Shaum Group Theory

^^,:v-.
,--.'.-...-
..,.-;.^. ---..-
..
iHAUM'S
OUTLINE
SERIES
THEORY and PROBLEMS

of
GROUP
THEORY
by
B.
BAUMSLAG and
B.
CHANDLER
including
600
problems
Complefely Solved
in
Detail
SCHAUM'S OUTLINE
SERIES
McGRAW-HlLL BOOK COMPANY
1 .'^
SCHAVM'S OVTLIISE OF
THEORY
AIVD
OF
PROBLEMS
GROrP THEORY
BY
BENJAMIN BAUMSLAG, Ph.D. BRUCE CHANDLER, Ph.D.

Department of Mathematics
New York
University
SCHAVM'S OUTLIBTE SERIES

McGRAW-HILL BOOK COMPANY
New
York,
St. Louis,
San Francisco, Toronto, Sydney
Copyright 1968 by McGraw-Hill, Inc. All Rights Reserved. Printed in the United States of America. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the
prior written permission of the publisher.
04124
34567890 MHUN
8 2
10
6 9
Typography by Signs and Symbols,
Inc.,
New
York, N. Y.
Preface
The study of groups arose early in the nineteenth century in connection with the solution of equations. Originally a group was a set of permutations with the property that the combination of any two permutations again belongs to the set. Subsequently this definition was generalized to the concept of an abstract group, which was defined to be a set, not necessarily of permutations, together with a method of combining its elements that is subject to a few simple laws.
The theory of abstract groups plays an important part in present day mathematics and science. Groups arise in a bewildering number of apparently unconnected subjects. Thus they appear in crystallography and quantum mechanics, in geometry and topology, in analysis and algebra, in physics, chemistry and even in biology.
One of the most important intuitive ideas in mathematics and science is symmetry. Groups can describe symmetry; indeed many of the groups that arose in mathematics and science were encountered in the study of symmetry. This explains to some extent why groups arise so frequently.
in connection with other disciplines, the study of groups is in Currently there is vigorous research in the subject, and it attracts the energies and imagination of a great many mathematicians.
itself exciting.
Although groups arose
designed for a first course in group theory. It is mainly intended for year graduate students. It is complete in itself and can be used for self-study or as a text for a formal course. Moreover, it could with advantage be used as a supplement to courses in group theory and modern algebra. Little prior knowledge is assumed. The reader should know the beginnings of elementary numbet theory, a summary of which appears in Appendix A. An acquaintance with complex numbers is needed for some problems. In short, a knowledge of high school mathematics should be a sufficient prerequisite, and highly motivated and bright high school students will be able to understand much of this book.
is
This book
college
and
first
The aim of this book is to make the study of group theory easier. Each chapter begins with a preview and ends with a summary, so that the reader may see the ideas as a whole. Each main idea appears in a section of its own, is motivated, is explained in great detail, and is made concrete by solved problems.
Chapter 1 presents the rudiments of set theory and the concept of binary operation, which are fundamental to the whole subject. Chapter 2, on groupoids, further explores the concept of binary operation. In most courses on group theory the concept of groupoid is usually treated briefly if at all. We have chosen to treat it more fully for the following reasons: (a) A thorough understanding of binary compositions is thereby obtained, (b) The important ideas of homomorphism, isomorphism and Cayley's theorem occur both in the chapter on groupoids and in the chapters on groups, and the repetition ensures familiarity.
Chapter 3 shows that the concept of group is natural by producing a large number of examples of groups that arise in different fields. Here are discussed groups of real and complex numbers, the symmetric groups, symmetry groups, dihedral groups, the group of Mobius transformations, automorphism groups of groupoids and fields, groups of matrices, and the full linear group. Chapter 4 is concerned with the homomorphism theorems and cyclic groups. The concept of homomorphism is fundamental, and thus the theorems of this chapter are indispensable for further study.
finite groups. The Sylow theorems are proved, the concept of external introduced, and groups up to order 15 are classified. The chapter concludes with the Jordan-Holder theorem and a proof that most alternating groups are simple.
Chapter 5
is
on
direct product
is
finitely
Chapter 6 is on abelian groups. Two important classes of abelian groups are treated: generated and divisible groups. Undergraduates will probably find their needs are met by the material through Section 6.3. Graduate students will certainly want to
continue.
Chapter 7 is on permutational representations and extensions. Chapter 8 is on free groups and presentations. Those who would like to study the theory of groups more deeply will find a guide to the literature at the end of the book.
Chapters 1-4 must be read in order, although, if desired, only the first three sections of Chapter 3 need be read at first (the other sections of Chapter 3 may be studied when they are needed). The order of reading Chapters 5-8 can be varied, although part of Chapter 7 is required for the last sections of Chapter 8.
The reader need not work all the solved problems; he should decide for himself how much practice he needs. Some of the problems are designed to clarify the immediately preceding text, and the reader will find that the solutions may overcome some of his
advisable to attempt the problems before reading their solutions. The numerous supplementary problems, some of which are very difficult, serve as a review of the material of each chapter.
obstacles.
On
the whole, however,
it is
thank Prof. Gilbert Baumslag for giving us access to several chapters of unpublished notes and for many useful suggestions. We thank Sister Weiss for reading two chapters of an early draft, Harold Brown for much helpful advice, Henry Hayden for typographical arrangement and art work, and Louise Baggot for the typing. Finally we express our appreciation to Daniel Schaum and Nicola Monti for their unfailing editorial
We
cooperation.
B.
B.
Baumslag Chandler
June 1968
CONTENTS
Page
Chapter
SETS,
MAPPINGS AND BINARY OPERATIONS

1
Preview of Chapter
1.1
1 1
SETS
a.
Basic notions,
b.
Union and
intersection.
6
c.
1.2
CARTESIAN PRODUCTS
a. Definition, d. b.
Equivalence relations,
Partitions and equivalence relations.
The
division notation.
11
b.
1.3
MAPPINGS
a.
Definition of mapping,
Formal
definition
of
mapping,
c.
Types of
17
mappings.
1.4 1.5
COMPOSITION OF MAPPINGS BINARY OPERATIONS

a. Definition,
19
table.
b.
The multiplication
1
A
Chapter
look back at Chapter
24
GROUPOIDS
Preview of Chapter 2
2.1
26
26
b.
GROUPOIDS
a.
Definition of a groupoid.
Equality of groupoids.
29
2.2
2.3
IDENTITIES
a.
COMMUTATIVE AND ASSOCIATIVE GROUPOIDS AND INVERSES IN GROUPOIDS

The identity of a groupoid.
Uniqueness of inverses, Notation for a mapping,
Definition of a
b. b.
30
Inverses in a groupoid.
33 of a set into itself,
2.4
SEMIGROUPS WITH AN IDENTITY ELEMENT

a.
c.
The semigroup of mappings d. The order in a product.

b.
2.5
HOMOMORPHISMS OF GROUPOIDS AND CAYLEY'S THEOREM

a.
40
homomorphism.
morphism, c. Properties of and semigroups.
Epimorphism, monomorphism, and isoepimorphisms. d. Naming and isomorphisms, e. Mx

47
A
Chapter
look back at Chapter 2
GROUPS AND SUBGROUPS

3.1
50
GROUPS
Definition.
50
Examples of groups of numbers.

54 56
c.
3.2
3.3
SUBGROUPS THE SYMMETRIC AND ALTERNATING GROUPS

The symmetric group on X. b. Even and odd permutations, nating groups, d. The order of A.
a.
The
alter-
3.4
GROUPS OP ISOMETRIES
Isometrics of the line. b. Two points determine an isometry. c. Isometries of the plane, d. Isometries are products of reflections, translations and rotations, e. Symmetry groups, f. The dihedral groups.
a.
64
3.5
THE GROUP OF MOBIUS TRANSFORMATIONS

a.
77
Defining the group,
b.
2 matrices. 83
3.6
SYMMETRIES OF AN ALGEBRAIC STRUCTURE

a.
Automorphisms of groupoids.
fields,
d.
b.
phisms of
group.
Vector spaces,
Fields of complex numbers, c. AutomorThe full linear e. Linear transformations.
91
CONTENTS
Chapter
(SOMORPHISM THEOREMS
4.1
Page
94 94
FUNDAMENTALS
a.
Preliminary remarks,
b.
More about subgroups,

b.
c.
Exponents.
101
4.2
CYCLIC GROUPS
a.
Fundamentals of
cyclic groups,
Subgroups of
cyclic groups.
4.3
COSETS
Introduction to the idea of coset. b. Cosets form a partition. Lagrange's theorem, c. Normal subgroups, d. Commutator subgroups, centralizers, normalizers. e. Factor groups.
a.
107
4.4
HOMOMORPHISM THEOREMS
factor groups: The homomorphism theorem, b. Correspondence theorem. Factor of a factor theorem, c. The subgroup isomorphism theorem, d. Homomorphisms of cyclic groups.
a.
117
Homomorphisms and
A
Chapter
127
FINITE
GROUPS
5
Preview of Chapter
5.1
130
130
THE SYLOW THEOREMS

a.
Statements of the Sylow theorems, b. Two lemmas used in the proof of the Sylow theorems, c. Proofs of the Sylow theorems.
5.2
THEORY OF p-GROUPS
a.
c.
139
in finite groups,
b.
The importance of p-groups The upper central series.
The center of a p-group.

143
5.3
DIRECT PRODUCTS AND GROUPS OF LOW ORDER

a.
c.
Direct products of groups, b. Groups of small order: orders p and 2p. Groups of small order: orders 8 and 9. d. Groups of small order: orders 12
15.
and
5.4
SOLVABLE GROUPS
a.
158
b.
Definition of solvable groups,
Properties of, and alternative definition
for, solvable groups.

5.5
COMPOSITION SERIES AND SIMPLE GROUPS

The Jordan-Holder theorem, b. Proof of Jordan-Holder theorem, c. Cycles and products of cycles, d. Transpositions, and even and odd permutations, e. The simplicity of A, n 5.
a.
163
A
Chapter
174
ABELIAN GROUPS
6.1
177
PRELIMINARIES
Additive notation and finite direct sums. b. Infinite direct sums. homomorphic property of direct sums and free abelian groups.
a.
c.
178
The
6.2
SIMPLE CLASSIFICATION OF ABELIAN GROUPS,
AND STRUCTURE OF TORSION GROUPS

Tentative classifications: torsion, torsion-free, and mixed, subgroup, c. Structure of torsion groups. Priifer groups, and rank.
a. b. d.
188
The torsion
Independence
196
6.3
a.
FINITELY GENERATED ABELIAN GROUPS Lemmas for finitely generated free abelian groups,
b.
Fundamental theorem
d.
of abelian groups, c. The type of a finitely generated abelian group, groups of finitely generated abelian groups.
6.4
Sub205
DIVISIBLE GROUPS
a.
groups. divisible groups.
p-Prufer
Divisible
subgroups,
b.
Decomposition
theorem for
211
CONTENTS
Chapter
PERMUTATION AL REPRESENTATIONS
Preview of Chapter
7.1
Page
214 214
b.
CAYLEY'S THEOREM
a.
Another proof of Cayley's theorem,
Cayley's theorem and examples of
groups.
7.2
7.3
PERMUTATIONAL REPRESENTATIONS DEGREE OF A REPRESENTATION AND FAITHFUL REPRESENTATIONS..

a.
216 217
Degree of a representation,
h.
Faithful representations.
218 222
7.4
7.5
PERMUTATIONAL REPRESENTATIONS ON COSETS FROBENIUS' VARIATION OF CAYLEY'S THEOREM

a.
The kernel
of a coset representation,
b.
Frobenius' theorem.
227
7.4.
7.6
APPLICATIONS TO FINITELY GENERATED GROUPS

a.
c.
Subg-roups of finite index, Marshall Hall's theorem, d.
Remarks about the proof of Theorem One consequence of Theorem 7.5.

b.
7.7
EXTENSIONS
a.
232
b.
General extension,
d.
The
splitting extension,
c.
An
analysis of splitting
extensions,
7.8
Direct product.
240
t is
THE TRANSFER
a. Definition,
b.
Proof that
of the choice of transversal,
a homomorphism. c. Proof that t d. A theorem of Schur.
is
independent
A
Chapter
243
FREE
8.1
GROUPS AND PRESENTATIONS

245
ELEMENTARY NOTIONS
a. Definition of a free group, b. Length of an element. Alternative description of a free group, c. Existence of free groups, d. Homomorphisms of free groups.
245
8.2
PRESENTATIONS OF GROUPS
a. Definitions, b.
253
Illustrations of presentations.
8.3
8.4
THE SUBGROUP THEOREM FOR FREE GROUPS: AN EXAMPLE PROOF OF THE SUBGROUP THEOREM FOR FREE GROUPS
Plan of the proof, b. Schreier transversals, c. A look at the elements a^^. d. The proof of the subgroup theorem, e. Subgroups of finite index, f. Intersection of finitely generated subgroups.
a.
259
260
A
Appendix
266
A
B
NUMBER THEORY
..
269
Appendix
A GUIDE TO THE Lil LITERATURE

INDEX
270
274
SYMBOLS AND NOTATIONS
278
chapter
Sets,
Preview of Chapter
Mappings and Binary Operations
This chapter begins with a few remarks about sets. A set is a collection of objects. For example, the real numbers form a set, the objects being the numbers. The real numbers have an operation called addition. Addition essentially involves two
numbers, for the addition of a single number

or
is
meaningless, while the addition of three
more numbers is repeated addition of two numbers. Because addition involves two numbers it is called a binary operation. The main object of this chapter is to define precisely the notion of a binary operation. The concept of binary operation is required to define the concept of group. We introduce the important ideas of cartesian product and mapping. Welding them
together gives rise to an explicit definition of a binary operation. Another important idea is that of equivalence relation, which is a generalization of the idea of equality. The reader will also pick up much useful notation.
1.1 a.
SETS
Basic notions
is
Set the set. denote

(i)
synonymous with
Usually
we
collection. The objects in a set are termed the elements of denote sets by capital Latin letters, for example, B, G, T. We shall
the set of positive integers the set of the set the set
the set
all
1, 2, 3, ...
by
P
by
(ii)
the set of nonnegative integers

integers
0, 1, 2, ...
(iii)
(iv)
(v)
(vi)
by Z of rational numbers by Q of real numbers by R of complex numbers by C.

of a set will usually be denoted "s
is
by small Latin letters such as s, t, u, etc. an element of S" or "s belongs to S". In particular, 2 GP. It s is not an element of S, we write s ^ S and read this as "s is not an element of S" or "s does not belong to S" or "s is not in S". For example, 1 P. In dealing with sets it is advantageous to abbreviate the phrase "the set whose elements are" by using braces. Thus, for example, we write {1,2} for the set whose elements are 1 and 2 and similarly we write {1, 0, 1, 2, ... } for the set whose elements are 1, 0, 1, 2, ... A variation of this notation is useful to describe a set in terms of a property which singles out its elements. Thus we write {x x has the property T} for the set of all those elements X which have the property 'P. Here "iP stands for some "understandable" property; to (T here is the property of illustrate: {a; a; is a real number} is R, the set of real numbers. being a real number.) Notice that we read {a; a; is a real number} as the set of all those elements x which have the property that a; is a real number, or the set of all those elements X such that a; is a real number.
The elements
s
By
G S we mean
SETS,
[CHAP.
We
(i)
can
now
introduce some useful notation.

{a, h)
If a
(a, 6)
and b are real numbers and a <b, then the open interval = {x\ X E.R and a <x <h}.
is
defined by
(ii)
Again if a and 6 are real numbers and a <h, then the dosed interval by [a,b] = {x\ x &R and a x h).
In coordinate geometry (0, 0) denotes the origin, In (0, 1) the point A, and (1, 0) the point B. general, if a, h are any two elements of a set, (a, h) is called the ordered pair formed from a and h. We will say (a, h) = (c, d) if and only if a = c and b d. (a, b) is called an ordered pair because the order of a and b matters; (a, 6) is not the same as {b,a) if a'b.
[a, h]
is
defined
(iii)
This idea enables us to define the Euclidean plane as {p\ p = {x, y) where x,y E R}, where the distance between {x,y) and {xi,yi) is defined by \/{x xi)^ + {y V\Y- We write R'^ = {p\ p = {x,y) with x,y G R}. R^ is not the Euclidean plane we normally think of. Rather it can be interpreted as the set of coordinates of the Euclidean plane. Having defined the Euclidean plane it is easy to define, for example, circles and discs.
(iv)
circle
with radius r and center at the origin
is
defined
by by
{p\
p p
= =
{x, y)
G R^ and G R^ and
(v)
The
x^
disc with radius r
and center at the origin
is
defined
{p\
{x, y)
+ y^ ~ r^}.
sets
to
S and T are equal, and write S=T, if every element of S belongs T and every element of T belongs to S. Thus {2,2,3,3} = (2,3} = {3,2}. If every element of the set S is also an element of the set T, we say S is a subset of T and
We say that two
express this briefly by writing
ScT.
ScT
T
if
means
ScT
if
but
S^T.
Thus
sets.
PcN
and
PgN. We
can use the notion of subset to give a criterion for the equality of
1.1:
Proposition
Set
is
equal to set
and only
SqT
and
TqS.
S belongs
to
Proof: every element of
SqT
T
and T
CS expresses in symbols "every element of belongs to S".

VêQ
(\/=l)2G2
T and
1.1.
Problems Are the following statements true?

(i)
{2}
(iv)
{a, 6}
(vii)
(viii)
(ii) (iii)
3e{2,4}
z
(v)
5ep
{a, b}, z =
a and
z = h
(vi)
feP
(ix)
(\/^)2ep
not a rational number.
is
Solution:
(i)
True
False
(iv)
False
(vii)
(viii)
False, since yfi
is
(ii) (iii)
(v)
(vi)
True
False
True, for
False,
True
(ix)
(V^)^ = -1 and -1 since -1 S P.

QCQ
{3, a, b, c}
an integer.
1.2.
Checli the truth of the following assertions,

(i)
(ii)
If
{2} is a subset of {2}. S is any set, S c S.
(v)
(vi)
(vii)
(viii)
PcQ
Z <zQ
(ix)
(x)
{3, a, b, 3, c, 6}
(iii)
{a}
{a, b},
(iv)
{2, 3}
{3, 4}
QcR RoC
{2}.
Solution:
(i)
True.
True.
The only element of
{2}
(ii)
Any
element in
is
is 2, and 2 G an element in S.
Sec. 1.1]
SETS
True, a is the only element of {a} and quently {a.} ^ {o, 5}.
False.
2
(iii)
{a, b}.
But
{a, 6}
and
{a}.
Conse-
(iv)
{2,3} and 2S{3,4}.

positive integer
is
(v)
True. True. True.
Any
a rational
number but not
all
rational
numbers are
positive
integers.
(vi)
(vii)
(viii)
All integers are rational numbers. All rational numbers are real numbers but
Q.
True.
False.
For
if
e
c
R, then
+ Ot G
C.
But
V-i ^ R
implies
C.
(ix)
Q =
3, a, b
Q.
(x)
True.
and
are the only elements of
{3, a, b, 3, c, 6}.
Therefore the sets are equal.
1.3.
Are
(i)
the following statements true?
{x\ X {x\ {x\ X {x\ {x\ X

{a;
I
is
(ii)
N =
Q = P = C = Z =
False.
xeQ
= =
is
(iii)
(iv)
xGN
(v)
(vi)
a;
and x > 0} and x^ 0} alb, where 6 # and a,b e Z} and x^^ 1} u + iv, where u,v G R and fi = 1} real and x^ S P}
real
Solution:
(i)
{x\
(ii)
False,
(iii)
True.
f S {x
and a; > 0} has no negative elements. and x ^ 0} but f N. Hence the sets are not equal. rational number is defined as the set of all numbers of the form a/b where
a;
is real
|
GQ
b =
(iv)
(y)
(vi)
and a,b & Z. True. For if x & P, then and x^ ^ 1. Thus xe.{x\ and x^ ^ 1}. Now if X G N and a;2 ^ i_ i^ a; G {a; a; G A^ and a;2 ss 1}, then x = 0. Hence a; G P, as the only element in A^ which is not in P is zero. True. The property that x ~ u + iv where u,v e. R and ? = 1 is the defining property for complex numbers. False, x^ G P implies x = 0. But G Z.
xGN
|
xGN
Union and intersection S and T be sets. Then the union of S and T, written SUT and read "S union T", is defined as the set whose elements are either in S or in T (or in both S and T). For example, (1,2,3) U {2,5,6} ={1,2,3,5,6} and PU{0} = N. Clearly, ScSUT and TcSUT. Indeed it follows from the definition of SuT that any set containing both S and T contains SU T, so we say SU T is the smallest set containing S and T. Similarly it [S,T,U , the union of S and .} is any set of sets, we define SuTU C7U T and U and to be the set whose elements are the elements that belong to at least one of the sets S,T,U, ...
b.
Let
SuTuUU
If
is
said to be the smallest set containing the sets
S,T,U,
To
illustrate,
{1,2}U{3,4}U{5,6}U
the common part or intersection of S and T. The and read as "S intersection T". For example, suppose S= {1,2,3} and 7= {2,5,6}. Then SnT={2}. Repeating the definition, SnT is the set of those elements which belong simultaneously to S and to T. Here the possibility arises that there are no elements of S which belong also to T. We shall agree to the convention that there is a set, which we denote by 0, with no elements. Again we shall agree to the convention that the empty set is a subset of every set. Two sets are termed disjoint if they have an empty intersection. Thus {1,2} and {3,4} are disjoint. This notion of intersection can be generalized to any number of sets in the same way that the notion of union was generalized from two to any number. To be precise, the intersection of sets S,T,U, ., written SnTnUn -, is the set of all those elements which belong simultaneously to S, to r, to C7, Notice that SnT can be thought of as the largest subset of S which is also a subset of T. Similarly SnTnUn is the largest subset of S which is contained in T and in U and in ...
intersection
is
S and T
are sets, denoted by
= P. we may consider
SnT
SETS,
[CHAP.
If
S qT we
define
T S, the difference between T and S, by T - S = {x\ X GT and x^S}

and
Thus
that
if
r=
{1,2,3,4}
S-
{1,2},
then
T-S=
{3,4}.
For any
sets
T and S such
(1.1)
TdS,
prove equation
(1.1)
T-(T-S) = S
by showing that
right side of equation {1.1)
We
and
c
C
left side of
equation
(1 .1)
left side of
equation
(1.1)
right side of equation (1.1)
this suppose x G S. Then clearly but x^T-S. So x G r-(r-S); in other words, the right side of equation {1.1) is contained in the left side of equation {1.1). {T S). Then and The reverse inclusion is obtained similarly. Suppose S. Therefore x and x GS, i.e. xGS. So the left side is contained in the X right side and we have proved equation {1.1) by virtue of Proposition 1.1.
To do
xGT
xGT
xGT
^T
GT
Problems
1.4.
Check that the following statements are correct:

(i)
{1,2}
{a,e}
(ii)
U {1,2,3,4} = {1,2,3,4} U {e,/} U {ff.fi.} = {.a,e,j,g,h)
(iii)
{...,-2,-1,0} U {0,1,2,...}
If
(iv)
(v)
= Z a<h, a,he.R, then [a, h\ - {a, b) U {a} U {6}. {p\ p = (x, y) e R^, a;2 + j/2 = 72} U {p p = {x, y) R^, yp.-\-yi ^ 72}, (In other words, the union of the circle
I
= {p\ P = (x,y) G R^, x^ + of radius 7 and the disc of radius 7
y^<l^
without
Solution:
(iv)
its
boundary,
is
the disc of radius 7
itself.)
Let
[a, 6].
x-h,
any x Hence
(v)
then x
(a, 6>
If a<x <h, we have x S (a, 6). If x- a or By definition, a - ^ {a) or x& {6} respectively. Therefore [a, 6] C (a, 6) U {a} U {b}. Now for G {6}; and in each case a^ x-h. U {a) U {6}, either x G (a, 6), x G {a}, or
a;
ft.
a;
{a,h)
U {a} U {6} C
[a, 6].
The equality follows from Proposition

x^ + y^^
1.1.
7^. x^ + y^ < 7^ implies p G disc p = {x, y) is any element of the disc, then and x^ + y'^ = 7^ implies p G boundary. Thus the disc C boundary U without its boundary; disc without its boundary. The reverse inclusion can be checked similarly. Proposition 1.1 then implies the sets are equal.
If
1.5.
Check the following statements:

(i)
{1,2} n {1,2,3,4}
{a, e}
{1,2,3}
(ii)
{e, /}
{g, h}
=
then
where
a, e, f, g,
h are
distinct.
(iii)
{...,-2,-1,0} n {0,1,2, ...}

If
{0}
(iv)
<
6,
a, 6
B,
[a, h]
{a, h)
{a)
{&}.
Solution
(i)
False.
{1,2,3,4} but 3
three sets
{1,2}.
(ii)
True.
The
{o, e}, {e, /}, {g, h}
have no elements
in
common.
(iii)
True, since True,
is the only element in the intersection

[a, 6]
and
is
the only element in {0}.

a, 6
(iv)
a and 6 are the only elements of

6 are the only elements of
{a, 6}
and
{a)
{6}.
Furthermore,
a and
1.6.
{a}u{6}.
correct:
Check that the following statements are

(i)
(ii)
{1,2,3,4} -{1,2} = {3,4} {1,2,3,4} -{1,2,3,4} = S3

If
o, 6
(iii)
G B and a <
6,
then
[a, 6]
- (a, b) =
{a, b}.
Sec. 1.1]
SETS
T and U
are any three sets, prove the following:
(viii)
1.7.
If S,
(i)
sur=rus
SnT=Tr\S
S
SnS = s
Sli(TuV) =
(li) (iii)
(ix)
SuTuU
(iv)
cSuT SnTQS
and S and
cTuS
c S
(x)
(xi)
(xii)
(xiii)
TnS
(v)
(vi)
(vii)
Su0=S
Sn0 = SuS = S
xGSdT. By
Sn(rn C7) = (Sn r)n Su(rn [/) = (Sur)n(Su t/) S-S = S - (S - S) = S
Solution:
(i)
Let
SuTcTuS.
TuS
(ii)
the definition of union of sets, Similarly if xGTuS, it follows that by Proposition 1.1.
x&S
or
T.
Hence
a;
TuScSuT.
Consequently
TuS and SuT =
a; S SnT, then, by the definition of intersection, x G S and x G T. ment of rnS and SnT C TnS. The reverse inclusion, TnS C SnT, same manner. The equality follows from Proposition 1.1.
If
a;
is
is
therefore an eleestablished in the
(iii)
By
X
the definition of union,
SLIT contains
(i)
all
elements of
also
(iv)
GSUT xGSnT
and S
implies
QSuT.
Using
and x
above,
we
S and of have S C TuS.

a;
T.
So
x&S
Part
(ii)
implies
xGS
e.
T.
In particular,
S.
Thus
SnTcS.
allows
us to write
(v)
TnS QS.
S,
ScSu0
By
(iv),
by (iii). If a;GSu0, then x e Hence by Proposition 1.1, Su0 = S.
for
a;
by
definition.
Therefore
Su0
S.
(vi)
(vii)
Sn0 C 0. Sn0 = 0. ScSuS, using

Hence
follows.
But we know that
is
a subset of any set and, in particular,
C Sn0.
(iii).
Now,
xGSuS
implies
implies
x&S.
Thus
Hence
SuScS
or
and the equality
(viii)
By
(iv),
SnS Q
S.
But
x&S
x&SnS.
ScSnS
a;
and the equality follows.

x
.
(ix)
Let xSSu{TuU). Then a; or x e. quently xGSuTuU. Hence Su(ru 17) C or a;Gf7. If a; G S, a;GSu(ruC7). If
eS
TuU. Thus
GS
or
a;
a;
[/.
Consea;
SuTU [7. If x^SuTuU, then G S or G T x&TorU, xGSu{TuU). Hence SuTuU Q
Su(TuU). The
(x)
result follows.
(xi)
e Sn(TnU) implies xGS and xe(TnU), which in turn implies x^T and a: G t7. G T it follows that a;G(Snr) and, as xÛ, xG(SnT)nU. ThereFrom a;GS and fore Sn{TnU) c{SnT)nU. Similarly (Snr)n ?7 c Sn(rn t/). GS Su(rnC7) C (Sur)n(Sul7). For, if a;GSu(rnC/), then x & S or xGTnU. Now, implies xeSuT, xGSuU, and consequently x G (Su T) n (Su U). If k G Tn I/, then xGT G U; hence arGSuT, x G SuU and, as before, G (Su T) n (Su U). (Su T) n (Su U) c and
X
a;
a;
a;
a;
Sn(7'U
(xii)
J7)
is
established in a similar manner.

a;,
If
SS
contains an element
then
a;
G S and
^S
which
is
impossible.
Hence S
must be empty.
(xiii)
S-S =
from
(xii).
Thus
S-{S-S) = S-0
and clearly S
S.
1.8.
Prove the following statements:

(i)
(ii)
(iii)
(iv)
(v)
(vi)
S C r and U is any set, then SuU cTuU. S c T and U is any set, then SnU cTn U. If S C r and T c t/, then S qU. S c r if and only if SnT = S. T cS if and only if S = TuS.
If If If
res,
then
(S-r)ur =
S.
Solution:
(i)
Let a;GSuf7. Then quently xeTuU. If
xGS
x&U,
or
a;
[/.
If
a;
GS
Thus
then
xGT
since
ScT,
and conse-
then a;GruJ7.
SuU qTuU.
SETS,

and
[CHAP.
(ii)
xGSnU
and
implies
xGS
if
xGU.
then
As S Q
T,
we
also
have x
T.
Therefore
SnV
xiTnU
x G.U.
Q TnU.
that
(iii)
S Q T means Hence S C U.
First,
S,
T;
and since
T qU,
this in turn implies
(iv)
assume
SnT =
By
S.
xGSnT implies x G T. xGT and so xGSnT.

Problem
(v)
The equality implies any element a; in S belongs to SnT. But Hence xGT and S C T. Secondly, let S Q T. If x G S, then Therefore S Q SnT. The reverse inclusion SnTcS is true by
1.1,
1.7(iv).
Proposition
S = SnT.
Assume S = TuS. If x G T, then xGTuS and, since S-TuS,xGS. Hence T Q S. On the other hand if we assume T Q S, then x G TuS implies that x G S. Consequently TuScS. By Problem 1.7(iii) we have SqTuS. Thus TuS = S.
follows from the definition that (S - T) c S. Using (i) above, (S-T)uT cSuT. But by implies S = TuS, and -we have {S-T)UTcS. To show Sc(S-T)uT, let xGS. Either or x T. If x G T, then a: (S - T) U T. If x ^ T, then a; e S - T and we also have {S-T)uT. Thus Sc;(S-r)ur. The equality follows by Proposition 1.1.
It
(v),
(vi)
TcS
xGT
xG
1.2
a.
CARTESIAN PRODUCTS
Definition
The plane R^
shall also denote
(see Section 1.1a) consists of all pairs {x,y) of real
R^ hy
R.
R x R; thus R X R
is
defined as the set of
all
numbers x and y. We ordered pairs (x, y) with

T:
G R and
There
is
a natural extension of this notation to any two sets,
S and
SxT
For example,
{1,2}
{p\
p^{s,t),
sGS, tGT}
X {1,2,3} =
{(1,1), (1,2), (1,3), (2,1), (2,2), (2,3)}
In words, is defined to be the set of all ordered pairs of elements (s, t), the first member of each pair always belonging to S and the second member always belonging to T. We term the cartesian product of S and T. It is worth pointing out that, just as in R^, two elements of are equal iff (if and only if) they are identical, i.e. (s, t) = (s', t') if and or T = 0, we interpret S x T as 0. only if s = s' and t = t'. If either S =
SxT
SxT
SxT
One
{n
defines similarly the cartesian product Si
x
,
S2
x
Si
.
.
x Sn of the n
S2GS2,
.
sets Si, S2,

s
.,Sn
<
00)
as the set of
all
%-tuples
(si, S2,
.
s)
with
(si
G Si,
.
. ,
S.
,
.
As with
.
the cartesian product of two sets,
(si, S2,
s)
sz ,
Sn ) iff Si
= Si
S2
= S2
.,Sn
= Sn
For example,
{1,2}
X {2,3} X {4,5} =
{(1,2,4), (1,2, 5), (1,3, 4), (1,3,5),
(2,2,4), (2,2,5), (2,3,4), (2,3,5)}
If
If
any one of the

each Si
sets Si, S2,
S,
then Si
S2 x
x Sn
0, then we shall interpret Si X S2 x is denoted by S".
x Sn to be 0.
We often are
geometry one investigates
interested in certain subsets of SxT. For example, in elementary analytic lines, circles (see Section 1.1a), ellipses and other figures in the
plane; these are subsets of R^.
Problems
1.9.
Let
(i)
S=
(1, 2, 3},
T-
{1, 5}.
Verify the following statements:
(ii)
(iii)
SX r = rXS = SX r #
{(1, 1), (1, 5), (2, 1), (2, 5), (3, 1), (3, 5)}
{(1, 1), (5, 1), (1, 2), (5, 2), (1, 3), (5, 3)}
TXS
Sec. 1.2]
CARTESIAN PRODUCTS
S2
(iv)
(v)
^ r2 (S2 = S X S and (SxTjXS - SX(TXS)
T^
^ TX
T)
Solution;
(i)
Clearly these are the only ordered pairs
{x, y)
with x with
e S and
e.
T.
(ii)
As
in
(i),
these are the only ordered pairs

(i)
{x, y)
xG T
(iii)
Looking- at
(3, 3)
and
(ii),
we
fz.
find
(5, 1)
S T X S and
(5, 1)
and y & S. S X T. Hence
SxT^ TxS.
e
(S
(iv)
S2 but
(3, 3)
(v)
An
element
(x, y)
oi (S
T)
X S has x G S X T and y ^
(1, 1)
S.
Thus
((1, 1), 5)
T)
S.
But
((1,1),5)SX{TXS),
1.10.
since
S.
Let S,
(i)
(ii)
(iii)
T and 1/ be any three sets. Prove the following: S X r = r X S ifF either S = T or at least one of the two If (x, y) e S2, then {y, x) e S^. Sxr^SxC/ iff either T - U or S = 0.
is
empty.
(iv)
{SXT)XU = SX(TXU)
iff
at least one of the sets S, T,
is
empty.
Solution
(i)
T implies and S = or T = jZ) implies, by the definition of the cartesian product of any set and the empty set, = = TxS. Therefore = whenever S - T, S = 0, or T = 0. To prove the converse, assume S X T = T X S. may also assume S and T -- 0. Let and (such elements exist since S ' and T - 0). Then {s,t)SSxT and, as = TxS, (s, t) e T x S. It follows, from the definition of T X S, that t S S and s e. T. Therefore T and S Q T. conclude, using Proposition 1.1, S = T.
S=
SxT=T2=TxS;
SxT
SxT
TxS
tGT
sGS
We
SxT
qS
We
(ii)
(iii)
(iv)
S S and ?/ S S. Hence {y, x) S S^. S = 0, we have S X T = S X U. Conversely, let SxT = SxU and S (if S = we have nothing to prove). If T # 0, let tGT. Then {s,t) e SxT for any s G S; and, as SxT ^ SXU, (s, t) G S x [/. But (s, <) G S X [/ means t G 17. Hence T Q U. A similar argument gives U C T, and we conclude T = U. If T = 0, then Sxr = and SXI7=0. U = follows from S X U = 0; for if U^0, then S X 17 would not be empty, by virtue of our assumption S = 0. Thus both T # and T = give r = C7. If either S,T or U is empty, (iS X T) x 17 = = S x (r X t7). Conversely, assume {S X T) X U = Sx{TxU). If S 7^ 0, r # and U ^ 0, there is at least one element (x, y) in (S X T) X U, X e Sx T and G ?7. But (x, y) must also be an element of S X (T X U). Hence x G S. This is a contradiction, for x cannot be both an element of S and an element of S X T. Therefore the assumption that S - 0, T ^ and U must be false and so either S, T or U is empty.
{x, y)
S2 means x
if
Clearly
T=U
or
2/
1.11.
(i)
If
A=
{p\ p
(x,y)
GP2
and x
<
y)
(recall that
is
the set of positive integers), prove
that:
(a)
(6)
(ii)
(x, x) if
Â
{p
for every x
if
P.
(x,j/)eA and
I
(y,z)
&
A, then (x,z)GA.
Let
(ct)
B =
Show
If (x,
p
if
that
(x, x)
(h)
(c)
(iii)
Prove that
J/)
G
I
B,
G P2, x ^ i/}. G B for every x G P. (x, !/) G B and if (y, z) G when is (2/, x) G B?
(x.j/)
(x, all
2/)
B, then
(, z)
B.
Let
(a) (6)
(c)
A =
(x,
a;)
{p
eA
for
G Z2 G Z.
with x
3/
divisible
by
3}.
Prove:
If
(x,y)&A, then (3/,x)GA.

(x,2/)GA and
{y,z)
If
C7
A, then (x,z)GA.
(iv)
Let
the
Put
{a) (6)
(c)
X V=muL2.
If If If
(a, a)
be the points of the plane J?2 on or above axis and let L be the points below the X axis. Notice that VqR^xR^. Prove:
[/
G i22 X i22, then (a, a) G V. (a, P) G y, then (^3, a) G y. (a, /3) G y and (,8, y) G V, then
(a, y)
y.
(d)
y?î22xi22.
SETS,
[CHAP.
Solution:
(i)
(a) (b)
{x, x) (x,
for every x
implies
P, since x
is
not less than
x.
v)&A
z,
X
(ii)
<
since x,y,zi P.
all
a;
<y, and (y, 2) S A implies y < z. Now x <y and By definition of set A, we have {x, z) e A.
Hence
(x, x)
<
imply
(a) (6)
(c)
e B. &B and z) S B imply x - y and and, by definition of set B, (x, z) e B. = For if {x, y)&B, x ~ (y, x) eB and y X iff x = y.
X for
P.
{x, y)
(2/,
iflF
2/
respectively.
It follows
that
a;
2/.
y;
and
if
(y, x)
&
B,
x.
But
(iii)
(a) (6)
For any
(x,
(c)
G Z, - x = 0, and zero is divisible by 3. Consequently (x, x) G A. G A means x y is divisible by 3, i.e. x y = 3q where q is some integer. Now 3/) y x = -(x-y) = -3q = 3{-q). Therefore y x is divisible by 3 and {y,x) G A. (x, 2/) G A means x y = 3q for some integer q, and (y, z) G A means y z 3r for
a;
some integer
r.
Thus
X
(x
2/)
(2/
z) =
a
3^
3r
3{q
+ r)
Hence
l.lO(ii),
and so X
(iv)
is
divisible
by 3 and
(x, z)
G A. G
Z7
(a)
(a, a)
(,a)
(6)
G 2 X J?2 G L2, and
implies
(a, a)
a G i?2_ which means G C/2uL2 = y. p)
or
L.
(a, a)
1/2
or
If
(y8,
(a,y8)Gy,
a)
then
{a,
(/3,
or
(a,y8)GL2
and by Problem
{p,a)&m
or
L2.
Hence
a)
V.
(c)
(a,
implies either a, p e U or a, p G L, but not both, since f/nl/ = 0. Now implies p,y G or L. If ;8, y G C7, then a, p & U since ;8 cannot be an element of both and L, and hence (a, y) G f72. Similarly if y8, y G L, we have (a, y) G L2. Therefore (a,y) G C/2uL2 = V.
/3)Gy
y)
(/3,
Gy
(d)
Let a
17
and
y8
L.
Then
(a,
p)
e R^ X
R^, but
(a, ;8)
V.
b.
Equivalence relations
Similarity of triangles
is
an example of an equivalence
relation.
This means that
if
s, t
(i)
and u are any
triangles, the followîng three conditions hold:
s is similar to s.
(ii)
If s is similar to If s is similar to
t,
t,
then
t is t is
similar to
s.
(iii)
and
similar to u, then s
is
similar to u.
Another example of an equivalence relation is congruence of triangles since (i), (ii) and (iii) above hold also if "similar" is replaced by "congruent". Continuing in this vein, if X is any non-empty set and ^ is a "relation on X", i.e. if for any pair of elements x,y & X either x is related to 2/ by B (written xRy and read "x is related to y by R") or not, then R is termed an equivalence relation in X if:
(i)
(ii)
If
(iii)
for all x G X. xRy, then yRx. If xRy and yRz, then xRz.
xRx
One
defined.
objection to this definition of equivalence relation is that "relation on X" is vaguely We shall therefore define the idea of equivalence relation by means of sets and
subsets.
Let us consider again the notion of similarity of triangles. Let T be the set of all Let S be the subset of T x r defined by (s, t) G S iflf s is similar to t. If t is a triangle, t is similar to itself, so {t, t) e S. If (s, t) and (t, u) S S, then s and t are similar and t and u are similar. Thus s and u are similar. Hence (s, u) G S. Similarly if (s, t) G S,
triangles.
it is
clear that
{t, s)
S.
Consequently, discarding our informal approach,
we have
the following
Sec. 1.2]
CARTESIAN PRODUCTS
Let
Definition:
be a non-empty set and let R he a subset of X^. Then R is called an if the following conditions are equivalence (or an equivalence relation) in
satisfied.
(i)
{x, x)
(ii) (iii)
If If
(x, y)
G R for e R,
all
Z
a;)
(reflexive property),
then
{y,
G i?
(symmetric property),
(transitive property).
{x,y)GR and {y,z):R, then {x,z)GR
Problem 1.12. Examine Problem

Solution:
(i)
1.11 for equivalence relations.
(ii)
A B
x
is is
not an equivalence relation in P, as
(x, x)
Â
for
all x.
not an equivalence relation in P, as we y. Thus though (1, 2) G B, (2, 1) g B.
know
(x, y)
& B and
reflexive,
{y, x) G.
occurs only if
(iii)
is an equivalence relation in Z^ as properties hold.
Ac ^2
and the
symmetric and transitive
(iv)
yc
in
i22
i?2
and the
reflexive,
symmetric and transitive properties hold, so
is
an equivalence
fi2.
c.
Partitions and equivalence relations
Suppose y by R, if
is
an equivalence relation in X.
We
Z= {1,2,3,4}
(x, y)
G R.
let
If
(x, y)
G R we
shall
say x is iJ-related to y, or x is related to sometimes write xRy. To illustrate let
and
R =
{(1,1), (2, 2), (3, 3), (4, 4), (1,3), (3,1), (3, 4), (1,4), (4, 3), (4,1)}
(1.2)
Then it is easy to check that R satisfies the three necessary conditions for it to be an equivalence relation. Now SRA since (3, 4) G R, but 2i24 is an incorrect assertion since
(2,4)
^R.
we
Note that we have used a notation that fits in with the notation of Section 1.2b where informally introduced an equivalence relation.
position of
equivalence relation in is intimately connected with a partition of X, i.e. a decominto disjoint subsets of such that every element of belongs to some subset. Examples of partitions of {1, 2, 3, 4, 5} are
An
{1}, {2,3}, {4,5}
and
{1,3,4}, {2}, {5}

is
On
the other hand {1,2}, {2}, {3,4,5}

If
not a partition of {1,2,3,4,5}.

{1,2,3,4}, then
all
is
the equivalence relation

i.e.,
(1.2) in
the elements of {1,3,4} are
i2-related to 1,
IRl, 1R3, 1R4:
This suggests a means of getting a partition of a set X. In order to explain, we need some Let R be an equivalence relation in a set X. If x G X, we define xR= {y\ y and {x,y) gR}. xR is thus a certain subset of X. This subset xR is called the R-class of x, or the R-equivalence class of x, or the R-block of x. A subset of will be called an R-class or R-block if it is the iZ-class or /2-block of some element x G X. To illustrate these terms, consider the equivalence relation R given by (1.2). Here
additional notation.
GX
li2
{1, 3, 4},
2R =
[2],
and BR
= 4R = IR
Thus the
iJ-classes
here are simply {1,3,4} and {2}.
Notice that these iZ-classes constitute
a partition of {1,2,3,4}.
More
generally,
we have
the following
10
SETS,

let
[CHAP.
Theorem
1.2:
Let
Z be a non-empty set and

(i)
be an equivalence relation in X.
x'R,
Then
(ii)
xRnx'Rv^^, then xR = X G xR for every x e X.

if
Thus the
J?-classes constitute
disjoint,
tinct /2-classes are
a partition of X, for (i) guarantees that diswhile (ii) shows that every element of X
i2-classes.
appears in at least one of the
Proof: First, we verify (i). Suppose xRnx'R - 0. Then there is an element y G xR which lies also in x'R, i.e. {x, y) G R and (x', y) G R. As R is an equivalence, it follows from the symmetric property that {y, x') G R. But (x, y) G R and (y, x') G R imply, by the transitive property, {x, x') G R. Now if z e x'R, then {x', z) G R; and hence by the transitive property applied to {x, x') and {x', z), we find {x, z) G R. This means, by the very definition of xR, that z G xR. Since z was any element of x'R, we have proved x'R c xR. The reverse inequality follows by a similar argument. Hence x'R = xR as required.
The
verification of
1.2.
(ii)
is trivial
since
(x, x)
G R means
xR.
This completes the
proof of Theorem
Problems
1.13.
(i)
Prove that
E =
{(0,0), (1,1), (2,2), (3,3), (0,2), (1,3), (2,0), (3,1)}
is
an equivalence relation
in
S=
(ii)
{0,1,2,3}.
(a)
Find the S-equivalence blocks
OE,
(b)
IE,
(c)
2E,
(d)
3E.
Solution:
(1)
The reflexive property holds, i.e. (x, x) e E for all xSS, since (0, 0), (1, 1), (2, 2), (3, 3) e E. To show that E is symmetric, let us examine all pairs (a;, y) where x = y. There are only four, namely (0, 2), (1, 3), (2, 0), (3, 1). Clearly if (, y) is any one of the four, so is (y, x). When x = y, {x,y) = (y,x). Thus (x,y)&E implies (y,x)e.E. S is also transitive. Let {x,y)eE and (y,z)eE. Suppose x - y. Then (a;, j/) can be (0,2), (1,3), (2,0) or (3, 1). If (x.y) = {0,2), then {y, z) = (2, 0) or (2, 2) and {x, z) = (0, 0) or (0, 2) respectively. Hence {x, z) e E. Simz) & E ilarly if (x, y) = (1, 3), (2, 0) or (3, 1), it can be shown that (x, z) S E. When x = y, (y, means {x, z) e E. Therefore for any (x, y) G E and {y, z) E, we have {x, z) e E and
is transitive.
(ii)
(a)
OE =
{0,2},
(6)
IE =
{1,3},
(c)
2E =
{2,0},
(d)
SE =
{3,1}.
Observe that OE
= 2E
and
IE = 3E.
Let
in
1.14.
A-
Z and
(x, y) G Z^ with {p\ find the K-equivalence classes.
p-
x-y
divisible
by
3}.
Prove that
is
Solution:
It
was shown
{3? g
in
equivalence relation.
(1) (2)
Problem l.ll(iii), page The i2-equivalance classes

if
(0,
7,
that
satisfies
the
three
conditions
of
an
are:
is divisible
OA =
1
zy, for
A,
lA.
a;
= -x
by
3.
Also,
(0,
3g)
G
If
A.
x
lA = {l-3g| gGZ};
a;
for
if
{l,x)eA,
(1, x)
l-x =
Zq
and hence
= l-Zq.
If
= l-Zq,
is
divisible
by
3;
hence
for
if
(3)
2A = {2-Zq\ q&Z};
is divisible
(2,a;)GA,
2-x =
Zq
and so x
= 2-Sq.
= 2-Sq, 2-x - 3g,

or 2
by
3;
hence
(2, x)
G A.
0A,1A,2A are
Consequently
the only fi-blocks, for any integer can be written as 3q, 1

Z.
- 3g.
0AulAu2A =
1.15.
all integers.) Let Z* be the set of nonzero integers and let S == Z X Z*. (Recall that Z is the set of = {p\ p = {(r,s), {t,u)) G S^ with ru = st}. Prove that E is an equivalence relation in S. Let E
Sec. 1.3]
MAPPINGS
11
Solution
7 is reflexive, for {r,s)BS implies ({r,s), {r,s)) S S^; and since rs = sr, {{r,s), (r,s)) G E. The symmetric property of E is established by noticing {{r,s), {t,w)) E means {r,s) and {t,w) S S^, and rw = st. But rw = st can be rewritten as ts = wr. Hence ((f w), (r, s)) e JB7. To show E is transitive, let ((r, s), (t, m)) S 7 and {{t,u), {v,w)) G E. Then rw = sf and tw = vu. Since m ^ 0,
,
and rw
= tw = vu =
sv.
Thus rw
sv
and so
((r.s),
(v.w))
E.
Hence
is
transitive.
Therefore
is
an equivalence relation.
in S.
1.16.
Prove that S X S
Solution:
is
SXS
is
reflexive since
(x, x)
and, by definition
of
SX
S,
{y, x)
x, y and z G {y, z) GS imply an equivalence relation on S.
S.
S S X S for all x & S. If {x, y) G S X S, then x and y G S G SxS. Hence SxS is symmetric. Now (x, y) G S and But then (x, z) S S X S and S X S is transitive. Thus S X S is
1.17.
Prove that a set

Solution:
is infinite if
and only
if
there are infinitely
many
equivalences in X.
(Hard.)
Assume there are an infinite number of equivalences in X. If X is finite, then there are at most number of distinct subsets of X^. Therefore when X is finite there are at most a finite number of equivalences in X, which contradicts our hypothesis. Hence X must be infinite. Conversely, assume X is infinite. We exhibit an infinite number of equivalences in X as follows: For
a
finite
each pair a,b

^(a,b)
B, a
= b,
we
(.
define
2/)
{p| P
X^, where either x
y, (x,y)
(a,b) or (x,y)
(b,a)}
Now
each
R(a,b)
infinite
number of
^(c,d) if and only if different pairs

{x,x)
{a,b}
a,bGX
i2(o,b)
= {c,d}. Therefore since X is infinite, we can find an each of which gives a distinct set iZ(a,b)- Furthermore,
i2(o,b) satisfies
Râ.b") is
an equivalence.
To prove that
the three conditions of an equivalence

definition of fi(a,b)-
relation,
we
first notice
Râ.b^
for all
xGX,
by the very
Secondly,
{y,x)
G R(a,b), y and then (x,y) = then {x,z) G R^^.b)- To see this, notice that (x,y) can only be (a, 6), (6, a) or ix,x). (x,y) = (a, b) implies (y,z) = (b, a) or (b,b); hence (x, z) = (a, a) or (a, 6), which are both elements of fi(a,b)Similarly,
{y,z) efi(
(x,j/)
,,)
R(o,b) is symmetric since (x, ?/) (b,a) or (a, 6) respectively, or x
means
(x,y)
(a,b) or (b, a), in which case (y,x). Thirdly, if (x,y) and (y,z)
(6, a)
implies
(x,z)
Rca.bi-
Finally,
{x,y)
(x,x)
means
{x,z)
d.
The
division notation
it
We
in a set
find
useful to introduce a notation for the i?-classes of an equivalence relation
X, namely X/R.
Problems
1.18.
What
is
S/E
in
Problem 1.13?
Solution:
S/E = {0E,1E}.
in
1.19.
What
is
Z/A
Problem 1.14?
Solution:
Z/A = {OA, lA, 2A}.
1.3
MAPPINGS
Definition of
a.
mapping
Assign to each even integer the value 1, and to each odd integer the value 1. Let us give the name a to this assignment; thus a assigns to each even element in Z, the set of all integers, the unique element +1 in the set {1, 1} and to each odd element in Z the unique element 1 in {1,-1}. In less detailed terms a assigns to each element in Z a unique element in (1,-1}. Such an assignment is termed a mapping from Z into {1,-1} or a map
12
SETS,
[CHAP.
from Z into {1,-1}. More generally, if S and T are any two non-empty sets, a mapping or a map from S into T is an assignment of a unique element of T to each element of S. For the most part we shall denote mappings by lower case Greek letters such as a, p, y. If a is a mapping from -S into T, we shall express this fact more briefly by writing a:S^T; this is read "a is a mapping from S into T". We call S the domain and T the codomain of a.
Suppose that a: S-* T. the element t in T, we write a: s^t and read this as "a sends s into t". We call t the image of s (under ) if a s -* i. It is convenient to have a number of notations for the image of an element s in S under a mapping a: S^T; thus we shall write For the most part we use the first notation. Sff or s", or even a{s) for the image of s under a.
find
it
We
useful to provide further notation and definitions.
If a assigns to s in
If t
e r and
a.
sa
t,
we
call s
a preimage of
t.
By Sa we mean
{sa
S).
We
call
Sa
the range of
Problems
1.20.
Suppose a
(i)
P
1
is
defined by
re
a: n-* n^ for
a a
:
all
(ii)
re
re
-*
w+
for
all
all
(iii)
-> 2re
for
eP we P re S P
(iv)
a a
Ji
-^ 1
for
re
all
n e
for
P
re
(v)
1 -^ 2,
-^ 1
all
P,
re
>
above, the mapping a is defined by describing its "action" on every element of P has a unique P. For example, in (i), la = 12 = 1, 2a = 4, Sa = 9, .... Note that each element of element assigned to it.
Afofe:
In
(i)-(v)
In each case determine:

of
(a) 2a, 5a, 6a;
(b)
P an image of some element of P How many preimages does 1 have in

Solution:
(1)
in (i)-(v)? How (iv)? in (v)?
a preimage (under a) of 2, 5, 6, 27. (c) Is every element many preimages (under a) does 2 have in (v)?
(a) (6)
(c)
2a
4;
is
5a
25; 6a
36.
2, 5,
There
no preimage of
of
6 or 27.
(ii)
(a) (6)
(c)
(iii)
(a)
(6)
(c)
P is an image, e.g. 2 is not an image. 2a = 3; 5a = 6; 6a = 7. la = 2; 4a = 5; 5a = 6; 26a = 27. Hence 1, 4, 5, 26 are the The only element of P which has no preimage is 1. 2a = 4; 5a = 10; 6a = 12.
Not every element
1 is 1
required preimages.
a preimage of
2; 5
has no preimage; 3
is
the preimage of
is
6;
27 has no preimage.
has no preimage, so not every element of
an image.
(iv)
(a) (6)
(c)
2a
1;
5a
1;
6a
1.
2, 5,
6 and 27 have no preimages.
has every element of
as a preimage and no other element of
has a preimage.
(v)
(a)
(6)
(c)
2a
1;
5a
1;
6a
1.
1 is
a preimage of
2; 5, 6
2 has one preimage,
namely
1
ment of
not equal to
and 7 have no preimages. In fact any ele1 has an infinite number of preimages. 1. 1 and 2 are the only elements of P which have is mapped onto 1.
preimages.
1.21.
Let
(i)
S=
{1, 2, 3},
T =
{1, 4, 5}.
(ii)
Write down a mapping of S into T. T an image of some Let a:S^T be defined by la = 4, 2a = 5, 3a = 4. Is every element of element in S? element in S under a? Is every element of T the image of more than one Give preimages of 1 and 4.
Solution:
(i)
(ii)
For example, a S -> T defined by la = 1, 2a = 4, 3a = 5. than Not every element of T is an image, for 1 has no preimage. Only 4 is an image of more has preimages 1 and 3. one element in S. 1 has no preimage, and 4
;
Sec. 1.3]
MAPPINGS
P
-*
13
1.22.
Suppose a:
(i)
is
defined by
{v)
a;
- a;2
a: x a: X a:
-> ix
-^
(ii)
a: X-* 2x a:
+ 27
where
z
.^
ix
.
(iii)
(iv)
a:
x^ z x^z
all
where
SC 6 C
is is
such that
z^ z^
such that
(vi)
x.
a;
ix
_^ i
+1
cc
- 1.
("^"^
x^ logjo
(a)
(6)
Do
of these descriptions really define mappings of
Is every
element in
a preimage of some element of
P P
into
in
C?
(vi), (vii)?
(i), (ii), (v),
Solution:
(a)
and (ii) define mappings since every x G P has a unique image, (iii) does not define a mapping; for example, either 2 or 2 could be taken as 4a. Similarly (iv) is not a mapping, since there are three complex cube roots of x 1, so that each x has three different images, (v), (vi)
(i)
and
(6)
(vii)
define mappings.
(v)
j
In
(i),
(ii),
and
(vii), i
has no preimage: for

