Taras Mel’nyk

Complex
Analysis
Complex Analysis
Taras Mel’nyk

Complex Analysis
Taras Mel’nyk
Faculty of Mathematics and Mechanics
Taras Shevchenko National University
of Kyiv
Kyiv, Ukraine

ISBN 978-3-031-39614-4 ISBN 978-3-031-39615-1 (eBook)

https://doi.org/10.1007/978-3-031-39615-1

Translation from the Ukrainian language edition: “Complex Analysis” by Taras Mel’nyk, © Mel’nyk
T.A. 2015. Published by Taras Shevchenko National University of Kyiv. All Rights Reserved.

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Paper in this product is recyclable.

To Warriors of Light
Preface

The proposed textbook contains theoretical material that corresponds to the edu-
cational program “Complex Analysis” of training specialists of the educational
qualification level “Bachelor” for students of the Faculty of Mechanics and Mathe-
matics of the Taras Shevchenko National University of Kyiv.
Being a student of Lviv State University, where the world-famous school of
functional analysis was founded, headed by Professor Stefan Banach, as well as a
graduate student of Moscow State University, where I attended Professor Shabbat’s
course on Complex Analysis of Several Variables [10], I was given the opportunity
to observe the best teaching traditions of these famous old universities. Taking them
into account, I developed a two-semester course “Complex Analysis” in 1993 and
since then I have been teaching this course regularly at the Faculty of Mechanics
and Mathematics of the Taras Shevchenko National University of Kyiv.
During this time, the prepared lectures have been expanded and modified. Since
2004, they have been presented in electronic form (in Ukrainian) on the website
of the Department of Mathematical Physics: http://www.matfiz.univ.kiev.ua/books.
Then two new chapters and many figures were added to this electronic version.
Several books [1, 2, 4, 9, 11] were useful guides in preparing lecture notes and then
for the Ukrainian version of the textbook [5].
Since then, this textbook has been read and downloaded from my Research Gate
webpage more than 2500 times by readers from different countries, although it was
published in Ukrainian. I was also encouraged by the positive feedback from many
readers and former students of mine who now teach at various universities (e.g.,
Professor Oleksandr Misiats used my lectures to teach a course on complex analysis
for masters at the Courant Institute of Mathematical Sciences), and many of them
recommended the publication of this book in English.
The English version is a substantial extension of the Ukrainian one. Some
important new theorems and their proofs are added, as well as many new examples,
exercises, and figures. I am very grateful to Springer Publisher for supporting my
proposal and publishing this textbook.

What Advantages Does This Book Offer over Competitive Titles

and What Is Unique About It?
• There are three concepts for constructing the theory of analytic functions, which
are associated with Cauchy, Riemann, and Weierstrass, respectively. In most

vii
viii Preface

textbooks, the theory is based on the Cauchy or Weierstrass approach. Cauchy’s

approach focuses on the integral calculus of functions of a complex variable,
and the key topics are contour integration and the Cauchy integral formula.
For Weierstrass’ approach, power series are central, and the key topic is spaces
of analytic functions represented as power series. As a result, the concept
of conformal mappings, which occupies a central place in complex analysis,
appears only in the middle or at the end of textbooks (courses).
In the textbook, I adhere to the Riemann concept, in which the differentiability
of complex-valued functions of a complex variable is central. This approach
makes it possible to introduce the concept of conformal mappings at the
beginning (Chap. 2) and to study properties of such mappings, including their
hydromechanical interpretation and geometric meaning of the modulus and
argument of the derivative. In addition, the material in the textbook is selected
in such a way as to simplify the proof of subsequent theorems. This leads to a
significant reduction in the volume of the textbook.
• I was surprised to find that many textbooks limit themselves to linear-fractional,
power, exponential, and logarithmic functions only. For example, the Joukowsky
function, which is used in many applications, is not considered at all. Chap-
ter 3 examines in detail the properties of many elementary analytic functions,
including trigonometric and hyperbolic functions, and their inverse functions. It
is shown how to construct Riemann surfaces of these inverse functions. Such a
detailed analysis of the properties of various elementary analytic functions allows
students to better understand the general properties of analytic functions.
• There are many topics in this book that are often missing in other texts,
namely the conformal mapping criterion and its proof; the equivalence of three
approaches to the construction of the theory of analytic functions; the notion
of an integral for an analytic function along an arbitrary curve (not necessarily
piecewise-smooth) is extended thanks to the introduction and study of properties
of an antiderivative along a curve; the theory of global analytic functions and their
Riemann surfaces; the inverse function theorem in the general case, Lagrange-
Bürmann and Puiseux series.
• The Complex Analysis course is a natural continuation of the theory of real func-
tions. Therefore, the tutorial contains many examples that compare properties of
analytic functions of a complex variable and real-valued functions, and also show
their differences.
• Many textbooks contain a list of problems and exercises after each section,
most of which are typical and standard. As there are many taskbooks for
this course (e.g., [3, 6, 8, 12, 14]), the author has taken a different approach.
Problems, many of them invented by the author, are presented in the textbook
either immediately after the definition of a new concept in order to better
understand it or immediately after the theorem that must be applied to solve
it. These are theoretical tasks and their purpose is to help students actively and
informally assimilate the material, as well as to illustrate the difficult points of
the theory. Their solutions do not require sophisticated calculations, but only a
good understanding. I like to call them “problems solved in three lines.”
Preface ix

• In addition, the textbook contains many full-color figures that successfully

illustrate the essence and basic properties of theoretical concepts.

Description of the Contents

The textbook contains a brief but fairly complete exposition of the main ideas
of the theory of functions of a complex variable, with clear and rigorous proofs,
the presence of which is mandatory in textbooks for mathematics departments. It
consists of nine chapters.
The first chapter is introductory and introduces complex numbers and their
various forms. We then look at the complex plane, the extended complex plane with
its geometric interpretation, and their basic topological concepts. Complex-valued
functions of a real variable and various curves in the complex plane are considered
in more detail.
The second chapter introduces analytic functions as functions that are dif-
ferentiable with respect to a complex variable. This leads to the proof of the
Cauchy-Riemann theorem and to the concept of harmonic conjugate functions. We
also define conformal mappings and study properties of such mappings, including
their hydromechanical interpretation and the geometric significance of the modulus
and argument of the derivative.
Chapter 3 examines in detail the properties of many elementary analytic func-
tions and their inverse. These inverse functions turn out to be multivalued functions.
Therefore, we first introduce the empirical concept of a Riemann surface for such
functions and show how to construct Riemann surfaces for these inverse functions.
A rigorous topological approach for Riemann surfaces is given in Chap. 8.
The theory of integration of complex-valued functions of a complex variable
along a curve is considered in Chap. 4. Here we prove the Cauchy-Goursat theorem
for triangles, the general Cauchy theorem for homotopic curves and corollaries to
this theorem, as well as Cauchy’s integral formula. In addition, the notion of an
integral for an analytic function along an arbitrary curve (not necessarily piecewise-
smooth) is expanded thanks to the introduction and study of properties of an
antiderivative along a curve. Theorems on the existence of a local antiderivative, an
antiderivative along a curve, and an antiderivative in the whole domain are proved.
Chapter 5 presents the most important application of the Cauchy integral
formula, namely the proof that an analytic function in a disk can be expanded in
a power series. As a result, we get simple and nice proofs of Liouville’s theorem,
the fundamental theorem of algebra, Morera’s theorem, and the equivalence of three
approaches to the construction of the theory of analytic functions. The theorem on
the uniqueness of analytic functions coinciding on a certain sequence is also proved.
This theorem allows one to characterize isolated zeros of an analytic function and
their concentration, and as a consequence, it is easy to prove the theorem about the
factorization of polynomials.
Chapter 6 deals with Laurent series. The properties of such series are studied
and it is proved that an analytic function in an annulus can be expanded into
a Laurent series. The connection between Laurent series and Fourier series is
also demonstrated. Isolated singularities of analytic functions are classified and
x Preface

the behavior of analytic functions in neighborhoods of singularities, including the

singularity at infinity, is studied. The chapter ends with the classification of analytic
functions with respect to their isolated singularities and the proof of theorem on a
meromorphic function.
Residue theory is considered in Chap. 7. Here various formulas for the calcu-
lation of residues are proved and a wide variety of applications of this theory are
demonstrated, namely different methods for calculating integrals, expansions of
meromorphic functions into the series of elementary fractions, and expansions of
entire functions into infinite products. In addition, logarithmic residues are defined,
the argument principle and Rouché’s theorem are proved. Using the latter, we derive
simple sufficient conditions for the conformity of a function, Hurwitz’s theorem, and
other corollaries.
The theory of analytic continuation is covered in Chap. 8. Here such important
theorems as the principle of analytic continuation by continuity, the Schwarz
reflection principle and the monodromy theorem are proved. The topological
approach is used to present the theory of global analytic functions (such a function is
a set of all analytic function elements obtained from some initial element by analytic
continuation along all possible curves) and their Riemann surfaces.
The purpose of Chap. 9, devoted to the qualitative properties of analytic func-
tions, is to prove the Riemann mapping theorem in the general case. Along the way,
we prove the open mapping theorem, the maximum modulus principle, Schwarz’s
lemma, the inverse function theorem, and Montel’s theorem. We also discuss
the Lagrange-Bürmann formula and Puiseux series, and deduce the theorem on
conformal automorphisms of canonical domains.

For Which Courses Would the Textbook Be Suitable?

The theory of complex-valued functions of a complex variable belongs in a
basic course of mathematics faculties as well as physics and some engineering
departments of many universities. Therefore, the book can be used in a two-semester
course for undergraduate mathematics majors, a one-semester course for physics or
engineering specialties, or a one-semester course for first-year graduate students in
mathematics. Essential prerequisites include basic courses in mathematical analysis
of real-valued functions of one and several variables as well as courses in linear
algebra and elementary topology.
There are several topics in this textbook that can be in one or another advanced
course; notably, the theory of global analytic functions and the general approach to
the theory of Riemann surfaces of global analytic functions (Sects. 8.3–8.6) can be
included in advanced course in topology; Lagrange-Bürmann formula and Puiseux
series (Sect. 9.2) can be in approximation theory or asymptotic analysis; conformal
isomorphisms and automorphisms and Montel’s theorem (Sects. 9.3 and 9.4) will
find their place in an advanced course in functional analysis.
Preface xi

There are, of course, many other interesting topics in the theory of complex
analysis. The interested reader can familiarize himself with them, for example, in
books [4, 10, 13].
Please send your feedback and suggestions on the content of the textbook to the
email address: [email protected]

Stuttgart, Germany Taras Mel’nyk

June 2023
Acknowledgements

It gives me great pleasure to thank those who have helped me to write this textbook.
First, I would like to thank Dr. Remi Lodh at Springer Heidelberg for his support
and cooperation, and for his motivational letters, which have been of more value to
me than he can know.
It took me more than two years to write this book, and much of that time
coincided with the full-scale war unleashed by the Russian regime. Some sections
were written in a bomb shelter in Kyiv during air and rocket attacks on the city. That
is why I dedicate this book to all those who are part of the fight against this brutal
and terrible aggression.
The writing of this book took place in parallel with my research at the University
of Stuttgart, following a persuasive invitation from Professor Christian Rohde in
May 2022, and supported first by the Humboldt Foundation until the end of February
2023, and then by the German Research Foundation (the research project SFB 1313,
Number 327154368). I am therefore very grateful to a remarkable number of people
who have made my family’s stay in Stuttgart safe, pleasant, and, for me, productive.
I am sincerely grateful to my former graduate student Andriy Popov for
transforming many of my figures into electronic format. Special thanks are due to
Professor Andriy Olenko of La Trobe University and Professor Oleksandr Misiats of
Virginia Commonwealth University, who read the first three chapters and provided
a number of helpful comments and suggestions.

xiii
Instructions for Readers

In the text, you will come across the following symbols:

 def 
The equality sign with “def ” above . = means that the left-hand side is being
defined by the right-hand side.
The equality sign with “.:” before .( := ) means that the left-hand side is the
designation for the right-hand side.
. indicates the end of the proof;

. indicates the end of a definition, remark, or example where appropriate.

Also, when a term is defined for the first time outside a formal “Definition,” the
word is italicized.
In addition to the generally accepted symbols

Symbol Meaning
.∃ There exists
.∀ For all
.! Unique
.⇒ Implies
.⇐⇒ Is equivalent to

which are often used when presenting mathematical arguments in statements,

definitions, and proofs of theorems, the following symbols are also often found in
the text:
 domain (an open and path-connected set) in the complex plane
 the closure of 
∂ the boundary of 
A() ring of analytic functions in 
.
Br (a) open disk of radius r > 0 centered at a point a
Br (a) closed disk
Eγ trace of a curve γ
M
fn ⇒ f uniform convergence on a set M of a function sequence {fn }n∈N

xv
Contents

1 Complex Numbers and Complex Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Sequences in the Complex Plane: Extended Complex Plane . . . . . . . . . 7
1.3 Complex-Valued Functions of a Real Variable . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Curves in the Complex Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Basic Topological Concepts of the Complex Plane . . . . . . . . . . . . . . . . . . . 17
2 Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1 Structure of Complex-Valued Functions of a Complex Variable . . . . . 22
2.2 Differentiability of Complex-Valued Functions
of a Complex Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Conjugate Harmonic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Hydrodynamic Interpretation of Analytical Functions . . . . . . . . . . . . . . . . 33
2.5 Conformal Mappings: Geometric Meaning of the Modulus
and Argument of the Derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Elementary Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.1 Linear and Fractional-Linear Functions and Their Simplest
Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Group and Circular Properties of Fractional-Linear Functions . . . . . . . 47
3.3 Preservation of Symmetric Points by Fractional-Linear Mappings. . . 52
3.4 Fractional-Linear Isomorphisms and Automorphisms . . . . . . . . . . . . . . . . 55
3.5 Power Functions with Natural Exponents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.6 The Inverse to a Power Function and Its Riemann Surface . . . . . . . . . . . 59
3.7 Exponential Function, Logarithmic Function and Its Riemann
Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.8 Joukowsky Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.9 Trigonometric and Hyperbolic Functions and Their Inverses. . . . . . . . . 72
4 Integration of Functions of a Complex Variable . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.1 Line Integrals and Their Simplest Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2 An Antiderivative: Cauchy-Goursat Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3 Local Existence of an Antiderivative: Antiderivative
Along a Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.4 The Cauchy Integral Theorem and Corollaries . . . . . . . . . . . . . . . . . . . . . . . . 96
4.5 The Cauchy Integral Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
xvii
xviii Contents

5 Complex Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

5.1 Basic Definitions and Properties of Function Series and Power
Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.2 Expansion of a Differentiated Function Into a Power Series. . . . . . . . . . 111
5.3 Analyticity of the Sum of a Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.4 Uniqueness of Power Series Expansions: Morera’s Theorem . . . . . . . . 115
5.5 Uniqueness Theorem for Analytic Functions: Zeros of
Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6 Laurent Series: Isolated Singularities of Analytic Functions . . . . . . . . . . . 131
6.1 Expansion of an Analytic Function Into a Laurent Series. . . . . . . . . . . . . 131
6.2 Relationship Between Laurent Series and Fourier Series . . . . . . . . . . . . . 137
6.3 Isolated Singularities of Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.4 Behavior of an Analytic Function Near Its Essential Singularity . . . . . 145
6.5 Classification of Analytic Functions with Respect to Their
Isolated Singularities: Theorem on a Meromorphic Function . . . . . . . . 147
7 Residue Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.1 Cauchy’s Residue Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.2 Formulas for Calculating Residues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.3 Methods for Calculating Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.4 Argument Principle: Rouché’s Theorem and Its Applications . . . . . . . . 167
7.5 Partial Fraction Decomposition of a Meromorphic Function . . . . . . . . . 176
7.6 Factorization of an Entire Function Into an Infinite Product . . . . . . . . . . 181
8 Analytic Continuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
8.1 Analytic Function Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
8.2 Methods of Analytic Continuation: Schwarz’s Reflection
Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.3 Analytic Continuation Along a Curve: The Monodromy
Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
8.4 Global Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
8.5 Riemann Surfaces of Global Analytic Functions . . . . . . . . . . . . . . . . . . . . . . 206
8.6 Singularities of Global Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
9 Qualitative Properties of Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.1 Open Mapping Theorem, Maximum Modulus Principle,
Schwarz Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.2 Inverse Function Theorem: Puiseux Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
9.3 Conformal Isomorphisms and Automorphisms . . . . . . . . . . . . . . . . . . . . . . . 227
9.4 Montel’s Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
9.5 Riemann Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Complex Numbers and Complex Plane
1

Abstract

In this chapter we recall some concepts from basic courses in mathematical

analysis of real-valued functions of one and several variables as well as from
a course in linear algebra, namely complex numbers and their various forms,
arithmetic operations on them (Sect. 1.1), and basic topological notions in the
vector space .R2 (Sect. 1.5). The novel notion of the stereographic projection
in Sect. 1.2 provides a geometric interpretation of the extended complex plane.
Complex-valued functions of a real variable and various curves in the complex
plane are considered in more detail in Sects. 1.3 and 1.4, respectively.

1.1 Complex Numbers

A number is the basic concept of mathematics, which evolved throughout the history
of humankind. The emergence and formation of this concept went hand in hand
with the emergence and development of mathematics. Practical human activities,
on the one hand, and internal needs of mathematics, on the other, determined the
development of the concept of numbers.
The necessity of counting objects led to the emergence of the concept of the set of
natural numbers .(N). Starting with natural numbers, the number system expanded in
response to the need to describe quantities that could not be accommodated within
the existing (previous) number system. As a result, sets of integers .(Z), rational
numbers .(Q), and real numbers .(R) appeared in mathematics such that .N ⊂ Z ⊂
Q ⊂ R.
Complex numbers arose from the need to find solutions of polynomial equations,
for example, .x 2 + 1 = 0. The first written mention of complex numbers as square
roots of negative numbers can be found in Girolamo Cardano’s book in 1545. For
nearly two centuries, complex numbers remained mysterious, had a poor reputation
and were generally not considered legitimate. The active use of complex numbers

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 1

T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_1
2 1 Complex Numbers and Complex Plane

in calculations began with works of Leonard Euler (1707–1783), in particular, with

his famous formula (1.9) introduced in 1748. The first systematic description of
complex numbers, arithmetic operations on them and their geometric interpretation
was conducted by Carl Gauss (1777–1855) in his memoir “Theoria residuorum
biquadraticorum” (1828, 1832). The term “complex number” is due to C. Gauss in
1831. The 2000-year and engaging history of complex numbers is presented in [7].

Definition 1.1 The set .C of complex numbers is the set of ordered pairs .(x, y)
of real numbers x and .y, equipped with algebraic operations of addition and
multiplication:

def
(x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ),
. (1.1)

def
.(x1 , y1 ) · (x2 , y2 ) = (x1 x2 − y1 y2 , x1 y2 + y1 x2 ). (1.2)

.

It is clear that two complex numbers .(x1 , y1 ) and .(x2 , y2 ) are equal if and only if
x1 = x2 and .y1 = y2 . From the definition it follows that
.

(x1 , 0) + (x2 , 0) = (x1 + x2 , 0),

. (x1 , 0) · (x2 , 0) = (x1 x2 , 0). (1.3)

Thus, addition and multiplication on complex numbers of the form .(x, 0) coincide
with the corresponding algebraic operations on real numbers. Therefore, we can
identify each real number x with the complex number .(x, 0), i.e., .R  x := (x, 0) ∈
C, and after this identification one can state that .R ⊂ C. In addition, one can verify
that for any real number a

a · (x, y) = (a, 0) · (x, y) = (ax, ay).

. (1.4)

The complex number .(0, 1) is called the imaginary unit and is denoted by the
Latin letter i. It is easy to check that

i 2 = (0, 1) · (0, 1) = (−1, 0) = −1

. and (0, y) = (0, 1) · (y, 0) = iy.

Based on these notations, any complex number can be represented as

.(x, y) = (x, 0) + (0, y) = x · (1, 0) + y · (0, 1) = x + iy,

which is called the algebraic form of a complex number. The algebraic form of a
complex number is usually denoted by one letter .z := x + iy. Moreover, the number
x is called the real part of the complex number .z and is denoted .Re(z), while the
number .y is called the imaginary part of z and is denoted .Im(z).
1.1 Complex Numbers 3

The conjugate of a complex number .z = x + iy is the complex number .z :=

x − iy. The modulus or absolute value of z is defined by

.|z| := x2 + y2.

Note that .|z| = |z| and .z · z = x 2 + y 2 = |z|2 .

Subtraction and division of two complex numbers .z1 = x1 +iy1 and .z2 = x2 +iy2
are defined as follows:
z1 z1 · z2
z1 − z2 = (x1 − x2 ) + i(y1 − y2 ),
. = (z2 = 0). (1.5)
z2 |z2 |2

The set of complex numbers with respect to the introduced operations forms a
field, i.e., it is an Abelian group1 with respect to addition with .0 = (0, 0) as the
additive identity; the nonzero elements in .C form an Abelian group with respect
to multiplication with .1 = (1, 0) as the multiplicative identity; and multiplication
distributes over addition.

Exercise 1.1 Prove that all these field properties are fulfilled.

From (1.3) it follows that the field of complex numbers includes the field of real
numbers as a subfield. The reader is invited to make sure that all extensions of the
field .R obtained by joining the root of the equation .x 2 + 1 = 0 to it are isomorphic
to the field .C.
Based on Definition 1.1, (1.1) and (1.4) we can assert that the set of complex
numbers is a real vector space,2 or more precisely, the vector space .R2 . This makes
the complex numbers a Cartesian plane (coordinate plane), called the complex plane.
Clearly that the real numbers lie on the horizontal x-axis, called the real axis, and
the y-axis is called the imaginary axis of the complex plane. This allows to give
the geometric interpretation of complex numbers and arithmetic operations defined
on .C and, conversely, to express some geometric properties and constructions in
terms of complex numbers. For instance, conjugation is the reflection symmetry
with respect to the real axis; multiplication by .−1 is the central symmetry about the
origin.

1 Recall that a group is a set of elements together with a binary operation on this set such that
the following three requirements, known as group axioms, are satisfied: the binary operation is
associative, there is a unique identity with respect to this operation, and every element of this
set has an inverse with respect to this operation. In an Abelian group, the binary operation is
additionally commutative.
2 It is a set of objects called vectors, which may be added together and multiplied by real numbers

(scalar multiplication). This set is an Abelian group under addition, and scalar multiplication has
the following properties: .x(u + v) = xu + xv, .(x + y)v = xv + yv, .(xy)v = x(yv), and .1v = v
for all .x, y ∈ R and all vectors .u and .v.
4 1 Complex Numbers and Complex Plane

Fig. 1.1 Geometric

interpretation of the sum and
subtraction two complex
numbers

Fig. 1.2 The argument of z

Furthermore, sum and subtraction of two complex numbers coincides with the
sum and subtraction of the corresponding vectors in .R2 (Fig. 1.1). The absolute
value of a complex number is the length of the corresponding vector (the usual
Euclidean norm) in the vector space .R2 .

Other Forms of Writing Complex Numbers

It is known that the position of the point .(x, y) is also determined by the pair
.(r, ϕ), where r is the distance of .(x, y) to the origin and .ϕ is the counterclockwise

angle (measured in radians) between the positive x-axis and the ray from the origin
through .(x, y); the values r and .ϕ are called the polar coordinates of .(x, y), and

x = r cos ϕ,
. y = r sin ϕ. (1.6)

Substituting these relations into the algebraic form of a complex number .z =

x + iy yields the trigonometric form:

z = |z| (cos ϕ + i sin ϕ),

. (1.7)

where the angle .ϕ is called the argument of z (Fig. 1.2).

Note that the argument of each nonzero complex number z is defined ambigu-
ously, and up to the term .2π k (k ∈ Z); in addition, there exists a unique angle
.ϕ0 ∈ (−π, π] such that .ϕ = ϕ0 + 2π k. This angle .ϕ0 is called the principal value
1.1 Complex Numbers 5

of the argument of z and is denoted .arg(z). The set of all arguments of z is denoted
by

Arg(z) := {ϕ0 + 2π k : k ∈ Z}.

.

The argument of 0 is not defined. The principal value of the argument of z can
be considered as a real-valued function defined on .C \ {0} and it can be expressed
from the formulas (1.6) in terms of the inverse trigonometric function .arctan :
⎧
⎪
⎪ arctan yx , if x > 0;
⎨
π + arctan yx , if x < 0, y > 0;
. arg(z) = (1.8)
⎪
⎪ π, if x < 0, y = 0;
⎩
−π + arctan yx , if x < 0, y < 0.

Example 1.1 It is easy to verify that

• the principal value of the argument of each positive number x .(y = 0) is zero,
and the set of all arguments of x is .Arg(x) = {2π k : k ∈ Z};
• .arg(1 − i) = − π4 , and .Arg(1 − i) = {− π4 + 2π k : k ∈ Z};
• .arg(−3) = π, and .Arg(−3) = {π + 2π k : k ∈ Z}.

Let us define the exponential function of an imaginary number .iα by the

following way:

def
. eiα = cos α + i sin α (α ∈ R), (1.9)

which is known as Euler’s formula (the proof is given in Example 5.4). From (1.9)
it is clear that .|eiα | = 1. In addition, it is easy to verify that

eiα1 · eiα2 = (cos α1 + i sin α1 )(cos α2 + i sin α2 )

.
  
= cos α1 cos α2 − sin α1 sin α2 + i sin α1 cos α2 + cos α1 sin α2 )
= cos(α1 + α2 ) + i sin(α1 + α2 )
= ei(α1 +α2 ) . (1.10)

In (1.10) we used the addition formulas for sine and cosine. Similarly, it is proved
that

n eiα1
. eiα = einα , = ei(α1 −α2 ) .
eiα2
6 1 Complex Numbers and Complex Plane

Using Euler’s formula (1.9), we get from (1.7) the exponential form of a complex
number: .z = |z| eiϕ . This form well illustrates the essence of multiplication and
division of complex numbers. If .z1 = |z1 | eiϕ1 and .z2 = |z2 | eiϕ2 , then

z1 |z1 | i(ϕ1 −ϕ2 )

z1 · z2 = |z1 ||z2 | ei(ϕ1 +ϕ2 ) ,
. = e (z2 = 0).
z2 |z2 |

Thus, when multiplying (respectively dividing) two complex numbers, their moduli
are multiplied (resp. divided):

z1 |z1 |
|z1 · z2 | = |z1 ||z2 |,
. = ,
z2 |z2 |

and arguments are added (resp. subtracted):

z1
ϕ1 + ϕ2 ∈ Arg(z1 · z2 ),
. ϕ1 − ϕ2 ∈ Arg( ).
z2

Definition 1.2 A complex number .z is called an .nth root of a complex number .a,
if .zn = a. Here, .n ∈ N and .a = 0. .

Let us derive a formula for finding .nth roots of a complex number .a = |a| eiθ
.(θ ∈ (−π, π )). If .z = |z| e
iϕ is an .nth root of .a, then according to the definition

|z|n = |a|,
|z|n einϕ = |a| eiθ ⇐⇒
.
nϕ = θ + 2π k, k ∈ Z,

whence
√
|z| = n
|a|,
.
ϕk = θ+2π k
n , k ∈ Z,

that is, the .nth roots of a are numbers

θ 2π k
i n+ n
zk =
.
n
|a| e , k ∈ Z. (1.11)

It is easy to see that among these complex numbers there are exactly n different
numbers. Indeed, the numbers .z0 , . . . , zn−1 are different since their arguments

θ θ + 2π θ + 2π(n − 1)
ϕ0 =
. , ϕ1 = , . . . , ϕn−1 =
n n n
1.2 Sequences in the Complex Plane: Extended Complex Plane 7

Fig. 1.3 The .nth roots of a

complex number a

are various and differ from each other less than .2π. For any other number .zk , .k ∈ /
{0, . . . , n − 1} there exist numbers .p ∈ Z and .q ∈ {0, 1, . . . , n − 1} such that
.k = pn + q. This means that .zk = zq .

Thus, the equation .zn = a has n different roots .z0 , . . . , zn−1 , defined by the
formula (1.11) and √located at the vertices of a regular n-sided polygon inscribed in
a circle of radius . n |a| centered at the point 0 (Fig. 1.3).

1.2 Sequences in the Complex Plane: Extended Complex Plane

Since the modulus of a complex number is just the usual Euclidean norm in the
vector space .R2 , it is natural to introduce the distance between two complex
numbers as follows

.d(z1 , z2 ) := |z1 − z2 | = (x1 − x2 )2 + (y1 − y2 )2 ,

where .z1 = x1 + iy1 , z2 = x2 + iy2 . In addition, we can naturally introduce

the convergence of a sequence of complex numbers as the convergence of a
sequence of the corresponding vectors in .R2 . We hope that the reader is familiar
with the properties of convergent sequences from real analysis of several variables.
Nevertheless, let us briefly recall the main definitions and properties.

Definition 1.3 A sequence .{zn = xn + iyn }n∈N of complex numbers is said to

converge to a complex number .a = α + iβ .(denoted as . lim zn = a), if
n→+∞

. lim |zn − a| = 0,
n→+∞

i.e., for every .ε > 0, there exists an integer N such that

|zn − a| < ε
. for all n ≥ N.

.
8 1 Complex Numbers and Complex Plane

From Definition 1.3 follows a statement, which is offered to the reader as an

exercise.

Exercise 1.2 Prove that a sequence .{zn = xn + iyn }n∈N converges to the complex
number .a = α + iβ if and only if

. lim xn = α and lim yn = β.

n→+∞ n→+∞

Definition 1.4 It is said that a sequence .{zn }n∈N of complex numbers converges to
infinity .( lim zn = ∞), if
n→+∞

. lim |zn | = +∞,

n→+∞

i.e., for every .R > 0, there exists an integer N such that

|zn | > R
. for all n ≥ N.


.

The symbol ."∞" is called the point at infinity.

Definition 1.5 The set .C := C ∪ {∞} is called the extended complex plane.

Obviously, that each sequence in .C contains a convergent subsequence. This is

called the principle of compactness in .C. The point at infinity does not participate
in algebraic operations, i.e. it cannot be multiplied or added to complex numbers. In
real analysis, points labeled .+∞ and .−∞ produce the two-point compactification
of the set of real numbers.

Geometric Interpretation of . C
Consider the space
 
R3 = (ξ, η, ζ ) : ξ ∈ R, η ∈ R, ζ ∈ R ,
.

in which the .ξ -axis coincides with the real axis, .η-axis coincides with the imaginary
axis, and .ζ -axis is perpendicular to the complex plane (Fig. 1.4). The sphere
 2 
S := (ξ, η, ζ ) ∈ R3 : ξ 2 + η2 + ζ −
.
1
2 = 1
4

is tangent to the complex plane at the origin. The point .N = (0, 0, 1), which lies
on the sphere, will be called the “north pole”. Define a mapping .p : C → S as
1.2 Sequences in the Complex Plane: Extended Complex Plane 9

Fig. 1.4 Riemann sphere

follows: to each point .z ∈ C we assign an intersection point .Z(ξ, η, ζ ) where the

line between z and N meets the sphere .S apart from N (Fig. 1.4), that is
p
C  z −→ Z(ξ, η, ζ ) := S ∩ [z, N ] \ {N }.
.

Obviously, if . lim zn = ∞, then the images .{Zn }n∈N on the sphere approach
n→+∞
to N . Therefore, it is naturally to determine .p at the point at infinity as follows:
p
.∞ −→ N. The mapping .p : C → S is called the stereographic projection.
Let us examine properties of p. Obviously, this is a one-to-one mapping. To
explicitly define the stereographic projection, we exclude the variable t from the
parametric equations of the segment .[N, z]: .ξ = tx, .η = ty, .ζ = 1 − t, where
.t ∈ [0, 1], and as a result we obtain formulas for the inverse mapping .p
−1 :

ξ η
x=
. , y= . (1.12)
1−ζ 1−ζ

Since the coordinates of the point .Z(ξ, η, ζ ) satisfy the relation

 
1 2 1
.ξ + η + ζ − = ⇐⇒ ξ 2 + η2 = ζ (1 − ζ ),
2 2
2 4

then

ξ 2 + η2 ζ x2 + y2
.x2 + y2 = = ⇒ ζ = .
(1 − ζ ) 2 1−ζ 1 + x2 + y2
10 1 Complex Numbers and Complex Plane

From the last equation and formulas (1.12) we get formulas for the stereographic
projection:

x y x2 + y2
ξ=
. , η= , ζ = . (1.13)
1 + x2 + y2 1 + x2 + y2 1 + x2 + y2

It follows from (1.12) and (1.13) that .p : C → S is a homeomorphism (by definition,

it is a one-to-one, onto continuous mapping with a continuous inverse).
Using the map p, we can identify the extended complex plane .C with the sphere
.S. After this identification, the sphere S is called the Riemann sphere, or the sphere

of complex numbers.

Exercise 1.3 Prove that under the stereographic projection an arbitrary circle or
straight line on .C maps to a circle on .S, and the angle between curves in .C is equal
to the angle between the images these curves on .S.

1.3 Complex-Valued Functions of a Real Variable

Consider a function .f : R → C. Such a complex-valued function of a real variable

can be represented as .f (t) = u(t) + iv(t), t ∈ R, where .u(t) := Re(f (t)) and
.v(t) := Im(f (t)). Thus, we see that each function .f : R → C can be viewed as a
 
vector-function . uv from .R in .R2 due to the geometrical interpretation of the set of
complex numbers. Therefore, such concepts as the limit of a function, continuity,
uniform continuity and many other properties of vector-functions of a real variable
are automatically transferred to such functions. Let us recall some of them.

Definition 1.6 A number .A = α + iβ is the limit of a function .f : R → C at a

point .t0 ∈ R .(denoted as . lim f (t) = A), if for each .ε > 0 there exists .δ > 0 such
t→t0
that

. |f (t) − A| = (u(t) − α)2 + (v(t) − β)2 < ε whenever |t − t0 | < δ.


.

As in the proof of the assertion in Exercise 1.2, the following statement can be
easily proved.

Proposition 1.1 The limit of a function .f : R → C exists at a point .t0 ∈ R and it

is equal to .A = α + iβ if and only if there exist the limits of its real and imaginary
parts and they are equal to .α and .β, respectively, i.e.,

. lim f (t) = A ⇐⇒ lim u(t) = α and lim v(t) = β.

t→t0 t→t0 t→t0
1.3 Complex-Valued Functions of a Real Variable 11

Definition 1.7 A function .f : [a, b] → C is called continuous on the closed

interval .[a, b] ⊂ R .(denoted as .f ∈ C([a, b]) ), if for all .t0 ∈ [a, b]

. lim f (t) = f (t0 ).

t→t0

The corresponding one-sided limits are considered at the endpoints a and b.

Definition 1.8 The derivative of a function .f : R → C at a point .t0 ∈ R .(denoted

by .f  (t0 )) is called the limit

f (t) − f (t0 )
. lim , (1.14)
t→t0 t − t0

provided that it exists. 

.

Suppose the limit (1.14) exists. Then, according to Proposition 1.1

 
u(t) − u(t0 ) v(t) − v(t0 )
f  (t0 ) = lim
. +i
t→t0 t − t0 t − t0
u(t) − u(t0 ) v(t) − v(t0 )
= lim + i lim = u (t0 ) + iv  (t0 ).
t→t0 t − t0 t→t0 t − t0

Obviously, the reverse chain of equalities is also true. Thus, the following statement
is correct.

Proposition 1.2 The derivative of a function .f : R → C at .t0 ∈ R exists if and

only if the derivatives of its real and imaginary parts exist at .t0 .

Example 1.2 The function .f (t) = exp(it), t ∈ R, has the derivative at each point
and .f  (t) = i exp(it). Indeed, for any . t ∈ R

(exp(it)) = (cos t + i sin t) = − sin t + i cos t = i(cos t + i sin t) = i exp(it).

.

Due to Proposition 1.2 the equality .f  (t0 ) = u (t0 ) + iv  (t0 ) can be taken as
an equivalent definition of the derivative of a complex-valued function of a real
variable. We will apply this approach to define the integral of a complex-valued
function of a real variable.
12 1 Complex Numbers and Complex Plane

Definition 1.9 Let .f (t) = u(t) + iv(t), t ∈ [a, b], and the functions .u and .v be
Riemann-integrable on the segment .[a, b].

b b b
def
. f (t) dt = u(t) dt + i v(t) dt.
a a a


.

Exercise 1.4 Prove that Definition 1.9 is equivalent to the definition of the integral
introduced through the limit of the Riemann sums of f , i.e.,

b 
n
. f (t) dt = lim f (τk ) Δtk ,
→0
a k=1

where .a = t0 < t1 < . . . < tn = b, .Δtk := tk − tk−1 , .tk−1 ≤ τk ≤ tk , .Δ =

maxk∈{1,...,n} Δtk .

It is easy to check the following properties of integrals of complex-valued

functions of a real variable:
 b  b  b
(1) .∀ λ, μ ∈ C (λf (t) + μg(t)) dt = λ f (t) dt + μ g(t) dt;
a b  c a
b
a

(2) .∀ c ∈ (a, b) f (t) dt = f (t) dt + f (t) dt;

a a c
(3) if .F is the antiderivative of .f, i.e., .F  (t) = f (t) for all .t ∈ [a, b], then
 b
. f (t) dt = F (b) − F (a);
a

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

Exercise 1.5 Using Exercise 1.4, prove the fourth property.

Remark 1.1 Not all properties of real-valued functions are automatically carried
over to complex-valued functions of a real argument. For instance, the statement
of the mean value theorem is incorrect. This fact is easy to check for such a
continuous function: .eit , t ∈ [0, 2π ]. Evidently that .eit = 0 for all .t ∈ [0, 2π ].
Therefore, on the one hand, assuming that the mean value theorem holds, we have
 2π
that . 0 eit dt = 0. On the other hand,
 2π  2π  2π
. e dt =
it
cos t dt + i sin t dt = 0.
0 0 0
1.4 Curves in the Complex Plane 13

Exercise 1.6 Show that the statements of Rolle’s theorem and Cauchy’s mean value
theorem are also incorrect for complex-valued functions of a real variable. Recall
that Rolle’s theorem states the following: if a real-valued function f is continuous
on a closed interval .[a, b], differentiable on .(a, b), and .f (a) = f (b), then there
exists a point .ξ ∈ (a, b) such that .f  (ξ ) = 0. The second theorem establishes
the relationship between the derivatives of two functions. Let functions f and g be
continuous on .[a, b], differentiable on .(a, b), and .g  (x) = 0 for all .x ∈ (a, b). Then
there is a point .ξ ∈ (a, b) such that

f (b) − f (a) f  (ξ )
. =  .
g(b) − g(a) g (ξ )

Remark 1.2 Since it is impossible to introduce an order relation for complex

numbers, the Weierstrass theorem for complex-valued functions of a real variable
is formulated as follows: for a continuous function .f : [a, b] → C, its modulus
reaches its largest and smallest value on the closed interval .[a, b].

1.4 Curves in the Complex Plane

A curve is a geometric concept, the exact and at the same time quite general
definition of which presents significant difficulties and is given in various branches
of mathematics and textbooks in different ways. For those branches in which
methods of the theory of functions dominate, the natural definition of a curve is
to define it by parametric equations. In this text, we will take this approach and give
the following definition of a curve and its elements.

Definition 1.10 A curve in .C .(in .C) is called a continuous complex-valued function

of a real variable: .z = γ(t), .t ∈ [a, b] ⊂ R.
Moreover, the points .γ(a) and .γ(b) are called the initial and end points of .γ,
respectively. The curve .γ is said to be closed if .γ(a) = γ(b). .

Remark 1.3 In the notation of a curve .z = γ(t), t ∈ [a, b], or .γ : [a, b] → C,

we will always mean that the closed interval .[a, b] is a real closed interval, i.e.,
.[a, b] ⊂ R.

The image of such a continuous function is also often called a curve. In the course
“Complex Analysis” it is convenient to distinguish between these concepts in order
to better understand
 some new definitions and theorem proofs. The image of .γ, i.e.,
the set .γ [a, b] , is called the trace of the curve .γ and is denoted by .Eγ .
Each curve specifies an orientation that can be interpreted as the direction of
movement of the point .γ(t) along the trace .Eγ from its initial point to its end as the
parameter t increases from .a to .b.
14 1 Complex Numbers and Complex Plane

Example 1.3 Let .z1 ∈ C, .z2 ∈ C, and .γ(t) = z1 + t (z2 − z1 ), t ∈ [0, 1]. The
initial point of this curve is .z1 = γ(0), the end point is .z2 = γ(1). We will denote
its trace by .[z1 , z2 ] and refer to it as the segment joining .z1 and .z2 .

Separating the real and imaginary parts in the equality .z = γ(t), we find the
parametric equations, which are called a parametrization of the curve .γ, namely
.x = Re(γ(t)), .y = Im(γ(t)), where the parameter .t ∈ [a, b].

Example 1.4 Let .z = γ1 (t) = eit , t ∈ [0, 2π ]. Then

x = cos t,
.x + iy = cos t + i sin t ⇐⇒ t ∈ [0, 2π ].
y = sin t,

The last two equations determine a parametrization of this curve, whose trace is the
unit circle centered at the origin. It is a closed curve that starts at point .(1, 0) and is
oriented counterclockwise.

It is easy to see that the curve .z = γ2 (t) = ei2π t , t ∈ [0, 1], has the same trace
and orientation as the curve .γ1 from Example 1.4. For such curves, we will give the
following definition.

Definition 1.11 Two curves

. z = γ1 (t), t ∈ [a1 , b1 ], and z = γ2 (τ ), τ ∈ [a2 , b2 ],

are called equivalent .(γ1 ∼ γ2 ), if there exists a real-valued function .τ = μ(t),

t ∈ [a1 , b1 ], such that
.

(1) .μ ∈ C([a1 , b1 ]) and it is strictly increasing on .[a1 , b1 ];

(2) .μ(a1 ) = a2 , .μ(b1 ) = b2 ;
(3) .γ1 (t) = γ2 (μ(t)) for all .t ∈ [a1 , b1 ].

Exercise 1.7 Prove that this relation between two curves is the equivalence relation,
i.e., it is reflexive, symmetric and transitive.

Therefore, a curve can be understood as the corresponding equivalence class. It

is clear that equivalent curves have the same trace and orientation.

Example 1.5 The curve .γ1 from Example 1.4 and the curve

z = γ2 (τ ) = ei2π τ , τ ∈ [0, 1],

.

are equivalent. To show this we need to take the function .τ = μ(t) = t/2π, .t ∈
[0, 2π ], and verify the conditions from Definition 1.11.
1.4 Curves in the Complex Plane 15

Exercise 1.8 Prove that the curve .γ1 from Example 1.4 and the curve .z = γ3 (τ ) =
e−iτ , τ ∈ [0, 2π ], are not equivalent.

Definition 1.12 A point .z0 is called a self-intersection point of a curve .z = γ(t),

t ∈ [a, b], if there are .t1 = t2 , .{t1 , t2 } ⊂ [a, b] such that
.

γ(t1 ) = γ(t2 ) = z0 .
.

If a curve .γ is closed, then the point .γ(a) = γ(b) is not considered a self-
intersection point. .

A curve without self-intersection points is called simple and a closed simple

curve is said to be a Jordan curve.
Let .z = γ(t), .t ∈ [a, b], be a Jordan curve in .C. Then the Jordan curve
theorem asserts that the trace .Eγ divides the complex plane into an “interior” region,
denoted by .int(γ), bounded by the trace, and an “exterior” region, denoted by .ext(γ)
(Fig. 1.5), i.e.,

C \ Eγ = int(γ) ∪ ext(γ).
.

Intuitively, the statement of this theorem is obvious and there is no trouble

verifying it when a curve is given explicitly. A rigorous proof of the general result
is rather difficult, and we refer the reader to a topology text, e.g., [15]. The proof of
the Jordan curve theorem for smooth Jordan curves can be found in [13, §4.8].
We will say that a Jordan curve .γ has the positive orientation (denoted by .γ + ) if
its interior remains on the left when traversing its trace for increasing values of the
parameter t from a to .b (Fig. 1.5). Otherwise, it is negatively oriented .(γ − ).

Definition 1.13 A curve .z = γ(t), .t ∈ [a, b], is called smooth, if .γ is continuously

differentiable on .[a, b], i.e., .γ ∈ C 1 ([a, b]) and

γ  (t) = 0 for all t ∈ [a, b].

. (1.15)

If .γ is a closed curve, the condition .γ  (a) = γ  (b) must also be satisfied. 

.

Fig. 1.5 The interior and

exterior of a Jordan curve
with the positive orientation
16 1 Complex Numbers and Complex Plane

Let us find out the geometric meaning of (1.15). It is equivalent to

x  (t) + iy  (t) = 0 for all .t ∈ [a, b]. Since . x  (t), y  (t) is the tangent vector
.

to .Eγ at the point .γ(t), then the condition (1.15) means that at each point of .Eγ
there is a nonzero tangent vector that changes continuously.

Definition 1.14 A curve .z = γ(t), .t ∈ [a, b], is called piecewise smooth, if there
is a partition .a = a0 < a1 < . . . < an = b of .[a, b] such that for each .k ∈
{0, 1, . . . , n − 1} the curve .z = γ(t), t ∈ [ak , ak+1 ], is smooth.

Definition 1.15 A curve .z = γ(t), .t ∈ [a, b], is called rectifiable, if

• .γ is differentiable on .[a, b] except, possibly, at a countable set of points and

• there exists a finite integral
 b  b 
γ :=
. |γ  (t)| dt = (x  (t))2 + (y  (t))2 dt.
a a

The value .γ is called the length of .γ.

An example of a piecewise smooth curve is a broken line. Note that a piecewise

smooth curve is rectifiable.

Example 1.6 The curve

z = γ5 (t) = t 3 + it 2 ,
. t ∈ [−1, 1],

is simple, however it is not smooth and piecewise smooth (Fig. 1.6).

The curve

z = γ6 (t) = cos 2t exp(it),

. t ∈ [0, 2π ] (four petal rose curve)

is closed non-Jordan smooth curve that has the self-intersection point at the origin.
The curve
   
z = γ7 (t) = t 1 + i sin 1t ,
. t ∈ − π1 , π1 ,

is simple and non-rectifiable, so it is not piecewise-smooth.

1.5 Basic Topological Concepts of the Complex Plane 17

Fig. 1.6 The traces of curves .γ5 (left), .γ6 (right) and .γ7 (from below)

1.5 Basic Topological Concepts of the Complex Plane

Let us now recall some topological concepts. An open disk of radius .r > 0 centered
at a point .a ∈ C is a set of all points of distance less than r from .a, i.e.,

Br (a) := {z ∈ C : |z − a| < r}.

.

It is also called the r-neighborhood of a. The R-neighborhood of the point at

infinitely is the set

BR (∞) := {z ∈ C : |z| > R} ∪ {∞}.

.

Definition 1.16 A set .D ⊂ C is called open, if each point of D is contained in D

together with some of its r-neighborhood, i.e.,

∀ z0 ∈ D ∃ r > 0 such that

. Br (z0 ) ⊂ D.

Definition 1.17 Let .D ⊂ C and .z0 ∈ C. The point .z0 is called a limit point of the
set .D if every r-neighborhood of .z0 contains at least one point of D different from
.z0 itself, i.e.,

∀r > 0 ∃z ∈ D
. such that z ∈ Br (z0 ) \ {z0 }.

Definition 1.18 A set .D ⊂ C is called closed if it contains all its limit points.
18 1 Complex Numbers and Complex Plane

Example 1.7 Let .D := Z. Then, the set D is closed in .C, since the set of its limit
points is empty and .∅ ⊂ D. However, it is not closed in .C since it does not contain
its limit point .∞.

The joining to a set .D ⊂ C all its limit points is called the closure of D and
denoted by .D. For example, the closure of the open disk .Br (a) is the closed disk

Br (a) := {z ∈ C : |z − a| ≤ r}.
.

A set .D ⊂ C is said to be path-connected if for any two distinct points in D there

is a curve whose trace belongs to D and connects these points (starting at one point
and ending at the other).
A set .D ⊂ C is called a domain if it is open and path-connected.

Definition 1.19 Let .D be a domain in .C. The set .∂D := D \ D is called the
boundary of D.

Exercise 1.9 Prove that the boundary of a domain is the closed set.

There are several approaches to introducing the concept of simply connected

domains: a domain is simply connected if its fundamental group is trivial; a domain
.D in .R is simply connected if any closed curve in .D is homotopic to a point in this
m

domain (see Definition 4.4 and Exercise 4.4); one can define simply connectedness
through the general concept of connectedness of a set in a topological space. In this
course an easier-to-understand definition of simply connectedness is proposed.

Definition 1.20 A domain D is said to be simply connected (also called 1-

connected) in .C if for any Jordan curve .γ, whose trace belongs to .D, the interior of
.γ is fully contained in D, i.e., .int(γ) ⊂ D.

A domain D is said to be simply connected in .C if for any Jordan curve .γ, whose
trace belongs to D and .∞ ∈ / Eγ , obligatorily either .int(γ) ⊂ D or .ext(γ) ⊂ D.
Domains that are not simply connected are called multiply connected. .

Intuitively, a simply connected domain is a domain “without holes”.

Definition 1.21 The connectedness order of a domain .D ⊂ C is the number of

path-connected closed components of the boundary .∂D, which do not intersect.

Example 1.8 Consider the domain .D1 = {z : |z| > 1}. Obviously, it is multiply
connected in .C, since the interior of the circle .{z : |z| = 2} is not a subset of .D1 . In
.C its connectedness order is 2, because the boundary of .D1 has two path-connected

closed components that do not intersect, namely .{z : |z| = 1} and .{∞}.

Example 1.9 Due to the second part of Definition 1.20 the domain .D2 = D1 ∪{∞}
is 1-connected in .C.
1.5 Basic Topological Concepts of the Complex Plane 19

Example 1.10 The connectedness order of the domain

  
N
 i 
D3 = z : |z| < 2 \
. z : z = x + k , x ∈ [ 14 , 34 ]
2
k=1

is equal to .N + 1, where .N ∈ N.
Analytic Functions
2

Abstract

In this chapter and onwards, we study properties of complex-valued functions of

a complex variable. It turns out that every complex-valued function is determined
by the corresponding vector function from .R2 into .R2 . This fact enables us
to obtain some properties of complex-valued functions from the first section.
Fundamentally new is the notion of differentiability of complex-valued functions,
which we introduce in Sect. 2.2, although it formally coincides with the standard
definition (from calculus) of the differentiability of real functions of one real
variable. Complex-valued differentiable functions, which we will call analytic
functions, have many remarkable and unexpected properties that do not exist
for real-valued differentiable functions. For example, a complex-valued differen-
tiable function necessarily has derivatives of all orders, and many of its properties
are determined by its values on arbitrary sets that have a limit point inside. These
functions are of great importance both in various branches of mathematics and
in many applications. The study of their properties is the main goal of complex
analysis. In this section, we prove a criterion for the differentiability of complex-
valued functions, which includes equivalence to the Cauchy–Riemann equations.
They are a system of two partial differential equations that relate the real and
imaginary parts of a complex-valued function. This leads to the concept of
conjugate harmonic functions in Sect. 2.3. In addition, using some properties
of conjugate harmonic functions, the hydrodynamic interpretation of analytic
functions is given in Sect. 2.4. The chapter ends with Sect. 2.5, which introduces
conformal mappings as analytic functions with a nonzero derivative. It turns out
that a conformal function at a point .z0 preserves angles between curves at .z0 and
equally stretches all curves starting at .z0 . These two properties of a conformal
function are characterized by the argument and the modulus of its derivative at
.z0 , respectively.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 21

T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_2
22 2 Analytic Functions

2.1 Structure of Complex-Valued Functions of a Complex

Variable

Let .Ω be a subset of .C. Consider a function .f : Ω → C. Such functions are called

complex-valued functions of a complex variable. For them, the notation .w = f (z),
.z ∈ Ω, will also be used. To visualize such mappings, we will consider two copies

of the complex plane: .Cz and .Cw . The relationship between the complex variables
.w and .z can be described by two real-valued functions of two real variables:

   
w = f (z) ⇐⇒ u + iv = Re f (x + iy) + i Im f (x + iy)
.

  
u = Re f (x + iy) =: u(x, y),
. ⇐⇒  
v = Im f (x + iy) =: v(x, y).

Example 2.1 Consider the function .ω = z2 − iz, z ∈ C. Separating the real and
imaginary parts in this equality, we get

.u + iv = (x + iy)2 − i(x − iy) = x 2 − y 2 − y + i(2xy − x).

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

Definition 2.1 Let .Ω ⊂ C, .f : Ω → C, .A = α + iβ ∈ C, and .z0 = x0 + iy0 be

a limit point of the set .Ω. We say that the limit of the function f at .z0 is equal to A
as z approaches .z0 , denoted by

. lim f (z) = A,
Ωz→z0

if for any .ε > 0 there is a .δ > 0 such that when .z ∈ Ω and .0 < |z − z0 | < δ, then
|f (z) − A| < ε, i.e.,
.

∀ε > 0 ∃δ > 0 ∀z ∈ Ω :
. 0 < |z − z0 | < δ ⇒ |f (z) − A| < ε.

.

It is very important to understand that .f (z) tends to A no matter what direction

z approaches .z0 . For example, it is easy to show that there is no limit of the
function.f (z) = zz as .z → 0. Indeed, if .z = x ∈ R \ {0}, then .f (x) = xx = 1,
so the limit of f as .z = x → 0 along the real axis is 1. On the other hand, on the
imaginary axis we have .f (iy) = −iy iy = −1, and the limit of f as .z = iy → 0 is
.−1. Since these two limits disagree, the limit of f as .z → 0 does not exist.

Now let us try to understand how the limit of a function f at .z0 is related to the
limits of its real and imaginary parts. The following statement holds.
2.1 Structure of Complex-Valued Functions of a Complex Variable 23

Proposition 2.1 Let .f (z) = u(x, y)+iv(x, y), .z = x +iy ∈ Ω, and .z0 = x0 +iy0
be a limit point of .Ω; .A = α + iβ.
There exists . lim f (z) = A if and only if there exist
Ωz→z0

. lim u(x, y) = α and lim v(x, y) = β.

x→x0 , y→y0 x→x0 , y→y0

The proof follows directly from Definition 2.1 and the equalities

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


|f (z) − A| = (u(x, y) − α)2 + (v(x, y) − β)2 .

Definition 2.2 Let every point of a set .Ω ⊂ C be a limit point of .Ω and .z0 ∈ Ω.

• A function .f : Ω → C is said to be continuous at .z0 if

. lim f (z) = f (z0 ).

Ωz→z0

• A function .f : Ω → C is called continuous in .Ω if it is continuous at every point

of .Ω.

The set of all continuous functions on .Ω is denoted by .C(Ω). 

.

Proposition 2.1 allows transferring some properties of limits of functions of

two real variables to functions of a complex variable. In particular, a function .f
is continuous at a point .z0 = x0 + iy0 if and only if the functions .u and .v are
continuous at the point .(x0 , y0 ). From this we obtain theorems on the continuity of
the sum, product, and division of two continuous functions of a complex variable.
 
Since each function .f : C → C can be viewed as a vector-function . uv from
.R in .R , many properties of such vector-functions are automatically transferred to
2 2

complex-valued functions of a complex variable. Let us recall some of them.

Theorem 2.1 Let .K be a path-connected and compact set (bounded and closed set)
in .C, .ω = f (z), z ∈ K. If .f ∈ C(K), then

1. the modulus of f is bounded on .K, i.e.,

∃M > 0 ∀z ∈ K :
. |f (z)| ≤ M;

2. its modulus takes its minimum and its maximum on .K, i.e.,

.∃ {z1 , z2 } ∈ K ∀z ∈ K : |f (z1 )| ≤ |f (z)| ≤ |f (z2 )|;

24 2 Analytic Functions

3. it is uniformly continuous on .K, i.e., for each .ε > 0 there exists a .δ > 0 such
that

|f (z1 ) − f (z2 )| < ε

. for all {z1 , z2 } ∈ K such that |z1 − z2 | < δ.

Remark 2.1 We will say that a complex-valued function is bounded on a set if its
modulus is bounded on that set.

2.2 Differentiability of Complex-Valued Functions

of a Complex Variable

Hereinafter, we denote by .Ω a domain (open and path-connected set) in the complex

plane .C. If .Ω is a domain in .C, this will be specified.

