Abel’s Theorem Through Problems

V. B. Alekseev
 Abel’s Theorem Through Problems

V. B. Alekseev

Translated by Yuliya Gorlina

and Igor Pavlovsky

Preface and Introduction translated

by Leonid Grinberg

Edited by Julian Gilbey

Translation Copyright c Mathematics Foundation of America, 2002–2015

 Contents

Translators’ Foreword v
Foreword by Prof. V. I. Arnold vii
Preface xi
Introduction xiii
Chapter 1. Groups 1
§1 Examples 1
§2 Groups of transformations 4
§3 Groups 6
§4 Cyclic groups 8
§5 Isomorphisms 10
§6 Subgroups 11
§7 Direct products 13
§8 Cosets and Lagrange’s Theorem 14
§9 Inner automorphisms 15
§10 Normal subgroups 17
§11 Quotient groups 18
§12 The commutator subgroup 20
§13 Homomorphisms 21
§14 Solvable groups 25
§15 Permutations 27
Chapter 2. Complex numbers 31
§1 Fields and polynomials 32
§2 The field of complex numbers 36
§3 The uniqueness of the field of complex numbers 39
§4 Geometric representation of complex numbers 41
§5 Trigonometric representation of complex numbers 43
§6 Continuity 46
§7 Continuous curves 48
§8 Maps of curves and the fundamental
√ theorem of algebra 53
§9 The Riemann surface for w = z 56
§10 Riemann surfaces for more complicated functions 64
§11 Functions expressible in radicals 71
§12 Galois groups of multi-valued functions 76
§13 Galois groups of functions expressible in radicals 78
§14 Abel’s theorem 79
Index 83

iii
 Translators’ Foreword

This book by V. B. Alekseev was originally published in Russian in 2001. It follows

a course given by Prof. V. I. Arnold in 1963 to a class of 14–16 year old students at
a new high school in Moscow. A short article by Arnold about the purpose of this
course can be found below.
The book was translated into English by Mathcamp staff for use at Mathcamp
2002. That summer, a self-selected group of about fifteen students worked on the
problems over the course of the five-week camp, presenting their solutions to their
classmates. Following the methodology of R. L. Moore, the students themselves
critiqued their classmates’ answers and arguments, and we only moved on when
they were all happy with the presented work. The post-docs who were running the
class (Mira Bernstein, David Savitt and Julian Gilbey) only stepped in to guide the
pace of work and to explain points which had been missed by the whole group. On
the very last day of classes, the group succeeded in solving problem number 352,
thereby completing the proof of Abel’s Theorem. The class developed a strong sense
of the nature of mathematical argument, and developed a facility with both group
theory and Riemann surfaces, as well as thoroughly enjoying the opportunity to fully
engage with real mathematics.
The original book has full solutions included; we did not have the time to translate
these. A translation of the whole book, including solutions and an extensive appendix
on differential Galois theory, has been published by Kluwer under the title Abel’s
Theorem in Problems and Solutions: Based on the Lectures of Professor V. I. Arnold.
Julian Gilbey
London, December 2007

v
 Foreword by Prof. V. I. Arnold

Abel’s Theorem and Modern Mathematics

This foreword is an edited version of Prof. Arnold’s article, which can be found at
http://www.institut.math.jussieu.fr/seminaires/singularites/abel.pdf.

The myth of the difficulty of mathematics and of its inaccessibility to school children
has caused harm similar to that of the physicists’ typical opinion, originating with
Landau, which regards mathematics as the continuation of the art of long multipli-
cation.
In contrast to this opinion, the greatest mathematicians are distinguished by their
inclination to replace blind calculations by clear ideas.† An interesting and sophisti-
cated example of the use of clear ideas in mathematics is provided by the topological
proof of Abel’s Theorem on the impossibility of solving algebraic equations of degree
five by radicals and rational functions.
It is very difficult to decide which parts of mathematics should and should not
be taught at a standard high school. In 1963, Kolmogorov asked me to provide a
one semester long mathematical course on any subject of my choice at the new high
school he had established in Moscow. This was an attempt to introduce something
new to replace the traditional Euclidean curriculum which was followed in all high
schools at the time. I chose to teach Abel’s Theorem, transforming its topological
proof into a series of accessible problems for beginners, trying to provide them with
several important ideas which usually remain outside the mathematical knowledge
of both schoolchildren and university students (and even professors).
Abel’s ideas presented in this way included, for instance, the geometrical theory
of complex numbers and the topology of Riemann surfaces, including notions such as
the fundamental group and ramified covering monodromies, normal subgroups and
exact sequences, and also the symmetry groups of the regular polyhedra (including
that of the dodecahedron and the Kepler cubes inscribed into it). All this “modern”
science might more commonly be found in quantum physics lectures than in for-
mal mathematics education; schoolchildren are more able to transform the difficult
trigonometric expressions with which they have been tortured in their mathematics
lessons than to think geometrically. My 14–16 year old students understood these
lectures on Abel’s ideas very well, solving many of the problems and exercises, and

†
In the Russian translation of Bourbaki’s “History of Mathematics”, these words of Dirichlet
appeared in the form “to replace clear ideas by blind calculations”; the Editor (Kolmogorov) ex-
plained to me that this (wrong) translation is a better description of Bourbaki’s activity than the
original words of Dirichlet.

vii
viii FOREWORD BY PROF. V. I. ARNOLD

were very successful in the final examination in which they were asked to think orig-
inally. I gave them far higher evaluations than typical university students, who are
only trained to apply the standard formal rules.
This formal application of formal rules, comprising the majority of standard
mathematics education, reminds me of Marat’s comment that the best mathemati-
cian was Laplace, since he calculated the solution to every problem, substituting
numbers in existing formulæ. Napoleon, however, dismissed Laplace from the min-
isterial post he had earlier granted him, accusing him of “introducing the spirit of
infinitesimals into the state administration”. (My French collaborators explained to
me that Laplace tried to verify every cent in all of the accounts.)
Abel, unlike Laplace, had never been close to government. His proof of his theo-
rem, presented by him to the Paris Academy of Sciences, had not been understood,
for it contained too many new ideas, including geometrical and topological ones, as
well as algebraic and combinatorial thinking. Moreover, his manuscript had been
lost or hidden by Cauchy and was published only many years after Abel’s death.
According to an article in a French newspaper, this mathematician was so unfor-
tunate that following the year in Paris “he returned by foot to his part of Siberia,
called Norway, over the Atlantic Ocean ice.” Abel’s own description of French math-
ematics was astonishingly enduring and modern: “Everyone wishes to teach and no
one wishes to study anything new, being expert in only his own particular domain
(be it heat theory or elasticity theory, celestial mechanics or number theory), and
having no interest in any more general questions.”
Abel’s approach to mathematics was very different. Early in his life, he had
studied the geometric theory of complex numbers with Wessel, later extending it
himself to the theory of Riemann surfaces and Abelian integrals (which are integrals
along curves on Riemann surfaces). (In this context, the explicit integrability of
the square root of quadratic polynomials depends on the topological fact that the
Riemann surface of a quadratic is a sphere, yielding both the technique of “Euler’
substitutions”, which reduces these integrals to integrals of rational functions, and
Pythagorean triples such as 32 + 42 = 52 and 122 + 52 = 132 , whose relationship to
the Riemann surface’s topology deserves to be shown at high school.) The unity of
all the mathematics appearing in these facts is similar to the discovery of the unity
of the theories of electricity and of magnetism in physics.
In a similar manner, Abel also understood the reason for the impossibility of
explicit integration of square roots of cubic polynomials (for instance, of the explicit
calculation of the length of an ellipse) as deriving from the toroidal topology of
the corresponding Riemann surfaces. (These integrals had already been studied by
Newton in his attempt to explain the ellipticity of the planets’ orbits from Kepler’s
law of areas swept out.)
The influence of Abel’s ideas, containing, among other things, the main points
of the the theory of Riemann surfaces (which were also developed simultaneously
by Jacobi), including their relationship to integrals of algebraic functions (today
 FOREWORD BY PROF. V. I. ARNOLD ix

known as Abelian integrals), is still exceptionally high both in mathematics and in

mathematical physics, which had both followed the approach proposed by Abel. It
is a pity to recall that some mathematicians, more successful than Abel socially,
proclaimed just the opposite opinion on his works. For example, Hardy wrote that
“Abel, Riemann, and Poincaré have contributed nothing essential for mankind.” My
opinion is that the ideas of Abel, of Riemann and of Poincaré are the real basis of
most of modern mathematics, on which are based most results of modern physics and
technology, and all the successes of modern human civilisation, be they intellectual or
material. This includes, for instance, the aerodynamics of aeroplanes and the space
shuttles, celestial mechanics and wave propagation theory, ecology and quantum
field theory. The quality and quantity of modern studies continuing these works of
Abel, Riemann and Poincaré are enormous, and the ideas contained in their works
influenced both the theoretical sciences and their useful applications far more than
the technical details of ready-made formulæ.
In my 1964 high school course on the topological proof of Abel’s Theorem, I had
attributed to Abel the topological proofs of some results, including the following
version of what I called “Topological Galois Theory”. My proof of Abel’s Theorem
extends his result on the unsolvability of degree 5 equations in radicals to all complex
functions which are topologically equivalent to the dependence of the roots of the
degree 5 equations on their coefficients. I considered it as the topological unsolvability
in radicals of equations of degree 5 (or higher). And I then attributed to Abel
the topological unintegrability theorem for Abelian integrals, including elliptic ones:
neither elliptic integrals nor the elliptic functions are topologically equivalent to any
elementary function (even when combined with any univalent function).
In 1964, I proposed to my students that they should publish the proofs of these
important results and it seems that forty years later they are finally attaining this
goal.
The return of mathematics education (both in high schools and in university)
from the sophisticated application of ready-made formulæ to real mathematical
thinking and to simple ideas seems to me the only way to preserve the past cen-
tury’s mathematical culture and to generate new discoveries and new applications
in the future.
 Preface

During secondary school, algebraic equations in one variable of degree 1 (linear equa-
tions) and of degree 2 (quadratic equations) are thoroughly studied. It is discovered
that for such equations, there exist general formulæ which express the roots of the
equation using its coefficients, using only the arithmetic operations and radicals
(square roots, cube roots and so forth). However, whether or not there exist similar
formulæ for the solution of algebraic equations of higher degrees is known to far
fewer people. It turns out that for equations of degree 3 or 4 such formulæ also
exist. We will examine methods of solving such equations in the Introduction. But
on examining a general algebraic equation in one variable with degree greater than 4,
one discovers that it is not solvable in radicals; that is, there does not exist a formula
which expresses the roots of such an equation in terms of its coefficients, using only
the arithmetic operations and radicals. That is the result known as Abel’s Theorem.
One of the goals of this book is to introduce the reader to the proof of Abel’s
theorem. We will not examine in detail the results obtained somewhat later by
the French mathematician Évariste Galois, who analysed not general, but specific
algebraic equations with fixed, numerical coefficients, and found for such equations
a rule by which the roots of the equation could be found using arithmetic operations
and radicals. For those who want to better acquaint themselves with Galois’ results,
the books of M. M. Postnikov (Foundations of Galois Theory) and B. L. van der
Waerden (Algebra) can be recommended.†
From Galois’ general results, one can deduce Abel’s theorem. However in this
book, we will follow a different route, which will allow the reader to acquaint himself
with two very different branches of modern mathematics: Group Theory and the
Theory of Functions of Complex Variables. The reader will learn what groups and
fields are, and what properties they possess. From there, we move on to discovering
what complex numbers are, with a brief discussion of their name. We will also find
out what a Riemann surface is and use them to prove The Fundamental Theorem of
Algebra.
The author will accompany and guide the reader on this path, but will allow him
a wide opportunity to test out his own abilities. For this purpose, the reader will be
presented with a large number of problems. These are, essentially, the main contents
of the text of this book. The problems are consistently numbered throughout the
book.

The book contains many terms which may be new to the reader. In order for
the reader to more easily navigate through them, there is an index of terms with the
page number on which they are defined.
†
Ian Stewart’s book Galois Theory is also highly recommended (ed.)

xi
xii PREFACE

The book is based on lectures presented over a number of years by Professor

Vladimir Igorovich Arnold and the author at the Faculty of Mechanics and Mathe-
matics of the Moscow State University. The author wishes to express his gratitude
to Prof. V. I. Arnold, who offered a large number of valuable comments during the
preparation of this book.
 Introduction

We begin this book by examining the question of how equations in one variable of
degree from 1 to 4 are solved. The methods of solving algebraic equations of degree 1
and 2 were known even to mathematicians of the ancient world; methods of solving
algebraic equations of degree 3 and 4 were only devised in the 16th century.
A general algebraic equation in one variable of degree n is an equation of the form
an xn + an−1 xn−1 + · · · + a1 x + a0 = 0
such that an 6= 0.†
When n = 1, we obtain the linear equation
a1 x + a0 = 0, a1 6= 0.
This equation obviously has the one solution
a0
x=−
a1
for any values of the coefficients.
If n = 2, we get the quadratic equation
ax2 + bx + c = 0, a 6= 0
(where instead of a2 , a1 , a0 , we write a, b and c, as is customary in school). Dividing
both sides of this equation by a, and setting p = b/a, q = c/a, we get the quadratic
equation
x2 + px + q = 0. (1)
We can manipulate this to obtain
p2 p2
x2 + px + = −q
4 4
so that
 p  2 p2
x+ = − q. (2)
2 4
2
In the course of middle school, only the case of p4 − q > 0 is examined further. If
p2
4
− q < 0, then it is said that the right hand side of equation (2) cannot be square-
rooted, and that equation (1) has no real roots. In order to avoid such exceptions,
it will be more convenient for us to study algebraic equations not in the field of real
numbers, but in the field of complex numbers.
We will examine complex numbers in detail (including their definition) in Chap-
ter 2. For now, it is sufficient for the reader to know or take on faith the following
statements about complex numbers:
†
The coefficients an , an−1 , . . . , a0 can for now be considered to be arbitrary real numbers.

xiii
xiv INTRODUCTION

• The set of complex numbers is an extension of the set of real numbers; that
is, real numbers are a subset of complex numbers in the same way that, for
example, whole numbers are a subset of real numbers.
• Complex numbers can be added, subtracted, multiplied, divided and raised
to a natural power, and these operations have the same properties as the
corresponding operations with real numbers.
• If z is a non-zero complex number and n is a natural number, then there
exist exactly n roots of the nth degree
√ of z; that is, n complex numbers w
n n
such that w = z. If z = 0, then 0 = 0. If w1 and w2 are the square roots
of z, then w2 = −w1 .
Below, we will be interested not only in real and complex roots of equations, but
also in the nature of the coefficients of such equations. At the same time, the above
discussion on linear and quadratic equations remains valid, because of the second
property of complex numbers just described.
Let us continue the examination of the quadratic equation. In the field of complex
numbers, equation (2), for any values of p and q, is equivalent to the equation
r
p p2
x+ =± − q,
2 4
p
where p2 /4 − q is understood to mean some particular value of the square root.
In this way the roots are
r
p p2
x1,2 = − ± − q. (3)
2 4
Introducing a, b and c, we obtain the well-known
√
−b ± b2 − 4ac
x1,2 = . (4)
2a
For the sequel, we will need two facts relating to quadratic equations:
• Viète’s Theorem†: Complex numbers x1 and x2 are roots of the equation
x2 +px+q = 0 if and only if x1 +x2 = −p and x1 x2 = q. Indeed, if x1 and x2
are roots of the equation x2 +px+q = 0, then equation (3) is satisfied. From
here, x1 + x2 = −p and x1 x2 = q. Conversely, if x1 + x2 = −p and x1 x2 = q,
then substituting for p and q in the equation x2 + px + q = 0, we obtain
x2 − (x1 + x2 )x + x1 x2 = (x − x1 )(x − x2 ) = 0, and therefore x1 and x2 are
the roots of the equation x2 + px + q = 0.
• The quadratic equation ax2 + bx + c = 0 is a perfect square (that is, ax2 +
√
2
bx + c = a(x − x0 ) for some complex number x0 ) if and only if the roots
of the equation ax2 + bx + c = 0 are both equal to x0 . This happens if and
only if (see equation (4)) b2 − 4ac = 0. The expression b2 − 4ac is known as
the discriminant of the quadratic equation.
†
François Viète (1540–1603): French mathematician
 INTRODUCTION xv

Let us now consider the monic cubic equation

x3 + ax2 + bx + c = 0. (5)
(The general cubic equation can be converted to this form on division by a3 .) Sub-
stituting x = y + d, where d will be assigned a value later, we obtain
(y + d)3 + a(y + d)2 + b(y + d) + c = 0.
Multiplying everything out, we get the equation,
y 3 + (3d + a)y 2 + (3d2 + 2ad + b)y + (d3 + ad2 + bd + c) = 0.
The coefficient of y 2 in this equation is equal to 3d + a. Therefore, if we set d = − a3 ,
the substitution x = y − a3 brings the equation into the form
y 3 + py + q = 0 (6)
where p and q are polynomials in a, b and c.
Let y0 be a root of equation (6). Writing this root as y0 = α + β (where the
values of α and β are for now unknown) and expanding y03 , we obtain
α3 + 3αβ(α + β) + β 3 + p(α + β) + q = 0,
which gives
α3 + β 3 + (α + β)(3αβ + p) + q = 0 (7)
Let us see if we can add another restriction to α and β:
p
αβ = −
3
In this case, we get two simultaneous equations for α and β:
α + β = y0
p
αβ = −
3
By Viète’s theorem, for any y0 , such α and β do actually exist (possibly as complex
numbers) and are the roots of the quadratic equation
p
w2 − y0 w − = 0.
3
If we take such α and β (for now, with still-unknown values), then equation (7) takes
the form
α3 + β 3 + q = 0. (8)
Raising αβ to the third power and combining the obtained equation with equa-
tion (8), we obtain
α3 + β 3 = −q
p3
α3 β 3 = −
27
xvi INTRODUCTION

whence by Viète’s theorem, α3 and β 3 are the roots of the quadratic equation
p3
w2 + qw − = 0.
27
In this way,
r r
q q 2 p 3 q q 2 p3
α3 = − + + and β3 = − − +
2 4 27 2 4 27
p
where again, q 2 /4 + p3 /27 is understood to mean a particular value of the square
root. From here, the roots of equation (6) are expressed by the formula
s r s r
3 q q 2 p 3 3 q q 2 p3
y1,2,3 = − + + + − − +
2 4 27 2 4 27
such that for each of the three values of the first cube root†, it is necessary to take
the corresponding value of the second root, such that the rule αβ = − p3 holds.
This formula is called Cardano’s Formula ‡. Replacing p and q by their values
in terms of a, b and c, and subtracting a3 , we obtain the formula for the roots of
equation (5). Then replacing a = aa23 , b = aa13 and c = aa03 , we obtain the formula for
the roots of the general cubic equation.
Let us now examine the monic quartic equation
x4 + ax3 + bx2 + cx + d = 0. (9)
(Again, the general equation reduces to the monic one upon dividing by a4 .) After
performing the substitution x = y − a4 , as in the cubic case, we reduce equation (9)
to the form
y 4 + py 2 + qy + r = 0, (10)
where p, q and r are polynomials in a, b, c and d.
We will solve equation (10) via a method known as Ferrari’s Method§. Let us
rewrite equation (10) in the following way. Since
p2
 

2 p 2
y + + qy + r − =0
2 4
it follows that
2   p2


2 p 2 p 2
y + + α − 2α y + + α − qy + −r =0 (11)
2 2 4
where α is an arbitrary number. Let us now attempt to pick α in such a way that
the quadratic in y in the second set of brackets
p2
 
2 2
2αy − qy + αp + α + −r
4
†
see the third of the above-mentioned properties of complex numbers
‡
G. Cardano (1501–1576), an Italian mathematician
§
L. Ferrari (1522–1565), an Italian mathematician and student of Cardano
 INTRODUCTION xvii

is actually a perfect square. As was mentioned earlier, in order for it to be a perfect

square, it is necessary and sufficient for the discriminant to be equal to 0, that is
p2
 
2 2
q − 8α αp + α + − r = 0. (12)
4
Expanding the brackets in this equation yields a cubic equation in α, which we now
know how to solve. If we then substitute α with one of the roots of equation (12)
back into (11), then the expression in brackets in (11) will be a perfect square. In
that case, the left side of equation (11) will be the difference of two squares, which
can thus be expressed as the product of two quadratic polynomials in y. Solving
these and using x = y − a4 gives us the solutions of the original quartic equation.
In this way, a quartic (degree 4 polynomial) can always be solved, and, further-
more, as in the case of cubics, it is possible to obtain a formula expressing the roots
of the general quartic equation in terms of the coefficients of the equation using only
the operations of addition, subtraction, multiplication, division, raising to a natural
power and extracting natural roots.
For a long time, mathematicians attempted to find a method of solving a general
quintic (degree 5) equation using only these same operations. However, in 1824,
the Norwegian mathematician Niels Henrik Abel (1802–1829) proved the following
theorem:
Abel’s Theorem. The general algebraic equation in one variable of degree greater
than four is unsolvable in radicals. That is, there does not exist a formula expressing
the roots of a general equation of degree greater than four using only the coefficients
and the operations of addition, subtraction, multiplication, division, raising to a
natural power and extracting natural roots.
We will be able to prove this theorem by the end of the book. However, in
order to achieve this, we will need to develop the mathematical concepts of groups,
solvable groups, functions of a complex variable, Riemann surfaces and so on. We
will acquaint the readers with these ideas in the remaining pages of this book. We
begin with the very important mathematical concept of a group.
 CHAPTER 1

Groups

Investigation of polynomial equations in the beginning of the nineteenth century

led mathematicians to a new concept—the concept of a group. This new concept
turned out to be so fruitful that not only did it enter and influence almost all areas
of mathematics, but also started playing an important role in other fields of science,
for example quantum mechanics and crystallography. Investigations connected to
the concept of a group grew to form a new branch of modern mathematics—Group
Theory. What is a “group” in mathematics? To answer that question, let us begin
with some examples.

§1. Examples
In arithmetic, we use operations where two numbers are put into correspondence
with a third number. For example, the operation of addition assigns the number 8
to the pair (3, 5) and the number 4 to the pair (2, 2). The operation of subtraction,
when viewed as acting on all integers, also places pairs of integers into correspondence
with single integers. In this situation, the order of the numbers is also important.
For example, the pair (5, 3) corresponds to 2 under subtraction, while the pair (3, 5)
corresponds to −2. Thus the pairs (5, 3) and (3, 5) must be viewed as different.
Pairs for which the order of elements is relevant will be called ordered pairs.
Definition 1. Let M be a set. A binary operation on M is an assignment of an
element of M to each ordered pair of elements of M .
Examples of binary operations include addition on the set of natural numbers or
integers, and subtraction on the set of integers. Subtraction is not a binary operation
on the set of natural numbers because, for example, the ordered pair (3, 5) does not
correspond to a natural number.
1. Consider the operations: (a) addition of numbers; (b) subtraction of numbers,
and (c) multiplication of numbers. For which of the following subsets of the set of
integers are they binary operations? (1) The set of all even integers; (2) the set of
all odd integers; (3) the set of all negative integers, and (4) the set of all positive
integers.†
†
Some of the suggested problems are of practical nature and serve to improve one’s understand-
ing of the new concepts using examples. Other problems are theoretical and their results are used
later. Therefore, if the reader is not able to solve a problem, he should familiarise himself with its
solution when it is discussed in class.