(v) ix
(i)
implies
27 x = P;
x^
implies
(vii)
x
if
yfi^ P;
(ii)
2a;
+ 27
10*
+ X-i
gives
+ t g P;
-:
logjo x
1.
i,
then
and thus x S P. In
(vi) 1
has no preimage because
IX 1.
1 implies 1
1.23.
What
(i)
is
Pa
in each of the cases in
Problem 1.20?
Solution:
(ii) (iii)
= Pa
Pa Pa
is
the set of squares {1,

{2, 3, 4, 5, ...
4, 9, 16,
.}.
(iv)
(v)
i.e.
Pa = Pa =
{1}
{1, 2}
{2,4,6,8,
.},
all
the even integers.
mapping The reader may ask whether our definition of mapping is precise. After all, it depends upon an English word, assignment, a word that is used in many different ways.
b.
Formal
definition of
comparison with Section 1.2b is valuable. In Section 1.2b we introduced the concept of equivalence relation in X, but as we felt uneasy about it, we redefined it in terms of a subset of X^. Here too we feel uneasy about our definition of mapping and so we shall redefine it in terms of sets.
is called a mapping of S into T if (s, ti) and (s, ^2) e a occurs only if and for each s GS there exists an element (s, t) G a. S is called the domain and T the codomain of a. If a is a mapping of S into T (written briefly as a: S^T) and (s, t) G a, we call t the image of s under a and write s -* t. We also write t = sa.
A subset a of S X r
ti,
U=
easy to see the relationship between the old definition and the new. definition the elements of S were assigned unique elements of T.
It is
In the old
Consider the subset of
SxT consisting
of the pairs
(s, t)
where
t is
assigned to
s.
The
two conditions of the new
definition are satisfied
by
this subset.
we
c.
In the sequel we will use the definition of Section 1.3a, being confident that if necessary could justify our arguments using the definition of a mapping in terms of a subset.
Types
of
mappings
We have talked of mappings without defining what is meant by the equality of two mappings. We will now remedy this. Suppose a: S ^ T and (S: S' -^ T'. Then we define a = ;8 if and only if S = S', T T', and, for every element s & S, sa = sa'. In other words, two mappings are equal if and only if they have the same domain, the same codomain, and the same "action" on each element of S. For example, let S {1,2,3,4}, r = {4,5,6}. Let a:S^T be defined by 1 = 4, 2a = 5, 3a = 6, 4a = 4; let /8 S - T be defined by Then a' 13 since 4a ^ 4:/3. 1/3 = 4, 213 = 5, 3j8 = 6, 4^8 = 5.
:
: :
14
SETS,
[CHAP.
we say a
important to distinguish certain types of mappings. Thus suppose a: S^ T. Then is a mapping from S onto T (notice that into has now been replaced by onto) if every element in T has at least one preimage in S, i.e. if for every t there is at least one element s G S for which Sa = t; in this case we call a an onto mapping.
It is
GT
=
On
of
-S
the other hand
we say
a is one-to-one if sa
a).
s'a implies s
s',
i.e.
distinct elements
have distinct images in T (under a matches S with T or is a bijeetion
Finally, a
is
we say
is
a matching of
S and T
or that
both onto and one-to-one. Two sets are termed equipotent or of the same cardinality if there exists a matching of the one with the other. If S is finite and a matches T, then we say S and T have the same number of elements. We denote the number of elements in a set S by |S|. If S is infinite this definition no longer makes sense unless one takes from all the sets which match S a single fixed set which we then term \S\, the cardinality of S+. A set which matches P is called enumerable, countable, or countably infinite. Important results are that the set of rational numbers is enumerable
if
and the
set of real
numbers
is
not+.
and suppose S' c <S. Then we define a mapping f rom <S" into T by simply restricting the domain of a to S'; this mapping is denoted by a^g. (read a restricted to S') and is called the restriction of a to S'. To be quite
We
require one further definition.
Suppose a:
S^T
explicit.
:
'IS'
S'-^T
is
defined
by
a^.
^ Sa
for
all
S'
Problems
1.24.
Which
of the
mappings defined
in
Problems 1.20-1.22 are
(a) onto,
(6)
one-to-one,
(c)
matchings?
Solution:
(a)
None
of the mappings defined in Problems 1.20 and 1.22 is onto by the solution already given to these problems. The mapping defined in Problem 1.21(ii) is hot onto, since 1 has no preimage.
1.20(i),
(ii),
(6)
Problems and so w
for in (i), if na = ma, then n^ = nfi only one positive square root of an element in P; in (ii), na = ma gives n-|-l=m+l or n = m; and in (iii), na = ma implies 2n = 2to or n = m. Clearly Problems 1.20(iv) and (v) do not define one-to-one mappings. The mapping a in Problem 1.21(ii) for in is not one-to-one since la = 3a. All the mappings defined in Problem 1.22 are one-to-one:
and
(iii)
define one-to-one mappings:
to,
since there
is
xa = x'a means 2x'-f27 or x = x';

(i),
a;2
x'"^
m
2ix
[v),
xa
and, since x,x' &V, x = x'; in (ii), xa = x'a implies 2a; = x'a implies ix + \~ix' + l or x = x'; {n)xa = x'a or
(iii)
a;
+ 27 =
means
and
tx+J.
tx
hence
(c)
_ 1^+1 1 = 10 =
ta;'
a;
^jj,
2ix'
x';
in
(vii),
xa
x'a
gives
logio^
logio x'
x'.
(Note that
in
and
(iv)
are not mappings.)

is
None
of the
mappings defined
Problems 1.20-1.22
a matching, since none of
them
is onto.
1.25.
Which mappings
Solution
in
Problems 1.20-1.22 are equal?
Problem 1.20 features a:P-^P; Problem 1.21, a Hence we need only compare the mappings in each exercise.
None.
->
T;
and Problem
1.22,
a:P-*C.
1.26.
Let
a-.P^Z
a
is
be defined by na
= -n
for
all
n e P.
Is a
onto?
One-to-one?
matching?
Solution
neither onto nor a matching, since 1 has no preimage.
a
is
one-to-one, for if
na
n'a,
then
n =
n' and hence n
n'.
tFor more
1953.
details see, for example, G. BirkhoflT
and
S.
MacLane,
Survey of Modern Algebra, Macmillan,
Sec. 1.3]
MAPPINGS
15
1.27.
Suppose n qm
m
+
1,
.
is
a fixed positive integer.

is
where the remainder r

every
m = 4,
4k, 4k
then
+
is
4k
.
+ 2,
4k
+
.
Every integer n can be written uniquely in the form .,m 1}. For example, if an element of the set {0, 1, integer can be expressed in precisely one of the following forms: 3. Let a Z - {0, 1, .,m 1} be defined by na = r, if n = qm, + r where
. .
:
{0, 1,
. ,
m 1}.
{0, 1,
.
Prove:
.,
(a)
(6)
a
If
onto
m 1}.
integers, then
TOj,
^2 ^re
any two
(n-^nâ
(n-ân2a)a.
Solution:
(a) (h)
is
onto because
=:
0, 1,
.,
7n
Let Ml
iW2
QiWi
ri
and n2
1 have preimages = 92 + '"2' Then

9192^
0, 1,
.,
tn
1
=
respectively.
(qim
n){q2Vi
+ rg) = +
rg,
'i92
^29ii
+ =
*'i'2
(9i92
+ -i92 + ^'agO'w + ^^2
Now
1.28.
letting
r^rj
qstn
we obtain
(wiW2)
''s
('i''2)'>'
(wia'K2)-
Check which of the following mappings are onto, one-to-one,

(i)
bijections:
a:
-^
defined by a: a
defined by a:
(ii)
a.Z-^P
a:
+ ib-â^+b^ n -^ n^ + 1
(iii)
P^Q
is
defined by a: n-^
r
a^+
b^
Solution :
(i)
Hence a
(ii)
neither onto nor one-to-one, because is not a bisection.

is
0,
for any
o, 6
i?,
and
la =
la
1.
As
1
3 has no preimage, a
neither onto nor a bijection.
Also a
is
not one-to-one, since
la
and la
(iii)
2.
has no preimage, since ^
x-r
71
implies
n.
= 1
.
P.
Therefore a
is
not onto and hence

n';
not a bijection.
rr ^ 2n' + 1 2n + l Thus each image has a unique preimage and a is one-to-one.

.,
,
na
n'a
means -
or
2n'n
+n =
2n'n
hence
n =
n'.
1.29.
Let
[a,b].
be the set of open intervals Define by {a,b) a
(a, b)
on the real line and

Is a one-to-one?
let
be the set of closed intervals
a.S^T
is
[a,b].
Onto?
Solution:
The mapping
equality holds
[a, b]
iff
one-to-one and onto.

b
a = a' and has a preimage {a, b).
b'.
is
For, if {a,b} a = {a',b') a then [a,b] = [a',b']. But this therefore one-to-one. a is also onto since a closed interval
1.30.
How many mappings are there from {1,2} into itself? From how many of these mappings are one-to-one? Onto?
Solution:
{1,2,3} into itself?
In each case,
There are four mappings of {1, 2} into itself, namely: aj defined by laj = 1 and 2q!i = 2; a2 = 1 and 2a2 = 1; a^ defined by las 2 and 2a3 = 1; a^ defined by la4 = 2 and 2a4 = 2. Only a^ and ag are one-to-one and onto.
defined by la2
To find the number of mappings of {1,2,3} into itself, we proceed as follows: 1 may have any of three images under such a mapping, i.e. 1 -> 1 or 1 -> 2 or 1 -> 3. Also, 2 may have any of 3 images, either 1, 2 or 3. So we have in all 3 X 3 possibilities for the actions of mappings on 1 and 2. Then 3 can be sent into 1, 2 or 3, giving 3 X 3 X 3 = 27 possible mappings of {1, 2, 3} into itself. There are 3X2X1 = 6 possible one-to-one and onto mappings; for when we once choose an image for 1 there are only two possible images for 2, and then the image of 3 is uniquely determined.
Let S = {1, 2, 3}, T = {3, 4, 5}, U = {4, 5, 6}, and let Let p and y be the mappings from T into U given by
1.31.
a:S^T
/8
:
be defined by
3 -^ 4, 4 -> 6, 5 - 4,
1 -^ 3, 2 -^ 3, 3 5. 3 -* 4, 4 -> 4, 5 -* 4.
Compute
Solution:
(la)y8
(la)/3, (2a)j8, (3a)y8, (la)y, (2a)y, (3a)y.
= 3y8 = 4;
(2a);8
= 3/3 = 4;
(3a);8
= 5,8 = 4;
(la)y
= 3y = 4;
(2a)y
= 3y = 4;
(3a)y
5y
= 4.
16
SETS,

into
[CHAP.
1.32.
Let
if
a, p,
y be the
mappings of Q
defined
by a: x^-^
.
^ x + 1, y:x^x-l.
'
-^^
Prove that
is
any element of Q, then
((xy)a)li
x + 1 =
What
is a^p ?
a^^.
"ip
i
"iz
Solution:
({xy)a)li
{(x
- l)a)p =
of
(3
^ +
_
1
_i_
^~
_i_
|P
is
'^
^ mapping of
into Q.
a,^ is a
a^^,
mapping
into Q.
a^p = a^^
since the
domain of
a^p
not the same as the
domain of
1.33.
If
of
S S
is
a non-empty set, prove that
is infinite if
and only
if
there are infinitely
many mappings
into S.
Solution:
First we show that if there are infinitely many mappings of S into S, then S is infinite. Assume, on the contrary, that S is finite and \S\ = n. Now each of the n elements can be mapped onto at most n images. Hence there are at most m" different mappings of S into S. This contradicts our assumption that there is an infinite number of such mappings. Hence S is infinite. Conversely, let S be infinite. We define, for each se.S, the mapping a^'- S -^ S by a^: x -^ s for all x S. s e S} is an infinite set since as = s' if and only if s = s', and we assumed S to be infinite. {qis Therefore we have found an infinite number of mappings of S into S.
I
1.34.
If
is
any non-empty
set,
verify that
S matches
a: s
-* s
S.
Solution:
Hence a
Define the mapping is a matching.
a:
-^
S by
for each
S.
is
clearly one-to-one
and onto.
1.35.
If
S matches
If
T, prove that
T matches
S.
Solution:
S matches
is
T, then there is a
mapping a: S
t
^T
which
is
is
Define a:
of
t
T -^S
as follows.
s.
Let
T.
Then there
an
sG S
such that sa
t.
The image
under a
defined to be
for some
is a matching. In the first place, 5 is a mapping. For if ta = s and ts = s', then by definition of a, sa = t and s'a = t. But a is one-to-one, so that sa = s'a implies s = s'. Thus the image of an element under a is unique. Secondly, 5 is one-to-one, because ta = t'a = s implies Sa = t and sa = t', which in turn implies, since a is a mapping, t = t'. Thirdly, if s G S, then sa = t for some t G T. By definition of a, ta = s. Hence every element of S has a preimage under a and a is onto.
We now
t
show a
T,
1.36.
If
S matches T and
if
T matches
U, prove that
S matches
U.
(Hard.)
Solution:
Let a:S-*T and p-.TÛ be matchings. Then a:SÛ, defined by sa = (sa)p, is a matching, a is a mapping of S into U; for sa S U, and if sa = u^ and sa = M2 then {sa)P u^ and {sa)p = M2, which implies u^ = 1*2 since sa has a unique image under p. a is one-to-one, for s'. a is also sa = s'a implies (sa)p = is'a)p. But a and /3 are one-to-one, so that sa = s'a and s onto, for if M e C7 then there is a t 6 T such that tp = u. Now t has a preimage s e S under
a.
Thus sa =
{sa)P =^ tp
u.
1.37.
Let a be the mapping of

4
-^ 7,
S =
5 -* 9, 6 -^ 10, 7 - 1, 8 -> 3, 9
into itself given by a : 1 {1, 2, 3, 4, 5, 6, 7, 8, 9, 10} 4, 10 -^ 5. Then there is a useful alternative
3,
4, 3 -* 5,
following a: we list the elements of S on one line (in any order) and on the each element of S its image under a, enclosing the entire description in parentheses as follows.
way of describing line we place under
123466789
34579
10
1345
10
Describe the following mappings of
into S, using this notation.
: :
Sec. 1.4]
COMPOSITION OF MAPPINGS
1 - 2, 2
17
(i)
/3
2,
2,
4 - 2, 5 - 2, 6 -^ 3, 7 -> 4, 8
4, 9 -^ 4,
10
(li)
y:1-*6, 2-^7, 3^8, 4-^9, 5^10,6-*1,7^1, 8->l, 9^1, 10^1.

to decide
Use these descriptions

Solution
1
(i)
whether
(iii)
y,
(iv)
/?
is
one-to-one, (v) y
is
onto.
2 2 2
7
3 2 3 8
4 2 4 9
8
4 8
1
9 4 9
1
10
5
2
1
(ii)
2
5
3 6
1
4
7
1
10
1
6.
6
(iii)
j8
10
1/3
7^ is
Y since
and ly
1/8
=
2(3
It is
2.
only necessary to compare the bottom rows. only necessary to find a repetition on the bottom row.
(iv)
/3
not one-to-one,
e.g.
It is
(v)
is
1, 2,
not onto, e.g. 2 has no preimage. 10 appear in the bottom row.

. . . ,
It is only
necessary to check whether
all
the integers
1.4
COMPOSITION OF MAPPINGS
Definition
This suggests Let a: S-^T, p-.T^lJ. Because Sa T, we can compute {sa)p. "composing the mappings a and ^", i.e. defining a mapping of S into U by performing a and p in succession on each of the elements in S. More precisely we define a op, the composition of a with p (in this order) as a mapping of S into JJ defined by
s{a P)
{sa)p,
for
all s in
S
\&
(Some authors use exactly the opposite order, so that their ap

let
our p o a.)
For example,
S =
and
let
{1,2},
T -
{3,4,5},
U =
{6,7}
a:S^T,
p T
:
-*
be defined by
a:
1^3,
2-*
5,
/3:3^6, 4-7,
(la)/3
(2a);8
5^6
Then
l(aoj8)
= =
=
=
3/3
5/3
6
6
2(ao/3)
Hence
ao/3: {1,2}
{6,7}
is
defined by
aop: l-6,
2^6
is
This notion of the composition of two mappings give the following drill problems.
Problems
1.38.
of
tremendous importance; hence we
Let
i2
a:
P-* C be
defined by
na
in
and
let
= 1. What
Let
do aOyS, (a/3) a, and Q:(/3a)
p C -* P map n G P
:
be defined by
to?
p: a
a
+ ib
13
<>
^b^,
a"!
where
Why
is
Solution
n&P.
and
Then n(a
(/3
in^
a p:
-^P
n{a while
fi) = {na)p {in + l)/8 = n^. Now n{{a p) a) (n{a p))a = nâ = = (na){p a) = {(in + V)P)a = nâ in^ + 1. Hence {a p)a = a {jS a). pa: C -* C. Hence a p = pa.
a))
1.39.
Let a:Q-*Q be defined by a: aâ^ + 2 and let P Q -^ Q be defined by pute a p, /3a. Are these mappings equal? Compute {a p)a, a{p a).
p:
a^ â-2.
Com-
: :
18
SETS,
[CHAP.
Solution:
a{a p) = (aa)/3 = {a^ + 2)/3 = A(a2 + 2) - 2 = la2 - 1. a(/3 a) = (a^)a = (|a - 2)a = + 2. Clearly a 13^13 a. Furthermore, a([ap)a) - (a(a p))a - (â^ - l)a (lo2-l)2 + 2 and a{a{lia)) = (aa)(/3a) = (a^ + Z)l3a = ((a2 + 2)/3)a = ((|(a2 + 2) - 2)a = (i(i2 - 1)2 + 2. Note then that {a li)a = a o (/3 a).
Let
aGQ.
(Aa
2)2
1.40.
Employing the notation of Problem

^.
1,37,
compute the following:

,...^
' '
/I
2
2 3 2
3
1
4 5 4
3
5\
/I
2 2 2
4
5
5\
/I
2 4
2
3 5
4
3
5\
/I
2 2
2
3 4
3
4
3
5 5
4/
5\
Vl
4
3
3/
5\
^2
/I
1/
5\
^1
/ 1
,..^
/I
2
14
! ?
5/123451/
(ii) '""'
/I
,.
112345/125341
(Hi) '"*'
Solution
(i)
('i
I) ,24135/
!
r! VS
!
2
!
5
! 4
f^ 1/
(lll'l^ 4 l/ ^2
3
5
/3)
/I
(iv) '"''
V2
2 5
^ 3
^ 4
^ 1
1.41.
Let a: S-* T, p-.T-Û,

Solution
If
s
y.U^V.
=
Prove that
(a
(/3
o y).
(Hard.)
S,
then
(a o
j8)
s{{ap)y)
o
(s{a j3))y
((sa)/3)y
and
s(ao(/3y))
(sa)(/3oy)
((sa)/?)y.
Consequently
a o
(/3
y).
1.42.
Prove that
Solution
if
a:
-^
T and p T ^ U and
:
a =
/3
is
onto, then
/3
is
onto.
Is a
onto?
(Hard.)
Let u & U. As a j8 is onto, we can find a preimage of u under a /3. Let s & S be a preimage M under a fS, i.e. s(a! o /3) = u. Thus s(a! ;3) = (sa)y8 = tt and so: is a preimage of u under p. Hence /3 is onto, a need not be onto, e.g. let S = {!}, T = {1,2}, [7 = {1}. Define a: S-*T by la = 1, and 13: T ^ U by 1/8 = 2,8 = 1. o /? is onto but a is not.
of
1.43.
Prove that (Hard.)

Solution
if
a:
S^
T,
T^
and a;8
is one-to-one,
then a
is
one-to-one.
Is
/3
one-to-one?
Let
!
y8
Si, S2
G S and
But a p
let
sâ
is
S2a.
Sjo:
820:
implies
Sj
{sia)p
{S2a)p
S2a ,8. one-to-one, e.g.
one-to-one, so that
82-
Hence a
is
is one-to-one.
S=
1
{1},
2/3
T =
P
1.44.
T ^ V by
:
1/3
and
{T
a'
:
=
is
1.
{1, 2} a o y8 is
and
U=
{1}.
one-to-one but p
Define not one-to-one.
and, by definition of a y8, p is not necessarily and by la = 1, a: S->T
Prove that p
pings
a: Solution
T -* U S ^ T and
</))
->
T,
a p
one-to-one if and only if for every set a' p implies a = a'. (Hard.)
S and every
pair of
map-
a p
First assume that for every set S and every pair of mappings a: S ^ T and a' S -^ T, = a' p implies a = a'. Under this hypothesis suppose p is not one-to-one. Then we can find U. Let S = {1, 2}, let a: S-*T be defined by la = t t,t' such that tp = t'p = ( ^ t') and 2a = t', and let a' S -^ T be defined by la' = t' and 2a' = t. Now a p a' p, since
:
GT
=
u.
uG
u, 2aop = (2a)y8 = t'p = u, la' o p = (la')p = t' p = u, and 2a' p = (2a')P = Hence the assumption that p is not one-to-one is false and p must be one-to-one. To prove the converse, let p be one-to-one. Say we can find a set S and a pair of mappings a and a' of S into T such that a p a p and a = a'. So there exists s & S such that sa = sa'. a p = a' p means that s{a P) = s(' p). Hence {sa)p = {sa')p. As p is one-to-one, sa sa'. Therefore we have a contradiction and a must be equal to a'.
la/8
(la)y8
tp
tp
But a
= a'.
1.45.
(T =0) is onto iff for every set U and every pair of mappings Prove that a: S and p': T ->U such that ap - ap', it follows that p = p'. (Hard.)
Solution:
^T
->
Let us assume that for every set U and every pair of mappings p and p' of T into U such that a p = a p', it follows that p = p'. Say a is not onto, and t^ is an element of T which has no
Sec. 1.5]
BINARY OPERATIONS
a.
19
for
{1, 2} and define the mappings /3 and p' of T into U as follows: tji = X for all t ?^ tj in T and tiyS = 2. Now if s&S, sap = (sa)p = 1 and sa p' = {sa)p' = 1, since sa # ij. Hence a 13 a p' and /3 t^ y8'. This contradicts the assumption that a p = tt p' implies p = /?'. Thus a must be onto.
preimage under
all
t
Let
1
r,
t/3'
Conversely, assume a is onto and we can find a set U and two mappings, p and /?', of T" into U such that ap = ap' and P - P'. P - p' means there is a tj S T such that *iy3 # ti,8'. Furthermore, since a is onto, we can find s S S such that sa = i,. But a y8 = a p' implies (sa)p = (sa)/3'
or
tiP
tiP'.
Here we have a contradiction because we chose
t^p ' t^p'
We
therefore conclude
P'.
1.5
a.
BINARY OPERATIONS
Definition
The idea
of a binary operation
is
illustrated
which may be analyzed
in the following
way.
by the usual operation of addition in Z, For every pair of integers (m, n) there is
therefore think of addition as being a brief the image of {m,n) G Z xZ is denoted by + n. Any mapping /3 ot into S, where S is any non-empty set, is called a binary operation in S. We shall sometimes write instead of (s, f)/3 (the image of (s, t) under (3) one of sot, s-t, st, s + t or sxt. We stress that in all these cases the meaning of the various expressions sot, S't, st, s + t and s x Ms simply the image of (s, t) under the given mapping
associated a unique integer description of a mapping ot
m + n. We may
ZxZ
into
Z where
SxS
These notations suppress the binary operation /3, so there is danger of However, we will work with binary operations and the various notations so frequently that the reader will become familiar with the pitfalls. Incidentally, we read
^ of
Sx S
into S.
confusion.
Sot
s
as "ess circle tee" as "ess dot tee" or "ess times tee"
st
as "ess tee" or "ess times tee" as "ess plus tee" as "ess times tee".
+t sxt
s
The notation s o Ms called the circle notation, the notations s t and st are termed multiand the notation s + 1 is termed additive. We sometimes refer to s-t or st as the product of s and t, and s + t as the sum of s and t. The following problems will help to
plicative,
make
the various notations clear.
Problems
1.46.
Convince yourself that the following mappings are binary operations in P.

(i)
a: a: a: a: a:
PXP ^ P
P XP P P X P -* P
-*
defined by a defined by a defined by a defined by a

defined
(i, j)
-* i^,
where
{i,
j)
P.
(ii)
{i, j) -* i
(i, j)
+ j,
j
where
{i, j)
P.
{i,
(iii)
-^
ix
(regular multiplication of integers), where

Sj,
1,
(i, j)
j)
P.
(iv)
PXP-*P
(i, j)
{i,
-^2i+
(v)
PXP ^ P
by a
j)î + j +
(i, j)
e P X P. e P X P.
{i,j)
1.47.
Which of the following are binary operations

(i)
in
(throughout
(v)
P^)
a
a
(i, j) - i {i, j)
(ii)
-* i
+i j
(iii)
a
a
:
{i, j)
{i, j)
î-^
(i, j) -^ j
(iv)
^ i + j + i^
Solution :
In (i), a is clearly a mapping from P X P into P. So a is a binary operation in P. (ii) and do not define binary operations in P because not every element in P X P has an image in P: e.g. in (ii), a: (1,2) -^ 1 -2 = -1 gP; and in (iii), a: (1,2) ^ 1 h- 2 = | g P. (iv) and (v) define mappings from P X P into P. Hence they define binary operations in P.
(iii)
20
SETS,

i
[CHAP.
1.48.
Interpret the following (abbreviated) definitions of binary operations in Z, where arbitrary elements of Z.
(i)
and
denote
i j
i
(ii)
= {i + i)^ +j = ij
(iii)
ij
ij
=
i
i-k-j
(v)
j
(vi)
1X3=
i' j
V
t
{ix
j)
(iv)
ix
27 j
Solution:
Throughout,
(i)
(i, j)
Z^.
(iii)
a: (i,j)^(t a
:
(ii)
+ ;)2 {i, j)î- j =
(i,
{i,
j)
(ix
j)
(iv)
a:
j) -* i
î +
j
i
(v)
a: (i,j)îi
a
:
(vi)
(i, j)
^ i+
27}
1.49.
Check that the following are binary operations

(i)
in the plane R^.
midpoint of the line joining (^, V) (*' y') (x',y'), define if {x,y) = (x',y'). If {.x,y)
(x, y)
(ii)
[x', y')
= = {x + x',y + y'),
{xx',yy'),
the point (x, y) e R^ to the point ix,y){x',y') = {x,y).
(x', y')
G B^
where
where where
(x, y), {x', y')
e
e.
jR2.
(iii)
(x,y)'(x',y')
(a;, 2/)
where
2/
(x,y),(x',y')
R^.
R^.
jR^.
(iv)
(a;'. 2/')
(v)
(x,
2/)
(x',
3/')
= (x x', = (x + x', x'2/ +
j/'), 2/'),
(x,y),(x',y')
(x,
2/),
(x', 2/')
(Notice here
Solution:
(i)
how we have
abbreviated the definitions of the binary operations considered.)
o is
a binary operation since two points determine a unique line and each line has a unique
midpoint.
(ii)-(v) are binary operations because each has a unique image by virtue of the fact that addition, multiplication and subtraction are binary operations in R.
1.50.
Let
S=
a
:
{1, 2, 3}
and
let a
and p be the following binary operations

(2, 2) -> 3, (2, 3) (2, 3)
in S:
(1, 1) -* 1,
(1, 2) -* 1, (1, 3) -* 2, (2, 1) -* 2,
3,
(3, 1)
8,
(3, 2) -* 2,
(3, 3) -* 1
p:
(i)
(1,1) -*1,
(1,2)^1, (1,3)^2, (2,1)^3, (2,2)->3,
-^3,
(3, 1) -^ 3,
(3,2)->l, (3,3)^2.
Is
pi
((l,l)a,l)/3, ((i, l)/3, l)a.
(ii)
Compute
(iii)
Compute
:
((l,2)a,3)a,
(1, (2,
3)a)a, ((1, 2)^, 3)^3, (1,(2,3)/?)^.
Solution
(i)
(ii)
(iii)
(2, l)a = 2 and = (l,l);8 = l; ((1, 2)a, 3)a = (1, 3)a = 2; (1, (2, 3)a)a = (1, 3)a = 2;
;8
for
(2, l)y8
3.
((l,l)a,l)/3
((l,l)/3,l)a
(l,l)a-l.
((1, 2)^8, 3)/3
(1, (2, 3)/3);8
= =
(1, 3)^8 (1, 3)/3
= =
of
2;
2.
1.51.
Let
(i)
(ii)
S =
{1,2,3}
\X\.
and
let
be the set of
all
mappings
into S.
Compute
a p
Verify that the composition

Is
o of
mappings
a,
is
a binary operation in X.
(iii)
a for
all
elements
S Xl
Solution :
(i)
|Z|
27
(see
Problem
1.30,
page
15).
(ii)
The composition of mappings

Therefore
ir:X^^X,
a: defined by
-*
S and p S
:
->
(a, P)t!-
a p,
{a,
p)
S &
is
X^,
is
again a mapping of a binary operation.
into S.
(iii)
No. For example, if a: S defined by sa sp - 2 for all s e S, are two mappings of spo a = (sp)a = 2a = 1. Hence a P ^ P a.
S^
=
S
for
all
s.S,
and
p S
:
into S, then
s{a p)
{sa)p
^ S defined by = ip 2 and
1.52.
Let Q* be the set of nonzero rational numbers. Make sense of the remark that division (denoted as usual by -^) is a binary operation in Q*. Check whether the following statements hold for all a,b,ce Q*.
(i)
H-
-=-
a.
(iv)
-rib-i-c).
If If
-^
(ii) (iii)
(a-.-6)- c
a d
(v)
6 -^
a
c
^
-i-
c,
then b then 6
c.
a,
{(a-^b)-^c)-.-
-(6
(c^d)).
Sec. 1.5]
BINARY OPERATIONS
21
Solution:
Division
integers, then
(i)
is
w/x
2
a binary operation in Q*; for -^ y/z = wz/xy S Q*. Hence

-^
if
-^ is
w/x G Q* and y/z&Q* where w,x,y,z are a mapping of Q* X Q* -* Q*.
False; e.g. False; e.g. False; e.g.
#
-=-
3 2
-^ 2.
(ii)
(2 h- 1)
((1
(iii)
(iv)
True, a
True,
-^
&
(v)
b -^
= 1 and 2 -h (1 ^ 2) = 2 ^ | = 4. ^ 2) -^ 2) -^ 2 = 1 and 1 h- (2 (2 ^ 2)) = |. = a ^ e implies ac = ab and, as a = 0, c = b. c ^ a implies ab ca. Hence b = c, since

-r-
= 0.
1.53.
Let
(i)
be the binary operation in
J?
defined by
ab = a+
+ ab.
Verify that:
(ii)
(iii)
a,b,cGR, {ab) o c = a {b c). For all a,b e. R, ab = ba. Prove that if a # 1, then ah = ac
For
all
iff
c.
Solution:
(i)
(ii)
(iii)
+ b + ab)c = a + b + ab + c + {a+ b + ab)c = a + b + c + bc + ab + ac + abc a{bc) = ao{b + c+bc) a + 6 + c + 6c + a(6 + c4- 6c) = a-\-b + c-\-bc-\-ab-\-ac + abc ab = a + b + ab = b + a + ba = b a If 6 = c, then a o 6 = a o c for any a. If ab = ac and a = 1, a + e + ac. Therefore 6 + a6 = c + ac, 6(1 + a) = c(l + a) and, since a =
(oofe)oc
(a
then 1, 6
+ 6 + a6 =
c.
1.54.
Let
all
(i)
be the binary operation in R^ defined by

(x,y)(x',y')
(x, y) (x', y')
(xx'
yy',
yx'
+ xy').
Verify that for
(x,y),(x',y')Ax",y")&Rh
(x',y')o(x,y)
(ii)
{(x, y) o (x', y'))
(x" y")
,
{x, y) {(x', y') (x",
y"))
Solution:
(i) (ii)
(x, y) o (x', y')
((, y) o (x', y')) (x", y")
- yy' = = {x,y)o{{x',y')o{x",y")) = = = =
(xx'
yx'
+ xy') = (x'x - y'y, y'x + x'y) = (x', y') o (x, y) ~ yy', yx' + xy') o (", y") {{xx' yy')x" {yx' + xy')y", {yx' + xy')x" + {xx' yy')y") {xx'x" yy'x" yx'y" xy'y", yx'x" + xy'x" + xx'y" yy'y")
{xx'
- y'y" y'x" + y"x') - y'y") - y{y'x" + y"x'), y{x'x" - y'y") + x{y'x" + y"x')) {xx'x" xy'y" yy'x" yy"x', yx'x" yy'y" + xy'x" + xy"x')
{x,y)o(x'x"
,
{x{x'x"
{{x,y)o{x',y'))o{x",y")
1.55.
Let

(a)
defined by
(b)
aob =
a-b + ab,
6 =
^-"'""^
(c)
a6=^!-t-^
3
Determine which of the above binary operations satisfy (a o 6) o c = o o (6 o c) for all a, 6, c S Q (i)
(ii)
6 o
a for
all
a, 6
SQ
(2o0)o2 = (2 + 0)o2 = 2o2 = e.g. = 2 - = 2- (-2) -4 = 0. (loO)oO = ôO = ^ and 1(00) = e.g.
e.g.
Solution:
(i)
(a)
(6)
{aob)c' ao(boc) for some a, 6, c G Q; 2-2 + 4 = 4 and 2 (0 2) = 2 (0-2 + 0) (oo6)oc ^ ao(6oc) for some a, 6, cGQ;
l0 =
(c)
f
#
ao(6c)
since
6
(ao6)oc
for some
a,
6,ceQ;
(loO)oO
= ôO =
and
lo(OoO)
loO =
(ii)
J.
(a)
/.
V
aob=ba
a6 =
.
10 =
and 01
a
=
.
-1.
(6)
+
-
0.6
=
+
3
I'
6a
= 6oa
/\
(c)
o6 =
I.
+ o
a
b
= ba
22
SETS,
[CHAP.
b.
The multiplication
table
selves with them.
So far we have introduced a number of definitions and notations and familiarized ourThe object of this section is to introduce a "table" as a convenient way of either defining a binary operation in a finite set S or tabulating the effect of a binary operation in a set S. To explain this procedure, suppose S = {1, 2, 3} and let /i be the binary operation in S defined by
^:
(1,1)^1, (1,2)^1, (1,3)^2, (2,1)^2, (2,2)

(2, 3)
- 3,
-3, (3,1)-1,
/x
(3, 2)
-3,
(3, 3)
-2
Then a
table
which sums up
this description of
is
112
2 3 2
3
12 13
We
put the number (2, 3)/^ = 3 in the square that is the intersection of the row facing 2 (on the left) and the column below 3 (on the top). More generally, in the {i,j)th square, i.e. the intersection of the ith row (the row labelled or faced by i) and the ith column (the column labelled by j), we put {i, y)/i.
A table
of this kind
is
termed a multiplication table because

calls
fi
plication tables.
One often
in a set S,
multiplication
/j,
we mean
a multiplication in S. that ft is a binary operation in S.
looks like the usual multiThus when we talk about a

it
There is a reverse procedure to the one described above. out with a table
For example, suppose we start
112
2 3
3 2
13
(2,3)^0
Then there
is
a natural
{i, j)ti
way
We simply define
We
to be the entry in the

(1,1)^1
of associating with this table a binary operation /x in {1,2,3}. For example, {i, i)th place in the table.
1,
3,
(3,2)ft
3
tables.
shall usually define multiplications in a finite set
by means of such
Problems
1.56.'
Write down the multiplication tables for the following binary operations in S
(i)
{1,2,3}.
(3, 3)
a
/3
(1, 1) -* 2,
(1, 2) -> 3,
(1, 3)
^
=
1,
(2, 1)
3,
(2, 2) -^ 1,
(2, 3)
^
^
2,
(3, 1)
1,
(3, 2) -> 2,
3.
(ii)
S2
-^
defined by
(i,
j)p
for all
(i, j)
S^.
(2, 3)
1,
(3, 2) -^ 1,
(iii)
(1, 1) -* 1,
(2, 2) -> 1,
(3,
3) -^1,
(1,3)^
2,
(3, 1) -^ 2,
(2, 1) -* 3,
(1, 2) - 3.
Solution:
12
2 3 3
3
1
12
1 1
(ii)
3
1
12 13
3 2
1
2
1
(i)
12 12
(iii)
11
Sec. 1.5]
BINARY OPERATIONS
in {1,2,3}?
23
1.57.
Does the following table define a binary operation
In {1,2,3,4}?
12
1
3 4
3
3
1
2
3
2
1
Solution:
The table does not give a binary operation in {1, 2, 3} since (1, 3) -^ 4 {1, 2, 3}. also does not define a binary operation in {1,2,3,4}, because (1,4), (2,4), (3,4), etc., have
The
table
no images.
1.58.
Write down
explicitly the binary operations in {1,2,3,4} defined

1 1
by the following
tables.
2 2 4
1
3 3
1
4 4 3 2
1
2
3
4
1
1111
2
3
2
3
4
1
2 3 4
2 3 4
2
3
2 3 4
(iii)
2
3
4
1
2 3
4 2
(ii)
2
(i)
Solution:
(i)
(I,:;)
î
(;'
1,2,3,4);
(;,
1)
^
3;
(j
=
=
2,3,4);
(2,2)^3;
(2, 3) -^ 4;
(3, 2) -^ 4;
(3,3)^1;
(3, 4) -^ 2;
(4, 3)
2;
(4, 4)
(4, 2) -* 1;
(2, 4)
1.
(ii)
(1, i)
^;
(i
1, 2, 3, 4);
(j, 1)
(j
2, 3, 4);
-> 3;
(2, 2) -^ 4;
(2, 3)
1;
(3, 2)
1;
(3, 3) -* 4;
(3,4)-2; (4,3)^2; (2,4)^3; (4,2)

(iii)
(t, j)
(4,4)->l.
î {i=
1, 2, 3,
and
1, 2, 3, 4).
1.59.
Rewrite the three binary operations in Problem

(a) (6)
(c)
1.58,
using
circle notation,
i.e.
write
i.e.
{i,
j)/3
as
t /,
additive notation,
write
i.e.
(i, i)/3
as
(i,
+ j,
i
multiplicative notation,
write
j)p as
j.
Solution:
(a)
Problem
1.58(i):
1i
i (i 2;
1, 2, 3, 4);
;"
o 1
(;
=
l.
1, 2, 3, 4);
2o2
3;
2o3
=
=
3o2
3o3 =
303 =
l;
3o4 = 4o3 =
1.58(ii):
4;
4o4 =
3;
4o2 = 2o4 =
jo\
3;
Problem
4;
1 i
2;
= =
(i
i (;
= =
1, 2, 3, 4);
=
=
304 = 403 =
1.58(iii):
j (j
1.
2,ZA); 2 o 2
4;
3 2
1;
204 = 402 =
i
(i
404 =
j
Problem
(6)
ioj
1, 2, 3,
4 and
;
1, 2, 3, 4).
Problem
3
+ 1 = j (; = 1, 2, 3, 4); 2 + 2 = 3; 2 + 3 = 3 + 2 = + 2 = 2 + 4 = 1. Problem 1.58(ii): 1 + i = i and i + 1 = i (i = 1, 2, 3, 4); 2 + 2 = 4; 2 + 3 = 3 + 2 = 1; 3 + 3 = 4; 3 + 4 = 4 + 3 = 2; 2 + 4 = 4 + 2 = 3; 4 + 4 = 1. Problem 1.58(iii): i + j = i (t = 1, 2, 3, 4 and j = 1, 2, 3, 4).

1.58(i):
+3 =
1;
+4 =
+i = i 4 + 3 =
1, 2, 3, 4);
4;
2;
+4 =
3;
(c)
Problem
1.58(i):
!}
j
3;
and
3.4 = 4.3 =
Problem
2;
4.4 =
4. 2
1 = y (/ = = 2-4 = 1.
1, 2, 3, 4);
2-2 =
2
3;
=
2
=
3
4;
1;
1.58(ii):
!}
= -
and
3;
(i
(j
1, 2, 3, 4);
4;
1;
3-3 =
4;
3-4 = 4-3 =
Problem
2;
2-4 = 4-2 =
i'j
i
4-4 =
1.
1.58(iii):
1, 2, 3,
and
1, 2, 3, 4).
24
SETS,
[CHAP.
began with a few remarks about sets. We then introduced the idea of cartesian This led to the idea of an equivalence relation on a set. Then the notion of a mapping was defined, followed by the definition of a binary operation.
products.
We
In this book we are mainly concerned with binary operations in sets. At this stage the reader may wonder what one could possibly say about binary operations in a set. Without some specialization we can say very little. In Chapter 2 we begin to place restrictions on binary operations.
Supplementary Problems
SETS
1.60.
The sets Sj (i = 1, SiUS'2U---uS.

Let
3},
(v)
2, ..., n)
are such that
Sj
C Sj+i
(i
1,2,
.,
w-1).
Find SinSaO
nS and
1.61.
and
E = {n\ nG Z and n even}, O = { nG Z F={m| nGZ and w divisible by 4}. Find

|
and n odd}, T = {n\ nG Z and n divisible by (i)EnT, (ii) EuT, (iii) TnFnEnO, (iv) TuF,
OnF,
(vi)
OnT.
0,
1.62.
Given
A =
{-3, -2, -1,
{1,2, 3,}}
and
B =
{-3, -1, {1,3}}.
Find
AuB
and
AnB.
1.63.
Prove
Let
Sn{TuU) = {SnT)D{SnU).
{-5,-4,-3, ...,3,4,5}, B = {-4,-2,0,2,4}, C = {-5,-3,-1,1,3,5}, D = {-4,4}, of X if (i) X C = 0, 0}, F = 0. Which, if any, of these sets take the place = C, (iii) X C C but X is not a subset of A, (iv) XqB and X is not a subset of E,
(vi)
1.64.
A =
E=
(ii)
{3, 2, 1,
XnB
(v)ZnCcA,
1.65.
Zu(BnD) = A?
if
Prove
ScT
if
and only
(TnC)uS = Tn(CuS)
for every set C.
CARTESIAN PRODUCTS Prove S X (TuW) = (SxT)u(SxW) 1.66.

1.67.
for any sets S,
T and W.
Let
r
P be the set of positive integers and S = + w = s + t} is an equivalence relation on S.

relation,
P^.
Show that
E=
{p\
p =
{{r,s), (t,w))
(4, 7).
G S^ and
Find the E-class determined by
1.68.
the equivalence classes of F are {n\ n = Aq for for g S Z}, and {n n = 3 + 4g for q G Z}; of a single integer; (iii) if the equivalence classes are (ii) if every equivalence class consists {q, q} for each q G Z.
Find the equivalence

{n\
E, on Z:
(i)
if
qeZ},
n=
+ 4q
for
Z}, {n\
n = 2 + 4q
1.69.
If
and
is
are equivalence relations on S,
is
(i)
EnF,
(ii)
EuF
a.n
equivalence relation on S?
1.70.
is a non-empty subset of S^ which has the symmetric the following argument: properties. If (a, b) S E, then by the symmetric property, (6, a) G E. But by the and transitive Therefore 7 is also reflexive. transitive property, {a,b)GE and (b,a)GE implies (a,a)GE.
What
wrong with
1.71.
(a)
the set of positive integers, show that reflexive and transitive but not symmetric.
If
is
E=
{(a, 6)
(a, 6)
G pa
and a divides 6}
is
(6)
(c)
not reflexive, Find an example of a subset E of P^ which is both symmetric and transitive but reflexive and symmetric but not transitive. Find an example of a subset E of P^ which is
CHAP.
1]
SUPPLEMENTARY PROBLEMS
25
MAPPINGS
1.72.
Show
(i)
that the following define mappings of
into Z.
a:
\x\
is
the absolute value of
x,
i.e.
Ixj
' '
yxl\x\
if
cc
y^
' '
\-x
if
a;
<
....
(ii)
^. X
if
_
jxl l^l
or 1
\ [(I)''
if
or 1,
and r
is
the
number
of distinct primes dividing x
f-l
(iii)
y:
-*
\
[
if X < ifx = if X >
(iv)
->
sin2 x
cos^ x
1.73.
Which
of the
mappings
a,/3,y,S of the preceding problem are equal?
1.74.
(ii)
Find subsets S(^0) and T(^0) of the real numbers i2 such that the mappings (i) a x -* cos x, X ^ sin X, and (iii) y: x -^ tan x are one-to-one mappings of S onto T. Do a, /?, y define /3
:
mappings
Let
of
into
Rt
and n even}. Define a: E -* E by na n for all n&E, and p E -^ E w S . Find an infinite number of mappings a', p' of P into P such that
:
1.75.
E =
M;8
{n\
nG P
for
all
by
[e
2w and
Z?;^
^.
1.76.
Suppose a
(ii)
a mapping of a set S into a set T and, for any subset of S, some s&W). If A and B are any subsets of S, show: (AnB)acAanPa; and (iii) A CB implies AaQBa.
is
Wa (i)
{t\
t^T
and
sa
for
(AuP)a = AaUBo:;
1.77.
(ii)
Let S be a subset of a set T, and onto implies a is onto. a
a:
T -^W.
Prove:
(i)
a one-to-one implies a
is
one-to-one;
COMPOSITION OF MAPPINGS 1.78. Suppose a: S ^ T is onto and

1.79.
T
p
->
is onto.
Show a p
is
an onto mapping.
Given
a:S-*T
X
- sin (x^),
is
one-to-one and
T^ U
is
one-to-one.
Prove a
/3
is
one-to-one.
1.80.
(i)
(ii)
/?
-^
sin (sin (x)),
and
(iii)
->
^/l-x^
define
mappings
of
non-empty
subsets of the real numbers R into R. First, find an appropriate subset in each case. Secondly, write a, p and y as the composition of two mappings, giving in each case the domain and codomain of each mapping defined.
BINARY OPERATIONS
1.81.
Let S be a set and gf the set of all subsets of S, union define binary operations on gf.
i.e.
gf
{A
A C S}. Show
that intersection and
1.82.
How many
different binary operations can be defined on a set of 3 elements?
1.83.
Consider the
by:
set,
F, of mappings
/j (i
1, 2,
6)
of
jB
{0, 1} into
defined for each x

/g
:
G ft - {0, 1}
that
fi-.x-^x;
f^-.x^--; f^-.x-^^^^; f^:x-^^; f^-.x-^^;

is
^ 1 - x. Show
composition of mappings
a binary operation on
and write a multiplication table for the operation.
chapter 2
Groupoids
In this chapter
we
define a set
G together with
1,
As we remarked without making

The
at the end of Chapter

restrictions.
there
is
a fixed binary operation o to be a groupoid. little one can say about binary operations
first restriction is
if
that of associativity.
(,92 gs)
groupoid with set

all gi, gi,
said to be associative
called a semigroup.
gs for In order of increasing specialization

gi
{gi
fl'2)
ga in G.
and operation o Such a groupoid
is is
we have
the concepts of groupoid,
semigroup, and group.
To
ideas.
define a
group we need the concepts of identity and inverse.
Hence we discuss these
The importance of into X. introduce the semigroup Mx of mappings of that, but for the names of the elements, each semigroup is contained in some Mx.
We
Mx
is
other important concepts we deal with are homomorphism and isomorphism. Homomorphism is a more general concept than isomorphism. There is an isomorphism between two groupoids if they are essentially the same but for the names of their elements.
Two
2.1
a.
GROUPOIDS
Definition of a groupoid
of integers. Z has two binary operations, addition (+) and multiplication (X). The set Z is one thing, a binary operation in Z is another; the two together constitute a groupoid. Repeating this definition in general terms.
Consider the set
Definition:
groupoid
is
a pair
(G,/x,)
[j.
consisting of a non-empty set G, called the carrier,

in G.
and a binary operation
when dealing with groupoids. Thus we This notation has been used in Chapter 1 or simply gh for {g, h)ti, g,hGG. {1,2,3} and let /x be the in our consideration of binary operation. As an example, let operation in G defined by the following table. binary
We
shall mostly use a multiplicative notation
write g-k
G=
13
3
211
2
3
Then the pair
(G,
f^)
is
a groupoid.
Suppose we use multiplicative notation
then
1-1
1,
1-2
3,
2-2 =
26
1,
3-2 =
2,
etc.
Sec. 2.1]
GROUPOIDS
recall that the notation
3,
27
These products look bizarre unless we

version of
{l,l)ix
employed
2,
is
a shorthand
l,
(1,2)m
(2,2)f.=
l,
(3,2)m
etc.
If
we
have already been given a binary operation

the groupoid (G,
/*).
use the expression "the groupoid G," where G is a set, it is understood that we in G, and that we have been talking about
i^.
Suppose now that
(G,
ix)
is
a groupoid.
If
we
use the circle notation for
/j.,
i.e.
we
write
goh
instead of
{g,h)ix,
),
ilarly
we
write (G,
1:
sometimes write (G,o) to refer to the groupoid (G,/x). (G, +), (G, x) if we employ g'h,g + h,gxh respectively for {g,
shall
we
Simh)ix.
Example
Let
G =
{1,2}
and
let n
defined as follows:
(2, 2)ix
(1,
1)m
1,
(1, 2)/i
2,
(2,
1)m
1,
If
we
use the circle notation,
we have
1,
11 =
The pair
Example
2:
lo2 =
2,
2ol =
i,
22 =
(G,
/j.)
or (G,
is
then a groupoid.
Let
where makes
be the set of all mappings of {1, 2, 3} into {1, 2, 3}. Then (S, o) is a groupoid, is interpreted as the usual composition of mappings. The composition sense, for if a, j3 G S, then
a: {1,2, 3} ^{1,2, 3}
and
/?
{1, 2, 3}
{1, 2, 3}
Therefore a p, the composition of a and /?, is defined by aa /3 = (aa)p, a G {1, 2, and is once again a mapping of {1, 2, 3} into itself. (S, ) is indeed a groupoid.
3},
Example
be the binary operation in Q, the rational numbers, defined by ab a + Then (Q, ) is a groupoid, because for every pair of rational numbers a and h, ab defines a unique rational number a + h + ab.
Let
b
+ ah.
Example
4:
Let B2 be the plane. Further, let there be a cartesian coordinate system in R^ and C be the disc of radius 1 with center at the point (2, 0) of the coordinate system. Consider the region R^ - C, the unshaded area in the diagram. We term any path beginning and ending at 0, which does not meet any point of C (i.e. it is entirely in R^ C), a loop in R'^ C. By a path we mean any line which can be traced out by a pencil without raising the point from the paper. For example, I and m are loops in R^ - C. Let L be the set of all such loops in R'^ - C. Then there is a natural binary operation in L which we denote by thus if l^, l^ e L, then l^ I2 is the loop obtained by first tracing out l^, followed by tracing 4This type of groupoid {L, ) is of considerable importance in modern topology.
let
;
28
GROUPOIDS
Example
[CHAP.
5:
Consider two Let F be the set of all mappings of R, the real numbers, into R. elements a,p & F. We define the mapping a + p R -^ R by a{a + p) aa + ap, aG R. a + /3 is clearly a unique mapping of R into R and hence is an element in F. Therefore + is a binary operation in F and (F, +) is a groupoid. Notice that {F, +) is not the groupoid with F as the carrier and the composition of mappings as the binary operation.
:
Problems
2.1.
Are the following groupoids?

(i)
(S,o)
{Z,
where S
{1,2, 3, 4}
and ij
for
and
elements of S.
(ii)
(iii)
(F,
),
the set of integers with the usual subtraction of integers as binary operation, the set of positive integers with the usual subtraction as binary operation.
),
(iv) (v)
{Q,
-^),
the set of rational numbers with the usual binary operation of division,
{Z, +), the set of integers
with the usual binary operation of division.
Solution:
(i)
(S, o) is
a groupoid since a groupoid; for
" is if
clearly a binary operation in S.
(ii)
{Z,
is
a, 6
Z,
then
ab
for
is
a unique element of Z.
a, 6
(iii)
(F,
) is not a groupoid since operation in P.