Definition 2.3 The derivative of a given function .f : Ω → C at a point .z0 ∈ Ω

(denoted by .f (z0 )) is called the limit
.

f (z) − f (z0 )
. lim ,
z→z0 z − z0

provided this limit exists. .

With derivatives there are several alternative notations, for example,

df (z)  f (z0 + Δz) − f (z0 )

f (z0 ) =
.  = lim
dz z=z 0 Δz→0 Δz

From Definition 2.3 and the limit properties, it follows that all formulas for
calculating derivatives that are known in the course of mathematical analysis
(derivative of the sum, product, division and superposition) are transferred to
complex-valued functions of a complex variable.
 
Example 2.2 Let us show that . zn = n zn−1 for any .z ∈ C and for any .n ∈ N.
Using the binomial formula, we deduce
n n n−k
(z + Δz)n − zn k=0 k z (Δz)k − zn
. =
Δz Δz
n(n − 1) n−2
= nzn−1 + z Δz + . . . + (Δz)n−1 .
2
Thus,

 n (z + Δz)n − zn
. z = lim = nzn−1 for all z ∈ C.
Δz→0 Δz
2.2 Differentiability of Complex-Valued Functions of a Complex Variable 25

Definition 2.4 A function .f : Ω → C is said to be differentiable at a point .z0 ∈ Ω

if its increment .Δf (z0 ) := f (z0 + Δz) − f (z0 ) can be represented as

Δf (z0 ) = A · Δz + o(Δz)
. as Δz → 0,

where A is a complex number; in this case, the linear part of the increment is called
the differential of .f at .z0 and is denoted by

def
.df (z0 ) = A · Δz =: A · dz,

and the second term has a higher order of smallness with respect to .Δz and is
denoted by .o(Δz). .

The symbol .o(Δz), pronounced “little oh of .Δz” is one of the Landau symbols
that are used to symbolically express the behavior of some function with respect to
another, in our case with respect to .Δz as .Δz → 0. By definition, one says that
.g(Δz) = o(Δz) as .Δz → 0, if

g(Δz)
. lim = 0. (2.1)
Δz→0 Δz

Remark 2.2 It is obvious that the limit (2.1) is equivalent to

|g(Δz)|
. lim = 0.
|Δz|→0 |Δz|

Thus, .g(Δz) = o(Δz) as .Δz → 0, if and only if .g(Δz) = o(|Δz|) as .|Δz| → 0.

Remark 2.3 Differentiable complex-valued functions of a complex value are

sometimes called complex-differentiable or .C-differentiable. The authors of many
textbooks, and I myself, are of the opinion that there is no need to complicate the
name of the concept when it is clear that complex-valued functions of a complex
variable are being considered.

In the same way as for real functions of a real variable, such a statement is proved.

Proposition 2.2 A function .f : Ω → C is differentiable at a point .z0 ∈ Ω if and

only if there exists .f (z0 ). In addition, .df (z0 ) = f (z0 )dz.

It is easy to see that if a function f is differentiable at a point .z0 , then f is

continuous at .z0 .
As noted above, the continuity of a complex-valued function at a point .z0 = x0 +
iy0 is equivalent to the continuity at the point .(x0 , y0 ) of its real and imaginary parts.
Such a statement does not exist for a differentiable function of a complex variable,
although this statement holds for a complex-valued function of a real variable (see
26 2 Analytic Functions

Proposition 1.2). The connection between a differentiable function of a complex

variable and the differentiability of its real and imaginary parts is established by the
following theorem.

Theorem 2.2 Let .f (z) = u(x, y) + iv(x, y), .z = x + iy ∈ Ω, and .z0 ∈ Ω. The
function f is differentiable at the point .z0 = x0 + iy0 if and only if

(1) the functions u and v are differentiable at the point .(x0 , y0 ),

(2) the Cauchy–Riemann equations are satisfied at .(x0 , y0 ) :

∂u ∂v ∂u ∂v
. (x0 , y0 ) = (x0 , y0 ), and (x0 , y0 ) = − (x0 , y0 ). (2.2)
∂x ∂y ∂y ∂x

Moreover, the derivative of f at .z0 is given by

∂u ∂v ∂v ∂u
f (z0 ) =
. (x0 , y0 ) + i (x0 , y0 ) = (x0 , y0 ) − i (x0 , y0 ). (2.3)
∂x ∂x ∂y ∂y

Proof

Necessity Due to Proposition 2.2 the function f has the derivative at the point .z0 .
Therefore, there exist the limit

f (z0 + Δz) − f (z0 )

f (z0 ) = lim
. , (2.4)
Δz→0 Δz

where .Δz = Δx + iΔy, and we can compute this limit by letting .Δz approach zero
from any direction in the complex plane.
Putting first .Δz = Δx in this limit and taking Proposition 2.1 into account, we
deduce
f (z0 + Δx) − f (z0 )
f (z0 ) = lim
.
Δx→0 Δx

u(x0 + Δx, y0 ) − u(x0 , y0 ) v(x0 + Δx, y0 ) − v(x0 , y0 )
= lim +i
Δx→0 Δx Δx
u(x0 + Δx, y0 ) − u(x0 , y0 ) v(x0 + Δx, y0 ) − v(x0 , y0 )
= lim + i lim
Δx→0 Δx Δx→0 Δx
∂u ∂v
= (x0 , y0 ) + i (x0 , y0 ). (2.5)
∂x ∂x
2.2 Differentiability of Complex-Valued Functions of a Complex Variable 27

On the other hand, if .Δz approaches zero vertically, i.e., .Δz = iΔy in the limit
(2.4), we find

f (z0 + iΔy) − f (z0 )

f (z0 ) = lim
.
iΔy→0 iΔy

u(x0 , y0 + Δy) − u(x0 , y0 ) v(x0 , y0 + Δy) − v(x0 , y0 )
= lim +
Δy→0 iΔy Δy
∂v ∂u
= (x0 , y0 ) − i (x0 , y0 ). (2.6)
∂y ∂y

From (2.5) and (2.6) follow the Cauchy–Riemann equations (2.2) and formulas
(2.3).
It remains to show that the functions u and v are differentiable at the point
.(x0 , y0 ). From Definition 2.4, Proposition 2.2 and (2.2) we get

Δf (z0 ) = Δu(x0 , y0 ) + iΔv(x0 , y0 ) = f (z0 ) · Δz + o(Δz)

.

∂u ∂v
= (x0 , y0 ) + i (x0 , y0 ) (Δx + iΔy) + ε1 + iε2
∂x ∂x

∂u ∂u
= (x0 , y0 ) Δx + (x0 , y0 ) Δy + o (Δx)2 + (Δy)2
∂x ∂y

∂v ∂v
+i (x0 , y0 )Δx + (x0 , y0 )Δy + o (Δx)2 + (Δy)2 ,
∂x ∂y
(2.7)
   
where .ε1 := Re o(Δz) , .ε2 := Im o(Δz) , and based on Remark 2.1 it is easy to
verify that

εk = o(|Δz|) = o
. (Δx)2 + (Δy)2 , k = 1, 2.

Equating the real and imaginary parts in (2.7), we obtain


∂u ∂u
.Δu(x0 , y0 ) = (x0 , y0 )Δx + (x0 , y0 )Δy + o (Δx)2 + (Δy)2 , (2.8)
∂x ∂y

∂v ∂v
Δv(x0 , y0 ) =
. (x0 , y0 )Δx + (x0 , y0 )Δy + o (Δx)2 + (Δy)2 , (2.9)
∂x ∂y

as .Δx → 0, Δy → 0. The relations (2.8) and (2.9) imply that u and v are
differentiable at the point .(x0 , y0 ).
28 2 Analytic Functions

Sufficiency Since the functions .u and .v are differentiable at .(x0 , y0 ), this means
that the relations (2.8) and (2.9) hold. Multiplying (2.9) by the imaginary unit .i and
summing with (2.8), and taking the Cauchy–Riemann equations and Remark 2.1
into account, we derive

Δu(x0 , y0 ) + iΔv(x0 , y0 )
.

∂u   ∂v  
= (x0 , y0 ) Δx + iΔy + (x0 , y0 ) −Δy + iΔx + o(|Δz|)
∂x ∂x
∂u ∂v
= (x0 , y0 ) + i (x0 , y0 ) Δz + o(Δz) as Δz → 0.
∂x ∂x

This relation means that the function .f is differentiable at the point .z0 . Formulas
(2.3) follow from Proposition 2.2 and (2.2).



Remark 2.4 For the first time, the Cauchy–Riemann equations were obtained in
the works of d’Alembert (1752) and Euler (1755) on fluid dynamics. However, the
implications of these conditions in terms of the differentiability of complex-valued
functions of a complex variable were not identified. About 70 years later, in papers
by Cauchy and then by Riemann, a clear definition of the differentiability of such
functions was given.
Theorem 2.2 shows that differentiable functions of a complex variable cannot be
identified with differentiable vector-valued functions from .R2 → R2 (differentia-
bility of the latter is equivalent to differentiability of each component). Obviously,
the set of differentiable functions of a complex variable is narrower than the set of
differentiable vector-valued functions from .R2 → R2 .
This distinction between the two concepts of differentiability leads to the fact that
differentiable functions of a complex variable have significantly different properties.
Because of these properties, the theory of differentiable functions of a complex
variable has wide applications both in various branches of mathematics and directly
in many other areas of natural science.
Theorem 2.2 also highlights that only existence of partial derivatives of real-
valued functions u and v satisfying the Cauchy–Riemann equations at .(x0 , y0 ) does
not ensure differentiability of the function .f = u + iv at .z0 = (x0 , y0 ). The
functions u and v are required to be differentiable at .(x0 , y0 ) as functions on .R2 .
This condition is stronger than existence of partial derivatives.

The Cauchy-Riemann equations provide us with a direct way to test the

differentiability of a function and calculate its derivative.

Example 2.3 Consider the function .f (z) = 32 z − 12 z̄, z ∈ C. It is easy to find that

u(x, y) = Ref (x + iy) = x,

. v(x, y) = Imf (x + iy) = 2y.
2.2 Differentiability of Complex-Valued Functions of a Complex Variable 29

∂u
Since the first Cauchy-Riemann equation is not satisfied .( ∂x =1=  ∂v
∂y = 2), the
function f is not differentiable at any point of the complex plane.
x
However, the corresponding vector-function . : R2 → R2 is differentiable
2y
at each point of .R2 .

Example 2.4 Using Euler’s formula (1.9), the exponential function of a complex
variable is defined as
def
ez = ex+iy = ex eiy = ex cos y + iex sin y
. for all z ∈ C.

The real and imaginary parts of .ez are as follows

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

These real-valued functions are differentiable at each point of .R2 and

∂u ∂v ∂u ∂v
. = ex cos y = , = −ex sin y = − for all (x, y) ∈ R2 .
∂x ∂y ∂y ∂x

Therefore, according to Theorem 2.2, the function .ez is differentiable in the complex
plane and

∂u ∂v
(ez ) =
. +i = ex cos y + iex sin y = ez for all z ∈ C. (2.10)
∂x ∂x

Definition 2.5 A function .f is called analytic in a domain .Ω ⊂ C, if it is

differentiable at every point of this domain.

Definition 2.6 A function f is called analytic at a point .z0 ∈ C, if it is

differentiable in some neighborhood of this point.

Definition 2.7 A function f is called analytic at .∞, if the function .g(z) := f ( 1z )

is analytic at zero.

This definition allows us to consider analytic functions on .C. Note that the notion
of a derivative in ."∞" is meaningless.
The set of all analytic functions in a domain .Ω is denoted by .A(Ω).

Exercise 2.1 Prove that the set .A(Ω) forms a ring with respect to the operations
of adding and multiplying two functions, i.e., it is an Abelian group with respect to
addition, and multiplication distributes over addition.
30 2 Analytic Functions

Example 2.5 Consider a function .f (z) = |z|2 , z ∈ C. Its real and imaginary
parts are .u(x, y) = x 2 + y 2 and .v(x, y) = 0. In addition, . ∂x
∂u
= 2x, . ∂u
∂y = 2y,
.
∂v
∂x = ∂v ∂y = 0. Thus, due to Theorem2.2, the function f is differentiable only at one
point .z = 0 and it is not analytic at any point of the complex plane.

Exercise 2.2 Prove that if f is analytic in a domain and if .|f | is constant there,
then f is constant.

Definition 2.8 An analytic function whose domain is the whole complex plane is
called an entire function.

Examples 2.2 and 2.4 show that .ez and .zn are entire functions.

2.3 Conjugate Harmonic Functions

Let .f ∈ A(Ω) and .f (z) = u(x, y) + iv(x, y), z = x + iy ∈ Ω. In Sect. 5.3 we

will prove that analytic functions have derivatives of all orders. This, in particular,
implies that the functions u and v have partial derivatives of all orders with respect
to x and y. Therefore, without loss of generality, we assume that .{u, v} ⊂ C 2 (Ω)
in this chapter. Then the following relations follow from the Cauchy-Riemann
equations (2.2):
⎧ ⎧
⎨ ∂u = ∂v in Ω, ⎨ ∂2u = ∂2v
∂ 2u ∂ 2u
∂x ∂y ∂x 2 ∂y∂x in Ω,
. ⇒ ⇒ + 2 = 0 in Ω.
⎩ ∂u = − ∂v in Ω, ⎩ ∂2u
= ∂2v
− ∂x∂y in Ω, ∂x 2 ∂y
∂y ∂x ∂y 2

∂ 2v ∂ 2v
Similar calculations show that . + = 0 in .Ω. In the mathematical literature,
∂x 2 ∂y 2
the sum of such second derivatives is called the Laplace operator or Laplacian, for
which the following notation is introduced:

∂ 2v ∂ 2v
Δv :=
. + .
∂x 2 ∂y 2

Definition 2.9 A real-valued function .g ∈ C 2 (Ω) is called harmonic in .Ω if it is a

solution the Laplace equation in .Ω, i.e., .Δg = 0 in .Ω.

Definition 2.10 An ordered pair .h, g of harmonic functions in .Ω satisfying the

Cauchy-Riemann equations is called a conjugate pair of harmonic functions in .Ω.

Thus, if a function .f is analytic in .Ω, then its real and imaginary parts form the
conjugate pair .u, v of harmonic functions in .Ω.
2.3 Conjugate Harmonic Functions 31

Obviously, if .u, v is a conjugate pair of harmonic functions in .Ω, then by

Theorem 2.2 the function .f (z) := u(x, y)+iv(x, y), .z = x +iy ∈ Ω, is analytical
in .Ω. As a result, the following statement follows from the above considerations.

Proposition 2.3 A function .f (z) = u(x, y) + iv(x, y), .z = x + iy ∈ Ω, for

which .{u, v} ⊂ C 2 (Ω), is analytic in .Ω if and only if .u, v is a conjugate pair of
harmonic functions in .Ω.

A natural question arises whether it is possible to restore an analytical function

of a complex variable for a given real or imaginary part, i.e., if u is a real-valued
and harmonic function in .Ω, does there exist an analytic function f in .Ω such that
.Ref = u in .Ω? The answer is given by Proposition 2.3 and the following theorem.

Theorem 2.3 Let u be a harmonic function in a simply connected domain .Ω. Then
there exists a function .v, which is determined up to an additive constant, such that
.u, v is a conjugate pair of harmonic functions in .Ω.

Before the proof, we recall some facts from the course on mathematical analysis
of several variables. A differential form is called an exact form if it is the exterior
derivative of another differential form. A differential form .w = g1 (x, y)dx +
g2 (x, y)dy with coefficients of class .C 1 is an exact form in a simply connected
∂g1 ∂g2
domain .Ω if and only if . = in .Ω. If the last condition is satisfied, then
∂y ∂x
there exists a function .v ∈ C 2 (Ω), which is defined up to an additive constant, such
that
∂v ∂v
dv = w ⇐⇒
. dx + dy = g1 dx + g2 dy.
∂x ∂y

∂u ∂u
Proof Consider the differential form .w = − dx + dy. Taking into account
∂y ∂x
that u is harmonic, we have

∂ ∂u ∂ 2u ∂ 2u ∂ ∂u
. − =− 2
= 2
= in Ω.
∂y ∂y ∂y ∂x ∂x ∂x

Since .Ω is simply connected, the differential form w is exact. This means that there
is a function .v ∈ C 2 (Ω), which is defined up to an additive constant, such that
.dv = w in .Ω, i.e.,


∂v ∂v ∂u ∂u ∂v
= − ∂u
. dx + dy = − dx + dy in Ω ⇐⇒ ∂x ∂y
in Ω.
∂x ∂y ∂y ∂x ∂v
∂y = ∂u
∂x
32 2 Analytic Functions

Thus, the Cauchy-Riemann equations in the domain .Ω are satisfied for the functions
u and .v. In addition, from the last relations it follows that
.

⎧
⎨ ∂2v = − ∂2u ,
.
∂x 2 ∂x∂y
⇒ Δv = 0 in Ω.
⎩ ∂ 2 v2 = ∂ 2 u ,
∂y ∂y∂x

Therefore, .u, v is a conjugate pair of harmonic functions in .Ω. 



Example 2.6 Find an analytic function in .C with the real part

u(x, y) = x 2 − y 2 − x.
.

Since u belongs to .C 2 (R2 ) and .Δu = 2 − 2 = 0 in the simply connected domain

.Ω = C, by Theorem 2.3, there is a harmonic function v such that .f = u + iv is

analytic in .C.
It follows from the first equation in (2.2) that

∂v ∂u
. = = 2x − 1.
∂y ∂x

Therefore,

v(x, y) = 2xy − y + ϕ(x),

.

∂v
where .ϕ is some differentiable function. Then, . ∂x = 2y +ϕ (x) and from the second
equation in (2.2) we get

∂v ∂u
. − 2y − ϕ (x) = − = = −2y ⇒ ϕ (x) = 0 ⇒ ϕ ≡ c,
∂x ∂y

where .c ∈ R. Thus, .v(x, y) = 2xy − y + c and

f (x + iy) = x 2 − y 2 − x + i(2xy − y + c) = (x + iy)2 − (x + iy) + ic = z2 − z + ic.

.

Exercise 2.3 Prove that if f and .f are analytic functions in .C, then f is constant
in .C.

Many properties of harmonic functions resemble those of analytic functions and

will be discussed in the following chapters of this textbook. Harmonic functions
occur in problems of electric, magnetic and gravitational potentials, in problems of
steady-state temperatures and in problems of hydrodynamics.
2.4 Hydrodynamic Interpretation of Analytical Functions 33

2.4 Hydrodynamic Interpretation of Analytical Functions

Let .f ∈ A(Ω) and .f (z) = u(x, y) + iv(x, y), z = x + iy ∈ Ω. From results

obtained in Sect. 2.3 it follows that .u, v is a conjugate pair of harmonic functions
in .Ω. We define a plane-parallel vector field

∂u ∂u
.V(x, y) := (x, y), (x, y), 0 , (x, y) ∈ Ω;
∂x ∂y

in this case, the function .u is called the potential of the vector field .V, and the
function .f is called the complex potential of .V.
It is easy to verify that

∂ ∂u ∂ ∂u
div V =
. + = Δu = 0 in Ω.
∂x ∂x ∂y ∂y

This means that the vector field .V is solenoidal (or incompressible) in .Ω (no sources
and drains).
Calculating
 
 i j k 
 
 ∂ ∂ ∂  ∂ ∂u ∂ ∂u
.curl V =  ∂x ∂y ∂μ  = 0, 0, − = 0 in Ω,
 ∂u ∂u  ∂x ∂y ∂y ∂x
 ∂x ∂y 0 

we conclude that the vector field .V is irrotational (or curl-free) in .Ω.

Thus, every analytical function f in a domain .Ω is the complex potential of

some plane-parallel, solenoidal, and irrotational vector field in .Ω.

Now let a solenoidal and irrotational vector field

 
V = ϕ1 (x, y), ϕ2 (x, y), 0
.

be given in a simply connected domain .Ω. We assume that the functions .ϕ1 and .ϕ2
belong to the space .C 1 (Ω) and .V = 0 in .Ω.
Since .V is irrotational,
 
 i j k 
 
 ∂ ∂ ∂  ∂ϕ1 ∂ϕ2
.curl V = 0 ⇐⇒  =0 ⇒ = in Ω.
 ∂x ∂y ∂μ  ∂y ∂x
 ϕ1 ϕ2 0 
34 2 Analytic Functions

Therefore, taking into account the simply connectedness of the domain .Ω, one can
assert that the differential form .w1 = ϕ1 dx + ϕ2 dy is an exact form in .Ω. Then, due
to Theorem 2.3 there exists a function .u ∈ C 2 (Ω) such that .du = w1 . This means
that
∂u ∂u
. = ϕ1 , = ϕ2 in Ω, (2.11)
∂x ∂y

i.e., u is the potential of the vector field .V.

Since .V is solenoidal,

∂ϕ1 ∂ϕ2 ∂ϕ2 ∂ϕ1

div V = 0
. ⇒ + = 0 ⇐⇒ − = in Ω.
∂x ∂y ∂y ∂x

Thus, the differential form .w2 = −ϕ2 dx + ϕ1 dy is an exact form in .Ω, which
means there is a function .v ∈ C 2 (Ω) such that .dv = w2 , whence

∂v ∂v
. = −ϕ2 , = ϕ1 in Ω. (2.12)
∂x ∂y

It follows from (2.11) and (2.12) that the functions u and v satisfy the Cauchy–
Riemann equations in .Ω. Therefore, one can determine a function .f := u + iv,
which, based on Theorem 2.2, will be analytic in .Ω.

Therefore, any plane-parallel, solenoidal, irrotational vector field in a simply

connected domain can be associated with an analytical function, which is the
complex potential for that field.

Let us see what v means in physical terms. Consider a curve that is implicitly
given by the equation .v(x, y) = const. Then according to (2.12),

∂v
dy ϕ2
. = − ∂x
∂v
= .
dx ∂y
ϕ1
   
dy
This means that the vector . 1, dx , 0 = 1, ϕϕ21 , 0 , which is the tangent vector to
the curve .v(x, y) = const, is collinear to .V = (ϕ1 , ϕ2 , 0). Consequently, the curve
.v(x, y) = const is the motion trajectory of particles of a fluid flow, the velocity

vector of which coincides with .V. The function .v is called the stream function of
the vector field .V.
2.5 Conformal Mappings 35

2.5 Conformal Mappings: Geometric Meaning of the Modulus

and Argument of the Derivative

Here, we take a closer look at the mapping properties of analytic functions.

Definition 2.11 A function .f : Ω → C at a point .z0 ∈ Ω is called conformal at .z0

if f is analytic at .z0 and .f (z0 ) = 0. The function f is said to be conformal in the
domain .Ω if it is analytic in .Ω and conformal at every point of .Ω. .

Let .f be conformal at a point .z0 ∈ C. Then for any smooth simple curve .z =
γ(t), t ∈ [a, b], with the origin at .z0 = γ(a), the limit

|f (z) − f (z0 )|
. lim = |f (z0 )| = 0 (2.13)
z→z0 , z∈Eγ |z − z0 |

exists, since f has the derivative at .z0 and the limit does not depend on how z tends
to .z0 . Thus, this limit is independent of .γ. On the other hand,

|f (z) − f (z0 )| |f (γ(t)) − f (γ(a))|

. lim = lim ,
z→z0 , z∈Eγ |z − z0 | t→a |γ(t) − γ(a)|

and this limit can be interpreted as a stretch coefficient (a scale factor) of the curve
γ at the point .z0 under the mapping f .
.

Hence, a conformal function at .z0 stretches equally any smooth simple curve
emanating from .z0 , and the equality (2.13) expresses the geometric meaning of the
modulus of the derivative .f (z0 ): this is the stretch coefficient at the point .z0 under
the mapping f .
This mapping property of the conformal function f can be commented as
follows: .f stretches small circles centered at the point .z0 of radius .|Δz| = |z − z0 |
in circles with center at .ω0 = f (z0 ) of radius .|f (z0 )| |Δz| up to a value .o(|Δz|)
(Fig. 2.1). Indeed,

|ω − ω0 | = |f (z) − f (z0 )| = |f (z0 ) Δz + o(Δz)| ∼ |f (z0 )| |Δz| as |Δz| → 0.

.

Definition 2.12 A mapping f is called a mapping with equal stretch at a point .z0 if
it stretches equally any smooth simple curve outgoing from .z0 , i.e., for any smooth
simple curve .z = γ(t), t ∈ [a, b], with the origin at .z0 = γ(a), the limit

|f (γ(t)) − f (γ(a))|
. lim (2.14)
t→a |γ(t) − γ(a)|

exists, is independent of .γ, and does not equal to zero.

36 2 Analytic Functions

Fig. 2.1 Mapping small circles by a conformal mapping f

Example 2.7 The function .f (z) = 2 z is not analytic at any point of the complex
plane. However, its stretch coefficient of any smooth simple curve equals 2. Indeed,

|f (γ(t)) − f (γ(a))| |γ(t) − γ(a)| |γ(t) − γ(a)|

. lim = 2 lim = 2 lim = 2.
t→a |γ(t) − γ(a)| t→a |γ(t) − γ(a)| t→a |γ(t) − γ(a)|

Example 2.8 It is easy to verify that the stretch coefficient of any smooth simple
curve under the mapping .f (z) = x + i2y at any point of the complex plane is
equal to 1 in the horizontal direction, and it is 2 in the vertical direction. Hence, this
function is not a mapping with equal stretch at any point of the complex plane.

Example 2.9 Let .f be conformal at a point .z0 = x0 + iy0 ∈ C .(f (z) = u(x, y) +
iv(x, y)). Then the Jacobian of the corresponding vector-valued function

u
. : Ω −→ R2
v

at the point .(x0 , y0 ) in view of the Cauchy-Riemann equations is equal to

 
 ∂u ∂u  2 2
  ∂u ∂v
J (u, v)|(x0 ,y0 )
. =  ∂x ∂y
= + = |f (z0 )|2 > 0. (2.15)
 ∂x
∂v ∂v
∂y
 ∂x ∂x

It
uis known from vector calculus that the Jacobian .J of the vector-valued function
.
v is the linear stretch coefficient of infinitesimal areas.

From example 2.9 and the theorem on the preservation of a domain under a
continuously differentiable mapping .f : Rn → Rn , whose Jacobian is not equal
to zero, the following statement follows.

Proposition 2.4 (Open Mapping Property) A conformal function f in a domain

Ω maps this domain into a domain in .C.
.
2.5 Conformal Mappings 37

A stronger statement is proved in Sect. 9.1. Example 2.9 and the inverse function
theorem for a continuously differentiable mapping .f : Rn → Rn , whose Jacobian is
not equal to zero, lead to the following statement.

Proposition 2.5 A conformal function f at a given point .z0 is a one-to-one

mapping in some neighborhood of .z0 . In addition, the inverse function is conformal
at the point .w0 = f (z0 ) and by the chain rule

 −1  1
.f (w0 ) = . (2.16)
f (z0 )

The question of how to explicitly find the inverse function is discussed in Sect. 9.2.
Let us now clarify the geometric meaning of the argument of the derivative of a
conformal function .f at a point .z0 .
It is known that for a smooth curve .z = γ(t) = x(t) + iy(t), t ∈ [a, b], with
.γ(a) = z0 , the value .γ (a) = (x (a), y (a)) is the tangent vector to the curve at .z0 ,

and .arg(γ (a)) is the angle of inclination of this vector to the real axis. Then for the
curve .γ(t) = f (γ(t)), .t ∈ [a, b], we have


.γ (a) = f (z0 ) · γ (a) ⇒ Arg
γ (a) = arg f (z0 ) + arg γ (a). (2.17)

Equality (2.17) is understood as follows: one of the arguments of the complex

number  .γ (a) is equal to the sum .arg(f (z0 )) + arg(γ (a)).

From (2.17) it follows that .arg(f (z0 )) is the angle by which you need to rotate
the tangent vector to .γ to get the angle of inclination of the tangent vector to the
image of this curve at the point .ω0 = f (z0 ), in other words: .arg(f (z0 )) is the angle
of rotation of an arbitrary smooth curve emanating from .z0 under the mapping .f
(Fig. 2.2).
Consider another smooth curve .z = μ(t), .t ∈ [a, b], emanating from .z0 . By
.
μ we denote its image under the mapping .f, i.e., . μ(t) = f (μ(t)), .t ∈ [a, b].
Similarly, we deduce


μ (a) = f (z0 ) · μ (a)
. ⇒ Arg 
μ (a) = arg f (z0 ) + arg μ (a). (2.18)

Fig. 2.2 Angle of rotation of a curve .γ under a conformal mapping f

38 2 Analytic Functions

It follows from (2.17) and (2.18) that

(
. μ,
γ)|ω0 := Arg 
μ (a) − Arg
γ (a)
= arg(μ (a)) − arg(γ (a)) =: (μ, γ)|z0 . (2.19)

Here, the symbol .(μ, γ)|z0 denotes the oriented angle between the curves .μ and .γ
at .z0 ; by definition, it is equal to the oriented angle between the vectors .μ (a) and
.γ (a), i.e., .arg(μ (a)) − arg(γ (a)).

Equality (2.19) means that the angle between the curves .μ and .γ at .z0 is equal to
the angle between their images under the conformal mapping f both in magnitude
and in the direction of readout. These properties of f are called angle-preserving
and orientation-preserving at the point .z0 .

Example 2.10 The function from Example 2.7 is angle-preserving, but is not
orientation-preserving. It reflects any smooth simple curve across the x-axis and
then stretches it by 2.
The function from Example 2.8 is not angle-preserving, but preserves the
x
orientation since the Jacobian of the vector-valued function . 2y : R2 → R2 is
positive.

Example 2.11 Let .f (z) = z2 . Then .f (z) = 2z and .f (0) = 0. Thus, f in not
conformal at 0. Let us show that f is not angle-preserving at 0.
Indeed, consider two curves (their traces are segments) emanating from the
origin:

γ1 (t) = t eiα
. and γ2 (t) = t eiβ , t ∈ [0, 1].

Then, .(γ2 , γ1 )|z=0 = β − α. But the angle between their images


.γ1 (t) = t e
2 i2α
and 
γ2 (t) = t 2 ei2β , t ∈ [0, 1],

is equal to .(
γ2 ,
γ1 )|ω=0 = 2(β − α).

Let us summarize the above, proving the main theorem characterizing conformal
mappings.

Theorem 2.4 (Conformal Mapping Criterion) Let .f : Ω → C, .f (z) =

u(x, y) + iv(x, y), .z = x + iy ∈ Ω. The function f is conformal in the domain .Ω
if and only if

(1) the real-valued functions u and v are differentiable in .Ω;

(2) .f is a mapping with equal stretch, angle-preserving and orientation-preserving
at any point of .Ω.
2.5 Conformal Mappings 39

Proof The necessity follows from the considerations above in this paragraph. Let
us prove the sufficiency. By Definition 2.12, the limit (2.14) exists, does not depend
on a curve and is not equal to zero for each point .z0 = x0 + iy0 of the domain .Ω.
Then the following limit

|Δf (z0 )|2 |Δu + iΔv|2 (Δu)2 + (Δv)2

. lim = lim = lim =: K = 0,
Δz→0 |Δz|2 Δz→0 |Δx + iΔy|2 Δz→0 (Δx)2 + (Δy)2

exists, where the symbol .Δ refer to corresponding increments (see (2.8), (2.9)).
Putting first .Δz = Δx .(Δy = 0) in this limit and taking into account the first
condition of the theorem, we find

(u(x0 + Δx, y0 ) − u(x0 , y0 ))2 + (v(x0 + Δx, y0 ) − v(x0 , y0 ))2

K = lim
.
Δx→0 (Δx)2
 ∂u 2  ∂v 2
= (x0 , y0 ) + (x0 , y0 ) . (2.20)
∂x ∂x
By the same way we find the limit in the case .Δx = 0:
 ∂u 2  ∂v 2
. K= (x0 , y0 ) + (x0 , y0 ) . (2.21)
∂y ∂y

In the case .Δx = Δy we get

∂u ∂u ∂v ∂v
K=K+
. (x0 , y0 ) (x0 , y0 ) + (x0 , y0 ) (x0 , y0 ). (2.22)
∂x ∂y ∂x ∂y

The Eqs. (2.20), (2.21), and (2.22) lead to the system

⎧       
⎨ ∂u 2 + ∂v 2 = ∂u 2 + ∂v 2
∂x ∂x ∂y ∂y
.
⎩ ∂u ∂u + ∂v ∂v
=0
∂x ∂y ∂x ∂y

Since .K = 0, one of the derivatives in this system is not equal to zero. We can
∂y (x0 , y0 ) = 0. Then
regard that . ∂v

∂u ∂u
∂v ∂x (x0 , y0 ) ∂y (x0 , y0 )
. (x0 , y0 ) = − ∂v
. (2.23)
∂x ∂y (x0 , y0 )

Substituting this expression in the first equality of the system, we find

∂u ∂v
. (x0 , y0 ) = ± (x0 , y0 ). (2.24)
∂x ∂y
40 2 Analytic Functions

Fig. 2.3 An angle-preserving map f

Thus, we conclude from (2.23) and (2.24) that either the Cauchy-Riemann
equations

∂u ∂v ∂u ∂v
. = and =− in Ω (2.25)
∂x ∂y ∂y ∂x

hold or opposite to them, namely

∂u ∂v ∂u ∂v
. =− and = in Ω. (2.26)
∂x ∂y ∂y ∂x

Let us draw from the point .z0 two segments parallel to the coordinate axes, which
belong to the domain .Ω (Fig. 2.3). Recall that .z0 = (x0 , y0 ) is arbitrary point from
.Ω.

Since .f is angle-preserving, the images of these segments will be perpendicular

curves emanating from .ω0 = f (z0 ). The vectors
 ∂u ∂v   ∂u ∂v 
. (x0 , y0 ), (x0 , y0 ) and (x0 , y0 ), (x0 , y0 )
∂x ∂x ∂y ∂y

are the tangent vectors to these curves at .ω0 , respectively.

Turning the vector .( ∂x , ∂x ) at the angle . π2 counterclockwise and taking into
∂u ∂v

account that f is orientation-preserving, we find that the directions of the vectors

∂v ∂u ∂u ∂v
.(−
∂x , ∂x ) and .( ∂y , ∂y ) coincide (obviously, these are nonzero vectors). Therefore,

∂v ∂u ∂u ∂v ∂u ∂v ∂v ∂u
0< −
. , · , = − = J (u, v). (2.27)
∂x ∂x ∂y ∂y ∂x ∂y ∂x ∂y

If Eq. (2.26) are satisfied, then from (2.27) we get a contradiction:

∂u ∂v ∂v ∂u
0<
. (x0 , y0 ) (x0 , y0 ) − (x0 , y0 ) (x0 , y0 )
∂x ∂y ∂x ∂y
2 2
∂v ∂v
=− (x0 , y0 ) − (x0 , y0 ) < 0.
∂y ∂x
2.5 Conformal Mappings 41

Thus, the Cauchy-Riemann equations (2.25) hold. Then, taking into account the
first condition of the theorem and Theorem 2.2, we conclude that .f ∈ A(Ω). From
(2.27), based on (2.15), it follows that

∂u ∂v ∂v ∂u  ∂u 2  ∂v 2
0 < J (u, v) =
. − = + = |f (z)|2 .
∂x ∂y ∂x ∂y ∂x ∂x

This means that f is conformal in .Ω. 



We now give a definition of the conformality of a function in the case when a

point or its image is the point at infinity.

Definition 2.13 Let .z0 = ∞ and  .f (z0 ) = ∞. A function .f is said to be conformal
at .z0 if the function .g(z) := f 1z is conformal at .0.
Let .z0 = ∞ and .f (z0 ) = ∞. A function .f is said to be conformal at .z0 if the
function .g(z) := f 1(z) is conformal at .z0 .
Let .z0 = ∞ and .f (z0 ) = ∞. A function .f is said to be conformal at .z0 if the
function .g(z) := 11 is conformal at .0.
f(z)

Definition 2.14 An analytic function .f : Ω → C is called univalent in .Ω if it is

injective in the domain .Ω, i.e., for any two different points .z1 and .z2 from .Ω we
have that .f (z1 ) = f (z2 ).
In this case, .Ω is referred to as the domain of univalence of f . .

One of the most important properties of a univalent function is the following: if

f : Ω → C is univalent, then the derivative of f is never zero, i.e., f is conformal
.

in .Ω. This statement will be proved in Theorem 7.8. It should be noted here that this
statement does not hold for real-valued smooth functions, e.g., .f (x) = x 3 , x ∈ R.
The main theorem on conformal mappings is the following Riemann theorem,
the proof of which will be presented in Sect. 9.5.

Theorem 2.5 (Riemann Mapping Theorem) Let .Ω and G be two arbitrary

simply connected domains in .C whose boundaries contain more than one point.
Then for arbitrary points .z0 ∈ Ω and .ω0 ∈ G and any real number .α ∈ (−π, π ]
there exists a unique bijective conformal mapping .f : Ω −→ G such that .ω0 =
f (z0 ) and .arg f (z0 ) = α.

At first glance, the statement of this theorem seems implausible. Simply con-
nected domains in the complex plane can be very complicated. For instance, there
are bounded domains such that the boundary is a nowhere-differentiable fractal
curve of infinite length. And the fact that such a domain can be mapped onto a
regular unit disk in an angle-preserving manner sounds counterintuitive.
Therefore, conformal mappings are invaluable for solving problems in engineer-
ing and physics that can be expressed in terms of functions of a complex variable,
42 2 Analytic Functions

such as for example, boundary value problems involving Laplace’s equation in

complicated domains. By choosing an appropriate mapping, the inconvenient
geometry of such a domain can be transformed into a much more convenient one.
To summarise what has been said so far in this section, conformal mappings
preserve the shape of any sufficiently small figure, possibly rotating and scaling it
(but not reflecting it).
Elementary Analytic Functions
3

Abstract

Conformal mappings are of immense importance in various branches of math-

ematics and in many applications. To solve many problems, one needs to be
able to construct a bijective conformal mapping from one domain onto another
in the complex plane. In this chapter we study how to construct such bijective
conformal mappings. We will consider various elementary analytic functions,
find domains of univalence and images of these domains. In addition, for many
elementary analytic functions in .C we find their inverses, which in some cases
turn out to be multivalued. We introduce the first (intuitive) concept of a Riemann
surface for multivalued functions and show how to construct Riemann surfaces
for the inverses of elementary analytic functions. As a result of these studies,
we will establish facts that are incorrect in real analysis, for example, we can
calculate the logarithm of negative numbers and solve the equation

. sin z = 2.

3.1 Linear and Fractional-Linear Functions and Their Simplest

Properties

Since the largest domain in the complex plane .C is .C itself, it is natural to start our
study of conformal mappings by considering the analytic functions from .C to itself,
which are one-to-one and onto.

Definition 3.1 A function of the form

w = az + b,
.

where .{a, b} ⊂ C and .a = 0, is called a linear function of a complex variable.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 43

T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_3
44 3 Elementary Analytic Functions

Since .w  = a = 0, a linear function is entire, conformal and univalent in the

complex plane .C. One can easily check that

w−b
z=
.
a

is the inverse linear mapping from .C onto .C. In Sect. 9.3, we will prove the
following statement: every conformal and univalent mapping from .C onto .C is
a linear function.
It is obvious that .limz→∞ (az + b) = ∞. This suggests that each linear function
can be determined at the point at infinity by setting it equal to .∞. Let us show
the conformality of a linear function in .∞. In accordance with Definition 2.13, one
should consider the function
1 z
g(z) =
. =
az−1 + b a + bz

and check its conformality at zero: .g  (0) = a1 = 0.

Having written the number .a in the exponential form .a = |a| · ei arg(a) , the
function .w = az + b can be represented as

w = ei arg(a) |a| z + b,
.

whence it is visible that each linear function is the composition of three mappings:

(1) .ξ = |a|z (homothety centered at the origin and with ratio .|a|);
(2) .τ = ei arg(a) ξ (rotation around the origin at the angle .arg(a));
(3) .w = τ + b (translation by the vector .b).

Example 3.1 Figure 3.1 shows that the square

K1 = {z = x + iy : x ∈ (0, 1), y ∈ (0, 1)}

.

is mapped by the linear function

√ √  √ π √
ω=
. 2 + i 2 z − i 2 = 2ei 4 z − i 2
√
onto the square .{ω = u + iv : |u| + |v| < 2}. .
3.1 Linear and Fractional-Linear Functions and Their Simplest Properties 45

Fig. 3.1 The image of the

square .K1

Also in Sect. 9.3, we will prove that conformal and univalent mappings from .C
onto .C are a special class of functions of the form

az + b
w=
. , {a, d, b, c} ⊂ C.
cz + d

One additional restriction .ad − bc = 0 is needed to ensure that such a function is

neither identically constant nor meaningless. It turns out that such functions have
many interesting properties. We will explore them in detail in this and the next three
sections.

Definition 3.2 A fractional-linear function is a mapping of the form

az + b
w=
. ,
cz + d

where .{a, d, b, c} ⊂ C and .ad − bc = 0.

When the coefficient .c = 0, then the fractional-linear function becomes linear, some
of the properties of which were studied above. Therefore, in this section we assume
that .c = 0. Since

az + b a az + b
. lim = and lim = ∞,
z→∞ cz + d c z→− dc cz + d
46 3 Elementary Analytic Functions

it is naturally to extend the fractional-linear function to .C by continuity as follows:

⎧ az + b
⎪
⎪ , z ∈ C \ {− dc },
⎪
⎨ cz + d
w = F(z) :=
. ∞, z = − dc ,
⎪
⎪
⎪
⎩a , z = ∞.
c

Remark 3.1 Thus, every fractional-linear function is a continuous mapping from

C in .C.
.

Theorem 3.1 A fractional-linear function is a homeomorphism from .C onto .C.

Proof Recall that a homeomorphism is, by definition, a one-to-one, onto continuous

mapping with a continuous inverse. First, let us show that .F is bijection, namely

∀w ∈ C ∃!z ∈ C :
. F(z) = w.

To prove this, it suffices to show that the equation .F(z) = w for .z has only one root
in .C. Indeed,

az + b −dw + b a
. =w ⇒ z= if w ∈ C \ ;
cz + d cw − a c

in addition, if .w = ∞, then .z = − dc ; and if .w = ac , then .z = ∞.

Thus, for each fractional-linear mapping there is an inverse which is also a
fractional-linear mapping determined by the formula
⎧
⎪ −dw + b
⎪
⎪ , w ∈ C \ { ac },
⎪
⎨ cw − a
−1
.F (w) := ∞, w = ac , (3.1)
⎪
⎪
⎪
⎪ d
⎩− , w = ∞.
c
Remark 3.1 completes the proof. 

Theorem 3.2 A fractional-linear function is conformal in .C.

Proof If .z ∈ {− dc , ∞}, then

a(cz + d) − c(az + b) ad − bc
.w = = = 0.
(cz + d)2 (cz + d)2
3.2 Group and Circular Properties of Fractional-Linear Functions 47

If .z = − dc , then according to Definition 2.13 one should consider the function

.g(z) =
F(z) = az+b and find its derivative at .z = − c :
1 cz+d d

cb − ad c2
g  (z) =
. = = 0.
(az + b)2 z=− dc cb − ad

If .z = ∞, then one should consider the function

1
a
z +b a + bz
g(z) := F
. = =
z c
z +d c + dz

and find its derivative at .z = 0 :

bc − ad bc − da
g  (z) =
. = = 0.
(c + dz)2 z=0 c2

3.2 Group and Circular Properties of Fractional-Linear

Functions

By .Λ we denote the set of all fractional-linear mappings and define a binary

operation on this set as the composition of fractional-linear mappings:
 
F2 ◦ F1 := F2 F1 .
.

Exercise 3.1 Let there be given two fractional-linear mappings

a1 z + b1 a2 ξ + b2
ξ = F 1 (z) =
. and w = F 2 (ξ ) = ,
c1 z + d1 c2 ξ + d2

where .a1 d1 − c1 b1 = 0 and .a2 d2 − c2 b2 = 0. Prove that their composition is a

fractional-linear mapping

  az + b
w = F2 ◦ F1 (z) =
. ,
cz + d

where

ab a b a b
. = 2 2 · 1 1 , ad − bc = 0.
cd c2 d2 c1 d1

Theorem 3.3 (The Group Property) The set .(Λ, ◦) is a noncommutative group.
48 3 Elementary Analytic Functions

Proof Let us verify the group axioms:

• associativity
   
∀ {F1 , F2 , F3 } ⊂ Λ :
. F3 ◦ F2 ◦ F1 = F3 ◦ F2 ◦ F1

follows from Exercise 3.1 and associativity of matrix multiplication;

• the identity mapping .E(z) = z is the identity element in this group:

F ◦ E = E ◦ F = F;
.

• the existence of the inverse element follows from (3.1).

To check non-commutativity, consider two fractional-linear mappings: .F1 (z) =

1
z and .F2 (z) = z + 2. Then

1 1
F2 ◦ F1 =
. + 2 = F1 ◦ F2 = .
z z+2

The theorem is proved. 

Theorem 3.4 (The Circular Property) Each fractional-linear function maps a

circle in .C onto a circle in .C.

It should be noted here that by a circle in .C we mean either a circle in .C or a line

in .C together with the point .{∞}.

Proof Each fractional-linear function .w = F(z) = az+b

cz+d , (ad − bc = 0, .c = 0)
.

can be represented as follows

a
c (cz + d) − ad
c +b a ad − bc B
F(z) =
. = − =: A + ,
cz + d c c2 (z + dc ) z+C

where
a ad − bc d
, B=−
A=
. , C= .
c c2 c
 
Thus, .F(z) = F3 F2 (F1 (z)) , where

1
F1 : z → z + C,
. F2 : z → , F3 : z → A + Bz, z ∈ C.
z
3.2 Group and Circular Properties of Fractional-Linear Functions 49

The functions .F1 and .F3 are linear and map circles onto circles, since every linear
function is a composition of homothety, rotation and translation (see Sect. 3.1). So,
it remains to prove the theorem statement for .F2 .
A general equation of a circle in the coordinate plane is

E(x 2 + y 2 ) + F1 x + F2 y + G = 0,
.

where .{E, F1 , F2 , G} ⊂ R, E 2 + F12 + F22 + G2 = 0. It covers both the equation

of a (standard) circle and the equation of a straight line, as well as a point (a circle
with zero radius) and the empty set.
Since .x = 12 (z + z) and .y = 2i1 (z − z) and .x 2 + y 2 = zz, the equation of a circle,
written in a complex variable, has the form

Ezz + F z + F z + G = 0,
.

where .F = 12 (F1 − iF2 ), .F = 12 (F1 + iF2 ). The mapping .w = 1

z transforms a
circle into a curve whose equation is

1 1 1 1
.E · + F + F + G = 0 ⇐⇒ Gw w + F w + F w + E = 0,
w w w w
but this is the equation of a circle in the complex plane. 

It turns out that a fractional-linear function is uniquely determined by the images

of any three different points in .C.

Theorem 3.5 There is only one fractional-linear function .F that maps three
different given points .{z1 , z2 , z3 } ⊂ C to three different given points .{w1 , w2 , w3 } ⊂
C, i.e., .F(zk ) = wk for .k ∈ {1, 2, 3}.
This fractional-linear mapping is defined by the formula

w − w1 w3 − w2 z − z 1 z 3 − z2
. · = · . (3.2)
w − w2 w3 − w1 z − z 2 z 3 − z1

Proof

1. Let us first consider the case when all the given complex numbers are finite, i.e.,
.{z1 , z2 , z3 } ⊂ C and .{w1 , w2 , w3 } ⊂ C.

Having found .w from (3.2), we check that (3.2) defines a fractional-linear

function .F. In addition, it is easy to verify that .F(zk ) = wk , .k ∈ {1, 2, 3}. Let us
prove the uniqueness of such a function.
50 3 Elementary Analytic Functions

Assume that there is another fractional-linear mapping .F1 such that .F1 (zk ) =
wk for .k ∈ {1, 2, 3}. Then, based on Theorem 3.3, we get the following three
relations:
  azk + b
F(zk ) = F1 (zk ) ⇐⇒ F−1
.
1 F(zk ) = zk ⇐⇒ = zk (3.3)
czk + d

for .k ∈ {1, 2, 3}. Here, .{a, b, c, d} ⊂ C and .ad − cd = 0. It follows from

(3.3) that the quadratic equation .cz2 + (d − a)z + b = 0 or the linear equation
.(d − a)z + b = 0 (if .c = 0) has three different roots .z1 , z2 , z3 :

czk2 + (d − a)zk + b = 0
. for all k ∈ {1, 2, 3}. (3.4)

This is possible only when .c = 0, d = a, b = 0, i.e.,

 
F−1
.
1 F(z) = z (∀ z ∈ C) ⇐⇒ F1 = F.

2. If one of the points .{z1 , z2 , z3 } and .{w1 , w2 , w3 } coincides with the point .{∞},
the corresponding numerator and denominator in (3.2), where this point appears,
must be replaced by .1 and then repeat the previous reasonings. For example,
if .z1 = ∞ and .w3 = ∞, then the corresponding fractional-linear function is
represented by the formula
w1
w − w1 1 − − 1 z 3 − z2
z
w3 z1 w − w1 z 3 − z2
. · = · ⇐⇒ = .
w − w2 1 − w1
w3 z − z2 zz31 − 1 w − w2 z − z2

The theorem is proved.



Remark 3.2 The expression

z − z1 z 3 − z2
. ·
z − z2 z 3 − z1

is called the cross-ratio of four points .z, .z1 , .z2 , .z3 , and the equality (3.2) means
invariance of the cross-ratio of four points under a fractional-linear mapping.

Remark 3.3 It follows from the proof of Theorem 3.5 that a fractional-linear map
.F = E can have no more than two fixed different points .z1 , z2 , i.e., .F(zk ) = zk ,
.k = 1, 2. In this case, thanks to (3.2) this mapping it is given by the formula

w − z1 z − z1
. =A
w − z2 z − z2
3.2 Group and Circular Properties of Fractional-Linear Functions 51

if .z1 = ∞ and .z2 = ∞, or .w − z1 = A(z − z1 ) if .z2 = ∞. Here the coefficient

A ∈ C \ {1}.
.

Depending on the coefficient .A, such mappings are called hyperbolic fractional-
linear mappings if .A is positive and .A = 1; elliptic if .A = eiθ , .θ = 2π n; and
loxodromic if .A = |A|eiα , .|A| = 1, .α = 2π n, where .n ∈ Z.
A fractional-linear function is called a parabolic fractional-linear mapping if it
has only one fixed point, i.e., the quadratic Eq. (3.4) has zero discriminant (fixed
points are coincided).

Exercise 3.2 Prove that every parabolic fractional-linear mapping can be repre-
sented in the form
1 1
. = +β
w − z1 z − z1

if .z1 = ∞, or .w = z + β if .z1 = ∞. Here, .β ∈ C, β = 0.

Corollary 3.1 Let .γ1 and .γ2 be two circles in .C. Then there exists a fractional-
linear mapping .F such that .γ2 = F(γ1 ).

To prove this, one needs to take three different points on one and the other circle,
and then use the formula (3.2) and Theorem 3.4.

Corollary 3.2 Let .B1 and .B2 be two disks in .C. Then there exists a fractional-linear
mapping .F such that .B2 = F(B1 ).

Proof Recall that by a disk in .C we mean either a disk in .C or its exterior in .C or a

half plane together with the point .{∞}.
Let us take three different finite points .z1 , z2 , z3 on the boundary of the disk
.B1 so that when they are successively traversed from .z1 trough .z2 to .z3 , the disk

remains to the left. By the same way we choose three different points .w1 , w2 , w3 on
the boundary of .B2 (Fig. 3.2). Then a fractional-linear mapping .F, given by (3.2),
is required. Let’s check it out.

Fig. 3.2 Disks .B1 and .B2

52 3 Elementary Analytic Functions

Since each linear fractional mapping is conformal in .C, then, based on statement
Preposition 2.4, Theorem 3.1 and Corollary 3.1, the image of .B1 under the mapping
.F can be either .B2 or the complement to .B2 in .C. Let us show that only the first

option is possible.
Obviously that the angle between the arc .(z1 , z2 , z3 ) of the circle .∂B1 and
the segment drawn from the point .z1 to the center of the disk .B1 is equal to . π2
(Fig. 3.2). Considering that each fractional-linear function is angle-preserving (it is
conformal), the image of this segment will be an circular arc outgoing from .w1 and
forming a right angle with the arc .(w1, w2 , w3 ) counterclockwise. This means that
.F(B1 ) = B2 . 

It follows from this proof that fractional-linear mappings preserve the orientation
of the boundaries. It turns out that this is also true for other conformal mappings.

Theorem 3.6 (Principle of Preserving Boundaries and Their Orientations [11])

Let .Ω and G be bounded simply connected domains with smooth boundaries. If a
function .f conformally and univalently maps .Ω onto .G, then

(1) the function f can be continuously extended to .Ω;

(2) this extended function is an orientation-preserving bijection between .∂Ω and
.∂G.

3.3 Preservation of Symmetric Points by Fractional-Linear

Mappings

In this section, we will discover another interesting property of fractional-linear

transformations: it turns out that they are symmetry preserving.

Definition 3.3 Two points .z1 and .z1∗ are called symmetric with respect to the circle
∗
.Γ = {z : |z − a| = R} (denoted by .z = InvΓ (z1 )) if
1

(1) .arg(z1∗ − a) = arg(z1 − a),

(2) .|z1∗ − a| · |z1 − a| = R 2 .

By definitionition, we assume that .∞ = InvΓ (a). .

The first condition in Definition 3.3 means that the points .z1 and .z1∗ lie on the
same ray emanating from the point .a, the second one means that these points are
on opposite sides of the circle .Γ, or on .Γ (in this case .z1∗ = z1 ∈ Γ ). Obviously, if
∗ ∗
.z = InvΓ (z1 ), then .z1 = InvΓ (z ).
1 1
From Definition 3.3 it follows that

z1∗ − a = k (z1 − a),

. (3.5)
3.3 Preservation of Symmetric Points by Fractional-Linear Mappings 53

where .k > 0. Therefore,

|z1∗ − a| R2 R2 R2
k=
. = = = .
|z1 − a| |z1 − a| 2
(z1 − a) · (z1 − a) (z1 − a) · (z1 − a)

Substituting this expression for .k in (3.5), we obtain

R2 R2
.z1∗ = a + ⇐⇒ z1 = a + .
z1 − a z1∗ − a

Using this formula, we define the function

⎧
⎪
⎪ R2
⎪
⎨ a + , if z ∈ C \ {a},
∗ z−a
.z = InvΓ (z) :=
⎪
⎪ ∞, if z = a,
⎪
⎩
a, if z = ∞,

which is called inversion (symmetry) with respect to the circle .Γ .

Example 3.2 Let .a = 0 and .R = 1. Then the inversion with respect to the unit
circle .Γ1 = {z : |z| = 1} is given with the formula

1
InvΓ1 (z) =
. .
z

From this equality it follows that the fractional-linear function

1 1
w=
. =
z z

is the composition of the inversion with respect to the circle .Γ1 and the symmetry
with respect to the real axis.

Lemma 3.1 Two different points .z1 and .z1∗ are symmetric with respect to the circle
.Γ = {z : |z − a| = R} if and only if every circle .γ in .C, passing through the points
∗
.z1 and .z , intersects .Γ orthogonally.
1

Proof

Necessity Let .z1∗ = InvΓ (z1 ) and .γ be a circle in .C passing through .z1 and .z1∗ . Let
us draw a tangent line to .γ from the point .a and denote the point of tangency by .P .
Then the intersecting chord theorem states that .|P − a|2 = |z1∗ − a| · |z1 − a|. Since
the points .z1 and .z1∗ are symmetric with respect to .Γ, we have that .|P − a| = R.
54 3 Elementary Analytic Functions

Thus, the point .P is on .Γ and the segment .[a, P ] is a radius of the circle .Γ . This
means that the circle .γ intersects .Γ orthogonally.

