z) = ∞ n=0 cn z is analytic inside |z| < R, show that
n

 2π ∞
1
|f (reiθ )|2 dθ = |cn |2 r 2n , 0 < r < R.
2π 0 n=0

5.24. Consider the following Taylor series

∞
 zn+1
.
n=1
n(n + 1)

Show that the circle of convergence is |z| < 1, and the series converges
to the sum function
S(z) = (1 − z)Log (1 − z) + z,
where Log 1 = 0. At z = 1, show that the corresponding series
∞
 1
n=1
n(n + 1)
 5.6 Problems 243

converges but S(z) fails to be analytic. Quote another example of a Taylor

series where the series is divergent at a point on the circle of convergence
but the sum function is analytic at that point.
.
5.25. Suppose f (z) = ∞ n=0 cn z
n
is analytic inside the domain |z| < 1,
together with Re f (z) ≥ 0 and f (0) = 1; show that
1 − |z| 1 + |z|
(a) |cn | ≤ 2, n ≥ 0; (b) ≤ |f (z)| ≤ .
1 + |z| 1 − |z|
5.26. Suppose the radii of convergence of the series
∞
 ∞

f (z) = an zn and g(z) = bn zn
n=0 n=0

are greater than 1. Show that

 2π ∞
1 −iθ
f (e )g(e ) dθ =
iθ
an bn .
2π 0 n=0

5.27. A function F (t, z) is called the generating function of the sequence of

complex functions {fn (z)} if F (t, z) admits a Taylor series expansion in
powers of t of the form
∞

F (t, z) = fn (z)t n for |t| < R.
n=0

Some useful properties of the sequence of functions {fn (z)} can be

derived via its generating function. Consider the generating function
of the Bernoulli polynomials Bn (z) defined by
∞
etz − 1  Bn (z) n
t = t .
et − 1 n=1
n!
Prove the following properties of the Bernoulli polynomials:
(a) Bn (z + 1) − Bn (z) = nzn−1 ;
(b) if m is a natural number, then
Bn+1 (m)
= 1 + 2n + 3n + · · · + (m − 1)n ;
n+1
n−1  
 n
(c) Bn (z) = Bk zn−k , where Bk are the Bernoulli numbers
k=0
k
defined in Example 5.3.4.
5.28. Find all the possible Laurent series expansions at the given point z0 of
each of the following functions:
244 Taylor and Laurent Series
1
(a) f (z) = , z0 = 0;
(z − 1)(z − 2)
1
(b) f (z) = 2 , z0 = i.
z (z − i)
5.29. For each of the following functions, find the Laurent series at z0 which
is convergent within the annulus A as indicated:
1
(a) where
(z − i)(z − 2)
(i) z0 = 0, A = {z : 1 < |z| < 2};
(ii) z0 = 0, A = {z : |z| > 2}; √
(iii) z0 = i, A = {z : 0 < |z − i| < 5}.
Then deduce the value of
,
1
dz,
C (z − i)(z − 2)
where C is a simple closed contour enclosing i but excluding 2.
1
(b) where z0 = 1 and A = {z : 1 < |z − 1| < 2}.
z(z − 1)
2
1
(c) ez+ z where z0 = 0 and A = C\{0}.
1
Hint: Find the Cauchy product of ez and e z .

5.30. Find the Laurent series expansion of

z
f (z) = sin ,
z−1
valid for the annulus: 0 < |z − 1| < ∞.

Hint: Consider
 
z 1 1 1
sin = sin 1 + = sin 1 cos + cos 1 sin .
z−1 z−1 z−1 z−1

5.31. Show that the coefficient cn in the Laurent expansion at z = 0 of

 
1
f (z) = cos z +
z
is given by
 2π
1
cn = cos(2 cos θ ) cos nθ dθ, n = 0, ±1, ±2, . . . .
2π 0
 5.6 Problems 245

Deduce the corresponding Laurent expansion at z = 0 of

 
1
g(z) = cosh z + .
z
5.32. Show that the function represented by the Laurent series
∞
1
n!z n
n=1

is analytic everywhere except at z = 0. Evaluate the contour integral of

the above function around the unit circle |z| = 1 in the positive sense.
5.33. Consider a uniform horizontal potential flow past the perturbed circle
r = 1 −  sin2 θ . We would like to solve for the stream function ψ(r, θ )
that is harmonic outside the perturbed circle and satisfies the boundary
conditions
∂ψ ∂ψ
(a) → 1 and → 0 as r → ∞,
∂y ∂x
(b) ψ(1 −  sin2 θ, θ ) = 0.

The perturbation series expansion of ψ(r, θ ) in powers of the param-

eter  is given by

ψ(r, θ ) = ψ0 (r, θ ) + ψ1 (r, θ ) +  2 ψ2 (r, θ ) + · · · .

Explain why the nth order solution ψn (r, θ ) takes the form

n
sin(2j + 1)θ
ψn (r, θ ) = Anj , n ≥ 1.
j =0
r 2j +1

Suppose we write
∞

(1 −  sin2 θ )−(2j +1) = Cj k sin2k θ ;
k=0
246 Taylor and Laurent Series

show that the surface boundary condition can be expressed in the form

ψ(1 −  sin2 θ, θ ) = (1 −  sin2 θ ) sin θ

∞
% ∞  ∞ &
  
j +
+ sin(2j + 1)θ Cj k sin θ2k
 Aj +,j .
j =0 k=0 =0

Using the trigonometric relation

j +k

sin(2j + 1)θ sin2k θ = Bmj k sin(2m + 1)θ,
m=0

show that the recurrence relations for the coefficients Anj are given by
n−j j +k

n 
n−1  
Anj Cj0 sin(2j + 1)θ = − Cj k An−k,j Bmj k sin(2m + 1)θ.
j =0 j =0 k=1 m=0

Use the above relations to show that

1 sin θ 9 sin 3θ 21 sin 5θ 7 sin 7θ
ψ3 (r, θ ) = − − + − .
16 r 16 r 3 32 r 5 32 r 7
z z
5.34. The two functions 1−z
and z−1
have the following series expansions:
z
= z + z2 + z3 + · · ·
1−z
z 1 1 1
= 1 + + 2 + 3 + ··· .
z−1 z z z
By adding them together, we obtain
z z 1 1 1
0= + = · · · + 3 + 2 + + 1 + z + z2 + z3 + · · · .
1−z z−1 z z z
What is wrong with the above argument?
1
5.35. (a) Show that z(z+1) is an analytic continuation of

∞  
1
f (z) = (−1) 1 − k+1 (z − 1)k
k

k=0
2
1 3 7
= − (z − 1) + (z − 1)2 − · · · + · · · .
2 4 8
1
(b) Show that z2
is an analytic continuation of
∞

f (z) = (k + 1)(z + 1)k .
k=0
 5.6 Problems 247

5.36. Show that the following two series are analytic continuations of each
other:
1 1
f (z) = z − z2 + z3 − · · · + · · ·
2 3
1 − z 1 (1 − z)2 1 (1 − z)3
g(z) = ln 2 − − 2
− − ··· .
2 2 2 3 23
5.37. For each of the following pairs of functions, show that they form analytic
continuations:
∞

(a) f1 (z) = zn , |z| < 1 and f2 (z) = 1
1−z
for Re z < 12 ;
n=0
(b) f1 (z) = Log z and f2 (z) = ln |z| + i arg z for 0 ≤ arg z < 2π .
5.38. Given that
sin2 x + cos2 x = 1
is valid for all real values x; using the corollary to Theorem 5.5.1, explain
why
sin2 z + cos2 z = 1
is valid for any complex number z in the whole (unextended) complex
plane.
5.39. Suppose
f (z) = f (z)
in a domain D which is symmetric about the real axis; show that the
coefficients of the Laurent series expansion of f (z) expanded about the
origin are real.
5.40. A function f (z) is analytic inside the annulus r < |z| < 1r , r < 1, and
satisfies
 
1
f = f (z).
z
Let the Laurent series expansion of f (z) at z = 0 be expressed in the
form
∞

f (z) = ak zk .
k=−∞

Show that ak = a−k and f (z) is real on the unit circle |z| = 1.
 6
Singularities and Calculus of Residues

This chapter begins with a discussion of the classification of isolated singu-

larities of complex functions. The classification can be done effectively by
examining the Laurent series expansion of a complex function in a deleted
neighborhood around an isolated singularity. An isolated singularity can be
either a pole, a removable singularity or an essential singularity. The various
forms of behavior of a complex function near an isolated singularity are exam-
ined. Next, we introduce the definition of the residue of a complex function at
an isolated singularity. We show how to apply residue calculus to the evaluation
of different types of integral. The Fourier transform and Fourier integrals are
considered, and the effective use of residue calculus for the analytic evaluation
of these integrals is illustrated. The concept of the Cauchy principal value of
an improper integral is introduced. We also consider the application of residue
calculus to solving fluid flow problems.

6.1 Classification of singular points

By definition, a singularity or a singular point of a function f (z) is a point at
which f (z) is not analytic. A point at which f (z) is analytic is called a regular
point of f (z). A point z0 is called an isolated singularity of f (z) if there
exists a neighborhood of z0 inside which z0 is the only singular point of f (z).
For example, ±i are isolated singularities of f (z) = z21+1 . More generally, the
P (z)
singularities of a rational function Q(z) are the zeros of Q(z) and they are all
isolated. The singularities of cosec z are all isolated and they are simply the
zeros of sin z, namely, z = kπ, k is any integer.
Certainly we may have non-isolated singularities. For example, every point
on the negative real axis (the branch cut of Log z) is a non-isolated singularity
of Log z. The function f (z) = z is nowhere analytic, so every point in the

248
 6.1 Classification of singular points 249

complex plane is a non-isolated singularity. An interesting example is the

function f (z) = cosec πz : all the points z = n1 , n = ±1, ±2, . . ., are isolated
singularities. Interestingly, the origin is a non-isolated singularity since every
neighborhood of the origin contains other singularities.
In this section, we confine our discussion to isolated singularities only.
Suppose z0 is an isolated singularity; then there exists a positive number r
such that f (z) is analytic inside the deleted neighborhood 0 < |z − z0 | < r.
This forms an annular domain within which the Laurent series theorem is
applicable. Suppose f (z) has the Laurent series expansion at z0 in the deleted
neighborhood of the form
∞ ∞

f (z) = an (z − z0 )n + bn (z − z0 )−n , 0 < |z − z0 | < r; (6.1.1)
n=0 n=1
the isolated singularity z0 is classified according to the principal part of the
Laurent series.

Removable singularity
In this case, the principal part vanishes altogether and the Laurent series is
essentially a Taylor series. The series represents an analytic function in the
solid disk |z − z0 | < r. As there is no negative power term in the Laurent
series, the limit lim f (z) exists and is equal to a0 . The singularity z0 is said to
z→z0
be removable since we can remove this singularity z0 by defining f (z0 ) = a0 .
For example, the function sinz z is undefined at z = 0. The Laurent series of
sin z
z
in a deleted neighborhood of z = 0 is given by
sin z z2 z4 z6
f (z) = =1− + − + ··· ,
z 3! 5! 7!
where the Laurent series has no negative power term. The singularity of f (z) =
sin z
z
at z = 0 can be removed by defining f (0) = 1.

Essential singularity
Here, the principal part has infinitely many nonzero terms. For example, z = 0
is an essential singularity of the function z2 e1/z . Inside the annular domain
0 < |z| < ∞, the Laurent series of z2 e1/z is found to be
1 1 1 1 1
f (z) = z2 + z + + + + ··· .
2! 3! z 4! z2

Pole of order k
In this case, the principal part has only a finite number of non-vanishing terms
and the last non-vanishing coefficient is bk ; that is, the Laurent series in the
250 Singularities and Calculus of Residues

deleted neighborhood of z0 takes the form

∞

f (z) = an (z − z0 )n + b1 (z − z0 )−1
n=0

+ b2 (z − z0 )−2 + · · · + bk (z − z0 )−k , bk = 0. (6.1.2)

It is called a simple pole when k = 1. For example, z21+1 has simple poles at
z = i and z = −i; sinz3 z has a pole of order 2 at z = 0 since

sin z 1 1 z2
3
= 2− + − ··· + ··· , 0 < |z| < ∞.
z z 3! 5!
The above three cases are mutually exclusive, that is, an isolated singularity
must be either a removable singularity, an essential singularity or a pole. A
complex function is said to be meromorphic if it is analytic everywhere in the
finite plane except at isolated poles. A meromorphic function has no essential
singularities in the finite plane, though it may have an essential singularity at
infinity. The number of poles in the finite complex plane can be infinite. An
example of a meromorphic function is tan z, where the isolated poles (infinitely
many) are the zeros of cos z.
It may be instructive to examine the behavior of f (z) around an isolated
singularity at z = z0 :

(i) When z0 is a removable singularity, lim f (z) exists and |f (z)| is

z→z0
bounded near z0 .
(ii) When z0 is a pole of finite order, it is obvious that

lim f (z) = ∞.
z→z0

(iii) When z0 is an essential singularity, f (z) has a complicated behavior

around z0 . In fact, in every neighborhood of an essential singularity z0 ,
the function f (z) comes arbitrarily close to any complex value. A more
precise description of the limiting behavior of f (z) at z0 is given by the
Weierstrass–Casorati theorem.

Theorem 6.1.1 (Weierstrass–Casorati theorem) Let z0 be an isolated

essential singularity of a function f (z) and let λ be any given complex
value. For any positive number , however small, there exists a point z,
distinct from z0 , in every neighborhood of z0 such that

|f (z) − λ| < .
 6.1 Classification of singular points 251

Proof Assuming the contrary, there exist  > 0 and δ > 0 such that

|f (z) − λ| ≥  for 0 < |z − z0 | < δ.

Define
1
g(z) = ;
f (z) − λ
by the above hypothesis, g is analytic inside the deleted δ-neighborhood at z0
and |g(z)| ≤ 1 . We can then conclude that z0 is a removable singularity of g(z).
Accordingly, we define

g(z0 ) = lim g(z).

z→z0

Now, g(z) admits a Taylor series expansion inside the neighborhood |z − z0 | <
δ, where
∞

g(z) = (z − z0 )m ak (z − z0 )k ,
k=m

where m is a finite non-negative integer and am = 0. Consider the following

two cases:

(i) When m = 0, g(z0 ) = a0 = 0 so

1
= f (z) − λ
g(z)
is analytic at z0 , a contradiction to “z0 being an isolated essential singu-
larity of f ”.
1
(ii) When m > 0, g(z) has a pole of finite order m. Again, a similar contra-
diction as in part (i) occurs.

Remark Indeed, there is a stronger result describing the behavior of f near an

isolated essential singularity beyond that of the Weierstrass–Casorati theorem.
The Picard theorem states that f (z) assumes every finite value, with one possible
exception, an infinite number of times in any neighborhood of an isolated
essential singularity.

As an example, we examine the behavior of the function f (z) = e1/z around

its isolated essential singularity at z = 0. We let z approach zero in different
directions. First, if z approaches zero along the positive x-axis, we observe that

lim e1/z = lim+ e1/x = ∞, z = x, x > 0.

z→0 x→0
252 Singularities and Calculus of Residues

However, when z approaches zero along the negative x-axis, we have

lim e1/z = lim− e1/x = 0, z = x, x < 0.

z→0 x→0

In the general case, if z approaches zero along the ray Arg z = θ , then

lim e1/z = lim ecos θ/r e−i sin θ/r , z = reiθ .

z→0 r→0

Given any value of r (equivalently, any neighborhood around z = 0) and any

complex number λ, we can find θ such that ecos θ/r e−i sin θ/r comes arbitrarily
close to λ.

Singularity at complex infinity

The behavior of a complex function f at complex infinity ∞ in theextended 
complex plane C can be reviewed by examining the behavior of f 1z in the
neighborhood of the point 0. For example, the complex exponential function
f (z) = ez is known to be entire. Its behavior at complex infinity can be explored
by considering f 1z = e1/z , which is known to have an isolated essential
singularity at z = 0. Hence, the complex exponential function has an isolated
essential singularity at complex infinity.

Simple method of finding the order of a pole

From eq. (6.1.2), we observe that if z0 is a pole of order k, then

lim (z − z0 )k f (z) = bk , bk = 0.
z→z0

In general, if we multiply f (z) by (z − z0 )m and take the limit z → z0 , we

obtain

 bk m = k
lim (z − z0 )m f (z) = 0 m>k . (6.1.3)
z→z0 
∞ m<k

This formula provides the basis for a simple method to find the
order k of a pole. We start by multiplying f (z) by a factor (z − z0 )m
for some integer m and check the limit of the product (z − z0 )m f (z) as z → z0 .
If m is too low, the limit does not exist; and if m is too high, the limit equals
zero. Only when the correct value m = k is hit does the limit exist and the
resulting limit is a finite nonzero value.
 6.1 Classification of singular points 253

Example 6.1.1 Suppose f (z) has a pole of order k at z0 . Find a polynomial

p(z) such that f (z) − (z−z
p(z)
0)
k is analytic (or at most has a removable singularity)

at z0 .

Solution Let D be a deleted neighborhood of z0 throughout which f (z) is

analytic. Since f (z) has a pole of order k at z0 , the Laurent series expansion is
of the form
∞

f (z) = fn (z − z0 )n , z ∈ D,
n=−k

where fn are the Laurent coefficients. We define the function

p(z)
g(z) = f (z) −
(z − z0 )k
for some polynomial p(z) yet to be determined. Since p(z) is analytic every-
where in the finite complex plane, g(z) is then analytic in the same deleted
neighborhood D. Suppose p(z) is a polynomial of degree N ≥ k − 1. The
Taylor series expansion of p(z) at z0 can be written as


N
p(n) (z0 )
p(z) = (z − z0 )n ,
n=0
n!

where p(n) (z0 ) is the nth-order derivative of p(z) evaluated at z0 . Inside the
deleted neighborhood D, the Laurent series expansion of g(z) at z0 is given by
(k−1)
f−k − p(z0 ) f−k+1 − p (z0 )
(z0 )
f−1 − p(k−1)!
g(z) = + + ··· +
(z − z0 )k (z − z0 )k−1 z − z0

N−k
p (n+k)
(z0 )

∞
+ fn − (z − z0 )n + fn (z − z0 )n .
n=0
(n + k)! n=N−k+1

Suppose p(z) is chosen such that it satisfies the following condition

p(n) (z0 )
= f−k+n , 0 ≤ n ≤ k − 1;
n!
then the principal part of the Laurent series expansion of g(z) in the deleted
neighborhood D vanishes. In other words, any polynomial p(z) of degree
N ≥ k − 1 together with the satisfaction of the above condition would have
a corresponding g(z) that is analytic (or at most has a removable singularity)
at z0 .
254 Singularities and Calculus of Residues

Example 6.1.2 The Laurent series

∞
 ∞
zn
z−n +
n=1 n=0
2n+1

contains infinitely many negative power terms of z. Check whether the point
z = 0 is an essential singular point of the function represented by the series.

Solution The first sum converges to

1
z 1
= for |z| > 1
1− 1
z
z−1

and the second sum converges to

1
1
2
= for |z| < 2.
1− z
2
2−z

Therefore, the Laurent series converges to the function

1 1 1
f (z) = + =− 2
z−1 2−z z − 3z + 2
inside the annulus A: 1 < |z| < 2. Obviously, A is not a deleted neighborhood
of z = 0. It would be incorrect to claim that z = 0 is an essential singularity
of f (z).

Example 6.1.3 The function

2
f (z) = e1/(1−z)

has an essential singularity at z = 1. Discuss the behavior of the function as

z → 1:

(i) along the circumference of the unit circle |z| = 1;

(ii) rectilinearly from the interior of the unit circle |z| = 1.

Solution
(i) On the circumference of the unit circle, we set z = eiθ , −π < θ ≤ π .
Now,
1 1 1 cos θ − i sin θ
=  2 = − ,
(1 − z) 2
−2i sin 2 cos 2 + i sin 2
θ θ θ 4 sin2 θ2
 6.2 Residues and the Residue Theorem 255

so
      
 1 cos θ i θ  1 cos θ
|f (z)| = exp − exp cot  = exp − .
4 sin2 θ2 2 2  4 sin2 θ2

As z → 1, we have θ → 0. We then observe that |f (z)| → 0, so

f (z) → 0.
(ii) We represent the radial rays from the interior of the circle towards z = 1
by z = 1 + reiθ , π2 < |θ | ≤ π . Along these rays,
2
f (z) = e(cos 2θ−i sin 2θ)/r ,

so
2
|f (z)| = ecos 2θ /r .

Next, we consider the following three separate cases. When π2 < |θ| <
3
4
π , |f (z)| → 0 as r → 0, implying that f (z) → 0. On the other hand,
when 34 π < |θ | ≤ π , |f (z)| is unbounded as r → 0, implying that the
limit of f (z) does not exist. In particular, when θ = ± 34 π , we have

f (z) = e∓i/r .
2

As r → 0, the limit of f (z) does not exist.

6.2 Residues and the Residue Theorem

Consider the contour integral
,
f (z) dz;
C

if f (z) has no singularity inside the closed contour C, then the value of the
integral is zero by virtue of the Cauchy–Goursat integral theorem. How do we
deal with the usual case where singularities of f are included inside the closed
contour C? In this section, we illustrate how to apply the method of residues
to the evaluation of the integral without resorting to direct integration.
Let z0 be an isolated singularity of f (z); then there exists a certain deleted
neighborhood N = {z: 0 < |z − z0 | < } such that f is analytic everywhere
inside N . The residue of f at z0 is defined by
,
1
Res(f, z0 ) = f (z) dz, (6.2.1)
2π i C
where C is any simple closed contour around z0 and inside N .
256 Singularities and Calculus of Residues

A simple method of evaluating the residue is to examine the Laurent series of

f (z) at z0 inside N . Suppose f (z) admits the Laurent series expansion inside
N , where

∞
 ∞

f (z) = an (z − z0 )n + bn (z − z0 )−n ; (6.2.2)
n=0 n=1

then by eq. (5.4.6),

,
1
b1 = f (z) dz = Res(f, z0 ). (6.2.3)
2π i C

Sometimes the residue value can be found in a straightforward manner.

Several examples are shown below.

  
1 if k = 1
(i) Res 1
, z0 = .
(z−z0 )k 0 if k = 1
1
(ii) Res(e z , 0) = 1 as deduced from the following Laurent series expansion

1 1 1 1 1 1 1
ez = 1 + + 2
+ + ··· , |z| > 0.
1! z 2! z 3! z3
 
(iii) Res 1
(z−1)(z−2)
,1 = 1
1−2
= −1 by the Cauchy integral formula.

Theorem 6.2.1 (Cauchy residue theorem) Let C be a simple closed contour

inside which f (z) is analytic everywhere except at the isolated singularities
z1 , z2 , . . . , zn (see Figure 6.1). The contour integral of f (z) around C is
given by
,
f (z) dz = 2π i [Res(f, z1 ) + Res(f, z2 ) + · · · + Res(f, zn )]. (6.2.4)
C

The validity of the theorem can be shown easily using Corollary 3 of the
Cauchy–Goursat integral theorem in Section 4.2.
In general the value of the residue of f (z) at an isolated singularity z0 is
computed either by direct integration or by finding the coefficient b1 in the
appropriate Laurent series of f expanded inside the deleted neighborhood N
of z0 . When the isolated singularity z0 is a pole, it is possible to derive some
efficient formulas for computing the residue at the pole.
 6.2 Residues and the Residue Theorem 257

y
C

z1
z2

zn

Figure 6.1. The closed contour C encircles isolated singularities of f (z) at

z1 , z2 , . . . , zn .

6.2.1 Computational formulas for evaluating residues

We now derive several efficient computational formulas for finding the residue
of f (z) at a pole z0 . First, it is necessary to know the exact order of the pole.
This can be found by the method discussed earlier [see eq. (6.1.3)].
Suppose z0 is a pole of order k of f (z). In a deleted neighborhood of z0 , the
Laurent series expansion takes the form
∞

f (z) = an (z − z0 )n + b1 (z − z0 )−1 + · · · + bk (z − z0 )−k , bk = 0.
n=0

Consider the function

g(z) = (z − z0 )k f (z)

whose principal part vanishes altogether. The power series expansion of g(z)
in the same deleted neighborhood of z0 is given by
∞

g(z) = bk + bk−1 (z − z0 ) + · · · + b1 (z − z0 )k−1 + an (z − z0 )n+k .
n=0
(6.2.5)

Now g(z) has a removable singularity at z0 , which can be removed by

defining g(z0 ) = bk . The series in eq. (6.2.5) is essentially the Taylor
series of g(z) inside a neighborhood of z0 . The value of Res(f, z0 ), which
equals b1 , can be seen to be
g (k−1) (z0 ) d k−1 [(z − z0 )k f (z)]
b1 = = lim .
(k − 1)! z→z0 dx k−1 (k − 1)!
258 Singularities and Calculus of Residues

When the pole z0 is a simple pole corresponding to k = 1, the above formula

is reduced to

Res(f, z0 ) = lim (z − z0 )f (z). (6.2.6)

z→z0

Suppose the function f (z) is a quotient of two functions p(z) and

q(z), both of which are analytic at z0 . Further, we assume p(z0 ) = 0 but
q(z0 ) = 0, q  (z0 ) = 0. The limit of (z−zq(z)
0 )p(z)
as z → z0 assumes the inde-
0
terminate form 0 . We then apply the L’Hospital rule to obtain
(z − z0 )p(z) p(z) + (z − z0 )p  (z)
lim = lim
z→z0 q(z) z→z0 q  (z)
p(z0 )
=  = 0.
q (z0 )
Hence, z0 is a simple pole of f (z). To find Res(f, z0 ), we expand p(z) and q(z)
in Taylor series at z0 . Using eq. (6.2.6) and observing that q(z0 ) = 0, we obtain

Res(f, z0 ) = lim (z − z0 )f (z)

z→z0
p(z0 ) + p (z0 )(z − z0 ) + · · ·
= lim (z − z0 )
z→z0 q  (z0 )(z − z0 ) + 12 q  (z0 )(z − z0 )2 + · · ·
p(z0 )
= . (6.2.7)
q  (z0 )

Example 6.2.1 Evaluate the integral

,
z+1
2
dz
|z|=1 z

using

(i) direct line integration,

(ii) the calculus of residues,
1
(iii) the primitive function log z − .
z

Solution
(i) On the unit circle, z = eiθ and dz = ieiθ dθ . We then have
,  2π
z+1
2
dz = (e−iθ + e−2iθ ) ieiθ dθ
|z|=1 z 0
 2π
=i (1 + e−iθ ) dθ = 2π i.
0
 6.2 Residues and the Residue Theorem 259

(ii) The integrand z+1z2

has a double pole at z = 0. The Laurent series expan-
sion in a deleted neighborhood of z = 0 is simply 1z + z12 , where the
coefficient of 1z is seen to be 1. We obtain
 
z+1
Res , 0 = 1,
z2

and so
,  
z+1 z+1
2
dz = 2π i Res , 0 = 2π i.
|z|=1 z z2

(iii) When a closed contour moves around the origin (which is the branch
point of the function log z) in the anticlockwise direction, the increase
in the value of arg z equals 2π . Therefore,
,
z+1
2
dz = change in value of ln |z| + i arg z − 1z in
|z|=1 z
traversing one complete loop around the origin
= 2π i.

Example 6.2.2 Evaluate the following integral

,
tanh z
dz.
|z|=2 z

Solution The isolated singularities of the integrand function tanh

z
z
inside the
closed contour |z| = 2 are z = 0, 2 and − 2 . The singularity at z = 0 is
iπ iπ

removable since

tanh z sinh z 1
lim = lim lim = 1.
z→0 z z→0 z z→0 cosh z

By the Cauchy residue theorem, we then have

,    

tanh z tanh z iπ tanh z iπ

dz = 2π i Res , + Res ,− .
|z|=2 z z 2 z 2

Since z = iπ2 is a simple zero of cosh z, it is a simple pole of tanh

z
z
; similarly,
z=− iπ
is a
 2 z iπ  simple pole of tanh z
z
. We use two different methods to compute
Res tanh
z
, 2
.
260 Singularities and Calculus of Residues

Method one
First, we observe that cosh z admits the following Taylor series at z = iπ
2
:
     
sinh z  iπ sinh z  iπ 3
cosh z = z− + z− + ···
1! z= iπ 2 3! z= iπ 2
% 2
  
2
&  
iπ 1 iπ 3  
=i z− + z− + ··· , z − iπ  < π.
2 3! 2  2 

Since z = iπ2 is a simple pole of tanh z, using eq. (6.2.6), we obtain

   
tanh z iπ iπ tanh z
Res , = lim z −
z 2 z→ iπ2 2 z
 
z − iπ2 sinh z
= lim '   1  3 (
iπ
z→ 2 i z− iπ
+ z− iπ
+ ··· z
2 3! 2
1 2
= iπ
= − i.
2
π

Method two
We write tanh z
as p(z) , where p(z) = sinh z
and q(z) = cosh z; and observe
 iπ  z   q(z) 
 iπ
 z
p 2 = 0, q 2 = 0 and q 2 = 0. Using eq. (6.2.7), we obtain
iπ

    
tanh z iπ p iπ2 1 2
Res , =   iπ  =  = − i.
z 2 q 2 z z= iπ π
2

In a similar manner, we obtain

  
tanh z iπ 1  2
Res ,− =  = i.
z 2 z z=− iπ π
2

The sum of residues at the two isolated singularities is seen to be zero, so

,
tanh z
dz = 0.
|z|=2 z

tanh z
Remark Since z
is an even function, part (b) of Problem 6.15 reveals that
   
tanh z iπ tanh z iπ
Res , = −Res ,− .
z 2 z 2
This is consistent with the result obtained in the above calculations.
 6.2 Residues and the Residue Theorem 261

Example 6.2.3 Suppose f (z) and g(z) are analytic in some domain containing
z = a, and z = a is a double zero of g(z) = 0 and f (a) = 0. Show that the
(z)
residue of fg(z) at z = a is given by
 
f (z) 6f  (a)g  (a) − 2f (a)g  (a)
Res ,a = .
g(z) 3[g  (a)]2

Solution Since z = a is a double zero of g(z) = 0, we then have

g(a) = g  (a) = 0. The Taylor series of f (z) and g(z) expanded in some neigh-
borhood centered at z = a are, respectively,
f  (a)
f (z) = f (a) + f  (a)(z − a) + (z − a)2 + · · · ,
2!

g  (a) g  (a) g  (a)

g(z) = (z − a)2 + (z − a)3 + (z − a)4 + · · · .
2! 3! 4!
Their quotient can be expressed in the form

f (z) f (a) + f  (a)(z − a) + f 2!(a) (z − a)2 + · · ·
= ' (
g(z) g  (a) g  (a) g  (a)
2
(z − a)2 1 + 3g  (a) (z − a) + 12g  (a) (z − a) + · · ·
2
' 
(
f (a) + f  (a) − f (a)g 
3g (a)
(a)
(z − a) + · · ·
= g  (a)
.
2
(z − a)2
1
From the above expansion, we deduce that the coefficient of z−a
in the Laurent
(z)
series expansion of fg(z) at z = a is given by
f (a)g  (a)
f  (a) − 3g  (a)
g  (a)
;
2

therefore,
 
f (z) 6f  (a)g  (a) − 2f (a)g  (a)
Res ,a = .
g(z) 3 [g  (a)]2

Example 6.2.4 Suppose pn (z) is a polynomial of degree n and α is a point

inside the unit circle centered at the origin; show that
,
pn (z)
dz = 0.
|z|=1 z
n+1 (z − α)

Solution Inside the unit circle |z| = 1, the integrand has two poles, namely,
z = 0 as a pole of order n + 1 and z = α as a simple pole. By the Cauchy
262 Singularities and Calculus of Residues

residue theorem, we have

,
pn (z)
dz
|z|=1 z
n+1 (z − α)
   

pn (z) pn (z)
= 2π i Res n+1 , 0 + Res n+1 ,α .
z (z − α) z (z − α)
To evaluate the residue at z = 0, it is necessary to expand the integrand in a
Laurent series inside a deleted neighborhood around z = 0. Suppose we write

pn (z) = a0 + a1 z + · · · + an zn .

The Laurent series expansion of the integrand takes the form

pn (z)
zn+1 (z − α)
   
a0 + a1 z + · · · + an zn 1 z zn
= − 1+ +· · ·+ n +· · · ,
zn+1 α α α
which is valid inside the annular domain 0 < |z| < |α|. The residue at z = 0 is
given by
 
pn (z) 1
Res n+1 , 0 = coefficient of in the above Laurent series
z (z − α) z
  
1 a0 a1
= − + + · · · + an
α αn α n−1
1
= − n+1 (a0 + a1 α + · · · + an α n )
α
1
= − n+1 pn (α).
α
Since z = α is a simple pole, the residue at α is found to be [see eq. (6.2.6)]
 
pn (z) pn (z) pn (α)
Res n+1 , α = lim n+1 = n+1 .
z (z − α) z→α z α
Combining the above results, we obtain
,

pn (z) pn (α) pn (α)

n+1 (z − α)
dz = 2π i − + = 0.
|z|=1 z α n+1 α n+1

Example 6.2.5 Suppose α is a simple pole of a meromorphic function f (z).

Show that the line integral

f (z) dz

 6.2 Residues and the Residue Theorem 263

Figure 6.2. The circular arc  subtends an angle φ around the center α with an infinites-
imal radius r.

equals iφ Res(f (z), α), where  is a circular arc subtended at an angle φ

around the point α with an infinitesimal radius.

Solution Since f (z) has a simple pole at z = α; in a deleted neighborhood

of α, f (z) admits a Laurent series expansion of the form

 ∞
a−1
f (z) = + an (z − α)n .
z − α n=0

On the circular arc around α, z = α + reiθ , φ0 ≤ θ ≤ φ0 + φ, where r is the

radius of the arc (see Figure 6.2). The contour integral becomes
  φ0 +φ
f (z) dz = f (α + reiθ )ireiθ dθ
 φ0
 φ0 +φ  ∞

a−1 
= + n inθ
an r e ireiθ dθ.
φ0 reiθ n=0

On taking the limit r → 0, we obtain

  φ0 +φ
lim f (z) dz = a−1 i dθ = iφa−1 = iφ Res(f (z), α).
r→0  φ0

α
Example 6.2.6 Find the residue of f (z) = (z−i) z
2 , 0 < α < 1, at each of its

isolated singularities, where the principal branch of the power function zα is

chosen.
264 Singularities and Calculus of Residues

z=i

branch cut
x

α
Figure 6.3. The function f (z) = (z−i)
z
2 has only one isolated singularity at z = i. The
branch cut of f (z) is the whole negative real axis, including the origin. All the points
along the branch cut of zα (including the end point z = 0) are non-isolated singularities.

Solution The only isolated singularity of

zα
f (z) = , 0 < α < 1,
(z − i)2
is at z = i; all the points along the branch cut of zα (the whole negative real axis,
including z = 0) are non-isolated singularities (see Figure 6.3). The principal
branch of zα is taken to be eαLog z , where Log z is the principal branch of the
logarithm function. Obviously, z = i is a pole of order 2 since
zα
lim(z − i)2 = lim eαLog z = eiαπ/2 ,
z→i (z − i)2 z→i

which is a finite nonzero value. The residue of f at z = i is then given by

 
zα
Res ,i
(z − i)2 ,
d α
= lim zα = lim eαLog z = αei(α−1)π/2 , 0 < α < 1.
z→i dz z→i z

Example 6.2.7 Let  f (z)  be analytic in a domain containing the whole real
axis, and let zn = n + 12 π, n is any integer. Show that
 
f (z) 
Res , zn = f (zn ), n is any integer.
cos2 z

 
Solution First, we observe that zn = n + 12 π is a double pole of
1
cos2 z
, for any integer n. In some deleted -neighborhood around the point
 6.2 Residues and the Residue Theorem 265
 
zn = n + 1
2
π, cos12 z admits the following Laurent series expansion:
 ∞
1 c−2 c−1
= + + ck (z − zn )k , 0 < |z − zn | < .
cos2 z (z − zn )2 z − zn k=0

On the other hand, since f (z) is analytic in a domain containing the whole
real axis, f (z) admits the following Taylor series expansion inside the -
neighborhood centered at z = zn :
∞

f (z) = am (z − zn )m , |z − zn | < .
m=0

Now, we consider the Laurent series expansion of

%∞ &% ∞
&
f (z)  c−2 c−1 
= am (z − zn ) m
+ + ck (z − zn ) .
k
cos2 z m=0
(z − zn )2 z − zn k=0
 
f (z) 1
Since Res cos 2 z , zn is equal to the coefficient of z−z n
in the above Laurent
series expansion, we obtain
 
f (z)
Res , zn = a1 c−2 + a0 c−1 ,
cos2 z
where a0 = f (zn ) and a1 = f  (zn ). The Taylor series of cos2 z at z = zn con-
tains only even power terms, so the Laurent series of cos12 z at z = zn has only
even power terms. This gives c−1 = 0. Lastly, we compute c−2 by evaluating
the following limit using the L’Hospital rule, where
(z − zn )2 2(z − zn ) 2
c−2 = lim = lim = lim = 1.
z→zn cos2 z z→zn −2 sin z cos z z→zn −2 cos 2z

Therefore, we have
 
f (z)
Res , zn = a1 = f  (zn ).
cos2 z

Example 6.2.8 Suppose f (z) is analytic everywhere in the finite complex

plane except for a finite number of isolated singularities interior to a positively
oriented simple closed contour C. Show that
,    
1 1
f (z) dz = 2π i Res 2 f ,0 .
C z z
Use the result to evaluate
,
6z + 4
dz.
|z|=3 z3 + 3z2 + 2z
266 Singularities and Calculus of Residues

y
Cout

c R
x

Figure 6.4. The function f (z) is analytic everywhere outside the simple closed contour
C. The contour C lies completely inside the circle |z| = R. The exterior simple closed
contour Cout encloses the circle |z| = R.

Solution Given the simple closed contour C that encircles all isolated sin-
gularities of f (z), we choose a circle |z| = R such that the contour C lies
completely inside the circle. An exterior simple closed contour Cout is chosen
such that it encloses the circle |z| = R (see Figure 6.4).
Let us consider the Laurent series expansion of f (z) in the annulus R <
|z| < ∞, where
∞

f (z) = cn zn , R < |z| < ∞.
n=−∞

From the Laurent series theorem, the Laurent coefficients are given by
,
1 f (z)
cn = dz, n = 0, ±1, ±2, . . . .
2π i Cout zn+1

In particular, we are interested in computing

,
1
c−1 = f (z) dz.
2π i Cout

The above annulus R < |z| < ∞ is not a deleted neighborhood about z = 0, so
c−1 is not equal to Res (f, 0). Suppose we replace z by 1z in the above Laurent
series expansion; we obtain
  ∞ ∞
1 1 cn cn−2
f = = ,
z2 z n=−∞
z n+2
n=−∞
zn
 6.2 Residues and the Residue Theorem 267

which is valid in the annulus 0 < |z| < R1 . This new annulus is now a deleted
neighborhood about z = 0. We then deduce that
    ,
1 1 1
Res 2 f , 0 = c−1 = f (z) dz.
z z 2π i Cout
Since there is no singularity lying between Cout and C, we can deform the
exterior contour Cout to C, so we obtain
,    
1 1
f (z) dz = 2π i Res 2 f ,0 .
C z z
The integrand function
6z + 4
f (z) =
z3 + 3z2 + z
has isolated singularities at z = 0, z = −1 and z = −2 inside the contour
|z| = 3. By the Cauchy residue theorem, we have
,
f (z) dz = 2π i[Res (f, 0) + Res (f, −1) + Res (f, −2)].
|z|=3

Instead of evaluating the above three residue

 values,
  a more
 efficient evaluation
method can be derived by computing Res z12 f 1z , 0 . Now, we consider
 
1 1 1
6
z
+4 4z + 6
f = 2 = ,
z2 z z 1
z3
+ z
z2
+ 2
z
2z2 + 3z + 1

which is seen to be analytic at z = 0, so

,    
1 1
f (z) dz = 2π i Res 2 f , 0 = 0.
|z|=3 z z

Remark By convention, we define the residue at complex infinity by

,    
1 1 1
Res (f, ∞) = − f (z) dz = −Res 2 f ,0 ,
2π i C z z
where the contour C is chosen such that all singularities of f (z) in the finite
plane are included inside C. The negative sign is deliberately chosen in the
above definition so that

Res(f, zn ) + Res (f, ∞) = 0,
all

where the summation is taken over all singularities zn inside C. That is, if
f (z) is analytic in the extended complex plane except for a finite number of
268 Singularities and Calculus of Residues

singularities, then the sum of all residues of f (z), including the residue at
complex infinity, is zero.

6.3 Evaluation of real integrals by residue calculus

A wide variety of real definite integrals can be evaluated effectively by the
calculus of residues. The techniques of evaluation can be split into different
categories as discussed below.