1
2 CHAPTER 1. GROUPS

Let us look at several more examples of binary operations. We will frequently

return to these examples.
Example 1. Let A, B and C be the vertices of an equilateral triangle ABC (fig. 1).
Let’s rotate the triangle 120◦ anticlockwise around its centre O. Then the vertex A
A

O
B C

Figure 1.

will go to B, B to C and C to A. This way, the triangle will be superimposed on

itself (if we ignore the names of the vertices), i.e., the 120◦ rotation around point O is
a transformation that takes this triangle
to itself. We will denote this transformation
by a. It can be written as a = B A B C , where the top row contains the list of all
C A
vertices and the bottom row shows where each of them goes. A 240◦ rotation in the
same direction around point O is also a transformation which
sends the triangle to
itself. We denote this transformation by b, so b = CA B C . There exists another
A B
rotation which sends the triangle to itself and which is different from
a and b—that is
the 0◦ rotation. Let us call that transformation e, with e = A B C . It is easy to see
A B C
that there are only three different rotations in the plane† which send the equilateral
triangle ABC to itself: e, a and b.
Let g1 and g2 be arbitrary transformations of the triangle. Then by g1 · g2 (or
simply g1 g2 ) we mean the transformation g3 which is obtained by first applying
transformation g2 and then g1 ; we will call g3 the composition of transformations g2
and g1 .
We can make a multiplication table (table 1), where each row and each column
corresponds to a rotation which sends the triangle ABC to itself. At the intersection

e a b
e
a e
b

Table 1.

of the row corresponding to g1 and the column corresponding to g2 we will put the
transformation equal to g1 · g2 . For example, in the highlighted cell of table 1, we
†
We mean that these are rotations which do not move into three dimensions, i.e., rotations
around only those axes which are perpendicular to the plane.
 1. EXAMPLES 3

put the transformation a · b which we get if we turn the triangle 240◦ and then 120◦ .
Therefore a · b is the 360◦ turn, so it is the same as e. We can get the same result
by reasoning this way: transformation b sends A to C and transformation a sends C
to A. This way, the transformation a · b sends vertex A to A. In exactly the
same
way, we can see that vertex B goes to B and C goes to C. So ab = A B C
A B C , i.e.,
ab = e.
2. Complete table 1.
Any transformation of a figure to itself, preserving the distances between all of
its points, is called a symmetry of that figure. In example 1, the rotations of the
triangle are its symmetries.
Example 2. In addition to rotations, an equilateral triangle has three other sym-
metries: reflections about the axes l1 , l2 and l3 (fig. 2).
A
l3 l2

B l1 C

Figure 2.


A B C ,
We will
call these transformations

respectively c, d and f , so that c = A C B
d = CA B B C and f = A B C . In this situation, there are two different ways to view
A B A C
the composition of two transformations. Let us look, for example, at the composition
of transformations c · d. We could assume that when we apply d, the axis l1 moves
to a new position (the position of the old axis l3 ), and view transformation c as a
reflection with respect to the new position of l1 (i.e., with respect to the old l3 ).
Alternatively, we could assume that the axes are not connected rigidly to the figure,
and do not transform with it, and therefore after applying d, we should apply c as a
reflection about the old axis l1 . The latter is the way we will view compositions of
transformations. When using this approach, reasoning analogous to that discussed
immediately before problem 2 turns out to work, and we discover that cd = a. This
type of reasoning is useful for calculating compositions of transformations.
3. Construct the multiplication table for all symmetries of an equilateral triangle.
Example 3. Let e, a, b and c represent rotations of a square by 0◦ , 180◦ , 90◦ and
270◦ anticlockwise, respectively (fig. 3).
4. Construct the multiplication table for rotations of the square.
Example 4. Let d, f , g and h represent reflections of the square about the axes
labelled in fig. 4.
5. Construct the multiplication table of all symmetries of the square.
4 CHAPTER 1. GROUPS

A D

O
B C

Figure 3.

d g f

Figure 4.

Example 5. Let ABCD be a rhombus that is not a square.

6. Find all symmetries of the given rhombus and construct their multiplication table.
Example 6. Let ABCD be a rectangle that is not a square.
7. Find all symmetries of the given rectangle and construct their multiplication
table.

§2. Groups of transformations

Let X and Y be two sets of elements and let every element x in X correspond to
precisely one element y in Y . Then we say that we have a map ϕ from the set X to
the set Y (written ϕ : X → Y ).† The element y is called the image of element x,
while x is called a preimage of the element y, and we write ϕ(x) = y. Note that each
element x in X has a unique image, but an element y in Y may have any number of
preimages.
Definition 2. The map ϕ : X → Y is called a surjective map if for every element y
in Y there exists an element x in X such that ϕ(x) = y, i.e., if every y in Y has a
preimage in X.
8. Let ϕ be the map from towns in the UK to the letters of the alphabet, which
takes every town to the first letter of its name. What would it mean for this map to
be surjective?
Definition 3. The map ϕ : X → Y is called a bijective map if for every element y
in Y , there exists a unique preimage in X.
†
A map is also known as a function.
 2. GROUPS OF TRANSFORMATIONS 5

9. Consider the following maps from the set of all integers to the set of all non-
negative integers:†
(a) ϕ(n) = n2 ,
(b) ϕ(n) = |n|,
(
2n if n > 0
(c) ϕ(n) =
2|n| − 1 if n < 0.
Which of these maps are surjective and which of them are bijective?
Let M be a set. A bijective map from the set M to itself, g : M → M , is called
a transformation of the set M .
Two transformations g1 and g2 will be considered equal if g1 (A) = g2 (A) for every
element A in M . Instead of the term transformation, the term permutation is often
used. In this book, we will only use this term when dealing with finite sets. In such
a case, the permutation can be written as
 
A1 A2 . . . An
,
Ai1 Ai2 . . . Ain
where the first row lists all elements of the set and the second row shows where the
permutation takes each of these elements.
Because a transformation is a bijective map, for every transformation g there
exists an inverse transformation g −1 , which is defined

as follows: if g(A) = B, then
−1 −1


A B C
g (B) = A. Thus in example 1, where a = B C A , it follows that a = CA B C
A B ,
i.e., a−1 = b.
10. Find the inverse transformations of all symmetries of an equilateral triangle
(examples 1 and 2).
11. Let g be the transformation of the set of real numbers defined by g(x) = 2x.
Find the inverse transformation g −1 .
The composition of the transformations g1 and g2 , written as g1 g2 , is defined
by (g1 g2 )(A) = g1 (g2 (A)): first apply transformation g2 , then g1 . If g1 and g2 are
transformations of a set M , then g1 g2 is also a transformation of the set M .
Definition 4. Let G be a non-empty set of transformations of a set M having the
following properties: (1) if transformations g1 and g2 are in G, then their composition
g3 = g1 g2 is in G; (2) if transformation g is in G, then the inverse transformation g −1
is in G. Such a set of transformations G will be called a group of transformations.
It is easy to check that the sets of transformations discussed in examples 1–6 are
groups of transformations.
12. Prove that every group of transformations contains an identity transformation e,
such that e(A) = A for every element A of the set M .
†
For any number x, the number |x| is the absolute value of x, meaning x with any minus sign
stripped off. For example, |−3| = |3| = 3.
6 CHAPTER 1. GROUPS

13. Prove that eg = ge = g for every transformation g.

14. Prove that for any three transformations g1 , g2 and g3 , the following equality
holds:†
g1 (g2 g3 ) = (g1 g2 )g3 .

§3. Groups
When solving problems 6 and 7, we constructed multiplication tables of the sym-
metries of a rhombus and of a rectangle. Check that it is possible to label these
symmetries in such a way that these tables are the same. For many purposes, it is
natural to consider such groups of transformations as being “the same”.
Because of this, we will turn our attention away from the nature of the elements
in the set (in this case, transformations) and the nature of the binary operation‡ (in
this case, composition of transformations). Instead, we will view a group simply as
an arbitrary set with a binary operation, as long as the main properties of a group of
transformations hold, as we now explain. We will usually call the binary operation
multiplication, and if a pair (a, b) corresponds to c, then we will call c the product of
a and b, and we write a · b = c or simply ab = c. In some cases, this operation will
be given a different name, such as composition, addition, etc.
Definition 5. A group is a set G of elements of arbitrary nature, together with a
binary operation a · b defined on G, such that the following conditions hold:
(1) the operation is associative, that is, (ab)c = a(bc) for every a, b and c in G;
(2) there exists an element e in G, such that ea = ae = a for every element a
in G; this element e is called the identity element of the group G;
(3) for each element a in G, there exists an element a−1 in G such that aa−1 =
a−1 a = e; this element is called the inverse of a.
From the results of problems 12–14, we see that any group of transformations is
a group (in some sense, the converse is also true—see problem 55).§Because of this,
we already have some examples of groups. All of these groups have a finite number
of elements, and such groups are called finite groups. The number of elements in a
finite group is called the order of the group. Groups that contain an infinite number
of elements are called infinite groups.
Let us look now at some examples of infinite groups.
Example 7. Consider the set of all integers. We will use addition as our binary
operation, which gives us a group. The identity element here is 0, because 0 + n =
†
This equality holds not only for transformations, but for any three maps g1 , g2 , g3 such that
g3 : M1 → M2 , g2 : M2 → M3 , g1 : M3 → M4 .
‡
For the definition of a binary operation, see page 1.
§
Actually, we also need to show that gg −1 = g −1 g = e for every g in the group of transforma-
tions. This is straightforward, though.
 3. GROUPS 7

n + 0 = n for every integer n. Also, for every n, there exists an inverse −n (which
is often called the additive inverse or negative when dealing with addition), because
n + (−n) = (−n) + n = 0. Associativity in this case follows from the laws of
arithmetic. The group we get this way is called the group of integers under addition.
15. Do the following sets form groups under multiplication? (a) All real numbers,
and (b) all non-zero real numbers?
16. Do the positive real numbers form a group under multiplication?
17. Do the natural numbers (0, 1, 2, . . . ) form a group: (a) under addition, (b) under
multiplication?
18. Prove that in any group, there is a unique identity element.
19. Prove that for any element a in a group, there is a unique inverse of a.
20. Prove that (a) e−1 = e; (b) for any element a in a group, (a−1 )−1 = a.
If a and b are elements of some group, then, by the definition of a binary operation,
a · b is an element of the group. Therefore, expressions of the form (a · b) · c, a · (b · c),
(a · b) · (c · d) and so on also give elements of the group. Any two elements defined
through multiplication can also be multiplied to get yet another element of the group,
and so on. To tell which operation was performed last, every time we multiply two
elements, we will place parentheses around them (expressions consisting of only one
element do not need parentheses, and neither do we need parentheses around the
final product). Those expressions that can be formed with this method will be called
well-formed products. For example, (a · b) · (c · (a · c)) is a well formed product, while
(a · b) · c · (a · c) is not, because the order in which the operations are performed is
not clear. When we look at products a1 · a2 · . . . · an of several real numbers a1 , a2 ,
. . . , an , we do not use parentheses at all, because it turns out that the result does not
depend on the order in which the operations are performed, i.e., with any placement
of parentheses that gives a well formed product, the result will be the same. It turns
out that this property is true for any group, which follows from the result of the
following problem.
21. Let a binary operation a · b be associative, i.e., (a · b) · c = a · (b · c) for any
three elements a, b, c. Prove that every well formed product that consists from left
to right of the elements a1 , a2 , . . . , an gives the same element as the product
(. . . ((a1 · a2 ) · a3 ) · . . . · an−1 ) · an .
Thus, if a1 , a2 , . . . , an are elements of a group, any well-formed product made
using these elements in this order will give the same element regardless of the place-
ment of the parentheses. We will denote this product by a1 · a2 · . . . · an or just
a1 a2 . . . an (without parentheses).
When multiplying real numbers, we have another very important property, namely
that the product a1 · a2 · . . . · an does not change when we change the order of the
terms. But in an arbitrary group, this property may not hold.
8 CHAPTER 1. GROUPS

Definition 6. Two elements a and b of a group are said to commute if ab = ba. If all
elements of a group commute with each other, then the group is called commutative
or abelian.†
There exist non-abelian groups. One example of a non-abelian group is the group
of symmetries of a triangle (see example 2, where ac = f but ca = d, and so ac 6= ca).
22. Determine whether the following groups are commutative (see problems 2 and
4–7): (a) the group of rotations of a triangle; (b) the group of rotations of a square;
(c) the group of symmetries of a square; (d) the group of symmetries of a rhombus,
and (e) the group of symmetries of a rectangle.
−1
23. Prove that in any group: (a) (ab)−1 = b−1 a−1 , (b) (a1 · . . . · an )−1 = a−1
n · . . . · a1 .
(An aide-mémoire: A jacket is put on after a shirt and is taken off before.)
If we have an equality a = b in an arbitrary group G (which means that the
left hand side and the right hand side denote the same element), then from it we
can deduce new equalities by multiplying both sides of the original equality by some
element c of the group G. However, because the product may depend on the order
of elements, we must either multiply both sides of the equality on the right to get
ac = bc, or multiply both sides on the left to give ca = cb.
24. Let a and b be any elements of a group G. Prove that the equations ax = b and
ya = b each have a unique solution in G.
The uniqueness of problem 24 can be expressed the following way: if ab1 = ab2
or b1 a = b2 a, then b1 = b2 .
25. Let a · a = e for every element a in a group G. Prove that the group G is
commutative.
By am , where m is an arbitrary positive integer and a an arbitrary element of a
group G, we will mean the product a · a · . . . · a, where the number of terms is m.
26. Prove that (a−1 )m = (am )−1 , where m is a positive integer.
Thus (a−1 )m and (am )−1 , where m is a positive integer, are the same element,
which we will denote by a−m . We also set a0 = e for every element a of the group.
27. Prove that am · an = am+n for any two integers m and n.
28. Prove that (am )n = amn for any two integers m and n.

§4. Cyclic groups

The simplest (yet very important) groups are the cyclic groups, which we will now
study.
†
Named after the Norwegian mathematician Niels Henrik Abel (1802–1829).
 4. CYCLIC GROUPS 9

Definition 7. Let a be an element of a group G. The smallest positive integer n

such that an = e is called the order of the element a. If such an n does not exist,
then we say that a is an element of infinite order .
29. Find the orders of all elements in the groups of symmetries of an equilateral
triangle, square and rhombus (see problems 3, 5 and 6).
30. Suppose an element a has order n. Prove that: (a) the elements e, a, a2 , . . . , an−1
are all distinct, and (b) for any integer m, the element am equals one of the elements
listed above.
Definition 8. If an element a has order n and all the elements of a group G are
equal to one of e, a, a2 , . . . , an−1 , then the group G is called a cyclic group of order n,
generated by a, and the element a is called a generator of the group.
Example 8. Suppose we have a regular n-gon in the plane. Let us consider all
rotations of the plane that send the regular n-gon to itself.
31. Prove that these rotations form a cyclic group of order n.
32. Find all the generators of the group of rotations of a triangle and of a square
(examples 1 and 3).
33. Suppose the element a has order n. Prove that am = e if and only if m = nd,
where d is an arbitrary integer.
34. Suppose that a has prime order p and let m be an arbitrary integer. Prove that
either am = e or am has order p.
35. Let the greatest common divisor (highest common factor) of two positive integers
m and n be d, and suppose that a has order n. Prove that am has order n/d.
36. Find all generators of the group of rotations of a regular 12-gon.
37. Let a be an element of infinite order. Prove that the elements . . . , a−2 , a−1 ,
a0 = e, a, a2 , . . . are all distinct.
Definition 9. If a is an element of infinite order in the group G, and all of the
elements of G are equal to one of . . . , a−2 , a−1 , e, a, a2 , . . . , then G is called an
infinite cyclic group, generated by a.
38. Prove that the group of integers under addition (example 7) is an infinite cyclic
group. Find its generators.
Example 9. Let n be a positive integer greater than 1. Let us consider all possible
remainders that we can get when dividing integers by n, i.e., the numbers 0, 1, 2,
. . . , n − 1. We define a binary operation on the set of these remainders as follows.
We add remainders as usual, then find the remainder of the sum when divided by n,
and define the result to be that remainder. We call this operation addition modulo n.
For example, modulo 4, we get 1 + 2 = 3 and 3 + 3 = 2.
10 CHAPTER 1. GROUPS

39. Make the tables of addition modulo: (a) 2, (b) 3, (c) 4.

40. Prove that the remainders with the operation of addition modulo n form a cyclic
group of order n.
Let us look again at an arbitrary cyclic group of order n with elements e, a, a2 ,
. . . , an−1 .
41. Prove that am · ar = ak , where 0 6 m < n, 0 6 r < n and 0 6 k < n, if and
only if the equality m + r = k holds modulo n.
It follows from the result of the last problem that, in some sense, the multiplica-
tion of elements in an arbitrary cyclic group of order n corresponds to the addition
of remainders modulo n. In the same way, multiplication of elements in an infinite
cyclic group corresponds to the addition of integers (see problem 27). Here we come
to an important concept in group theory, that of an isomorphism.

§5. Isomorphisms
Definition 10. Suppose we are given two groups G1 and G2 , and suppose there
exists a bijective map ϕ from the elements of the group G1 to the elements of the
group G2 (see section 2), such that multiplication in G1 corresponds to multiplication
in G2 , i.e., if ϕ(a) = a0 , ϕ(b) = b0 , ϕ(c) = c0 and ab = c in the group G1 , then in
the group G2 , a0 b0 = c0 . Then ϕ is called an isomorphism from the group G1 to
the group G2 , and groups between which we can define an isomorphism are called
isomorphic. The condition that the bijective map ϕ is an isomorphism can be written
the following way: ϕ(ab) = ϕ(a) · ϕ(b) for any elements a and b of the group G1 ;
here the multiplication ab is performed in G1 and the multiplication ϕ(a) · ϕ(b) is
performed in G2 .
42. Which of the following groups are isomorphic to each other? (1) The group of
rotations of a square; (2) the group of symmetries of a rhombus; (3) the group of
symmetries of a rectangle; (4) the group of remainders under addition modulo 4.
43. Let ϕ : G1 → G2 be an isomorphism. Prove that the inverse map ϕ−1 : G2 → G1
is also an isomorphism.
44. Let ϕ1 : G1 → G2 and ϕ2 : G2 → G3 be isomorphisms. Prove that ϕ2 ϕ1 : G1 →
G3 is also an isomorphism.
It follows from the last two problems that if two groups are both isomorphic to
a third, then they are isomorphic to each other. (Why?)
45. Prove that any cyclic group of order n is isomorphic to the group of remainders
under addition modulo n.
46. Prove that any infinite cyclic group is isomorphic to the group of integers under
addition.
 6. SUBGROUPS 11

47. Let ϕ : G → F be an isomorphism. Prove that ϕ(eG ) = eF , where eG and eF

are the identity elements of G and F , respectively.
48. Let ϕ : G → F be an isomorphism. Prove that ϕ(g −1 ) = (ϕ(g))−1 for every
element g of the group G.
49. Let ϕ : G → F be an isomorphism and ϕ(g) = h. Prove that the orders of g
and h are the same.
When we are studying the group operation itself, and the nature of the elements
of the group does not play any role, then isomorphic groups may be considered to be
the same. For example, we say that there exists only one cyclic group of order n up
to isomorphism (see problem 45), which we will denote by Cn , and that there is only
one infinite cyclic group up to isomorphism (see problem 46), which we will denote
by C∞ .
If a group G1 is isomorphic to a group G2 , then we will write G1 ∼
= G2 .
50. Find, up to isomorphism, all groups that have: (a) 2 elements, (b) 3 elements.
51. Give an example of two groups that have the same number of elements and are
non-isomorphic.
52. Prove that the group of all real numbers under addition is isomorphic to the
group of all positive real numbers under multiplication.
53. Let a be an arbitrary element of the group G. Consider the map ϕa from the
set of elements of the group G to itself, defined by ϕa (x) = ax for any element x
in G. Prove that ϕa is a transformation of the set of elements of the group G (i.e.,
it is a bijective map from the set of elements of G to itself).
54. We can define the transformation ϕa (see previous problem) for every element a
of the group G. Prove that the set of these transformations ϕa forms a group with
the usual operation of composition of transformations.
55. Prove that the group G is isomorphic to the group of transformations defined
in the previous problem. (This result is known as Cayley’s Theorem.)

§6. Subgroups
Let us consider a subset H of the group G. It may turn out that H itself is a group
under the same binary operation as G.
In that case, H is called a subgroup of the group G. For example, the group of
rotations of a regular n-gon is a subgroup of the group of all symmetries of a regular
n-gon.
If a is an element of the group G, then the set of all elements of the form am is
a subgroup of the group G. (This subgroup is a cyclic group, like those we studied
in section 4.)
12 CHAPTER 1. GROUPS

56. Let H be a subgroup of the group G. Prove that: (a) the identity elements of
G and H are the same; (b) if a is an element of the subgroup H, then the inverse
of a is the same in both G and H.
57. For H to be a subgroup of G (under the same binary operation) it is necessary
and sufficient for following conditions to be satisfied: (1) if a and b are both in H,
then the element ab (the multiplication performed in G) is also in H; (2) e (the
identity element of G) is in H; (3) if a is in H, then a−1 (inversion performed in G)
is also in H. Prove this.
Note. Condition (2) follows from conditions (1) and (3) if H is assumed to be non-
empty.
58. Find all subgroups of the group of: (a) symmetries of an equilateral triangle;
(b) symmetries of a square.
59. Find all subgroups of the cyclic groups: (a) C5 , (b) C8 , (c) C15 .
n
60. Prove that all subgroups of Cn are of the form {e, ad , a2d , . . . , a( d −1)d }, where
d is a factor of n and a is a generator of Cn .
61. Prove that all subgroups of the infinite cyclic group C∞ are of the form {. . . , a−2r , a−r ,
e, ar , a2r , . . . }, where a is a generator and r is an arbitrary non-negative integer.

62. Prove that in any infinite group there are an infinite number of subgroups.
63. Prove that the intersection of any number of subgroups† of a group G is also a
subgroup of the group G.
Example 10. Let us look at a regular tetrahedron with vertices labelled A, B, C
and D. If we look from the side of the vertex D at the triangle ABC, then the points
A, B, C can go either clockwise or anticlockwise (see fig. 5). This is the way we will
distinguish the two orientations of the tetrahedron.

D D

B C C B

A A

Figure 5.

†
The intersection of several sets is the set of all elements that belong to all of the given sets at
the same time.
 7. DIRECT PRODUCTS 13

64. Do the
following transformations preserve orientation of the

tetrahedron: a =
A B C D , rotation by 120◦ about the altitude; b = A B C D , rotation by 180◦
B C A D D C B A

about the axis that goes through the centres of the edges AD and BC; c = A B C D
A C B D ,
reflection about the plane

that goes through the edge AD and the centre of the
edge BC; d = B A B C D , the transformation that permutes the vertices cyclically?
C D A

All symmetries of the regular tetrahedron clearly form a group, which is called
the group of symmetries of a tetrahedron.
65. How many elements are there in the group of symmetries of a tetrahedron?
66. In the group of symmetries of a tetrahedron, find subgroups isomorphic to:
(a) the group of symmetries of a triangle; (b) the cyclic group C4 .
67. Prove that all symmetries of a tetrahedron which preserve orientation form a
group. How many elements are in that group?
The group of symmetries of a tetrahedron which preserve orientation is called
the group of rotations of a tetrahedron.
68. In the group of rotations of a tetrahedron, find subgroups isomorphic to the
cyclic groups: (a) C2 , (b) C3 .

§7. Direct products

Given two groups, we can form a new group in the following way.
Definition 11. The direct product of two groups G and H, denoted by G × H (and
usually pronounced “G cross H”), is the set of all ordered pairs (g, h), where g is
an arbitrary element of G and h is an arbitrary element of H, with the following
binary operation: (g1 , h1 ) · (g2 , h2 ) = (g1 g2 , h1 h2 ), where the multiplication g1 g2 is
performed in the group G and h1 h2 in H.
69. Prove that G × H is a group.
70. Suppose that the group G has n elements and the group H has k elements. How
many elements are in the group G × H?
71. Prove that the groups G × H and H × G are isomorphic.
72. Find subgroups of G × H which are isomorphic to G and to H respectively.
73. Suppose that G and H are commutative. Prove that G×H is also commutative.
74. Let G1 be a subgroup of G and H1 be a subgroup of H. Prove that G1 × H1 is
a subgroup of G × H.
75. Let G and H be arbitrary groups. Is it necessarily true that every subgroup of
the group G × H can be written as G1 × H1 where G1 is a subgroup of G and H1 is
a subgroup of H?
14 CHAPTER 1. GROUPS

76. Prove that the group of symmetries of a rhombus is isomorphic to the group
C2 × C2 .
77. Are the following pairs of groups isomorphic? (a) C2 × C3 and C6 ; (b) C2 × C4
and C8 .
78. Prove that Cm × Cn ∼ = Cmn if and only if the numbers m and n are relatively
†
prime. (The “if” part of this result is often known as the Chinese Remainder
Theorem, and can be also be stated in terms of modular arithmetic as follows (see
problem 45): If m and n are relatively prime, then the simultaneous equations x ≡ a
(mod m) and x ≡ b (mod n) have a unique solution modulo mn.)