Sufficiency Let an arbitrary circle in .C, passing through the points .z1 and .z1∗ ,
intersects .Γ orthogonally. Then the line (a special case of circles in .C), which
passes through these points, must also intersect the circle .Γ at a right angle. This
is possible only when this line passes through the center of the circle .Γ, that is,
through the point .a.
Moreover, the points .z1 and .z1∗ lie on a ray emanating from the point .a and on
opposite sides of .Γ, since otherwise the circle of radius . 12 |z1∗ − z1 |, which passes
through these points, cannot intersect the circle .Γ orthogonally. Thus, .arg(z1∗ −a) =
arg(z1 − a).
It remains to verify the second condition of Definition 3.3. Now let .γ be a circle
in .C that passes through the points .z1 and .z1∗ and intersects the circle .Γ at a right
angle. Let us denote by .P one of the intersection points. Then the radius .[a, P ] is
a segment of the tangent line to the circle .γ , and hence, based on the intersecting
chord theorem, we have

R 2 = |P − a|2 = |z1∗ − a| · |z1 − a|.

.

The lemma is proved.



Remark 3.4 The right part of the statement of Lemma 3.1 can be taken as a new
definition of symmetric points with respect to a circle. This definition is more
general because it includes the case when .Γ is a straight line (then it is the symmetry
of points relative to this line).

Theorem 3.7 (Symmetric Points Preservation Property) Let .Γ be a circle in .C

and .z1∗ = InvΓ (z1 ). Then for any fractional-linear mapping .F we have that

w1∗ = InvF(Γ ) (w1 ),

.

where .w1 = F(z1 ) and .w1∗ = F(z1∗ ); that is, symmetric points with respect to the
circle .Γ under a fractional-linear mapping become symmetrical points relative to
the image of this circle.

Proof Note that based on Theorem3.4, .F(Γ ) is a circle in .C. Let .γ be a circle in
C passing through the points .w1∗ and .w1 . Then .γ := F−1 (γ ) is a circle that passes
.

through the points .z1 and .z1∗ , and therefore, according to the preliminary lemma, the
circle .γ intersects .Γ at a right angle.
Considering the conformality of fractional-linear mappings, the circle .F(γ ) = γ
must intersect the circle .F(Γ ) also orthogonally. Then, by Lemma 3.1, we conclude
that .w1∗ = InvF(Γ ) (w1 ). 
3.4 Fractional-Linear Isomorphisms and Automorphisms 55

3.4 Fractional-Linear Isomorphisms and Automorphisms

Definition 3.4 Two domains Ω and Ω ∗ in C are called fractional-linear isomorphic

if there exists a fractional-linear mapping F such that Ω ∗ = F(Ω). In this case, the
mapping F : Ω → Ω ∗ is called the fractional-linear isomorphism of Ω onto Ω ∗ . 

Obviously, that the inverse function F−1 maps the domain Ω ∗ onto Ω.

Proposition 3.1 Each fractional-linear isomorphism of the half-plane

{z : Im z > 0} onto the unit disk B1 := {w : |w| < 1} can be represented in the form

z−a
w = eiα
. , (3.6)
z−a

where Im a > 0 and α is a real number.

Proof Let F be a fractional-linear isomorphism of {z : Im z > 0} onto B2 . The

existence of at least one such isomorphism follows from Corollary 3.2. Then there
exists a unique point a ∈ B1 such that F(a) = 0. Due to Theorem 3.7 we have that
F(a) = ∞.
Putting z1 = a, z2 = a, w1 = 0, w2 = ∞ in (3.2), we get

w 1 z − a z3 − a z−a
. · = · ⇒ w=A , (3.7)
1 w3 z − a z3 − a z−a

where A is a complex number.

Let us show that |A| = 1. Since points of the real axis under the mapping F go
to points on the circle {w : |w| = 1}, we have for all x ∈ R that

|x − a|
1 = |F(x)| ⇐⇒ 1 = |A|
. = |A| ⇒ A = eiα (α ∈ R).
|x − a|

It is easy to check that a mapping of the form (3.6) is a fractional-linear

isomorphism of the upper half-plane onto B1 . 

Definition 3.5 Fractional-linear isomorphism of a domain Ω onto itself is called a

fractional-linear automorphism of Ω.

Obviously, the set of all fractional-linear automorphisms of a domain Ω forms a

subgroup of the group (Λ, ◦).
56 3 Elementary Analytic Functions

Proposition 3.2 Each fractional-linear automorphism of the unit disk B1 = {z ∈

C : |z| < 1} can be represented in the form

z−b
w = eiβ
. , (3.8)
1 − zb

where |b| < 1 and β is a real number.

Proof Let F be a fractional-linear automorphism of the disk B1 . Then there exists

a unique point b ∈ B1 such that F(b) = 0. Due to Theorem 3.7 we have that
F (b∗ ) = ∞, where

1
. b∗ = Inv∂B1 (b) = .
b

Substituting z1 = b, z2 = 1 , w1 = 0, w2 = ∞ in (3.2), we get

b

z−b z−b
w=A
. = A1 .
z− 1
1 − zb
b

Since F(1) ∈ ∂B1 , we have the following

|1 − b|
.|F(1)| = 1 ⇐⇒ 1 = |A1 | = |A1 | ⇒ A1 = eiβ (β ∈ R).
|1 − b|

It is easy to check that a mapping of the form (3.8) is a fractional-linear

automorphism of the disk B1 . 

Exercise 3.3 Let F be a fractional-linear automorphism of the unit disk B1 . It is

known that
i 1i
.F =− . +
2 2 8
√ π
Reasonably find the value of F 2 ei 4 .

Exercise 3.4 Prove that each fractional-linear automorphism of the half-plane {z :

Im z > 0} can be represented in the form

az + b
w=
. ,
cz + d

where a, b, c, d are real numbers and ad − cb > 0.

For more properties of fractional-linear automorphisms, see Sect. 9.3.

3.5 Power Functions with Natural Exponents 57

3.5 Power Functions with Natural Exponents

Consider the power function

ω = f (z) := zn ,
. z ∈ C (n ∈ N, n ≥ 2).

It follows from Example 2.2 that .f  (z) = nzn−1 for any .z ∈ C. Thus, .f ∈ A(C)
and .f is conformal in .C \ {0}.
Let us find domains of univalence (see Definition 2.14) of f . Let .z1 = z2 and
.z = z . Then
n n
1 2

|z1 | = |z2 |,
|z1 |n einα1 = |z2 |n einα2
. ⇒
α1 = α2 + 2π k
n , k ∈ Z,

where .α1 ∈ Arg(z1 ), .α2 ∈ Arg(z2 ).

Thus, a domain .Ω is the domain of univalence for the function .zn if it does not
contain any pair of complex numbers with the same moduli and with the principal
n . In particular, the function .ω = z is univalent in the
arguments that differ by . 2π n

corner
 2π 
Kβ := z ∈ C : β < Arg(z) < β +
. (β ∈ R).
n

Here, by .Arg(z) we mean one of arguments of .z, which satisfies this inequality. Let
us find the image of .Kβ under the power function f .
It is easy to check that f maps the ray

.z = t eiα , t ∈ (0, +∞),

into the ray .z = t n einα , .t ∈ (0, +∞) (Fig. 3.3).

Now if .α changes from .β to .β + 2π n , then these rays will fill out the domain .C \
{ω : Arg(w) = nβ} (Fig. 3.4). This means that .ω = zn conformally and univalently
maps the corner .Kβ onto .C \ {ω : Arg(ω) = nβ}.
In particular, if .β = 0 and .n = 2, then the image of the upper half-plane under
.w = z is the complex plane without the positive real half-axis.
2

Fig. 3.3 The image of a ray under the power function .zn
58 3 Elementary Analytic Functions

Fig. 3.4 The image of the corner .Kβ under the power function .zn

Fig. 3.5 The image of the square .K1 under .w = z2

Example 3.3 Let us find the image of the unit square .K1 from Example 3.1 under
the function .ω = z2 .
Since the square .K1 is symmetric with respect to the straight line .y = x, its
image will be symmetric with respect to the imaginary axis, because .ω = z2 sends
the points .|a| exp(iϕ) and .|a| exp(i( π2 − ϕ)) .(here .ϕ ∈ (0, π2 )) respectively to the
points

|a|2 exp(i2ϕ)
. and |a|2 exp(i(π − 2ϕ)).

It is easy to verify that the function .ω = z2 maps the segment .[0, 1] onto itself,
and the segment .[0, 1 + i] onto .[0, 2i].
Substituting the parameterization .z = 1+iy, y ∈ [0, 1], of the segment .[1, 1+i]
in .u + iv = z2 , we find its image that is a part of the parabola

v2
u=1−
. , v ∈ [0, 2].
4

So the square .K1 is mapped by .ω = z2 onto a domain bounded by parabolas

v2 v2
u=1−
. and u = −1 + ,
4 4
and the real axis (Fig. 3.5).
3.6 The Inverse to a Power Function and Its Riemann Surface 59

3.6 The Inverse to a Power Function and Its Riemann Surface

First, consider the function .w = z2 . It is clear that there is no single-valued

inverse function for .z2 : C → C, because every point from .C, except √ for the
points .0 and .∞, has two inverse images, i.e., the function .z = w is two-
valued. It is inconvenient to work with multi-valued functions because it is not clear,
for example, how to introduce the concept of continuity, differentiation, etc. The
German mathematician B. Riemann (1826–1866) in his dissertation proposed the
following approach for transforming a multi-valued function into a single-valued
one:

it is necessary to separate points that have more than two images and consider
them on different sheets of a surface, which is now called a Riemann surface.

In this section and next one we will get acquainted with the concept of a Riemann
surface on examples of some functions, and the abstract approach will be considered
√ Sect. 8.5. First, let us implement this approach for the two-valued function .z =
in
w. Consider the upper and lower half-planes

D0 = {z ∈ C : 0 < arg z < π}

. and D1 = {z ∈ C : −π < arg z < 0} .

The previous section shows that the function .w = f (z) = z2 conformally and
univalently maps .D0 and .D1 onto the domain .E = {w ∈ C : 0 < Arg w < 2π}.
Therefore, for each of these mappings there is a unique inverse function
−1
.f
k : E −→ Dk .(k ∈ {0, 1}; Fig. 3.6), i.e.,

.f (fk−1 (w)) = w for all w ∈ E, and fk−1 (f (z)) = z for all z ∈ Dk .

√
Fig. 3.6 Branches .f0−1 and .f1−1 of the inverse function .z = w
60 3 Elementary Analytic Functions

√
Fig. 3.7 Construction of the Riemann surface of the function . w

Using (1.11), we get that

 Arg w
fk−1 (w) =
. |w| ei( 2 +π k) for all w ∈ E (0 < Arg w < 2π ).

Since

dfk−1 (w) 1 1
. =  = = 0 for all w ∈ E,
dw f (z) z=fk−1 (w) 2fk−1 (w)

the function .fk−1 is conformal and univalent in .E .(k ∈ {0, 1}).

Take two sheets (instances) of the domain .E and place them one above the other
(Fig. 3.7). The sheet from which the function .f0−1 acts is called the 0th-sheet; and
the sheet from which the function.f1−1 acts is called the 1st-sheet. The point .w = 0
from the 0th-sheet is identified with the point .w = 0 from the 1st-sheet, because
they have the same inverse image under the mapping .w = f (z) = z2 ; the same for
.w = ∞.

To understand how to glue (identify) points from different edges of the cuts of
these two sheets, one should find the limits of the functions .f0−1 and .f1−1 as .w tends
from each sheet to positive x from above and below. Since
√
. lim f −1 (w) = x= lim f −1 (w),
w→x>0, Im w>0 0 w→x>0, Im w<0 1

points of the upper edge of the 0th-sheet cut must be glued (identified) with the
corresponding points of the lower edge of the 1st-sheet cut. Since
√
. lim f −1 (w) =− x= lim f −1 (w),
w→x>0, Im w<0 0 w→x>0, Im w>0 1

points of the lower edge of the 0th-sheet cut must be glued with the corresponding
points of the upper edge of the 1st-sheet cut.
3.6 The Inverse to a Power Function and Its Riemann Surface 61

√
The surface thus glued is called √ a Riemann surface of . w and is denoted
by .R√w (Fig. 3.8). The function . w becomes single-valued on its Riemann
surface and conformal everywhere except for the points 0 and .∞, because
the two values that the root assigns to each nonzero complex number are
now images of two different points lying on different sheets above this
number. Positive numbers are no exception, since over them there is no self-
intersection of the sheets of the Riemann surface.

Example 3.4 Consider, for example, the complex number .−2. Over it there are
two different
√ points .−20 and .−21 lying on √
the 0th-sheet and 1st-sheet, respectively.
Since . w = f0−1 (w) on the 0th-sheet and . w = f1−1 (w) on the 1th-sheet, we find
  √
−20 = f0−1 (−2) =
π
. | − 2| ei 2 = 2 i,
  √
−21 = f1−1 (−2) =
π
. | − 2| ei( 2 +π ) = − 2 i.


.

Each of the points 0 and .∞ is called a first-order branch point. The order of
a branch point indicates the additional number of sheets of the Riemann surface
that must be traversed around this point in order to return to the original position
(Fig. 3.8).
−1
The functions
√ .f
0 and .f1−1 are called analytic branches of the single-valued
function . w given on its Riemann surface .R√w . In 3D-space, we will not able to
√
draw the Riemann surface of .z = w without self-intersection, but schematically
it can be represented as shown in Fig. 3.8. √
Similarly, the Riemann surface of the function . n w .(n ∈ N, n > 2) is
constructed. To do this, we consider the angles
 
2π k 2π(k + 1)
.Dk = z ∈ C : < Arg z < , k ∈ {0, 1, . . . , n − 1},
n n

that are domains of univalence for the function .w = f (z) = zn . It con-

formally and univalently maps each of these corners onto the domain .E =

Fig. 3.8 The Riemann

surface .R√w
62 3 Elementary Analytic Functions

Fig. 3.9 The image of .Dk by the function .w = zn

{w ∈ C : 0 < Arg w < 2π} (Fig. 3.9). Therefore, for each of these mappings there
is a unique inverse function .fk−1 : E → Dk , i.e.,

f (fk−1 (w)) = w
. for all w ∈ E, and fk−1 (f (z)) = z for all z ∈ Dk ;

and
 Arg w 2π k
fk−1 (w) =
.
n
|w| ei( n + n ) for all w ∈ E (0 < Arg w < 2π ).

Since

dfk−1 (w) 1 fk−1 (w)

. =  = = 0 for all w ∈ E,
dw f (z) z=fk−1 (w) nw

the function .fk−1 is conformal and univalent in .E .(k ∈ {0, 1, . . . , n − 1}).

We take n sheets of E and place them on top of each other above the complex
plane. The sheet from which the function .fk−1 acts is called the kth-sheet. Using
the functions .{fk−1 }n−1
k=0 , these sheets must be glued together to form a continuous
function.
The points .w = 0 from these sheets are identified because they have the same
inverse image under the mapping .w = zn ; the same for .w = ∞. Since for each
.k ∈ {1, . . . , n − 2}

√
f −1 (w) f −1 (w)
2π k
. lim = n
x ei n = lim
w→x>0, Im w>0 k w→x>0, Im w<0 k−1

and
√ 2π(k+1)
. lim f −1 (w) = n
x ei n = lim f −1 (w),
w→x>0, Im w<0 k w→x>0, Im w>0 k+1

points of the upper edge of the kth-sheet cut must be glued with the corresponding
points of the lower edge of the .(k − 1)th-sheet cut, and points of the lower edge of
3.7 Exponential Function, Logarithmic Function and Its Riemann Surface 63

Fig. 3.10 Riemann

surface .R √
4w

the kth-sheet cut must be glued with the corresponding points of the upper edge of
the .(k + 1)th-sheet cut. Since
−1 √
. lim fn−1 (w) = n
x= lim f −1 (w),
w→x>0, Im w<0 w→x>0, Im w>0 0

points of the lower edge of the .(n − 1)th-sheet must be glued with the corresponding
points of the upper edge of the 0-sheet. √
The surface thus glued is called a Riemann√surface of the function . n w and
is denoted by .R √ n
n w (Fig. 3.10). The function . w becomes single-valued on its

Riemann surface and conformal everywhere except for the points 0 and .∞ that are
branch points of order .n − 1. √The functions .{fk−1 }n−1
k=0 are called analytic branches
of the single-valued function . n w given on its Riemann surface .R √ n w . Now above

each nonzero complex number there are .n different points lying on different sheets
of the Riemann surface .R √n w.

3.7 Exponential Function, Logarithmic Function and Its

Riemann Surface

The exponential function is already defined in Example 2.4:

def
w = ez = ex+iy = ex (cos y + i sin y),
. z ∈ C,

which also shows that it is an entire function and .(ez ) = ez = 0 for all .z ∈ C. Due
to (1.10) we get .ez1 +z2 = ez1 ez2 .
A new property of the exponential function is its periodicity with the main period
.2π i. Indeed, .e
z+2π i = ez e2π i = ez for all .z ∈ C. On the other hand, if we assume

that there is another period .T = T1 + iT2 , then for all .z ∈ C

ez+T = ez
. ⇒ eT1 eiT2 = 1 ⇒ T1 = 0 and T2 = 2π k, k ∈ Z,

i.e., .T = i2π k, k ∈ Z \ {0}.

 64 3 Elementary Analytic Functions

Fig. 3.11 The image of a horizontal line by the function .ez

Fig. 3.12 The image of a

horizontal strip by the
function .ez

Consequently, a domain is a domain of univalence for .ez if it does not contain any
pair of points connected by the relation .z1 − z2 = 2π ik, .k ∈ Z \ {0}. In particular,
the horizontal strips

Dk := {z : −π + 2kπ < Im z < π + 2π k},

. k ∈ Z.

are domains of univalence for the exponential function.

Let us find the image of the horizontal strip .{z : α < Im z < β} .(0 < β − α <
2π ) under the mapping .ez . It is easy to check that the exponential function maps a
horizontal line

z = x + ib,
. x ∈ (−∞, +∞),

into the ray .w = ex eib , .x ∈ (−∞, +∞) (see Fig. 3.11).

Now if b changes from .α to .β, then .ez conformally and univalently maps the
strip .{z : α < Im z < β} onto the corner .{w : α < Arg w < β} (Fig. 3.12).

Example 3.5 Similarly, we verify that the image of a vertical segment .z = a + iy,
.y ∈ [−π, π ], by .w = ez is the circle .w = ea ey , .y ∈ [−π, π ], of radius .ea
centered at the origin (Fig. 3.13); here .a ∈ R. Thus, the image of the closed rectangle
.[a, b] × [−iπ, iπ] by .w = e is the annulus .{w : e < |w| < e }.
z a b

From the above it follows that for all .k ∈ Z the exponential function .f (z) := ez
conformally and univalently maps the horizontal strip .Dk onto the domain

E1 = {w : − π < arg(w) < π },

.
3.7 Exponential Function, Logarithmic Function and Its Riemann Surface 65

Fig. 3.13 Images of a

segment and rectangle by the
function .ez

Fig. 3.14 The image of the strip .Dk by the function .ez

moreover, the lower part of the boundary of .Dk is mapped onto the lower edge
of the cut .(−∞, 0], and the upper one is mapped onto the upper edge of this cut
(Fig. 3.14).
This means that for each such mapping there is an inverse .fk−1 : E1 → Dk , i.e.,
for all integer k

f (fk−1 (w)) = w
. for all w ∈ E1 , and fk−1 (f (z)) = z for all z ∈ Dk .

In addition, since

d fk−1 (w) 1 1
. =   = = 0 for all w ∈ E1 , (3.9)
dw ez z=f −1 (w) w
k

the function .fk−1 is conformal and univalent in .E1 ; .k ∈ Z.

66 3 Elementary Analytic Functions

To deduce formula for the inverse function .fk−1 , one should find the unique root
.z ∈ Dk of the equation .e = w, where .w ∈ E1 . From the equality of two complex
z

numbers we derive

ez = w ⇐⇒ ex eiy = |w| ei arg(w)

. ⇒ x = log |w| and y = arg(w) + 2π k
 
. ⇒ z = x + iy = log |w| + i arg(w) + 2π k .

Thus,

fk−1 (w) = log |w| + i arg(w) + 2π ik,

. w ∈ E1 .

The functions .{fk−1 }k∈Z are called analytical branches of the multi-valued logarith-
mic function

Log (w) := log |w| + iArg(w),

. w ∈ C \ 0.

Recall that .Arg w is the set of all arguments of a complex number w.

The analytical branch .f0−1 (w) = log |w|+i arg(w) is called the principal branch
of .Log.

Remark 3.5 As follows from (1.8), the function .arg(w), w = u + iv ∈ E1 , is

differentiable in .E1 as a function of two real variables u and .v, and it has a jump .2π
as w crosses the negative real axis:

. lim arg(w) = π and lim arg(w) = −π.

w→x<0, Im w>0 w→x<0, Im w<0

Thus, the values of each branch .fk−1 jump by .2π i when crossing the negative real
axis.

Exercise 3.5 By using the Cauchy–Riemann Theorem 2.2, check the differentia-
bility of each branch .fk−1 in .E1 .

To construct the Riemann surface of .Log we take a denumerable set of .E1 -sheets
and place them on top of each other over the complex plane. The sheet from which
the function .fk−1 acts is called the kth-sheet. Using the functions .{fk−1 }k∈Z , these
sheets must be glued together to form a continuous function. Since for each .k ∈ Z

. lim f −1 (w) = log |x| + i(π + 2π k) = lim f −1 (w)

w→x<0, Im w>0 k w→x<0, Im w<0 k+1

and

. lim f −1 (w) = log |x| + i(−π + 2π k) = lim f −1 (w),

w→x<0, Im w<0 k w→x<0, Im w>0 k−1
3.7 Exponential Function, Logarithmic Function and Its Riemann Surface 67

Fig. 3.15 Construction of the Riemann surface of the function .Log

Fig. 3.16 The Riemann

surface .RLog

points of the upper edge of the kth-sheet cut must be glued with the corresponding
points of the lower edge of the .(k + 1)th-sheet cut, and points of the lower edge of
the kth-sheet cut must be glued with the corresponding points of the upper edge of
the .(k − 1)th-sheet cut (Fig. 3.15).
At the points .w = 0 and .w = ∞ the logarithm is not defined and, therefore, its
Riemann surface is not contain points above .w = 0 and .w = ∞. The surface thus
glued is a Riemann surface of the function .Log and is denoted by .RLog (infinite
helical surface, Fig. 3.16). The function .Log w becomes single-value on .RLog and
conformal everywhere except for the points 0 and .∞ that are branch points of
infinite order (or logarithmic branch points).

Example 3.6 Consider the complex number .−2. There are countably many differ-
ent points .{−2k }k∈Z above .−2, each of them lies on the corresponding kth-sheet.
Since .Log w = fk−1 (w) on the kth-sheet, we find

Log(−2k ) = fk−1 (−2) = log | − 2| + i arg(−2) + 2π ik = log 2 + i π + 2π ik.

.
68 3 Elementary Analytic Functions

3.8 Joukowsky Function

Here we study an interesting function named after Nikolai Joukowsky (1847–1921),

the founder of hydro and aeromechanics. He used this function to derive the formula
for aircraft wing lift (1906), to develop the vortex theory of propellers, and to
determine optimal wing and propeller blade profiles. This function was used to
theoretically calculate the possibility of performing the “loop” aerobatic manoeuvre,
first performed by P. Nesterov in Kyiv (1913).
The Joukowsky (or Zhukovsky) function is a mapping of the form

1 1
.w = J (z) := z+ , z ∈ C \ {0}.
2 z

By continuity, we define J at zero, namely, since . lim J (z) = ∞, we set

z→0

def
J (0) = ∞.
.

Obviously, .J (z) = J (1/z) for all .z ∈ C. Since

1 1
J  (z) =
. 1− 2 (z = 0)
2 z

and the derivative of the function .g(z) = 1

J (z) = 2z
1+z2
at 0 is equal to .2, the
Joukowsky function is conformal in .C \ {±1}.
Let there exist two different numbers .z1 = z2 such that .J (z1 ) = J (z2 ), i.e.,
.z1 − z2 +
z1 − z2 = 0. From here we deduce
1 1

1
(z1 − z2 ) 1 −
. =0 ⇒ z1 z2 = 1. (3.10)
z1 z2

Thus, some domain is a domain of univalence for the Joukowsky function if it does
not contain any pair of points that satisfy the relation (3.10). It is easy to understand
that a domain of univalence of J cannot contain points .±1, because in arbitrary
neighborhoods of these points there are always different points satisfying (3.10).
Obviously, the following domains: .{z : |z| > 1}, .{z : |z| < 1}, .{z : Im z > 0} and
.{z : Im z < 0} are domains of univalence for the Joukowsky function.

Let us derive the so-called transition formulas for the Joukowsky function, which
will help us to find images of various curves and regions. If the function J sends a
point .z = r eiϕ to a point .w = u + iv, then

1 1 1 1
w = J (z) ⇐⇒ u + iv =
. r+ cos ϕ + i r − sin ϕ
2 r 2 r

1 1 1 1
. ⇐⇒ u = r+ cos ϕ and v= r− sin ϕ. (3.11)
2 r 2 r
3.8 Joukowsky Function 69

Fig. 3.17 The image of the

unit circle

Consider the circle

KR := {z : z = Reiϕ , ϕ ∈ [0, 2π ]}.

.

If .R = 1, then it follows from (3.11) that .u = cos ϕ and .v = 0. Thus, the image of
the unit circle is the segment .[−1, 1] that is traversed twice when .ϕ changes from 0
to .2π (Fig. 3.17).
If .R = 1, the image of .KR under the mapping J is the ellipse
⎧ ⎫
⎪
⎨ ⎪
⎬
u2 v2
. w = u + iv :    +    = 1
⎪
⎩
2 2 ⎪
⎭
1
R + R −1
2
1
2R − R −1

with the semi-major axes

1  1
a=
. R + R −1 and b= R − R −1 ,
2 2
and with foci at .±1. However, for .R > 1, the orientation of the ellipse is preserved
(this means that the upper semicircle is mapped to the upper half-ellipse, and the
lower semicircle is mapped to the lower half-ellipse (see the upper part of Fig. 3.18),
and when .R < 1, the orientation is reversed (this means that the upper semicircle

Fig. 3.18 Images of circles

by the Joukowsky function
70 3 Elementary Analytic Functions

Fig. 3.19 Images of the

exterior and interior of the
unit disk by J

is mapped to the lower half-ellipse, and the lower semicircle into the upper half-
ellipse).
Let’s have a look at where the Joukowsky function maps its univalence domains:
.{z : |z| > 1}, .{z : |z| < 1}, .{z : Im z > 0} and .{z : Im z < 0}.

Since .{z : |z| > 1} = R∈(1,+∞) KR , the image of the exterior of the unit disk by
the Joukowsky function is .C\[−1, 1]; moreover, knowing the images of semicircles
(see Fig. 3.18), we have

J
{z : |z| > 1, Im z > 0} −→ {w : Im w > 0}
.

and
J
{z : |z| > 1, Im z < 0} −→ {w : Im w < 0}
.

(see the upper part of Fig. 3.19).

Since .J (z) = J ( 1z ), the image of the unit disk .{z : |z| < 1} by the Joukowsky
function is also .C \ [−1, 1], but now (see the lower part of Fig. 3.19)

J
{z : |z| < 1, Im z > 0} −→ {w : Im w < 0}
.

and
J
{z : |z| < 1, Im z < 0} −→ {w : Im w > 0}.
.

To find the image of the upper half-plane, we present it as a union of three sets,
the images of which were found above: .{z : |z| > 1, Im z > 0}, .{z : |z| <
1, Im z > 0} and .{z : |z| = 1, Im z > 0}. Thus,

J 
.{z : Im z > 0} −→ C \ (−∞, −1] ∪ [1, +∞) =: E2 (3.12)
 3.8 Joukowsky Function 71

Fig. 3.20 Images of the upper and lower half-planes by J

(the complex plane with cuts along the rays .(−∞, −1] and .[1, +∞); see the upper
part of Fig. 3.20). Similarly, we find that (the lower part of Fig. 3.20)

J
. {z : Im z < 0} −→ E2 . (3.13)

Exercise 3.6 Find a conformal and univalent function that maps the domain
{z : |z| < 1, 0 < arg(z) < π2 } onto the unit disk .B1 (0).
.

The mappings (3.12) and (3.13) are one-to-one. The inverse mappings

J1−1 : E2 −→ {z : Im z > 0}

. and J2−1 : E2 −→ {z : Im z < 0}

are called analytical branches of the multi-valued inverse function


J −1 (w) = w +
. w 2 − 1.

Exercise 3.7 With the help of .J1−1 and .J2−1 construct the Riemann surface of the
√
multi-value function .J −1 (w) = w + w 2 − 1.

Now we find where Joukowsky’s function maps a ray

{z = r eiα : r ∈ (0, +∞)}.

.

By (3.11), we conclude that the image of this ray is the curve

 
1 1 1 1
. w = u + iv = r+ cos α + i r − sin α : r ∈ (0, +∞) .
2 r 2 r
72 3 Elementary Analytic Functions

Fig. 3.21 Images of rays under Joukowsky function

Fig. 3.22 Joukowsky wing profile (on the right)

Here, .α is a fixed value from .(0, π ) This is the right branch of the hyperbola

u2 v2
. − =1
cos2 α sin2 α

with the foci at the points .±1 if .α ∈ (0, π2 ); this is the left branch of this hyperbola
if .α ∈ ( π2 , π ); and this is the imaginary axis if .α = π2 . It should be noted that if the
parameter r changes from 0 to .+∞, then the point on these branches moves from
the bottom to top (Fig. 3.21).

Exercise 3.8 Suppose that two circles .K1 and .K2 pass through the point .z = 1 at
an angle .α to the real axis, and the circle .K1 also passes through the point .z = −1
and lies inside the circle .K2 (Fig. 3.22).
Prove that the image of the domain .int(K2 ) ∩ ext(K1 ) under the Joukowsky
function is a so called “Joukowsky wing profile” (Fig. 3.22).

3.9 Trigonometric and Hyperbolic Functions and Their

Inverses

From the Euler formula (1.9) we derive

eix + e−ix eix − e−ix

. cos x = , sin x = .
2 2i
3.9 Trigonometric and Hyperbolic Functions and Their Inverses 73

These relations are used to define the trigonometric functions of a complex variable:

def eiz + e−iz eiz − e−iz

def
. cos z = , sin z = , z ∈ C, . (3.14)
2 2i
def sin z ei2z − 1 π 
tan z = = −i i2z , z∈C\ + πn : n ∈ Z ,. (3.15)
cos z e +1 2
def cos z ei2z + 1 
cot z = = i i2z , z ∈ C \ πn : n ∈ Z . (3.16)
sin z e −1

From these formulas and properties of the exponential function it follows that .cos
and .sin are .2π -periodic functions, and .tan and .cot are .π -periodic. Using (3.14)–
(3.16), it can be verified that all formulas for the trigonometric functions of a real
argument remain true for a complex argument as well.
Applying (2.10), one finds the derivatives of .sin z and .cos z, namely

(sin z) = cos z,

. (cos z) = − sin z for all z ∈ C.

Therefore, .sin z and .cos z are entire functions. Then the quotient rule states

1 π 
(tan z) =
. for all z ∈ C \ + πn : n ∈ Z ,
cos2 z 2
1 
(cot z) = − 2
for all z ∈ C \ π n : n ∈ Z .
sin z

It is easy to see that the trigonometric functions are connected with the hyperbolic
functions

def ez + e−z def ez − e−z

. cosh z = , sinh z = ,
2 2
def sinh z e2z − 1 def cosh z e2z + 1
tanh z = = 2z , coth z = = 2z (3.17)
cosh z e +1 sinh z e −1

through the following relations:

. cosh(iz) = cos z, cos (iz) = cosh z,

. sinh(iz) = i sin z, sin (iz) = i sinh z,

. tanh(iz) = i tan z, tan (iz) = i tanh z.

Hence, .cosh and .sinh are .2π i-periodic functions, and .tanh and .coth are .π i-periodic.
74 3 Elementary Analytic Functions

A new property of the trigonometric functions .sin z and .cos z is that they are
unbounded, which means that their moduli are unbounded functions. Indeed,

. sin z = sin (x + iy) = sin x cosh y + i cos x sinh y



sin2 x 1 |y|
. ⇒ | sin z| = sin x + sinh y = | sinh y| 1 +
2 2
∼ e ,
sinh2 y 2

1 |y|
as .|y| → +∞. Similarly, we show that .| cos z| ∼ 2e as .|y| → +∞.

Remark 3.6 It turns out that every nonconstant entire function is unbounded. We
will prove this statement in Sect. 5.2.

Since .cos z = sin(z + π2 ), we continue to consider the function

w = f (z) := sin z = sin x cosh y + i cos x sinh y,

. z = x + iy ∈ C. (3.18)

Let there exist two different numbers .z1 = z2 such that .sin z1 = sin z2 . Then
2 sin z1 −z
.
2
2 cos 2
z1 +z2
= 0, which means

z1 − z2 = 2π n, n ∈ Z \ {0},
. or z1 + z2 = π + 2π k, k ∈ Z. (3.19)

Recalling the geometric interpretation of the sum and subtraction of complex

numbers, it follows from (3.19) that the domains
 π π 
Dk := z : − + π k < Re z < + π k ,
. k ∈ Z,
2 2
are domains of univalence for the sine. Let us find their images.
If .sin sends a point .z = x + iy to a point .w = u + iv, we derive from (3.18) the
transition formulas

u = sin x cosh y,
. v = cos x sinh y. (3.20)

By (3.20), we conclude that the image of a vertical straight line

.lα = {z = α + iy : y ∈ (−∞, +∞)},

where .α ∈ (0, π2 ) (see Fig. 3.23), is the right branch of the hyperbola

u = sin α cosh y, u2 v2
. y ∈ R, ⇒ − = 1,
v = cos α sinh y, (sin α) 2 (cos α)2
3.9 Trigonometric and Hyperbolic Functions and Their Inverses 75

Fig. 3.23 The image of a vertical line by mapping the sine function

Fig. 3.24 The image of the strip .D0 by mapping the sine function

and the left branch if .α ∈ (− π2 , 0), and this is the imaginary axis if .α = 0. Moreover,
the upper part of the line is mapped onto the upper part of the corresponding branch
of the hyperbola,and the lower part into the lower one.
Since .D0 = α∈(− π , π ) lα and the right (left) branch of the hyperbola is shrunk
2 2
into the cut .[1, +∞) .((−∞, −1]) as .α → π2 − 0 .(α → − π2 + 0), the image of the
vertical strip .D0 by the function .sin is the complex plane with the cuts .(−∞, −1]
and .[1, +∞) along the real axis (Fig. 3.24).
Such a domain was faced in the previous paragraph and was denoted by .E2
(see (3.12)). Also note that the part of the strip .D0 lying in the upper half-plane
.{z : Im z > 0} is mapped onto the upper half-plane .{w : Im w > 0}, and the lower

one is mapped onto the lower half-plane, respectively.

Using the formula .sin z = (−1)k sin(z + kπ ), we can assert that for each .k ∈ Z
the function .f (z) = sin z conformally and univalently maps the vertical strip .Dk
onto .E2 . However,

• for .k = 2p the part of the strip .Dk lying in the upper half-plane .{z : Im z >
0} is mapped onto the upper half-plane .{w : Im w > 0}, and the lower one,
respectively, onto the lower half-plane;
• for .k = 2p − 1 the part of .Dk lying in the upper half-plane is mapped onto
the lower half-plane, and the lower one, respectively, onto the upper half-plane
(Fig. 3.25).
76 3 Elementary Analytic Functions

Fig. 3.25 The images of the strips .Dk by mapping the sine function

Thus, for each .k ∈ Z there exists an inverse .fk−1 : E2 → Dk such that

f (fk−1 (w)) = w for all w ∈ E2 ,

. and fk−1 (f (z)) = z for all z ∈ Dk ,

dfk−1 (w) 1 1
. = =√ = 0 for all w ∈ E2 .
dw cos z z=fk−1 (w) 1 − w2

Therefore, the function .fk−1 is conformal and univalent in .E2 . Obviously, that
−1 −1
.f
k (0) = π k. The functions .{fk }k∈Z are called analytical branches of the multi-
valued function .z = Arcsin w. To find the formula for .Arcsin, you need to solve the
equation

eiz − e−iz
. sin z = w ⇐⇒ = w ⇐⇒ (eiz )2 − 2iweiz − 1 = 0
2i
 
. ⇒ eiz = iw + 1 − w 2 ⇒ z = −i Ln(iw + 1 − w 2 ) =: Arcsin w.

In the same way, one can find other functions that are inverse to both trigonometric
and hyperbolic functions, for example,
 1 1 + iw
Arccos w = −i Ln(w +
. w 2 − 1), Arctan w = Ln ,
2i 1 − iw
 1 1+w
Arcsinh w = Ln(w +
. 1 + w 2 ), Arctanh w = Ln .
2 1−w
3.9 Trigonometric and Hyperbolic Functions and Their Inverses 77

To construct the Riemann surface of .Arcsin we take a denumerable set of .E2 -

sheets and place them on top of each other over the complex plane. The sheet from
which the function .fk−1 acts is called the kth sheet. Using the functions .{fk−1 }k∈Z ,
these sheets must be glued together to form a continuous function.
Take a sheet with an odd number .k = 2p − 1. Considering how the sine maps
the upper and lower parts of the vertical strip .Dk (see Fig. 3.25), we get the rule for
gluing the .E2 -sheets. Since

. lim f −1 (w) = lim f −1 (w)

w→x<−1, Im w<0 k w→x<−1, Im w>0 k+1

and

. lim f −1 (w) = lim f −1 (w),

w→x<−1, Im w>0 k w→x<−1, Im w<0 k+1

points of the lower edge of the left kth sheet cut must be glued with the correspond-
ing points of the upper edge of the left .(k + 1)th sheet cut, and points of the upper
edge of the left kth sheet cut must be glued with the corresponding points of the
lower edge of the left .(k + 1)th sheet cut.
Similarly, since

. lim f −1 (w) = lim f −1 (w)

w→x>1, Im w<0 k w→x>1, Im w>0 k−1

and

. lim f −1 (w) = lim f −1 (w),

w→x>1, Im w>0 k w→x>1, Im w<0 k−1

points of the lower edge of the right kth sheet cut must be glued with the
corresponding points of the upper edge of the right .(k − 1)th sheet cut, and point
of the upper edge of the right kth sheet cut must be glued with the corresponding
points of the lower edge of the right .(k − 1)th sheet cut.
The surface thus glued is the Riemann surface of .Arcsin and is denoted by
.RArcsin . The function .Arcsin w becomes single-value on .RArcsin and conformal

everywhere except for the points .±1 and .∞. Over the points 1 and .−1 of the
complex plane there is a denumerable set of first-order branch points, respectively.
The point .∞ is a logarithmic branch point. The scheme of the transition between
the sheets of the Riemann surface is shown in Fig. 3.26, and a part of the Riemann
surface is shown in Fig. 3.27.
It turns out that the trigonometric functions .cos and .sin and the hyperbolic
functions .cosh and .sinh can be represented as the corresponding compositions of
the Joukowsky function and the exponential function, namely

. cos z = J (eiz ), sin z = −J (i eiz ),

cosh z = J (ez ), sinh z = −iJ (i ez ).
78 3 Elementary Analytic Functions

Fig. 3.26 Transition scheme between sheets of the Riemann surface of .Arcsin

Fig. 3.27 A part of the Riemann surface of .Arcsin

Therefore, to find the images of domains mapped by these functions, the properties
of the exponential function and the Joukowsky function are used.
To find the images of domains when mapped by the functions .tan, cot, tanh and
.coth, the formulas (3.15), (3.16) and (3.17) are used.

Example 3.7 The function

e2iz − 1
w = tan z = −i
.
e2iz + 1

is the composition of the linear function .ξ = 2iz, the exponential function .η = eξ ,

and the fractional-linear function
η−1
w = −i
. .
η+1
3.9 Trigonometric and Hyperbolic Functions and Their Inverses 79

Fig. 3.28 The image of a vertical strip by mapping the function .w = tan z

Therefore, the tangent function maps the vertical strip

 π π
.z : − < Re z <
4 4
onto the unit disk .B1 (0), as shown in Fig. 3.28.
Integration of Functions of a Complex Variable
4

Abstract

In the previous two chapters, it was shown that analytic complex-valued func-
tions enjoy excellent differentiability properties that their real counterparts do
not share. It is well known that differentiation and integration are mutually
inverse operations and they are the main concerns of calculus. To continue on, the
next logical step is to consider the integration in the complex plane, as initiated
by the French mathematician Augustin-Louis Cauchy (1789–1857). Integration
is impossible without the concept of an antiderivative, which becomes much
more complicated in complex analysis. For example, it turns out that there are
analytic functions in some domains that have no antiderivatives. In this chapter,
we will introduce a new concept of an antiderivative along a curve and study its
properties. We will also show that the beauty of complex integration also goes
far beyond real analysis and prove very important and interesting theorems.

4.1 Line Integrals and Their Simplest Properties

Definition 4.1 Let f be a continuous function in a domain Ω and let z = γ(t),

t ∈ [a, b], be a piecewise smooth curve, whose trace belongs to Ω, i.e., Eγ ⊂ Ω.
The integral of the function f along the curve γ is defined as follows
  b
def
. f (z) dz = f (γ(t)) γ  (t) dt. (4.1)
γ a

The integrand function in the right-hand side of (4.1) is a complex-valued

function of a real variable. Under the definition hypotheses, its real and imaginary

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 81

T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_4
82 4 Integration of Functions of a Complex Variable

parts have only a finite number of first-order breakpoints of the kind, and are
therefore Riemann-integrable on [a, b].
Let’s analyze several examples.

Example 4.1 Compute the integral γ z dz, where

(1) the curve γ is the straight line from 0 to 1 − i,

(2) γ is the broken line connecting in order the numbers 0, 1 and 1 − i.

Solution 1 We take the parametrization γ(t) = t (1−i), t ∈ [0, 1]. Then γ  = 1−i
and the line integral is
  1
. z dz = t (1 + i)(1 − i) dt = 1.
γ 0

Solution 2 Now we parameterize the the broken line by


t, t ∈ [0, 1],
. γ(t) =
1 − i(t − 1), t ∈ [1, 2].

So,

1, t ∈ [0, 1],
γ  (t) =
.
−i, t ∈ (1, 2],

and the integral becomes

  1  2 1 1
. z dz = t dt + (1 + i(t − 1))(−i) dt = + − i = 1 − i.
γ 0 1 2 2

Example 4.2 Let γ : [a, b] → C be a piecewise smooth curve; n ∈ Z \ {−1}, and

moreover, if n < 0 then we also assume that {0} ∈
/ Eγ . Then
  b 1  n+1 
. zn dz = (γ(t))n γ  (t) dt = γ (b) − γ n+1 (a) .
γ a n+1

Example 4.3 Let z = γ(t) = a + reit , t ∈ [0, 2π ], where a ∈ C, r > 0; n ∈ Z.

By Definition 4.1, we find
  
2π
2π i, n = −1,
. (z − a) dz = r
n n+1
i e i(n+1)t
dt =
γ 0 0, n = −1.
4.1 Line Integrals and Their Simplest Properties 83

Important 
Example 4.1 shows that the value of γ z dz depends on the curve along which
the integration takes place.
On the other hand, the integral of the function zn (n = −1) in Example 4.2
is not depend on the integration curve
 and is determined only by its beginning
and end. And if γ is closed, then γ zn dz = 0.
From Example 4.3 it follows that there exists a function whose integral,
even over a closed curve, is not equal to zero.
Why this happens, and when line integrals do not depend on the integration
curve, but only on its start and end points, is what we need to find out in this
chapter.

Remark 4.1 Separating the real and imaginary parts in

   
.γ  (t) = x  (t) + iy  (t) and f (γ(t)) = u x(t), y(t) + iv x(t), y(t) ,

the integral (4.1) can be represented as the sum of two curvilinear integrals of the
second kind
  
. f (z) dz = u dx − v dy + i u dy + v dx. (4.2)
γ γ γ

Now we list the main properties of line integrals of complex-valued functions.

Later in this section, we assume that integrand functions are continuous and curves
are piecewise smooth.
1. Linearity. For any {λ, μ} ∈ C
  
 
. λ f (z) + μ g(z) dz = λ f (z) dz + μ g(z) dz.
γ γ γ

The proof follows directly from Definition 4.1 and the linearity of the Riemann
integral.
2. Additivity. Let z = γ1 (t), t ∈ [a, b], and z = γ2 (t), t ∈ [b, c], are piecewise
smooth curves, for which γ1 (b) = γ2 (b). The union of these curves is called the
curve

γ1 (t), t ∈ [a, b],
.γ1 ∪ γ2 := (4.3)
γ2 (t), t ∈ [b, c].
84 4 Integration of Functions of a Complex Variable

Then
  
. f (z) dz = f (z) dz + f (z) dz.
γ1 ∪ γ2 γ1 γ2

The proof follows directly from Definition 4.1 and the additivity of the Riemann
integral.
3. Orientability. Let z = γ(t), t ∈ [a, b], is a given curve. Denote by γ − the
curve

γ − (τ ) := γ(a + b − τ ),
. τ ∈ [a, b].

It is easy to see that the curves γ and γ− have the same trace, but opposite
orientations (the initial and end points are switched).
Then
 
. f (z) dz = − f (z) dz. (4.4)
γ− γ

Proof Since (γ − )τ (τ ) = γt (a + b − τ ) · (−1),

  b  
. f (z) dz = f γ − (τ ) (γ − )τ (τ ) dτ
γ− a
 b
= f (γ(a + b − τ )) · γt (a + b − τ ) · (−1) dτ = t = a + b − τ
a
 b 
=− f (γ(t)) · γt (t) dt =− f (z) dz.
a γ

This property also follows from (4.2) and the fact that curvilinear integrals of the
second kind change sign when changing the orientation of the curve.
4. Invariance. If a curve z = γ1 (t), t ∈ [a1 , b1 ], is equivalent to a curve z =
γ2 (τ ), τ ∈ [a2 , b2 ] (see Definition 1.11; moreover, the function μ, realizing this
equivalence is assumed to be continuously differentiable), then
 
. f (z) dz = f (z) dz.
γ1 γ2

Proof By the definition of the integral,

  b2
. f (z) dz = f (γ2 (τ ))(γ2 )τ (τ ) dτ.
γ2 a2
4.1 Line Integrals and Their Simplest Properties 85

Substituting τ = μ(t) in the integral and taking into account that γ2 (μ(t)) =
γ1 (t), t ∈ [a1 , b1 ], we get
 
b2 b1    
. f (γ2 (τ ))(γ2 )τ (τ ) dτ = f γ2 (μ(t) (γ2 )τ μ(t) μt (t) dt
a2 a1 

γ1 (t)

 b1 
= f (γ1 (t)) · γ1 (t) dt = f (z) dz.
a1 γ1

Remark 4.2 Since equivalent curves have the same trace and the same orientation,
it is often, if it does not cause misunderstanding, one speaks of the integral over the
trace and indicates its orientation. So, the result of Example 4.3 can be rewritten as
 
2π i, n = −1,
. (z − a) dz =
n
0, n = −1.
{|z−a|=r}+

Here “+” under the integral sign indicates the positive orientation of the circle.

5. Estimation of the integral. The following inequality holds:

  
 
 f (z) dz ≤ |f (z)| dl, (4.5)
.
 
γ γ


where dl = (x  (t))2 + (y  (t))2 dt = |γ  (t)|dt is the arc length differential; on the
right in (4.5) there is a curvilinear integral of the first kind of the function |f | along
the curve γ.

Proof Let us denote by I := γ f (z) dz and write this number in the exponent form
I = |I|eiθ . Then
  b
| I | = e−iθ I =
. e−iθ f (z) dz = e−iθ f (γ(t)) · γ  (t) dt.
γ a

Since the integral on the right is a real nonnegative number and |Rez| ≤ |z|, then
 b    b 
|I | =
. Re e−iθ f (γ(t)) · γ  (t) dt ≤ |f (γ(t))| · |γ  (t)| dt = |f | dl.
a a γ
86 4 Integration of Functions of a Complex Variable

The estimate (4.5) implies the corollary.

Corollary 4.1 If the modulus of a function f is bounded by a constant M on the

trace of a curve γ, i.e., ∃ M > 0 ∀ z ∈ Eγ : |f (z)| ≤ M, then
   
 
 f dz ≤ |f | dl ≤ M dl = M γ ,
.
 
γ γ γ

where γ is the length of γ.

4.2 An Antiderivative: Cauchy-Goursat Theorem

As with functions of a real variable, we introduce the following definition of an

antiderivative (a primitive) for a function f defined in a domain .Ω ⊂ C.

Definition 4.2 Let .F ∈ A(Ω) and .f ∈ C(Ω). The function .F is called an

antiderivative of .f in .Ω if

F  (z) = f (z) for all z ∈ Ω.

.

.

Obviously, if .F is an antiderivative of .f in .Ω, then for any complex number .c

the function .F + c is also an antiderivative of .f in .Ω.
Let .F1 and .F2 be two antiderivatives of .f in .Ω. Then .F1 − F2 =: G ∈ A(Ω)
and certainly .G (z) = 0 for all .z ∈ Ω. Denote by .u and .v the real and respectively
the imaginary part of the function .G. Using Theorem 2.2, the last equality can be
rewritten as
∂u(x, y) ∂v(x, y) ∂v(x, y) ∂u(x, y)
0 = G (z) =
. +i = −i ,
∂x ∂x ∂y ∂y

whence
⎧
⎪ ∂u ∂u 
⎪
⎨ (x, y) = (x, y) = 0 for all (x, y) ∈ Ω,
∂x ∂y u ≡ c1 ,
. ⇒
⎪ ∂v
⎪ ∂v v ≡ c2 .
⎩ (x, y) = (x, y) = 0 for all (x, y) ∈ Ω,
∂x ∂y

Thus, .G ≡ c := c1 + ic2 .
4.2 An Antiderivative: Cauchy-Goursat Theorem 87

Theorem 4.1 (The Cauchy–Goursat Theorem for Triangles) If .f ∈ A(Ω), then

for any triangle ., which together with its closure belongs to the domain .Ω (we
write this fact as follows . ⊆ Ω), we have

. f (z) dz = 0,
∂+

where .∂ +  is the positively oriented triangle boundary.

Proof The proof is by contradiction. Suppose that there exist a positive number M
and a triangle .0 ⊆ Ω such that
 
 
 f (z) dz = M > 0. (4.6)
.

∂ + 0

Divide the triangle .0 by the middle lines into four triangles .1 , .2 , .3 , .4 with
positively oriented boundaries as in Fig. 4.1. Then

 4 

. f (z) dz = f (z) dz, (4.7)
∂ + 0 +
k=1 ∂ k

because on the right side of (4.7) the integrals along the middle lines are taken twice,
but in opposite directions, and by (4.4) their sum is zero.
It follows from (4.6) and (4.7) that there exists a number .k ∈ {1, 2, 3, 4} (let for
definiteness .k = 1) such that
 
  M
 
.
 + f (z) dz ≥ 4 .
∂ 1

Fig. 4.1 Subdividing of a

triangle
88 4 Integration of Functions of a Complex Variable

Let us do the same procedure with triangle .1 and subdivide it into four triangles
11 , .12 , .13 , .14 . It possible to find one of them, call it .11 , for which
.

 
  M
 f (z) dz ≥ 2 .
.
 4
∂ + 11

Evidently, that .11 ⊂ 1 .

Continuing the same considerations, we obtain a sequence of nested triangles

0 ⊃ 1 ⊃ 11 ⊃ 111 . . .
.

such that for the nth triangle .(n) := 11 . . . 1 the inequality

n

 
  M

. f (z) dz ≥ n (4.8)
∂ + (n) 4

holds and the intersection of their closures is some point

+∞

{z0 } =
. (n) ,
n=0

which obviously belongs to the domain .Ω. Moreover, the perimeter .((n) ) is equal
to
1 1
((n) ) =
. ((n−1) ) = . . . = n (0 ). (4.9)
2 2
Since .f is differentiable at .z0 , the following statement is satisfied: for any .ε > 0
there are a number .δ > 0 and a function .α such that for all .z ∈ Bδ (z0 )

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

. and |α(z)| < ε. (4.10)

From the sequence .{(n) }n∈N0 we choose a triangle .(k) that belongs to the disk
.Bδ (z0 ). Then, on the one hand, the inequality (4.8) holds for this triangle. On the

other hand, based on (4.10) and (4.9) we have

    
M   
. ≤  
f (z) dz = f (z0 ) 
dz + f (z0 ) (z − z0 ) dz
4k ∂ + (k) ∂ + (k) ∂ + (k)
   
  
+ α(z)(z − z0 )dz =  α(z)(z − z0 )dz
∂ + (k) + (k) ∂ 
4.3 Local Existence of an Antiderivative: Antiderivative Along a Curve 89

  2  2
 2 ((k) ) ((k) )
≤ ε ( (k)
) dl = ε ( (k)
) =ε =ε .
∂(k) 2k 4k

Here, in the first line, the integrals of .1 and .z − z0 along the boundary of .(k) are
equal to zero (see Example 4.2). Thus,
 2
0 < M ≤ ε (∂0 )
.

for any .ε > 0. It means that .M = 0. But this is a contradiction.

4.3 Local Existence of an Antiderivative: Antiderivative Along

a Curve

The question on the existence of an antiderivative in the whole domain is more

complicated and will be considered in the next section. For now, we just note that not
every analytic function has an antiderivative in the whole domain (see Remark 4.6).
This statement does not agree with the fact from real analysis. Indeed, it is known
x
that any continuous function .ϕ on an interval has the primitive .F (x) = x0 ϕ(t) dt.

Theorem 4.2 (On the Local Existence of an Antiderivative) Let .f ∈ A(Ω).

Then there is an antiderivative of f in an arbitrary disk .Br (a) ⊂ Ω, which is
determined by the formula

F (z) :=
. f (ξ ) dξ, z ∈ Br (a). (4.11)
[a,z]

In (4.11) the integral is taken along the segment .[a, z] from a to z.

Proof Let us consider an arbitrary disk .Br (a) ⊂ Ω, then fix a complex number
z ∈ Br (a) and take any complex number .z such that .z +z ∈ Br (a).
.

Evidently, the closure of the triangle .a,z,z+z with vertices at the points .a, .z
and .z +z belongs to the domain .Ω. Therefore, due to Theorem 4.1,

. f (ξ ) dξ = 0,
∂ + a,z,z+z

wherefrom
  
. f (ξ ) dξ + f (ξ ) dξ + f (ξ ) dξ = 0.
[a,z] [z,z+z] [z+z,a]
90 4 Integration of Functions of a Complex Variable

According to the notation (4.11) and the orientability property, the last equality can
be rewritten as

.F (z+  z) − F (z) = f (ξ ) dξ.
[z,z+z]

Dividing this equality by .z and subtracting .f (z), we get

   
 F (z+  z) − F (z)   
 − f (z) = 1 ·  (f (ξ ) − f (z)) dξ  . (4.12)
.
 z  | z |
[z,z+z]

Here we use the identity


1
f (z) =
. f (z) dξ.
z [z,z+z]

Since f is continuous, the following statement holds: for any .ε > 0 there is a
positive number .δ that for all .z ∈ C such that .| z | < δ and for all .ξ ∈ [z, z + z]
we have

|f (ξ ) − f (z)| < ε.
.

Therefore, taking (4.12) into account, we get

  
 F (z+  z) − F (z)  1  
 
− f (z) ≤ f (ξ ) − f (z) dl < ε.
.
 z |  z | [z,z+z]

This means that

F (z+  z) − F (z)
. lim = f (z).
z→0 z

In view of the fact that z is an arbitrary point in the disk .Br (a), the function F
defined by the formula (4.11) is a primitive of f in .Br (a).

Remark 4.3 The formula (4.11) cannot always be extended for the whole domain
Ω, for example, if .Ω is not simply connected. Therefore, the equality (4.11) is said
.

to define only the local antiderivative for the function f .

Remark 4.4 When proving this theorem, two facts were used, namely

• continuity of .f, 
• and for an arbitrary triangle . that . ⊆ Ω, it is necessary . ∂ +  f dξ = 0.
4.3 Local Existence of an Antiderivative: Antiderivative Along a Curve 91

Therefore, instead of the analyticity of f in Theorem 4.2, we can require the above
properties for the function .f to prove this theorem. .

Now consider the intermediate concept of antiderivative between the notions of

local antiderivative and antiderivative in the whole domain, namely the antideriva-
tive along a curve.