6.3.1 Integrals of trigonometric functions over [0, 2π ]

We consider a real integral involving trigonometric functions of the form
 2π
R(cos θ, sin θ ) dθ,
0

where R(x, y) is a rational function defined inside the unit circle |z| = 1, z =
x + iy. The real integral can be converted into a contour integral around the
unit circle by the following substitutions:

z = eiθ , dz = ieiθ dθ = iz dθ,

 
eiθ + e−iθ 1 1
cos θ = = z+ ,
2 2 z
 
eiθ − e−iθ 1 1
sin θ = = z− .
2i 2i z

The real integral can then be transformed:

 2π
R(cos θ, sin θ) dθ
0
,  
1 z + 1z z − 1z
= R , dz
|z|=1 iz 2 2i
%   &
1 z + 1z z − 1z
= 2π i sum of residues of R , inside|z| = 1 . (6.3.1)
iz 2 2i

Example 6.3.1 Evaluate the real definite integral

 π
1
I= dθ, a > b > 0.
0 a − b cos θ
 6.3 Evaluation of real integrals by residue calculus 269

Solution Since the integrand is symmetric about θ = π , it is desirable to

extend the integration interval to [0, 2π ] so that
 2π  2π
1 1 eiθ
I= dθ = dθ.
2 0 a − b cos θ 0 2aeiθ − b(e2iθ + 1)

We let z = eiθ and as θ increases from 0 to 2π , z moves around the unit circle
|z| = 1. By eq. (6.3.1), the real integral can be transformed into the following
contour integral
,
1 1
I =− dz,
i |z|=1 bz2 − 2az + b

where the path of integration is the unit circle |z| = 1. The integrand has two
simple poles, which are given by the zeros of the denominator. The product of
the poles is seen to be 1. Let α denote the pole that is inside the unit circle; then
the other pole will be α1 . The two poles are found to be
√ √
a− a 2 − b2 1 a+ a 2 − b2
α= and = .
b α b
Since a > b > 0, the two poles are distinct. We then have
,
1 1
I =− dz
ib |z|=1 (z − α) (z − α1 )
 
2π i 1
=− Res ,α
ib (z − α) (z − α1 )
2π i π
=−  =√ .
ib α − α1
a 2 − b2

6.3.2 Integrals of rational functions

We consider a real integral involving a rational function of the form
 ∞
f (x) dx,
−∞

where

(i) f (z) is a rational function with no singularity on the real axis,

(ii) lim zf (z) = 0.
z→∞
270 Singularities and Calculus of Residues

CR
>

> x
R R
Figure 6.5. The contour of integration is the union of the upper semi-circle CR and the
line segment from −R to R along the real axis.

It can be shown that

 ∞
f (x)dx
−∞
= 2π i [sum of residues at the poles of f in the upper half-plane]. (6.3.2)
To verify the claim, we integrate f (z) around a closed contour C that consists
of the upper semi-circle CR and the line segment from −R to R along the real
axis (see Figure 6.5). By the Cauchy residue theorem, we have
,  R 
f (z) dz = f (x) dx + f (z) dz
C −R CR
= 2π i [sum of residues at the poles of f inside C].
As R → ∞, all the poles of f in the upper half-plane will be enclosed inside
C. To establish eq. (6.3.2), it suffices to show that as R → ∞,

lim f (z) dz = 0. (6.3.3)
R→∞ C
R

The modulus of the above integral is estimated by the modulus inequality

(4.1.6) as follows:
   π
 
 f (z) dz ≤ |f (Reiθ )| R dθ

CR 0
 π
≤ max |f (Re )| R
iθ
dθ
0≤θ≤π 0
= max |zf (z)|π,
z∈CR

which goes to zero as R → ∞, since lim zf (z) = 0. Hence, the result in

z→∞
eq. (6.3.3) is established.
 6.3 Evaluation of real integrals by residue calculus 271

Example 6.3.2 Evaluate the real integral

 ∞
x4
dx
−∞ 1 + x
6

by the residue method.

Solution The complex function

z4
f (z) =
1 + z6
√ √
has simple poles at i, 3+i
2
and − 23+i in the upper half-plane, and it has no
singularity on the real axis. Furthermore, the integrand observes the property
lim zf (z) = 0. By eq. (6.3.2), we have
z→∞
 % 
√   √ &
∞
3+i − 3+i
f (x) dx = 2π i Res(f, i) + Res f, + Res f, .
−∞ 2 2

By eq. (6.2.7), the residue values at the simple poles are found to be
 
z4  1  i
Res(f, i) = d  =  =− ,
6 
(1 + z ) z=i 6z 6
dz z=i

√   √
3+i 1  3−i
Res f, =  √ = ,
2 6z z= 3+i 12
2

and
 √   √
− 3+i 1  3+i
Res f, =  √ =− .
2 6z z= − 3+i 12
2

Combining the results, we obtain

 ∞  √ √ 
x4 i 3−i 3+i 2π
dx = 2π i − + − = .
−∞ 1 + x 6 6 12 12 3

6.3.3 Integrals involving multi-valued functions

We would like to evaluate a real integral involving a fractional power function
of the form
 ∞
f (x)
dx, where 0 < α < 1.
0 xα
272 Singularities and Calculus of Residues

The required properties of f (z) are:

1. f (z) is a rational function with no singularity on the positive real axis,

including the origin;
2. lim f (z) = 0.
z→∞

Remark We may relax the first requirement to allow the function f (z) to have
a finite number of simple poles along the positive real axis. In this case, the
Cauchy principal value of the integral should be considered (see Section 6.5).

We integrate φ(z) = fz(z)

α along the closed contour C as shown in Figure 6.6.

The contour C consists of four parts:

(i) the line segment from  to R along the upper side of the positive real
axis, that is, z = x,  ≤ x ≤ R;
(ii) the large circle CR traversing in the anticlockwise direction, that is,
z = Reiθ , 0 < θ < 2π;
(iii) the line segment from R to  along the lower side of the positive real
axis, that is, z = xe2πi ,  ≤ x ≤ R;
(iv) the infinitesimal circle C traversing in the clockwise direction, that is,
z = eiθ , 0 < θ < 2π.

Subsequently, we take the limits R → ∞ and  → 0. We first establish the

following two results:
 
lim φ(z) = 0 and lim φ(z) = 0. (6.3.4)
R→∞ C →0 C
R 

The bounds on the moduli of the above two integrals are found to be
   2π
 
 φ(z) dz ≤ |φ(Reiθ )| R dθ ≤ 2π max |zφ(z)|,
  z∈CR
CR 0

and
  
  2π
 φ(z) dz ≤ |φ(eiθ )|  dθ ≤ 2π max |zφ(z)|.
  z∈C
C 0

It suffices to show that

zφ(z) → 0 as either z→∞ or z → 0.

As lim f (z) = 0 and f (z) is a rational function, the degree of the denominator
z→∞
of f (z) is at least one higher than that of its numerator. Since 1 − α < 1,
we have zφ(z) = z1−α f (z) → 0 as z → ∞. Also, we deduce that f (z) is
 6.3 Evaluation of real integrals by residue calculus 273

Figure 6.6. The closed contour C consists of an infinitely large circle and an infinitesi-
mal circle joined by line segments along the positive x-axis.

continuous at z = 0 since f (z) has no singularity at the origin. Therefore,

zφ(z) = z1−α f (z) ∼ 0 · f (0) = 0 as z → 0.
Lastly, we observe that the term zα in the integrand is multi-valued since

zα = eα log z = eα(ln |z| + iθ) = |z|α eiαθ , where θ = arg z.

The choice of the contour of integration as shown in Figure 6.6 dictates that
the argument of the principal branch of zα is chosen to be 0 ≤ θ < 2π . Note
that the origin is a branch point and the positive real axis is chosen to be the
branch cut.
Now the contour integral around C can be expressed as the sum of four
contour integrals:
,  
φ(z) dz = φ(z) dz + φ(z) dz
C CR C
 R  
f (x) f (xe2πi )
+ dx + α 2απi
dx
 xα R x e
= 2π i [sum of residues of all the isolated singularities
of f enclosed inside the closed contour C]. (6.3.5)

By taking the limits  → 0 and R → ∞, the first two integrals vanish. The last
integral can be expressed as
 ∞  ∞
f (x) f (x)
− dx = −e−2απi dx.
0 x e2απi
α
0 xα
274 Singularities and Calculus of Residues

Combining the results, the real integral can be evaluated by the formula:
 ∞
f (x) 2π i
dx = [sum of residues at all the isolated singularities
0 x α 1 − e−2απi
of f in the finite complex plane]. (6.3.6)

Example 6.3.3 Evaluate the integral

 ∞
1
dx, 0 < α < 1.
0 (1 + x)x α

Solution The integrand

1
f (z) =
(1 + z)zα
is multi-valued and has an isolated singularity (a simple pole) at z = −1. We
choose the branch cut to be along the positive real axis with the origin as a
branch point. We now consider the contour integral
,
1
dz
C (1 + z)z
α

around the closed contour C as shown in Figure 6.6.

The contour integrals along the upper line segment and the lower line segment
along the positive real axis can be expressed as
 R  R
1 −2απi 1
dx and −e dx,
 (1 + x)x  (1 + x)x
α α

respectively. By the Cauchy residue theorem, the contour integral is found to

be
,
1
dz
C (1 + z)z
α
 R  
dx dz dz
= (1 − e−2απi ) + +
 (1 + x)x α
CR (1 + z)z α
C (1 + z)zα
 
1 2π i
= 2π i Res , −1 = απi ,
(1 + z)zα e
where CR and C denote the outer large circle and the inner small circle,
respectively. Given 0 < α < 1, the moduli of the second and third integrals can
be estimated as follows:
 
 1  2π R
 dz ≤ ∼ R −α → 0 as R → ∞;
 (1 + z)z α (R − 1)R α
CR
 
 1  2π 
 
 (1 + z)zα dz ≤ (1 − ) α ∼  →0 as  → 0.
1−α
C
 6.3 Evaluation of real integrals by residue calculus 275

Figure 6.7. The closed rectangular contour encloses only one simple pole at z = π i.

On taking the limits R → ∞ and  → 0, we obtain

 ∞
−2απi 1 2π i
(1 − e ) dx = απi ;
0 (1 + x)x α e
so
 ∞
1 2π i π
dx = απi = .
0 (1 + x)x α e (1 − e−2απi ) sin απ

6.3.4 Miscellaneous types of integral

The success of applying the residue calculus techniques to evaluation of differ-
ent types of real integral relies on the ingenious choice of the contour, possibly
together with an appropriate modification of the integrand. Additional exam-
ples are presented below to illustrate the high level of ingenuity exhibited in
the residue calculus approach.

Example 6.3.4 Evaluate the real integral

 ∞
eαx
dx, 0 < α < 1.
−∞ 1 + e
x

Solution The integrand function in its complex extension has infinitely

many poles in the complex plane, namely, at z = (2k + 1)π i, where k
is any integer. Therefore, it is not advisable to choose a contour that
includes an infinitely large outer circle. Instead, we choose the rectangu-
lar contour as depicted in Figure 6.7. The contour consists of four line
276 Singularities and Calculus of Residues

segments:

l1 : y = 0, − R ≤ x ≤ R,
l2 : x = R, 0 ≤ y ≤ 2π,
l3 : y = 2π, − R ≤ x ≤ R,
l4 : x = −R, 0 ≤ y ≤ 2π.

Let C = l1 ∪ l2 ∪ l3 ∪ l4 , which is chosen to be oriented in the anticlockwise

sense. The only simple pole that is enclosed inside the closed contour C is
z = π i. By the Cauchy residue theorem, we have
,  R  2π
eαz eαx eα(R+iy)
dz = dx + idy
C 1 + ez −R 1 + ex 0 1 + eR+iy
 −R α(x+2πi)  0 α(−R+iy)
e e
+ dx + −R+iy
idy
1 + ex+2πi 2π 1 + e
R
 αz 
e
= 2π i Res , πi
1 + ez

αz 
e
= 2π i z  = −2π ieαπi .
e z=πi

The above residue is evaluated using the formula in eq. (6.2.7). We estimate
the bounds on the moduli of the second and the fourth integrals as follows:
   2π
 2π
eα(R+iy)  eαR
 idy ≤ dy ∼ O(e−(1−α)R ),
 1 + eR+iy  eR − 1
0 0

   2π
 0
eα(−R+iy)  e−αR
 idy ≤ dy ∼ O(e−αR ).
 1 + e−R+iy  1 − e−R
2π 0

As 0 < α < 1, both e−(1−α)R and e−αR tend to zero as R → ∞. Therefore,

the second and the fourth integrals tend to zero as R → ∞. Taking the limit
R → ∞, the sum of the first and the third integrals becomes
 ∞
eαx
(1 − e2απi ) dx = −2π ieαπi ;
−∞ 1 + e
x

so
 ∞
eαx 2π i π
dx = απi = .
−∞ 1 + ex e − e−απi sin απ
 6.3 Evaluation of real integrals by residue calculus 277

Example 6.3.5 Evaluate the real integral

 ∞
1
dx.
0 1+x
3

Solution If the interval of integration is (−∞, ∞) instead of (0, ∞), then

the corresponding integral can be evaluated using the technique discussed in
Subsection 6.3.2. Here, the integrand is not an even function, so it serves no
purpose to extend the interval of integration to (−∞, ∞). Alternatively, we
consider the branch cut integral
,
log z
dz,
C 1+z
3

where C is the closed contour shown in Figure 6.6. The inner small circle C
has radius ,  → 0, and the outer large circle CR has radius R, R → ∞. The
positive real axis is taken to be the branch cut of log z so that 0 ≤ arg z < 2π .
From the relation log z = ln |z| + i arg z, we have

log z = ln |z| along the upper side of the positive x-axis,

log z = ln |z| + 2π i along the lower side of the positive x-axis.

The asymptotic order of the modulus of the contour integral around the inner
small circle C is estimated to be
, 
 log z 
 
 1 + z3 dz = O(2π  · ln ) → 0 as  → 0.
C

Therefore, the contribution from the contour integral around C is zero. Sim-
ilarly, the contribution from the contour integral around the outer large circle
CR is also zero since
,   
 log z  ln R
 
dz = O 2π R · 3 → 0 as R → ∞.

CR 1 + z
3 R
Taking the limits  → 0 and R → ∞, the contour integral around C becomes
,  ∞  0
log z ln x ln x + 2π i
lim dz = dx + dx
R→∞
C 1 + z 3
0 1 + x 3
∞ 1 + x3
 ∞
→0

1
= −2π i dx
0 1+x
3

log z
= 2π i sum of residues of at the
1 + z3

poles in the finite complex plane. .

278 Singularities and Calculus of Residues
log z
Interestingly, the two terms with ln x cancel. The integrand 1+z3 has simple
poles at α = eiπ /3 , β = eiπ and γ = e5πi/3 . The sum of the residues at the
three poles is found to be
     
log z log z log z
Res , α + Res , β + Res , γ
1 + z3 1 + z3 1 + z3
log α log β log γ
= + +
(α − β) (α − γ ) (β − α) (β − γ ) (γ − α) (γ − β)
π 
(β − γ ) + π(γ − α) + 5π (α − β) 2π
= −i 3 3
=− √ .
(α − β) (β − γ ) (γ − α) 3 3
Combining the results, we obtain
 ∞
1 2π
dx = √ .
0 1 + x3 3 3

6.4 Fourier transforms

A real function is said to be piecewise continuous in an interval if it
is continuous everywhere in the interval except at a finite number of
points where the function is allowed to have jump discontinuities. Let
u(x) be a piecewise continuous real function over (−∞, ∞) satisfying the
condition
 ∞
|u(x)| dx < ∞.
−∞

The Fourier transform of u(x) is defined by

 ∞
F{u(x)} = U (ω) = eiωx u(x) dx, −∞ < ω < ∞. (6.4.1)
−∞

The following properties of Fourier transforms can be deduced immediately

from the basic rules in calculus:

(i) F{α1 u1 (x) + α2 u2 (x)} = α1 F{u1 (x)} + α2 F{u2 (x)},

where α1 and α2 are complex constants;
(ii) F{u(x − a)} = eiaω F{u(x)}, a is real;
1 ω
(iii) F{u(bx)} = U ;
b b
−iω0 x
(iv) F{e u(x)} = U (ω − ω0 );
d
(v) F{u(x)} = iωF{u(x)};
dω 
(vi) F{u (x)} = −iωF{u(x)}.
 6.4 Fourier transforms 279

Convolution product
The convolution product of two functions f (x) and g(x) is defined by
 ∞
(f ∗ g)(x) = f (ξ )g(x − ξ ) dξ. (6.4.2)
−∞

The Fourier transform of f ∗ g is given by

 ∞
F{f ∗ g} = eiωx (f ∗ g)(x) dx
−∞
 ∞  ∞
= eiωx f (ξ )g(x − ξ ) dξ dx. (6.4.3)
−∞ −∞

Suppose f and g are absolutely integrable; we apply the interchange of the

order of integration to give
 ∞  ∞
F{f ∗ g} = eiω(x−ξ ) g(x − ξ ) dx eiωξ f (ξ ) dξ.
−∞ −∞

The first and second integrals are recognized as F{g(x)} and F{f (x)}, respec-
tively, so we obtain
F{f ∗ g} = F{f }F {g}. (6.4.4)

6.4.1 Fourier inversion formula

We would like to establish the relation
 ∞  ∞
1
u(x) = e−iωx eiωt u(t) dtdω, (6.4.5)
2π −∞ −∞

from which we can deduce the Fourier inversion formula

 ∞
1
u(x) = e−iωx U (ω) dω. (6.4.6)
2π −∞
This is an elegant result since the Fourier transform formula and its Fourier
inversion resemble each other closely, except with the flip of the sign in the
1
exponent of the integration kernel and the appearance of the factor 2π .
To establish eq. (6.4.5), we start from the Fourier series representation of
a periodic function u(x) defined over the finite interval [−L, L]. The Fourier
series expansion of u(x) takes the form
∞ ∞
a0  nπx  nπx
u(x) = + an cos + bn sin . (6.4.7)
2 n=1
L n=1
L

The validity of the above expansion requires f (x) to have only a finite number
of finite discontinuities and only a finite number of maxima and minima. These
280 Singularities and Calculus of Residues

conditions are sufficient but not necessary, and they are commonly called the
Dirichlet conditions. By applying the orthogonality properties of the component
functions over [−L, L], the Fourier coefficients an and bn are given by

1 L nπt
an = u(t) cos dt, n = 0, 1, . . .
L −L L

1 L nπt
bn = u(t) sin dt, n = 1, 2, . . . .
L −L L
Writing out all the terms in full, eq. (6.4.7) can be written as
 L ∞ 
1 1 nπx L nπt
u(x) = u(t) dt + cos u(t) cos dt
2L −L L n=1 L −L L
∞ 
1 nπx L nπt
+ sin u(t) sin dt
L n=1 L −L L
 L ∞ 
1 1 L nπ
= u(t) dt + u(t) cos (t − x) dt. (6.4.8)
2L −L L n=1 −L L

Suppose we let L approach infinity so that [−L, L] becomes (−∞, ∞). Further,
we set
nπ π
= ω and = ω.
L L
As L → ∞, the first term in eq. (6.4.8) vanishes since the value of the integral
is finite given that u(t) is absolutely integrable. The second term becomes the
series
∞  ∞
1
ω u(t) cos ω(t − x) dt.
π n=1 −∞

In the limit ω → 0, the above infinite sum is replaced by an integral with

integration variable ω. We thus formally obtain
 
1 ∞ ∞
u(x) = u(t) cos ω(t − x) dtdω.
π 0 −∞

To arrive at the form shown in eq. (6.4.5), we observe that cos ω(t − x) is an
even function of ω and sin ω(t − x) is an odd function of ω. We then have
 ∞ ∞
1
u(x) = u(t) cos ω(t − x) dt dω (6.4.9a)
2π −∞ −∞
and
 ∞  ∞
1
0= u(t) sin ω(t − x) dt dω. (6.4.9b)
2π −∞ −∞
 6.4 Fourier transforms 281

Adding eq. (6.4.9a) and i times eq. (6.4.9b) together, the result in eq. (6.4.5) is
finally established.

Remark Suppose ue (x) is an even function, where ue (−x) = ue (x), x > 0.

The Fourier cosine transform of ue (x) is defined by
 ∞
Uc (ω) = Fc {ue (x)} = cos ωx ue (x) dx, x > 0. (6.4.10)
0
The corresponding inversion formula can be deduced to be

2 ∞
ue (x) = cos ωx Uc (ω) dω, ω > 0. (6.4.11)
π 0
Hints for the proof of eq. (6.4.11) are given in Problem 6.24.
Similarly, suppose uo (x) is an odd function, where uo (−x) = −uo (x),
x > 0. The Fourier sine transform of uo (x) and its inversion formula are
given by
 ∞
Us (ω) = Fs {uo (x)} = sin ωx uo (x) dx, x > 0 (6.4.12)
0
and
 ∞
2
uo (x) = sin ωx Us (ω) dω, ω > 0, (6.4.13)
π 0
respectively.

Parseval identity
The Parseval identity states that
 ∞  ∞
1
|u| (x) dx =
2
|U (ω)|2 dω. (6.4.14)
−∞ 2π −∞
To prove the identity, we apply the Fourier inversion formula to F{f ∗ g} so
that
 ∞  ∞
1
(f ∗ g)(x) = f (ξ )g(x − ξ ) dξ = e−iωx F{f ∗ g} dω.
−∞ 2π −∞
Suppose we choose f (ξ ) = u(ξ ) and g(ξ ) = u(−ξ ), and set x = 0. We then
obtain
 ∞  ∞
1
u(ξ )u(ξ ) dξ = F{u(x)}F {u(−x)} dω
−∞ 2π −∞
 ∞
1
= U (ω)U (ω) dω;
2π −∞
hence the result.
282 Singularities and Calculus of Residues

Dirac function
The construction of the Dirac function may be motivated by attempts to find a
function ψ(x) such that the convolution of ψ with any function f always gives
f . That is,
 ∞
f (ξ )ψ(x − ξ ) dξ = f ∗ ψ = ψ ∗ f
−∞
 ∞
= ψ(ξ )f (x − ξ ) dξ = f (x). (6.4.15)
−∞

If we set x = 0, then ψ(x) has to observe

 ∞
ψ(ξ )f (−ξ ) dξ = f (0). (6.4.16)
−∞

The above result indicates that the values of f (x) for x = 0 do not have any
impact on the value of the integral of the product ψ(ξ )f (−ξ ). This occurs
only when ψ(x) = 0 for all nonzero values of x. On the other hand, by taking
f (x) = 1, we obtain
 ∞
ψ(ξ ) dξ = 1. (6.4.17)
−∞

Apparently, there is no function defined in the usual pointwise sense that

possesses the above two properties. Actually, ψ(x) belongs to the class of
generalized functions.
Since no such function exists in the usual pointwise sense, can we approxi-
mate the generalized function ψ(x) in some form as the limit of a sequence of
functions? Consider the sequence of functions δn (x), n = 1, 2, . . . , defined by

n − 2n 1
< x < 2n1
δn (x) = . (6.4.18)
0 otherwise
As n increases, δn (x) resembles a high and narrow spike, like an impulse with
infinite magnitude. Also, {δn (x)} is a sequence of functions that satisfy the
normalized condition:
 ∞
δn (x) dx = 1.
−∞

The limit
lim δn (x)
n→∞

does not exist; however, the sequence of integrals

 ∞
δn (x)f (x) dx
−∞
 6.4 Fourier transforms 283

for any continuous function f (x) exists and the corresponding limit is equal to
 ∞
lim δn (x)f (x) dx = f (0).
n→∞ −∞

The Dirac function δ(x) is formally defined by

 ∞  ∞
δ(x)f (x) dx = lim δn (x)f (x) dx = f (0), (6.4.19a)
−∞ n→∞ −∞

where the left-hand integral is the limit of a sequence of integrals. Here, δ(x)
is a distribution, which means that it has effect on continuous functions but is
not itself a function. With f (x) = 1, we obtain
 ∞
δ(x) dx = 1. (6.4.19b)
−∞

Though δ(x) cannot be defined in the usual pointwise sense, sometimes it may
be convenient to think of δ(x) as a function of x with the property

∞ x=0
δ(x) = . (6.4.19c)
0 x = 0

The choice of the sequence of limiting functions {δn (x)} is not unique. Some
other possible choices are
n
δn(1) (x) = √ e−n x ,
2 2
n = 1, 2, . . . ; (6.4.20a)
π
n 1
δn(2) (x) = , n = 1, 2, . . . ; (6.4.20b)
π 1 + n2 x 2
 n
sin nx 1
δn(3) (x) = = e−ixt dt, n = 1, 2, . . . . (6.4.20c)
πx 2π −n
These sequences of functions all satisfy the properties:
 ∞  ∞
lim δn (x)f (x) dx = f (0) and δn (x) dx = 1. (6.4.21)
n→∞ −∞ −∞

They invariably become infinitely high and narrow spikes as n increases to

infinity. Readers should be wary that we may have a sequence of functions that
satisfy the normalized condition and exhibit the property

∞ x = 0
lim δn (x) = ;
n→∞ 0 x=0

however the limit of which is not the distribution of δ(x).

284 Singularities and Calculus of Residues

Suppose we translate δ(x) by x0 to give another generalized function defined

as δ(x − x0 ), and this new generalized function observes the property:
 ∞
δ(x − x0 )f (x) dx = f (x0 ). (6.4.22)
−∞

In a loose sense, we may think of δ(x − x0 ) as a generalized function which

assumes zero value when x = x0 and infinite value when x = x0 .
The Dirac function is closely related to the Heaviside function H (x) defined
in eq. (3.3.13). By observing
 x
δ(x) dx = 0, x < 0,
−∞
 ∞
δ(x) dx = 0, x > 0,
x

we deduce that
 x 
1 if x > 0
δ(ξ ) dξ = .
−∞ 0 if x < 0
By formally differentiating with respect to x on both sides of the above equation,
the relation between H (x) and δ(x) is found to be
d
H (x) = δ(x). (6.4.23)
dx
Suppose we take f (x) = eiωx in eq. (6.4.22); we obtain
 ∞
eiωx δ(x − x0 ) dx = eiωx0 .
−∞

Setting x0 = 0 in the above equation, we obtain the following formula for the
Fourier transform of δ(x):

F{δ(x)} = 1. (6.4.24)

Taking the Fourier inversion of U (ω) = 1, we have

 ∞
1
δ(x) = e−iωx dω. (6.4.25)
2π −∞
Suppose we rewrite the above integral formally as
 ∞  n
1 −iωx 1 −iωx sin nx
e dω = lim e dω = lim . (6.4.26)
2π −∞ n→∞ −n 2π n→∞ π x

Therefore, the sequence {δn(3) (x)} defined in eq. (6.4.20c) is seen to be a sequence
of functions that approximate the distribution of δ(x).
 6.4 Fourier transforms 285

Interestingly, eq. (6.4.5) can be established readily using eqs. (6.4.22) and
(6.4.25), assuming that interchanging order of integration is valid. The proof is
shown as follows:
 ∞
u(x) = u(t)δ(x − t) dt
−∞
 ∞   ∞ 
1
= u(t) e−iω(x−t) dω dt
−∞ 2π −∞
 ∞  ∞
1
= e−iωx eiωt u(t) dtdω.
2π −∞ −∞

6.4.2 Evaluation of Fourier integrals

We would like to evaluate a Fourier integral of the form
 ∞
eimx f (x) dx, m > 0,
−∞

where
(i) f (z) has a finite number of isolated singularities,
(ii) lim f (z) = 0,
z→∞
(iii) f (z) has no singularity along the real axis.

Remark
(i) The assumption m > 0 is not strictly essential. With some minor modifi-
cations, the method proposed below also works even when m is negative
or imaginary.
(ii) When f (z) has singularities on the real axis, the Cauchy principal value
of the integral is considered (see Section 6.5 and Example 6.5.3).

Jordan lemma
Suppose f (z) → 0 as z → ∞; the Jordan lemma states that

lim eiλz f (z) dz = 0, (6.4.27)
R→∞ C
R

where λ is a positive real number and the contour CR is the upper semi-
circle with radius R, R → ∞. The proof of the lemma requires the following
inequality:
2θ π
sin θ ≥ for 0≤θ ≤ . (6.4.28)
π 2
286 Singularities and Calculus of Residues

Figure 6.8. Geometric visualization of inequality (6.4.28).

The inequality can be visualized by observing that the curve for y = sin θ is
concave over the interval [0, π2 ], so it always lies on or above the line segment
y = 2θ
π
(see Figure 6.8).
We consider the modulus of the integral
   π
 
 f (z)eiλz dz ≤ iθ
|f (Reiθ )| |eiλRe | R dθ
 
CR 0
 π
≤ max |f (z)| R e−λR sin θ dθ
z∈CR 0
 π
2
= 2R max |f (z)| e−λR sin θ dθ
z∈CR 0
 π
2
e−λR π dθ
2θ
≤ 2R max |f (z)|
z∈CR 0
π
= 2R max |f (z)| (1 − e−λR ),
z∈CR 2Rλ
which tends to 0 as R → ∞, given that f (z) → 0 uniformly as R → ∞ for
z ∈ CR .

Remark We may relax the restriction that λ > 0. For λ < 0, the same result
can be obtained if CR is chosen to be the lower semi-circle. When λ = im, m >
0, we choose CR to be the right semi-circle; and when λ = −im, m > 0, CR
is chosen to be the left semi-circle.

To evaluate the Fourier integral, we integrate eimz f (z) along the closed
contour C that consists of the upper semi-circle CR and the line segment from
 6.4 Fourier transforms 287

−R to R along the real axis (see Figure 6.8). We then have

,  R 
eimz f (z) dz = eimx f (x) dx + eimz f (z) dz. (6.4.29)
C −R CR

Taking the limit R → ∞, the integral along CR vanishes by virtue of the Jordan
lemma. Lastly, we apply the Cauchy residue theorem to obtain
 ∞
eimx f (x) dx = 2π i [sum of residues at all the isolated
−∞
singularities of f in the upper half-plane (6.4.30)
since C encloses all the singularities of f in the upper half-plane as R → ∞.

Example 6.4.1 Evaluate the Fourier integral

 ∞
sin 2x
2+x +1
dx.
−∞ x

Solution It is easy to check that

1
f (z) =
z2 +z+1
has no singularity along the real axis and
1
lim = 0.
z→∞ z2 + z + 1
The integrand has two simple poles, namely, z = e2πi/3 in the upper half-plane
and e−2πi/3 in the lower half-plane. By virtue of the Jordan lemma, we have
 ∞  ∞ ,
sin 2x e2ix e2iz
dx = Im dx = Im dz,
−∞ x + x + 1 −∞ x + x + 1 Cz +z+1
2 2 2

where C is the union of the infinitely large upper semi-circle and its diameter
along the real axis. By eq. (6.4.30), we have
,  
e2iz e2iz 2πi
dz = 2π i Res , e 3

Cz +z+1 z2 + z + 1
2

 2πi
e2iz  e2ie 3
= 2π i = 2π i .
2z + 1 z=e 2πi3 2πi
2e 3 + 1
Hence, the Fourier integral is found to be
 
 ∞ 2πi
2ie 3 √
sin 2x 2π i e2πi  = − √2 π e− 3 sin 1.
dx = Im
−∞ x + x + 1
2
2e 3 + 1 3
288 Singularities and Calculus of Residues

6.5 Cauchy principal value of an improper integral

Suppose a real function f (x) is continuous everywhere in the interval [a, b]
except at a point x0 inside the interval. The integral of f (x) over the interval
[a, b] is an improper integral, which may be defined as
 b  x0 −1  b

f (x) dx = lim f (x) dx + f (x) dx , 1 , 2 > 0.

a 1 ,2 →0 a x0 +2
(6.5.1)
In many cases, the above limit exists only when 1 = 2 , and does not exist
otherwise. For example, the function y = x1 is not continuous at x = 0, and in
a strict sense the integral
 2
1
dx
−1 x

does not exist. If we define the improper integral according to eq. (6.5.1), we
obtain
 2
1
dx
−1 x
 −1  2
1 1
= lim dx + lim dx, 1 > 0 and 2 > 0
1 →0 −1 x 2 →0  x
2
 
1
= lim (ln 1 + ln 2 − ln 2 ) = ln 2 + lim ln . (6.5.2)
1 ,2 →0 1 ,2 →0 2

The limit does not exist if 1 → 0 and 2 → 0 in an arbitrary manner.

The Cauchy principal value of the improper integral in eq. (6.5.1) (symbol-
ized by the letter P ) is defined by taking 1 = 2 =  so that
 b  x0 −  b

P f (x) dx = lim f (x) dx + f (x) dx ,  > 0, (6.5.3)

a →0 a x0 +

provided that the limit exists. For the improper integral in eq. (6.5.2), the
corresponding Cauchy principal value is found to be
 2  
1
P dx = ln 2 + lim ln = ln 2. (6.5.4)
−1 x →0 

In contour integration, consideration of the principal value of an improper

integral is necessary when the integration contour passes through an isolated
singularity of the integrand. We illustrate the evaluation of the Cauchy principal
value of an improper integral through the following examples.
 6.5 Cauchy principal value of an improper integral 289

Example 6.5.1 Find the principal value of the following definite inte-
gral:
 π
cos nφ
In = dφ, 0 < θ < π and n is any non-negative integer.
0 cos φ − cos θ

Solution By definition, the principal value of In is given by

 θ−  π 
cos nφ cos nφ
In = lim dφ + dφ .
→0 0 cos φ − cos θ θ+ cos φ − cos θ

When n = 0, we have
 θ−
θ−
1 1 sin 12 (θ + φ) 
dφ = ln 
0 cos φ − cos θ sin θ sin 12 (θ − φ) 
0

and
 π
π
1 1 sin 12 (φ + θ ) 
dφ = ln  ;
θ+ cos φ − cos θ sin θ sin 12 (φ − θ ) 
θ+
so
 
1 sin θ − 2
I0 = lim ln   = 0.
→0 sin θ sin θ + 2
When n = 1, by observing I0 = 0, we obtain
 π
cos φ − cos θ
I1 = I1 − I0 cos θ = dφ = π.
0 cos φ − cos θ

To evaluate In , n > 1, we write

 π
cos nφ − cos nθ
In = In − I0 cos nθ = dφ.
0 cos φ − cos θ
Using the trigonometric relation
cos (n + 1)φ − cos (n + 1)θ + cos (n − 1)φ − cos (n − 1)θ
= 2 cos nφ(cos φ − cos θ ) + 2 cos θ (cos nφ − cos nθ ),
and observing
 π
cos nφ dφ = 0,
0

we divide each term by cos φ − cos θ and integrate with resepct to φ from 0 to
π. This gives the reduction formula
In+1 + In−1 = 2 cos θ In .
290 Singularities and Calculus of Residues

The general solution to the above reduction formula is given by

In = A sin nθ + B cos nθ,
where A and B are arbitrary constants. The constants A and B can be determined
using the initial values I0 = 0 and I1 = π . We obtain A = sinπ θ and B = 0. The
final result is
 π 
cos nφ sin nθ
P dφ = π , n ≥ 0.
0 cos φ − cos θ sin θ

Example 6.5.2 Show that

 ∞ px
e − eqx
dx = π(cot pπ − cot qπ),
−∞ 1 − e
x

provided that 0 < p < 1 and 0 < q < 1. Hence, deduce that
 ∞
sinh ax π aπ
dx = tan , if b > |a|.
−∞ sinh bx b 2b

Solution The point z = 0 is a simple pole of the two functions

epz eqz
and ;
1 − ez 1 − ez
however it is not so for the integrand
epz − eqz
.
1 − ez
Therefore,
 ∞  ∞
epx eqx
dx and dx
−∞ 1 − ex −∞ 1 − ex
are improper integrals, and it can be shown readily that their principal values
exist. To simplify the valuation procedure, we choose to split the integral as
follows:
 ∞ px  ∞ px  ∞ qx
e − eqx e e
dx = P dx − P dx.
−∞ 1 − e x
−∞ 1 − e x
−∞ − e
1 x

The integrand functions

epz eqz
and
1 − ez 1 − ez
have infinitely many simple poles at z = 2nπi, where n is any integer. The
two simple poles that are of interest in the present problem are z = 0 and
z = 2π i. The following closed rectangular contour C is chosen: (i) it has two
 6.5 Cauchy principal value of an improper integral 291

Figure 6.9. The contour is a closed rectangle with infinitesimal indentations at z = 0

and z = 2π i.

infinitesimal indentations at z = 0 and z = 2π i; (ii) the left and right vertical

lines are, respectively, z = −L + iy and z = L + iy, where 0 ≤ y ≤ 2π ; (iii)
the upper and lower horizontal lines are, respectively, z = x + 2π i and z = x,
where x ∈ [−L, L]\(−, ) (see Figure 6.9). The indentations are infinitesimal
semi-circles with common radius . They are adopted here so as to avoid the
passage of the integration path through the poles at z = 0 and z = 2π i. As a
result, there is no inclusion of the poles inside the contour. The contour C is
seen to consist of eight parts, so the contour integral around C can be split into
epz
eight line integrals. Consider the integration of 1−e z around the closed contour

C; since there is no pole enclosed inside the contour C, we have

,
epz
0= dz
C1−e
z
 − px  pz 
e 2π i e
= dx − Res ,0
−L 1 − e 1 − ez
x 2
 L px  2π p(L+iy)
e e
+ dx + i dy
 1−e 1 − eL+iy
x
0
  p(x+2πi)  pz 
e 2π i e
+ dx − Res , 2π i
L 1−e 1 − ez
x+2πi 2
 −L p(x+2πi)  0 p(−L+iy)
e e
+ dx + −L+iy
i dy.
− 1 − e x+2πi
2π − e
1

Subsequently, we take the limits  → 0 and L → ∞ in the above integrals.

The second and sixth terms in the right-hand expression arise from the contour
292 Singularities and Calculus of Residues

integration in the negative sense around the infinitesimal semi-circular inden-

tations at z = 0 and z = 2π i, respectively. Since z = 0 and z = 2π i are simple
epz
poles of 1−e z , by the result in Example 6.2.5, the value of the contour integral

around the infinitesimal semi-circle is equal to half of that around the whole
circle. For 0 < p < 1, it can be shown that the modulus values of the fourth
and the eighth integrals have the following asymptotic properties:
 
 2π
ep(L+iy) 
 i dy  ∼ O(e−(1−p)L ) → 0 as L → ∞,
 1 − eL+iy
0

 
 0
ep(−L+iy) 
 i dy  ∼ O(e−pL ) → 0 as L → ∞.
 −L+iy
2π 1 − e

Moreover, the respective residue values at the simple poles z = 0 and z = 2π i

are found to be
 pz   pz 
e e
Res , 0 = −1 and Res , 2π i = −e2pπi .
1 − ez 1 − ez

Combining the results, we have

 −  L

epx epx
(1 − e2pπi ) lim dx + dx
→0
−L 1 − ex  1 − ex

L→∞
∞
epx
= (1 − e2pπi )P dx
−∞ 1 − e
x
 pz   pz 

e e
= π i Res , 0 + Res , 2π i = −π i(1 + e2pπi ),
1 − ez 1 − ez

and so
 ∞
epx (epπi + e−pπi )/2 cos pπ
P dx = π −pπi
=π = π cot pπ.
−∞ 1 − e (e − e
x pπi )/2i sin pπ

For 0 < q < 1, we deduce a similar result

 ∞ qx
e
P dx = π cot qπ.
−∞ 1 − e
x

Finally, we obtain
 ∞
epx − eqx
dx = π(cot pπ − cot qπ).
−∞ 1 − e
x
 6.5 Cauchy principal value of an improper integral 293

To evaluate the second integral in the problem, we first rewrite the integrand
into the form
sinh ax eax − e−ax −e(a+b)x e(b−a)x
= bx = + .
sinh bx e − e−bx 1 − e2bx 1 − e2bx
When b > |a|, we have
b−a a+b
0< <1 and 0< < 1.
2b 2b
By taking p = b−a
2b
and q = a+b
2b
, we have
 ∞  ∞ (b−a)x  ∞ (a+b)x
sinh ax e e
dx = P dx − P dx
−∞ sinh bx −∞ 1 − e 2bx
−∞ − e
1 2bx
 ∞ b−a x  ∞ b+a x
1 e 2b 1 e 2b
= P dx − P dx
2b −∞ 1 − e x 2b −∞ 1 − ex
   

π b−a a+b
= cot π − cot π
2b 2b 2b
π '  π aπ   π aπ (
= cot − − cot +
2b 2 2b 2 2b
π aπ
= tan .
b 2b

Example 6.5.3 Find the Cauchy principal value of

 ∞ ix
e
dx,
−∞ x

and use the result to deduce that

 ∞
sin x π
dx = .
0 x 2

eix
Solution The integrand x
has an isolated singularity at x = 0. The Cauchy
principal value of
 ∞
eix
dx
−∞ x
is defined as
 ∞  −  R

eix eix eix

P dx = lim dx + dx .
−∞ x R→∞ −R x  x
→0

We consider the contour integral

,
eiz
dz,
C z
294 Singularities and Calculus of Residues

Figure 6.10. The closed contour C consists of the infinitely large upper semi-circle CR
of radius R, the line segments (−R, −) and (, R) along the x-axis, and the infinitesimal
upper semi-circle C of radius  around z = 0.

where C is the closed contour shown in Figure 6.10. An infinitesimal upper

semi-circle C around z = 0 is appended so that the integration path does not
cross the pole of the integrand. With this construction of the closed contour C,
the pole z = 0 is not included inside the contour.
iz
Since ez has no singularity inside C, we have
, iz
e
dz = 0.
C z

On the other hand, the above contour integral can be split into three parts,
namely,
, iz   ∞ ix 
e eiz e eiz
dz = dz + P dx + dz.
C z CR z −∞ x C z
iz
Since z = 0 is a simple pole of ez and C is looped in the negative sense, using
the result in Example 6.2.5, we obtain
  iz 
eiz e
dz = −π i Res , 0 = −π i.
C z z
By virtue of the Jordan lemma, the first integral

eiz
dz
CR z

is seen to be zero as R → ∞. Combining the results, we obtain

 ∞ ix 
e eiz
P dx = − dz = π i.
−∞ x C z
 6.6 Hydrodynamics in potential fluid flows 295

For the integral

 ∞
sin x
dx,
0 x

the point x = 0 is a removable singularity of the integrand since sin x

x
→ 1 as
x → 0. We observe that
  ∞   −  R 
eix sin x sin x
Im P dx = lim dx + dx
−∞ x R→∞
→0 −R x  x
 R  ∞
sin x sin x
= 2 lim dx = 2 dx,
R→∞
→0  x 0 x

so
 ∞   ∞ ix   
sin x 1 e iπ π
dx = Im P dx = Im = .
0 x 2 −∞ x 2 2

6.6 Hydrodynamics in potential fluid flows

In this section, we would like to illustrate the application of contour integration
and residue calculus to compute the hydrodynamic lifting force and the moment
on a body immersed in a potential flow field.
Consider a two-dimensional body immersed in a potential flow field whose
complex potential is given by w = f (z), z = x + iy. Let the boundary of the
body be represented by the closed curve γb , where the boundary itself is a
streamline. As the velocity values V along the boundary of the body are not
uniform, this causes the pressure P to vary along the boundary. This physical
property follows from the Bernoulli law of hydrodynamics, which states that
P + ρ2 |V |2 is constant along any streamline. Here, ρ is the density of the fluid,
which is assumed to be constant in the present discussion. Bernoulli’s law is
a rephrasing of the principle of conservation of energy. The non-uniformity of
pressure values along the boundary of the body leads to hydrodynamic force
and moment exerted on the body by the potential flow motion.