§8. Cosets and Lagrange’s Theorem

Each subgroup H of a group G gives rise to a partition of the set of elements of G,
as we now show.
For any element x in G, consider the set of all elements of the form xh, where
h is an element of the subgroup H. This set, denoted by xH, is called the left coset
of H generated by the element x.
79. Find all left cosets of the following subgroups of the group of symmetries of a
triangle: (a) rotations of a triangle; (b) reflections with respect to one axis {e, c}
(see examples 1 and 2).
80. Let H be a subgroup of a group G. Prove that every element of G belongs to
some left coset of H.
81. Suppose that the element y belongs to the left coset of H generated by the
element x. Prove that the left cosets of H generated by the elements x and y
coincide. (In symbols: if y is an element of xH, then xH = yH.)
82. Suppose that the two left cosets of H generated by elements x and y have an
element in common. Prove that these cosets coincide.
Therefore the left cosets generated by any two elements either do not intersect or
they coincide, giving us a partition of all elements of the group G into non-intersecting
classes. This partition is called the left coset decomposition of the group G by the
subgroup H.
Assume G to be finite. The number of elements in a subgroup of G is called the
order of that subgroup. Let m be the order of the subgroup H. If h1 6= h2 , then
xh1 6= xh2 , so each left coset also has m elements. Therefore, if n is the order of the
group G and r is the number of left cosets in the coset decomposition of G by H, we
have mr = n, which proves
†
Two numbers are called relatively prime or coprime if their greatest common divisor (highest
common factor) is 1.
 9. INNER AUTOMORPHISMS 15

Theorem 1 (Lagrange’s† Theorem). The order of a subgroup divides the order of

the group.
83. Prove that the order of any element (see page 9) divides the order of the group.
84. Prove that any group of prime order is cyclic and any element in it except e is
a generator.
85. The group G contains 31 elements. How many subgroups can the group G have?
86. Prove that all groups of prime order p are isomorphic to each other.
87. Suppose that m divides n. Construct a group of order n which contains a
subgroup isomorphic to a given group G of order m.
88. Suppose that m divides n. Must a group of order n have a subgroup of order m?
We can also construct the right cosets Hx and the right coset decomposition of
the group G by the subgroup H. If the order of the subgroup H is equal to m, then
as before, all right cosets also have m elements and their number is equal to the
integer n/m, where n is the order of the group. So the number of right cosets is
equal to the number of left cosets.
Note. In practice, to find the coset decomposition of a finite group, it is not necessary
to construct the coset of each element, because this way we will have repeating
classes. Rather, we should take elements that are not in cosets already constructed.
Because eH = He = H, the subgroup itself is always a right and a left coset.
89. Construct the left and right coset decompositions of the group of symmetries of
a triangle by the subgroup of: (a) rotations {e, a, b}, (b) reflections with respect to
one axis {e, c}.
90. Construct the left and right coset decompositions of the group of symmetries of
a square by the subgroup of: (a) reflections with respect to the central reflections
{e, a}, (b) reflections with respect to a diagonal {e, d}.
91. Construct the coset decomposition of the group of integers under addition by
the subgroup of numbers divisible by 3.‡
92. Find, up to isomorphism, all groups of order: (a) 4, (b) 6, (c) 8.

§9. Inner automorphisms

Let us begin with an example. We will study the group of symmetries of an equi-
lateral triangle. If we denote the vertices by A, B, and C, then every element of
†
Langrange, Joseph Louis (1736–1813) was a French mathematician and mechanical physicist.
‡
We do not specify here which decomposition to construct, left or right, because in a commu-
tative group these two decompositions obviously coincide.
16 CHAPTER 1. GROUPS

the group can be defined by a permutation of the three letters A, B and C. For
example, the reflection of the triangle with
respect to its altitude from the vertex A
A B C
to the side BC will be written as A C B . To multiply two elements of the group of
symmetries of the triangle, it is sufficient to perform the two permutations one after
another. This way, we get an isomorphism of the group of symmetries of a triangle
to the group of permutations of three letters, A, B, and C. Note, though, that the
isomorphism is not unique: it depends on which vertex we denote by A, which by B
and which by C. Relabelling the vertices can itself be
viewed as a permutation of the
three letters A, B and C. For example, g = B A B C corresponds to the following
C A
relabelling of the vertices:
old name of vertex A B C
new name of vertex B C A

With these new names, every element of the group of symmetries of the triangle
will get a new name as a permutation of the letters A, B, and C. For example, the
reflection of the triangle with respect to the vertical altitude (fig. 6) is denoted in
the following way:


old name of permutation: A B C
A C B


new name of permutation: CA B B A
C

A (B)

B (C) C (A)

Figure 6.

93. Consider an element of the group of symmetries of a triangle, which, under some
naming of the vertices, corresponds to the permutation h. Which permutation will
correspond to the same element of the group of symmetries of a triangle when the
vertices are renamed by g?
Note that this relabelling using g turns the element h of the group of transforma-
tions into ghg −1 . This is the case not only in this example of the group of symmetries
of a triangle, but also in general. An element of the form ghg −1 is called a conjugate
of h and the process is called conjugation. Doing this relabelling to an entire group,
though, is so important that it warrants the following definition.
Definition 12. Let G be a group and g an element of G. Define a map ϕg from the
group G to itself by the formula ϕg (h) = ghg −1 (where h is any group element). This
map is called the inner automorphism of the group G generated by the element g.
 10. NORMAL SUBGROUPS 17

94. Prove that an inner automorphism of a group is an isomorphism from the group
to itself. (The word automorphism means an isomorphism from the object to itself.)
95. To what does the reflection with respect to an altitude get sent by all possible
inner automorphisms of the group of symmetries of a triangle?
96. To what does the rotation of the triangle by 120◦ get sent by all possible inner
automorphisms of the group of symmetries of a triangle?
97. Which pairs of elements of the group of symmetries of a tetrahedron can be sent
to each other by inner automorphisms and which cannot? Answer the same question
for the group of rotations of a tetrahedron.
98. Prove that the orders of elements ab and ba are equal in any group.
We note that under any inner automorphism of a group (as well as with any
other isomorphism of the group to itself), a subgroup is generally sent to a different
subgroup. For example, reflections with respect to one altitude are sent to reflections
with respect to another altitude. However, some highly symmetric subgroups stay
in place under all inner automorphisms, for example, the subgroup of rotations of
the group of symmetries of a triangle. We will now look at such subgroups.

§10. Normal subgroups

Definition 13. A subgroup of a group is called a normal subgroup if it is sent to
itself by every inner automorphism of the group. In other words, a subgroup N of
the group G is called a normal subgroup of G if for any element a in N and any
element g in G, the element gag −1 is in N .
Thus, the subgroup of rotations is a normal subgroup of the group of symme-
tries of a triangle, while the subgroup of reflections with respect to the altitude
from A to BC (consisting of two elements) is not a normal subgroup of the group of
symmetries of a triangle.
99. Prove that in a commutative group, every subgroup is normal.
100. Is the subgroup of central symmetries of the group of symmetries of a square
consisting of elements {e, a} normal (examples 3 and 4)?
Theorem 2. A subgroup N of a group G is normal if and only if the left and right
coset decompositions (see section 8) of the group G by the subgroup N coincide.†
101. Prove the theorem stated above.
102. Let G be a group of order n and H a subgroup with order m = n/2. Prove
that H is a normal subgroup of G.
†
In this case, the decompositions will simply be called decompositions by a normal subgroup.
18 CHAPTER 1. GROUPS

103. Prove that an intersection (see the footnote on page 12) of any number of
normal subgroups of a group G is a normal subgroup of the group G.
104. The set of elements of the group G that commute with all elements of the
group is called the centre of the group G. Prove that the centre is a subgroup, in
fact, a normal subgroup, of the group G.
105. Let N1 and N2 be normal subgroups of the groups G1 and G2 respectively.
Prove that N1 × N2 is a normal subgroup of the group G1 × G2 .
The following example shows that a normal subgroup of a normal subgroup of a
group G does not have to be a normal subgroup of the group G itself.
Example 11. We examine the subgroup of the group of symmetries of a square
which consists of reflections with respect to the diagonals and the centre (see exam-
ples 3 and 4), that is, {e, a, d, f }. This subgroup contains half of the elements of the
group and hence is a normal subgroup (see problem 102 above). The subgroup {e, d},
which consists of reflections with respect to one of the diagonals, contains half of the
elements of the subgroup {e, a, d, f } and is, therefore, a normal subgroup of it. On
the other hand, the subgroup {e, d} is not a normal subgroup of the entire group of
symmetries of a square, because d goes to bdb−1 = f under the inner automorphism
of the group generated by reflection with respect to another diagonal.

§11. Quotient groups

We begin with an example. We will look at the decomposition of the group of
symmetries of a square by the normal subgroup consisting of the central symmetries
e and a (see examples 3 and 4). It is an easy result that the decomposition will
have four cosets and be as described in table 2. Let us denote each coset by a letter;

e b d g
a c f h
E A B C
Table 2.

we have used E, A, B, and C in our table. If we multiply any element from the
class A by any element in the class B, then the result turns out to be in the class C,
independently of which elements were chosen from A and B. From the solution of
the following problem, it follows that this isn’t by accident.
106. Suppose we have a decomposition of the group G by a normal subgroup N ,
and suppose elements x1 and x2 lie in the same coset and elements y1 and y2 also lie
in the same coset. Prove that the elements x1 y1 and x2 y2 both lie in the same coset.
 11. QUOTIENT GROUPS 19

So by taking a representative from each of the two cosets and multiplying them
in a given order, we arrive at a coset which does not depend on which representatives
we have chosen. Therefore, when we decompose a group by a normal subgroup N
into a set of cosets, we can define a binary operation in the following way: if A = xN
and B = yN , then let A · B = (xy)N . The result of problem 106 shows that this
operation is well defined and does not depend on the choice of elements x and y,
which generate the classes A and B. So, in the example above, A · B = C.
Problems 107–109 deal with decompositions of a group G by a normal sub-
group N .
107. Let T1 , T2 and T3 be cosets of N . Prove that (T1 T2 )T3 = T1 (T2 T3 ).
108. Prove that N T = T N = T for every coset T of N .
109. Prove that for any coset T of N , there exists a coset T −1 such that T T −1 =
T −1 T = N .
From the statements of problems 107–109, it follows that the set of cosets with
the binary operation described above is a group. This group is called the quotient
group of G by the normal subgroup N and is denoted G/N . It is also sometimes
called the factor group of G by N .
Clearly, G/{e} ∼= G and G/G ∼ = {e}. It is also clear that in the finite case, the
order of the quotient group is equal to n/m, where n is the order of the group G
and m is the order of the normal subgroup N . For example, the quotient group of
the group of symmetries of a square by the subgroup of central symmetries has four
elements.
110. Determine whether the quotient group of the group of symmetries of a square
by the subgroup of central symmetries is isomorphic to the group of rotations of a
square or to the group of symmetries of a rhombus.
111. Find all normal subgroups and the corresponding quotient groups† by them in
the following groups: (a) the group of symmetries of a triangle, (b) C2 × C3 , (c) the
group of symmetries of a square, (d) the group of quaternions (see the solution of
problem 92).
112. Find all normal subgroups and quotient groups of the groups: (a) Cn , (b) C∞ .
113. Find all normal subgroups and quotient groups of the group of rotations of a
tetrahedron.
114. In the direct product of groups G1 × G2 , consider the subgroup G1 × {e}.
Prove this is a normal subgroup and that the quotient group by it is isomorphic to
the group G2 .
†
Subsequently, to find a quotient group means to point out some group, studied earlier, to
which the sought quotient group is isomorphic.
20 CHAPTER 1. GROUPS

§12. The commutator subgroup

Recall that two elements a and b of a group G are said to commute if ab = ba. The
extent to which two elements of a group do not commute can be measured using the
product aba−1 b−1 , which equals the identity if and only if a and b commute (prove
this).
Definition 14. The element aba−1 b−1 is called the commutator of the elements
a and b. The commutator subgroup K(G) of the group G is the set of all products
of a finite number of commutators in the group G.
115. Prove that the commutator subgroup is actually a subgroup.
116. Prove that the commutator subgroup is a normal subgroup.
117. Prove that the commutator subgroup equals the identity subgroup {e} if and
only if the group is commutative.
118. Find the commutator subgroup of the groups: (a) of symmetries of a triangle,
(b) of symmetries of a square, (c) of quaternions (see the solution of problem 92).
119. Prove that the commutator subgroup of the group of symmetries of a regular
n-gon is isomorphic to the group Cn for n odd and the group Cn/2 for n even.
120. Find the commutator subgroup of the group of rotations of a tetrahedron.
121. Prove that if a normal subgroup of the group of rotations or the group of
symmetries of a tetrahedron contains at least one rotation around an axis that goes
through a vertex, then it contains the entire group of rotations of a tetrahedron.
122. Find the commutator subgroup of the group of symmetries of a tetrahedron.
Let us examine two more groups, the group of rotations of a cube and the group
of rotations of a regular octahedron (fig. 7).

Figure 7.

123. How many elements are there in each of the above groups? List the elements
of the group of rotations of a cube.
124. Prove that the groups of rotations of a cube and an octahedron are isomorphic.
125. How many different ways can we paint the faces of a cube with six colours
(each face a different colour), if two paintings are considered different only if they
cannot be superimposed with one another by a rotation of a cube? Answer the same
question for a matchbox.
 13. HOMOMORPHISMS 21

126. To which of the groups we have already studied is the group of rotations of a
matchbox isomorphic?
The computation of the commutator subgroup of the group of rotations of a cube
can be done conveniently by inscribing a tetrahedron into the cube as shown in the
figure (fig. 8). If we connect the remaining vertices B, D, A1 and C1 , then we will

D1 C1

A1 B1

D
C

A B

Figure 8.
get a second tetrahedron. Any rotation of the cube either sends each tetrahedron to
itself, or swaps them.
127. Prove that the set of all rotations of the cube that send each tetrahedron to
itself forms: (a) a subgroup, (b) a normal subgroup of the group of rotations of a
cube.
128. Prove that the commutator subgroup of the group of rotations of a cube is
isomorphic to the group of rotations of a tetrahedron.
We now prove the following three properties of the commutator subgroup, which
we will need subsequently.
129. Prove that the quotient group of an arbitrary group G by its commutator
subgroup is commutative.
130. Suppose N is a normal subgroup of the group G and with the quotient group G/N
commutative. Prove that N contains the commutator subgroup of the group G.
131. Suppose N is a normal subgroup of the group G, and let K(N ) be the com-
mutator subgroup of the group N . Prove that K(N ) is a normal subgroup of G.
(Compare this with the example on page 18.)

§13. Homomorphisms
A homomorphism from the group G to the group F is a map ϕ : G → F such that
ϕ(ab) = ϕ(a) · ϕ(b) for any elements a and b of the group G. (Here the multiplica-
tion ab is performed in the group G and the multiplication ϕ(a) · ϕ(b) is performed
22 CHAPTER 1. GROUPS

in F .) A homomorphism is different from an isomorphism because a homomorphism

does not have to be bijective.
Example 12. Let G be the group of rotations of a cube and C2 be the group of
permutations of the two tetrahedra inscribed into it (see page 21). Every rotation
of the cube corresponds to a permutation of the tetrahedra; when two rotations
are performed successively, the corresponding permutations of the tetrahedra are
multiplied. This way, the map from the group of rotations of the cube to the group
of permutations of the tetrahedra is a homomorphism.
132. Suppose ϕ : G → F is a surjective homomorphism from the group G to the
group F . If the group G is commutative, then F is also commutative. Prove this.
Is the converse true?
133. Prove that a homomorphism from the group G to the group F sends the
identity of the group G to the identity of the group F .
134. Prove that ϕ(a−1 ) = (ϕ(a))−1 , where ϕ : G → F is a homomorphism, on the
left hand side the inverse is taken in the group G and on the right hand side the
inverse is taken in the group F .
135. Suppose that ϕ1 : G → F and ϕ2 : F → H are homomorphisms. Prove that
ϕ2 ϕ1 : G → H is a homomorphism.
We can obtain some important examples of homomorphisms by the following
construction of the natural homomorphism.
Let N be a normal subgroup of the group G. Consider the following map ϕ from
the group G to the quotient group G/N : let every element g in the group G be
mapped by ϕ to the coset T of N which contains g.
136. Prove that ϕ : G → G/N is a surjective homomorphism.
The map ϕ is called the natural homomorphism from the group G to the quotient
group G/N , and is surjective. We have thus shown that there is a homomorphism
associated to every normal subgroup. We will also show the converse, that every
surjective homomorphism from a group G to a group F can be viewed as a natural
homomorphism from G to the quotient group G/N for a suitable normal subgroup N .
If ϕ : G → F is a homomorphism, then the set of elements g such that ϕ(g) = eF
is called the kernel of the homomorphism ϕ, and is denoted by ker ϕ.
137. Prove that ker ϕ is a subgroup of the group G.
138. Prove that ker ϕ is a normal subgroup of the group G.
Let’s look at the decomposition of the group G with respect to the kernel ker ϕ.
139. Prove that g1 and g2 lie in the same coset if and only if ϕ(g1 ) = ϕ(g2 ).
 13. HOMOMORPHISMS 23

Theorem 3 (The First Isomorphism Theorem). Suppose ϕ : G → F is a surjective

homomorphism from the group G to the group F . Then the map ψ : G/ ker ϕ → F
which sends each coset of ker ϕ to the value of ϕ(g) for any element g in the coset is
an isomorphism. (This is well-defined by problem 139.)
The proof of this theorem is contained in the solutions to the following problems.
140. Prove that ψ is a surjective map.
141. Prove that ψ is a bijective map.
142. Prove that ψ is an isomorphism.
Let us give an example of an application of the above theorem.
Example 13. In problem 110, we had to determine whether the quotient group
of the symmetries of a square by the normal subgroup of central reflections was
isomorphic to the group of rotations of a square or the group of symmetries of a
rhombus. Every element of the group corresponds to a certain permutation of the
axes l1 , l2 , l3 , l4 (see figure 9). Note that the diagonals l1 and l3 can only be sent to

l4

l1 l2 l3

Figure 9.
each other, and the axes l2 and l4 can also only be sent to each other. In this way,
we get a map from the group of symmetries of a square to a group of permutations
of four elements: l1 , l2 , l3 and l4 . This map is a surjective homomorphism to the
group of permutations that sends l1 and l3 to l1 and l3 and sends l2 and l4 to l2 and l4
(check this). This group consists of four permutations and is isomorphic to the group
of symmetries of the rhombus L1 L2 L3 L4 (figure 10). The kernel of the constructed
L4

L1 L3

L2

Figure 10.
homomorphism consists of all symmetries of the square that send each one the four
axes to itself. It is easy to check that the only such transformations are e and the
central symmetry a. Therefore, by Theorem 3, the subgroup of reflections {e, a} with
respect to the centre is a normal subgroup in the group of symmetries of a square
and its quotient group is isomorphic to the group of symmetries of a rhombus.
24 CHAPTER 1. GROUPS

The following problems can be solved in a similar way.

143. Prove that the rotations of a tetrahedron by 180◦ around the axes that go
through the centres of opposite edges, together with the identity transformation,
form a normal subgroup of the group of symmetries of a tetrahedron, and find the
quotient group.
144. Prove that the rotations of a cube by 180◦ around axes going through the
centres of opposite faces, together with the identity transformation, form a normal
subgroup of the group of rotations of a cube, and find the quotient group.
145. Suppose we have a regular n-gon centred at O in the plane. Let R be the group
of all rotations of the plane around the point O. Let us look at the subgroup Cn of
all rotations of the plane that send the regular n-gon to itself. Prove that this is a
normal subgroup of the group R and that R/Cn ∼ = R.
146. Let N1 and N2 be normal subgroups of the groups G1 and G2 respectively.
Prove that N1 × N2 is a normal subgroup of G1 × G2 , and that G1 × G2 /N1 × N2 ∼
=
G1 /N1 × G2 /N2 .
147. Can two non-isomorphic groups have isomorphic normal subgroups with iso-
morphic quotient groups?
148. Can a group have two isomorphic normal subgroups with non-isomorphic quo-
tient groups?
149. Can a group have non-isomorphic normal subgroups with isomorphic quotient
groups?
Let us now consider what happens to subgroups, normal subgroups and commu-
tator subgroups under homomorphisms. If ϕ : G → F is a homomorphism and we
choose a subset M of G, then we call the set of all elements of F that have at least
one preimage in M the image of the set M under the homomorphism ϕ (denoted
by ϕ(M )). Conversely, if P is a subset of F , then we call the set of all elements
of G whose image is in P the preimage of the set P (denoted by ϕ−1 (P )). Note that
ϕ−1 by itself, without specifying P , does not have a meaning: in general, there does
not exist an inverse map for a homomorphism. Note also that if ϕ(M ) = P , then
ϕ−1 (P ) contains M but does not necessarily coincide with it (see figure 11).
150. Prove that under the homomorphism ϕ : G → F , the image of a subgroup H
of the group G is a subgroup of the group F .
151. Suppose H is a subgroup of F and ϕ : G → F is a homomorphism. Prove that
ϕ−1 (H) is a subgroup of G.
152. Suppose that N is a normal subgroup of F and ϕ : G → F is a homomorphism.
Prove that ϕ−1 (N ) is a normal subgroup of G.
 14. SOLVABLE GROUPS 25

ϕ−1 (P )

M
ϕ(M ) = P

ϕ
G F

Figure 11.

153. Suppose that ϕ : G → F is a homomorphism and let K1 and K2 be the

commutator subgroups of the groups G and F respectively. Prove that ϕ(K1 ) is a
subset of K2 and that K1 is a subset of ϕ−1 (K2 ).
154. Suppose that N is a normal subgroup of the group G and that ϕ is a surjective
homomorphism from the group G to the group F . Prove that ϕ(N ) is a normal
subgroup of the group F .
155. Let K1 and K2 be the commutator subgroups of the groups G and F respec-
tively and suppose that ϕ is a surjective homomorphism from the group G to the
group F . Prove that ϕ(K1 ) = K2 . Is it necessarily true that K1 = ϕ−1 (K2 )?

§14. Solvable groups

There exists an important class of groups, related to commutative groups: the so-
called solvable groups. They are called solvable, because, as we will see, the ability
to solve a polynomial equation in radicals depends on whether a certain group is
solvable.
Let G be a group and K(G) its commutator subgroup. The commutator sub-
group K(G) is itself a group and so it itself has a commutator subgroup K(K(G)).
That commutator subgroup also has a commutator subgroup, and so on. The group
K(K(. . . (K(G)) . . . )) with r Ks will be denoted by Kr (G), so Kr+1 (G) = K(Kr (G)).
Definition 15. A group G is called solvable if the chain of groups G, K(G), K2 (G),
K3 (G), . . . ends for some finite n with the identity group, i.e., if for some n we have
Kn (G) = {e}.
For example, any commutative group is solvable, because if G is a commutative
group, then at the first step we get K(G) = {e} (see problem 117). Similarly, a group
G is solvable if its commutator subgroup is commutative, because then K2 (G) = {e}.
156. Determine whether or not the following groups are solvable: (a) the cyclic
group Cn , (b) the group of symmetries of a triangle, (c) the group of quaternions
(see the solution to problem 92), (d) the group of rotations of a tetrahedron, (e) the
group of symmetries of a tetrahedron, (f) the group of rotations of a cube.
26 CHAPTER 1. GROUPS

We will be exploring the solvability of other groups later. In particular, we need to

show that the group of rotations of a regular dodecahedron (figure 12) is non-solvable.

Figure 12.