(Q, -^) is not a groupoid operation in Q.
(Z, -V) is
is
a-b^P
-^
all
F.
Therefore
is
not a binary
(iv)
because a
is
not defined for any
aS Q
e.g.
and hence
-^
-^ is
not a binary
(v)
not a groupoid since not a binary operation in Z.
a^bZ
(i)
a.
for all
a, b
Z,
Z.
Therefore division
2.2.
Is {Z, ) a
(iv)
o 6
groupoid if o is defined as o 6 = a? 0, (v) a
= ^/a+b
(ii)
ab = {a+
6)2,
(iii)
aob-a-b-ab,
Solution:
All but
(i)
define a binary operation in Z.

in
(i)
Therefore
(Z, ) is a
grou poid
in
(ii)
through
(v).
The multiplication
integer.
2.3.
does not define a binary operation in
since y/a
b is
not always an
Let S be any non-empty set and T the set of

Solution
all
subsets of S.
Are
(T,
n) and (T, U) groupoids?
Both intersection n and union U are binary operations on T, for the intersection or union of two subsets of S is again a unique subset of S. Thus (T, n) and {T, U) are groupoids.
b.
Equality of groupoids
groupoids are equal if and only if they have the same carriers and the same binary operation. Remember, a binary operation was defined as a mapping and two mappings are equal if and only if they have the same domain and codomain, and the image of each 1-5 element is the same under both mappings. Thus the groupoids described in Examples
Two
are
all different.
Problems
2.4.
Are any two

Solution:
of the groupoids in Problems 2.1-2.3 equal?
No.
2.5.
Which
(i)
of the following pairs define equal groupoids?
(ii)
(iii)
=a+ (Z, ) and (Z, ), where a^b = ab. (Z, o) where ab - a for all a and
(Z,
+) and
(Z, m),
where
(a, b)ii
b.
6 in Z,
and
(Z,
X) where a X 6
&
for all a and 6 in Z.
Solution:
The groupoids in (i) are clearly the same. So the same as (Z, X); for if a # 6, ab - aX b.
too are the groupoids in
(ii).
In
(iii)
(Z, o) is
not
Sec. 2.2]
COMMUTATIVE AND ASSOCIATIVE GROUPOIDS
29
2.2

Let
(Z,
Definition of commutative and associative groupoids
+) be the groupoid of integers under the usual operation of addition. a
Then
{2.1)
{2.2)
and
for
all
{a
h)-\- c
= a-^{h + c)
a,b,c
Z.
)
Similarly
if (Z,
is
the groupoid of integers under multiplication,
= =
'
{2.3) {2.A)
and
for
all
{a-b)- c
a- {h-c)
a,b,c
Z.
The analog
of {2.1) and {2.3) in an arbitrary groupoid {G,o)
is
aob boa
for
all
{2.5) {2.4) in (G, o) is {2.6)
a,b
GG.
Similarly the analog of {2.2) and
{aob)oc = ao{boc)
for all a,b,c G G. We term a groupoid satisfying {2.5) commutative or abelian, and a groupoid satisfying {2.6) associative or a semigroup. Thus a semigroup is an associative groupoid. Of course it is not clear that there are non-commutative groupoids, i.e. groupoids which are definitely not commutative, and similarly it is not clear that there are nonassociative groupoids. We settle the issue now. Let G = {1,2} and let o be the following binary operation in G:
1
1
2
1
Then
i.e.
(G,o) is a groupoid.
is
Furthermore,
2 = 2 not a semigroup.
(2 o 1) o
Observe that lo2 = 1 but 2! = 2, so o 2 = 1 but 2 o (1 o 2) = (2 o 1) = 2, so G
is
not commutative.
non-associative,
is also
For the most part we shall use the multiplicative notation for a groupoid {G,^) and simply talk about the groupoid G. If the groupoid is commutative we will use the additive notation instead of the multiplicative notation, since we are accustomed to addition as a commutative binary operation, e.g. in the integers.
The order of a groupoid
(G,
/x)
is
the
number
of elements in
is finite.
G and
is
denoted by
|G[;
(G,/x) is infinite if \G\ is infinite,
and
fi,nite if \G\
Problems
2.6.
Which
of the groupoids in
Examples
1, 2,
and
5 are
commutative and which are associative?
Solution
The groupoid of Example To

(a)
(6)
(c)
1 is
shovir (G, o) is
associative
i^ 1, i_
we must examine
i
i
not commutative, since 1 o 2 = 2 and 2 the following 8 cases:

(e)
(/)
o 1
i,
but
is associative.
(d)
= ioi = 2(11) = 21 = l(2ol) = ioi = lo(lo2) = lo2 =

lo(loi)
2,
= ioi = (2ol)l = loi = (lo2)ol = 2oi = (lol)o2 = lo2 =

(loi)oi
{g) {h)
= 2o(lo2) = 2o2 = lo(2o2) = io2 =

2o(2oi) = 2oi 2o(2o2) = 2o2 =
i,
2, 2,
= 2ol = (2 ol) 02 = 102 = (lo2)2 = 22 =

(2o2)ol
2 2
2
2,
(2o2)o2
= 2o2 =
30
GROUPOIDS
The groupoid
is
[CHAP. 2
of Example 2 is not commutative; for if a is defined by la = 1, 2a = 3 and 3a = 2, by 1^ = 2, 2/3 = 1 and 3/3 = 1, then la ^ = (la),8 = 1/3 = 2, 1/3 a = (l/3)a = 2a = 3. Hence a p = 13 a. Since the binary composition of mappings is an associative binary operation (Problem 1.41, page 18), (S, ) is an associative groupoid. The groupoid in Example 3 is both commutative and associative. ah = a + h + ah = h + a + ba ba, since addition and multiplication are commutative binary operations in Q. Also, a o (6 o c) = a o (6 + c + 6c) = (a + (6 + c + 6c)) + a(6 + c + 6c) = a + 6 + c + 6c + a6 + ac + a6c and (ao6)oc = (a+6 + a6)c = a+6 + a6 + c + (a + 6 + a6)c = a+6 + a6 + c + ac+6c + a6c. Using the associative and commutative properties of addition and multiplication in the rationals,
and p
defined
we
a(p
see
o (6 o c)
(a o 6) o
c.
In Example 5 the groupoid (F, +)
+ a), a,pe.F
and
aG R
binary operation in R). (a/3 + ay) = aa + a(P +
a(a + p) = aa + ap = ap + aa = is commutative because use the fact that aa, ap G R and addition is a commutative {F, +) is also a semigroup, for a{{a + P) + y) = a{a + p) + ay = aa + y) = a{a + (/3 + y)) (here we use the associativity of addition in R).
(here
we
2.7.
Construct an example of a commutative groupoid of order

Solution:
3.
Let
S=
{a, 6, c}
and the binary operation

a a a
b
c
be defined by the multiplication table

6 6 c
c
a
c
(S, ) is clearly
a commutative groupoid.
2.8.
Show that the set Q* of nonzero rational numbers with binary operation the usual division of rational numbers, is a groupoid. Is it commutative? Is it associative?
Solution:
If
and
since
are any two elements of Q*, then

a,
"^
| = |^
is
^ unique element in
Q*
/^TÔ
is
6,c,d#oy
(e.g.
Therefore division
is
a binary operation in Q*.
However, division
neither commutative
i 2.9.
(i
|--î
2#|
and
in
= i-|)
Problems
nor associative
(e.g.
"^
(i
i)
i)
i).
Which
of the groupoids in
Examples
1-3, 5
2.7
and
2.8 are finite?
Solution:
The groupoid of Example 1 is clearly finite of order 3. In Example 2 the set S of all mappings of {1,2,3}
is finite.
into itself contains 27 elements,
and so
In Example
of
3, (Q, ) is
not
finite as
In Example mappings of R rîGR and ipj
5 the set into
is infinite.
there are an To show that

:
as follows. Let pj iJ -* Clearly p^ = pj iff i = j. different elements in F. Notice the pj are not all the elements of F. In Problem 2.7 the groupoid has only three elements and is therefore
. .
0.
number of rational numbers. we construct an infinite number be defined by rpi = r for all fi (i = 1, 2, 3, .) Therefore we have found an infinite number of
infinite
is
not finite
finite.
In Problem 2.8, since there
is
an
infinite
number of nonzero rational numbers, Q*
is
not
finite.
2^
a.
IDENTITIES AND INVERSES IN GROUPOIDS

The
identity of a groupoid
e in
Let G be a groupoid written multiplicatively. An element element of G if eg ^ ge = g
is called
an identity
Sec. 2.3]
IDENTITIES AND INVERSES IN GROUPOIDS
31
for every g &G. For example in the multiplicative groupoid Z of integers, 1 is an identity element. less natural example is the groupoid {1,2,3} with multiplication table given by
2 3
1
3
1
2 3
2 3
The element 3
is
an identity element of
since
3-1 =
1- 3,
3-2 = 2 = 2-3, 3-3 = 3 = 3-3

identity element.
One might ask vs^hether or not a groupoid can have more than one The following theorem settles this question.
Theorem
2.1:
If a groupoid G has an identity element, it has precisely one identity element. In other words if e and e' are identity elements of G, then e = e'.
is
Proof:
of G.
Since e
ee'
Hence
an identity element of G, and so e = e'.
ee'
e'.
But
e' is
also
an identity element
reformulate the proof of Theorem 2.1 in a different notation. Thus revert to the notation (G./x) and instead of the multiplicative notation gh we write (g, h)ix.^ Suppose e and e' are identity elements. Then as e is an identity element, (e, e')fx = e'. But e' is also an identity element. Hence {e, e')fi - e; and because the image of any element under the mapping is unique, e = e'.
It is instructive to
we
2.10.
Problem Which
(i)
of the following groupoids have identity elements?

(Z,
The groupoid The groupoid The groupoid
+) under the usual operation of addition.
(ii) (iii)
of nonzero rational of complex

all
numbers under
division.
numbers under multiplication.

{1, 2, 3, 4} into itself
(iv) (v)
The groupoid of
mappings of
{1, 2}
under the composition of mappings.
The groupoid with carrier
and multiplication table
1
(a)
1
(6)
2
(c)
2
1
2
(d)
2
1
2
Solution:
(i)
(Z,
+) has the identity element

e
0,
since
+z = z+
for
all
z e. Z.
(ii)
If this groupoid
particular,
(iii)
-=-
had an identity element e, then e^q = q for all 9 in the groupoid. In = e and so e = 1. But 1-4-2 = 1/2^2. Hence there is no identity element.
Recall
^
that a complex
number
is
any number of the form

since
a+bi where
a, 6
fi
and
Oai
(iv)
= V-i- 1 + (K is the identity element of this groupoid, 06i2 + bi = a + bi = {1 + Oi){a + bi). The identity mapping defined by ji = j, j G {1, 2, 3, 4}, is old since i(<i o = (ja)i = ja = (ji)a = i(i o ).
1,
(a 4- bi)(l -f Oi)
= a+
the identity element of the group-
i)
(v)
Only
(d)
has an identity element, namely
1.
b.
Inverses in a groupoid
If
we
use multiplicative notation for groupoids,
we
shall for the
most part reserve the
symbol
1 for the identity element.
32
GROUPOIDS
[CHAP.
In the multiplicative groupoid of nonzero rational numbers, one talks about inverses; for example, i is the inverse of 2, the determining factor being 2x| = l = |x2. In general if G is any groupoid with an identity 1, we term h an inverse of g {h,g G G) if
gh ^
Clearly
if
= hg
h.
is
an inverse of
6:
g,
then g
is
an inverse of
Examples
follow.
Example
Let
G=
{a, b,c}
be the groupoid with multiplication table a a

b
h
h
a
b
Then a
be
is
=
is
and a
Furthermore, the identity element of G, as inspection of the table shows. and aa = a = aa, so that c is an inverse of b, b is an inverse of c, a, b and c have no other inverses. its own inverse,
cb
Example
7:
be the groupoid of mappings of {1, 2, 3} into itself. Then the identity mapping by ji. = j {) = 1, 2, 3) is the identity element of the groupoid (see Problem 2.10(iv)). Now let <7 e G be defined by la = 2, 2or = 3, 3cr = 1. Then r e G, defined by It = 3, 2t = 1, 3t = 2, is an inverse of a because
Let
I
defined
la
o T
(U)t
=
i.
2t
1,
2(T o T
{2a)T
i.
3t
2,
Scr o
(3<t)t
It
which implies ar
Similarly, ra
Unlike identities, inverses in groupoids are not always unique. {1,2,3,4} be the groupoid with multiplication table
For example,
let
12
2
3
111 114
2
3
element in G. Moreover, 2 2 = 1 = 2 2 and 2 3 = 1 = 3 2; thus 2 has two inverses, 2 and 3. Notice that in this groupoid, 4 is not an inverse of 2 even though 2-4 = 1. The definition of an inverse requires both 2 4 and 4 2 to equal 1.
Here
1 is the identity
Problems
2.11.
Consider the groupoids in Problem 2.10 which have identity elements. the groupoids have inverses?
Solution:
(i)
What
elements in each of
Any
i.e.
integer z has an additive inverse, namely z; for z
( z)
( z)
z.
(iii)
All nonzero elements in the groupoid of complex
bi {a
and
b not both zero) is

1
numbers under multiplication have any element in the groupoid, then

bi bi
inverses;
a
bi
a
is
bi
\a +
bi
62
62
62'
an element in the groupoid and

(0
a
O-i).
-{a+bi)
bi
(a+bi)-
bV
(0
+ Oi)
has no in-
verse, since
Oi)(a
bi)
(0
Sec. 2.4]

Only mappings
t
33
(iv)
which are one-to-one and onto have inverses. Say t is a one-to-one mapping itself, defined by It = a, 2t = b, St = c and 4t = d, vi^here a,b,c,d are the elements of {1, 2, 3, 4}. We define the mapping a which will be an inverse of r, by aa 1, her = 2, Co- = 3 and d<r = 4. ct {a, b, c, d} = is a mapping of {1, 2, 3, 4} onto {1, 2, 3, 4}, since It o (T = (1t)ct ~ CU7 = 1, 2t a = b<T = 2, 3t o <t = ca = 3, and 4t a = d<r = 4. Hence {1, 2, 3, 4}. Ta We must also show a t = i. Now au t = It = a, ba t 2t = b, cct o t = 3t = c, dtr T = 4t = d. As {a, 6, c, d} is {1, 2, 3, 4}, a t = If t is not one-to-one then there are at least two elements, a and 6 {a = b) in {1, 2, 3, 4} which are mapped onto the same element, c, by T, i.e. ar c and 6t = c. Now if o- is an inverse of t, then a(T cr) = a and 6(t o o-) = 6 or (aT)(r = ca = a and (6t)<7 = cct = 6; thus ca = a and cct = 6. But under a mapping each element has a unique image. Hence we have a contradiction. So t has no inverse.
of {1,2,3,4}
onto
I.
i.
(v)
Both
and 2 have inverses since
and 2 2
2.
2.12.
What
Find the identity of the groupoid elements have inverses?
(Q,
where
a^b = a +
ab
(see
Example
3,
page
27).
Solution:
is the identity of {Q, ), since Qoa O + a + Oa = o and a0 = a + + Oa = a. To find an inverse for an element aG Q, we must find an x G Q such that aoa; = = a;oa. Now a X a + X + ax, so that x must satisfy the equation a + x + ax = 0. If a = 1, we obtain 1 + a; a; = 1 = 0; thus if a = 1, there is no x such that ax = 0. If a ^ 1, the equa-
tion
ax
1
=
^
o
can be solved, giving
=
i.
commutative,
+a
so that
ir',
have inverses.
+a
~T
thus
1.
-r
0.
Since (Q,
is
is
an inverse of
a.
Hence
all
elements of Q, except 1,
-*>*->
2.13.
Find the identity of the groupoid

Solution:
(F,
+) in Example 5 and show that every element has an inverse.
defined by roo = for all r G R, is the identity of (F, +). For aa = + aa = aa + aa = a{a + a) for all a & R implies a + a = a = a + a. If /3 is an inverse of a, then a + 13 = a and a(a + /?) = a<o for all a G R. But a(a + P) = aa + ap and aa = 0. Thus aa + afl = or (oa) = a/3. Consequently if p is an inverse of a, the image of a G jB under /3 must be the negative of the image of a under a. We therefore define an inverse p for the mapping a by a/S = -{aa), a G R. /? is a mapping of R into R, since -{aa) is a unique element of R. Furthermore, a{a + 13) = aa + af3 - aa + ( (oa)) = = aa implies a + /3 = a. {F, +) is a commutative groupoid. Hence a + l3 = u = l3 + a and /3 is an inverse of a.
The mapping
aa
a:
-^
a{a
+ a) =
+ aa =
aa
R =
2.4
a.
Uniqueness of inverses Suppose G is a groupoid with an identity 1 and suppose h is the inverse ot g: gh = l hg. It is tempting to employ the notation used vsrhen dealing with real numbers and write g~^ for an inverse of g. The trouble with this notation is that in a groupoid an element may have more than one inverse, as we have already seen in Section 2.3b. However, the associative law on a groupoid prohibits this as we see from
Theorem
2.2:
Let G be a semigroup with an identity element 1. li g G G has an inverse, it has precisely one; i.e. if h and h' are inverses of g, then h h'.
Proof:
= = = =
hi
h(gh')
(since hi (gh'
h for
all
is
h G G)
g)
=
=
1,
since h'
an inverse of
(hg)h'
Ih'
h'
(by associativity) {hg

1,
since h
is
an inverse of
g)
34
GROUPOIDS
Theorem
[CHAP.
multiplicatively,
2.2 entitles us to denote the inverse of an element g in a semigroup, written hy g~^. Note that if g and h have inverses, then gh has an inverse, namely
h~^g~^.
(Notice the reversed order.)

(gh){h-'g-')
For,
= =
1.
iigh)h-')g-'
{g(hh~'))g-'
(S'l)fi'"'
99~
Similarly, {h~'g-'){gh)
Problems
2.14.
Let
G=
{1, 2, 3, 4}.
The binary operations on G given by the following
tables
make G
into
groupoid.
1234
1
1234
3 2
(6)
2
4 2 4
3 2
1
4 4 4 4
2
(a)
3 4
12 14 12
3 3
1234 1111
2
(c)
2 3 4
2 3 4
2
3
Which
of the groupoids are semigroups?
Which have an
identity?
Which elements have
inverses?
Solution:
(a.)
In order to see that in this case G is a semigroup, we must check associativity, i.e. we must show a{bc) = (ab)c for every a,b,c&G. Notice that when either a, 6 or c is 1, a{bc) is clearly equal to (ab)c; e.g. if c = 1, {ab)X - ab = a(bl). Since 4^ = 4 = 9-4 for any g e. G, then (ab)c = 4 = a(bc) if either a, 6 or c is 4. Therefore we need only check the products when a, b and c have values 2 or 3. If two of the three elements a, b, c are equal to 2 and the other and 2 3 = 2 = 3 2. The followis equal to 2 or 3, then {ab)c = 4 = a(bc) because 2 2 = 4
ing calculations take care of the remaining cases:

3(3 2(3
1 is
3) 3)
= =
3 2
1 1
3 2
= =
3 3
= =
(3 (2
3)3
3(2
3(3
3)
3 3
2 2
2 2
(3
2)3
3)3
2)
(1-2)
1
(3-3)2
3;
the identity element of G. are unique.

(6)
The only elements which have inverses are
and
these inverses
The
identity,
associative law does not hold, since 4(2 3) = 4 4 = 1 and (4 2)3 = 2 3 = 4. 1 is the and 1, 2, 3 and 4 have inverses. Notice that the inverses are unique even though G is
not a semigroup.
(c)
is
a semigroup.
(ab)c
a{bc)
Associativity follows from the fact that ab = a for all a,b G. G; hence for any a, b, c in G. G has no identity element; therefore no element has an
inverse.
2.15.
aob
be the groupoid with carrier Q, the set of rational numbers, and binary operation o defined by a+b ab. Is the groupoid (Q,o) a semigroup? Is there an identity element in (Q, o) ? Which elements of the groupoid have inverses?
Let
Solution:
Notice
a+ b ab
in (Q,
),
is
a unique rational number, so that (Q,
o)
is
a groupoid.
Now
if a, b, e
are
any elements
(aob)c
= (a+b ab)c = a + b ab + = a+b + c ab ac bc + abc

=
ao(b
a
{a
+ b ab)c
be)
and
a(bc)
+ c be) = + 6 + c 6c
a
a.6
-\-
-k-
be
a{To \- c
a.c
abc
Hence (Q, ) + o Oa =
inverse of a
is a.
is
a semigroup.
The
Using an analysis similar

r a-\
identity of (Q, ) is to that in
since
Problem
aoO = a + aO = a and a # 2.12, we find that for
the
and that a
has no inverse. To check that
a-\
is
the inverse of
a.
Sec. 2.4]

a r a1
35
a1
\
)
\a 1/
-{
a(a 1) ^ o
^^
Similarly,
0.
2.16.
Let G be the groupoid with carrier Q and binary operation defined by ab = ab + ab. Is the groupoid (Q,) a semigroup? Is there an identity element in (Q,)l Which elements of the groupoid have inverses?
Solution:
(Q, ) is clearly a groupoid.
o (6 o c)
= =
ao
(6
c + 6c) =
(b
c+
+
ab
be)
a(b
c+
be)
+ c
ab
bc
+ +
ab
ac
+ +
abc
and
(a o 6) o c
(a.
6 + ab) c a
c+(a~b + ab)c = a~b +

all
ac
bc
abe
values of a, b and c. For example, (0 0) o 1 = o 1 = 1. 1 = 1 while (0 o 1) = Hence (Q, ) is not a semigroup. Furthermore, {Q, ) has no identity element; for if e were an identity, then 6! = 1 and eoO = 0, since 1,0 G Q. But eol = e 1 + 6 = 1 implies 2e = 2 or e = 1, and = e0 = e + implies e = 0. This is clearly impossible. Therefore (Q, ) has no identity element.
These two expressions are not the same for
2.17.
If Q is replaced by Z, the integers, as the carrier for the groupoids in Problem 2.15, are the solutions the same? (Hard.)
Solution:
(Z,o)
ment for the

and
o defined as in Problem 2.15 is a semigroup with an identity, since the arguassociativity depended only upon the associativity and commutativity of addition multiplication in Q. These laws also hold in Z. The same is true for the proof that is the
with
identity of (Z,
).
However, the inverse of an element a(#

a
1)
G Z would
be
""
^ "-"^
Z.
Our problem
then
1 an integer? Let ^^3-j- = r, an integer. Then a = r(a- 1). Clearly r = 1 is impossible, (a) Assume first that a > 1. Then r must be positive, and so r ^ 2. Hence o ^ 2(a 1) and thus - o - 2. Therefore a = 2. If a = 2, then = 2 is an integer. (6) Now assume a ^ 0. If a = 0, then 1 a ^_. =0 is an integer. If a < 0, r must be positive and - 2. Then a = r(a. 1) =s 2(a - 1) and O a 2, which is impossible. Thus and 2 are the only elements with inverses.
is:
for which integers
is
~
ct
2.18.
Let G be the mappings of P, the positive integers, into P. Determine whether G is a semigroup with an identity element if the binary operation in P is (i) the composition of mappings, (ii) the addition of mappings, where a + /3 is defined by a{a + p) aa + aji, a,peG and o 6 F.
Solution:
(i)
The composition of mappings is an associative binary operation (see Problem 1.41, Page 18). Then G with the binary operation of composition of mappings is a semigroup. The mapping defined by ii = j for all j e P is the identity of G; for if Te.G, then /(i t) = (;i)t =
I
JT
(ii)
(;V)i
i(T o
,).
that the addition of mappings in the set F of all mappings of R an associative binary operation. The argument here is similar. Thus G is a semigroup. If a were an identity element in G and if p e G, then a + p p. Thus if j e P, then j{a + P) - ja + ip = j/3; hence ja = 0. But a & G, a P -> P and, since P, this is a contradiction. Thus (G, +) has no identity. (Compare with Problem 2.13.)
In Problem 2.6
into
we showed
is
2.19.
Let
a, b, c
and that
Solution
be three elements in a semigroup this inverse is (c~i6~i)a~i.
which have inverses.
Prove that a(bc) has an inverse
We
need only verify that
{a(6c)}{(c-i6-i)a-i}
a{(6c)[(c-i6-i)a-i]}
a{[(6c)(c-i6-i)]a-i}
a{la-i}
aa-i
and similarly that {(c-i6-i)a-i}{a(bc)}
to
prove the result.
Note our use of the associative law.
36
GROUPOIDS
The semigroup
of
[CHAP. 2
b.
mappings
of a set into itself
is the set Mx of all mappings of a given non-emptywhere the binary operation is composition of mappings. We repeat the definition of the composition of mappings in this special case. Suppose fx.: X ^ X and y: X -* X, i.e. /x, y S Mx. We define ft o y to be the mapping of X into X given by
particularly important semigroup

into itself,
set
x{tx o y)
{xix)y
for
all
X
/x
It is clear that o is
a binary operation in Mx.

/i
or simply
ixy,
instead of
If
o y.
We now
show
set,
is
We shall use the multiplicative notation Mx is a semigroup with an identity.

is
y,
Theorem
Proof:
2.3:
X is any non-empty
Mx
Mx
a semigroup with an identity element.

Let
jn,
We
begin by proving
a semigroup.
y, p
G Mx and
let
G X.
Then using the
definition of composition of
3:idyp))
mappings,
(a;fx)(yp)
((a;/x)y)p
{x{i,y))p
x{{ii.y)p)
Since x
is
Thus
(juy)p
any element of X, = iîyp), and so Mx
{ixy)p
and
iU,(yp)
have the same
effect
on every element of X.
is
a semigroup.
i:
To show that Mx has an xGX. Then if ^ G Mx,

Thus
identity element, let
X X
-*
be defined by
xi
for
all
x{ifji)
(Xt)/i
=
i
Xfn.
(Xix)i
X{fii)
tp,
=
is
p,
theorem
= for complete.
/x,t
all
G Mx;
hence
is
an identity element of Mx.
The proof of the
a
1
Not every element in Mx necessarily has an inverse. For example, if X = {1,2, 3} and G Mx is defined by la = 1, 2(t = 1, 3<7 = 1, then a has no inverse; for if ay we have = It = l(CTy) = (lor)y ly and 2 = 2t = 2(ory) = (2cr)y = ly, SO that 1 would have two disi,
tinct
Mx consisting of
Theorem
Proof:
(a;y)p
which contradicts the assumption that y is a mapping. The subset of those elements which have inverses is very important. We characterize these elements in the following theorem.
images under
y,
all
2.4:
An
element in
Mx has
an inverse
if
and only
if it is
Suppose p has an inverse is a preimage of and xy
GX
y;
then p is onto. Moreover, p x.

X
For
is
if
GX,
=
then x
=
XiJ.
xl
a;(yp)
one-to-one.
For
2/'
if
y\i.,
then
(a;p)y
ic(py)
xi
and
(i/p)y
3/(py)
2/
Therefore, since y
is
a mapping, xp
j/p
implies x
is
y; in other
words, p
Define
is
one-to-one.
will be element xGX, we define xy-y where y is that of which is the preimage of x under p. To check that the definition of y is meaningful, such that i/p = a;. observe that as p is onto there certainly is at least one element y is the unique element But p is one-to-one, i.e. distinct elements have distinct images. So y and xp = y. such that Vix. x. To conclude we show y is the inverse of p. Let x G = (a;p)y yy x by the definition of y, and so py = t. Then a;(py)
To prove the converse, we assume p shown to be an inverse of p, as follows: if
y,
which
GX
Since p
is
onto, each x
^{y\>)
GX
must have a preimage y GX,
i.e.
^p
x.
Then
x
iViAiyi^)
((î")y)M
(2/(/^y))^
(^')m
yi^
and so yp
i.
Thus
y is the inverse of p
and the proof of Theorem
2.4 is complete.
Sec. 2.4]
37
c.
Notation for a mapping
When
a
is
finite,
say
{ai,
.,a}, there is
a convenient
way
of denoting any
G Mx, namely
That
is,
we
place below each element
ai
1<t
of
X
2
1
(i
X=
{1,2,3} and a G
is
defined
by
=
1
1, 2(t
= 1, 2, = 1 and
.
its
2,
image under
<j
a; e.g.
if
3<r
then
is
represented by
3'
Notice that every element of has a unique image under an element a in Mx; therefore the /^ ^2 as ... a\ ^^jj (.qju^^jj^ î\ ^^g elements of and under each element will top row
appear
as
its
we
see
unique image under from the example above.

<t.
All elements of
need not appear in the bottom row,
Problems
2.20.
(a)
Write the element a in Mx, defined by

(i)
X=
4ct
{1, 2, 3, 4, 5, 6},
in the notation introduced above,
when
a is
1<7
(ii)
la la
= = =
1, 1, 6,
2a
2a 2a
= = =
4,
1, 5,
3ct
5, 1,
= = =
6,
1,
5a 5a 5a
3a 3a
(iii)
= =
4a
4a
4,
3,
= 2, = 1, = 2,
6<r
= = =
6
1 1
6a 6a
(6)
What
,.,
elements of
b c c
M^, X
e\
e
,..,
{a, b, c, d, e},
are represented by the following?
/a
d d
\b
Solution:
/a \a
b
e
d
c
e\
bJ
/a \e
d
b
/I
<"^
2 4
3
5
4 6
5 2
6\
/I
(")
2
1
3
1
4
1
5
1
6\
.....
^Hl
(i)
ej
(l
ij
<"')
(654321
/I
(6)
The mappings defined by

a a
-> 6, -* a,
e,
a, c -^ c,
- d,
e -> e
(ii)
6-e, c-*d, d-c,

b -* a, e -*
c,
e^6
e -*
(iii)
a->
-* b,
2.21.
Exhibit
all
elements of
M^ when
(i)
X=
,
{1},
(ii)
X=
1
{1, 2},
(iii)
Z=
{oj, dj}.
Solution:
(i)
There
is
only one element in M,j>
namely
'
^
o
1,
the identity mapping.
(ii)
The following are the elements
/I
of M^^_ 2)
{
-,
2\
I
'
/I
{
2\
1
'
/I
(
2\
/I
i)'(
.
Ml
\i
<H\
12/
/a,
a,\
/a,
a,\
Vi
i/
\2
i/
J W
a,
a,\
a^J
^ê the only elements of M. , "'''"^'
2.22.
Write out the multiplication tables of the three semigroups
in
Problem
2.21(i)
and
(ii).
(i)
(ii)
The
identity of M.^,^..
is
/I
(
38
GROUPOIDS
1
[CHAP. 2
Let
"1
"2
and
ffs
Then the multiplication

2
"z
t'S
table is
2
"2
(
"1 "1 "1 "1
"2 "3
I
"1
"3
<'3
"2
"2 "3
"1
"3
The multiplication
Affj 2}
3
is
calculated as follows.
first
Since
is
the identity,
leaves every element of

ja^
unchanged; hence the
row and
first
column are easily written down. Since
1,
1,2,
and
if
cr
M,j
a^,
ji
then k(aai)
(k<j)ai
1;
thus aa^
ai-
Hence the second column

still
consists of
CTj.
Similarly
aa^
so the last
and
each element of {1, 2} takes every element to 2; hence aj^^ (202)02 ^"2 2; hence ct2<'^2 ' ^^
CTgcrj.
Now
column consists of is taken to 1 by
a^.
We
and
2
must
1 is
calculate
ai<72, <T2<r2
a^,
taken to 2 by
ctj,
so
(Tjcra
= ^3. Now lcr2<'2 then JCT3 = ^ ^ {!> 2},
ii<^2)''2
^"2
^nd
2^2
2o'2CT2
and so
^(73(72
1.
Thus
<'3<'2
''l-
2.23.
Let
X=
{1,2,
.,n}
and a
G Mx- Show
that a has an inverse
if
and only
if
the bottom
row
in
/I
the representation of a given on page 37, viz. and only once.
Solution:
(
\
)
>
2a
contains every element of
once
no} = has an inverse then a is one-to-one and onto. Hence {la, 2a, Sa, /I 2 n \ contains all elements of X once and only {1,2, ...,m.} and the bottom row of \la 2a ... naj once. Conversely if the bottom row of the representation has all the elements of X once and only once, then each element in X has a unique image under a, namely the entry under that particular element. Therefore a is one-to-one. a is also onto, for if j & X it must be one of the elements in the bottom row and j is then an image of the element of X appearing above it in the representation. As <T is a one-to-one and onto mapping it has, by Theorem 2.4, an inverse.
By Theorem
2.4, if
ct
. ,
'
2.24.
List all elements of M,^^}

find all representations of elements of
Solution :
Problem 2.23 shows us that we must
M,^2\
'
"with
bottom
/
row containing
and
2.
Using the result of Problem
2.21 (ii),
we have
/I
I
2\
)
/I
(
and
2\
1
*^ *"
only elements of M,i2)
which have inverses. In the notation of Problem
2.22,
and
02 are the only
elements which have inverses.

2.25.
List all elements of M,j g^j
Solution
The
possible representations of elements of

1
M^^^.m which have inverses are

2
3\
12
1
"4
3/ 3\ 1/
"^"^13
_ /I
2
_ /I
3\ 27
3\
"^""(^213
_ /I "^""13
2
2
3 1
/I
2 3
,2
"'^"(312/
2.26.
What
are the inverses of the elements in Problem 2.25?
Solution:
Theorem 2.4 explains how to find the inverse of a mapping which is one-to-one and onto. For example, to find the inverse of ag, we note a^ takes 1^2. Hence a~i 2 -^ 1. (T3 3 -> 3, hence <J~' = 05, "^^ = "4' <T2, a ' = ai, a^ 1 ff-i 3 - 3. But then cg is its own inverse. Similarly
: : :
.Tg.
Sec. 2.4]
39
2.27.
Prove that the subset of elements of the usual compositions of mappings.

Solution:
Mx
which have inverses
is
a semigroup with identity under
Let S be the elements of Mx which have inverses. Is the usual composition of mappings, a binary operation in SI In other words, is (a, /3) ^ a /3 a mapping of S X S ^ S"! If ao p e S, the answer is yes. So we ask: if a, 13 have inverses, does a^? Note that (a /3) (/3-i oa-i) = ao(ô^-i) Off-i = aoioa-i = I. Hence a /? has the inverse I3^â~^. As is its own inverse, S S. As Mx satisfies the associative law, so does S. Hence S is a semigroup.
,
d.
The order
in
a product
There is one way in which the associative law makes it easier to work in a semigroup than in a non-associative groupoid. Suppose S is a semigroup and let ai, a2, as G S. There are two ways in which one can multiply ai, ai and as together (in this order): {aia2)a3 and ai{a2a3). The point of the associative law is that these products coincide. Suppose now ai, ai, as, at G S. Then we can multiply ai, a^, az, ai together (in this order) in the following five ways:
ai{a2iazai)), {aia2){azai), ((aia2)a3)a4,
{ai{a^z))ai, ai{{ai<as)ai)
some of these products give rise to different elements of S. However, the associative law, which of course involves products of only three elements, prohibits this. To see this consider first {aia2){azai). By the associative law, {aia2){asai) = ai{a2{a3ai)); hence the second product coincides with the first. In fact all of the products equal the first. As a second illustration consider ((aia2)a3)a4. Here we have, as desired,
It is conceivable that
{{aia2)az)a4,
(ai(a2a3))a4
ai{{a2a3)ai)
01(02(0304))
In general,
we have
Let S be a semigroup and
ai, 02,
. .
Theorem
2.5:
let
. ,
On coincide,
last.
when
.,a e S. Then any two products of 01,02, the Oi appears first in each product, 02 sec. .
ond
Proof: are equal.
and On
the contrary, that not all possible products of Oi, , a in that order that n is the first integer for which two different products give rise to different elements. Let x and y be these two different products. Now x = uv and y - uivi for some u, v and some Ui, vu Suppose u is the product of the elements Oi, Or
. . .
Assume
We may assume
and V the product of the elements Or+i, .,a, while ui is the product of oi, .,0, and Vi is the product of Os+i, .,a. Without loss of generality we may suppose that s^r. If s = r then u = Ui and v = Vi, since n is the first integer for which there exist two unequal products of the same elements. If s < r, then u = (oi. .Os)(as+i. .Or) while vi = (tts+i. .ar)(ar+i. .On). Hence x = uv = {(oi. .Os)(os+i. .Or)}(or+i. .a) while y = uiVi = (oi. .as){(as+i. .aT)(Or+i. .o)). By the associative law for the three elements (oi. .Os), (as+i...aT) and (ar+i...a), we have x = y, contradicting the assumption that not all possible products are equal. Hence the result follows.
. . .
It follows from Theorem 2.5 that in a semigroup, if we are given the order of a product, the bracketing is immaterial. Thus we write simply 0102. .a, without brackets, for the product of oi, 02, o in this order. For example, if is any positive integer, we write â a for the product of o's; a useful abbreviation for such a product is o*". If n is a second positive integer, then a"* o" = 0*"+" since a"" o" is simply the product of 4- n
. . . . ,
o's.
Similarly,
(o"")"
a""".
40
GROUPOIDS
[CHAP. 2
For example, if S = {1,2,3} following table, then 1^ = 3, 2^ =
is 3,
the semigroup with binary operation given by the
3^
3.
12
2
3
1
2 3
12 12
3
2.5
a.

Definition of a
homomorphism
homomorphism
of one groupoid into another,
Before
we
give the formal definition of a
Let {R, ) be the groupoid of positive real numbers with the binary operation of ordinary multiplication. Let {R,+) be the groupoid of real numbers with binary operation the usual addition inside R. Let us define a mapping e.R-^R by 6 is an a;6i = logioa;. Recall that logio(a;2/) = logioa; + logiot/; hence {xy)e = xO + yO. is example of a homomorphism. The formal definition
we
will give
an example.
Definition:
A
e:
homomorphism from a groupoid

G-*
{G,a) into a groupoid {H,p) is a
mapping
which satisfies the condition

{{9u92)<x)e
{gi9,g29)p
for
all gi,
g% in G.
Usually groupoids are written in multiplicative notation. form: A homomorphism of (G, ) into {H, ) is a mapping
{gig2)B
The
definition then takes the
e.G^H
such that
(^-T")
gi6g29
for
all gi,
g^ in G.
{2.7) is
often expressed as: "6 preserves multiplication."
Problems
2.28.
of
G be the semigroup of integers under the usual addition of integers and even integers under the usual addition. Verify that the mapping e: 9 ^2g for all ff G Gf is a homomorphism of G into H.
Let
First
let e
:
H be the semigroup G -* H defined by

a
Solution:
mapping.
Then gê + QiB = 29^ preserves multiplication and is a homomorphism.

Let
we must check
to see if e is
a mapping.
Since 2g is a unique integer, e
is clearly
0^,92^ G.
ig^,
2{gi
+ g^) =
(gx
+ 92)6.
Therefore*
2.29.
be the Let G be the semigroup of integers under the usual multiplication of integers, and let defined by a: g -^2g of even integers under the usual multiplication. Is a: semigroup for all g G G a homomorphism of G into Ht
G^H
Solution:
As
then
in the
preceding problem,
2ffiff2
<r
is
a mapping.
(ffifTzV
^.nd
gy<jg^<j
^2g^g2 =
But a is not a homomorphism; for if for all ^^g^di- Hence {9192)" ^ (gi<')iff2<')
g^, 92
e G, g^gi^G.
2.30.
(H, +) which
Let (G,+) and (H,+) be the semigroups of Problem 2.28. is not equal to e.
Find a homomorphism of (G,+) into
Solution
Define
since
Tj,:
ng &
by t^: g ^ ng where m is a fixed even integer. t is a mapping of G into H, even integer n and ng is unique. For each n, t is a homomorphism because for any
G^H
Sec. 2.5]
41
9i,g-i^G implies (ffi + sr2)'-n = '^(ffi + ^2) = ^S'l + S'2 = S'l^n + fl'2Tn- r^^ 9, since U = It^ if n # 2. Therefore there is an infinite number of different homomorphisms between (<?,+) and {H,+). Notice r is an onto mapping if n = 2, because any element in H is of the form 2q, q an integer, and qT2 = 2q. But when n ^ 2, t is not onto since 2 has no preimage under t. For if qr^ 2, q&G, then nq = 2 or q- 2/n (if n ^ 0) and 2/n g G.
2.31.
Let
all
G
g
and
H be
is
&
G,
as in Problem 2.28. Verify that the mapping a: not a homomorphism.
G^H
defined
by a-g^g^, for
Solution:
Let
Hence a
2.32.
is
gi.g^SG. Then (srj + ffz)" = (9'i + S'2P = gl + gl+2gig2 and not a homomorphism for (g^ + g^)/! ' gâ + g^ for all gi,gi S
a of
gâ
(3.
g2<'
gf
+ gl
Is the mapping homomorphi sm ?
into i?,
G and
as in Problem 2.28, defined by
ga
for
all
G,
Solution:
If
*'i.
92
G> then
(ffi
+ g2)<' =
and
gi<r
+ gâ =
0.
Hence
<r
is
a homomorphism.
2.33.
Let G be the semigroup of positive integers P under the usual addition, and let of positive integers P under the usual multiplication. Show that the mapping ri gtl = 29 for all g G G is a homomorphism.
Solution:
V is clearly
H
:
be the semigroup
G^H
defined by
a mapping.
Let gi,g2
G.
Then
(ffi
+ ^2)') =
91
+ B2
2'2
'
givg2V-
Hence
1;
is
a homomorphism.
Let G = ({1,2, 3},a) and by the multiplication tables
2.34.
H=
{{a,b,c},p)
be the groupoids with binary operations a and ^ defined

3 3
1
12
1
a
c
2 3
1
a
5
ff
2 3
2
3
6
c c
a
b
Which
(a)
(6)
(c)
of the following
a,
mappings are homomorphisms?

(d)
(e)
(/)
^ ^
- 6,
1 - a, 2 1
a,
a,
-* 6,
^c 3 ^ a 3 ^ 6
3
1 -* 6, 2 1
^ ^
c,
^
^
6,
2 2
-^ 6,
3 -* 6 &
1 -> c,
o, 3
Solution
We
(o)
use a to indicate the mapping in each case.

lff2<r
ab = a and (1 2)ff = 2a = b. a is not a homomorphism. (6) la2a = aa = c and a. is not a homomorphism. (c) l(r2CT = ab = a and (1 2)ff = b. a is not a homomorphism. (d) luStr = be = a and (1 3)<t = 3a = c. a is not a homomorphism. (e) a is a homomorphism since the image of 1, 2 and 3 is b, so that (yV = b for any j G {1, 2, 3}, and iaja = bb = b. = a. a is not a homomorphism. 2)a = (1 if) lo2a = ca = c and
(1
2)a
2a
<r
i,
2(i-
2.35.
Let
of
(i)
G be the semigroup of positive G into G are homomorphisms:

<r
:
integers under the usual addition.
Determine which mappings
2n -t-
1,
(ii)
a: n-* 2v?,
(iii)
a:
- 1.
Solution:
(i)
(ii)
(iii)
+ m2)(T = 2(wi + W2) + 1 and Wja + nâ = 2wi + 1 + 2n2 + 1 = 2(mi + rig) + 2. Hence + ti2)<' ^ Wi<7 + ii2<r, and so a is not a homomorphism. (Wi + re2)a = 2(Wi + 712)^ = 2Wi + 2w2 + 4tciW2. TOja + n^ = 2wf + 22Then nâ + ^2(7 ^ (wj + n2)<'> and so a is not a homomorphism. nâ + n^ = 1 + 1 = 2. Thus TOja + nja y= (ii + n2)<T and hence a is not a (ii + TC2)<' =1.
(%!
(tox
homomorphism.
42
GROUPOIDS
Epimorphism, monomorphism, and isomorphism Three special types of homomorphism arise naturally.
[CHAP.
b.
1.
2. 3.
A homomorphism of groupoid G into groupoid H may be an onto mapping. A homomorphism of a groupoid G into a groupoid H may be a one-to-one mapping. A homomorphism of a groupoid G into a groupoid H may be both onto and one-to-one.
We
give these three types of
homomorphisms
special names.
Definition:
Let
1.
be a homomorphism of a groupoid
called
into a groupoid
H.
Then
6 is
H,
2.
i.e.
an epimorphism if Ge = H. (See Section

a
maps
the carrier of
onto the carrier of
1.3a,
page
12, for the definition of Ge.)
9 is called
monomorphism
H.
if ^ is
a one-to-one
mapping
of the carrier of
G
3.
into the carrier of

is
called a
isomorphism
if 6 is
both an epimorphism and a monomorphism,
i.e. 9 is
an isomorphism from groupoid G onto the groupoid H, then we say are isomorphic, or G is isomorphic to H, and write G = H.
If there is
G and
Problems
2.36.
the even integers with Let G be the groupoid of integers with addition as binary operation, and addition as the binary operation. Let t for n an even integer be the homomorphism (Problem 2.30) defined by gr^ = ng, for g G G. When is t an isomorphism, monomorphism or epimorphism?
Solution:
If
one, so
it is
g
is
2/n
T is one-to-one since gr^ = g'T implies ng = ng' and so g = g'. tq is not one-tosuch that gr^ = ng = 2. Then not a monomorphism. If t is onto, there exists g and n = 2. Hence t2 are the only epimorphisms. Thus t2 are isomorphisms and t
0,
&G
a monomorphism
when n
= Q.
2.37.
and
are finite groupoids and
\G\ ' \H\.
Show
that
cannot be isomorphic to H.
Solution :
Let
.
e:G^H
.
9i6, gê, Hence \H\
.,
gyfi
be an isomorphism and let gi,...,gn ^^ the (distinct) elements of G. Then (since is onto). are distinct (since 9 is one-to-one) and are all the elements of n, which contradicts \G\ # 1H|. Thus there exists no isomorphism e: G-^ H.
2.38.
Prove that if G, ff, and H = K, then

Solution:
(i)
if are groupoids, then:
(i)
G=
is
G;
G^
K. In other words " = "
(ii) if G = H, then an equivalence relation.
G;
(iii)
if
G =
(Hard.)
Let
r.
G^G
it is
be the mapping defined by
gi.
g for
all
fir
G.
is
a one-to-one epimorphism.
Hence
(ii)
an isomorphism, and so
G =
G.
ji
:
Let h e H.
a-G^H
As
yS
a is
be an isomorphism. Then we define a mapping one-to-one and onto, there exists a unique g G
h/Sa
H -* G
as follows:
Let
such that
ga
=
is
h.
Put
hp
g.
Note that
is
h.
Now
one; for if
and hp = g onto G, for if g G G, ga = hG hi/3 = h^P, then Aj/Ja = /i2/8a and so h^ = /i2/3
by
definition.
Also
j8
one-to-
Finally
is
hiP
(iii)
-* K be isomorphisms. Let y = ayS. We shall prove that y is an Let fi isomorphism. First, 7 G ^ K. Secondly, y is onto; because if k e. K, there exists hG such that hp = k, and there exists g G G such that ga = h, so gia/}) = hp = k. Next y is one-to-one; for if g^y - giy, {g\a)P = (fi'2)y8, and as p is one-to-one, gâ = fir2. which Finally, we must show that y is a homomorphism. implies, since a is one-to-one, g^ = '2= ((fl'ifl'2))y8 = (S'ifl'2)i8 = {Sia)p{g2a)p = gi(ap)g2{ap) = gagij. Hence G s K. (.9if2)y
: :
= 9u ^2^ = 92a: G ^ H and
a homomorphism. Let /ij, /ij S H. Suppose fifja = hi, g2a = /i2- Then Note that (firiff2)a = Qragâ = /ii?i2- Hence (h^h^fi = ^1^2 = KPh2P-
Sec. 2.5]

Is the
43
2.39.
(a)
of nonzero real
(6)
semigroup of integers under the usual addition isomorphic to the multiplicative groupoid numbers? {Hint: The integers and the reals are not equipotent.)
Is the
groupoid of nonzero rational numbers under division isomorphic to the groupoid of nonzero rational numbers under multiplication?
Is the
(c)
semigroup of integers under the usual addition isomorphic to the semigroup of rational
e
numbers under the usual addition? (Hint: Show that under any homomorphism integers under addition into the rationals under addition, re = r(l9).)
Solution:
(a)
of the
No. Because if the integers were isomorphic to the reals, the isomorphism between them would constitute a matching and hence the reals and the integers would be equipotent. Let a (Q*, -^) -> (Q*, ) be a homomorphism between the nonzero rationals under division and the nonzero rationals under multiplication. Now 1 = 1 -M. Thus la = (1 ^ l)a = la la and so la = 1. Now 1 = la = (2 -V 2)a = 2a2a and so 2a = 1. Similarly, 3a = 1. Hence a is not one-to-one. In particular, a is not an isomorphism. Thus there is no isomorphism between the two groupoids.
:
(6)
(c)
Let e {Z, +) -^ (Q, -I-) be any homomorphism. Let le = q. We shall show by induction on r that re = rq for all r G N. Now q = le = (0 + l)e = Qe + le = Os + q and so Oe - 0. Suppose re = rq for r = n. Consider r = n + 1. {n+ l)e = ne + le = nq + q = {n + l)q. Hence re = rq for all r G N, by induction. If r G Z and r is negative, then r e N. Then since = (r + -r)e = re + {-r)e = re + (-r)q, re = rq. Hence re = rq for all r G Z.
:
not onto. If g = m/n, with m, n integers and m, w t^ 0, then l/2n G Q. If r were an integer such that re = l/2n, then rq - l/2n and so r = l/2m. But 1/2'm is not an integer. Thus e is not an epimorphism and there is no isomorphism.
If
0,
ff
is
Has l/2n a pre-image?
2.40.
Let
{Z, )
be the groupoid with binary operation
defined
by
ab =
and
Is
let (Z, *)
+ +
ab
be the groupoid with binary operation * defined by
(Z,o)
a*b =
(Z,*)?
ab
Solution:
(a
Let a (Z, o) + b + ab)a (o

:
(Z, *)
-l-
-1-
a6)
be defined by aa and aa * fea

a
is onto,
= a for aG Z. a = a * 5 = (a) +
is clearly
(6)
Hence
aa
<r
is
a homomorphism.
for if
ba
implies
a = b and
Is Mjj,^}
b.
o G (Z, *), then a G Therefore a is one-to-one and
a mapping, (a o b)a ( a)( 6) = (a + 6 -f- ah). (Z, ) and ( a)(T = a. Now
(Z, )
(Z, *).
2.4L
Is
Mj,j^M(j_,j?
^Mj,_2_3j
Solution:
l^wl =
and
\M {1,2}
~^
j;
(^
Problem
2.21,
page
37).
Therefore, by Problem 2.37, M^^y
cannot be isomorphic to Mjj

inverses has 6 elements.
From Problem
|Af.j
2}
2.25, the subset of
elements in M^j
,
2,3)
which have
is
Thus
is less
morphic to M,j
then the order of M,{1.2,3}
and so
Af. {1,2}
not
iso-
^ ^
2.42.
Give an example of two groupoids of order two which are not isomorphic.
Solution:
a a a
c
c c
d d
c
baa
e
:
These two groupoids are not isomorphic, since there are only two one-to-one mappings, namely -> c, b -> d and a ^ d, b -^ c. Now e is not an isomorphism, for (ab)e = ae = c while aebe = cd = d. is not an isomorphism, for (ao)^ = a>// d while aâxi^ = dd = c. a
\//
:
vJ-
2.43.
Prove that the mapping e a + ib numbers under addition with itself.

:
-^
ib
is
an isomorphism from the groupoid
of complex
44
GROUPOIDS
Solution
e is
[CHAP.
onto; for
if
(!
-I-
ib{)e
(ttj
ib2)e
C, x = a + ib. implies a^ ibi
Now
=
a^
(a
i{b))e
ib
=
ttj
x.
Also
6i
fl
is
62-
one-to-one, since
tfig
and hence a^
and
=
(a2
Finally e
is [(ai
a homomorphism, for
+ i6i) +
to the
(02
162)] 9
(ffli
+ 02) -
tCfti
62)
(!
+ i6i)9 +
+ ^^2)*
2.44.
Is the
homomorphism
e of the
complex numbers
groupoid C of complex numbers under the usual multiplication of groupoid of real numbers under the usual multiplication defined by
9
:
ib -^
]a,
ibj
= + y/a?- +
b^
an epimorphism or monomorphism?
Solution:
It is well
known
that
if
xîX^ are two complex numbers, then
|a;ia;2l
l^î!
\^2\-
The
calcula-
tions are
|ai
+ i6i||a2 + i62l =
V(a?
6?)(a|
6|)
V'(aia2
[aiOa
- 6162)^ +
6162
(6ia2
^2^1)'^
= =
Then
exists
+
+
(6i2+
i62)|
62i)l
](ai-|-ibi)(a2
(x^xê
Xytxê
and so
is
so 9 is not one-to-one.
no x such that
Finally e xe = 1.
is
a homomorphism. On the other hand (a -1- ib)e = (a ib)e, not onto, for it is always the case that \x\ 0, and thus there
2.45.
Let
oid
e
:
) be the groupoid P under the usual multiplication of positive integers and {R, +) the groupunder the usual addition of real numbers. Is the mapping B of (P, ) into {R, ) defined by a - login o- &!> epimorphism, monomorphism or isomorphism?
(P,
Solution
homomorphism. If aa = be, then logjo a = logio b and hence a = b. = logjo 1 < logio 2 < logjo 3 < a monomorphism. Since there is no integer such that logjo X = 1. Hence e is not onto. Therefore 9 is a monomorphism but not an epimorphism nor an isomorphism.
As
in Section 2.5a, * is a
Thus
e is
c.
Properties of epimorphisms
We
will
show
in this section that if 6 is
an epimorphism from the groupoid
G
H.
to the
groupoid H, then
shares some of the properties of G.

6
if
Theorem
2.6:
Let
(a)
be an epimorphism from the groupoid
G
is
to the groupoid
Then
is
a groupoid with an identity

if
1,
so
and 1^
is
H.
(b)
(c)
Furthermore ge in H.
if
is
an inverse of g
in G, then fe is
the identity of an inverse of
if
G is commutative, so is H. G is a semigroup, so is H.
i.e.
fl'
Proof:
(a)
Let hGH. We shall prove 16 is the identity of H, an epimorphism, 6 is onto and we can find an element
h-19
in
h-ie = ge-ie = (g-i)e = ge
G such = h
19-h. As ^ is that gO = h. Then
and
19 -h
IS
is
16 'g9
=
/.
{l'g)9
ge
= h
identity.
1
Thus
the identity of
H and H is a groupoid with an

Then gf =
{fg)e
Now
suppose g
GG
has an inverse
fg.
Therefore
ge-fe
{gf)e
^19 =
H.
f9-ge
which means
f9 is the inverse of ge in
Sec. 2.5]

Suppose Because
45
(6)
is
6 is onto,
commutative. We show H is commutative. we can find g,g' G G for which gO

hh'
To
this
end
let
h, h'
h'.
G H.
ge-g'd
{g'g')e
{g'
g)e
h and g'6 = = g'O-gd = h'h
Hence
and so
(c)
To
Let g"e
H is commutative as claimed. show H is a semigroup, we must prove

h, h',
h" G H.
Since
Then we can
is
find
h".
associative
we
that multiplication is associative in H. G such that ge - h, g'9 = h' and have, as required,
g, g',
g" G
{hh')h"
= =
{ge'g'6)g"e
{9{g'9"))e
[{gg')9]g"e
{{gg')g")e
fl'e[(fl''fl'")e]
ge{g'e-g"9)
h{h'h")
d.
Naming and isomorphisms

In our study of groupoids we will take isomorphic groupoids to be essentially the same. will describe a "naming process," beginning with an example.
To explain why, we
Let
G be
the groupoid with binary operation
elements
and
a,
and multiplication table
a
1
We
a,
define a new groupoid G by relabeling the elements of G. p and have multiplication table
a
Let
consist of the elements
/3
What we have done
is to call
the elements of
G by
different names.
In general if G is any groupoid, we can form a new groupoid G by renaming the elements of G. Thus for each g G G we take a new element g, ensuring only that fy^g if f = g, i.e. don't use the_ same name twice. If fg = h, then we define multiplication of elements of Ghy fog -h. It is easy to prove that G is a groupoid with this multiplication.
are not interested in distinguishing between groupoids which differ only because their elements have different names. Considering groupoids to be the same if they are
We
isomorphic overcomes this snag. To see this we will show that the G constructed from G above by renaming is isomorphic with G. We must find a one-to-one onto homomorphism e. Define gO = g, i.e. the image of g under is the new name of g. 9 is one-to-one onto, as one and only one g corresponds to each g. Also {fg)e - h where fg = h. But fdoge =
fog =
h,
by
definition of the multiplication of G.
Hence
{fg)0
feoge,
and
is
iso-
morphic with G. Thus isomorphism gets rid of the on simply renaming.
difficulty of
obtaining a
new groupoid
We look at the problem from another point of view. Suppose F and G are two isomorphic groupoids, and that 9 is an isomorphism between F and G. Then we will apply our renaming process to show that G and F, F suitably renamed, cannot be distinguished
either as regards their elements or the
way they
multiply.
Let us as before renarne each element f GF, f. But we shall choose / to be f9. This is a proper renaming since f = g means that f9 = g9, and, as 9 is one-to-one, f = g. So we have not used the same name twice. As before, if fg = h we define / o g to be Ti. Thus F becomes a groupoid with respect to the binary operation o
46
GROUPOIDS
[CHAP.
compare with G? F has the same elements as G. But do these elements multiply the same way? Suppose /i and /a are two elements of F. Then /i was previously called /i, and fi was called /2. If /1/2 = /a, then we defined /i o /a = /g.
does
How
Now
of
/i,/2
are also elements of G; in fact,
therefore fiOfzO
=
in
(f if 2)0
faO
/a,
since 6
is
The product /1/2 in G is a homomorphism. Hence the product /10/2

fiO, /2
A=
/2^.
two elements
F is
the
same element as the product
/1/2 inside G.
Thus a groupoid isomorphic with a groupoid F is indistinguishable from a suitable renaming of F as far as the elements and the way they multiply are concerned. For this reason we do not distinguish between groupoids that are isomorphic.
6.
Mx and
semigroups
iso-
The importance of Mx is explained by the following theorem, which says that an morphic copy of any semigroup S is contained in some Mx.
Theorem
2.7:
Let S be a semigroup with identity. Then there is into M^. (The semigroup <S is an abbreviation for the semigroup (S, /x) where û is a binary operation. Mg is the semigroup of all mappings of the set S into itself, with binary operation the composition of mappings.)
(Cayley's Theorem):
monomorphism
of
Proof: Let s G S and let p/. product of X and s in S, i.e. {x, s)[i.
S^ S
It is clear
be defined by xp^ = xs, x that p^ is a mapping of

^ is
G
S
S.
Here xs
i.e.
is
the
into S,
p^
e M^.
First
Let
S-^
Mg
we have
to check that
be defined by s9 = p^. We shall show that is a homomorphism. Let s, s' G S.