Definition 4.3 Let .f ∈ C(Ω) and .γ : [a, b] → C be a curve, whose trace .Eγ
belongs to the domain .Ω.
A continuous function .Ψ : [a, b] → C is called an antiderivative of the function
f along the curve .γ, if in some neighborhood of each point on .Eγ there is an
antiderivative of .f, whose restriction to the corresponding part of the trace .Eγ
coincides with .Ψ, i.e.,
 
∀ t0 ∈ [a, b] ∃ δ > 0 ∃ r > 0 ∃ Ft0 ∈ A Br (γ(t0 )) such that
.


Ft0 (z) = f (z) for all z ∈ Br (γ(t0 )) ⊂ Ω,
.
Ft0 (γ(t)) = Ψ (t) for all t ∈ (t0 − δ, t0 + δ) ∩ [a, b].

Important
This definition does not require the existence of an antiderivative of f in the
whole domain .Ω. Moreover, the antiderivative .Ψ along the curve .γ does not
necessarily have a derivative (it is only continuous).

Remark 4.5 It is easy to understand that if there is an antiderivative .F of f in .Ω,

then for an arbitrary curve .γ : [a, b] → Ω the function .F (γ(t)), t ∈ [a, b], is an
antiderivative of f along the curve .γ.

Example 4.4 Let us find an antiderivative of .cos z along the curve

γ(t) = t + i|t|,
. t ∈ [−1, 1].

Since .sin z is an antiderivative of .cos z in .C, the function

Ψ (t) = sin(t + i|t|) = sin t cosh |t| + i cos t sinh |t|,

. t ∈ [−1, 1],

is an antiderivative of .cos z along .γ. Evidently, .Ψ ∈

/ C 1 ([−1, 1]).

Theorem 4.3 (The Existence of an Antiderivative Along a Curve) If .f ∈ A(Ω),

then there exists an antiderivative of the function f along any curve .γ whose trace
belongs to the domain .Ω.
92 4 Integration of Functions of a Complex Variable

Fig. 4.2 Construction an antiderivative along a curve

This antiderivative is determined up to an additive constant.

Proof Let .z = γ(t), t ∈ [a, b], be a curve, whose trace .Eγ is in .Ω. Then, due to
Theorem 4.2, for every point on .Eγ there exists a disk centered at that point, which
belongs to .Ω and in which there exists a primitive of f . Thus, we have obtained an
infinite cover of .Eγ by disks. Since .Eγ is a compact, from this cover one can choose
a finite subcover .{K1 , .. . , Kn }. According
n to this subcover, we divide the segment
.[a, b] by the segments . Im := [tm , tm ]
 as follows:
m=1

.t1 = a, tn = b, 
tm < tm+1 < tm , γ(Im ) = Eγ ∩ Km

(see Fig. 4.2). A more general case, for example, when a curve has self-intersection
points, is considered in [4, 9].
Now we fix an antiderivative .F1 of f in the disk .K1 . If we consider an arbitrary
antiderivative of f in the disk .K2 , then at the intersection of .K1 ∩K2 these primitives
can differ only by some constant, since they are two antiderivatives of the function
f in .K1 ∩ K2 . Therefore, there exists a unique antiderivative .F2 of f in .K2 such that
.F2 ≡ F1 in .K2 ∩ K1 . Continuing these considerations, in each disk .Km we choose

the unique antiderivative .Fm of .f such that

Fm ≡ Fm−1
. in Km ∩ Km−1 , m ∈ {2, . . . , n}.

Now we can determine the function

Ψ (t) := Fm (γ(t)),
. t ∈ Im , m ∈ {1, . . . , n}. (4.13)
4.3 Local Existence of an Antiderivative: Antiderivative Along a Curve 93

Then .Ψ is an antiderivative of the function f along the curve .γ. In fact, .Ψ ∈

C([a, b]) by construction. Also, for every .τ0 ∈ [a, b] there is a disk .Br (γ(τ0 ))
belonging to some disk .Km0 from the subcover .{K1 , . . . , Kn }, and there is an
antiderivative .Fm0 of f in .Br (γ(τ0 )) for which, thanks to (4.13), we have

Fm0 (γ(t)) = Ψ (t),

. t ∈ (τ0 − δ, τ0 + δ) ∩ [a, b]

for some .δ > 0.

Now we prove the second part. Let .Ψ1 and .Ψ2 be two antiderivative of .f along
.γ. Consider the function .ϕ(t) := Ψ1 (t) − Ψ2 (t), t ∈ [a, b]. Let us show that it is

locally constant. Then, taking into account its continuity on .[a, b], this means that
.ϕ ≡ const on .[a, b].

Based on Definition 4.3, we have the following statement:

(1) (2)
. ∀ t0 ∈ [a, b] ∃ δ > 0 ∃ r > 0 ∃ antiderivatives Ft0 and Ft0

of .f in the disk . Br (γ(t0 )) such that

Ψ1 (t) = F (1) (γ(t)),

. Ψ2 (t) = F (2) (γ(t)), t ∈ Iδ := (t0 − δ, t0 + δ) ∩ [a, b].

(1) (2)
Since .Ft0 = Ft0 + C in .Br (γ(t0 )), the function .ϕ is constant in .Iδ .

Theorem 4.4 (Analog of the Newton-Leibniz Formula) Let .γ : [a, b] → Ω be a

piecewise smooth curve. If a function f is continuous in the domain .Ω and has an
antiderivative .Ψ along .γ, then

. f (z) dz = Ψ (b) − Ψ (a). (4.14)
γ

Proof Let us first consider the case when .γ is a smooth curve and its trace belongs
to some disk .K ⊂ Ω, in which there exists an antiderivative F of f . Then the
composition .F (γ(t)), .t ∈ [a, b], is an antiderivative of the function .f along the
curve .γ, and from the second part of Theorem 4.3 it follows that

.Ψ (t) = F (γ(t)) + C, t ∈ [a, b].

Since .F  = f in K and .γ is a smooth curve,

Ψ  (t) = F  (γ(t)) · γ  (t) = f (γ(t)) · γ  (t) for all t ∈ [a, b].

.

Therefore,
  b  b

. f (z) dz = f (γ(t)) · γ (t) dt = Ψ  (t) dt = Ψ (b) − Ψ (a). (4.15)
γ a a
94 4 Integration of Functions of a Complex Variable

In the general case, by virtue of the assumptions of the theorem and Defini-
tion 4.3, the curve .γ can be decomposed into a finite number of smooth curves

. γm : [μm , μm+1 ] → Ω, m ∈ {0, 1, . . . , n − 1} (a = μ0 < . . . < μn = b),

so that the trace of each curve .γm belongs to a disk .Km in which there exists an
antiderivative of f .
Then, due to (4.15) we get

 n−1 
 n−1 
 
. f (z) dz = f (z) dz = Ψ (μm+1 ) − Ψ (μm ) = Ψ (b) − Ψ (a).
γ m=0 γm m=0

Remark 4.6 This theorem makes it possible to verify that not every analytic
function in a domain has an antiderivative in that domain.
Consider, for example, the function

1
.f (z) = , z ∈ Ω = {z : 0 < |z| < 2}.
z

Obviously, .f ∈ A(Ω). Assume that there is an antiderivative F of .f in .Ω.  

Take the circle .γ(t) = eit , t ∈ [−π, π ] (.Eγ ⊂ Ω). Then the function .F γ(t) ,
.t ∈ [−π, π ], is an antiderivative of .f along .γ, and according to Theorem 4.4,


. z−1 dz = F (eiπ ) − F (e−iπ ) = F (−1) − F (−1) = 0.
γ

On the other hand, given the result of Example 4.3, we have


. z−1 dz = 2π i. (4.16)
γ

This contradiction indicates that there is no antiderivative of f in .Ω.

However, the function f has an antiderivative along .γ thanks to Theorem 4.3.
Due to (3.9) it is equal to

Ψ (t) = ln |eit | + i arg(eit ) + i2π k = it + i2π k,

. t ∈ [−π, π ],

for any fixed .k ∈ Z. Therefore, using (4.14), we get the result


. z−1 dz = Ψ (π ) − Ψ (−π ) = 2π i
γ

that coincides with (4.16).

4.3 Local Existence of an Antiderivative: Antiderivative Along a Curve 95

Remark 4.7 The Newton-Leibniz formula (4.14) and Theorem 4.3 make it possible
to generalize the concept of the integral for an analytic function f in a domain .Ω
along an arbitrary curve .γ : [a, b] → Ω. Recall that Definition 4.1 requires the
piecewise smoothness of .γ. Since every analytic function f has an antiderivative
.Ψ along any continuous curve .γ, the integral of f along .γ can be determined as

follows

def
. f (z) dz = Ψ (b) − Ψ (a). (4.17)
γ

Based on Remark 4.5 and Theorem 4.4, we have the statement.

Proposition 4.1 If a function .f : Ω → C is continuous in .Ω and has an

antiderivative F in .Ω, then for any two points .{z1 , z2 } ⊂ Ω

. f (ξ ) dξ = F (z2 ) − F (z1 ),
z
1 ,z2

where .z1 , z2 is an arbitrary piecewise smooth curve with initial point .z1 and the end
point .z2 , and its trace is in the domain .Ω.

Taking Remark 4.7 and Proposition 4.1 into account and using the equalities

 
.fg  = (f g) − f  g, f (ξ )g(ξ ) dξ = f (z2 )g(z2 ) − f (z1 )g(z1 ),
z
1 z2

we deduce the statement.

Proposition 4.2 If .f, g ∈ A(Ω), then for any .{z1 , z2 } ⊂ Ω

 

. f (ξ ) g (ξ ) dξ = f (z2 )g(z2 ) − f (z1 )g(z1 ) − f  (ξ ) g(ξ ) dξ,
z
1 ,z2 z
1 ,z2

where .z1 , z2 is an arbitrary curve with initial point .z1 and the end point .z2 , and its
trace is in .Ω.

It follows from the above statements that the methods and formulas for inte-
grating complex-valued functions of a complex variable remain the same as for
functions of a real variable.

Example 4.5 For any .z1 , z2 ∈ C


. eξ dξ = ez2 − ez1 .
z
1 ,z2
96 4 Integration of Functions of a Complex Variable

Sometimes curvilinear integrals of the second kind can be calculated using the
Newton-Leibniz formula (4.14).

Example 4.6 Let .γ(t), .t ∈ [a, b], be a piecewise smooth curve in .R2 with initial
point .(0, 0) and the end point .(1, 1). Then due to (4.2)
  
. sin x cosh y dx − cos x sinh y dy = Re sin z dz
γ γ
   
. = Re − cos(γ(b)) + cos(γ(a)) = Re − cos(1 + i) + cos 0 = 1 − cos 1 cosh 1.

Exercise 4.1 Find an antiderivative for the function .cosh z along the segment
[0, 1 + i] with the orientation from 0 to .1 + i, and by the Newton-Leibniz formula
.

calculate
 
.Re cosh z dz .
[0,1+i]

4.4 The Cauchy Integral Theorem and Corollaries

From the formula (4.17) and proving Theorem 4.3 it is visible that the integral of
an analytical function along a curve will not change when the curve is continuously
deformed so that its ends remain in place and its trace remains in the subcover
.{K1 , . . . , Kn }. How to understand the continuous deformation of a curve? For this

purpose, we recall some definitions and facts from differential geometry. We will
assume that curves considered in this section are defined on the closed interval .I :=
[0, 1]; this can always be done with the admissible change of a variable without
leaving the curve equivalence class.

Definition 4.4 (Homotopic Curves)

• Two curves

γ0 : I → Ω
. and γ1 : I → Ω

with the same start point .a = γ0 (0) = γ1 (0) and end point .b = γ0 (1) = γ1 (1)
.(a = b) are called homotopic in a domain .Ω (denoted as .γ0 ≈ γ1 in .Ω) if there
exists a continuous map .ϕ : I ×I → Ω with the following properties (Fig. 4.3):

(1) .ϕ(0, t) = γ0 (t) and ϕ(1, t) = γ1 (t) for all t ∈ I ;

(2) .ϕ(s, 0) = a and ϕ(s, 1) = b for all s ∈ I .
4.4 The Cauchy Integral Theorem and Corollaries 97

Fig. 4.3 Homotopic curves

• Similarly, two closed curves .γ0 : I → Ω and .γ1 : I → Ω are said to be

homotopic in a domain .Ω, if the previous first condition is satisfied and the
second one reads as follows:

ϕ(s, 0) = ϕ(s, 1)
. for all s ∈ I.

If .γ1 ≡ const = γ1 (0) ∈ Ω .(γ1 is a constant curve), then we write that .γ0 ≈ 0
in .Ω.

Remark 4.8 The first part of Definition 4.4 means that for each .s ∈ I the curve

γs := ϕ(s, t), t ∈ I,
. (in red color in Fig. 4.3) (4.18)

has the same initial and end points (.γs (0) = a and .γs = b), its trace belongs to the
domain .Ω, and in addition, the family.{γs }s∈I is a continuous deformation of .γ0 into
.γ1 inside of .Ω such that the endpoints are fixed during deformation.

The same interpretation applies to the second part of Definition 4.4, but now all
curves in the family.{γs }s∈I must be closed, and the start and end points, which now
coincide, can move in .Ω without leaving .Ω.
If .γ1 is a constant curve, i.e., its trace is a point in the domain .Ω, then we say
that the curve .γ0 can be continuously deformed to this point while remaining in .Ω,
and .γ0 is said to be null-homotopic in .Ω.

Example 4.7 Let .Ω be a convex domain, i.e, any two points in .Ω can be connected
by a segment that entirely belongs to .Ω. Then, any two closed curves .γ0 and .γ1 ,
whose traces belong to .Ω, are homotopic in .Ω.
Indeed, the homotopic function that deforms .γ0 into .γ1 is as follows
 
ϕ(s, t) = γ0 (t) + s γ1 (t) − γ0 (t) ,
. (s, t) ∈ I × I.
98 4 Integration of Functions of a Complex Variable

Example 4.8 Any closed curve .γ : I → C is homotopic in .C to a point .a ∈ C via

the homotopy

ϕ(s, t) = s a + (1 − s) γ(t),
. (s, t) ∈ I × I.

Example 4.9 The curve .γ(t) = ei2π t , t ∈ I, is not null-homotopic in .C \ {0}.

Exercise 4.2 Prove that the homotopy relation of curves is an equivalence relation,
that is, it is reflexive, symmetric, and transitive.

Exercise 4.3 Prove if .γ1 ∼ γ2 (see Definition 1.11) and .γ2 ≈ γ3 in .Ω, then .γ1 ≈
γ3 in .Ω.

Exercises 4.2 and 4.3 show that all curves in a domain .Ω with the same endpoints
(or closed curves) can be divided into homotopy classes, and equivalent curves fall
into the same homotopy class.

Exercise 4.4 Prove that a domain in .C is simply connected (see Definition 1.20) if
and only if an arbitrary closed curve is null-homotopic in this domain.

Exercise 4.5 Prove that an arbitrary closed curve is null-homotopic in a domain if

and only if two arbitrary curves with the same initial and end points are homotopic
in that domain.

Theorem 4.5 (Homotopy Version of the Cauchy Integral Theorem) If a func-

tion .f : Ω → C is analytic in a domain .Ω and .γ0 ≈ γ1 in .Ω, then
 
. f (z) dz = f (z) dz. (4.19)
γ0 γ1

Proof Let .ϕ : I × I → Ω be a homotopy between .γ0 and .γ1 . By Theorem 4.3, for
any .s0 ∈ I there is an antiderivative .Ψs0 of the function f along the curve .γs0 (t) :=
ϕ(s0 , t), t ∈ I . In addition, it follows from the construction of the antiderivative
.Ψs0 that there is an .ε0 > 0 such that a curvilinear strip

 
Uε0 (γs0 ) := {z ∈ C : dist z; Eγs0 < ε0 }
.

belongs to the union  of thedisks .{K1 , . . . , Kn } (see the Proof of Theorem 4.3 and
Fig. 4.2). Here, .dist z; Eγs0 is the distance from z to the trace .Eγs0 of .γs0 .
Due to the uniform continuity of .ϕ, there is a positive number .δ0 such that for all
.s ∈ (s0 − δ0 , s0 + δ0 ) ∩ I

|γs (t) − γs0 (t)| < ε0

. for all t ∈ I,
4.4 The Cauchy Integral Theorem and Corollaries 99

i.e., the trace .Eγs belongs to the curvilinear strip .Uε0 (γs0 ). Moreover, this means that
we can use the same functions .{F1 , . . . , Fn } to determine an antiderivative .Ψs of the
function f along the curve .γs as for defining .Ψs0 (see the Proof of Theorem 4.3).
As a result, by using the formula (4.17), we get
 
. f (z) dz = f (z) dz for all s ∈ (s0 − δ0 , s0 + δ0 ) ∩ I. (4.20)
γs0 γs

Since .I = [0, 1] is compact, it can be covered by a finite number of intervals, on

each of which equalities (4.20) hold. Then, starting from .s0 = 0, in a finite number
of steps we arrive at the equality (4.19).

Corollary 4.2 If .f ∈ A(Ω) and .γ ≈ 0 in .Ω, then . γ f (z) dz = 0.

Proof Since .γ ≈ 0 in .Ω, there exists a constant curve .γ1 ≡ const = a ∈ Ω such
that .γ ≈ γ1 in .Ω. Then due to Theorem 4.5, we have
   1
. f (z) dz = f (z) dz = f (γ1 (t)) γ1 (t) dt = 0.
γ γ1 0 

0

Corollary 4.3 If .f ∈ A(Ω) and the domain .Ω is simply connected, then for an
arbitrary closed curve .γ : I → Ω

. f (z) dz = 0.
γ

The proof follows directly from Exercise 4.4 and Corollary 4.2.

Definition 4.5 A function f is analytic in the closure of a domain .Ω (.f ∈ A(Ω)),

if there is a domain .G ⊃ Ω such that .f ∈ A(G).

Corollary 4.4 If .f ∈ A(Ω) and the domain .Ω is bounded and simply connected,
then

. f (z) dz = 0.
∂+Ω

Proof By Definition 4.5, there is a domain .G ⊃ Ω where .f ∈ A(G). Since .Ω is

simply connected and bounded, the curve, whose trace coincides with .∂Ω, is null-
homotopic in G. Next we should apply Corollary 4.2.
100 4 Integration of Functions of a Complex Variable

Fig. 4.4 Cuts of a domain .Ω


n
Corollary 4.5 Let .Ω be a bounded domain and .∂Ω = Eγk , where
k=0

• .{γk }nk=0 are Jordan curves whose traces are pairwise disjoint,
• for any .k ∈ {1, . . . , n} the trace .Eγk belongs to the interior of .γ0 ,
• for any .k ∈ {0, 1, . . . , n} the orientation of .γk coincides with the positive
orientation of the boundary .∂Ω.

If .f ∈ A(Ω), then

. f (z) dz = 0.
∂+Ω

Remark 4.9 The positive orientation of the boundary .∂ + Ω of a bounded domain

.Ω is such an orientation of the closed curves whose traces form the domain

boundary, in which the domain .Ω always remains on the left when traversing the
traces.
Figure 4.4 shows a domain .Ω with the positive orientation of the boundary,
which is described in conditions of Corollary 4.5. Note that the curve .γ0 is oriented
counterclockwise, and all other curves .{γk }nk=1 are clockwise.

Proof There exists a domain .G ⊃ Ω such that .f ∈ A(G). Let .λ− k and .λk
+

be oppositely oriented curves whose traces coincide and connect .Eγk with .Eγ0
(Fig. 4.4); here .k ∈ {1, . . . , n}.
Then with the help of the curves .{γk }nk=0 and .{λ±
k }k=1 we construct the closed
n

curve

Λ := γ0 ∪ (λ−
.
+ − + − +
1 ∪ γ1 ∪ λ1 ) ∪ (λ2 ∪ γ2 ∪ λ2 ) ∪ . . . ∪ (λn ∪ γn ∪ λn )
4.4 The Cauchy Integral Theorem and Corollaries 101

with the positive orientation. It is easy to see that the curve .Λ is null-homotopic in
G. Therefore, considering Corollary 4.2, we have
  n 
   
0=
. f dz = f (z) dz + f (z) dz + f (z) dz = f (z) dz.
∂+Ω k=1
Λ λ+ λ− ∂+Ω

k k


0

The corollary is proved.

Remark 4.10 The condition of Corollary 4.5 for the function f can be weaken,
namely .f ∈ A(Ω)∩C(Ω).
 However, in this case the curves .{γk } must be piecewise
smooth. Then also . ∂ + Ω f (z) dz = 0.

Theorem 4.6 (On the Global Existence of an Antiderivative) If .f ∈ A(Ω) and

the domain .Ω is simply connected, then there exists an antiderivative of f in .Ω.

Proof Since .Ω is simply connected, an arbitrary closed curve whose trace belongs
to .Ω is null-homotopic in .Ω. According to Exercise 4.5, this means that any curves
with the same endpoints, whose traces are in .Ω, are homotopic.
Then, by Theorem 4.5, the integral of f along a curve depends only on the initial
and end points of the curve, but not on the curve itself. Therefore, we can determine
the single-valued function

F (z) :=
. f (ξ ) dξ, z ∈ Ω, (4.21)

a,z

where .a , z is an arbitrary curve with initial point .a ∈ Ω and the end point .z and its
trace is in .Ω.
Now we fix arbitrary .z ∈ Ω and take any complex number . z such that the
segment .[z, z+  z] ⊂ Ω. Then the curve

Γ := (
. 
a , z) ∪ [z, z +  z] ∪ (z +  z, a)

is closed and its trace belongs to .Ω (Fig. 4.5).

Corollary 4.3 gives that

. f (ξ ) dξ = 0,
Γ

or
  
. f (ξ ) dξ + f (ξ ) dξ + f (ξ ) dξ = 0,

a,z [z,z+z] 
z+z,a
102 4 Integration of Functions of a Complex Variable

Fig. 4.5 The curve .Γ

whence, using (4.21), we get


F (z + Δz) − F (z) =
. f (ξ ) dξ.
[z,z+Δz]

Further, as in Theorem 4.2 (on the local existence of an antiderivative), we prove

that the derivative of F exists and .F  (z) = f (z).

When proving the theorem, we used only the continuity of the function f and the
vanishing of the integral along any closed curve. Therefore, the corollary is true.

Corollary 4.6 If a function f is continuous in a domain .Ω and the integral of f

along any closed piecewise smooth curve, whose trace belongs to .Ω, is equal to
zero, then the function

F (z) :=
. f (ξ ) dξ, z ∈ Ω,

a,z

is an antiderivative of f in .Ω. Here .a

, z is an arbitrary piecewise smooth curve with
initial point .a ∈ Ω and the end point .z and its trace is in .Ω.

It is clear that if a continuous function f in a domain .Ω has an antiderivative in

Ω, then the integral of f along an arbitrary closed piecewise smooth curve whose
.

trace lies in .Ω is equal to zero. Thus, we have the following theorem on necessary
and sufficient conditions for the existence of an antiderivative in the whole domain.

Theorem 4.7 Let a function f be continuous in a domain .Ω. There exists an

antiderivative of f in .Ω if and only if the integral of f along an arbitrary closed
piecewise smooth curve whose trace lies in .Ω is equal to zero.

Returning to Remark 4.6, we see that the sufficient condition of Theorem 4.7
does not hold for the function .z−1 in the domain .{z : 0 < |z| < 2}, since the integral
(4.16) does not vanish.
4.5 The Cauchy Integral Formula 103

4.5 The Cauchy Integral Formula

From the Cauchy integral theorem we derive an integral representation for an

analytic function, which has many important consequences.

Theorem
 4.8 (The Cauchy Integral Formula) Let .Ω be a bounded domain and
∂Ω = nk=0 Eγk , where
.

• .{γk }nk=0 are Jordan curves whose traces are pairwise disjoint,
• for any .k ∈ {1, . . . , n} the trace .Eγk belongs to the interior of .γ0 ,
• for any .k ∈ {0, 1, . . . , n} the orientation of .γk coincides with the positive
orientation of the boundary .∂Ω.

If .f ∈ A(Ω), then the following integral representation takes place:


1 f (ξ )
f (z) =
. dξ for any z ∈ Ω. (4.22)
2π i ∂+Ω ξ −z

Important
The formula (4.22) expresses a very interesting fact: the value of an analytic
function in a domain is completely determined by its values on the boundary
of that domain. Therefore, this representation is very often used in both
theoretical and applied problems.

Proof Let us fix any .z ∈ Ω. Clearly, that there is a positive number .r0 such that for
all .r ∈ (0, r0 ) the closed disk .Br (z) belongs to the domain .Ω. Consider the function

f (ξ )
. , ξ ∈ Ωr := Ω\Br (z).
ξ −z

Obviously, that this function is analytic in .Ωr . Then Corollary 4.5 yields
  
f (ξ ) f (ξ ) f (ξ )
.0= dξ = dξ − dξ. (4.23)
∂ + Ωr ξ −z ∂+Ω ξ −z ∂ + Br (z) ξ −z

Here the positive orientation of .∂Ωr means that the circle .∂Br (z) is oriented
clockwise, and this orientation is opposite to the positive orientation of the boundary
of the disk .Br (z). Therefore, the minus appeared before the last integral in (4.23). It
follows from (4.23) that for all .r ∈ (0, r0 )
 
1 f (ξ ) 1 f (ξ )
. dξ = dξ. (4.24)
2π i ∂+Ω ξ −z 2π i ∂ + Br (z) ξ −z
104 4 Integration of Functions of a Complex Variable

Since .f ∈ C(Ω), we have that

∀ ε > 0 ∃ δ > 0 ∀ ξ ∈ Bδ (z) :

. |f (ξ ) − f (z)| < ε.

Then, thanks to Example 4.3, for an arbitrary .r ∈ (0, min(δ, r0 ))

     
 1 f (ξ )   1 f (z) − f (ξ ) 

. f (z) − 
dξ =  dξ 
 2π i ∂ + Br (z) ξ − z   2π i ∂ + Br (z) ξ −z

1 |f (ξ ) − f (z)| 1 1
. ≤ dl < ε · · 2π r = ε.
2π ∂Br (z) |ξ − z| 2π r

This means that


1 f (ξ )
. lim dξ = f (z).
r→0 2π i ∂ + Br (z) ξ −z

Taking this fact into account and passing to the limit in (4.24), we obtain the Cauchy
integral formula (4.22).

The Cauchy formula remains correct if .f ∈ A(Ω) ∩ C(Ω) and the boundary of
the domain .Ω consists of a finite number of piecewise smooth closed curves.
It follows from Corollary 4.5 that

1 f (ξ )
. dξ = 0 for all z ∈ Ω. (4.25)
2π i ∂+Ω ξ −z

Using the Cauchy integral formula, it is sometimes possible to calculate integrals

from analytic functions along closed curves.

Example 4.10 Let us find the integral


sin ξ
I :=
. dξ,
γ ξ2 + 4

where .z = γ(t) = i + 2 exp(it), t ∈ [0, 2π ].

The integral can be rewritten as follows
    
1 1 1 1 sin ξ 1 sin ξ
I=
. − sin ξ dξ = dξ − dξ.
4i ξ − 2i ξ + 2i 4i ξ − 2i 4i ξ + 2i
γ γ γ
4.5 The Cauchy Integral Formula 105

The point 2i belongs to the interior of .γ and .−2i ∈ int(γ). Thus, thanks to (4.22)
and (4.25) we have
π π
I=
. sin(2i) = i sinh 2.
2 2
Theorem 4.9 (Mean Value Theorem) Let .f ∈ A(Ω). Then for any disk .BR (a),
which together with its closure belongs to the domain .Ω, the value of f at the
center of this disk is equal to the mean value of this function taken around the disk
boundary .∂BR (a), i.e.
 2π
1
.f (a) = f (a + R eit ) dt. (4.26)
2π 0

Proof For any .a ∈ Ω there is a positive number R such that .BR (a) ⊂ Ω. Then, by
virtue of the Cauchy integral formula, we have
 !
1 f (ξ )
.f (a) = dξ = ξ = a + Re , t ∈ [0, 2π ]
it
2π i∂ + BR (a) ξ − a
 2π  2π
1 f (a + R eit ) 1
= · Rie it
dt = f (a + R eit ) dt.
2π i 0 R eit 2π 0

The theorem is proved.

This theorem shows once again that analytic functions are very “nice functions”
(the value of an analytic function at each point is closely related to the values of this
function at neighboring points). In the next sections, using this fact, we will prove
many of their other remarkable properties.

Exercise 4.6 Does there exist an entire function that

• on the circle .{z : |z − 1 + i| = 4} takes the value .3i,

• and on the circle .{z : |z − 1 + i| = 3} takes the value .1 + i ?

Exercise 4.7 Let f be an analytic function in .{z : Im z ≥ 0} and .f (x) = 2 − i for

all .x ∈ R. Find the value of f at the points .1 + i and .2 + i.

Exercise 4.8 Using Theorem 4.9 and Exercise 2.2, prove the maximum modulus
principle, which states that if f is an analytic function, then its modulus cannot
have a strict local maximum in the domain of f .
In a different way, this principle, as well as other properties of the modulus of an
analytic function, will be proved in Sect. 9.1.
Complex Power Series
5

Abstract

The main goal of this chapter is to show that analytic functions can be represented
as infinite power series. The key to proving this theorem is the Cauchy integral
formula established in the previous section. Here we generalize this formula for
derivatives and prove the surprising fact that derivatives of analytic functions
can be calculated by integration. Conversely, we will establish that the sum of a
complex power series is an analytic function in the open disk where this series
converges. This fact is then used to prove that an analytic function is infinitely
differentiable. In addition, the reader can familiarize himself with the proofs of
such remarkable statements as Liouville’s theorem, Maurer’s theorem, and the
equivalence of three approaches to the construction of the theory of analytic
functions. In the last section, applications of power series representations lead us
to a statement about the coincidence of analytic functions when they coincide on
some sequence, and to a statement characterising the isolated zeros of an analytic
function and their concentration.

5.1 Basic Definitions and Properties of Function Series and

Power Series

The main properties of infinite complex series are the same as real ones. Neverthe-
less, we briefly recall the fundamental definitions and some properties.

Definition 5.1 Two sequences of complex numbers, one of which is a sequence

{an }n∈N , and the second is determined from the previous one as follows
.

 
n 
.Sn := ak
n∈N
k=1

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 107
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_5
108 5 Complex Power Series

are called a series of complex numbers and are denoted by one symbol

+∞

. ak . (5.1)
k=1

The sequence .{Sn }n∈N is said to be the sequence of partial sums of the series (5.1). If
{Sn }n∈N converges to a number .S ∈ C, then we say that the series (5.1) is convergent
.

and the number S is the sum of (5.1); in this case, one also writes

+∞

S=
. ak .
k=1

If there is no finite limit of .{Sn }n∈N , then the series (5.1) is called divergent.
  i k
Example 5.1 The series . +∞
k=1 2 is convergent since

n   i n
 i k
i −1 i 1 2
. lim = lim 2
= = − + i,
n→+∞ 2 n→+∞ 2 i
2 −1 2−i 5 5
k=1

and its sum is equal to .− 15 + 25 i.

From Exercise 1.2 it follows that a series is convergent if and only if the
corresponding series of real and imaginary parts are convergent. Similarly, as for
real series, we prove that the sum and difference  of two convergent series are
convergent, and the necessary condition for . +∞ k=1 ak to be convergent is that
.limk→+∞ ak = 0.
 +∞
A series . +∞k=1 ak is said to be absolutely convergent if . k=1 |ak | converges; a
series is said to be conditionally convergent if it converges, but it does not converge
absolutely. It is easy to verify that an absolutely convergent series converges.
Let .{fn : Ω → C}n∈N be a sequence of functions. For each .z ∈ Ω consider the
series
+∞

. fn (z) (5.2)
n=1
n
and its sequence of partial sums .{Sn (z) := k=1 fn (z)}n∈N .

Definition 5.2 The set of such numbers .z ∈ Ω for which the series (5.2) converges
is called the convergence set of the function series (5.2).
5.1 Basic Definitions and Properties of Function Series and Power Series 109


Definition 5.3 A function series . +∞
n=1 fn (z) is called uniformly convergent on a
set .M ⊂ Ω to a function .f : M → C, if the sequence of its partial sums .{Sn (z), z ∈
M
Ω}n∈N uniformly converges to the function f on M .(Sn ⇒ f ), i.e.,

+∞

∀ ε > 0 ∃ n0 ∈ N ∀ n ≥ n0 ∀z ∈ M : |f (z) − Sn (z)| =
. fk (z) < ε.
k=n+1

As in real analysis, we establish the theorem.

Theorem 5.1 (Weierstrass Criterion for Uniform Convergence) Let for a

sequence of functions .{fn : Ω → C}n∈N the following conditions are satisfied:

(1) .∀ n ∈ N ∃ an ∈ R+ : sup |fn (z)| ≤ an ;

 z∈Ω
(2) the series . +∞
n=1 an is convergent.


+∞
Then the function series . fn (z) converges uniformly and absolutely on .Ω.
n=1

Also, without any changes one can prove the theorem about the continuity of
the sum of a function series and theorem about the term-by-term integration of a
uniformly convergent function series.

Definition 5.4 Let .{cn }n∈N be a sequence of complex numbers and .z0 ∈ C. A
power series centered at .z0 is called a function series of the form

+∞

. cn (z − z0 )n . (5.3)
n=0

The constants .{cn }n∈N are called the coefficients of this power series. 
.

For the power series, we define the value

1
. := lim sup n
|cn |. (5.4)
R n→+∞

Obviously, that .0 ≤ R ≤ +∞.

Theorem 5.2 (Cauchy–Hadamard Theorem) If .R = +∞, then the series (5.3)

is absolutely convergent in .C; when .R = 0 the series (5.3) converges only at the
point .z = z0 and diverges if .z = z0 .
110 5 Complex Power Series

If .R ∈ (0, +∞), the following statements hold:

(1) the series (5.3) converges absolutely at every point

z ∈ BR (z0 ) = {z ∈ C : |z − z0 | < R};

.

(2) the series (5.3) diverges at every point .z ∈ BR (z0 );

(3) the series (5.3) converges uniformly on any compact set .K ⊂ BR (z0 ).

The proof is carried out in exactly the same way as for real power series. The disk
BR (z0 ) is called the disk of convergence of the power series (5.3) and the number .R
.

is called the convergence radius of (5.3).

Note that the convergence set of a power series may also contain points that lie
on the boundary of the disk of convergence.

Example 5.2 Let us consider the following power series:

+∞
 +∞ n
 +∞ n

z z
(A)
. zn , (B) , (C) .
n n2
n=1 n=1 n=1

By using (5.4), we conclude that .B1 (0) = {z : |z| < 1} is the disk of convergence
for these series. However,

• the series .(A) diverges for all .z ∈ ∂B1 (0) (the necessary convergence condition
is not fulfilled);
• the series .(B) diverges at the point .z = 1 (this is the harmonic series) and for the
other points .z = eit ∈ ∂B1 (0) \ {1}, t ∈ (0, 2π ), this series

+∞ int
 +∞
 +∞

e cos(nt) sin(nt)
. = +i
n n n
n=1 n=1 n=1

converges only conditionally since its real and imaginary parts conditionally
converge due to the Dirichlet criterion:

. the sequence .{ n1 }n∈N converges monotonically to zero,

+∞
. and the sequence of partial sums of the series .
+∞ n=1 cos(nt) is bounded
for each .t ∈ (0, 2π ) (the same for .( n=1 sin(nt) );

• the series .(C) absolutely converges for all .z ∈ ∂B1 (0) based on the Weierstrass
 n 
criterion . nz 2 ≤ n12 for all .z ∈ ∂B1 (0) .
5.2 Expansion of a Differentiated Function Into a Power Series 111

5.2 Expansion of a Differentiated Function Into a Power Series

One of the main theorems of complex analysis is the following theorem.

Theorem 5.3 Let .Ω be a domain in .C and .f ∈ A(Ω). Then, in arbitrary disk

K := BR (z0 ) ⊂ Ω, the function f can be represented as the sum of the power
.

series
+∞

f (z) =
. cn (z − z0 )n for all z ∈ K, (5.5)
n=0

where
1 f (ξ )
cn =
. dξ , ∀ r ∈ (0, R). (5.6)
2π i (ξ − z0 )n+1
{|ξ −z0 |=r}+

Proof Take any point .z0 ∈ Ω and a such positive number R that the disk .K =
BR (z0 ) belongs to .Ω. Then, taking into account the Cauchy integral formula (4.22),
we have for any .z ∈ K that

1 f (ξ )
.f (z) = dξ ,
2π i γr ξ −z

where.γr = z0 + r exp(it), .t ∈ [0, 2π ], and r is a number from the interval . |z −
z0 |, R .
The function .(ξ − z)−1 , ξ ∈ Eγr , can be represented as sum of the following
series (the sum of an infinite geometric series):

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

This series uniformly converges on .Eγr , since

+∞

|z − z0 | |z − z0 | 1
. = =: q < 1 for all ξ ∈ Eγr , and qn = .
|ξ − z0 | r 1−q
n=0

Then the series

 1 +∞
1 f (ξ ) f (ξ )
. = · (z − z0 )n , (5.7)
2π i ξ − z 2π i (ξ − z0 )n+1
n=0
112 5 Complex Power Series

which is obtained from the previous one by the multiplication with the bounded
function . 2π1 i f (ξ ), .ξ ∈ Eγr , also converges uniformly on .Eγr .
Therefore, according to the theorem on the term-by-term integration of a function
series, we can integrate the series (5.7) term by term. As a result, we obtain the
representation (5.5), whose coefficients are defined by (5.6). It should be noted here
that integrals in (5.6) are independent of .r ∈ (0, R) due to the Cauchy integral
Theorem 4.5.

Remark 5.1 The series (5.5) is called a power series representation of f around
the point .z0 . It is clear that the radius of convergence of (5.5) is not less than the
distance from .z0 to the boundary of .Ω.

Remark 5.2 It follows from Theorem 5.3 that the expansion of an analytic function
in a power series around a given point is a necessary condition for the analyticity of
this function at this point (see Definition 2.6).

Corollary 5.1 (Cauchy’s Inequalities for the Coefficients) Let .f ∈ A(Br (z0 )). If
f is bounded by a constant M on .∂Br (z0 ), then the coefficients of the power series
(5.5) satisfy the inequalities

M
|cn | ≤
. for all n ∈ N0 . (5.8)
rn
Proof From (5.6), due to the boundedness of f and (4.5), we get

1 f (ξ ) 1 |f (ξ )| M
|cn | =
. dξ ≤ dl ≤ n .
2π i (ξ − z0 )n+1 2π |ξ − z0 |n+1 r
∂ + Br (z0 ) ∂Br (z0 )

Theorem 5.4 (Liouville’s Theorem) If an entire function f is bounded, then it is

constant.

Proof Using Theorem 5.3, for any .R > 0

+∞

f (z) =
. cn zn for all z ∈ BR (0).
n=0

Since f is bounded in .C, i.e. .∃ M > 0 .∀ z ∈ C : .|f (z)| ≤ M, it follows from (5.8)
that
M
|cn | ≤
. for all n ∈ N0 .
Rn
5.3 Analyticity of the Sum of a Power Series 113

Seeing that R can be taken arbitrarily large and .limR→+∞ RMn = 0 if .n ∈ N, we

have .cn = 0 for all .n ∈ N. This means that .f (z) ≡ c0 for all .z ∈ C.

One of the elegant applications of Liouville’s theorem is the following proof of

the fundamental theorem of algebra.

Theorem 5.5 Every polynomial .Pn (z) := cn zn + cn−1 zn−1 + . . . + c1 z + c0 with

complex coefficient, where .n ∈ N and .cn = 0, has at least one root.

Proof Assume that .Pn (z) = 0 for all .z ∈ C. Then, .f (z) := Pn1(z) is an
entire function. Moreover, f is bounded since .lim|z|→+∞ |Pn (z)| = +∞. So, by
Liouville’s theorem, the function f is constant, which yields a contradiction because
.Pn is not constant. Hence, .Pn must have a root in .C.

Exercise 5.1 Prove that if .f ∈ A(C), then .f is constant.

Exercise 5.2 Let .f ∈ A(C). Prove that

(1) if .limz→∞ f (z) = ∞, then the set .{z ∈ C : f (z) = 0} = ∅;

(2) if .Im f (z) > 0 for all .z ∈ C, then .f is constant.

5.3 Analyticity of the Sum of a Power Series

It turns out that the converse claim to Theorem 5.3 also holds.

Theorem 5.6 The sum

+∞

f (z) :=
. cn (z − a)n (5.9)
n=0

is an analytic function in the convergence disk .BR (a) of the series (5.9).
Moreover, the derivative of f is calculated by the formula

+∞


.f (z) = n cn (z − a)n−1 for all z ∈ BR (a). (5.10)
n=1

Proof Since .BR (a) is the disk of convergence of the series (5.9), the radius R is
determined with the formula (5.4). It is easy to verify that this disk is also the disk
of convergence for the following series:

+∞

.φ(z) := n cn (z − a)n−1 . (5.11)
n=1
114 5 Complex Power Series

Due to the Cauchy–Hadamard Theorem 5.2 and theorem on the continuity of

the sum of a function series, the function .φ is continuous in .BR (a). In addition,
the series (5.11) uniformly converges on .∂, where . is an arbitrary triangle
that, together with its closure, belongs to .BR (a). Using the theorem on termwise
integration of a function series and the Cauchy-Goursat Theorem 4.1 for triangles,
we get

+∞

. φ(z)dz = n cn (z − a)n−1 dz = 0.
∂+ n=1 ∂+

Then, according to Remark 4.4, the function .φ has

 an antiderivative .Ψ in the disk
BR (a), which is defined by the formula .Ψ (z) = [a,z] φ(ξ )dξ, .z ∈ BR (a).
.

On the other hand,

+∞
 +∞

. φ(ξ ) dξ = ncn (ξ − a) n−1
dξ = cn (z − a)n = f (z) − c0
[a,z] n=1 [a,z] n=1

for all .z ∈ BR (a), whence we obtain that .f = c0 + Ψ in .BR (a). Thus, the function
f is also an antiderivative for .φ in .BR (a), i.e., .f ∈ A(BR (a)) and

+∞

f  (z) = φ(z) =
. n cn (z − a)n−1 for all z ∈ BR (a).
n=1

Corollary 5.2 The derivative of an analytic function in a domain .Ω is also analytic

in this domain.

Proof Let .f ∈ A(Ω). Consider an arbitrary point .z0 ∈ Ω and a disk .Br (z0 ) that
belongs to .Ω. Due to Theorem 5.3 the function f is represented as the sum of a
power series

+∞

.f (z) = cn (z − z0 )n , z ∈ Br (z0 ). (5.12)
n=0

Using the second claim of Theorem 5.6, the function f has the derivative that
is obtained from (5.12) by termwise differentiation, i.e., the derivative .f  is also
represented as a power series in the same disk. Then we apply the first assertion of
Theorem 5.6 to .f  and obtain that .f  ∈ A(Br (z0 )). Since .z0 is an arbitrary point of

.Ω, the function .f is analytic in .Ω.

Applying Corollary 5.2 to .f  gives the statement.

5.4 Uniqueness of Power Series Expansions: Morera’s Theorem 115

Corollary 5.3 Any analytic function in a domain .Ω has derivatives of all orders in
Ω.
.

Corollary 5.2 implies the following necessary condition for the existence of an
antiderivative.

Corollary 5.4 If a continuous function .f in a domain .Ω has an antiderivative, then

f is analytic in .Ω.
.

Remark 5.3 From Theorems 5.3 and 5.6 it follows that the representation of a
function as the sum of a convergent power series in some disk is a necessary and
sufficient condition for the analyticity of this function in this disk.

However, the convergence of a power series at points on the boundary of its disk
of convergence is not related to the analyticity of its sum at those points. Indeed, let
us return to Example 5.2.
z
The sum of the series .(A) is equal to . 1−z in the disk .B1 (0) and it diverges at
z
each point of the boundary .∂B1 (0). But the function . 1−z is analytic in the domain
.C \ {1}.

The series .(C) absolutely converges to some function f in the closed disk .B1 (0).
Assuming that .f is analytic at the point .z = 1, by Corollary 5.2 the derivative .f 
must be also analytic at .z = 1. Thus, there must be a finite limit

. lim f  (z) = f  (1).

z→1−0, Im z=0

However, since
+∞ n−1
 z
.f  (z) = for all z ∈ B1 (0),
n
n=1

+∞
1  (−1)n−1 (−x)n 1
. lim f  (x) = − lim = − lim ln(1 − x) = +∞.
x→1−0 x→1−0 x n x→1−0 x
n=1

This contradiction indicates that .f can not be analytic at .z = 1.

5.4 Uniqueness of Power Series Expansions: Morera’s

Theorem

A natural question that arises is: if for a given function we somehow get a kind of
power series expansion about a given point, will this be the only expansion? For
example, is the right-hand side of the identity

z2 + 1 = 2 + 2(z − 1) + (z − 1)2
. (5.13)
116 5 Complex Power Series

the unique power representation of the function .z2 + 1 around the point 1? The
answer is given by the following theorem.

Theorem 5.7 If a function f is equal to the sum of a power series in a disk .Br (z0 ),
then this series is its Taylor series, i.e., if

+∞

f (z) =
. cn (z − z0 )n for all z ∈ Br (z0 ), (5.14)
n=0

then

f (n) (z0 )
cn =
. ∀ n ∈ N0 := N ∪ {0}, (5.15)
n!
n
where .f (n) denotes the n-th order derivative . ddzfn .

Proof Theorem 5.6 implies that f in analytic in the disk .Br (z0 ) and Corollary 5.3
implies that f has derivatives of all orders in this disk. In addition, from the formula
(5.10) it follows that for any .k ∈ N the kth order derivative

+∞
 n!
f (k) (z) =
. cn (z − z0 )n−k for all z ∈ Br (z0 ). (5.16)
(n − k)!
n=k

Taking .z = z0 in (5.16), we find .f (k) (z0 ) = k! ck . That is, we get the formula (5.15)
for the power series coefficients.

Remark 5.4 Sometimes one can find such formulations of this theorem:

• a power series is the Taylor series of its sum;

• a function can be represented by a power series only in one way.

Example 5.3 Using (5.15), it is easy to check that the coefficients of the power
representation of .z2 + 1 around 1 are as follows: .c0 = c1 = 2, .c2 = 1, and .cn = 0
for .n > 2 (compare with (5.13)).

It is often inexpedient to calculate coefficients using (5.15) or impractical

formulas (5.6). Based on Theorem 5.7 and with the help of known formulas for the
coefficients of the Taylor series of elementary real functions, it is possible to write
down the power representations of the corresponding elementary complex-valued
entire functions:
+∞ n
 +∞
 +∞

z z2n−1 z2n
. ez = , sin z = (−1)n−1 , cos z = (−1)n .
n! (2n − 1) ! (2n) !
n=0 n=1 n=0
5.4 Uniqueness of Power Series Expansions: Morera’s Theorem 117

Example 5.4 From the representations above, the following identity follows:

+∞
 +∞

z2n z2n−1
. cos z + i sin z = (−1)n +i (−1)n−1
(2n) ! (2n − 1) !
n=0 n=1

z2 z3 z4 z5
=1+iz− −i + +i − ...
2! 3! 4! 5!
(iz)2 (iz)3 (iz)4 (iz)5
=1+iz+ + + + + . . . = eiz
2! 3! 4! 5!

for all .z ∈ C. This identity coincides with Euler’s formula (1.9) when z is a real
number. .

Comparing formulas (5.6) and (5.15) for finding power series coefficients, we
get

n! f (ξ )
f (n) (z0 ) =
. dξ, n ∈ N0 ; (5.17)
2π i ∂ + Br (z0 ) (ξ − z0 )n+1

here .r ∈ (0, R) and R is the convergence radius of (5.14).

What is interesting is that equalities (5.17) make it possible to estimate the
derivative of an arbitrary order of an analytic function through the value of this
function on the boundary of a disk.

Proposition 5.1 Let .f ∈ A(Br (z0 )). Then

n! |f (ξ )| n!
|f (n) (z0 )| ≤
. dl ≤ n max |f (ξ )| for all n ∈ N.
2π ∂Br (z0 ) r n+1 r ξ ∈∂Br (z0 )

Exercise 5.3 Let .f ∈ A(C) and there exist numbers .A > 0, .B > 0 and .n ∈ N
such that for all .r ∈ (0, +∞) :

. max |f (z)| ≤ A r n + B.
|z|=r

Show that .f is a polynomial whose degree is not greater than .n.

Exercise 5.4 Let .Pn (z) = an zn + an−1 zn−1 + . . . + a0 , where .an = 0, and

. max |Pn (z)| ≤ M.

|z|=1

Show that

|Pn (z)| ≤ M|z|n

. for all z, |z| ≥ 1.
118 5 Complex Power Series

Lemma 5.1 (The Cauchy Integral Formula for Derivatives) Let .f ∈ A(Ω)
and the other conditions of Theorem 4.8 be satisfied. Then for any .n ∈ N and any
.z0 ∈ Ω :

n! f (ξ )
f (n) (z0 ) =
. dξ. (5.18)
2π i ∂+Ω (ξ − z0 )n+1

Proof Since .f ∈ A(Ω), there exists a domain .G ⊃ Ω such that .f ∈ A(G). Fix
any point .z0 ∈ Ω and take any circle .βr = z0 + r eit , .t ∈ [0, 2π ], whose interior
belongs in .Ω. Obviously, the curve .βr is null-homotopic in G (see Definition 4.4).
Similarly, as in Corollary 4.5, we construct the closed curve .Λ with the positive
orientation (see Sect. 1.4), which is null-homotopic in G and whose trace does not
contain the point .z0 . Then .Λ ≈ βr in .G \ {z0 }.
Now, based on the Cauchy integral Theorem 4.5 and the equality (5.17), just as
in Corollary 4.5, we derive

n! f (ξ ) n! f (ξ )
f (n) (z0 ) =
. dξ = dξ
2π i βr (ξ − z0 )n+1 2π i Λ (ξ − z0 )n+1
n! f (ξ )
= dξ.
2π i ∂+Ω (ξ − z0 )n+1

Again, as in the case the Cauchy integral formula (see Example 4.1), it is
sometimes possible to calculate integrals from analytic functions along closed
curves with the help of the formula (5.18).

cos iz
Example 5.5 Compute . dz.
{|z|=2}+ (z − 1)
3

Since .f (z) = cos(iz) is analytic in .B2 (0) and .z0 = 1 ∈ B2 (0), we use (5.18)
with .n = 2 to get

cos iz
. dz = π i f  (1) = π i cos i = π i cosh 1.
{|z|=2}+ (z − 1)3

From mathematical analysis, it is known that any continuous function on a

segment can be uniformly approximated by a polynomial with a given accuracy. A
similar problem arises in complex analysis: can an analytic function on a compact
set be approximated by a polynomial with a given accuracy? It is clear that if this
compact set is a closed disk, then the problem is solved by Taylor polynomials.
However, power series converge in disks, so Taylor polynomials are not suitable
for approximating analytic functions in general domains. For simply connected
domains, this problem was solved by the German mathematician Karl Runge (1856–
1927) in 1885.
5.4 Uniqueness of Power Series Expansions: Morera’s Theorem 119

Theorem 5.8 (Runge’s Theorem) Let f be an analytic function in a simply

connected domain .Ω. Then for any compact set .K ⊂ Ω and any .ε > 0 there is
a polynomial .Pn (z) = an zn + an−1 zn−1 + . . . + a0 such that

. max |f (z) − Pn (z)| < ε.

z∈K

We do not present the proof of the theorem. The reader can find it in [4, Vol. 3].
Runge’s theorem has many applications in various branches of complex analysis; its
generalizations are Walsh’s theorem, Mergelyan’s theorem and Keldysh–Lavrent’ev
theorem (see [4]).
Now we can prove a converse statement of the Cauchy-Goursat Theorem 4.1.
Surprisingly, the property described there is almost equivalent to analyticity.

Theorem 5.9 (Morera’s Theorem) Let f be a continuous function in a domain

Ω. If the integral of f over the boundary of an arbitrary triangle, which together
.

with its closure belongs to .Ω, is zero, then the function .f is analytic in .Ω.

Proof Take any point .a ∈ Ω and any number .r > 0 such that .Br (a) ⊂ Ω. Based
on Remark 4.4, the function

F (z) :=
. f (ξ )dξ, z ∈ Br (a),
[a,z]

is an antiderivative of f in .Br (a). Corollary 5.4 implies that .f ∈ A(BR (a)). Since
the point a was chosen arbitrarily, .f ∈ A(Ω).

As we will see further on, Morera’s theorem is useful in proving many important
facts. Now consider an example.

Example 5.6 Show that the function

d
f (z) =
. e−t t z−1 dt z ∈ Ωr := {z ∈ C : Im z > 0},
c

where .0 < c < d < +∞, is analytic in the right half-plane .Ωr .
The first thing to note is that the function f is well defined and continuous in
.Ωr because the integrand is a continuous function of two variables t and z in the

domain of definition .(t z−1 = e(z−1) log t ).

Now take an arbitrary triangle . ⊆ Ωr and calculate the integral
 d
. f (z) dz = e−t t z−1 dt dz
∂+ ∂+ c
d  d
= e−t t z−1 dz dt = 0 dt = 0.
c ∂+ c
120 5 Complex Power Series

Here we used the analyticity of the function .t z−1 = e(z−1) log t with respect to the
variable z and the Cauchy-Goursat Theorem 4.1. Then, Morera’s theorem implies
that .f ∈ A(Ωr ).

Exercise 5.5 Using the same approach, show that

+∞ t z−1
g(z) =
. dt (5.19)
1 et − 1

is an entire function. 
.

It is well known from mathematical analysis that the term-by-term differentiation

of a functional series requires its convergence at some point and the uniform
convergence of the corresponding series of derivatives. In complex analysis, the
situation is simplified for functional series consisting of analytic functions. The
following theorem holds.

Theorem 5.10 Let .{fn }n∈N be a sequence of analytic functions in a domain .Ω. If
for any compact set .K ⊂ Ω the series

+∞

. fn (z)
n=1

converges uniformly on .K to a function .f, then .f ∈ A(Ω) and for all .k ∈ N

+∞

f (k) (z) =
. fn(k) (z) for all z ∈ Ω.
n=1

+∞ Fix a point a in .Ω and consider a closed disk .BR (a) ⊂ Ω. Since the series
Proof
.
n=1 fn (z) converges uniformly on .BR(a) to .
f and each term of this series is a
continuous function, the function .f ∈ C BR (a) .
Now take any triangle . ⊆ BR (a). Then, due to the uniform convergence and
the Cauchy-Goursat Theorem 4.1 we have

+∞
 +∞

. f (z) dz = fn (z) dz = fn (z) dz = 0.
∂+Δ ∂ + Δ n=1 +
n=1  ∂ Δ  

0

Based on Morera’s Theorem 5.9, .f ∈ A(BR (a)), and hence .f ∈ A(Ω), since a is
an arbitrary point from .Ω.
5.4 Uniqueness of Power Series Expansions: Morera’s Theorem 121

To prove the second statement, consider a circle .γr = a + reit , .t ∈ [0, 2π ], such
that .Eγr ⊂ BR (a). Since the series

+∞
k !  fn (z)
. (k ∈ N)
2π i (z − a)k+1
n=1

f (z)
converges uniformly on .Eγr to . (z−a) k+1 , it can be integrated term by term. Recalling
the formulas (5.17), we get

 k! +∞  +∞
k! f (z) fn (z)
f
.
(k)
(a) = dz = dz = fn(k) (a).
2π i (z − a)k+1 2π i (z − a)k+1
γr n=1 γr n=1

The theorem is proved.

It follows from this theorem that analyticity is preserved in uniform limits, in

contrast to differentiability in real analysis, where the uniform limit of differentiable
functions may be nowhere differentiable.

Corollary 5.5 Let .{fn }n∈N be a sequence of analytic functions in a domain .Ω. If
for any compact set .K ⊂ Ω the sequence .{fn }n∈N converges uniformly on K to a
function .f, then .f ∈ A(Ω) and for all .k ∈ N

f (k) (z) = lim fn(k) (z) for all z ∈ Ω.