157. How many elements are there in the group of rotations of a dodecahedron?
The rotations of a dodecahedron can be broken down into four classes: (1) the
identity transformation; (2) rotations around axes going through the centres of op-
posite faces; (3) rotations around axes going through opposite vertices; (4) rotations
around axes going through the centres of opposite edges.
158. How many elements are there in each class? (The identity transformation does
not belong to classes (2)–(4).)
159. Suppose that N is a normal subgroup in the group of rotations of a dodecahe-
dron, and suppose that N has some element of one of the four classes (1)–(4). Prove
that N then contains that entire class.
So each of the classes (1)–(4) either is entirely in N or entirely not in N .
160. Prove that the group of rotations of a dodecahedron has no normal subgroups
besides {e} and the entire group.
161. Let a group G be non-commutative and have no normal subgroups besides {e}
and G. Prove that the group G is non-solvable.
From problems 160 and 161, it follows that the group of rotations of a dodeca-
hedron is non-solvable.
Let us look at several more problems whose results we will need later.
162. Prove that any subgroup of a solvable group is solvable.
163. Let ϕ : G → F be a surjective homomorphism from the group G to the group F
and suppose G is solvable; prove that the group F is also solvable.
164. In the setup of the previous problem, give an example where the group F is
solvable, while the group G is non-solvable.
 15. PERMUTATIONS 27

165. Suppose that the group G is solvable and that N is a normal subgroup of G.
Prove that the quotient group G/N is solvable.

166. Suppose that N is a normal subgroup of the group G. Prove that if both N
and G/N are solvable, then the group G is also solvable.

167. Suppose that the groups G and F are solvable. Prove that the group G × F is
also solvable.

168. Suppose that the group G is solvable. Prove that there exists a chain of
subgroups G0 , G1 , . . . , Gn of G such that: (1) G0 = G; (2) each group Gi with
1 6 i 6 n is a normal subgroup of the group Gi−1 , and each of the quotient groups
Gi−1 /Gi is abelian; (3) the group Gn is abelian.

169. Suppose that for a group G there exists a chain of subgroups with the properties
described in the previous problem. Prove that the group G is solvable.

The results of problems 168 and 169 show that, for a group G, the existence of a
chain of subgroups as described in problem 168 is equivalent to the solvability of G
and so could be taken as the definition of solvability. Another equivalent definition
can be derived from the results of the following two problems.

170. Suppose that the group G is solvable. Prove that there exists a chain of groups
G0 , G1 , . . . , Gn such that: (1) G0 = G; (2) each group Gi with 1 6 i 6 n−1 contains
a commutative normal subgroup Ni such that Gi /Ni ∼ = Gi+1 ; (3) the group Gn is
commutative.

171. Suppose that for a group G there exists a chain of groups with the properties
described in the previous problem. Prove that the group G is solvable.

§15. Permutations
Let us look now in more detail at permutations (i.e., transformations) on the set of
the first n positive integers 1, 2, . . . , n; we will call these permutations permutations
of degree n. Note that permutations of an arbitrary set with n elements can be
reduced to the permutations described above by just numbering the elements of
that set by positive integers
1, 2, . . . , n. An arbitrary permutation of degree n
can be written as i11 i22 ... n
... in , where im is the image of the element m under this
permutation. Recall that a permutation is a bijective map, so all the elements in the
second row are distinct.

172. How many different permutations of degree n are there?

28 CHAPTER 1. GROUPS

Definition 16. The group of all permutations of degree n with the usual operation
of multiplication (i.e., composition)† is called the symmetric group of degree n and
is denoted by Sn .
173. Prove that for n > 3, the group Sn is non-commutative.
A permutation can move some elements and leave others fixed; it can turn out
that the moving elements are moving as if in a circle. For example, the permutation
 
1 2 3 4 5 6 7
4 2 6 3 5 1 7
leaves the elements 2, 5 and 7 fixed, and moves the rest of the elements in a circle:
1 → 4, 4 → 3, 3 → 6, 6 → 1. Permutations of this sort are called cyclic permutations
or simply cycles. For cyclic permutations, we will use also a different notation: the
expression (1 4 3 6) will mean the permutation that sends 1 → 4, 4 → 3, 3 → 6,
6 → 1 and leaves the rest of the elements fixed. So, if this is a permutation of
degree 7, then it coincides with the permutation described above.
Not all permutations are cyclic. For example, the permutation
 
1 2 3 4 5 6
3 5 4 1 2 6
is not cyclic, but it can be expressed as a product of two cycles:
 
1 2 3 4 5 6
= (1 3 4) · (2 5).
3 5 4 1 2 6
The two cycles move different elements, and such cycles are called independent. It is
easy to see that the product of independent cycles does not depend on the order in
which they are performed. If we do not distinguish products of independent cycles
differing only in the order in which they are performed, then the following statement
will be true.
174. Any permutation splits uniquely (up to the order of the terms) into several
independent cycles. Prove this.
Cycles of the form (i j) that move only two elements are called transpositions.
175. Prove that any cycle can be split into a product of transpositions (not neces-
sarily independent).
The transpositions (1 2), (2 3), . . . , (n−1, n) are called elementary transpositions.
176. Prove that an arbitrary transposition can be expressed as a product of elemen-
tary transpositions.
†
With our definitions, we will view products of transformations (see page 5) as being performed
from right to left. Sometimes products of transformations are considered as being performed from
left to right. The groups produced in the two cases are isomorphic.
 15. PERMUTATIONS 29

From the results of problems 174–176, it follows that an arbitrary permutation

of degree n can be expressed as a product of elementary transpositions. In other
words, the following theorem is true.
Theorem 4. If a subgroup of the symmetric group Sn contains all elementary trans-
positions, then this subgroup coincides with the entire group Sn .
Let the numbers 1, 2, . . . , n be written in a line in an arbitrary order. We say
that a pair of numbers i, j forms an inversion in this line if i < j but j is earlier in
the line than i. The number of inversions characterises the disorder of the line with
respect to the usual order of numbers 1, 2, . . . , n.
177. Find the number of inversions in the line 3, 2, 5, 4, 1.
Subsequently we will be interested not in the actual number of inversions, but in
its parity.
178. Prove that the parity of the number of inversions in a line changes if two
arbitrary numbers switch places.


Definition 17. A permutation i11 i22 ...
... n
in is called even or odd according to whether
the number of inversions in the second line is even or odd.


For example, the identity permutation 11 22 ... n
... n is an even permutation, because
the number of inversions in the second line is equal to zero.


179. Find the parity of the permutation 12 25 34 41 53 .
180. Prove that when an even permutation is multiplied by a transposition, either
on the left or the right, we get an odd permutation, and, vice versa, when an odd
permutation is multiplied by a transposition we get an even permutation.
181. Prove that an even permutation can only be split into a product of an even
number of transpositions, and an odd permutation can only be split into a product
of an odd number of transpositions.
182. Find the parity of an arbitrary cycle of length: (a) 3, (b) 4, (c) m.
183. Prove that when two permutations of the same parity are multiplied, we get
an even permutation, and when two permutations of different parities are multiplied,
we get an odd permutation.
184. Prove that the permutations a and a−1 have the same parity, where a is an
arbitrary permutation.
From the results of problems 183 and 184, it follows that the set of all even
permutations forms a subgroup of the group Sn .
Definition 18. The group of all even permutations of degree n is called the alter-
nating group of degree n and is denoted by An .
30 CHAPTER 1. GROUPS

185. Prove that for n > 4, the group An is non-commutative. (Compare prob-
lem 173.)
186. Prove that the alternating group An is a normal subgroup of the group Sn , and
construct the coset decomposition of the group Sn with respect to the subgroup An .
187. Find the number of elements in the group An .
188. Prove that the groups S2 , S3 and S4 are solvable.
Let us prove now that the alternating group A5 is non-solvable. One of the
methods for proving this consists of the following. Inscribe five tetrahedra numbered
1, 2, 3, 4 and 5 into a dodecahedron. Then we can check that every rotation of
the dodecahedron corresponds to an even permutation of the tetrahedra and that
different rotations correspond to different permutations. In this way, we establish
an isomorphism between the group of rotations of a dodecahedron and the group A5
of even permutations of degree five. Then the non-solvability of the group A5 will
follow from the non-solvability of the group of rotations of a dodecahedron.
189. Inscribe five tetrahedra into a dodecahedron using the method described above.
Another method of proving the non-solvability of the group A5 amounts to a
repetition of the idea in the proof of the non-solvability of the group of rotations of
a dodecahedron. For that, the following problems need to be solved.
190. Prove that any non-identity even permutation of degree five can be split into
independent cycles in precisely one of the following three ways: (1) (i1 i2 i3 i4 i5 ),
(2) (i1 i2 i3 ), (3) (i1 i2 )(i3 i4 ).
191. Suppose that N is a normal subgroup of the group A5 . Prove that if N contains
at least one permutation that splits into independent cycles in one of the ways
described in the previous problem, then N contains all permutations that split in
the same way into independent cycles.
192. Prove that the group A5 has no normal subgroups except for the identity
subgroup and the entire group.
From the statements of problems 192 and 161, and from the fact that the group A5
is non-commutative, it follows that the group A5 is non-solvable.
193. Prove that the symmetric group Sn contains a subgroup isomorphic to the
group A5 for n > 5.
From the statements of problems 193 and 162 we get the following theorem.
Theorem 5. For n > 5, the symmetric group Sn is non-solvable.
The above theorem, as well as other results from this chapter, will be needed
in the next chapter to prove the non-solvability in radicals of general polynomial
equations of degree greater than four.
 CHAPTER 2

Complex numbers

As we progress through secondary school, we gradually expand the set of numbers

that we study. In doing this, the fact most stressed is that such expansion allows us
to operate on the numbers more freely. So in moving from the natural numbers to
the integers, it becomes possible to subtract any numbers; in moving to the rationals,
it becomes possible to divide any numbers, and so on. A more significant feature of
such expansions turns out to be the fact that the properties of the expanded systems
often allow us to get new results about the original systems. For example, many
difficult problems of number theory relating only to positive integers were solved
using real and complex numbers.
Historically, complex numbers appeared specifically to solve certain problems
about the real numbers. For example, the Italian mathematician Cardano (1501–
1576), when solving cubic equations, found the real roots of the equation using
intermediate “unreal” square roots of negative numbers.
Over time, complex numbers began taking on a more prominent position in math-
ematics and its applications. First of all, they deeply penetrated into the theory of
polynomial equations, because over the complex numbers, the study of such equa-
tions turned out to be much more convenient. For example, any polynomial equation
of degree n (n > 1) with real or complex coefficients has at least one complex root
(see “the fundamental theorem of algebra” below, page 55). On the other hand, not
every polynomial equation with real coefficients has at least one real root.
After an interpretation of the complex numbers appeared using points in a plane
and vectors in a plane, it became possible to apply geometric concepts such as
continuity and geometric transformations to the study of the complex numbers. The
connection between complex numbers and vectors allowed the reduction of many
problems in mechanics, especially in hydro- and aerodynamics, as well as the theory
of electricity, the theory of heat, etc., to problems about complex numbers and
equations in them.
The study of complex numbers has now developed into a large and important
branch of modern mathematics known as the theory of functions of a complex vari-
able.
The reader can expect to gain a reasonably deep understanding of complex num-
bers and functions of a complex variable from this chapter.

31
32 CHAPTER 2. COMPLEX NUMBERS

§1. Fields and polynomials

Real numbers can be summed and multiplied, and it is also possible to perform the
inverse operations, subtraction and division. In sums, we can arbitrarily switch sum-
mands and arbitrarily distribute parentheses. We can treat the terms of a product
in the same way. All of these properties, as well as the connection between addition
and multiplication, can be expressed concisely in the following way.
The real numbers have the following three properties:
(1) They form an abelian group (see chapter 1, section 3) under addition; the
identity element of this group is denoted by 0 and is called the zero.
(2) If we ignore 0, then the rest of the numbers form an abelian group under
multiplication.
(3) Addition and multiplication are connected through the distributive law: for
any numbers a, b and c,
a(b + c) = ab + ac.
Having these three properties is very important, because they allow us to simplify
arithmetic and algebraic expressions and to solve many equations. The set of real
numbers is not the only set that has these three properties. We introduce a special
term to describe such sets.
Definition 19. If on a set we have two binary operations defined, called addition
and multiplication, satisfying the three properties described above, then such a set
is called a field .
194. Determine whether the following subsets of the real numbers under the usual
operations of addition and multiplication are fields: (a) all positive integers;
√ (b) all
integers; (c) all rational numbers; (d) all numbers of the form r1 + r2 2, where r1
and r2 are arbitrary rational numbers.
195. Prove that in any field we have a · 0 = 0 · a = 0 for any element a in the field.
196. Prove that in any field we have (−a)·b = a·(−b) = −(a·b) and (−a)·(−b) = ab
for any elements a and b in the field.
197. Suppose that a and b are elements of an arbitrary field, with a · b = 0. Prove
that a = 0 or b = 0.
Example 14. Consider the set {0, 1, . . . , n − 1} of remainders modulo n, with the
operations of addition modulo n (see example 9 on page 9) and multiplication mod-
ulo n, where the result of the multiplication of two numbers is the remainder when
their product is divided by n.
198. Construct the multiplication tables for multiplication modulo 2, 3 and 4.
199. Prove that this set of remainders under the operations of addition and multi-
plication modulo n forms a field if and only if n is a prime.
 1. FIELDS AND POLYNOMIALS 33

Definition 20. The difference of elements b and a in an arbitrary field (denoted by

b − a) is the solution to the equation x + a = b (or to a + x = b). If a 6= 0, then
the quotient of the element b by a (denoted by b/a) is the solution to the equation
ya = b (or to ay = b).
From the result of problem 24 and the fact that in a field addition and multipli-
cation are commutative, it follows that the elements b − a and b/a (with a 6= 0) are
defined uniquely in any field.
Because a field is a group under addition and, without the zero, is a group under
multiplication, the equality x + a = b is equivalent to the equality x = b + (−a),
and the equality ya = b with a 6= 0 is equivalent to the equality y = ba−1 . Thus,
b − a = b + (−a) and b/a = ba−1 .
The reader can easily prove that the operations of addition, subtraction, mul-
tiplication and division have all of the basic properties of these operations in the
field of real numbers. In particular, in any field both sides of any equality can be
multiplied or divided by any non-zero element, any term can be moved from one
side of an equality to the other with the opposite sign, etc. As an example, let us
consider the distributive property which connects subtraction and multiplication.
200. Prove that in any field, (a − b)c = ac − bc for any elements a, b, and c in the
field.
If K is a field, then we can consider polynomials with coefficients from the field
K, also known as polynomials over the field K, in the same way as we consider
polynomials over the field of real numbers.
Definition 21. Let n be a positive integer. A polynomial of degree n in one vari-
able x over the field K is any expression of the form
an xn + an−1 xn−1 + · · · + a1 x + a0 , (1.1)
where an , an−1 , . . . , a0 are elements of the field K, and an 6= 0. If a is an element of
the field K, then the expression a is also considered a polynomial over the field K;
if a 6= 0, then this is a polynomial of degree 0, while if a = 0, then the degree of the
polynomial is undefined.
The elements an , an−1 , . . . , a0 are called the coefficients of the polynomial (1.1),
and an is called the leading coefficient.
Two polynomials in the variable x are considered equal if and only if all of their
corresponding coefficients are pairwise equal.
Let
P (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 .
If we plug in some element a of the field K in the right hand side of this equality
in place of x and perform the necessary operations, taking the operations of addition
and multiplication as operations in the field K, then as a result we will get some
element b of the field K. Then we write P (a) = b. If P (a) = 0, where 0 is the zero
34 CHAPTER 2. COMPLEX NUMBERS

element of the field K, then we say that a is a root of the equation P (x) = 0; in this
case we also say that a is a root of the polynomial P (x).
Polynomials over an arbitrary field K can be added, subtracted and multiplied.
The sum of the polynomials P (x) and Q(x) is the polynomial R(x) where the
coefficient of xk in R(x) (k = 0, 1, 2, . . . ) is equal to the sum (in the field K)
of the coefficients of xk in the polynomials P (x) and Q(x). The difference of two
polynomials is defined in an analogous way. Obviously, the degree of the sum or the
difference is no larger than the greater of the degrees of the given polynomials.
To calculate the product of the polynomials P (x) and Q(x), we need multiply
each term axk of the polynomial P (x) by each term bxl of the polynomial Q(x) using
the rule: axk · bxl = abxk+l , where ab is a product in the field K and k + l is the
usual sum of non-negative integers. After this, all of the resulting expressions have
to be added and like terms collected, that is, all terms involving the same degree r
of the variable x are collected, and the sum d1 xr + d2 xr + · · · + ds xr is changed to
the expression (d1 + d2 + · · · + ds )xr .
If
P (x) = an xn + an−1 xn−1 + an−2 xn−2 + · · · + a0
and
Q(x) = bm xm + bm−1 xm−1 + bm−2 xm−2 + · · · + b0 ,
then†
P (x) · Q(x) = an bm xn+m + (an bm−1 + an−1 bm )xn+m−1 +
(an bm−2 + an−1 bm−1 + an−2 bm )xn+m−2 + · · · +
(a1 b0 + a0 b1 )x + a0 b0 .
Because a0 6= 0 and b0 6= 0, we have a0 b0 6= 0 (see problem 197), so the degree of
the polynomial P (x) · Q(x) is equal to n + m, i.e., the degree of the product of two
non-zero polynomials is equal to the sum of the degrees of the given polynomials.
Using the fact that the operations of addition and multiplication of elements in
a field K have the properties of commutativity, associativity and distributivity, it is
easy to show that the operations we have introduced, of addition and multiplication of
polynomials over the field K, also have the properties of commutativity, associativity
and distributivity.
If
P (x) + Q(x) = R1 (x), P (x) − Q(x) = R2 (x), P (x) · Q(x) = R3 (x),
and a is an arbitrary element of the field K, then it is also easy to show that
P (a) + Q(a) = R1 (a), P (a) − Q(a) = R2 (a), P (a) · Q(a) = R3 (a).

The coefficient of xk in the product P (x) · Q(x) is equal to a0 bk + a1 bk−1 + · · · + ak−1 b1 + ak b0 ,

†

where we take ai = 0 for i > n and bj = 0 for j > m.

 1. FIELDS AND POLYNOMIALS 35

Polynomials over an arbitrary field K can be divided by each other with a re-
mainder. To divide a polynomial P (x) by the polynomial Q(x) with a remainder
means to find polynomials S(x) (the quotient) and R(x) (the remainder) such that
P (x) = S(x) · Q(x) + R(x),
and either the degree of the polynomial R(x) is less than the degree of the polyno-
mial Q(x), or R(x) = 0.
Let P (x) and Q(x) be arbitrary polynomials over the field K, with Q(x) 6= 0.
We show that we can divide the polynomial P (x) by the polynomial Q(x) with a
remainder.
Suppose that
P (x) = an xn + an−1 xn−1 + an−2 xn−2 + · · · + a0
and
Q(x) = bm xm + bm−1 xm−1 + bn−2 xm−2 + · · · + b0 .
If n < m, then letting S(x) = 0 and R(x) = P (x) we get the required quotient and
remainder. If n > m, then let us look at the polynomial
an n−m
P (x) − x Q(x) = R1 (x).
bm
The polynomial R1 (x) does not contain an xn term, so either its degree is no more
than n − 1 or R1 (x) = 0. If
R1 (x) = ck xk + ck−1
k−1 + · · · + c0
and k > m, then we look at the polynomial
ck k−m
R1 (x) − x Q(x) = R2 (x),
bm
and so on. Because the degree of the resulting polynomial is strictly less than the
degree of the previous polynomial, this process will have to end, i.e., at some step
we will get
dl l−m
Rs−1 (x) − x Q(x) = Rs (x),
bm
where the degree of the polynomial Rs (x) is less than m, or Rs (x) = 0. Then we get
an n−m
P (x) = x Q(x) + R1 (x)
bm
an n−m ck k−m
= x Q(x) + x Q(x) + R2 (x)
bm bm
= ···
an n−m ck k−m dl l−m
= x Q(x) + x Q(x) + · · · + x Q(x) + Rs (x)
bm bm bm
a ck k−m dl l−m 
n n−m
= x + x + ··· + x · Q(x) + Rs (x).
bm bm bm
36 CHAPTER 2. COMPLEX NUMBERS

Here, the expression in the parentheses is the quotient of the division of the poly-
nomial P (x) by Q(x) and Rs (x) is the remainder. The process of division of a
polynomial by another polynomial described here is basically the process of long
division.
The following problem shows that if P (x) and Q(x) are two polynomials with
Q(x) 6= 0, then, no matter what method we use for dividing P (x) by Q(x) with a
remainder, the quotient and the remainder are defined uniquely.
201. Suppose that
P (x) = S1 (x) · Q(x) + R1 (x),
and also that
P (x) = S2 (x) · Q(x) + R2 (x),
where the degrees of polynomials R1 (x) and R2 (x) are less than the degree of the
polynomial Q(x) (allowing R1 (x) = 0 or R2 (x) = 0). Prove that
S1 (x) = S2 (x), R1 (x) = R2 (x).

§2. The field of complex numbers

From the solution to problem 194, it follows that there exist fields smaller than
the field of real numbers, for example the field of rational numbers. We will now
construct a field larger than the field of real numbers, namely the field of complex
numbers.
Let us look at all ordered pairs of real numbers, i.e., pairs of the form (a, b),
where a and b are arbitrary real numbers. We say that (a, b) = (c, d) if and only if
a = c and b = d. On the set of all such pairs we define two operations, addition and
multiplication, in the following way:
(a, b) + (c, d) = (a + c, b + d), (2.1)
(a, b) · (c, d) = (ac − bd, ad + bc). (2.2)
(In the terms on the right hand sides of the equalities, the operations are the usual
ones on real numbers.) For example, we have
√ √ √
( 2, 3) + ( 2, −1) = (2 2, 2),
(0, 1) · (0, 1) = (−1, 0).
Definition 22. The set of all possible ordered pairs of real numbers, with the oper-
ations of addition and multiplication defined by (2.1) and (2.2), is called the set of
complex numbers.
From this definition it is clear that in the set of complex numbers there isn’t
anything “supernatural”: complex numbers are just pairs of real numbers. However,
one may question whether it is right to call such objects numbers. We will discuss
 2. THE FIELD OF COMPLEX NUMBERS 37

this question at the end of this section. Another question that the reader may have
is why we defined the operations of addition and multiplication of complex numbers,
especially the odd-looking multiplication, in this way and not in some other way. We
will answer this question in section 3.
We will now explore useful properties possessed by the complex numbers.
202. Prove that the complex numbers form an abelian group under addition. Which
complex number is the identity element (the zero) of this group?
Subsequently, it will be convenient to denote complex numbers with a single
letter, for example z or w.
203. Prove that the operation of multiplication of complex numbers is commutative
and associative, i.e., z1 · z2 = z2 · z1 and (z1 · z2 ) · z3 = z1 · (z2 · z3 ) for any complex
numbers z1 , z2 and z3 .
It is easy to check that
(a, b) · (1, 0) = (1, 0) · (a, b) = (a, b)
for any complex number (a, b). Therefore, the complex number (1, 0) is an identity
element in the set of complex numbers under multiplication.
204. Let z be an arbitrary complex number with z 6= (0, 0). Prove that there exists
a complex number z −1 such that
z · z −1 = z −1 · z = (1, 0).
The results of problems 203 and 204 show that the non-zero complex numbers
form an abelian group under multiplication.
205. Prove that the operations of addition and multiplication satisfy the distributive
law, i.e., (z1 + z2 ) · z3 = z1 · z3 + z2 · z3 for any complex numbers z1 , z2 and z3 .
It follows from the results of problems 202–205 that the complex numbers together
with the operations of addition and multiplication defined in (2.1) and (2.2) form a
field. This is called the field of complex numbers.
For complex numbers of the form (a, 0), where a is an arbitrary real number,
equations (2.1) and (2.2) give
(a, 0) + (b, 0) = (a + b, 0),
(a, 0) · (b, 0) = (a · b, 0).
Thus, if we make a correspondence between each complex number of the form (a, 0)
and the real number a, then the operations on numbers of the form (a, 0) will cor-
respond to the usual operations on real numbers. Therefore, we will simply identify
the complex number (a, 0) with the real number a,† and we will say that the field of
complex numbers contains the field of real numbers.
†
In the same way, for example, that the rational number n/1 is identified with the integer n.
38 CHAPTER 2. COMPLEX NUMBERS

The complex number (0, 1) is not real (with our identification), and we will denote
it by i, i.e., i = (0, 1). Because the field of complex numbers contains all real numbers
and the number i, it will also contain all numbers of the form b · i and a + b · i, where
a and b are arbitrary real numbers, and the operations of addition and multiplication
are understood as operations on complex numbers.
206. Let (a, b) be a complex number. Prove that
(a, b) = a + b · i.
From the result of problem 206, it is clear that we have a + bi = c + di if and
only if a = c and b = d.
In this way, any complex number can be expressed in a unique way in the form
a + bi, where a and b are real numbers. If z = a + bi, then, following historical
tradition, we call a the real part of the complex number z, we call bi the imaginary
part, and b is called the coefficient of the imaginary part.
We call the expression of a complex number z in the form z = a + bi the algebraic
form of the complex number z.
For complex numbers in algebraic form, formulæ (2.1) and (2.2) can be written
in the following way:
(a + bi) + (c + di) = (a + c) + (b + d)i, (2.3)
(a + bi) · (c + di) = (ac − bd) + (ad + bc)i. (2.4)
207. Solve the equation (find the formula for a difference):
(a + bi) + (x + yi) = (c + di).
208. Solve the equation (find the formula for a quotient):
(a + bi) · (x + yi) = (c + di), where a + bi 6= 0.
It is easy to check that
i · i = (0, 1) · (0, 1) = (−1, 0) = −1,
2
i.e., i = −1. Hence, in the field of complex numbers, we can take square roots of
some negative real numbers.
209. Calculate: (a) i3 , (b) i4 , (c) in .
210. Find all complex numbers z = x + yi such that: (a) z 2 = 1, (b) z 2 = −1,
(c) z 2 = a2 , (d) z 2 = −a2 (where a is some real number).
Definition 23. The complex number a − bi is called the complex conjugate of the
complex number z = a + bi, and is denoted by z̄. It is easy to check that
z + z̄ = 2a, z · z̄ = a2 + b2 .
211. Let z1 and z2 be arbitrary complex numbers. Prove the following identities:
(a) z1 + z2 = z̄1 + z̄2 , (b) z1 − z2 = z̄1 − z̄2 , (c) z1 · z2 = z̄1 · z̄2 , (d) z1 /z2 = z̄1 /z̄2 if
z2 is non-zero.
 3. THE UNIQUENESS OF THE FIELD OF COMPLEX NUMBERS 39

212. Let
P (z) = an z n + an−1 z n−1 + · · · + a1 z + a0 ,
with z a complex number and all the ai real numbers. Prove that P (z) = P (z̄).
Moving to the complex numbers is the next step in the sequence: positive
integers–integers–rational numbers–real numbers. However, the reader may be of
the opinion that real numbers are in fact numbers, while complex numbers are not
numbers but objects of more a complicated nature. Of course, any terminology can
be accepted, but in reality, complex numbers do deserve to be called numbers.
The first objection may be that they are not numbers, but pairs of numbers.
Remember, however, that this is the same way that the rational numbers are formally
introduced. A rational number is a class of equivalent fractions, and fractions are
pairs of integers written in the form m/n (where n 6= 0), where the operations on
these rational numbers are simply operations on pairs of integers. Therefore this
first objection is not well-founded. Another objection may be that numbers are
something with which we can measure. If we take this to mean that numbers are
what we can use to measure anything we want, then we should forbid, for example,
negative numbers, because there are no segments of length −3 cm, and a train cannot
travel for −4 days. If we consider numbers as something that can be used (or are
convenient) for measuring at least something, then complex numbers turn out to
be no worse than other numbers; they can be used to conveniently write down, for
example, the current and resistance in electrical circuits of alternating current, and
this technique is widely used in electronics.
Thus, in moving from real numbers to complex numbers is as natural as, for
example, the transition from integers to rationals.