{ss')9
a monomorphism.
p,,.
and
s9s'9
=
As
p^p,.
Now
P^P^,.
if
S, xp^p^,
{ss')9
(aspjpj,
{xs)p^.
(xs)s'
Xp^^,.
this is true for all
S,
p,,,
Hence
s9s'9.
Secondly
we must show
is
that 9
is
one-to-one.
Suppose
Ip^
s9
=
'
s'9;
then
s
p^
p^,
particular, if 1
the identity of S,
^ is
Ip^
Ip^..
But then
= p,.. In = 1 s' - s'
and so
s'
and
a monomorphism.
let
This completes the proof of Theorem

(oi, a^, as}
2.7.
As an illustration of the proof, ing multiplication table:
S=
be the semigroup given by the followas 3

Oj
0-2
Oi
ai
0-2
a^
ttj
aa
ttg
as
(Xi
ag
Notice that ai
is
the identity element of

Ui
tti
S and
tti
that
a2
is
a semigroup.
tti
Now
a2
ai
a2
Ua
aa
ai
9a^
as a2
a2
as a^9
P'^2
"2^
0,2
aa
as
The mapping 9 is defined by ^ is a monomorphism.
Paj'
Paj'
'^3'^
P03
It
can be checked directly that
CHAP.
2]
47
Let Se = {se\ s G S} where is the monomorphism of Theorem 2.7; then S ^ S0 and hence Ms contains an isomorphic copy of S. In Section 2.5d we pointed out that, but for naming, S and an isomorphic copy were the same. Thus we see that in a rough way every semigroup with identity appears in some Mx. Hence the importance of Mx.
We We We
defined a groupoid, associative groupoid (called a semigroup),
and commutative
groupoid.
showed that
defined a
in a
groupoid the identity
is
unique, while inverses are unique in a
semigroup.
be a homomorphism if [(gi, g2)a]e = it an isomorphism of (G,a) onto (G,/3). We proved Cayley's theorem, that each semigroup has an isomorphic copy in Mx for some suitable X.
mapping
{G, a) -* {G, p)
to
[(910,926)]^.
If e is one-to-one
and
onto,
we
called
2.46.
GROUPOIDS Let G =
Show
{1,
1}.
Is (G,
a groupoid
if
is
the usual multiplication of integers?
2.47.
that (G,
is
a groupoid when
G =
{l,-l,i,i},
- y/^,
and
is
the usual multiplication
of complex numbers.
2.48.
Suppose G is the set of all integers divisible by the integer n. multiplication of integers, a groupoid?
For which n
is ((?,
).
with the usual
2.49.
Let
/i, i
1, 2, 3, 4, 5, 6,
be the set
of
mappings of
R- {0, 1} into R defined for each
a;
G fi -
{0, 1}
by fi-.x-^x;
{fofj}
(i)
fi-
^ rr^; fs'-x^^; f^:x-^-; f^-.x-^-^; f^-.x^l-x.

the composition of mappings.
(iii)(Gi,4,),
Suppose Gy
and that
o is
Determine which of the following are groupoids:

(v)(Gi,6,o),
(vi) (G5,6,.).
(Gi.2,),
(ii)(Gi,3,),
(iv)(Gi,5,),
2.50.
Let
F=
(F,
{fvfz'fa}
o) is
Prove
a"^ = {/i.^.A./e) where /; are the mappings defined in Problem 2.49. a groupoid while {H, o) is not a groupoid (o is the composition of mappings).

2.51.
Let R* be the set of nonzero real numbers. Define the binary operation o on K* by aob = \a\ b for a,b G R*. Prove (R*, o) is an associative groupoid but not a commutative groupoid. Hint:
\a\ \b\
\ab\.
2.52.
tive or
Define the binary operation -on an associative groupoid?
G =
RXR
as
(a, 6)
(c,
d)
(oc, be
+ d).
Is
(G,
a commuta-
48
GROUPOIDS
The binary operation a on
[CHAP.
2.53.
is
defined by
: {a,b)-*\a
b\
for
a,h
&
R.
Show that
(R,a)
is
commutative but not associative.

Suppose we define a binary operation
2.54.
on
fi
by a
the
minimum
of a and &
(a, 6
R).
Show
that (R, +) is both associative and commutative.
2.55.
Let G = {a a: R-^ R}. For a, p ^ G define the mapping a* fi = a ji fia where is the usual composition of mappings and x{a 13 /} a) = x{a P) ~ x{/3 a) for all x G R. Prove: (iii) {a* fi) *a = a* {/} * a) (ii) (G, *) is neither associative nor commutative; (i) (G, *) is a groupoid; for all a,/3 e G; (iv) a * /3 = ( /3) * a where /3 is the element of G defined by p-.x-^ (xp) for all xGR. (Hard.)
I
2.56.
Let
be the set in Problem 2.55.
For
a,
p &
define
a' p
= P
,
composition of
is
mappmgs and x
(ii)
a P
+ Pa ~ o
x(a ~^
P) + ^~2
x(p
a)
P where is x e R. td ^ d Prove:
(a
the usual
/\
(i)
try
(G,
a groupoid;
(G,
is
commutative but not associative;
(iii)
^)
(/3
a)
for
all
a,PGG.
INVERSES IN GROUPOIDS
2.57.
Let
B+
be the set of
is
all
non-negative real numbers.
Define
a*6 =
\/a2
+P
for
all
a,
6eK +
iVoF+b^
2.58.
the positive square root).
Find an identity
in (i2 + ,*).
What
elements have inverses?
where all a, p & G = {a\ a: Z ^ Z}, let a X ^ be the mapping defined by x{a X p) = xa'xp G Z and is the usual multiplication of integers. Is (G, X) a groupoid? Does it have an identity? What elements have inverses?
For
2.59.
Let
G =
{a
I
a:
Z^
Q,
(G, X)
have an identity?
What
the rational numbers}. Define a elements have inverses?
,8
as in the preceding problem.
Does
2.60.
Which
of the groupoids in Problems 2.57, 2.58
and 2.59 are commutative and which are associative?
2.61.
of a
Define the following binary operation + in i2 + the non-negative real numbers: a -I- 6 = the + Does (i2 + ,-|-) have an identity? What elements have inverses? and 6, a,
,
maximum
6GB
2.62.
Let (G,
*)
be the groupoid of Problem 2.56.
What
is
the identity of (G,
*) ?
Find an
infinite
number
of elements which
have inverses.
2.63.
Show
that (G,
*),
the groupoid of Problem 2.55 has no identity.
SEMIGROUPS WITH AN IDENTITY

2.64.
Which elements
of (G,
),
where
G=
{l,-l,i, -i),
- V-l,
and
the usual multiplication of com-
plex numbers, have inverses?
2.65.
Let G = {/i./z./s.A./s./e} of Problem 2.49. Prove that G with the binary operation of composition of mappings, is a semigroup with an identity. Find the inverse of each element in (G,).
2.66.
Let
G =
{(a, 6, c, d)
a,b,c,d. Z}.
Define
(a, b, c, d)
(a', b', c', d')
(aa'
+ cb',
ba'
+ db',
ae'
cd' , be'
dd')
and
- {(\, 0, 0, 1), (-1, 0, 0, -1)} a semigroup with an identity. Show that the subsets and F respectively, c G Z}, with the binary operation of (G, ) restricted to and F. are semigroups with an identity. Find the inverses of the elements of
Show
that (G,
)
is
FG=
{(1, 0, c, 1)
2.67.
Let
{a\ a
mapping
position of mappings.
Is (0,)
of {1,2,3,4,5} into {1,2}}. For a semigroup with an identity?
a.peG,
let
a p be the usual com-
CHAP.
2]
:
49
2.68.
For a, p e G = {a\ a {1, 2, 3,4, 5} ^ {1, -1}}, define the mapping a X ^8 by n(a X p) = na "np Is {G, X) a semigroup with an where n S {1,2,3,4,5} and the usual multiplication of integers. identity? If so, what elements have inverses?
2.69.
HOMOMORPHISMS OF GROUPOIDS AND CAYLEY'S THEOREM Let H = {(1, 0, 0, 1), (-1, 0, 0, -1)} be the groupoid defined isomorphic to the groupoid (G, where G = {1, 1} and is
)
is Problem 2.66. Show that the usual multiplication of integers.
in
2.70.
U F = {(1, 0, c, 1) c e Z} is the groupoid defined in Problem 2.66 and under addition, prove the two groupoids are isomorphic.
I
{Z,
+) the groupoid of integers
2.71.
Which
Let
of the groupoids of Problem 2.49 are isomorphic?
2.72.
= {1,-1} {/i, A, /a. /4, /s, /e) be the groupoid of Problem 2.65, (H,') the groupoid with the usual multiplication of integers. Find all possible homomorphisms of G into H. Show that there is no homomorphism of (G, o) into (F, o) where F = {/j, /a, f^} and the usual composiand
G=
tion of mappings.
(Hard.)
2.73.
Can the groupoid

is
(G, *) of Problem 2.55 be a homomorphic image of the groupoid (Mr, the usual composition of mappings?
where
2.74.
Suppose G = {a\ a: Z Z}, X the binary operation defined in Problem 2.58 and the usual composition of mappings. Show that * (G, X) -> (G, ) defined by *:-* is not a homomorphism. (See Theorem 2.6, page 44.)
:
chapter 3
Groups and Subgroups

We define a group as a semigroup with an identity in which every element has an inverse.
The
object of this chapter is to show that the concept of a group is natural. This is done by providing illustrations of groups which arise in various branches of mathematics. The most important concepts of this chapter are group and subgroup.
3.1
GROUPS
As we remarked
in the preview, a
Definition
semigroup with an identity in which every element

repeat the definition in more detail:
has an inverse
Definition:
is
termed a group.
set
iS
We
A
(i)
non-empty
words,
together with a binary operation in
S
c.
is called
a group
if
there exists an identity element (usually denoted by)

.
S;
in other
1
is
= a =
j;
for
all
^ aGS
2.1,
Recall that the identity

(ii)
unique by Theorem
page
31.
for every choice of the elements a,b,c

{a'b)
'
E
a
S,
(6
c)
)
Thus
(i)
and
(ii)
are the conditions for
(S,
to be a
semigroup with
identity.
(iii)
every element such that
aGS
is
has an inverse in S,
i.e.
there
is
an element b
GS
a'b =
I.
-,
= o*a
T
This element b
2.2,
often denoted by
a"^
The inverse
is
unique by Theorem
page
33.
Whenever we
I.
define a group
we
shall follow the pattern:
Define a set
(?^
P).
II.
Define a binary operation in S.
III.
Verify that the groupoid Verify that the groupoid
(-S, )
contains an identity element.

is
IV.
(S,
a semigroup,
i.e. is
associative.
V.
Verify that every element of S has an inverse.
The number
page
29.)
We
of elements of S, \S\, is called the order of the group. (Compare Section 2.2, will exhibit groups of infinite and finite order. (See following examples and
problems.)
50
Sec. 3.1]
GROUPS
of
51
Examples
groups of numbers
1:
Example
The additive group of integers.

I.
Let Let
Z +
be the set of integers.
II.
be the binary operation of addition in /.
III.
M+
If
I,
=n=
+n
for every
n. Z.
Thus
(Z,
+) has an identity element.
IV.
m, n are integers,
{l
+ vi) + n =
+ {m + n)
i.e.
{Z,
+)
is
a semigroup.
V.
If
nG
Z, then
in
has the property
n + ( n) =
i.e.
( n)
is
an inverse of n
in {Z, +).
is
Thus we have shown that the groupoid (Z, +) referred to as the additive group of integers.
Example
2:
a group.
This group
is
usually
The additive group of

I.
rationale.
Let Let
be the set of rational numbers.
II.
be the binary operation of addition in Q.
III.
o+0 =
If
If
<i
for every a
(a in
&
Q, so
is
an identity element for
(Q, +).
IV.
a,b,cG Q, then
a
+ Q
b)
(b
+ c).
V.
Q, then
has the property a
{a)
( o)
a.
Example
3:
The additive group of complex numbers.
The description of
Example
4:
this
group
is left to
the reader.
The multiplicative group of nonzero rationale.

I.
Let Q* be the set of nonzero rational numbers. Let
II.
be the binary operation of multiplication,
i.e.
the usual multiplication of
rational numbers.
III.
The rational number

If
o, 6, c
1 is clearly
an identity
in the
groupoid (Q*,
).
IV.
Q*, then
(a
6)
=
1
(6
c)
V.
If
Q*, so
is
1/a and
a Thus every element of Q* has an inverse.
'
= 'a a
Example
5:
The multiplicative group of nonzero complex numbers. This group is very similar to that in Example 4. We five stages in setting up and describing the group.
I.
shall
go through the usual
Let C* be the set of
all
nonzero complex numbers.
Thus
and a,b R}
C*
Recall that
II.
i^
{x\ X
+ ib
where x
j^
+ iO
-1.
We
This
define multiplication of complex

(a
is
numbers as
(ac
follows:
+ ib){c + id) =
ment
III.
in
a binary operation in C* since C* (not both ac bd and ad + be can be
bd) + i(ad + 6c) (ac 6d) + i(ad + 6c)

zero).
).
is
a unique ele-
i*0
G C*
and
it is
clearly an identity in (C*,
52

IV.
[CHAP.
Suppose
+ ib, c + id, e + if G C*. Then [{a + ib){c + id)](e + if) = [{ac - bd) + i{bc + ad)]{e + if) = [{ac - bd)e - (be + ad)/] + i[(bc + ad)e +
a
the other hand,
(a
{ac
bd)f]
On
+ ib)[{c + id){e + if)] = = + ib)[{c +
(a
+ ib)[{ce - df) + i{de + cf)] [a{ce - df) - b{de + cf)] + i[b{ce - df) +
+ if)] =
[{a
a(de
+ cf)]
It follows
from these two computations that

{a
id){e
+ ib){c + id)]{e + if)

Thus suppose a + and so
ib
V.
We
have to check the existence of inverses. not both a and 6 are zero. Hence a^ + 6^ ^
C*;
then
e
Moreover,
"'
C"
62
'2
"
\{a 62y V-
+ ib) ->
'
= -
1 -
= -
(a VU-
+ ib) ->
1
^^^^+1:2
'(,2+62
Thus we have proved (C*, ) is a group and we term group of nonzero complex numbers.
this
group the multiplicative
Problems
3.1.
Is (S,) a
(i)
group
if
(ii)
(iii)
SZ S Q S = {g
S =
{
j
and and
I
is the is the
usual multiplication of integers?

usual multiplication in
q
z
Q?
(iv)
G Q and zG Z and
q
>
0}
and
is
the usual multiplication of rational numbers?
yf2}
and
o is
the usual multiplication
Z1
(v)
(vi)
S R S=Z
The
and
and
is
the usual addition of real numbers?
o is defined
by a
for
all a, b in Z"!
Solutions
(i)
identity element is the integer integer z in Z such that z5 = 5z
1.
(S,. )
is
not a group because
GZ
1.
but there
is
no
1.
(ii)
Again the
{S,
)
identity
is
the
number
1.
There
is
no
GQ
such that
q0 =
Hence
(S,) is
not a group.
(iii)
is
S;
a group. Clearly S # hence 1 is an identity.
and is a binary operation onS. Multiplication of rational numbers

if
q'l = ! q q for
is
all
associative
q
and every
element in
(iv)
has an inverse; for

\/2
S,
then
G S and
9 q
(/)
since
g
S
2.
'
Therefore
(S, o) is
is
not a group.
(v)
(S, ) is
and r
(vi)
a group. ( r) =
and addition
an associative binary operation onS. r
( r)
+r
for all
is
G S.
(S, ) is
not a group because there
no identity element in S.
3.2.
Let
be the set of even integers.
Show
that
is
a group under addition of integers.
Solution:
Let
2(ii
and
26i
in S; thus addition is a binary operation on S.
sociativity of addition in Z.
2ai
implies
a 2( a^).
be any two elements in S. a + b = 2(ai + 61) is a unique element Associativity of addition in S follows from the as= 2-0 is an identity element in S. If a G S, then a S since Hence a has an inverse in S, as o + (a) {a) + a = 0.
3.3.
Let
zero.
be the set of real numbers of the form a
Show
that
S becomes
+ 6\/2 where a,b G Q and are not simultaneously a group under the usual multiplication of real numbers.
Sec. 3.1]
GROUPS
53
Solution:
(0
= {ae + 2bd) + (cb + ad)\f2 If neither a + 6^2 nor c + rfV2 is zero, i.e. (a + 6V2 )(c + dV2 + 0\/2), then their product cannot be zero. Hence the product of elements in S belongs to S.
)
.
^ + 0v2
is
an identity for
1
S.
Multiplying
laby
by/2.
6-v/2
,
(2
we
obtain
6\/2
6
'^-bV^
"^
a
^2
a+
Hence
bV2
262
262
_ 262)
.^
S.
The associativity holds because
it is
a+
3.4.
true for multiplication of real numbers.
byjt
Let S be the set of complex numbers of the form o + 6\/ 5 where a,b & Q and are not both simultaneously zero. Show that S becomes a group under the usual multiplication of complex numbers.
Solution:
{a
b^/E){c
+ d^/^) =
{ac
5bd) +
+
{be
+ ad) \/ 5
.
and cannot be zero

1
if
its
factors
are
S.
not zero; hence the product of two elements of

is
belongs to S.
+ 0^/^
is
an identity for
6/^
certainly an inverse for o
b^/5
Multiplying top and bottom by a
6\/ 5
we
get
a,
f,^fli
a2
6v + 562
^2
552
^
S
^2
5^2
and
so
a+ bV^
S.
The
associativity of multiplication in
follows
from
associativity of multi-
plication of complex numbers.
3.5.
Let
be any fixed positive integer and

a.o6
let
S =
{0, 1, 2,
if
. ,
to
1).
Define a binary operation in
S by
= a+6
if
a+ b < m
ab
Prove that
Solution
If
a, 6
a+b = m + r,
r<m
(S, ) is
a group of order m.
(Hard.)
S,
then a
o 6 is
Note that a6
5i is
or
1.
+ 6 Sm where 6 is or a (6 o c) = o + (6 o c) - S^'m where =

a
uniquely defined and belongs to yS. a,b 1, for any

82 is
or
1.
a0 = 0a = a, so is an identity. G S. So 6oc = 6 + c Sitn where Then a (b c) = a+ b + e {Si+ b^m

i;,
where both
Sj
and
Sg could be
or
1.
Hence
a
Similarly
(b c)
o c
a a
+ +
b
b b
+ +
c
c
c
tjjWi
where where
is
or 1 or 2 or 1 or 2
(aob)
v^m
V2 is
Now
because
a.o(6oc)<m and
a
77J
{aob)
a
< m.
Suppose vi>-n2' then

a
(bc)
7)2
is
at least
1.
a{bc) <
0;
this contradicts
contradiction.
Thus
(m
vi
=
0;
(tu
a) =
a) a
vi and the above equation implies that a(bc). Hence vi V2- V2 > Vi leads in a similar way to a and a{bc) = {ab)c. If a& S, then m ~ a & S and V2 hence w a is an inverse to a. Thus S is a group.
+ + c {v2 + l)i = a+ b + c tj^m < But
vgm
3.6.
Let
qC
positive integer.
{C the set of complex numbers) be the set of all roth roots of unity, where is a fixed Prove that under the usual multiplication of complex numbers, S becomes a group
of order m.
Solution:
Recall that a complex
number x
Since
is
an ith root of unity
m
1
distinct roots of unity, viz.

is
e*2'rT/m^ t
1,2, ...,m;
1, is
also,
is
then ab
uniquely defined.
1
a,
so 1 is S,
(a6) a^b'" = an identity. Associativity

(l/o.)
ab
= 1 and that there are exactly = cos + i sin . If a, 6 G S, an mth root of unity and hence ab G S.
if a;
e'
a;
true, since it is true for
complex numbers in
a.
general.
If
aG
then
1/a
1;
thus
1/a
G S and
1/a
is
the inverse of
54

Let
[CHAP.
3.7.
S qC
:
(C the set of complex numbers) be the set of

into a group.
all
roots of unity.
Describe one
way
of
making S
Solution
Use as binary operation the usual multiplication of complex numbers. If a, 6 G S and a is an mth root of unity, b an wth root of unity, then ab is an mwth root of unity because (a6)" = a'""6'"" = 1 G S and acts as an (a"*)" (6") = l"! = 1. Hence a6 G iS and is, of course, uniquely defined. identity. 1/a G S and is the inverse of a. Associativity holds for multiplication of complex numbers. Thus S is a group with respect to the usual multiplication of complex numbers.
3.8.
The following
table defines a binary operation.
Is the resultant
1
groupoid a group?
2 2
1
Solution:
We
(a)
need to check only

associativity,
(a) associativity,
(6)
existence of identity and inverse.
To check
(a)
(b)
(c)
we have
the following possible questions:
Does
(1
1)
equal
(l-l)-l?
(1
(e)
(/)
Does 2
Does 2 Does 2
(1 (1
1)
equal equal equal equal
(2-l)-l? (2'l)-2?
(2
Does 1-(1'2) equal
1)
2?
2)
2)
1)
Does l'(2'l) equal (1-2) -1?

Does l'(2'2) equal (l'2)-2?
all
(g) (h)
(2
(2
-2)
-2?
(d)
Does 2
(2-2) -1?
Checking
(6)
these products,
we
see that associativity holds.

1,
1 acts
as an identity. The
inverse of 1 is
the inverse of 2 is
2.
Hence the table
defines a group.
3.9.
Write the multiplication table for the group of Problem

Solution :
If
3.5
with
m=
and
m = 4.
m = 3,
the multiplication table
is
2
1
If
m ~ 4,
the multiplication table
is
1
1
2
3
2 3
2 3
3.2
SUBGROUPS
Let (G,
)
Definition
Then we say H is a subgroup of G H which makes H into a group.
and let fl^ be a non-empty subset of G. be a group with binary operation is a binary operation in restricted to if the operation
Sec. 3.2]
SUBGROUPS
55
is
For example, if a subgroup of G.

is
is
the group with

it is
m=4
o
of
Problem
3.9,
then the subset
H=
{0, 2}
For when the operation
in G, as defined in the multiplication table
2oO =
a binary operation in H, i.e. OoO = OeH, 02 = 2eif, = 0; 0, the identity, is in H; is & group because: the operation o restricted to is an associative binary operation (since the operation in G is associative); and every element in has an inverse in H.
for G,
restricted to H,
2GH,
and 2o2
= 0G/f.
The following lemma
facilitates
proving a subset of a group
is
a subgroup.
Lemma
3.1:
Let (G,
be a group.
(i)
Then a subset
and
(ii)
^ of G
if
is
a subgroup of
iff
H^^
H
a, b
G H,
then ab^^
Proof:
eration.
in
If
H satisfies these conditions,

if
jfif
For
then 16-1
5-1
then is a group with respect to the binary op0, then there exists a G H. Hence aa-' = 1 G H. Also, if b G Hence a,b G implies a(b"i)-i - ab G H. Associativity is true
H, as it is true in G. Thus is an associative binary operation on H, 1 G H, inverse of every element of is an element of H. Therefore {H, ) is a subgroup.
and the
(i)
Conversely
(ii)
if
is
a group with respect to
then clearly
satisfies conditions
and
above.
Problems
3.10.
Let (Q,+) be the additive group of rationals.

Solution
Is
a subgroup of
Q?
Is
a subgroup of
Q?
(Q, +} then the binary operation is +, and the inverse accordance with Lemma 3.1, whether a + (6) = a 6 G Z. It is, and so Z is a subgroup of {Q, +). Clearly P ^ and PqQ. If a,b e P, is a. + (-6) = a bGPI No, for P does not contain negative numbers; and if a = 1 and 6 = 2, then a 6 is
If
Clearly
is
Z^
0, and
a, 6
Z CQ.
a,be
of 6
6.
So, if
Z,
we ask
in
negative.
3.11.
Is
a subgroup of (C, +), the additive group of complex numbers?
Solution:
Q# Q
3.12.
is
0.
QqC;
for if
Q,
+ OiGC.
If
a, 6
Q,
then
a + (-6)
= a-bGQ.
Hence
a subgroup of (C, +).

{0} a
Is
Z
1
subgroup of (Q*,
),
the multiplicative group of nonzero rational numbers?
Solution:
is
the identity.
iQ*,').
GZ-
{0},
but 3 has no inverse in
Z- {0}.
Therefore
Z-
{0} is not a
subgroup of
3.13.
Is Q'*, as above, a
subgroup of (C*,
),
the multiplicative group of nonzero complex numbers?
Solution:
Q*
3.14.
0.
a, 6
G Q*
implies ab~^
is
a nonzero rational.
Thus Q*
is
a subgroup of (C*,
).
Is Q* a subgroup of the group of real numbers of the form a taneously zero, under multiplication?
byf2
a,b
& Q and
a, b
not simul-
Solution:
Q* = 0. Ii a 6 Q*, then a = a + 0^/2 e {a+ Thus Q* C {a+b\2 a, 6 G Q and not both 0}. Hence Q* is a subgroup.
|
by/2
a, 6
a,b
GQ
and not both
zero}, since
# 0.
G Q*
implies ab"^ is a nonzero rational.
3.15.
Prove that the intersection of two subgroups

Solution:
1
and
if of
a group
(3 is
a subgroup.
G H,
identity
g,heH
for as is not empty, there is an element h G H. of G. Similarly, 1 G K. Hence and and gh-^GH. Also, gh-^GK. Thus y/i-i G
IGHnK
But then H contains hh~^ = 1, the HnK ?^ 0. If g,hGHnK, then HnK and HnK is a subgroup of G.
56

By considering the group of Problem 3.5 with not necessarily a subgroup.
Solution:
[CHAP.
3.16.
m=
6,
show that the union
of
two subgroups
is
is
That {0,3} and {0,2,4} are subgroups is easily verified. But U = {0,2,3,4} not a subgroup as 3 4 = 1 g C/. Therefore o is not a binary operation in U.
{0,
3}u{0,2, 4}
3.3
THE SYMMETRIC AND ALTERNATING GROUPS The symmetric group on X a. Let X be any non-empty set. A very important group one-to-one mappings of X onto X, called the symmetric group
in the usual five steps.
I.
arises
from the
on X.
We
set Sx of all describe this group
Sx
the set of all matchings of the non-empty set (see Section 2.4b, page 36).
is
CT,
X with
itself.
Clearly,
Sx C
Mx
XL
and t. G Sx, then we define ,7t to be the composition of the mappings of X with itself. Suppose x G X; then Here we must verify that ar is a matching is also onto, so we can as T is onto, we can find x" G X such that x"t = x. But = x". Consequently find X' G X such that x'a
If
T
<r <t
X'(o-t)
(x'o-)t
X"t
(x(t)t
X
{y(j)T
and hence
<7t
is
onto.
If
xiar)
y{(TT),
then
by the definition of the

xa
is
composition of mappings; this means, since turn implies, since a is one-to-one, that x = v. Therefore Thus composition of mappings is a binary operation in Sx.
t is
one-to-one, that
<jr
ya.
This in
also one-to-one.
III.
Clearly the identity
mapping r.x^x
is
is
in
Sx and
is
an identity element of Sx.

is
IV.
The groupoid
(Theorem
2.3,
(Sx,
a semigroup, since composition of mappings
associative.
page
36.)
V.
2.4, page 36, implies a has an an inverse of t. By Theorem 2.4 inverse, t, in Mx. Now, (tt the only elements in Mx which have inverses are those which are one-to-one onto mappings. Therefore t G Sx and r is the required inverse of a. The proof that Sx is a group is complete.
Let a
Sx.
Since a
is
Theorem
<t
= =
t
t(t
means
is
We will
call
an element of Sx a permutation of X,
or, for short,
a permutation.
In the particular case where symmetric group of degree n.

\Sn\ is
X-
{1,2,
.,n],
we
write
Sx
Sn.
is
called the
can be one of n elements. 2o- can be calculated as follows. If <t G S, then chosen and a must be one-to-one; so 2cr ^ la. one of 1 elements, as la has been 3a can be one of n - 2 elements, as la and 2a have been chosen and a must be one-to-one; so 3a is not equal to 2a or la. Continuing in this way, we conclude that there are
n -
n-{n-l){n-2)
elements of
Sn,
i.e.
2-1
\Sn\
=n
The elements
of S3, for example, are
Sec. 3.3]
THE SYMMETRIC AND ALTERNATING GROUPS
57
12 12 12
2
3
table,
3
3
3
1
12 3 3 12 12 3 13 2
12
3 2
3
1
12 3 13
find the multiplication
a^r^.
Here we are using the notation of Section 2.4c, page 37. To we must compute the products. As an example, we calculate
12
l<TjTj
3
3orjTj
12
2Tj 3tj
Itj
3
1
2<r,Tj
To do
this calculation mentally,
we think
as follows:
1 -* 2 (in a,),
2-^3
3
(in r^)
Write down
12
3
2-^3
(in
(Tj),
3^2
(in t^).
Put 2 beneath 2 to get
12
3 2
1.
Only one other number can appear, namely
Thus
3
1
12
3 2
The multiplication
table for Ss is
t
"i
"1
<'2
Tl Tl
'2
'2
T3 ^3
"2
I
^2
'"3
"1 "2
Ti
"1
"2
i
'"I
"1
"1
n
i
^1
^2
"1
n
'"2
^3
Tl ^2
^2
T3 'l
"2
I
"1
"2
I
'3
'3
"2
"1
CTjTj
The reader should check some of the entries. = TjtTj. Hence S^ is not commutative.
Note that
<TjTj
r^
and
Tj(rj
Tg,
so that
Problems
3.17.
Calculate aP, Pa, a-i, p-i, iaP)-i, and (/3a)-l
if
12
2 3
3 6
4
5 3
and
4 4 2
1
Solution
aP
12
3 5
x(aa~'^)
5 6
Pa
1
12 3 4 5 6 13 5 6 2 4 12 3 4 5 6 2 6 4 13 5
Now we la = 2,
determine 2a~^ = 1.
To
find a~i,
we
note that
which X is taken onto 1. 6a = 1, Proceeding in this way, we obtain
so
we must have
'1
and hence a~i must carry xa to . la-i = 6. Next, since

2
3
6N
.612543/
58

An
easy method of calculating the answer mechanically follows.
[CHAP.
Take
12
2 3
3
6
4
5
6
1
Interchange the rows,

2
3
6
3
5 4
4
5
12
Rearrange the columns so that the top row reads
12
3
3 4 5
6,
'12
This method
is
,612543^
conceptually the same as the
first.
To
6
find P"'^, interchange the
rows
to obtain
/I
3 2
5 3
2 5
4 6
\l
and rearrange
to get
^"'
/12345
- (l
5 2 6 3
6\
4
/12345
= (e
4
1
6\
5
,,,
^^"^
(^^"'
=415362
/I
23456
3.18.
Verify that
a{l3y)
{aP)y
where
4 4 5 5
12 3 12 3
Solution:
12
3 2
3
5 4
12
4
3
15
3
15
4
2 4 2 5 5
M)y
12 13
3
4
5
12
4
3
15
3
12 3 13
(i8y)
12 3 12
5 5
12 13
4
2
5 5
12 3 13 4
5 5
3.19.
Is the subset
R=
{i, ctj,
02}
a subgroup of 1S3?
(For this notation see above.)
Solution:
R R
is
0.
From
Since
again in R.
the multiplication table of S3, page 57, the product of any two elements in R is <r~* = 1T2 and (t~i = cti, it follows that xy^^ & R for any x,y&R. Hence
a subgroup of 53.
3.20.
Find
all
elements of S, and Sj and exhibit multiplication tables for these groups.

'1-
Solution
Si contains one element

1
is
a multiplication table for Si.
There are two
ele-
2\
ments in
82'-
and
J
/8
/I
{
2\
^
)
The multiplication
table for S2 is
/3
Sec. 3.3]

S4.
59
3.21.
Find the elements of

Solution
2
'
3
(Tg
= -
= = = = = =
3 4 3 2
3
4
TS
'1
2
2
3 4 3
1
4
1
2
1
3 3
4
4 4 4
2 2
:)
CTy
(a
2 4
1
3
1
2 2
2
4
^2
"1
G
/I
4
3
1
I)
'\ 2;
"8
G
/I
2
2
3
1
4 2
1
4
1
3
1
2 4
4 3 4
'"e
4
as
1
2 2 2
3 2 4 2
3 3 3 2 3 3 3
4
1
(72
(s
(2
4
3
1
2
1
4
2
1
3
3
4
1
"3
=
=
3
CTg
4
"4
(l
:)
Tl
G G G
3 2
3
2,
4
1
3 3
1
2 4
5
1 1
1 1
4
4 2 4 3
3
1
3 4 3 2
4 2 4
1"8
2
3
"4
G G
4 2 3
:)
T"2
2
1
4
4
6
"h
2 4
2
1
:)
We
give the multiplication table for future reference.
THE SYMMETRIC GROUP OF DEGREE

1
4
Tg rs
ffi
<Jl
a2 "2
0-3
0^3
CT4
r5
CTg
ffy
tJg
(Tg
Tj
T2
T3 T3
"^4
T4 T4 1
l-l
T5
Tg
i-s
Tj
'7
a2 2
^8
3
3
T7
4 4
l^S
<*5
"3
I
"4
i-e
^S
^6
^2
CT7
"S
'9
ri
'2
's
"I
'"1
5
"5
"8
T5
<'l
"2 '3
1
2
"8 5
''e
T4 1
Tl
T-e
"5
''S
rs
a7
3
r?
"9
^8
"4
's
"6
T2
6
i-s
"2
"2 "3 "4

"5
"l
"4
^3
3
T?
/
6
T5
T-S
U
1
Te
"1
'4
^"2
5
"5
's
"8
T2
"S
CT4
2
'S
"9
^8
T-?
"3
"4 "5
"1
"2
^1
2 4 "2 3
"9
t
"6
"9 T3 1
"^l
4
1
'"2
(77
6
"3
^1
'4
3
(77
''g
iê
'8
3
"8 4
''s
2
T-g
''I
"6
T7
(77
"5
T\
"l
"2
U
'I
5
'4 'e
2
T5
<'8
<^4
6
T7 "8
1 ^2
t
T8
T6
"9
T3
"3
l"!
4 "2
Tg
0-4
"3
<'8
(77
"e
"6 "7
^4 1
"5
^4
"3
fg
''s
5 "4
T5
"2
"6
6
5
'3
T8
^3
T6 "5
1
''2
6
"5 1 T8
1-3
'7
Tl
"3
T8
(74
2
^2
"4
T\
"2
^2
'5
O'S
"8 "9
1-1
2
^2
5
i-s
3
T8 1
"1
"2 6
1-4
4
'"5
"9
I
0-7
re
T7
u
4
^7 "8 "2 ^2 re
I
6
<f5
"'I
"6
^4
"3
T-l
<'9
(77
<'8
2
I
"5 "2
1-8
''e
<'l
"3
"8 ^5
Tl
<f5
3
^z
T6
'l
<'9
^1
"4
"I
3
<^9
"'7
2
''a
TS
^3
T2
t
^6
"5
^4
"1
<74
"6
<77
"6
<*5
4 6
(77
5
4 "3
'2
T2 TS
Te
2
<'3
T7 T6
rs
'l
<3
Tl
"5
I
"2
"&
"3
"9 "4
"6
T2 TS
1
<'6
5
0f7
T7
Tl
"4
"4
"1
"5
'7 T3
<'8
'"s
T4
I
n
"s
t
ê
S 3
<'4
2
8
1
^1
T4 T5
^6
T4 ^5
i-e
"5 "3
"2
"5
TS
i-s
T6
^2
t^S
<^3
"^
<'7
<'4
2
"1
U
'7
i-e
4
"7
"9
^8 T2
6
"1
2
"4
"8
Tl
'7
(72
"5
3
"1 "9
tê
6 "7
<*3
2
"3
"4 3
"1
T3
'2
"2 "8 '2

<'4
TS
Tl
'"4
"3
4
'9
5
O'l
"3 2 1
Ti '7
"6
"4
'"T
^"7
6
(T4
'"S
S
'g
"6
<'2
T4 "8 "3
CT4
T5
CT5
"5
'3
7-8
1 4 '7
t
'"8
'"8
1-1
"6
T4
T5
0-7
^4
"5
t'2
T6
ffe
"2
<*5
'"7
2
t
"7
T5 TS
1
"3
0'8
^3
"9
1-2
Tj
"5
's '2
"6
"^S
"1
3
"7
2
4
4
1
^S
'"6
2 3 4 "5 6
2 3 04 5
e
"5
^1
5
"4 "6
"8
^2 "2
^8
"1
^6
l-l
<'9
"3
6
2 "3
''G
"6
1
T8 T6
"2 TS
l-l
'4
n
'4
's
4 3
"3
<f9
"8 "2
'"3
"6
CT4
'7
^6
''I
(77
6
CT7
"5
"1
"9
1
t'4
'4
^S
T3
'"2
^5 "S '7
^3
Tg
"5
"6
"6
"9 "3
<'3
"2
(77
T7
7-5
^8
"5
T6
1"4
2
<T7
T?
'"5
"1
'"4
"4 5
"8 2
"1
"2
Ta
T2
Tl
'l
*
1-6
'8
"2
''S
"3
"4
"X
"6
"8
T2
60

Even and odd permutations
[CHAP.
b.
We are interested in a special subgroup of S, the alternating group of degree n, usually denoted by An. This subgroup An is obtained from Sn by singling out certain elements.
To begin
with, consider S3.
Let a
=
3<T
-.
Then
2 1 1'
2(7
1(7
3<T
Iff
2<r
2-1
If
r
3^n^
3-2
3 ~ 2-
2
1'
3-
1 3 3-2 =
! 2
then
1^
2r
It
3t
It
3t
- 2t
2-1
We
say
<r
3-1
odd.
3-2
G
2-3 1-3 1 2-1 3-1 3-2

if
is
even and
t is
More
generally, let us call a
Sn even (or an even permutation)

-Ho-
2(7
Ict
3(7
we
1<7
3<t
2(7
G
lc7
no'
2g
n^
2-1
On
2(7
3-1
call
3(7
3-2
a
3(7
nW(7
_
w-2
n(7
(n 1)<7 _ w-(-l)
the other hand,
S odd (or an odd permutation)
if n(7
1(7
1(7
2(7
1(7
'
2(7
(n
1)(7
2-1
The
3-1
3-2
odd
is
n-1
n-2
-(n-l)
_ _ "
definition of even or
written more briefly as

IS
(7
even
i<fc
I
?-
(7
is
odd
if
-Y
= 1
i.e.
,1,1
,
We shall
show that an element
in S is either
even or odd,
_^ =
1.
Corresponding to each factor k im the denominator, we will find a factor in the numerator which is either k-i or {k-i). As cr is a permutation, there exist unique integers I, m such that 1<t = k, ma = i. U l> m, the factor hj-m<T = k-i appears in the factor m(j-l(T = -{k-i) appears in the numerator. The the numerator. If
Km,
quotient
^~ '^
or
^^ ~
f''
is
thus 1.
Note that
distinct factors
k-i,k'-i'
in the
in the
denominator give rise to distinct associated factors tor. For if 1 = 1', m = m', we have k = k' and i =
{l(T
i'.
m<T),
{l'<T
m'<j)
numera-
Thus regrouping the factors in the numerator, the product becomes the product of factors 1 and hence is itself 1. Therefore every permutation of Sn is either even or odd.
There
is
an easy way of determining whether a permutation
a is
even or odd.
If
we
of integers, we call the number of integers in the row smaller than the Thus for example the number of inversions in the first integer, the number of inversions. row 7, 4, 3, 2, 1, 6, 8 is 5. We will use this concept of inversion to find out if a given
are given a
row
permutation
'1
1(7
2 2a
n
n<T
...
is
even or odd.
Sec. 3.3]
61
To do
as
this
we must
calculate
Tl (-^
Much
of the calculation
is
redundant
we have proved that the result is always 1 or -1. We must only determine the sign. In the denominator we always have positive numbers. In the numerator a negative number ka - i<j arises if in > ka. For fixed i and varying k, the number of negative factors that arise is the number of A; > i for which > fca. But this is the number of inversions in the
row
The total number of negative factors is the number of inversions in la, 2(7, .. .,71(7 + the number of inversions in 2(r, So-, n^ + the number of inversions in {n - 1)<7, na. Let this total be /. The product of I negative numbers is positive if / is even, and negative if / is odd. So a- is even or odd according as / is even or odd.
ia, (i
. .
+ l)a,
.,na.
Example
6:
Is
even or odd?
in 1, 2 is 0.
The number of inversions in 3, 1, 2 is 2. The number of inversions Thus the total number of inversions is 2, and hence a is even.
Problem
3.22.
Is a
(i)
even or odd?
'12
3
(iv)
(ii)
(V)
(iii)
Solution:
(i)
Number
of inversions in
14 14
3
3
4 3
Total
number
a
is
of inversions
Hence
(ii)
even.
5 3 2 4 1 3 2 4 1
Number
of inversions in
=
= =
2
1 1
2 4 1
4 1
Total
number of
<r
inversions
Hence
(iii)
is
even.
Odd.
(iv)
Odd.
(v)
Even.
c.
The alternating groups
We
1.
shall
show that the
set of even permutations
forms a subgroup of
=
S.
Let An be the set of even permutations in S.

tion
I
Then A
fc
0, since the identity permuta-
A:
Tl
i<fc
ki
iL i _ ~
T-r
i<k
i i
2.
If
<T
and
G Sn,
then
TT
*<"
kjar)
{(ar) i
JJot) i _ TT (fco-)T (to)T *<fc ka ia

TT
kjar)
i<k
~ ja ~ i<j "TT fcg to ^<k k i

ku ka
(3.1)
62
[CHAP.
We
will
prove that
rr
(fco-)T
fco-
{i(j)T iff
rr
m<i
It
I
mr m
{3.2)
To do
this
we
will
show that each
of the factors
-^j-^
corresponds to one and only
one of the factors ^^-^
v^
Corresponding to each of the factors
in the
that
right-hand side of {3.2), there exists, since a is & permutation, unique integers p, q such If p > q, then a factor Vff I and qa = m.
(Po-)t
p<T
(g(r)T
qa
If
_ "
q,
It
I
niT
m
_ ~

niT
appears on the left-hand side of

{qa-)T
{3.2).
p <
then a factor
It
I
qa
(Po-)t pa
_ ~
{P(t)t
p<j
(g(r)T
q(T
m
,
appears on the left-hand side of

factor
(^"^"
{3.2).
We
\
associate with the factor

^^"^"
_
if
^
P
the (equal)
^^")"
if
p<Tq(T
p>q, 1,
r-
and the
(equal) factor -1 /
q^
~ ?"^" _ p^
<
q.
It
is
easy to see that each of the factors of the right-hand side of {3.2) corresponds in this way to one and only one of the (equal) factors of the left-hand side of {3.2). Hence {3.2) holds. From {3.1) we therefore obtain
Yl
i<fc
fc((TT)
- i{(rT) i k
"<'
Yl
^^""^^
l
~ n ^^ k
<fc
^^
{3 3)
i
It
follows
3.2:
from
{3.3)
and the rules for multiplying +1 and -1, that we have

of
is
Lemma
The product
(i)
two even permutations two odd permutations
even
(ii)
is
even
is
is
(iii)
an odd permutation and an even permutation

an even permutation and an odd permutation
odd
odd.
(iv)
Accordingly
aT~^
if
G An. An
is
aa"^ = t is even. Thus if <t,t G A, cr is even, then ct~^ is even too, since therefore a subgroup of Sn. It is called the alternating group of degree n.
As an
illustration let us find the multiplication table for A4.
From
4\
1/
the
list
of elements
of Si given in Problem 3.21 we determine that the even permutations are

1
4\
""^
1234/
1
_ ~
/I
2
1
4\
\2
/I
(yl
4
3
3/
4\
^^
^ /I ~
1^3
2 2 2 2
^/l
^-^
2
1
3 3 3
1
4
2 4 4 4
4
3
1
\4
4\
^1
3
1
4 2
12/
3
_ ~ ^ ~
2
3
4
3 2
2/
4\
"*
_ ~
^
/I
(^4
4\
_/l
^^"1^2
2 3 2
3/
4\
4321/
The
3.21.
4\
/I
2
4
/I
1^2
^
^'
/I
^^
[1
3)
^'
431/
\3
124
multiplication table
is
easily written
down from
the multiplication table of Problem
Sec. 3.3]

i
63
"2
CT2
"5
''o
('g
^1
^2
^2
T3
^4 '4
-^5
7-6
Tj
7-8
''S
^1
^3
^6
Ts
'e '3
r?
^2
7-8
"2 "S
"2 "5
f^s
''S
"5 "2
^4
"^7
Tl
n
^1
T5 '4
"8 "5
i-s
'"S
I^S
^2 ^7
TS
'7
7-6
"8
Tl
<'2
i-s
''e
n
7-6
T4 '7
^2
'"a
'"1
U
T2
T4
TS T7 T6
^"2
"2
''S
"s
^5
7"3
'2
^3
rs
^"2
^6 ^7
'"1
Ti
i-s
''S
"s
^"4
"2 "8
76 '2
1"3
'S
's
U
I
"2
^2
^8
l^l
U
T5
^4
's
^4
^8
^"7
^2
<f5
^3
"%
"5
"'2
^8
^1
'"S
"s
^"7
7-6
Te
^7
^6
'7 ^6
'1
^2
T3 '2
T4
'S
T-8
'"I
^2 "8 '2
^5
7-4
"5
I
l-T
'"S
''2
^4
'a
"5
T3
'l
7-8
's
'8
T5
l-fi
"8
"5
"2
77
Problems
3.23.
Write out a multiplication table for

Solution:
(i)
(i)
Aj,
(ii)
Aj,
(iii)
A3.
There
is
only one element in
S^ namely
i
L)
is
and
it is
an even permutation.
Sj.
Hence a
multiplication table for
is
[Tj.
Notice Aj
the
same group as
/I
2
(ii)
S2
^^
1 " 1
2\
"
I
/I
'
2 "
1
H
1 1
1 " 1
2 "
2/'V2
: is
.. --. , an even permutation and
2
is
an odd permuta-
tion.
Therefore
A2 =
2
I
for
(iii)
A 2.
and
ijTj
where
'
^j
is
a multiplication table
S3 contains six elements (see example in Section 3.3a). permutations, and a multiplication table for A3 is
"2
I
The elements
.,
o^
and
a. are the
"1
"2
i
"1
"1
"2
i
"2
"2
"1
3.24.
Prove A
Solution:
else fixed
=
>
S implies n
1.
S must contain a permutation which interchanges 1 and 2 and leaves everything = 2, 2r = 1, and ir = i (i = S, .,n). r A, since r is an odd permutation, and therefore A - S. By Problem 3.23(i), Aj = Sj. Hence A = S implies n = 1.
If
71
1,
i.e.
Ir
3.25.
Show
If
that the set
S=
{i,,j2.,<'s,<'s}
is
a subgroup of A4.
(For notation see page
62.)
Solution:
we
look at the multiplication table for A4,
we
again in S. It is clear that Hence S is a subgroup of A4.

IS
<Xj,u^&S
implies
see that the product of any two elements in -i (7 = a.a-i e S because a. o;. 8) \J " ,
25
.
64

The order
of
[CHAP.
d.
An
Suppose that n
We now
-PC*-
count the elements of A.

t:
^ 2.
Let
be the following element
1^2, 2^1, 3-^3
n-^n
We
claim
t is
odd because
2t
It
'
3t
2-1
Now
are
let
tj, 2
It
1
'
3t-2t
Wt(i.e.
It
Wt
(w
3-2
fjT, C^T,
n-1
.
.
- 1)t _ _
Then
(5.^)
n-(w-l)
be the even permutations
elements of A).
., fcT
all
odd; moreover,
e,
if
e^T
c^t,
then
e,.
.,(tt-')
(.,t)t-'
(e.T)T-l
c-Itt"')
e..
.,
Notice that this means there are at least as many odd permutations as even ones. there are exactly the same number of each; for if is odd, then
o)
Indeed
(')''
every odd permutation is listed in {3.A). Thus there are precisely the same number of even permutations as odd ones. Consequently, as k is the number of even permutations, the number of odd permutations But every permutation of is also k, and the number of odd and even permutations is 2fc. S is either even or odd, and |S| = n\ Therefore k = n !/2.
since
t^
t.
Since
<ot
is
even by
Lemma
3.2,
Putting our results together,
we have proved
integer
Theorem
Problem
3.26.
3.3:
If
n is any positive
>
1,
then S
is
of order
n \, and An
is of
order
n Ml.
Check
to see
whether
\A^\
U=
2, 3, 4)
agrees with Theorem
3.3.
Solution:
IA2I
by Problem by Problem
3.23, 3.23,
3.3c,
and 2!/2 and 3!/2 and 4!/2
= = =
1. 3.
IA3I IA4I
= =
12 by Section
12.
3.4
a.
GROUPS OF ISOMETRIES
Isometries of the line
We begin with certain subgroups of Sr, the symmetric group on R, the set of real numbers. We think of the elements of R arranged as points on the real line. Then if a,b GR
it is
clear
of
a b. We define
what we mean by the distance between a and b, namely the absolute We denote the distance between a and b by d{a, b).
value, \a-b\,
a group I{R), the group of isometries of R, as a subgroup of Sr in the follow-
ing way.
1.
Let I(R) be the set of all elements of Sr which preserve distance. The elements of this To put this definition more explicitly, an element set will be called isometries of R. a G Sr is termed an isometry if and only if
d{a, b)
d{aa, ba)
i
for every pair of elements
a.bGR.
Since the identity mapping
I{R), I{R) = 0.
Sec. 3.4]
65
2.
Suppose
pose a,b
(T
I{R).
R.
Then Then
of course a'^
ftcr-i)
Sr.
We
claim
<7-
7(72).
To
see this, sup-
d(a(7-S
d{(a<7-i)(r, (b<r-)<T)
d(a, 6)
as
CT
is
an isometry.
is
Hence
d{a, b)
=
cr->
d(ao-"S &o-"')
which and
precisely
what we need. Thus
I{R).
Suppose a,T
G I{R);
then t~'
G J{R)
d(a(<7T->), b(<Tr-i))
d((a<7)r-i, (&(7)t-)
.Sr.
d(a<T, ba)
d(a, b)
its
Thus
I{R)
CTT~>
/(i2)
is called
and /(/?) is a subgroup of the group of isometries of R.
Considered as a group in
own
right,
Problems
3.27.
Is a
(1)
an isometry?
a;
(In the following,

(ii)
R.)
7^ 1
a:
a;
+ 2.
a: x
->
nx,
n an integer
or 1.
(iii)
-> x^.
(iv)
-^
x.
Solution:
(i)
First note that a
Sr.
ct
is
an isometry, since
d{x
d{x, x')
|(x
d(xa,x'a)
(ii)
<r
+ 2,x' + 2) =
= \x x'\ and + 2) - (a;' + 2)| =
\x-x'\
is
not an isometry, since d{x,x')
d{xa, x'a)
so that, for
x
\n\.
x',
d{x, x') =
x'\ and = d(nx,nx') = = d{xa, x'a), e.g.

=
\x
a;
|n(a;
a;')|
2,
x'
implies
d(x, x')
and
d{xa,x'a)
(iii)
a is not
an isometry, since
d(x, x')
d(xa,x'a)
so that, for
(iv)
= x', d(x, x') =
x'\ and = d(x2,x'^) = |a;2-x'2| = 2, x' = 0. d{xa, x'a), e.g.