6.6.1 Blasius laws of hydrodynamic force and moment

The Blasius laws of hydrodynamic force F and the moment M exerted on
a body immersed in the potential flow field as characterized by the complex
296 Singularities and Calculus of Residues

Pidz

dz

Figure 6.11. The differential pressure force P i dz acts normal to the differential segment
dz on the boundary of an immersed body.

potential w = f (z) are given by

,
ρi
F = [f  (z)]2 dz, (6.6.1)
2 γ
 , 
ρ 
M = Re − 2
z[f (z)] dz , (6.6.2)
2 γ
respectively. Here, γ is any curve surrounding the body, provided that there is
no singularity included between γ and the body’s boundary γb .
To prove the above formulas, we consider the differential pressure force
exerted on the body normal to the differential length dz on the body’s boundary
γb . Since dz is along the tangent to the body, i dz lies along the interior normal
to γb (see Figure 6.11); so the differential pressure force dF is given by P i dz.
Let B denote the constant in the Bernoulli law. The differential pressure
force dF acting along the boundary of the body γb is then given by
' ρ (
dF = P i dz = B − |V (z)|2 i dz,
2
where V (z) = f  (z) is the velocity of the flow at the point z = x + iy. The
total force F = Fx + iFy exerted on the immersed body is given by the contour
integral around γb , where
,
F = Fx + iFy = P i dz
γb
, ,
ρi
= Bi dz − |V (z)|2 dz. (6.6.3)
γb γb 2

The first integral is zero since the integrand Bi is a constant. It is more preferable
to express the second integral in terms of the complex potential f (z). Since
the body is a streamline, the velocity along the body is tangential to the body.
When V (z) and dz are treated as vector quantities, they should have the same
 6.6 Hydrodynamics in potential fluid flows 297

angle of inclination. We then obtain

Arg V (z) = Arg f  (z) = Arg dz.
Let θ denote the common value of the above argument quantities. The integrand
in the second integral can be expressed as
ρi  ρi ρi
− [f (z)]2 e−2iθ dz = − [f  (z)]2 dz = [f  (z)]2 dz,
2 2 2
so
,
ρi
F = Fx − iFy = [f  (z)]2 dz. (6.6.4)
2 γb

By the Cauchy–Goursat integral theorem, we can replace γb by any simple

closed contour γ surrounding the body, provided that there is no singularity
enclosed between γ and γb .
Next, we would like to compute the hydrodynamic moment exerted on
the body immersed in the potential flow field. Let us consider the differen-
tial moment dM due to the differential pressure force dF = dFx + i dFy at
z = x + iy. Since the pressure force acts along the interior normal to γb , we
have
π π
Arg dF = + Arg dz = + θ.
2 2
The differential moment dM can then be expressed as
dM = x dFy − y dFx
' π  π (
= P |dz| x sin + θ − y cos +θ
2 2
= P |dz|(x cos θ + y sin θ ) = P Re(ze−2iθ dz).
The total moment exerted on the body is given by
, ' (
ρ
M= B − |V (z)|2 Re(ze−2iθ dz)
γb 2
,
ρ 
= Re − [f (z)]2 e2iθ ze−2iθ dz
γb 2
 , 
ρ
= Re − z[f  (z)]2 dz . (6.6.5)
2 γb
Again, provided that there is no singularity included between γ and γb , we may
replace γb by γ . Thus we obtain the desired result in (6.6.2).

Example 6.6.1 Consider the potential flow field in the right half-plane,
Re z > 0, due to a field source of strength m located at z = a, a > 0. An
298 Singularities and Calculus of Residues

fictitious
source
• z=a x
z = −a

wall
Figure 6.12. The combination of the real source at z = a and the fictitious source at
z = −a produces a flow field with the y-axis as a streamline.

infinite wall is placed along the y-axis. Find the total pressure force exerted on
the wall due to the fluid source.

Solution Consider the effect of adding a fictitious image source of source

strength m at z = −a. The complex potential of the flow field in the whole
complex plane due to the two sources is given by
m
f (z) = [Log(z − a) + Log(z + a)].
2π
When z assumes either real values or imaginary values, Im f (z) becomes
constant, indicating that both the real and imaginary axes are streamlines of the
flow field. Effectively, we may replace the streamlines by non-penetrable walls
without distorting the flow field since the fluid particles flow tangentially along
the streamlines, as they would flow along the walls. Since we are seeking the
solution in the right half-plane, it is appropriate to place the wall along the y-
axis. By limiting the domain of definition to Re z > 0, the above f (z) is already
the solution for the complex potential of the given flow field. This method is
called the method of images since the fictitious source can be considered as the
mirror image of the physical source with respect to the wall (considered as the
mirror). Several streamlines of the flow field are sketched in Figure 6.12.
The square of the speed of the fluid flow is given by
  
df df m2 1 1 1 1
= + +
dz dz 4π 2 z + a z−a z+a z−a
m2 x2 + y2
= 2 2 .
π (x + y 2 )2 − 2a 2 (x 2 − y 2 ) + a 4
 6.6 Hydrodynamics in potential fluid flows 299

According to the Blasius law of hydrodynamic force, the pressure force exerted
on the wall is obtained by integrating along the y-axis from y = ∞ to y = −∞
(following the anticlockwise sense of the integration contour) ρi2 df df
dz dz
evaluated
at x = 0. This gives
  
ρi −∞ df df 
F = i dy, dz = i dy
2 ∞ dz dz x=0

ρ ∞ m2 y2 ρm2
= dy = .
2 −∞ π 2 (y 2 + a 2 )2 4π a
As F is a real quantity, the direction of the net force is horizontal. The positive
sign indicates that the net force is to the right, that is, the wall is sucked towards
the source.

6.6.2 Kutta–Joukowski’s lifting force theorem

Consider an airfoil immersed in an otherwise uniform potential flow with the
free stream velocity U∞ (see Figure 5.3), and a circulation of magnitude  (can
be either positive or negative) developed around the airfoil. Using the Blasius
law of hydrodynamic force, it can be shown that the airfoil experiences a lift
of magnitude ρ|U∞ | exerted by the potential flow field with fluid density
ρ. The direction of the lifting force vector is obtained by rotating the free
stream velocity vector through a right angle in the sense opposite to that of the
circulation.
First, we observe that the general representation of the complex potential
f (z) of the flow field due to the presence of the airfoil together with circulation
 around the airfoil is given by
 b1 b2
f (z) = U∞ z + Log z + + 2 + ··· , (6.6.6)
2π i z z
given that the origin is placed inside the airfoil. To justify the above claim, the
first term is seen to correspond to the free stream U∞ at |z| → ∞ since
lim f  (z) = U∞ . (6.6.7)
|z|→∞

The second term is seen to account for the circulation around the airfoil. This
is because the circulation around the airfoil is given by
,
 
Re f  (z) dz = Re 2π i Res(f  (z), 0)
γb
 

= Re 2π i = . (6.6.8)
2π i
300 Singularities and Calculus of Residues

Applying the Blasius law of hydrodynamic lifting force, we use eq. (6.6.1)
to compute the lifting force exerted on the airfoil by the potential fluid flow. To
evaluate the contour integral, it suffices to consider the Laurent series expansion
of the integrand function [f  (z)]2 up to the first negative power term. The
Laurent series expansion of [f  (z)]2 is seen to be

2 U∞ 1
[f  (z)]2 = U∞ + + ··· .
πi z
Accordingly, the lifting force exerted on the airfoil is given by
, % &
ρi 2 U∞ 1
F = U∞ + + · · · dz
2 γb πi z
 ρ 
= 2π i U ∞ = −iρU∞ . (6.6.9)
2π
The magnitude of the lifting force equals ρ|U∞ |. Due to the factor −i, the
direction of the lifting force is at right-angles to U∞ , rotating in the sense
opposite to that of the circulation.

6.7 Problems
6.1. Suppose the radius of convergence of the Taylor series
∞

cn zn , cn ≥ 0,
n=0

equals 1; explain why z = 1 is a singular point.

6.2. Discuss the nature of the isolated singularity z = 0 of the following
functions:

sin z z = 0
(a) f1 (z) = ;
1 z=0

sin z z = 0
(b) f2 (z) = ;
∞ z=0

sin 1z z = 0
(c) f3 (z) = .
1 z=0
6.3. Suppose z = 0 is a pole of orders n and m, respectively, of the functions
f1 (z) and f2 (z). For the following functions, find the possible order of
the pole z = 0:

(a) f1 (z) + f2 (z); (b) f1 (z)f2 (z); (c) f1 (z)/f2 (z).

 6.7 Problems 301

6.4. Suppose z = z0 is an essential singularity of f1 (z) and f2 (z). Can the

point z = z0 become a removable singularity or a pole of the function
f1 (z) + f2 (z)? If yes, construct such examples.
6.5. Find all the isolated singularities of each of the following functions in the
unextended complex plane and classify the nature of these singularities:

z2 + 1 1 1 1
(a) ; (b) z − ; (c) e−z cos ; (d) sin sin1 1 .
e z e −1 z z z

6.6. The function

1
f (z) =
z(z − 1)2
has a pole of order 2 at z = 1, and it admits the following Laurent series
expansion:
1 1 1 1
= − + − · · · + · · · , |z − 1| > 1.
z(z − 1)2 (z − 1)3 (z − 1)4 (z − 1)5
We then conclude that z = 1 is an essential singularity since the Laurent
series expansion has infinitely many negative power terms. Also, since
1
the coefficient of z−1 is zero in the above expansion, we claim that

Res(f (z), 1) = 0.

Explain why the above claims on the nature of the singularity and residue
value are incorrect.
6.7. For each of the following functions, find all the isolated singularities in
the unextended complex plane and evaluate the residue value at each of
these singularities.

z2 − 1 tan z eiαx
(a) ; (b) ; (c) , α and β are real;
z3 (z2 + 1) 1 − ez z4 + β 4
1 e1/z
(d) ; (e) .
z sin z z
  
(z)
6.8. Evaluate Res ff (z) , α if

(a) α is a zero of order n of f (z),

(b) α is a pole of order n of f (z).

(z)
Let g(z) be analytic at α; evaluate Res(g(z) ff (z) , α) subject to the same
conditions as in (a) and (b).
302 Singularities and Calculus of Residues

6.9. Classify the nature of the isolated singularity of each of the following
functions:
1 1
(a) sin at z = 1; (b) 3 z3 at z = 0.
z−1 z (e − 1)
6.10. Suppose f (z) and g(z) are analytic at z = z0 , f(z0 ) = 0, and z = z0 is
(z)
a zero of order 2 of g(z). Express Res fg(z) , z0 in terms of the Taylor
coefficients of f (z) and g(z) at z = z0 .
6.11. Locate the isolated singularities of the function f (z) = π cot π z, and
determine whether each is a removable singularity, a pole or an essential
singularity. If the singularity is removable, give the limit of the function
at the point. If the singularity is a pole, give the order of the pole, and
compute the residue at the singularity.
6.12. Compute the residue at each of the isolated singularities of
cos z
f (z) = .
z2 (z− π )3
6.13. What is the order of the pole of
1
f (z) =
(2 cos z − 2 + z2 )2
at z = 0? Compute Res (f, 0).
6.14. Compute
   
(a) Res tan z, k + 12 π , k is any integer;
 2n 
z
(b) Res (z−1) n,1 .

6.15. Suppose α is an isolated singularity of f (z); show that

(a) Res(f, α) = Res(f, −α) if f (z) is odd;

(b) Res(f, α) = −Res(f, −α) if f (z) is even.

6.16. Let a1 , a2 , . . . , an be distinct points in the complex plane and suppose

all of them lie within the circle |z| = R. Suppose f (z) is analytic on and
inside the circle; evaluate the integral
,
f (z)
dz.
|z|=R (z − a1 )(z − a2 ) · · · (z − an )
6.17. A proper rational function f (z) with only poles of first order can be
represented by
b0 zn + · · · + bn−1 z + bn
f (z) = ,
(z − z1 )(z − z2 ) · · · (z − zk )
 6.7 Problems 303

where n < k, and z1 , z2 , . . . , zk are distinct. Show that the corresponding

partial fraction decomposition of f (z) takes the form
c1 c2 ck
f (z) = + + ··· + ,
z − z1 z − z2 z − zk

where cj = Res(f, zj ), j = 1, 2, . . . , k. Apply the result to find

the partial fraction decomposition of each of the following functions:

z+1 z2 1
(a) ; (b) ; (c) .
(z − 1)(z − 2)(z − 3) z5 + 1 zn −1

6.18. Show that the value of each of the following integrals is zero.
, ,
2z
(a) ze dz; (b) tanh z dz;
|z|=1 |z|=1
, ,
1 1
(c) dz; (d) dz;
|z|=1 cos z |z|=3 z2 +1
,
tan z
(e) dz.
|z|=2 z

6.19. Use the Cauchy residue theorem to evaluate the following integrals:
, ,
z4 + z z3 + 3z + 1
(a) dz; (b) dz;
|z|=2 (z − 1) |z|=2 z − 5z
2 4 2
, ,
sinh2 z ez + z
(c) dz; (d) dz;
|z−i|=2 (z − 1)
z 4 4
|z|=2
,
e−z sin z
(e) dz;
|z−i|=2 z2
,
sin z
(f) dz, n is any positive integer.
|z−i|=2 (z − i)
n

6.20. Evaluate the following definite integrals. When the integral is improper,
find its principal value if it exists.
 2π
dθ
(a) , a > b > 0;
0 (a + b cos θ )2
 2π
dθ
(b) , a > b > 0;
0 (a + b cos2 θ )2
 2π
dθ
(c) , a is a complex number and a = ±1.
0 1 − 2a cos θ + a 2
304 Singularities and Calculus of Residues

6.21. Evaluate the following integrals:

 ∞  ∞
x x2
(a) dx; (b) dx, a > 0;
−∞ (x + 4x + 13) (x 2 + a 2 )2
2 2
0
 ∞
1
(c) dx, n is a positive integer;
0 (x + 1)n
2
 ∞  ∞ 2
1 x +1
(d) 2 + a 2 )(x 2 + b2 )
dx, a > 0, b > 0; (e) dx;
−∞ (x 0 x4 + 1
 ∞
1
(f) dx, n ≥ 2 and n is a positive integer;
0 1 + xn
 ∞
x4
(g) dx.
−∞ 1 + x
8

6.22. Show that

 2π
ln |eiθ − 1| dθ = 0,
0

and use the result to deduce

 π
ln sin θ dθ = −π ln 2.
0

6.23. Find the Fourier transform of each of the following functions:


1 −t 2 1 |t| < β
(a) u(t) = ; (b) u(t) = e ; (c) u(t) = .
1+t 2 0 |t| > β

6.24. Suppose ue (x) is an even function, where ue (−x) = ue (x), x > 0. Show
that
 ∞
F{ue (x)} = 2Uc (w) = 2 cos wx ue (x) dx,
0

and hence deduce the Fourier cosine transform inversion formula given
in eq. (6.4.11).
6.25. Use the Parseval identity to show that
 ∞ −iaω
1 |e − e−ibω |2
dω = b − a,
2π −∞ ω2
and deduce that

1 ∞ sin aω sin bω
dω = min(a, b), a > 0 and b > 0.
π −∞ ω2
 6.7 Problems 305

6.26. Show that each of the following functions satisfies the defining properties
of the Dirac function δ(x) when the limit  → 0 is taken:

1 
(a) √ e−x / ;
2
(b) .
π π(x 2 + 2)

6.27. Evaluate the following Fourier integrals:

 ∞
x cos x
(a) dx;
−∞ − 2x + 10
x2
 ∞
x sin x
(b) 2 + 4x + 20
dx;
−∞ x
 ∞
cos ax
(c) 2 + b2
dx, a and b are positive;
−∞ x
 ∞
x sin ax
(d) dx, a and b are positive;
−∞ x + b
2 2
 ∞
sin ax
(e) dx.
0 x(x 2 + 1)

6.28. Let f (z) = eimz F (z), m > 0, and the function F (z) have the following
properties:

(a) it has a finite number of isolated singularities z1 , z2 , . . . , zn

in the upper half-plane;
(b) it is analytic at all points on the real axis, except at the points a1 ,
a2 , . . . , am , which are simple poles;
(c) F (z) → 0 if z → ∞ and Im z ≥ 0.

Show that

 % &
∞ 
n
1
m
P f (x) dx = 2π i Res(f (z), zk ) + Res(f (z), ak ) .
−∞ k=1
2 k=1

6.29. Evaluate
 ∞
xex
dx.
−∞ e4x + 1
306 Singularities and Calculus of Residues

Hint: Choose the closed rectangular contour whose four sides are

1 : {(x, y) : −R ≤ x ≤ R, y = 0}
 π
2 : (x, y) : x = R, 0 ≤ y ≤
 2 
π
3 : (x, y) : −R ≤ x ≤ R, y =
 2
π
4 : (x, y) : x = −R, 0 ≤ y ≤ .
2

6.30. Evaluate
 ∞
ln x
dx.
0 x2 + 4
Note that ln x has a singularity at x = 0. The improper integral is thus
defined as
 R
ln x
lim dx.
→0
R→∞ 
x +4
2

6.31. Show that

 ∞
ln(x 2 + 1)
dx = π ln 2.
0 x2 + 1

Hint: Use the relation

Log(i − x) + Log(i + x) = Log(i 2 − x 2 ) = ln(x 2 + 1) + π i.

6.32. Suppose a > 0; show that

 a+i∞ zt 
1 e 1 t >0
dz = .
2π i a−i∞ z 0 t <0
The path of contour integration is along the infinite vertical line Re z = a.

6.33. The following integral occurs in the quantum theory of atomic collisions:
 ∞
sin t ipt
I= e dt,
−∞ t

where p is real. Show that


0 |p| > 1
I= .
π |p| < 1
Discuss the cases when p = ±1.
 6.7 Problems 307

6.34. Suppose the analytic functions P (z) and Q(z) both have a zero at z0 ;
prove that the L’Hospital rule is given by
P (z) P  (z)
lim = lim  .
z→z0 Q(z) z→z0 Q (z)
6.35. Show that
 ∞  
x 2p − x 2q π 2p + 1 2q + 1
dx = cos π − cos π ,
−∞ 1−x 2r r r r
where p, q, r are non-negative integers, p < r and q < r.

Hint: Choose the closed contour of integration that includes the

infinitely large semi-circle in the upper half-plane.

6.36. Suppose f (z) has poles α1 , α2 , . . . , αn , none of which lies on the positive
real axis or equals zero, and there exists a real number p such that
   
lim zp+1 f (z) = lim zp+1 f (z) = 0.
z→0 z→∞

Prove that
 ∞
x p f (x) dx
0


 π n

 − e −pπi
Res(zp f (z), αk ) p is not an integer
 sin pπ
= n
k=1



−
p
 Res(z f (z) log z, αk ) p is an integer.
k=1
2
6.37. Suppose we integrate eiz along the closed wedge-shaped contour shown
in the figure below, and take the limit R → ∞ subsequently; show that
 ∞  ∞ √
2π
cos x dx =
2
sin x dx =
2
.
0 0 4

Re i
>

>

4 >
x
0 R
308 Singularities and Calculus of Residues

6.38. Show that

 ∞ √ 
2π

−ax 2
e cos bx dx = √
2
a + a 2 + b2 , a > 0.
0 4 a +b
2 2

Hint: Use a wedge-shaped closed contour similar to that in

Problem 6.37, where the angle of the wedge should be chosen
appropriately according to the parameter values a and b.

6.39. Evaluate the Cauchy principal value of each of the following integrals:
 ∞  ∞
xeix eimx
(a) dx; (b) dx, m > 0.
−∞ x − π −∞ (x − 1)(x − 2)
2 2

6.40. Show that

 ∞
cos(ln x) π
dx = .
0 1+x 2 2 cosh π2

y
Ri
<
<

i
x
i

Ri

Hint: Integrate the function

zi
f (z) =
−1 z2
along the closed semi-circle with an infinite radius and an
infinitesimal indentation at z = 0 in the right-hand plane.
az
e
6.41. Suppose we integrate sinh πz
, a is real, along the indented closed rect-
angular contour shown below, and take the limits L → ∞ and  → 0;
show that  ∞
sinh ax 1 a
dx = tan , |a| < π.
0 sinh π x 2 2
 6.7 Problems 309

L+i > L+i

>
>

> x
L L
6.42. By considering the contour integral
,
1 − e2iz
dz,
C z2
where C is the closed contour depicted in Figure 6.10, show that
 ∞
sin2 x π
2
dx = .
0 x 2
Use a similar technique to evaluate
 ∞
sin3 x
dx.
0 x3
6.43. Consider the potential flow over the circle |z| = a, where the free stream
velocity U∞ is aligned with the positive x-axis and a circulation  > 0
is developed around the circle. Show that the complex potential of the
flow field is given by
 
a2  z
f (z) = U∞ z + + Log .
z 2π i a
310 Singularities and Calculus of Residues

Using the Blasius laws of hydrodynamic force and moment, show that
the circle experiences a downward force of magnitude ρU∞ , where ρ
is the density of the fluid, and the hydrodynamic moment acting on the
circle is zero.
6.44. Show that the Blasius laws of hydrodynamic force and moment can be
expressed in the alternative forms

,  
∂ψ 2
F = −2ρi dz,
γ ∂z
,  
∂ψ 2
M = Re 2ρz dz,
γ ∂z
where ψ(z) is the stream function of the fluid flow.

Hint: Use the relation

∂ψ
u − iv = −2i
∂z
and
∂ψ
u + iv = 2i ,
∂z
where u and v are the x- and y-components of velocity,
respectively. Further, on the boundary of the body, we have
∂ψ ∂ψ
0 = dψ = dz + dz.
∂z ∂z
By combining these relations, show that
 
∂ψ 2
dF = −2ρi dz.
∂z
 7
Boundary Value Problems and
Initial-Boundary Value Problems

In the earlier chapters, we have analyzed several prototype potential field prob-
lems, including potential fluid flows, steady state temperature distribution, elec-
trostatics problems and gravitational potential problems. All of these potential
field problems are governed by the Laplace equation. There is no time variable
in these problems, and the characterization of individual physical problems is
exhibited by the corresponding prescribed boundary conditions. The mathe-
matical problem of finding the solution of a partial differential equation that
satisfies the prescribed boundary conditions is called a boundary value prob-
lem, of which there are two main types: Dirichlet problems where the boundary
values of the solution function are prescribed, and Neumann problems where
the values of the normal derivative of the solution function along the bound-
ary are prescribed. In other physical problems, like the heat conduction and
wave propagation models, the time variable is also involved in the model.
To describe fully the partial differential equations modeling these problems,
one needs to prescribe both the associated boundary conditions and the initial
conditions. The latter class is called an initial-boundary value problem. This
chapter discusses some of the solution methodologies for solving boundary
value problems and initial-boundary value problems using complex variables
methods.
The link between analytic functions and harmonic functions is exhibited by
the fact that both the real and imaginary parts of a complex function that is
analytic inside a domain satisfy the Laplace equation in the same domain. The
Gauss mean value theorem (see Subsection 4.3.3) states that the value of a
harmonic function at the center of any circle inside the domain of harmonicity
equals the average of the values of the function along the boundary of the cir-
cle. In Section 7.1, we generalize the above result by establishing two forms of
integral representation of harmonic functions: the Poisson integral formula and
the Schwarz integral formula. We also discuss the properties of solutions to the

311
312 Boundary Value Problems and Initial-Boundary Value Problems

Dirichlet and Neumann problems. In particular, the compatibility conditions

required for the existence of solutions to Neumann problems are examined.
In Section 7.2, we discuss the Laplace transform and its inversion method.
Though the Laplace transform is applied to functions that are real, its inversion
formula involves the evaluation of the Bromwich contour integral in the com-
plex plane. The last section is devoted to the discussion of application of the
Laplace transform method to solutions of two prototype initial-boundary value
problems, namely, heat conduction in a thin rod and longitudinal oscillations
of an elastic thin rod.

7.1 Integral formulas of harmonic functions

Recall that the Cauchy integral formula gives the value of an analytic function
f inside a contour C in terms of its values on the contour:
,
1 f (ζ )
f (z) = dζ, z lies inside C. (7.1.1)
2π i C ζ − z
This harmonic-analytic dualism leads us to expect that the solution to the
Laplace equation inside a domain D can be expressed in terms of the values of
the function along the boundary of D in the form of an integral. In this section,
we derive the integral representation formulas of harmonic functions with the
Dirichlet-type and Neumann-type boundary conditions. The types of domain
are limited to circles and the upper half-plane.

7.1.1 Poisson integral formula

We would like to find the integral representation of a function u(x, y)
that is harmonic on and inside the circle |z| = R together with the pre-
scribed boundary condition u(R, θ ) along the boundary of the circle. Let
f (z) = u(x, y) + iv(x, y), z = x + iy, be analytic on and inside the circle
|z| = R. By the Cauchy integral formula, we have
,
1 f (ζ )
f (z) = dζ. (7.1.2)
2π i |ζ |=R ζ − z
The symmetry point of z with respect to the circle |z| = R is given by (see
Subsection 1.3.2)
R2 ζζ
zs = = ,
z z
where ζ is a point on the circle (see Figure 7.1). Since the symmetry point zs is
(ζ )
outside the circle, the function ζf−z s
is analytic on and inside |ζ | = R; we then
 7.1 Integral formulas of harmonic functions 313

ζ
×
zs
φ ×
θz x

Figure 7.1. The pair of symmetry points z and zs with respect to the circle |ζ | = R
are on the same ray through the center. The polar forms of z and ζ are reiθ and Reiφ ,
respectively, and |ζ − z|2 = R 2 − 2Rr cos(φ − θ ) + r 2 .

have
,  
1 1 1
f (z) = f (ζ ) − dζ, (7.1.3)
2π i |ζ |=R ζ − z ζ − zs

by virtue of the Cauchy–Goursat integral theorem. Consider the quantity

ζ ζ ζ ζ ζ z |ζ |2 − |z|2
− = − = + = ,
ζ − z ζ − zs ζ −z ζ − ζζ ζ −z ζ −z |ζ − z|2
z

which is manifestly real. We write ζ = Reiφ and z = reiθ ; then the above
quantity can be expressed as

|ζ |2 − |z|2 R2 − r 2 R2 − r 2
= = .
|ζ − z|2 ζ ζ − (ζ z + ζ z) + zz R 2 − 2Rr cos(φ − θ ) + r 2

By observing that

dζ = iReiφ dφ = iζ dφ,

the contour integral in eq. (7.1.3) can be expressed as


1 2π
R2 − r 2
f (z) = f (Reiφ ) dφ. (7.1.4)
2π 0 R2 − 2Rr cos(φ − θ) + r 2
Finally, we take the real parts of both sides of the above equation. As a result,
we obtain the Poisson integral formula for the solution of the Dirichlet problem
314 Boundary Value Problems and Initial-Boundary Value Problems

in the circular domain |z| < R, which takes the form

 2π
1
u(r, θ ) = P (R, r, φ − θ )u(R, φ) dφ, r < R. (7.1.5)
2π 0

The kernel

R2 − r 2
P (R, r, φ − θ) = (7.1.6)
R2 − 2Rr cos(φ − θ ) + r 2

is called the Poisson kernel. The Poisson integral formula resembles an integral
transform of the boundary value u(R, φ), with 2π 1
P (R, r, φ − θ ) as the trans-
form kernel. This integral formula implies that u(r, θ ), r < R, is determined
completely by its boundary value u(R, φ) on the circle.
In the above derivation of the Poisson integral formula, it has been assumed
that f is analytic on and inside the circle |z| = R. This implies implicitly that
the real part u(R, θ ) would be a continuous function of θ on the circle. However,
the continuity of the boundary value function u(R, θ ) can be relaxed. It can be
proved rigorously that the Poisson integral formula holds even when u(R, θ ) is
piecewise continuous.

Fourier series expansion

The Poisson kernel can be expressed as

 
ζ z z ζ 1 ζ +z ζ +z
P (R, r, φ − θ ) = + = + = +
ζ −z ζ −z ζ −z ζ −z 2 ζ −z ζ −z
   
ζ +z 2
= Re = Re −1 +
ζ −z 1 − ζz
 
2
= Re −1 + .
1 − Rr e−i(φ−θ)

Since r < R, the Poisson kernel can be decomposed into the following Fourier
cosine series:

∞  
 r n
P (R, r, φ − θ ) = 1 + 2 cos n(φ − θ ).
n=1
R
 7.1 Integral formulas of harmonic functions 315

The harmonic function u(r, θ ) can then be expressed as

 2π
1
u(r, θ ) = u(R, φ) dφ
2π 0
∞   r n
1  2π
+ u(R, φ) cos n(φ − θ ) dφ
π n=1 0 R
 2π
1
= u(R, φ) dφ
2π 0
∞    
1 2π r n
+ u(R, φ) cos nφ dφ cos nθ
n=1
π 0 R
∞    
1 2π r n
+ u(R, φ) sin nφ dφ sin nθ.
n=1
π 0 R

We write

1 2π
an = u(R, φ) cos nφ dφ, n = 0, 1, 2, . . . ,
π 0

1 2π
bn = u(R, φ) sin nφ dφ, n = 1, 2, . . . .
π 0
Here, an and bn are recognized as the Fourier coefficients in the Fourier expan-
sion of the boundary value function u(R, φ). The Fourier series expansion of
u(r, θ ) can then be expressed as

a0   r n   r n
∞ ∞
u(r, θ ) = + an cos nθ + bn sin nθ. (7.1.7)
2 n=1
R n=1
R

Properties of the Poisson kernel

(i) Since ζ = Reiφ , the function
ζ +z
ψ(z) =
ζ −z
is an analytic function of z inside the circle |z| = R. Therefore, its real
part P (R, r, φ − θ) is harmonic inside the same circle.
(ii) Note that the function u(r, θ ) = 1 is harmonic inside the domain |z| < R
and satisfies the boundary condition u(R, θ ) = 1. By setting u(r, θ ) = 1
and u(R, θ ) = 1 in the Poisson integral formula, we then obtain
 2π
1
P (R, r, φ − θ ) dφ = 1. (7.1.8)
2π 0
316 Boundary Value Problems and Initial-Boundary Value Problems

(iii) For fixed values of R, r and θ , the maximum value of P (R, r, φ −

θ) over φ ∈ [0, 2π ] occurs when its denominator attains its minimum
value. This occurs at φ = θ , for any value of r. The corresponding
maximum value is found to be
R+r
max P (R, r, φ − θ ) = ,
0≤φ≤2π R−r
which tends to infinity as r → R.
Similarly, the minimum value of P (R, r, φ − θ ) occurs at either
φ = θ + π or φ = θ − π (only one of these two values falls inside
[0, 2π]), corresponding to the points where R 2 − 2Rr cos(φ − θ ) + r 2
attains its maximum value. We then have
R−r
min P (R, r, φ − θ) = → 0 as r → R.
0≤φ≤2π R+r
(iv) Let f (θ ) be piecewise continuous over [0, 2π ]. For each θ ∈ [0, 2π ],
we state without proof the following result:
 2π
1
lim P (R, r, φ − θ )f (φ) dφ
r→R 2π 0

f (θ ) if f is continuous at θ
=
αf (θ − ) + (1 − α)f (θ + ), 0 < α < 1 if f has a finite jump at θ .
(7.1.9)
The value α depends on the angle of approach to the boundary point
ζ = Reiφ from the interior of the circle (see Problem 7.5 and Exam-
ple 7.1.1). In particular, α equals 12 when the boundary point ζ = Reiφ
is approached radially.

Uniqueness of solution for Dirichlet problems

The integral representation of the solution to the Dirichlet problem for a circular
domain was established in eq. (7.1.5). Is that solution unique? The uniqueness
of the solution for Dirichlet problems with arbitrary domain can be established
quite readily using the maximum principle for harmonic functions.
Let φ(z) be the boundary value prescribed along the boundary of the domain
for a Dirichlet problem. Suppose the Dirichlet problem admits two solutions
u1 (z) and u2 (z) inside the domain D, and both solutions satisfy the prescribed
boundary condition along the boundary ∂D. Define u = u1 − u2 ; then by lin-
earity of the Laplace equation, u is also harmonic in D since u1 and u2 are both
harmonic in D. The boundary condition for u becomes
u(z) = φ(z) − φ(z) = 0, for z ∈ ∂D.
 7.1 Integral formulas of harmonic functions 317

Using the maximum principle for harmonic functions (see Subsection 4.3.3), the
maximum value of u(z) for any point z inside D must be zero since its boundary
value is zero. We then have u(z) = 0, or equivalently u1 = u2 , throughout D.
The uniqueness of the solution is then established. An alternative proof using
Green’s theorem is outlined in Problem 7.10.

Example 7.1.1 An infinitely long metal cylinder of radius R is cut into two
halves. The upper half is grounded and the lower half is maintained at a fixed
potential K. Find the electrostatic potential at points inside the cylinder.

Solution The infinite length of the metal cylinder fits well the two-
dimensional consideration of the problem. The posed problem is equivalent to
the Dirichlet problem of finding the harmonic function in the domain |z| < R
with boundary condition

0 0<φ<π
f (φ) = .
K π < φ < 2π

There is a discontinuity at both φ = 0 and φ = π . Using the Poisson integral

formula, the solution to the electrostatic potential u(r, θ ) is given by

1 2π
R2 − r 2
u(r, θ ) = K dφ.
2π π R2 − 2Rr cos(φ − θ) + r 2

To evaluate the integral, we set t = tan φ−θ

2
so that dφ = 2
1+t 2
dt. The integral
is then reduced to
 tan(π − θ2 )
K R2 − r 2 2
u(r, θ ) = dt
+ r2 1 + t2
2
2π tan( π2 − θ2 ) R2 − 2Rr 1−t
1+t 2
 tan( π − θ2 )
K (R + r)/(R − r)
= dt
π tan( ) 1
π
2
+ (R + r)2 t 2 /(R − r)2
− θ2
 

K R+r φ − θ 2π
= tan−1 tan
π R−r 2 π
 2 2 
K R − r
= tan−1 .
π 2Rr sin θ

The equation of the equipotential line u(r, θ ) = u0 , 0 < u0 < K, is given by

 
K R2 − r 2
u0 = tan−1 , r < R.
π 2Rr sin θ
318 Boundary Value Problems and Initial-Boundary Value Problems

[0]

−R [K/2] R

[K]

Figure 7.2. The equipotential lines of the charged cylinder are circular arcs passing
through the two points z = R and z = −R.

By substituting r 2 = x 2 + y 2 , r sin θ = y, and rearranging the terms, we obtain

 π u0 
x 2 + y 2 + 2R tan y = R2, x 2 + y 2 < R2.
K
The equipotential lines are circular arcs passing through the two points z = R
and z = −R. These two points exhibit a jump of discontinuity in the boundary
value. The equipotential lines of the charged cylinder are shown in Figure 7.2.
The pattern of the equipotential lines reveals that the electric potential values
at points close to z = R and z = −R depend on the direction of approach to
that boundary point [see also eq. (7.1.9)].

Example 7.1.2 Find the function u(r, θ ) that is harmonic inside the domain
{z : |z| < R and Im z > 0} and satisfies the boundary condition

f (θ ) 0 < θ < π, r=R
u(R, θ ) = .
0 θ = 0 or θ = π, r < R

Solution The boundary value function f (θ ) has a discontinuity at θ = 0 and

θ = π. Since u(R, 0) = u(R, π ) = 0, we may try to consider the odd extension
of f (θ ) over (π, 2π) by defining

f (θ ) = −f (2π − θ ) for π < θ < 2π.

Now, with the boundary value of the extended f (θ ) prescribed along the whole
circumference of the circle, the solution to the Dirichlet problem in the upper
 7.1 Integral formulas of harmonic functions 319

u ( R, ) = f ( )

( R, ) ( R, )
x
u ( r, ) = 0 u ( r, ) = 0
Figure 7.3. The Dirichlet problem in a semi-circular domain is formulated by prescrib-
ing the boundary condition along the upper semi-circle and the diameter.

half-disk is given by
 π
1
u(r, θ ) = P (R, r, φ − θ )f (φ) dφ
2π 0
 2π
1
− P (R, r, φ − θ )f (2π − φ) dφ
2π
 π
π
1
= [P (R, r, φ − θ ) − P (R, r, φ + θ )]f (φ) dφ,
2π 0
r < R, 0 < θ < π,
where P (R, r, φ − θ ) is the Poisson kernel.
Since P (R, r, φ − θ ) and P (R, r, φ + θ ) are harmonic inside |z| <
R, u(r, θ ) should also be harmonic in the same domain. One can check easily
that
u(r, 0) = u(r, π ) = 0, r < R,
hence the solution satisfies the boundary condition along the real axis.
Also, the boundary condition on the upper half-disc is satisfied, by virtue
of eq. (7.1.9). Hence, u(r, θ ) as defined above solves the given Dirichlet
problem.

7.1.2 Schwarz integral formula

Next we consider the problem of finding the real function that is harmonic in
the upper half-plane Im z > 0 and satisfies the prescribed Dirichlet boundary
condition along the real axis. Again, we derive the integral representation of
the solution in terms of the boundary values.
320 Boundary Value Problems and Initial-Boundary Value Problems

Figure 7.4. The point z lies outside the closed contour C since it is the mirror image of
z with respect to the real axis.

Let f (z) be analytic and satisfy the order property |zk f (z)| < M, for some
positive constants k and M, in the closed upper half-plane Im z ≥ 0. Let C
denote the positively oriented closed contour which consists of the upper half
semi-circle CR with radius R and the line segment from −R to R along the
real axis (see Figure 7.4). Let z be any point inside C. By the Cauchy integral
formula, we have
,  
1 1 1
f (z) = − f (ζ ) dζ
2π i C ζ − z ζ − z
  
1 1 1
= − f (ζ ) dζ
2π i CR ζ − z ζ − z
 R 

1 1
+ − f (ζ ) dζ . (7.1.10)
−R ζ − z ζ −z
Note that the contribution from the second term in each of the above integrals
is zero since z lies outside the contour C.
Since |f (ζ )| < RMk for points ζ lying on CR , the asymptotic order of the
modulus of the integral satisfies
      
 1 1 1  1
 − 
f (ζ ) dζ  ∼ O , k > 0, (7.1.11)
 2π i ζ − z ζ − z R k
CR

which tends to zero as R goes to infinity. Combining eqs. (7.1.10) and (7.1.11),
and taking the limit R → ∞, we then obtain
 ∞ 
1 1 1
f (z) = − f (t) dt
2π i −∞ t − z t − z
 ∞
1 z−z
= f (t) dt, Im z > 0.
2π i −∞ |t − z|2
 7.1 Integral formulas of harmonic functions 321

Suppose we write f (z) = u(x, y) + iv(x, y), z = x + iy. By taking the real
parts on both sides of the above equation, we obtain

1 ∞ y
u(x, y) = u(t, 0) dt
π −∞ (t − x)2 + y 2
 ∞
1
= S(t − x, y) u(t, 0) dt, y > 0, (7.1.12)
π −∞
where the kernel function
y
S(t − x, y) =
(t − x)2 + y 2
is called the Schwarz kernel. The above integral representation formula is called
the Schwarz integral formula.
In the above proof, we require f (z) to be analytic and satisfy the order
property |zk f (z)| < M in the closed region Im z ≥ 0. However, it can be
proved rigorously that the Schwarz integral formula remains valid with less
stringent requirements on the boundary value function u(x, 0). The precise
requirements are that u(x, 0) is bounded for all x and continuous except for at
most a finite number of finite jumps.
The Schwarz kernel possesses properties similar to those of the Poisson
kernel. Some of the obvious ones are:

(i) S(t − x, y) is harmonic inside the domain y > 0.