§3. The uniqueness of the field of complex numbers

We now move on to consider the question of why we defined the complex numbers
in exactly the way we did, and not in another way. The answer to the question is
that we wanted to get a field that was an extension of the field of real numbers. Is
it possible to construct a different field which is also an extension of the field of real
numbers? We will answer this very question in this section.
Definition 24. An isomorphic map (or simply an isomorphism) from one field to
another is a bijective map ϕ which is an isomorphism with respect to both addition
and multiplication, i.e., ϕ(a + b) = ϕ(a) + ϕ(b) and ϕ(ab) = ϕ(a)ϕ(b) for all a, b
in the first field. Fields between which we can define an isomorphism are called
isomorphic.
If we only study the operations of addition and multiplication in the field, then
in isomorphic fields, all essential properties turn out to be the same. Therefore, in
the same way as with groups, we will not distinguish between isomorphic fields.
40 CHAPTER 2. COMPLEX NUMBERS

As we saw in the previous section, in the field of complex numbers there is an

element i such that i2 = −1. The following problem shows that adjoining such an
element to the field of real numbers necessarily gives the field of complex numbers.
213. Suppose that M is a field containing both the field of real numbers and an
element i0 satisfying i20 = −1. Prove that M contains a field M 0 that is isomorphic
to the field of complex numbers.
We will say that a field is a minimal field with given properties if it has the given
properties and does not contain any smaller field with the same properties.
For example, the result of problem 213 can be stated the following way: the
minimal field that contains the field of real numbers and an element i0 such that
i20 = −1 is the field of complex numbers. This proves, in some sense, the uniqueness
of the field of complex numbers. However, there exists a stronger result. Namely,
let’s remove the requirement that M needs to contain an element i0 such that i20 = −1
and pose the problem of finding all fields that are minimal non-trivial extensions of
the field of real numbers. It turns out that there are only two such extensions (up to
isomorphism), and one of them is the field of complex numbers. We now show this.
Suppose that a field M contains the field of real numbers, i.e., M contains all real
numbers and operations on them in the field M coincide with the usual operations
on real numbers. Let the field M also contain an element j, distinct from all real
numbers. Then for any real numbers an−1 , an−2 , . . . , a0 there exists an element in M
equal to
j n + an−1 j n−1 + an−2 j n−2 + · · · + a0 . (3.1)
We will call n the degree of the expression (3.1).
There are two possible situations:
(1) there exists an expression of the form (3.1) with n > 1 that gives an element
equal to 0;
(2) there does not exist an expression of the form (3.1) with n > 1 that gives
an element equal to 0.
Suppose first that we have situation (1).
Definition 25. A polynomial with coefficients from a field K is called reducible over
the field K if it can be expressed as a product of two polynomials of smaller degree
with coefficients in K. In the other case it is called irreducible over the field K.†
For example, the polynomials x3 − 1 and x2 − x − 1 are reducible over
√

the field
√

of
3 2 2 1+ 5 1− 5
real numbers, because x −1 = (x−1)(x +x+1) and x −x−1 = x− 2 x− 2 ,
2 2
whereas the polynomials x + 1 and x + x + 1 are irreducible over the field of real
numbers. Clearly, polynomials of degree one over any field are irreducible.

†
Irreducible polynomials over the field K are analogous to prime numbers in the set of integers.
 4. GEOMETRIC REPRESENTATION OF COMPLEX NUMBERS 41

214. We can select, from among all expressions of the form (3.1) which are equal
to 0, an expression with the smallest degree n (n > 1). Suppose this expression is
j n + an−1 j n−1 + an−2 j n−2 + · · · + a0 = 0.
Prove that the polynomial
xn + an−1 xn−1 + an−2 xn−2 + · · · + a0
is irreducible over the field of real numbers.
Later, we will show that any polynomial of degree greater than two with real
coefficients is reducible over the field of real numbers (see problem 272). Therefore
the smallest degree n in problem 214 has to be no greater than 2. And since n 6= 1
(or we would have j + a = 0, so that j is equal to the real number −a, contrary to
our assumption), we must have n = 2.
Thus, in situation (1), the following equality has to be true for some real numbers
p and q in the field M :
j 2 + pj + q = 0,
where the polynomial x2 + px + q is irreducible over the field of real numbers.
215. Prove that in situation (1), the field M contains an element i0 such that i20 =
−1.
From the results of problems 215 and 213, it follows that in situation (1), the
field M contains a field M 0 , isomorphic to the field of complex numbers. Therefore,
if the field M is a minimal non-trivial extension of the field of real numbers, then
the field M must coincide with the field M 0 . Thus, in situation (1), any minimal
field which is a non-trivial extension of the field of real numbers coincides with (i.e.,
is isomorphic to) the field of complex numbers. So, in situation (1), there exists a
unique field, up to isomorphism, which is a minimal non-trivial extension of the field
of real numbers, namely the field of complex numbers.
216. The field of rational functions in the variable x over the reals consists of all
elements of the form P (x)/Q(x), where P (x) and Q(x) are polynomials over the
reals, with Q(x) 6= 0. Prove that if a field M is an extension of the field of real
numbers, and M is in situation (2) above, then M contains a subfield isomorphic to
the field of rational functions in the variable x over the reals.

§4. Geometric representation of complex numbers

We introduce a rectangular coordinate system XOY , and associate with each com-
plex number a + bi the point (a, b) in the coordinate plane. We thus get a bijective
map between all complex numbers and all points of the plane. This gives us the first
geometric representation of the complex numbers.
42 CHAPTER 2. COMPLEX NUMBERS

217. Which complex numbers correspond to the points in figure 13?

y

B 3

2 A

−2 −1 0 1 2 x
C −1

−2

Figure 13.

218. Let complex numbers be represented by points in the plane. What is the
geometric meaning of the map ϕ, if for any complex number z: (a) ϕ(z) = −z,
(b) ϕ(z) = 2z, (c) ϕ(z) = z̄ (where z̄ is the complex conjugate of z).
Let A = (xA , yA ) and B = (xB , yB ) be two points in the plane (see figure 14).
−→
The segment AB with the noted direction from A to B is called the vector AB. The
y

yB B

yA A
xB xA x

Figure 14.
−→
coordinates of the vector AB are, by definition, computed in the following way:
x−→ = xB − xA , y−→ = yB − yA .
AB AB
Two vectors are considered equal if they are parallel, pointing in the same direction
and equal in length.
 5. TRIGONOMETRIC REPRESENTATION OF COMPLEX NUMBERS 43

219. Prove that two vectors are equal if and only if their coordinates are equal.
Equal vectors are usually considered to be the same one vector, which is char-
acterised only by its coordinates—this is the so called free vector . If we create a
correspondence such that to each complex number a + bi we associate a free vector
with coordinates (a, b), we get the second geometric representation of the complex
numbers.
220. Suppose that the complex numbers z1 , z2 and z3 correspond to the free vectors
u, v and w. Prove that z1 + z2 = z3 if and only if u + v = w, where the sum of
vectors is calculated using the parallelogram rule.
221. Prove the following relationship between the two geometric representations of
the complex numbers: if zA , zB , and z−→ are the complex numbers corresponding to
AB
−→
the points A and B and the vector AB, then z− → = zB − zA .
AB

From the definition of equal vectors, we see that equal vectors have the same
length. This length is also taken to be the length of the corresponding free vector.
Definition 26. The modulus of a complex number z (denoted by |z|) is defined to
be the length of the corresponding free vector.†
222. Suppose that z = a + bi. Prove that
|z|2 = a2 + b2 = z · z̄.
223. Prove the following inequalities:
(a) |z1 + z2 | 6 |z1 | + |z2 |,
(b) |z1 − z2 | > |z1 | − |z2 | ,
where z1 and z2 are arbitrary complex numbers. In what cases does equality hold?
224. Prove, using the complex numbers, that in an arbitrary parallelogram, the sum
of the squares of the lengths of the diagonals is equal to the sum of the squares of
the lengths of all of the sides.

§5. Trigonometric representation of complex numbers

Recall that the angle between the rays OA and OB is the angle by which we need
to rotate the ray OA around the point O anticlockwise to get the ray OB. (If the
rotation is clockwise, it is denoted by a “minus” sign.) Note that the angle is not
defined uniquely, but only up to addition of 2kπ, where k is any integer.
†
For real numbers (as a special case of complex numbers), the concept introduced here coincides
with the concept of the absolute value. In fact, the real number a + 0i corresponds to the vector
with coordinates (a, 0), parallel to the x-axis, and its length is equal to |a|, the absolute value of
the number a.
44 CHAPTER 2. COMPLEX NUMBERS

Definition 27. Let the point O be the origin and the vector OA with coordinates
(a, b) correspond to the complex number z = a + bi (see figure 15). The argument of
the complex number z (denoted by Arg z) is the angle between the positive direction
of the axis OX and the ray OA (angle ϕ in figure 15). If z = 0, then Arg z is
undefined.
y

A b

a 0 x

Figure 15.

Since, for a given number z 6= 0, the angle is not defined uniquely, when we write
Arg z, we will mean a multi-valued function that for each z 6= 0 takes an infinite
set of values, the difference between each of which is a multiple of 2π. By writing
Arg z = ϕ, we will mean that one of the values of the argument of z is equal to ϕ.
−→
Suppose that z = a + bi 6= 0 and let r = |z|. The vector OA with coordinates
(a, b) corresponds to the complex number a + bi, and therefore its length is equal
to r. Suppose too that Arg z = ϕ. Then, by definition of the trigonometric functions
(see figure 15), we have
a b
cos ϕ = , sin ϕ = .
r r
From this we get
z = a + bi = r cos ϕ + ir sin ϕ = r(cos ϕ + i sin ϕ),
where r = |z| and ϕ = Arg z. This is called the trigonometric representation of the
complex number z. √ √
For example, if z√ = −1 + 3 i, then |z| = 1 + 3 = 2 (see problem 222) and
√
cos ϕ = − 21 , sin ϕ = 23 . We can take ϕ = 2π
3
, so z = −1 + 3 i = 2(cos 2π
3
+ i sin 2π
3
).
225. Represent
√ the following complex numbers using their trigonometric form: (a) 1+
i, (b) − 3 − i, (c) 3i, (d) −5, (e) 1 + 2i.
 5. TRIGONOMETRIC REPRESENTATION OF COMPLEX NUMBERS 45

226. Let z1 = r1 (cos ϕ1 + i sin ϕ1 ) and z2 = r2 (cos ϕ2 + i sin ϕ2 ). Prove that

z1 z2 = r1 r2 (cos(ϕ1 + ϕ2 ) + i sin(ϕ1 + ϕ2 )),
z1 r1
= (cos(ϕ1 − ϕ2 ) + i sin(ϕ1 − ϕ2 )) if z2 6= 0.
z2 r2
So, when multiplying complex numbers, their moduli are multiplied and their
arguments are added; when dividing, the moduli are divided and the arguments
subtracted, that is, |z1 z2 | = |z1 | · |z2 |, Arg(z1 z2 ) = Arg z1 + Arg z2 , and so on.
227. Prove De Moivre’s formula †:
(r(cos ϕ + i sin ϕ))n = rn (cos nϕ + i sin nϕ)
for any positive integer n.
√
228. Calculate (1 − 3 i)100 /2100 .
229. Let z = r(cos ϕ + i sin ϕ) be a fixed non-zero complex number and n a positive
integer. Find all complex numbers w satisfying the equality
wn = z. (5.1)
√
Definition 28. By z (the nth degree root of z), we mean a multi-valued function
n

that puts each complex number z √ 6= 0 in correspondence with all n solutions to

equation (5.1). When z = 0, we let n 0 = 0.
√ √ √
230.√Find all values of the following: (a) −1, (b) 3 8,(c) 4 cos 100◦ + i sin 100◦ ,
(d) 3 1 + i.
It is convenient to introduce the following notation for subsequent use:
2π 2π
εn = cos + i sin .
n n
√
231. Prove that all values of 1 are 1, εn , ε2n , . . . , εnn−1 .
n

Note. Because εnn = 1, the set of elements 1, εn , ε2n , . . . , εn−1

n forms a cyclic group
under multiplication.
√ √
232. Let z1 be one of the values of n z0 . Find all values of n z0 .
Henceforth, we will usually use the representation of complex numbers as points
in the plane, i.e., the complex number z = a + bi will correspond to the point
with coordinates (a, b). Also, instead of saying “the point which corresponds to the
complex number z”, we will simply say “the point z”.
233. Let the complex numbers be represented by points in the plane. What are the
geometric meanings of the expressions: (a) |z|, (b) Arg z, (c) |z1 −z2 |, (d) Arg(z1 /z2 )?
†
De Moivre (1667–1754), an English mathematician.
46 CHAPTER 2. COMPLEX NUMBERS

234. Find the locus of points z satisfying the following conditions, where z0 , z1
and z2 are fixed complex numbers, and R is a fixed positive real number:
(a) |z| = 1, (b) |z| = R, (c) |z − z0 | = R, (d) |z − z0 | 6 R, (e) |z − z1 | = |z − z2 |,
(f) Arg z = π, (g) Arg z = 9π
4
, (h) Arg z = ϕ.
√
235. How are the values of n z positioned in the plane, where z is a fixed complex
number?

§6. Continuity
In what follows, the concept of continuity, and in particular the concept of a con-
tinuous curve, will play an important role for us. If the reader does not know the
formal definitions of these concepts, then he or she will probably still intuitively
understand what a continuous curve is, as well as what a continuous function of a
real variable is. (On an intuitive level, we can say that a continuous function is a
function whose graph is a continuous curve.) However, if a function of a real variable
is sufficiently complicated (for example, consider f (x) = (x3 − 2x)/(x2 − sin x + 1)),
it can be rather difficult to determine whether it is continuous or not simply by using
the intuitive concept of continuity. Therefore we will give a rigorous definition of
continuity, and use it to prove several basic results about continuous functions. We
will give the definition of continuity not only for functions of a real variable, but also
for functions of a complex variable.
If we look at the graph of a function of a real variable, then this graph can
be continuous at some points and yet have discontinuities at some other points.
Therefore, it is natural to first introduce a definition, not of continuity of a function,
but of continuity at a point.
If we try to describe more precisely our intuitive understanding of the continuity
of a function f (x) at a given point x0 , we will see that continuity means the following:
with only small changes to the variable x near the point x0 , the function also changes
by only a relatively small amount around the value f (x0 ). Note that we can make the
change in the function as small as we want by choosing a sufficiently small interval of
changing the variable x around x0 . More formally, this can be stated in the following
way.
Definition 29. Let f (z) be a function of a real or complex variable z. We say that
the function f (z) is continuous at the point z0 if for any real number ε > 0, we
can find a real number δ > 0 (depending on z0 and ε) such that, for all numbers z
satisfying the condition |z − z0 | < δ, the inequality |f (z) − f (z0 )| < ε holds.†
Example 15. We shall prove that the function of a complex variable f (z) = 2z is
continuous at any point z0 . Suppose we are given a point z0 and an arbitrary real
†
For the geometric meaning of the inequalities |z − z0 | < δ and |f (z) − f (z0 )| < ε, see prob-
lems 233(c) and 234(d).
 6. CONTINUITY 47

number ε > 0. We need to pick a real number δ > 0 such that, for all numbers z
satisfying the condition |z − z0 | < δ, the inequality |f (z) − f (z0 )| = |2z − 2z0 | < ε
holds. It is not difficult to see that we can pick δ = ε/2 (independently of the
point z0 ). Indeed, from the condition |z − z0 | < δ, it follows that:
|2z − 2z0 | = |2(z − z0 )| = |2| · |z − z0 | < 2δ = ε,
using 226, so |2z−2z0 | < ε. Thus the function f (z) = 2z is continuous at any point z0 .
In particular, it is continuous for all real values of the variable z. Therefore, if we
restrict ourselves to real values of the variable, we see that the function of a real
variable f (x) = 2x is continuous at all real values of x.
236. Let a be a fixed complex (or, as a special case, real) number. Prove that the
constant function of a complex (or real) variable, f (z) = a, is continuous for all
values of the variable z.
237. Prove that the function of a complex variable f (z) = z and the function of a
real variable f (x) = x are continuous for all values of the variable.
238. Prove that the function of a complex variable f (z) = z 2 is continuous for all
values of the variable.
Definition 30. Let f (z) and g(z) be two functions of a complex (or real) variable.
The function of complex (or real) variable h(z) is the sum of the functions f (z) and
g(z) if for every z0 , the equality h(z0 ) = f (z0 )+g(z0 ) holds. Also, if the value of f (z0 )
or of g(z0 ) is undefined, then the value of h(z0 ) is also undefined. In the same way,
we can define the difference, product and quotient of two functions.
239. Suppose that the functions f (z) and g(z) of a complex or real variable are
continuous at the point z0 . Prove that at the point z0 , the following functions are
also continuous: (a) h(z) = f (z) + g(z), (b) h(z) = f (z) − g(z), (c) h(z) = f (z)g(z).
From the result of problem 239(c) we see, in particular, that if the function f (z)
is continuous at a point z0 , and n is a positive integer, then the function (f (z))n is
also continuous at the point z0 .
240. Suppose that the functions f (z) and g(z) of a complex or real variable are
continuous at the point z0 , and that g(z0 ) 6= 0. Prove that at the point z0 , the
following functions are also continuous: (a) h(z) = 1/g(z), (b) h(z) = f (z)/g(z).
Definition 31. Let f (z) and g(z) be two functions of a complex or real variable. The
function h(z) is the composition of the functions f (z) and g(z) if at every point z0 ,
the equality h(z0 ) = f (g(z0 )) holds. Also, if g(z0 ) is undefined or if the function f (z)
is not defined at the point g(z0 ), then h(z0 ) is also not defined.
241. Let f (z) and g(z) be functions of a complex or real variable. Suppose that
g(z0 ) = z1 , that the function g(z) is continuous at the point z0 , and that the func-
tion f (z) is continuous at the point z1 . Prove that the function h(z) = f (g(z)) is
continuous at the point z0 .
48 CHAPTER 2. COMPLEX NUMBERS

From the results of problems 239–241, it follows in particular that if we construct

an expression from several functions of a complex (or real) variable, each of which is
continuous for all values of the variable, using the operations of addition, subtraction,
multiplication, division, raising to a positive integer power and composition, then the
resulting expression will be a function which is continuous everywhere that none of
the denominators is zero.
For example, considering the results of problems 236 and 237, we see that the
functions f (z) = z n , f (z) = az n , and, more generally, f (z) = an z n + an−1 zn−1 +
· · · + a0 , are continuous functions of the variable z for any complex numbers a, an ,
an−1 , . . . , a0 .
242. Prove that the functions of real variables f (x) = sin x and f (x) = cos x are
continuous for all values of the variable x.
√
243. Consider the function
√ f (x) = n
x for all real values x > 0, where n is some
positive integer, and x is taken to be the non-negative nth degree root. Prove that
n

this function is continuous for all x > 0.

When studying continuity, we often encounter statements that intuitively seem
completely obvious, but for which constructing a formal proof presents great techni-
cal difficulties and requires a more formal definition of the real numbers than seen in
high school, as well as a study of the basics of set theory and topology. One example
of such a statement is this: if a function of a real variable f (x) is continuous on some
interval and on that interval it only takes integer values, then on that interval it takes
only one value. Intuitively, it seems obvious that when the point x is moving along
the interval, the value of the function f (x) must change continuously and cannot
“jump” from one integral value to another. However, it is rather difficult to prove
this statement formally.
For this reason, in the sequel, we will often rely on the intuition of the reader, and
assume without proof several “intuitively obvious” statements related to continuity.
In particular, we will assume without proof the statement above that was used as
an example. A formal proof of this statement can be found in standard analysis
textbooks.

§7. Continuous curves

Let the variable t take real values on the interval 0 6 t 6 1, and let every such value
of t correspond to some complex number
z(t) = x(t) + iy(t).
The plane in which we illustrate the values of z will subsequently be called simply
“the z-plane”. If the functions x(t) and y(t) are continuous for 0 6 t 6 1, then when
t varies between 0 and 1, the point z(t) will describe some continuous curve in the z-
plane. We will consider this curve together with its direction, taking the point z(0)
 7. CONTINUOUS CURVES 49

to be the beginning of the curve and the point z(1) to be its end. We call the
function z(t) a parametric equation of this curve.
Example 16. Let z(t) = t + it2 . Then x(t) = t and y(t) = t2 . Therefore, y(t) =
(x(t))2 for every t, i.e., the point z(t) lies on the parabola y = x2 for every t. When
t varies from 0 to 1, then x(t) also varies from 0 to 1, and the point z(t) runs along
the parabola y = x2 from the point z(0) = 0 to the point z(1) = 1 + i (see figure 16).

−1 0 1 x

Figure 16.