=
\x
a;
a is an isometry, since
d{xa, x'a)
G Sr and = d(-x, -x') =

a
a;a
\-x
{-x')\
\-x
x'\
=
Is a
|a;
a;'|
d{x, x')
3.28.
Set
la
2,
2a
and
=
1|
a;
for
all
GR
excepting
and
2.
an isometry?
Solution:
No, since d(l,3)
|3
and d{la,3a)
d{2,3)
1.
b.
Two
points determine an isometry
The following the same.
lemma
gives us a
method of determining whether the two isometries are
Lemma
3.4:
If a,T
Oct
I{R) have the
same
effect
on two distinct real numbers a and
b, i.e.
ar and ba
br,
then a
t.
Proof:
Let
be any element of R.
d(c,a)
Then
d{c<T,aa)
|co-
and hence
ca
aa
a(r| = (i(cT, ar) = (ct ar)

ar)
implies
|ct
ar] =
Assume
cct
Ct,
ca ct. Since a<7 = ar, ca contrary to our assumption.
aa = +{ct Therefore
Ca
Ct
aa
ot
0,
i.e.
Ca
Similarly, ca + ct aa = ba. But a is a permutation. Consequently a = b, contrary to the hypothesis that a and b are distinct real numbers. We conclude that ca = ct and, since c is any element in i2, a t.
= (Ct ar) = Ct + ttr, = 2(b<T). Hence 2a<j = 25<r and

a(T
i.e.
Ca
Ct
2(a<T)
66

Using
this
[CHAP.
Qa
a.
Now
lemma it is possible to describe the elements of I{R). d{0(r, U) = d{a, U) = \a- ltr| = d(0, 1) = 1. Hence
a
Let
<t
I(R)
and
let
l<r
= 1
or
l<r
(i)
Itr
r
(ii)
= a + 1 and Oct = a. Let ct* be the member of I{R) defined by mapping + a. CT* is clearly an isometry. Then ct* and a agree on and 1. Hence =
a
GR
to
ct
== ct*.
Iff
1 and
a
CT
Oct
a.
ct*
r + a. Then
Thus
if
and
ra
Let ct* be the agree on and
member
1.
of I{R) defined
ct
by mapping r G
to
Hence
a*.
G I{R),
tr
+ Oct where
is
either 1 or -1.
rcr
Geometrically,
the real line
is
if ra r + a, it "moves" the real line a units to the right. If inverted about the origin and then moved a units to the right. is
Oct
= r + a,
We come now to an interesting subset of / = I{R) which we will prove Let {I:Z) = {,T\aGl,7i(jGZ whenever nGZ}. Let a,rG(I:Z). Let a and b must be integers. The effects of a and t are
ra
vt
a subgroup.
a,
Or
6.
iT
rjr
+a
where where
rijfj.)
=
=
or
or
1
r){rir
e-q
Let
jj.:
r^
r(/xT)
TjT
ilarly
rjb. Then = r. Hence

integer,
Thus
if
is
an
t~\ = n{ar~^) G Z.
fx
for
(?;)r
r{aT~^)
i?(a
= {rjr + b)ix = 6). Clearly

ctt~^
+
(/
b)
is
:
Therefore
G{I: Z) and
rjb r and sim1, and q{a b)GZ. Z) is a subgroup of /.
Problem
3.29.
Determine whether the following

groups.
(i)
sets of
mappings, with composition as the binary operation, are
The
set of all
mappings of the plane of the form

To
:
(x, y)
{x
a,
+ a)
a.
with
(x, y)
R^,
a&R
Notice that t^
(ii)
is
defined for each real
number
The The The
set of all t^ with set of all t with set of all
a&Z.
a
(iii)
Q.
real
(iv)
mappings of
numbers of the form

X
-^
Ha'.
ax
with
7^ 0,
aS^
(v)
The
set of all
mappings of the plane R^ of the form

/"a
(^) y)
~*
(*'*^>
"-y)
with
a
= Q.
a& Q
(vi)
The
set of
mappings
of (v) but with
Q",
i.e.
Solution:
(i)
a permutation of R^. t,, is a matching of R^ -* R^, for if {x, y)Ta = Ii {x,y) G R^, there exists {x a,y a)GR2 and (x a, y a)Tn = {x, y). Hence r is a one-to-one onto mapping and so t is a permutation of R"^. T_ = T-i. The set of all t^ is not empty, and Tar^^ = Tar^^, ~ Ta-h since {x,y)TaT^^ Thus the set of all Xa is {x,y)TaT^t, = {x + a,y + a)T-,, = {x + a b, y + a b) = (x,y)Ta-i,-
We
show
first
that r^
{x,y)
is
(i,2/i)i"a,
then
{xi,yi).
a subgroup of 8^2 and so

(ii)
it
constitutes a group.
Using
(i),
we need
only consider whether
ttî
T^-t,
belongs to the given
set,
which
it
clearly does.
(iii)
This follows by arguing as in

It is
(ii).
(iv)
a permutation of R. Also, fia/ay (mo)"'- Now a'oA'^' Hence the set of all /t, a ^^ 0, a 6 Q, constitutes a subgroup of the group of all tâ/bAaMi/b permutations of R and itself forms a group. easy to verify that each
/j.^
is
'
Sec. 3.4]
67
(v)
The set of Ma >s not a group, as ;Uo has no inverse. For /iq maps all points onto the point (0, 0), and by Theorem 2.4, page 36, the only elements of Mjj2 which have an inverse are the one-to-one and onto mappings. The
set of all
/i
(vi)
in this case
forms a group.
The
set is non-empty.
If Ha is
an element, then
I'-al'-^^
(x,y) and (x,y){iii/alâ) (.ax,ay)in,a Ml/a M^' because (, 2/)(Mai"i/a) 2.4 M is a one-to-one onto mapping and thus Mo is a permutation of R^.
Ma/b
= (;.?/) By Theorem ~ MaMi/b Now

set of all
/i,,
since
{x,y)tiali-^
{ax,ay)n-i
group of
iSr2.
"(f'^'l^^ )~ ^
-^
^^'V^fâ/b-
Hence the
is
a sub-
c.
Isometries of the plane
Let
be the set R^
Rx R.
A
and
If
(xA,yA)
= +
A, {xB,yB)
=B
are two elements of E,
we
define the distance
between
as
\/(a;A
ccb)^
{va
- ysf
it by d{A, B). Recall that if we interpret A and B as points of the Euclidean plane with coordinates (xa, Va) and {xb, Vb) relative to a given cartesian coordinate system, then the formula for d{A, B) is the ordinary distance between A and B. The interpretation of E as the Euclidean plane is not an essential part of our argument. Logically we do without it and work abstractly with the ordered pairs {x,y).
and denote
<j
Se,
d{A,B)
the symmetric group on E, d{A,Ba).
is
called
an isometry
if
for any A,
points of E,
Theorem
3.5:
The
set / of all isometries of
forms a subgroup of Se.
Proof: I is not empty, since the identity mapping is an isometry. We need to show only that <tt~^ G I whenever (t,tGI. Consider the effect of t-"'. As t G Se, there exist, for each pair of points A,B gE, A', B' G E such that A'r = A, B'r - B. Then
d{A',B')
since t is an isometry.
diA'r.B'r)
= d{A,B)
But At~^
A',
Bt"^
= B\ Hence = d{A,B) = d{A,B)
d(AT-,Br-i)
and
T~' is
an isometry.
Consequently
d{AaT-\ Bar-')
d{Aa, Ba)
and so aT~^ G
I.
Hence
/ is a subgroup of Se.
In the next four paragraphs
we
shall
argue informally.
Let us imagine that the Euclidean plane E is covered by an infinite rigid metal lamina If we move S so that it still covers E, we can define an isometry induced by that movement. Let the points of E be denoted P,Q,R, and the points of S be denoted A,B,C,
S.
. . . . . .
In the initial position suppose that A lies on top of P, B on top of Q, C on top of i2, After the sheet S is moved, the point A is on top of another point of E, say Pi; the point is on top of, say, the point Qi; C is on top of, say, Ri;
. .
. .
ABC
>
< -"^
Metal lamina ~^_-_ ^^'*p

Q
(6)
ABC
Q
(a)
R
Initial position.
^^-Euclidean plane
Pj
Qj
iij
After a movement.
Fig.
1.
The isometry induced by a movement.
68
[CHAP.
Let e-.EÊ be defined by Pe = Pi, Q9 = Qi, Re = Ri, Then we assert 6 is an isometry. For if P, Q are any two points of E, with, in the initial position, A of S on top of P, and B of S on top of Q, then we have d{A, B) d{P, Q). After S is moved, A lies on top of, say, Pi, and B on top of Qi. Hence d{A, B) = d{Pi, Qi) and so d{Pi, Qi) = d{P, Q).
.
. . .
Using
1.
this informal approach,
we now
describe three particular isometries:
Rotation about a point. Let O be any point of S. Rotate S about through an angle ir. Then the isometry induced by this movement of S is called the rotation about O through an angle *.
A on top of P
"7
Metal lamina
,S
Knclidean plane
(a)
Initial position.
(b)
rotation through an angle *.
Fig. 2
2.
Let us choose a line in E and turn S over this line and back to E. The isometry obtained in this way is called the reflection in XY.
Reflection in a line.
p u
i;
/
Initial position.
'.'
/
Rotating about XY.
I'
QR
RiQi Pi
(c)
(a)
(6)
Final position.
Fig.
3.
reflection in line
XY.
3.
Let us choose a line XY.
Let
which XeYe, the
line joining
Xe
be an isometry corresponding to a movement of S for and Ye, is parallel to XY. Then 9 is called a translation.
As a simpliflcation describe formally these three types of isometries. describe only rotations about the origin and reflections about the axis OX.
We now
we
A translation t^^ is the mapping defined by

can be shown that for each a,b, Problem 3.30 for details.)
It
t_,^ is
an isometry and that
(t ^)~^
t_^
(See
counterclockwise rotation about the origin through an angle
9
^
is
the
mapping
p,
defined
by
{x, y)p^
{x cos 9
y sm9, x sine
y cos
9)
3.32
For each 9, p^ is an isometry and (pj-^ = p_,. (See Problem 3.31 for proof. shows why p is called a rotation through 9.)
Problem
Sec. 3.4]
69
3.
Define a reflection in
OX
to be the
map
a^
where
{x, y)a^
{x,
-y)
{<ry)~^
This
3.33).
is
readily seen to be an isometry
and
it is
easy to show that
a^
(Problem
Since
is
the product of
two
reflections,
we
call it
a reflection.
This
is
a convenience
which
will simplify the statement of
Theorem
3.8.
Problems
3.30.
Show
that t^^
is
(Ta,},)"^
'^-a.-t-
Solution
First
we must show
that
To_
^,
Se,
so
we must show
that
it is
(x,y) (x,y)GE, (x\y'). If {^'>y')ra,h clearly implies i'^yVha.b {x,y) and so T(,j, is onto. Hence Tab ^ Sg. Is T_b an isometry? If
a one-to-one onto mapping. then {x - a, y - h)Ta,b =

(oj^.j/a)
A =
and
B=
(x^yy^),
then
d(A,B)
and
To.b is
y/(xA
Xb)^
(va
- VbV =
+
b)T-a,-b
d{Ar_^, Br^^,,)
Similarly,
an isometry. (x,y)Ta,b-râ.-b - {x + a, y t_ _bTo_b = and so (xab)"' = t ___(,.

'
(.3/)-
Hence
Tq (,T_a,_b
'
3.31.
Show
that p
e
is
(p
6
)~i
(Hard.)
Solution:
We
must show
first
that p
is
an element of S^.
{x, y)p
(x', y')p
implies
x cos x
Multiply the
first
y sin y cos
e
sin e
x' cos
a;'
y' sin
(5.5)
j/'
sin 9 sin e
-\-
cos 9
equation by cos
a;(cos2 e
and the second by
and add to obtain

e sin e)
-|-
sin2 e)
a;'(cos2 e
+ +
{y cos
sin2 0)
sine
y cos
=
Since
cos^ 9
-(-
{y' cos 9 sin 9
y' cos
sin 9)
sin^ 9
1,
x'
the second by cos

Is
it
and subtracting the
Similarly by multiplying the first equation of (3.5) by sin 9 and first from the second, we find y ~ y'. Hence p is one-one.
{.x,y)
That
is,
onto? In other words, can we find does there exist a solution to
such that
(x,y)p
-{a,b)
for any
(a,b)
G R^l
x cos 9 X
for x,y
y sin y cos
= =
a
b
i.e.
sm s +
S Rt solve these equations for x and y using the same stratagem as above, the top equation by cos 9 and the bottom by sin 9 and add to obtain
a;(cos2 9
We
multiply
sin2 e)
a cos
-(-
6 sin 9
or
=
9
a cos
-I-
6 sin 9
first
Multiply the first equation of second to obtain
(3.6)
by sin
2/(sin2 9
-I-
and the second by cos

cos2 e)
and subtract the

9
from the
b cos 9
a sin
On
substituting these values of x and y in
(3.6), it is
easily seen that they satisfy the equation. cos 9

(3.6) is
(The reader who knows that the condition for existence of a solution to can verify the existence of a solution to (3.6) immediately.)
Finally,
is
sin
cos
sm
0,
an isometry?
If -4
(a;^, 2/a)
and
B=
(xB.ys),
then
70

d(Ap
d
[CHAP.
Bp
6
\/[(A COS e ^/{(xa
Va sin
e)
{xg cos e
sin
y^ sin
e)]^
[(â sin e
-\-
Va cos
e)
{xg sin e
yg cos
9)]^
- Xb)
cos e
- (y^- Vb)
e^
[(â
~ ^b)
sin 9
(2/^
- j/g)
cos
ef
and thus
is
an isometry.
y)p p
(x,
= =
=
(x cos e
[(x
COS
ff
y
+
sin
9,
a;
sin e
y cos
9)p
2/
fl
sin e) cos (9)
(a;
sin e
+ =
y cos
e)
sin ( ),
tf)
(a cos
[a:(sin2 e
2/
sin 9) sin (9) y{sin^ 6
+
e)]
(a;
sin e
y cos
cos (9)]
cos^
i.
e),
cos^
{x, y)
and
3.32.
so
p_
=1.
Similarly p_ p
equidistant from
Show
that
(0, 0) is
{x, y)
and
(x, y)p
and that the smallest angle between the

(0,0),
is
line
Lj joining (x,y)p
2jr 9 when
it
and
6
(0,0),
and the
line
Lg joining (x,y) and
when
=s
tf
jr,
or
Iv.
Solution:
As we have shown in Problem 3.31, d((0, 0), (x, 2/)) = d((0, 0)p^, {x, y)p^) = d{{0, 0), (x, y)pj. We remind the reader of the formula for calculating the angle between two lines meeting at the origin. If the one line Li has end point {a^, b^) and the other Lg has end point (02, 62)' t^^n the cosine of the angle between L^ and L^ is given by
aia2
6162
\/(a2+6f)(a|+6|)
Put
(!, 61)
(a,y),
(a2, 62)
C*'.
J/)p
Then
if
/i
is
'
the smallest angle between the lines Lj and L^,

cos e
cos M
= =
2-n-
/
-e
=
(x2
V(a;2
+ ij2) cos e " = + 2/2)(a;2 + y^)

it
Hence
3.33.
/t
if
O^^s^tt, and
p.
if
lir.
Show
that
CTj,
is

and
<r
i.
Solution:
If
A =
a,,
(wa.I/a)
B=
(a;B.2/B)>
d(A<rjBa)
VW^ bP + (-J/A -
(-2/b))'
d{A,B)
and so
2
ffj,
clearly onto, for
_ -
preserves distance. Obviously {x,y)Oy = {%' ,v')ay implies (x.j/) = (ac'.y')- And a^ is (x,y)a^ = (a;, 2/)- Hence a^ is an isometry. {x,y)ayay = {x,y)ay - (x,y). Thus
I.
3.34.
Show
that rotations about the origin form a subgroup of
I.
Solution: identity; hence the set of rotations is not empty. Po, the rotation through zero degrees, is the )~' = Put -0 = *, as it is annoying to carry the minus sign. If p ,P are two rotations, (p p( ,) Is p (p )~i = p p^ a rotation?
(x, y)p p
= = = =
(a;
cos
sin 9 + cos 9) sin *, * + sin 9 + cos 9) cos *)) sin 9 sin *) 2^(sin 9 cos * + cos 9 sin *), (a;(cos 9 cos 4' a;(cos 9 sin 4' + sin 9 cos *) + 2/(cos 9 cos * sin 9 sin *)) cos (9 + *)) sin (9 + *) + sin (9 + *), (x cos (9 + *)
((a;
2/
cos 9
(a;
2/
sin
9,
x sine
sin
9)
+ cos *
9)
y cos e)p^
(a;
2/
cos e
y sin
sin
(a;
2/
2/
a;
2/
(. 2/)p
Hence
p p
and
p (p )"> is
a rotation. Thus the rotations form a subgroup of
/.
Sec. 3.4]
that the reflections about
71
3.35.
Show
OX
form a subgroup of
/.
Solution:
have only two elements, for there are only the two reflections and Oy. Since we have shown <r""i, we quickly verify that the two possible products of a reflection and the inverse of a reflection is again a reflection. Hence the set of reflections forms a subgroup of 1 and is of order 2.
i
We
(Tj,
that
3.36.
Show
that the set of translations forms a subgroup of
/.
Solution:
T'a.b
('cd)"'
(a;>
'a,b''-c,
-d
=
(a;
''.i-cb-d
is
easily verified as
2/)Ta,b'--c,-d
o,
3/
+ 6)r_e,-d =
{x-^
/.
a- e,y ^h - d) =
{x,y)Ta-c.h-A
and thus the
set of translations
forms a subgroup of
3.37.
Find an element of Sg that

Solution :
is
is
not an isometry.
Let ctGSe be defined by easy to verify that a e Se-
(x,y)a
=
if
{x,y)
except that
(0,0),
(0,
0)a
(l,0)
and
(0,1),
(1, 0)ct
=
1
(0, 0).
It
Now
A =
is
B=
(1,0)
and
C=
then
d(A,C).
d(Aa, Ca)
d{B, C)
= ^2
d{A, C).
Hence a
not an isometry.
d.
Isometries are products of reflections, translations and rotations
an isometry is determined uniquely by its action on on a straight line. This enables us to prove that every isometry is expressible as the product of a reflection, a translation and a rotation. (Note, however, that there are isometries that are neither reflections, rotations, nor translations.)
We
will prove in this section that

all
any three points not
Lemma
3.6:
Let a
d{A', B')
G /.
Let A,
Ba = B' and Ca =
Proof:
and
Then A'B'C
be three noncollinear points in E. Let A<j is a triangle congruent to ABC.

d{A, C)
A',
d{A, B),
d{A',
C) =
two triangles with corresponding
sides equal,
and d{B', C) = d{B, C). Thus we have and so AABC is congruent to AA'B'C.
Lemma
3.7:
If a
and
t,
do not
Proof:
lie
elements of /, have the same effect on three points on a straight line, then a r.
A,B,C which
Let
be any point of E.
Let d{D,A)
a,
d{D,B)
and d{D, C)
c.
Let
A'
Then the distance of Da from A' is a, from B' is 6, and from C is c. Similarly Dr is a distance a from A', h from B', and c from C. Geometrically it is possible to see that Da - Dt, for both Da and Dr lie on the intersection of three circles: Oi with center A' and radius a; Oz with center B' and radius h; and O3 with center and radius c.
C'
Ct.
= Act =
At, B'
= Ba = Br
and
= Ca =
two points at most. Hence Oi and O2 determine two points. two points. But we shall show that as Da and Dt must lie on O3, they must be the same point. If this is not so, then Oi, O2 and O3 have two points in common. We shall prove geometrically then that A, B, C must lie on the same
circles intersect in
It is possible
Two
for
Da and Dt
to be these
straight line.
Let three circles with centers A', B' and C" intersect in
erality
two points
and Q.
Without
loss of gen-
we may assume that B' lies between A' and C. A'P = A'Q and B'P = B'Q; hence A'B' lies on
the perpendicular bisector of PQ. Similarly B'C lies on the perpendicular bisector of PQ. Therefore A'B'C is a straight line. But aAjPC is congruent
to
straight line.
aA'B'C by Lemma 3.6. Hence ABC lies in a But we assumed this was not so.
This contradiction proves
Lemma
3.7.
72

3.8:
[CHAP.
Theorem
Every isometry of and a translation.
E is
first,
expressible as the product of a reflection, a rotation
We
steps.
present an intuitive proof
The formal proof below follows exactly the same
and C = (0, 1) as shown in (0, 0), B = (1, 0) product, *, of a translation, a rotation and a as a reflection. It is easy then to prove a is the product of a reflection, rotation and translation. So we must find a ^ which is the product of a translation, rotation and reflection such that To check that <t* = t, we need only prove that for the three noncollinear points <T^ = L. A, B, C the effect of and <t* is the same (Lemma 3.7). We build * in three stages so that we bring (a) Aa to A; (6) Ba to B; (c) Ca to C. Let Aa = (a, h).
Intuitive proof:
1.
Let
o-
/.
Let
A=
Fig.
Our
idea
is first to find <t~*
Bai
Aa
Barâ
B'
/ B
B=
^'""-a,
.^
-bP-e
-bP-6
Ac
= Ao
^"'^-a.-bP-e
C'
Fig. 1
Fig. 2
Fig. 3
line joining
$,
Apply the translation which moves each point a distance d{A, Atj) parallel to the Then B'A is at an angle Aa and A, so that AaT_ a, ~b = A. Let B' Be say, to OX and, since t_^_^ and a are isometries, is of length 1. See Fig. 2.
Apply the rotation through an angle of Q that takes B' onto B. Let C = CorT__ _5 p_^. is at a distance 1 from A and a distance Then since t_ _^ and p_^ are isometries, V2 from B. Hence C is either as in Fig. 3, in which case let ^ be the reflection in OX, or
<r,
else
C'
C,
in
which case
let
ft
be the identity reflection.
Let
*=
/x-a.-bP-^M-
Since
A,
B and C are mapped to A, B and C by a^, <t* = t by remark on the inverse of a product in Section 2.4, page 34,
=
*"
f*
Lemma
f*P<,Vb
3.7.
Hence, using our
-HP-XKr-a.-.r
and the theorem
is true.
Formal
proof:
We
(a,
f>).
follow the
(0, 1)
A=
(a) (b)
(0, 0),
B=
(1, 0),
C=
same three steps and use the same notation. and a G 7.

A(jt_^ _^
Thus
Suppose
B'
Acr
Then
A.
must be a distance 1 from A, since 1 = d{A,B) d(A(TT_,_b, BaT_,_i,) = d{A, B'). Hence B' is of the form (cos 6, sin 5) for some angle 0. [This is a well known fact of coordinate geometry. All we must check is that if B' = {d, e) with d'^ + e^ = 1, the equations cos $ = d and sin 9 = e can be solved for 0.] Then
Bo-T_
_(,
B' P-o
(cos
e,
sin ^)p_,
(cos e (cos -61)

(cos^ e
sin 9 (sin -6i), cos 9 sin (-9)
sin 9 cos (-e))
sin2
e,
0)
(1,0)
= B
Also,
Ap_ A.
Sec. 3.4]
73
(c)
C
i-
Co-i-_ _(,p_^
is
a distance
from
let
/.
A B
and \/2 from B, since
A<tt_ _bP_^
=
ix
BaT___t,p_,
S,
and
<rT___6p_^ is
(0,1),
an isometry. and
Thus
C<r*
C=
=
C.
(0, 1)
or (0,-1).
C'
Let
A, be
the identity
if
and
'
be the reflection
a^
a,
, b
u. p e^ "
Then A<t^ = A, 5<t* =

/xt^.^p,.
= (0,-1). Put by Lemma Hence c^ =

if
t
3.7,
and a
= *- =
e.
Symmetry groups
Given a figure in the Euclidean plane,
we
shall define
a group which
we
will call the
symmetry group
of the figure.
is
not sufficiently precise for the needs of Chemistry and Physics; we talk about symmetry groups. In comparing two figures, it usually turns out that what one would normally think of as the more symmetrical figure has a symmetry group of greater order than the less symmetrical figure. Also, the symmetry group of what we would normally call a non-symmetrical figure usually turns out to be of order 1. More generally, we will be concerned with subsets of the Euclidean
The word symmetry
so instead of talking about symmetry,
plane.
(Clearly, a figure is a subset of the Euclidean plane.)

3.9:
Theorem
Let S be any subset of the Euclidean plane. The set, denoted by h, of all such that (i) s gS implies sa G S and (ii) ta G S implies t G S, (J Gl forms a subgroup of /, called the symmetry group of S. (An element of Is, therefore, is characterized by its mapping elements of S, and only elements
of S, into S.)
Proof:
Is
^ 0,
as the identity
mapping
of the Euclidean plane into itself

If s
is in Is.
If
a,TG
(ii)
Is,
is
(TT-i
implies st~^
e /s? We first prove r"* G Is. G S. Thus t~^ satisfies (i), i.e.
tr-^
S,
{st-^)t
GS
Then
implies
{tT-^)T
= sGS. st"^ G S.
t
Because
G h,
To show T-i also satisfies (ii), let Hence t~^ satisfies (ii), and t~* G Is.
S.
S,
since
t satisfies (i).
Now we show
in Is.
<tt-*
h.
Let
(i).
S.
Then
tar'^
t
sa
S,
G S and
sorr"^
S,
since a
and
t~^ are
Therefore trr"' Furthermore, since <t
satisfies
If
G
/.
G h,
(ii)
implies
GS
since t~^ satisfies (ii), we have ta and consequently aT~^ also satisfies
S.
(ii).
Hence
ctt"'
Is
and
Is
forms a subgroup of
Problems
3.38.
Find a plane figure S such that
U =
does not form a subgroup of
Solution:
/.
{<r
I
<r
and for
all
S, Sa
3.9
S}
condition
(ii).)
(In other words, In
Theorem
we cannot drop
Let
be the infinite half-line starting at (1,0),
i.e.
S =
{(a;,0)
a;
1}.
Then
ti,o
U.
Now
T-J
T_i,o.
But T_i,o moves
(1,0)
to
(0,0)
S.
Hence
is
not a subgroup.
3.39.
Find the orders of the symmetry group of

(i)
(ii)
a'
[E
\^
H
1
\o 4l^_J
I
74
[CHAP.
{Hint: Use Theorem 3.8 and argue intuitively. A useful intuitive approach for this problem is to cut a cardboard figure corresponding to each figure, label its vertices on both sides, and draw its perimeter onto a sheet of paper. The isometries are obtained on moving each cardboard figure so that it lies on the drawn perimeter.)
Solution
(i)
Let S = ABCDEF. Let the images of A,B,C,D,E and F under <t e /, be denoted by A',B',C',D',E' and F'. By Theorem 3.8 it is easy to see intuitively that the image of the plane figure S will be the congruent figure A'B'C'D'E'F', since Theorem 3.8 states that every element of 7 is a product of a reflection, rotation and translation. But if a rigid body is rotated, translated or reflected it retains the same shape. Now A'B' must lie along AB, as all other sides of S are either smaller or larger than d{A',B') d{A,B), which means F' must lie on F and hence E' on E, etc. Thus a must be the identity. Accordingly I^ = {i} and is of order 1. Let
(ii)
S = UGH. Now
Sj:
|/s| is
at least 8, for
we have
as
members
of /g:
A
A
reflection in the diagonal GI,
Gsi
S2.
= = = =
G, Hsi
= =
J,
Isi
= =
/.
reflection in the diagonal

I,
HJ,
Gs2
S3:
Hs2.
H,
Is2
G.
A A
A A
reflection in
J,
OX,
I,
Gs^
S4:
HS3
Isg
H.
reflection in
OY,
G, Is^
Gsi
S5:
H,
Hsi =
J.
clockwise rotation about
of 0,
I.
Gsg
Sg-.
=
=
G, Hs^
H,
IS5
= =
O
O
of 90,
Gsg
s^.
H, Hsg
=
= -
/,
Ise
J.
of 180,
Gs^
Sg-.
I,
Hs.,
J,
Is^
G.
A clockwise rotation
Gsg
J,
about
of 270,
Hsg
G, Isg
H.
Could
.
That
all
of these are distinct is clear.

s
|/sl
be greater than 8?
Let
Ig.
d{G,
I)
d{Gs, Is)
= Vl
Only two pairs of points of S are a distance
^2
apart, namely G, I and H, J. Therefore the line Gsis is one of the diagonals of S. Similarly the line HsJs is a diagonal of S. As s is a permutation of S, distinct points of S are mapped by 8 to distinct points. This means that at most the following possibilities arise:
(1) (2) (3)
GsIs
is
is is
GI,
HsJs
is
HJ
HJ
or or
JH
GsIs GsIs GsIs
IG,
HsJs
is
is
is
JH
JH, HsJs HJ, HsJs
GI or IG GI or IG
(4)
is
Since these represent eight distinct cases and each involves the movement of three points not in a straight line, by Lemma 3.7 at most one isometry could correspond to any one case. Then \Is\ 8. But we have already exhibited eight elements of /g. Hence |/s| = 8.
(iii)
Let
be the triangle
(Tj,
KLM. Then
mapping
L" ^t"!
Ig contains the following elements:
(a)
the identity
of
Kai
(6)
<72,
=
2.
K, Lai K,
= =
^
L.
\J2
I,
the reflection in
1,(72
KN,
M<'2
Ka2 Hence
|/s|
= ^'
Let a
distance
\/2^
apart are
S Ig. Since d(M,L) = L and M, then either
and the only two points of
KLM which
are a
Sec. 3.4]
(a)
75
Ma M, Ma =
Lop
L.
Then as
must be a distance
from both
and L, Ka
K.
or
(6)
M
f.
L, La = M. Then as K must be a distance 1 from both Ma and Lit, i.e. from and L, and K is the only point of S which is a distance 1 from both L and M, Ka = K. Hence |/sl - 2, by Lemma 3.7. Therefore |/s| = 2.
The dihedral groups

Let
iS
be a regular n-gon, n> 2, e.g. one of the figures below. We will show that in S, vertices are taken to vertices. This will make it easy to determine the order of the symmetry group of a regular n-gon, n> 2.
any isometry of
A
We
Lemma Lemma
will take the following geometrical
3.10:
lemma
for granted.
circle.
Every regular n-gon can be circumscribed by one and only one

its center.
We call
the center of the circumscribing circle of an n-gon
3.11:
The center
of a regular n-gon
is
taken onto
itself
by any element of
Is.
Proof: Since every point of S is within a distance r, say, from the center 0, and <j is an isometry, then every element of Sir is within a distance r from Oct. Also, there are points of Sa which are exactly a distance r from 0(r, as there are points of S which are exactly a distance r from 0. But Sd = S. Hence the circle with radius r and center Oa is a circumscribing circle of S. But by the previous lemma there is only one circumscribing circle of S.
Thus Oa = O.
Lemma
3.12:
If
is
a regular n-gon and a
G h,
then vertices of
are taken onto ver-
tices of
S by
<r.
Proof: If A is a vertex of S and is the center of S, Oa = by Lemma 3.11. d{Oa,A(j) = d{0,Av) = d{0,A). Hence Act is a distance r from O, where r is the radius of the circumscribing circle C. The only points of S on the circumference of C are vertices. But Act is an element of S on the circumference of C. Thus Act is a vertex.
The symmetry group

can
now
of the regular K-gon is called the dihedral group of degree n. calculate the orders of the dihedral groups.
We
Let the vertices of a regular n-gon S with center

direction).
be Ai, ...,A
(in
a clockwise
^^-^
Let
radians
CT.,
l^j^n,
360
1)
is
rotate
S about O
\
in
a clockwise direction through an angle
'
(= "^0'-
^
CTj
degrees j so that
A^ct.
A..
As an example,
the effect of
on the
regular pentagon
shown below.
sâ
76

Let
T
[CHAP.
The
effect of t
be the reflection about the line through Aj and O, so that on the regular pentagon is shown.
A^-r
A^, A^t
A^.
The following diagram

lowed by the rotation
a^.
illustrates the effect
on a regular pentagon of the reflection
t fol-
Z'ÎOZ
(A2r)ff3
(Air)(r3
The elements
Aâ.
CTj,
T<7
are
all distinct.
Aj(j.
^ A^<j^,
Ttr.
^ k.
If
then
AjTo-^
Aj(T^.
For certainly Thus
a, -
<j^,
j - k, j
as
k.
implies
But
a.
implies t
<j^,
the identity, contrary to assumption.
Finally, tq.
to-^.
implies
least 2m possible elements of the dihedral group of degree n. But we can that there are no more than 2n. For if a G 7s, S the regular n-gon, then there easily show .,A^. are n possibilities for A^<t. As vertices are taken to vertices, Aâ is one of A^, Aâ has only two possibilities once Aâ is determined as d{Aâ,Aâ) = d{A^,A^, and A^tr ti are also determust also be a vertex. Once Aâ and Aâ are determined, A.a, i = 3, 4, mined. Hence there are at most two elements <r e /g which map Aâ to A.. Thus there are at most 2n elements of 7^, and so \h\ = 2w-
So there are at
Let D^ denote the dihedral group of degree n.

Problems
3.40.
Find D^ and
Solution:
its
multiplication table.
The elements of Dg are the (Tj.toj above. Note that aja^ = ffj + i if Also note that t~i = t and a(r t^ctjt = ttoit. Now tvit = a^, since a^ as AiTtT2T = AiagT = ^2'^ = -^3; *"<^ A2Tr2T = '13<'2T = AlT == -^l- So CT3T
cordingly the multiplication table
is
1
is
2,
and
tcts
0302
toji"
the identity;
rCT2
= =
ai"3,
and
a2r.
Ac-
as follows:
^2
''s
n
"1
T T
Tffg
7'a2
TO'3
"2
"3
"1
T<T2
<f3
TIT2
TCT3
"2 "3 T
T(r2
CT2
"l
r
TCT3
Tff2
"3
"2
T<r3
T(72
T "3
T
T(T2
"1
''3
<'2
rag
T
Ta2
!
<'3
"2
"1
Tffg
T(r3
"2
Sec. 3.5]

that the following are subgroups of Dg: in the preceding problem.)
(i)
77
3.41.
Show
the
{a,},
(ii)
{<ri, ctz.
"s).
("i)
{roztOi}-
(Notation
is
same as
It is
Solution:
gh~^ belongs
only necessary to check in each case that the set is not empty and if g, h belong to the set, to the set. It is easy to calculate gh-^ from the multiplication table of Problem 3.40.
3.42.
Find D^, the sj^mmetry group of the square, and

Solution:
its
multiplication table.
Notice that we have already found the elements of D4 in Problem 3.39(ii). We will, however, use the notation of Section 3.4f, i.e. Vj, j = 1,2,3,4, for the rotations and t for the reflection. Accordingly the elements of D^ are ctj and tctj, j = 1, 2, 3, 4. Now <rj<j2 = Cj + i for 1 j 3, C4IT2 = oTj, and aiaj = "jî for all i and j. Also t~^ = t, ctjt ttoit, and to^t = <7j.
Aâ^r
show To^r = Oi- AiT<T2T = Aj(T2T = A.2T ^= A4, A2T(T2T = A^(r2''' ~ A^T = Aj, and A3tct2''' A^T = A2. Furthermore Aâ^ = A4, A2CT4 = Aj, and Aâ^^ = A2. Since t<t2t and (T4 have the same effect on the three points Aj, A2 and A3, t(T2t = a^ by Lemma 3.7.
We
The following calculations

implies
"sî
facilitate the construction of a multiplication table for D4.
a\
ag
to^t
'"I''
{''o'2T)(T<r2'')
''4
(T3,
and
tto^t
a^
<T3a2
implies
t<74t
Ta3a2T
(Ta3T)(TCT2''")
=
to
''2*
Hence
a2T
tto2t
O"!
t(T4,
agr
=
'^3
=
^4
Tffg,
and
"^
mT =
T'^2
TT(r4T
TCT2-
It is
now easy
construct the table:

0^2
^^3 TO3
"^^4
"1
a2
o's
"4
r(r2
T(f4
"2
''2
(^s
"4
"1
T<74
Ta2
T
TCT3
03
"3
"4
"1
a2
Ta3
r<74
T(T2
"4,
"4 T
"1
"2
<'3
Ta2
^3
<'2
Ta^
T<r2
Ta3
T<r4
'I
"3
<'4
ra2
TCT2
TOg
T<r4
"4
<'l
"2
"1
"S
"2
T<T3
''^s
T<r4
Tff2
"3
<'4
Tff4
TCT4
Tff2
Tffg
"2
<'3
"4
"1
3.5
a.

Defining the group
The complex numbers can be represented as points of the Euclidean plane E, the comnumber z = x-\-iy corresponds to the point with coordinates (a;, y). Instead of inquiring (as we did in Section 3.4c) what are the permutations of E that preserve distance, we inquire what are the permutations of E that preserve both angles and their orientation. These are called conformal mappings. It can be shown (see Ford, L. R., Automorphic Functions, Chelsea, 1951) that the mappings ,
plex
.
.(a,b,c.d):
az + z--^^:^
where
If
(i)
a, b, c,
orientation.
d are fixed complex numbers such that ad But (T(a, b, c, d) is not always a mapping of
if
bc=0,
E to
E.
Two
preserve angles and their things can go wrong.

0.
cÔ,
z<T is
then:
not defined
is
d/c, as then the denominator becomes
(ii)
There
no complex number that maps to a/c. For suppose za = ale, then az + 6 = (c2 + d)alc and b adic = and hence be ad 0, the very condition we assumed
did not hold.
78

It
[CHAP.
seems as if we have been cheated in our efforts to argue analogously to Section 3.4c prove the a for various a, b, c, d form a group, because not all the <t are permutations of E.
in order to
However, by adding an extra element o to E and forming Eu{'x>} = E we can overcome these difficulties. > is any object outside E. It is customary to write for historical reasons. The reader is cautioned that just as the symbol x can have different meanings (e.g. X is sometimes a number and sometimes an element of a group or a groupoid, etc.), so 00 has different meanings. The > we introduce should not be confused with the < in such
expressions as lim
-=
0;
it is
logically distinct.
Our
<7
idea
is to
extend
utoEin
order to patch up the
difficulties
(i)
and
(ii)
above so that
Sjg,
the symmetric group on E.
(a)
If
0,
define
z<t
=
"^^
-r +
'''
''
,
for any complex
number
z,
and put
<t
>.
(3.7)
--
(6)
If c 7^ 0,
put za
for z
-die, z
gC,
and
the complex numbers. Put

ooor
{d/c)a
00
a/c
we have no real problem. Having had to add an extra element, we just let it map to Itself. In (6) we neatly get rid of both difficulties (i) and (ii) above, for we have both defined (d/c)a and found an element to map to a/c.
In
(a)
is
will denote the set of all mappings of E to E defined by (3.7). We will show that a subgroup of -S^, leaving most of the checking of details to the problems.
First, each of the (7(a,6,c,d) is a

<T{a,b,c,d) is given
member
of
Sg (Problem
Finally,
3.45).
Next, the inverse of
by
(r(-d, 6, c,
-a) (Problem 3.44).

^2)
<T{ai, bi, Ci, di) (T{a2, 62, C2,
^(as, 63, Cs, da)
Note that <r(l, 0, 0, 1) is the identity mapping and the inverse of an element in (denoted by i). Hence the product of an element of is a subgroup of S^. It is called the group of Mobius transformations. belongs to M. Thus
for some choice of
ag, 63, Cs,
dz (Problem 3.46).
Problems
3.43.
Determine the image of

Solutions
(1)
(i)
i,
(ii)
+ 2t,
(iii)
=,
and
(iv)
-1/3 under
ct(2,
1,3,
1).
2i 3i
+ + ~
10
2(1
10*
1
(iii)
ff
2/3
*'"
3(1
+ 2i) + + 2i) +
a)
"
13
"26^
(iv)
-l/3a
3.44.
Show
that a(d,
b, e,
is
the inverse of
<r(a, 6, c, d),
given ad
bc=
0.
Solution:
Case (a)
0.
z a(a, b, 0, d) <r(d, b, 0,
00
a)
=
-a)
(-d)
/(-a)
a(a, b, 0, d) a{-d, b, 0,
" a(-d,
b, 0,
-a)
Case (b) :
(i)
c - 0.
= d/c.
za(a,b,c,d)a(-d,b,c,-a)
a(-d,b,c,-a)
-die
Sec. 3.5]

2
79
(ii)
die or .
Then
za{(i,b,c,d) a(d,h,c,a)
=
bcz
az
ez
+ +
i-d)
azc
=
+ +
bd
be
adz
aez
bd ad
+ b +d /[( {be ad)z be ad

f az
cz
a{a, b, c, d)
c,
=
i.
ale
and
{ale)
a{d,
b, e,
a)
Hence
<r{a, b, c,
d) a(d, b,
a)
Similarly
we can show
b, c,
that
<r{d, b, c,
a)
o{a, b, e, d)
i.
Hence a{d,
a)
a{a, b, c, d)~^.
3.45.
Prove
<j(a, b, e,
d)
G Sg
for any choice of a, b,
c,
d such that
ad
bc
= 0.
Solution:
In Problem 3.44 we have seen that each a{a, b, c, d) has an inverse. By Theorem 2.4, page 36, any mapping of a set into itself which has an inverse is a one-to-one onto mapping. Therefore (r{a,b,c,d) is one-to-one and onto, and so a{a,b,c,d) is an element of Sg.
3.46.
Show
that
"(oi, &i, Ci, di) cr(a2, 62.
''a. ^^2)
=
d2<'i>
'("s-
63, C3, ^3)
where a^
a^dg
1102
62''i>
^3
''2^1
is
b^d^, c^
aiC2
SJ^d
^3
61C2
^1^2.
Prove also that
&3C3 7^ 0,
(Hint:
This
very much an endurance
test.)
Solution:
Let
a(ai, hi, Cj, di)

Ogrfs
ai,
<t(o2, 62>
<'2>
^2)
"2'
and
^(as, 63, Cg, dg)
ag.
Note that
63C3
(ajOj
(ajdi
62<'l)(^l<'2
+ ^1^2)
(ajd,
6iCi)a2d2
(61O2 + 62dl)(OlC2 + d^Ci) 6iCi)&2C2 = (oidi 6iCi)(a2d2 62^2) ^
Hence
^3 is
it
a Mobius transformation.
z
Now
&
if
E
Ci
satisfies
(A)
(B)
(C)
^ 0, # 0,
z= -djci
ZCTi
if
if
C2 = 0,
C3
d^lc^
?=
-dg/ca
(D)
z ?*
M
ajZ
CjZ
then
Zaia2
+ 6i + d, ^2 + aiz + 61 C2 +
62
(OjOj
(aiC2 -f d2Ci)z
+ 62''l)^ + +
(<*2^1
+
-I-
^2dl)
Zor3
{^\e%
did2)
Thus except for

ffg
restrictions (A), (B), (C)
are permutations,
z,
we may
and (D), there is nothing more to pi-ove. Since oâ^ and ignore one of these cases, say (D). This obtains because if for all comthen
z
< o
plex numbers
zaitr2
zo's,
can only be mapped to one element of

in the following case
possibilities:
(i)
by
0^0^
and
wg.
Ac-
cordingly
we
(a),
0.
shall not consider

Cj
by case
(ii)
analysis.
Case
(i)
0.
We
=
"3-
have two
C2
0,
Cj - 0.
C2
Then
cti(T2
Cg
aiC2
Cid2
0.
Thus
(A),
(B)
and
(C)
do not restrict
z,
and we can
conclude
(ii)
C2 v^ 0.
We
first
show
=
+
d^lcg
61
if
and only
if
ztri
= d2/c2.
-aj(6iC2
d^/e^
implies
d2
"2
ai( d3/c3)
+ d^dî +
a^c^di
aiC26i
Cgdi
A
Zffi
zct]
simple computation shows that z<rj = (ajZ + 6i)/di = d^lc^, then zai<T2 = (d^lcâ^ = and, as z
= =
d^h'^,
d^c^,
aâ^
implies z = d^c^. za-^ = . z = d^lc^

"s-
Now
if
implies
dile^.
Thus
z<rjjr2
ZCT3
in this case,
and so
80

Case
(i)
[CHAP.
(b),
0.
cj = 0.
Again there are two
possibilities:
(i)
c^
0,
(ii)
Cj
0.
C2
Then
Note also that z = dje^. If

(ii)
C2
^0.
(a)
Cg
and, as a^d^ cja^ # implies d^ - 0, it follows that c^ = 0. = didj/îda d^/c-^. Hence we need consider only the possibility = 0, xctiCT2 = '"<'2 " while ZCTg = 2 = dje^, then, as Hence ctiCT2 = dj/cj. = aiC2 + 01^2 = if and only if ajc-^ =
Cg
e^d.
o's-
C3
0.
If
djci, then
z<Ti(r2
= +
ooo-g
=
+
a2/''2 '^1^2)
while
Zffg
( di/ci)(aja2
C162)
(6i2
ai( aid,
Ci(6iC2
61C2
did2
+ 6,0,) + did2)
From ajcx = djc^ we have
d2
za^
{0,-^lc^c,^
and
02( didi
61C1)
C2(0iCi fflidi)
=0,
a^
C2
Now
ZCTj
=
Cg
d^/c^
Oj/cj
only
if
and we need not consider

If
z
this case.
Hence
oia^
ag.
(/3)
^
rfi
0.
Then
ai/cj 7^
djc^.
d^lc,^,,
za-â^,
<ct2
02/02-
(aia2
(aiC2
C\
+ 62Ci)( 'î/ci) + (02*1 + ^2^1) + d2Ci)( dj/ci) + (61C2 + did2)
<^%{~0'\d^
C2( ajdj
+ îCi) + 6iC])
<H
Cg
Finally
61C2
z
if zai = d^lc^, then z = dg/cg; for (ajz + bi)/(ciZ + dj) = djc^ implies a^c^z + = Cid22 d2d2. Hence (aiC2 + Cid2)2 = (61C2 + 'î'^2) and, as Cg = aiC2 + d2Cj = 0, = (61C2 + d^d^l{a,<fi^ + d2Ci) = djc^. Therefore
dg
Cg
^^2
'^s
C2
Cg
3.47.
Let
a{a, b,
c,
d)
a^
be a Mobius transformation and k a nonzero complex number.
Show
that
a{a, b, c, d)
a{ka, kb, kc, kd).
Solution:
kaz
-,
kcz
+ +
Denote a(ka,kb,kc,kd) by kb
;T
fed
fee
5.
If
z =
>
or
z
d/c
(if
c = 0),
then, as
c
= 0,
za^
cz
-jr
Then
a,
zai
= =
za.
0,
To
treat the special cases

worj
or d/c (when Secondly, if
+
c
=
0.
0),
we
then
first
assume
and
>
=
<T,
m5
by
definition.
c = 0,
fee = 0.
Therefore
z,
==-;=
c
kc
" S and
e
(kd/kc)a.
Thus for
all
possible
choices of
we have
za and hence a^
a.
3.48.
Show
that the set of
all
Mobius transformations
{a{a,b,G,d)
adbc
l).
Solution:
Let
M
d
{aia, b, c, d)\
adbc
= 0.
1}.
If
a(a, b, c, d)
is
any Mobius transformation, then,

that a
by
definition,
D=
b
ad
be
c
From Problem
1.
3.47 above
-
we know
a( =,
a But zr
=:
VdVd
ad-bc = 7 1= = = ad-bc ^/Dy/D

;
\Vd VD
z: zr
,
Hence
& M.
all
Furthermore, any element
VdVdJ of M obis
zz
viously a Mobius transformation.
Thus
il? is
the set of
Mobius transformations.
3.49.
Suppose ad
Solution :
If
be
= 0.
Prove
a(a, b,c,d)
iff
d and
0.
a(a,b,c,d)
I,
then
b/d
1
za{a,b,c,d).
6
Hence
'â{a,b,c,d)
implies
0.
=
1
Oa(a, b,
c,
d)
=
c
implies
=
a
0.
la(a, b,c,d)
implies
0,
d.
It
(r(l,
0,0,1)
=d =
and
then, using the results of
Problem
3.47,
a(a, b,
c,
d)
a(a, 0, 0, a)
I.
Sec. 3.5]

matrices
81
b,
2x2
In this section we will deiine the group of two by two matrices and indicate ship to the group of Mobius transformations.
its relation-
An
array
'I
I)
(^*>
of complex
a two by two (2 x 2) matrix. (Since we will only deal with 2x2 matrices, we usually omit the adjective 2 x 2.) a, b, c and d are called the Two matrices are equal if and only if their entries are the entries of matrix (3.8).
numbers
a, b, c,
is called
same,
i.e.
"
(
w) =
f'
^^
d')
^^^ ^^^
^^
aâ',b =
b',
c^c' and d =
d'.
We
define
the product of two matrices as follows:
a
b
c\/a' d)\b'
is
c'\
d'j
_ ~
faa' \ba'
+ +
cb'
ac' be'
db'
+ +
cd'
dd'
3.51)
The product of two matrices

multiplication
is
clearly a matrix.
calculation
shows (Problem
matrix
an associative binary operation.

I
The matrix
a
b
is
the identity matrix, since
c$l
dl\0
0\ ij
_ ~
+
O
[b + O
+ c\ + dj
(a [b
c\ d)
(1 \0
0\( a
\]\b
In order to determine which matrices have inverses,
we
define the determinant of the matrix

It is If
A =
then
J
J
to be the
complex number I>(A)
ad
bc.
easy to show D{A) D{B)

(
D{AB)
for any
two matrices
and
(see
,
Problem
3.52).
A =
M
^1
and D{A)
- 0,
\^
d Id{A)
/
1
-c D{A)
-b
_a
D{A)I
\d{A)
is
the inverse of
(see
Problem
3.53(i)).
If
D{A) -
0,
then
has no inverse
(see
Problem
3.53(ii)).
We
claim that the set
a
b
a, b, c,
d complex numbers, ad be
is
with the operation of matrix multiplication
a group.
For
if
A,B G M,
of
I
then,
as
D{A) D{B) = D{AB),
D{AB)
and
AB G IM.
The determinant
D{I)
Furthermore if A E S'^, then A~'^ G M since is in 31. D{A~^) - 0. Therefore ,?l/ is a group. We call iM the group of rices over the complex numbers.
Hence the identity /
is 1[q $ D{A) D{A~^) =
1 implies
2x2
mat-
The
relationship between the group of

is
matrices
now
evident.
For
in
Mobius transformations and the group of 2 x 2 Problem 3.46 we found
(T(ai,bi,Ci,di)(T(a2,&2, 62,^2)
^(as, &3, cs, ds)
82

as
[CHAP.
where
aiOz
+ 62C1,
bs
azbi
+ &2di, _ ~
cs
aiC2
+ d2Ci, and
da
biCz
+ didi. But by the def/as

Ca
inition of multiplication,
tti
fei
Ci \/a2 di/\&2
C2 \
/aia2
^2/
\6ia2
+ +
C1&2
aiC2+Cid2\
hiC2
di62
didi.j
_ "
\ba
ds
This does not mean the group of 2 x 2 matrices is identical with the group of Mobius transformations. For we have seen in Problem 3.47 that a(o, b, c, d) = a{ka, kb, kc, kd) for
any complex number

tionship between the
A;
9^ 0.
But
L
I
ft
f*
^j
Tea
Jcf*
\
,
[j^^
j^^j
e.g. if
A;
-1.
The
precise rela-
two groups
will be given in
Problem
4.81,
page 120.
Problems
(ii)
Find the inverse of
M
2i\
(I _l)
(J
J),
(c)
(d)
("^
Solution:
8i
(i)
-3 +
2i
(a)
20
(6)
(c)
10
y
\
'
(^
^ c
(d)
a
6
Vd
'
14
(ii)
+
3
3i
14
+ +
3i
(6)
^0
(
i\
(c)
(a)
-7
3i
(d)
fl/a
\
1/d J
0/
14
14
3i
3.51.
Show
that matrix multiplication
is
an associative binary operation.
Solution:
Let
A = f'
"'),
B =
("'
J), C = /"3
"3
f>3
"').
Then
(AB)C
=
61
a-2
^2
diJ\b2
dg
CsX dg/
/aia2+
\6i02
C162
0-102+ Cid2\/ 13
6iC2
"3
+ di62
+ did2/\63 + Cid2)d3 + did2)d3
dg
/(OlOj
\ (61O2
+ <;i62)3 + + di62)o3 +
("l2 (biC2
+ Cld2)63 + did2)bs
(l'2
{bâz
+ Ci62)<'3 + + dib2)c3 +
(<lC2
(61C2
/"'
\bi
c,\ /a.2a3+ C2*3 a2C3+<'2d3\
di/ybza^
+ dibs
bzCa
+ d^gJ
ai(a2C3
hiia^c^
/ai(a^3
\ 6i(a23
+ C263) + 61(6203 + dgfcg) + C263) + di(62a3 + d263)

{AB)C
+ C2d3) + + c^^ +
Ciibê^
di(62<'3
+ d^^ + d^d^)
check of entries shows
A(BC).
3.52.
Prove D(A)D(B)
Solution:
= D(AB) = D(B)D(A),
c\
for any two matrices
and B.
ac'
6c'
/a
^'
/ a'
6'
c' \
^ =
d)' ^ =
d')-
^^^"
^^ =
/aa'
(6a'
+ eb' + d6'
+ cd' + dd'
'
^"'^
Sec. 3.6]

D(A) D(B)
83
- bc)(a'd' - b'c') = aa'dd' + bb'cc' - adb'c' = (ao' + cb')(bc' + dd') - (6a' + db')(ac' + ed')
=
{ad
bca'd'
= D{AB)
Because multiplication of complex numbers
is
commutative, D{A) D(B)
= D(B) D(A).
3.53.
Let
A =
'a
I
c\
)
be a matrix.
/
>
Prove:
(i)
If
D{A)
0,
d D(A)
-I
D(A)
is
the inverse of A.
^DiA) \'
(ii)
DiA),
If
D{A)
0,
has no inverse.
Solution:
a
(i)
d c\\ D(A)
<^/l
ad be ac
D(A)
-\-
ac \
\
D{A)
a
V.
,
'
D(A)
1
,
D(A)
1
D(A)\f^
-b
>.
bd db be + da
\D(A)
(ii)
D(A)j
D{A)
D(A)
-b ^D(A)
rc:)
D{A)i
If
A' is an inverse of A, then D(A') D(A) = D(I) where / is the identity matrix. But D{I) and D(A) = 0. Since zero times any number is zero, A cannot have an inverse.
3.54.
Show
(i)
that the following sets of matrices are subgroups of the group c3/ of 2
2 matrices.
%
-U
a
b
c
a, b, c,
d real numbers, ad
6c t^
(ii)
=
b
a, 6, c,
d complex numbers, ad 5c
(iii)
"^
-{[^
j)
o,
<^,
complex numbers, ad
>
Solution:
(i)
%Q^
are
all
and
= (; '^)g%. VO 1/
A-^
1,
If
real numbers,
e%
A^(l dj \b 1)g%
"3^
then,
as
-5_,
D{A)' D{A)' D(A)' D{A)

easy to check that
-^. ^_.
^_
%
is
and
is
a subgroup of ^M, as
it is
closed with respect to products.

(ii)
VC,M
D(A-i)
and
!(/)
so
ZGU
Let
(^^,
ASV.
as
it is
Then
Z)(A)
D(A-i)
1.
Hence
X( is
a subgroup of
easy to check that
is
jD(7) = 1 implies closed with respect
to products.
(iii)
'J'Q.M
and 7
<P.
The inverse of
^ =
(a
G'J'.
Thus
is
a subgroup of .^M, as
it is
H"<= A-i G "P if 1/d ) easily checked that "P is closed with respect to
d)
*^
o"
products.
3.6
a.

Automorphisms
of groupoids
figures.
We
mapping
have discussed isometrics of plane of a groupoid is defined as follows.

Let
of
The corresponding one-to-one onto

is
Definition:
G be a groupoid. Then an automorphism of G G onto G such that (ab)a oaba for all a,b GG.
a one-to-one mapping
84
[CHAP.
groupoid multiplication.
Note that isometries preserve distance whereas automorphisms of groupoids preserve As the analog to Theorem 3.5, page 67, we have
3.13:
Theorem
Proof:
I.
I,
The
set A of all automorphisms of a groupoid symmetric group on G.
is
a subgroup of Sg, the
the identity mapping, belongs to A; hence
A^ 0.
=
[a(aP)][b{a(i)]
II.
If a,p
GA
and a,b
GG
then
(a6)(a/3)
{{ab)a)li
is
[{aa){ba)]fi
Thus the composition of mappings

III.
a binary operation in A.
The
identity
mapping
is
an automorphism, and so
contains an identity.
IV. (A,
is associative.
V.
If a
e A,
let
choose
a', b'
GA
(aa""')(6a~^).
a~^ be the inverse of ; since a G Sg, a"^* makes sense. Let a,b so that a'a = a, b'a = b. Then {a'b')a = ab. Hence {ab)a-^ Thus is a group, and hence a subgroup of Sg.
GG
=
and
a'b'
=
by
We
aut (G).
Problems
3.55.
call
the automorphism group of the groupoid
G and
sometimes denote
it
Find the automorphism group of

(a)
(&)
(G,
where
G =
{a, 6}
and
is
defined by the multiplication table
a
b
a a
b b
Solution:
(a)
I,
the identity mapping, is the only automorphism for the only other possibility is the mapping a defined by aa = b and ba = a. But {bb)a = aa = b and baba = aa = a; hence {bb)a = {})a) {Jod). Thus a is not an automorphism.
(6)
Define a by
aa = b and 6a = a. Note that xy = y. Hence {xy)a -yamorphism group therefore contains the two elements and a. Notice aa =
i
{xa)iya).
i.
The auto-
3.56.
Find the automorphism group of Ag.