(ii) Since the Schwarz integral formula is satisfied by setting u(x, y) = 1
and u(t, 0) = 1, we have

1 ∞
S(t − x, y) dt = 1, y > 0. (7.1.13)
π −∞
(iii) The maximum value of S(t − x, y) as a function of t over (−∞, ∞)
occurs at t = x, and
1
max S(t − x, y) = S(0, y) = → ∞ as y → 0. (7.1.14)
−∞<t<∞ y

Example 7.1.3 Find the function u(x, y) that is harmonic inside the first
quadrant x > 0 and y > 0, and satisfies the boundary conditions

u(x, 0) = f (x), 0 < x < ∞
,
u(0, y) = g(y), 0 < y < ∞

where f (x) and g(y) are bounded and continuous except for at most a finite
number of finite jumps (see Figure 7.5).
322 Boundary Value Problems and Initial-Boundary Value Problems

u (0, y ) = g ( y)

x
u ( x , 0) = f (x )
Figure 7.5. Dirichlet problem in the first quadrant, x > 0 and y > 0, with prescribed
boundary conditions: u(x, 0) = f (x) and u(0, y) = g(y).

Solution First, we find the harmonic functions u1 (x, y) and u2 (x, y) such
that:

(i) u1 (x, y) is harmonic in the first quadrant x > 0 and y > 0, and satisfies
the boundary conditions

u1 (x, 0) = f (x), 0<x<∞
;
u1 (0, y) = 0, 0<y<∞

(ii) u2 (x, y) is harmonic in the first quadrant x > 0 and y > 0, and satisfies
the boundary conditions

u2 (x, 0) = 0, 0<x<∞
.
u2 (0, y) = g(y), 0<y<∞

Since the Laplace equation is linear, by virtue of the principle of superpo-

sition of solutions, u1 + u2 is also harmonic. As the boundary conditions for
u1 (x, y) + u2 (x, y) are the same as those for u(x, y), by virtue of uniqueness
of solution, we then have

u(x, y) = u1 (x, y) + u2 (x, y).

To solve for u1 (x, y), given the zero boundary condition u1 (0, y) = 0 along
the positive y-axis, we perform the odd extension of the boundary value function
f (x) over the interval (−∞, 0) by defining

f (x) = −f (−x), −∞ < x < 0.

 7.1 Integral formulas of harmonic functions 323

By the Schwarz integral formula (7.1.12), the solution for u1 (x, y) is given by
 
1 ∞ y 1 0 y
u1 (x, y) = f (t) dt − f (−t) dt (i)
π 0 (t − x) + y
2 2 π −∞ (t − x)2 + y 2


y ∞ 1 1
= − f (t) dt, x > 0 and y > 0.
π 0 (t − x)2 + y 2 (t + x)2 + y 2

Obviously, the boundary condition u1 (0, y) = 0 is satisfied by the above solu-

tion; thus the procedure of performing the odd extension of the boundary value
function f (x) is justified.
Next, in order to solve for u2 (x, y), we first derive the modified Schwarz
integral formula for the Dirichlet problem in the right half-plane x > 0. In the
new derivation procedure, we choose the corresponding closed contour to be
the infinitely large semi-circle in the right half-plane plus the vertical diameter.
The chosen image point outside the closed contour is −x + iy, which is the
image of x + iy with the y-axis as the mirror. The modified Schwarz integral
formula for the domain x > 0 becomes
 −∞

1 1 1
u(x, y) = − u(0, t) i dt
2π i ∞ it − (x + iy) it − (−x + iy)

1 ∞ x
= u(0, t) dt, x > 0.
π −∞ (t − y)2 + x 2

We need to cope with the zero boundary condition u2 (x, 0) = 0 along the
positive x-axis. By following the same procedure as above, we perform the odd
extension of g(y) over the interval (−∞, 0) by defining

g(y) = −g(−y), −∞ < y < 0.

By the modified Schwarz integral formula derived above, the solution for
u2 (x, y) is found to be
 
1 ∞ x 1 0 x
u2 (x, y) = g(t) dt − g(−t) dt (ii)
π 0 (t − y) + x
2 2 π −∞ (t − y)2 + x 2
 ∞

x 1 1
= − g(t) dt, x > 0 and y > 0.
π 0 (t − y) + x
2 2 (t + y)2 + x 2

Note that the boundary condition u2 (x, 0) = 0 is satisfied by the above solution.
Finally, the solution for u(x, y) is given by the sum of u1 (x, y) and u2 (x, y) as
given in eq. (i) and eq. (ii), respectively.
324 Boundary Value Problems and Initial-Boundary Value Problems

7.1.3 Neumann problems

We would like to find a solution u(x, y) that is harmonic in a bounded domain D
and satisfies the prescribed boundary values taken by ∂u
∂n
along the boundary ∂D
of the domain. For simplicity, the domain D is assumed to be simply connected.
In order that a solution exists for the Neumann problem, the boundary values
taken by ∂u
∂n
must satisfy the compatibility condition
,
∂u
ds = 0, (7.1.15)
∂D ∂n

where s is the arc length along the boundary. The proof for the necessity of the
compatibility condition is constructed as follows:

0= ∇ 2 u dxdy since u is harmonic
D

= ∇· ∇u dxdy
D
,
∂u
= ds by virtue of Green’s theorem.
∂D ∂n

The following physical interpretation may be helpful to appreciate why the

compatibility condition is necessary. Suppose u(x, y) is the solution to the
steady state temperature distribution inside D; then the normal temperature
gradient ∂u
∂n
on ∂D is proportional to the heat flux across the boundary. In order
that steady state temperature prevails, simple physical intuition dictates that
the net heat flux across the whole boundary must be zero. This is precisely the
compatibility condition imposed on ∂u ∂n
along the boundary (assuming constant
conductivity of the material).
The solution to a Neumann problem is unique to within an additive constant
since the addition of a constant to u does not alter the value of ∂u
∂n
.
Dirichlet problems and Neumann problems are closely related, as we now
show. Let v(x, y) denote a harmonic conjugate to the harmonic solution u(x, y)
of the given Neumann problem with prescribed normal derivative

∂u
= g(s) along ∂D. (7.1.16)
∂n

By the Cauchy–Riemann relations, we deduce that (see Example 2.6.1)

∂v ∂u
= = g(s) along ∂D. (7.1.17)
∂s ∂n
 7.1 Integral formulas of harmonic functions 325

To obtain the Dirichlet boundary condition for v(x, y) along ∂D, we integrate
the above equation with respect to s to obtain
 s
v = G(s) = G0 + g(s) ds on ∂D, (7.1.18)
0

where G0 is an arbitrary constant. Now, v(x, y) is harmonic with G(s) as the

corresponding Dirichlet boundary condition along ∂D.
The final goal in this subsection is to find the integral representation of the
function u(r, θ ) that is harmonic inside the circular domain |z| < R with normal
derivative ∂u
∂n
along the boundary |z| = R, where

∂u 
= g(s), s = Reiθ . (7.1.19)
∂n |z|=R

According to eq. (7.1.18), we compute G(s), s = Reiθ , by

 θ
G(Re ) = G0 +
iθ
g(Reiφ )iReiφ dφ, (7.1.20)
0

where G0 = G(Reiθ )|θ=0 . We expect

G(Reiθ )|θ=0 = G(Reiθ )|θ=2π ,

and this condition would implicitly require

 2π
g(Reiφ )iReiφ dφ = 0.
0

This is precisely the compatibility condition on g(s) as stated in eq. (7.1.15).

As an intermediate step, we solve the Dirichlet problem with the prescribed
Dirichlet boundary condition G(s) along the boundary |z| = R. Let v(r, θ )
denote the solution to this Dirichlet problem in the domain |z| < R. The integral
representation of v(r, θ ) is given by the Poisson integral formula (7.1.5), which
takes the form
 2π
1
v(r, θ ) = P (R, r, φ − θ )G(Reiφ ) dφ. (7.1.21)
2π 0
To find u(r, θ ), it suffices to find a harmonic conjugate to −v(r, θ ).
Next we find the harmonic  conjugate
 of the Poisson kernel P (R, r, φ − θ ).
ζ +z
Since P (R, r, φ − θ ) = Re ζ −z , where ζ = Reiφ and z = reiθ , its harmonic
conjugate is
 
ζ +z 2Rr sin(φ − θ)
Q(R, r, φ − θ ) = Im =− 2 .
ζ −z R − 2Rr cos(φ − θ ) + r 2
326 Boundary Value Problems and Initial-Boundary Value Problems

A harmonic conjugate to v(r, θ ) can be obtained readily by replacing the ker-

nel P (R, r, φ − θ) by Q(R, r, φ − θ ). Since −u(r, θ ) is a harmonic conjugate
to v(r, θ ), the integral representation of u(r, θ ) is given by

1 2π
2Rr sin(φ − θ)
u(r, θ ) = U0 + G(Reiφ ) dφ. (7.1.22)
2π 0 R2 − 2Rr cos(φ − θ ) + r 2

An additive constant U0 is added here since the solution to a Neumann problem

is unique to within an additive constant. By setting r = 0 in eq. (7.1.22), we
see that U0 equals the value of u at the origin.
Now, G(Reiθ ) is unique to within an additive constant. The representa-
tion of- G(Reiθ ) in eq. (7.1.20) contains the arbitrary constant G0 . However,
2π
since 0 Q(R, r, φ − θ) dφ = 0 (see Problem 7.3), the value of the inte-
gral in eq. (7.1.22) is not affected by any choice of the value of the arbi-
trary constant G0 . For simplicity, we may take G0 = 0, and correspondingly,
G(Reiθ )|θ=2π = 0.
It is more desirable to express the solution in terms of g(s). Integrating the
integral in eq. (7.1.22) by parts and observing

∂ 2Rr sin(φ − θ)
ln(R 2 − 2Rr cos(φ − θ ) + r 2 ) = 2
∂φ R − 2Rr cos(φ − θ ) + r 2

and G(Reiθ )|θ=0 = G(Reiθ )|θ=2π = 0, we obtain

 2π
R
u(r, θ ) = U0 − ln(R 2 − 2Rr cos(φ − θ ) + r 2 )g(Reiφ ) dφ, r < R.
2π 0
(7.1.23)
This is the integral representation of the harmonic function u(r, θ ), which
satisfies the Neumann boundary condition as specified in eq. (7.1.19).

7.2 The Laplace transform and its inversion

Given a function f (t) defined for t ≥ 0, and a complex parameter s = σ + iτ ,
the Laplace transform of f (t) is defined by
 ∞
F (s) = L{f (t)} = e−st f (t) dt, (7.2.1)
0

provided that the integral exists.

To guarantee the existence of the integral over any finite interval [0, b],
it suffices to assume that the integrand is piecewise continuous over [0, b],
where b takes any finite value. A possibility that leads to the non-existence of
 7.2 The Laplace transform and its inversion 327

the Laplace transform is that the integrand e−st f (t) may diverge over large t.
Suppose there exist constants M and a such that

|f (t)| ≤ Meat , t > 0; (7.2.2)

then

|f (t)e−st | ≤ Meat e−σ t = Me(a−σ )t .

If s is chosen such that Re s = σ > a, then the integral is absolutely convergent

for Re s > a. A function f (t) that satisfies condition (7.2.2) is said to be of
exponential order (with growth exponent a).
In summary, any piecewise continuous function of exponential order has
a Laplace transform. These conditions are sufficient but not necessary. For
example, the function f (t) = √1t has an infinite singularity at t = 0 and so it is
not piecewise continuous, but its Laplace transform exists (see Example 7.2.2).
It can be shown that the Laplace transform is analytic in the domain Re s > a
in the s-plane. Also, the linearity property is observed, where

L{αf (t) + βg(t)} = αL{f (t)} + βL{g(t)},

provided that the individual Laplace transforms exist.

Some basic transform formulas

The Laplace transforms of some elementary functions are shown below:
 ∞
1
L{eat } = e−st eat dt = , Re s > a. (7.2.3a)
0 s−a
 ∞
L{t n } = e−st t n dt
0
 ∞
n n!
= e−st t n−1 dt = · · · = , Re s > 0. (7.2.3b)
s s n+1
0 ∞
L{sin wt} = Im e−st eiwt dt
0
1 w
= Im = 2 , Re s > 0. (7.2.3c)
s − iw s + w2
 ∞
eβt + e−βt
L{cosh βt} = e−st dt
2
0
 
1 1 1 s
= + = 2 , s > |β|. (7.2.3d)
2 s−β s+β s − β2
328 Boundary Value Problems and Initial-Boundary Value Problems

Step function
Recall that the Heaviside step function H (x) is defined by

1 t >0
H (t) = . (7.2.4a)
0 t <0
The Laplace transform of H (t − τ ), τ > 0, is found to be
 ∞
e−sτ
L{H (t − τ )} = e−st dt = . (7.2.4b)
τ s

Shifting
Consider the shifting of the function f (t) as defined by

f (t − τ ) t ≥ τ
fτ (t) = , τ > 0;
0 t <τ
then its Laplace transform is found to be
 ∞
L{fτ (t)} = e−st fτ (t) dt
0 ∞  ∞
= e−st f (t − τ ) dt = e−sτ e−s(t−τ ) f (t − τ ) d(t − τ )
τ
 ∞ τ

= e−sτ e−st f (t  ) dt  = e−sτ L{f (t)}. (7.2.5)
0

Dirac function
The Laplace transform of the Dirac function δ(t − t0 ), t0 > 0, is found to be
 ∞
L{δ(t − t0 )} = e−st δ(t − t0 ) dt = e−st0 . (7.2.6)
0

Derivatives
Suppose f (t), f  (t), . . . , f (n) (t) are piecewise continuous and of exponential
order with growth exponent a; then
 ∞  ∞
 −st  −st ∞
L{f (t)} = e f (t) dt = [e f (t)]0 + s e−st f (t) dt.
0 0
−st −st
For Re s > a, we have lim e f (t) = 0 and lim e f (t) = f (0); so
t→∞ t→0

L{f  (t)} = sL{f (t)} − f (0). (7.2.7a)

Deductively,
L{f  (t)} = sL{f  (t)} − f  (0) = s 2 L{f (t)} − sf (0) − f  (0);
 7.2 The Laplace transform and its inversion 329

and, in general,

L{f (n) (t)} = s n L{f (t)} − s n−1 f (0) − s n−2 f  (0) − · · · − f (n−1) (0).
(7.2.7b)

Convolution property
Suppose L{f (t)} and L{g(t)} exist for Re s > a, and recall that the convolution
of f (t) and g(t) is given by
 t
(f ∗ g)(t) = f (τ )g(t − τ ) dτ.
0

The Laplace transform of the convolution is given by

 ∞  t
L{(f ∗ g)(t)} = e−st f (τ )g(t − τ ) dτ dt
0 ∞ 0 ∞
= f (τ ) e−st g(t − τ ) dt dτ,
0 τ

where the last integral is obtained by changing the order of integration. By

setting t − τ = t  , we obtain
 ∞  ∞

L{(f ∗ g)(t)} = e−sτ f (τ ) dτ e−st g(t  ) dt 
0 0
= L{f (t)}L{g(t)}. (7.2.8)

Example 7.2.1 Solve each of the following equations using the Laplace
transform method:
 x
(a) f (u)(x − u) du + f (x) = sin 2x, x > 0;
0 2

 d y dy
 2 +2 + 2y = δ(t − 1) + f (t)
(b) dt dt .



y(0) = y  (0) = 1

Solution
(a) This is an integral equation with the unknown function f (x). The integral
takes the form of the convolution of f (x) and x. Taking the Laplace
transform on both sides of the equation and applying the convolution
formula (7.2.8), we obtain
L{f (x)} 2
+ L{f (x)} = 2 .
s2 s +4
330 Boundary Value Problems and Initial-Boundary Value Problems

Solving for L{f (x)}, we obtain

2s 2 4 2 2 1
L{f (x)} = = − .
s 4 + 5s 2 + 4 3 s2 + 4 3 s2 + 1
The inversion of the above Laplace transform is easily seen to be
4 2
f (x) = sin 2x − sin x, x > 0.
3 3
(b) Let Y (s) and F (s) denote the Laplace transforms of y(t) and f (t),
respectively. By taking the Laplace transform of the equation, we obtain

s 2 Y (s) − sy(0) − y  (0) + 2[sY (s) − y(0)] + 2Y (s) = e−s + F (s).

Applying the given initial conditions and solving for Y (s), we have
e−s (s + 1) + 2 F (s)
Y (s) = + + .
(s + 1) + 1 (s + 1) + 1 (s + 1)2 + 1
2 2

The Laplace transform inversions of the individual terms are

 
−1 e−s
L = H (t − 1)e−(t−1) sin(t − 1),
(s + 1)2 + 1
 
−1 (s + 1) + 2
L = e−t (cos t + 2 sin t),
(s + 1)2 + 1
   t
−1 F (s)
L = f (t)e−(t−τ ) sin(t − τ ) dτ.
(s + 1)2 + 1 0

The solution is then given by

y(t) = H (t − 1)e−(t−1) sin(t − 1) + e−t (cos t + 2 sin t)

 t
+ f (t)e−(t−τ ) sin(t − τ ) dτ.
0

Note that y(t) is continuous but y  (t) is discontinuous at t = 1, due to the

impulsive effect caused by the term δ(t − 1) in the differential equation.

7.2.1 Bromwich integrals

The inversion of the Laplace transform of a function can be effected by inte-
grating the corresponding Bromwich integral, the details of which are stated in
the following theorem.

Theorem 7.2.1 Assume f (t) to be a piecewise continuous function of expo-

nential order defined for t ≥ 0, and that there exists a real constant a > 0 such
 7.2 The Laplace transform and its inversion 331
- ∞ at
that the real integral 0 e |f (t)| dt exists. Let F (s) = L{f (t)} denote the
Laplace transform of f (t). The inverse Laplace transform of F (s) is given by
 a+iR
1
f (t) = lim est F (s) ds, t > 0. (7.2.9)
2π i R→∞ a−iR

Proof We apply the Fourier transform inversion formula to the function

e−at f (t), assuming f (t) = 0 for t < 0 [see eq. (6.4.5)], to obtain
 R  ∞

−at 1 −iut iut  −at   

e f (t) = lim e e e f (t ) dt du.
2π R→∞ −R 0

The above result is reformulated as

 R  ∞

1 
f (t) = lim e(a−iu)t e−(a−iu)t f (t  ) dt  du, t > 0, (7.2.10)
2π R→∞ −R 0

where the inner integral is recognized as the Laplace transform of f (t) if a − iu

becomes the transform variable s. Changing the variable of integration from u
to s = a − iu, the above formula becomes
 a+iR
1
f (t) = lim est F (s) ds, t > 0, (7.2.11)
2π i R→∞ a−iR

where the integration path is the infinite vertical line Re s = a. Recall that
the-constant a > 0 is chosen to be sufficiently large to ensure the existence
∞
of 0 e−at |f (t)| dt. The vertical line Re s = a is called the Bromwich line,
and subsequently, the inversion integral in eq. (7.2.11) is called the Bromwich
integral.

Remark
(i) At a point with a finite jump, f (t) is assigned the average of its left-hand
and right-hand limits.
(ii) Suppose F (s) is analytic except for a finite number of poles in the half-
plane to the left of the vertical line Re s = a. We construct a positively
oriented closed contour CB which consists of the vertical Bromwich line
B from a − iR to a + iR, the upper horizontal line u from a + iR
to iR, the left semi-circle CR: z = Reiθ , π2 ≤ θ ≤ 3π 2
, and the lower
horizontal line d from −iR to a − iR (see Figure 7.6).
332 Boundary Value Problems and Initial-Boundary Value Problems

y
iR u
a + iR
<

R
B
x

iR d
a iR

Figure 7.6. A closed contour CB is constructed which consists of the vertical Bromwich
line B , the horizontal lines u and d , and the left semi-circle CR .

Suppose F (s) tends uniformly to zero as R → ∞ when s assumes values on

u ∪ CR ∪ d . It can be shown that
 a+iR
1
f (t) = lim est F (s) ds
2π i R→∞ a−iR
,
1
= lim est F (s) ds, (7.2.12)
2π i R→∞ CB
since the contour integrals along the line segments u and d and the left semi-
circle CR all tend to zero as R → ∞ (hints about proving these claims are
given in Problem 7.20). Now the Laplace transform inversion formula becomes
a contour integral with a closed contour. As R → ∞, the closed contour CB
contains all the poles of F (s) in the half-plane to the left of B . By the Residue
Theorem, it then follows that

f (t) = Res(est F (s), sn ), (7.2.13)
n

where the points sn denote the poles of F (s) that lie on the left side of B , and
the summation of residue values is taken over all these poles.

Example 7.2.2 Find the inverse Laplace transform of

√
s 1 e−λ s
(a) F (s) = ; (b) F (s) = √ ; (c) F (s) = , λ > 0.
(s + 1)3 s s

Solution
(a) Note that F (s) = (s+1)
s
3 has one pole of order 3 at s = −1. One may

choose the Bromwich line to be Re s = a, a > −1. By observing that

 7.2 The Laplace transform and its inversion 333

the pole s = −1 lies to the left of the Bromwich line and (s+1)
s
3 tends to

zero uniformly as |s| → ∞, we then apply eq. (7.2.13) to obtain

 
s
f (t) = Res est , −1 .
(s + 1)3
The most effective method for evaluating the above residue is to find
1 est s
the coefficient of s+1 in the Laurent series expansion of (s+1) 3 valid in a

deleted neighborhood of s = −1. We obtain

s
est
(s + 1)3
(s + 1) − 1
= e(s+1)t e−t
(s + 1)3

(s + 1)2 t 2 1 1
= 1 + (s + 1)t + + ··· − e−t
2! (s + 1)2 (s + 1)3
% 2
&
1 1−t t − t2
= − + + + · · · e−t ;
(s + 1)3 (s + 1)2 s+1
so
1
f (t) = coefficient of in the above Laurent expansion
s+1
 
t 2 −t
= t− e , t > 0.
2
 
Remark There is a more direct method of finding L−1 s
(s+1)3
. By
making use of the inversion formulas

 ∞
L{eat f (t)} = e−st eat f (t) dt = F (s − a), F (s) = L{f (t)},
0

and
n!
L{t n } = , n is a non-negative integer,
s n+1
we deduce that
     
s 1 1 t 2 −t
L−1 = L−1
− = t − e .
(s + 1)3 (s + 1)2 (s + 1)3 2
(b) We choose the branch cut of F (s) = √1s to be along the negative real
axis. The Bromwich line can be any vertical line in the right half-plane.
The closed contour with the Bromwich line as the vertical side has to be
334 Boundary Value Problems and Initial-Boundary Value Problems

Figure 7.7. To avoid crossing the branch cut along the negative real axis, the closed
contour CB is modified to include the extra excursion to the branch point s = 0 along
the upper and lower sides of the branch cut.

modified to include the extra excursion to the branch point s = 0 along

the upper and lower sides of the branch cut, plus an infinitesimal circle
C looped in the clockwise sense around the branch point s = 0 (see
Figure 7.7). The rationale for such a construction is that the closed con-
tour CB should not cross the branch cut along the negative real axis.
Note that F (s) = √1s tends uniformly to zero as |s| → ∞. Following
analogous arguments that establish eq. (7.2.12), it is seen that the contri-
butions along the outer quarter circles and the line segments joining the
outer quarter circles with the Bromwich line are zero (see Problem 7.20).
The remaining contributions to the contour integral around the closed
contour CB come from the Bromwich line B , the upper line u and the
lower line d along the branch cut and the small circle C (looped in the
negative sense). Note that the contour integration along B gives f (t).
est
On the other hand, since the closed contour CB contains no pole of √ s
,
the contour integral around CB is zero. We then have
,
1 est
0= √ ds
2π i CB s

1 est
= f (t) + √ ds
2π i u s
 ,
1 est 1 est
+ √ ds − √ ds.
2π i d s 2π i C s

Along the upper side of the branch cut u , s = eiπ ξ , where ξ runs
from R to , and along the lower side of the branch cut d , s = e−iπ ξ ,
 7.2 The Laplace transform and its inversion 335

where ξ runs from  to R. On the infinitesimal circle C , we have

s = eiθ , −π < θ ≤ π . Taking the limits R → ∞ and  → 0, we obtain
 ∞  ∞
1 1 1 1
f (t) = e−ξ t √ (−dξ ) − e−ξ t √ (−dξ )
2π i 0 i ξ 2π i 0 −i ξ
 π
1 iθ 1
+ lim ete √ iθ/2 ieiθ dθ.
2π i →0 −π e
√
The modulus of the last integral is O( ) and so it becomes zero as
 → 0. The above expression can be simplified as
 ∞
1 e−ξ t
f (t) = √ dξ
π 0 ξ
 ∞
2
e−y dy,
2
= √ y2 = ξ t
π t 0
1
=√ .
πt
√ √
−λ s
(c) With the appearance of s in the exponent of F (s) = e s , the closed
contour is √chosen to be the same as that shown in Figure 7.7. Since
−λ s
F (s) = e s also tends uniformly to zero as |s| → ∞, the contributions
to the contour integral come only from the Bromwich line, the upper
and lower line segments along the branch cut and the infinitesimal circle
around the branch point s = 0. Again, there is no pole enclosed within
the closed contour CB . Following a similar approach to part (b), we
obtain
 ∞ √
−λ ξ i  ∞ √
−(−λ ξ i)
1 −ξ t e 1 −ξ t e
f (t) = e (−dξ ) − e (−dξ )
2π i 0 −ξ 2π i 0 −ξ
 π √
−λ eiθ/2
1 iθ e
+ lim ete ieiθ dθ.
2π i →0 −π eiθ

The last integral becomes 1 as  → 0. The above expression can be

simplified as
 √
1 ∞ −ξ t sin λ ξ
f (t) = 1 − e dξ
π 0 ξ
 ∞  λ 
2 −tu2


=1− e cos αu dα du, u = ξ
π 0 0
  ∞ 
2 λ −tu2
=1− e cos αu du dα.
π 0 0
336 Boundary Value Problems and Initial-Boundary Value Problems

It is shown in Example 4.2.4 that

 ∞ 
1 π −α2 /4t
e−tu cos αu du =
2
e ,
0 2 t
so we obtain
 λ
1
e−α
2
f (t) = 1 − √ /4t
dα.
πt 0

The above integral can be expressed in terms of the complementary error

function erfc(x) defined by
 x
2
e−η dη.
2
erfc(x) = 1 − √
π 0
α λ
By taking η = √ and x = √ , we finally obtain
2 t 2 t
 
λ
f (t) = erfc √ .
2 t

Remark By differentiating the Laplace transform formula

   √
λ e−λ s
L erfc √ =
2 t s

with respect to the parameter λ repeatedly, we obtain the following

related Laplace transform formulas:
  √
1 −λ2 /4t e−λ s
L √ e = √ ,
πt s
  √
λ
e−λ /4t = e−λ s .
2
L √
2 πt 3

These formulas will be used in solving the heat conduction problems

discussed in the next section.

7.3 Initial-boundary value problems

The Laplace transform method is known to be an effective tool for solv-
ing initial-boundary value problems arising from mathematical physics. Two
classical initial-boundary value problems are examined in this section: heat
conduction and wave propagation. Here we show how the complex variables
techniques are applied in the analytic inversion of the Laplace transform to
obtain the solutions. Though the governing equations for heat conduction and
 7.3 Initial-boundary value problems 337

wave propagation are derived in many texts on partial differential equations,

for the purpose of making this text self-contained, we spare a few paragraphs
to examine how these governing equations arise from the modeling of the
underlying physical phenomena.

7.3.1 Heat conduction

The heat conduction problem is closely related to the steady state temperature
distribution problems discussed in the earlier chapters. Consider the same con-
trol volume inside a two-dimensional conducting body as shown in Figure 2.7.
Let c and ρ denote the specific heat of the material and density of the conduct-
ing body, respectively. The material parameters are assumed to be constant.
The heat content H (t) contained in the control volume at time t is given by

H (t) = cρT (x, y, t)xy,

where the temperature T (x, y, t) is a function of x, y and t. The rate of change

of the heat content is given by
dH ∂T
= cρ xy.
dt ∂t
The net accumulation of heat energy per unit time inside the control volume
through conduction along its four sides is given by
 2 
∂ T ∂ 2T
Kxy + ,
∂x 2 ∂y 2
where K is the thermal conductivity of the material (see Section 2.6.2). By the
law of conservation of energy, the rate of change of heat energy in the control
volume should equal the heat flux across its boundary through conduction.
From the above two expressions, we then obtain
 
∂T K ∂ 2T ∂ 2T
= + . (7.3.1)
∂t cρ ∂x 2 ∂y 2
This is the two-dimensional heat conduction equation. When steady state con-
ditions prevail, the temporal rate of change of temperature ∂T
∂t
becomes zero.
The above equation then reduces to
∂ 2T ∂ 2T
+ = 0. (7.3.2)
∂x 2 ∂y 2
This is the familiar Laplace equation that governs steady state temperature
distribution. The one-dimensional version of the heat conduction equation is
338 Boundary Value Problems and Initial-Boundary Value Problems

given by
∂T ∂ 2T K
= a2 2 , a2 = , (7.3.3)
∂t ∂x cρ
where T = T (x, t). The above equation governs the heat conduction process
inside a thin longitudinal rod.

Heat conduction in an infinite rod

Consider the heat conduction problem in an infinitely long rod with initial
condition T (x, 0) = f (x), −∞ < x < ∞. The boundary conditions at the
infinite ends are not specified except that the temperature values are assumed
to remain bounded as |x| → ∞. The set of governing equations for the initial
value problem can be stated as


 ∂T ∂ 2T
 = a 2 2 , −∞ < x < ∞, t > 0
∂t ∂x . (7.3.4)



T (x, 0) = f (x)
First we try to find the fundamental solution F (x, t) that satisfies


 ∂F ∂ 2F
 = a 2 2 , −∞ < x < ∞
∂t ∂x , (7.3.5)



F (x, 0) = δ(x)
where δ(x) is the Dirac function. By the defining property of the Dirac function,
the initial condition f (x) can be expressed as
 ∞
f (x) = f (ξ )δ(x − ξ ) dξ. (7.3.6)
−∞

The above relation can be interpreted as the decomposition of f (x) into

impulses distributed over (−∞, ∞) with magnitude f (ξ ) at the position ξ . The
solution to the heat conduction problem with initial condition f (ξ )δ(x − ξ ) is
recognized to be f (ξ )F (x − ξ, t). Since the heat equation is linear, the prin-
ciple of superposition of solutions applies, so the solution to eq. (7.3.4) can
be obtained by integrating f (ξ )F (x − ξ, t) with respect to ξ from −∞ to ∞.
Hence, the solution to eq. (7.3.4) can be represented as
 ∞
T (x, t) = f (ξ )F (x − ξ, t) dξ. (7.3.7)
−∞

To solve for F (x, t), we take the Laplace transform with respect to t on both
sides of eq. (7.3.5). Let F̂ (x, s) denote the Laplace transform of F (x, t). We
 7.3 Initial-boundary value problems 339

obtain the following ordinary differential equation for F̂ (x, s):

d 2 F̂
s F̂ (x, s) − a 2 (x, s) = δ(x), −∞ < x < ∞. (7.3.8)
dx 2
The solution to the above equation takes the form
 √ √
A+ (s)e−√sx/a + B+ (s)e√sx/a x>0
F̂ (x, s) = .
A− (s)e− sx/a + B− (s)e sx/a x<0
To ensure that the solution remains bounded as |x| → ∞, we must set
B+ (s) = A− (s) = 0. To determine A+ (s) and B− (s), we observe that F̂ (x, s)
is continuous across x = 0 and there is a jump of ∂∂xF̂ across x = 0 due to the
term δ(x) in eq. (7.3.8) [see also Example 7.2.1, part (b)]. The continuity of
F̂ (x, s) across x = 0 gives the following condition on F̂ :

F̂ (0+ , s) − F̂ (0− , s) = 0.

By integrating eq. (7.3.8) with respect to x from x = 0− to x = 0+ , we obtain

another condition on F̂ :
∂ F̂ + ∂ F̂ − 1
(0 , s) − (0 , s) = − 2 .
∂x ∂x a
Using these two conditions, the solutions to A+ (s) and B− (s) are found to be
1
A+ (s) = B− (s) = √ ;
2a s
so
1 √
F̂ (x, s) = √ e− s|x|/a . (7.3.9)
2a s
The inverse of the above Laplace transform is found to be (see Example 7.2.2)
1
e−x
2
/4a 2 t
F (x, t) = √ .
4π a 2 t
Finally, by eq. (7.3.7), the solution to eq. (7.3.4) is found to be
 ∞
1
f (ξ )e−(x−ξ ) /4a t dξ.
2 2
T (x, t) = √ (7.3.10)
2
4π a t −∞

Heat conduction in a semi-infinite rod

Next we consider the heat conduction problem in a semi-infinite rod x ≥ 0.
For simplicity, we assume the initial temperature distribution to be zero. At the
finite end x = 0, the temperature is maintained at T (0, t) = g(t). The problem
340 Boundary Value Problems and Initial-Boundary Value Problems

can then be formulated as



 ∂T ∂ 2T
 = a 2 2 , x > 0, t > 0
∂t ∂x . (7.3.11)



T (x, 0) = 0 and T (0, t) = g(t)

We follow a similar technique of taking the Laplace transform with respect

to t. Let T̂ (x, s) and G(s) denote the Laplace transforms of T (x, t) and g(t),
respectively. We obtain the following ordinary differential equation for T̂ (x, s):


 ∂ 2 T̂
 s T̂ (x, s) = a 2 2 (x, s), x > 0
∂x .



T̂ (0, s) = G(s)
√
Following a similar procedure as before, the term e sx/a in the general solution
is discarded since the solution remains bounded as x → ∞. The solution to the
above equation is then found to be
√
T̂ (x, s) = G(s)e− sx/a
, x > 0.

Using the Laplace transform inversion formula derived in Example 7.2.2

√ x
L−1 {e− e−x /4a t ,
2 2
sx/a
}= √
2
2 πa t 3/2

and the convolution formula (7.2.8), the solution to eq. (7.3.11) is found to be
 t
x 1
e−x /4a (t−τ ) dτ.
2 2
T (x, t) = √ g(τ ) (7.3.12)
4π a 2 0 (t − τ ) 3/2

Example 7.3.1 Consider the heat conduction problem within a rod of unit
length. Suppose the two ends are kept at zero temperature, and the initial
temperature is kept at the constant value T (x, 0) = 1, 0 < x < 1; find the
temperature along the rod at any time t > 0.

Solution
The heat conduction problem can be stated as


 ∂T ∂ 2T
 = , 0 < x < 1, t > 0
∂t ∂x 2 ,



T (x, 0) = 1, T (0, t) = T (1, t) = 0
 7.3 Initial-boundary value problems 341

where the coefficient a 2 has been taken to be unity for simplicity. The differ-
ential equation governing the Laplace transform T̂ (x, s) is given by
 2

 ∂ T̂
 (x, s) − s T̂ (x, s) = −1
∂x 2 .



T̂ (0, s) = T̂ (1, s) = 0
The solution to the above differential equation is found to be
1 √ √ √ √ √ √
T̂ (x, s) = √
−
√ [e
s
− e− s + (e− s − 1)e sx − (e s − 1)e− sx ].
s(e s −e s )
The direct analytic valuation of the Bromwich inversion integral for T̂ (x, s)
is intractable. Alternatively, we expand the transform function in series for
large s and perform the inverse Laplace transforms term by term. It can be
shown that the “large s” approximation in T̂ (x, s) corresponds to the “small t”
approximation in T (x, t) (see Problem 7.21). We expand T̂ (x, s) in the “large
s” approximation as follows:
√
1 e− s √ √
T̂ (x, s) = + (1 + e−2 s + e−4 s + · · · )
s s √ √ √ √
(e− s(1−x) − e s(1−x) − e sx + e− sx )
∞ ∞
1 1 √  √
= + (−1)n e− s(n−x) − (−1)n e− s(n+x) .
s s n=1 n=0

We apply the inversion formula (see Example 7.2.2)

 √
  
−1 1 −λ s λ
L e = erfc √ ,
s 2 t
so the termwise inversion of the above series gives the following solution to
the temperature distribution:
∞    ∞  
n−x n+x
T (x, t) = 1 + (−1)n erfc √ − (−1)n erfc √ .
n=1
2 t n=0
2 t

7.3.2 Longitudinal oscillations of an elastic thin rod

Let us consider the phenomenon of wave propagation in a thin elastic rod under
longitudinal oscillations. The longitudinal movement of a segment of the rod
leads to contraction or elongation of its neighboring portions, which in turn
affect the other parts of the rod in sequence. Subsequently, the longitudinal
oscillations are propagated to the whole rod at a certain speed of propagation.
342 Boundary Value Problems and Initial-Boundary Value Problems

x x+ x

u u+ u

Figure 7.8. Longitudinal oscillations of a differential segment of length x along an

infinite rod.

Let u(x, t) denote the displacement of the rod at the position x and time
t. Consider the differential segment [x, x + x] of length x along the rod.
The displacements of the rod at positions x and x + x are u and u + ∂x ∂u
x,
respectively (see Figure 7.8). The magnitude of the strain at position x is given
by
u+ ∂u
x −u ∂u
∂x
= .
x ∂x
Assuming that the material of the rod observes the law of linear elasticity, the
∂u
stress at position x is given by E ∂x (x, t), where E is the Young modulus of the
material. Let A denote the uniform cross-sectional area of the rod. The forces at
the left and right ends of the segment [x, x + x] due to the elastic stresses are
EA ∂x∂u ∂u
(x, t) and EA ∂x (x + x, t), respectively. The mass of the differential
segment is ρAx, where ρ is the mass per unit length of the rod.
Using Newton’s second law of motion, the equation that governs the dynam-
ics of the longitudinal oscillations can be expressed as
∂ 2u ∂u ∂u
ρAx 2
(x, t) = EA (x + x, t) − EA (x, t).
∂t ∂x ∂x
Taking the limit x → 0, we obtain

∂ 2u 2
2∂ u E
= c , c= > 0. (7.3.13)
∂t 2 ∂x 2 ρ

This is the wave propagation equation that governs the longitudinal oscilla-
tions of a thin elastic rod. The parameter c gives the speed of propagation of
longitudinal oscillations along the rod [see eq. (7.3.15)].
 7.3 Initial-boundary value problems 343

Wave propagation along a semi-infinite elastic rod

Since the wave equation involves the second-order time derivative, the full
prescription of the set of initial conditions requires the specification of the initial
displacement and velocity. Consider the wave propagation along a semi-infinite
elastic rod where the time-dependent movement at x = 0 is specified to be
f (t), and the initial displacement and velocity are zero. Also, the displacement
vanishes at the far end, x → ∞. The longitudinal oscillation problem can be
formulated as
 2 2

 ∂ u 2∂ y

 ∂t 2 = c , 0 < x < ∞, t > 0
∂x 2
. (7.3.14)



u(0, t) = f (t), u(x, 0) = ∂u (x, 0) = 0
∂t
Let U (x, s) and F (s) denote the Laplace transform of u(x, t) and f (t) with
respect to time, respectively. The governing differential equation for U (x, s) is
given by


 ∂ 2U
 s 2 U (x, s) = c2 2 (x, s)
∂x .



U (0, s) = F (s)
Solving the above equation, we obtain

U (x, s) = A(s)e−sx/c + B(s)esx/c ,

where the arbitrary functions A(s) and B(s) are determined by the boundary
conditions. Since u(x, t) → 0 as x → ∞, we have B(s) = 0. Incorporating the
boundary condition U (0, s) = F (s), U (x, s) is found to be

U (x, s) = F (s)e−sx/c .

Inversion of the above Laplace transform gives [see eq. (7.2.5)]

 x  x
u(x, t) = H t − f t− . (7.3.15)
c c
The waveform represented by the solution remains unchanged if t − x
c
is kept
constant, indicating that the waveform propagates at the speed c.