244. Construct the curves that are given by the following parametric equations
in the z-plane: (a) z(t) = 2t; (b) z(t) = it; (c) z(t) = it2 ; (d) z(t) = t − it;
(e) z(t) = t2 + it; (f) z(t) = R(cos 2πt + i sin 2πt); (g) z(t) = R(cos 4πt + i sin 4πt);
(h) z(t) = R(cos πt + i sin πt);
(
cos 2πt + i sin 2πt when 0 6 t 6 21 ,
(i) z(t) =
4t − 3 when 12 < t 6 1.
245. Write a parametric equation for the line segment connecting z0 = a0 + b0 i and
z1 = a1 + b1 i.
Note. In subsequent problems, some of the parametric equations have indices. These
indices simply denote the number of the curve, but we consider all of these curves
to be in the same z-plane.
246. What geometric transformations need to be performed on the curve C1 satis-
fying the equation z1 (t) to get the curve C2 satisfying the equation z2 (t) if:
(a) z2 (t) = z1 (t) + z0 , where z0 is a fixed complex number;
(b) z2 (t) = a · z1 (t), where a is a positive real number;
(c) z2 (t) = z0 · z1 (t), where |z0 | = 1;
(d) z2 (t) = z0 · z1 (t), where z0 is an arbitrary complex number?
247. Suppose that z1 (t) is a parametric equation of the curve C. What curve is
described by the equation z2 (t), if z2 (t) = z1 (1 − t)?
50 CHAPTER 2. COMPLEX NUMBERS

248. Suppose that z1 (t) and z2 (t) are the parametric equations of the curves C1 and C2 ,
and that z1 (1) = z2 (0). What curve is described by the equation z3 (t), if
(
z1 (2t) when 0 6 t 6 12 ,
z3 (t) =
z2 (2t − 1) when 21 < t 6 1?
249. Let z(t) = cos πt + i sin πt (see figure 17). For each t, find all values of Arg z(t)
in terms of t.

−1 0 1

Figure 17.

250. Let z(t) = cos πt+i sin πt. How should we pick one value for Arg z(t) for each t
so that the selected values vary continuously with t varying from 0 to 1, given that
Arg z(0) is picked to be: (a) 0, (b) 2π, (c) −4π, (d) 2πk, where k is an integer?
The following statement seems intuitive enough that we will give it without proof.
Theorem 6. Suppose that a continuous curve C with parametric equation z(t) does
not pass through the origin (i.e., z(t) 6= 0 for 0 6 t 6 1), and let the argument of
the starting point of the curve C (i.e., Arg z(0)) be chosen to equal ϕ0 . Then we can
pick one value for the argument for all points of the curve C so that while moving
along the curve the argument changes continuously, starting from the value ϕ0 .
In other words, there is a function ϕ(t) so that ϕ(t) is one of the values of Arg z(t)
for every t, the function ϕ(t) is continuous for 0 6 t 6 1, and ϕ(0) = ϕ0 .†
251. Let ϕ1 (t) and ϕ2 (t) be two continuous functions describing the varying of
Arg z(t) along the curve C. Prove that ϕ1 (t) − ϕ2 (t) = 2πk, where k is some fixed
integer not depending on t.
252. Prove that if we pick some value ϕ(0) = ϕ0 , then the function ϕ(t) describing
a continuous varying of Arg z(t) along the curve C is determined uniquely.
253. Suppose that the function ϕ(t) describes a continuous varying of Arg z(t).
Prove that the function ψ(t) = ϕ(t) − ϕ(0) is uniquely defined by z(t) and does not
depend on the choice of ϕ(0).
†
It is possible to give a formal definition of an angle swept out by a given curve. Using this
concept, it is easy to obtain the statement of Theorem 6: we need only take ϕ(t) = ϕ0 + ϕ1 (t),
where ϕ1 (t) is the angle swept out by the part of the curve from z(0) to z(t).
 7. CONTINUOUS CURVES 51

From the statement of problem 253 it follows by taking t = 1 that on a given

continuous curve C, not passing through the point z = 0, the value of ϕ(1) − ϕ(0) is
uniquely determined by the condition of that ϕ(t) is continuous.
Definition 32. The value ϕ(1) − ϕ(0) will be called the change in argument along
the curve C.
254. What is the value of the change in argument along the curves with the following
parametric equations:
(a) z(t) = cos πt + i sin πt,
(b) z(t) = cos 2πt + i sin 2πt,
(c) z(t) = cos 4πt + i sin 4πt,
(d) z(t) = (1 − t) + it?
255. What is the value of the change in argument along the curves in figure 18?

1
1

−2i −2i

Figure 18.

If a continuous curve C is closed, i.e., z(1) = z(0), then the value ϕ(1) − ϕ(0)
has the form 2πk, where k is an integer, said to be the number of times the curve C
goes around the point z = 0.
256. How many times do the following curves go around the point z = 0:
(a) z(t) = 2 cos 2πt + 2i sin 2πt (see figure 19(a)),
(b) z(t) = 12 cos 4πt − 21 i sin 4πt (see figure 19(b)),
(c) the curve in figure 19(c),
(d) the curve in figure 19(d)?
257. Prove that the number of times a closed curve goes around the point z = 0
does not depend on the choice of starting point, but only on the direction of the
curve.
258. Suppose that the closed curve C with equation z1 (t) goes k times around the
point z = 0. How many times does the curve with equation z2 (t) go around the
52 CHAPTER 2. COMPLEX NUMBERS

0 1 0 1

Figure (a) Figure (b)

A
0 1 0 1

Figure (c) Figure (d)

Figure 19.

point z = 0 if: (a) z2 (t) = 2z1 (t); (b) z2 (t) = −z1 (t); (c) z2 (t) = z0 · z1 (t), where
z0 6= 0 is a fixed complex number; (d) z2 (t) = z1 (t)?
Definition 33. Suppose that a continuous closed curve C with equation z1 (t) does
not pass through the point z = z0 . Then we say that the curve C goes k times
around the point z0 if the curve with equation z2 (t) = z1 (t) − z0 goes k times around
the point z = 0 (see figure 20).
Thus, to determine the number of times the curve goes around the point z = z0 ,
we simply have to observe the rotation of the vector z1 (t) − z0 , which can be viewed
as the vector connecting the points z0 and z1 (t) (see problem 221).
259. How many times do the curves in problem 256 go around the point z = 1?
260. Let z1 (t) and z2 (t) be the equations of curves C1 and C2 , which do not pass
through the point z = 0. Let ϕ1 and ϕ2 be the respective changes in argument along
 8. MAPS OF CURVES AND THE FUNDAMENTAL THEOREM OF ALGEBRA 53

z1 (t)

z0

z2 (t) = z1 (t) − z0

Figure 20.

these curves. What is the value of the change in argument along the curve C with
equation z(t) if (a) z(t) = z1 (t)z2 (t), (b) z(t) = z1 (t)/z2 (t)?

§8. Maps of curves and the fundamental theorem of

algebra
Suppose we are given two planes of complex numbers: a z-plane and a w-plane,
and that we are given a function w = f (z) which associates to each value of z a
uniquely determined value w. If we have a continuous curve C in the z-plane with
equation z(t), then every point of this curve is sent to some point in the w-plane by
the function w = f (z). If this function f (z) is continuous, then in the w-plane we
will also get a continuous curve with equation w0 (t) = f (z(t)). This curve will be
denoted by f (C).
261. What does the curve f (C) look like if w = f (z) = z 2 and the curve C is:
(a) a quarter circle:
 
πt πt
z(t) = R cos + i sin ,
2 2
(b) a semicircle: z(t) = R(cos πt + i sin πt),
(c) a circle: z(t) = R(cos 2πt + i sin 2πt)?
262. Suppose that the change in argument along the curve C is ϕ. What is the
change in argument along the curve f (C) if: (a) f (z) = z 2 , (b) f (z) = z 3 , (c) f (z) =
z n , where n is an arbitrary positive integer?
54 CHAPTER 2. COMPLEX NUMBERS

263. Suppose that the closed curve C goes k times around the point z = z0 . How
many times does the curve f (C) go around the point w = 0 if f (z) = (z − z0 )n ?
264. Suppose that the number of times that the closed curve C goes around the
points z = 0, z = 1, z = i and z = −i is k1 , k2 , k3 and k4 , respectively. How
many times does the curve f (C) go around the point w = 0 if: (a) f (z) = z 2 − z,
(b) f (z) = z 2 + 1, (c) f (z) = (z 2 + iz)4 , (d) f (z) = z 3 − z 2 + z − 1?
Let us consider the equation
an z n + an−1 z n−1 + · · · + a1 z + a0 = 0
with n > 1, where the ai are arbitrary complex numbers and an 6= 0. Our next goal
is to show that this equation has at least one complex root. If a0 = 0, then the
equation has the root z = 0. We will therefore assume that a0 6= 0.
Let us denote the greatest of the numbers |a0 |, |a1 |, . . . , |an | by A. Because
an 6= 0, we have A > 0. Pick positive real numbers R1 and R2 with R1 small enough
for the two inequalities R1 6 1 and R1 < |a0 |/10An to hold, and R2 large enough
for the two inequalities R2 > 1 and R2 > 10An/|an | to hold.
265. Let |z| = R1 . Prove that
|a0 |
|an z n + · · · + a1 z| < .
10
266. Let |z| = R2 . Prove that
an−1 a0 |an |
+ ··· + n < .
z z 10
We will denote the curve with equation z(t) = R(cos 2πt + i sin 2πt), (i.e., the
circle with radius R drawn anticlockwise) by CR . Because the curve CR is closed
(z(1) = z(0)), the curve f (CR ), where f (z) = an z n +· · ·+a0 , is also closed (f (z(1)) =
f (z(0))). Let ν(R) be the number of times that the curve f (CR ) goes around the
point w = 0 if f (CR ) does not pass through the point w = 0.
267. What are the values of ν(R1 ) and ν(R2 )?
Let us now change the radius R continuously from R1 to R2 . The curve f (CR )
will be continuously deformed from the curve f (CR1 ) to the curve f (CR2 ). If for
some value of R the curve f (CR ) does not pass through the point w = 0, then for
sufficiently small changes of R, the curve f (CR ) will be deformed so little that the
number of times it goes around the point w = 0 will not change, i.e., for this value
of R, the function ν(R) is continuous. If none of the curves f (CR ) with R1 6 R 6 R2
pass through the point w = 0, then ν(R) will be continuous for all R1 6 R 6 R2 .
Because the function ν(R) only takes on integer values, it can only be continuous if
for all values of R in the interval R1 6 R 6 R2 it takes the same value; in particular,
it follows that ν(R1 ) = ν(R2 ). But from the solution to problem 267 it follows that
ν(R1 ) = 0 while ν(R2 ) = n. Therefore the assumption that none of the curves f (CR )
 8. MAPS OF CURVES AND THE FUNDAMENTAL THEOREM OF ALGEBRA 55

pass through the point w = 0 for R1 6 R 6 R2 is false, so for some z it has to be

true that f (z) = 0. So, we have deduced the following theorem†
Theorem 7 (The fundamental theorem of algebra‡). Any equation
an z n + an−1 z n−1 + · · · + a1 z + a0 = 0,
where n > 1 and the ai are arbitrary complex numbers with a0 6= 0, has at least one
complex root.
268. Prove Bézout’s theorem § that if z0 is a root of the equation an z n +· · ·+a1 z+a0 =
0, then the polynomial an z n +· · ·+a1 z +a0 is divisible by z −z0 without a remainder.
269. Prove that the polynomial an z n +· · ·+a1 z +a0 , where an 6= 0, can be expressed
in the form an z n + · · · + a1 z + a0 = an (z − z1 )(z − z2 ) · · · (z − zn ).
Note. Suppose that the polynomial P (z) is factored into factors:
P (z) = an (z − z1 )(z − z2 ) · · · (z − zn ).
The right part is equal to 0 if and only if at least one of the factors is equal to 0 (see
problems 195 and 197). Therefore the roots of the equation P (z) = 0 are precisely
z1 , z2 , . . . , zn .
270. Let z0 be a root of the equation
an z n + · · · + a1 z + a0 = 0,
where all of the ai are real numbers. Prove that z̄0 is also a root of this equation.
271. Suppose that the equation with real coefficients
an z n + · · · + a1 z + a0 = 0
has a complex root z0 which is not a real number. Prove that the polynomial
an z n +· · ·+a1 z+a0 is divisible by some polynomial of degree two with real coefficients.
272. Prove that any polynomial with real coefficients can be expressed as a product
of polynomials of degree one and two with real coefficients.
Note. From the result of problem 272, it follows that the irreducible polynomials (see
page 40) over the field of real numbers are precisely the polynomials of degree one
and the polynomials of degree two which do not have real roots. We used this result
earlier in section 3 of this chapter. Over the field of complex numbers, it follows
from problem 269 that the only irreducible polynomials are those of degree one.
Let us return now to polynomials with arbitrary complex coefficients.
†
Our discussion is somewhat informal and should be viewed only as an outline of a proof. This
discussion can (although not easily) be made absolutely rigorous.
‡
This theorem was proven in 1799 by the German mathematician Gauss (1777–1855).
§
Bézout (1730–1783) was a French mathematician
56 CHAPTER 2. COMPLEX NUMBERS

Definition 34. Suppose that z0 is a root of the equation

an z n + · · · + a1 z + a0 = 0.
We say that z0 is a root of multiplicity k if the polynomial an z n + · · · + a1 z + a0 is
divisible by (z − z0 )k but not divisible by (z − z0 )k+1 .
273. What are the multiplicities of the roots z = 1 and z = −1 in the equation
z 5 − z 4 − 2z 3 + 2z 2 + z − 1 = 0?
Definition 35. The derivative of the polynomial
P (z) = an z n + an−1 z n−1 + · · · + ak z k + · · · + a1 z + a0
is the polynomial
P 0 (z) = an nz n−1 + an−1 (n − 1)z n−2 + · · · + ak kz k−1 + · · · + a1 .
The derivative is usually denoted with a prime.
274. Let P (z) and Q(z) be two polynomials. Prove the following equalities:
(a) (P (z) + Q(z))0 = P 0 (z) + Q0 (z)
(b) (cP (Z))0 = cP 0 (z), where c is an arbitrary complex number;
(c) (P (z)Q(z))0 = P 0 (z)Q(z) + P (z)Q0 (z).
275. Let P (z) = (z − z0 )n with n > 1 an integer. Prove that P 0 (z) = n(z − z0 )n−1 .
276. Prove that if the equation P (z) = 0 has a root z0 of multiplicity k > 1, then
the equation P 0 (z) = 0 has the root z0 with multiplicity k − 1, and if the equation
P (z) = 0 has a root z0 of multiplicity 1, then P 0 (z0 ) 6= 0.

√
§9. The Riemann surface for w = z
Earlier we looked at single-valued functions for which every value of the variable
corresponded to only one value of the function. Henceforth, however, we will be
especially interested in multi-valued functions for which some values of the variable
correspond to several values of the function.† Our interest in these functions is
easily explained. The final goal of this book is to prove Abel’s theorem, that the
function that describes the roots of the general fifth degree polynomial in terms of
its coefficients cannot be expressed in terms of radicals. But this function is multi-
valued, because a fifth degree polynomial with fixed coefficients has, in general, five
roots. Functions that can be expressed in terms of radicals are also multi-valued.
The general idea behind our proof of Abel’s theorem consists of the following.
We will associate to a multi-valued function of a complex variable some group—the
so-called Galois ‡ group of the function. It will turn out that Galois group of the
†
In cases where it does not cause confusion, we will often omit the word “multi-valued”.
‡
Évariste Galois (1811–1832), a French mathematician who determined general conditions of
solvability of equations in radicals and who established the fundamentals of group theory.
 √
9. THE RIEMANN SURFACE FOR w = z 57

function that expresses the roots of the general fifth degree polynomial cannot be
the Galois group of a function expressed in radicals, and therefore the function itself
cannot be expressed in radicals.
To introduce the concept of a Galois group, we will first introduce another im-
portant concept from the theory of functions of a complex variable—the concept of
a Riemann† surface for a multi-valued function. We will begin with the construction
of a Riemann surface for √ one of the simplest examples of a multi-valued function,
namely the function w = z. √
We know that the function w = z takes on only one value w = 0 at z = 0
and two values for all z 6= 0 (see problem 229). Note that if w0 is one of the values
√ √
of z0 , then the other value of z0 is −w0 .
√ √ √ p √ √
277. Find all values of: (a) 1, (b) −1, (c) i, (d) 1 + i 3 (where by 3 we
mean the positive value of the root).
Let us make a cut in the z-plane along the negative part of √ the real axis from 0
to −∞. For all z not on the cut, we will pick the value w = z that lies on the
right half of the w-plane. In this way, we get a function that is single-valued and
continuous
√ on the entire z-plane excluding the cut. We will denote this function
by 1 z. This function gives a single-valued and continuous map of the z-plane,
excluding the cut, to the right half of the w-plane (see figure 21).

A0

A
B

B0

Figure 21.
√
Note. If √we pick Arg z such that −π < Arg z < π, then for the function 1 z√we will
get Arg 1 z = 12 Arg z (see problem 229). Therefore under the map w = 1 z, the
z-plane is folded back as a fan towards the positive side of the real axis by decreasing
the angle of the “fan” by a factor of two and by changing lengths along the rays of
the “fan”.
√
If we now pick the value of w = z that lies in the left half of the w-plane for
all z that do not lie on the cut, then we will get another function, also continuous and
single-valued on the entire z-plane excluding the cut. This function, which we will
†
Named after B. Riemann (1826–1866), a German mathematician.
58 CHAPTER 2. COMPLEX NUMBERS

√
denote by 2 z, gives a single-valued and continuous map of √ the z-plane,
√ excluding
the cut, to the left half of the w-plane (see figure 22). Here 2 z = −1 z.

D0

C
D

C0

Figure 22.
√ √
The constructed functions√1 z and 2 z are called the single-valued continuous
branches of the function w = z (with the given cut).
We now take two copies of the z-plane, which we will also call sheets, and on
each we make a cut along the negative part of the√real axis from 0 to −∞ (see
figure 23). On the first sheet we define the function 1 z and on the second sheet the

A 1
O
B

C 2
O
D

Figure 23.
√ √ √
function 2 z. Then we can view the functions 1 z and 2 z together as one single-
valued function, defined not on the z-plane, but on a more complicated surface
consisting of two separate sheets.
If a point z is moving continuously on the first sheet (or on the second sheet), not
crossing the cut, then the single-valued function we constructed changes continuously.
If, however, the point z moving on the first sheet goes across the cut, continuity is
broken. This can be seen, for example, from the fact that the points
√ A and B, which
are close together in the z-plane, are sent under the map w = 1 z to the two points
A0 and B 0 , which are far apart from each other (see figure 21).
On the other hand, from figures√21 and 22, it is easy to see that the image A0 of
the point A under the map w√= 1 z turns out to be close to the image D0 of the
point D under the map w = 2 z.
 √
9. THE RIEMANN SURFACE FOR w = z 59

Thus, if when crossing the cut, the point z goes from the upper side of the
cut on one sheet to the lower side of the cut on the other sheet, then the single-
valued function we constructed will change continuously. To guarantee the necessary
movement of the point z, we will consider the upper side of the cut on the first sheet
to be glued to the lower side of the cut on the second sheet, and the upper side of
the cut on the second sheet to be glued to the lower side of the cut on the first sheet
(see figure 24). Furthermore, during the gluing, between the edges of the cuts that

Figure 24.

are being glued we will add √ a ray of points from 0 to −∞. During the first gluing,
we will pick values of w = z that lie on the positive
√ part of the imaginary axis, and
during the second gluing, we pick values of w = z that lie on the negative part of
the imaginary axis.
After performing
√ the described gluing, what we find is that the double-valued
function w = z is changed to another function which is single-valued and continu-
ous, not on the z-plane, but on some other new, more complicated
√ surface. We will
call this surface the Riemann surface of the function w = z.
Attempts to glue this surface in the way described without intersections (and
without turning one plane over) are unsuccessful. Despite this, we consider
√ figure 24
to be an illustration of the Riemann surface of the function w = z, assuming an
additional condition, namely that the intersection along the negative part of the real
axis is not actually present. For comparison, consider the following example. In
figure 7 (page 20), we have a drawing of the frame of a cube. Although some of the
segments in the figure intersect, we can easily agree that these intersections are only
present in the illustration, and this allows us to avoid mistakes.
The Riemann surface of an arbitrary multi-valued function w(z) can be con-
structed
√ similarly to the way we constructed the Riemann surface of the function
w = z. To do this, we first need to separate the single-valued continuous branches
of the function w(z), with some points z (the cuts) being excluded from considera-
tion. After that, the branches are glued, restoring the cuts in the process, so that
we get a single-valued continuous function on the constructed surface. We call the
resulting surface the Riemann surface of the multi-valued function w(z).†
†
Such constructions cannot be performed for every multi-valued function, but for those functions
which we will be considering in the sequel, such constructions really can be performed.
60 CHAPTER 2. COMPLEX NUMBERS

So it remains for us to determine how to separate the continuous single-valued

branches of an arbitrary function w(z) and how to glue them together afterwards.
To obtain
√ answers to these questions, let us look again in more detail at the function
w = z.
Let w(z) be a multi-valued function, and suppose we fix one of the values w0 of
the function w(z) at some point z0 . Let w∗ (z) be a continuous and single-valued
branch of the function w(z) defined on some region of the z-plane (for example, on
the entire plane excluding some cuts),† and such that w∗ (z0 ) = w0 . Also, let C be a
continuous curve, lying entirely within this region, going from the point z0 to some
other point z1 . Then when the point z moves along the curve C, the function w∗ (z)
will change continuously from w∗ (z0 ) to w∗ (z1 ).
This property can also be used in reverse, namely to determine the function w∗ (z).
For suppose that at some point z0 we fix one of the values w0 of w(z), and we let
C be a continuous curve going from the point z0 to some point z1 . As we move along
the curve C, we can pick a value of w(z) for each point z on C such that the values
change continuously as we move along the curve C, beginning with the value w0 .
When we reach the point z1 , we will have a uniquely determined value w1 = w(z1 ).
We will say that w1 is the value of w(z1 ) determined by continuity along the curve C
given the condition that w(z0 ) = w0 . If the chosen values of the function w(z) along
the curve C are plotted in the w-plane, then we will get a continuous curve which
begins at the point w0 and ends at the point w1 . This curve is one of the continuous
images of the curve C under the map w = w(z).
√ √
278. Suppose we pick √ the value w(1) = 1 = 1 for the function w(z) = z.
Determine w(−1) = −1 by continuity along: (a) the upper semicircle of radius 1
centred at the origin, (b) the lower semicircle (see figure 25).

−1 0 1

−i

Figure 25.

It turns out that when defining functions by continuity along a curve, we can
encounter some difficulties. Consider the following example.
†
We also require this region to be path-connected, that is, given any two points in the region,
there is a continuous curve between them which lies entirely in the region. This will always be true
in the cases we will be considering.
 √
9. THE RIEMANN SURFACE FOR w = z 61

279. Find all continuous images w0 (t) of the

√ curve C with parametric equation
z(t) = 2t − 1 (figure 26) under the map w = z, beginning: (a) at the point i, (b) at
the point −i.

−1 0 1

Figure 26.