Solution:
G',
(For table of
Ag
see
Problem
3.23(iii).)
Theorem 2.6, page 44, showed that for any homomorphism a of a groupoid G into a groupoid a mapping of G into G' such that (ffiSTa)" - 'i'2 for a^ 9i, 92 ^ <?, the image of an identity in G is an Identity in G' and the image of an inverse of g & G is an inverse of ga, i.e. la = 1' (1 an identity of G and 1' an identity of G') and if gh = 1 = hg, gaha = V - haga.
i.e.
Now an automorphism of a group G is a one-to-one homomorphism of G onto G. Therefore if Also <ria is either ag or ''i. as a is a one-to-one onto mapping. an automorphism of A3, la = Hence there are at most two automorphisms of A3.
is
i.
aÂ
Let I be the identity mapping, i.e. J = i, aj = a^, a^l = O" checking the homomorphism property, 0^2, 0-2^ = "v phisms. Note that A^ = /. Thus the multiplication table is
a^.
Let
we
see that /
be the mapping lA = i, and A are automor-
A A
I
Sec. 3.6]

G
into
85
3.57.
Let a be any element of a group G. Define the mapping p^ of that Pa is an automorphism of G and that pa.Pb = Pab where
G by
Pa-
'*
a^^ga.
Prove
p^pj, is
the usual multiplication of
mappings.
Solution:
Pa is clearly a
mapping.
Also p
If
is
fore Pa
is
one-to-one.
g^p^ g^Pa, onto, for if g
then a'^gâ = a~^g2a and hence G, g has as pre-image aga~^.

o~>fiaa-ir2a
ffi
fif2.
There-
(9i92)Pa
-~M9'ifi'2)a
p,,
=
is
giPa92Pa
and so
p,,
is
a homomorphism. Hence
is
an automorphism. Note that we have used the associative

not a semigroup.
If
law, so that our argument would not apply to a groupoid which

9{PaPb)
&
G,
{9Pa)Pb
(a-~^
9a)pb
h~â-^gab = (ab)-^g(ah) = gpa
and thus
Pab
PaPb-
3.58.
Find the automorphism group of S3. (Hint: Use Problem 3.57 to find six automorphisms. prove that there are no other automorphisms. This problem is difficult.)
Solution:
Then
Refer to the multiplication table of S3 given in Section

Pt
,
3.3(a).
By Problem
3.57, p
p^
p^
p^
automorphisms of S3. We use the notation of Section 2.4(c), page 37, to denote the effect of these mappings. We use the multiplication table of Section 3.3(a) to calculate the images under the automorphisms.
p-
are
all
ffi
ff2
''1
''2
'3
(Tj
(72 (72
Tj
T2
'"1
Tg
''2
aj
02
<T2
Ti
T2
T3
Pa2
(Tl
(T2
Tj
''S
T2
'^2
T3
'"1
(Tl
'3
CT2
''I
CTi
'^1
'^2
^S
Tl
aj
"'2
^2
"1
''1
'^2
'3
''2
(Tj
CT2
''1
''2
'"3
(Ti
02
T2
T3
'
''l
'3
(72
(Tl
T2
Tl
T3
If p
were another automorphism, then
ip
=
ip
1.
Once
1.
a^p is given,
(tjp is
known
<tip<tip
as
<t2P
=
i;
((r]<Ti)p
oipoip.
Now
(Tip
must be
either a^ or
(Tg,
for
if
say a^p
ti,
then
a^p
titi
but this
cTip.
contradicts p a one-to-one mapping, since
Hence there are two possible choices for

Tip
Tip.
Now Tip must be one of tj, t2 or T3 for if for example, a contradiction. Hence there are 3 possible choices for effect of p on all the elements of S3 is known, since
'"i"!
a^,
then
ip
(titi)p
(Tjai
(tj
i,
But once
a^p
and
Tip are
known, the
Ts
and
Tiff2
T2
So this means that there are at most six possible automorphisms.
To
find the multiplication table

Pi
we
use the result of Problem 3.57, that ppb

Per,
Pab-
Pt,
Pt,
Pi.
Pi
P<ri
P-2
P-1
P-2
P-3
Per,
P-1
P<^2
Pl
P-2
"-3
P-l
Pa-y
Po-2
Pl
PCTI
P-3
P-l
P-2
Pt,
P-1
Prs
P-2
Pi
P-2
P-1
Pt,
P-2
Pri
P^3
Po-l
Pi
Pct2
P-3
P-2
P-1
P<T2
Pa,
Pl
86

Fields of complex numbers
[CHAP.
b.
plication.
The complex numbers C have customarily two binary operations, addition and multiNotice that if a,h G C, then a-h&C; and if v^ 0, afe-i e C. We are often interested in a subset of C that satisfies the same conditions. This leads us to a field of
ft
complex numbers.
Definition:
A
(a)
subset
F of C IGF.
is
called a field of
complex numbers
if
(h)
(c)
Whenever a, b G F, then also a-b gF. Whenever a,b GF and b -^ 0, then ab^^ GF.
field
(The definition of
bers. See, for example, Birkhoflf
can be extended to sets which are not contained in the complex numand MacLane, A Survey of Modern Algebra, Macmillan, 1953.)
field.
Of course the complex numbers themselves form a

Recall that the set of complex
addition, denoted
Let
be a
field.
numbers is a group under the usual binary operation of by (C, +), and that C* = C {0} is a group (C*, x) under the usual multiplication of complex numbers (see Example 5, page 51). Therefore, using Lemma 3.1, page 55, for a subset of a group to be a subgroup, parts (a) and (&) of the definition of a field imply (F, +) is a subgroup of (C, -I-), and parts (a.) and (c) imply (F*, x), F* = F- {0}, is a subgroup of (C*, x). In view of these remarks the definition of a field is equivalent to:
3.14:
Lemma
A
(1)
(2)
subset
of
is
field of
complex numbers
-I-),
if
(F, +) is a
subgroup of
(C,
(F*, X)
is
a subgroup of (C*, x) where
F* =
F- {0},
C*
= C-
{0}.
Problems
3.59.
Show
1
that
and
are fields of complex numbers.
Solution:
fi.
If
a, 6
J?,
then
ah & R
and
ab~^
& R whenever 6^0.
The same argument
applies for Q.
3.60.
Which
(i)
of the following sets are fields?

|
(ii)
(iii)
F = {a + 6v/2 a,h & Q} F = {a+ hi a,h & Q}, i = yf^ F = {a+bi a.beZ}, i = \/^
\ \
Solution:
(i)
+ 0\/2 e F.
Let a
6\A2
and
{a'
a'
6'\/2
be two elements in F.
(a
(a+bV2) and
if
+
/^v
b'yf2)
=
'
a')
{b
b')V2
G F
a'
+ b'y/2 ^
,
0,
T
rs^w
1,
266'
a'b
ab'
Therefore
(ii)
is
field.
i^ is
field.
+ Oi= IGF.
(a+bi)
(a'
+ 6'i)
,
aa'
(a
-a')
66'
,
(b
b')i
G F
.
and,
if
a'
b'i > 0,
a'h
ab'
(iii)
is
not a
field,
since
+i#
0,
+ iSF
but
(1
+ i)-> =
1/2
l/2i
F.
Sec. 3.6]

of fields
87
c.
Automorphisms
We have discussed isometries of the plane and automorphisms of groupoids. responding one-to-one onto mapping of a field is defined as follows.
Definition:
The
cor-
A
if
(i)
one-to-one
mapping
a of a field
onto
itself is
termed an automorphism
(ii)
+ &) = act + ha for all a,h &F. {ah)a = {aa){ba) for all a,b GF.
(a
Note that the automorphisms of fields preserve both the operations of addition and multiplication.
Theorem
Proof:
I.
3.15:
The
set
of automorphisms of a field
forms a subgroup of the sym-
metric group Sf.
We
must prove
is
A
If
?^
0; this
true since the identity
mapping
A.
II.
a,l3GA, then
(a
-I-
b)(al3)
=
=
{{a
+ &))/? = =
and
{aa
ba)p
{aa)p
{ba)j3
a{ap)
b{ap)
and
for
III.
all
(.ab){ap)
{(ab)a)p
[(aa){ba)](3
[{aa)p][(ba)p]
is
[aial3)][b{al3)]
a,b
GF. Thus
mapping
composition of mappings
in
a binary operation in A.
The
identity
is
is
is
an identity element.
is associative.
IV. (A,')
a semigroup, since composition of mappings
V.
If a
A, then a
claim a~^ a is onto,
a
a'
+ 6 =: + b' =
Let "' be the inverse of a. We G A and so a will have an inverse in A as desired. Let a,b G F. Then as we can find a', b' G F such that a = a'a, b = b'a. Then ab = {a'b')a and {a' + b')a. Consequently (a6)a- = a'b' - {aa-'){ba-^) and {a + b)a~^ = aa-^ + ba-\ Thus a-^ e A as desired.
the symmetric group on F.
G Sf
We have proved that the automorphisms of a field form a group. This group is extremely useful. For additional pertinent remarks and references, see Section 5.4a, page 158.
Problems
3.61.
Find the automorphism group of Q.

Solution:
We will use the fact that (Q, +) is a group and (Q*, X), Q* = Q - {o}, is a group. Notice that X)is a groupoid (not a group, since has no inverse) and, because of part (ii) of the definition, the automorphism of the field Q is also an epimorphism (see Section 2.5b, page 42) of groupoid {Q, X) onto (Q, X). Hence by Theorem 2.6, page 44, 1, the multiplicative identity of (Q, X), is mapped onto
(Q,
1
by any automorohism of Q.
Let a be an auotmorphism of Q; then
all positive
for
ka
la
+
+)
la = 1. Using mathematical induction integers n. \a = 1. Assume ka - k for some integer fc ^ 1. 1, by the automorphism property of a. conclude that na = n
we show na = n Then (k + l)a =

for
all
We
positive
integers n.
is a group and, by definition, any automorphism of Q is an epimorphism of the Hence by Theorem 2.6, inverses are mapped onto inverses and the identity, 0, of (Q, +) is mapped onto 0. Therefore (n)a - n for all positive integers n, since na = n and -n is the additive inverse of n. Furthermore, Oa = 0. Hence ra = r for all integers. But the automorphism
Now
(Q,
group
(Q, +).
is
also
an epimorphism of the group (Q*, X) onto
itself so
that (r)-ia
facts
a = r
n
for
all posi-
tive integers r, because
element in Q, then
and
is
\ J the only possible automorphism of Q.
is the inverse of r. Collecting these m r_ (m--)a = ma a = m- = 1\ 1 1 a / n n n
we
see that if
is
(n
?^ 0) is
any
n n The automorphism group
Therefore a
of
the identity
mapping
is of
order one.
88

Find the automorphism group of
Solution:
[CHAP.
3.62.
F =
{a
b\/2
a,b
Q}.
Any
since 2 is
rational
number q
Q,
is
an element of F, since q
arguing as in Problem
only
then for any q
qa
an element of Q. But
that
+ 0\/2 = q. It a is an automorphism of F, Now \/2 e F and (\/2^/2)a = 2a 2, \/2aV2a = (^/2^)a-2a = 2, so that (^2 a )2 = 2 or V2a = y/2.
3.61.
We
(a
in
conclude
Vs
,
has
(a+b>/2)a
+ 6\/2 )a
= aa + (b-\/2)a = = a+ 6\/2 in which

{(a
da
two possible images under an automorphism of F. Hence + baV2 a = a+b(-â). There are two possibilities: (1)
i;
case a is the identity automorphism

is
(2) {a
which case we must check to see whether a
an automorphism.
If
a:
a+
+ 6V2 )a a+ b{y/2 then 6\/2 -> a 6-\/2

,
),
+ 6\/2 )(a' +
b'yf2 )}a
= =
{aa'
aa'
(a
+ 266' + (a'6 + 6'a)V2 )a + 266' - (a'6 + b'a)y/2

)
and
(a
+ 6\/2)a(a' + 6'V2)a: = =
6\/2 )(a'
- 6\/2 )(a' - 6'V2 aa' + 266' - (a'6 + b'a)yf2

(o
Hence
Also,
{(a+
+ +
6'V2 )}a
6'\/2)}a
6\/2 )a(a'
6'V2
)a
{(o+6\/2)
(a'
= =
= =
{(a
o
(a (a
+ a') + (6 + 6')V2 }a + a' - (6 + 6')v^ - 6 VI + (a' - 6'V2 + a') - (6 + b')^f2

)
)
and
(a
+ +
6\/2 )a
(o'
6'v/2 )a
Hence
{(a
6\/2
(a'
6'\/2 )}a
= (a+ 6V2)a +
(a'
6'V2)a
Thus a
a:
is
an automorphism of F.
.
The automorphism group of

i.
has two elements
and
a+ 6\/2 ^ o 6\/2
Notice aa
d.
Vector spaces
In Physics we represent a force a; by a straight line pointing in the direction the force acting and of length proportional to the magnitude of the force. We shall assume for and act in the Euclidean plane E. the moment that all the forces act on a fixed point Any It is then possible to represent a force by its endpoint, as we know it begins at 0. point of course can be represented by its coordinates, so a force can be represented by the
is
coordinates of
its
endpoint.
force
/a;
can talk of increasing the force x in magnitude by a factor 3, say. The resultant written as 3x. It x ^ {fi,f2), then Sx = (3/1,3/2). Similarly if / is any real number, we define fx to be the force x increased in magnitude by a factor / and we can prove
We
is
(//i,//2).
parallelogram law.
The sum of two forces x = (/i, /a) and 2/ - {gi, 92) is a third force z computed by the Again it can be shown that z = (fi+gi, f^ + gi). We write z = x + y. The set of all 2-tuples {fufz) is called a vector space of dimension 2 over the field of real numbers (because we can multiply the 2-tuples by real numbers).
We
(i)
shall generalize the concept of
We
shall
two dimensional physical forces in two ways: deal with arbitrary dimensions and not only 2 or 3. (We must therefore
the real numbers. of
relinquish our contact with the real world.)

(ii)
Let
If
;u
:
We shall consider vectors that involve fields other than F be any field. Let V = F"" be the cartesian product
(/i,
ji.,
. .
copies of F.
Then
consists of the w-tuples

a;
.,
fn)
where
/i
F.
/
(/i,
...,/)
(i.e.
ja
and y
F X F -> y
is
.,gn) are two elements of V, and {gu a binary operation in V) and oi-.FxV^V by
e F, we
define
Sec. 3.6]

{x, y)ix
89
(/i
+ gi, fi + Qi,
by
fx.
...,/
+ On)
{f,x).
{ffu...,ffn)
We
denote
{x, y)ix
/t
hy x + y, and
and
<
(/, a;)<o
V together with
The elements
Problems
3.63.
is called
the vector space of dimension
n over
the field F.
of
V
+
3
are called vectors.
Find
(i)
(1, 2, 3)
(6, 7, 8),
(ii)
4(6,
-2,
0, 3).
Solution:
(i)
(1
+ 6, 2 +
.
7,
+ 8) =
(7, 9, 11)
(ii)
(4
6,
(-2), 4
0,
3)
(24,
-8,
0,
12)
3.64.
Prove that
(x
if x,
y and
+ y) +
If
are elements of a vector space of dimension n, then
iy
+ z).
y
+ y = y-^x
and
Solution:
a;
=
, fn
(fi,
...,f),
=
.
(ffi,
(/i
+ ffi,
(x
+ y)-\-z
+ Sn) as = ((/i +
addition
is
. .
and z = {hi,...,hj, ff) commutative in any field.

. .
.
then
fifi)
+ /ii,
(/
flf)
+ K) =
{y
+ z)
by associativity of
addition.
3.65.
Prove that if V is a vector space of dimension under the operation /i.

Solution:
n, then the elements of
form an abelian group
is
an abelian groupoid by the preceding problem.

(/>, /g,
.
. .
(0,0,
.,0) is the identity element.
inverse of
The
/) is (-/i, -f^, ..., -/).
3.66.
of
Prove that if î = (1, 0, 0), e^ = (0, 1, V can be represented uniquely in the form
.
. .
0),
. ,
(0, 0,
. ,
1),
then every element x
a;
/iî
/2e2
/e
Solution:
Suppose
(/..., /).
Then indeed x
then
is
Jf
In
giei
+ g2e2+---+gê,
ifi,
+ /êj + ...,/) = iffi,

/je,
+ /e.
Hence
fi
9n and the representation
.,ffn)-
9i,
unique.
h = g^,
.,
e.
Linear transformations.
Let
The
full linear
group
F
.
be a vector space and

{x
a:V-^V.
(ii)
Then
is
of
said to be a linear transformation
if
(i)
+ y)a =
let
xa + yu
{fx)a
/(ica),
for
all
x,y&V
and
/GF
(gi,g2)a
For example,
{{fuf2)
(/i,/2)a
(/2,/i).
Then
ifiji)
{gug2)}a
{f2
+ g2,fi+gi) = =
(//2,//l)
+ {g2,gi) =
=
ifi,f2h
^^'
(/(/l,/2)}
/(/2,A)
/((/l,/2)a)
Note that linear transformations preserve both the additive and the multiplicative structures
of V.
Now we have
the analog of
sist of all one-to-one linear
Theorems 3.13 and 3.15. First let us define L(7,F) to contransformations of V, the vector space of dimension n over F.
clearly.
L(V,F) CSv, the symmetric group of V,
90

3.16:
L
[CHAP.
Theorem
Proof:
Ln{V,F)
Ln{V, F) as
is
a subgroup of Sv.
preserves both addition and multiplication.
"'
Hence Ln{V, F)
0.
If a G L^{V,F), we transformation? Let x,y and f xâ = X and yâ = y. Of course x^
ask whether "'
GV
{x^
e L^{V,F). Is it a linear is one-to-one onto. G F. Since a is onto, there exists x^ and y^ such that = xa~^, y^ ya~^; and (a;^ + j/Jo: = icâ + i/ja =
(a;^
a; -I- 1/.
Hence
+ y^ =
+ 2/Jaa:~> =
{x
+ y)a-^
fx;
if
and so
i.e.
ica""' -l-i/a~'
{x
/(-!)
ifx)a~K
Also, Accordingly -'
+ y)a^^.
(/a;j)a
= /(iKjtt) =
so
a,;8
((/a;()a)a"i
G LJV,F). Thus
L(F,F),
= (fx)a~^, r' ^ L^{V,F)

fx^
and we ask whether a^~^ G L^{V,F).

(x
We
((a;
have
+ 2/)a;8-i = = =
+ 2/)a)/3-> = +
(a;a
+ 2/a);8-i
X{ap-')
{Xa)l3-'
(ya)^-'
lj{ap-')
and
ifx)al3-'
is
{{fx)a)r'
S^,.
{fi^cc))r'
f{iXa)r')
f{x{ar'))
and thus L^(F, i^)
a subgroup of
LJF, iF)
is
called the full linear
group of dimension
n.
Problems
3.67.
Show
is
that if a is a linear transformation of V, a vector space of dimension n, then the effect of a .,e of Problem 3.66. uniquely determined by its effect on the elements Ci,
.
.
Solution:
By Problem
/i(eia)+

3.66 each element of
is
of the
form
x
its
= Aei +
/e.
Then
ej,
.
.
xa
+/(ea).
Hence the
effect of a is
known once
effect on the elements
.,6 is
known.
3.68.
Show that
5:
->
if a is any mapping of such that eâ = e^S, j = 1,
{ei,
. .
...,eJ^V,
then there exists a linear transformation
.,n.
Solution:
Each element of
is
uniquely of the form fê^

(/iBi
-I-
-I-
fe.

Define
a:V^V
by
+ /e) 5 =
/i(e,a)
/(ea)
Then 3
is
a linear transformation, since
{(/iSi
+ frfin) +
(SiCi
+ sr6)} a ^ =
(/i
+ S'i)(eia) ++(/ + Sr)(ea)

-I
(/ll
+ /nn) +
-I
(fi'l^'l
"f
+ On^n)^
and
{/(/i6i
+ /e)}a = =
(//i)(eia)
-h (//)(e)
/{(/lei
+ /e)a} =
=
3.69.
Is
aGl/(V,F)
if
is
a linear
transformation
and
eâ
e^,
eja
eg,
.,
e_ia
and
ea
ei?
Solution:
Yes. All
we must prove

is
that a
has /2ei
+ /3e2+
+/e_i
+ /ie
is one-to-one and onto. as a pre-image. Also,
An
arbitrary element /iCj
+ /e
(/iî
+-!- /e)a
(flîej
+-!-
Sfe)a
is one-to-one.
implies
/i
= S'i, /2 = S'2.
/n
= S'n
by Problem
3.66.
Hence a
Thus
aGZ-(V,i^).
CHAP.
3]
91
We have met many important groups, including groups of real and complex numbers, the symmetric group S, symmetry groups, the dihedral groups, the automorphism groups of groupoids and fields, and the full linear group.
Groups thus arise in many different branches of mathematics, and hence general theorems about groups can be useful in apparently unrelated topics.
In subsequent chapters
we
will derive general
theorems for groups.
GROUPS
3.70.
Let n be any positive integer and let Prove that with respect to addition G
G^ = {a + by/n
is
a, b e.
Z}
where
is
the set of integers.
a group.
When
does
= Zl
3.71.
Let n be any positive integer. Let G = {a + i6 Vm a,beZ} where i = ^/^ and Z is the set of Is G a group with respect to addition? Is G a group with respect to multiplication of complex numbers?
|
integers.
3.72.
Let D = ZxZ, Z the set of integers. Define a group with respect to this operation o
(a, 5) = (c, d)
(a.
+ c, (-lyb + d).
Prove that
is
3.73.
Prove that the group

Let
{a
of Problem 3.72
is
not abelian.
3.74.
G = ZxQ, + c, 2'=6 + d).
where Z is the set of integers and Q the set of rationals. Prove G is a group with respect to this operation *
.
Define
{a, b) * (c, d)
3.75.
Is the
group of Problem 3.74 abelian?
3.76.
If Is
we
define {a,b)o(e,d) - {a+ c,2-':b + d), is G (of Problem 3.74) a group with respect to o? a group with respect to the operation defined by (a, b) (c,<) = (a + c, 2=5 - c?)?
3.77.
Let
{m
B = + n, ^)
{e\
e:Z^Z}.
Let
W = ZXB.
Z,
We
define
where for each
zG
z^
- (z- n)e + z0.
Prove that PT
a multiplication on by (m, e){n, 0) is a group. (Hard.)
SUBGROUPS 3.78. Let G be

G.
a group and Gi c Gg c be subgroups of G. Show that GjUGzU is a subgroup of Find a group G and two subgroups Gj and Gj of G such that GiUGg is not a subgroup of G.

3.79.
Let Gi, G2,
be subgroups of G.
Prove
GinG2n
is
a subgroup of G.
3.80.
Let G be an abelian group. Let be a subgroup of G. Prove that S(H) is a subgroup of G. Let D be the group of Problem 3.72. Determine whether are subgroups of D. Let G be the group of Problem 3.74. Determine whether are subgroups of G.
Let
S{H)
{x
G G and xx & H}.
3.81.
H= H=
{(a,0)
aSZ}
and
K= K=
{(0,a)
ae.Z}
3.82.
{{a, 0)
oG Z}
and
{(0, q)
G Q}
92

Let
z).
[CHAP.
3.83.
W
3.84.
be as in Problem 3.77. Let C = {e = Let B' - {x\ x = (0, 6), 6 B}, and is a subgroup of B'. (Hard.)
Z
|
-^
a;
Z, z
z c
for
all
{a;
(0, c),
C}.
but a finite number of integers Prove that B' is a subgroup of
Prove
Using the notation of the preceding problem, is a subgroup of W. (Hard.)
let
{x\ x
(m,
c),
where
ttiG.
Z and
C}.
SYMMETRIC GROUPS AND ALTERNATING GROUPS 3.85. Let a: Z^ Z be defined by za = z + 1 for all z&Z. Let (2m + 1)/? = 2n + 3 for all integers n. Let y: Z -^ Z be defined
for
3.86.
all
integers n.
p Z ^ Z be defined by 2tc/3 = 2n, by 2wy = 2(n + 1), (2w + l)y = 2n + 1 Prove that a, p,y & S^ and show that aa Py yp.
:
Let
G=
Sp,
where
is
the set of positive integers.

P}.
{e
e
Let
Sp =
{e
& Sp and
ze =^ z
for all
but a
3.87.
finite
number
of
Prove that Sp
is
a subgroup of Sp.
z
Let G = Sp. Let G = G = GiUGgU--.
S Sp and
ze
for
all
SP
such that
z>
is
n}.
Prove that
3.88.
Let
jfiT
H =
{e
I
{tf
I
Sg,
Iff =:
1}.
S5, I9
or
19
What Prove that is a subgroup of S5. 2}. Prove that K is not a subgroup of S5.
Let
iff
I
its
order?
Let
3.89.
Let n and r be positive integers.
H =
Prove that
3.90.
{1,2, ...,r} for all
ie
{1,2
r}
and
G SJ
H
{$
\
is
a subgroup of S and find \H\.
Y a proper subset of X. Let H = {e e e. Sx and G F}. Prove that H and if & Sx and ye G Y for all thatXDH. Prove that if |y| - 2, H - K.
Let Let
be a set and
e
K =
j/
ye = y for all are subgroups of
2/
F}.
S^ and
3.91.
Let
H =
{e
A5, 1
|A|
= =
1}.
Prove that
is
a subgroup of A5 and find

b
its order.
3.92.
Let A, B be sets with
1.
Prove that S^
^3.
(Hard.)
3.93.
Prove that
if
IXj
jFl,
Sx = Sy
GROUPS OF ISOMETRIES 3.94. Let S be the even integers

subgroup of
3.95.
and
(I
S)
{e
G
G
I(R), se
GS
for
all
S}.
Prove
(7
S)
is
I(R).
Let
(7
Q)
9
I
7(72)
and qe
GQ
for all q
Q}.
Prove
(7
Q)
is
a subgroup of
7(72).
3.96.
Find the symmetry group of the figure W.

Find the symmetry group of the figure
8.
3.97.
3.98.
What is
the symmetry group of the graph of

circle.
j/
sin
a;?
3.99.
Determine the symmetry group of the Prove that

e
if
3.100.
is
any subspace of the plane and

S',
S' is a
congruent figure,
i.e.
there is an isometry
such that Se
then 7s
Is-.
(Hard.)

3.101.
Prove that
if
is
the group of Mobius transformations, then the only element
m G Af
, .
for which
mn = nm
3.102.
for
all
n& M
is
m,
i.
Let a(a,h,c,d)
a(a, b, c, d)
be the
Mobius transformation defined by

only
a'd'
if
a{a,b,c,d)
c
z->
or
= c',
<r{a', b', c', d') if and d = d', given ad 6c =
either
= a',
= b',
= c',
= d'
Prove that a = a', b = 6',
b'c'
1.
(Hard.)
CHAP.
3]
93
3.103.
Let
N be the
set of
Mobius transformations
(r(a, b, c,
d)
with
0.
Prove that
AT is a
subgroup of
M, the group
3.104.
of all Mobius transformations.
Prove that
if
m&N
{N defined as
in
Problem
3.103),
then there exists s
e.
such that ss
m.
3.105.
Let T/ be the set of
all
matrices
where
a, b, e,
d are integers such that
od 6c
1.
Find the
set of all matrices such that
/a
\c
b\/a 6\ _ dj\c d) "
/I
\0

3.106.
Prove that the automorphism group of a
finite
group
is finite.
3.107.
Find a
finite
groupoid
with
\G\
>
2,
whose automorphism groupoid
is
of order
1.
3.108.
Let
be a non-abelian group.
Prove that the automorphism group of
is
not of order
1.
3.109.
Prove that the subset
of the symmetric group S4 defined by

'1
^
is
2
1
3 3
4\
/I
'
2 2
" 2
4/
Vl
4\ /I 3^' V2
(Hard.)
2
1
3 4
3/
' '
a subgroup of S4.
Find the automorphism group of K.
3.110.
Let = {a + ib^/l^\ a,b rational numbers, where i = V^}. Verify that is a field under the usual operations of addition and multiplication of complex numbers. Determine the automorphism group of F.
3.111.
Let
be the vector space over the rationals of dimension n
1.
Let
(l,0, ...,0)}
S =
Prove that
{a\
aeL + i(V,F),
(1,0, ...,0)a
is
a subgroup of L + i(y,F) and that
S=
L^(V,F).
chapter 4
Isomorphism Theorems
We say that two groups are isomorphic if they are isomorphic groupoids. Here we shall prove three theorems which provide a means of determining whether two groups are isomorphic. The main concepts that arise are those of subgroups generated by a set, cosets, and normal subgroups. We find a structure theorem for cyclic groups. The contents of this chapter are indispensable for any further understanding of group theory.
4.1
a.
FUNDAMENTALS
Preliminary remarks
in
We begin by reminding the reader of our previous results. A group is a semigroup which every element has an inverse. Consequently we have the following.
(1)
(2)
The
identity
is
unique.
(Theorem
2.1,
page
31.)
2.2,
The inverse
g.heG,
(3)
an element is unique (Theorem then {gh)'^ = h-^g'K

of
page
is
33),
and
if
is
any group and
The product
(Theorem
of
n elements
39).
ai,
a, in
that order,
independent of the bracketing

defined as they are
2.5,
page
(4)
Homomorphisms, monomorphisms and isomorphisms for groups are

for groupoids.
(Section 2.5, page 40.)
If Gr is a group,
(5)
and
any homomorphism of
into a groupoid, then
Ge =
{x\
x^ go,
&G}
For by Theorem 2.6, page 44, GO has an identity, is associative, and each is a group. element has an inverse. Note that 6 maps the identity of G to the identity of G9 and that {g~^)e = {ge)-^ for each g in G.
(6)
If
of
e-.G^K is a homomorphism from the group G to the group K, and if i? is a subgroup G, then He is a subgroup of K. For 0|h is a homomorphism of H into K and, by
He
is
(5),
a group.
(See
(7)
Isomorphic groups are roughly the same except for the names of their elements. Section 2.5d, page 45.)
The following theorem
is useful.
a and 6 are two elements of a group G, then there exist unique elements X and y such that ax b and ya = b. Proof: We consider first the solution of the equation ax = b. If we put x = a~^b, then a(a~*6) = (aa~*)6 = b. Hence the equation ax = b has a solution. Suppose axi = b and ax2 = b; then axi = ax2. Multiplying both sides of the equation on the left by a"*, we have
Theorem
4,1:
If
a-âxi)
a~^{ax2),
(a-â)xi
is
{a~â)x2
or
Xi
X2
The argument for solving ya = b yia = b and y2a = b, then yia = yza.
{yia)a-^
ba~^ is a solution. Also, if Multiplying both sides by a~^ on the right, we get
similar; in fact
yi
94
(2/2a)a-
2/2
Sec. 4.1]
FUNDAMENTALS
95
Problems
4.1.
Prove that the groups given by the following multiplication tables are isomorphic.
-11
1
1 1
:
-1
G:
1
-1
1
-1
Solution:
Let
also a
e-.G^H
ffi
homomorphism
sible choices of
be defined by le = 0, le ~ 1; then 9 is a one-to-one onto mapping. If it is it will be an isomorphism. We must check that (fififfz)* = Qi^gê for all posand g^ in G, i.e. we must check
(i)
(1-1)9
lel9
(iii)
(-1
-1)9
(-19)(-19)
(ii)
(-1-1)9
(i):
= {-U)(U)
(iv)
(1--1)9
(19)
'(-19)
l9
(i)
to (iv) hold.
(i)
Hence
(Thus for holds.) Therefore
1-1 G = H.
by the multiplication
table.
0.
l9
19
0-0 =
0.
to choose?
Of course 9 had to be some mapping of G to H. How did we know which was the right mapping Examining the multiplication table for G, it is obvious that 1 is the identity for G. is the identity for H. We remarked that any homomorphism must map an identity to an identity. Thus the choice for 9 was quite clear.
4.2.
Prove that Sz, the symmetric group of degree 2, is isomorphic to G, where 3.5, page 53, with = 2. Prove S3 s D^ the dihedral group of degree
is
3, i.e.
the group of Problem the symmetry group
of the equilateral triangle.
(Difficult.)
Solution:
The multiplication
table for
is
1 1
1
while the multiplication table for S2
is
(Problem
3.20,
page
58):
/8
i8
Let
S2
As
it is
-> G be defined by iS = 0, pe one-to-one and onto, ^2 = G.
1.
Then
it
can be checked that
9 is
a homomorphism.
The multiplication table for S3 is on page 57, that of D^ is in Problem 3.40, page 76. As we have used the same Greek symbols for S3 and O3, we face the risk of not knowing whether a, for example, refers to an element of S3 or to an element of ZJj. To avoid such ambiguities, we will rename the elements of D^, replacing a <r by an s and a t by a *. The multiplication table then becomes
Sj
S2 S2
S3
S3
tS2
tS2
tS3
Si
tss
S2
S3
Sl
tSg
tS2
S3
Sl
S2
tS2
tea
tS2
t83
Sl
S2
tS2
tea
tSg
S3
Sl
H
Sl
tSg
tS3
tS2
82
S3
96
ISOMORPHISM THEOREMS
Note that an element
j
[CHAP.
TjSTje
satisfies tjTj = i. It e is an isomorphism from S3 to D3, then can only map the tj, j = 1,2,3, among the elements t,ts2,tss since these are the only non-identity elements of Dg which have the property that their squares are Sj. As 9 must map to Sj, it maps ffj, (12 onto the elements Sg, S3.
tj,
1,2,3,
{TjTj)e
=
i
(9
Si.
So
If we know the effect of e on ctj, since the effect of 9 on tj, then, because T2 = elements of 1S3.
o-ici
02,
we know
T3
tiCT2
^.nd
=
:
Tjax,
the effect of e on trj- Also if we know we know the effect of 6 on all the
So we have to experiment.
Tê
suitable
mapping
S3-* Dg must satisfy
ctxA
S2
or S3
while
t,
tSj or fSg.
We
026
try the following definition.
S3, T29
isg,
T39
iS2
if
is to
Let lO = Sj, ciB = Sg be an isomorphism.
and
Tiff
t.
Then we must have

a
To check whether this mapping is an isomorphism, we must check whether this mapping homomorphism. As a mechanical procedure of doing this we use the following table.
"1
<'2
is
^2
'3
\
S2
S2
\S3
S3
X
tSs
X*S3 X*S2
tS2^v
\
"i S2
<'2
X
X
\
t
X
X*S3 ^\*S2
S2
\S3 \si
S3 Si
\
\
*
\
X^
Sj
t82
X
X^Sg
tSs
X^s X^Sl
S3 Si
\S2
S2
^\*S2
*S2
X
t
X \
X
X
\.
\
\
fS2
\*S3
*S3
\
Si S2
\^S3
S3
\S2
S2
*S2^\
\ \
\
Si
\.*S2
*S2
\S2 \si \s3
*S3\
X^*S2
^3
\
S2
S3
tSg
\
t
\83
S3
\
S2
Xsi
Si
iS2X *S3\
X \
calculated as follows: In the in the second row and third column, for example, is homomorphism, In the top corner we place (aiaa)^. If e is a bottom corner we place .rê-aê. same in each square of the table. Hence these two entries should be the {<r,c2)e. (cr,e)(a,e) square are equal. Hence a is a homoChecking through this table, we see that the entries in each
The entry
morphism and as
it is
one-to-one onto, e
is
an isomorphism.
4.3.
Prove that on F and
if
<pe
F -* G
and
identity
<p:G^F are two homomorphisms such that are isomorphisms of mapping on G, then e and
e
But
e<p
is
the identity on F.
Hence x
y.
Similarly
is one-to-one.
Next
so 6 is
let
g&G;
then
94,
F.
9 is
g^-e
an onto mapping. Thus
= g; hence g is the image of an element of is an isomorphism. an isomorphism. Similarly
under
and
4.4.
Prove that
solution.
if
group G, then the equation gi,92.93 are elements of a
g^xg^^
gz
has a unique
Solution:
If
we put X
the left and
g:^'^g^~2
find
gtxg2
g3.
If
ffiifl'2
9\
g'^ on the right we have
^ fl-f Mfl'iiKi'2)fl'2
= ~
fl'ia'2'2
'
ffs,
then on multiplying by
or
a!i
î
(ffi2^2)fl'2 '
X2.
Sec. 4.1]
FUNDAMENTALS
if
97
4.5.
Prove that
Solution:
If
finite
is
finite
group and
H is
an
infinite
group, then
and
are not isomorphic.
G = H,
there
is
and
H is infinite.
s S^
if
a one-to-one mapping from
onto H.
But
this is not possible since
is
4.6.
Prove that S
Solution:
and only
if
n = m.
S has order n\ and S^ has order m! Now if S = S^, then there is a one-to-one mapping of S onto S^. So S and S have the same order, i.e. n! = m!, and this implies n m. On the other hand every group G is isomorphic to itself. In fact the identity mapping of G onto G is an isomorphism. Hence n m implies S = S.
4.7.
Prove that
Solution:
if
G=
H, then
H=
G.
Let
H=
4.8.
be an isomorphism from
onto H.
Then
i?-i
is
an isomorphism from
to G,
and so
G.
(See Problem 2.38, page 42.)
Prove that
Solution:
ii
G = H
and
H=
K, then
Gs
X.
Suppose 9 is an isomorphism from G to and ^ an isomorphism from isomorphism from G to if, i.e. G = K. (See Problem 2.38, page 42.)
4.9.
to
X.
Then
e<f>
is
an
Prove that there are

Solution:
infinitely
many
groups, no two of which are isomorphic.
Consider the symmetric groups Si,S2,

isomorphic.
Then by Problem
4.6,
no pair of these groups
is
4.10.
Prove that
Solution:
if
is
finite
group and
H is a subgroup of G, H # G,
then
and
H are not isomorphic. H
observed in the solution of Problem 4.6 that if two finite groups are isomorphic, they have order. Since the order of is less than that of G, it follows that G and are not isomorphic.
the
We
same
b.
More about subgroups

Let
G=
S^ and let
X=
{a^, t^}
where
3
"^-[2145)
(See Problem 3.21, page 59.)
<'8'"7<'8
/l
4\
^"^
^^=(2314
i^
/l
''f ''7~^'^8
It will
Suppose we wish to refer to a product such as ^^r, or rââ^ or be convenient to have some general notation. We will write
x\^
x'^,
where
1,
x^GX
to represent the product of n elements chosen from or the set of inverses of the elements of X, where x\ will mean x^, and x'"- will mean the inverse of x.. For example, if ^^ = c^ = 1,
*3
^4
^5
^6
=
1:
-1. and X,
Xg
Xg
cr^,
x^
x^
x^
t then x\^ x\^ xl^ x\^ x^^ xl^ stands
for a^râg-lr-lr-la-l.
Example
li
x\^
x'",
then ff-i
h where h
. . .
x~'^
x'^K
n
Proof:
gh
-I 1 I Similarly hg
x\^
^
x'^ x-'ri
Ttn
^-H = 1
a;î
1
...a;'"-i.l.a;"'"-i---a;-'i
Q-
as'ia;-'!
1.
11
98
[CHAP.
x^x^^
proved in Lemma 3.1, page 55, that if ff is a subgroup e H. We now generalize this and prove Lemma 4.2: If ^ is a subgroup of G and XcH, then
We
of
G and
x^, x^
G H,
then
H
Proof:
a?!/
[xl^
^1"
I
x.GX,
1, n a positive integer}
is a group, i e and x,y G H implies y-^ GH, xy~^ G H, prove the lemma by induction on n. Let n- 1. Then x^^ since a^j G H. Hence iCj* G if. Assume, by induction, that x = x^^ x^^ G for n = k. Let x^+jGZ. Since x,xlW^GH -where ^^^ = 1,
Recall that as
and
f.
We
GH
/yt/y* Îc "T ft
'"^fc
1 +l +1
_ ~
r'l
/yi
*''
X^k +
iff
Hence
aj^' .J
a;^"
/T for all n.
This proves the lemma.
Now
if
is
"large enough", e.g.

i?
X = H,
x.gX,
I
we may have
^= 1,
Ji
=
if
{Xj*
.X"
^n"
a positive integer}
We
is
ask what happens
is
not "large enough",

i.e. if
X is
a subset of
H and
S
S a subgroup
4.3:
{xl^
x^"
I
a;.
G Z,
e;
of
G?
Lemma
Let
G be
a group and
let
Z be a non-empty subset of G.
x^^
\
Let
S =
Then S
Proof:
(i)
{a;['
x.G X,
is
a subgroup of G.
that:
If
H is any subgroup containing X, Hd S.

as
55).
We
must prove
(ii)
S'0; this is true because there exists x^G X If f,gGS, then fg~^GS (Lemma 3.1, page
f,g
is
non-empty,
G S means
f
xl^
xl"
(c.
= 1)
and
^
flr
y'l
V^
('?;=
1)
where
x.
and w are elements of X.

fg-
Hence g
=
"1
1
y~
"1
and x\+i n+1
<' <"2/;
2/,
w, *'
= x^
1
.
x'"
n
*Ti x'' + m n
where x^j
is
...,x^ a subgroup of G. If
2/,
Therefore /flr i G S and and e^^ = -,?^, .,e+. = -r/^. H D Z, we use the previous lemma to conclude H'dS.
.
We
If a
denote
S by gp{X) and
call <S the
subgroup generated by X.
group can be generated by a

2:
finite set,
we
call it
finitely
generated group.
Example
What
is ffp({l}) in the group of Problem 3.5, page 53, where the multiplication table is
m = 3?
Recall that
2
1
S =
gp{{l})
=
S.
{ajj'
a;^"
î
{1},
ej
1,
to
a positive integer}
is
Now
e S and
Also,
1*2 =
S.
Hence gp({l})
the whole group.
reader that, for example, in the multiplicative group of nonzero rationals the inverse of a, which we have denoted in this section by a~S is 1/a, i.e. in this case a~^ has the meaning usually associated with it when a is a number. But in the additive group of rationals a~Ms a.
We remind the
Sec. 4.1]
FUNDAMENTALS
99
Problems
4.11.
Let
G =
Z,
the additive group of integers.
What
is S'p({l})?
Solution:
gp{{l})
D1 + 1 + 1+---+1
n times
(w-1).
Hence gp{{l}) contains
all positive integers.
gp{{l}) contains
l-i
+ l-i+---+l-i = -1 +
n times
(-1)
(-1)
(n
1).
Hence gp{{l}) con-
n times
tains all negative integers. Also,
gpiW)
contains
l"'
(-1)
0.
Thus gp({l})
Z.
4.12.
Let
G=Q
the additive group of rationals.
Find gp{{l}).
Solution:
Exactly as in the last problem, ^^({1})

differences of
1,
Z.
Since no other elements can arise as sums or
gp({l}}
Q-
4.13.
Find the subgroup of the multiplicative group of rationals generated by

Solution:
2""i
{2}.
^.
The elements of gp{{2}) are
either of the
form 2" or 2"", n a
positive integer.
4.14.
Determine the subgroup

Solution:
H of S3
generated by a^ and
tj
of Section 3.3a, page 57.
We
H contains H = S3.
4.15.
use the multiplication table for S3 shown in page 57. <t^* = 02', hence contains a2- ^^ rj, it contains r^ = tjo-x and t^ = râ^. Thus contains all the elements of S3, and so
Determine the subgroup of the symmetry group of a square generated by those isometries that leave two vertices fixed. (Hard.) (Hint: To see what is happening, cut out a square from a piece of cardboard and label the four vertices. Perform the isometries on the figure.)
Solution:
all vertices fixed.
refer to Problem 3.39(ii), page 73. Sj leaves G and / fixed; 82 leaves and / fixed; 85 leaves Hence we require S = gpiis^, Sj, Sj}), and this must contain Sy, since 81S2 = 87. It is easy to prove SiSg = Sy, for the effect of 87 and SjS2 is the same on three points not on a single straight line and this is sufficient by Lemma 3.7, page 71. Note that the inverses of 82, Sj, 87 are 82, 81, 87 respectively.
We
Let
implies
T=
{85, 87, 82, 81}.
We
tit^^GT. Since
tGT
assert that T is a subgroup. All we must check is that ti, ^2 ^ ^ implies t~^ = t, all we must check is that t^ti&T for t^jt^^T.
As 85 is the identity, t^t^ G 2" if 85 is either t^ or t^. If ti = t^, then t^t^ = 85 G T. Therefore we need only consider the following cases: s^s^ = S2S1 = 87 G T\ SiSj = s^s^ = 82 S T. Finally, 82S7 = 8782 = 81 G T. Then T is a subgroup of the symmetry group of the square. But Sd T, and T 3 {85, 82, 81}. Hence T D gpis^, s^, s{) - S, by Lemma 4.3. Thus T = S.
4.16.
Find the subgroup of M, the group of Mobius transformations by -> z -> z (z ),

1)
:
(see Section 3.5a,

>
page
78),
generated
77
t:
z-2
{z = <),
>
->
(In the notation of Section 3.5a,
i)
ct( I, 0, 0, 1)
and t
a(l, 1, 0, 1).
This
is
a difficult problem.)
Solution:
Let
fffn.^)
be the mapping defined by

In other words,
a^^^ ^^
z-*
ez
+n
for
z t^ ,
and
-> 00^
where
= 1
and
is
any
integer.
^(e,
.,
0, 1).
100
[CHAP.
We
c
will
1.
Let 2
show that the subgroup generated by = {or(^) n any integer, e = 1}.

I
i;
and
t consists of all
(t, ^j,
n any
integer,
We

claim
of the
form
o-j^
^j
Now
zt~i
= z
gp(v,T)
2.
Any
t^'-
element of
is
1.
Also,
zr
= 2 + re
ci^^
and
zt-^
n
= zn
so a(^,,
for any
positive integer
re.
Hence for any arbitrary integer

difficulty.)
t).
re,
j,
gp(ii,r).
(We must check what hapand

e
pens to
re
<,
but this presents no
Also,
s-pCv, t)
<i(,i)''
"(-n.-i)
S'pI''''')
1,
any
integer, belongs to gpi-n,
Thus
2.
Note that 17 = a(o__i) and r = "d belongs to 2. of the group of Mobius transformations to conclude that need only show that <r(_,, cr-J_,, e 2, 5 = 1. But
So we need only show that 2 is a subgroup 2 D gp(ri, t). To show 2 is a subgroup, we
oT^'.s,
(32
+ m)8 -8m -
^-a^.s).
since
^.r^^.a,
S^z
+ Sm-Sm =
<T(,
S^z
z.
Then
2(T(, j, <7~^' g,
(e2
+ n) (i(_aâ) =
^^2
^-a^.s, + reS - Sm,
and we conclude that
a-Jg,
a(5_a_, 5)
2.
c.
Exponents
We have
a's (to
> 0),
seen in the previous section that we often are forced to consider the product of e.g. a a. (Note that as we are dealing with groups, it is not necessary
V
m to indicate in which order the multiplication is performed. See Section 4.1a.) It is convenient to introduce the notation a for the product of a's, > 0. Then a*" a" is the product of m a's followed by n a's, i.e. a^-a" = a'"+ (Section 2.4d, page 39). Our idea is to extend the exponent notation in a sensible way to zero and negative exponents. We would naturally like the law ,.4. /, ^x
when and n are arbitrary integers. Now if it were true that aPa/^ = a"*, then multiplication by a" leaves a unchanged. Hence we have only one choice in extending the exponent notation and retaining the law (i.l), namely putting a" = 1, the identity. Now
to be true
if
m = n
(i)
where
j_g_
n> 0, m + n
^g
j^yg^.
0.
gm+n
-go-i^
py^ ^m
if
(a)-i.
Because we want (4.i) to be satisfied, we must have Note that (a")~ = (a"')" = a-". Thus we have
defined a to be
the product of
1 if
m a's
m
m > 0, m < 0,
and
n.
all to
(ii)
(iii)
mO,
a~^'s if
(i.l) for
the product of
hoping thus to satisfy
(i-.l) is true if m,n are both positive. If both are nonnegative, again by running through the possible cases {4.1) holds. If both and n are negative, then
a"a"
(a"i)<-'"''(a~i)<""^
(a-i)<- +
~''>
(a,-i)-<+> If
a"+"
If m and n are nonpositive, again the result is easily verified. by checking the various possibilities m > n, m n and
a'"(a-)<~">
to
>
n,
and n <0, then
to
<
we
find
aâ"
a'"+".
is
a"""
{Jt.2)
Another result which holds for exponents

(a")"
We already know that {4.2) holds when n - -1. If m, n are positive, (a*")" is the product n elements, each of which is the product of to a's. Hence (a)" is the product of tow a's. TO is negative, n positive,
(a)"
of
If
((a-i)-")"
= =
(a-i)-"
as
-to
>
0,
w>0
=
(a-')-"'
a" as
WW <
=
((a-")-')-"
If
now %
is
negative,
(a-")"
= _ =
(a')(-i'(-"'
(j(-m)-(-n)
(by our previous remarks)
vtvx
Hence
(.4.^) is
proved.
Sec. 4.2]
CYCLIC GROUPS
101
In the study of groups there are two main notations for the binary operation. One is the multiplicative notation we have employed up until now. The other is the additive notation. We denote the binary composition by + in this case. The identity is denoted by zero, compositions of the same The result of performing 0, and the inverse of a by a.
n>
element,
i.e.
of taking a
V
+
n
a,
J
we
denote by na.
{n
The law
(i-.l)
becomes
na +
while the law
(4.2)
ma =
=
+ m)a
becomes
n{ma)
{nm)a
In other words, translation takes place according to the following dictionary:

Multiplicative notation
Additive notation
ah
1
+
a
na
a~^
a"
It is immaterial which notation one uses. But additive notation is most often used for a group in which the order of the composition of two elements is irrelevant, i.e. in which a + h = h + a for all a, h in the group. Such a group is called abelian, after the Norwegian mathematician Niels Henrik Abel, or commutative (Section 2.2, page 29).
Problems
4.17.
Find
1^,
1* where
Q, the additive group of rationals.
Solution:
In the additive group of rationals the binary operation
is
the usual addition.
Then
1^
means
lolol where is the binary operation in Q. Hence 1* = 1 + 1 + 1 = 3. Also l"* means (l~i)*, i.e. l-ioi-ioi-ioi-i where o is the binary operation under discussion. Now l""i = 1 in
(Q,+). Thus
1-4
(-1)
(-1)
+ (-1) + (-1) =
-4.
4.18.
Find
2^,
2~^ where 2
Q*, the multiplicative group of nonzero rationals.
Solution:
22
and 2-3
(2-i)3
(i)3
4.19.
Find
<fi,
j,an =
element of S4.
Solution:
<7^
and so
ct^
z= ,
<,.
4.20.
Find
t",
where
t is as defined in
Problem
4.16.
Solution:
t"
<'(.i).
See Problem 4.16.
4^
a.
CYCLIC GROUPS
Fundamentals of
cyclic
groups
If gp{X) = H, we say H is generated by X. To get an understanding of groups, a good plan is to investigate the simpler groups first. So we begin by considering groups which can be generated by a single element. We call such groups cyclic. Thus a group H is cyclic if we can find an element x G H such that H = gp{{x}). We will usually write
gp{x) instead of gp{{x}).
102
4.4:
[CHAP.
Lemma
gp{x)
{t\
t^x", ran
integer).

Cyclic groups are abelian.

x.
Proof:
gp{z)
= =
(ajji
a;^"
{x},
^.
1,
0}
n>
0}
{x^^ (
x^"
I
1,
>
5
r
I
^0
=
If a, 6
{'
any integer}
e fl'p(a;),
then a
a;'',
x,
a&
a;''x''
x-'+s
6a
x'x''
a;""*".
any two elements of a
cyclic group.
Thus we have shown that
cyclic
Hence ah = 6a for groups are abelian.
Suppose now that - gp{x) and \H\ = (m < >). Then we know that the elements of are of the form x" for various integers r. The x' cannot be distinct for all integers i. Consider x" = l,x, .,x^-^ and suppose these are distinct but that x' = x" for some k<l,
kÔ;
then
this
x'(a;'=)-i
a;'-"
1.
If
assumed
was not
so.
Hence k =
k-^0, and
. .
m = l-k<l
x'-
and
a;""
is
equal to
a;".
But we
x"
1.
We will show that {l,x,x^, .,a;'-'} is actually H. This is easy. that as a;' = 1, every positive power of x is in S. Furthermore, a;"^ = a;'"'. negative power of x lies in 5. But = {x" r any integer). Therefore S= as stated.
S=
First notice
Hence every HqS and so
Thus we have proved
Lemma
4.5:
Let
be cyclic of order
m generated by the element x.