.
n→+∞

Example 5.7 Using the Weierstrass criterion (see Theorem 5.1), it is easy to
verify that the function series . +∞
n=1 n
−z converges absolutely and uniformly in

.{z ∈ C : Re ≥ 1+δ}, where .δ is an arbitrary positive number. To apply this criterion,

you need the inequality

1
|n−z | = e−(x+iy) log n = e−x log n ≤
. ,
n1+δ

and the well-known fact that the numerical series . +∞n=1 n
−(1+δ) converges.
−z
Since for every .n ∈ N the function .n = e −z log n is entire, Theorem 5.10 says
that the sum
+∞
 1
ζ (z) :=
. (5.20)
nz
n=1

is an analytic function in the half-plane .{z ∈ C : Re z > 1}. This function is called
the Riemann zeta function. Note that .ζ has a singularity at .z = 1 and .limε→0 ζ (1 +
122 5 Complex Power Series

 −1 of the harmonic series is equivalent to

ε) = +∞ since the partial sum . Nn=1 n
.log N as .N → +∞.

Example 5.8 The real-valued gamma function

+∞
Γ (x) =
. e−t t x−1 dt, x ∈ (0, +∞), (5.21)
0

is studied in mathematical analysis, and it is well known that .Γ ∈ C ∞ (0, +∞).

We now define the gamma function in the right half-plane .Ωr by the formula
+∞
Γ (z) =
. e−t t z−1 dt, z ∈ Ωr = {ξ ∈ C : Im ξ > 0}. (5.22)
0

Since

|e−t t z−1 | = |e−t e(z−1) log t | = e−t t x−1 ,

. (5.23)

the inequality
+∞
.|Γ (z)| ≤ e−t t x−1 dt < +∞
0

holds for all .z ∈ Ωr , where .Re z = x > 0. Thus, the gamma function is correctly
defined in .Ωr . In addition,
n
Γ (z) = lim fn (z),
. where fn (z) = e−t t z−1 dt.
n→+∞ 1
n

Every function .fn is analytic in .Ωr (see Example 5.6). Using (5.23), it is easy
to show that the sequence .{fn }n∈N converges uniformly on any compact .K ⊂ Ωr
(even on every vertical strip .{z : 0 < α ≤ Re z ≤ β < +∞}). By Corollary 5.5, the
gamma function is analytic in the right half-plane .Ωr .
Replacing the integration variable .(nt to t), where .n ∈ N, we get
+∞ 1 +∞ Γ (z)
. e−nt t z−1 dt = e−t t z−1 dt = ,
0 nz 0 nz

from where, we deduce for any .z ∈ {ξ ∈ C : Re ξ > 1} that

+∞
 +∞ 
N +∞
1 t z−1
Γ (z)
. = t z−1 lim e−nt dt = dt.
nz 0 N→+∞ 0 et − 1
n=1 n=1
5.4 Uniqueness of Power Series Expansions: Morera’s Theorem 123

As a result, we obtain the relation

+∞ t z−1
Γ (z) ζ (z) =
. dt, z ∈ {ξ : Re ξ > 1}, (5.24)
0 et − 1

between the gamma and zeta functions. These functions have many applications in
physics, probability theory, and applied statistics; the Riemann zeta function plays a
central role in analytic number theory. We will continue our study of these functions
in Sect. 8.2. .

In this section, we have proved several equivalent statements about the analyticity
of a function. We summarise them in the following theorem.

Theorem 5.11 The following three statements are equivalent:

(R)
. a function .f is differentiable in a disk .Br (a);
(C)
. f ∈ C(Br (a)) and for any triangle . ⊆ Br (a) : .
. f (z) dz = 0;
∂+

+∞
(W )
. f (z) =
. cn (z − a)n for all z ∈ Br (a).
n=0

Remark 5.5 These three statements reflect three concepts in the development of
the theory of analytic functions:

(1) functions are called Riemann analytic or simply analytic if they satisfy
the .(R) condition;
(2) functions are called Cauchy analytic or holomorphic if they satisfy the .(C) con-
dition;
(3) functions are called Weierstrass analytic or regular if they satisfy the .(W ) con-
dition.

Thus there were three different starting points in the development of the theory of
complex-valued functions of a complex variable, and this explains why, even today,
different words are used, such as “analytic”, “regular” and “holomorphic”. Now we
see that these are equivalent concepts and we prefer the term “analytic”. .

The proof of Theorem 5.11 is based on already proved theorems. The equivalence
of the statements .(R) and .(C) is ensured by the Cauchy-Goursat Theorem 4.1 and
Morera’s Theorem 5.9, and the equivalence of statements .(R) and .(W ) is provided
by Theorem 5.3 (on the expansion of a analitic function into a power series) and
Theorem 5.6 (on the sum of a power series).
124 5 Complex Power Series

5.5 Uniqueness Theorem for Analytic Functions: Zeros of

Analytic Functions

Let us now consider some applications of power series representations. We begin

with the uniqueness theorem for analytic functions.

Theorem 5.12 Suppose that f and g are analytic functions in a domain .Ω and
there exists a sequence of distinct points .{zn }n∈N ⊂ Ω such that

• . lim zn = a ∈ Ω and .f (zn ) = g(zn ) for all .n ∈ N.

n→+∞

Then .f ≡ g in .Ω.

This theorem shows one more difference between differentiable complex-

valued functions and differentiable real-valued functions. It is easy to give
many examples where two infinitely differentiable functions of a real variable
may coincide on a certain segment of the domain of definition, but not be
identically equal in the entire domain of definition. According to this theorem,
if two analytic functions coincide on an arbitrary set that has a limit point
in the domain of analyticity of these two functions, then they necessarily
coincide identically in the entire domain.

Proof

1. Let us first show that these functions coincide in some neighborhood of the point
a. Since .a ∈ Ω, there is a positive number .R such that the disk .BR (a) belongs
to the domain .Ω. By Theorem 5.3, we have

+∞
 +∞

f (z) =
. ck (z − a)k and g(z) = dk (z − a)k
k=0 k=0

for all .z ∈ BR (a). Consider the difference

+∞

.h(z) := f (z) − g(z) = (ck − dk )(z − a)k , z ∈ BR (a).
k=0

Since .limn→+∞ zn = a, for any .r0 ∈ (0, R) there exists a number .n0 ∈ N
such that

.zn ∈ Br0 (a) ⊂ BR (a) for all n ≥ n0 .

5.5 Uniqueness Theorem for Analytic Functions: Zeros of Analytic Functions 125

By the Cauchy-Hadamard Theorem 5.2, the series

+∞

. (ck − dk )(z − a)k
k=0

converges uniformly on .Br0 (a). Therefore, passing to the limit .(n → +∞) in the
equality

+∞

0 = h(zn ) =
. (ck − dk )(zn − a)k (n ≥ n0 ), (5.25)
k=0

we find that .0 = h(a) = c0 − d0 , whence .c0 = d0 .

Now the equality (5.25) can be rewritten in the form

+∞

0 = h(zn ) =
. (ck − dk )(zn − a)k−1 . (5.26)
k=1

Passing to the limit in (5.26), we obtain .c1 = d1 . Continuing this process, we find
.ck = dk for all .k ∈ N ∪ {0}. This means that for all .z ∈ Br0 (a)

h(z) = 0 ⇐⇒ f (z) = g(z).

.

2. Consider now an arbitrary point .b ∈ Ω. Let .γ(t), .t ∈ [0, 1], be a broken line
whose trace .Eγ is contained in .Ω, starting at the point .a = γ(0) and ending at
the point .b = γ(1).
Denote by .δ := dist(Eγ , ∂Ω) (the distance from .Eγ to the boundary of .Ω). It
is obvious that .δ > 0. In what follows, we assume that the number .r0 chosen in
the first item of the proof satisfies the inequality .r0 < δ.
Since the set .Eγ is compact, there are finitely many disks .{Br0 (aj )}m j =0
covering .Eγ , and we can assume that the center of the next disk is contained
in the previous one and .a0 = a, am = b.
According to the just proved .h(z) ≡ 0 for all .z ∈ Br0 (a0 ). Since the point
(1)
.a1 ∈ Br0 (a0 ), there exists an infinite sequence .{zn }n∈N of distinct points in
(1)
.Br0 (a1 ) ∩ Br0 (a0 ) converging to .a1 and .h(zn ) = 0 for all .n ∈ N. Similarly, as

in the previous item, we show that .h ≡ 0 in .Br0 (a1 ).

Continuing this procedure, after a finite number of steps we get .h(z) = 0 for
all .z ∈ Br0 (b).

Based on this theorem, we can immediately conclude about the structure of the
set of zeros .Z := {z ∈ Ω : f (z) = 0} of an analytic function .f : Ω → C.
126 5 Complex Power Series

Corollary 5.6 Let .f ∈ A(Ω) and .f ≡ 0. Then the following two cases are
possible:

(1) the set .Z consists only of a finite number of isolated points;

(2) the set .Z is countable, and if .a is the limit point of .E, then .a ∈ ∂Ω.

To prove this corollary, we should apply Theorem 5.12 to the functions f and .g ≡ 0.
Let us consider an example showing that the second assertion of Corollary 5.6 is
incorrect in real analysis.

Example 5.9 (Counterexample from Mathematical Analysis) Consider the

function

x 2 sin x1 , x ∈ (−1, 1) \ {0};
.f (x) =
0, x = 0.

This function is differentiable on the interval .(−1, 1) and is not identical to zero.
However, the set of its zeros is countable and the limit point of the zeros, the point
0, belongs to .(−1, 1). .

Using the uniqueness theorem, one can prove functional identities in the complex
plane, which are valid for real numbers. For example, let us show that .sin2 z +
cos2 z = 1 for all .z ∈ C. Consider the entire functions

f (z) = sin2 z + cos2 z − 1

. and g ≡ 0.

Since .f (x) = 0 for all .x ∈ R, then by Theorem 5.12 .f (z) = 0 for all .z ∈ C.
The next theorem shows the behavior of an analytic function in a neighborhood
of its isolated zero.

Theorem 5.13 Let .f ∈ A(Ω), .f ≡ 0, and .z0 ∈ Ω. If the point .z0 is a zero of .f,
then in some neighborhood of it the function f can be uniquely represented as

f (z) = (z − z0 )m ϕ(z),
. (5.27)

where .m ∈ N and the function .ϕ is analytic and not equal to zero in this
neighborhood.

Proof Consider a disk .BR (z0 ) ⊂ Ω. By Theorem 5.3,

+∞

f (z) =
. ck (z − z0 )k for all z ∈ BR (z0 ). (5.28)
k=0
5.5 Uniqueness Theorem for Analytic Functions: Zeros of Analytic Functions 127

Since .f ≡ 0, not all coefficients in (5.28) are zero. Hence, there is a unique natural
number m such that

.c0 = c1 = . . . = cm−1 = 0 and cm = 0. (5.29)

Then
+∞
 +∞

f (z) =
. ck (z − z0 ) = (z − z0 )
k m
cn+m (z − z0 )n , z ∈ BR (z0 ).
k=m n=0

Define the function

+∞

ϕ(z) :=
. cn+m (z − z0 )n , z ∈ BR (z0 ).
n=0

It is clear that as the sum of the power series the function .ϕ ∈ A(BR (z0 )) and
.ϕ(z0 ) = cm = 0. This means that there exists such a number .r ∈ (0, R) that
.ϕ(z) = 0 for all .z ∈ Br (z0 ).

Thus, the function f has the form (5.27) in the disk .Br (z0 ).

Definition 5.5 The number .m in (5.27) is called the order (multiplicity) of the zero
z0 of the function f . If .m = 1, then .z0 is called a simple zero of f .
.

The representation (5.27) shows that an analytic function can be factorized

in a neighbourhood of its isolated zero like polynomials (see Theorem 5.14).
Factorization of entire functions in the complex plane will be discussed in
Sect. 7.6.

The following consequences follow from the proof of Theorem 5.13, namely
from (5.29) and formulas (5.15).

Corollary 5.7 Let .z0 ∈ Ω be a zero of multiplicity .m for an analytic function f in

Ω. Then
.

f (k) (z0 ) = 0 for all k ∈ {0, 1, . . . , m − 1}

. and f (m) (z0 ) = 0.

Corollary 5.8 An analytic function that is not identical to zero cannot have zeros
of infinite order (a point .z0 is a zero of infinite order of an analytic function .f, if
.f
(k) (z ) = 0 for all .k ∈ N ).
0 0
128 5 Complex Power Series

Example 5.10 (Counterexample) Consider the following function:

 1
−
f (x) =
.
e x2 , x = 0;
0, x = 0.

It is easy to verify that .f ∈ C ∞ (R), .f ≡ 0, but for all .n ∈ N0 : .f (n) (0) = 0. This
counterexample shows that an infinitely differentiable non-zero real-valued function
can have zeros of infinite order.

Exercise 5.6 Let .f, g : Ω → C be two analytical functions. Prove that if there is a
point .z0 in the domain .Ω such that

d nf d ng
. (z 0 ) = (z0 ) for all n ∈ N0 ,
dzn dzn

then .f ≡ g in .Ω.

Corollary 5.7 is useful for finding multiplicity of zeros of analytic functions.

Example 5.11 Find the order of the zero .z0 = 1 of the function

f (z) = sin (z − 1) − z + 1.
.

Direct calculations give that .f (1) = 0, .f  (1) = 0, .f  (1) = 0, .f  (1) = −1 = 0.

So the multiplicity of this zero is 3.
In this case, it is easy to factorize this function

+∞
 (z − 1)2n−1
.f (z) = (−1)n−1 − (z − 1) = (z − 1)3 ϕ(z),
(2n − 1) !
n=1

+∞
 (z − 1)2n−4 1
where .ϕ(z) = (−1)n−1 , ϕ(1) = − = 0. Thus, .z0 = 1 is indeed
(2n − 1) ! 6
n=2
the third order zero of the function f .

Now we will prove a stronger assertion than in the fundamental theorem of

algebra (Theorem 5.5).

Theorem 5.14 The polynomial

.Pn (z) = zn + cn−1 zn−1 + . . . + c1 z + c0 ,

where .{cj }n−1

j =0 ⊂ C, n ∈ N, has, counted with multiplicity, exactly n zeros.
5.5 Uniqueness Theorem for Analytic Functions: Zeros of Analytic Functions 129

In addition, the following factorization of .Pn is valid:

Pn (z) = (z − a1 )m1 · . . . · (z − ak )mk

. for all z ∈ C, (5.30)

where .{a1 , . . . , ak } are distinct zeros of the polynomial .Pn of multiplicity .m1 , . . . ,
mk , respectively, and .m1 + . . . + mk = n.
.

Proof Theorem 5.5 implies that .Pn has as least one zero; denote it by .a1 . Then,
according to Theorem 5.13 and the Euclidean division of polynomials, there exists
a unique number .m1 ∈ N and a polynomial .Pn−m1 of degree .n − m1 such that

Pn (z) = (z − a1 )m1 Pn−m1 (z),

. and Pn−m1 (a1 ) = 0.

If .n − m1 ≥ 1, then again thanks to Theorem 5.5 there is a zero .a2 .(a2 = a1 ) of

the polynomial .Pn−m1 . Repeating the previous considerations , we get

.Pn (z) = (z − a1 )m1 (z − a2 )m2 Pn−m1 −m2 (z),

where .Pn−m1 −m2 (ap ) = 0 for .p ∈ {1, 2}. Continuing these considerations, we
obtain (5.30).

Definition 5.6 Let f be an analytic function at .∞ (see Definition 2.7). The point at
 .m of the function .f, if .z0 = 0 is a zero of order m
infinity is called a zero of order
of the function .g(z) = f 1z .

Recall that by definition .f (∞) = limz→∞ f (z).

Theorem 5.15 Let f be an analytic function at .∞ and let f have a zero  of order

m at .∞. Then there exists a number .r1 > 0 and a unique function .ψ ∈ A Br1 (∞) ,
.ψ(z) = 0 for all .z ∈ Br1 (∞), such that

ψ(z)
f (z) =
. for all z, |z| > r1 .
zm

Proof By Definition
 5.6, the point .z0 = 0 is a zero of multiplicity .m of the function
.g(z) = f
1
z . Then, due to Theorem 5.13, there exists a disk .Br (0) and a unique
function .ϕ ∈ A(Br (0)), .ϕ(z) = 0 for all .z ∈ Br (0), such that

.g(z) = zm ϕ(z) for all z ∈ Br (0).

130 5 Complex Power Series

Returning to the function .f, we find

1 1 1
f (z) = g
.
z = ϕ for all z, |z| > r1 , where r1 = 1r .
zm z

It remains now to denote .ψ(z) := ϕ 1z .
Laurent Series: Isolated Singularities of
Analytic Functions 6

Abstract

In this chapter, we continue the study of power series, but already their
generalizations, namely power series containing terms .(z − z0 )n with a negative
integer n. These series were introduced by the French mathematician Pierre
Laurent (1813–1854) in 1843. Laurent series are a valuable tool for studying the
behavior of analytic functions near their isolated singularities, a classification
of which is given here. It is noteworthy that, knowing the behavior of an
analytic function near its singular points, one can determine its behavior in
the entire domain, as well as calculate other characteristics associated with that
function. As a result, it became possible to classify analytic functions according
to their isolated singularities (Sect. 6.5). Interestingly, Laurent series have an
equivalent relationship to Fourier series (Sect. 6.2), which have real applications
in engineering (signal processing, spectroscopy, computer tomography, and
many others).

6.1 Expansion of an Analytic Function Into a Laurent Series

It is often possible to find a series representation for a function involving both

positive and negative powers of z. Consider, for example, the function

1
f (z) =
. , z ∈ Ω := {z : 1 < |z| < 2}.
(1 − z)(z + 2)

It can be presented as
 
1 1 1
.f (z) = + , z ∈ Ω := {z : 1 < |z| < 2}.
3 1−z z+2

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 131
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_6
132 6 Laurent Series: Isolated Singularities of Analytic Functions

If .|z| > 1, then the formula for the sum of an infinite geometric progression gives
us
+∞
 +∞
 −∞

1 1 1 1 −n
. =− · =− =− z =− zn .
1−z z 1− 1
z
zn+1
n=0 n=1 n=−1

If .|z| < 2, then in the same way we derive

+∞

1 1 1 (−1)n
. = = zn .
z+2 2 1+ z
2 2n+1
n=0

As a result,
+∞ −∞
 +∞
1  (−1)n n  
.f (z) = − =: for all z ∈ Ω,
n
z z cn zn (6.1)
3 2n+1 n=−∞
n=0 n=−1

where
⎧
⎪
⎪ (−1)n
⎨ , n ∈ Z, n ≥ 0;
.cn =
3 · 2n+1
⎪
⎪ 1
⎩− , n ∈ Z, n < 0.
3

Thus, the function f , which is analytic in .Ω, has been expanded into a series in
integer powers of z. Such series are often called generalized power series or Laurent
series. Let us give a rigorous definition of such a series.

Definition 6.1 Let .z0 ∈ C and .{cn }n∈Z ⊂ C. Laurent’s series around the point .z0 is
called the following expression:

+∞

(L) :
. cn (z − z0 )n .
n=−∞

The series .(L) is called convergent at a point .z, if the functional series

+∞

(R) :
. cn (z − z0 )n
n=0

and
−∞
 +∞
 1
.(P) : cn (z − z0 )n = c−n
(z − z0 )n
n=−1 n=1
6.1 Expansion of an Analytic Function Into a Laurent Series 133

converge at .z, and the sum of the Laurent series .(L) is defined as follows:

+∞
 +∞
 +∞
 1
. cn (z − z0 )n := cn (z − z0 )n + c−n .
n=−∞
(z − z0 )n
n=0 n=1

The series .(R) is named the regular part, and the series .(P) is named the principal
part of the Laurent series .(L). .

Let us determine the sets of convergence of the series .(R) and .(P). The series .(R)
is a power series, so it is convergent in the disk .{z : |z −z0 | < R}, where the number
R is determined with the formula (5.4).
If we make the substitution .η = z−z 1
0
, then the series .(P) is reduced to
+∞ n
the power series . n=1 c−n η , and therefore √ the series .(P) converges in the set
.{z : |z − z0 | > r} , where .r = lim supn→+∞ |c−n |. Thus,
n

• if .r > R, then by Definition 6.1 the series .(L) is divergent;

• if .r = R, then the series .(L) can be both convergent and divergent at points on
the circle .{z : |z − z0 | = r};
• if the inequality .r < R holds, then the series .(L) converges in the annulus .{z : r <
|z − z0 | < R}.

Definition 6.2 The annulus .A := {z : 0 ≤ r < |z − z0 | < R}, the inner radius of
which is determined by the formula

r = lim sup
.
n
|c−n |, (6.2)
n→+∞

and the outer radius by the formula

 −1
.R = |cn |
n
lim sup (6.3)
n→+∞

is called the annulus of convergence of the Laurent series .(L). .

The following statement follows from the Cauchy–Hadamard Theorem 5.2.

Proposition 6.1 For the Laurent series .(L) the following assertions hold:

(1) the series (L) absolutely converges at each point of the annulus .A;
(2) the series (L) diverges at each point of the set .C \ A;
(3) the series (L) can be both convergent and divergent at points on .∂A;
(4) for any compact set .K ⊂ A the series (L) converges uniformly on K.
134 6 Laurent Series: Isolated Singularities of Analytic Functions

Again using Cauchy’s integral formula, as in Theorem 5.3, we prove that an

analytic function in an annulus can be represented as a Laurent series.

Theorem 6.1 Let .0 ≤ ρ1 < ρ2 ≤ +∞ and .Aρ1 ,ρ2 := {z : ρ1 < |z − z0 | < ρ2 },

where .z0 ∈ C. If .f is an analytic function in the annulus .Aρ1 ,ρ2 , then

+∞

.f (z) = cn (z − z0 )n for all z ∈ Aρ1 ,ρ2 , (6.4)
n=−∞

where
1 f (z)
. cn = dz, n ∈ Z, ∀ ρ ∈ (ρ1 , ρ2 ). (6.5)
2π i {|z−z0 |=ρ}+ (z − z0 )n+1

Proof Fix any .z ∈ Aρ1 ,ρ2 . Obviously, there are numbers .r1 > 0 and .R1 > 0 such
that .ρ1 < r1 < |z − z0 | < R1 < ρ2 . Then .Ar1 ,R1 ⊂ Aρ1 ,ρ2 and the function
.f ∈ A Ar1 ,R1 . By Cauchy’s integral formula (Theorem 4.8)

1 f (ξ )
f (z) =
. dξ
2π i ∂ + Ar1 ,R1 ξ −z
1 f (ξ ) 1 f (ξ )
= dξ − dξ. (6.6)
2π i ∂ + BR1 (z0 ) ξ −z 2π i ∂ + Br1 (z0 ) ξ −z

The appearance of a minus before the last integral is explained in the same way as
in the equality (4.23).

1. For any .ξ ∈ ∂BR1 (z0 )

|z − z0 | |z − z0 |
. = =: q < 1.
|ξ − z0 | R1

Therefore,

 (z − z0 )n+∞
1 1 1
. = · = ,
ξ −z ξ − z0 1 − ξz−z
−z0
0 (ξ − z0 )n+1
n=0

moreover, this series converges uniformly with respect to .ξ ∈ ∂BR1 (z0 ). This
means that the series

 1 +∞
1 f (ξ ) f (ξ )
. = · (z − z0 )n ,
2π i (ξ − z) 2π i (ξ − z0 )n+1
n=0
6.1 Expansion of an Analytic Function Into a Laurent Series 135

which obtained from the previous one by the multiplication with the bounded
function . 2π1 i f (ξ ), .ξ ∈ ∂BR1 (z0 ), converges uniformly on the same circle as
well. Hence, this series can be integrated term by term and we obtain

+∞

1 f (ξ )
. dξ = cn (z − z0 )n , (6.7)
2π i ∂ + BR1 (z0 ) ξ −z
n=0

where
1 f (ξ )
cn =
. dξ, n ∈ Z, n ≥ 0. (6.8)
2π i ∂ + BR1 (z0 ) (ξ − z0 )n+1

2. For .ξ ∈ ∂Br1 (z0 ) we have

|ξ − z0 | r1
. = =: q1 < 1.
|z − z0 | |z − z0 |

And therefore,
+∞

1 1 1 (ξ − z0 )n
. =− · = − ,
ξ −z z − z0 1 − ξ −z0 (z − z0 )n+1
z−z0 n=0

and this series converges uniformly on .∂Br1 (z0 ), as well as the series

 1 +∞
1 f (ξ )
. =− f (ξ ) · (ξ − z0 )n−1 · (z − z0 )−n ;
2π i (ξ − z) 2π i
n=1

here we shifted the summation index. Integrating this equality term by term and
changing the summation index to the opposite, we get

−∞

1 f (ξ
. − dξ = cn (z − z0 )n , (6.9)
2π i ∂ + Br1 (z0 ) ξ −z
n=−1

where
1 f (ξ )
cn =
. dξ, n ∈ Z, n < 0. (6.10)
2π i ∂ + Br1 (z0 ) (ξ − z0 )n+1

From (6.6), (6.7) and (6.9) follows the representation (6.4). Based on Theo-
rem 4.5, the equalities (6.8) and (6.10) imply (6.5).
136 6 Laurent Series: Isolated Singularities of Analytic Functions

Just as the Cauchy inequalities were proved for the coefficients of power series,
we prove the Cauchy inequalities for the coefficients of Laurent series.

Corollary 6.1 Let .f ∈ A(Aρ1 ,ρ2 ) and

∃ ρ0 ∈ (ρ1 , ρ2 ),
. ∃ M > 0 such that max |f (z)| ≤ M.
z∈{ξ : |ξ −z0 |=ρ0 }

M
Then, .|cn | ≤ for all .n ∈ Z.
ρ0n

In general, the integral formulas (6.5) are not practical for calculating Laurent
coefficients. Instead, various algebraic techniques are used, such as those described
at the beginning of this section. However, to justify the fact that the result is a
Laurent series, we need a theorem on the uniqueness of the expansion, which is
proved below.

Theorem 6.2 Let

+∞

f (z) =
. cn (z − z0 )n for all z ∈ A = {z : 0 ≤ r < |z − z0 | < R}.
n=−∞

Then f is an analytic function in the annulus A and the coefficients are

determined with the formulas

1 f (ξ )
cn =
. dξ, n ∈ Z, (6.11)
2π i {|ξ −a|=ρ}+ (ξ − z0 )n+1

where .ρ is an arbitrary number from the interval .(r, R).

Proof The analyticity of f in the annulus A follows from the definition of the sum
of a Laurent series (see Definition 6.1) and the theorem on the analyticity of the sum
of a power series (see Theorem 5.6).
Take any number .ρ ∈ (r, R). Then, according to the fourth item of Proposi-
tion 6.1, the series
+∞

f (z) =
. cn (z − z0 )n
n=−∞

converges uniformly on the circle .{z : |z − z0 | = ρ}.

Multiplying this series by . (z−z1)m+1 , where m is an arbitrary integer, we obtain
0
the series
+∞

f (z)
. = cn (z − z0 )n−m−1 ,
(z − a)m+1 n=−∞
6.2 Relationship Between Laurent Series and Fourier Series 137

that converges uniformly on the same circle. Integrating over the positively oriented
circle and taking into account the results of Example 4.3, we get

+∞

f (z)
. dz = cn (z − z0 )n−m−1 dz = cm 2π i
(z − z0 )m+1 n=−∞
{|z−z0 |=ρ}+ {|z−z0 |=ρ}+

from which the formulas (6.11) follow.

Based on this theorem we can state that the expansion (6.1) is indeed a Laurent
1
series for the function . (1−z)(z+2) in the annulus .{z : 1 < |z| < 2}.

6.2 Relationship Between Laurent Series and Fourier Series

Let .ϕ ∈ C 1 (R) and .ϕ be a .2π -periodic function. It is known from calculus that such
a function can be expanded into a Fourier series

+∞
a0   
.ϕ(t) = + an cos (nt) + bn sin (nt) , t ∈ R, (6.12)
2
n=1

moreover, this series converges uniformly on .R, and its coefficients are determined
by the formulas

1 π 1 π
an =
. ϕ(t) cos (nt) dt, bn = ϕ(t) sin (nt) dt.
π −π π −π

By using the identities

eint + e−int eint − e−int

. cos (nt) = and sin (nt) = , (6.13)
2 2i
we re-write the series (6.12) as follows

+∞   +∞
a0  an − ibn int an + ibn −int 
ϕ(t) =
. + e + e = cn eint , (6.14)
2 2 2 n=−∞
n=1

where
⎧ π
⎪
⎪
⎪ an − ibn = 1
⎪ ϕ(t)e−int dt, n ∈ Z, n ≥ 0;
⎪
⎪
⎨ 2 2π
−π
cn =
. π
⎪
⎪ a−n + ib−n
⎪
⎪ 1
⎪
⎪ = ϕ(t)e−int dt, n ∈ Z, n < 0.
⎩ 2 2π
−π
138 6 Laurent Series: Isolated Singularities of Analytic Functions

The series (6.14) is the Fourier series (6.12) written in complex form.
Let us introduce a new variable .z = eit , .t ∈ [−π, π ], in (6.14). Then .t = −i ln z.
Denoting .f (z) := ϕ(−i ln z), we get

+∞

f (z) =
. cn zn for all z, |z| = 1,
n=−∞

where
π
1 1 f (z)
.cn = f (eit )e−int dt = dz.
2π 2π i zn+1
−π {|z|=1}+

This definition of the coefficients .{cn } is consistent with the formulas (6.5) for
Laurent coefficients. So, we can make the following conclusion.

The Fourier series of a function .ϕ written in complex form is the Laurent

series of the function .f (z) := ϕ(−i ln z) on the unit circle centered at
the point .z = 0. Conversely, the Laurent series of any analytic function f
restricted to the unit circle (if the convergence annulus contains this circle) is
the Fourier series of the function .ϕ(t) := f (eit ), .t ∈ [−π, π ].

Sometimes a Fourier series expansion is easier to obtain through a Laurent series.

Consider the following example.

Example 6.1 Expand the function

a sin t
.ϕ(t) = , t ∈ R, (a ∈ (−1, 1))
1 − 2a cos t + a 2

into a Fourier series.

Solution Obviously, .ϕ is .2π -periodic. Using (6.13) for .n = 1, we obtain

1 a(eit − e−it )
ϕ(t) =
. .
2i 1 − a(eit + e−it ) + a 2

Introducing a new variable .z = eit , .t ∈ [−π, π ], we deduce

 
a z − 1  
1 z 1 1 1
.f (z) := ϕ(−i ln z) =   = − .
2i 1 − a z + 1 + a 2 2i 1 − az 1 − a/z
z
6.3 Isolated Singularities of Analytic Functions 139

Using the formula for the sum of an infinite geometric progression, we get

+∞ +∞
 an
1 1 1
. = a n zn for |z| < , and = for |z| > |a|.
1 − az |a| 1 − a/z zn
n=0 n=0

Hence, the Laurent series for f looks as follows

+∞
 1 n 1
f (z) =
. an (z − z−n ) for all z, |a| < |z| < .
2i |a|
n=0

Its restriction to the unit circle gives the Fourier series

+∞
  +∞
eint − e−int
ϕ(t) = f (eit ) =
. an = a n sin (nt), t ∈ [−π, π ].
2i
n=0 n=0

6.3 Isolated Singularities of Analytic Functions

It is well known that discontinuities of a real-valued function of a single real variable

can be classified as removable, jump, infinite, or mixed. In this section, we give a
classification of isolated singular points of analytic functions, show how one can
characterize their type using Laurent series, and study the behavior of analytic
functions near their singularities.

Definition 6.3 A point .z0 ∈ C is a called an isolated singular point (or isolated
singularity) of an analytic function .f, if there exists a number .R > 0 such that .f is
analytic in the punctured disk

{z : 0 < |z − z0 | < R}, if z0 = ∞;
B̆R (z0 ) :=
.
{z : |z| > R}, if z0 = ∞;

Depending on the behavior of the function .f in a punctured neighborhood of the

point .z0 , three types of isolated singularities are distinguished.

Definition 6.4 (Classification of Singularities) An isolated singularity .z0 of an

analytic function .f is

• a removable singularity if there exists a finite .limz→z0 f (z);

• a pole if .limz→z0 f (z) = ∞;
• an essential singularity if the limit of f as .z → z0 does not exist.
140 6 Laurent Series: Isolated Singularities of Analytic Functions

Example 6.2 Consider examples of different types of isolated singular points:

sin z
• the point .z0 = 0 is removable for .f (z) = since . lim f (z) = 1;
z z→0
z
• the point .z0 = −1 is a pole for .f (z) = since . lim f (z) = ∞;
1+z z→−1
• the point .z0 = ∞ is essential for .ez . Indeed, since

. lim ex = +∞ and lim ex = 0,

x→+∞ x→−∞

the limit .limz→∞ ez does not exist.

Important
Note that an analytic function can also have non-isolated singular points. For
example, the function
 π −1
f (z) = sin
. (6.15)
z
 
has poles at the points . an = n1 n∈Z\{0} , whose limit point is 0. Thus, the
point .0 is a non-isolated singular point of f .

Next, we prove theorems showing the connection between the type of an isolated
singular point of f and the form of the Laurent series for f around that point.

Theorem 6.3 Let .f ∈ A(B̆R (z0 )), .z0 ∈ C.

The point .z0 is removable for the function f if and only if the Laurent series of
f around .z0 has no principal part, i.e.,

+∞

f (z) =
. cn (z − z0 )n for all z ∈ B̆R (z0 ). (6.16)
n=0

Proof

Necessity Theorem 6.1 implies that f expands into a Laurent series in .B̆R (z0 ),
whose coefficients are determined by the formulas (6.5).
If .z0 ∈ C is removable for .f, then there exists finite .limz→z0 f (z), and this means
that .f is bounded in a punctured neighborhood of the point .z0 , i.e.,

∃ R0 ∈ (0, R)
. ∃M > 0 ∀ z ∈ B̆R0 (z0 ) : |f (z)| ≤ M.
6.3 Isolated Singularities of Analytic Functions 141

In virtue of Corollary 6.1, the coefficients of this Laurent series satisfy the
inequalities

M
|cn | ≤
.
ρn

for all .ρ ∈ (0, R0 ) and .n ∈ Z. But if .n is negative, then

M
|cn | ≤
. →0 as ρ → 0.
ρn

Hence, .cn = 0 for .n ∈ Z, n < 0.

Sufficiency If the Laurent series of f around .z0 has no principal part (see (6.16)),
then it is a power series and its sum is an analytic function in .BR (z0 ). This means
that . lim f (z) = c0 .
z→z0

The next corollary follows directly from the proof of Theorem 6.3.

Corollary 6.2 A point .z0 ∈ C is removable for an analytic function f if and only if
f is bounded in a punctured neighborhood of .z0 .
.

In addition, the function f can be extended by continuity at its removable

singular point .z0 , namely we set .f (z0 ) := limz→z0 f (z), and as a result, f will
be analytic in the whole disk .Br (z0 ).

Example 6.3 From the first item of Example 6.2 and Corollary 6.2 it follows that
the function
⎧
⎨ sin z
, if z = 0;
.f (z) = z (6.17)
⎩ 1, if z = 0,

is analytic in .C. In addition, its Laurent series is

+∞
 (−1)n−1 2n−2
f (z) =
. z for all z ∈ C.
(2n − 1)!
n=1

Theorem 6.4 (About a Pole) Let .f ∈ A(B̆R (z0 )) and .z0 ∈ C.

The point .z0 is a pole for the function f if and only if the principal part of the
Laurent series of f around .z0 contains a finite number of nonzero terms, i.e., there
exists a unique positive integer N such that

+∞

f (z) =
. cn (z − z0 )n for all z ∈ B̆R (z0 ), and c−N = 0. (6.18)
n=−N
142 6 Laurent Series: Isolated Singularities of Analytic Functions

Proof

Necessity We get from Theorem 6.1 that f is expanded into a Laurent series in
.B̆R (z0 ), whose coefficients are determined by the formulas (6.5).
If .z0 is a pole, then .limz→z0 f (z) = ∞, and this means that there exists a number
.r0 ∈ (0, R) such that .f (z) = 0 for all .z ∈ B̆r0 (z0 ). Hence,

1
. =: ϕ ∈ A(B̆r0 (z0 )) and lim ϕ(z) = 0.
f z→z0

Thus, .z0 is a removable point for the function .ϕ, which can be extended by
continuity at the point .z0 , and as a result, we get the analytic function in the whole
disk .Br0 (z0 ); in addition, .ϕ(z0 ) = 0 and .ϕ(z) = 0 for .z ∈ B̆r0 (z0 ).
From Theorem 5.13 (about a zero of an analytic function) it follows that there
exists a unique number .N ∈ N and a unique function .ψ ∈ A(Br0 (z0 )) such that

.ϕ(z) = (z − z0 )N ψ(z) and ψ(z) = 0 for all z ∈ Br0 (z0 ). (6.19)

Now consider the function . ψ1 that is also analytic in .Br0 (z0 ), and therefore, by
Theorem 5.3, it can be expanded into the power series

+∞

1
. = bn (z − a)n for all z ∈ Br0 (z0 ), and b0 = 0. (6.20)
ψ(z)
n=0

From (6.19) and (6.20) we have that

+∞

1 1
f (z) =
. = cn (z − z0 )n , z ∈ B̆r0 (z0 ), (6.21)
(z − z0 )N ψ(z)
n=−N

where .cn = bn+N , c−N = b0 = 0. Due to the uniqueness of the expansion into a
Laurent series (Theorem 6.2), the representation (6.21) holds in .B̆R (z0 ).

Sufficiency Let the Laurent series of f around .z0 be of the form (6.18). Then

ϕ(z) := (z − z0 )N f (z) = c−N + c−N +1 (z − z0 ) + . . . ,

. z ∈ B̆R (z0 ).

By Theorem 6.3, this means that .z0 is removable for the function .ϕ and
limz→z0 ϕ(z) = c−N = 0. Thus,
.

ϕ(z)
. lim f (z) = lim = ∞,
z→z0 z→z0 (z − z0 )N

that is, the point .z0 is a pole of the function f .

6.3 Isolated Singularities of Analytic Functions 143

From the proof of Theorem 6.4 it follows the corollary.

Corollary 6.3 The point .z0 ∈ C is a pole of an analytic function .f in .B̆R (z0 )
if and only if there exists a unique positive integer N and a unique function .ϕ ∈
A(BR (z0 )), .ϕ = 0 in .BR (z0 ), such that

ϕ(z)
f (z) =
. for all z ∈ B̆R (z0 ). (6.22)
(z − z0 )N

Definition 6.5 The number N in (6.22) is named the order (multiplicity) of the pole
z0 .
.

Comparing Corollary 6.3 and Theorem 5.13, we get the statement.

Corollary 6.4 A point .z0 ∈ C is a pole of order N of an analytic function .f if and

only if .z0 is a zero of order N for the function . f1 .

Remark 6.1 Simple zero and simple pole are terms used for zeroes and poles of
order .N = 1.
sin z
Example 6.4 The function . has a pole of order 3 at the point .z0 = 0. Indeed,
z4
it can be represented as

sin z f (z)
.
4
= 3 for all z ∈ B̆ π2 (0),
z z

where the analytic function f is determined in (6.17) and .f = 0 in .B π2 (0). In

addition, its Laurent series is of the form
+∞

sin z 1 1 1 (−1)n−1 2n−5
. = − + z for all z ∈ C \ {0}.
z4 3
z  6 z (2n − 1)!
n=3
the principal part

Theorem 6.3 and Theorem 6.4 provide the following theorem.

Theorem 6.5 (About an Essential Singularity) Let .f ∈ A(B̆R (z0 )) and .z0 ∈ C.
The point .z0 is essential for the function f if and only if the principal part of the
Laurent series of f around .z0 contains infinitely many nonzero terms.
144 6 Laurent Series: Isolated Singularities of Analytic Functions

1 1
Example 6.5 Clearly, the point .z0 = 0 is essential for .e z .(limx→0− e x = 0 and
1
.limx→0+ e x = +∞). Its Laurent series around 0 is of the form

+∞

1 1 1
e =
. z +1 for all z ∈ C \ {0}. (6.23)
n! zn
n=1
  
the principal part

Isolated Singularity at Infinity

Theorems 6.3, 6.4, 6.5 need tobeclarified when .z0 = ∞. Let .f ∈ A(B̆R (∞)).
Then the function .ϕ(ω) := f ω1 is analytic in .B̆ 1 (0) and expands there into a
R
Laurent series
+∞
 −1
 +∞

ϕ(ω) =
. bn ω =n
bn ω n
+ bn ωn , ω ∈ B̆ 1 (0).
R
n=−∞ n=−∞ n=0
     
the principal part the regular part

Returning to the variable .z = 1

ω, we get the Laurent series for .f around .∞ :

  −1
 +∞

1 1 1
f (z) = ϕ
. = n
+
bn bn n
z n=−∞
z z
n=0
     
the principal part the regular part

+∞
 
0 +∞

= b−n zn + b−n zn = cn zn , (6.24)
n=1 n=−∞ n=−∞
     
the principal part the regular part

where .cn = b−n . But now, the series . +∞ n

n=1 cn z with positive powers of z is
principal part of the Laurent series of f in .B̆R (∞), and . 0n=−∞ cn zn is the regular
one.

Summing up, we can state that the principal part of a Laurent series contains
terms that become unbounded when approaching the isolated singularity (it
can be both finite and the point at infinity).
6.4 Behavior of an Analytic Function Near Its Essential Singularity 145

Therefore, the Theorems 6.3, 6.4 and 6.5 are reformulated with precision up to
the definition of the principal part of a Laurent series around .∞. For example,
the point .∞ is a pole of an analytic function .f if and only if the principal part of its
Laurent series around .∞ contains a finite number of nonzero terms.
1
Example 6.6 Let us clarify the type of isolated singular points of the function .e z .
It is easy to see that these are the points 0 and .∞. In Example 6.5 it was shown
1
that 0 is an essential singularity for .e z . The series (6.23) is also the Laurent series
1
of .e z around .∞. But now the principal part of the Laurent series (6.23) around
1
.∞ is absent, so .∞ is removable for the function .e z . It is also easy to check that
1
.limz→∞ e z = 1.

6.4 Behavior of an Analytic Function Near Its Essential

Singularity

The behavior of .f (z) is clear as z approaches a removable singularity or a pole of

f . But the behavior of f near its essential singularity is “horrible” and needs further
study.

Theorem 6.6 (Casorati–Sokhotskyi–Weierstrass Theorem) Let .f ∈ A(B̆R (z0 ))

and .z0 ∈ C. If .z0 is essential for .f, then for every .α ∈ C there exists a sequence
.{zn }n∈N such that

. lim zn = z0 and lim f (zn ) = α,

n→+∞ n→+∞

i.e., the function f approaches any complex number, including .∞, in any neighbor-
hood of .z0 .

Proof Let first .α = ∞. Since the function f cannot be bounded in an arbitrary

punctured disk centered at .z0 (see Corollary 6.2), then there exists a point

z1 ∈ B̆R (z0 )
. such that |f (z1 )| > 1.

We use the same arguments to justify the existence of a point

. z2 ∈ B̆ R (z0 ) at which |f (z2 )| > 2.

2

Continuing in the same way, we find .zn ∈ B̆ R (z0 ) at which .|f (zn )| > n, .n ∈ N.
n
Thus,

. lim zn = z0 and lim f (zn ) = ∞. (6.25)

n→+∞ n→+∞
146 6 Laurent Series: Isolated Singularities of Analytic Functions

Now let .α = ∞. Then two cases are possible, namely

1. for any .n ∈ N there is a point .zn ∈ B̆ 1 (z0 ) such that .f (zn ) = α, but this is what
n
needs to be proven;
2. otherwise, there is a number .r ∈ (0, R) such that .f (z) = α for all .z ∈ B̆r (z0 ).
Thus, we can determine the analytic function

1
ϕ(z) :=
. , z ∈ B̆r (z0 ),
f (z) − α

for which the point .z0 is also an essential singularity. Then, according to the first
item of the proof, there exists a sequence .{zn }n∈N such that the limits (6.25) hold
(the second one for the function .ϕ). But then also

1
. lim f (zn ) = α + lim = α.
n→+∞ n→+∞ ϕ(zn )

The theorem is proved.

Exercise 6.1 Prove the Casorati–Sokhotskyi–Weierstrass theorem in the case when

z0 = ∞.
.

Exercise 6.2 Prove the Casorati–Sokhotskyi–Weierstrass theorem in the case when

z0 is a limit point of the poles, i.e., .z0 is a non-isolated singular point (see, e.g., the
.

function (6.15)).

Let us present, without proof, a theorem that additionally characterizes the

sophisticated behavior of an analytic function in a punctured neighborhood of an
essential singular point (in fact, the first case from the proof of Theorem 6.6 takes
place).

Theorem 6.7 (Great Picard’s Theorem) Let .z0 ∈ C be an essential singular point
of an analytic function f . Then, in an arbitrary punctured neighborhood of .z0 , the
function f takes an infinitely times arbitrary complex number, except perhaps one.

We demonstrate the Great Picard Theorem using the following example.

Example 6.7 The point .∞ is essential for the function .ez (see Example 6.2). It is
easy to verify that for every complex number .A = 0

ez = A
. ⇐⇒ zk = ln |A| + i(arg A + 2π k), k ∈ Z.

Thus, .limk→+∞ zk = ∞ and .ezk = A for all .k ∈ Z.

6.5 Classification of Analytic Functions 147

6.5 Classification of Analytic Functions with Respect to Their

Isolated Singularities: Theorem on a Meromorphic
Function

Obviously, the point at infinity is an isolated singular point for any entire function.
It turns out that by knowing the type of singularity at .∞ for an entire function, one
can determine its form and many other properties.

Lemma 6.1 If .∞ is removable for an entire function .f, then f is constant.

Proof By Theorem 5.3, the function f can be represented as the sum of the power
series centered at zero:
+∞

f (z) =
. cn zn for all z ∈ C. (6.26)
n=0

Due to the uniqueness of the Laurent series expansion (Theorem 6.2), the series
(6.26) is also the Laurent series of f around .∞ and . +∞ n
n=1 cn z is its principal part.
Since .∞ is removable for .f, the principal part must be absent, i.e., .cn = 0 for all
.n ∈ N. Thus, .f ≡ c0 .

Lemma 6.2 If .∞ is a pole for an entire function .f, then f is a polynomial.

Proof Since .∞ is a pole for .f, the principal part of the Laurent series (6.26) of f
around .∞ must contain a finite number of nonzero terms, i.e.,

f (z) = c0 + c1 z + c2 z2 + . . . + cm zm ,
. cm = 0 and cn = 0 for all n > m.

The lemma is proved.

Definition 6.6 An entire function is called a transcendental entire function if .∞ is

its essential singular point.

Example 6.8 Obviously, the following functions are entire transcendental func-
tions: .ez , .sin z, .cos z, .sinh z, .cosh z.

From Liouville’s Theorem 5.4 it follows that the image of an entire non-
constant function must be unbounded. Based on the Great Picard Theorem 6.7 and
the fundamental theorem of algebra (Theorem 5.5), the following more stronger
statement becomes obvious.

Theorem 6.8 (Little Picard’s Theorem) The domain of values of a non-constant

entire function is the whole complex plane, except perhaps for only one point.
148 6 Laurent Series: Isolated Singularities of Analytic Functions

Exercise 6.3 Give a reasoned answer to the question:

Is there a non-constant analytic function f defined on .C \ {i} which maps .C \ {i}
into .{z : 0 < |z| < 1}?

Now consider an interesting class of analytic functions whose only isolated

singularities are poles.

Definition 6.7 A function f is said to be meromorphic in a domain .Ω if f is

analytic in .Ω except for the poles.

Clearly the class of meromorphic functions includes both analytic functions (the
set of poles is empty) and rational functions (a ratio of two polynomials).

Remark 6.2 Taking into account that poles are isolated singularities, it follows
from Definition 6.7 that

• the set of poles of a meromorphic function is at most countable;

• if the set of poles of a meromorphic function in .Ω is countable, then the
limit points of this set lie on the boundary of .Ω;
• the limit point of poles of a meromorphic function in .C can only be .∞;
• a meromorphic function in .C has a finite number of poles, and .∞ is either
its removable point or a pole.

Example 6.9 Obviously, the functions

. tan(z) and cot(z)

are meromorphic in the complex plane .C, each of which has a countable set of
 −1
poles. The function .f (z) = sin πz is not meromorphic in .C, since the origin is
its non-isolated singularity (see (6.15)).

Similar as for entire functions, one can sometimes establish the form of a
meromorphic function, knowing the structure of the set of its poles.

Theorem 6.9 If f is a meromorphic function in .C, then f is rational.

Proof It follows from the theorem condition and the fourth point of Remark 6.2 that
f has a finite number of poles .{a1 , . . . , an } and .∞ is either a removable point or a
pole for it.
6.5 Classification of Analytic Functions 149

By Theorem 6.4, the principal part of the Laurent series of f around the pole .ak
has a finite number of nonzero terms; we denote this principal part as follows

(k) (k)
c−Nk c−1 (k)
pk (z) :=
. + ... + , c−N = 0.
(z − ak )Nk (z − ak ) k

If .∞ is a pole, then we introduce the following notation for of the principal part of
the Laurent series of f around .∞:

p0 (z) := c1 z + c2 z2 + . . . + cN zN ,
. cN = 0.

Now consider the function .ϕ := f − nk=0 pn . Since the principal part of

the Laurent series of .ϕ around each point of the set .{a1 , . . . , an , ∞} is absent,
Theorem 6.3 implies that the points .a1 , . . . , an , ∞ are removable for .ϕ. Extending
the function .ϕ by continuity at these points (see Corollary 6.2), we obtain an analytic
function in .C. But then it follows from Lemma 6.1 that .ϕ ≡ c0 , whence
 

n (k)
c−Nk (k)
c−1
f (z) = c0 + c1 z + . . . + cN z +
.
N
+ ... + , (6.27)
(z − ak )Nk (z − ak )
k=1

i.e., f is rational.

Remark 6.3 The formula (6.27) obtained in the proof shows that an arbitrary
rational function can be decomposed into its integer part (polynomial) and the sum
of simple fractions.

Now we complete the picture of the behavior of a meromorphic function with a

theorem, which is given here without proof.

Theorem 6.10 (Little Picard’s Theorem for Meromorphic Functions) A non-

constant meromorphic function f in .C attains every complex number except maybe
one or two.

A number that is not assumed by a meromorphic function is called a Picard

exceptional value for that function.

Example 6.10 It is easy to check that the Picard exceptional values of the mero-
morphic function .tan z are the numbers .±i.
1
Zero is just one Picard exceptional value of the meromorphic function . 1−z .

In Sect. 7.6 we will show that every meromorphic function in .C is a ratio of two
entire functions.
Residue Calculus
7

Abstract

Just as a person’s character is manifested in extreme situations, so the properties

of analytic functions are determined by their behavior in isolated singularities. In
this chapter, we will illustrate this claim with examples of integral calculations.
It turns out that in order to calculate the integral of an analytic function along a
curve, it is necessary to determine some values, called residues, of that function
at its singularities. The reader can appreciate both the power and the simplicity
of the residue theory developed by Cauchy for calculating complicated integrals,
including integrals of real-valued functions. This theory helps to deduce amazing
formulas both for the decomposition of a meromorphic function, e.g. .cot z, into
an infinite sum of simple fractions that are responsible for its poles, and for
the factorization of an entire function, e.g. .sin z, into an infinite product of
factors that are responsible for its zeros. The theory also supplies a ready-made
framework for counting zeros and poles of a given meromorphic function or
zeros of an analytic function, in particular, we prove the argument principle and
Rouché’s theorem.

7.1 Cauchy’s Residue Theorem

One of the main theorems of complex analysis is proved here, which opens the door
to the calculation of integrals of various kinds.

Definition 7.1 Let .f be an analytic function in a punctured disk .B̆R (a), where
a ∈ C. The residue of .f at the point .a is the number
.


1
. Res f (z) := f (z) dz, (7.1)
z=a 2π i
{|z−a|=ρ}+

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 151
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_7
152 7 Residue Calculus

where .ρ is an arbitrary number from the interval .(0, R).

Definition 7.2 Let .f ∈ A(B̆R (∞)). The residue of .f at .∞ is the number


1
. Res f (z) := − f (z) dz, (7.2)
z=∞ 2π i
{|z|=ρ}+

where .ρ is an arbitrary number from the interval .(R, +∞).

Remark 7.1 The integrals (7.1) and (7.2) are independent of .ρ due to Cauchy’s
integral Theorem 4.5. By this theorem, .Resz=a g(z) = 0 if g is an analytic function
in a. Since integration is a linear operation, finding the residue is also a linear
operation:
 
. Res λ f (z) + μ g(z) = λ Res f (z) + μ Res g(z) for all λ, μ ∈ C.
z=a z=a z=a

Theorem 7.1 (Cauchy’s Residue Theorem) Let

• the conditions of Theorem 4.8 be satisfied for a bounded domain .Ω,

• a function f be analytic in the closure of .Ω except for a finite number of points
.a1 , . . . , ap from .Ω .(we denote this as follows: .f ∈ A(Ω \ {a1 , ..., ap }) ).

Then
 
p
. f (z) dz = 2π i Res f (z).
∂+Ω z=ak
k=1

Proof There is a positive number r such that for all .k ∈ {1, . . . , p}

Br (ak ) ⊂ Ω,
. and Br (ak ) ∩ Br (am ) = ∅ for k = m.

Denote by


p
Ωr := Ω \
. Br (ak ).
k=1

By virtue of the theorem

 conditions, the function .f ∈ A(Ω r ). Then it follows
from Corollary 4.5 that . ∂ + Ωr f (z) dz = 0 or, taking into account that

  p 

. f (z) dz = f (z) dz − f (z) dz,
∂ + Ωr ∂+Ω +
k=1 ∂ Br (ak )
7.2 Formulas for Calculating Residues 153

we get
 p 
 
p
by (7.1)
. f (z) dz = f (z) dz = 2π i Res f (z).
∂+Ω + z=ak
k=1 ∂ Br (ak ) k=1

The theorem is proved. 

Theorem 7.2 (On the Full Sum of Residues) Let a function f be analytic in .C
except for a finite number of points .{a1 , a2 , . . . , ap }, i.e.,
.f ∈ A(C\{a1 , a2 , . . . , ap }). Then


p
. Res f (z) + Res f (z) = 0.
z=ak z=∞
k=1

Proof It is clear that one can pick a positive number R such that .ak ∈ BR (0) for all
k ∈ {1, . . . , p}. According to the previous theorem,
.

 
p
. f (z) dz = 2π i Res f (z),
∂ + BR (0) z=ak
k=1

or


p 
1
. Res f (z) − f (z) dz = 0,
z=ak 2π i ∂ + BR (0)
k=1

from where, remembering Definition 7.2, we have


p
. Res f (z) + Res f (z) = 0.
z=ak z=∞
k=1

7.2 Formulas for Calculating Residues

To apply the theorems from the previous section, we need to be able to calculate
residues. Below we derive the main formulas.

1. Let .f ∈ A(B̆R (a)) and .a ∈ C. According to Theorem 6.1, the function f

expands into a Laurent series in .B̆R (a), whose coefficients .{cn }n∈Z are deter-
mined by the formulas (6.5). Recalling Definition 7.1, we conclude that

. Res f (z) = c−1 . (7.3)

z=a
154 7 Residue Calculus

2. Let .f ∈ A(B̆R (∞)). The coefficients of the Laurent series of f in .B̆R (∞) are
determined as follows

1 f (z)
.cn = dz, n ∈ Z, ∀ρ ∈ (R, +∞).
2π i {|z|=ρ}+ zn+1

Based on Definition 7.2, we conclude that

. Res f (z) = −c−1 . (7.4)

z=∞

3. Let .f ∈ A(B̆R (a)) and .a ∈ C. If .a is removable for the function f , then the
principal part of the Laurent series of f around a is absent, i.e., .cn = 0 for all
negative integer n. Thus, due to (7.3) we have .Resz=a f (z) = 0.

Important
If .∞ is a removable singular point of f , then it cannot be asserted that
.Resz=∞ f (z) = 0, because the coefficient .c−1 is not included in the principal

part of the Laurent series of f around .∞ (see Sect. 6.3).

But if .∞ is a zero of order m for f (see Theorem 5.15) and .m ≥ 2, then
.Resz=∞ f (z) = 0.