From the solution to problem 279, we see that even with a fixed image of the
beginning
√ point of the curve C, the continuous image of the curve C under the map
w = z may not be uniquely defined. The uniqueness is broken where the curve C
passes through the point z = 0. It√ turns out that the only place where the uniqueness
of the image of the function w = z is broken is where two images of the point z(t)
come close to each other, blending into one point.
To avoid the
√ problem of non-uniqueness of continuous images of curves under
the map w = z, we can remove the point z = 0 and forbid curves from passing
through this point. But even this restriction does not give
√ us the ability to separate
single-valued continuous branches of the function w = z. For if we fix one of the
values w0 = w(z0 ) at some point z0 and determine w(z1 ) at some other point z1
by continuity of w(z) along various curves from z0 to z1 , we may still get different
values for w(z1 ) (see, for example, problem 278). Let us see how we can avoid such
non-uniqueness issues.
280. Suppose that the change in argument of z(t) along the curve C is ϕ. Find the
change in
√ argument along any continuous image w0 (t) of the curve C under the map
w(z) = z.
√ √
281. Let√ w(z) = z and suppose we pick w(1) = 1 = −1. Find the value of
w(i) = i by continuity along: (a) the line segment connecting z = 1 and z = i;
(b) the curve with parametric equation z(t) = cos 23 πt − i sin 32 πt; (c) the curve with
parametric equation z(t) = cos 52 πt + i sin 25 πt.
√
282. Let w(z) =√ z and let C be a closed curve beginning at z = 1. Suppose that
we pick √w(1) = 1 = 1. Determine by continuity along the curve C the value of
w(1) = 1 at the ending point of C, if the curve C has the equation: (a) z(t) =
cos 2πt + i sin 2πt, (b) z(t) = cos 4πt − i sin 4πt, (c) z(t) = 2 − cos 2πt − i sin 2πt.
283. Let C be a closed curve in the z-plane√which does not pass through the point
z = 0. Prove that the value of the function z at the ending point of the curve C,
62 CHAPTER 2. COMPLEX NUMBERS

determined by continuity, will coincide with the value at the beginning point if and
only if the curve C goes around the point z = 0 an even number of times.
For later use, it is convenient to introduce the following notation.
Definition 36. Let C be a continuous curve with parametric equation z(t). We
will denote the curve that geometrically coincides with C but goes in the opposite
direction by C −1 , which has parametric equation z1 (t) = z(1 − t) (see problem 247).
Definition 37. Let the beginning point of the curve C2 coincide with the ending
point of the curve C1 . Then by C1 C2 we will mean the curve we get if we first walk
along C1 and then along C2 (see problem 248).
284. Suppose that C1 and C2 are two curves connecting the point z0 to the point z1 ,
neither one passing through the point z = 0, and suppose we pick one of the values
√ √
w0 = z0 . Prove that the values of z1 determined by continuity along the curves
C1 and C2 will be the same if and only if the curve C1−1 C2 (see figure 27) goes around
the point z = 0 an even number of times.

z0
C1−1

z1 C2

Figure 27.

From the statement of the last problem, it follows in particular that if √the curve
C1−1 C2 goes 0 times around the point z = 0, then the value of the function z at the
ending points of the curves C1 and C2 determined by continuity will be the same if
their beginning point values are the same. √
Thus to separate single-valued continuous branches of the function w = z, it is
sufficient to ensure that the curve C1−1 C2 cannot go around the point z = 0 at all.
For that, it is sufficient to make a cut from the point z = 0 to infinity and to forbid
curves from intersecting the cut. This is exactly what we did above, making the cut
from the point z = 0 to −∞ along the negative part of the real axis.
If, after making the cut, we fix at some point z0 not on the cut one of the values
0 √
w0 = z0 , then the value at any other point z1 will be determined by continuity along
any curve C going from z0 to z1 and √ not passing through the
√ cut. This determines
a single-valued continuous branch 1 z of the function w = z. If at the point z0 we
00 √
now fix the other
√ value w 0 = z0 , then
√ from that we can similarly determine the
other branch 2 z of the function w = z.
 √
9. THE RIEMANN SURFACE FOR w = z 63

√√
285. Prove that for any point z not on the cut, we have z 6= 2 z
1
√
286. Let us fix at some point z√0 not on the cut the value w0 = 1 z 0 and determine
√ w = z at all other points of the z-plane (excluding the
the value of the function
cut) by continuity of z along curves from the point z 0 and not going across the cut.
Prove that√ we will get a single-valued continuous branch which coincides with the
function 1 z as defined above from the point z0 .
When we are trying to separate the single-valued continuous branches, it follows
from the result of problem 286 that choosing as a beginning point different points of
the z-plane will result in the same collection of single-valued continuous branches,
depending only on where we put the cut.
287. Suppose that the points z0 and z1 do not lie on the cut, and suppose that a
curve C connecting z0 to z1 crosses the cut exactly once (see figure 28). Suppose we
√
pick a value w0 = z0 , and by continuity along the curve C we determine the value
√
w1 = z1 . Prove
√ that the values w0 and w1 correspond to different branches of the
function w = z.

z0

z1

Figure 28.
√
Thus when crossing the cut, we go from one branch of the function w = z to the
other branch, i.e., the branches connect in exactly the same way we connected√them
earlier (see figure 24). Hence we get the Riemann surface of the function w = z as
before.
We will say that some property holds when walking around the point z0 if it holds
when we walk once anticlockwise around any circle centred at z0 of small enough
radius.†
288. Prove that when walking around√the point z0 we stay on the same sheet of the
Riemann surface of the function w = z if z0 6= 0, and we move to the other sheet
when z0 = 0.
The following concept is very important in what follows.
†
More formally, this means the following: we can find a real number δ > 0 such that the
mentioned property holds when we walk once around any circle centred at z0 whose radius is less
than δ.
64 CHAPTER 2. COMPLEX NUMBERS

Definition 38. If walking around a point moves us from one sheet to another (i.e.,
it changes the value of the function), then this point is called a branch point of the
given multi-valued function.
√
The Riemann surface of the function w = z can be drawn as a schematic
diagram, as in figure 29. The diagram shows that this Riemann surface has two

Figure 29.

√
sheets, that the point z = 0 is a branch point of the function w = z, and that
walking around the point z = 0 we go from the first sheet to the second sheet, and
also walking around the same point takes us from the second sheet to the first. The
arrows at the point z = 0 are used to indicate that we move from the first sheet to
the second and back again not only when walking around the point z = 0, but also
at any crossing of the cut going from the point z = 0 to infinity. We will see below
that the connection between branch points and the cuts made starting at them is
not coincidental. (Note that if there were three sheets and moving around z = 0
were to take us from the first to the second, from the second to the third and from
the third back to the first, then we would draw single-headed arrows from the first
to the second, from the second to the third and from the third back to the first.)
Henceforth, we will generally draw not the Riemann surface of a multi-valued
function itself but its schematic diagram instead.

§10. Riemann surfaces for more complicated functions

√
Let us now look at at the multi-valued function w = 3
z.

289. Assume that the curve C with parametric equation z(t) does not pass through
z = 0. Suppose that the change in the argument along the curve C√is ϕ, and let
w0 (t) be a continuous image of the curve z(t) under the map w = 3 z. Find the
change in argument along the curve w0 (t).
√
290. Find the branch points of the function w = 3 z.

291. Suppose a cut is made from the point z = 0 to −∞ along the negative
√ part of
the real axis, and let the single-valued branches of the function w = z be given by
3
 10. RIEMANN SURFACES FOR MORE COMPLICATED FUNCTIONS 65

the following conditions: f1 (1) = 1,

√
2π 2π 1 3
f2 (1) = cos + i sin =− + i,
3 3 2 √2
4π 4π 1 3
f3 (1) = cos + i sin =− − i.
3 3 2 2
Find: (a) f1 (i), (b) f2 (i), (c) f1 (8), (d) f3 (8), (e) f3 (−i).
292. √Construct the Riemann surface and its schematic diagram for the function
w = 3 z.
293. Let C be a pcontinuous curve with parametric equation z(t) and let w0 be one
n
of the values of z(0). Prove √
that there is at least one continuous image of the
curve C under the map w(z) = n z starting at the point w0 .
294. Assume that the curve z(t) does not pass through the point √ z = 0, and let
w0 (t) be a continuous image of the curve z(t) under the map w = z. Letting ϕ be
n

the change in argument along the curve C, find the change in argument along the
curve w0 (t).
√
295. Find the branch points of the function n z.
Earlier (see page 45), we introduced the notation
2π 2π
εn = cos + i sin .
n n
There we also looked at some properties of this complex number.
296. Assume that the curve z(t) does not pass through the point z = 0, and√let
w0 (t) be one of the continuous images of the curve z(t) under the
√ map w = n z.
Find all continuous images of the curve z(t) under the map w = n z.
Suppose that two continuous curves √ C1 and C2 go from some point z0 to another
point z1 . Just as for the function w = z (see problem 284), we can prove that if
the curve
√ C1−1 C2 does not go around the point z = 0, then the value of the function
w = n z at z1 determined by continuity along the curves C1 and C2 will be the same
if the√values at z0 are the same. Therefore, in the same way as for the function √
w = z, if we make a cut from the point z = 0 to infinity, then the function w = n z
will separate into continuous single-valued branches.
297. Let us make a cut from the point z = 0 to infinity not passing through √ the
point z = 1, and define continuous single-valued branches of the function n z by the
conditions fi (1) = εin , where i runs from 0 to n − 1. Express fi (z) in terms of f0 (z).
√
298. Construct the schematic diagram for the Riemann surface of the function n z.
√
299. For the function z − 1, find the branch points and construct the schematic
diagram of the Riemann surface.
66 CHAPTER 2. COMPLEX NUMBERS

300. Find the branch points

√ and construct the schematic diagram of the Riemann
surface for the function n z + i.

In those situations where the multi-valued function has several branch points, to
separate the continuous single-valued branches we will make cuts from each branch
point to infinity using non-intersecting lines. In such a case, the schematic diagram
of the Riemann surface of the given function may depend on the actual lines along
which the cuts were made. (An example of this will be considered below in problems
327 and 328.) When this happens, we will make a note of how the cuts are made,
otherwise we will not.
Schematic diagrams of Riemann surfaces constructed by the reader while solving
the problems below may differ from the schematic diagrams in the solutions by the
numbering of the sheets. If the sheets are re-numbered accordingly, the schematic
diagrams should coincide.

301. Let f (z) be a single-valued continuous function and C a continuous

p curve in the
z-plane starting at the point z0 . Let w0 be one of the values of n f (z0 ). Prove
p that
n
there exists at least one continuous image of the curve C under the map w = f (z)
starting at the point w0 .

From the solution

p to the problem 301, we have the possibility of determining the
n
function w = f (z) by continuity along any curve not passing through points at
which the single-valuedness of continuous images of curves is violated (that is, where
f (z) = 0).

302. Let f (z) be a single-valued continuous function and w0 (z) be one of the p
contin-
uous single-valued branches (with corresponding cuts) of the function w(z) = n f (z).
Find all single-valued continuous branches (with the same cuts) of the function w(z).

303. Find all branch points and

p construct the
√ schematic diagrams of the Riemann
surfaces of the functions: (a) z(z − i), (b) z 2 + 1.

304. Construct
√ schematic
p diagrams of the
p Riemann surfaces of the following func-
3
tions: (a) z − 1, (b) (z − 1) z, (c) (z 2 + 1)2 .
2 3 2 3

305. Find the continuous single-valued √

branches and construct a schematic diagram
of the Riemann surface of the function z 2 .

Note. From the solution

√ to problem 305 we see that the point z = 0 is not a branch
point of the function z 2 . At the same time, the images of curves passing through
the point z = 0 are not determined uniquely. For example, the continuous
√ images
of the piecewise-linear curve AOB in figure 30 under the map w = z are the2

piecewise-linear curves COD, COF , EOD, and EOF . When moving through the
point z = 0, we can stay on the same sheet (as in curves COD and EOF ) or move to
 10. RIEMANN SURFACES FOR MORE COMPLICATED FUNCTIONS 67

the other
√ sheet (as in curves COF and EOD). The Riemann surface of the function
2 †
w(z) = z has the form shown in figure 31.

B z D w

O A E O C

Figure 30.

Figure 31.

Definition 39. Points at which the single-valuedness of continuous images of curves

is violated are called singular points of the given function.
When constructing a schematic diagram of such a Riemann surface, we do not
need to make cuts to infinity, but simply to remove these points, that is to not allow
curves to pass through them.
306. Construct
√ schematic
√ diagrams
p of the Riemann surfaces p of the following func-
4
tions: (a) 4 z 2 + 2, (b) z 2 , (c) 4 (z − 1)2 (z + 1)3 , (d) 4 z(z 3 − 1).
Below we will also consider functions that are not defined at some points. It may
turn out that those points are branch points.
p
307. Construct a schematic diagram of the Riemann surface of the function 1/z.
308. Construct schematic diagrams of the Riemann surfaces of the following func-
tions: s
r r
1 z − 1 (z + i)2
(a) , (b) 3 , (c) 4 .
z−i z+1 z(z − 1)3
†
Editor’s footnote: In modern mathematics, Riemann√ surfaces are not permitted to have singu-
lar points such as these in them, so the function w = z 2 would not be considered.
√ This is similar
to the problems exhibited by the derivative of the real function y = |x| = x2 at x = 0.
68 CHAPTER 2. COMPLEX NUMBERS

When solving the problems in this section, we always discover that after making
non-intersecting cuts from all branch points to infinity, the function considered sep-
arates into single-valued continuous branches which are then connected in a certain
way along the cuts. It turns out that a rather large class of multi-valued functions has
this property. In particular, all of the functions considered below have this property,
namely functions that can be expressed in radicals (section 11) and the algebraic
functions (section 14).†
The proof of this statement is beyond the scope of this book. We could easily
refer the reader to existing literature on this matter‡ and accept the statement for-
mulated above without proof. (The reader may wish to do this and to skip ahead
to section 11.) However, the reader may then be left with some feeling of dissatis-
faction. Although we cannot free the reader completely from that feeling, we can
still show that the statement formulated above follows from another property, the
so-called monodromy property, which seems more intuitively obvious.
We know that to separate single-valued continuous branches of a multi-valued
function w(z) in some region of the z-plane, we need the function w(z) to be defined
by continuity in the same way along any two curves C1 and C2 lying in that region
and going from some point z0 to another point z1 . The monodromy property is
connected to this condition.
Definition 40. Suppose that w(z) is a multi-valued function such that if any of
its values w0 is fixed at an arbitrary point z0 , then the function can be determined
(perhaps non-uniquely) by continuity along any continuous curve which begins at
the point z0 and does not pass through any of the points at which w(z) is not defined.
We say that w(z) has the monodromy property if it satisfies the following condition.
Let C1 and C2 be continuous curves in the z-plane which both start at some
point z0 , end at some point z1 , and do not pass through any of the branch points or
singular points of the function w(z). Suppose that the curve C1 can be continuously
deformed into the curve C2 in such a way that the curves we get during the defor-
mation process do not pass through any of the branch points of the function w(z)
and such that the endpoints of the curve remain fixed. (See figure 32 as an ex-
ample, where a and b are branch points.) The intermediate curves may, however,
pass through non-branch singular points. Then if we fix the value of w(z0 ), the
value of w(z1 ) determined by continuity along the curve C1 will be the same as that
determined by continuity along the curve C2 .
Let us determine what follows from the property of monodromy.
309. Suppose that the function w(z) has the monodromy property. Let us make
non-intersecting cuts in the z-plane from all branch points of the function w(z) to

†
Both types of functions are special cases of a larger class of functions, the so-called analytic
functions, which also have these properties.
‡
See, for example, George Springer, Introduction to Riemann Surfaces, chapter 3.
 10. RIEMANN SURFACES FOR MORE COMPLICATED FUNCTIONS 69

z1
C2

a b

C1
z0

Figure 32.

infinity and also remove any singular points of the function. Prove that then the
function w(z) separates into single-valued continuous branches.
310. Suppose that, under the conditions of the previous problem, the cuts do not
pass through the singular points of the function w(z) and that w(z) has a finite
number of branch points. Prove that when crossing a cut (in a given direction), we
move from any fixed branch of the function w(z) to exactly the same other branch
independently of where we cross the cut.
Note 1. When walking around a branch point we cross the cut going from that point
to infinity precisely once. Therefore, because of problem 310, the movements from
one branch to another when crossing a cut in an arbitrary place coincide with the
movements that we get when walking around the branch point from which that cut
is made (in the corresponding direction), which therefore coincide with movements
noted at that point in the schematic diagram of the Riemann surface.
Note 2. From the results of problems 309 and 310, it follows that if a multi-valued
function w(z) has the monodromy property, then we can construct a Riemann surface
for w(z). Also, to find out the structure of this surface, it suffices to find the branch
points of the function w(z) and to determine the movements between the branches
of the function w(z) when walking around these points.
All functions which we will consider below have the monodromy property. We
cannot prove this statement formally here, because for that we would need the notion
of an analytic function. However, we can give an outline of the proof that a multi-
valued function w(z) has the property of monodromy, assuming that the function is
“sufficiently good”. What that means will be clear from the idea of the proof.
So suppose that we have satisfied the conditions which follow from the mon-
odromy property. Let C10 and C20 be continuous images of the curves C1 and C2
under the map w(z), starting at the same point w0 = w(z0 ). We need to prove that
the curves C10 and C20 end at the same point.
Suppose first that curves that we get by deforming C1 to C2 not only do not pass
through any branch points, but also do not pass through any singular points of the
function w(z) (see page 67). Let C be any of these curves. Then there exists a unique
70 CHAPTER 2. COMPLEX NUMBERS

continuous image C 0 of the curve C under the map w(z) which begins at w0 = w(z0 ).
If the function w(z) is “sufficiently good”,† then under continuous deformation of the
curve C from position C1 to position C2 , the curves C 0 are continuously deformed
from position C10 to position C20 . Thus the ending point of the curve C 0 is also
deformed continuously. But the curve C ends at the point z1 , so the ending point
of the curve C 0 must coincide with one of the images w(z1 ) of the point z1 . If
the function w(z) takes on only a finite number of values at every point z, and in
particular at z1 (and we are only considering such functions), then the ending point
of the curve C 0 cannot jump from one image of the point z1 to another, because then
the deformation would not be continuous. Therefore the ending points of all of the
curves C 0 , and in particular of the curves C10 and C20 , coincide.
We can now look at what happens when the curve C passes through a singular
point of the function w(z) which is not a branch point. We need only consider the
special case where the curve is changing only near the singular point a (see figure 33).
If at the point z0 we fix the value w0 = w(z0 ), then by continuity we can determine

z1
D E
a
A B

z0

Figure 33.

the value of w(z) at the point A. After that, we can determine the value of w(z) at
the point E by continuity along either of the curves ADE or ABE. These yield the
same value, because otherwise when walking around the curve EDABE, the value
of the function w(z) would change, and the point a would be a branch point of the
function w(z). As the value of w(z) at the point E is then determined, the value
of w(z) at the point z1 can now be determined by continuity along the curve Ez1 .
Thus the two curves z0 ABEz1 and z0 ADEz1 determine the same value of w(z1 ).
So the missing piece in our exposition is a justification of the fact that the func-
tions studied below are “sufficiently good”. This the reader has to take on trust or
turn to a deeper study of analytic functions.‡

†
The monodromy theorem is usually proved for arbitrary analytic functions. See, for example,
George Springer, Introduction to Riemann Surfaces, chapter 4.
‡
See, for example, George Springer, Introduction to Riemann Surfaces, chapters 4 and 5.
 11. FUNCTIONS EXPRESSIBLE IN RADICALS 71

§11. Functions expressible in radicals

Definition 41. Let f (z) and g(z) be two multi-valued functions. By f (z) + g(z),
we mean the multi-valued function, all of whose values at a point z0 are given by
adding each value of f (z0 ) to each value of g(z0 ). In a similar way, we define the
functions f (z) − g(z), f (z)g(z) and f (z)/g(z).
By (f (z))n , where n is a positive integer, we mean the function, all of whose
values atp the point z0 are given by raising each value of f (z0 ) to the nth power.
n
By f (z), where n is a positive integer, we mean the function, p all of whose
n
values at the point z0 are given by calculating all possible values of f (z0 ) for every
value of f (z0 ).
√
3
√ √ √ p √
311.pFind all values √ of: (a)
√ −8 + 2i, (b) (1 − −2i)/( −4), (c) i + −1,
(d) ( 4 (1 + i)2 )2 , (e) ( i + i)2 .
Definition 42. We say that a multi-valued function h(z) can be expressed in radicals
if we can produce it from the function f (z) = z and the constant functions fa (z) =
a (with a an arbitrary fixed complex number) using the operations of addition,
subtraction, multiplication, division, raising to a positive integer power and taking
a root of a positive integer degree.
p√ √
4
For example, the function h(z) = 3 z + 3z 2 − i/ z is expressed in radicals.
Some functions that can be expressed in radicals have already been studied above.
312. Suppose that the function h(z) can be expressed in radicals and that C is a
continuous curve in the z-plane, starting at the point z0 and not passing through
points z where h(z) is undefined. Prove that if w0 is one of the values of h(z0 ), then
there exists at least one continuous image of the curve C under the map w = h(z)
starting at the point w0 .
From the result of problem 312 we deduce that an arbitrary function h(z) that can
be expressed in radicals can be defined by continuity along any continuous curve C
that does not go through points where the function h(z) is undefined. If the curve C
does not pass through any branch points or singular points (see page 67) of the
function h(z), then the function h(z) is defined uniquely by continuity along the
curve C.
We already noted in previous section that functions that can be expressed in
radicals are “sufficiently good”,† i.e., they have the monodromy property. Therefore,
for any function that can be expressed in radicals, we can construct a Riemann
surface (see 309 and 310).‡ We will now determine the structure of these Riemann
surfaces.
From now on, we will assume that all functions being considered can be expressed
in radicals.
†
Functions that can be expressed in radicals are analytic.
‡
Any function that can be expressed in radicals only has a finite number of branch points.
72 CHAPTER 2. COMPLEX NUMBERS

313. Let h(z) = f (z) + g(z). Remove from the plane all singular points of the
function h(z) and make non-intersecting cuts to infinity from all points that are
branch points of at least one of the functions f (z) and g(z). Let f1 (z), . . . , fn (z)
and g1 (z), . . . , gm (z) be the continuous single-valued branches of the functions f (z)
and g(z) on the surface with cuts. Find the continuous single-valued branches of the
function h(z).
If when going around the point z0 we go from the branch fi1 (z) to the branch
fi2 (z) and from the branch gj1 (z) to the branch gj2 (z), then it is clear that from the
branch hi1 ji (z) = fi1 (z) + gj1 (z) we will go to the branch hi2 j2 (z) = fi2 (z) + gj2 (z).
This is suggested by the following formal method of constructing the schematic
diagram of the Riemann surface of the function h(z) = f (z) + g(z) given that the
schematic diagrams of f (z) and g(z) have already been constructed (with the same
cuts). For every pair of branches fi (z) and gj (z) we put into correspondence a sheet
where we define the branch hi,j (z) = fi (z) + gj (z). If in the schematic diagrams of
the Riemann surfaces of the functions f (z) and g(z) we have denoted movements at
the point z0 from the branch fi1 (z) to the branch fi2 (z) and from the branch gj1 (z)
to the branch gj2 (z), then in the schematic diagram of the Riemann surface of the
function h(z) we will denote at the point z0 a movement from the branch hi1 ,j1 (z) to
the branch hi2 ,j2 (z).
314.√ Construct
√ the schematic
√ diagrams
p of √
the Riemann
√ surfaces of√the functions:
√
3
(a) z + z − 1, (b) z − 1 + 1/z, (c) z + z, (d) z 2 − 1 + 4 z − 1.
2 3

The formal method of constructing a schematic diagram of the Riemann surface

of the function h(z) = f (z) + g(z) described above does not always give the right
result, because it does not account for the fact that some of the branches hi,j (z) may
turn out to be equal. For simplicity we will suppose that the cuts do not go through
the singular points of the function h(z). In that case, crossing any cut we will go from
sheets that correspond to equal branches of the function h(z) to other sheets also
corresponding to equal branches, because of uniqueness. Therefore, if we identify
sheets that correspond to equal branches of the function h(z), i.e., replace each such
set of sheets with just one sheet, then we will have uniquely defined movements
between the resultant sheets when going around any branch point z0 .
√ √ √ √
4 √
315.
√ Find all values of f (1) if: (a) f (z) = z + z, (b) f (z) = z + z 2 , (c) 3 z +
3
z.
316. For the following functions, construct both the schematic diagram of the Rie-
mann surface using the formal method and the correct √ schematic diagram of the
√ √ √ 4 √ √
Riemann surface: (a) f (z) = z + z, (b) f (z) = z + z 2 , (c) 3 z + 3 z.
Finally we see that to construct a schematic diagram of the Riemann surface of
a function h(z) = f (z) + g(z) using the schematic diagrams of the functions f (z)
and g(z) (constructed with the same cuts), it is sufficient to construct a schematic
 11. FUNCTIONS EXPRESSIBLE IN RADICALS 73

diagram using the formal method described above and then perform some identifi-
cations.
It is easy to see how this algorithm can also be used for constructing the schematic
diagrams of the Riemann surfaces of the functions h(z) = f (z) − g(z), h(z) =
f (z)g(z) and h(z) = f (z)/g(z).
317. Construct the schematic
√ diagrams of the √
√ Riemann√surfaces of the following
√ 4 √ √ √

functions: (a) i z − z , (b) z − 1 · z, (c) z 2 − 1/ 4 z + 1, (d)
2 4
z+ z
p3
z(z − 1).
318. Let f1 (z), f2 (z), . . . , fm (z) be all of the continuous single-valued branches of
the function f (z). Find, with the same cuts, all continuous single-valued branches
of the function h(z) = (f (z))n , where n is some positive integer.
From the result of the last problem, it follows that the schematic diagram of the
Riemann surface of the function h(z) = (f (z))n would coincide with the schematic
diagram of the Riemann surface of the function f (z) if all of the branches hi (z) =
(fi (z))n were different. However, this is not always the case. If we have equal
branches in h(z), then because of uniqueness we will always go from equal branches
to equal branches, as before.
Thus to construct the schematic diagram of the Riemann surface of the function
h(z) = (f (z))n , it is sufficient to look at the schematic diagram of the Riemann sur-
face of the function f (z), from the branches fi (z) getting branches hi (z) = (fi (z))n .
If some branches turn out to be equal, then we simply identify the corresponding
sheets.
319. Construct√the schematic √ diagrams
√ of√the √Riemann surfaces of the following
functions: (a) ( 4 z)2 , (b) ( z + z)2 , (c) ( z · 3 z − 1)3 .
p
Let us now study how the schematic diagram of the function n f (z) is connected
to the schematic diagram of the Riemann surface of the function f (z).
p
320. What are the possible branch points of the function n f (z)?
We introduce cuts in the z-plane from the branch points of the function f (z) to
infinity such that they do not pass through the points where one of the values of f (z)
is 0. Choose continuous single-valued branches of the function f (z); suppose these
are the (single-valued) functions f1 (z), f2 (z), . . . , fm (z). Let us make additional
cuts to infinity from the points where one of the values ofpf (z) is 0. Let g(z) be one
of the continuous single-valued branches of the function n f (z) under these cuts.
321. Prove that the function (g(z))n coincides with one of the functions fi (z) ev-
erywhere except for the cuts.