Example 7.3.2 An elastic rod of length L is fixed at one end and the other end
is subject to the periodic force F sin ωt. The initial displacement and velocity
of the rod are zero. Determine the subsequent motion of the elastic rod under
longitudinal oscillations.
344 Boundary Value Problems and Initial-Boundary Value Problems

Solution The present problem can be formulated as

 2

 ∂ u ∂ 2y

 = c2 2 , 0 < x < L, t > 0

 ∂t 2 ∂x




∂u
 u(x, 0) = (x, 0) = 0 ,

 ∂t





 ∂u F
 u(0, t) = 0 and (L, t) = sin ωt
∂x E

where E is the Young modulus of the rod material, c = Eρ , and ρ is
the linear density of the rod. Let U (x, s) denote the Laplace transform of
u(x, t) with respect to t. By taking the Laplace transform of the govern-
ing equation with respect to t and incorporating the initial conditions, we
obtain
∂ 2U
s 2 U (x, s) = c2 (x, s),
∂x 2
subject to the boundary conditions
∂U F ω
U (0, s) = 0 and (L, s) = .
∂x E s 2 + ω2
The solution is found to be
F cω sinh sxc
U (x, s) = .
E s(s 2 + ω2 ) cosh sLc

The poles of est U (x, s) are

 
c 1
s = 0, ±ωi and ± k− π i, k = 1, 2, 3, . . . .
L 2
It can be checked easily that U (x, s) tends uniformly to zero as R → ∞ when
s assumes values on u ∪ CR ∪ d (see Figure 7.6). By the Laplace transform
inversion formula (7.2.13), the solution to the displacement function is given
by
 
 F cω sinh sx
u(x, t) = Res est 2 + ω2 )
c
sL
, si ,
i
E s(s cosh c

where si are the poles of est U (x, s). The poles are all simple and so their
residues are given by

est F cω sinh sxc 
Res(est U (x, s), si ) = d    .
ds
Es(s 2 + ω2 ) cosh c s=si
sL
 7.3 Initial-boundary value problems 345

The respective residues are found to be:

(i) Res(est U (x, s), 0) = 0;  

F cω ±iωt
i sin ωx
(ii) Res(e U (x, s), ±iω) =
st
∓e c
;
E 2ω2 cos ωL
    c
c 1
(iii) Res est U (x, s), ± k− πi
 L 2 5
F cω 8i(−1)k L2 sin (2k−1)πx ±i(2k−1)cπt/(2L)
= ± L
e .
E (2k − 1)π [4L2 ω2 − (2k − 1)2 c2 π 2 ]

Therefore, the displacement of the elastic rod is given by

F c sin ωx
c
u(x, t)= sin ωt
E ω cos ωL
c
∞
16 F L2  (−1)k−1 sin (2k−1)πx (2k − 1)cπ t
+ cω 2L
sin .
π E k=1
(2k − 1)[4L 2 ω2 − (2k − 1)2 c2 π 2 ] 2L

Example 7.3.3 Use the Fourier transform method to solve the following wave
equation with the non-homogeneous term f (x, t):
 2 2

 ∂ u 2∂ u

 ∂t 2 − c = f (x, t), −∞ < x < ∞ and t > 0
∂x 2



u(x, 0) = 0 and ∂u (x, 0) = 0.
∂t

Solution First, we take the Fourier transform of the equation and the initial
conditions with respect to x. Let the Fourier transform of u(x, t) and f (x, t)
be denoted by
 ∞  ∞
U (t, ω) = eiωx u(x, t) dx and F (t, ω) = eiωx f (x, t) dx,
−∞ −∞

respectively. By observing
 2 
∂ u
F (x, t) = (−iω)2 F{u(x, t)},
∂x 2
the resulting equation for U (t, ω) becomes
 2

 ∂ U (t, ω) + c2 ω2 U (t, ω) = F (t, ω)

 ∂t 2
.



U (0, ω) = 0 and ∂U
(0, ω) = 0
∂t
346 Boundary Value Problems and Initial-Boundary Value Problems

The solution to the above differential equation for U (t, ω) can be obtained
by using the Laplace transform method and applying the Laplace inversion
formula for the convolution of two functions [see Example 7.2.1, part (b)].
(s, ω) and F
Let U (s, ω) denote the Laplace transforms of U (t, ω) and F (t, ω),
respectively. Taking the Laplace transform of the above equation, we obtain
(s, ω)
F (s, ω)  1
F 1


U (s, ω) = 2 = − .
s + c2 ω2 2cωi s − cωi s + cωi
The inversion of the above Laplace transform gives
 t
1
U (t, ω) = F (τ, ω)[eicω(t−τ ) − e−icω(t−τ ) ] dτ.
2cωi 0
Lastly, by taking the inversion of the Fourier transform U (t, ω), the solution
u(x, t) can be expressed as
 t  ∞

1 1 F (τ, ω) icω(t−τ ) −iωx

u(x, t) = e e dω dτ
2c 0 2π −∞ iω
 t  ∞

1 1 F (τ, ω) −icω(t−τ ) −iωx

− e e dω dτ.
2c 0 2π −∞ iω
Recall the Fourier transform rules

F{u (x)} = −iωF{u(x)} and F{u(x − a)} = eiaω F{u(x)};

correspondingly, we define
 x
g(x, τ ) = f (ξ, τ ) dξ, where x0 is some constant.
x0

We then see that

F (τ, ω) ∓icω(t−τ )
F{g(x ± c(t − τ ), τ )} = − e ,
iω
so
 t  t
1 1
u(x, t) = g(x + c(t − τ ), τ ) dτ − g(x − c(t − τ ), τ ) dτ
2c 0 2c 0
 t  x+c(t−τ )
1
= f (ξ, τ ) dξ dτ.
2c 0 x−c(t−τ )

7.4 Problems
7.1. Consider the steady state temperature distribution inside the domain
|z| < 1. Given that the boundary temperature value T (1, θ ) along the
 7.4 Problems 347

circumference of the unit circle is prescribed by


1 0 < θ < π2
T (1, θ ) = ,
0 otherwise

show that the temperature T (r, θ ) inside the unit circle is given by

1 1 − r2
T (r, θ ) = tan−1 2 , r < 1,
π r − 2r(cos θ + sin θ) + 1
where the inverse tangent function assumes values in [0, π ].
7.2. Let u(r, θ ) and U (r, θ ) be harmonic in the respective domains that are
interior and exterior to the circle |z| = R, and let both functions satisfy
the same boundary condition on |z| = R. Show that u(r, θ ) and U (r, θ )
are related by
 2 
R
U (r, θ ) = u , θ − 2π .
r
Hence show that the integral representation of U (r, θ ) is given by
 2π
1 r 2 − R2
U (r, θ ) = − U (R, φ) dφ, r > R.
2π 0 r − 2rR cos(φ − θ ) + R 2
2

7.3. Let u(r, θ ) be the harmonic function in the domain |z| < R that satisfies
the boundary value f (θ ), 0 ≤ θ < 2π, on the circumference of the circle.
Let v(r, θ ) be the harmonic conjugate of u(r, θ ) satisfying the condition
v(0, θ ) = 0. Show that v(r, θ ) is given by
 2π
1 2Rr sin(φ − θ)
v(r, θ ) = − f (φ) dφ, r < R.
2π 0 R − 2Rr cos(φ − θ ) + r 2
2

Further, show that

 2π
2Rr sin(φ − θ )
dφ = 0.
0 R 2 − 2Rr cos(φ − θ ) + r 2

Hint: Show that

 
2Rr sin(φ − θ) ζ +z
− = Im
R 2 − 2Rr cos(φ − θ ) + r 2 ζ −z
  r n
∞
=2 sin n(φ − θ ),
n=1
R

where ζ = Reiφ and z = reiθ .

348 Boundary Value Problems and Initial-Boundary Value Problems

7.4. Let f (z) = u(z) + iv(z) be analytic on and inside the circle |z| = R.
Show that
,
1 ζ + z u(ζ )
f (z) = dζ + iv(0), |z| < R.
2π i |ζ |=R ζ − z ζ
7.5. Consider the Dirichlet problem of finding the function that is harmonic
inside the unit disk |z| < 1. The boundary values are prescribed as 1 on
the upper semi-circle and −1 on the lower semi-circle of the boundary.
Show that the solution to the Dirichlet problem is given by
2 1+z
u(z) = Arg .
π 1−z
From the form of the solution, explain why the value of u(z), when z
approaches 1 or −1, depends on the direction of approach to these two
points.

Hint: The value of Arg(z − 1) = tan−1 x−1 y

, z = x + iy, depends on
the angle of inclination of the line segment joining (x, y) and
(1, 0).

7.6. Show that the Poisson kernel P (R, r, φ − θ) can be expressed as

R2 − r 2
.
|Reiφ − reiθ |2
Explain why

R − r ≤ |Reiφ − reiθ | ≤ R + r,

and hence deduce

R−r R+r
≤ P (R, r, φ − θ ) ≤ .
R+r R−r
Suppose u is a positive harmonic function inside the circle |z| = R; use
the above inequalities to show that
R−r R+r
u(0) ≤ u(reiθ ) ≤ u(0), r < R and − π < θ ≤ π.
R+r R−r
7.7. Suppose u(r, θ ) is harmonic inside the disk |z| < R and assumes the
following boundary values on r = R:

1 θ1 < θ < θ2
u(R, θ ) = .
0 otherwise
 7.4 Problems 349

Let P1 and P2 denote the two points Reiθ1 and Reiθ2 on the boundary
|z| = R. Take any arbitrary point z inside the disk and denote it by P .
Draw the two chords P P1 and P P2 which intersect the circle at P1 and
P2 , respectively. Let S denote the arc length of P1 P2 (see the figure).
Show that the value of u(r, θ ) at the point P is given by
S
u(r, θ ) = .
2π R

P2 [1]
P1

P x 0

P1'
S
P2'

7.8. Use the Schwarz integral formula to show that

U z − x2
u(x, y) = Arg , z = x + iy,
π z − x1
is harmonic in the upper half-plane Im z > 0, and satisfies the boundary
condition

U x 1 < x < x2
u(x, 0) = .
0 otherwise

Remark: Compare this with a similar result in eq. (3.3.14).

7.9. Let f (z) = u(x, y) + iv(x, y), z = x + iy, be analytic in the closed
upper half-plane Im z ≥ 0. Show that

1 ∞ x−t
v(x, y) = u(t, 0) dt, Im z > 0.
π −∞ (t − x)2 + y 2
7.10. Let u be harmonic inside a domain D and assume the zero value along
the boundary ∂D of the domain; that is,

∇ 2 u = 0 in D and u = 0 on ∂D.
350 Boundary Value Problems and Initial-Boundary Value Problems

Using Green’s theorem in the form

,  
∂u ∂u
u dx − dy
∂D ∂y ∂x
 %  2   2  2 &
∂ u ∂ 2u ∂u ∂u
= − u 2
+ 2 + + dxdy,
D ∂x ∂y ∂x ∂y

show that u = 0 throughout the domain D.

7.11. Following a similar approach as used in Problem 7.10, given that a
solution exists for a Neumann problem where the normal derivative is
prescribed along the boundary of the domain, show that the solution is
unique to within an arbitrary additive constant.
7.12. Suppose u(r, θ ) is harmonic inside the disk |z| < R and satisfies the
prescribed Neumann boundary condition along the boundary
∂u
(R, θ ) = g(θ ),
∂r
where g(θ ) is piecewise continuous over [0, 2π ]. Show that the series
representation of u(r, θ ) can be expressed as

a0   r n
∞
u(r, θ ) = + (an cos nθ + bn sin nθ ),
2 n=1
R

where a0 is arbitrary, and the coefficients an and bn are given by

 2π
R
an = g(φ) cos nφ dφ, n = 1, 2, . . . ,
nπ 0
 2π
R
bn = g(φ) sin nφ dφ, n = 1, 2, . . . .
nπ 0
7.13. Find the function u(x, y) that is harmonic in the upper half-plane and
satisfies the following Neumann boundary condition along the x-axis:

∂u 0 if |x| > 1
(x, 0) = .
∂y K if |x| < 1
7.14. Find the function that is harmonic inside the half-disk |z| < R and
Im z > 0, and satisfies the following mixed Dirichlet–Neumann bound-
ary conditions:


u(R, θ ) = f (θ ), 0 < θ < π

.

 ∂u ∂u
 (R, 0) = 0 and (R, π ) = 0
∂θ ∂θ
 7.4 Problems 351

7.15. Show that

 ∞
1 (t − x)2 + y 2
u(x, y) = ln g(t) dt, x > 0 and y > 0,
2π 0 (t + x)2 + y 2
is harmonic inside the domain {z: Re z > 0 and Im z > 0} and satisfies
the boundary conditions


 u(0, y) = 0, y>0

.

 ∂u
 lim (x, y) = g(x), x > 0
y→0 ∂y

7.16. Show that

dn
L{t n f (t)} = (−1)n L{F (s)}, where n is any positive integer.
ds n
Let Y (s) denote the Laplace transform of y(t). In each of the following
ordinary differential equations, find the corresponding governing equa-
tion for Y (s):
d 2y dy
(a) 2
− 2t + λy = 0, y(0) = y0 and y  (0) = y0 ;
dt dt
d 2y dy
(b) t 2 + (1 − t) + λy = 0, y(0) = 0 and y  (0) = 0.
dt dt
7.17. Let F (s) denote the Laplace transform of f (t), t > 0, and define
 T
FT (s) = e−st f (t) dt.
0

Show that

L{f (t + T )} = esT [F (s) − FT (s)].

Illustrate the  with the function f (t) = cos(t + θ ).

 result
7.18. Suppose L f t(t) = G(s) and L{f (t)} = F (s); show that
 ∞
G(s) = F (z) dz.
s

Use the formula to find

   
sin t cos t
(a) L ; (b) L .
t t
7.19. Show that
 t 
1
L f (ξ ) dξ = L{f (t)}.
0 s
352 Boundary Value Problems and Initial-Boundary Value Problems

Use the formula to find

 ∞   ∞ 
sin u cos u
(a) L du ; (b) L du .
t u t u

7.20. Suppose F (s) tends uniformly to zero as R → ∞ when s assumes values

on u ∪ CR ∪ d (see Figure 7.6); that is, |F (s)| ≤ KR on u ∪ CR ∪ d ,
where KR → 0 as R → ∞. Show that
 
(a) lim e F (s) ds = 0; (b) lim
st
est F (s) ds = 0.
R→∞  R→∞ C
u R

Hint: For part (a), s = x + Ri on u and so

|est | = |et(x+Ri) | = ext ≤ eat , since 0 < x < a.

For part (b), s = Reiθ on CR , π

2
≤θ ≤ 3π
2
. Consider
  
  π
 est F (s) ds  ≤ 2RKR etR cos θ dθ
 π
CR 2
 π
2
= 2RKR e−tR sin θ dθ,
0

then apply the Jordan lemma (see Subsection 6.4.2).

7.21. Use the Laplace transform formula

L{f  (t)} = sF (s) − f (0+ ),

to show

lim sF (s) = lim+ f (t).

s→∞ t→0

Verify the result with the following functions:

sin t
(a) f (t) = sin t; (b) f (t) = .
t
7.22. Find the inversion of the following Laplace transform functions:

s+1 1
(a) F (s) = Log ; (b) F (s) = 2 , a > 0.
s (s + a 2 )1/2
 7.4 Problems 353

Hint: For part (a), consider the closed contour shown below:

x
z= 1 z=0

For part (b), the function F (s) = (s 2 +a1 2 )1/2 has branch points
at z = ±ai. Suppose we choose the branch cut to be the line
segment joining z = −ai and z = ai; modify the above closed
contour so that it includes an excursion along the branch cut with
indentations around the two branch points.

7.23. Solve the following integral equation for y(t):

 t
y(τ )
φ(t) = dτ.
0 (t − τ )1/2

Hint: Use the convolution formula and apply the relation

 ∞
sL{φ(t)} = e−st φ  (t) dt.
0

7.24. Consider the differential equation

t 2 y  (t) + ty  (t) + t 2 y(t) = 0
with initial conditions: y(0) = 1 and y  (0) = 0. Let Y (s) denote the
Laplace transform of y(t). Show that
dF (s) s
=− 2 F (s).
ds s +1
By expanding F (s) in negative powers of s, show that the solution is
given by
∞
 (−1)n t 2n
y(t) = .
n=0
(2n n!)2
354 Boundary Value Problems and Initial-Boundary Value Problems

7.25. Prove the following Laplace transform formulas:

s+λ
(a) L{e−λt cos ωt} = ;
(s + λ)2 + ω2
 √
  
1 1 2 a
(b) L √ e−2a t = √ ea /s erfc √ ;
πt s s
 
1 √ 1 −a/s
(c) L √ sin 2 at = √ e ;
πa s s
√
√ a
(d) L{erf( at)} = √ ;
s s+a
√ 1
(e) L{et erfc( t)} = √ ;
s+ s
 
1 √ 1
(f) L √ − e erfc( t) =
t
√ ;
πt 1+ s
 
1 1 1 √ √
(g) L √ sin = √ e− s sin s;
πt 2t s
 
1 1 1 −√s √
(h) L √ cos =√ e cos s.
πt 2t s
7.26. Solve the heat conduction equation

∂T ∂ 2T
= , 0 < x < 1 and t > 0,
∂t ∂x 2
with boundary conditions T (0, t) = T (1, t) = 2, and initial condition
T (x, 0) = 2 + sin π x.
7.27. Solve the heat conduction problem in a semi-infinite rod with insulated
end as modeled by
 2
 ∂T 2∂ T

 = a x > 0, t > 0

 ∂t ∂x 2



∂T .

 (0, t) = 0 t >0

 ∂x




T (x, 0) = f (x) x > 0

7.28. Suppose the temperature on the earth’s surface (assumed to be an infinite

flat surface) has seasonal variations. Let the x-axis be along the direc-
tion normal to the earth’s surface. We would like to find the seasonal
variations of the temperature in the free space above the earth’s surface.
For simplicity, we assume the temperature function T depends on time t
and spatial coordinate x only. The governing equation for T (x, t) can be
 7.4 Problems 355

described by
∂T ∂ 2T
= a2 2 , x > 0 and t > 0.
∂t ∂x
Let the seasonal temperature variations on the earth’s surface be modeled
by the boundary condition T (0, t) = sin ωt, and the initial temperature
is assumed to be zero. Find the solution for T (x, t).
7.29. Consider the heat conduction problem in a semi-infinite rod subject to
the heat flux q(t) incident at the free end x = 0; that is,

∂T 
−k = q(t),
∂x x=0
where k is the conductivity of the material. For simplicity, we assume
that the heat conduction equation takes the form
∂T ∂ 2T
= , x > 0, t > 0,
∂t ∂x 2
and the initial temperature is T (x, 0) = 0, x > 0. Solve the problem
under the following two special cases:

(a) q(t) = q0 ; (b) q(t) = q0 sin ωt.

Hint: The temperature is obtained by evaluating the Bromwich integral

, √ Q(s)
1
T (x, t) = esτ − sx √ ds,
2π ik CB s
where Q(s) = L{q(t)}. In part (a) where q(t) = q0 , it is neces-
sary to consider the derivative
, √
∂T q0 1 esτ − sx
=− ds
∂x k 2π i CB s
so that the contour integral around the infinitesimal circle |z| = 
(see Figure 7.7) becomes zero.

7.30. Use the Laplace transform method to solve the heat conduction problem
within a finite rod of length L. The governing equation is given by
∂T ∂ 2T
= a 2 2 , 0 < x < L, t > 0,
∂t ∂x
subject to the boundary conditions T (0, t) = T1 and T (L, t) = T2 , where
T1 and T2 are constants, and initial condition T (x, 0) = 0, 0 < x < L.
356 Boundary Value Problems and Initial-Boundary Value Problems

Hint:
 √  ∞
−1 sinh x s x 2  (−1)n −n2 π 2 t/L2 nπx
L √ = + e sin .
s sinh L s L π n=1
n L

7.31. The heat conduction problem in an infinite rod with the non-
homogeneous term f (x, t) and initial condition φ(x) is posed as


 ∂T ∂ 2T
 = a 2 2 + f (x, t), −∞ < x < ∞, t > 0
∂t ∂x .



T (x, 0) = φ(x)

Show that the solution is given by

 ∞
1 (x−ξ )2
u(x, t) = √ φ(ξ )e− 4a2 t dξ
2
4π a t −∞
 t ∞ )2
1 f (ξ, τ ) − 4a(x−ξ
+ √ √ e 2 (t−τ ) dξ dτ.
2a π 0 −∞ t − τ

7.32. Use the Fourier transform method to solve the following wave equation:
 2

 ∂ u ∂ 2u

 ∂t 2 = + t sin x, −∞ < x < ∞ and t > 0
∂x 2
.



u(x, 0) = 0 and ∂u (x, 0) = sin x.
∂t

7.33. Use the Laplace transform method to solve the following wave equation:
 2
 ∂ u ∂ 2u

 = + sin π x, 0 < x < 1 and t > 0

 ∂t 2 ∂x 2



∂u .

 u(x, 0) = 0 and (x, 0) = 0

 ∂t




u(0, t) = u(1, t) = 0

7.34. Consider the wave propagation along an infinite elastic rod with ini-
tial displacement and velocity defined by f (x) and g(x), respectively,
and suppose the displacement and velocity die off at the far ends. Let
u(x, t) denote the displacement function. The wave propagation problem
 7.4 Problems 357

is formulated as follows:
 2 2

 ∂ u 2∂ u

 ∂t 2 = c , −∞ < x < ∞, t > 0
∂x 2
.



 u(x, 0) = f (x) and ∂u
(x, 0) = g(x)
∂t
By taking the Laplace transform of the wave equation with respect to t,
show that the governing equation for U (x, s) = L{u(x, t)} is given by
∂ 2u
s 2 U (x, s) − sf (x) − g(x) − c2 (x, s) = 0.
∂x 2
Solve the above equation and explain the steps that lead to
 ∞ −s(ξ −x)/c  x −s(x−ξ )/c
1 e 1 e
U (x, s) = g(ξ ) dξ + g(ξ ) dξ
2c x s 2c −∞ s
 ∞ −s(ξ −x)/c  x −s(x−ξ )/c
1 e 1 e
+ sf (ξ ) dξ + sf (ξ ) dξ.
2c x s 2c −∞ s
By performing the inversion of the Laplace transform U (x, s), show that
 x+ct
1 1
u(x, t) = g(ξ ) dξ + [f (x + ct) + f (x − ct)].
2c x−ct 2
The result is called the D’Alembert formula.

Hint: Using the Laplace inversion formula

 −s(ξ −x)/c   
−1 e ξ −x
L =H t− ,
s c
show that
  ∞ −s(ξ −x)/c   x+ct
1 e 1
L−1 g(ξ ) dξ = g(ξ ) dξ.
2c x s 2c x

7.35. Use the Laplace transform method to solve

∂u ∂u
x + = x, x > 0, t > 0,
∂t ∂x
subject to the auxiliary conditions u(x, 0) = 0 for x > 0 and u(0, t) = 0
for t > 0.
 8
Conformal Mappings and Applications

A complex function w = f (z) can be regarded as a mapping from its domain

in the z-plane to its range in the w-plane. In this chapter, we go beyond the
previous chapters by analyzing in greater depth the geometric properties asso-
ciated with mappings represented by complex functions. First, we examine the
linkage between the analyticity of a complex function and the conformality of
a mapping. A mapping is said to be conformal at a point if it preserves the
angle of intersection between a pair of smooth arcs through that point. The
invariance of the Laplace equation under a conformal mapping is also estab-
lished. This invariance property allows us to use conformal mappings to solve
various types of physical problem, like steady state temperature distribution,
electrostatics and fluid flows, where problems with complicated configurations
can be transformed into those with simple geometries.
First, we introduce various techniques for effecting the mappings of regions.
Two special classes of transformation, the bilinear transformations and the
Schwarz–Christoffel transformations, are discussed fully. A bilinear transfor-
mation maps the class of circles and lines to the same class, and it is conformal
at every point except at its pole. The Schwarz–Christoffel transformations take
half-planes onto polygonal regions. These polygonal regions can be unbounded
with one or more of their vertices at infinity. We also consider the class of
hodograph transformations, where the roles of the dependent and independent
variables are reversed.

8.1 Conformal mappings

Let f be a complex function that is analytic at a point z0 and f  (z0 ) = 0. Let γ
denote a smooth arc through z0 , whose parametric form is given by

z(t) = x(t) + iy(t), a ≤ t ≤ b, (8.1.1)

358
 8.1 Conformal mappings 359

w = f (z)
y v

>
arg w'(t0)

>
arg z' (t0) w0
z0 x

x u

Figure 8.1. The curve γ through z0 is carried under the mapping w = f (z) to the curve
 through w0 = f (z0 ).

and z(t0 ) = z0 . Recall that the arc represented by z(t) is smooth if x(t) and y(t)
are both continuously differentiable and x  (t) and y  (t) do not both vanish at
the same value of t. The image of γ under the mapping w = f (z), denoted by
, can be represented by

w(t) = f (z(t)), a ≤ t ≤ b. (8.1.2)

We would like to examine the change in the direction of a curve through the
point z0 under the mapping w = f (z). The chain rule of differentiation gives

w  (t) = f  (z(t))z (t), a ≤ t ≤ b,

in particular,

w (t0 ) = f  (z0 )z (t0 ).

Given that z (t0 ) = 0 (corresponding to γ being a smooth curve) and f  (z0 ) =
0, it then follows that w (t0 ) is nonzero. Hence we have
arg w  (t0 ) = arg f  (z0 ) + arg z (t0 ). (8.1.3)

Here, arg z (t0 ) and arg w  (t0 ) give the angles of inclination of the tangent vectors
to the curve γ at z0 and the image curve  at w0 = f (z0 ), respectively (see
Figure 8.1). By virtue of eq. (8.1.3), the tangent vector to γ at z0 is rotated in the
anticlockwise sense through the angle arg f  (z0 ) under the mapping w = f (z).
The magnitude and sense of this rotation are dependent only on the mapping
f (z), but independent of the curve γ .
Let γ1 and γ2 be a pair of smooth arcs intersecting at z0 , and let α be the
angle between the tangent vectors to the two arcs at z0 . Since the two tangent
vectors are rotated in the same sense and by the same magnitude under the
360 Conformal Mappings and Applications

w = f (z)

y v

w0= f (z0)
z0

x u

Figure 8.2. The mapping w = f (z) is conformal at z0 since it preserves the angle of
intersection between a pair of smooth arcs through the point.

mapping w = f (z), the angle between the tangent vectors to the image curves
1 and 2 at w0 = f (z0 ) is also α (see Figure 8.2). This property of angle
preservation under a mapping is formally defined as follows:

Definition 8.1.1 A mapping that preserves the magnitude and sense of the
angle between any two smooth arcs passing through a given point is said to be
conformal at that point.
The relation between analyticity and the above angle-preserving property of
a complex function is stated in the following theorem.

Theorem 8.1.1 An analytic function f is conformal at every point z0 for

which f  (z0 ) = 0.

Proof In the z-plane, we draw two smooth arcs γ1 and γ2 through z0 . The
smooth arcs are parametrized by
z1 (t) = x1 (t) + iy1 (t), a1 ≤ t ≤ b1 ,
z2 (s) = x2 (s) + iy2 (s), a2 ≤ s ≤ b2 ,
with z1 (t0 ) = z2 (s0 ) = z0 . The included angle α, measured in the anti-clockwise
sense, from the tangent vector z1 (t0 ) to the tangent vector z2 (s0 ) is given by
(see Figure 8.2)
α = arg z2 (s0 ) − arg z1 (t0 ).
Sometimes, special care may be needed to adjust appropriately the argument
of one of the tangent vectors by adding or subtracting 2π . The image curves 1
 8.1 Conformal mappings 361

and 2 under the mapping w = f (z) are represented by w1 (t) = f (z1 (t)) and
w2 (s) = f (z2 (s)), respectively, and they intersect at w0 = f (z0 ) (see Figure
8.2). The angle, measured in the anticlockwise sense, from the tangent vector
w1 (t0 ) to the tangent vector w2 (s0 ) is given by arg w2 (s0 ) − arg w1 (t0 ). By virtue
of eq. (8.1.3), this included angle between the image curves is the same as α
since
arg w2 (s0 ) − arg w1 (t0 ) = [arg f  (z0 ) + arg z2 (s0 )] − [arg f  (z0 ) + arg z1 (t0 )]
= arg z2 (s0 ) − arg z1 (t0 ) = α. (8.1.4)
Note that analyticity of f at z0 and non-vanishing of f  (z0 ) are required to
establish eq. (8.1.4). The mapping is seen to preserve the magnitude and sense
of the angle between any two smooth arcs passing through z0 , and so the
function f is conformal at z0 .

Scale factor
We have seen that arg f  (z0 ) gives the angle of rotation of the tangent vector to
the curve γ at z0 under the mapping w = f (z). Can we attach any geometric
meaning to the modulus |f  (z0 )|? Let z − z0 = reiθ and w − w0 = ρeiφ , where
z and w are points on the curve γ and the image curve , respectively. Let s
be the arc length between z0 and z along γ , and σ be the arc length between
w0 and w along . We see that as z tends to z0 , w tends to w0 . Consider the
ratio
w − w0 f (z) − f (z0 ) ρeiφ σ ρ s i(φ−θ)
= = iθ = e ,
z − z0 z − z0 re s σ r
and observe that
ρ s
lim =1 and lim = 1.
w→w0 σ z→z0 r
We obtain
 
 f (z) − f (z0 )  σ |f (z) − f (z0 )|
 
|f (z0 )| =  lim = lim = lim . (8.1.5)
z→z0 z − z0  z→z0 s z→z0 |z − z0 |
As z → z0 , |f (z) − f (z0 )| ≈ |f  (z0 )| |z − z0 |, so the quantity |f  (z0 )| is visu-
alized as the scale factor of the conformal transformation f (z) at z0 (see also
Example 8.1.2).
There is a close link between the one-to-one property of a mapping and
non-vanishing of the derivative value of the mapping function. We state the
following result without proof: if f is analytic at z0 and f  (z0 ) = 0, then there
exists a neighborhood of z0 such that f is one-to-one inside that neighborhood.
362 Conformal Mappings and Applications
w = z2
y v

i
γ1
>
Γ1
<

γ2
<

> x > u
0 γ3 1 −1 Γ2 0 Γ3 1

Figure 8.3. The mapping w = z2 takes the quarter circle |z| ≤ 1 and 0 ≤ Arg z ≤ π2 in
the z-plane onto the upper semi-circle |w| ≤ 1 and 0 ≤ Arg w ≤ π in the w-plane.

Example 8.1.1 Discuss the conformal property of w = f (z) = z2 inside the

domain
 π
D = z : 0 < |z| < 1, 0 < Arg z < .
2

Solution The analytic function w = f (z) = z2 is seen to be conformal at all

points inside D since f  vanishes only at z = 0 (which does not lie inside D).
The image of D in the w-plane under the given mapping is the upper semi-circle

D = {w : 0 < |w| < 1, 0 < Arg w < π }.

Let γ = γ1 ∪ γ2 ∪ γ3 be the boundary of D, where γ1 is the arc of the unit

circle, γ2 is the line segment along the imaginary axis, and γ3 is the line segment
along the real axis. The corresponding image curves of γ1 , γ2 and γ3 are denoted
by 1 , 2 and 3 , respectively (see Figure 8.3).
The curves γ1 and γ2 intersect orthogonally at z = i, as do the image curves
1 and 2 at w = −1. The same phenomenon holds for the curves γ3 and γ1
at z = 1 and the image curves 3 and 1 at w = 1. These observations agree
with the conformal property of w = z2 at z = i and z = 1.
However, though the curves γ2 and γ3 intersect orthogonally at z = 0, the
angle of intersection of the image curves 2 and 3 at w = 0 is zero. This
is not surprising since the mapping w = z2 fails to be conformal at z = 0.
Actually, the magnification of the included angle at a critical point of the
mapping function depends on the order of the critical point (see discussion
below).
 8.1 Conformal mappings 363

Mapping behavior around a critical point

Suppose z0 is a critical point of an analytic function f , that is, f  (z0 ) = 0.
In this case, f fails to be conformal at z0 . Let m be the lowest order such
that f (m) (z0 ) = 0, that is, f  (z0 ) = · · · = f (m−1) (z0 ) = 0. What would be the
mapping property of w = f (z) around z = z0 ?
The Taylor series for f (z) at z0 becomes
f (m) (z0 ) f (m+1) (z0 )
w = f (z) = f (z0 ) + (z − z0 )m + (z − z0 )m+1 + · · · .
m! (m + 1)!
At points close to z0 , we may neglect the higher-order terms and obtain the
approximation
f (m) (z0 )
w − w0 ≈ (z − z0 )m , w0 = f (z0 ). (8.1.6)
m!
The arguments and moduli of w − w0 and z − z0 are then approximately related
by

arg(w − w0 ) ≈ arg f (m) (z0 ) + m arg(z − z0 ) (8.1.7a)

and
 (m) 
 f (z0 ) 

|w − w0 | ≈  |z − z0 |m . (8.1.7b)
m! 
Suppose we write z − z0 = reiθ and w − w0 = ρeiφ , where z and w are
points on the curve γ and the image curve , respectively. From eq. (8.1.7a),
we deduce that

φ ≈ arg f (m) (z0 ) + mθ. (8.1.8)

As z → z0 , θ and φ tend to the angle of inclination of the tangent vector to

the curve γ at z0 and the image curve  at w0 , respectively. Let α be the angle
between the two smooth arcs γ1 and γ2 intersecting at z0 ; then α = θ2 − θ1 ,
where θ1 and θ2 are the respective angles of inclination of the tangent vectors
to γ1 and γ2 at z0 . The corresponding angle between the tangent vectors to the
image arcs 1 and 2 intersecting at w0 is given by

[arg f (m) (z0 ) + mθ2 ] − [arg f (m) (z0 ) + mθ1 ] = m(θ2 − θ1 ) = mα. (8.1.9)

That is, the included angle between the image arcs in the w-plane is m times
that between the smooth arcs in the z-plane.
Referring to the mapping function w = z2 considered in Example 8.1.1,
z = 0 is a critical point with m = 2. Consider the pair of curves z = t, t ≥ 0,
and z = is, s ≥ 0, where they intersect at z = 0 in the z-plane. The image
364 Conformal Mappings and Applications

curves under the mapping w = z2 in the z-plane become w = t 2 , t ≥ 0, and

w = −s 2 , s ≥ 0. These image curves intersect at w = 0. The included angle
between the pair of curves in the z-plane is π2 while that between the image
curves in the w-plane is π. This is expected since m = 2, so the included angle
is magnified by a factor of 2 under the mapping.

Example 8.1.2 Suppose the conformal transformation w = f (z) maps a

domain D in the z-plane onto the domain  in the w-plane. Show that the
area of  is given by
 
A= |f  (z)|2 dx dy.
D

Solution The area of the domain  in the w-plane is given by

   
∂(u, v)
A= du dv = dx dy, w = u + iv, z = x + iy,
 D ∂(x, y)
∂(u,v)
where ∂(x,y)
is the Jacobian of the transformation. One can establish
 2  2
∂(u, v) ∂u ∂v ∂v ∂u ∂u ∂u
= − = + ,
∂(x, y) ∂x ∂y ∂x ∂y ∂x ∂y
by virtue of the Cauchy–Riemann relations. On the other hand, we have
∂u ∂v ∂u ∂u
f  (z) = +i = −i ,
∂x ∂x ∂x ∂y
so
 2  2
∂u ∂u
|f  (z)|2 = + ;
∂x ∂y
thus the area formula is established. The result is not surprising since |f  (z)|
gives the local scale factor of the conformal transformation at z [see eq. (8.1.5)].

8.1.1 Invariance of the Laplace equation

In the previous chapters, we have obtained various forms of the solution to the
Laplace equation where the domain of the problem has a simple configuration,
like the unit circle and the upper half-plane. However, most real life physical
problems involve configurations with complicated geometries. Suppose a con-
formal mapping can be found that maps a complicated domain onto a simple
domain; can the solution to the Laplace equation in the mapped domain be
carried over to the original domain? In other words, given that φ(x, y) is har-
monic in a domain D in the z-plane, z = x + iy, and that w = u + iv = f (z)
 8.1 Conformal mappings 365

is a conformal mapping in D which takes the domain D onto the domain  in

the w-plane, does φ(u, v) remain harmonic in ?
Fortunately, the answer to the above question is ‘yes’. This property is termed
the invariance of the Laplace equation, where
 2 
∂ 2φ ∂ 2φ  2 ∂ φ ∂ 2φ
+ 2 = |f (z)| + 2 . (8.1.10)
∂x 2 ∂y ∂u2 ∂v
Note that φ(x, y) is transformed into the function φ(x(u, v), y(u, v)) by the
transformation w = u + iv = f (z), z = x + iy. Suppose f (z) is conformal in
D so that f  (z) = 0 in D; eq. (8.1.10) dictates that
∂ 2φ ∂ 2φ
+ = 0 in D
∂x 2 ∂y 2
if and only if
∂ 2φ ∂ 2φ
2
+ 2 = 0 in .
∂u ∂v
The mathematical steps required to show eq. (8.1.10) amount to straightfor-
ward differentiation. First, we observe that
∂φ ∂φ ∂u ∂φ ∂v ∂φ ∂φ ∂u ∂φ ∂v
= + and = + ,
∂x ∂u ∂x ∂v ∂x ∂y ∂u ∂y ∂v ∂y
so the second derivatives are found to be
 2 
∂ 2φ ∂φ ∂ 2 u ∂ φ ∂u ∂ 2 φ ∂v ∂u ∂φ ∂ 2 v
= + + +
∂x 2 ∂u ∂x 2 ∂u2 ∂x ∂v∂u ∂x ∂x ∂v ∂x 2
 2 
∂ φ ∂u ∂ 2 φ ∂v ∂v
+ + 2 ,
∂u∂v ∂x ∂v ∂x ∂x
 2 
∂ 2φ ∂φ ∂ 2 u ∂ φ ∂u ∂ 2 φ ∂v ∂u ∂φ ∂ 2 v
= + + +
∂y 2 ∂u ∂y 2 ∂u2 ∂y ∂v∂u ∂y ∂y ∂v ∂y 2
 2 2 
∂ φ ∂u ∂ φ ∂v ∂v
+ + 2 .
∂u∂v ∂y ∂v ∂y ∂y
Adding the above two equations together, we obtain
  %   2 &
∂ 2φ ∂ 2φ ∂φ ∂ 2 u ∂ 2 u ∂ 2φ ∂u 2 ∂u
+ 2 = + 2 + 2 +
∂x 2 ∂y ∂u ∂x 2 ∂y ∂u ∂x ∂y
   
∂ 2 φ ∂u ∂v ∂u ∂v ∂φ ∂ 2 v ∂ 2v
+2 + + + 2
∂v∂u ∂x ∂x ∂y ∂y ∂v ∂x 2 ∂y
%    &
2 2
∂ 2φ ∂v ∂v
+ 2 + .
∂v ∂x ∂y
366 Conformal Mappings and Applications

Since w = u + iv = f (z) is analytic, both u and v are harmonic and so the first
and fourth terms in the above equation are zero. Further, since u and v satisfy
the Cauchy–Riemann relations, the third term also vanishes. Hence, the above
equation reduces to
%   2 & 2 %   2 & 2
∂ 2φ ∂ 2φ ∂u 2 ∂u ∂ φ ∂v 2 ∂v ∂ φ
2
+ 2 = + 2
+ +
∂x ∂y ∂x ∂y ∂u ∂x ∂y ∂v 2
 2 
∂ φ ∂ 2φ
= |f  (z)|2 + ,
∂u2 ∂v 2
and thus eq. (8.1.10) is established.
Given the invariance property of the Laplace equation, a wide range of
physical problems with complicated configurations that are governed by the
Laplace equation can be solved by finding an appropriate conformal mapping
that maps the given domain onto either the unit circle or the upper half-plane.
Such a solution approach is illustrated in the following examples, which include
applications in electrostatics, steady state temperature distribution and fluid
flows.

Example 8.1.3 Find the electric

+ potential inside the semi-infinite strip {z = x
+ iy : − 2 < x < 2 , y > 0 , where the potential value along the bottom
π π

boundary is kept constant at φ0 , and the potential values along the vertical
boundaries are zero.

Solution Let φ(x, y) denote the electric potential inside the strip. The elec-
tric potential function is known to satisfy the Laplace equation (see Sub-
section 4.4.2). It was shown in Subsection 3.2.1 that the mapping function
w = sin z carriesthe above strip onto the upper half-plane conformally, with
the base − π2 , π2 of the strip going to the segment [−1, 1] and the vertical
sides to the horizontal rays (−∞, −1] and [1, ∞) (see Figure 8.4).
In the transformed plane, by virtue of the invariance property of the Laplace
equation, the electric potential φ(u, v) remains harmonic. The corresponding
boundary potential values along the u-axis are

φ0 −1 < u < 1
φ(u, 0) = .
0 u > 1 or u < −1
Using the techniques developed in Subsection 3.3.2, the solution to the electric
potential in the w-plane is readily found to be
φ0 w−1
φ(u, v) = Arg , w = u + iv.
π w+1
 8.1 Conformal mappings 367

Figure 8.4. The mapping w = sin z carries the semi-infinite strip onto the upper half-
plane. The curves inside the strip are the equipotential lines.

Referring back to the z-plane, the electric potential inside the strip is given by

φ0 sin z − 1 φ0 2 Im(sin z)
φ(x, y) = Arg = tan−1
π sin z + 1 π | sin2 z| − 1
φ0 2 cos x sinh y φ0 2 cos x/ sinh y
= tan−1 = tan−1
π sinh2 y − cos2 x π 1 − (cos x/ sinh y)2
2α
= φ0 ,
π

where tan α = sinh

cos x
and tan2 α = 1−tan
2 tan α
2 α . Since cos x assumes value within
 yπ π 
[0, 1] when x ∈ − 2 , 2 and sinh y assumes a positive value for y > 0, α can
assume any value within the semi-infinite strip. The equipotential lines inside
the strip correspond to the level curves sinhcos x
y
= k, for some positive constant k.