4. Let .f ∈ A(B̆R (a)), .a ∈ C, and the point .a be a pole of order m for f ; .m ≥ 2. In

this case, in virtue of Theorem 6.4 the Laurent series of f around a has the form

 +∞
c−m c−1
f (z) =
. + ... + + cn (z − a)n , z ∈ B̆R (a).
(z − a) m z−a
n=0

Multiplying this equality by .(z − a)m , we get

+∞

(z − a)m f (z) = c−m + c−m (z − a) + . . . + c−1 (z − a)m−1 +
. cn (z − a)n+m .
n=0

Then differentiating the previous equality .(m − 1) times, we obtain

d m−1 
+∞
 (n + m)!
. (z−a) m
f (z) = (m−1)! c−1 + cn (z−a)n+1 , z ∈ B̆R (a).
dzm−1 (n + 1)!
n=0
7.2 Formulas for Calculating Residues 155

Passing here to the limit as .z → a, we find the coefficient .c−1 . Taking (7.3) into
account, we conclude that the residue of .f at the mth order pole .a is calculated
by the formula

1 d m−1 
. Res f (z) = lim m−1 (z − a)m f (z) . (7.5)
z=a (m − 1)! z→a dz

Example 7.1 By Corollary 6.3, the point 1 is a pole of order 2 for the function

cos 2z
f (z) =
. .
(z − 1)2

Using (7.5) with .m = 2, we obtain

d  d 
. Res f (z) = lim (z − 1)2 f (z) = lim cos 2z = −2 sin 2.
z=1 z→1 dz z→1 dz

5. If .a ∈ C is a simple pole of f , then the Laurent series of f is as follows

 +∞
c−1
f (z) =
. + cn (z − a)n , z ∈ B̆R (a).
z−a
n=0

Finding the coefficient .c−1 from this equality, we have

. Res f (z) = lim (z − a)f (z). (7.6)

z=a z→a

It should be noted that the formula (7.6) coincides with (7.5) for .m = 1.
6. Let
ψ(z)
f (z) =
. , z ∈ B̆R (a),
ϕ(z)

where the functions .ϕ and .ψ are analytic in the disk .BR (a), the point .a ∈ C is a
simple zero of .ϕ and .ψ(a) = 0. Then, obviously, the point a is a simple pole of
the function f . Applying the formula (7.6), we find

ψ(z) ψ(z) ψ(a)

. Res f (z) = lim (z − a) = lim ϕ(z)−ϕ(a)
= . (7.7)
z=a z→a ϕ(z) z→a ϕ (a)
z−a
156 7 Residue Calculus

Example 7.2 The function .cot z = cos z

sin z has isolated singularities at the points
.zk = π k, .k ∈ Z. By Corollary 6.3 these points are simple poles since for each

.k ∈ Z the representation

Ψ (z)
. cot z = , z ∈ B̆ π2 (π k),
z − πk

holds, where the function

(−1)k cos z · z−π k

sin(z−π k) , if z ∈ B̆ π2 (π k);
Ψ (z) =
.
1, if z = π k,

is analytic in the disk .B π2 (π k) and is not equal to zero there (see Example 6.3).
By using (7.7), we have

cos π k
. Res cot z = =1 for all k ∈ Z. (7.8)
z=π k (sin z) |z=π k

7. Let .f ∈ A(B̆R (∞)) and .∞ be a pole of order m for f . Then the Laurent series
of the function f around .∞ is given by


0
f (z) =
. cn zn + c1 z + ... + cm zm , z ∈ B̆R (∞)

n=−∞ the principal part

(see Sect. 6.3). Consider the function

−2

1 c1 cm
ϕ(z) = f
.
z = cn z−n + c−1 z + c0 + + ... + m , z ∈ B̆ 1 (0).
n=−∞
z z R

Similarly, as in the fourth item above, we find the coefficient .c−1 :

d m+1  m 
. lim m+1
z ϕ(z) = (m + 1)! c−1 .
z→0 dz

Considering the formula (7.4), we get

1 d m+1   
. Res f (z) = − lim m+1 zm f 1z . (7.9)
z=∞ (m + 1)! z→0 dz
7.2 Formulas for Calculating Residues 157

Example 7.3 Consider the function

2
.f (z) = + 1 + z2 (7.10)
z

for which .∞ is a pole of order 2.

Due to Theorem 6.2 (on the uniqueness of a Laurent expansion), the right-
hand side of (7.10) is the Laurent series of f around .∞, whose principal part
equals .z2 , and the coefficient .c−1 = 2. Therefore, by using (7.4), we get

. Res f (z) = −2.

z=∞

The formula (7.9) gives the same result:

 
1 d3  1
. Res f (z) = − lim 3 z2 2z + 1 + 2
z=∞ 3! z→0 dz z
1 d3 
= − lim 3 2z3 + z2 + 1 = −2.
3! z→0 dz

8. If an analytic function f is even and the points 0 and .∞ are its isolated singular
points, then

. Res f (z) = 0 and Res f (z) = 0.

z=0 z=∞

Indeed, if we consider, for example, the point 0, then for the function f the
following representations hold:

+∞
 +∞

f (z) =
. cn zn and f (−z) = cn (−z)n , z ∈ B̆R (0),
n=−∞ n=−∞

wherefrom, taking into account the evenness of .f, we get

+∞

0=
. c2k+1 z2k+1 , z ∈ B̆R (0). (7.11)
k=−∞

By Theorem 6.2, we get from (7.11) that .c2k+1 = 0 for all .k ∈ Z; so .c−1 = 0.
158 7 Residue Calculus

7.3 Methods for Calculating Integrals

The purpose of this section is to present a collection of methods and examples that
can be used to calculate various types of integrals, including improper real integrals
that often cannot be calculated using methods of real analysis.

Integrals Over Closed Curves

When calculating integrals over closed curves, Cauchy’s residue Theorems 7.1 and
7.2 (on the full sum of residues) are used.

cos z
Example 7.4 Compute . 3
dz =: I .
{|z|=2}+ z

Solution For the function

cos z
f (z) :=
.
z3

there is only one isolated singularity .z = 0, which is inside the integration contour—
the circle .{|z| = 2}. By Theorem 7.1, the integral .I = 2π i Resz=0 f (z).
There are a number of ways to find the residue of f at 0. It is ease to see that
the point 0 is a third order pole for f , so we can use the formula (7.5) with .m = 3.
However, if it is easy to expand a function into a Laurent series, then it leads to the
result faster. In our case
+∞ +∞
cos z 1  (−1)n z2n −3 z−1  (−1)n 2n−3
. = = z − + z for all z ∈ C \ {0}.
z3 z3 (2n)! 2 (2n)!
n=0 n=2

So, the coefficient .c−1 = − 12 . By (7.3) we have

I = 2π i Res f (z) = −π i.
.
z=0

dz
Example 7.5 Compute . =: I .
{|z|=2}+ (z7 + 1)3

Solution The function

1
f (z) :=
.
(z7 + 1)3

has 7 poles, namely

π 2π k
ak = ei( 7 +
. 7 ) , k ∈ {0, 1, . . . , 6},
7.3 Methods for Calculating Integrals 159

inside the circle .{|z| = 2}, and all these poles have order 3. By Theorem 7.1


6
I = 2π i
. Res f (z).
z=ak
k=0

In this case, to avoid tedious calculations, it is better to use Theorem 7.2, which
gives


6
. Res f (z) = − Res f (z).
z=ak z=∞
k=0

Since
1 1
f (z) =
. ∼ 21 as z → ∞,
(z7 + 1) 3 z

the point at infinity is a zero of order 21 for f . Therefore, the coefficient .c−1 of the
Laurent series of f around .∞ is equal to 0 (see Important in Sect. 7.2). This means
that .I = 0.

Trigonometric Integrals

Here we show how to calculate definite trigonometric integrals of the form

 2π  
I :=
. R cos (nϕ), sin (mϕ) dϕ, (7.12)
0

where .R(u, v) is a rational function of two real variables u and .v, i.e., it can be
written as a ratio of two polynomials

Q(u, v)
R(u, v) =
. for all (u, v) ∈ R2 ,
P (u, v)

and .P (u, v) = 0 for all .(u, v) such that .u2 + v 2 = 1.

The trick is to use the substitution .z = eiϕ , ϕ ∈ [0, 2π ], to transform the integral
(7.12) into a complex integral over the unit circle, to which one can use Cauchy’s
residue theorem.
It is easy to see that the parametrization .z = eiϕ , ϕ ∈ [0, 2π ], give us the
positively oriented circle .{z : |z| = 1}. In addition,

dz = i eiϕ dϕ ⇒ dϕ = −iz−1 dz,

.

einϕ + e−inϕ zn + z−n zm − z−m

. cos (nϕ) = = , sin (mϕ) = .
2 2 2i
160 7 Residue Calculus

As a result, we get the integral


I=
. R1 (z)dz,
{|z|=1}+

where
1 1 n
R1 (z) =
. R 2 (z + z−n ), 2i1 (zm − z−m )
iz

is a new rational function of a complex variable, which can be evaluated using the
residue theorem.

Example 7.6 Compute the integral

 2π dϕ
I :=
. ,
0 1 − 2a cos ϕ + a 2

where .a ∈ R and .|a| = 1.

Solution For .a = 0, the answer is obvious. Therefore, we further assume that

a = 0. After the substitution .z = eiϕ , ϕ ∈ [0, 2π ], we get
.

 
−iz−1 dz dz
I=
. = −i
1 − a(z + z−1 ) + a 2 (a − z)(az − 1)
{|z|=1}+ {|z|=1}+

i dz
= .
a (z − a)(z − a1 )
{|z|=1}+

We see that the integrand

1
.f (z) =
(z − a)(z − a1 )

has two simple poles at .z = a and .z = a1 , and only one of which is inside of the
circle .{|z| = 1} (it depends on a). Thus, by Theorem 7.1, we have

• if .|a| < 1, then

i 2π 2π 1 2π
I=
. 2π i Res f (z) = − lim (z − a)f (z) = − lim = ;
a z=a a z→a a z→a z − 1
a
1 − a2
7.3 Methods for Calculating Integrals 161

• if .|a| > 1, then

2π 2π   2π
I =−
. Res f (z) = − lim z − a1 f (z) = 2 .
a z= a
1 a z→ a1 a −1

Integrals Along the Real Line

First, let us prove the lemma.

Lemma 7.1 Let the following conditions be satisfied:

 
1. .f ∈ A {z : Im z ≥ 0}\{a1 , . . . , am } , where .Im ak > 0, .k ∈ {1, 2, . . . , m};
2. there exist positive numbers .M, .R0 and .δ > 0 such that for all z from the set
.{ξ : |ξ | > R0 , Im ξ ≥ 0}

M
|f (z)| ≤
. . (7.13)
|z|1+δ

Then

. lim f (z) dz = 0, where γr = reit , t ∈ [0, π ].
r→+∞ γ
r

Proof It can be considered that .r > max{R0 , |a1 |, . . . , |am |}. Then, taking (4.5)
and (7.13) into account, we deduce
   
  M M Mπ
 f (z) dz ≤ dl = dl = δ → 0 as r → +∞.
.
 
γr |z|
1+δ r 1+δ r
γr γr

The lemma is proved. 

Theorem 7.3 Let the conditions of Lemma 7.1 be satisfied. Then

 +∞ 
m
. f (x) dx = 2π i Res f (z). (7.14)
−∞ z=ak
k=1

Proof First, we note

 +∞ that, based on the second condition of Lemma 7.1, the
improper integral . −∞ f (x) dx converges.
Now let us take .r > max{R0 , |a1 |, . . . , |am |} and consider the positively oriented
closed contour .[−r, r]∪γr , where .γr = reit , t ∈ [0, π ]. The interior of this contour
162 7 Residue Calculus

contains all the isolated singularities .{a1 , . . . , am } of f that are in the upper half-
plane. Under the Cauchy residue Theorem 7.1 we have

r   
m
. f (x) dx + f (z) dz = f (z) dz = 2π i Res f (z).
z=ak
−r γr [−r,r]∪γr k=1

Letting r tend to infinity in this equality and using Lemma 7.1, we arrive at the
formula (7.14). 
 +∞ x2 + 1
Example 7.7 Compute the integral . dx =: I .
0 x4 + 1

Solution Since the integrand is even,

 +∞
1 x2 + 1
.I = dx.
2 −∞ x4 + 1

Let us verify the conditions of Lemma 7.1 for the function

z2 + 1 ψ(z)
f (z) =
. = , where ψ(z) = z2 + 1, ϕ(z) = z4 + 1.
z4 + 1 ϕ(z)

The function f is rational and has poles at the points where the function .ϕ vanishes;
they are
π 3π 5π 7π
a1 = ei 4 ,
. a2 = ei 4 , a3 = ei 4 , a4 = ei 4 .

Only .a1 and .a2 lie in the upper half-plane .{z : Im z > 0}. They are simple poles
because

ψ(ak ) = 0 and
. ϕ (ak ) = 4 ak3 = 0, k ∈ {1, 2, 3, 4}.

If .|z| > 2, then

1 + 1  1 1+
1
1  z2  |z|2 4
.|f (z)| =  ≤ 2 ≤ .
|z|2 1+ 1 |z| 1 − 1 |z|2
z4 |z|4

Hence the second condition of Lemma 7.1 is satisfied .(M = 4, R0 = 2, δ = 1).

By using (7.14) and (7.7), we find
 +∞
1 x2 + 1  
. I= dx = π i Res f (z) + Res f (z)
2 −∞ x +1
4 z=a1 z=a2
7.3 Methods for Calculating Integrals 163

 ψ(a ) ψ(a2 )  e i 2π + 1 e i 3π2 +1

1
= πi + = πi i 3π
+ i 9π
ϕ (a1 ) ϕ (a2 ) 4e 4 4e 4
πi  i π iπ
√
= − e 4 − e− 4 = π sin π4 = π 22 .
2

 +∞
Fourier Transform Type Integrals f (x) eiλx dx (λ > 0)
−∞

Such integrals occur in physical and engineering applications, and Jordan’s Lemma
plays a fundamental role in their computation.

Lemma 7.2 (Jordan’s Lemma) Let the following conditions be satisfied:

  
• .f ∈ A {z : Im z ≥ 0}\ {a1 , . . . , am } ∪ {x1 , . . . , xn } , where .Im ak > 0 for .k ∈
{1, . . . , m}, and .{x1 , . . . , xn } ⊂ R;
•

. max |f (z)| =: Mr → 0 as r → +∞, (7.15)

z∈Eγr

where .γr (t) = reit , t ∈ [0, π ].

Then

. lim f (z) eiλz dz = 0 (λ > 0). (7.16)
r→+∞ γ
r

Proof Take .r > max{|a1 |, . . . , |am |, |x1 |, . . . , |xn |}. Then the statement of this
lemma follows from the following considerations:
       
   iλz 
π
 iλr(cos t+i sin t) 
 f (z)eiλz dz ≤ |f (z)| e  dl ≤ M e  r dt
.
  r
γr γr 0
 π  π
2
= Mr r e−λr sin t dt = 2Mr r e−λr sin t dt
0 0
 π
2 2 Mr π
≤ 2Mr r e−λr π t dt = (1 − e−λr ) → 0 as r → +∞.
0 λ

In the last line, we used the obvious inequality . π2 t ≤ sin t for .t ∈ [0, π2 ]. 
164 7 Residue Calculus

Remark 7.2 Comparing the second condition of the Jordan lemma with the second
condition of Lemma 7.1, we see that f can tend to zero at infinity more slowly. This
is due to the presence of the factor .eiλz near f .

Theorem 7.4 Let the conditions of Lemma 7.2 be satisfied and .{x1 , . . . , xn } be
simple poles of f and .x1 < x2 < . . . < xn . Then

+∞ 
m 
n
.p.v. f (x)eiλx dx = 2π i Res f (z)eiλz + π i Res f (z)eiλz . (7.17)
z=ak z=xk
−∞ k=1 k=1

Before proving, we recall that the principal value of an improper integral, which
 +∞
is divergent, is the way in which we assign a finite value to it. For example, . −∞ dx
x
is divergent, but
 −ε dx
 r dx
. lim + = 0.
r→+∞, ε→0 −r x ε x
 +∞
So, .p.v. dx
−∞ x = 0. In (7.17)

+∞ 
.p.v. f (x)eiλx dx := lim f (x)eiλx dx,
r → +∞
−∞ ε → 0 Jr,ε

n 
where .Jr,ε := [−r, r] \ k=1 (xk − ε, xk + ε) .

Proof We choose r as in Jordan’s lemma. Let

ε0 :=
. min |xk − al |, ε1 := min |xk+1 − xk |.
k ∈ {1, . . . , n} k∈{1,...,n−1}
l ∈ {1, . . . , m}

Then for any .ε ∈ (0, ε2 ), where .ε2 = min{ε0 , ε1 /2}, we have

Eγkε ∩ Eγjε = ∅ for k = j,

.

where .γkε (t) = xk − εe−it , t ∈ [0, π ], k ∈ {1, . . . , n}.

Now consider the closed positively oriented contour

  
n
Cr,ε := γr
. Jr,ε γkε ,
k=1
7.3 Methods for Calculating Integrals 165

where the upper semicircle .γr is defined in Jordan’s lemma. The interior of .Cr,ε
contains only the isolated singularities .{a1 , . . . , am } of the function f ; the poles
.{x1 , . . . , xn } are in the exterior.

The Cauchy residue Theorem 7.1 says that

 
m
. f (z)eiλz dz = 2π i Res f (z)eiλz , (7.18)
Cr,ε z=ak
k=1

Considering (7.16) and passing to the limit in (7.18) as .r → +∞ and .ε → 0, we

find

+∞ 
m 
n 
.p.v. f (x)e dx = 2π i
iλx
Res f (z)e −
iλz
lim f (z)eiλz dz. (7.19)
z=ak ε→0
−∞ k=1 k=1 γkε

It remains to find the limits in the right-hand side of (7.19). Since for any .k ∈
{1, . . . , n} the point .xk is a simple pole of the function .f (z)eiλz , its Laurent series
is as follows
(k) +∞

c−1 (k)
.f (z)eiλz = + cj (z − xk )j , z ∈ B̆δk (xk ),
z − xk
j =0

where .δk is a positive number. Denote by .gk the sum of the power series
+∞ (k)
j =0 cj (z − xk ) . Due to Theorem 5.6 the function .gk can be considered analytic
.
j

in the closed disk .Bδk (xk ).

In what follows, we assume that .ε < min{δ1 , . . . , δn }. Using (7.3), we get

  (k)  π (k) 
c−1 c−1
. f (z)e iλz
dz = dz + gk (z) dz = εie−it dt + gk (z) dz
z − xk −εe−it
γkε γkε γkε 0 γkε

 
(k)
. = −π i c−1 + gk (z) dz = −π i Res f (z)eiλz + gk (z) dz. (7.20)
z=xk
γkε γkε

Since for any .k ∈ {1, . . . , n} the function .gk ∈ A(Bδk (xk )), there exists a positive
constant .Mk such that for all .z ∈ Bδk (xk ) the inequality .|gk (z)| ≤ Mk holds. Then
  
 
 
. gk (z) dz ≤ |gk (z)| dl ≤ Mk π ε → 0 as ε → 0. (7.21)
 γε  γε
k k

Based on (7.20) and (7.21), from (7.19) the equality (7.17) follows. 
166 7 Residue Calculus

Example 7.8 Compute the integral

 +∞ eiλx
p.v.
. dx =: I (λ > 0).
−∞ x

Solution It is easy to check that the function .f (z) = 1z satisfies the conditions of
Jordan’s lemma and the point 0 is its only simple pole. Therefore, from (7.17) and
(7.6), we have

eiλz
I = π i Res
. = π i lim eiλz = π i.
z=0 z z→0

Using this result, we simply evaluate the Dirichlet integral

 +∞  +∞   +∞ iλx 
sin λx 1 sin λx 1 e π
. dx = dx = Im p.v. dx = .
0 x 2 −∞ x 2 −∞ x 2

A key point here is the replacement .sin λx by .eiλx and the identity

1
. sin λx = Im eiλx .
2
Remark 7.3 It is known from mathematical analysis that an antiderivative of the
integrand in the Dirichlet integral is not an elementary function. Nevertheless, the
value of the integral can be obtained in various (not always short and simple)
ways, including double integration and differentiation under the integral sign. In
Example 7.8, the calculation of the Dirichlet integral takes one line!
 +∞ (x − 1) cos 5x
Example 7.9 Compute . dx =: I.
−∞ x 2 − 2x + 5

Solution As before, using the Euler formula, we get

  +∞ 
(x − 1)
.I = Re p.v.
5ix
e dx .
−∞ x 2 − 2x + 5

The function
z−1 1
f (z) =
. ∼ as z → ∞,
z2 − 2z + 5 z
7.4 Argument Principle: Rouché’s Theorem and Its Applications 167

it has two singular points .a1 = 1 + 2i and .a2 = 1 − 2i. They are simple poles and
only .a1 lies in the upper half-plane. So, all conditions of Theorem 7.4 are satisfied
and it implies that
 
z−1
I = Re 2π i Res
. e5iz
z=a1 (z − a1 )(z − a2 )
 
a1 − 1 5ia1
= 2π Re i e = −e−10 π sin 5.
a1 − a2

7.4 Argument Principle: Rouché’s Theorem and Its

Applications

In this section, we consider applications that help to count the zeros and poles of a
given meromorphic function. Let .f be meromorphic in a domain D.

f
Definition 7.3 The function .
f , where it is defined, is called the logarithmic
derivative of f .

Remark 7.4 This definition is justified by the fact that

d   f (z)
. logk f (z) = ,
dz f (z)

where .logk is any branch of the multi-valued function .Log (see (3.9)).

Let .a be a zero of order .n for the function .f. According to Theorem 5.13 (on
the zero of an analytic function), there is a positive number .δ and a unique function
.ϕ ∈ A(Bδ (a)) such that

f (z) = (z − a)n ϕ(z)

. and ϕ(z) = 0 for all z ∈ Bδ (a).

Using this representation, we find that for all .z ∈ B̆δ (a)

f (z) n(z − a)n−1 ϕ(z) + (z − a)n ϕ (z) n ϕ (z)

. = = + ,
f (z) (z − a) ϕ(z)
n z−a ϕ(z)

whence, remembering Remark 7.1, we get

f (z)
. Res = n. (7.22)
z=a f (z)
168 7 Residue Calculus

Let .b be a pole of order p for .f. Corollary 6.3 (on the pole of an analytic function)
says there is a positive number .δ1 and a unique analytic nonzero function .ψ in
.Bδ1 (b) such that

ψ(z)
f (z) =
. for all z ∈ B̆δ1 (b).
(z − b)p

As before, we find for all .z ∈ B̆δ1 (b)

p
f (z) − (z−b) p+1 ψ(z) + (z−b)p ψ (z)
1
p ψ (z)
. = =− + ,
f (z) 1
(z−b)p ψ(z)
z − b ψ(z)

from where

f (z)
. Res = −p. (7.23)
z=b f (z)

Let us now denote by .{ak } the set of zeros, and by .{bk } the set of poles of the
function f . Consider a bounded domain .Ω, whose boundary is the union of a finite
number of pairwise disjoint Jordan curves, such that

Ω ⊂ D,
.

∂Ω ∩ {ak } = ∅, ∂Ω ∩ {bk } = ∅,
Ω ∩ {ak } = {a1 , . . . , aq }, Ω ∩ {bk } = {b1 , . . . , bm }.

Under these assumptions,

f  
. ∈ A Ω \ ({a1 , . . . , aq } ∪ {b1 , . . . , bm }) . (7.24)
f

Hereinafter we consider that there are neither zeros nor poles of f on .∂Ω.

Definition 7.4 Let the above assumptions hold. The logarithmic residue of f with
respect to the positively oriented boundary of .Ω is the integral

1 f (z)
. dz.
2π i ∂+Ω f (z)

Theorem 7.5 (On a Logarithmic Residue) Let the above assumptions be satis-
fied. Then

1 f (z)
. dz = Z − P , (7.25)
2π i ∂+Ω f (z)
7.4 Argument Principle: Rouché’s Theorem and Its Applications 169

where .Z := n1 + n2 + . . . + nq , .P := p1 + p2 + . . . + pm , and .ni is the order of

the zero .ai and .pi is the order of the pole .bi .

Proof Since the inclusion (7.24) holds, the Cauchy residue Theorem 7.1 and
formulas (7.22) and (7.23) yield
  f (z)  f (z)  
q m q m
1 f (z)
. dz = Res + Res = nk − pk = Z − P .
2π i f (z) z=ak f (z) z=bk f (z)
∂+Ω k=1 k=1 k=1 k=1

Example 7.10 The formula (7.25) can be used for calculation of integrals, for
example,
 
dz 1 (tan z)
. = dz = π i(1 − 2) = −π i.
∂ + B3 (0) sin 2z 2 ∂ + B3 (0) tan z

Here it was easy to see that .tan has one zero and two poles in the disk .B3 (0).

Theorem 7.6 (Argument Principle) Let the conditions of Theorem 7.5 for a
meromorphic function f be satisfied and .Ω be a simply connected domain.
Then the difference between the number of its zeros .Z and the number of its poles
.P inside .Ω is equal to the increment of the argument of .f (z) when z passes once

along the positively oriented boundary .∂ + Ω, divided by .2π :

1
Z−P =
. Δ∂ + Ω Argf (z). (7.26)
2π
Before the proof, we explain what the increment of the argument along a curve
is. Let .z = γ(t), .t ∈ [α, β], be a curve whose trace does not contain the origin, i.e.,
.Eγ ∩ {0} = ∅. The angle of rotation of the vector z when the point z moves along

the trace of .γ from its initial point to the end point is called the increment of the
argument z along .γ and is denoted by .Δγ Arg z (see Fig. 7.1).

Fig. 7.1 The increment of

the argument along a curve .γ
170 7 Residue Calculus

Example 7.11 It is easy to verify that

• if .z = γ(t) = 1 + it, .t ∈ [−1, 1], then .Δγ Arg z = π2 ;

• if .z = γ(t) = e−it , .t ∈ [0, π ], then .Δγ Arg z = −π ;
• if .z = γ(t) = 2eit , .t ∈ [−π, π ], then .Δγ Arg z = 2π ;
• If .z = γ(t) = 2 + eit , .t ∈ [−π, π ], then .Δγ Arg z = 0.

This example shows that the increment of the argument along a curve does not
depend on the continuous branch of the argument that we choose at the beginning
of the movement of the point z. Therefore, in this example, in the equality (7.26)
and further in this section, any continuous branch of the multi-valued function .Arg
is assumed.

Proof Since the inclusion (7.24) holds, there is a positive number .δ > 0 such that
f
.
f ∈ A(Uδ (∂Ω)), where .Uδ (∂Ω) = {z : dist(z, ∂Ω) < δ}.
Because the domain .Ω is simply connected, its boundary coincides with the trace
of a positively oriented Jordan curve .z = γ(t), .t ∈ [α, β], i.e., .Eγ = ∂Ω. By
Theorem 4.3 (on an antiderivative along a curve), there exists an antiderivative .Ψ of
the function . ff along .γ. It follows from Theorem 4.3 and Remark 7.4 that

Ψ (t) = ln |f (γ(t))| + i Arg(f (γ(t)) + C,

. t ∈ [α, β].

Then by the Newton-Leibnitz formula (see Theorem 4.4), we have


1 f (z) 1  
. dz = Ψ (β) − Ψ (α)
2π i ∂+Ω f (z) 2π i
1 
= lim Argf (γ(t)) − lim Argf (γ(t))
2π t→β−0 t→α+0

1
=: Δ∂ + Ω Argf (z). (7.27)
2π
From (7.25) and (7.27) it follows (7.26). 

Example 7.12 Let’s check the formula (7.26) for the function .f (z) = z3 in the
domain .Ω := B1 (0); .γ = eit , t ∈ [0, 2π ], is a positively oriented Jordan curve
whose trace coincides with .∂Ω.
On one hand it is easy to see that .Z = 3 and .P = 0 for f , and on the other hand

1  2π
1  1
. Δ∂ + Ω Argf (z) = Arg e3it  = (6π − 0) = 3.
2π 2π 0 2π

Here we have fixed such a continuous branch that .Arg e3it |t=0 = 0.
7.4 Argument Principle: Rouché’s Theorem and Its Applications 171

Important
If f is analytic, i.e., .P = 0, then the formula (7.26) counts zeros (with their
multiplicities) of f in .Ω, i.e.,

1
. Z= Δ∂ + Ω Argf (z). (7.28)
2π
Informally, this formula can be explained as follows: an analytic function
f has as many zeros in a simply connected domain .Ω as many times the
radius vector .f (z) rotates around the origin when the point z passes once the
boundary of .Ω counterclockwise.

In the same way, we can count how many times an analytic function f takes the
value .w0 (such points are called .w0 -points of .f ).

Definition 7.5 An analytic function f is said to take the value .w0 at a point .z0 ∈ Ω
with multiplicity .n ∈ N if

f (z) = w0 + (z − z0 )n g(z),
. z ∈ Bδ (z0 ), (7.29)

where .g ∈ A(Bδ (z0 )) and .g(z0 ) = 0. 

.

Similarly to how (7.25) was proved, we obtain


1 f (z)
. dz = Zf (w0 ), (7.30)
2π i ∂+Ω f (z) − w0

where .Zf (w0 ) is the number of .w0 -points of f in .Ω counting multiplicity. If .Ω is

simply connected, then

1
Zf (w0 ) =
. Δ∂ + Ω Arg(f (z) − w0 ).
2π
The argument principle is used indirectly through Rouche’s theorem, which has
many important applications, some of which are proved in this subsection and others
in Chap. 9.

Theorem 7.7 (Rouché’s Theorem) Let .Ω be a simply connected domain and

f, g ∈ A(Ω). If
.

|g(z)| < |f (z)| for all z ∈ ∂Ω,

. (7.31)
172 7 Residue Calculus

then f and .f + g have the same number of zeros (counting multiplicity) in the
domain .Ω, i.e.,

Zf = Zf +g
. in Ω,

where .Zf is the number of zeros (counting multiplicity) of f .

Proof From the condition (7.31) it follows that

|f (z)| = 0 and
. |f (z) + g(z)| = 0 for all z ∈ ∂Ω.

To argue the last relation we also used the inequality

|f (z) + g(z)| ≥ |f (z)| − |g(z)| > 0

. for all z ∈ ∂Ω.

Then, recalling the properties of the argument (see Sect. 1.1), we obtain
 
g(z)
.Arg(f (z) + g(z)) = Argf (z) + Arg 1 + for all z ∈ ∂Ω. (7.32)
f (z)

Here, as in Theorem 7.6, .Arg(f + g) is any continuous branch of the multi-valued

function .Arg. Again using (7.31), we get
  
 g(z)  |g(z)|
 
.
 1 + f (z) − 1 = |f (z)| < 1 for all z ∈ ∂Ω.

g(z)
This means that .1 + remains in the disk .B1 (1) for all .z ∈ ∂Ω. Therefore,
f (z)
 
g(z)
Δ∂ + Ω Arg 1 +
. = 0.
f (z)

Considering this and (7.32), Theorem 7.6 implies

1 1
Zf +g =
. Δ∂ + Ω Arg(f + g) = Δ∂ + Ω f = Zf .
2π 2π
The theorem is proved. 

This theorem is often used to find the number of roots of an equation in a given
domain.

Example 7.13 Find the number of roots of the equation

z9 − 6z4 + 3z − 1 = 0
.
7.4 Argument Principle: Rouché’s Theorem and Its Applications 173

in the unit disk .B1 (0).

Solution Let .f (z) = −6z4 and .g(z) = z9 + 3z − 1. It is easy to see that

|g(z)| ≤ |z|9 + 3|z| + 1 = 5 < 6 = |f (z)| for all z ∈ ∂B1 (0).

.

Thus, all conditions of the Rouché theorem are satisfied for the functions f and g
in .B1 (0). Therefore, .Zf +g = Zf = 4 in .B1 (0).
Recall that zeros are counted taking into account their multiplicities. Clearly, the
function f has one zero in .B1 (0) and this is a zero of order 4.

Exercise 7.1 Prove the fundamental theorem of algebra using Rouché’s theorem.

Exercise 7.2 Find the number of roots of the equation .z4 + 10z + 1 = 0 in the
annulus .{z : 1 < |z| < 2}.

Theorem 7.8 (Sufficient Conditions of Conformality) If a function f is analytic

and univalent in a domain .Ω, then f is conformal in .Ω.

Proof Let us prove the theorem by contradiction. Suppose there exists a point .z0 ∈
Ω such that .f (z0 ) = 0. Then .z0 is an isolated and finite multiple zero of the
derivative .f . If this is not the case, then, according to Corollaries 5.6 and 5.8,
.f ≡ const, and this contradicts the univalence of f . Thus, there is a positive number

.δ, such that

f (z) = 0 for all z ∈ Bδ (z0 ) \ {z0 }.

. (7.33)

By Theorem 5.3, we have

+∞

f (z) = c0 +
. ck (z − z0 )k for all z ∈ Bδ (z0 ), (7.34)
k=n

(n)
where .n ≥ 2 and .cn = f n!(z0 ) = 0. It is easy to understand that .(n − 1) is the order
of the zero .z0 of the derivative .f . Since .cn = 0, there exists .δ1 ∈ (0, δ) such that

+∞

. ck (z − z0 )k−n = 0 for all z ∈ Bδ1 (z0 ). (7.35)
k=n

Define the function

+∞
 +∞

ϕ(z) :=
. ck (z − z0 )k = (z − z0 )n ck (z − z0 )k−n , z ∈ Bδ1 (z0 ).
k=n k=n
174 7 Residue Calculus


Obviously, .ϕ ∈ A Bδ1 (z0 ) as the sum of a power series, and .z0 is a zero of
multiplicity n of the function .ϕ. Due to (7.35)
+∞ 
 
 
.m0 := min |ϕ(z)| = δn1 min  ck (z − z0 )k−n  = 0.
z∈∂Bδ1 (z0 ) z∈∂Bδ1 (z0 )  
k=n

Let .ψ(z) = −α for all .z ∈ Bδ1 (z0 ), where .α is a fixed positive number less than
.m0 . Since .|ψ(z)| = α < m0 ≤ |ϕ(z)| for all .z ∈ ∂Bδ1 (z0 ), Rouché’s theorem yields
.Zϕ+ψ = Zϕ = n in .Bδ1 (z0 ). But

ϕ(z) + ψ(z) = f (z) − c0 − α for all z ∈ Bδ1 (z0 ).

.

Thus,

.Zf −c0 −α = n in Bδ1 (z0 ),

 
or, more precisely, in .B̆δ1 (z0 ), because . f (z) − c0 − α |z=z0 = −α.
Since
d  
. f (z) − c0 − α = f (z) = 0 for all z ∈ Bδ1 (z0 ) \ {z0 }
dz

(see (7.33)), the function .f − c0 − α has .n (n ≥ 2) distinct simple zeros in .B̆δ1 (z0 ).
This means that there are two points .z1 = z2 , z1 , z2 ∈ B̆δ1 (z0 ) such that .f (z1 ) =
f (z2 ). But this contradicts to the univalence of the function f .
Hence, .f (z) = 0 for all .z ∈ Ω, i.e., the function f is conformal in .Ω. 

Remark 7.5 The conditions of Theorem 7.8 are not necessary for conformality.
Indeed, the function .f (z) = ez is conformal in .C, but is not univalent in .C (see
Sect. 3.7).

The next statement shows how the zeros of analytic functions that form a
uniformly convergent sequence are related to the zero of the limit function.

Theorem 7.9 (Hurwitz’s Theorem) Let .{fn }n∈N be a sequence of analytic func-
tions in a domain .Ω, which converges uniformly on any compact set .K ⊂ Ω to a
function f that is not identically equal to a constant.
If .z0 ∈ Ω is a zero of .f, then for any disk .Br (z0 ) ⊂ Ω there exists .n0 ∈ N such
that

Zfn > 0
. in Br (z0 ) for any n ≥ n0 .
7.4 Argument Principle: Rouché’s Theorem and Its Applications 175

Proof The function f is continuous in .Ω as the limit of uniformly convergent

continuous functions. Take any triangle . which, together with its closure, belongs
to .Ω. Then
 
. f (z) dz = lim fn (z) dz = 0,
∂+ n→+∞ ∂ + 

and due to Morera’s Theorem 5.9 the function .f ∈ A(Ω).

Since .f = const, its zeros are isolated and finitely multiple (see Corollaries 5.6
and 5.8). If .z0 ∈ Ω is a zero of .f, then for any .r > 0 such that .Br (z0 ) ⊂ Ω there
exists a number .δ ∈ (0, r) that

f (z) = 0 for all z ∈ Bδ (z0 ) \ {z0 }.

.

Denote by .μ := min |f (z)|. It is clear that .μ > 0. Since the sequence

z∈∂Bδ (z0 )
{fn }n∈N converges uniformly on .Bδ (z0 ) to .f, there exists .n0 ∈ N such that for all
.

integer .n ≥ n0 :

|fn (z) − f (z)| < μ ≤ |f (z)| for all z ∈ ∂Bδ (z0 ).

.

This means that the Rouché theorem can be applied to the functions .fn − f and f .
As a result, .Zfn = Z(fn −f )+f = Zf > 0 in .Bδ (z0 ). 

Example 7.14 (Counterexample from Mathematical Analysis) For real func-

tions, the statement of Hurwitz’s theorem is incorrect. Indeed, consider the function
sequence

1
fn (x) = x 2 +
. , x ∈ (−1, 1), n ∈ N.
n
Then, it is easy to verify that

(−1,1)
fn ⇒ f = x 2
. as n → +∞,

and .f (0) = 0. However, for all .n ∈ N

fn (x) = 0 for all x ∈ (−1, 1).

.

Corollary 7.1 If .{fn }n∈N is a sequence of analytic and univalent functions in a

domain .Ω, which converges uniformly on any compact set .K ⊂ Ω to an analytic
function .f, then either f is univalent or .f ≡ const in .Ω.

Proof As in Theorem 7.9, we show that .f ∈ A(Ω). Next, we go on to prove the

corollary by contradicting it.
176 7 Residue Calculus

Suppose .f ≡ const and there are two points .z1 = z2 in .Ω such that .f (z1 ) =
f (z2 ). Consider the function sequence

gn (z) := fn (z) − fn (z1 ),

. z ∈ Br (z2 ), n ∈ N,

where .r < |z1 − z2 | and .Br (z2 ) ⊂ Ω. It is obvious that .gn ∈ A(Br (z2 )),

Br (z2 )
gn (z)
. ⇒ g(z) = f (z) − f (z1 ) as n → +∞,

g(z2 ) = 0, and .g ≡ const. Using Hurwitz’s theorem for .{gn }n∈N , we get
.

∃ n0 ∈ N ∀ n ≥ n0 :
. Zgn > 0 in Br (z2 ).

This means that for any integer .n ≥ n0 there is a point .zn∗ ∈ Br (z2 ) such that

gn (zn∗ ) = fn (zn∗ ) − fn (z1 ) = 0.

.

This contradicts the univalence condition of .fn , because .zn∗ = z1 . 

Example 7.15 (Counterexample from Mathematical Analysis) Consider the

function sequence

x3, x ∈ [0, 1),

fn (x) =
. n ∈ N.
1 3
nx , x ∈ (−1, 0),
 
It is easy to check that .fn ∈ C 1 (−1, 1) , .fn is an injection and

(−1,1) x3, x ∈ [0, 1),
fn (x) ⇒ f (x) :=
.
0, x ∈ (−1, 0),

But, .f ≡ const and f is not univalent in .(−1, 1).

Exercise 7.3 Prove that if .{fn }n∈N is a sequence of analytic functions in a domain
Ω, which converges uniformly on any compact .K ⊂ Ω to an analytic function .f,
.

and each .fn is nonzero everywhere in .Ω, then either .f ≡ 0 or f is also nowhere
zero in .Ω.

7.5 Partial Fraction Decomposition of a Meromorphic

Function

For a meromorphic function with a finite number of poles, the formula (6.27) about
its decomposition into the sum of a polynomial and the sum of simple fractions was
7.5 Partial Fraction Decomposition of a Meromorphic Function 177

proved. In this section, we consider the case when a meromorphic function in .C

has a countable number of simple poles. As was noted in Remark 6.2 the point at
infinity is the limit point of its poles in this case.

Definition 7.6 A sequence of positively oriented Jordan curves .{γn }n∈N is said to
be regular if the following conditions are satisfied:

(1) for any .n ∈ N : .{0} ⊂ int(γn ) and .int(γn ) ⊂ int(γn+1 );

(2) .dn := min |z| → +∞ as .n → +∞;
z∈Eγn

(3) there exists a positive constant .C such that . dγnn ≤ C for all .n ∈ N.

Remark 7.6 The first condition in Definition 7.6 means that the interior of the
curve .γn with its closure belongs to the interior of .γn+1 , and the origin is inside
all these Jordan curves. The second condition says that the traces of these curves
expand to infinity in any direction as .n → +∞; and the last one means that the
traces expand uniformly in all directions.
 
Example 7.16 The sequence of the circles . γn (t) = neit , .t ∈ [0, 2π ] n∈N is
regular because for it: .dn = n, γn = 2 π n, C = 2π .

Example 7.17 If the trace of .γn coincides with the boundary of the rectangle
[−n2 , n2 ] × [−ni, ni], then the sequence .{γn }n∈N is not regular. Indeed, for this
.

sequence .dn = n and .γn = 2 n + 2 n2 , but the value . dγnn = 2(1 + n) is unbounded
as .n → +∞.

Theorem 7.10 Let

• f be a meromorphic function in .C and its poles .{an }n∈N be simple and

0 < |a1 | ≤ |a2 | ≤ . . . ≤ |an | ≤ . . . → +∞ as n → +∞;

.

• there be a regular sequence of positively oriented Jordan curves .{γn }n∈N and a
positive constant M such that

|f (z)| ≤ M
. for all z ∈ Eγn and for all n ∈ N. (7.36)

Then for all .z ∈ C \ {an }n∈N

+∞
  1 1
f (z) = f (0) +
. Ak + , (7.37)
z − ak ak
k=1
178 7 Residue Calculus

where .Ak := Resz=ak f (z), and moreover, for an arbitrary bounded domain .Ω, the
series (7.37) converges uniformly on .Ω\{an }n∈N .

Proof For each .n ∈ N consider the integral


1 zf (ξ )
In (z) :=
. dξ, z ∈ int(γn )\{ak }k∈N .
2π i γn ξ(ξ − z)

zf (ξ )
Due to the theorem conditions, the integrand .F (ξ ) := has only simple
ξ(ξ − z)
poles .0, z, a1 , . . . , amn in the interior of the curve .γn .
Then, by Cauchy’s residue Theorem 7.1 we have


mn
In (z) = Res F (ξ ) + Res F (ξ ) +
. Res F (ξ )
ξ =0 ξ =z ξ =ak
k=1


mn
z
= − f (0) + f (z) + Ak . (7.38)
ak (ak − z)
k=1

Since
z  1 1
. =− + ,
ak (ak − z) z − ak ak

we get from (7.38) that


mn  1 1
f (z) = f (0) +
. Ak + + In (z) (7.39)
z − ak ak
k=1

for all .z ∈ int(γn )\{ak }mn

k=1 .
Let us estimate .In (z). Take an arbitrary bounded domain .Ω. Since .{γn }n∈N is a
regular sequence of Jordan curves, there exist numbers .R > 0 and .n0 ∈ N such that

Ω ⊂ BR (0)
. and BR (0) ⊂ int(γn ) for all n ≥ n0 .

Then, for all .z ∈ int(γn )\{ak }mn

k=1

1 |z||f (ξ )| M R lγn MRC
.|In (z)| ≤ dl ≤ ≤ −→ 0
2π γn |ξ ||ξ − z| 2 π dn (dn − R) 2 π(dn − R)
(7.40)

as .n → +∞. Taking (7.40) into account and passing to the limit in (7.39) as .n →
+∞, we get (7.37). 
7.5 Partial Fraction Decomposition of a Meromorphic Function 179

Remark 7.7 The summation in the formula (7.37) proceeds as follows: first, we
sum the terms related to the poles from the interior of .int(γ1 ), then from .int(γ2 ) \
int(γ1 ) and so on.

Remark 7.8 The condition (7.41) can be weaken and replaced by

|f (z)| ≤ M |z|p
. for all z ∈ Eγn and for all n ∈ N, (7.41)

where .p ∈ N. Then the following partial fraction decomposition will be valid:

+∞
 

p
f (m) (0) m  1  zl
p
.f (z) = z + Ak + . (7.42)
m! z − ak akl+1
m=0 k=1 l=0

The decomposition (7.42) coincides with (7.37) if .p = 0.

Example 7.18 Perform the partial fraction decomposition of the cotangent.

Solution We cannot directly apply the formula (7.37) since 0 is a pole of .cot z.
Therefore we first expand the following meromorphic function:
⎧
⎨ 1
cot z − , z = {π n}n∈Z ,
f (z) =
. z
⎩ 0, z = 0.

Using L’Hopital’s rule twice, we find

 1 z cos z − sin z
. lim cot z − = lim = 0.
z→0 z z→0 z sin z

Thus, poles of f are .{π n}n∈Z\{0} and all of them are simple (Example 7.2).
Consider a sequence of positively oriented Jordan curves .{γn }n∈N , whose
traces coincide with the boundaries of squares .{An Bn Cn Dn }n∈N (see Fig. 7.2),
respectively, where .αn = π2 + π n. It is easy to verify that this sequence is regular.
Now we show that the inequality (7.41) holds for the function f . First take any
.z ∈ [Dn , Cn ], i.e., .z = x + iαn , .x ∈ [−αn , αn ]. Using (3.16), we obtain

 2iz   
 e + 1   e−2αn e2ix + 1  1 + e−2αn 1 + e−π
.| cot z| =   =  ≤ ≤ .
 e2iz − 1   e−2αn e2ix − 1  1 − e−2αn 1 − e−π

Let us now consider any .z ∈ [Bn , Cn ], i.e., .z = αn + iy, .y ∈ [−αn , αn ]. Then

 π  e|y| − e−|y|
 
.| cot z| = cot + π n + iy  = | tan (iy)| = | tanh y| ≤ |y| ≤ 1.
2 e + e−|y|
180 7 Residue Calculus

Fig. 7.2 The regular

sequence of Jordan curves
.{γn }

Because .| cot(−z)| = | cot(z)|, then

1 + e−π
|f (z)| ≤
. + 1 for all z ∈ Eγn and for all n ∈ N.
1 − e−π

In virtue of Remark 7.1 and (7.8) we have that

Ak = Res f (z) = 1.
.
z=π k

In .int(γk ) \ int(γk−1 ) there are two poles .π k and .−π k of the function f . Therefore,
according to the formula (7.37) and Remark 7.7, we get

+∞  
1  1 1 1 1
. cot z = + − + +
z z+kπ kπ z−kπ kπ
k=1
+∞

1 2z
= + for all z ∈ C\{π k}k∈Z . (7.43)
z z2 − k 2 π 2
k=1

 1
Example 7.19 Find the sum . +∞ , a ∈ R \ {0}.
k=1
k2 + a2

Solution In the formula (7.43) we put .z = i a π . Then

+∞
i 2ai  1
. cot(iaπ ) = − − ,
aπ π k2 + a2
k=1
7.6 Factorization of an Entire Function Into an Infinite Product 181

or
+∞
 1 π 1
. = coth(π a) − 2 .
k2 + a2 2a 2a
k=1

Exercise 7.4 Perform the partial fraction decomposition of the following functions:
1 1
tan z,
. 2 , ez −1 .
sin z

 1
Exercise 7.5 Find the sum . +∞ , a ∈ R \ {0}.
k=1
(k 2 + a 2 )2

7.6 Factorization of an Entire Function Into an Infinite Product

It is known (see Theorem 5.14) that every polynomial can be factorized into a
product of elementary factors. In this section, we will show that, under some
additional assumptions, every entire function can also be decomposed into a product
(possibly infinite) of elementary factors.
Let an entire function f have a finite number of zeros .a1 , . . . , am of order
.n1 , . . . , nm , respectively. Then the function

f (z)
. Φ(z) = (7.44)
(z − a1 )n 1 · . . . · (z − am )nm

has isolated singularities at the points .a1 , . . . , am . By Theorem 5.13, they are
removable, and therefore the function .Φ can be extended by continuity at these
points (see Corollary
  6.2). As a result, .Φ is an entire function without zeros. Hence
.F (z) = log Φ(z) is also an entire function, where .log is the principal branch of

the logarithm. From the last relation and (7.44), we obtain

.f (z) = eF (z) (z − a1 )n1 · . . . · (z − am )nm , z ∈ C. (7.45)

A natural question is: what decomposition does an entire function have if it has
a countable number of zeros, e.g. .sin z? To answer this question, we recall some
definitions from mathematical analysis and rewrite them in terms of complex values.
  
Definition 7.7 An infinite product . +∞
k=1 1 + fk (z) is said to converge to a
function f in a domain .Ω if .1 + fk (z) = 0 for all .k ∈ N and for all .z ∈ Ω,
and


n
 
. lim 1 + fk (z) = f (z). (7.46)
n→+∞
k=1
182 7 Residue Calculus

  
If the limit (7.46) is uniform in .z ∈ Ω, then the product . +∞
k=1 1 + fk (z) is
called uniformly convergent in .Ω to the function f . .

  
If the product . +∞
k=1 1 + fk (z) uniformly in .Ω converges to f and .fk ∈ A(Ω),
then, based on Corollary 5.5, the function f is analytic in .Ω.

Theorem 7.11 (Weierstrass’s Factorization Theorem) Let

• f be an entire function, .f (0) = 0, and moduli of its zeros .{ak }k∈N form a non-
decreasing sequence:

0 < |a1 | ≤ |a2 | ≤ . . . ≤ . . . ≤ |ak | ≤ . . . ,

. and lim |ak | = +∞;
k→+∞

f
• the function be uniformly bounded on a regular sequence of positively
.
f
oriented Jordan curves .{γk }k∈N .

Then for all .z ∈ C we have

f (0) 

+∞
z nk  nk 
f (z) = f (0) exp
.
f (0) z 1− exp ak z , (7.47)
ak
k=1

where .nk is the order of zero .ak . Moreover, for any bounded domain .Ω the product
(7.47) is uniformly convergent in .Ω.

Proof It is easy to see that the function . ff is meromorphic in .C and has poles only at
the points .{ak }k∈N ; in addition, they are simple poles and due to the formula (7.22)
f
.Resz=ak
f = nk . Considering the theorem conditions, we can apply the formula
(7.37) to . ff . As a result,

f (0)   1
+∞
f (z) 1
. = + nk + for all z ∈ C \ {ak }k∈N . (7.48)
f (z) f (0) z − ak ak
k=1

Furthermore, the series (7.48) is uniformly convergent in .Ω\{ak }k∈N , where .Ω is

an arbitrary bounded domain.
Fix any point .ξ ∈ C \ {ak }k∈N and take a curve .z = γ(t), t ∈ [α, β], whose trace
does not pass through the zeros of .f, i.e., .Eγ ∩ {ak }k∈N = ∅, and whose endpoints
are .0 = γ(α) and .ξ = γ(β). Integrating (7.48) along the curve .γ, we obtain
  +∞
  
f (z) f (0) 1 1
. dz = dz + nk + dz.
γ f (z) f (0) γ γ z − ak ak
k=1
7.6 Factorization of an Entire Function Into an Infinite Product 183

With the help of the Newton-Leibnitz formula (see (4.17)) we get

+∞
   
f (ξ ) f (0) ξ ξ
. log = ξ+ nk log 1 − + , (7.49)
f (0) f (0) ak ak
k=1

whence
 +∞   
f (0)  ξ ξ
.f (ξ ) = f (0) exp ξ+ nk log 1 − +
f (0) ak ak
k=1
  +∞

f (0) ξ nk  nk 
= f (0) exp ξ 1− exp ak ξ . (7.50)
f (0) ak
k=1

Obviously, we have the identity .0 = 0 for .ξ = ak in the product (7.50). Therefore,

(7.50) holds for all .ξ ∈ C. The uniform convergence of (7.50) in any bounded
domain .Ω is equivalent to the uniform convergence of the series

+∞
   
z z
. nk log 1 − +
ak ak
k=1

in .Ω \ {ak }k∈N . But this is a consequence of the uniform convergence of the series
(7.48). 

Remark 7.9 By using the formula (7.47), one can construct entire functions that
have zeros of a given multiplicity at given points.

Exercise 7.6 Prove that a meromorphic function in .C can be represented as a ratio

of two entire functions.

Example 7.20 Factorize the function .sin z into an infinite product.

Solution Obviously, the function

sin z
z , z = 0,
f (z) =
.
1, z = 0.

is an entire function, .f (0) = 0 and .{π k}k∈Z\{0} are simple poles of .f . In addition,
the function
z cos z−sin z
f (z) z2 1
. = sin z
= cot z −
f (z) z
z
184 7 Residue Calculus

is uniformly bounded on the regular sequence of Jordan curves from Example 7.18,
and .f (0) = 0.
Thus, all conditions of Theorem 7.11 are satisfied for the function f and the
formula (7.47) gives


+∞
z  z  z  z
. sin z = z 1+ exp − 1− exp
πk πk πk πk
k=1


+∞
z2

=z 1− 2 2 for all z ∈ C. (7.51)
π k
k=1

Taking .z = π
2 in (7.51), we get the Wallis formula

+∞   +∞
π  1 π  (2k)2
.1 = 1− ⇐⇒ =
2 (2k)2 2 (2k − 1)(2k + 1)
k=1 k=1

or
π 2·2 4·4 6·6
. = · · ···
2 1·3 3·5 5·7

Exercise 7.7 Factorize the function .ez − 1 into an infinite product.

Analytic Continuations
8

Abstract

Analytic functions have surprised us in previous chapters with their properties. In

this chapter we will learn about another interesting property of analytic functions,
namely the possibility of continuing an analytic function from the domain in
which it is defined to a larger one. An important property of the analytic
continuation procedure is that it is unique. In essence, this means that knowing
the values of an analytic function in a small domain, or even on a curve, uniquely
determines the value of the function at any other point where that function
can be continued. This is a rather surprising and extremely strong assertion.
Moreover, for some analytic functions, the analytic continuation can lead to a
new concept of a function, which we have already experienced empirically in
Sects. 3.6 and 3.7. Here we will find out under which conditions the analytic
continuation leads to a multi-valued function, and under which conditions the
newly extended function is single-valued. Along the way we will be introduced
to various continuation techniques and other fundamental concepts of complex
analysis such as monodromy, global analytic functions, their singularities and
Riemann surfaces.

8.1 Analytic Function Elements

Thanks to the uniqueness theorem (see Theorem 5.12), the usual way to define
analytic functions is to first specify the function only in a small domain, and then
extend it by analytic continuation to the largest possible domain. Therefore, we now
give some basic definitions of continuation.

Definition 8.1 Let .Ω be a domain in .C and E be a subset of .Ω. An analytic

function .F : Ω → C is called an analytic continuation of a function .f : E → C
from the set E into the domain .Ω if .F (z) = f (z) for all .z ∈ E.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 185
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_8
186 8 Analytic Continuations

In other words, the restriction of F to the set E is the function f .

Definition 8.2 Let G be a domain in .C and .f ∈ A(G). An ordered pair .(f, G) is

called an analytic function element.
 
Definition 8.3 An analytic function element . f, Br (a) is said to be a canonic
analytic function element if the radius of the disk .Br (a) is
⎛  ⎞−1
|f (n) (a)| ⎠
r = ⎝lim sup
n
. , (8.1)
n→+∞ n!

i.e., .Br (a) is the largest disk in which the function f can be expanded as the sum of
a power series centered at a.

Definition 8.4 Two analytic function elements .(f1 , G1 ) and .(f2 , G2 ) are called
direct analytic continuations of each other if .G1 ∩ G2 =: D = ∅, .D is a domain
and .f1 (z) = f2 (z) for all .z ∈ D (Fig. 8.1).

Lemma 8.1 Let .(f2 , G2 ) and .(f3 , G2 ) be two direct analytic continuations of an
analytic function element .(f1 , G1 ). Then .(f2 , G2 ) = (f3 , G2 ), i.e., no more than
one function can be analytic in .G2 and coincides with .f1 in .D = G1 ∩ G2 .

Proof By Definition 8.4, .f1 (z) = f2 (z) = f3 (z) for all .z ∈ D. Then, due to
Theorem 5.12 we have that .f2 ≡ f3 in .G2 .

Example 8.1 Consider two canonic analytic function elements .(f1 , G1 ) and
(f2 , G2 ), where
.