p The result of the previous problem implies that every branch of the function
n
f (z) corresponds to some branch of the function f (z).
74 CHAPTER 2. COMPLEX NUMBERS

p
322. Let g(z) be a continuous single-valued branch of the function n f (z) corre-
sponding to the branch fip
(z) of the function f (z). Find all continuous single-valued
branches of the function n f (z) corresponding to the branch fi (z).
It follows from the result of the preceding problem that for every branch fi (z) of
the
p function f (z), there is a corresponding “packet” of n branches of the function
n
f (z). We can enumerate the branches in this packet fi,0 (z), fi,1 (z), . . . , fi,n−1 (z)
so that for every k we have fi,k (z) = εkn · fi,0 (z).
Let z0 be a branch point of the function f (z), and suppose that walking around z0
takes
p us from the branch fi (z) to the branch fj (z). Then clearly for the function
n
f (z), walking around z0 starting on a sheet in the packet for the branch fi (z) will
move us to a sheet in the packet for the branch fj (z).
323. Let C be a curve in the z-plane given parametrically by the function z(t).
Suppose that the curve in the w-plane given parametrically
p by w0 (t) is a continuous
n
image of the curve C under the map w = f (z). Prove that the curve given
k
k (t) = εn · w0 (t) also is a continuous image of the curve C under the map
by wp
w = n f (z).
324. Suppose that the curve C in the p z-plane passes through neither branch points
nor singular points of the function n f (z). Prove that if moving along the curve C
takes us from the branch fi,s (z) to the branch fj,r (z), then it will also take us from
the branch fi,s+k (z) to the branch fj,r+k (z). Here the sums s + k and r + k are
evaluated modulo n (see example 9 on page 9).
p
Therefore to determine where a walk around a given branch point of n f (z) will
take us from any sheet of a given packet, it is sufficient to determine where such a
walk would take us from a single sheet in this packet, for walks from other sheets in
the same packet are then completely determined by the result of problem 324.
p√
325. Construct a schematic Riemann surface diagram for the function z − 1.
326.pConstruct schematic Riemann surface diagrams for the following functions:
3
√ p√
(a) z − 2, (b) 3
z − 1.
The following two problems consider cases where the Riemann surface diagram
of a function depends on the cuts made.
√
327. Construct the Riemann surface diagrams for the function f (z) = z 2 + 1 − 2
with the cuts pictured in figure 34 (a) and (b). In each case, determine whether the
points z with f (z) = 0 lie on the same or different sheets.
328. Using the pprevious problem, construct the Riemann surface diagrams for the
√
function h(z) = z 2 + 1 − 2 with the cuts pictured in figure 35 (a) and (b).
Let us restate the results of this section which we will need later on.
 11. FUNCTIONS EXPRESSIBLE IN RADICALS 75

i i
√ √ √ √
− 3 −i 3 − 3 −i 3

Figure (a) Figure (b)

Figure 34.

i i
√ √ √ √
− 3 −i 3 − 3 −i 3

Figure (a) Figure (b)

Figure 35.

Theorem 8. Given Riemann surface diagrams for the functions f (z) and g(z),
the following construction will yield Riemann surface diagrams for the functions
h(z) = f (z) + g(z), h(z) = f (z) − g(z), h(z) = f (z)g(z), and h(z) = f (z)/g(z) under
the same cuts as the original diagrams:
(a) for every pair of branches fi (z) and gj (z), define a sheet with the value of
the branch hi,j (z) on it equal to fi (z) + gj (z), fi (z) − gj (z), fi (z)gj (z) and
fi (z)/gj (z) respectively;
(b) if the walk around z0 takes us from the branch fi1 (z) to the branch fi2 (z)
and from the branch gi1 (z) to the branch gi2 (z), then, for the function h(z),
let the same walk take us from the branch hi1 ,j1 (z) to the branch hi2 ,j2 (z);
(c) identify the sheets where the branches hi,j (z) are equal.
Theorem 9. Given a Riemann surface diagram for the function f (z), the following
construction will yield a Riemann surface diagram for the function h(z) = (f (z))n
with the same cuts as the original diagram:
76 CHAPTER 2. COMPLEX NUMBERS

(a) for every branch fi (z), define a sheet with the value of the branch hi (z) on
it equal to (fi (z))n .
(b) identify the sheets where the branches hi (z) are equal.
Theorem 10. Given a Riemann surface diagram for the functionp f (z), the following
hold for the Riemann surface diagram for the function h(z) = n f (z) constructed
with the same cuts as the original diagram and additional cuts introduced as neces-
sary from the points where 0 is a value of f (z):
(a) every sheet of the Riemann surface diagram for f (z) is replaced by a packet
of n sheets;
(b) every walk around a branch point of the function h(z) takes us from packets
to packets consistently, that is, the packet we end on depends only upon the
starting packet and not on the particular sheet in it;
(c) these transitions from one packet to another coincide with the transitions
between the corresponding sheets of the Riemann surface diagram for f (z);
(d) if the sheets in the packets are numbered so that fi,k (z) = εkn · fi,0 (z), then
the transitions from one packet to another preserve the order of the sheets
up to a cyclic shift (see problem 324).

§12. Galois groups of multi-valued functions

We will now associate a certain permutation group with every Riemann surface
diagram.
329. Suppose that the curve C in the z-plane does not pass through any branch
point or singular point of the function w(z). Prove that if we start on different
sheets of the Riemann surface diagram and move along the continuous image of the
curve C under the function w(z), we will move to different sheets of the diagram.
Because of the result of this problem, it follows that to each anticlockwise walk
around a branch point of the function w(z), there is a corresponding permutation
of the sheets of the Riemann surface diagram for w(z), specifying for each sheet the
new sheet we walk to.
330. Consider the Riemann surface diagrams for the functions in problem 314. Sup-
pose that the sheets of these diagrams are numbered 1, 2, . . . from top to bottom.
For each of these functions, write down the permutation corresponding to a walk
about each branch point.
331. Let g1 , . . . , gs be elements of an arbitrary group G. Consider all elements of G
which may be obtained from g1 , . . . , gs by repeated applications of multiplication
and taking the inverse. Prove that the resulting collection of elements is a subgroup
in the group G.
Definition 43. The subgroup constructed in problem 331 is called the subgroup
generated by the elements g1 , . . . , gs .
 12. GALOIS GROUPS OF MULTI-VALUED FUNCTIONS 77

Definition 44. Let g1 , . . . , gs be the sheet permutations of a Riemann surface

diagram corresponding to anticlockwise walks around each of the branch points.
The subgroup of all possible sheet permutations generated by g1 , . . . , gs will be
called the sheet permutation group of the given Riemann surface diagram.
Note 1. If the number of sheets in a diagram is finite (and we only consider such
diagrams in this book), then for the construction of the sheet permutation group for
this diagram it is sufficient to use multiplication alone, without resorting to inverses.
Indeed in this case every sheet permutation has some finite order k: g k = e, thus
g −1 = g k−1 = g · g · . . . · g (k − 1 times).
Note 2. We will only consider the sheet permutation groups constructed below up
to isomorphism. Hence the sheet numbering will be unimportant since with different
numberings, the resulting subgroups of Sn are still isomorphic, even if different.
332. Which familiar groups are isomorphic to the sheet permutation
√ groups
√ of √
the
Riemann
√ surface diagrams of the p following functions: (a) z, (b) 3
z, (c) n
z,
3 2 2 3
(d) z − 1 (see problem 304),(e) (z − 1) (z + 1) (see problem 306)?
4

333. Which familiar groups are isomorphic to the sheet permutation groups of the
Riemann surface diagrams of the functions appearing in problems: (a) 314, (b) 317,
(c) 319?
334. Describe the
p√sheet permutation groups for the two Riemann surface diagrams
of the function z 2 + 1 − 2 constructed in problem 328.
Suppose that the point z0 is neither a branch point nor a singular point of the
multi-valued function w(z). Let w1 , w2 , . . . , wn be the values of w(z) at z0 . Consider
a continuous curve C starting and ending at the point z0 and passing through neither
branch points nor singular points of the function w(z). If we choose a value wi =
w(z0 ) and determine the value of w(z0 ) by continuity along the curve C, we will
obtain a new value wj = w(z0 ). Thus if we begin with different values wi , we
will obtain different values wj (otherwise the uniqueness of w(z) determined by
continuity along the curve C −1 would be violated). Consequently the curve C is
associated with a certain permutation of the values w1 , w2 , . . . , wn . Furthermore,
if a curve C is associated with a permutation g, then the curve C −1 is associated
with the permutation g −1 , and if the curves C1 and C2 (both with endpoints at z0 )
are associated with the permutations g1 and g2 respectively, then the curve C1 C2 is
associated with g2 g1 (recalling that we multiply permutations from right to left).
Therefore the associated permutations of all possible curves starting and ending
at z0 form a permutation group of the values of w(z0 ).
335. Let G1 be the permutation group of the values of w(z0 ) and let G2 be the sheet
permutation group of a Riemann surface diagram for the function w(z). Prove that
the groups G1 and G2 are isomorphic.
78 CHAPTER 2. COMPLEX NUMBERS

Note that to define the permutation group of the values of w(z0 ), we did not use
any Riemann surface diagrams for the function w(z). Therefore, by problem 335, it
follows that the permutation group of the values of w(z0 ) for an arbitrary point z0
and the sheet permutation group for an arbitrary Riemann surface diagram of the
function w(z) are isomorphic. Consequently, the permutation groups of the values
of w(z0 ) for all points z0 and the sheet permutation groups for all Riemann surface
diagrams of the function w(z) are isomorphic, and we may view them as the same
group. We call this group the Galois group of the multi-valued function w(z).†

§13. Galois groups of functions expressible in radicals

We have now come to one of the main results of this book.
Theorem 11. If a multi-valued function h(z) can be expressed in radicals, then its
Galois group is solvable (see section 14 of chapter 1).
The proof of Theorem 11 is contained in the solutions to the following problems.
336. Suppose that f (z) and g(z) are multi-valued functions, and that either h(z) =
f (z) + g(z) or h(z) = f (z) − g(z) or h(z) = f (z)g(z) or h(z) = f (z)/g(z). Con-
sider the formally constructed Riemann surface diagram of the function h(z), as in
Theorem 8 (a) and (b), page 75. Prove that if F and G are the sheet permutation
groups of the Riemann surface diagrams of the functions f (z) and g(z), then the
sheet permutation group of the formally constructed diagram is isomorphic to some
subgroup of the direct product F × G (see section 7 of chapter 1).
337. Under the conditions of the previous problem, let H1 be the sheet permutation
group of the formally constructed diagram and let H2 be the sheet permutation
group of the true Riemann surface diagram for the function h(z). Prove that there
is a surjective homomorphism (see section 13 of chapter 1) from the group H1 onto
the group H2 .
338. Suppose that the Galois groups of the functions f (z) and g(z) are solvable.
Prove that the Galois groups of the following functions are also solvable: h(z) =
f (z) + g(z), h(z) = f (z) − g(z), h(z) = f (z) · g(z) and h(z) = f (z)/g(z).
339. If the Galois group of a function f (z) is solvable, prove that the Galois group
of h(z) = (f (z))n is also solvable.
340. Let H be the p sheet permutation group of the Riemann surface diagram for
the function h(z) = n f (z). Let F be the sheet permutation group of the Riemann
surface diagram for the function f (z), constructed with the same cuts. Construct a
surjective homomorphism from the group H onto the group F .
†
This group is also called the monodromy group.
 14. ABEL’S THEOREM 79

341. Prove that the kernel of the homomorphism (see section 13 of chapter 1) con-
structed in the previous problem is commutative.
342. Suppose the Galois group of the
p function f (z) is solvable. Prove that the
n
Galois group of the function h(z) = f (z) is also solvable.
The constant functions fa (z) = a and the function f (z) = z are single-valued
and continuous in the entire z-plane, therefore their Riemann surfaces have only
one sheet, and consequently the corresponding Galois groups are trivial (just {e})
and hence solvable. From this, recalling the definition of functions expressible in
radicals (page 71), and the results of problems 338, 339, and 342, we have proven
Theorem 11.
Note. Let us point out the following fact to a reader familiar with the theory of
analytic functions. If the Galois group is defined as the permutation group of the
values of h(z) at some point z0 (see page 78), then Theorem 11 will hold for a wider
class of functions. For example, in constructing the function h(z), in addition to
constants, the identity function, the arithmetic operations and radicals, we may use
any single-valued analytic functions (e.g., ez , sin z, etc.), the multi-valued function
log z, and certain other functions. The Galois group of h(z) will still be solvable,
though not necessarily finite.

§14. Abel’s theorem

Consider the equation
3w5 − 25w3 + 60w − z = 0. (14.1)
We will consider z to be a parameter, and for every complex z, will look for all
complex roots w of this equation. By problem 269, this equation has 5 roots (counting
multiplicities) for every z.
343. Which values of w may be multiple roots (roots of multiplicity greater than 1,
see page 56) of the equation 3w5 − 25w3 + 60w − z = 0? For which values of z are
they multiple roots?
From the solution of the above problem, we know that equation (14.1) has 4 dif-
ferent roots when z = ±38 and z = ±16, and it has 5 different roots for other
values of z. Thus the function w(z) expressing the roots of (14.1) in terms of the
parameter z takes 4 different values at z = ±38 and z = ±16, and takes 5 different
values at other z. Let us study this function w(z). First, we will prove that small
variations of z result in only small variations of the roots of (14.1). This property is
more precisely expressed in the following problem.
344. Let z0 be an arbitrary complex number and let w0 be one of the roots of (14.1)
at z = z0 . Consider a disc of a small radius r centred at w0 . Prove that there is a
real number ρ > 0 such that if |z00 − z0 | < ρ, then there is at least one root of (14.1)
at z = z00 lying inside the disc.
80 CHAPTER 2. COMPLEX NUMBERS

Let w(z) express the roots of equation (14.1) in terms of the parameter z, and
let w0 be one of the values w(z0 ). Then the result of problem 344 implies that if
z moves continuously along a curve from z0 , then it is possible to keep choosing
one of the values of w(z) so that the point w also moves continuously along a curve
from w0 . In other words, w(z) may be defined via continuity along any curve C. If,
in addition, the curve C does not pass through any branch points or singular points
of the function w(z) (see page 67), then w(z) will be uniquely defined along C.
345. Prove that if z 6= ±38 and z 6= ±16, then z is neither a branch point nor a
singular point for the function w(z).
The function w(z) is algebraic† and therefore “well-behaved” (see section 10,
page 69), i.e., it has the monodromy property. Therefore it is possible to construct a
Riemann surface for w(z) (see problems 309 and 310). This surface will clearly have
5 sheets.
By problem 345, z = ±38 and z = ±16 are the only possible candidates for
branch points or singular points of the function w(z). Nevertheless, it is not yet
clear whether these points are indeed branch points or singular points.
346. Suppose it is known that z0 = +38 (or z0 = −38, or z0 = ±16) is a branch
point for the function w(z). How are the sheets of the Riemann surface of w(z)
glued at z0 ? (To be more precise, how are the sheets glued along the cut from z0 to
infinity? See note 2 on page 69.)
347. Suppose that z0 and z1 are arbitrary points different from z = ±38 and z =
±16, and let w0 and w1 be any of their respective images under the map w(z). Prove
that there is a continuous curve from z0 to z1 which does not pass through z = ±38
and z = ±16 and whose continuous image, beginning at w0 , ends at w1 .
348. Prove that all four points z = ±38 and z = ±16 are branch points for the
function w(z). What can a Riemann surface diagram for w(z) look like? Sketch all
possible cases. (We consider two diagrams different if one cannot be obtained from
the other by a permutation of sheets and of branch points.)
349. Find the Galois group for the function w(z).
350. Prove that the function w(z) cannot be expressed in radicals.
351. Prove that the general polynomial equation of degree five
a5 w5 + a4 w4 + a3 w3 + a2 w2 + a1 w + a0 = 0
(a0 , a1 , a2 , a3 , a4 , a5 are complex parameters, a5 6= 0) is not solvable in radicals, i.e.,
there is no formula which expresses the roots of the above equation in terms of the
†
A multi-valued function w(z) is called algebraic if it expresses, in terms of the parameter z,
all roots of some equation
an (z)wn + an−1 (z)wn−1 + · · · + a0 (z) = 0,
where each ai (z) is a polynomial in z. All algebraic functions are analytic.
 14. ABEL’S THEOREM 81

coefficients using the operations of addition, subtraction, multiplication, division,

raising to the power of a natural number and taking radicals of the degree of a
natural number.
352. Considering the equation
(3w5 − 25w3 + 60w − z)wn−5 = 0,
prove that for n > 5, the general polynomial equation of degree n is not solvable in
radicals.
The results of problems 351 and 352 comprise the central statement of this book.
We have proven the following theorem.
Abel’s Theorem. When n > 5, the general polynomial equation of degree n,
an wn + an−1 wn−1 + · · · + a1 w + a0 = 0,
is not solvable in radicals.
Note 1. In the introduction, we obtained Cardano’s formula for solving the general
cubic equation. Recall that among the values given by the formula, not all were
roots of the equation, but only those for which some additional condition held. It is
therefore reasonable to ask whether, for the general polynomial equation of degree n,
there is a formula in radicals whose values include the roots of the equation as a
subset. Let us show that no such formula exists even for the equation (14.1).
As before, let w(z) express the roots of equation (14.1) in terms of the parame-
ter z. If the values of w(z) form a subset of values of some function w1 (z) expressible
in radicals, then the Riemann surface for the function w(z) is a separate piece of the
Riemann surface for w1 (z). If G is the Galois group of the function w1 (z), then each
permutation in G yields a permutation of the five sheets corresponding to w(z). This
map is a homomorphism from the group G onto the group S5 . Since the group S5
is not solvable, the group G is also not solvable (see problem 163). On the other
hand, G must be solvable as the Galois group of a function expressible in radicals,
and thus we have a contradiction.
Note 2. The note on page 79 implies that Abel’s theorem remains true if, in addition
to radicals, we allow the use of certain other functions, such as any single-valued
analytic functions (ez , sin z, etc.), the function log z and some others.
Note 3. Consider equation (14.1) over the real numbers only. Suppose the func-
tion y(x) expresses the real roots of equation
3y 5 − 25y 3 + 60y − x = 0
in terms of the real parameter x. Perhaps it is possible to express the function y(x)
in radicals? The answer again turns out to be no. For the reader familiar with the
theory of analytic functions, we note that this fact follows from analytic continuation.
The function w(z) expressing the roots of equation (14.1) in terms of the parame-
ter z is an analytic function. Therefore, had the function y(x) been expressible in
82 CHAPTER 2. COMPLEX NUMBERS

radicals, then the corresponding formula, viewed over the complex numbers, would
have defined the function w(z) by the uniqueness of analytic continuation. But this
would mean that w(z) is expressible in radicals.
Therefore Abel’s theorem remains true even if we only consider the real roots of
the general nth degree equation (n > 5) with real coefficients. In addition, as in
note 2, the theorem remains true even if we allow the use of some other functions,
for example, any functions which admit a single-valued analytic continuation (ex ,
sin x, etc.), log x and others.
Note 4. The class of algebraic functions (see the footnote on page 80) is a very rich
and interesting class. In particular, one can show that all functions expressible in
radicals are algebraic. We have proved that any function expressible in radicals has
a solvable Galois group (Theorem 11, page 78). It turns out that if we limit our
attention to algebraic functions, the converse is also true: if a Galois group of an
algebraic function is solvable, then the function is expressible in radicals. Therefore
an algebraic function is expressible in radicals if and only if its Galois group is
solvable. This statement is one of the results of general Galois Theory.
 Index

Abel’s Theorem, xvii, 81 difference of polynomials, 34

abelian, 8 direct product, 13
addition modulo n, 9 discriminant, xiv
algebraic form of complex number, 38
algebraic function, 80 elementary transpositions, 28
alternating group of degree n, 29 even permutation, 29
argument of complex number, 44 expressed in radicals, 71
associative, 6
factor group, 19
bijective, 4 field, 32
binary operation, 1 field of complex numbers, 37
branch point, 64 field of rational functions, 41
finite, 6
Cardano’s Formula, xvi free vector, 43
change in argument along a curve, 51 Fundamental Theorem of Algebra, 55
Chinese Remainder Theorem, 14
Galois group, 78
coefficient of the imaginary part, 38
general algebraic equation, xiii
coefficients of polynomial, 33
generator, 9
commutative, 8
geometric representation of complex
commutator, 20
numbers, 41
commutator subgroup, 20
group, 6
commute, 8, 20
group of integers, 7
complex numbers, 37
group of rotations of a tetrahedron, 13
complex conjugate, 38
group of symmetries of a tetrahedron, 13
composition, 2, 5
group of transformations, 5
composition of functions, 47
conjugate, 16 homomorphism, 21
conjugation, 16
continuous curve, 48 identity element, 6
continuous function, 46 image, 4, 24
coprime, 14 imaginary part, 38
cycles, 28 independent cycles, 28
cyclic group, 9 infinite, 6
cyclic permutations, 28 infinite cyclic group, 9
infinite order, 9
degree of polynomial, 40 inner automorphism, 16
derivative, 56 inverse, 6
determined by continuity, 60 inverse transformation, 5
difference of elements, 33 inversion, 29
difference of functions, 47 irreducible over a field, 40

83
84 INDEX

isomorphic, 10, 39 singular points, 67

isomorphic map, 39 solvable, 25
isomorphism, 10, 39 subgroup, 11
subgroup generated by elements, 76
kernel of a homomorphism, 22 sum of functions, 47
sum of polynomials, 34
leading coefficient, 33
surjective, 4
left coset, 14
symmetric group of degree n, 28
left coset decomposition, 14
symmetry, 3
map, 4
transformation, 5
minimal field, 40
transpositions, 28
modulus, 43
trigonometric representation of complex
monodromy group, 78
numbers, 44
monodromy property, 68
multiplication modulo n, 32 vector, 42
natural homomorphism, 22 well-formed products, 7
normal subgroup, 17

odd permutation, 29
order, 6, 9
order of subgroup, 14
ordered pairs, 1

parametric equation, 49
permutation, 5
permutations of degree n, 27
polynomial, 33
preimage, 4
product, 6
product of functions, 47
product of polynomials, 34

quotient group, 19
quotient of elements, 33
quotient of functions, 47

real part, 38
reducible over a field, 40
relatively prime, 14
Riemann surface, 59
right coset decomposition, 15
right cosets, 15
root of equation, 34
root of multiplicity k, 56
root of polynomial, 34

set of complex numbers, 36

sheet permutation group, 77
sheets, 58
single-valued continuous branches, 58