Example 8.1.4 Find the steady state temperature T (x, y) in the upper half-
plane y > 0 if the boundary temperature along the x-axis is kept at

T0 −∞ < x < −1
T (x, 0) = ,
T1 1<x<∞

and the wall is insulated within |x| < 1; that is,

∂T
(x, 0) = 0, |x| < 1.
∂y
368 Conformal Mappings and Applications
z= cos w

y
v

[ T 0] [ T 1]

x u
[ T 0] 1 [insulated] 1 [ T 1] 0 [insulated] 1

Figure 8.5. The mapping z = − cos π w takes the upper half z-plane onto the semi-
infinite strip in the w-plane. The isothermal lines are orthogonal to the insulated
boundaries.

This is a mixed Dirichlet–Neumann problem where values of the dependent

variable are prescribed on some part of the boundary and derivative values are
prescribed on the remaining part of the boundary.

Solution In the last example, we facilitate the solution by finding a mapping

that takes the given strip onto the upper half-plane. Though the configuration
in the present problem is already the upper half-plane, the solution procedure
is complicated by the presence of the Neumann condition along a part of the
boundary. Interestingly, we solve this problem by considering a mapping that
takes the upper half-plane onto a strip in the transformed plane, just the reverse
of the usual procedure. This is because the Neumann boundary condition can
be treated easier under the configuration of a strip.
We choose the mapping z = − cos π w which takes the upper half z-plane
onto the semi-infinite strip {w = u + iv : 0 < u < 1 and 0 < v < ∞} (see
Figure 8.5).
In the transformed w-plane, the boundary conditions are:

(i) T = T0 along the left vertical edge u = 0, v > 0;

(ii) ∂T
∂v
= 0 along the bottom end v = 0, 0 < u < 1;
(iii) T = T1 along the right vertical edge u = 1, v > 0.

Since the normal derivative of the temperature at points along an insulated

boundary must be zero, the isothermal lines must be orthogonal to an insulated
surface. In the w-plane, the isothermal lines within the strip must be parallel
to the v-axis. One then naturally deduces that the temperature T (u, v) in the
 8.1 Conformal mappings 369

w-plane is given by

T (u, v) = T0 + (T1 − T0 )u.

To verify that this is the solution, we check that T (u, v) is harmonic inside the
strip and all boundary conditions along the boundaries of the strip are satisfied.
By the uniqueness property of solutions to the Laplace equation, this is the
solution to the mixed Dirichlet–Neumann problem.
The next step is to express u in terms of x and y. Noting that

z = − cos π(u + iv) = − cos π u cosh π v + i sin π u sinh π v,

we have

(cosh π v − cos π u)2 = cosh2 π v − 2 cosh π v cos π u + cos2 π u

= x 2 + 2x + 1 + y 2 .

By taking the positive square root on both sides, we have

cosh π v − cos π u = (x + 1)2 + y 2 .

Similarly, we obtain


cosh π v + cos π u = (x − 1)2 + y 2 .

In the original z-plane, the temperature is then given by

 


T1 − T0 −1 1
T (x, y) = T0 + cos (x − 1) + y − (x + 1) + y
2 2 2 2 .
π 2
The isothermal lines in the z-plane are given by the level curves:



(x − 1)2 + y 2 − (x + 1)2 + y 2 = k,

where k is some value lying between T0 and T1 . These isothermal lines can
be shown to be circles that cut orthogonally the insulated boundary along the
segment [−1, 1] on the x-axis. The temperature value along the boundary y = 0
is found to be



T 1 − T0 −1 (x − 1)2 − (x + 1)2
T (x, 0) = T0 + cos
π 2


 T if x < −1
 0
T1 − T0 −1
= T0 + cos (−x) if − 1 ≤ x ≤ 1 .

 π
T if x > 1
1
370 Conformal Mappings and Applications

T(x, 0)

T1

T0

x
−1 1
Figure 8.6. The boundary temperature T (x, 0) takes constant value outside [−1, 1] and
assumes the value of an inverse cosine function within [−1, 1].

The plot of T (x, 0) against x is shown in Figure 8.6. Interestingly, the tem-
perature assumes the value of an inverse cosine function along the insulated
surface.

Example 8.1.5 Find the complex potential for the potential flow around a flat
plate of length 2a that is oriented to be perpendicular to the incoming uniform
flow of velocity U .

Solution The configuration of a uniform flow past a perpendicular obstacle

is shown schematically in Figure 8.7. By symmetry, the x-axis is a streamline
dividing the flow field into two equal halves. It suffices to seek the solution to
the flow field in the upper half-plane.
We try to find a mapping that carries the upper half z-plane (excluding the
x-axis) minus the segment from z = 0 to z = ai along the imaginary axis to
the upper half w-plane. First, we take w1 = z2 which maps the above domain
onto the whole w1 -plane minus the segment [−a 2 , ∞) along the real axis. Next,
we take the translation w2 = w1 + a 2 so that the mapped domain is the whole
w2 -plane minus
√ √ the segment [0, ∞) along the real axis. √ Finally, we choose
w = w2 = z2 + a 2 (the branch which takes z = 2ai to w = ai). This
mapping carries the given domain in the z-plane to the upper half w-plane. The
sequence of mappings is shown in Figure 8.8.
In the w-plane, the flow configuration becomes uniform and rectilinear. The
speed of the flow stays at the same value U since
√
w z2 + a 2
lim = lim = 1.
z→∞ z z→∞ z
 8.1 Conformal mappings 371

(0, a)

x
(0, a)

Figure 8.7. The configuration of a uniform flow past a perpendicular obstacle of length
2a.

y w1 = z2 v1
z-plane w 1- plane

(0, a )

( a 2, 0) u1
x

w 2 = w1 + a 2
w- plane w 2- plane
v v2

u u2

w = w2 = z 2 + a 2

Figure 8.8. The sequence of mappings that takes the upper half-plane minus the vertical
segment onto the whole upper half-plane.

The complex potential of the flow in the w-plane is given by

f (w) = U w.

Referring back to the z-plane, the complex potential then becomes

f (z) = U z2 + a 2 .
372 Conformal Mappings and Applications

The complex velocity is given by

Uz
V (z) = f  (z) = √ .
z2 + a 2
At the tips z = ±ai of the obstacle, the values of the velocity become infinite.

8.1.2 Hodograph transformations

Suppose we would like to seek the complex potential f = φ + iψ of a potential
flow problem, where values of φ are prescribed on some parts of the bound-
ary of the physical domain and values of ψ are prescribed on the remaining
parts. The solution approach then amounts to finding a mapping function that
carries the physical plane onto the φ-ψ plane. In this case, the desired complex
potential f = φ + iψ can be viewed as a mapping. In other cases, the depen-
dent variables may be the x- and y-component velocities, denoted by u and v,
respectively (see Problem 8.20). We attempt to solve for a mapping function
that carries the physical plane onto the u-v plane. These types of mapping
are called the hodograph transformations, where the dependent variables in
the physical problem are employed as the independent variables in the new
formulation. To illustrate the technique, we examine the use of a hodograph
transformation in the study of the seepage flow under a gravity dam.

Seepage flow under a gravity dam

We would like to study the effect of water seepage through the soil from a
waterhead upstream on one side of the dam to downstream on the other side
(see Figure 8.9). The waterheads upstream and downstream are taken to be h1
and h2 (h1 > h2 ), respectively. The dam has a base of width 2L. The coordinate
axes are chosen such that the soil level coincides with the x-axis and the base
of the dam lies between −L ≤ x ≤ L.
A rather ideal form of seepage flow through soil and flow configuration is
assumed here. The soil underneath the water and the dam is homogeneous and
has an infinite depth so that the domain of the seepage is the whole lower
half-plane. The velocity of the flow is so slow that viscosity is negligible, and
so potential flow can be assumed for the seepage. The dam is assumed to be
completely impermeable.
The flow problem is considered solved if the complex potential of the seepage
is known. The complex potential is of the form

w = f (z) = φ(x, y) + iψ(x, y), z = x + iy, (8.1.11)

 8.1 Conformal mappings 373

Figure 8.9. Seepage flow through the soil from the waterhead upstream on one side of
the dam to downstream on the other side.

where φ and ψ are the velocity potential and stream function, respectively. We
adopt Darcy’s law for fluid flow inside a porous medium, which states that the
velocity of fluid flow through fine sands and soils is directly proportional to the
pressure gradient. That is

|∇P | ∝ |∇φ|, (8.1.12)

where P is the pressure. With reference to the material properties associated

with flow in a porous medium, Darcy’s law becomes
η
P =− φ, (8.1.13)
K
where η is the coefficient of viscosity of water and K is the coefficient of
permeability of the soil. Here, both coefficients are assumed to be constant.
Using this empirical law of seepage flow, we manage to relate the waterheads
with the velocity potential.
The boundary conditions along the x-axis for the seepage are prescribed as
follows. The impervious wall is a streamline so that ψ(x, 0) is constant for
−L ≤ x ≤ L. Without loss of generality, we take

ψ(x, 0) = 0, −L ≤ x ≤ L. (8.1.14)

Along the soil surface to the left side of the dam, x ≤ −L and y = 0, the
pressure head is given by P (x, 0) = ρgh1 , where ρ is the density of water and
g is the acceleration due to gravity. From eq. (8.1.13), we then have
K
φ(x, 0) = − ρgh1 = φ1 , x ≤ −L. (8.1.15a)
η
374 Conformal Mappings and Applications

Figure 8.10. The boundary values assumed by φ and ψ along the soil surface dictate
the mapping that carries the lower half x-y plane onto a semi-infinite strip in the φ-ψ
plane.

Similarly, the boundary condition along the soil surface to the right side of the
dam, x ≥ L and y = 0, is given by

K
φ(x, 0) = − ρgh2 = φ2 , x ≥ L. (8.1.15b)
η

Both φ1 and φ2 are negative quantities. Now, the hodograph transformation w =

f (z) = φ(x, y) + iψ(x, y), z = x + iy, maps the lower half z-plane onto the
semi-infinite strip φ1 ≤ φ ≤ φ2 and ψ ≤ 0 in the w-plane (see Figure 8.10). The
stream function values in the lower half z-plane are negative since ψ(x, 0) = 0
along a part of the real axis and the seepage runs from the left side to the right
side of the dam, implying ∂ψ ∂y
> 0.
Recall that the mapping ζ = cos−1 ( Lz ) maps the lower half z-plane onto
the strip {ζ : 0 ≤ Re ζ ≤ π and Im ζ ≥ 0} in the ζ -plane. Hence, the required
transformation which effects the mapping shown in Figure 8.10 is deduced
to be

φ1 − φ2 z
w = f (z) = cos−1 + φ2 . (8.1.16)
π L

It is not possible to express φ and ψ explicitly as functions of x and y. Rather,

the reverse is possible, where x and y can be expressed in terms of φ and ψ.
From eq. (8.1.16), we can obtain

x π(φ − φ2 ) πψ
= cos cosh
L φ 1 − φ2 φ 1 − φ2
 8.2 Bilinear transformations 375

and
y π(φ − φ2 ) πψ
= − sin sinh .
L φ 1 − φ2 φ 1 − φ2
Once the complex potential f (z) is known, the properties of the seepage flow
field can be analyzed. For example, the streamline ψ = ψ0 (< 0) is the lower
2 2
half-portion of the ellipse xa 2 + yb2 = 1, where
π ψ0 π ψ0
a = L cosh and b = L sinh .
(φ1 − φ2 ) (φ1 − φ2 )
The equipotential line φ = φ0 (φ1 ≤ φ0 ≤ φ2 ) is the lower half-portion of the
2 2
hyperbola Ax 2 − By 2 = 1, where
cos π (φ0 − φ2 ) sin π (φ0 − φ2 )
A=L and B=L .
φ 1 − φ2 φ 1 − φ2
A civil engineer would be interested in finding the upward lift exerted by
the seepage on the base of the dam. It turns out that this upward lift is given
by the product of the width of the base and the average pressure due to the
waterheads upstream and downstream. To prove the claim, we consider the
pressure P (x, 0) along the base of the dam, −L ≤ x ≤ L. From Darcy’s law,
P (x, 0) = − Kη φ(x, 0). Further, since ψ(x, 0) = 0, the complex potential f (x)
is real when z = x and correspondingly, φ(x, 0) = f (x). Using eq. (8.1.16),
we obtain
 
η φ 1 − φ2 −1 x
P (x, 0) = − cos + φ2
K π L
P1 − P2 x
= cos−1 + P2 . (8.1.17)
π L
The total uplift exerted on the base of the dam is given by
 L  L
P1 − P2 x P1 + P2
P (x, 0) dx = cos−1 + P2 dx = (2L). (8.1.18)
−L −L π L 2

8.2 Bilinear transformations

The transformation defined by
az + b
w = f (z) = , (8.2.1)
cz + d
where a, b, c and d are complex numbers and ad − bc = 0 is called a bilinear
transformation (or linear fractional transformation). It is so named because it
takes the form of the ratio of two linear functions. The bilinear transformation
376 Conformal Mappings and Applications

can be decomposed as
a bc − ad 1
w = f (z) = + , c = 0. (8.2.2)
c c cz + d
Clearly, the restriction ad − bc = 0 is essential in order to ensure that f (z) is
not a constant function. When c = 0, f (z) becomes a linear transformation.

One-to-one mapping
The bilinear transformation is a one-to-one mapping of the extended plane
onto itself. In other words, the bilinear transformation maps distinct points
onto distinct images. This means
f (z1 ) = f (z2 ) if and only if z 1 = z2 .
The “if” part is obvious. To show the “only if” part, we assume
az1 + b az2 + b
= f (z1 ) = f (z2 ) = .
cz1 + d cz2 + d
After some manipulation, we obtain
(ad − bc)z1 = (ad − bc)z2 ,
which implies z1 = z2 since ad − bc = 0.
Since a bilinear transformation is one-to-one, its inverse always exists. The
inverse transformation is obtained by solving eq. (8.2.1) for z. This gives
−dw + b
z= . (8.2.3)
cw − a
 
When c = 0, w = f (z) has a simple pole at − dc so that f − dc = ∞. Sim-
ilarly, the inverse transformation has a simple pole at ac so that f (∞) = ac .
When c = 0, we have f (∞) = ∞.
Consider the derivative of f in the form
ad − bc
f  (z) = , (8.2.4)
(cz + d)2
which is well defined everywhere except at the pole − dc . Also, f  (z) never
assumes the zero value in the finite complex plane, provided that ad − bc =
0. Hence, a bilinear transformation is conformal at every point in the finite
complex plane except at its pole.

Triples to triples
Apparently, we have four coefficients in the bilinear transformation, but only
three of them are independent. There exists a unique bilinear transformation
 8.2 Bilinear transformations 377

that maps three distinct points z1 , z2 , z3 in the z-plane onto three distinct points
w1 , w2 , w3 in the w-plane. First, we assume that the six points are all finite.
Since zj is mapped to wj , j = 1, 2, 3, it follows that
azj + b
wj = , j = 1, 2, 3.
czj + d
Using the relations
(ad − bc)(z − zj )
w − wj = , j = 1, 2,
(cz + d)(czj + d)
and
(ad − bc)(z3 − zj )
w3 − wj = , j = 1, 2,
(cz3 + d)(czj + d)
we obtain the following formula for the required bilinear transformation:
$ $
w − w1 w3 − w1 z − z1 z3 − z1
= . (8.2.5)
w − w 2 w 3 − w2 z − z 2 z 3 − z2
What happens when some of these points are not finite? For example, when
z1 → ∞, the right-hand side of eq. (8.2.5) is then replaced by
$
z − z1 z3 − z1 z 3 − z2
lim = .
z1 →∞ z − z2 z3 − z2 z − z2
This technique can be applied to other limiting cases, like z2 → ∞, w1 → ∞,
etc., to find the corresponding reduced form of the bilinear transformation
formula.

Example 8.2.1 Find the bilinear transformation that carries the points
−1, ∞, i onto the points
(a) i, 1, 1 + i; (b) ∞, i, 1.

Solution Write z1 = −1, z2 = ∞ and z3 = i.

(a) Here, w1 = i, w2 = 1 and w3 = 1 + i. Taking the limit z2 → ∞,

eq. (8.2.5) becomes
$
w−i 1+i−i z+1
= .
w−1 1+i−1 i+1
Rearranging the terms, we obtain
z+2+i
w= .
z+2−i
378 Conformal Mappings and Applications

(b) Now, w1 = ∞, w2 = i and w3 = 1. Taking the limits z2 → ∞ and

w1 → ∞ in eq. (8.2.5), we obtain
1−i z+1 iz + 2 + i
= or w= .
w−i i+1 z+1

8.2.1 Circle-preserving property

The decomposition of f (z) in eq. (8.2.2) reveals that a bilinear transformation
can be expressed as a composition of three successive transformations, namely,
(i) linear transformation:
w1 = cz + d; (8.2.6a)
(ii) inversion:
1
w2 = ; (8.2.6b)
w1
(iii) linear transformation:
a bc − ad
w = f (z) = + w2 . (8.2.6c)
c c
Expressing c in polar form, where c = reiθ , the linear transformation defined
in eq. (8.2.6a) can be decomposed into (i) rotation: z1 = eiθ z; (ii) magnifica-
tion: z2 = rz1 ; and (iii) translation: w1 = z2 + d. Combining these transfor-
mations, a bilinear transformation can then be visualized as the composition of
a sequence of translation, rotation, magnification and inversion.
The four types of transformation – translation, rotation, magnification and
inversion – all share the circle-preserving property. They map the class of circles
and lines to the same class. Here, we treat a straight line as a circle with infinite
radius. It is quite obvious that translation, rotation and magnification do preserve
circles. A few steps are required to show that the inversion transformation also
shares the circle-preserving property. Note that the general form of a circle in
the complex plane can be expressed as
Azz + Ez + Ez + D = 0, A and D are real, (8.2.7)
where A = 0 and |E|2 − AD > 0. When A = 0, the above equation represents
a straight line. Consider the inversion transformation w = 1z . Substitute z = w1
into eq. (8.2.7) to obtain
Dww + Ew + Ew + A = 0. (8.2.8)
This is an equation of a circle if D = 0 or a straight line if D = 0.
 8.2 Bilinear transformations 379

Under what condition is a circle mapped to a straight line by the bilinear

transformation defined in
 eq. (8.2.1)? Since a straight line passes through the
point of infinity and f − c = ∞, we deduce that a circle or a straight line
d

that passes through the pole z = − dc of the bilinear transformation is mapped

onto a straight line.

Example 8.2.2 A bilinear transformation is defined by

2iz − 2
w = f (z) = .
2z − i
(a) Determine the invariant points of the transformation.
(b) Find the point ξ for which the equation f (z) = ξ has no solution for z
in the finite complex plane.
(c) Show that the imaginary axis is mapped onto itself.
(d) Determine the image of the disc |z| < 1.

Solution
(a) The invariant points are found by solving
2iz − 2
z = f (z) = ⇔ 2z2 − 3iz + 2 = 0.
2z − i
This gives z = − 2i and z = 2i.
(b) The point ξ is given by

ξ = lim f (z) = i.
z→∞

In other words, the equation f (z) = i is satisfied only by z = ∞.

(c) Along the imaginary axis, z = it, where t is real. The image point of
z = it is
2i (it) − 2 1+t
f (it) = = ,
2 it − i t − 12
which is purely imaginary. Hence, the imaginary axis is mapped onto
itself.
(d) We rearrange w = 2iz−2
2z−i
to obtain z = 2w−2i
iw−2
. Now,
 
 iw − 2 
|z| < 1 ⇔  < 1 ⇔ ww − 2iw + 2iw > 0 ⇔ |w − 2i| > 2.
2w − 2i 
The mapped region in the w-plane is the exterior of the circle centered
at 2i and with radius 2.
380 Conformal Mappings and Applications

Figure 8.11. The lens is mapped to the infinite wedge bounded within the lines Arg w =
3π
4
and Arg w = − 3π4 .

Example 8.2.3 Consider √ the two circles whose centers are at z = −1 and
z = 1, both with radius 2. They intersect at z = −i and z = i, and their
overlapping region is in the shape of a lens. Find the image of this lens under
the bilinear transformation
z−i
w = f (z) = .
z+i

Solution Since z = −i is a pole of the bilinear transformation, any circle that

passes through z = −i is mapped onto a straight line. Hence, the two given
circles are mapped onto two straight lines. These two straight lines also pass
through w = 0 since both circles pass through z = i and f (i) = 0.
Next, we try to find the inclination of these two mapped straight lines in
the w-plane. Let Ar (A ) be the right (left) edge of the lens corresponding
to the circle with center at z = −1 (z = 1), and L be the line segment along
the imaginary axis z = iy, −1 < y < 1 (see Figure 8.11). The image of L is
the negative real axis in the w-plane since f (iy) = y−1
y+1
< 0. Suppose we move
along the two arcs Ar and A and the vertical line L from z = i to z = −i; the
image points move along the mapped straight lines Ar , A and L , respectively,
from w = 0 to w = ∞ (see the direction arrows shown in Figure 8.11).
Since the bilinear transformation is conformal at z = i, the angle included
between A and L is preserved under the transformation. It is straightforward
to show that the line joining z = −1 and z = i is tangent to the arc A at
z = i. With regard to the directed curves shown in Figure 8.11, the arc A
 8.2 Bilinear transformations 381

is turned π4 radians in the anticlockwise sense at z = i to the line L. By the

conformal property of the bilinear transformation, the image line A should
correspondingly turn the same angle in the same sense to the image line L .
Hence, we deduce that A is the line Arg w = 3π 4
. By following the same
argument, or alternatively using the principle of symmetry, one deduces that
the arc Ar is mapped onto the line Ar : Arg w = − 3π 4
.
The images of the boundary curves of the lens are now known. The final
step is to determine whether the interior of the lens is mapped onto the infinite
wedge bounded within the lines Arg w = 3π 4
and Arg w = − 3π 4
, or to the region
exterior to the wedge. A simple rule of thumb is that the left (right) region of
a directed curve is mapped onto the left (right) region of the directed image
curve. Note that the line segment L is left of the directed curve A . By the
conformal property of the bilinear transformation, the image line segment L
should also lie to the left of the directed image curve A . By connectivity, one
can conclude that the left region is mapped onto the left region. Hence, in this
case, the lens is mapped onto the infinite wedge bounded within A and Ar (see
Figure 8.11).

8.2.2 Symmetry-preserving property

In the preceding subsection, we explored the circle-preserving property of bilin-
ear transformations. This subsection further examines the symmetry-preserving
property of bilinear transformations. First, we present the definition of sym-
metric points of a circle.

Definition 8.2.1 Given the circle C : |z − α| = R in the z-plane, two points

z1 and z2 are said to be symmetric with respect to the circle C if z1 and z2
satisfy

Arg(z1 − α) = Arg(z2 − α), (8.2.9a)

|z1 − α| |z2 − α| = R .2
(8.2.9b)

By convention, the center α of the circle C and the complex infinity ∞ are
symmetric with respect to C.

Geometrically, by virtue of relation (8.2.9a), the two symmetric points z1

and z2 and the center α of the circle are on the same ray through α, and
their distances to the center satisfy the relation (8.2.9b). Given a point z1 , its
symmetric point with respect to the circle C can be found using the same
method of construction shown in Figure 1.4 (that figure shows the special case
382 Conformal Mappings and Applications

where α = 0). When z1 is on the circle, the symmetric point is simply itself. The
two relations (8.2.9a,b) can be combined into the following single equation:
(z1 − α)(z2 − α) = R 2 . (8.2.10)

Representation of the equation of a circle in terms of symmetric points

Consider the circle C : |z − α| = R, and let z1 and z2 be a pair of symmetric
points with respect to C. We write z1 = α + d1 eiφ , where d1 = |z1 − α| and
φ = Arg (z1 − α). Since z1 , z2 and α lie on the same ray, by virtue of relation
2
(8.2.9b), we obtain z2 = α + d2 eiφ , where d2 = |z2 − α| = Rd1 . Suppose z is a
point on the circle C, and write z = α + Reiθ . We then have
     
 z − z1   Reiθ − d1 eiφ  d1  Reiθ − d1 eiφ 
 =  =  
 z − z   Reiθ − d eiφ  R  d eiθ − Reiφ .
2 2 1

By symmetry, the two moduli |Reiθ − d1 eiφ | and |d1 eiθ − Reiθ | are equal so
that
 
 z − z1  d1 R
 
 z − z = R = d .
2 2

This gives an alternative representation of the equation of the circle C in terms

of symmetric points z1 and z2 .
Conversely, given an equation of the form
 
 z − z1 
 
 z − z  = k, z1 = z2 and k is real and non-negative, (8.2.11)
2

it can be shown that this equation represents a circle with z1 and z2 as a pair of
symmetric points. We wish to find the center and radius of the circle in terms
of z1 , z2 and k. Using the relation |z − z1 | = k|z − z2 |, it is straightforward to
show that
|(z − z1 ) − k 2 (z − z2 )| = k|(z − z1 ) − (z − z2 )|.
Upon rearranging, we obtain
 
 2 
z − z1 − k z2 = k|z1 − z2 | .
 1 − k2  |1 − k 2 |
The above equation is in the form |z − α| = R, where
z1 − k 2 z2 k|z1 − z2 |
α= and R= . (8.2.12)
1 − k2 |1 − k 2 |
Further, since we have
k2 1
z1 − α = (z2 − z1 ) and z2 − α = (z2 − z1 ),
1 − k2 1 − k2
 8.2 Bilinear transformations 383

it follows that z1 and z2 satisfy the relation

R2
z2 − α = . (8.2.13)
z1 − α
Therefore, the two points z1 and z2 are symmetric with respect to the circle
whose center and radius are given in eq. (8.2.12).
Consider the two special cases k = 0 and k = 1. When k = 0, the circle
reduces to a point since the radius becomes zero. When k = 1, the radius
becomes infinite and so the circle becomes a straight line. In fact, the equation
|z − z1 | = |z − z2 | represents the perpendicular bisector of the line joining z1
and z2 . The two symmetric points are mirror images of each other with the
perpendicular bisector as the mirror.

Apollonius’ family of circles

The above result shows that any circle can be represented in the form
 
 z − z1 
 
 z − z  = k, k > 0, (8.2.14)
2
with z1 and z2 as its symmetric points. When k = 1, the center and radius
of the circle are given by eq. (8.2.12). The family of circles with the same
pair of symmetric points z1 and z2 is called the Apollonius family of circles.
Alternatively, given any two circles C1 and C2 , we can find the pair of points
z1 and z2 that are symmetric with respect to both circles. As a result, C1 and C2
belong to the same Apollonius family. Another interesting property that can be
deduced immediately is that the bilinear transformation
z − z1
w= (8.2.15)
z − z2
maps the Apollonius family of circles defined by eq. (8.2.14) onto the family
of concentric circles |w| = k. These properties can be used together to find a
bilinear transformation that maps a given pair of circles onto two concentric
circles. The details of the technique are shown in the following two examples.

Example 8.2.4 Find the bilinear transformation that maps the region bounded
within the circles |z| = 1 and |z − 1| = 52 onto the annular region 1 < |w| < r,
for some r to be determined, and the unit circle |z| = 1 in the z-plane onto the
unit circle |w| = 1 in the w-plane.

Solution First, we find the pair of points z1 and z2 which are symmetric points
of both circles in the z-plane. Since the centers of these two circles are on the
real axis, this pair of symmetric points z1 and z2 must also lie on the real axis.
384 Conformal Mappings and Applications

By substituting α = 0, R = 1 and α = 1, R = 5
2
successively into eq. (8.2.9b),
we deduce that z1 and z2 satisfy
 2
5
z1 z2 = 1 and (z1 − 1)(z2 − 1) = .
2
The solutions to the above simultaneous equations give z1 = − 14 and z2 = −4.
By applying eq. (8.2.11), we deduce that the two given circles in the z-plane
belong to the following Apollonius family of circles:
 
 z + 14 
 
 z + 4  = k, for some real positive number k.
The center of the unit circle |z| = 1 is α = 0. Using eq. (8.2.12), the
corresponding value of k for the above unit circle can be found by
solving
z1 − k 2 z2 − 14 + 4k 2
0=α= = ,
1 − k2 1 − k2
which gives k = 14 . Therefore, an alternative representation of the unit circle is
z+ 1
| z+44 | = 14 . Following a similar procedure, the other circle |z − 1| = 5
2
can be
z+ 14
represented as | z+4 | = 1
2
.
z+ 1
Since |z| = 1 is equivalent to | z+44 | = 14 , for any z on the unit circle |z| = 1,
we have
 
 iθ z + 14 
4e  = 1, for any real θ .
 z + 4
We deduce that the circle |z| = 1 is mapped onto the circle |w| = 1 by the
bilinear transformation
z + 14
w = 4eiθ , θ is real.
z+4
The determination of the required bilinear transformation is unique up to the
 1  the other circle |z − 1| = 2 , which has
5
multiplicative constant eiθ . Likewise,
 z+ 
the alternative representation 2  z+44  = 1, is mapped onto the circle |w| = 2 by
the above bilinear transformation. The region bounded within the two circles
|z| = 1 and |z − 1| = 52 is then mapped conformally onto the annular region
1 < |w| < 2 (see Figure 8.12).

Example 8.2.5 A circular pipe of radius a lies below the earth’s surface at a
depth h (h > a). Assuming that the surface of the buried pipe is kept at constant
 8.2 Bilinear transformations 385

z +1
w = 4 ei 4
z+4
y v

z1 u
x
z2 3 7
2 2

Figure 8.12. The region bounded within the circles |z| = 1 and |z − 1| = 52 is mapped
conformally onto the annular region 1 < |w| < 2 by the bilinear transformation w =
z+ 14
4eiθ z+4
.

y
[0]

[T0]
surface

buried
pipe × x
(h, 0) (h + a, 0)
earth’s
[0]

Figure 8.13. The isothermal lines lie below the earth’s surface (right half-plane) sur-
rounding the pipe and form a system of coaxial circles along the x-axis.

temperature T0 and the temperature of the earth’s surface is zero, find the
steady state temperature distribution below the earth’s surface surrounding the
pipe.

Solution For convenience, we assign the earth’s surface to be the vertical

imaginary axis and the center of the buried pipe to be placed at the point
(h, 0). Both the pipe and the earth’s surface can be recognized as circles in the
z-plane (see Figure 8.13). We seek a bilinear transformation that maps the pipe
and the earth’s surface onto concentric circles in the w-plane. Under the new
386 Conformal Mappings and Applications

configuration, it then becomes easier to solve for the steady state temperature
distribution in an annular region bounded within concentric circles.
Similar to the last example, we find a pair of symmetric points for the buried
pipe and the earth’s surface in the z-plane. Since the earth’s surface is a straight
line, the pair of symmetric points must be mirror images of each other with
the imaginary axis (earth’s surface) as the mirror. Also, since the center of the
pipe is on the real axis, the pair of symmetric points also lie on the real axis as
these three points must lie on the same ray. The pair of symmetric points can
be represented by z1 = −c and z2 = c, where c is some real number. Without
loss of generality, we may assume c > 0. The bilinear transformation
z+c
w = f (z) =
z−c
then maps the two circles Re z = 0 and |z − h| = a onto a pair of concentric
circles centered at w = 0 in the w-plane.
Since any point z = iy along the imaginary axis is equidistant from the points
z = −c and z = c on the real axis, we then have
 
 iy + c 
 
 iy − c  = 1.

The above relation implies that the imaginary axis Re z = 0 in the z-plane is
mapped onto the unit circle |w| = 1 in the w-plane. To determine c, we use
the property that z1 = −c and z2 = c are symmetric points of the pipe circle
|z − h| = a. Substituting α = h and R = a into eq. (8.2.9b), we obtain


(c − h)(−c − h) = a 2 , so c = h2 − a 2 > 0.

To find the image of the pipe circle in the w-plane, we first rewrite the equation
of the pipe circle into the form
 
z + c
 
 z − c  = k.

From eq. (8.2.12), k is governed by

−c − k 2 c
h=
1 − k2
so that

h+c
k= > 1.
h−c

Hence, the pipe circle is mapped onto the circle |w| = h+c
h−c
in the w-plane.
 8.2 Bilinear transformations 387

At this stage, we have found the bilinear transformation

z+c

w= , c = h2 − a 2 > 0,
z−c
which maps the region outside the buried pipe and below the  earth’s surface
in the z-plane onto the annular region 1 < |w| < k, k = h−c h+c
> 1, in the
w-plane.
To find the steady state temperature distribution within the annular region in
the w-plane, we observe that the function
T0
T (w) = Re(Log w)
ln k
is harmonic inside 1 < |w| < k. Also, T (w) becomes 0 on the circle |w| = 1
and equals T0 on the circle |w| = k. Hence, this is the solution to the present
temperature distribution problem. Referring back to the z-plane, the solution
to the temperature distribution outside the pipe and below the earth’s surface is
given by
  √
T0 z + c T0 (x + h2 − a 2 )2 + y 2
T (z) =  ln  = √ ln √ .
ln h+c z − c  ln h+√h2 −a 2 (x − h2 − a 2 )2 + y 2
h−c h− h2 −a 2

The isothermal lines are given by

√
(x + h2 − a 2 )2 + y 2
√ = µ, µ is some real constant.
(x − h2 − a 2 )2 + y 2
They lie below the earth’s surface surrounding the pipe and form a system of
coaxial circles along the x-axis (see Figure 8.13). One isothermal line T (x, y) =
T0 is the buried pipe and another isothermal line T (x, y) = 0 is the earth’s
surface. The other isothermal lines assume temperature values that lie between
0 and T0 .

Mapping of symmetric points by a bilinear transformation

Consider the bilinear transformation
az + b
w = f (z) = , (8.2.16)
cz + d
which maps the circle Cz in the z-plane onto the circle Cw in the w-plane.
Suppose z1 and z2 are a pair of symmetric points of Cz , z1 = z2 , and w1 = f (z1 )
and w2 = f (z2 ) are image points of z1 and z2 , respectively. Are w1 and w2 a
pair of symmetric points of Cw ? With the property of conformality of a bilinear
transformation in mind, the answer to the above question is “yes”. This is called
388 Conformal Mappings and Applications

the symmetry-preserving property of bilinear transformations. The validity of

this property can be shown using either the geometric or algebraic approach.
The algebraic proof is relatively straightforward. A circle Cz that has z1 and
z2 as a pair of symmetric points can be represented by
 
 z − z1 
 
 z − z  = k, k > 0. (8.2.17)
2

The inverse of the bilinear transformation (8.2.16) is found to be

−dw + b
z= . (8.2.18)
cw − a
Substituting eq. (8.2.18) into eq. (8.2.17), we obtain
 
 (cz1 + d)w − (az1 + b) 
 
 (cz + d)w − (az + b)  = k. (8.2.19)
2 2

Suppose both z1 and z2 differ from − dc (the pole of the bilinear transformation);
the above equation can be rewritten as
   
 w − w1   
  = k  cz2 + d  = k  , k  > 0,
w − w   cz + d 
2 1

indicating that w1 and w2 are a pair of symmetric points  of Cw .

What happens when z1 = − dc ? We then have w1 = f − dc = ∞. By observ-
ing cz1 + d = 0, cz2 + d = 0, and az1 + b = 0, eq. (8.2.19) becomes
 
1  az1 + b 
|w − w2 | =  ,
k cz2 + d 
which reveals that Cw is a circle with w2 as the center. Recall that the center
of a circle and the complex infinity are symmetric with respect to the circle.
Hence, w1 and w2 are symmetric with respect to Cw . A similar argument can
be applied when z2 = − dc .
The alternative proof using the geometric approach employs the conformal
property of bilinear transformations. First, we establish a basic geometric prop-
erty about symmetric points: any circle passing through a pair of symmetric
points z1 and z2 of Cz always cuts Cz orthogonally (see Figure 8.14).
To prove the claim, we draw a circle γ that passes through the pair of
symmetric points z1 and z2 of Cz . Through the center of Cz , a tangent is drawn
that touches γ at the point z . Let α and R denote the center and radius of Cz ,
respectively. The three points α, z1 and z2 are collinear. Here, |z − α| is the
length of the tangent to the circle γ through the external point α. We recall the
well-known result in elementary geometry that |z − α| satisfies
|z − α|2 = |z − z1 | |z − z2 |,
 8.2 Bilinear transformations 389

z2
×
z′

γ
×
z1

Cz

Figure 8.14. The circle γ passes through the symmetric points z1 and z2 of Cz . A
tangent is drawn from α (the center of Cz ) to γ that touches γ at the point z . The line
segment joining α and z can be shown to be the radius of Cz .

where the three points α, z1 and z2 all lie on the same straight line, z1 and z2 are
points on the circle γ . On the other hand, since z1 and z2 are a pair of symmetric
points of Cz , we know from eq. (8.2.9b) that |z − z1 | |z − z2 | = R 2 . We then
obtain the result |z − α| = R so that the line joining z and α is a radius of Cz ;
thus γ and Cz cut each other orthogonally at z .
Reversing the argument, one can show the converse of the above result: if
every circle passing through the two points z1 and z2 intersects Cz orthogonally,
then z1 and z2 are a pair of symmetric points of Cz .
The proof of the symmetry-preserving property follows by using arguments
that combine the above geometric property about symmetric points of a circle,
the conformal property and the circle-preserving property of bilinear trans-
formations. Let  be the image of a circle γ that passes through the pair of
symmetric points z1 and z2 of the circle Cz . Since a bilinear transformation is
circle-preserving, the image curve  is a circle that passes through w1 and w2 ,
which are the image points of z1 and z2 , respectively. The question is whether
w1 and w2 are a pair of symmetric points of Cw , the image circle of Cz under the
bilinear transformation. Since γ and Cz cut each other orthogonally, we argue
that the respective image circles  and Cw also cut each other orthogonally,
by virtue of the conformality of the bilinear transformation. If this is true for
any image circle  that passes through w1 and w2 , then w1 and w2 must be a
pair of symmetric points of Cw . Hence, we can conclude that bilinear trans-
formations are symmetry-preserving. As shown in the examples below, this
symmetry-preserving property is a versatile technique which can be employed
to find the appropriate bilinear transformation that maps one given region onto
another region.
390 Conformal Mappings and Applications

8.2.3 Some special bilinear transformations

We now examine the properties of some special bilinear transformations that
map half-planes onto half-planes, half-planes onto circles, and circles onto
circles.

Upper half-plane onto upper half-plane

The required bilinear transformation should map the real axis in the z-plane
onto the real axis in the w-plane. More precisely, the transformation maps
three distinct points x1 < x2 < x3 on the real axis onto three distinct points
w1 < w2 < w3 on the real axis. As we move in the positive direction along
the two real axes, the left-hand region (upper half z-plane) is mapped onto the
left-hand region (upper half w-plane). The angle of turning under the bilinear
transformation at xi , i = 1, 2, 3, is zero, that is,
Arg f  (xi ) = 0 so that f  (xi ) > 0, i = 1, 2, 3. (8.2.20)
The coefficients in the bilinear transformation can be determined (up to a
multiplicative constant) by the three conditions
axi + b
wi = f (xi ) = , i = 1, 2, 3.
cxi + d
Since x1 , x2 , x3 and w1 , w2 , w3 are all real, these coefficients must be real.
Further, we have
ad − bc
f  (xi ) = , i = 1, 2, 3,
(cxi + d)2
which are known to be real and positive by virtue of eq. (8.2.20). We then
deduce that the coefficients must observe the condition ad − bc > 0.
Conversely, consider a bilinear transformation
az + b
w = f (z) = ,
cz + d
where the coefficients a, b, c and d are real, and ad − bc > 0. With real coeffi-
cients, the transformation then maps the real axis onto the real axis. In addition,
we observe that
ad − bc
f  (x) = >0
(cx + d)2
for all x since ad − bc > 0, thus Arg f  (x) = 0. Hence, the positive direction
along the real axis remains the same direction along the image real axis under
the bilinear transformation. The upper half z-plane is then mapped onto the
upper half w-plane.
 8.2 Bilinear transformations 391

Remark Supposing instead the sign of ad − bc is negative, the bilinear trans-

formation with real coefficients maps the upper (lower) half-plane onto the
lower (upper) half-plane.

Example 8.2.6 Find the bilinear transformation that maps the upper half
z-plane onto the upper half w-plane, the point z = 0 to w = 0, and the point
z = i to w = 1 + i.

Solution Since the required bilinear transformation maps the real axis onto
the real axis, it should take the form
z+b
w = f (z) = ,
cz + d
where the coefficients b, c and d are real. Since f (0) = 0, we obtain b = 0.
Further, from f (i) = 1 + i, we have

i
1+i = or (d − c) + i(d + c) = i.
ci + d
Equating the real and imaginary parts, we obtain the simultaneous equations

d − c = 0 and d + c = 1.