+∞
f1 (z) =
. zn , z ∈ G1 := {z ∈ C : |z| < 1},
n=0

+∞ n √
1 z−i
f2 (z) =
. , z ∈ G2 := {z ∈ C : |z − i| < 2}.
1−i 1−i
n=0

Fig. 8.1 Analytic function

elements .(f1 , G1 ) and
.(f2 , G2 ) that are direct
analytic continuations of each
other
8.1 Analytic Function Elements 187

Since the sums of these series are

1 1 1 1
f1 (z) =
. , z ∈ G1 , and f2 (z) = = , z ∈ G2 ,
1−z 1 − i 1 − 1−i
z−i 1−z

the function elements .(f1 , G1 ) and .(f2 , G2 ) are direct analytic continuations of
each other. Obviously, each of these analytic function elements is the direct analytic
continuation of the analytic function element .(f3 , G3 ), where

1
f3 (z) =
. , z ∈ G3 := C \ {1}.
1−z

The question naturally arises: how to construct an analytic continuation of a given

analytic function, defined on some domain, into a larger domain? The mathemati-
cians Weierstrass and Riemann proposed a method of analytic continuation called
the method
 of re-expansion
 of power series. Next, we have a look at how it works.
Let . f, Br (a) be an analytic function element, where

+∞
. f (z) = cn (z − a)n , z ∈ Br (a).
n=0

Take any point .b ∈ Br (a) and rewrite this representation as follows

+∞ +∞
 n
f (z) =
. cn (z − a)n = cn (z − b) + (b − a)
n=0 n=0
+∞ n +∞
k
= cn (z − b)k (b − a)n−k = dn (z − b)n , (8.2)
n
n=0 k=0 n=0

(n)
where the coefficients .dn = f n!(b) , n ∈ N0 , are uniquely determined by the
theorem on the uniqueness of the expansion of an analytic function in a power series
(see Theorem 5.7).
Next we find the convergence radius of the power series . +∞
n=0 dn (z − b)
n


−1
n |f (n) (b)|
r1 =
. lim sup
n→+∞ n!

and determine the function

+∞
f1 (z) =
. dn (z − b)n , z ∈ Br1 (b).
n=0
188 8 Analytic Continuations

Fig. 8.2 The case

.r1> r − |b − a|

 
Then . f1 , Br1 (b) is a new canonic analytic function element.
Due Theorem 5.3, the radius .r1 ≥ r − |b − a|. Two cases are possible, namely
 
• if .r1 > r − |b − a| (see Fig. 8.2), then . f1 , Br1 (b) is the direct analytic
continuation of the analytic function element . f, Br (a) ;
• if .r1 = r − |b − a|, then it is said that through the point .z0 =∂Br (a) ∩ ∂Br1 (b)
it is not possible to analytically
 extend
 the function element . f, Br (a) and the
point .z0 is singular for . f, Br (a) .

It is easy to see that the point .z0 = 1 is singular for the analytic function element
(f1 , G1 ) from Example 8.1. We may apply this approach to any domain on which
.

f is analytic.

Definition 8.5 Let .(f, Ω) be an analytic function element and .z0 ∈ ∂Ω.
We say that the analytic function element.(f, Ω) is continued through the point
.z0 if there is an analytic function element . F, Br (z0 ) which is a direct analytic

continuation of .(f, Ω).

If such a continuation does not exist, then the point .z0 is called a singular point
of the analytic function element .(f, Ω).

Definition 8.6 A domain .Ω is called a domain of analyticity (domain of holomor-

phy or natural domain) of an analytic function f if every point on .∂Ω is a singular
point of the analytic function element .(f, Ω).

Example 8.2 Consider the function

+∞
k
f (z) =
. z2
k=1

By the formula (5.4), the convergence radius of this power series is 1, so f is analytic
in the unit circle .B1 (0). If .z from .B1 (0) approaches .z0 , where .z0 is a root of one
k
of the equations .z2 = 1, k ∈ N, then .f (z) → ∞. Therefore, f has a singularity
8.1 Analytic Function Elements 189

at every .2k th root of 1. Since these roots are dense on .∂B1 (0), the disk .B1 (0) is a
domain of analyticity of f .

Remark 8.1 Every domain in .C is a domain of analyticity of some analytic

function. Indeed, it possible to define an analytic function with isolated zeros
accumulating everywhere on the boundary of a given domain, which must then be
its domain of analyticity.
 
Theorem 8.1 (On a Singularity of a Canonic Analytic Element) Let . f, Br (a)
be a canonic analytic function element. Then it has a singular point .z0 ∈ ∂Br (a).
 
Proof Since . f, Br (a) is a canonic analytic function element, the function f can
be represented as the sum of the power series

+∞  −1
f (z) =
. cn (z − a)n in Br (a), where r = lim sup n
|cn | .
n→+∞
n=0
 
Assume that the canonic analytic function element . f, Br (a) has no singular
points on .∂Br (a). Then for any point .z0 ∈ ∂Br (a) there is an analytic  function

element . fz0 , Bδ(z0 ) (z0 ) which is the direct analytic continuation of . f, Br (a) .
As .∂Br (a) is compact, it is possible to choose a finite subcover (see Fig. 8.3)

{Brn (zn )}N

. n=1 of the cover {Bδ(z0 ) (z0 )}z0 ∈∂Br (a)
N
such that .∂Br (a) ⊂ n=1 Brn (zn ). Denote by


N 
G :=
. Brn (zn ) ∪ Br (a).
n=1

Fig. 8.3 Thesubcover

. {Brn (zn )}n=1
N
190 8 Analytic Continuations

In the domain G we define the function


f (z), z ∈ Br (a),
.F (z) := (8.3)
fn (z), z ∈ Brn (zn ), n ∈ {1, . . . , N}.
 N
Since . fn , Brn (zn ) n=1 are direct analytic continuations of the analytic function
 
element . f, Br (a) , we have

fn (z) = fn+1 (z) = f (z)

. for all z ∈ Brn (zn ) ∩ Brn+1 (zn+1 ) ∩ Br (a)

and for all .n ∈ {1, . . . , N − 1}, and

fN (z) = f1 (z) = f (z)

. for all z ∈ BrN (zN ) ∩ Br1 (z1 ) ∩ Br (a).

Therefore, according to Theorem 5.12, for any .n ∈ {1, . . . , N − 1}

fn (z) = fn+1 (z)

. for all z ∈ Brn (zn ) ∩ Brn+1 (zn+1 ),
fN (z) = f1 (z) for all z ∈ BrN (zN ) ∩ Br1 (z1 ).

Consequently, the formula (8.3) correctly defines the analytic function F in the
domain G.
It is easy to recognise (Fig.
 8.3) that the distance from .∂G to .Br (a) is positive;
let us denote it by .α := dist ∂G, Br (a) > 0. In virtue of the analyticity of F in the
disk .Br+α (a), it can be expanded in the power series

+∞
F (z) =
. dn (z − a)n for all z ∈ Br+α (a),
n=0

and it follows from (5.4) and (8.3) that

 
1  n |F (n) (a)| n |f (n) (a)| 1
. ≥ lim sup n |dn | = lim sup = lim sup = .
r +α n→+∞ n→+∞ n! n→+∞ n! r

Since .α > 0, this inequality is a contradiction, which completes the proof.

Remark 8.2 Theorem 8.1 can be used to find radii of convergence of power series.
For example, the radius of convergence of the power series

+∞
. tan z = cn zn , z ∈ Br (0),
n=0
8.2 Methods of Analytic Continuation: Schwarz’s Reflection Principle 191

is equal to the distance from the point 0 to the nearest singular point of the function
tan z, i.e. .r = π2 .
.

It is clear that the method of re-expansion of power series is inefficient for

constructing analytic continuations. In the next section, we will get to know more
effective methods of analytic continuation, which can be used to construct new
important non-elementary functions.

8.2 Methods of Analytic Continuation: Schwarz’s Reflection

Principle

To feel the power statement of the following theorem, let us first consider an
example. We define two real-valued functions

.g1 (x1 , x2 ) = −x1 , (x1 , x2 ) ∈ (−1, 0) × (0, 1),

and

g2 (x1 , x2 ) = x1 ,
. (x1 , x2 ) ∈ (0, 1) × (0, 1).

These functions are smooth in their domains and

. lim g1 (x1 , x2 ) = lim g2 (x1 , x2 ) for all x2 ∈ (0, 1).

x1 →0− x1 →0+

But, the function


g1 (x1 , x2 ), (x1 , x2 ) ∈ (−1, 0] × (0, 1),
g(x1 , x2 ) =
.
g2 (x1 , x2 ), (x1 , x2 ) ∈ (0, 1) × (0, 1),

is only continuous in the rectangle .(−1, 1) × (0, 1).

For complex-valued analytic functions, the situation is quite different. The
following theorem is true.

Theorem 8.2 (Analytic Continuation by Continuity) Assume that the following

conditions are satisfied:

(1) domains .Ω1 and .Ω2 do not intersect, however the intersection of their closures
is .Γ := Ω 1 ∩ Ω 2 , and it is the trace of some smooth curve;
(2) in each domain .Ωk , an analytic function .fk is given, which is also defined on .Γ
and is continuous on .Ωk ∪ Γ, i.e., .fk ∈ A(Ωk ) ∩ C(Ωk ∪ Γ ), k ∈ {1, 2};
(3) for all .z ∈ Γ

f1 (z) = f2 (z).
.
192 8 Analytic Continuations

Then the function


f1 (z), z ∈ Ω1 ∪ Γ,
f (z) :=
.
f2 (z), z ∈ Ω2 ,

is analytic in the domain D, where .D = Ω 1 ∪ Ω 2 .

Proof From the theorem’s conditions follows that .f ∈ C(D). Consider an arbitrary
triangle . which, together with its closure, belongs to the domain D. Then two
cases are possible: either . ∩ Γ = ∅, or . ∩ Γ = ∅.
If . ∩ Γ = ∅, then based on the Cauchy-Goursat theorem for triangles
(Theorem 4.1)

. f dz = 0.
∂+

Consider the case of . ∩ Γ = ∅. We assume that .Γ divides the triangle . into

two open sets .Ξ1 and .Ξ2 (see Fig. 8.4); other cases are either treated similarly or
are obviously simplified, for example, when .Γ intersects only a side or a vertex of
.. Then

  
. f dz = f1 dz + f2 dz. (8.4)
∂+ ∂ + Ξ1 ∂ + Ξ2

Since .fk ∈ A(Ξk ) ∩ C(Ξk ) for .k ∈ {1, 2} and taking into account Remark 4.10,
each of the integrals on the right-hand side of (8.4) is equal to zero.
Thus, for an arbitrary triangle . which, together with its closure, belongs to
the domain .D, . ∂ +  f dz = 0. Then Morera’s theorem (Theorem 5.9) says that
.f ∈ A(D).

Remark 8.3 The theorem remains valid when .Γ is the union of at most a countable
number of smooth curves.
 
Exercise 8.1 Let .f ∈ A B1 (0) ∩ C(B1 (0) ∪ Γ0 ), where .Γ0 is an arc of the circle
.{z ∈ C : |z| = 1}, and .f (z) = 0 for all .z ∈ Γ0 . Prove that .f = 0 in .B1 (0).

Fig. 8.4 Intersection of .Γ

with a triangle .
8.2 Methods of Analytic Continuation: Schwarz’s Reflection Principle 193

Example 8.3 Example 5.8 shows that the gamma function

 +∞
Γ (z) =
. e−t t z−1 dt, z ∈ Ωr = {ξ : Re ξ > 0}, (8.5)
0

is analytic in the right half-plane .Ωr . Let us show that this function can be continued
in the complex plane with the exception of simple poles.

Solution Integration by parts in the integral (8.5) gives

Γ (z + 1)
.Γ (z + 1) = z Γ (z) ⇐⇒ Γ (z) = for all z ∈ Ωr . (8.6)
z

The last identity in (8.6) makes it possible to continue the gamma function in the half
plane .{z : Re z > −1}, except at the origin. Obviously, this extension is analytic in
the vertical strip .{z : −1 < Re z < 0}. It is continuous at the points of the imaginary
axis, except at the origin. Indeed,

Γ (z + 1) Γ (iy + 1)
. lim Γ (z) = lim = = Γ (iy) for all y ∈ R \ {0}.
z→iy z→iy z iy

Considering that .Γ (n + 1) = n! for all .n ∈ N0 , we get

Γ (z + 1) Γ (1) 1
Γ (z) =
. ∼ = as z → 0. (8.7)
z z z

Thus, Theorem 8.2 says that .Γ is analytic in .{z : Re z > −1} \ {0}, and the
asymptotic relation (8.7) indicates that .Γ has a simple pole at the origin and

.Resz=0 Γ (z) = 1.

Continuing in the same way, we define

Γ (z + 1) Γ (z + 2)
Γ (z) =
. = for all z ∈ {ξ : Re ξ > −2} \ {0, −1},
z z(z + 1)

Γ (z + k + 1)
.Γ (z) = (8.8)
z(z + 1) · . . . · (z + k)

for all .z ∈ {ξ : Re ξ > −k − 1} \ {0, −1, . . . , −k}. It follows from (8.8) that

(−1)k
Γ (z) ∼
. as z → −k. (8.9)
k! (z + k)

Thus, the analytic continuation of the gamma function from the right half-
plane gives us a meromorphic function with simple poles at non-positive integers
194 8 Analytic Continuations

k
.{−k}k∈N0 at which the residue of .Γ is equal to . (−1)
k! , respectively. In addition, its
restriction on the positive real axis coincides with the real-valued gamma function
(5.21), where it is positive, and due to (8.8) it is also positive on the intervals
.(−k − 1, −k), .k ∈ N0 .

Using the uniqueness Theorem 5.12, as in Sect. 5.5, one can prove functional
identities for the gamma function in the complex plane that are valid for real
numbers. For example, it is well known that
π
Γ (x) Γ (1 − x) =
. for all x ∈ (0, 1).
sin π x
Consequently,
π
Γ (z) Γ (1 − z) =
. for all z ∈ C \ Z. (8.10)
sin π z

An immediate consequence of the identity (8.10) is the absence of zeros of the

gamma function. This means that . Γ1 is an entire function that has simple zeros at
non-positive integers.

Example 8.4 Example 5.7 shows that the Riemann zeta function is analytic in the
half-plane .{z : Re z > 1}. From the functional Eq. (5.24) and the fact that . Γ1 is
entire, it follows that
 1
1 t z−1
ζ (z) =
. dt + g(z) , (8.11)
Γ (z) 0 et − 1

where the entire function g is defined in (5.19). Taking into account the power
expansion for the function .z/(ez − 1) around .z = 0, namely

+∞
z 1
. =1− z+ ck zk for all z ∈ B2π (0)
e −1
z 2
k=2

(to find the radius of convergence, Remark 8.2 is used), and the third statement of
Theorem 5.2, we deduce
 1  1 +∞ 
t z−1 1 1
. dt = t z−1 − + ck t k−1 dt
0 et − 1 0 t 2
k=2
+∞
1 1 ck+1
= − +
z − 1 2z z+k
k=1
8.2 Methods of Analytic Continuation: Schwarz’s Reflection Principle 195

According to (8.11), we get the formula

+∞
1 1 1 1 ck+1
ζ (z) =
. · + − + + g(z) (8.12)
Γ (z) z − 1 Γ (z) 2z z+k
k=1

which provides the analytic continuation of .ζ in .C \ {1}. Indeed, despite the fact that
the bracketed expression in (8.12) has a simple pole at each non-positive integer, all
these poles are cancelled by the zeros of . Γ1 based on the formula (8.9); moreover,
.ζ (0) = − , and .ζ has a simple pole at .z = 1 with residue 1. .
1
2

Remark 8.4 By modifying the contour integration, Riemann deduced from the
identity (5.24) the surprisingly mysterious functional equation
πz
ζ (z) = Γ (1 − z) ζ (1 − z) 2z π z−1 sin
. (8.13)
2
that relates values of the zeta function at the points z and .1 − z. In particular, it
follows from (8.13) that the zeta function has a simple zero at the points .{zn =
−2n}n∈N , known as the trivial zeros of .ζ . When .z = 2m, .m ∈ N, the limit of the
product .Γ (1 − z) sin π2z is equal to .(−1)m π/2(2m − 1)! because of (8.9); due to
(8.12) the limit .ζ (1 − z) sin π2z is .− π2 as z tends to zero.
In 1859 Riemann famously conjectured that there are infinitely many non-trivial
zeros of the zeta function and that all of these zeros lie on the line .Re z = 12 . The
Riemann hypothesis has been confirmed by many theoretical and numerical studies,
but still remains unproven. The Clay Mathematical Institute in 2000 announced a
US$1 million prize for the first correct solution of this conjecture.

Theorem 8.3 (Schwarz’s Reflection Principle) Let .Ω be a domain in .C which

is symmetric with respect to the real axis. We set .Ω + = {z ∈ Ω : Imz > 0},
.Ω
− = {z ∈ Ω : Imz < 0} and .J = Ω ∩ R (see Fig. 8.5).

If f is an analytic function in .Ω + , continuous in .Ω + ∪ J and takes real values

on .J, then f has an analytic extension .f to .Ω which determined by the formula

f (z), z ∈ Ω + ∪ J,
(z) =
.f
f (z), z ∈ Ω −.

Proof Take any point .z0 ∈ Ω − . Then

f(z) − f(z0 ) f (z) − f (z0 ) f (z) − f (z0 ) 

. = = −→ f (z0 ) as z → z0 .
z − z0 z − z0 z − z0
196 8 Analytic Continuations

Fig. 8.5 Domain

.Ω = Ω+ ∪ Ω− ∪ J

This means that there is a derivative of .f at .z0 and

d f 
. (z0 ) = f (z0 ).
dz

Hence, .f ∈ A(Ω − ).

Since f is continuous in .Ω + ∩ J and .f (x) ∈ R for all .x ∈ J ,

. lim f(z) = lim f (z) = f (x)

Ω + z→x∈J Ω + z→x∈J

and

. lim f(z) = lim f (z) = f (x) = f (x).

Ω − z→x∈J Ω − z→x∈J

Then .f ∈ A(Ω) according to Theorem 8.2.

Corollary 8.1 Let conditions of Theorem 8.3 be satisfied and, moreover,

• let the function f univalently map the domain .Ω + onto a domain .G+ that is
located in the upper half-plane .{w ∈ C : Im w > 0} and
• the mapping .f : J → I be a bijection, where .I := ∂G+ ∩ R.

Then the function


f (z), z ∈ Ω + ∪ J,
f(z) =
.
f (z), z ∈ Ω −,
8.2 Methods of Analytic Continuation: Schwarz’s Reflection Principle 197

is a univalent and conformal mapping of the domain .Ω onto the domain .G, where
G = G+ ∪ I ∪ G− and .G− is the symmetric image of .G+ with respect to the real
.

axis.

Proof Based on Theorem 8.3, we have that .f ∈ A(Ω). It follows from the corollary
conditions that the mapping .f: Ω → G is a bijection. Then, from Theorem 7.8 (on
sufficient conditions of conformality), we have that .f is conformal in .Ω.

Example 8.5 Find a univalent and conformal

 mapping of the domain
.Ω := C \ [−1, 1] ∪ [−i, i] onto the domain .G := C \ [−1, 1].

Solution The domains .Ω and G are symmetric with respect to the real axis and

.Ω = Ω + ∪ Ω − ∪ J, G = G+ ∪ G− ∪ I,

where .J := Ω ∩ R = (−∞, −1) ∪ (1, +∞), .I := G ∩ R = (−∞, −1) ∪ (1, +∞),

Ω + := {z ∈ C : Im z > 0} \ [0, i],

. Ω − := {z ∈ C : Im z < 0} \ [−i, 0],
G+ := {w ∈ C : Im w > 0}, G− := {w ∈ C : Im w < 0}.

To apply Corollary 8.1, we must find a univalent and conformal mapping of .Ω +

onto .G+ . It is easy to see that this is the function

1 
z ∈ Ω +,
0
f (z) := √ z2 + 1,
.
2
√
where . 0 · is the 0th-branch of the square root (see Sect. 3.6); in addition, it is a
bijection of the interval .(1, +∞) onto .(1, +∞). Since
√ √
1 
0 x 2 + 1 iπ x2 + 1
. lim √ z2 + 1 = √ e =− √ ,
Ω + z→x<−1 2 2 2

f is also a bijection of .(−∞, −1) onto .(−∞, −1).

Thus, all the conditions of Corollary 8.1 are satisfied, which means that the
function
⎧ 
⎪ 1 0 2
⎨ √2 z + 1, z ∈ Ω + ∪ (−∞, −1) ∪ (1, +∞),
(z) =
.f
⎪ 
⎩ √1 0 (z)2 + 1 , z ∈ Ω − ,
2

is the desired mapping.

198 8 Analytic Continuations

8.3 Analytic Continuation Along a Curve: The Monodromy

Theorem

In the previous section we introduced two specific methods of analytic continuation.

Here we will study all the possible analytic extensions of a canonic analytic function
element.
 n
Definition 8.7 A finite set of analytic function elements . (fk , Dk ) k=0 is called
an analytic chain if consecutive pairs are analytic continuations of each other, i.e.,
.(fk−1 , Dk−1 ) and .(fk , Dk ) are direct analytic continuations of each other for all

.k ∈ {1, . . . , n}. Hereby .(fn , Dn ) is referred to as an analytic continuation of the

analytic function element .(f0 , D0 ) along the given chain.

According to Lemma 8.1 the analytic continuation along an analytic chain is

unique.

Example 8.6 Let .(f0 , B0 ), .(f1 , B1 ), .(f2 , B2 ) be analytic function elements, where
B0 := B1 (0), .B1 := B1 (i), .B2 := B1 (−1),
.


f0 (z) =
. |z| exp(i ϕ2 ), z ∈ B0 , ϕ ∈ Arg(z), ϕ ∈ (− π2 , π2 );

f1 (z) = |z| exp(i ϕ2 ), z ∈ B1 , ϕ ∈ Arg(z), ϕ ∈ (0, π );

f2 (z) = |z| exp(i ϕ2 ), z ∈ B2 , ϕ ∈ Arg(z), ϕ ∈ ( π2 , 3π
2 ).

It is easy to verify that .{(f0 , B0 ), (f1 , B1 ), (f2 , B2 )} is an analytic chain, and

(f2 , B2 ) is the analytic continuation of the analytic function element .(f0 , B0 ) along
.

this chain.
Similarly, we check that .{(f0 , B0 ), (f−1 , B−1 ), (f−2 , B−2 )} is also an analytic
chain, where .B−1 := B1 (−i), .B−2 := B1 (−1),

f−1 (z) = |z| exp(i ϕ2 ), z ∈ B−1 , ϕ ∈ Arg(z), ϕ ∈ (−π, 0);
.


f−2 (z) = |z| exp(i ϕ2 ), z ∈ B−2 , ϕ ∈ Arg(z), ϕ ∈ (− 3π
2 , − 2 ).
π

However, .f2 (z) = f−2 (z) for all .z ∈ B2 = B−2 . Thus, the results obtained by
continuing .(f0 , B0 ) along these two chains are different. And this is an illustration of
how repeated continuation of a single-valued analytic function element can produce
a multi-valued function or a single-value function given on the corresponding
Riemann surface (see Sects. 3.6 and 3.7). .

From this example, a natural question arises: under what conditions can it be
guaranteed that the result of an analytic continuation along different analytic chains
will be the same?
8.3 Analytic Continuation Along a Curve: The Monodromy Theorem 199

Before answering this question, let us give a somewhat similar but more
convenient definition of analytic continuation. As in Sect. 4.4, we assume that curves
are given on the segment .I = [0, 1].
 
Definition 8.8 Let . f0 , Br0 (a) be a canonic analytic function element, and let
.γ : I → C be a curve with the origin .a = γ(0) and the endpoint .b = γ(1).
 
We say that . f0 , Br0 (a) is analytically continued along .γ if there exists a
partition .{0 = t
 0 < t1 < . . . < tn = 1} of the segment I and an analytic chain
n
. fk , Brk (γ(tk ))
k=0
such that
   
(1) .γ [tk−1
 , tk ] ⊂ B  rk−1 γ(tk−1 ) for all .k ∈ {1, . . . , n},
(2) and . fn , Brn (b) is a canonic analytic function element.
   
Hereby, . fn , Brn (b) is called an analytic continuation of . f0 , Br0 (a) along the
curve .γ ; this is denoted as follows
  a.c.  
.f0 , Br0 (a) −→ fn , Brn (b) .
γ

 
Exercise 8.2 Prove that the analytic continuation of . f0 , Br0 (a) along .γ does not
depend on a partition of the segment I .

Remark 8.5 The reader may wonder why, in this definition, we need the canonic
analytic function elements at the beginning and the end of the analytic continuation
 
along .γ. There is no loss generality, since an analytic function element . f, Br (a)
can easily be made a canonic one. To do this, we need to use Theorem 5.3 (on the
expansion of an analytic function in a power series) and find the convergence radius
using the formula (8.1). Also, without losing generality, we can replace the domains
in an analytic chain with disks.

Example 8.7 The analytic chain .{(f0 , B0 ), (f1 , B1 ), (f2 , B2 )} from Example 8.6
can be replaced with the analytic continuation of .(f0 , B0 ) along the curve
.γ(t) = e
iπ t , t ∈ [0, 1]. The corresponding analytic chain is as follows
 4
. fk , B1 (γ(tk ))
k=0
, where the partition

{0 = t0 < t1 =
.
1
4 < t2 = 1
2 < t3 = 3
4 < t4 = 1}

and

fk (z) =
. |z| exp(i ϕ2 ), z ∈ B1 (γ(tk )), ϕ ∈ Arg(z), ϕ ∈ (− π2 + πk π
4 , 2 + πk
4 ).

Thus,
a.c.
(f0 , B0 ) −→ (f2 , B2 ).
. (8.14)
γ
200 8 Analytic Continuations

Fig. 8.6 Curvilinear strip .Uε (γ)

Clearly (see Example 8.6),

a.c.
(f0 , B0 ) −→ (f−2 , B−2 ),
. (8.15)
γ1

where .γ1 (t) = e−iπ t , t ∈ [0, 1]. .

 n
Remark 8.6 For the analytical chain . fk , Brk (γ(tk )) k=0 from Definition 8.8, the
following properties hold:

(1) there is a positive number .ε such that .rk ≥ ε for all .k ∈ {0, 1, . . . , n};
(2) for any curve .γ1 : I → C such that .γ1 (0) = a, .γ1 (1) = b and whose trace .Eγ1
belongs to the curvilinear strip

Uε (γ) := {z ∈ C : dist(z; Eγ ) < ε},

. (8.16)
 
(see Fig. 8.6) there exists an analytic continuation of . f0 , Br0 (a) along .γ1 and
  a.c.  
.f0 , Br0 (a) −→ fn , Brn (b) .
γ1

 
The same result of the analytic continuation of . f0 , Br0 (a) along .γ1 is guaran-
teed by Theorem 5.12. .

It turns out that if an analytic function element can be analytically continued

along homotopic curves (see Definition 4.4), then the results of the extension
coincide. The following theorem proves this.

Theorem 8.4 (Monodromy Theorem) Let .ϕ : I ×I → Ω be a homotopy between

two curve

γ0 : I → Ω
. and γ1 : I → Ω,
8.3 Analytic Continuation Along a Curve: The Monodromy Theorem 201

i.e., .γ0 ≈ γ1 in .Ω, and let .a = γ0 (0) = γ1 (0) and .b = γ0 (1) = γ1 (1). For each
s ∈ I , denote by
.

γs (t) := ϕ(s, t),

. t ∈ I.
 
If for every .s ∈ I a given canonic analytic function element . f0 , Br0 (a) is
analytically continued along .γs and
  γs  
.f0 , Br0 (a) −−→ fn(s)
s
, Br (s) (b) ,
a.c. ns

then
 (s)   
. fns , Br (s) (b) = fn(0)
0
, Br (0) (b) for all s ∈ I. (8.17)
ns n0

Proof Thanks to Definition 8.8, for every .s ∈ I there is an analytic chain

   (s) (s)   
. f0 , Br0 (b) , f (s) , Br (s) (γs (t1 )) , . . . , fn(s)
s
, Br (s) (b)
r1 1 ns

 (s) 
which leads to the canonic analytic function element . fns , Br (s) (b) as a result of
  ns
the analytic continuation of . f0 , Br0 (a) along .γs .
From the first point of Remark 8.6 we see that there exists .εs > 0 such that
(s)
.r
k ≥ εs for all .k ∈ {0, 1, . . . , ns }.
Due to the uniform continuity of .ϕ, there is a positive number .δs such that for all
.μ ∈ Υδs := (s − δs , s + δs ) ∩ I

. max |γs (t) − γμ (t)| = max |ϕ(s, t) − ϕ(μ, t)| < εs .

t∈[0,1] t∈[0,1]

This means that the trace .Eγμ belongs to .Uε (γs ) (see (8.16)) for all .μ ∈ Υδs .
By the second point of Remark 8.6,
 (μ)   
. fnμ , Br (μ) (b) = fn(s)
s
, Br (s) (b) for all μ ∈ Υδs . (8.18)
nμ ns

Since .I = [0, 1] is compact, it can be covered by a finite number of intervals

{Υδs0 , . . . , Υδsp }, where .s0 = 0, sp = 1, on each of which equality (8.18) holds for
.

any .μ ∈ Υδk , .k ∈ {1, . . . , p}. Then, starting from .s0 = 0 and taking a finite number
of steps, we arrive at .sp = 1, which means that the identity (8.17) is satisfied.

The following corollary follows from the monodromy theorem and from the fact
that arbitrary curves with a common origin and end are homotopic in a simply
connected domain.

Corollary 8.2 Let .Ω be a simply connected domain and .a ∈ Ω.

202 8 Analytic Continuations

If for any curve .γ starting at a and such that .Eγ ⊂ Ω there

 exists an
 analytic
continuation of a given canonic analytic function element . f, Br (a) along .γ,
then the result of the analytic continuation does not depend on .γ but is uniquely
determined by its end.

Remark 8.7 If the conditions of Corollary 8.2 hold, then a single-valued analytic
function F is determined in .Ω such that the analytic function element .(F, Ω) is a
direct analytic continuation of . f, Br (a) .

Example 8.8 The curves .γ and .γ1 from Example 8.7 are homotopic in .C. However,
the analytic continuations of the canonic analytic function element .(f0 , B0 ) along
these curves yields different results (see (8.14) and (8.15)). The reason is that along
a curve whose trace contains the origin and connects the points 1 and .−1, it is
impossible to analytically continue .(f0 , B0 ) (see Sect. 3.6). This means that the
condition of Theorem 8.4 is not satisfied.

8.4 Global Analytic Functions

Example 8.8 shows that all possible analytic continuations of a canonic analytic
function element can lead to a new object, which is not necessarily a single-valued
function. Such an object is called a global analytic function which appears as a
collection of canonic analytic function elements related to each other in a prescribed
way.

Definition 8.9 A set of canonic analytic function elements

 
F := (fα , Bα ) α∈Ξ
.

is called a global analytic function if the following conditions are met:

(1) for any .(fα1 , Bα1 ) ∈ F and .(fα2 , Bα2 ) ∈ F there is a curve .γ such that

a.c.
.(fα1 , Bα1 ) −→ (fα2 , Bα2 );
γ

(2) if for some canonic analytic function element .(g, B) there exists an element
.(fα3 , Bα3 ) ∈ F and a curve .γ0 such that

a.c.
(fα3 , Bα3 ) −→ (g, B),
.
γ0

then necessarily .(g, B) ∈ F. .

It follows from Definition 8.9 that two global analytic functions .F1 and .F2 are
equal if they have at least one element in common.
8.4 Global Analytic Functions 203

a.c.
Exercise 8.3 Check that the relation ." −→ " between canonic analytic function
γ
elements is an equivalence relation, i.e., it is reflexive, symmetric and transitive. .

Therefore, the set of all canonic analytic function elements is partitioned into
a.c.
equivalence classes by the equivalence relation ." −→ ", and these equivalence
γ
classes are different global analytic functions.

Working with such bulky objects is not convenient. Next, a new definition is
introduced which characterises the concept of a global analytic function in more
detail.  
For each global analytic function .F = (fα , Bα ) α∈Ξ we define the set

 :=
. Bα . (8.19)
α∈Ξ

It is easy to see that for each point .a ∈  there exists a canonic analytic function
element .(f, Br (a)) which belongs to the global analytic function .F. It also means
that .Br (a) ∈ , and therefore . is an open set.
Consider two arbitrary points a and b from .. Then there are canonic analytic
function elements .(f1 , Br1 (a)) and .(f2 , Br2 (b)) from .F. By Definition 8.9, there
exists a curve .γ such that
a.c.
(f1 , Br1 (a)) −→ (f2 , Br2 (b)),
.
γ

and this means (see Definition 8.8) that the trace of .γ belongs to .. Thus, . is a
domain.
If there is an analytic function element .(g, ) which is the direct analytic
continuation of each .(fα , Bα ) ∈ F (see examples below), then the global analytic
function .F bijectively specifies the single-valued analytic function .g :  → C for
which . is its domain of analyticity.
Therefore, we also call the domain . defined by (8.19) a domain of analyticity
of the global analytic function .F.
 
Definition 8.10 Let.F = (fα , Bα ) α∈Ξ be a global analytic function, and D be a
subdomain of . = α∈Ξ Bα .
A single-valued continuous function g defined in D is called a branch of the
global analytic function .F if for every .z0 ∈ D there exists a canonic analytic
function element .(fα1 , Bα1 ) ∈ F such that .z0 ∈ Bα1 and

g(z) = fα1 (z)

. for all z ∈ D ∩ Bα1 .


.
204 8 Analytic Continuations

It follows from this definition that g is an analytic function in D. It is also obvious

that for any index .α ∈ Ξ , the function .fα defined in .Bα is a branch of the global
analytic function .F in the disk .Bα . Depending on the domain ., the following cases
are possible.

1. If . is a simply connected domain, then by Corollary 8.2 a unique branch is

defined in . and the domain . is its domain of analyticity. In this case, the global
analytic function is a single-valued
  analytic function. For example, for the global
analytic function .F = (sin z, C) the domain . = C and .g(z) = sin z, .z ∈ C.
2. If . is multiply connected, then two cases are possible:

• In the domain . one can still pick out a unique branch of the global analytic
function .F, and then . is a domain of analyticity for this branch. For example,
for the global analytic function
 1 
F=
.
z , B|a| (a) a∈C\{0}
,

the domain . = C \ {0} and .g(z) = 1z , .z ∈ C \ {0}.

• A branch of .F cannot be selected in . (see Example 8.9 below).

Example 8.9 Consider the global analytic function

√   
. z = f, B|a| (a) a∈C\{0} , (8.20)

√
where .f (z) = |z| exp(i Arg z
2 ), and .Arg z is a continuous in .B|a| (a) branch of the
multi-valued function .Arg. From Example 8.7 it can be seen that for this global
analytic function it is not possible to select a single-valued branch in the domain
√
. = C \ {0}. For any point .a ∈ C \ {0}, the global analytic function .{ z} has two

different canonic analytic function elements, namely .(f, B|a| (a)) and .(−f, B|a| (a)).

Remark 8.8 The French mathematician Henri Poincaré (1954–1912) and the
Italian mathematician Vito Volterra (1860–1940) independently proved that a global
analytic function can have at most a countable number of different canonic analytic
function elements centred at the same point. .

If it is impossible to select a single-valued branch for a global analytic function

F in its domain . of analyticity, then we cut . by connecting the components of
.

its boundary to obtain a simply connected domain . . Next, we take an arbitrary

canonic analytic function element of .F and, based on Corollary 8.2, select a branch
of .F in .. According to Remark 8.8 there can be at most a countable number of
such branches in . . Thus, instead of a global analytic function .F, one can consider
the set of all its branches in the domain ..
8.4 Global Analytic Functions 205

√
Example 8.10 Consider the global analytic function .{ z} (see Example 8.9). For
this function, . = C \ {0}. If we connect the points 0 and .∞, the simplest and most
convenient way to do this is to use a real positive semi-axis, then we get a simply
connected domain


 := C \ {z : Im z = 0, Re z ≥ 0}.
.

In .
 it is possible to select two different branches for the global analytic function
√
.{ z}, namely

 arg z
gk (z) =
. |z| ei( 2 +π k) , k ∈ {0, 1};
√
here .0 < arg z < 2π . For every canonic analytic function element of .{ z}, there
exists a branch that  is its direct
 analytic continuation
√ onto some larger domain.
For example, for . f, B1 (1) , where .f (z) = |z| exp(i ϕ2 ) and .ϕ ∈ Arg(z), .ϕ ∈
(− π2 , π2 ), the direct analytic continuation onto the upper plane .D0 := {z : Im z > 0}
is the analytic function element .(g0 , D0 ).
√
Therefore, the specification of the global analytic function .{ z} is equivalent to
the specification of its two branches: .g0 and .g1 in . (see Sect. 3.6).

Example 8.11 Consider the following global analytic function

 
{Log} =
. f, B|a| (a) a∈C\{0} ,

where .f (z) = log |z| + iArg z, and .Arg z is any continuous in .B|a| (a) branch of the
multi-valued function .Arg. For this function, . = C \ {0}, and there it is impossible
to select a single-valued branch of .{Log}. However, by making a cut along the real
negative semi-axis, we obtain the simply connected domain .  = C \ {z : Im z =
0, Re z ≤ 0}, in which it is possible to pick out the following branches:

gk (z) = log |z| + i arg z + i2π k,

. k ∈ Z; (8.21)

here .−π < arg z < π (see Sect. 3.7). For every canonic analytic function element
of {Log}, there exists a branch which is the  direct analytic
 continuation of it on a
somewhat larger domain. For example, for . f, B1 (−1) , where .f (z) = log |z| + iϕ
and .ϕ ∈ Arg(z), .ϕ ∈ ( π2 , 3π
2 ), the direct analytic continuation onto the upper plane
.D0 is the analytic function element .(g0 , D0 ).

Therefore, specifying the global analytic function .{Log} is equivalent to specify-

ing its branches (8.21) in .
 (see Sect. 3.7). .
206 8 Analytic Continuations

The following operations can be performed on global analytic functions:

(1) multiplication by an arbitrary complex number or by an arbitrary single-valued

analytic function (of course, it must be defined in the corresponding domain .);
for example:
 
. sin z · F = sin z · fα , Bα α∈Ξ
,
 
where .F = (fα , Bα ) α∈Ξ is a global analytic function;
(2) superposition with any single-valued analytic function; for example:
   
. exp F = exp(fα ), Bα α∈Ξ ;

(3) when integrating (differentiating) a global analytic function, one selects its
branches in the corresponding domain . , and then integrates (differentiates)
over the required branch.

But, it is impossible to correctly determine the addition or multiplication of

√ √
global analytic functions. For example, if we assume that the sum .{ z} + { z}
(see (8.20)) is uniquely determined as the sum of all possible canonic analytic
function elements of these functions, then, on the one hand, the sum of the elements
.(f, B|a| (a)) and .(−f, B|a| (a)) defines a function that is identically equal to zero;

on the other hand, the sum of the elements .(f, B|a| (a)) and .(f, B|a| (a)) defines the
√
global analytic function .{2 z}.

8.5 Riemann Surfaces of Global Analytic Functions

In Sects. 3.6 and 3.7 we showed√ how to construct Riemann surfaces of the multi-
valued functions .Log and . n · in a simple way. The aim of such constructions is
to make a single-valued function out of a multi-valued one. The purpose of this
section is to present a unified topological approach to the construction of Riemann
surfaces for global analytic functions, in which the above examples are special cases
and which shows that Riemann surfaces can be considered as mathematical objects
worthy of independent study.
In what follows, we will use the abbreviation .fa for a given canonic analytic
function element .(f, Br (a)) indicating the function f itself and, as an index, the
point a (the center of this element).
Next, we construct a set .R, in which elements (points) are ordered pairs .A :=
(a, fa ), where a is an arbitrary point from .C, and .fa is an arbitrary canonic analytic
function element centered at a.
8.5 Riemann Surfaces of Global Analytic Functions 207

.τ generated by a set of .ε-neighbourhoods of

On .R, we introduce a topology 1
 
each point .A = a, (f, Br (a)) ∈ R, where .ε is any positive number from .(0, r).
An .ε-neighbourhood of a point .A ∈ R is defined as follows

Uε (A) := B = (b, gb ) ∈ R : |a − b| < ε
.

and gb is a direct analytic continuation of fa . (8.22)

It is easy to verify that

• for every .A ∈ R, there is an .ε-neighbourhood such that .A ∈ Uε (A);

• if .C = (c, hc ) ∈ Uε1 (A) ∩ Uε2 (B), then there is .Uε3 (C) such that

C ∈ Uε3 (C) ⊂ Uε1 (A) ∩ Uε2 (B).

.

To do this, we need to take .ε3 such that the disk .Bε3 (c) ⊂ Bε1 (a) ∩ Bε2 (b).

This means that the set of .ε-neighbourhoods of each point .A ∈ R is a basis of a

topology, and a subset in .R is defined to be open if it is a union of .ε-neighbourhoods.
Thus, the collection .τ of all open sets, including .∅ by definition, is a topology on .R,
and .(R, τ) is a topological space.

Lemma 8.2 The space .(R, τ) is a Hausdorff space (for each pair of distinct points,
there are their non-intersecting neighborhoods), i.e., for all .A = B ∈ R there exists
a number .ε > 0 such that .Uε (A) ∩ Uε (B) = ∅.

Proof Consider two distinct points .A = (a, fa ) and .B = (b, gb ) lying in .R. There
are two possible cases. The first one is .a = b. In this case, we set .ε = |a−b| 4 .
Then, based on the definition  of the .ε-neighbourhood of a point in the space .R (see
(8.22)), it is clear that .Uε (A) Uε (B) = ∅.
In the second case, we have .a = b and .fa = ga , i.e.,

(f, Br1 (a)) = (g, Br2 (a)).

.

The last relation means that .f =  g at the intersection .Br1 (a) ∩ Br2 (a). Taking
ε < min{r1 , r2 }, we get that .Uε (A) Uε (B) = ∅.
.

Recall that a topological space Y is connected if there are no nonempty open sets
U, V ⊂ Y such that
.

.U ∩V =∅ and U ∪ V = Y.

1A topology .τ on a set Y is a collection of subsets of Y such that .∅ and Y lie in .τ, the finite
intersection of subsets of .τ is in .τ, and the union of arbitrarily many subsets of .τ is also in .τ. An
element of .τ is called an open set, and Y with .τ is called a topological space.
208 8 Analytic Continuations

Lemma 8.3 The topological space .(R, τ) is not connected.

  
Proof Consider the set .D = a, (1, C) : a ∈ C . It is an open set, since every
point of D lies in D together with its .ε-neighbourhood
 
Uε a, (1, C) ⊂ D,
.

where .ε > 0.
Now let us show that D is a closed set. Let .C = (c, hc ) be a limit point of .D,
i.e., for every .ε > 0 there is a point .(b, (1, C)) which belongs to .Uε (C). This means
that .|c − b| < ε and the canonic analytic function elements .(h, Br (c)) and .(1, C)
are direct analytic continuations of each other. This implies that .h ≡ 1 in .C. Thus,
.C ∈ D.

Therefore, .R can be represented as a union of two open sets D and .R \ D, which

do not intersect.
 

Let .F = (fα )aα α∈Ξ be a global analytic function, where

 

(f
. α )aα := fα , Brα (aα ) .

The set
  
. 
RF := Aα = aα , (fα )aα α∈Ξ

is said to be a Riemann surface of .F.

Lemma 8.4 The set .RF is a domain in the topological space .(R, τ).
 

Proof Let .F = (fα )aα α∈Ξ be a global analytic function. Consider any point

 
. 
Aα = aα , (fα )a α ∈ R F .

Then .Uε (Aα ) ⊂ RF , where .ε ∈ (0, rα ). Indeed, if a point

B = (b, gb ) ∈ Uε (Aα ),

.

then .|b − aα | < ε and the canonic analytic function elements .gb and .(f  α )aα are
direct analytic continuations of each other. According to Definition 8.9, the element
b ∈ F, and therefore .B ∈ RF . Thus, .Uε (Aα ) ⊂ RF , i.e., .RF is an open set.
.g

Now let us show that .RF is path-connected. Consider two arbitrary points
   
. 
Aα1 = aα1 , (fα1 )a α ∈ R F and 
Aα2 = aα2 , (fα2 )aα ∈ RF .
1 2
8.5 Riemann Surfaces of Global Analytic Functions 209


Since .(f 
α1 )aα ∈ F and .(f α2 )aα ∈ F, based on Definition 8.9, there exists a curve
1 2
.γ : I → C such that .γ(0) = aα1 , .γ(1) = aα2 and


(f
.
a.c.
α1 )aα −→ (f
α2 )a α .
1 γ 2

This means (see Definition 8.8) that there exists an analytic chain

. 
(f 
α1 )aα , (f 
1 )γ(t1 ) , . . . , (f 
n−1 )γ(tn−1 ) , (fα2 )a α
1 2

 
in which each analytic function element .(f  k )γ(tk ) = fk , Brk (γ(tk )) can be
considered canonic.
According to the second part of Definition 8.9, the point
 
. 
γ(tk ), (fk )γ(tk )

belongs to .RF , where .k ∈ {1, . . . , n − 1}. In addition, for every .t ∈ I = [0, 1] there
is a number .k ∈ {1, . . . , n} and a unique canonic analytic function element .(f )γ(t)
 
such that .t ∈ [tk−1 , tk ] and the elements .(f )γ(t) and .(fk )γ(tk ) are direct analytic
continuations of each other. Then, thanks to Definitions 8.9 and 8.11, we have
 
. )γ(t) ∈ RF
γ(t), (f for all t ∈ [0, 1]. (8.23)

Since .γ : I → C is a curve, based on (8.23) and the definition of an .ε-neighbourhood

of a point .A ∈ R, we can state that the function
 
. )γ(t) ,
Γ (t) := γ(t), (f t ∈ [0, 1],

is continuous as a function acting from the segment I into .RF . This means that .Γ
is a curve whose trace lies in .RF with the starting point .Γ (0) = Aα1 and ending
point .Γ (1) = Aα2 .

The definition of a Riemann surface given before Lemma 8.4 was slightly
simplified in the notation in order to make it easier to understand the proof of the
lemma and not to clutter it up with additional notation. By Remark 8.8, in general,
a global analytic function can have a countable number of canonic analytic function
elements centered at the same point, i.e.
 

(f
. k,α )aα = fk,α , Br(α,k) (aα ) , k ∈ N;

if the number of such elements is finite, we show how the index k changes. So, the
next definition is proposed.
210 8 Analytic Continuations

Definition 8.11 Let

 
. 
F = (fk,α )aα α∈Ξ, k∈N

be a global analytic function. The set

 !
 
RF := A(k)
. α = aα , fk,α a α
α∈Ξ k∈N

is called a Riemann surface of .F.

Important
Summarizing the results proved in this section and recalling Definitions 8.9
and 8.11, we can make the following conclusions:

(1) for every global analytic function .F one can uniquely determine its
Riemann surface .RF , which is a domain in the topological space .(R, τ);
(2) if .F1 = F2 , then .RF1 ∩ RF2 = ∅, i.e., the Riemann surfaces of two
different global analytic functions cannot intersect;
(3) each domain in the topological space .(R, τ) corresponds to a unique
global analytic function;
(4) every global analytic function .F can be associated with a single-valued
function on its Riemann surface, namely the function .F : RF → C that
maps each point
 
A(k)
.α = aα , fk,α a ∈ RF
α

to the complex number .fk,α (aα ).

Example 8.12 The global analytic function from Example 8.11 can be rewritten as
follows
 
{Log} =
. fk , B|a| (a) k∈Z, a∈C\{0} ,

where .fk (z) = log |z| + i arg z + i2π k, and we assume that either .arg z ∈ (−π, π ]
or .arg z ∈ [0, 2π ] (it depends on a disk .B|a| (a)).
Its Riemann surface (see Definition 8.11) is the set
  
RLog =
. a, fk , B|a| (a) k∈Z, a∈C\{0} ;
8.6 Singularities of Global Analytic Functions 211

     
and if .k = m, then . a, fk , B|a| (a) and . a, fm , B|a| (a) are two distinct points
of the Riemann surface .RLog .
The associated function .Log : RLog → C is single-valued because for each fixed
.k ∈ Z and .a ∈ C \ {0} the point

  
.a, fk , B|a| (a) ∈ RLog

is uniquely mapped to the complex number .fk (a) (see also Sect. 3.7).

Exercise 8.4 For the global analytic function from Example 8.10, write down its
Riemann surface and the single-valued function associated with it.

8.6 Singularities of Global Analytic Functions

√
As shown in Sects. 3.6 and 3.7, the multi-valued functions . n · and .Log have special
isolated singularities called branch points (now we can talk about points around
which a multiple-valued function has nontrivial monodromy). So in this section, we
give a classification of singularities of global analytic functions using the abstract
approach introduced in the previous sections.
  
Definition 8.12 Let .F = fk,α a α∈Ξ, k∈N be a global analytic function.
α
A point .b ∈ C is called an isolated singular point of .F if there exists a positive
number r such that, for any point .aα ∈ B̆r (b), every canonic analytic function

element .(fk,α )aα is analytically continued along every curve .γ that starts at the point
.aα and whose trace belongs to .B̆r (b). .

Let us prove a lemma which allows to classify singularities of global analytic

functions.
  
Lemma 8.5 Let .F =  fk,α a α∈Ξ, k∈N be a global analytic function, let a point
α
b be an isolated singular point of .F, and let the following conditions be satisfied:


(1) there exists a canonic analytic function element .(fk,α1 )aα ∈ F, where the point
1
. aα1 ∈ B̆r (b), and a closed curve .γ0 , whose trace .Eγ0 ⊂ B̆r (b) and .γ0 (0) =
γ0 (1) = aα1 (see Fig. 8.7), such that


(f
.
a.c.
k,α1 )aα −
−→ (f
k,α1 )aα ;
1 γ0 1
212 8 Analytic Continuations

Fig. 8.7 Isolated singular

point b of .F


(2) there exists a canonic analytic function element .(fm,α2 )aα ∈ F, where the
2
point .aα2 ∈ B̆r (b), and a curve .μ, whose trace .Eμ ⊂ B̆r (b) and .μ(0) = aα1
and .μ(1) = aα2 , such that

. 
(f
a.c.
k,α1 )aα −
−→ (f
m,α2 )aα .
1 μ 2

Then for any closed curve .γ for which .γ(0) = γ(1) = aα2 and which is homotopic
to the curve .γ0 in .B̆r (b) .(γ ≈ γ0 ), we have


(f
.
a.c.
m,α2 )aα −
−→ (f
m,α2 )aα . (8.24)
2 γ 2

Proof Consider the curve  .γ := μ

−1 ∪ γ ∪ μ (see (4.3)). It is easy to see (Fig. 8.7)
0
that 
.γ ≈ γ0 in .B̆r (b) and

. 
(f
a.c.
m,α2 )aα −→ (f
m,α2 )aα .
2 
γ 2

Since by the condition .γ ≈ γ0 in .B̆r (b), the curve .γ ≈  γ in .B̆r (b).


According to Definition 8.12, the canonic analytic function element .(fm,α2 )aα is
2
analytically continued along any curve whose trace belongs to .B̆r (b). Therefore, by
Theorem 8.4, we have (8.24).

Remark 8.9 If a closed curve .γ ≈ 0 in .B̆r (b) and it starts at .aα3 ∈ B̆r (b), then the
analytic continuation of any canonic global function element .(f k,α3 ) ∈ F along
aα3
γ results the same function element. In fact, the curve .γ is continuously deformed
.
8.6 Singularities of Global Analytic Functions 213

into a curve belonging to the disk of this canonic global function element, and the
continuation along such a curve obviously does not change it. Next we have to apply
Lemma 8.5.

Definition 8.13 Let .F be a global analytic function and let a point b be its isolated
singularity.
The point .b is called an isolated single-valued singular point of .F if there exists a
Jordan curve .γ whose trace .Eγ belongs to .B̆r (b) and the point b lies in the interior of
.γ so that the analytic continuation along .γ of any canonic analytic function element

of .F centered at .γ(0) does not change it. .

Remark 8.10 If b is a single-valued singular point of a global analytic function

.F, then, based on Lemma 8.5, the analytic continuation along any closed curve
.γ, whose trace belongs to .B̆r (b), of any canonic analytic function element of .F

centered at a point in .B̆r (b) will result in the same element. Thus, single-valued
branches of .F are distinguished in .B̆r (b), and the point b for them can be either a
pole, or removable, or essential (Definition 6.4).
  
Definition 8.14 Let .F =  fk,α a α∈Ξ, k∈N be a global analytic function and let a
α
point b be its isolated singularity.
The point .b is called a branch point of .F if there exists a Jordan curve .γ,
.Eγ ⊂ B̆r (b) and .b ∈ int(γ), and two different canonic analytic function elements

.(f 
k )γ(0) ∈ F and .(f m )γ(0) ∈ F such that


(f
.
a.c.
k )γ(0) −→ (f
m )γ(0) .
γ

Moreover,

(1) if there exists an integer .n ≥ 2 (the smallest selected) such that


(f
.
a.c.
k )γ(0) −→ 
(fk )γ(0) ,
nγ

where .γ n := γ ∪ . . . ∪ γ, the point .b is said to be a branch point of finite order,

" #$ %
n
namely of order .n − 1;
(2) if such a number does not exist, then the point b is called a logarithmic branch
point (or branch point of infinite order). .

Remark 8.11 Thus, a branch point of a global analytic function is its isolated
singular point such that the analytic continuation of a global analytic function
element of that function along a Jordan curve enclosing this point leads to a new
function element. Branch points are divided into two categories: If, after the analytic
continuation along the specified Jordan curve (based on Lemma 8.5 it suffices to
214 8 Analytic Continuations

take one such curve) n times, we obtain the original function element again, then
this point is a branch point of finite order; if this does not happen, then the point is
a logarithmic branch point (branch point of infinite order).

Example
 8.13 Find isolated singular points of the global analytic function
√
{ 1 + z} and characterize them.
.

Solution It known that the point 0 is a first-order branch point for the global
√ √
analytic function .{ z} (see Sect. 3.6) and two branches of .{ z} are selected (see
√ √ arg z
Example 8.10). Let us denote . 0 z := |z| ei 2 the zeroth branch, where .arg z ∈
(−π, π ]. Then in the disk .B 1 ( 12 ), according to Corollary 8.2, four branches can be
2  √
selected for the global analytic function .{ 1 + z}, namely
& &
√ √
f0 (z) := 1 + 0 z, f1 (z) := 1 − 0 z,
0 0
.

& &
√ √
f2 (z) := − 1 − 0 z, f3 (z) := − 1 + 0 z.
0 0

It is clear that the points .0, 1 and .∞ are suspected to be branching points.
Let .γ1 (t) = 12 exp(it), t ∈ [0, 2π ]. Then the values of the outer roots in these
branches are not changed by analytic continuations along .γ1 . They only change the
values of the inner root. Therefore,

.
(f
a.c.
0 ) 1 −→ (f a.c.
1 ) 1 −→ (f0) 1 , (8.25)
2 γ1 2 γ1 2

and

. 
(f
a.c.
2 ) 1 −→ (f a.c.
3 ) 1 −→ (f2) 1 . (8.26)
2 γ1 2 γ1 2

By Definition
 8.14, the relations (8.25) and (8.25) mean that the global analytic
√
function .{ 1 + z} has two first-order branch points at the origin. This is shown
schematically in Fig. 8.8.

 √
Fig. 8.8 Diagram of branch points of the global analytic function .{ 1 + z}
8.6 Singularities of Global Analytic Functions 215

Now consider the Jordan curve .γ2 (t) = 1 − 12 exp(it), .t ∈ [0, 2π ]. Then
the analytic continuation along .γ2 doesn’t change the canonic analytic function

elements .(f 
0 ) 1 and .(f3 ) 1 . But,
2 2


(f
.
a.c.
1 ) 1 −→ (f a.c.
2 ) 1 −→ (f1) 1 .
2 γ2 2 γ2 2

 √
Thus, at the point 1, the global analytic function .{ 1 + z} has only one first-
order branch point (Fig. 8.8). For the branches .f0 and .f3 the point 1 is a removable
singular point.
√ To study the point .∞, consider the curve .γ3 (t) = 9 exp(it), .t ∈ [0, 2π ]. Then
. γ3 (t) = 3 exp(i ), t ∈ [0, 2π ]. For each value .k ∈ {0, 1, 2, 3} we consider the
0 t
2
restriction .fk (γ3 (t)), t ∈ [0, 2π ], and find its values at the points .t = 0 and .t = 2π .
Considering these values, we conclude that


(f
.
a.c.
0 )9 −→ (f a.c.
1 )9 −→ (f a.c.
3 )9 −→ (f a.c.
2 )9 −→ (f0 )9 .
γ3 γ3 γ3 γ3

Thus,
 the point .∞ is a branch point of order 3 for the global analytic function
√
{ 1 + z} (see Fig. 8.8).
.