=
1,
a;m-ij
Furthermore
x""
and
a;"*
is
the least positive
Then G = {x", x^, power of x that is

.
.,
1.
We
w>
ask a simple question: do cyclic groups of order exist for all finite integers Yes Let us consider in the symmetric group Sm of degree the element
!
<'
...
3...
2
...
m1 m m 1 m2 m1 m m 1 2 m m H = gv{<J^
is cyclic
Then
4...
2 2
... ...
1 1
and so the elements
i,a^,
.,<rJJ~'
are distinct and
of order m.
Hence
there exist cyclic groups of order
m for each w > 0.
And now we ask

order
w?
another question: are there two essentially different cyclic groups of Rephrasing the question, we ask: are two cyclic groups of order m isomorphic?
Lemma
4.6:
Let
G=
gp(x),
H = gp{y)
a;*"-!),
be each of order m.
{y\ y\ ..., y""'^}.
Then
G = H.
be defined by
Proof:
G=
{x",
x,x^,...,
H=
y'
Let
e:G^H
x'e
{i-Q,l, ...,m-l)
it is
Then
is
one-to-one onto H.
To prove
an isomorphism we must show
it is
a homo-
morphism.
Consider
Sec. 4.2]
CYCLIC GROUPS
103
Now 0î,j^m-l. Then Oî + j ^2{m-l) = 2m-2 and so i + j = em + r r TO 1 and or 1. Hence {x*-^')9 = (a;+'-)^ = {x'"'x')e - (x')e = V (x'e)(a;'5) = i/V = y^+' = ?/+'' = ^"'y' = r But
e
where
Hence
is
((a;')(a;'))9
{x^6){x'e).
Thus
is
homomorphism
and, as
it is
one-to-one onto,
it
an isomorphism.
We now
infinite cyclic
ask the obvious question: groups isomorphic?
are there any infinite cyclic groups and are two
Consider the element a in the symmetric group Sz on Z, the set of integers, defined by
Za
1,
GZ
- H,
say,
As
Z(t"
= 2 + n,
elements and so
Recall that
a;""
implies m = n. Then a"* = H is an infinite cyclic group.

ct"
\
g-p{<i)
has an
infinite
number
of
such G = gp{x) = {x" n any integer}. If there exists an integer m > = 1, then G will consist of only a finite number of elements (see the remarks prefor which a;"" = 1. ceding Lemma 4.5). Consequently if G is infinite, there exists no m t^ For we have already shown that there can exist no to > for which a; = 1; while if a;"' = l for to < 0, then a;^-"" = 1 and (-to) > 0. If x^ = x", n=l, then a;""' = 1. But this contradicts the condition that there exists no to such that a; = 1. Hence the elements of G are simply the powers a;" of x, and two such powers x"* and x" are equal if and only if m n. Now we can easily prove that two infinite cyclic groups are isomorphic. Let G = gp{x), H = gp{y) both be infinite cyclic groups. Then each element of G is uniquely of the form
that
Define and each element of is uniquely of the form i/", n an integer. a one-to-one onto mapping. Furthermore, (x"x'")S = (a;"+'")e = y"*"* and (x"e){x"'9) = y^y"" = 2/"+. Hence x"ex'"9 = (x''x)^. Therefore 9 is an isomorphism and G and are isomorphic groups.
X",
n an
integer,
e is
(x")^
I/",
Collecting our results,
we have proved
Theorem
4.7:
There exist cyclic groups of all orders, finite and infinite. Any two cyclic groups of the same order are isomorphic. (We therefore often talk about the cyclic group of order to, or the infinite cyclic group, or sometimes the
infinite cycle.)
an element of a group G, then we define the order of x as the order of gp{x). Note of order to < oo, then x"" = 1 and to is the first positive integer r for which x*" =1. If X is of infinite order, then x'" = 1 implies to = 0. If x is of order m, to < <, we say x is of finite order.
If
is
that
if
is
Lemma
4.8:
Let x be of order
w<
>.
If
x*"
1,
then to divides
r. x'.
Proof: Put r qm + s where integer greater than for which x"*
^s <
=
1,
m.
0.
Then 1 = x*" = x^^x' = Hence w divides r.
As
to is the first
Problems
4.21.
Prove that the additive group of integers

Solution:
is infinite cyclic.
Z=
4.22.
gp(l).
As Z
is infinite, it is infinite cyclic.
Prove that the group of Problem

Solution:
3.5,
page
53, is cyclic of order
m.
The group
is gp{i.),
and
its
order
is
m.
104
[CHAP.
4.23.
Prove that {Z,+) and the subgroup of M, the group of Mobius transformations, generated by the
mapping
Solution:
rj
^ z+1,
4.7 all
=07;
are isomorphic.
By Theorem
{Z,
we need prove
But since
;
is
that
srp(i))
ij
is infinite,
);
+)
is infinite cyclic.
= z + n,
as we know from Problem 4.21 that implies n = m. Thus gp{v) is infinite and
so
gpin)
(Z, +).
4.24.
Find the order

defined by
v
of
(i)
<t
z,
<^ri
= ^.
2
/I
3\
)
S3,
(ii)
<t
/I
(
4\
)
S4,
(iii)
the
map
i;
of
Solution:
(i)
a
er,
and
or2
=
1,
/I
f
3\ /I
3\
q/fo
o-*
q)~''
Thus
it
Hence
4.
!?
is
of order 2.
(ii)
ct2^ ff3
are not
but
i.
is
of order
(iii)
i;2
and so
v is of
order
2.
4.25.
Let G be abelian. Let x,y are co-prime, i.e. have no
G G be of orders r, s respectively. Show common prime divisors.
that ay
is
of order rs if r
and
Solution
Note that since G is abelian, (xy)" x^y" for any integer n. Since (xyY^ = a;"2/" = 1, the order of xy divides rs, by Lemma 4.8. If (xy)'^ = 1, i.e. x^y^ ~ 1, then x'" = 2/~" and 1 = (x'^y y-mr_ Therefore s, the order of y, divides mr. Since s does not divide r, s must divide m. Similarly is the order of xy, is divisible by rs and we can show r divides m. Hence rs divides m. So if also divides rs. Thus the order of xy is rs.
4.26.
Show
with
implies
< and s is co-prime to m, then a = 6* (a, 6 S G) that if G is a cyclic group of order a = b. Find a group G and a nonzero integer n such that there are two elements a,b e G a" = 6" but a = h.
Solution:
is abelian, so (ab~^Y = a^{b^^Y 1- Since G = gp(x) and the order of G is m, then for some r, and (x^Y = 1. Hence a;''^ 1 and divides rs. But s and are co-prime; 6. then w, divides r, say r = gw.. Now a6~i = x""" = 1 and so o
Since
ab~^
x^
In S3, \&t ^
= (^
^2
^
1
^^ 3/
, '
(^
1^1
^ 3
^\. Then 2/
o2
62
but
6.
4.27.
Show
that if
G=
gp{x)
and
is
of finite order r and s
is
co-prime to
r,
then
gp{x^)
G.
Solution:
The
distinct elements of gp(x^) are
X,x^,x^
a;'""!'"
where
(x)"
Since a"^ = 1 and G is of order r, r divides ns. r divides w. Hence there are at least r distinct elements in gp{x^). But as has only r elements, gp(x^) = G.
least such positive integer.
a;" = 1 and n is the As r and s are co-prime, G 3 rp(a;^) and G itself
4.28.
Find a group which

Solution:
is
not abelian.
(Hint.
Consider
S3.)
See Section 3.3a, page 57, where
we
pointed out that a^ri = tiOi.
Hence S3
is
not abelian.
4.29.
Prove that a subgroup

Solution:
H of S3 is
cyclic ii
H = S3.
H H
S3, then is survey of the subgroups of S3 shows that if i? is a subgroup of S3 and either cyclic of order 3 or cyclic of order 2 or cyclic of order 1. To obtain all the subgroups of S3, we refer to the multiplication table for S3 in Section 3.3a, page 57, and list all the subsets of S3. Then we check which subsets are subgroups. Of course since a subgroup must contain the identity,
there
this
is
no need to go through the process of finding

will suffice.
all
subgroups quite so crudely.
Nevertheless
method
Sec. 4.2]
CYCLIC GROUPS
105
4.30.
Prove that
Solution:
If (Q,
(Q,
+)
is
not cyclic.
+)
is cyclic,
Q.
there exists
m/n,
and n integers
Of course
m =
Each element
(= 0)
of
Q would
(n # 0), such that gp(mln) = Q. then be of the form q- q+ + q with some
suitable choice of the positive integer
r,
or else of the
form
qq
q.
But But r and

r terms in
l/2n
Q.
l/2n
+
r
implies
l/2n
rm/n,
i.e.
2rm;
then
2rm =
0.
m are integers; hence

all,
the equation 1 2rm = is not true. If l/2w = qqq--- q, a similar argument leads to a contradiction. Therefore (Q, +) is not cyclic.
4.31.
Prove that an abelian group generated by a

Solution:
finite
number
of elements of finite order is finite.
Let
the form
G=
9p({xi,
.,x}),
ji
<
g
>,
and suppose
xll
G
{X,.
is
abelian.
Then every element
fir
in
is of
xll
X,
e,.
1)
Since
is
abelian,
we can
rewrite g in the form
ji
if
/"
if
(yi, 72,
>
Y integers)
in
(i.S)
To
see this
we need only observe that

g
^ = = Xi
for s
Xi
<
the first expression for g, then
Xf
i.e.
we can always
1,2
n,
"collect" all occurrences of
any
in a product.
Now
if a;i,a;2'
finite order,
then the number of distinct elements given by

the distinct powers of
(i.S) is at
a;;
(4,5) is finite.
For
if k^ is
>n ^^^ ^H of the order of Kj,

distinct ele-
i-
are l,Xi,xf,
.,a;^~'.
Thus the number of
ments given by
most
fcifcg.
.fc,
and
so
is finite.
b.
Subgroups of
cyclic
groups
tions.
(i)
Before beginning the study of a new section it is a good idea to list the natural quesIf we want to know something about the subgroups of cyclic groups, we might ask:
(ii)
Are subgroups of cyclic groups cyclic? Does there exist a subgroup of any given order?
(iii)
How many
distinct
subgroups of a cyclic group
(less
than or equal to the order of the
group) are there?

(iv)
How many
subgroups of a given order are there?
We
tackle each of these questions.

4.9:
(i)
Theorem
be a subgroup of G = gp(x). is cyclic and either Then where x' is the least positive power of x which lies in or flfp(a;') = {1}. If the order of G is else < <x>, then l\ and the order of = {1}. is m/l. If the order of G is infinite, is infinite or
Let
H= H
(ii)
Conversely if of order m/l.
any positive integer dividing m, then S = gp{x'^) is Consequently there is a subgroup of order q for any q
is
that divides m.
(iii)
The number
There
is
of distinct subgroups of
is
the same as the
number
of
distinct divisors of
(iv)
m=
\G\
<
o.
at
most one subgroup of
of any given order for
finite.
106
Proof:
[CHAP.
(i)
If
H =
{1}, there exists "
i7.
As
i? is a subgroup, x""
e H. Now
one of
a;'
n,
is positive.
Hence we can
talk meaningfully about the smallest positive

x""
Clearly,
HdS = gp{x'^).
Suppose
i?;
then r
ql
+ s, 0^ s
power <l, and
n G H.
x'-ix^")-^
x"
GH
Thus
s
But and
a;'
is
the least positive power of x that belongs to H.
and so r
gt
{x")
(a;')'
S.
Hence S
is
=
1
H.
If the
order of
to
<
<,
then
m = ql +
=
a;""+^
s,
^s
G
<l.
if
Now
=
a;'
a;"
a;'"a;''
G H. Then s = 0, as is the least positive power of x that lies in H. Hence and w = Zg. Clearly (a;')" = 1, and q is the least positive integer for which this occurs. Then, by Lemma 4.5, writing a = x\ we have H = gp{a) = {a", aS a""'} = = m/l. If the order of G is infinite, all the powers of x are distinct, and hence
and so
a;*
I
divides
to,
|fl^|
gf
and so
(ii)
x', a;^',
... is
an
infinite set of distinct
elements of H.
Thus
is infinite.
Let
a"
l\in,l>0.
a;"
Put
m/l
and
a;'
a.
Then
S=
gpix^)
{l,a,
.,a''-^},
as
the least positive power of a which is 1 (for if a"' = 1, q' < q, then a;'' = 1 and Iq' < to, contradicting the fact that a; is the least positive power of x which is 1). Thus S is a subgroup of order q. Consequently if we start out with a positive integer q which divides to and we put I = m/q, then S is a subgroup of G of order q.
is
(iii)
.,lnhe the distinct divisors of to. Then put Hi = gv{x^'), ., H- gpix^'). m/U. These are n distinct subgroups of G (because their orders are |Hi| of G will have to different). Are there any more subgroups? By (i) any subgroup be generated by a;' where I is a positive integer dividing to. Hence I = k, say. There Hi. Thus the subgroups of G are simply Hi, H2, ., Hn, as desired. fore
Let
h, h,
We know
(iv)
If
H and K are two subgroups of G with

But \Hi\ i and j. H = Hi = Hj K.
\H\
\K\,
then
H and K are Hi and Hj of part

Since
\H\
above, for some therefore i = / and

(iii)
Tn/k, \Hj\
m/lj.
lîCI,
ii
J^
and
The reader will perceive that our knowledge of the cyclic groups is in some ways quite comprehensive. We know in the case of finite cyclic groups what the distinct subgroups In the case are, we know they are cyclic and we know which cyclic subgroups appear. of infinite cyclic groups we can easily prove there are an infinite number of subgroups. We will distinguish between them in Theorem 4.24, page 126, using the concept of index which will be introduced in Section 4.3b.
The reader might naturally be led to consider now groups generated by two elements, hoping that similar powerful conclusions can be obtained, e.g. that every subgroup of a two generator group is a two generator group. But in going from one to two generators we lose control. It has been shown that every countable group is a subgroup of a two generator group, so we can never hope for a simple account of two generator groups.
Problems
4.37.
subgroup
of a
order a prime p.
Solution:
group G Prove that
is
= G and called proper it has no proper subgroups.
H#
{1}.
Let
be a cyclic group of
We know
{1}
distinct divisors of p,
and
from Theorem 4.9 that the number of subgroups of G is the same as the number of which are p and 1. Hence the number of distinct subgroups of G is two. As itself are two distinct subgroups, the number of proper subgroups is zero.
Sec. 4.3]
COSETS
107
4.38.
Prove that the only groups which have no proper subgroups are the cyclic groups of order p and the group consisting of the identity alone.
Solution:
Let g ^ G, g = 1. Then S = gp{g) is a has no proper subgroups. Hence G is cyclic. If G is cyclic of order mn, m,n- 1, then, by Theorem 4.9, G has a subgroup of order m. But this is a proper subgroup. Hence G is cyclic of prime order or else possibly infinite cyclic, say G = = gp(x^) is a subgroup not equal to {1}, and not gp(x) = {. .,x-^,x-^,x'^,x^,x^, But .}. equal to G since a; S H. Hence G can only be cyclic of order p, a prime.
Let G be a group with no proper subgroups, G subgroup by Lemma 4.3. Since g ^ S, S = G as
= {1}.
4.39.
Find a group with two distinct subgroups both of the same order.
Solution :
\Hint. Consider S3.]
^^*
^'
"
(2
3)
'
"^^
"
(1
2)
"^^^^
'*'^^"'^'
"
'^^^^'^'
"
^-
4.40.
of
Find a group which is of infinite order but has a subgroup of Mobius transformations, Af, of Section 3.5, page 77.)
finite order.
{Hint.
Try the group
Solution:
Let
rf.
Z-*
1/z,
00
-> 00.
Then
gp{v) is of order 2, but
is infinite.
4.41.
Let // be a subgroup of G. Let g & G. Prove that the set S = {g~^hg h S H] is a subgroup -* S, defined by of G. Prove that s: he = g-^hg, is an isomorphism of onto S. If is a finite cyclic subgroup of G which contains both = S. (Hard.) and S, prove that
Solution:
Since
H -
0,
0.
Let g~^hig,g-^h^
S.
Then
(g-^h^gHg-^h^g)-^
because
he
= g-^higg-ih-^g = g-HKhî)g e S
is a subgroup implies hih^^ G H. Thus S is a subgroup, e is an onto map, since g-^hg. If hê = hê, then g-^h^g = g-^h^ff. Pre-multiply by g and post-multiply by g~^ g{g~^hig)g~^ g{g~^h^)g-'^. Hence h^ = h^ and so e is one-to-one onto. We need only check that is a homomorphism to conclude the proof:
hêhê
g~^hig - g-^hig
g^^hih^g
{h^Jiê
Thus
H and S are isomorphic. If X is a finite cyclic subgroup

|H|
since they are isomorphic, = S. order. Hence
|S|.
containing and S, then But, by Theorem 4.9,
and S are both of finite order; and has only one subgroup of any given
4JS
a.
COSETS
Introduction to the idea of coset
In this section we propose a natural question which introduces the idea of a coset. Cosets are important for other reasons: (i) With cosets we can perform useful counting arguments for finite groups, (ii) Cosets of a subgroup sometimes enable us to construct a
new group from an

subgroups
old. We can also see how a group G is built up from one of its and the group constructed from the cosets of H. (iii) The fundamental idea of a homomorphism can be re-interpreted in terms of the idea of a group constructed from
cosets.
the natural question we ask? In Section 3.4c, page 67, and Section 3.4e, defined the group / of isometries of the plane and the isometry group of a given figure S in E. Recall that an isometry <t of " belongs to Is if for each tGE, t<T G S implies t G S, and s G S implies sa G S. ask: which eleSuppose a Gl h.
is
What
73,
page
we
We
108
[CHAP.
ments 6 of I are such that Sd = Scrl In the case where S is an equilateral triangle we know that Sa is a congruent equilateral triangle by Lemma 3.6, page 71. So the question we are asking is this: which other elements 9 oi I send the equilateral triangle S onto the equilateral triangle Sal
and a related? If 9 i. If 9 and o- were widely different, we would expect 9(r~^ to be any 9<j~^. We will show that ^ is thing but I. Let an element of Is. To do so observe that ^ is an isometry of the plane E, since it is a product of isometries of E. Now S9 Sa implies that for each s &S there is a t &S such that s9 = ta, and
Sa.
Let S9
How
are
were equal
to
a,
then
9^"^
<j>
conversely.
Therefore for each
G
t
S,
s9o
(ta)c
and x<l> G S. We must show that a; G S to complete the Suppose now that x that ^ G h. Suppose x = tGS, i.e. x{9<t-^) = t. Now there exists an s G S such proof that ta-s9. Hence x9 = {{x9)a-^)(T = t<7 = s9 and so x9 = s9. As ^ is one-to-one, x = s. G.Is. Hence a; G S. This means
<j>
gE
We have of course 9 = ^<r. If we write Iâ = G Iâ. We have thus shown that every the form
in the set of isometries Iga.
Conversely
/^ct
if
t G h), we may put our deduction in isometry 9 of E satisfying S9 - S<r lies Iâ, then 9 - <j><t for some ^ G 7s; then
{t<t|
S9 =
9 of E for which S9 - Sa. isometries of E associated with the subgroup Is of I Such a subset /gtr of the group / of all More generally we have the following is called a right coset of Is in 7.
S<j)(j ~
Sa.
This means that
consists of all the isometries
Definition:
Let G be a group and 77 a subgroup of G. Then a right coset of 7f in G is a subset of the form Hg = {x\ x = hg, h G H} for some g in G. We define H}. inG to be a subset of the form gH {x\ x = gh, a left coset of
hG
Note that a coset

the left of
77.
is
a right or
left coset
according as the element g
is
on the right or
In the case where the group G operation, a right coset is written Problems
4.42.
is
written additively,
i.e.
H + g.
Of
course,
77
flr
+ is used to denote the binary = {x\ x = h + g, h G H}.
H in
= gp(aP). Find all right cosets of be the cyclic group of order 4 generated by {a}. Let G. Show that two cosets are either equal or else have no elements in common, and prove that the union of these cosets is G.
Let
Solution:
H-l =
Ha^ =
H=
-
{l,a^
{a^, a}
is
a right coset.
Ha =
=
{a, a^}
is
{a^, a^}
Ha. Thus the
distinct cosets of
a right coset. Hw^ = {a^, a*} jFiT in and Ha. G are
{1,
a^ = H.
HnHa =
4.43.
0, and
HuHa =
{\,a^,a,a^)
G.
Let
be the trivial subgroup of a group G,
i.e.
{1}.
Determine the distinct right cosets of
in G.
Solution:
If
G,
Hg =
{Ig}
{g}.
Thus the cosets are the
sets consisting of single elements of G.
Sec. 4.3]
COSETS
109
4.44.
= gpira) in Sg with the notation of Section 3.3a, page 57. Find the right cosets of = gp(ai) in S3? the right cosets of
What
are
Solution:
H=
K=
4.45.
(i,
T3),
Hi
H, Hgi
{ai, Tj),
Ha2
{<Ti, cTj. '}
The
cosets of
are Ki
= {c2> Ti)= K and
These are Ktj
all
the cosets of
in S3.
{tj, t2, T3}.
Let A,B, C be subsets of a group G. If X' and F are subsets of G, we define and y S F}. Prove that A(BC) = {AB)C. Hence conclude that if X e f,geG, then (i) (Jg)H = f(gH), (ii) H(/sr) = {Hf)g, (iii) (/H)fir = /(Hfr).
XY =
is
{g
xy,
a subset of G,
Solution:
Let X
6
GB
and
& A{BC); then c & C; hence

{g}
a;
a;
= =
ad where
a(6c)
a&A
(ii)
(ab)c
and d & BC. But d.BC implies d = &c, where S (AB)C and so A{BC)Q(AB)C. Similarly (AB)C C
A.(BC).
Therefore A{BC)
A =
4.46.
{/},
B=
and C
= (AB)C. = H.
(i),
and
(iii)
follow immediately,
e.g.
(i)
is
the case where
Let
be a group with a subset H.
Show that /H = Hf
implies
f'^H =
/f/-i.
Solution:
fH =
Hf.
Hence
f~HfH)
Thus H/-1
f-HHf),
(f-^f)H
= {f-mf
in
and
H=
(f-m)f
{(f-m)f)f--^
= f-m.
Problem
4.45.
Note that we have used the "associative law" proved
b.
Cosets form a partition.
Lagrange's Tiieorem
In the problems above it is clear that any two right (or left) cosets are either disjoint or inG is G. We recall exactly the same and that the union of all the right (or left) cosets oi that a family of subsets of a set G is a partition of G if they are disjoint and their union is G. The examples above point to the following:
Theorem
4.10:
Let
form a partition
H be a subgroup of a of H in
G
is
H
Proof:
since 1
It is
in
Then the right (left) cosets of i? in G G, i.e. the union of all the right (left) cosets of itself and any pair of distinct cosets has empty intersection.
group G.
(The proof for
easy to show that each element of G occurs in at least one right coset. and is not included here.) For if g GG, then g G Hg and ! g = g.
left cosets is similar
Suppose now that

exists g
Ha and Hb
G HanHb. Then g^h'a =

two elements of
are two cosets of in G and that Ha n Hb - 0, i.e. there h"b, h',h" G H. Hence c = h'-''h"b = h"'b, h'" G H,
since the product of
H belongs
\
to
H.
Therefore
as hh"'b
Ha ^ {ha\hGH} =
Similarly
{h{h"'b)
G H} c Hb
hib, hi
Hb c Ha.
Thus Ha = Hb and any two
cosets are either disjoint or identical.
In Section 4.3a
follows
we mentioned
4.10.
from Theorem
the cosets of i? in G each coset. We use this fact in proving
that cosets are useful for counting arguments; this a subgroup of G, then, since of finite order and are disjoint, the order of G is the sum of the number of elements in
in
(i)
If
is
Theorem
Proof:
4.11
(Lagrange's Theorem):
The order of a subgroup divides the order of G.
of a finite group
Let the distinct cosets of i? in

,^, \G\
be Hgi, Hg^,
,
.,
Hgn.
Since these form a

/
.
partition of G,
= ,..,,.. + \Hgi\ + \Hg2\
,.r
,^
\Hgn\
(^.-4)
110
is IHg]"!
[CHAP.
= hg is a oneWe will show that the mapping dg-. H -* Hg defined by = \Hg\. Clearly Og is onto by the definition of right mapping and hence coset. If hig h^g, then multiplying by fl'"' on the right we conclude that hi = hi and so hiOg = hÔg implies hi = hi. Therefore 9g is one-to-one and onto and \H\ = \Hg\. Thus for each i, \H\ = \Hgi\. Hence \G\ = n\H\ by (^.^).
What
to-one onto
|fl^|
Corollary
4.12:
Let
be a
\G\.
finite
group and g an element of

is
of order m.
Then
divides
Proof:
The order
4.13:
of g
is
the order of gv{9) which
a subgroup of G.
\G\.
Then by Theorem
4.11, \gv{g)\ divides \G\.
But
m=
\gpig)\.
Hence
divides
Corollary
If
is
of finite order n, and g
&G,
\G\
then g"
1.
Proof: Every element of by the preceding corollary, and the result follows.
Definition:
G must
be of
finite order.
divides \G\
and so
be of order m. Then, Let g qm. Hence ?'i = g""" = (5"*)" = 1
gG
Let the number of right cosets of it by [G:H].

:
H in G be called the index of H in G. H

\H\
Denote
Note that [G H] is read as "the index of and H appear in [G:H].

4.14:
in G",
i.e.
in the opposite order to
which
Corollary
If
is
finite
group,
4.11
i.e.
\G\
[G H].
:
Proof: In the proof of
Theorem
is
the
number
of cosets of
in G,
we conclude with "Hence n=[G:H], we have \G\ =
\G\
n\H\ ...".
Since
\H\
[G:H].
Problems
4.47.
Show
and
that:
g''
t,
S^ has no subgroup of order 11; = 1. then

(i)
fir
(ii)
D4 has no subgroup
of order 3;
(iii)
if
Sf
e A5
Solution:
(i)
\S^\
7!
32
24.
Theorem
there
(ii)
4.11, 11 divides IS7I.
But
11.
8,
in the
It H were a subgroup of S^ of order 11, then, by prime decomposition of [S;] there is no 11. Hence
is 8.
no subgroup of order
ID4I
=
g''
Since 3 does not divide

1,
Theorem
4.11 tells us there is
no subgroup of order
3.
(iii)
If
then either
ff
divides m.
As
m=
and
it
Now
by Corollary 4.12
is of order 7, since or s- = 1 implies that the order of g is that the order of g is 7. a prime, the only possibility if g would follow that 7 divides IA5I, which is not true. Hence g - i.
i
fif
is
4.48.
Prove that
Solution:
if
is
a group of prime order, then
is cyclic.
its
\G\
Hence is a subgroup of G. nothing to prove. If 1 ^ ff e G, gp(g) = divides the prime |G|. As \H\ # 1, |H| = |G| since the only divisors of order, by Theorem 4.11, = G, as H cG and and G have the same number of elements. are 1 and |G|. Thus
If
G=
{1},
there
is
4.49.
= gp{{v}) where v Give the right and left cosets of of Mobius transformations, Section 3.5a, page 77.
Solution:
If
<7(a, 6, c,
"(0, 1, 1, 0)
is
an element of M, the group
d)
G M,
then
Ha{a,b,c,d)
{i
'
<r(a,b,C,d), v<r{a,b,c,d)}
is
{<r(o, 6,
c,d), a(6,a,d,c)}
by the rule for multiplication which
obtained in Problem 3.46, page 79.

a, b)},
Now
ff(a, 6, c,
d)H
{a{a, b,
c, d), a(c, d,
as
is
easily checked.
Thus we know what the
right and left cosets are in terms of
a, b, c, d.
Sec. 4.3]
COSETS
111
c.
Normal subgroups
of G.
We discussed left and right cosets of a subgroup H in G. Each gives rise to a partition How do these partitions compare ? In particular, are they the same ? Sometimes yes.
In Problem 4.43, the right cosets are just the elements of G; the left cosets can similarly be shown to be just the elements of G. Sometimes no. In Problem 4.49 the right coset containing a(a, 6, c, d) also contains a{b, a, d, c). But the left coset containing <r{a, h, c, d) contains (t(c, d, a, b). With a suitable choice of a,b,c,d we can ensure that <T{b a, d, c) a{c, d, a, h), e.g. if a = & = c = 1, d = 0, then cr(l, 1, 0, 1) - u{l, 0, 1, 1). Thus the left and right cosets of in G do not coincide.
,
We ask: when do the right and left cosets of a subgroup H in a group G coincide? Suppose every left coset of H is also a right coset of H in G. Let a gG. aH contains a, as does Ha. Since the right cosets form a partition, the only right coset containing a is Ha. But we have assumed that there is some right coset which is the same as aH. Hence it must be Ha, and so aH Ha. In other words if every left coset of H in G is a right coset, then for every a e H we must have aH = Ha.
Proposition 4.15:
necessary and sufficient condition for the left cosets of provide the same partition as the right is that for each a G G,
if is
in G to aH Ha.
Proof:
Ha = aH
for all aGG. fore every right coset of
H in G is
If
every left coset is a right coset. Ha = aH. Let the set of all the right cosets of in G. Therefl^ in G is a left coset of in G. Similarly every left coset of a right coset of in G. This completes the proof.
We
have proved above that {Ha a G G}

\
a~^ha
If
Ha = aH for G H for each
if Hence Ha C aH. hi belonging to H, ha = ahi. aH and assuming x'^hx G H for all x G G and all h GH, then ah = ah{a~â) = {{a~^y^ha~^)a G Ha. Accordingly HaDaH and aH = Ha. Thus we have Proposition 4.16: aH = Ha for all aGG if and only if a'^ha G H for all hGH and
Conversely
aGG, then for GH. a^^ha hi, for some

all
each h
G H, ha =
ahi
for some
hiG H.
Hence
ah G
all
aGG.
if
Definition:
for
subgroup fi of a group G is normal (also called invariant) in G all g GG and all hGH. We write H < G and read it normal subgroup of G".
as:
g'^hgGH "H is a
(equiva-
By
lently,
Proposition 4.16,
is
normal
(ii)
in
if
and only
if
Hg - gH
for
all
GG
g-'Hg = H).
In Section 4.3a
cosets
we gave
in
and
(iii)
reasons for the importance of some cosets.
The
we had
in
mind are those arising from normal subgroups.
Problems
4.50.
Prove that every subgroup of an abelian group

Solution:
is
a normal subgroup.
Let
be abelian and
is
abelian,
g^^hg S H.
H any subgroup of G. Then ii g G G and h S H, g~^hg = h; for since gh = hg and hence multiplying by g~^ on the left, h = g~^hg. Then if h & H, Thus H is a normal subgroup of G.
<]
4.51.
Prove that A
Solution:
S for each positive integer
n.
T~i
T~âT e A? Now t is either odd or even. If t is even, then t and A. If r is odd, then t~i is also odd, and hence T~'â is odd, for a G A.. Since T~â and t are odd, their product T~i<rr is even. Hence t~1o't e A. Note that we used
Let a e A, G A and so
3.2,
S. t~i<7t
Is
Lemma
page
62.
112
Let
that
[CHAP. 4
4.52.
H be
K
Let
<1
a finite cyclic subgroup of G, and let G. (Hard.)
H <i
G.
Let if be a proper subgroup of H.
Prove
Solution:
Note that if y is of order m, and g G G, then g~^yg is also of order w, since l and (g~^ygY = l implies g^^y^g = 1 and therefore y^ 1. Hence m divides r, and the order of g~'^yg is w. Since i? <1 G, g~^yg SH. Therefore gp{g~^yg) is a subgroup of of order m. Then by Theorem 4.9(iv), gpig'^yg) = X. In particular, g~^kg G K for any k G G. Hence X <1 G.
(g'^yg)"^
K = gp{y).
g~^V^g =
d.
Commutator subgroups,
centralizers, normalizers
We
1.
will
is
now
introduce some subgroups which are normal.
If
a group,
we
define the center of G, denoted

{z\ z
by Z{G),
gz
to be
GG
and for
all
&G,
zg)
Z{G) turns out to be a normal subgroup of

2.
(see
problems below).
briefly,
a group and x,y &G, then x~^y~^xy is called the commutator of x and y or, more a commutator. We often write [x,y] for the commutator a;~ii/">a;i/. The subgroup of G generated by all commutators is called the commutator subgroup (also called the derived group) of G and is denoted by G'. Again G' turns out to be normal in G.
If
is
Proceeding along somewhat different

1.
lines, let
be a subset of a group G.
The
centralizer C{A) of
(in
G)
is
defined
by
all
C{A)
{c\
cGG
and for
G A,
ca
= ac}
in
2.
C(A) is a subgroup of G (see problems below). C{A) (see problems below).

A'^(A)
If
is
an abelian subgroup,
is
normal
The normalizer
of
in
is
defined
by
and
An = nA} N{A) is a subgroup of G and, if A is a subgroup of G, A is normal in N{A). Furthermore, These facts will be if A is a subgroup of G, A is normal in G if and only if N{A) = G.
N{A) =
{n\
nGG
proved in the problems below.
The
details concerning the
In Chapter 5
Problems
4.53.
we
will use the concepts
groups Z{G), G', C{A), N{A) appear in the problems below. we have just introduced.
Prove that the inverse of a commutator

Solution:
[x,y]
is
a commutator.
x-^y-^xy
z,
say.
So z-^
= y-^x'^yx =
[y,x].
4.54.
Prove that G'

Solution:
is
normal
in G.
We
h
must show that

g-^hg
if
GG
and
G',
then
g'^hg G
G'.
If
/i
is
a commutator, say
x~^y~^xy, then
g-^x-^gg-^y~^gg-^xgg-^yg
= =
x^'^y^'^x-^y^
[xi,yi]
where
x^
g~^xg
(consequently a;->
g^^x~^g) and
j/i
g~^yg.
any element h of G' is a product of commutators and their inverses; and as an inverse of c^ of commutators. Therefore a commutator is a commutator, every element h of G' is a product Ci
Now
g-ihg
where
d;
= g-Hci---Ck)g = g-ê^gg-^c^
dj is
g-^c^g
did2
d^
g~^Cig.
But we have just shown that
a commutator.
Hence
if
hGG', g~^hgGG'.
Sec. 4.3]
COSETS
G
abelian
113
4.55.
Show
that
:
is
if
and only
if
G'
{1}.
G is abelian and that x,y G G. As a; and y commute (i.e. xy = yx), [x,y] = x'^y-^xy = Then G' is the subgroup of G generated by 1, and G' = 1. Now if G' - {1}, then in x-'^x particular any commutator [x,y] = x'^y-'^xy - 1. Hence x(x-'y-^xy) = x and y(y-^xy) = yx, xy = xy. Thus G is abelian. i.e.
Suppose
Solution
1.
4.56.
Show
that every element in
A3
is
a commutator of elements in S3. Hence show that S3
Ag.
Solution:
We
ments
use the table of Section 3.3a, page 57.
tj
t~'.
t^'ct^^ticti
ti<t2Ti<ti
aâ^
02-
r-i<T-iTi<T2
"i^z in S3 (A3 is listed in

'jffiTiCTa
A3, then A3 the result.
S3'.
This
is
Thus every element of A3 is a commutator of eleProblem 3.23, page 63). If we can show that all commutators belong to a matter of trying all possibilities, e.g. T-'r-iTiTg = TiTariTj = cti. Hence
oi-
T-it-rii
I.
4.57.
Show
finite cyclic
of Section 3.5a, page 77, that the commutator subgroup of subgroup generated by a commutator.]
is infinite.
[Hint.
Find an
in-
Solution:
[a(2, 0, 0, 1), a(l, 1, 0,
for each n.
D] = (t(1, -1, 0, 1) = a, say. Now gp(a) is infinite cyclic, as a" contains fl'p(a), it is infinite. Since the commutator subgroup of
(1,
-w,
0, 1)
4.58.
Prove that
Solution:
1
if
is
any group, Z{G)
is
a normal subgroup of G.
Z(G),
since
Ig
g\
for
all
g&G.
Consequently
Z(G)
0.
If
9-1,
02
e Z{G)
and
g&G, then g^g^g-^) {ggi)g-^ = gi(gg~^) = diff^^^ since gg2 = g2ff implies g-^g-gg-^. It = hg and so ^ = g-^hg. e Z(G), then follows that Z{G) is a subgroup of G. If g e G and S 2'(G). Thus Z{G) is normal in G. S G and all Hence g-^hg e Z(G) for all
=
/i
fiffe
fl^
/i
4.59.
Show
1
that C(A)
is
a subgroup of
and,
if
is
an abelian subgroup of G,
A<
C(A).
Solution:
e C(A) and so C(A) - 0. If g^, gz S C(A), and a A, then f2a = affa and hence ^^ affifl-a"' = ffiÂ'a"' = S'iS'2"^" ^'^'^ ^^ ffiff^^ ^ C(A) if aâ"' = fl'2'''*' i-^- ffa"^ ^ <^(^)S C(A). Therefore C(A) is a subgroup of G. If A is an abelian subgroup, then each o G A ^1,^2 belongs to C(A). Now ii g & C(A), then for each a G A, ga - ag, i.e. fir-las' = a G A. Accordingly A < C(A).
4.60.
Show
that
if
A
<l
is
of G, then
A ^
a subset of G, then N(A) if and only if N{A) = G.
is
a subgroup of G.
Show
that
if
is
a subgroup
Solution:
Ar(A)
0,
since
e N(A).
Let
gA = Ag
and
/A = A/
implies
Hence /, G iV(A) implies fg'^ G subgroup, A c N(A) and A < iV(A). If A is a subgroup and A < G, then for each f G G, gA = Ag. Hence g G Ar(A) G c ]V(A). Therefore G = Ar(A). If A is a subgroup and N(A) = G, then since A < ;V(A),
Find
f,gGN(A). Using the results of Problems 4.45 and 4.46, (fg-^}A = f(g-iA) = /(Afl--!) = (fA)g~i = (A/)5r-i = A(fg-i). iV(A). Therefore iV(A) is a subgroup of G. Clearly, if A is a
and so
A <
G.
4.61.
all
normal subgroups of
S3.
Solution:
We use the notation and multiplication table in page 57. Clearly {i} and S3 are both normal subgroups of S3. There are no normal subgroups of S3 containing elements of order 2 except S3. The elements of order 2 in S3 are, as we readily check by using the multiplication table for S3, ti, t2 and T3. Suppose for example that a normal subgroup of S3 contains tj; then iT]r'''i<î = <'2''i<'i = r2 & N. Similarly T3 G A^. Hence titj = (72 G
and
or|
CTi
G N, and
so
it
follows in this
way
that
N=
S3.
114
[CHAP.
a normal subgroup of S3, then if A^ contains elements of order 2, then must contain elements of order 3 (there are only elements of order 1, 2, or 3 in S3). Now aj and aj = af are the only elements of S3 of order 3. In fact {1,01,02} is a normal subgroup of S3. For example, t-ôiti = ti<titi = 02 B {i,<ti, crj}. Accordingly S3 has precisely three distinct normal subgroups.
if iV Is
We
S3.
N=
have shown that Therefore if N
= {1},
4.62.
Show
that
if
is
a subgroup of
G and B <
G, then
AB =
Solution:
{x
ab,
AB is ae A,
then
-1
a subgroup of G, where
b
e B}
AB '
bi&B.
0,
as
= I'le AB.
,
If
gi,g2&AB,
,_, _, "2
g^
a^b^,
g^S
B)
a^b^
where aiG
and
Now
S'iS'2
~ =
'î^'iâ
i''302
(where 63
aia'âjftga^*
ab,
(Iia2""'(a2"*)-63(a2"')
=
where a-a-fl.~ê.A, and group of G.
4.63.
say,
aafega^"'
eB
as
BOG.
is
Thus g^g'^ &
AB
and
AB
is
a sub-
Show
that the intersection of two normal subgroups
a normal subgroup.
Solution:
We
subgroups
cGHnK K < G. Thus

e.
refer the reader to Problem 3.15, page 55, in which we proved that the intersection of two is a subgroup. If H,K are normal subgroups of G, then is a subgroup of G. If and g&G, then g^^cg . and as c E. < G, and g~^cg e as c G and
HnK
g-^cg
GHnK,
and
H so WnX
is
normal
in G.
Factor groups
In Section 4.3a we mentioned that the concept of a coset sometimes gives rise to a group. This occurs when, and only when, the group is normal.
new
Let G be a group and N <i G. Let us denote by G/N (read as "G over N", or "G factor N", or "the factor group of G by N") the set of right cosets of N in G. We turn G/N into a groupoid by defining a binary operation as follows.
Na and Nb to be the coset Nab. This definition of multiBut it is conceivable that if Na = Nai and Nb = Nbi, that Naibi - Nab. In such case, what would we take for the "product" of the two cosets, Naibi or Nab ? What we must show is that the product of two cosets is uniquely defined by the formula NaNb = Nab when N < G. If Nai Na and Nbi = Nb, then ai = na for some and 61 = mb for some m G N. Accordingly, n
Define a product of two cosets
plication depends on a
and
b.
EN
= namb n{ama~^)ab = lab n(ama'^). Since N < G, ama~^ G N and hence IGN. Thus aibiGNab. where Since the cosets form a partition, it follows that Naibi = Nab. But this is just what we
ai&i
l
wanted to prove.
Thus G/N with
this binary operation is a groupoid.
Is it
an associative groupoid?
{{Na){Nb)){Nc)
= {Nab)Nc = =
N{ab)c
Na{bc)
(as
is
associative)
{Na){Nbc)
{Na){{Nb){Nc))
and so G/N
is
an associative groupoid.
Theorem
Proof:
4.17:
G/N
is
a group.
of
The mapping
onto G/N.
G -* G/N
defined by gv
Ng
is
a homo-
morphism
First,
an associative groupoid. Nl = N is an identity, for Na Nl = N{a !) = Na and N-Na N{1 a) = Na. Next, each element Na has an inverse, for NaNa-^ - N(aa-^) = N while Na-^Na = Nia-â) = N. Thus G/2V is a group.
G/N
is
Clearly v is a mapping of G onto G/N. Since (g^g^v = N{g^g^) - Ng^g^, (g^g^v = {g^v){g2v) and hence v is a homomorphism. NffiNg^
and
{giv){g2v)
Sec. 4.3]
COSETS
called the natural
115
V is
homomorphism
of
onto
its
factor group G/N.
Note that in the case where G is a group whose binary operation is +, as we remarked at the end of Section 4.3a, the elements of G/N are of the form N + g. Instead of using the multiplicative notation for G/N, we use additive notation. Our definition of the product of two cosets is written as the sum: {N + gi) + {N + g2) = N + {gi + g^).
Problems
4.64.
Prove that ZIE

Solution:
C^,
where
is
the set of even integers and C2
is
the cyclic group of order
2.
As Z
form
is
abelian,
JE7
is
a normal subgroup of
Z and Z/E makes
sense.
coset of
E -\-z, z&Z. E + Q
+
1)})
consists of all the even integers,
E -\-l
2.
oi all the odd.
is of the Since an integer is
either odd or even, these are all the cosets.
gp({(E
+ 1)})
order
contains
E+\
and {E
+ 1) + {E + V) =
E,
and so gp(,{{E
ZIE. Thus ZIE
is cyclic of
Hence the
result.
4.65.
Is QIZ s Q, where ments of Q/Z.)
is
the additive group of rationals?
{Hint.
Examine the order
of the ele-
Solution:
coset of
in
is
of the
form Z +
q,
Q.
As
q (Z
min
for some integers m, n,
it
follows
that nq
Z.
Then
{Z+q) +
{Z
+ q)+
in all
+ q) = Z
n terms
and therefore Z + q is oi finite order in QIZ. Thus every element of QIZ is of finite order. On the other hand no element other than of Q is of finite order, for nq = if q = and n = 0. It is this fact that we will utilize to prove that QIZ is not isomorphic to Q. Suppose it were and that QIZ - Q was such an isomorphism. Choose q & Q {q = 0); then there exists an element Z + r e of QIZ such that (Z + r)e = q. Now Z + r is oi finite order n, say. Then {n(Z + r))$ nq and now n(Z + r) = Z. Since Z is the identity of QIZ and e is an isomorphism, e takes Z to the identity of Q, namely 0; hence nq = 0. But q = 0, so q # 0. This contradiction proves that there exists no isomorphism of QIZ onto Q.
:
4.66.
Find Sg/Sg.
Solution:
Prom Problem
^^sTi
4.56,
1S3
A3.
{ti, T2, T3}.
Thus S3/S3
consists of these
Hence the cosets of S3/S3 are A3 = A31 = {i, <ri, two cosets. The multiplication table is
is'-i
ffg}
and
A3
As
^STl
A3T1
A 3^-1
A3
4.67.
Prove that
Solution:
if
G=
H, then G'
= H' and
Z{G)
Z(H).
Let e be an isomorphism from G to H. both one-to-one and a homomorphism.
Then
-^
e^Q
is
a monomorphism into H, since e
is
What
Cj is
inverse of a commutator
The elements of G' are products of commutators and their is a commutator, each element of G' is of the form Cj c^*. a commutator. Then the images of G' under * are of the form Cj*
is G''*?
inverses.

As the
c^,
where each
If
[a, 6]
a~i6~io6,
G'-ir
then
c*
{ae)~Hb0)~âebe
[ae,be]
Thus the e^^ are commutators and so
c H'.
116
[CHAP. 4
To show G'* = H', we need only show that all possible commutators [h^, h^ are images under *. As e is onto, there exists 9^,92 such that gê = h^, gê = h^. Hence [f/i.ffj* = ['ii.'i2]- Therefore G'* = H' and consequently G' = H'.
Let = 6i|z(c). Then is a monomorphism of Z{G) into H, and we need only show that it maps onto Z{H) to prove Z{G) = Z(H). Let z S Z(G). For each there exists ^ G with ge = h. Hence zeh = zege = (zg)e = (gz)e = geze = hzs and so Z(G)^QZ{H). Let x e Z(H). As 9 is onto, there exists 2/ G G with 2/ = : Let S G. Then {gy)e = geye - yege - (yg)e. As e is one-to-
heH
is
fl^
yg. result follows.
one,
gy
Hence y e Z(G), and every element of Z{H)
an image of an element of Z{G). The
4.68.
(i)
If
is
abelian
show that A/A' = A.

is
(ii)
(iii)
Show Show
that G/G'
abelian.
that
G/N
abelian implies
N 2 G'.
Solution:
(i)
A'
x-^y~'^xy
the subgroup generated by the commutators x-^y-^xy, x,y & A. As A is abelian, y-^x-^xy = y^^ly = 1 and A' is the subgroup generated by 1. Since any product of 1 and its inverse is again 1, A' = {!}. Let v be the natural homomorphism of A to
is
A/A'. To show
i.e.
V is
an isomorphism we need only show that

i-e.
it is
one-to-one.
Suppose a^v
a2v,
{l}ai
{l}a2.
Wi}
{a2i-
Then of course
Oj
ttg,
so v is one-to-one.
Thus
c is
an
isomorphism.
(ii)
Let G'x and G'y be two elements of GIG'. Then G'xG'y - G'xy while G'yG'x = G'yx. Now (x-^)'^(y~^)'-^x-^y~^ = ajj/a;~ij/-i G G', and so G'j/a; contains the element xyx~^y^^ yx = xy. But G'xy contains xy; hence G'xy = G'2/a;. Therefore (G'x)(G'y) = {G'y)(G'x) and G/G' is
abelian.
(iii)
We
need only show that contains every commutator. For then generated by the commutators, which is G' by definition.
If
contains the subgroup
GIN
is
xy
= yxn
where
abelian and x, y are any elements of G, {xy)N = xNyN = yNxN {yx)N. Hence n G N. Multiplying on the left by j/-i and then by x~^, we obtain
4.69.
Show that G, then

Solution:
if H H< G
is
a subgroup of G containing and GIH is cyclic of order 2.
G',
then
H<
G.
Show
that
if
is
of index 2 in
Let h thus to H.
If
and consider g~^hg. Therefore g-^hgh-^ h
Now g'~^hgh~^ is = g-^hg GH and

where
a commutator, and hence belongs to G' and
H<
0.
G.
now
jy is of index 2,
G = HuHg
HnHg =
k~^hk
Let
hGH.
If
fe
G,
then
hi
or k
h^g {h^
H).
Hence
k-^hk
h^^hhi G
or
= g-^h^^hhig =
g~^h2g
where h^GH. If g-'^h^gGH, we are through. Otherwise g~^h^GHg, so that g~^h2g = h'g {h' G H) and thus gr-i/ig = A'. Hence g=:h2h~^GH. But this contradicts the assumption HnHg = 0. Thus we are forced to conclude that fe-iJifc G H for every hG H and every k G G, i.e. that H < G. The two cosets oi in G are f/ and Hi/ for some g H. Accordingly gp(Hg) GIH and so G/H is cyclic of order 2.
4.70.
Show by considering a suitable non-normal subgroup be defined as on page 114 without ambiguity.
Solution:
of S3 that the product of two cosets cannot
use the notation of Section 3.3a, page 57, in dealing with = gpiri) = {I.tJ. Ha^ = {<ti,t3} = Ht^ while we defined the product of the coset Ha^ and Ha^ as Haô^ H. Ht2 would be Ht^t^. -^"i '^ ^- This means that the product of
3.
We
Let us take
Ha2
the symmetric group S3 of degree = {a2,T2i Hr^- In page 114 However, the product of Ht^ and
cosets is not uniquely defined.
Sec. 4.4]
117
4.4
a.
Homomorphisms and factor groups: The homomorphism theorem We now consider the connection between homomorphisms and factor groups. We have already established in Theorem 4.17 that corresponding to every factor group G/N there What about the converse? Suppose is a homomorphism v G -* G/N such that Gv = G/N. now that 9 is a homomorphism of G into a group H. We ask: is there a normal subgroup N of G such that G/N = Ge1 Let us define the
:
kernel of 9 (denoted Ker6i) by Ker6' do the trick.
{g
&G,
gO
1).
We
will
show that Ker6i
will
Theorem
4.18
(Homomorphism Theorem, also called the First Isomorphism Theorem): e:G-*H is a homomorphism of a group G into a group H, then N = Ker 9 is a normal subgroup of G, and r,:ge^Ng defines an isomorphism of Gd onto G/N.
If
Proof:
First
we
will
show that
AT^
is
a subgroup.
If
gi, g2
(9,92^ =
and so gig-^ G N.
{9ie){g^'e)
igie)ig2e)-'
G N, = 1.1 =
then
1
is
To prove
g-^ng
Also IGN, so ^ 0, the empty set. Thus iV and g gG. a normal subgroup of G, let N; this will hold if {g-^ng)e = 1. But
that
it is
nGN
We
=
a subgroup of G. must show that
(9~'ng)e
(ge)-\ne){g9)
{ge)-n{ge)
{ge)-\g6)
Hence g^^ng G
exist
y^
and
iV is a
normal subgroup of G.
rj:
Next we must show that

fifi
g9
^ Ng
defines a mapping.
It is
conceivable that there

rj
is not a mapping as For if not, g26. We ask: is Ngi - Ng2l it is not uniquely defined. Now gig^'^ G N, for {gig2^)0 - (g\e){g26)~^ = 1 as gxO = g2e. = ng2 where nGN. Thus Ngi and Ng2 have in common and, Hence gWz^ G N and
g2 with giO
gfi
gfi
as the right cosets form a partition, Ngi

Is
it
Ng2.
Consequently
1;
is
a well defined mapping.
homomorphism?
{gx9g2e)-n
{{gig2)e)r,
N{gig2)
= NgiNgz =
{gie)r,{g29)rj
and so
77
is
a homomorphism.
is
1;
Finally,
one-to-one?
If
{gi9)T)
{g29)rj,
then Ngi
nGN,
Problems
4.71.
and
g29
{ngi)9
n9gi9 =
l'gi9
gi9.
Thus
17
is
Ng2. Then ngi - fif2 for some one-to-one and hence is an
isomorphism.
Let e be the homomorphism of Z into the multiplicative group of nonzero rational numbers defined by a;e = 1 if a; is even, and xe = 1 if a; is odd. Find the kernel of e and examine the claim G/(Ker e) = Ge.
Solution:
{x\ X is even} and G/(Ker e) = {Ker e, Ker e + 1}, so Ker e + 1 (Ker + 1) + (Ker e + 1) = Kerff; hence Ker9 + 1 is of order 2 and G/(Ker e) is cyclic of order 2. Now Ge = {1, -1}. Ge = gp{{l}) and, since (1) (-1) = 1, Ge is also cyclic of order 2. Therefore G/(Ker e) and Ge are isomorphic by Theorem 4.7, page 103.
Ker e = {x\ xe generates G/(Ker).
1}
We
have
fl
4.72.
Check that the kernel of the natural homomorphism of the additive group of integers Z/2Z is 2Z, where 2Z = {x x = 2z, z G Z}, i.e. 2Z is the set of even integers.
\
(Z,
+) onto
Solution:
The natural homomorphism v is defined by zp 2Z + z. The identity of Z/2Z Consequently z e Ker c if and only if 2Z + z = 2Z and hence if and only if
is
the coset 2Z. S 2Z. Thus
Ker f
2Z.
118
[CHAP. 4
4.73.
Verify that if G is any group, the subgroup generated by the squares of the elements of G, elements of the form gg = g^, is a normal subgroup of G.
Solution
i.e.
is
Let S denote the subgroup generated by the squares of the elements of G. Let a; G S. Then x a product s-fi^ s^ of elements of G each of which is a square or the inverse of a square. Since the inverse of a square is also a square, we may assume each Si, s^. is a square. If g S G, then
' . . . ,
g-^xg
where
tj
g-Higg-^s^g
tj
g-^s^g
tit^-'-t^
Sj
g~^Sig.
We
assert that each
is
a square.
Since
<i
r? for some
rj,
k and hence the

tj
g~^Sig
g-'^r^gg-^Tig
(g-îg)^
G.
are squares as asserted and g~^xg
S.
Thus S
4.74.
Let a be the homomorphism of Q*, the multiplicative group of nonzero rationals into Q* defined by xa = \x\. Find the kernel and image of a. Verify the homomorphism theorem directly in this case.
Solution:
(r
is
a homomorphism, since (xy)a
Ker<T
Q*<T
{x\ xa
{
a;
I
l}
3/
=
V
J/0-,
= {x \x\ e Q*} = {x
\
= \xy\ \x\ \y\ = = 1} = {1,-1}.

\
(xa)(ya).
Q*, x
is
positive}
Let
be the mapping of Q*/(Ker(7) into Q*a that takes
(Ker a)q
->
qa
is
{q,~q}, and so the only other "representation" for the coset (Kera)g Hence the other possibility for c as far as (Ker <7)g is concerned is that ((Ker a)q)i' ( g)(T = \q\ qa. Thus is a well defined mapping of G/Kera into Q*<t. Since
Now
(Ker a)q
(Keri7)( g).
{q)(r.
But
i-
[(Ker ir)gi' (Ker
(7)92]''
= =
((Ker
a)giq'2)''
kiggl
=
then
19111921
((KerCT)gi)j''((Ker(T)q'2)'
is
a homomorphism.
Iîl
Is one-to-one?
If
(Ker a)qip
(Ker
0)92",
we have
that
ir
i.e.
and
it
Thus if gj ^ g2> 9i ~Q2- Therefore (Kera)gi follows that Q*/(Ker <r) = Q*a.
Igji-
(Kera)g2.
Hence
is
qâ = q2<r, one-to-one
4.75.
Let G be the group of mappings of the real line R onto itself of the form aa,^: x -* ax + b, a =f^ 0, a,b real numbers, x G R. Prove that the map * ao,b ~* <*a,o is a homomorphism of G into G. Find the kernel and the image of this homomorphism and exhibit the isomorphism described in the homomorphism theorem.
:
Solution:
Then and ( bW. d = c(^ + + aa,bc,d = oe, bc+d' î'l^ _{, = aa; + b = Kc.bc+J = "ac.O' (<â.b)i<^c.d) = a,oc,o = ac,0' <! SO 9 is a homomorphism. If (a,bc,d) 6 any real the identity mapping. Hence {aj a(,eKerfl, we have a^^ff = ajo as a^^ = number} = Ker. The image of e = {ao,o a any nonzero real number}.
Note that
'')
<*
c,
[,
typical coset
is
(Ker
)(.
(j
=
= =
^"i.b
{(.
^ +d
^
I
any real number} ttc,

^
(J
bc
I
any
real
number}
{c.e
any real number}

is
The isomorphism
17
between G/{Ker
e)
and Ge as given by the homomorphism theorem
the one that
takes the coset (Ker e)a^,i to ac.o-
4.76.
Prove that if G is cyclic of order n and p divides n, then there group of order p. What is the kernel of this homomorphism?
Solution:
is
a homomorphism of
onto a cyclic
Let
the
G =
gp(x)
and
let
n = pm. Let
be cyclic of order p,
H = gp(y). We
define e to be
mapping
Sec. 4.4]
as
119
e is well defined
we
established in
let
if
i i
Lemma
n.
4.6 that the elements
a;',
n 1,
are
all
e
the dis-
tinct elements of G.
i
+ j n~l
while
Now = 1
and
be less than n.
We
have
(*')
(* +'-'")
where
if
+j ~
Then
{x'Z')e
=
x'
yi + i-^ri
yiyiy-en
Since the order of j/ divides n, j/"" = 1. phism. The kernel of 9 is the set of all = 1 if and only if p divides i, J/'
Hence
(x'x')e
2/'2/^
(x*0){x'e)
and so
1.
is
a homomor-
such that
x'9
=
.
1,
w =
Since
a;'fl
y*
and
Kere =
{a;*
|
p divides
i}
{x",
x'^",
.,
oj^^-Dp}
flrp(xP)
4.77.
Let (R'^
be the multiplicative group of positive real numbers.

:
into {R, +), the additive is the kernel of el
Prove that the mapping of i2 + group of real numbers defined by e x ^ logjoX, is a homomorphism. What What is the image? Using the homomorphism theorem, prove that
{R+,')^{R,+).
Solution
$ is certainly
a mapping of jR+ into R.

(xy)e
Furthermore,
logi(xy)
logioa;
+
a;
logic?/
x9
ye
and so
9 is
a homomorphism.
Kerff
=
is
{a;
I
logiox
0,
G K+} =
{1}
We
assert that
10":
R'ê
x.
R.
if
To
y
see this observe that if x is
Moreover, log
do
Then
any element of R, IQê
log 10"
any real number, then 10^ e fl+. = y and hence R+e = R.
The homomorphism theorem states that R + /(Ker e) = iZ+s = R. As Kere = {1}, all we must show that ft+/{l} s JJ + We can indeed show this in general: if G is any group, G/{1} = G. To do this we exhibit the isomorphism. Let r be the natural homomorphism of <? onto G/{1}. Then we need only show that n is one-to-one. Suppose ffiv = g^v, i.e. {l}fl'i = {l}fl'2. Then {gi} = {g^ and consequently gi g^- Hence v is one-to-one and thus an isomorphism. Accordingly R+ = +/{!} = R and so jB+ = R.
is
.
4.78.
Prove that the mapping 9 a; -> group of positive real numbers.