Solving the above equations, the bilinear transformation is found to be

2z
w = f (z) = .
z+1

Upper half-plane onto unit circle

We would like to determine the general form of bilinear transformations that
map the upper half-plane Im z > 0 onto the unit circle |w| < 1.
Since Im z > 0 is mapped to |w| < 1, there exists a particular point
α, Im α > 0, that is mapped onto w = 0. The symmetric point of α with
respect to the real axis is α. By virtue of the symmetry-preserving property, the
point α will be mapped to the symmetric point of w = 0 with respect to the
unit circle |w| = 1. Since w = 0 is the center of the unit circle, its symmetric
point is w = ∞. Let the bilinear transformation be represented by

az + b
w = f (z) = .
cz + d
392 Conformal Mappings and Applications

From the deduced properties f (α) = 0 and f (α) = ∞, we obtain aα + b = 0

and cα + d = 0, thus giving α = −b/a and α = −d/c. The bilinear transfor-
mation can be rewritten as
 
a z + ab az−α
w = f (z) =   = .
c z+ cd c z−α
Recall that the real axis Im z = 0 is mapped onto the circle |w| = 1. By setting
z = x in the above equation, we obtain
    
a x − α  a  x − α 
 =  
 c x − α   c   x − α  = 1.
 
Since |x − α| = |x − α|, we obtain a/c= 1 and so a/c = eiθ for some real
θ. As a summary, the required bilinear transformation takes the general form
z−α
w = f (z) = eiθ , θ is any real value. (8.2.21)
z−α

Example 8.2.7 Find the function u(x, y) that is harmonic in the upper half-
plane and assumes the boundary value
x
u(x, 0) = 2
x +1
along the real axis.

Solution The obvious choice

z
u(z) = Re , z = x + iy,
z2 +1
fails since the function has a singularity z = i in the upper half-plane. One may
try to solve the problem by transforming the domain of the problem from the
upper half-plane to the unit disk. Suppose we choose the bilinear transformation
i−z
w=
i+z
that maps the upper half-plane Im z ≥ 0 onto the unit disk |w| ≤ 1. For any
point on |w| = 1, we write w = eiφ ; and for any point on Im z = 0, we write
z = x. Since the inverse transformation is
i(1 − w)
z= ,
(1 + w)
the quantities eiφ and x are related by
1 − eiφ
x=i .
1 + eiφ
 8.2 Bilinear transformations 393

The boundary value u(x, 0) = x 2x+1 along the real axis in the z-plane is trans-
formed into the boundary value u(eiφ ) along |w| = 1 in the w-plane. Since
x and φ are related by the above equation, the boundary value u(eiφ ) can be
expressed as

1 − eiφ
i
1 + eiφ i 1 − e2iφ sin φ
u(eiφ ) =   = = .
iφ 2 iφ
1−e 4 e 2
1+i 2
1 + eiφ
The function u(w) that is harmonic inside |w| < 1 and satisfying the boundary
condition u(eiφ ) = 12 sin φ is easily seen to be

1
u(w) = Im w.
2
Since the function
i−z
w=
i+z
is analytic in the upper z-plane, its imaginary part
1 i−z x
u(x, y) = Im = 2 , z = x + iy,
2 i+z x + (y + 1)2
would be harmonic in the upper half-plane. Also, it satisfies the prescribed
boundary condition
x
u(x, 0) = .
x2 + 1
Hence, it is the desired solution.

Remark
(i) The final solution
1 i−z
u(z) = Im
2 i+z
has its singularity at z = −i, which is outside the upper half-plane. The
true solution can be obtained only when we transfer the domain of the
problem from the upper half-plane to the unit disk.
(ii) Suppose the boundary condition along the real axis remains the same,
but the domain of the problem is changed from the upper to the lower
half-plane. The solution to this new problem can be judiciously deduced
394 Conformal Mappings and Applications

w = f (z)
y v

B
β
A
α A′ B′
x u
w= w=0

8
Figure 8.15. The circular arc eiα to eiβ on the unit circle is mapped onto the negative
real axis.

to be
1 z+i x
u(z) = Im = 2 .
2 z−i x + (1 − y)2
It can be checked easily that the boundary condition
x
u(x, 0) =
x2 + 1
is satisfied. The singularity of the above solution is now at z = i, which
is outside the lower half-plane.

Example 8.2.8 Find a bilinear transformation w = f (z) that carries the inte-
rior of the unit circle |z| < 1 onto the upper half-plane Im w > 0, such that the
circular arc eiα to eiβ (α < β) in the z-plane is mapped onto the negative real
axis in the w-plane.

Solution As deduced from eq. (8.2.21), the required bilinear transformation

takes the form
w − w0
z = F (w) = eiθ ,
w − w0
where θ and w0 (Im w0 > 0) are to be determined. To satisfy the require-
ments that the circular arc eiα to eiβ is mapped onto the negative real axis
and the interior of the unit circle is mapped onto the upper half-plane, we
impose two mapping conditions: z = eiβ is mapped to w = 0 and z = eiα is
mapped to w = ∞ (see Figure 8.15). Note that the left side of the directed
 8.2 Bilinear transformations 395

circular arc (interior of the unit circle) is mapped onto the left side of
the directed line segment along the negative real axis (upper half-plane).
Since
w − w0
eiα = F (∞) = lim eiθ = eiθ ,
w→∞ w − w0

this leads to θ = α. Further, given that

w0
eiβ = F (0) = eiα ,
w0

we obtain Arg w0 = β−α 2

. We expect w0 to lie in the upper half-plane since 0 <
Arg w0 < π , given that 0 < β − α < 2π. The modulus |w0 | is not fixed, and it
may take any real positive value. Suppose we denote |w0 | by r; correspondingly,
we write w0 = rei(β−α)/2 . Putting the results together, the required bilinear
transformation becomes

w − rei(β−α)/2 z − eiβ
z = eiα or w = rei(α−β)/2 ,
w − re−i(β−α)/2 z − eiα
where r is some undetermined real constant. The bilinear transformation leaves
one undetermined constant since the problem prescribes only two rather than
three mapping pairs.

Example 8.2.9 Find a bilinear transformation that carries the unit circle
|z| = 1 to a line segment parallel to the imaginary axis, takes the point z = 9
to the point w = 0, and leaves the circle |z| = 3 invariant.

Solution We are only given one mapping pair: z = 9 is mapped to w = 0.

The symmetry-preserving property can be used to find an additional mapping
pair. We know that the circle |z| = 3 is mapped onto the circle |w| = 3. The
symmetric point of z = 9 with respect to the circle |z| = 3 is z = 1, while
the symmetric point of w = 0 with respect to the circle |w| = 3 is w = ∞.
By virtue of the symmetry-preserving property, the point z = 1 is mapped
to w = ∞. Using the information about the two mapping pairs, the bilinear
transformation takes the form
z−9
w = f (z) = β ,
z−1
where β is some complex constant to be determined. For a point on the circle
|z| = 3, we write z = 3eiθ . Since the circle |z| = 3 is mapped onto the circle
396 Conformal Mappings and Applications

|w| = 3, we have
 iθ 
 3e − 9 
3 = |f (3eiθ )| = |β|  iθ 
3e − 1 
 
 1 − 3e−iθ 
= |β| | − 3eiθ |  = 3|β|;
1 − 3eiθ 
thus giving |β| = 1. The image of an arbitrary point z = eiθ on the unit circle
|z| = 1 is given by
 
eiθ − 9 θ
f (e ) = β iθ
iθ
= β 5 + 4i cot .
e −1 2
We know that this image point always lies on a line segment parallel to the
imaginary axis; that is,
θ
Re f (eiθ ) = Re (β(5 + 4i cot ))
2
which is a constant, independent of θ . We deduce that β must be real in order
to satisfy the above property.
To satisfy both conditions, β has to be real and |β| = 1, we then have β = 1
or −1. Hence, the two bilinear transformations
z−9 z−9
w= and w = −
z−1 z−1
both satisfy all the mapping requirements. They map the unit circle |z| = 1
onto the vertical lines Re w = 5 and Re w = −5, respectively.

Circles to circles
We wish to find a bilinear transformation that maps the unit disk |z| ≤ 1 onto the
unit disk |w| ≤ 1; in particular, the point z = α (|α| < 1) is mapped to w = 0.
By the symmetry-preserving property, the point z = 1/α (the symmetric point
of z = α with respect to the circle |z| = 1) is mapped to the complex infinity
w = ∞ (the symmetric point of w = 0 with respect to the circle |w| = 1). The
bilinear transformation is seen to assume the form
z−α
w = f (z) = β ,
1 − αz
where β is some complex constant to be determined. Let z = eiθ be a point on
the circle |z| = 1; the corresponding image point f (eiθ ) in the w-plane always
lies on the circle |w| = 1. This mapping property can be represented by
 iθ   iθ 
 e −α  
−iθ  e − α 

1 = |f (e )| = |β| 
iθ   = |β| |e |  −iθ = |β|;
1 − αe iθ  e − α
 8.2 Bilinear transformations 397

thus giving |β| = 1. Suppose we write β = eiθ0 , then the bilinear transformation
can be expressed as
z−α
w = f (z) = eiθ0 , (8.2.22)
1 − αz
where θ0 is real and |α| < 1.

Example 8.2.10 Find the bilinear transformation

  w = f (z) that carries |z| <
1 to |w| < 1, z = 13 to w = 3i , and f  3i > 0.

Solution In this problem, we need to construct a sequence of two bilinear

transformations; one maps z = 13 to ζ = 0 and the other one maps ζ = 0 to
w = 3i . First, we try tofind  transformation ζ = g(z) that carries
 the bilinear
 1
|z| < 1 onto |ζ | < 1, g 3 = 0 and g 3 > 0. From eq. (8.2.22), the bilinear
1

transformation takes the form

z − 13
ζ = g(z) = eiθ0 , θ0 is real.
1 − 3z

The derivative of g(z) at z = 1

3
is found to be
     
1 1 − 3z + 13 z − 13 
g = eiθ0    = 9 eiθ0 .
3 z 2  1 8
1− 3 z= 3
1
The condition g  3
> 0 implies eiθ0 = 1, so
3z − 1
ζ = g(z) = .
3−z
Next, we find another bilinear transformation ζ = h(w)
  that carries |w| < 1
onto |ζ | < 1, maps w = 3i to ζ = 0 and satisfies h 3i > 0. By judiciously
setting z = wi in g(z), the required bilinear transformation is deduced to be
3w − i
ζ = h(w) = .
3 + iw
Recall that the inverse of a bilinear transformation remains bilinear and the
composition of two bilinear transformations is also bilinear. Since we have

ζ = h(w) = g(z),

where both h and g are bilinear, the mapping defined by

w = h−1 (g(z))
398 Conformal Mappings and Applications

remains bilinear, and carries |z| < 1 onto |w| < 1. Further, we have
  
1 i
h−1 g = h−1 (0) = .
3 3
Observing that
dw
h (w) = g  (z),
dz
we then have
  
d −1  g  13
h (g(z)) =   i  > 0,
dz z= 13 h 3
so all the mapping requirements are satisfied. To express w explicitly in terms
of z, we solve for w from the relation
3w − i 3z − 1
=
3 + iw 3−z
and obtain
3(i − 1) + (9 − i)z
w= .
(9 + i) − 3(1 + i)z
This is the required bilinear transformation.

Riemann mapping theorem

We have seen various examples of conformal mappings between the upper half-
plane and the unit disk. The natural question is: what are the types of domain
that can be mapped conformally onto the unit disk? The Riemann mapping
theorem gives the sufficient conditions on the domain that can be mapped
conformally onto the unit disk. We state this renowned theorem without proof.

Theorem 8.2.2 (Riemann mapping theorem) Let D be any simply con-

nected domain in the complex plane other than the entire plane. There exists
a one-to-one analytic function w = f (z) that maps D in the z-plane onto the
interior of the unit disk |w| < 1 in the w-plane. The mapping can be chosen to
carry an arbitrary point of the domain and a direction through that point onto
the center of the unit disc and the direction of the positive real axis, respectively.
Under these conditions, the mapping is unique.

The Riemann mapping theorem guarantees the existence of the mapping

function. However, it does not inform us how to derive the required mapping.
We can deduce from the theorem that any simply connected domain D can be
 8.3 Schwarz–Christoffel transformations 399

Figure 8.16. The Schwarz–Christoffel transformation carries the upper half-plane onto
the interior of an n-sided polygon, where f (xk ) = wk , k = 1, 2, . . . , n. The angle of
inclination of the kth side joining the neighboring vertices wk−1 and wk is denoted by
βk . The exterior angle αk at the vertex wk is given by βk+1 − βk . For convenience of
notation, we also write wn as w0 and treat βn+1 as β1 .

analytically mapped in a one-to-one manner onto another simply connected

domain, provided that neither domain is the whole complex plane.

8.3 Schwarz–Christoffel transformations

We often encounter physical problems in potential flows, electrostatic fields
and steady state temperature distributions where the domains of interest are
polygons (or open polygons with one or more of the vertices at infinity). The
solution of these problems is greatly facilitated if the domain of the problem can
be mapped onto the upper half-plane. The first step in the solution procedure
is to construct a conformal mapping that carries the upper half-plane onto the
polygonal domain (see Figure 8.16). This class of conformal mappings has
come to be known as the Schwarz–Christoffel transformations.
Consider a closed n-sided polygon with vertices w1 , w2 , . . . , wn in the finite
w-plane. The angle of inclination of the kth side joining wk−1 and wk is denoted
by βk , k = 1, 2, . . . , n. For convenience of notation, we also write wn as w0 so
that the first side of the polygon refers to the line segment joining wn and w1 .
The angle of turning of the inclination angles of the two neighboring sides at
the vertex wk is called the exterior angle at wk . Let αk denote the exterior angle
at wk ; then αk is given by βk+1 − βk , k = 1, 2, . . . , n. Here, βn+1 is taken to be
β1 . It is known from elementary geometry that the sum of the exterior angles
of a polygon always equals 2π .
400 Conformal Mappings and Applications

We choose n distinct points x1 , x2 , . . . , xn on the real axis in the z-plane satis-

fying x1 < x2 < · · · < xn such that they are mapped to the vertices of a n-sided
polygon in the w-plane. Let w = f (z) denote the corresponding Schwarz–
Christoffel transformation to be found, where f (xk ) = wk , k = 1, 2, . . . , n.
From the relation dw
dz
= f  (z), we deduce that

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

Note that arg dz becomes zero when z moves along the real axis, so

arg dw = arg f  (z), when z is real.

Specifically, when z moves along the segment (xk−1 , xk ) on the real axis, the
image point w moves along the kth side of the polygon and arg dw assumes the
constant value βk . However, when z moves across the point xk along the real
axis, arg dw jumps by an amount αk to assume a new constant value βk+1 . In
summary, arg dw = βk when z assumes real value in (xk−1 , xk ), k = 2, 3, . . . , n
and arg dw = β1 when z assumes real value in (−∞, x1 ) or (xn , ∞).
Now, the problem of finding arg f  (z) is equivalent to the steady state temper-
ature distribution problem in the upper half-plane with discrete constant values
along the real axis [see Problem 3.25]. The solution for arg f  (z) is given by
β1 β2 z − x2 βn z − xn
arg f  (z) = arg(z − x1 ) + arg + ··· + arg
π π z − x1 π z − xn−1
βn+1
+ [π − arg(z − xn )]
π
n
αk
= β1 − arg(z − xk ), (8.3.1a)
k=1
π

where βn+1 = β1 and αk = βk+1 − βk , k = 1, 2, . . . , n. We then deduce that

6
n
f  (z) = Keiβ1 (z − xk )−αk /π , (8.3.1b)
k=1

where K is a real constant to be determined. Integrating with respect to z, the

Schwarz–Christoffel transformation formula is found to be
 z6n
f (z) = w1 + Keiβ1 (ζ − xk )−αk /π dζ,
x1 k=1

where the integration is performed along a simple path joining x1 and z. To

find the real constant K, we set z = x2 and observe that w2 = f (x2 ). Taking
 8.3 Schwarz–Christoffel transformations 401

the modulus value on both sides of the equation, we then obtain

 x2 n 
 6 
|w2 − w1 | = K  (ζ − xk ) −αk /π
dζ . (8.3.2)
x1 k=1

Remarks
(i) For convenience, xn is often chosen to be at infinity. Correspondingly,
eq. (8.3.1a) has to be modified when the last term − απn arg(z − xn )
disappears; that is,

n−1
αk
arg f  (z) = βn − arg(z − xk ). (8.3.3)
k=1
π

In this case, the mapping formula reduces to

 z n−1
6
f (z) = w1 + Keiβn (ζ − xk )−αk /π dζ. (8.3.4)
x1 k=1

(ii) Since only the angles of turning at the vertices are involved in the
mapping formula, the polygon can be an unbounded polygon with one
or more of the vertices at infinity. Since the angles of inclination of
the neighboring sides of a vertex (even at infinity) are always known,
the corresponding αk can be found as the difference of the two angles of
inclination of the neighboring sides of the polygon.

Example 8.3.1 Find the Schwarz–Christoffel transformation that maps the

upper half z-plane to the triangle with vertices w1 , w2 and w3 as shown in
Figure 8.17, where w1 = 0, w2 = 1, and θ1 and θ2 are some given angles. Take
x1 = 0, x2 = 1 and x3 = ∞ for ease of computation.

Solution The corresponding mapping pairs of points and the respective turn-
ing angles at the vertices are tabulated below:

k xk wk αk
1 0 0 π − θ1
2 1 1 π − θ2
3 ∞ w3 θ1 + θ2

Note that β3 is the angle of inclination of the side joining w2 and w3 , and its
value equals π − θ2 . As a check, the sum of the exterior angles at the vertices
402 Conformal Mappings and Applications

Figure 8.17. The remaining vertex w3 of the triangle can be determined once w1 , w2 , θ1
and θ2 are known.

equals 2π . Using eq. (8.3.4), we obtain

 z
θ1 θ2
f (z) = Ke i(π−θ2 )
ζ π −1 (ζ − 1) π −1 dζ,
0

where K is a real constant to be determined. Since f (1) = 1, the constant K is

given by
 1
θ1 θ2
1=K ζ π −1 (1 − ζ ) π −1 dζ,
0
i(π−θ2 )
where the factor e has been absorbed into the integral. The above integral
is related to the beta function defined by
 1
β(p, q) = x p−1 (1 − x)q−1 dx.
0

By setting p = θ1
π
and q = θ2
π
, it is seen that
1
K=  θ1 θ2  .
β ,
π π

Finally, the required Schwarz–Christoffel transformation is found to be

 z
1 θ1 θ2
f (z) =  θ1 θ2  ζ π −1 (1 − ζ ) π −1 dζ.
β π, π 0

Example 8.3.2 Find the Schwarz–Christoffel transformation that maps the

upper half z-plane Im z > 0 onto the semi-infinite strip −π/2 < Re w < π/2,
Im w > 0.
 8.3 Schwarz–Christoffel transformations 403

y w = f (z) v
w3 =

x u
x1=−1 x2=1 x3= w1= − π
2 w2= π2

Figure 8.18. The upper half z-plane is mapped onto the semi-infinite strip − π2 <
Re w < π2 , Im w > 0 (considered as an open triangle with one of the vertices at infinity)
in the w-plane.

Solution The semi-infinite strip can be considered as an unbounded tri-

angle with one of the vertices w3 at infinity. For convenience, we assign
x1 = −1, x2 = 1 and x3 = ∞. The corresponding mapping pairs of points
and the respective angles of turning at the vertices are summarized below (also
see Figure 8.18):

k xk wk αk
1 −1 − π2 π
2
π π
2 1 2 2

3 ∞ ∞ π

Note that β3 = π2 since the side of the unbounded triangle joining w2 and w3
is the vertical line Re w = π2 . Using the Schwarz–Christoffel formula (8.3.4),
we have
 z
π π 1
f (z) = − + Kei 2 dζ
−1 (ζ + 1) (ζ − 1)1/2
2 1/2
 z
π 1
=− +K
dζ
2 −1 1 − ζ2
π
= K[sin−1 z − sin−1 (−1)] − .
2
Applying the condition f (1) = π
2
, K is found to be 1. The required mapping
is then given by

f (z) = sin−1 z.
404 Conformal Mappings and Applications

Figure 8.19. The bottom edges of the potential flow field are along the horizontal lines
v = H for u < 0 and v = 0 for u > 0. The vertical step of height H is placed along
u = 0 and 0 < v < H .

The result agrees with the mapping properties of w = sin z discussed in Sub-
section 3.2.1.

Example 8.3.3 Consider the potential flow over a vertical step of height H ,
the configuration of which is shown in Figure 8.19. The flow upstream far
from the step is uniform with constant speed U and parallel to the floor bottom.
Find the speed along the bottom edges of the flow field.

Solution First, we construct the Schwarz–Christoffel transformation w =

f (z) that maps the upper half z-plane onto the domain of the given step flow
(shown in Figure 8.19) in the w-plane. The flow domain can be visualized as
an unbounded polygon with true vertices at (0, 0) and (0, H ), and a virtual
vertex at infinity. We assign x1 = −1, x2 = 1 and x3 = ∞; correspondingly,
w1 = iH, w2 = 0 and w3 = ∞. The angles of turning at w1 and w2 are − π2
and π2 , respectively, and β3 = 0. Using eq. (8.3.1b), we obtain
f  (z) = K(z + 1)1/2 (z − 1)−1/2 ;
and the constant K is determined by [see eq. (8.3.2)]
H H H
K= = 1 = .
- 1     π
 (ζ + 1)1/2 (ζ − 1)−1/2 dζ   −1  
 −1  (z − 1) + cosh z
2 1/2

−1

The required transformation is then found to be

H  2 
w = f (z) = (z − 1)1/2 + cosh−1 z .
π
 8.3 Schwarz–Christoffel transformations 405

Let F (w) and V (w) denote the complex potential and complex velocity of
the potential flow field in the w-plane, respectively, where V (w) = F  (w). The
complex potential of the corresponding flow field in the z-plane is given by
F (w(z)), where w(z) is the above Schwarz–Christoffel transformation. By the
chain rule of differentiation, the two complex velocities dFdw(w) and dFdz(z) are
related by
dF (w) dF (z) dz
= .
dw dz dw
When w → ∞, dFdw(w) tends to U ; and when z → ∞, dw dz
tends to H
π
. Using the
above relation between the complex velocities, we deduce that
dF (z) H
lim = U.
z→∞ dz π
Since the flow domain in the z-plane is the whole upper half-plane, the flow
velocity is uniform throughout. In other words, the complex velocity dF dz
of the
flow field in the z-plane is equal to the constant value Hπ U . Also, we see that
 
dz 1 π z − 1 1/2
=  = ,
dw f (z) H z+1
so the complex velocity of the flow field in the w-plane is given by
 
dF (w) z − 1 1/2
V (w) = =U .
dw z+1
The points along the bottom edges and the vertical step of the flow field in the
w-plane are the image points of z = x along the real axis in the z-plane. In
terms of x, the speed of the step flow along the bottom edges is given by
 
x − 1
|V | = U   .
x + 1
The speed becomes infinite at x = x1 = −1 and zero at x = x2 = 1. These
two points correspond to the upper and lower corners of the vertical step,
respectively.

Example 8.3.4 Consider the steady state temperature distribution within a

thin wall near the corner of a building. When the thickness a of the wall is small
compared to its span, the wall can be visualized as having two perpendicular
semi-infinite channels (see the configuration shown in Figure 8.20). Let the
outer surface be maintained at zero temperature, while the inner surface is kept
at constant temperature T0 . Find the steady state temperature distribution inside
the wall.
406 Conformal Mappings and Applications

Figure 8.20. The thin wall at the corner of a building is visualized as having semi-
infinite channels along both the positive u-axis and v-axis. The domain of interest can
be recognized as an unbounded polygon with vertices at (0, 0) and (a, a) and virtual
vertices at the two far ends of the channels.

Solution We recognize the semi-infinite L-shaped thin wall as an unbounded

polygon and attempt to construct a Schwarz–Christoffel transformation
w = f (z) that carries the upper half z-plane onto the unbounded L-shaped
polygon in the w-plane. The two true vertices of the unbounded polygon are
w1 = a(1 + i) and w3 = 0, while the two virtual vertices w2 and w4 are at
the far ends of the channels (see Figure 8.20). For convenience, we assign
x1 = −1, x2 = 0, x3 = 1 and x4 = ∞, and they are mapped to the vertices
w1 , w2 , w3 and w4 , respectively.
The angles of turning at w1 , w2 and w3 are − π2 , π and π2 , respectively, and
β4 = 0. By eq. (8.3.4), the corresponding Schwarz–Christoffel transformation
is given by
 z
w = f (z) = a(1 + i) + K (ζ + 1)1/2 ζ −1 (ζ − 1)−1/2 dζ.
−1

The appropriate integration variable is

 1/2
ζ +1 η2 + 1
η= or ζ =
ζ −1 η2 − 1

so that

4η
dζ = − dη.
(η2 − 1)2
 8.3 Schwarz–Christoffel transformations 407

Upon integration, we obtain

 ξ
4η2
w = a(1 + i) + K dη
0 1−η
4
   1/2
1+ξ −1 z+1
= a(1 + i) + K Log − 2 tan ξ , ξ= ,
1−ξ z−1
where the principal branches for Log and tan−1 are chosen so that the corre-
sponding values tend to zero as ξ → 0. To determine K, we apply the condition
w3 = f (1) = 0. This gives
'  π (
0 = a(1 + i) + K −iπ − 2 ,
2
and leads to K = a
.
The required mapping function is then found to be
π
   
2a 1 1+ξ z + 1 1/2
w = f (z) = a(1 + i) + Log − tan−1 ξ , ξ = .
π 2 1−ξ z−1
In the z-plane, the boundary temperature values along the x-axis are given by

T0 x < 0
T (x, 0) = .
0 x>0
The steady state temperature function T (z) is seen to be
T0
T (z) = Arg z, Im z > 0.
π
Referring back to the w-plane, the temperature function T (w) can be expressed
as
T0
T (w) = Arg f −1 (w),
π
where f −1 (w) is the inverse of the above Schwarz–Christoffel mapping func-
tion.

Example 8.3.5 Suppose two semi-infinite charged rods with electric potential
φ1 and φ2 are placed in the configuration shown in Figure 8.21. Find the electric
potential of the electrostatic field induced by the two charged rods.

Solution It does require a certain amount of imagination to visualize the

whole u-v plane minus the two charged rods as an unbounded polygon. The
two true vertices are at w1 = ih and w3 = h, and the two virtual vertices w2
and w4 are at the far bottom left and far top right corners of the w-plane,
respectively (see Figure 8.21).
408 Conformal Mappings and Applications

Figure 8.21. The horizontal semi-infinite rod with electric potential φ1 is placed along
the positive u-axis with the free end at (h, 0) while the vertical semi-infinite rod with
electric potential φ2 is placed along the positive v-axis with the free end at (0, h).

We assign x1 = −1, x2 = 0, x3 = 1 and x4 = ∞ along the real axis in the

z-plane, and these points are mapped by a Schwarz–Christoffel transformation
w = f (z) to w1 , w2 , w3 and w4 in the w-plane, respectively. The angles of
turning at the vertices w1 and w3 are easily seen to be both equal to −π .
Interestingly, the angle of turning at the virtual vertex w2 is 5π 2
since the
orientation starts northward, then turns one complete revolution plus a final
right angle turn to settle in the westward direction. The angle of inclination β4
of the virtual line segment joining w3 and w4 is seen to be zero.
By eq. (8.3.4), the corresponding Schwarz–Christoffel transformation is
given by
 z
(ζ − 1)(ζ + 1)
w = f (z) = ih + K dζ
−1 ζ 5/2
 
2 8
= ih + K 2z1/2 + z−3/2 − i .
3 3

The constant K is determined by f (1) = h, and this gives

 
2 8 3
h = ih + K 2 + − i or K = h.
3 3 8

The required Schwarz–Christoffel transformation is found to be

 
h 1/2 1
w = f (z) = z 3+ 2 .
4 z
 8.4 Problems 409

In the z-plane, the boundary values for the electric potential φ(z) along the
x-axis are given by

φ2 x < 0
φ(x, 0) = .
φ1 x > 0
The solution for φ(z) is easily found to be
φ 2 − φ1
φ(z) = φ1 + Arg z.
π
Using the relation between z and w as defined by the above Schwarz–Christoffel
mapping function, the electric potential φ(w) at any arbitrary point in the w-
plane can be obtained.

8.4 Problems
8.1. For each of the following functions, find the points at which the function
is not conformal:
1
(a) w = z2 + 2 ; (b) w = sin 2z.
z
8.2. A mapping that preserves the magnitude of the angle between any two
smooth arcs passing through any point in a domain but not the sense is
called an isogonal mapping in that domain. Suppose f (z) is conformal
in a domain D; show that f (z) is isogonal in the same domain D.
8.3. For each of the following curves in the z-plane, find the corresponding
image curve under the mapping w = 1z in the w-plane:

(a) a line through the point z = α = 0;

(b) the hyperbola x 2 − y 2 = 1;
(c) the parabola y 2 − 2px = 0.

8.4. Show that the transformation w = 1z maps the common part of the two
disks |z − 1| < 1 and |z + i| < 1 onto the quarter-plane Re w > 12 and
Im w > 12 .
8.5. Find a transformation that maps the cycloid as defined by

x = a(t − sin t), y = a(1 − cos t), 0 ≤ t ≤ 2π,

onto a line through the origin with slope m.

8.6. Find the curvature of the image contour of the unit circle |z| = 1 under
the conformal mapping w = f (z).
8.7. Show that the function w = 2 Log z − z2 maps the upper half z-plane
conformally onto the w-plane minus the two lines v = 2π, u < −1 and
v = 0, u < −1.
410 Conformal Mappings and Applications

8.8. Find the area of the *closed region in the w-plane which
+ is the image
of the closed region z : 1 ≤ |z| ≤ 2, − π4 ≤ Arg z ≤ π4 in the z-plane
under the mapping w = f (z) = z2 .
8.9. Find a conformal mapping w = f (z) that maps the region outside the
hyperbola

x2 y2
− > 4,
cos2 α sin2 α
where 0 < α < π2 , in the z-plane onto the upper half w-plane.
8.10. Show that the mapping function


z
1 − ζ4
w= dζ, z0 = 0,
z0 ζ2

maps the disk |z| < 1 in the z-plane onto the exterior of a square in the
w-plane.
8.11. Find the image in the w-plane of the domain − π2 < x < π2 , y > 0 under
the mapping w = (sin z)1/4 .
8.12. Find the steady state temperature inside the sectoral domain 0 < Arg z <
π
4
, where the ray Arg z = 0 is maintained at constant temperature K and
the ray Arg = π4 is at constant temperature K2 .
8.13. Find the steady state temperature T (x, y) inside the first quadrant where
the temperature values along the x-axis satisfy

 T (x, 0) = 1, x>1
∂T ,
 (x, 0) = 0, x < 1
∂y

and the temperature values along the y-axis satisfy


 T (0, y) = −1, y > 1
∂T .
 (0, y) = 0, y < 1
∂x

8.14. Find the steady state temperature T (x, y) inside the first quadrant where
the temperature along the y-axis is maintained at T1 . The boundary
temperature values along the x-axis satisfy

 T (x, 0) = T2 , x > x0
∂T .
 (x, 0) = 0, x < x0
∂y
 8.4 Problems 411

8.15. Find the steady state temperature T (x, y) inside the infinite slab 0 < y <
1, where the boundary temperature values are given by

T (x, 0) = T0 , x < 0; T (x, 1) = −T0 , x < 0;

∂T ∂T
(x, 0) = (x, 1) = 0, x > 0.
∂y ∂y
8.16. Find the complex potential of the uniform flow U∞ parallel to the x-axis
that streams past the elliptical cylinder
x2 y2
+ = 1.
a2 b2
√
z+ z2 −c2
Hint: Consider the Joukowski mapping ζ = 2
, c2 = a 2 − b2 .

8.17. Find the complex potential of the flow field with uniform upstream flow
U∞ > 0 that streams past an infinite obstacle of parabolic shape defined
by

y 2 = 2px, p > 0.

U >0 2
y = 2 px, p > 0
x

Hint: Let D denote the domain of the flow field

 that is external to
the parabolic obstacle. The mapping ζ = z − p2 takes
 D onto
the first quadrant above the line segment ζ = p2 i. The
subsequent mapping
  2
p

w = g(ζ (z)) = ζ − i = z − p + p2 − 2pz
2
takes D onto the upper half w-plane. Check that g  (∞) = 1.
412 Conformal Mappings and Applications

8.18. A potential flow field is confined between the parallel lines y = ±1,
with the source and sink of equal strength m placed at z = −1 and
z = 1, respectively. Find the complex potential of the flow field.
8.19. Consider the seepage problem discussed in Subsection 8.1.2. Show that
the total amount of seepage from upstream to downstream of the dam is
given by
 ∞
φ 1 − φ2 x
cosh−1 dx.
L π L

Hint: The flux across the line segment joining x = L to x = x0

(x0 > L) is given by
 x0
ψ(x, 0) dx.
L

8.20. This problem investigates the potential flow field of a fluid jet coming
through a slot of width 2a in a two-dimensional plane. We wish to find
the shape of the jet, given that the amount of fluid flowing through the
slot per unit time is Q. The coordinate axes are assigned so that the slot
lies along the x-axis, between −a and a, and the y-axis is the axis of
symmetry of the jet flow (see the figure). The fluid flows from the upper
half-plane to the lower half-plane through the slot.

Let AB define the free boundary of the jet flow on the right-hand
plane and BC be the wall along the x-axis. The point B corresponds
to x = a and the point A is on the free jet boundary far from the slot.
The free boundary of the fluid jet is not known in advance but has to be
determined as part of the solution procedure. Fortunately, the velocity
 8.4 Problems 413

values along the axis of symmetry, the bounding wall and the free jet
boundary can be readily deduced.

(a) Let U∞ denote the uniform flow speed of the free jet at infinitely far
distance from the slot. Explain why
Q
U∞ = ,
2b
where 2b is the width of the jet at infinity (b will be determined later).
(b) Let u and v denote the x-component and y-component of the fluid
velocity, respectively. Show that
(i) u = 0, −U∞ ≤ v ≤ 0 for x = 0 (along the axis of symme-
try);
(ii) u2 + v 2 = U∞2
along the free jet boundary;
(iii) −U∞ ≤ u ≤ 0, v = 0 for a ≤ x < ∞, y = 0 (along the
bounding wall).
(c) Let w = −u + iv, and treat w as a mapping. Show that the
flow field in the right half z-plane, z = x + iy, is mapped onto
the quarter circle Cw = {w : |w| ≤ U∞ , − π2 ≤ Arg w ≤ 0} in the
lower right quadrant in the w-plane. Check that the free boundary
is mapped onto the circular boundary of the quarter circle in the
w-plane.
(d) Show that the mapping



w = U∞ ( ζ − ζ − 1)

carries the upper half ζ -plane onto the quarter circle Cw in the
w-plane. Check that the negative real axis is mapped to the vertical
segment of Cw , the line segment (0, 1) is mapped onto the circular
arc of Cw and the line segment (1, ∞) is mapped onto the horizontal
segment of Cw .

Hint: Consider the Joukowski mapping discussed in Subsec-

tion 3.6.1.

(e) By relating the mappings together, we see that the streamline along
ABC (the free boundary of the jet plus the bounding wall along
the positive x-axis) in the z-plane is mapped onto the positive
real axis in the ζ -plane, and the streamline along the y-axis (the
axis of symmetry) in the z-plane is mapped onto the negative
real axis in the ζ -plane. Without loss of generality, let the stream
function value on the streamline along the y-axis assume zero
414 Conformal Mappings and Applications

value. Explain why the stream function value on the streamline

along ABC assumes the value Q2 . Show that the complex potential
f (z) = φ(x, y) + iψ(x, y), z = x + iy, is given by
Qi Q
f (z) = − Log ζ.
2 2π

Hint: Note that when ζ moves along the positive part and negative
part of the real axis, Im f = ψ equals Q2 and zero,
respectively.

(f) Using the following relations

df (z)


= u − iv = −w and w = U∞ ( ζ − ζ − 1),
dz
show that
Q '


(
z= ζ + ζ − 1 + tan−1 ζ − 1 − 1 + a,
π U∞
where the branch of the inverse tangent is chosen such that
√
tan−1 ζ − 1 becomes zero as ζ → 1.

Hint: The point z = a is mapped to the point ζ = 1.

(g) Verify that the parametric representation of the free boundary of the
jet is given by
Q √
x= ( t − 1) + a,
π U∞
 √ 
Q √ 1 1+ 1+t
y= 1 − t − ln √ , 0 ≤ t ≤ 1.
π U∞ 2 1− 1−t

Deduce that the width of the jet at far distance from the slot is given by
2a
2b = .
1 + π2
8.21. Given the straight line whose foot of perpendicular from the origin is
ξ = 0, show that its image under the inversion transformation
1
w = f (z) =
z
is a circle whose center is at w = 1
2ξ
and which passes through the origin.
 8.4 Problems 415

8.22. Show that the composite of two bilinear transformations remains bilinear.
8.23. In each of the following cases, find the corresponding bilinear
transformation that maps z1 , z2 and z3 to w1 , w2 and w3 , respectively:
(a) z1 = 1, z2 = i, z3 = −1; w1 = 1, w2 = 0, w3 = i.
(b) z1 = ∞, z2 = 0, z3 = 1; w1 = 2, w2 = ∞, w3 = 0.
(c) z1 = −1, z2 = i, z3 = 1 + i; w1 = i, w2 = ∞, w3 = 1.
8.24. Find the bilinear transformation that has two invariant points 1 and −1,
and maps z = eiπ/3 to w = e2iπ/3 .
8.25. Find a mapping function that maps the upper half of the unit circle
Im z > 0, |z| < 1 in the z-plane onto the interior of the unit circle
|w| < 1 in the w-plane.
8.26. Find a bilinear transformation that maps the circle |z| < 1 in the z-plane
onto the circle |w − 1| < 1 in the w-plane, and takes the points z = −1
and z = 1 to w = 2 and w = 0, respectively.
8.27. Find the conditions for the coefficients such that a bilinear mapping of
the form w = az+bcz+d
maps the unit circle |z| = 1 in the z-plane onto a
straight line in the w-plane.
8.28. In each of the following cases, find a transformation that carries
(a) the circle |z| = 1 onto the line Re (1 + i)w = 0;
(b) the circle |z − z0 | = r onto the circle |w| = 1.
8.29. Given the circle C : |z − (1 + i)| = 4, find the inversion point of
z1 = 2(1 + i) with respect to C. Express the equation of the circle in
terms of z1 and its inversion point.
8.30. Show that the necessary and sufficient condition for the two points z1
and z2 to be a pair of symmetric points of the circle
Azz + Bz + Bz + D = 0
416 Conformal Mappings and Applications

is given by
Az1 z2 + Bz2 + Bz1 + D = 0.
8.31. Find a bilinear transformation that carries the region between the two
circles |z − 3| = 9 and |z − 8| = 16 in the z-plane onto the annular
region ρ < |w| < 1 in the w-plane. Also, find the value of ρ.
8.32. Find the bilinear transformation w = f (z) that carries the upper half
z-plane onto |w − w0 | < R, and satisfies f (i) = w0 and f  (i) > 0.
8.33. Find the bilinear transformation w = f (z) that carries |z| < 1 onto
|w| < 1, and satisfies f (α) = α, Arg f  (α) = α, |α| < 1 and α is real.
8.34. Find the bilinear transformation w = f (z) that takes
 |z| ≤ 1 onto
|w| ≤ 1, where z = 12 is mapped to w = 0 and f  12 > 0.
8.35. Find the Schwarz–Christoffel transformation that maps the upper
half z-plane conformally onto the interior of the rectangle with
vertices w1 = −1 + i, w2 = −1, w3 = 1 and w4 = 1 + i. The
preimages of w1 , w2 , w3 and w4 along the z-axis are, respectively,
x1 = − 1k , x2 = −1, x3 = 1 and x4 = k1 , 0 < k < 1.
8.36. Find the Schwarz–Christoffel transformation that carries Im z > 0
onto the upper half w-plane minus the semi-infinite strip
{w : Re w ≥ 0 and 0 ≤ Im w ≤ H } (see the figure). Here, we choose
x1 = 0, x2 = 1, x3 = ∞ and w1 = 0, w2 = iH, w3 = ∞.
y w = f (z) v
z-plane w-plane

w2 = iH

w3=

x =0
. .
x2 = 1 x 3=
x
w1 = 0
u
1

8.37. Consider the domain defined by

{w: Im w > 0}\{w: Re w = a, 0 ≤ Im w ≤ h},
which is the upper half w-plane minus a vertical line segment (see the
figure). By considering the above domain as a degenerate quadrilateral
with vertices w1 = a, w2 = a + ih, w3 = a and w4 = ∞, find the
Schwarz–Christoffel transformation w = f (z) that carries the upper half
z-plane onto the degenerate quadrilateral. Here, we choose x1 = −1,
x2 = 0, x3 = 1 and x4 = ∞.
 8.4 Problems 417

w = f ( z)
y v
z-plane w-plane

w2= a + ih

w4 =

. . .
x 1= 1 x2= 0 x3 = 1 x4 =
x
w1 = w3 = a
u

8.38. Find the transformation w = f (z) that maps the infinite strip
−π < Im z < π in the z-plane onto the polygonal domain in the w-plane
bounded within the lines v = ±h, u < 0 and Arg(w − ±ih) = ±απ ,
u > 0, where 0 < α ≤ 1 (see the figure).

w = f (z)
v
y
i ih

x u

ih
i

8.39. Find the function T (w), w = u + iv, that is harmonic in the semi-
infinite strip {w : u > 0, 0 < v < H }, subject to the following Dirichlet
boundary conditions (see the figure):
(i) T (iv) = K, 0 < v < H ;
(ii) T (u) = T (u + iH ) = 0, u > 0.