 
Exercise 8.5 Let .f ∈ A B̆r (b) and b is a simple zero or pole of f .
Prove that
√
• b is a branch point of order .n − 1 of the global analytic function .{ n f };
• b is a branch point of infinite order of the global analytic function .{Log(f )}.

Exercise 8.6 Find isolated singular points of the global analytic function
√ √
{ z z − 1} and characterize them.
.
Qualitative Properties of Analytic Functions
9

Abstract

Inspired by the properties of analytic functions proved in the previous sections, in

the last section we are ready to explore new, no less amazing properties of such
functions. In Sect. 9.1 we show that analyticity is sufficient for a nonconstant
function being an open map. This property indicates that the modulus of a
non-constant analytic function cannot have a strict local maximum. A direct
application of the maximum modulus principle is Schwarz’s Lemma, established
by the German mathematician K. A. Schwarz (1943–1921) in 1869, which is
important in the theory of bounded analytic functions, where it is fundamental
to most estimates. Sect. 9.2 shows how methods of complex analysis can be
used to efficiently find inverse functions and expand them into Lagrange series
(for single-valued inverse functions) and Puiseux series (for multi-valued inverse
functions). Sections 9.3 and 9.4 are a preparation for the proof of Riemann’s
theorem, namely here we are interested in the conformal classification of
domains of the complex plane and the finding of a sufficient condition for the
precompactness of a family of analytic functions (Montel’s theorem). In the last
section there is a proof of the Riemann mapping theorem, which is undoubtedly
one of the most beautiful theorems in mathematics.

9.1 Open Mapping Theorem, Maximum Modulus Principle,

Schwarz Lemma

In real analysis, there are many examples of differential functions that are not open
mappings, for example, the function .f (x) = x 2 maps the open interval .(−2, 2) onto
the half-open interval .[0, 4). The following open mapping theorem once again points
to the essential difference between the properties of analytic functions in complex
analysis and smooth functions in real analysis.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 217
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1_9
218 9 Qualitative Properties of Analytic Functions

Theorem 9.1 (Open Mapping Theorem) Let f be a non-constant analytic func-

tion in a domain .Ω. Then the image of .Ω under the mapping f is also a domain
in .C.

Proof Denote by .Ω ∗ := f (Ω). Clearly, .Ω ∗ is path-connected as a continuous

image of a connected set. Indeed, for any two distinct points .w1 , w2 from .Ω ∗ there
are two points .z1 = z2 from .Ω such that

f (z1 ) = w1
. and f (z2 ) = w2 .

Since .Ω is a domain, there is a curve .z = γ(t), t ∈ [0, 1], such that .γ(0) = z1 ,
γ(1) = z2 , and its trace .Eγ ⊂ Ω. Then .w = f γ(t) , t ∈ [0, 1], is a curve that
.

connects .w1 and .w2 , whose trace belongs to .Ω ∗ .

Let us show that .Ω ∗ is open. For any point .w0 ∈ Ω ∗ there is a point .z0 ∈ Ω
such that .f (z0 ) = w0 . It follows from the theorem conditions that the .w0 -point of
the function f is isolated, i.e., there is an .r > 0 such that

Br (z0 ) ⊂ Ω
. and f (z) = w0 for all z ∈ Br (z0 ) \ {z0 }. (9.1)

If this is not the case, then, by Theorem 5.12, the function .f ≡ w0 in .Ω.
Denote by

μ :=
. min |f (z) − w0 | > 0. (9.2)
z∈∂Br (z0 )

Since .|f (z) − w0 | is a positive (see (9.1)) and continuous function on .∂Br (z0 ), the
extreme value theorem guarantees the existence of its positive minimum.
Consider any .w1 in the disk .Bμ (w0 ) and two functions

F (z) := f (z) − w0
. and G(z) := w0 − w1 in Br (z0 ).

It is easy to see that .F, G ∈ A(Br (z0 )) and

|F (z)| = |f (z) − w0 | ≥ μ > |w1 − w0 | = |G(z)| for all z ∈ ∂Br (z0 ).

.

Then by Rouché’s Theorem 7.7, these functions have the same number of zeros
(counted with multiplicity), i.e., .ZF = ZF +G in .Br (z0 ) or

Zf (z)−w1 = Zf (z)−w0 > 0

. in Br (z0 ), (9.3)

because .Zf (z)−w0 > 0 and .F (z) + G(z) = f (z) − w1 .

This means that there is a point .z1 ∈ Br (z0 ) such that .f (z1 ) = w1 . Hence, .w1 ∈
Ω ∗ . But .w1 was arbitrary from .Bμ (w0 ). Therefore, the disk .Bμ (w0 ) is contained
in .Ω ∗ . As .w0 was also any point in .Ω ∗ , Definition 1.16 says that .Ω ∗ is an open
set. 
9.1 Open Mapping Theorem, Maximum Modulus Principle, Schwarz Lemma 219

We can now gracefully prove a very interesting property of analytic functions,

namely that the modulus of an analytic non-constant function f cannot have a strict
local maximum within the domain of f .

Theorem 9.2 (Maximum Modulus Principle) Let f be an analytic function in a

domain .Ω and let there be a disk .BR (a) in .Ω such that

. maxz∈BR (a) |f (z)| = |f (a)|. (9.4)

Then .f ≡ const in .Ω.

Proof Let .w0 := f (a) and assume that .f = const in .Ω. Since in this case the
w0 -point of the function f is isolated, there exists a number .r ∈ (0, R) such that
.

.f (z) = w0 for all z ∈ Br (a) \ {a}.

Then by the previous theorem .f (Ω) =: Ω ∗ is a domain in .C and the disk .Bμ (w0 )
belongs to .Ω ∗ , where .μ is determined by the formula (9.2).
Obviously, there is a point .w1 ∈ Bμ (w0 ) such that . |w0 | < |w1 |. And this means
that there exists a point .z1 ∈ Br (a) ⊂ BR (a) such that .f (z1 ) = w1 . But this
contradicts to (9.4) because

|f (a)| = |w0 | < |w1 | = |f (z1 )|.

.

Thus, f is constant throughout .Ω. 

Corollary 9.1 Let .f ∈ A(Ω) ∩ C(Ω) and .Ω be a bounded domain in .C. Then

. maxz∈Ω |f (z)| = maxz∈∂Ω |f (z)|.

Proof If .f ≡ const, then the statement is obvious. Therefore, assume that .f ≡

const. Since .|f | ∈ C(Ω), the extreme value theorem assures the existence of .z0 ∈ Ω
such that

|f (z0 )| = maxz∈Ω |f (z)|.

.

However, if .z0 ∈ Ω, then by Theorem 9.2 the function f is constant in .Ω, which
contradicts the assumption. Therefore, .z0 ∈ ∂Ω. 

Example 9.1 It is easy to see that the function .f (z) = z, z ∈ B1 (0), is analytic
and non-constant in the closed disk .B1 (0) and

. minz∈B1 (0) |f (z)| = |f (0)|.

220 9 Qualitative Properties of Analytic Functions

This example shows that both the assertion of Theorem 9.2 and the assertion of
Corollary 9.1 fail for the minimum modulus of an analytic function. However, the
following statements hold.

Exercise 9.1 If a functionf is analytic, non-constant, and non-zero in a domain .Ω,

then its modulus cannot have a strict local minimum in .Ω.

Exercise 9.2 Let .f ∈ A(Ω) ∩ C(Ω), .Ω be an bounded domain and .f (z) = 0 for
all .z ∈ Ω. Prove that

. minz∈Ω |f (z)| = minz∈∂Ω |f (z)|.

Exercise 9.3 Let .f ∈ A(Ω), .f = u + iv, and

 
. sup u(x, y) = u(x1 , y1 ) or inf u(x, y) = u(x1 , y1 ) .
(x,y)∈Ω (x,y)∈Ω

Prove that .u ≡ const in the domain .Ω.

A direct application of the maximum modulus principle is Schwarz’s lemma

which helps to prove the Riemann mapping theorem (see Sect. 9.5).

Lemma 9.1 (Schwarz’s Lemma) Let the following conditions be satisfied:

1. a function f is analytic in the unit disk .B1 (0);

2. .|f (z)| ≤ 1 for all . z ∈ B1 (0);
3. .f (0) = 0.

Then

.|f (z)| ≤ |z| for all z ∈ B1 (0), and |f (0)| ≤ 1.

Moreover, if there is a point .z1 ∈ B1 (0) \ {0} such that .|f (z1 )| = |z1 |, then f is a
rotation, i.e., .f (z) = eiα z, where .α is some real number. The same statement holds
if .|f (0)| = 1.

Proof Consider the function .ϕ(z) = f (z)z , z ∈ B1 (0) \ {0}. By virtue of the first
and third conditions, the point 0 is removable for .ϕ and

. lim ϕ(z) = f (0).

z→0

By Corollary 6.2, .ϕ can be extended by continuity at .0, i.e., .ϕ(0) = f (0), and as a
result, .ϕ ∈ A B1 (0) .
9.2 Inverse Function Theorem: Puiseux Series 221

Fix any point .z0 ∈ B1 (0). Applying the maximum modulus principle to the
function .ϕ in the disk .Br (0), where .r ∈ (|z0 |, 1), we get

|f (z)| 1
. max |ϕ(z)| = max |ϕ(z)| = max ≤ ,
z∈Br (0) z∈∂Br (0) z∈∂Br (0) r r

where the second condition is used. By letting .r → 1, we find that .|ϕ(z0 )| ≤ 1 for
all .z0 ∈ B1 (0). Thus, .|f (z0 )| ≤ |z0 | for all .z0 ∈ B1 (0), and .|f (0)| = |ϕ(0)| ≤ 1.
We now prove the second assertion of this lemma. Let there be a point .z1 ∈
B1 (0) \ {0} such that .|f (z1 )| = |z1 |. This means that .|ϕ(z1 )| = 1, i.e., the modulus
of .ϕ has a local maximum within the disk .B1 (0), and therefore by Theorem 9.2

ϕ ≡ const
. in B1 (0), and |ϕ| ≡ 1.

Consequently, .ϕ(z) = eiα for some .α ∈ R, or .f (z) = eiα z. The same reasoning
can be applied to the case .|f (0)| = 1. 

Exercise 9.4 Instead of the third condition of the Schwarz lemma, let the following
condition be satisfied:

f (0) = f (0) = . . . = f (n−1) (0) = 0.

.

Then prove that . |f (z)| ≤ |z|n for all .z ∈ B1 (0).

Exercise 9.5 (Schwarz–Pick Theorem) Let f be an analytic function in .B1 (0)

1 − |b|2
and .|f (z)| ≤ 1 for all .z ∈ B1 (0). Prove that . |f (a)| ≤ for all . a ∈ B1 (0),
1 − |a|2
where .b = f (a).

9.2 Inverse Function Theorem: Puiseux Series

Assume that a function .f (z) = u(x, y) + iv(x, y) is analytic in a domain .Ω and

.f (z0 ) = 0, where .z0 = x0 + iy0 ∈ Ω. By Theorem 9.1, the image .Ω ∗ := f (Ω)
is
 ua domain in .C. In addition, based on (2.15), the Jacobian of the vector mapping
v : R → R at the point .(x0 , y0 ) is equal to
.
2 2

 u 

J
.  = |f (z0 )|2 = 0.
v (x0 ,y0 )

It is well known from multivariable calculus that this relation is a sufficient condition
for the existence of an inverse vector function in some neighborhood of the point
.w0 = f (z0 ). However, only methods of complex analysis make it possible to
222 9 Qualitative Properties of Analytic Functions

effectively determine this neighborhood and find the inverse function. We will now
show how this can be done.
By the same considerations as in the proof of Theorem 9.1, we obtain that there
exists a number .r > 0 such that .Br (z0 ) ⊂ Ω and

f (z) = w0
. ∀ z ∈ Br (z0 ) \ {z0 }, and f (z) = 0 ∀ z ∈ Br (z0 ). (9.5)

Take any .w1 ∈ Bμ (w0 ), where the radius .μ is determined by the formula (9.2).
According to (9.3) and (9.5),

Zf −w1 = Zf −w0 = 1
. in Br (z0 ), (9.6)

Thus, there exists a unique point .z1 ∈ Br (z0 ) such that .f (z1 ) = w1 , i.e., the inverse
mapping .f −1 : Bμ (w0 ) → Br (z0 ) is defined, for which, due to (2.16) and the right
relation in (9.5), we have

  1 
. f −1 (w) =   for all w ∈ Bμ (w0 ).
f (z) z=f −1 (w)
 
Since .f −1 ∈ A Bμ (w0 ) , it can be represented as the sum of the power series

+∞

f −1 (w) =
. dn (w − w0 )n for all w ∈ Bμ (w0 ). (9.7)
n=0

The series (9.7) is called Bürmann-Lagrange series.

To find the coefficients .dn in (9.7), we fix a point .w ∈ Bμ (w0 ) and define the
function

ξ f (ξ )
h(ξ ) =
. , ξ ∈ Br (z0 ).
f (ξ ) − w
 
It is easy to see that .h ∈ A Br (z0 ) \ {f −1 (w)} , and the point .z = f −1 (w) is
a simple pole of the function .h, because .f (z) = 0 (see (9.5)). Then applying
Cauchy’s residue theorem, we get

1 ξ f (ξ )
. h(ξ ) dξ = Resξ =z h(ξ ) = lim (ξ − z) = z
2π i ξ →z f (ξ ) − w
∂ + Br (z0 )

or

1 ξ f (ξ )
f −1 (w) =
.   w−w0  dξ. (9.8)
2π i f (ξ ) − w0 1 − f (ξ )−w0
∂ + Br (z0 )
9.2 Inverse Function Theorem: Puiseux Series 223

Since
|w − w0 | |w − w0 |
. ≤ =: q < 1 for all ξ ∈ ∂Br (z0 ),
|f (ξ ) − w0 | μ

we obtain from (9.8) by the infinite geometric progression formula that

+∞

f −1 (w) =
. dn (w − w0 )n ,
n=0

where

1 ξ f (ξ )
dn =
.  n+1 dξ, n ≥ 0. (9.9)
2π i ∂ + Br (z0 ) f (ξ ) − w0

Clearly, .d0 = f −1 (w0 ) = z0 . For .n ≥ 1 we integrate by parts in (9.9) and use

the Cauchy residue theorem, namely

1 1
dn = −
. ξd  n
2π i n ∂ + Br (z0 ) f (ξ ) − w0

1 dξ
=  
2π i n ∂ + Br (z0 ) f (ξ ) − w0 n
n 
1 d n−1 z − z0 
=  . (9.10)
n! dzn−1 f (z) − w0 z=z0

Thus, the following theorem has been proved.

Theorem 9.3 (Inverse Function Theorem) Let .f ∈ A(Ω), and let .f (z0 ) = 0,
where .z0 ∈ Ω, and .w0 = f (z0 ).
Then, there are numbers .r > 0 and .μ, where .μ is defined by (9.2), and an inverse
function .f −1 : Bμ (w0 ) → Br (z0 ), which is the sum of the Bürmann-Lagrange
series (9.7) whose coefficients are determined by formulas (9.10).

From Theorem 9.3 and the theorem on sufficient conditions for conformality
(see Theorem 7.8) we get immediately necessary and sufficient conditions for local
univalence.

Lemma 9.2 Let .f ∈ A(Ω) and .z0 ∈ Ω. The function f is univalent in a

neighborhood of .z0 if and only if .f (z0 ) = 0.

Example 9.2 (Counterexample From Mathematical   Analysis) For differenti-

ated mappings in real analysis, the condition .J uv |(x0 ,y0 ) = 0 is not necessary
224 9 Qualitative Properties of Analytic Functions

 
for the local injectivity of the mapping . uv . Indeed, for injective mapping

x3
. : R2 → R2 ,
y

we have
  2 
x 3   3x 0 
J =  = 0.
y (0,0)  0 1 (0,0)
.

Example 9.3 Find the inverse function of the function .f (z) = ze−az in a
neighborhood of the point .0 = f (0), where a is a positive number.

Solution Since .f (0) = 1 = 0, then according to the formula (9.10) we have


1 d n−1 z n  1 n−1 n−1
d0 = 0, dn =  = a n for n ∈ N.
.
n! dzn−1 ze−az  n!
z=0

Thus,
+∞ n−1 n−1
 a n
z = f −1 (w) =
. wn for all w ∈ BR (0),
n!
n=1

dn
where .R = limn→+∞ dn+1 = 1
ae.

Bürmann-Lagrange series are often used in the search for the asymptotics of
solutions to various transcendental equations.

Example 9.4 Consider the transcendental equation

1
. tan x = , x ∈ (0, +∞). (9.11)
x

It can be seen from Fig. 9.1 that each interval .(πp, πp + π2 ) contains only one root
of the Eq. (9.11), i.e.

1
∀ p ∈ N ∪ {0}
. ∃ ! xp ∈ (πp, πp + π2 ) : tan xp = ,
xp

and, moreover, . xp − πp → 0 as .p → +∞.

After the substitution .t = xp − πp, the equation becomes as follows

1 1 sin t
. tan t = ⇐⇒ = . (9.12)
t + πp πp cos t − t sin t
9.2 Inverse Function Theorem: Puiseux Series 225

Fig. 9.1 Roots of the transcendental Eq. (9.11)

Now consider the complex-valued function

sin z
w = f (z) :=
. .
cos z − z sin z

It is easy to verify that .f (0) = 0 and .f (0) = 1. Then, Theorem 9.3 says that the
inverse function .f −1 exists in a neighborhood of the point .0 = f (0) and

+∞

z = f −1 (w) =
. dn w n , (9.13)
n=1

where

1 d n−1 z(cos z − z sin z) n 
.dn =  , n ∈ N.
n! dzn−1 sin z 
z=0

The function in these equalities is even. Therefore its odd derivative is odd, meaning
dn = 0 for all even .n ≥ 2. For .n = 1, we directly calculate .d1 = 1.
.

The last equation in (9.12) can be rewritten with the help of the function f as
follows
1  1 
f (t) =
. , or t = f −1 .
πp πp

Keeping in mind that .t = xp − πp and using (9.13), we find from the last equation
that
1 1 1
xp − πp =
. + d3 + d5 + ...,
πp (πp)3 (πp)5
226 9 Qualitative Properties of Analytic Functions

whence we obtain the following asymptotics of the roots of the transcendental

Eq. (9.11):

1  1 
xp = πp +
. +O 3 as p → +∞.
πp p


.

Theorem 9.3 states that an analytic function f is invertible in some neighborhood

of a given point .z0 if .f (z0 ) = 0. But what can be said if this condition is not
satisfied and we have the following: there exists a unique number .p ∈ N (p ≥ 2)
such that

f (z0 ) = . . . = f (p−1) (z0 ) = 0 and

. f (p) (z0 ) = 0.

Then, as in (9.5), we can conclude that there exists a number .r > 0 such that

f (z) = w0
. and f (z) = 0 for all z ∈ Br (z0 ) \ {z0 }.

But now the Eq. (9.6) becomes

Zf −w1 = Zf −w0 = p
. in Br (z0 ). (9.14)

This means there exists a set of distinct points .{z1 = . . . = zp } ⊂ Br (z0 ) \ {z0 } such
that

f (zk ) = w1
. for all k ∈ {1, . . . , p}. (9.15)

So we can speak of a p-valued inverse function .f −1 in the disk .Bμ (w0 ), where .μ
is defined by (9.2).
To find its representation, we write the Taylor series of the function f around the
point .z0

+∞ (n)
 f (z0 )
f (z) = f (z0 ) + (z − z0 )p
. (z − z0 )n−p , z ∈ Br (z0 ).
n=p
n!

Denoting

+∞ (n)
 f (z0 )
ϕ(z) :=
. (z − z0 )n−p , z ∈ Br (z0 ),
n=p
n!
9.3 Conformal Isomorphisms and Automorphisms 227

 from the previous equation that .f (z) = w0 + (z − z0 ) ϕ(z).(p)

we obtain p Obviously,
.ϕ ∈ A Br (z0 ) and can regard that .ϕ(z) = 0 for all .z ∈ Br (z0 ), since .f (z0 ) = 0.
Thus, for all .z ∈ Br (z0 )
√ −i 2π k
w − w0 = (z − z0 )p ϕ(z) ⇒
. w − w0 e p = ψ(z)
p|0
(9.16)
√
for .k ∈ {0, 1, . . . , p − 1}, where .ψ(z) := (z − z0 ) p|0 ϕ(z) and
√ i arg z
.
p|0
z := p
|z| e p

√
is the 0th-branch of the global analytic function .{ p z}.
Since .ϕ = 0 in .Br (z0 ), the function .ψ is single-valued and analytic in .Br (z0 ),
and, moreover, .ψ(z0 ) = 0 and .ψ (z0 ) = 0. Therefore, according to Theorem 9.3,
the inverse function .ψ −1 : Bμ1 (0) → Br (z0 ) exists and

+∞

−1
ψ
. (ξ ) = dn ξ n for all ξ ∈ Bμ1 (0),
n=0

where .d0 = z0 and

1 d n−1 (z − z0 )n
dn = lim
. , n ∈ N.
z→z0 n! dzn−1 ψ n (z)

Then, it follows from (9.16) that

 p|0 +∞
  √ n
√ −i 2πp k  −i 2π k
zk = ψ −1
. w − w0 e = dn p|0
w − w0 e p ,
n=0

for all .k ∈ {0, 1, . . . , p − 1}. Combining these branches into a global analytic
function, the previous equalities can be rewritten as a series

+∞
 n
z=
. dn (w − w0 ) p , w ∈ Bμ2 (w0 ),
n=0

p
which is called Puiseux series. Here, .μ2 := μ1 < μ.

9.3 Conformal Isomorphisms and Automorphisms

Section 3.4 studied fractional-linear isomorphisms and automorphisms. Here we

generalize and continue these studies.
228 9 Qualitative Properties of Analytic Functions

Definition 9.1 Domains .Ω1 and .Ω2 in .C are said to be conformally equivalent if
there exists a conformal univalent mapping f that maps .Ω1 onto .Ω2 . In this case,
the mapping .f : Ω1 → Ω2 is called a conformal isomorphism of .Ω1 onto .Ω2 . .

It is clear that if .f : Ω1 → Ω2 is a conformal isomorphism of .Ω1 onto .Ω2 , then

there exists an inverse mapping that is a conformal isomorphism of .Ω2 onto .Ω1 .

Definition 9.2 A conformal isomorphism of a domain .Ω onto itself is called a

conformal automorphism of .Ω. .

The set of all conformal automorphisms of a domain .Ω is a group concerning

the composition of mappings. It is denoted by .Aut(Ω).

Lemma 9.3 (On the Set of Conformal Isomorphisms) Let .f0 be a conformal
isomorphism of .Ω1 onto .Ω2 . Then for any conformal isomorphism .f : Ω1 → Ω2
there is a conformal automorphism .ϕ ∈ Aut(Ω2 ) such that

f = ϕ ◦ f0 .
.

Proof For any conformal isomorphism .f : Ω1 → Ω2 , we define the following

mapping .ϕ := f ◦ f0−1 : Ω2 → Ω2 , which obviously belongs to .Aut(Ω2 ). Then

ϕ ◦ f0 = (f ◦ f0−1 ) ◦ f0 = f ◦ (f0−1 ◦ f0 ) = f
.

that had to be proved. 

Exercise 9.6 Let .Ω1 and .Ω2 be conformally equivalent domains. Prove that the
groups .Aut(Ω1 ) and .Aut(Ω2 ) are isomorphic (i.e., there is a one-to-one correspon-
dence between the elements of the groups that preserves the given group operations).

The domains .C, .C, and .B1 := B1 (0) are called canonical domains. Next, for
each canonical domain, we find the group of its conformal automorphisms.
It follows from Theorems 3.2 and 3.3 that the set of all fractional-linear mappings
 az + b 
Λ = F(z) =
. : a, b, c, d ∈ C, ad − bc = 0
cz + d
 
is a subgroup of .Aut C , and from Proposition 3.2 that the set
 z−b 
. F(z) = eiβ : b ∈ B1 , β ∈ R
1−b z
 
is a subgroup of .Aut B1 . It is also clear that the set of all linear mappings
 
.w = az + b : a, b ∈ C, a = 0
9.3 Conformal Isomorphisms and Automorphisms 229

 
is a subgroup of .Aut C . In fact, it is possible to put equal signs in these inclusions,
as the following theorem shows.

Theorem 9.4 (On Conformal Automorphisms of Canonical Domains) Every

conformal automorphism of a canonical domain is its fractional-linear automor-
phism.

Proof
 
1. Let .ϕ ∈ Aut C . Then, there is a unique point .z0 ∈ C such that .ϕ(z0 ) = ∞. It
can be assumed that .z0 ∈ C. In the opposite case, it is necessary to consider the
function .ϕ( 1z ).
Thus .ϕ is a meromorphic function and .z0 is its unique pole. Suppose that it is
of order n and .n ≥ 2. Then .z0 is a zero of multiplicity n for the function .f := ϕ1 .
Therefore,

f (z)|z=z0 = . . . = f (n−1) (z)|z=z0 = 0,

. f (n) (z)|z=z0 = 0,

and this means (see (9.15)) that f is not univalent in a neighborhood of .z0 , which
contradicts the univalence of the function .ϕ. Thus, .z0 is a simple pole of .ϕ. Then,
by Theorem 6.9 (or more exactly by the formula (6.27)),

A
ϕ(z) =
. + B, if z0 = ∞ (A = 0),
z − z0
ϕ(z) = Az + B, if z0 = ∞ (A = 0).

Hence, .ϕ is afractional-linear
 mapping.
2. Let .ϕ ∈ Aut C . Then, the point at infinity is an isolated point of the function
.ϕ. If .∞ is removable, then due to Lemma 6.1, .ϕ ≡ const. If .∞ is essential for

.ϕ, then by Picard’s great theorem (Theorem 6.7) the function .ϕ is not univalent,

which cannot be the case. Thus, .∞ is a pole. Similar to the first point of the proof,
we show that .∞ is a simple pole of .ϕ and

.ϕ(z) = Az + B (A = 0).
 
3. Let .ϕ ∈ Aut B1 . Then .ϕ(0) = w0 ∈ B1 . 
Consider the function .ζ = f (z) := F ϕ(z) , z ∈ B1 , where

w − w0
F(w) =
. (9.17)
1 − w0 w

is a fractional-linear automorphism of .B1 , which maps

 the
 point .w0 to the origin.
Clearly, .f (0) = 0. Thus, the function .f ∈ Aut B1 , and it satisfies all the
conditions of Schwarz’s Lemma 9.1. Therefore, .|f (z)| ≤ |z| for all .z ∈ B1 .
230 9 Qualitative Properties of Analytic Functions

Obviously, the inverse function

 
z = f −1 (ζ ) = ϕ −1 F−1 (ζ ) ,
. ζ ∈ B1 ,

also satisfies the conditions of Schwartz’s lemma, and hence .|f −1 (ζ )| ≤ |ζ |

for all .ζ ∈ B1 . It follows from this inequality that .|z| ≤ |f (z)| for all .z ∈
B1 . Given this inequality and considering the inverse inequality proved in the
previous paragraph, we obtain that .|f (z)| = |z| for all .z ∈ B1 .
Then, according to the second statement of the Schwarz lemma, we get

f (z) = eiβ z
. for all z ∈ B1 ,

where .β ∈ R, or .ϕ(z) = F−1 (eiβ z) for all .z ∈ B1 . Thus, .ϕ is a fractional-linear

automorphism of .B1 .


Now we want to find out: Can canonical domains be conformally equivalent?

If there is a conformal isomorphism .f : C → B1 , then .f ∈ A(C) and .|f (z)| < 1
for all .z ∈ C. According to Liouville’s theorem (Theorem 5.4), the function f is
constant, which cannot be the case. For the same reasons, there is no conformal
isomorphism .f : C → B1 . Now let .f : C → C be a conformal isomorphism. Then
.f ∈ A(C) and .|f (∞)| < +∞, from which again .f ≡ const. This is why the

canonical domains cannot be conformally isomorphic to each other.

Let’s find out when a domain .Ω is conformally equivalent to one of the canonical
domains. If .∂Ω = ∅, then .Ω = C. If .∂Ω = {z0 }, then .Ω = C \ {z0 } and the
1
fractional-linear function . z−z 0
maps it onto .C. If .Ω is a simply connected domain
whose boundary contains more than one point, then according to Riemann’s theorem
(see Theorem 2.5), which we are preparing to prove in Sect. 9.5, the domain .Ω is
conformally equivalent to .B1 .

9.4 Montel’s Theorem

Corollary 5.5 has introduced to us one of the characteristics of convergent analytic

functions. Another fundamental convergence property is Montel’s theorem on
compactness conditions for a family of analytic functions, obtained by the French
mathematician Paul Montel (1876–1975).

Definition 9.3 Let .M be a set of functions defined on a domain .Ω. The set .M is
said to be locally uniformly bounded on .Ω if for any bounded domain G, which
together with its closure, belongs to .Ω, there exists a constant C such that

|f (z)| ≤ C
. for all f ∈ M and for all z ∈ G.
9.4 Montel’s Theorem 231

Definition 9.4 Let .M be a set of functions defined on a domain .Ω. The set .M is
called locally equicontinuous on .Ω if for any bounded domain G, which together
with its closure, belongs to .Ω and for an arbitrary .ε > 0 there is a .δ = δ(ε, G) > 0
such that

|f (z1 ) − f (z2 )| < ε

. for all f ∈ M and for all z1 , z2 ∈ G with |z1 − z2 | < δ.

Lemma 9.4 Let .M be some set of analytic functions in a domain .Ω. If .M is locally
uniformly bounded on .Ω, then it is locally equicontinuous on .Ω.

Proof Take any bounded domain G, which together with its closure belongs to
Ω. We denote the distance from G to the boundary .∂Ω by .2ρ. Then, the .ρ
.

neighborhood .Gρ := ∪z∈G Bρ (z) of G also belongs to .Ω with its closure.

It follows from the theorem’s condition and Definition 9.3 that there is a constant
C such that

|f (z)| ≤ C
. for all f ∈ M and for all z ∈ Gρ .

Then, for all .z1 ∈ G, for all .z ∈ Bρ (z1 ) and for all .f ∈ M we have

|f (z) − f (z1 )| ≤ |f (z)| + |f (z1 )| ≤ 2 C.

. (9.18)

For an arbitrary function .f ∈ M define the function

1  
g(ξ ) :=
. f (ρ ξ + z1 ) − f (z1 ) for all ξ ∈ B1 . (9.19)
2C
It is easy to see that the linear function .z = ρ ξ + z1 maps the unit disk .B1 onto the
disk .Bρ (z1 ), and thanks to (9.18) the function g satisfies the conditions

1) g ∈ A(B1 ),
. 2) g : B1 → B1 , 3) g(0) = 0.

Therefore, by Schwartz’s lemma, we have .|g(ξ )| ≤ |ξ | for all .ξ ∈ B1 , or

 1   
g (z − z1 )  ≤  1 (z − z1 ) for all z ∈ Bρ (z1 ).
.
ρ ρ

In view of (9.19), the last inequality can be rewritten as

2C
|f (z) − f (z1 )| ≤
. |z − z1 | for all z ∈ Bρ (z1 ). (9.20)
ρ

Thus, for any .ε > 0 one can choose

 ερ
δ := min ρ,
.
2C
232 9 Qualitative Properties of Analytic Functions

so that for all .f ∈ M and for all .z1 , z2 ∈ G with .|z1 − z2 | < δ based on (9.20) we
have .|f (z1 ) − f (z2 )| < ε. This means that the set .M is locally equicontinuous on
.Ω. 

Definition 9.5 A set .M of functions defined on a domain .Ω is called locally

precompact if, for any compact set (bounded and closed set) from .Ω, every
sequence from .M contains a subsequence that converges uniformly on that compact.

Theorem 9.5 (Montel’s Theorem) Let .M be a set of analytic functions defined

on a domain .Ω. If . M is locally uniformly bounded, then .M is locally precompact.

Proof

1. Consider the set

E := Ω ∩ {Q × Q} = {zj }j ∈N ;
.

it is clear that it is a countable and dense subset of .Ω.

Take an arbitrary sequence .{fn }n∈N from .M.
According to the condition of the theorem, the sequence .{fn (z1 )}n∈N is
bounded. Therefore, there exists a convergent subsequence .{fnk (z1 )}k∈N , which
we denote by .{fn,1 (z1 )}n∈N . Now consider the sequence .{fn,1 (z2 )}n∈N that is
also bounded, hence there is a convergent subsequence

{fnk ,1 (z2 )}k∈N =: {fn,2 (z2 )}n∈N .

.

Thus, the sequence .{fn,2 (·)}n∈N already converges at two points .z1 and .z2 .
Continuing this process, we get a subsequence .{fn,k (·)}n∈N which is conver-
gent at the points .z1 , . . . , zk from the set E .(k ∈ N).
Now consider the diagonal sequence .{fn,n }n∈N . It converges at each point
.zp ∈ E since .{fn,n (zp )}n∈N is a subsequence of .{fn,p (zp )}n∈N for .n ≥ p.

2. Now take an arbitrary compact G from the domain .Ω and any .ε > 0. Thanks
to Lemma 9.4, the set .M is locally equicontinuous on .Ω. Therefore, according
to Definition 9.3, one can choose .δ > 0 such that for all .f ∈ M and for all
.z1 , z2 ∈ G with .|z1 − z2 | < δ we have

ε
|f (z1 ) − f (z2 )| <
. . (9.21)
3

Obviously, .Ω ⊂ +∞ j =1 Bδ (zj ). The compactness of G implies the existence
of a finite subcover of G. We can assume, without loss of generality, p that there
exists a finite number of points .{z1 , . . . , zp } ⊂ G such that .G ⊂ j =1 Bδ (zj ).
According to what was proved in the first point, the sequences

.{fn,n (z1 )}n∈N , . . . , {fn,n (zp )}n∈N

9.4 Montel’s Theorem 233

are Cauchy sequences. Therefore,

∃ n0 ∈ N ∀ n ≥ n0 , ∀m ≥ n0 ∀ j ∈ {1, 2, . . . , p} :
.

ε
|fn,n (zj ) − fm,m (zj )| < . (9.22)
3
Now take any .z ∈ G. Then there exists a .j0 ∈ {1, . . . , p} such that the point
z ∈ Bδ (zj0 ), and using (9.21) and (9.22) we deduce that
.

.|fn,n (z) − fm,m (z)| ≤ |fn,n (z) − fn,n (zj0 )| + |fn,n (zj0 ) − fm,m (zj0 )|
+ |fm,m (zj0 ) − fm,m (z)| < ε.

Thus the Cauchy criterion for uniform convergence of .{fn,n (·)}n∈N on G is

satisfied. According to Definition 9.5, the set .M is locally precompact.


Definition 9.6 Let .M be a set of functions defined on a domain .Ω.

A functional .J : M −→ C is said to be continuous on .M if for any sequence
.{fn }n∈N ⊂ M, which converges uniformly on any compact .G ⊂ Ω to a function

.f0 ∈ M,

. lim J (fn ) = J (f0 ).

n→+∞

Example 9.5 Let .M = A(Ω). Consider the functional

1 d p f (z) 
J (f ) =
.  , f ∈ M,
p! dzp z=a

where a is a point from .Ω and p is a fixed positive integer. Let us show that J is
continuous on .M.

Solution Take any sequence .{fn }n∈N ⊂ M, which converges uniformly on any
compact .G ⊂ Ω to a function .f0 ∈ M. Then, for a closed disk .Br (a) ⊂ Ω and for
any .ε > 0 there exists a .n0 ∈ N such that

|fn (z) − f0 (z)| < ε

. for all n ≥ n0 and for all z ∈ Br (a).

Then, using Cauchy’s inequalities for the coefficients of a power series (see
Corollary 5.1), we have that for all .n ≥ n0
 1 d p (f − f )  
 n 0   ε
|J (fn ) − J (f0 )| = 
.
p   ≤ p.
p! dz z=a r

This means that .limn→+∞ J (fn ) = J (f0 ). 

.
234 9 Qualitative Properties of Analytic Functions

Definition 9.7 Let .M be a locally precompact set of functions defined on a domain

Ω. The set .M is said to be locally compact if, for any sequence .{fn }n∈N ⊂ M that
.

converges uniformly to .f0 on any compact set .G ⊂ Ω, the function .f0 necessarily
belongs to .M.

Lemma 9.5 Let .M be a locally compact set of functions defined on a domain .Ω,
and let .J : M −→ C be a continuous functional.
Then there exists a function .f0 ∈ M such that

. sup |J (f )| = |J (f0 )|.

f ∈M

Proof Denote by .A := supf ∈M |J (f )|. Then there exists a maximising sequence

{fn }n∈N ⊂ M such that .limn→+∞ |J (fn )| = A. Since the set .M is locally compact,
.

there exists a subsequence .{fnk }k∈N ⊂ {fn }n∈N and .f0 ∈ M such that .fnk converges
uniformly to .f0 on every compact set .G ⊂ Ω. The continuity of the functional J
implies that

A = lim J (fnk ) = J (f0 ),

.
k→+∞

which had to be proved. 

9.5 Riemann Mapping Theorem

We are now in a position to demonstrate the proof of Riemann’s mapping theorem

(see Theorem 2.5). It was first formulated and proved in 1851 by B. Riemann, using
potential theoretical methods that were not yet fully substantiated at that time. For
most modern proofs, Montel’s theorem is used. This is the fastest way to prove this
theorem. We will separate the formulation of Theorem 2.5 and carry out its proof in
several steps.

Theorem 9.6 An arbitrary simply connected domain .Ω ⊂ C whose boundary

contains more than one point is conformally equivalent to the unit disk .B1 .

Proof

1. The first thing we wand to do is to show that there is a univalent analytic function
.f1 which maps .Ω into .B1 .

From the conditions of the theorem it follows that there are two different points
.α = β that lie on the boundary .∂Ω. Since .Ω is simply connected, two single-
9.5 Riemann Mapping Theorem 235

valued analytic branches .ϕ1 and .ϕ2 can be distinguished in .Ω for the global
analytic function
 z − α 
. .
z−β

Clearly, .ϕ1 (z) = −ϕ2 (z) for all .z ∈ Ω. Moreover, they are univalent. Indeed, if
there are .z1 , z2 ∈ Ω such that .ϕj (z1 ) = ϕj (z2 ) (j ∈ {1, 2}), then

z1 − α z2 − α
. = ,
z1 − β z2 − β

from which, based on the univalence of a fractional-linear mapping, we get .z1 =

z2 .
Theorem 9.1 says that .Ω1∗ := ϕ1 (Ω) and .Ω2∗ := ϕ2 (Ω) are domains. Assume
that .Ω1∗ ∩ Ω2∗ = ∅. Then there are two points .z1 , z2 ∈ Ω such that .ϕ1 (z1 ) =
ϕ2 (z2 ). Hence

z1 − α z2 − α
. = ,
z1 − β z2 − β

which means .z1 = z2 . So .ϕ1 (z1 ) = ϕ2 (z1 ) = −ϕ1 (z1 ), whence .ϕ1 (z1 ) = 0. This
means that it is necessary .z1 = α. As a result, we have a contradiction, because
/ Ω. Thus, .Ω1∗ ∩ Ω2∗ = ∅.
.α ∈

Since .Ω2∗ is a domain, for any .w0 ∈ Ω2∗ there exists a .ρ > 0 such that
∗
.Bρ (w0 ) ⊂ Ω . Define the function
2

ρ
f1 (z) :=
. , z ∈ Ω.
ϕ1 (z) − w0

Here the denominator is not equal to zero, because .Ω1∗ ∩ Ω2∗ = ∅. Then it is easy
to see that .f1 ∈ A(Ω), .|f1 (z)| < 1 for all .z ∈ Ω, and .f1 is univalent in .Ω as a
composition of two univalent functions.
2. Fix any point .a ∈ Ω. By Theorem 7.8 (on sufficient conditions of conformality),
.|f (a)| > 0. Now define the following function set
1

M := f ∈ A(Ω) : f is univalent in Ω,
. |f (a)| ≥ |f1 (a)|, and

|f (z)| < 1 for all z ∈ Ω .

The set .M = ∅, since the function .f1 defined in the first item of the proof belongs
to .M. It follows from the Montel Theorem 9.5 that .M is a locally precompact set.
Let us show that .M is locally compact.
Consider any sequence .{gn }n∈N ⊂ M that converges uniformly to a function
g on any compact set .G ⊂ Ω. By Corollary 5.5, the function .g ∈ A(Ω).
In addition, from Example 9.5 it follows that .limn→+∞ gn (a) = g (a). Since
236 9 Qualitative Properties of Analytic Functions

|gn (a)| ≥ |f1 (a)| > 0, then and .|g (a)| ≥ |f1 (a)| > 0. By Corollary 7.1 from
.

the Hurwitz theorem, either the function .g ≡ const in .Ω (which is impossible

because .|g (a)| > 0) or g is univalent in .Ω. Next, since

|gn (z)| < 1 for all z ∈ Ω

. and for all n ∈ N,

and .|g(z)| ≤ 1. But the maximum modulus principle gives (see Theorem 9.2)
that .|g(z)| < 1 for all .z ∈ Ω. Thus, .g ∈ M and .M is locally compact.
3. Now define the functional .J (f ) := |f (a)| for all .f ∈ M. According to
Example 9.5, the functional J is continuous, and due to Lemma 9.5 there exists
a function .f0 ∈ M such that

. sup |f (a)| = |f0 (a)|. (9.23)

f ∈M

The next step will be a demonstration that .f0 is the conformal isomorphism of
interest.
First, we prove that .f0 (a) = 0. Assume that is not the case, i.e., .f0 (a) = 0,
and define the function
f0 (z) − f0 (a)
.g0 (z) := , z ∈ Ω.
1 − f0 (a) f0 (z)

Since .g0 is a composition of a fractional-linear automorphism of .B1 (see (9.18))

and .f0 , the function .g0 is analytic and univalent, and .|g(z)| < 1 for all .z ∈ Ω. It
is easy to calculate that
 
 f0 (a)  |f0 (a)|
.|g0 (a)| =  
 1 − |f (a)|2  = 1 − |f (a)|2 > |f0 (a)| ≥ |f1 (a)| > 0.
0 0

Thus, .g ∈ M and .|g (a)| > |f0 (a)|. However, the last inequality is in
contradiction with the equality (9.23). Therefore, .f0 (a) = 0.
Now we show that the function .f0 : Ω → B1 is surjective. If this is not the
case, then there exists a point .b ∈ B1 such that its preimage is empty, i.e., .f0 (z) =
b for all .z ∈ Ω. Since .f0 (a) = 0, the point .b = 0. We can therefore define the
function

f0 (z) − b
.ψ(z) := , z ∈ Ω.
1 − b f0 (z)
√ √
Here by . · we mean one of the branches of the global analytic function   .{ · },
which is uniquely defined in .Ω because .f0 (z) = b and .f0 (z) = 1 , ( 1  > 1) for
b b
all .z ∈ Ω. Thus, .ψ is a single-valued, analytic, and univalent function in .Ω, and
also .|ψ(z)| < 1 for all .z ∈ Ω (see the explanation above in this item).
9.5 Riemann Mapping Theorem 237

√
Since .ψ(a) = −b, then .|ψ(a)|2 = |b|. Next we find

1 f0 (a) − f0 (a) |b|2 1 − |b|2

ψ (a) = √
. = f0 (a) √ . (9.24)
2 −b 1 2 −b

Now define this function

ψ(z) − ψ(a)
h(z) :=
. , z ∈ Ω,
1 − ψ(a) ψ(z)

which is single-valued and analytic in .Ω, and .|h(z)| < 1 for all .z ∈ Ω. In
addition, using (9.24), we find

|ψ (a)| 1 − |b|2 1 1 + |b|

|h (a)| =
. = |f0 (a)| √ = |f0 (a)| √ .
1 − |ψ(a)| 2 2 |b| 1 − |b| 2 |b|

Since .b = 0, . 21+|b|
√
|b|
> 1. And therefore,

|h (a)| > |f0 (a)| ≥ |f1 (a)| > 0.

.

Thus, .h ∈ M and .|h (a)| > |f0 (a)|. However, the last inequality contradicts the
equality (9.23). This means that the mapping .f0 : Ω → B1 is surjective.
On the basis of sufficient conformality conditions (see Theorem 7.8), the
function .f0 is conformal in the domain .Ω. Thereby, .f0 is conformal isomorphism
of .Ω onto .B1 .


From this theorem follows immediately the following statement.

Corollary 9.2 Any two simply connected domains in .C whose boundaries contain
more than one point are conformally equivalent.

It is clear that there is no unique conformal isomorphism of the domain .Ω

onto the unit disk .B1 . This follows from the fact that the set of all conformal
automorphisms is infinite (see Theorem 9.4 (part 3) and Proposition 3.2). However,
taking into account the form of all conformal automorphisms of .B1 , the following
statement gives additional conditions under which there exists a unique conformal
isomorphism .Ω onto .B1 .
238 9 Qualitative Properties of Analytic Functions

Theorem 9.7 (On Uniqueness of Conformal Isomorphism) Let .Ω be a simply

connected domain .Ω ⊂ C whose boundary contains more than one point. Then for
any .z0 ∈ Ω and any real number .α ∈ (−π, π ] there exists a unique conformal
isomorphism .f : Ω → B1 such that

f (z0 ) = 0 and
. arg f (z0 ) = α. (9.25)

Proof Let g be any conformal isomorphism of the domain .Ω onto .B1 which exists
by Theorem 9.6. Then, by Lemma 9.3 and Theorem 9.4 (part 3), an arbitrary
conformal isomorphism .f : Ω → B1 can be represented as .f = F ◦ g, where
.F ∈ Aut B1 , i.e.,

g(z) − a
f (z) = eiθ
. , z ∈ Ω,
1 − a g(z)

where a is an arbitrary point from .B1 and .θ ∈ R.

Take .a = g(z0 ) and consider the following conformal isomorphism

g(z) − g(z0 )
f1 (z) = eiθ
. , z ∈ Ω.
1 − g(z0 ) g(z)

It is easy to verify that .f1 (z0 ) = 0 and

g (z0 )
f1 (z)|z=z0 = eiθ
. , Arg f1 (z0 ) = θ + arg g (z0 ).
1 − |g(z0 )|2

Let us set .θ = α − arg g (z0 ). Thus there exists a conformal isomorphism .Ω onto
.B1 which satisfies the relations (9.25). Next, we will show that it is unique.
Suppose there is another conformal isomorphism .f2 : Ω → B1 which satisfies
the relations (9.25). Then .ϕ := f1 ◦ f2−1 ∈ Aut B1 . Theorem 9.4 (part 3) says that
.ϕ is a fractional-linear automorphism of .B1 , and moreover, .ϕ(0) = 0,

f1 (z0 )
.ϕ (0) = and hence arg ϕ (0) = 0.
f2 (z0 )

These equalities mean that .ϕ(w) = w for all .w ∈ B1 , whence we get that .f1 (z) =
f2 (z) for all .z ∈ Ω. 
References

1. Golberg, A.A., Sheremeta, M.M., Zabolotsky, M.V., Skaskiv, O.B.: Complex Analysis. Afisha,
Lviv (2002) (in Ukrainian)
2. Greene, R.E., Krantz, S.G.: Function Theory of One Complex Variable. Graduate Studies in
Mathematics, vol. 40, 2nd edn. AMS, Providence (2002)
3. Grishchenko, A.E., Nagnibida, N.I., Nastasiev, P.P.: Theory of Functions of a Complex
Variable. Solving Problems. Vyshcha shkola, Kyiv (1986) (in Russian)
4. Markushevich, A.I.: Theory of Analytic Functions of a Complex Variable, vol. I, II, III, 2nd
English edn. Chelsea Publishing, New York (1977)
5. Mel’nyk, T.A.: Complex Analysis. University Press “Kyiv University”, Kyiv (2015) (in
Ukrainian)
6. Milewski, E.G.: The Complex Variables Problem Solver. Research & Education Association,
Piscataway (1987)
7. Nahin, P.J.: An Imaginary Tale: The Story of the Square of Minus One. Princeton University
Press, Princeton (2016)
8. Pap, E.: Complex Analysis Through Examples and Exercises. Kluwer, Amsterdam (1999)
9. Shabat, B.V.: Introduction á l’analyse complexe. Tome I “Fonctions d’une variable”. Mir,
Moscow (1990)
10. Shabat, B.V.: Introduction to Complex Analysis. Part II “Functions of Several Variables”.
Translation of Mathematical Monographs, vol. 110. AMS, Providence (1992)
11. Shabunin, M.I., Sidorov, Y.V., Fedoryuk, M.V.: Lectures on the Theory of Functions of a
Complex Variable. Mir, Moscow (1985)
12. Shakarchi, R.: Problems and Solutions for Complex Analysis. Springer, Berlin (1999)
13. Simon, B.: Basic complex analysis. A Comprehensive Course in Analysis, part 2A. AMS,
Providence (2015)
14. Volkovyskii, L.I., Lunts, G.L., Aramanovich, I.G.: A Collection of Problems on Complex
Analysis. Dover Publications, Mineola (2011)
15. Whyburn, G.: Analytic Topology. AMS, Providence (1942)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 239
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1
Index

A Closure of a set, 18
Abelian group, 3 Complex numbers, 2
Analytic branches, 61, 63, 66, 203, 205 Complex plane, 3
Analytic continuation Complex potential, 33
along a chain, 198 Conformal mapping, 35
along a curve, 199 angle-preserving, 38
direct, 186 conformal automorphism, 228
Schwarz’s reflection, 195 conformal isomorphism, 228
Analytic continuation by continuity, 191 criterion, 38
Analytic function element, 186 equal stretch at a point, 35
canonic, 186 at infinity, 41
Antiderivative, 86 orientation-preserving, 38
along a curve, 91 sufficient conditions, 173
global, 101, 102 Conjugate pair of harmonic functions, 30
local, 89 Criterion of local univalence, 223
Argument, 4, 66 Curves
increment along a curve, 169 closed, 13
equivalent .(γ1 ∼ γ2 ), 14
homotopic .(γ0 ≈ γ1 ), 96
B Jordan, 15
Boundary of a domain, 18 null-homotopic .(γ0 ≈ 0), 97
Boundary with positive orientation .∂ + Ω, 100 piecewise smooth, 16
Branch point rectifiable, 16
first-order, 61 simple, 15
logarithmic, 67, 213 smooth, 15
.n − 1 order, 63, 213
Bürmann-Lagrange series, 222

D
C Derivative, 24
Casorati–Sokhotskyi–Weierstrass theorem, geometric meaning of the argument, 37
145 geometric meaning of the modulus, 35
Cauchy–Goursat theorem, 87 Domains, 18
Cauchy-Riemann equations, 26, 28 of analyticity, 188
Cauchy’s integral formula, 103 canonical, 228
for derivatives, 118 conformally equivalent, 228
Cauchy’s residue theorem, 152 convex, 97
Cauchy’s theorem, 98 fractional-linear isomorphic, 55
Classification of isolated singularities multiply connected, 18
removable, pole, essential, 139 simply connected, 18

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 241
T. Mel’nyk, Complex Analysis, https://doi.org/10.1007/978-3-031-39615-1
242 Index

E I
Euler’s formula, 5, 117 Imaginary axis, 3
Exact differential form, 31 Improper integral, 161
Extended complex plane, 8 principal value, 164
Infinite product, 181
Integral
F along a curve, 81, 95
Field, 3 curvilinear of the first kind, 85
Fractional-linear automorphism of a domain, curvilinear of the second kind, 83, 96
55 Riemann, 12
Fractional-linear mappings hyperbolic, elliptic Interior .int(γ) of a Jordan curve, 15
loxodromic, parabolic, 51 Inverse function theorem, 223
Function Inversion with respect to a circle, 53
analytic at a point, 29
analytic at infinity, 29
analytic, holomorphic, regular, 123
J
analytic in a domain, 29
Jordan curve theorem, 15
continuous at a point, 23
Jordan’s lemma, 163
continuous on a set, 23
Joukowsky wing profile, 72
differentiable, 25
entire, 30, 147
exponential, 29, 63
fractional-linear, 45 L
hyperbolic, 73 Landau symbol .o(z), 25
Joukowsky, 68 Laplace operator, 30
linear, 43 Laurent series, 132
meromorphic, 148 analysity of the sum, 136
power .zn , 57 annulus of convergence, 133
rational, 148 around .∞, 145
trigonometric, 73 Cauchy inequalities, 136
uniformly continuous, 24 connection to Fourier series, 137
univalent, 41, 223 expansion of an analytic function, 134
Fundamental theorem of algebra, 113, 129, 173 regular and principal parts, 133
Length of a curve, 16
Limit point, 17
G Liouville’s theorem, 112
Gamma function, 122, 193 Little Picard’s theorem, 147, 149
Global
√ analytic
 function, 202 Logarithmic derivative, 167
. z , 204, 205
.{Log}, 205
branch, 203 M
domain of analyticity, 203 Mean value theorem, 105
isolated singular point, 211 Modulus, 3
branch point, 213 Monodromy theorem, 200
single-valued, 213 Montel’s theorem, 232
Riemann surface, 208, 210 Morera’s theorem, 119
Great Picard’s theorem, 146

N
H Neighborhood
Harmonic function, 30 .Br (a) of a point a, 17
Homeomorphism, 10 of .∞, 17
Hurwitz’s theorem, 174 punctured, 139
Index 243

Newton-Leibniz formula (analog), 93 Riemann zeta function, 121, 194, 195

Non-isolated singular point, 140 Rouché’s theorem, 171
Runge’s theorem, 119

O
S
Open mapping theorem, 218
Schwarz–Pick theorem, 221
Orientation of a curve, 13
Schwarz’s lemma, 220
negative, 15
Sequence of functions
positive, 15
analycity of the limit function, 121
term-by-term differentiation, 121
uniformly convergent, 109
P
univalence of the limit function, 175
Parametrization of a curve, 14
Sequence of numbers convergence, 7
Partial fraction decomposition of a
Series of functions, 108
meromorphic function, 148, 177
power series (109 (see also Power series))
Picard exceptional value, 149
term-by-term differentiation, 120
Point
uniformly convergent, 109
.w0 -point of f , 171
Weierstrass criterion, 109
at infinity .∞, 8
Series of numbers, 107
isolated, 139
absolutely convergent, 108
Polar coordinates, 4
conditionally convergent, 108
Pole of order N , 143
Set
Power series, 109
closed, 17
analyticity of the sum, 113
compact, 23
Bürmann-Lagrange series, 222
open, 17
Cauchy–Hadamard theorem, 109
path-connected, 18
Cauchy’s inequalities, 112
Set of functions
disk of convergence, 110
locally compact, 234
expansion of an analytic function, 111
locally equicontinuous, 231
radius of convergence, 109
locally precompact, 232
Taylor series, 116
locally uniformly bounded, 230
Principal branch of .Log, 66
Singular point of an analytic function element,
Principle
188
argument, 169
Stereographic projection, 9
maximum modulus, 105, 219
Stream function, 34
preserving boundaries and their
Symmetric points with respect to a circle, 52
orientations, 52
Schwarz’s reflection, 195
Puiseux series, 227
T
Taylor series, 116
R Topological space .(R, τ), 207
Real axis, 3 Trace of a curve, 13
Regular sequence of Jordan curves, 177
Residues, 151, 152
formulas, 153 U
logarithmic, 168 Uniqueness
Riemann mapping theorem, 41, 234 for analytic functions, 124
Riemann sphere, 10
Riemann surface, 59, 208, 210
.RArcsin , 77 V
.RLog , 67, 210 Vector field potential, solenoidal irrotational,
.R √n w , 63 33
.R√w , 61 Vector space, 3
244 Index

W Z
Wallis formula, 184 Zero of order m, 127
Weierstrass’s factorization theorem, 182 Zero of order m at infinity, 129