Hint: Find the Schwarz–Christoffel transformation that carries

the upper half z-plane onto the above semi-infinite strip (con-
sidered as a degenerate triangle) in the w-plane. For convenience,
we choose x1 = −1, x2 = 1, x3 = ∞ and w1 = iH, w2 = 0,
w3 = ∞. The solution of the Dirichlet problem in the upper
half z-plane is given by
K z−1
Arg .
π z+1
418 Conformal Mappings and Applications

[0]

[K]

u
[0]

8.40. Find the complex potential f (z) of the electrostatic field produced by
two parallel semi-infinite charged rods that are a distance 2a apart (see
the configuration in the figure). We choose a coordinate system so that
the rods lie along the positive and negative x-axis, with end points at
z = a and z = −a, respectively. The electric potentials of the left rod
and right rod are −V and V , respectively.
y

[ V] [V ]
x
z= a z=a

8.41. Find the Schwarz–Christoffel transformation that carries Im z > 0

conformally onto the domain {w : −h1 < Im w < h2 , h1 > 0 and
h2 > 0}\{w : Re w < 0 and Im w = 0} (see the figure).

v
ih2

ih1
 Answers to Problems

Chapter 1
1.1. (a) −11 − 2i; (b) i; (c) 2(n+2)/2 cos nπ 4
;
3 4 3 i
(d) − − i; (e) − .
5 5 2 2
√
1.2. (a) 5 2; (b) 1; (c) 2.
 
1 x −y
1.8. .
x2 + y2 y x
 1/2
n 
n  n
1.14. r =  rj2 + 2 rk r cos(θk − θ ) , k =  in the summation;
j =1 k=1 =1
 
n $
n
θ = tan−1  rj sin θj rj cos θj .
j =1 j =1
1.15. When −π < θ < 0, the real and imaginary parts of z1/2 have opposite
signs. In this case
  
√ 1 + cos θ 1 − cos θ
z =± r
1/2
−i .
2 2

The two square roots of 3 − 4i are ±(2 − i).

1.20. Equality holds when (i) z1 = z2 = · · · = zn = 0; (ii) for nonzero
z1 , zk /z1 are all real positive, valid for k = 2, . . . , n.
z1 − z3
1.22. Im = 0.
z 2 − z3
1.23. Since |z1 | = |z2 | = |z3 | = 1, we have z2 = eiα z1 and z3 = eiβ z1 .
Together with z1 + z2 + z3 = 0, we obtain eiα + eiβ + 1 = 0. One then
deduces that α = −β and cos α = − 12 .

419
420 Answers to Problems

1.24. As an intermediate step, show that

   
   
  
|z − 1| ≤ z − |z| + |z| − 1
   
 θ   
= |z|2 sin  + |z| − 1,
 where θ = Arg z.
2
θ1 − θ2
1.26. p = 2 sin , where Arg z1 = θ1 and Arg z2 = θ2 .
2

1.27. |z|max = |β|, |z|min = |β|2 − |α|2 .



a 2 + 4|b| + a a 2 + 4|b| − a
1.29. |z|max = , |z|min = .
2 2
1.30. (a) The region is on or within the two branches of the hyperbola
x 2 − y 2 = 1.
(b) The region is a wedge with its vertex at the origin and the angle
subtended equal to 2π 3
; the axis of symmetry lies on the real
axis.
(c) The region is outside the circle |z| = 13 .
(d) The region lies on the left side of the right branch of the hyperbola
12x 2 − 4y 2 = 3.
(e) The region is outside the circle x 2 + (y − 1)2 = 2 and lying in the
upper half-plane.
1.31. (a) The family of circles where the ratio of the distances of any point on
these circles to z1 and z2 is equal to λ.
(b) The family of circular arcs with end points at z1 and z2 , including
also the straight line segment joining z1 and z2 .
n
1.36. 0, .
ω−1
1.39. b = 0, d − adb + cb2 = 0 or b = d = 0, a 2 − 4c ≥ 0.
2

1.41. First, establish the relation

1 + cos(Arg w2 − Arg w)
1 + cos(Arg w1 − Arg w)
|w| + |w2 | − |w1 | cos(Arg w1 − Arg w2 )
= ,
|w| + |w2 | cos(Arg w1 − Arg w2 ) − |w1 |
then apply the cosine rule
|w1 |2 + |w2 |2 − |w|2
cos(Arg w1 − Arg w2 ) =
2|w1 ||w2 |
to obtain
1 + cos(Arg w2 − Arg w) |w1 | + |w2 | + |w| |w1 |
= .
1 + cos(Arg w1 − Arg w) |w1 | + |w2 | − |w| |w2 |
 Answers to Problems 421

1.42. (a) no limit point; (b) 0;

1+i −1 + i −1 − i 1−i
(c) √ , i, √ , −1, √ , −i, √ , 1;
2 2 2 2
(d) same as (c).
1.43. Both are closed sets.
1.44. (a) Interior points: {z = x + iy : −3 < y < 0},
exterior points: {z = x + iy : y < −3 or y > 0},
boundary points: {z = x + iy : y = −3 or y = 0},
limit points: {z = x + iy : −3 ≤ y ≤ 0}.
(b) Write z = r(cos θ + i sin θ ), −π < θ ≤ π ;
interior points: {z : 0 < θ < π/4 and r > 2},
exterior points: {z : π/4 < θ < 2π} ∪ {z : r < 2},
boundary points: {z : θ = 0 and r ≥ 2} ∪
{z : θ = π/4 and r ≥ 2} ∪
{z : 0 < θ < π/4 and r = 2},
limit points: {z : 0 ≤ θ ≤ π/4 and r ≥ 2}.
1.45. (a) closed and unbounded; (b) open and unbounded;
(c) bounded, neither open nor closed.
1.46. z = 0 is a limit point.
1.47. (a) (1/2, 0, 1/2); (b) (0, 1/2, 1/2); (c) (−1/2, 0, 1/2);
(d) (0, −1/2, 1/2).
1.48. z1 = 1/z2 .
F cω

1.51. xp (t) = cos(ωt − δ), where δ = sin−1 ,  = (k 2 − ω2 )2 + c2 w 2 .
 
1
1.52. ωL = .
ωC

Chapter 2
2.1. (a) u(x, y) = 2x 3 − 6xy 2 − 3x, v(x, y) = 6x 2 y − 2y 3 − 3y;
x y
(b) u(x, y) = 2 , v(x, y) = − 2 ;
x +y 2 x + y2
−x 2 − y 2 + 1 2x
(c) u(x, y) = 2 , v(x, y) = − 2 .
x + (y − 1) 2 x + (y − 1)2
3 z
2.2. (a) z2 ; (b) z ; (c) .
2z − 1
u2 v2
2.3.  2 +  2 = 1;
r0 + r10 r0 − r10
x(x 2 + y 2 + 1) = α(x 2 + y 2 ), y(x 2 + y 2 − 1) = β(x 2 + y 2 ).
ik
2.5. V (z) = , k is the vortex strength.
z−α
422 Answers to Problems

2.8. (a) not continuous at z = 0; (b) continuous at z = 0.

2.12. Along the line segment joining z1 and z2 with period π .
x2 y2
2.14. 2 + 2 = 1.
a b
2.15. (a) the whole complex plane;
(b) the whole complex plane except at z = 0;
(c) z = 0; (d) the whole complex plane.
2.16. (a) z = 0; (b) z = 0; (c) empty set.
2.17. 0.
2.19. (a) nowhere analytic; (b) entire, f  (z) = 2z − 3;
1+i
(c) analytic everywhere except at z = 0, f  (z) = − 2 .
z
2.20. (b) 2(1 + i); f  (z) exists only along the line x − y = 1.
2.22. No. 
2z for |z| < 1
2.23. f  (z) = .
1 for |z| > 1
f is not differentiable on the circle |z| = 1. Define D = {z : |z| = 1},
which is an open set since its complement D c = {z : |z| = 1} is closed.
Since every point inside D is an interior point and f is differentiable
throughout D, f is analytic in D.
2.25. (a) no; (b) yes, the  harmonic
 conjugate is also constant.
1
2.30. (a) iz + i; (b) i
3
− 1 ; (c) (1 − i)z3 + ic, c is any real number.
z
3 x4 + y4
2.31. (a) x 2 y 2 − = β; (b) 2e−x cos y + 2xy = β;
2 4
(c) (r 2 − 1) sin θ = βr.
2.32. (a) a = 2, b = −1, c = −1, d = 2; (b) k = 1; (c)  = 1.
2.33. v(x, y) is not a harmonic conjugate of u(x, y).
2.34. u2 is not harmonic.
2.35. No.
2.36. Yes.
1 (x + a)2 + y 2
2.38. v = ln + c, c is a real constant,
2 (x − a)2 + y 2
2ay
θ = tan−1 2 .
x + y 2 − a2
∂ 2 u 1 ∂u 1 ∂ 2u
2.39. 2
+ + 2 2 = 0.
∂r r ∂r r ∂θ
r 4 cos θ
2.40. (a) u(x, y) = (y − 2)xey ; (b) u(r, θ ) = .
15
2.44. T (r, θ ) = k ln r, r = 0, k is real. The flux lines are y = βx, β is
constant.
 Answers to Problems 423

Chapter 3  
1
3.4. (a) ln 2 + + 2k π i, k is any integer,
2
 
1
(b) + 2k π ± 3i, k is any integer.
2
3.6. (a) Im(sin z) = 0 if Re z = (2k + 1)π/2 or Im z = 0, k is any
integer,
Re(sin z) = 0 if Re z = kπ, k is any integer;
(b) Im(tan z) = 0 if Im z = 0,
Re(tan z) = 0 if Re z = kπ
2
, k is any integer;
(c) Im(coth z) = 0 if Im z = kπ 2
, k is any integer,
Re(coth z) = 0 if Re z = 0.


sin2 2x + sinh2 2y
3.7. (a) | tan z| = ,
cos 2x + cosh 2y


sinh2 2x + sin2 2y
(b) | tanh z| = .
cosh 2x + cos 2y
3.10. 2i. √
π √ 3π 3± 7
3.15. (a) + 2kπ − i ln( 2 ± 1); (b) + 2kπ − i ln ;
4 √ 4 √ 2
π 3−1 3π 3+1
(c) + 2kπ − i ln √ and − + 2kπ − i ln √ ;
4 2 4 2
(d) 2kπ
 i; (e)  − ln 2 + (2k + 1)π  i; 
1 1
(f) 2k + π i and − ln 3 + 2k − π i;
2 2
(2k + 1)π
(g) kπ(1 ± i); (h) kπ(1 + i) and ;
1+i
(4k + 1)π (4k − 1)π
(i) and .
2(1 + 2i) 2(1 − 2i)
In all cases, k is any integer.
3.16. (a) limit does not exist;
(b) (i) unbounded as y → ±∞; (ii) oscillatory;


(c) | sin z| ≤ 1 + sinh2 α.
1 3
3.17. (a) ln 13 − tan−1 i;
2 2

1 3
(b) ln 13 + (2k + 1)π − tan−1 i, k is any integer;
2 2
sinh 4 − i sin 2 40 + 9i
(c) cos 2 cosh 1 − i sin 2 sinh 1; (d) ; (e) ;
cosh 4 − cos 2 41
π √ π √
(f) + 2kπ − i ln( 2 + 1) and − + 2kπ − i ln( 2 − 1), k is any
2 2
integer;
424 Answers to Problems
√ √
(g) 2kπ + i ln( 2 + 1) and (2k + 1)π − i ln( 2 − 1), k is any integer;
(h) not defined.
π
3.18. (a) analytic everywhere except at 0, + kπ, k is any integer;
2
π
(b) analytic everywhere except at − + kπ, k is any integer;
4
ln 3 2k + 1
(c) analytic everywhere except at − +i π , k is any positive
2 2
integer.
3.21. The image is the whole complex plane except the origin. The inverse
1
function is Log z.
2π
3.23. Log z is multi-valued.
3.24. The heat source and heat sink are at z1 and z2 , respectively. The temper-
ature value is zero along the perpendicular bisector of the line segment
joining z1 and z2 .
U1 U2 z − x2
3.25. T (z) = Arg(z − x1 ) + Arg + ···
π π z − x1
Un z − xn Un+1
+ Arg + [π − Arg(z − xn )].
π z − xn−1 π

U 2 Im(sin z)
3.26. T (z) = tan−1
π | sin2 z| − 1
2U cos x
or T (x, y) = tan−1 , z = x + iy.
√ π sinh y
3.27. − 3 2.
3.28. Both z = 0 and z = 1 are branch points of order 3; z = ∞ is not
a branch point. Choose the line segment joining the two branch
points as the branch cut. The Riemann surface consists of four
sheets.
3.30. Both z = i and z = −i are branch points, while z = ∞ is not a branch
point. The function is not defined at z = ±i, while tan−1 ∞ = π2 + kπ, k
is any integer. The branch cut of the Riemann surface is the line segment
joining z = −i and z = i.
1
3.31. .
1 − z2
3.32. (b) ln 3 + iπ ; (c) No, the starting point of the semi-infinite branch
cut can be chosen to be any point that lies on the branch cut joining
z = −1 and z = 1.
3.33. No. [z(z + 1)]1/2 is a double-valued function while z1/2 (z + 1)1/2 is the
product of two double-valued functions.
x2 y2
3.37. One branch of the hyperbola 2
− = 1.
cos θ0 sin2 θ0
 Answers to Problems 425

Chapter 4
1 −5 + 7i
4.2. (a) e + ; (b) .
e 3
4.3. (a) 1; (b) 1 + i.
4.4. The length of the line segment joining any two points is always less than
or equal to the arc length of any curve joining the same two points.
4.6. 6π (π + ln 3).
4.7. 2π e.
4.8. (a) 7π/2; (b) π cosh 1; (c) 2/e.
πi πi
4.11. (a) 0; (b) 6π i; (c) √ ; (d) −π e−i ; (e) − √ ; (f) 0;
2 2
(g)
√ 0; (h) 0; (i) π i; (j) 8.
2 2
4.12. i.
3
4.18. 2π ie , 2π ie−iπ/3 , 0.
iπ/3

4.20. (a) 4π i; (b) 0 . g(2) does not , exist.

1 1
4.26. No; for example, for n = 1, n
dz = 0 but n is not analytic at
|z|=r<1 z z
z√= 0.
4.30.  13 at z = ±i.
π
4.31. ,1
2  
K1  z − α  K2 z−α
4.36. φ(x, y) = ln  + Arg ,
2π z−β 2π z−β
 
K2  z − α  K1 z−α
ψ(x, y) = − ln + Arg ;
2π  z − β  2π z−β
the streamlines are families of logarithmic spirals.
√ c2
4.40. f (z) = U∞ z2 + 1; the streamlines are y 2 = c2 + 2 , c is any
x + c2
constant.
n

K  rj +1
4.41. u(x, y) = Sj ln − Cj (θj +1 − θj ) ,
2π j =1 rj

K 
n
rj +1
v(x, y) = − Cj ln − Sj (θj +1 − θj ) ,
2π j =1 rj
ηj − y
where rj2 = (ξj − x)2 + (ηj − y)2 , θj = tan−1 , j = 1,
ξj − x
2, . . . , n,
[(ηj +1 − ηj )(ξj − x) − (ξj +1 − ξj )(ηj − y)](ξj +1 − ξj )
Cj = ,
(ξj +1 − ξj )2 + (ηj +1 − ηj )2
[(nj +1 − ηj )(ξj − x) − (ξj +1 − ξj )(ηj − y)](ηj +1 − ηj )
Sj = .
(ξj +1 − ξj )2 + (ηj +1 − ηj )2
426 Answers to Problems

4.43. The greatest speed occurs at (a, π2 ) and (a, − π2 ). The equipotential lines
are c(x 2 + y 2 ) = x(x 2 + y 2 + a 2 ), c is any constant.
ρ
4.45. (z) = − [Log(z − z1 ) − Log(z − z1 )]; the equipotential lines
2π   
 z − z1 
form the family of coaxial circles defined by    = k, k is any
z − z1 
constant.
φb − φa φa ln b − φb ln a
4.46. φ(r) = ln r + .
ln ab, ln ab
r(ω0 ζ + r)
4.48. gh = Gρ Re dζ, ω0 = ξ0 + iζ0 ,
|ζ |=1 ζ (ω0 + rζ )
2Gρπ r 2 ζ0
= 2 , |ω0 | > r.
ξ0 + ζ02

Chapter 5

5.4. (a) absolute convergence; (b) divergence;

(c) absolute convergence.
5.5. (a) when |z| < 1, the series converges to −1;
(b) when z = 1, the series converges to 0;
(c) when z = −1, the series diverges;
(d) when z = eiθ , θ = 0, the series diverges;
(e) when |z| > 1, the series diverges.
5.8. (a) R; (b) R/2; (c) when R = 0, the new radius of convergence
becomes infinite; when R > 0, the new radius of convergence can
be any value; (d) R k .
5.9. Divergence on |z| = 1.
5.11. (a) ∞, (b) 1, (c) 1/4, (d) e.
z3 z z2 z4
5.14. (a) z − z2 + , (b) ln 2 + + − .
3 2 8 192
5.15. (a) 1 − 2z + z2 + z4 − 2z5 + z6 + z8 − 2z9 + z10 + · · · ,
7 47 z9 z15
(b) z − z3 + z5 − · · · + · · · , (c) z3 + + + ···,
6 40 3! 15!
z2 z4 z6
(d) − + − ···.
4 96 4320
π
5.16. E0 = 1, E2 = −1, E4 = 5, E6 = −61; circle of convergence is |z| < .
2
5.17. |z| = 1.  
 
√ n+1  
√ n+1 √
1  1+ 5 1− 5  5−1
5.18. cn = √ − , n ≥ 0, R = .
5 2 2 2
5.19. sin−1 z = z + z3
6
+ 3z5
40
+ · · · , valid for |z| < 1.
 Answers to Problems 427

a0 b0 a1 − b1 a0 b02 a2 − b0 b1 a1 + (b12 − b0 b2 )a0

5.20. c0 = , c1 = , c2 = ;
b0 b02 b03
b0 cn + b1 cn−1 + b2 cn−2 + b3 cn−3 = 0, n = 3, 4, . . . .
5.22. The complex extension of the function, f (z) = 1+z 1
2 , has singularities at

z = ±i.
1
5.24. 1−z = 1 + z + z2 + · · · , valid for |z| < 1; the series diverges at z = −1
but the sum function 1−z1
is analytic at z = −1.
 
1 z z2
5.28. (a) 0 < |z| < 1, (1 + z + z + · · · ) −
2
1+ + + ··· ;
2 2 4
   
1 1 1 1 z z2
1 < |z| < 2, − 1 + + 2 + ··· − 1+ + + ··· ;
z z z 2 2 4
   
1 2 4 1 1 1
2 < |z| < ∞, 1 + + 2 + ··· − 1 + + 2 + ··· ;
z z z z z z
∞
n(z − i) n−2
(b) 0 < |z − i| < 1, (−1)n−1 ;
n=1
i n+1
∞
(n + 1)i n
1 < |z − i| < ∞, (−1)n .
n=0
(z − i)n+3
1 i n−1
5.29. (a) (i) an = , n ≥ 0; bn = , n ≥ 1;
2n+1 (i − 2) i−2
i n−1 − 2n−1
(ii) an = 0, n ≥ 0; bn = , n ≥ 1;
i−2
−1 1
(iii) an = , n ≥ 0; b1 = , bn = 0 for n ≥ 2;
(2 − i)n+2 i−2
2π i
value of the integral = .
 n+2 i −2
−1 1
(b) an = , n ≥ 0; b1 = − , bn = (−1)n for n ≥ 2;
2 2
∞ ∞
1 1
(c) an = bn = , n ≥ 1; a0 = .
k!(n + k)! (k!)2
k=0 k=0

1 1 n 1 1
5.30. sin 1 1 − + · · · + (−1) + · · · + cos 1
2! (z − 1)2 (2n)! (z − 1)2n

1 1 1 1 1
× − + · · · + (−1) n
+ ··· .
z − 1 3! (z − 1)3 (2n + 1)! (z − 1)2n+1
5.32. 2π i.
5.34. The two power series expansions are valid within their respective circles
of convergence which are non-overlapping. Therefore, it is meaningless
to add these two series together.
428 Answers to Problems

Chapter 6
6.2. (a) removable singularity; (b) removable singularity;
(c) essential singularity.
6.3. (a) When n = m, z = 0 can be a removable singularity, for example,
f1 (z) = −f2 (z); it can be a pole of order k, k ≤ m, for example,
f1 (z) = −f2 (z) + z1k .
(b) It is a pole of order m + n.
(c) When n > m, z = 0 is a pole of order n − m; when n ≤ m, it is a
zero of order m − n.
6.4. It becomes a removable singularity if f1 = −f2 , and a pole if f1 =
−f2 + (z − z0 )−k , k is a positive integer.
6.5. (a) The function is entire;
(b) z = 2kπ i is a simple pole, k is any nonzero integer;
(c) z = 0 is an essential singularity;
(d) z = kπ 1
is an essential singularity, k is any nonzero integer.
6.6. The classification of an isolated singularity and the computation of its
residue should be done with reference to the Laurent series expansion of
the function
 2in a deleted  neighborhood  of2 the singularity.

z −1 z −1
6.7. (a) Res 3 2 , i = −1, Res 3 2 , −i = −1,
z (z + 1) z (z + 1)
 2 
z −1
Res 3 2 , 0 = 2;
z (z + 1)
 
tan z π 1
(b) Res , nπ + = nπ+ π , n is any integer;
1−e z 2 e 2 − 1
 iαz  √
e β(1 ± i) 2 √
(c) Res 4 , √ = ∓ (1 ± i)e αβ(i∓1)/ 2
;
z + β4 2 8β 3
 
1 1
(d) Res , nπ = (−1)n , n is any non-zero integer,
z sin z nπ
 
1
Res , 0 = 0;
z sin z
 1 
ez
(e) Res , 0 = 1.
z
6.8. (a) n, (b) −n; (a) ng(α), (b) −ng(α).
6.9. (a) essential singularity; (b) pole of order 6.
∞ ∞

6.10. Write f (z) = fn (z − z0 )n and g(z) = gn (z − z0 )n ,
n=0 n=2
 
f (z) f1 g2 − f0 g3
Res , z0 = .
g(z) g22
 Answers to Problems 429

6.11. z = 0, ±1, ±2, · · · , represent poles of order one of f (z) = π cot π z;

Res (π cot π z, n) = 1.
6.12. The function has a pole of order 2 at z = 0 and a pole of order 3 at
3 π2 − 6
z = π ; Res (f, 0) = − 4 , Res (f, π ) = .
π 2π 4
6.13. 8.
(2n)!
6.14. (a) −1; (b) .
(n − 1)!(n + 1)!
n
f (ak )
6.16. 2π i .
k=1
(ak − a 1 ) · · · (ak − ak−1 )(ak − ak+1 ) · · · (ak − an )
1 3 2
6.17. (a) − + ;
z−1 z−2 z−3
 
1 z2 z5 z3 z1 z4
(b) − + + + + ,
5 z − z1 z − z2 z − z3 z − z4 z − z5
where zk = ekπi , k = 1, 3, 5, 7, 9;
 
1 z1 zn
(c) + ··· + , zk = e2kπi , k = 1, 2, . . . , n.
n z − z1 z − zn
6.19. (a) 10π i; (b) − 65 π i; (c) 0; (d) 3e π i; (e) 2π i;
2π i  π
(f) sin i + (n − 1) , n is any positive integer.
(n − 1)! 2
2π a (2a + b)π
6.20. (a) 2 ; (b) ;
(a − b2 )3/2 [a(a + b)]3/2
2π 2π
(c) if |a| < 1, 2 if |a| > 1, 0 (principal value) if |a| = 1
1 − a2 a −1
but a = ±1. When a = ±1, the principal value of the integral does
not exist.
π π 1 · 3 · 5 · · · (2n − 3) π π
6.21. (a) − ; (b) ; (c) if n > 1 and
27 4a 2 · 4 · 6 · · · (2n − 2) 2 2
if n = 1;
π π π π π
(d) ; (e) √ ; (f) π ; (g) √ sin .
ab(a + b) 2 n sin n 2 8
√ 2 sin βω
6.23. (a) π e−|ω| , ω is real; (b) π e−ω /4 ; (c)
2
.
ω
π π π e−ab
6.27. (a) (cos 1 − cos 3); (b) (2 cos 2 + sin 2); (c) ;
3e3 2e4 2b
π
(d) π e−ab ; (e) (1 − e−a ) .
2
π2
6.29. − √ .
8 2
6.30. π4 ln 2.
430 Answers to Problems
π
6.33. I = when p = ±1.
2
6.39. (a) −iπ; (b) π i(e2mi − emi ).
3π
6.42. .
4

Chapter 7
 1

K
7.13. u(x, y) = ln((x − t) + y ) dt − 2 ln(x + y ) .
2 2 2 2
2π −1
 π
1
7.14. u(r, θ ) = [P (R, r, φ − θ ) + P (R, r, φ + θ )]f (φ) dφ.
2π 0
dY (s)
7.16. (a) 2s + (s 2 + λ + 2)Y (s) = sy0 + y0 ;
ds
dY (s)
(b) s(s − 1) + (s − λ − 1)Y (s) = 0.
ds
1 1
7.18. (a) tan−1 ; (b) Log √ .
s s +1
2
√
tan−1 s π Log s 2 + 1
7.19. (a) − ; (b) .
s 2s s

1 − et 2 a cos ty
7.22. (a) ; (b)
dy.
t π 0 a2 − y 2

1 t φ  (τ )
7.23. y(t) = dτ .
π 0 (t − τ )1/2
7.26. T (x, t) = 2 + e−π t sin π x.
2

 ∞ −(x−ξ )2 /4a 2 t
+ e−(x+ξ ) /4a t
2 2
e
7.27. T (x, t) = √ f (ξ ) dξ .
0 4π a 2 t
√ω   
−x ω
7.28. T (x, t) = e 2a 2
sin ωt − x .
2a 2
 ∞  
q0 ξ
7.29. (a) T (x, t) = erfc √ dξ
k x 2 t
%   &
q0 t −x 2 /4t x
= 2 e − x erfc √ ;
k π 2 t
 t
q0 sin ω(t − τ )
(b) T (0, t) = √ √ dτ,
k π 0 τ
 t
x 1
e−x /4(t−τ ) dτ .
2
T (x, t) = √ T (0, τ )
4π 0 (t − τ )3/2
 Answers to Problems 431
 x x
7.30. T (x, t) = T1 1 − + T2
L L
∞
21 nπx
(T2 cos nπ − T1 )e−n π a t/L sin
2 2 2 2
+ .
π n=1 n L
7.32. u(x, t) = t sin x.
1
7.33. u(x, t) = 2 (1 − cos π t) sin π x.
π
 2
t if t < x2
7.35. u(x, t) = x 2 2 .

2
if t ≥ x2

Chapter 8
 
8.1. (a) z = ±1, ±i; (b) z = nπ, n + 12 π, n is any integer.
8.3. (a) a circle that passes through w = 0 and w = α1 ;
(b) (u2 + v 2 )2 = u2 − v 2 ; (c) (1 − 2pu)v 2 = 2pu3 .
8.5. One possible choice of the mapping function is
 
w
+ i − ie−i 1+im .
w
z = x + iy = a
1 + im
 
zf  (z)
Re 1 + f  (z)
8.6. .
|zf  (z)|
15
8.8. π.
2  π/(π−2α)
√
−iα z + z2 − 4
8.9. w = e .
2
π
8.11. An infinite wedge: 0 < Arg w < .
4
K π
8.12. T (z) = K − 4
Arg z , 0 < Arg z < .
2π 4
2 −1 2
8.13. T (z) = Re (sin z ).
π   2 
T 1 + T2 T2 − T1 2z
8.14. T (x, y) = + Re sin−1 − 1 , z = x + iy.
2 π x02
2T0
8.15. T (z) = − Re(sin−1 (−e−πz )).
π

(a + b)2
8.16. f (z) = U∞ ζ + ,
4ζ
√
z + z2 − c 2
where ζ = and c2 = a 2 − b2 .
2

8.17. f (z) = U∞ (z − p + p2 − 2pz).
432 Answers to Problems
 π (z+1) 
e −1
8.18. f (z) = m ln π (z−1) .
e −1
(1 + i)(z − i) 2(z − 1) iz + 3
8.23. (a) w = ; (b) w = ; (c) w = .
2z z (2 + i)(z − i)
1 − 2z
8.24. w = iπ/3 .
e z−1
z2 + 2iz + 1
8.25. w = .
z2 − 2iz + 1
z−1
8.26. w = 3 .
z−2
8.27. |d| = |c|.
z−1 z − z0
8.28. (a) w = (1 − i) ; (b) w = .
z+1 r
 
 z − 2(1 + i)  √

8.29. Inversion point is 5(1 + i);   = 2 2.
z − 5(1 + i) 
2z 2
8.31. w = eiθ , θ is real; ρ = .
z + 24 3
z−i
8.32. w = w0 + Ri .
z+i
w−α z−α
8.33. = eiα .
1 − αw 1 − αz
2z − 1
8.34. w= .
2−z
 z
1

dz
0 (1 − z )(1 − k 2 z2 )
2
8.35. w= 1 .
k 1

dz
1 (z2 − 1)(1 − k 2 z2 )
2H √

8.36. w=i [sin−1 z + z(1 − z)].
π
√
8.37. w = a + h z2 − 1.
 z

h
8.38. w= πi + (eζ + 1)α dζ ;
π πi
h z
in particular, when α = 1, w = (e + z + 1).
π
 Answers to Problems 433

8.39. The required Schwarz–Christoffel transformation is

 
1 1
w = iH − sin−1 z ,
2 π
or, equivalently,
π
z = cosh w.
H
K cosh Hπ w − 1 K  π 
T (w) = Arg = Arg tanh2 w .
π cosh H w + 1
π
π 2H
2V

8.40. f (z) = Log(z + z2 − a 2 ).
π  

h1 h2 h1
8.41. w = Log(1 − z) + Log 1 + z .
π π h2
 Index

Absolute convergence, 197, 201 Cauchy inequality, 155

Accumulation point, 23 Cauchy integral formula, 151
Alternating current circuit, 33 generalized form, 154
phasor, 34 Cauchy integral theorem, 142
Analytic continuation, 233, 246 Cauchy principal value, 285, 288,
Analytic functions, 70 294, 308
Analyticity in a domain, 70 Cauchy product, 215, 230
Apollonius family of circles, 383 Cauchy residue theorem, 256
Argument, 2 Cauchy-Riemann relations, 64, 90
polar coordinates, 68
Bernuolli law, 295 Chebyshev polynomial, 39
Bernoulli numbers, 221, 243 Circle of convergence, 206
Bernoulli polynomials, 243 Circle theorem, 192
Bessel functions, 213, 229 Circulation, 165
Bilinear transformations, 375, 415 Closed region, 27
circle-preserving property, 378 Closed set, 24
circle-to-circle, 396 Closure, 24
mapping of symmetry points, 387 Compact set, 27
one-to-one mapping, 376 Comparison test, 197, 238
symmetry-preserving property, 381 Complex acceleration, 64
triples to triples, 376 Complex conjugate, 3
upper-half plane onto circle, 391 Complex functions, 46
upper-half plane onto itself, 390 conformal property, 63, 360
Blasius laws, 295, 310 continuity, 58
Boundary point, 24 differentiation, 61
Boundary value problems, 311 domain of definition, 46
Dirichlet problems, 316, 319, 348 limit, 54
Neumann problems, 324, 350 range, 46
Bounded set, 27 Complex infinity, 29
Branch cut, 120, 353 Complex numbers, 1
Branches of a multi-valued function, additive inverse, 6
47 argument, 2
principal branch, 47, 105 complex conjugate, 3
Branch points, 119 complex infinity, 29
Bromwich integrals, 330, 355 Gaussian integer, 11
Bromwich line, 331 imaginary part, 2
modulus, 2
Cauchy criterion, 195 multiplicative inverse, 6
Cauchy-Goursat theorem, 143 polar representation, 6, 38

434
 Index 435

real part, 2 Derivatives of contour integrals, 153

root of unity, 16 Differentiation, 61
vectorial representation, 3 Dirac function, 282, 305, 328
Complex number field, 1 Dirichlet condition, 280
Complex plane, 2 Dirichlet problems, 316, 319, 348
extended complex plane, 29 circular arc, 349
imaginary axis, 3 Domain, 27
real axis, 3 Domain of univalence, 118, 131
Complex potential, 168, 178 Doublet, 49, 173
fluid, 168, 404, 412
electrostatic, 178, 407, 418 Electric circuit, 33
Complex sequences, 194 capacitor, 35
absolute convergence, 197 inductor, 35
Cauchy criterion, 195 resistor, 34
conditional convergence, 197, Electric potential, 177
240 Electrostatic fields, 176
convergence tests, 197 complex potential, 178, 407, 418
Complex series, 200 cylindrical conductors, 181, 317
region of convergence, 200 electric potential, 177
uniform convergence, 202, 239 equipotential lines, 167
Weierstrass M test, 202 flux function, 178
Complex velocity, 63, 169 line charge, 178
Conditional convergence, 197, 240 semi-circle, 317
Conformal mappings, 358 semi-infinite strip, 366
angle-preserving property, 360 two semi-infinite rods, 407, 418
behavior around a critical point, 363 Entire function, 71
bilinear transformations, 375, 415 Equipotential lines, 167
one-to-one property, 362 Essential singularity, 249, 254
scale factor, 361 Euler formula, 7
Schwarz-Christoffel transformations, Euler numbers, 241
399, 416 Exponential function, 53, 93
Conjugate complex variables, 69, 181 generalized function, 115
Connected domain, 28 mapping properties, 97
Connected set, 27 Exponential order, 327
Conservative fields, 166 Extended complex plane, 29
potential function, 166 Exterior point, 24
Continuity equation, 163
Continuity of a function, 58 Fluid flows, 162
uniform continuity, 60 Blasius laws, 295, 310
Contour, 135 circle theorem, 192
Contour integral, 136 circulation, 165
Control volume, 81 complex potential, 168, 178
Convergence tests, 197 complex velocity, 63, 169
comparison test, 197 continuity equation, 163
Gauss test, 197 doublet, 49, 173
ratio test, 197, 208 hydrodynamic force, 295, 310
root test, 197, 208 hydrodynamic moment, 295, 310
Convolution, 279, 329 incompressibility condition, 167
Fourier transform, 279 irrationality condition, 164
Laplace transform, 329 Kutta-Joukowski theorem, 299
Cosine function, 98 seepage flow, 372, 412
Coulomb law, 175 sink, 49
source, 48, 171, 298
D’Alembert formula, 357 stream function, 310
Darcy law, 373 streamlines, 167
De Moivre theorem, 7, 37, 39 velocity potential, 167
Deleted neighborhood, 23 vertical step, 404
436 Index

Fluid flows (cont.) modulus inequality, 137

vortex, 85, 172 primitive function, 144
vortex patch, 191 real variable of integration, 134
vorticity, 166 Interior point, 24
Fluid source, 48, 171, 298 Inverse Laplace transform, 332
velocity function, 48 Irrationality condition, 164
Flux lines, 83 Isolated singularity, 248, 300
Fourier cosine transform, 281, 304 Isometry, 85
Fourier integrals, 285, 305 Isothermal curves, 82, 92, 369
Fourier series, 315
Fourier sine transform, 281 Jacobian of transformation, 92, 364
Fourier transforms, 278, 304, 345, Jet flow, 412
356 Jordan arc, 28
convolution, 279 Jordan lemma, 285, 294, 352
Dirichlet condition, 280 Joukowski mapping, 123, 411, 413
inversion formula, 279 inverse, 125, 132
Parseval identity, 281, 304
Kutta-Joukowski theorem, 299
Gauss mean value theorem, 158, 190
Gauss test, 197 Laplace equation, 74
Gaussian integer, 11 invariance property, 111, 364
Goursat theorem, 143 Laplace transforms, 326, 351, 355
Gravitational fields, 179 convolution, 329
potential, 180 derivatives, 328
vertical gravity, 180 Dirac function, 328
Gravity dam, 372 exponential order, 327
Green theorem, 150 inversion, 332, 352
conjugate complex variable shifting, 328
formulation, 181 step function, 328
Laurent series, 221, 244, 265
Harmonic conjugate, 75, 90, 325, 347 principal part, 222
Harmonic function, 74, 190, 312, 318, Laurent series theorem, 222
322, 347, 350 Lemniscate, 20
maximum and minimum value principles, Limit of a function, 54
161, 317 limit at infinity, 56
Harmonic series, 196 Limit point 23, 44
Heat conduction, 69, 337, 354 Limit superior, 196
earth’s surface, 354 Line charge, 178
finite rod, 339, 355 Liouville theorem, 159
infinite rod, 338, 356 Logarithmic function, 104, 120, 130
semi-infinite rod, 339, 355 branch, 120
Heat flux, 83, 106 Riemann surface, 121
Heat source, 94, 115 Longitudinal oscillations, 341
Heaviside function, 109, 284, 328
Hodographic transformations, 372 Maclaurin series, 215, 219
Holomorphic function, 70 Mapping properties, 50
Hydrodynamic force, 295, 310 conformal, 358
Hydrodynamic moment, 295, 310 isogonal, 409
Hyperbolic functions, 99 Maximum modulus theorem, 159, 162,
inverse function, 111 189
Maxwell equation, 177
Impedance, 34, 45 Mean value theorem, 158, 190
Incompressibility condition, 164 Meromorphic function, 250
Initial-boundary value problems, 336 Method of images 191, 298
Integration, 133 Mixed Dirichlet-Neumann problem, 350,
contour integral, 136 368
line integral, 135 Modulus, 2, 36
 Index 437

Modulus inequality, 137 Residue at complex infinity, 267

Morera theorem, 157 Residue theorem, 256
Riemann mapping theorem, 398
Neighborhood, 23 Riemann sphere, 29, 44
complex infinity, 30 stereographic projection, 30
Neumann problems, 324, 350 Riemann surfaces, 118
compatibility condition, 324 logarithmic function, 121
Newton’s law of universal gravitation, power function, 119
179 Riemann zeta function, 205
Root of unity, 16
Open region, 27 Root test, 197, 208
Open set, 27
Order field, 7 Schwarz-Christoffel transformations,
399, 416
Parseval identity, 281, 304 Schwarz integral formula, 319, 349
Partial sums, 196 first quadrant, 321
Perturbation series, 232 Schwarz kernel, 319, 321
Phasor, 33 Seepage flows, 372, 412
Picard theorem, 251 Simply connected, 24
Pointwise convergence, 200 Sine function, 98
Poisson equation, 84, 91 mapping properties, 102
Poisson integral, 148 inverse function, 111, 113
Poisson integral formula, 312 Singular point, 70
Fourier series expansion, 314 essential singularity, 249,
Poisson kernel, 38, 314 254
Pole, 249, 292, 300 isolated singularity, 248
order, 252 pole, 249, 292, 300
Potential flows, removable singularity, 249
jet flow, 412 Smooth curve, 135
past a circle, 175, 309 Stereographic projection, 30
past a perturbed circle, 232, Stokes theorem, 166
245 Stream function, 167
past an obstacle, 230, 411 Streamlines, 167
right half-plane, 298 Symmetric points, 17, 387
seepage, 372, 412 Symmetry-preserving property,
vertical step, 404 381
Power function, 116, 241, 263
Power series, 211 Tangential acceleration, 64
circle of convergence, 207, 211 Taylor series, 215
Maclaurin series, 215, 219 Taylor series theorem, 216
radius of convergence, 207 Temperature distributions, 80
Taylor series expansion, 214 buried pipe, 384
Principal branch, 47, 105 first quadrant, 110
infinite strip, 129, 411
Radial acceleration, 64 inside a channel, 405
Radius of convergence, 207 inside a circle, 308
Ratio test, 197, 208 sectorial domain, 369
Real number field, 1 source, 106
Reflection principle, 100, 236 upper-half plane, 108
Region, 27 Temperature fields, 82
Region of convergence, 200 flux lines, 83, 92
Regular function, 70 isothermal curves, 82, 92,
Regular point, 248 369
Removable singularity, 249 Triangle inequalities, 14, 40
Residues, 255 Trigonometric functions, 53,
computational formulas, 257 97
evaluation of real integrals, 268 inverse functions, 112
438 Index

Uniform continuity, 60 Vortex, 85, 172

Uniform convergence, 202, 239 Vortex patch, 191
power series 211 Vorticity, 166
Univalent, 118, 131
Wave propagation, 341
Velocity potential, 167 D’Alembert formula, 357
Vertical gravity, 180 finite rod, 343
circular cylinder, 193 semi-infinite rod, 343
horizontal cylinder, 181, 193 Weierstrass-Casorati theorem, 250
Von Neumann polynomial, 11 Weierstrass M test, 202