:
e*
defines
an isomorphism of (R,+) onto (R*
),
the multiplicative
Solution:
and so 9 is a homomorphism. If xe ye, then e^ = e and and x = y, so 9 is one-to-one. Is e onto? Yes, for if y is any positive real number, the equation e^ = y has a solution x G R. Thus e is an isomorphism between {R,+) and (B + ,').
(x-'ry)e
e^+w
eêw
(xe)(ye)
ex-y
1^
from which x y
4.79.
Prove that [fg,a] = g-^[f,a]g[g,a] for any /, jr and a in a given group. Suppose center of G. Let a be a fixed element of G. Prove that the mapping e g -* [g,a] phism of G into G. WhatisKers? (Diflicult.)
:
is
G'cZ(G), the a homomor-
Solution:
Observe that [fg,a]
g-^f~â-^fga.
9~'^[f,a]g{9,a\
On
the other hand,

g~^(f''â,~^fa')g(,g-â-^ga)
= =
g-if-ia-^f{agg-â-^)ga
g-^f-â-ifga
[fg,a]
[fg,a]
G'qZ(G), then [f,a]eZ{G) and g-'^[f,a]g = [f,a.]g~ig = [f,a]. Therefore Hence (fg)e = fege and e is a homomorphism of G into G. Kertf = {g g G G}. Thus Ker e = C(gp{a)), the centralizer in G of gp{a).
If
=
1,
[f,a]{g,a].
[g,a]
4.80.
Prove that
Solution:
if
G -* K
is
a homomorphism and
|G|
<
then \Ge\ divides
\G\.
By
the
homomorphism theorem, G/(Ker

Therefore
\Ge\
e)
s
'
Ge.
By Lagrange's
theorem, IKer
g\
divides the
order of G.
IGl
divides IGl.
|Ker|
120
Show
that the group
[CHAP. 4
4.81.
M of Mobius transformations
11
is
a homomorphic image of the group
a,b,c,d complex numbers, ad
be
Find the kernel of the homomorphism.

Solution:
In Problem 3.48, page 80,
we showed
\
M
We
define the
=
fi:
{a{a,b,c,d)
a,b,c,d complex numbers,
adbc =
/x
1}
mapping
TA -^
hy
3.46,
/a
I
c\
j
(r(a, b, c, d).
Clearly
is
a mapping.
Further-
more, using the results of Problem

1
page
79,
CiWttg
C2
\6ia2
+ di62 + Ci62>
6iC2
byd-i
did2/
a'iC2
a(aja2
+ djb2>
+ Ci(Z2, =
(
biC2
+ d^d2)
,
)
"{O'x, bi, Cj,
di)a(a2, 62, Cg, ^2)
\bi
,)/'( dj \b2 dj
(
,
Thus
and
a
/I
is
a homomorphism. Kernel
d)
/'="i(
3.49,
i/'(
M)
'1
if
jf' t)ecause
/i
=
6
(T(a, 6, c,
d)
CT(a, b, c,
(t(1, 0, 0,
1)
0,
(the identity of
and only
if
and
or
= 1
and
by Problem
page
80.
Therefore
0\
/-I
0'
"
'"'
-0
l/'V
-1
b.
Correspondence Theorem.
Factor of a factor theorem
Let 0:G-^K be a homomorphism of subgroup of JT. If H is normal in G, He
G
is
onto
iiT.
If
fl^
is
a subgroup of G, then ffS
is
normal
in K.
(See Problem 4.82 below.)
the preimage of a subgroup S of K, gO S}, be a subgroup of G? We know that if S = {1}, then the {g g preimage of S is Ker9, and this is indeed a subgroup of G. We generalize this result in the following theorem. the reverse of this procedure?
What about
\
Would
i.e.
the set
&G,
G.
Theorem
4.19
(Correspondence Theorem): Let 6:G-^K be a homomorphism of G onto K. The preimage H of any subgroup S of K i?, a subgroup of G containing Ker e. If S <1 K, then H <i G. Furthermore if Hi is any other subgroup of G containing Ker d such that HiO = S, then Hi = H.
{g
\
Since 16 is the identity of K, and S contains the identity the empty set. Also, if g,heH, {gh-^)9 = g9ih9)-K As and is, a. subgroup of gd G S and he G S, it follows that {gh~^)6 G S. Hence gh~^ G G. Since Ker = {x\ xe = \] and 1 G S, Ker9 cH.
Proof: of .K, then
H = IG^
g6
S}.
and so
H^
gy,
If
S
he
<l
Now
Hi
jK", we must show that H <1 G. Let h G H, g G G; then (g-^hg)e = (ge)-^(he){ge). gS, ge G K, and S <\ K implies {g~^hg)9 G S. Then g-^hg G H and so H < G.
Let ff 1 be a subgroup of G containing Ker d and suppose Hie = S. We will show that = {g ge G S}. Therefore hi G and H. Let Ai G Hi; then Ai^ G S. Now hence Hi C ff On the other hand if h G H, then he = s G S. Choose hi G Hi such that = Hi. hi9 = s. Then hhi^ GKerdcHi and so /i G /î and HcHi. Thus
Sec. 4.4]
4.20:
121
Corollary
Let 6: index n
G^ K
<
.
.
oo.
be an onto homomorphism. Let be the preimage of S.
Let S be a subgroup of K of Then is of index n in G.
Proof: Let Ski, Sfe, Skn, where fcj e K, be the distinct cosets of S an onto homomorphism, there are elements gi, .,gnot G such that giO = fe.
.
in K.
As
is
We claim that
(-4.5)
Hgi,
are the distinct cosets of
Hgn
H
?'.
in G.
Suppose that Hgi which iSfej = S/Cj and
=
i
Hgj; then
Qigr^GH. Hence giOigjO)'^ G S, Thus we have shown that all the cosets
i.e.
hki^GS, from
(.4.5)
in
are distinct.
Let g gG. Then gO G K and so g9 G Sh for some integer i. Hence gO = sk, with S. Consider x = gg^^- x6 = gOigiOy^ - shkr^ = s. Consequently gg^^ is in the preimage of S, so that ggr^ g H. This means that g G Hgi. We have thus shown that every
G is a member of one of the cosets in the cosets of /f in G.

element of
{i.5).
It follows that (4.5) consists of all
Of course we can always reformulate results about homomorphisms with the homomorphism theorem as results about factor groups. Thus we have
Corollary
4.21:
aid of the
Let
N <\
G.
where
i? is a
If Hi/N = Hi = H.
L be a subgroup of G/N. Then we can write L = H/N subgroup of G containing N. If L <] G/N, then H <i G. H/N where Hi and H are subgroups of G containing N, then
Let
Proof: Let v be the natural homomorphism of G ^ G/N. Let H = {g gv GL). Then by the correspondence theorem H is a subgroup of G; and if L O G/N, H < G. Also, Hv = L. Since v:g-^ Ng, Hv consists of all cosets Nh, hGH. Because Ker by the correspondence theorem above, H/N makes sense and consists of all the cosets Nh, hGH. Hence H/N = Hv = L.
\
H^N-
Now if HiDN
theorem that Hi
and Hi/N
= H/N,
may
then Hiv
L.
It
then follows by the correspondence
= H.
be studied before reading Theorem 4.22.)
(Problems 4.82-4.89 below
The reader may very well wonder what happens when we take a factor group of a For example, if iV <l G and G/N is a group containing a normal subgroup M/N, then what is {G/N)/{M/N)^ The next theorem tells us that this is isomorphic to a single factor group, i.e. a factor G by one of its normal subgroups.
factor group.
Theorem
4.22
If in the factor
(Factor of a Factor Theorem, also called the Third Isomorphism Theorem): group G/N there is a normal subgroup M/N, M^N, then
M<i G and
:
G/M ^
{G/N)/{M/N)
:
Proof: Let v G ^ G/N be the natural homomorphism of G ^ G/N. Let p G/N -^ {G/N)/{M/N) be the natural homomorphism of G/N -* {G/N)/{M/N). Put = vp. Then is a homomorphism of G -* {G/N)/ {M/N); and since v is onto G/N and p is onto {G/N)/ {M/N), vp is onto {G/N)/{M/N). Therefore G/Ker (vp) = {G/N)/{M/N), by the homomorphism theorem. If g GG, fifv = Ng and {Ng)p = {M/N){Ng); note that here {M/N){Ng) is a coset of the normal subgroup M/N in {G/N), i.e. an element in the group {G/N)/{M/N). Now the elements of M/N are all the cosets Nm, The identity of {G/N)/ {M/N) is {M/N). We ask, what is the kernel of vp? It will be all g such that gvp = {M/N)Ng = M/N. But in
mGM.
GG
122
[CHAP. 4
Ng G M/N, i.e. Ng = Nm for some m G M. Hence g = nm where n G N. But D N. Therefore g G and so Ker vp C M. Note that if m G M, m{vp) = {M/N)Nin = M/N. Then Ker vp = M. Thus as a kernel of a homomorphism is normal in G and G/M = {G/N)/(M/N), which is the required result.
that case
Problems
4.82.
Prove that if subgroup of K.

Solution:
It
G -* K is a homomorphism of G onto H < G, prove that He < K.

=
K, and
is
a subgroup of G, then
He
is
Since le
S He, He
0.
If
x^,
x^
G He,
x^
hyO,
x^
hê
t<yi
some
h^, Aj
S H.
Then
where
fir
Hence He is a subgroup of K. If now H <3 G, then g~^hg G H for all any element of K, say fe, is of the form ge for some g & G; and any element of Hfi is of the form Afl. Is (ge)~^hege e. Het Yes, because {ge)-^(he)g$ = (g'-^hg)e and as g-^hgG H, (ge)-Hhe)ge G He. Thus He < H.
/i
= h]h2^^H.
all
e G and
h&H. Now
4.83.
Let
Let K = ^^(6) be cyclic of order 2. Then e G -^ K is a homomorphism of G onto K. (Take this as a fact.) Find all the subgroups of K and all their preimages. Check that the assertions of Theorem 4.19 and Corollary 4.20 hold. (See Problem 3.42, page 77, for the multiplication table.)
G = D^
by
aê
{ai, ag, as,

1,
(T4,
T, T(72, TCTg, ro'4}.
defined
(raê
h, i
1,2, 3, 4,
Solution:
The subgroups of
G2
(a)
(6)
(c)
K are Ky- K
K2
{g
\
and K^
=
e
{!}.
a2.
<'i>
Gi
the preimage of K,
{g\ ge
&K) -
G.
the preimage of
ge
1}
{tri,
"ii-
Gj,
G2 are subgroups of
in
containing Ker
G2.
Kj and K2 are normal
K, and Gj and G2 are normal in G.

G2.
The subgroups of G containing Ker 9 are Gi and where 1 i, 2. I = i

7'
We
note then that
Gj9
Gj9
implies
(d)
Xj
is
of index 1 in K.
Gj
<r2>
is
of index 1 in G.
K^
is
of index 2 in K.
''<'4)'
G^
is of
index 2 in G,
its
cosets being
Gx
{ai,
"S) "4}
and TG2
{r, tct2>
'''<'3'
Thus
(a), (6), (c)
and
(d)
agree with Theorem 4.19 and Corollary 4.20.
4.84.
Let
G =
gp{a)
be cyclic of order 12 and

. .
let
defined
by aê = b\ i = 0,1,2, fact.) Find all the subgroups of 4.19 and Corollary 4.20 hold.
Solution:
.,11.
Then
all
and
be be cyclic of order 4. Let e G -* a homomorphism of G onto K. (Take this as their preimages. Check that the assertions of Theorem
:
K = gp{b)
e is
{x\
The subgroups of K are Xi = X, K^ = xe:K} = G. G2 = the preimage of K2
{1, b^},
=
a^
{x\
09 = 6 that G2
(a)
(6)
(c)
#
=
62^
hence
G2.
oê
b^,
so
(j^.
Kg = {1}. Gj = the preimage of K^ = Clearly 1 G Gj. xe = 1 or xe = b^}. Continuing in this fashion we conclude
{l,a^,a*,a^,a,a^<>}.
all
Finally,
Gg
the preimage of
9
X3 =
{a;
a;9
1}
{l,a.*,a}.
Gi,
G2 and G3 are
subgroups of
in K,
containing Ker
G3.
in
Xi,
X2 and K3 are normal
and
Gi,
G2 and G3 are normal
(trivially, as
is abelian).
The subgroups of G containing the kernel of e are those containing a*. Hence the subgroups of G containing Ker e are Gj, Gj and G3. Note GjS = Gj implies i = i (1 - i i - 3).
(d)
Xi is of index 1 in X. Gi is of index 1 in G. Xg is of index 2 in X, its cosets being X2 = {1, 62} and X26 = {6, 53}. Gj is of index 2 in G, its cosets being G2 = {1, a\ a*, a^, a, fti"}, and Gâ = {a, a3, a^, a', a^, a"}. X3 is of index 4 in X. G3 is of index 4 in G, the distinct cosets being G3 = {1. aS a}, Gâ = {a, a\ a^}, Gâ^ = {a^ a^ aW}, Ggd^ = {a^, a'', a"}.
Thus
(a), (6), (c)
and
(d)
agree with Theorem 4.19 and Corollary
4.20.
Sec. 4.4]
:
123
4.85.
Let e G -* K be an onto homomorphism. subgroups of index 2 and 5 and 10.

Solution :
of
Let
be cyclic of order
10.
Prove that
has normal
= gp(k). Then the subgroups K^ = {1}, K^ - gp{k^), Kg = gp(k^) are normal subgroups of index 10, 5 and 2 respectively. Consequently their preimages, by Theorem 4.19 and Corollary 4.20, are normal of index 10, 5 and 2 respectively. Hence the result.
Let
4.86.
Let
of
N <\ G and suppose G/N is cyclic of order 6. Let G/N = G/N and express them in the form of Corollary 4.21. (Hard.)
gp{Nx).
Find
all
the subgroups
Solution:
Let G/N -= K. Let K^ = {AT}, K^ = {N, Nx^, K^ = {N, Nx^, Nx*} and K^ = K. These are the subgroups of K. To find the corresponding subgroups of Corollary 4.21 we let v. G -^ GIN be the natural homomorphism, i.e. gv - Ng. Then let G^ be the preimage of Ki, i = 1, 2, 3, 4.
all
Gi
Gz G3
G4
= = = =
{g\
{g
{g
{g\
= N} = {g\ Ng = N} = {g\ g^N} = AT = N or g - Nx^} = {g Ng = N or Ng = Nx^ = NuNx^ gv = N or gv = Nx^ or gv = Nx*} = NuNxûNx* gvGK} = G. Then G/iV = Xj for t = 1, 2, 3, 4.
g,'
gv
4.87.
Use the correspondence theorem to prove that if i/ is a subgroup of G containing G' group of G), then H <i G (i.e. prove Problem 4.69 by another method).
Solution:
(the derived
Let v. G -^ G/G' be Any subgroup of G/G'
the natural homomorphism; then v is onto. G/G' is abelian by Problem 4.68. therefore normal, and thus Hv = S, say, is normal in G/G'. By the correspondence theorem, is the preimage of S. Hence using the correspondence theorem once more, is normal in G.
is
4.88.
Let Hhe a subgroup of index h in G. Let of index n in X if D Ker tf.
-^
be a homomorphism onto K.
Prove that
He
is
Solution:
from Corollary
only necessary to prove that S - He is of finite index in K, for then the result follows 4.20. If Hg^, ..., Hg^ are the cosets of in G, then we claim that {Sigê), ..., S(gê)} is the set of all the cosets of S in K. We need only show that if k e K, then k G SigiS) for some i = l, .. .,n. As 9 is onto, there is a g such that ge = k. Let g = hgi. Then k - ge = heigiS) e S(gie). The result now follows from Corollary 4.20.
It is
GG
.
Alternatively
Then
(srj)(fl'j9)-'
preimage of S.
we can show that Sigê), .,S(gê) are all distinct. Suppose S(ffifi) = (g^~^)e S S. Hence giQ^^ &H as H, by the correspondence theorem, Accordingly i - j and the index of S in K is n.
.
S(ff^s).
is
the
4.89.
Let G be a group and let he a normal subgroup of G. Suppose further that L and are subgroups of G/N. Then show that we can write L in the form H/N, and in the form K/N, where and K are subgroups of G containing N. Show also that if LqM, qK; and ii L <i < K Show that if and [M L] = < , then [K:H]=n.
LqM
M.H
Solution:
This
is
and^
M = K/N
- {g\ gvGL} and K = {g\ gvGM}. Hence if I-cM, H (ZK follows immediately. Now if L < M, we consider the homomorphism $ K ^ K/N defined by ke = fee for all S Jf i.e. e = v,KClearly Ke = K/N and the preimage of L is H, the preimage of M is K. We can then conclude from the correspondence theorem that H < K.
:
just an application of Corollary 4.21 to Corollary 4.20 and Theorem 4.19. follows from Corollary 4.21. If v is the natural homomorphism,
That
L = H/N
we
recall that
fc
4.90.
Let
G=
Di.
Let
M = {ai,<rs,Ta3,T}, N = {cri.as}.
explicitly
Then accept
G/N and M/N. Find {G/N)/(M/N)

Solution:
N<G
and
M<
G.
and check that
words, check agreement with the factor of a factor theorem.
isomorphic to G/M. (Use table on page 77.)

it is
Consider In other
G/N
A4 -
consists of the cosets
(t(72)A'^
{t(72, r(74}.
Now
M = {<,i,a3,Tff3,r},
Ai =
N =
{ai,a3},
A^ = a^N = U2,<rd, A^
hence
= tN =
elements
{r,r<7s},
M/N
consists
of the
AÂg.
124
[CHAP. 4
MIN <
(M/N)
in (GIN).
GIN. Therefore we can talk of (GIN)I(M/N). The elements of this group are the cosets of These cosets are B^ = (MIN)A^ = {î, A3} and Bg = (MIN)A2 = {AyA^.AÂ^} =
(A3A2, for example,
is
{A2,A^}.
plication in the
group (GIN)I(MIN)
calculated as follows: A3A2 = (riV)(<r2^) = ('"<'2)^ is calculated in the usual way for cosets, e.g.,
^4-)
Multi-
B2B2
as
A2A2 =
(aaiVXtrjiV)
= {MIN)A2(MIN)A2 = {MIN){A2A2) = {MIN)Ai = = a^N = N = Ai. It is clear then that (GIM)I{MIN) =

GIM
is.
B^
{Bj, B2}
is
the cyclic
03, rag, t}
group of order 2 generated by B2-
Now
let
us decide what
The elements of
C2C2 = Ma^Ma^, (GIN)I(MIN).
and C2 = Ma2 group of order
=
2.
{<t2,
^4, Ta4, TCTg}-
As
GIM are the cosets - M03 = M, GIM =
Cj
M=
{<ti,
{Cj, C2}
is
the cyclic
Therefore
GIM =
4.91.
= gp{a^), N - gp(a^). Consider GIN Let G be the cyclic group of order 12, say G = gp{a). Let and MIN. Find (GIN)I(MIN) explicitly and check that it is isomorphic to GIM. In other words, check agreement with the factor of a factor theorem. (Difficult.)
Solution:
GIN
Na3
Aj = AT = {l,a}, Aj = Nâ = {a,a''), A3 = Na^ = {a^,a}, A4 = consists of the cosets = {a3, a9}, Ag = iVa* = {a*, ai"}, Ag = Na^ = {a^, a"}. Now M = {1, a2, a*, a, a^, aio}. Hence MIN consists of the elements Aj, A3 and Ag. MIN < GIN,
this
and we can talk of (GIN) I {MIN). The elements of

cosets are
group are the cosets of
MIN
in
GIN.
These
Bi
Multiplication
is
= {MIN)Ai =
B2B2
{Ai,A3,A5}
and
B^
= (MIN)A2 =
(M/2V)(A2A2)
{A2,A4,A6}
calculated in the usual way,
= {MIN)A2(MIN)A2 =
Now
order
the product of
2.
Ag and A2
Bi.
It is
B2B2 = iMIN)As =
is also the product of cosets. clear then that (GIN)I(MIN)
As Aj = Na, AÂ^ = Na^ -
A3.
Hence
{B^, B^}
gp(B2)
is
a cyclic group of
Now
Clearly
let
us decide what Ci
GIM
is.
The elements of GIM are the

C2
cosets
{1, a2,
aS
a6, a8, aiO}

2.
GIM =
gpifii)
is cyclic
of order
= Ma = {a, aS, 0.=, a^ a**, a"} and Hence GIM = (GIM)I{MIN).

iî+i(G)
is
4.92.
Let
{AH
be any group. Let Pi,(G) be all possible nonisomorphic factor groups of G. Let In the particular case that G possible factor groups of the groups in Fi(G)}.
00
cyclic of
order
25, find all
nonisomorphic groups in
KJ
Fi{G),
i.e.
Fi(G)uF2(G) Vi
(Hard.)
Solution:
It is sufficient to
G FîG). But then consider Fi(G). For if L&F2(G), L = MIN where Thus L is a factor of a factor group. Hence by the factor of a GIK, by definition of FîG). factor theorem it is isomorphic to a factor group of G. We must therefore find the number of factor groups of G. All subgroups of G are known. G has unique subgroups of orders 1,2,22,23,2* and 25 by Theorem 4.9, page 105. The factor groups
will therefore
if
G=
be of orders 25, 2<, 23, 2^, 2 and 1. In each case the factor groups will be gp{a) and iV is a subgroup of G, gp{Na) = GIN. Hence the result.
cyclic.
For
The subgroup isomorphism theorem In the homomorphism theorem we were able to say that the image of a homomorphism 9:G-*K was essentially a factor group of G. What can we say about the effect of 6 on subgroups? Let Hh%& subgroup of G. Let Bi = dâ, i-e. ^1 is the mapping of i? to K deThen 61 is a homomorphism of H^K, and so HOi = He ^ fined by h9i = he, hG H. H/{KerOi). Now if Ker^ = N= {x\ xGG, x9 = 1}, then Kerî = {x\xGH and So He = Hdi^H/{HnN). On the other hand, we know H9 is x0i = xe = l} = HnN. a subset of GB and G0 = G/N. Our question is: what has H/{HnN) got to do with G/NI It must be isomorphic to some subgroup of GIN. But which? This is what the subgroup isomorphism theorem is
c.
about.
Sec. 4.4]
4.23
125
Theorem
(Subgroup Isomorphism Theorem, also called the Second Isomorphism Theorem): Let N <i G and let H he a subgroup of G. Then <i H, HN is a subgroup of G, and
HnN
H/{HnN) = HN/N
hGH, nG N}) Proof: If nGHnN and hGH, then h~^nh G N nGH. Therefore k-^nhGHnN and HnN < H.
(HN =
{hn\
as
N<
G,
and
h'^nhGH
Xz
as
HN is a subgroup, for
xix^^
it is
not empty; and
if
xi, X2
G HN,
then Xi
hiiti,
hiTii
and
hinin^^h^^
hirizhi^
hikz^h^nahz^
as
hui
G.
where nz-nmz^ GN, hJii^-hGH and and HiV is a subgroup of G.

:
hinJii^^mGN
N <\
Hence
ajiaja"^
ifiV
Let ^ ff - HWA^ be defined by ft^ = iV/t. Then ^ is clearly onto, i.e. H = HN/N. Also is a homomorphism: (hih2)<j> = N{hih2) = NhiNhz = /ii^fe^. By the homomorphism theorem, H4,^H/{Ker<i,). Ker = {x\ x G H, x,j> = 1} ^ {x\ x G H, Nx = N}. If Nx = N, then Ix =:= a; G iV; and if xGN, Nx = N. Therefore Ker ^ = {a; xGH, xGN) == HnN.
</>
|
Hence
HN/N ^ H/{HnN).
Let Q* be the multiplicative group of rationals. Let = {1, 1}. Let be the subgroup generated ii)- Find HN, HN/N and thereby verify the assertion of the subgroup isomorphism theorem
Problems
4.93.
by
that
HN/N s H/HnN.
Solution:
The elements of
H are
all
of the form {^Y, r various integers.

hn,
HN
A
X
coset of
{x\ x
hGH, n&N} =
HN/N
is
of the
form
is
= h or x = -h, h e H} = {x\ X = {^y for all integers r} Nx = {1, ~l}x = {x, -z} where x G HN. Now
{x\ x
(^)^ Hence each coset
of the
form {(^)^ -(\y).
Since N{\)
N {^)
N(^)
G HN, if = N{^y and

a;
e Ar(^)r^ each coset of HN/N is a power of N(^). Thus 'p({A^(^)}) = HiVW; and iV for r'0, HN/N is the infinite cyclic group. i^y Now HnN = {x\ x = {^y for some r and x = 1} = {1}. Hence H/(HnN) = H. But infinite cyclic. Thus we have verified that H/(HnN) = HN/N.
(i)*"
since
is
4.94.
Let<r=^
Solution:
*^j
and
H = gp{{a}).
is
Prove that
HAM
is cyclic of
order
2.
HA/A = H/{HnAJ. Now
HnA =
of order
{*}.
Therefore
HAJA = ff/{,} ^ ff.
an odd permutation, hence a A. Also H = {<r, i}, so But is cyclic of order 2. Thus AM is cyclic
2.
4.95.
Let a group
contain two normal subgroups

if
HM/M = HN/N
Solution:
and N.
Let
be a subgroup of G.
Prove that
Hnilf
:=
HniV.
By
the
HN/N = HM/M
4.96.
subgroup isomorphism theorem, by Problem 4.8, page 97.
HM/M = H/HnM = H/HnN = HN/N.
Hence
If in the preceding
H/{HnN) has
Solution:
problem we know that G/M has every element of order a power of every element a power of 2.
2,
show that
H/{H nN)= H/(H n M) =
HM/M q G/M.
Hence the
result.
126
Let
[CHAP.
4.97.
and
be subgroups of G,
<i
G and
HN
= KN.
Prove that
H/(HnN) = K/(KnN).
97,
Solution :
H/(Hr\N) =
HN/N = KN/N = K/{KnN).
Then by Problem
4.8,
page
H/{HnN) = K/{KnN).
4.98.
Let Gd GiD {1} and Gj O G. Suppose G/Gi and Gj are abelian and H is any subgroup of G. Prove that there exists a subgroup H^ of H such that Hi <1 H, and ///Hi and Hi are abelian.
(Hard.)
Solution :
Let Hi = HnGi. Then by the subgroup isomorphism theorem. Hi <1 H, and H/H^ = HGi/Gi. But HGi/Gj C G/Gi and G/Gi is abelian. Therefore H/H^ is abelian. As Hi c Gi and Gi is abelian,
we
Let
conclude that
Hi
is
abelian.
4.99.
{1}. Let Gi < G, G2 < Gi, and suppose G/G^, GJG^ and G^ are abelian. Prove any subgroup, then it has subgroups Hi and Hj such that Hi <1 H, Hj <1 Hi and H/Hi, H1/H2 and Hj are abelian. (Hard.)
GD
if
that
Gi 2 G2 2
is
Solution:
Let H,
= HnGi.
Then, as
in
Problem
4.95,
H^ <
and H/Hi
is abelian.
Now Hi as a subgroup of Gi. As G2 <3 Gi, by the subgroup isomorphism theorem HinG2 <] Hi and Hi/iHinG^) = H1G2/G2 C Gi/Gj. Since G1/G2 is abelian, so is HiKH^r^Gi). Consequently we put H2 = H^nGiconsider
Finally as
H2 C G2 and G2
is
abelian, so
is
H2 and
the result follows.
Homomorphisms of cyclic groups We return now to study cyclic groups. In Section 4.2b we could state that in a finite cyclic group there was at most one subgroup of any given order. An analogy for the infinite cyclic group would have been awkward to formulate without the concept of index which we
d.
have had at our disposal since Section
4.3a.
Recall that a subgroup oi a. group G is of index in G if there are exactly n distinct in G. In the case of finite cyclic groups we have proved that there is one right cosets of dividing \G\. Since [G:H] = \G\/m, there is one of any order and only one subgroup and only one subgroup of any given index dividing the order of G. This gives the clue to
Theorem
4.24:
There
is
one and only one subgroup of any given
finite
index n
>
in
the infinite cyclic group.
Proof: Let G = gp{x) where G is infinite cyclic. Let Hn = gp{x''). Then Hn = {a;"'.,HnX"-'^ are all distinct and are all the cosets Hence the cosets Hn,HnX, all integers r}. of Hn in G. Hence Hn is of index n.
. .
Next if r is a subgroup of index n, we already know that H positive power x"^ G H (Theorem 4.9, page 105), and this means Hence r = n and Hr H. This concludes the proof. r.
find all
is
H = Hr.
generated by the smallest But Hr is of index
With our knowledge homomorphisms

4.25:
of cyclic groups of cyclic groups.
it is
easy to apply the homomorphism theorem to
Theorem
Let
if
6*
be a homomorphism of a cyclic group G.
Then Ge
is
cyclic;
and
|G|
<
=0,
\Ge\ divides \G\.

if
Furthermore
is
is
any
cyclic
group such that
\H\ divides \G\,
there
a homomorphism of
If
onto H.
there
is
is infinite cyclic,
homomorphism
of
onto any cyclic
group.
Proof:
If
G=
gp{x), then
G=
I
{x''\
all
integers r}, and
Gd =
[x^'e
all r}
{(xoy
all r)
gpixO)
CHAP.
4]
is cyclic.
127
Thus Ge
If
\G\
<
>,
normal subgroup
Say,
N of G.
Hence
then by the homomorphism theorem 09 = \Ge\ = |G|/|A^| (see Lagrange's theorem) and
G/N
for some
\Ge\ divides \G\.
Suppose H is cyclic, H = gviv), and the order m of H divides the order of G = gpix). |G| = rm. Let A^ = gpix^). Then A^ = {l,x'-, (a;'")'-i}, \N\ = r, and N is of index ., m. Because G is abelian, G/N makes sense and is of order m. But G is cyclic and consequently so is G/N. Hence G/N - if, and is a homomorphic image of G. If G is infinite
.
cyclic, then, as
Theorem 4.24, [G G] = w if G = rp(a;"). As G <1 G, G/Gn is of order w. But G/Gn is the homomorphic image of a cyclic group; hence it is cyclic. Thus G has as homomorphic image any cyclic group of order n, n> 0. Obviously it also has the infinite cyclic group as a homomorphic image.
savir
we
in the proof of
In this chapter we have thoroughly investigated the simplest class of groups, the cyclic groups. We know that there are cyclic groups of all orders, we know their subgroups, we know that they have as homomorphic images only cyclic groups, and we know whether any
group G has as homomorphic image a given cyclic group. groups of cyclic groups are again cyclic.
cyclic
Furthermore the sub-
introduced the concept of coset. The cosets form a partition of the group. obtained Lagrange's theorem which states that the order of a subgroup divides the order of a finite group. This enables us to eliminate certain groups as possible subgroups of a given group. We will see later on that it also enables us to find more quickly the groups of a given order.
We have also
this fact
Using
we
Next we have introduced the idea of a homomorphisms, namely as The subgroup isomorphism theorem tells subgroup iy of G in a factor group G/N is
of looking at
H/{HnN).
normal subgroup. This gives rise to a new way factor groups (see the homomorphism theorem). us that the subgroup corresponding to a given isomorphic to a factor group of itself, namelv
The factor of a factor theorem tells us that a factor group of a factor group G/N is just a factor group of G of the form G/M. Finally, the correspondence theorem associates with each subgroup of the image of a homomorphism e.G-^K a unique subgroup of G
itself.
FUNDAMENTALS
4.100.
Prove that
if
a and
6 are elements of
a group
and
if
a-^bâ
ha,
then 6
a.
4.101.
Suppose a and
6 are elements of
a group G.
If a^
and a-^hâ
b^,
prove 6^
(Hard.)
4.102.
Suppose a and 6 are elements of a group G. (Very hard.) Suppose G and are groups. Suppose that Prove G and are not isomorphic.
If
a-^bâ
b^
and b-^^b
a^,
prove
b.
4.103.
cannot be generated by two elements but that
H can.
if
4.104.
Let
.S:
be a non-empty set and
let
Y=
{y}
be disjoint from X.
Prove Sx
= Sx u y
if
and only
is infinite.
128
[CHAP. 4
CYCLIC GROUPS
4.105.
Let
be a cyclic group.
Prove that
if A^ is
a subgroup of
such that
G/N =
G,
then
N=
{1}.
4.106.
Let
G=ZXZ
(fc,
where
is
the set of integers.

(k,l)o(m.,n)
Define a binary operation in
G by G
is
(k
+ m,l+n)
Prove that
not cyclic.
where
l),{m,,n)
G.
is
a group with respect to this composition.
4.107.
Let Gj, G2,
...
G1UG2U
4.108.
is
be subgroups of a group G. not a cyclic group. (Hard.)
If
GjCGj+i, Gji^Gj+i for
l,2,...,
prove that
Let
ce
e.
G be a group and C for all c&C.
let
It
is
-> F be a homomorphism. Let C be a cyclic subgroup of G. any subgroup of C, prove that he G H for all h & H.
Let
4.109.
Let Q be the additive group of rationals with respect to addition. subgroup (# 0) of Q is infinite cyclic.
Prove that every two-generator
COSETS
4.110.
H be a subgroup of G. Prove that H in G with the left cosets of H in G.

Let
Let
of
Hg ^ g-^H
is
a matching of the right cosets of
4.111.
H and K be subgroups of a group

D
G.
Show that a
coset of
H intersection
zGZ}.
a coset of
K is
a coset
HnK.
be the group of Problem 3.72, page 91.
4.112.
Let
Let
N=
{(0,z)
Prove that
N< D N< G
and
D/N
4.113.
is infinite cyclic.
Let
G/N
4.114.
is infinite cyclic.
- {(0, q) q e. Q}. Prove that be the group of Problem 3.74, page 91. Let is isomorphic with the additive group of rationals. Show that
\
and
Let
W be
the group of Problem 3.77, page 91.
Let Af
{(0, 6)
(0, 6)
G W}.
Show that
M<
and that
G/M
is infinite cyclic.
4.115.
be a subgroup of G, and let g be an element of G. Prove that if N{H) is the Let G be a group, let and N(g-^Hg) the normalizer of g-^Hg, then g-^N(H)g - N(g-^Hg). g'^Hg normalizer of
{g-^hg\
hGH}.
4.116.
Let
g = If""
a,b,c,dGz\-.
Prove that g forms a group with respect to the operation
defined by
*\
/! \ci
&i\
/a +
oi
ci
dj
/a
(
di)
\c +
d.
+ 61 + di
8 is
b\
,
Let
e:
-^
be defined by
j e
a+
Prove that
a homomorphism of
g onto
the additive group of integers and find its kernel. Consider ff/(Ker a) and prove that in accordance with the homomorphism theorem it is isomorphic with the additive group of integers.
4.117.
Let
e:
g -
("'
ad
be
9i 0,
a, b, c,
d real numbers
{
with operation matrix multiplication. Let

] e
^c
g-* R*, the nonzero real numbers, be defined by
ad
be.
Prove that
# is
a homo-
\c
dJ
and
find its kernel.
morphism from g onto the

that
g/{KeT
e)
= R*
multiplicative group of nonzero real numbers in accordance with the homomorphism theorem.
Prove
CHAP.
4]
G
:
129
4.118.
Let
defined by
be any subgroup of S, the symmetric group of degree n. Let 8 G - {1, -1} be the mapping Prove that a;e = 1 if k is an even permutation and xe = 1 if a; is an odd permutation. 9 is a homomorphism of G into the group {1,-1} with operation multiplication of integers. Using the homomorphism theorem, prove that the even permutations of G form a normal subgroup of G.
4.119.
Let G be a group and N a normal subgroup of G. Suppose that = HiOH^^ dH where [Hj: jffj + i] = i + 1, for groups 2 G such that [Gj G has a sequence of subgroups G = Gj D G2 3
H=
i
:
GIN
1,2,
has a sequence of sub...,%- 1. Prove that

i
Gi+ 1]
= i + 1,
1,
. ,
n-
1.
4.120.
Let
M and JV be normal subgroups of G with MdN.

G
Prove that
G/N
is finite if
G/M and M/N
are
finite.
4.121.
Let
cyclic.
be a group and a normal subgroup of G. Suppose G/N has a factor group which Prove that G has a normal subgroup of index n for each positive integer n.
is infinite
4.122.
Let
be a
finite
the orders of
group with normal subgroups M and N. Let H be a subgroup of G. Suppose that M and H and those of N and H are co-prime. Prove that HM/M = HN/N.
4.123.
Let N,
be normal subgroups of G,
iV
D M. Suppose GIN
is cyclic
and
\NIM\
2.
Prove that
GIM
4.124.
is
abelian.
Find a group
abelian.
with normal subgroups
Ai'
and M,
Nd M,
G/N
cyclic,
N/M
cyclic
but
G/M
not
chapter 5
Finite
Groups
The most important result of this chapter is a theorem of Sylow which guarantees the existence of subgroups of prime power order. We prove two other theorems of Sylow concerning subgroups of prime power order and then examine groups of prime power order. One result is that groups of prime power order always have non-trivial centers.
In order to construct a new group from any two groups G and H, we define a binary operation on the cartesian product of G and H. The resultant group is called the direct product of G and H. A simple condition enables us to conclude that a group is a direct product.
The concept
order
15.
of direct product together with general theorems about subgroups, e.g. the
classify finite groups.
Sylow theorems, help us to
In this chapter
we
find all
groups up to
We study a class of groups called solvable groups. Solvable groups are used in Galois theory to determine whether an equation is solvable in terms of nth roots.
ambitious plan for studying finite groups is to find all simple groups, i.e. groups without proper normal subgroups, and then see how groups are built from simple groups. The Jordan-Holder theorem shows that in a sense a group is built from simple groups in only one way. As yet the task of finding all simple groups is far from complete. We conclude the chapter by exhibiting a class of simple groups, namely A, for n^5.
An
5.1
a.
THE SYLOW THEOREMS

Statements of the Sylow Theorems
Lagrange's theorem (Theorem 4.11, page 109) tells us that the order of a subgroup divides the order of a finite group. Conversely one might ask: if G is a finite group and n\ \G\, is there always a subgroup of order nl
The answer to this question is no: Ai is of order 12 but has no subgroup of order 6 (see Problem 5.1 below). The following important theorem, however, ensures the existence of subgroups of prime power order. In the following p will denote a prime.
Theorem
5.1
(First
Sylow Theorem):
Let G be a finite group, p a prime, and p"" the highest power of p dividing the order of G. Then there is a subgroup of G of order p".
Suppose is a subgroup of G of order a power of a prime p, and \H\ is the highest power of p that divides \G\. Then is called a Sylow p-subgroup of G. By Theorem 5.1 every finite group has a Sylow p-subgroup.
In general a group of order a power of the prime p is called a p-group. of a group G is a maximal p-group in G, i.e. if where then F = (see Problem 5.4).
group
H qF qG
A F
Sylow p-subis
a p-group,
130
Sec. 5.1]
THE SYLOW THEOREMS

illustration
3.
131
As an
for p
we
1
find the
Sylow j^-subgroups of the symmetric group S3 on {1,2,3},

of S3 given in Section 3.3a, page 57, are
2 and
The elements
2
3\
12
1
3/
3\
''^
_ " _
/I
2
1
3\
\3
/I
2/
3\
^2
_ -
/I
(^3
3
1
2
2
1
2 3
1/
^^"(^13
is
2/
^'^
_ /I ~ \2
3 3
The order
l-Sgl
2 and the order of any Sylow 3-subgroup is 3, since so the sets {t, xj, {hrj and {t, T3} are all subgroups of order 2 and therefore, by definition, Sylow 2-subgroups of S3. There are no other Sylow 2-subgroups of S3 because a^ a^, a\ a^ implies S3 has no other elements of order 2. {i, ffj, <j^ is the only subgroup of order three in S3, so it is the only Sylow 3-subgroup of S3.
of
any Sylow 2-subgroup

Iijow
Tf
2-3.
T2
T^
i,
Theorem
5.2
(Second Sylow Theorem):

If if is in a
a subgroup of a finite group Sylow p-subgroup of G.
and
H is a p-group, then B. is contained

if
g-^Sg
Two =
subgroups S and T of a group T. Recall g-^Sg = {g'^sg s

|
G
e
are called conjugate

S}.
there
is
a g
&G
such that
Theorem
5.3
(Third Sylow Theorem):
Any two Sylow

Sp of distinct
divides
\G\.
p-subgroups of a finite group G are conjugate. The number Sylow p-subgroups of G is congruent to 1 modulo p and Sp (Sp is congruent to 1 modulo p if Sp = 1 + kp for some integer A;.)
Before proving the Sylow theorems, we will use them to show that, up to isomorphism, is one and only one group of order 15. If \G\ = 15 then, by Theorem 5.1, G has at least one subgroup of order 3 and at least one of order 5. Now Theorem 5.3 implies that there are S3 = l + 3fe subgroups of order 3 and ss] \G\. But (l + 3fe)| 15 implies k = 0. Therefore G has one and only one subgroup of order 3. Similarly G has one and only one subgroup of order 5. These subgroups must be cyclic (Problem 4.48, page 110). Let Hi = {1, a, a^} be the subgroup of order 3 and H2 = {1, b, h^, h^, 6*} the subgroup of order 5. Hir\H2= {!}, because an element ^1 cannot have order 3 and 5 simultaneously. We look at the order of ab in G which must be either 1, 3, 5 or 15. If the order of ab is 1, then ab = l and a = 6- which is impossible, for HinH2= {1}. If the order of ab is 3, then gp{ab) = Hi, since H\ is unique. In this case ab = a* {i = 0, 1 or 2) and & = a*"' which is impossible. If the order of ab is 5, gp(ab) = H2. Hence ab = b' (i = 0, 1, 2, 3 or 4) and a = 6*-i which is impossible. Therefore the order of ab is 15 and G is the cyclic group of order 15 generated by ab.
there
Further applications of the Sylow theorems are given

Section 5.3.
in the
problems below and in
Problems
5.1.
Show
that the alternating group
A4
has no subgroup of order
6.
Solution:
The elements of A^ were given

1
in Section 3.3c,
page
'"
62.
We
2
1
repeat them here for convenience:

3 3 3
1
12
1
3 3
4\ 4/ 4\
/I
'^
2 2
2
4\
/I
4\
/I
"^
2 4 2
3
3
1
V3
4
3
1
1/ 4\
V4
/I
'"'
2/
4\
Vs
2 4
1
_
'
/I
2 3
13
1
_
"^
/I
3 2 3
4
3 2
2/
4\
V4
2
2
3/
4\
^
V2
4/
4\
V^
2 4
^^
i.l
3/
_ -
/I
3 3
1/
_ /I -
2
1
3 2
4/
"^"12143
_ /I
132
FINITE GROUPS
Suppose that A4 has a subgroup Also Tj is of order 3 for j = 1,2,
. .
.
[CHAP.
6. a\=^ a\ a\ = so a^, a^ and wg are of order 2. Hence the elements of order 2 are ag. "5 and "s. and the elements of order 3 are tj, t2, Tg. Now H is of order 6, so it contains a subgroup of order 3, by Theorem 5.1. Therefore tj G H for some i, say tj S fl^; then 1, ti, rf = t2 S H. H must also contain an element of order 2, by Theorem 5.1. Hence H contains a a;, say H contains 02- Because OiTi = T4 and T1O2, = Tg, if H contains (rg it also contains T4, T4 = T3, rg and Tg = T7. This would mean H has at least 8 distinct elements, which contradicts the assumption that \H\ 6. Thus 0-2 /f. A similar argument shows a^ and wg i?. This means that H does not contain subgroups of order 2, contradicting Theorem 5.1. Therefore our initial assumption is invalid and A4 does not contain a subgroup of order 6.
H
.
of order
.,8.
i,
5.2.
Find
all
Sylow p-subgroups of A^ for p
and
3.
Solution:
The elements of A^ are given in Problem 5.1. The order of a Sylow 2-subgroup is 4, since power of two dividing 12, the order of A4. Consequently by Lagrange's theorem none of the t's can be elements of a Sylow 2-subgroup because they are all of order 3 (see Problem 5.1) and 3 does not divide 4. Now aja^ = where i, j, k e {2, 5, 8} and a? = for i 2, 5, 8. Hence P {'. "21 "5' "} is a subgroup of A4 of order 4. P is the only possible Sylow 2-subgroup as there are only four elements having order dividing 4, viz. The o-g, (75, ag, and these elements are in P. order of a Sylow 3-subgroup is 3. The sets {i, ti.Tj}, {i, Tg, t|}, {1, T5, Tj} and {1, T7, t^} are all subgroups of order 3. These are all the possible Sylow 3-subgroups, as they include all the elements of
22 is the highest
<rj^
1
t,
order
3.
Alternately we may use Theorem 5.3: S3 = 1 -I- 3A; must divide 12. Clearly fc 7^ (we already have four subgroups); and if A; > 1, Sg does not divide 12. Hence fc = 1 and there are exactly four Sylow 3-subgroups.
5.3.
If ff is a subset of a gfroup
and g G G, then \g-^Hg\
\H\,
where g~^Hg
{g~^hg
H}.
Solution:
We define a matching a:H-*g~^Hg by a:h-*g--^hg for hGH. a is clearly an onto mapping. To show a is also one-to-one, we must prove hi h2 (hi,h2&H) if and only if fi'-'ftifl' jr-ifejjsr. Let h^ = Aj- Then by multiplying on the left by g~^ and on the right by g we get g~^hig = g~^h2g. Similarly g~^hig = g~^fi2g implies hi h2. Hence a is a matching and
\g-iHg\
\H\.
5.4.
Let \G\ p^m (r 1 p-group such that P

Solution:
and
qH cG,
p/ m)
then
and
let
be a Sylow p-subgroup of G.
Prove that
if
is
H = P.
|
Suppose
\H\
p', t
t
and
\P\
p*-.
Hence
Q. By = r and
Lagrange's theorem, p'

\P\
p'^m.
Since
p/w-,
r.
But
PcH
\H\,
and so
P - H.
Sylow p-subgroup of G.
5.5.
If
is
a Sylow p-subgroup of G, then
g~^Hg
is
also a
Solution:
Suppose
that subgroup.
Hence g~'^Hg
- p^m (r ^ and p/w); then \H\ - p^ But \g-^Hg\ = \H\ by Problem 5.3. a Sylow p-subgroup of G if it is a subgroup. To prove g~^Hg is a subgroup, observe (g~^hig)(g-ih2g)~'^ g~^hih-^g G g-^Hg. From Lemma 3.1, page 55, g~^Hg is therefore a
|G|
is
5.6.
Prove that a
Solution:
finite
group
is
a p-group
if
and only
if
every element of
has order a power of
p.
If |G| = p'' then, as every element of G must have order dividing the order of the group, every element has order a power of p. To prove the converse let every element of G have order a power of p and assume the order of G is not a power of p. Then there is some prime q, q '^ p, such that q |G1. But by the first Sylow theorem, G has a subgroup of order a nonzero power of contains an element g = 1. By Lagrange's theorem, the order of r is a nonzero power of q. So q and hence the order of g is not a power of p. This contradicts the assumption that all elements have order a power of p. Hence \G\ = p*" for some r 0.
Sec. 5.1]
THE SYLOW THEOREMS

G
has only one Sylow p-subgroup H, then
133
5.7.
If
H<
G.
Solution:
If g e.G, g~^Hg is a p-subgroup. Thus g~^Hg =
Sylow p-subgroup by Problem for all g G G and H < G.
5.5.
But G has only one Sylow
5.8.
If
[G\
pq,
Furthermore
Solution:
if
where p and q primes and p < q, then q^ \ + kp for any integer A;, then G
G
is
has one and only one subgroup of order the cyclic group of order pq.
q.
By Theorem 5.3 G has s, = 1 + fcg Sylow g-subgroups of order q, with fe - 0. Also 1 + kq divides pq. There can only be four possibilities for 1 + kq as the expression of pq as a product of primes is unique: X + kq = q or 1 + kq p or l + kq = pq or 1 -I- fcg = 1. As g does not divide 1 -f- kq, we are left with the possibilities that 1 + kq = p or 1 + kq = 1. Since q > p, 1 + kq = p
and hence fe = 0. Thus there is only one subgroup of order q, say H. There are Sp = 1 + fep subgroups of order p. Again we have the possibilities 1 + fcp = 1, 1 + kp = p, 1 + kp = q, or 1 + kp pq as Sp divides |G|. Clearly p does not divide 1 + fcp, so 1 + fep = 1 or 1 + kp = q. The last is not true by assumption, so again there is only one subgroup
K of
order p.
It follows
H are of order
from Problem 5.7 that H < G, K q and those of K are of order p.

h-^k-^hk
<\
G.
If
Also he. and k
HnK =
as as
{1}
= =
h-Hk-ihk)eH
{h--^k-^h)k
GK
= 1.
& K (h H< G K< G
as the nonunit elements of = 1, k^ 1), then
Hence h~^k~^hk 1 and h and k commute. But {hk)" = hPkP as h and k commute, so hk is of order pq, and so G is cyclic.
Instead of the .,fep~i}. {l,k,k^,
.
By Lagrange's
{hk)p
h^
theorem, the order of hk is Similarly (/ifc) = fe" ?^ 1.
p,
q or pq.
Therefore
= {l,h,h^, ...,fc9-i}^ K = last paragraph, we note that an element of order l,p,q or pq. If hk is of order p, since there is only one subgroup of order p, gp{hk) = K, i.e. hk = fe' for some i, i p 1. But then = {!}. Similarly gp(hk) is not of order q, nor of order 1. Thus hG K, which contradicts gp(hk) is of order pq, and so G is cyclic.
argument of the
.
Now
hk
is
HnK
5.9.
Show
that
if
+ kp
in
Problem
5.8,
is
not necessarily cyclic.
Solution:
Consider S3, the symmetric group on

cyclic group.
{1, 2, 3}.
|iSg|
3'
2,
+ 1'2
and S3
is
not a
5.10.
If \G\ 2p, p an odd prime, then G has one and only one subgroup of order p and either exactly p subgroups of order 2 or it has exactly one subgroup of order 2.
has
Solution:
From Problem 5.8 we know G has one and only one Sylow p-subgroup. Because p is itself the highest power of p dividing \G\, the Sylow p-subgroup of G is of order p. Thus there is precisely one subgroup of G of order p. The number of Sylow 2-subgroups of G is Sj = 1 + fe2 for some integer k. Again 1 4- 2fc = 1, 2, p or 2p. As 2 does not divide 1 -I- 2fe, then 1 -I- 2fe = 1 or 1 -I- 2fc = p and the number of Sylow 2-subgroups is either 1 or p.
b.
Two lemmas used Lemmas 5.4 and

Throughout
in the proof of the
Sylow theorems
5.6 will
provide the tools for proving the Sylow theorems (see

will
etc.
Section 5.1c).
this section
be a fixed group and
a subgroup of G.
As
usual
we
denote subsets of
G by
A, B, C,
flr
A generalization of the concept of normalizer as defined in Section 4.3d, page 112, will be essential. We point out once more that if A is a non-empty subset of a group G and G G, then g^Âg - {g^âg a G A}.
\
Definition:
Let
is called
be a non-empty subset of a group G. The set {h\ h~Âh the normalizer of A in and is written Nh{A).
A, h
G H}
134
FINITE GROUPS
easy to prove that Nh{A)
[CHAP.
It is
is
is
a subgroup of
4.
(Problem
5.11).
When
H ^ G,
Ng{A)
the normalizer of
as defined in Chapter
Definition:
Let
if
and
h-Âh =
B be non-empty subsets of G. B is said to be an H-conjugate of A B for some hG H. (Note that if H = G, then A and B are concalculating the
jugate as defined in Section 5.1a.)
The next lemma gives us a formula for

which are i?-conjugates of A.
number
of distinct subsets of
Lemma
5.4:
If
is
finite
group
viîth
subgroup
is
distinct H-conjugates of
:
and non-empty subset A, the number of the index of A^h(A) in H, i.e. [H A7^h(A)].
:
Proof: Since [H Nh{A)] is the number of distinct right cosets of Nh{A) in H, we need only define a one-to-one mapping, a, of the right cosets of NuiA) in onto the distinct fl'-conjugates of A. Let a be defined by
a:
N{A)h^h-Âh
and only
{h
G H)
that for
hi,
To show
that
is
a one-to-one mapping,
we must prove
if if
hz
H,
NH{A)hi
(i)
= NH{A)h2
hrÂhi
= hzÂhi
hiÂhi = h2Âh2. Then A = hihzÂhzhi^ = (kihr^yÂih^hr^). Hence hzhr^ G NniA) and so /12 G NH{A)ht. Since two right cosets are equal or disjoint, we conclude NH{A)hi = N{A)h2. Thus hTÂhi = h^Âhi implies NH(A)hi = A^h(A)/i2.
Let
If
(ii)
A^h(A)Ai
NH{A)h2,
then
hi
NH{A)h2,
i.e.
hi
nhz
for some
n G Nh{A).
hiÂhi
because
{nh2)"'Ânh2
definition of
= hz^n^Ânhi = h2Âh2
Nh{A).
n~Ân =
A
a
is
by
Hence
NH{A)hi
NH{A)h2
implies
hrÂhi
h2Âh2.
clearly onto, so the proof is complete.
Most of our arguments are concerned with sets whose elements are subsets of G. denote such sets by script letters cA, "B, etc. For example,
for example, subsets of G set would be
is
We
let
be the cyclic group of order
{\,a},
{a^,a^,a^},
C=
the set whose elements are 'B = {A, B, C).
Subsets of G are, {a}. An example of a set whose elements are and B. We write cA {A,B}. Another such
6,
G=
{l,a,
.,a^}.
Proposition
5.5:
Let of be a set of subsets of G. We define for A,B G cA, A ~ B ifi?is such that an fl'-con jugate of A (i.e. if there exists an element h G B). Then ~ is an equivalence relation on cA (see Problem 5.16 h~Âh for the proof).
Recall that equivalence class containing
it is an equivalence is the A), i.e. {X| and A (see Section 1.2c, page 9). Recall that the distinct equivalence classes are disjoint and that their union is qA (Theorem 1.2, page 10).
We
will
make a few
observations about ~, which follow because

if
relation on cA.
AGqA,
A~
XgA
X~
By a set of representatives of the equivalence classes we mean a set % which contains one and only one element from each of the distinct equivalence classes. It follows that cA is the disjoint union of the sets R~, R G%. Hence \cA\= |^~1- We are now in Reîi. a position to prove our main lemma.
Lemma
5.6:
Let cA {' 0) be a set of subsets of G. Suppose that for each A G cA and each h GH, h~ÂhGcA. Let ~ denote the equivalence relation defined by A ~ S if S is an ^-conjugate of A. Let '7^ be a set of representatives of the
equivalence classes.
Then
Ic^l
\H:Nh(R)]
Sec. 5.1]
THE SYLOW THEOREMS
135
Proof:
We know
from the remarks above that

Reg;.
But
R~
is
R~
= {X\ X^h-'^Rh for some h the set of H-conjugates of R. The

Hence
\c4\
& H)
since
number
of such H-conjugates
h^^Rh G cA for every hGH. is, by Lemma
So
5.4,
[H:N{R)].
Corollary
Keg;.
5.7:
[^:A^H(i2)],
as claimed.
Let P (?^ 0) be a subset of G. be as in Lemma 5.6. Then

\^\
Let c^
{flr-iPflr
\
g gG}.
Let
'R, i?
and
[H:Nh{R)]
rG:iVG(P)]
Proof:
Clearly
\cA\ is
the
number
of G-conjugates of P, and the result follows
from
Lemma
5.4.
Corollary
5.8:
Let cA = {A\ A is a subset of G and A has precisely one element). Let ^ be the equivalence relation in cA when = G, and let '5^ be a set of representatives of the equivalence classes. Let %* {R\ RnZ{G) = ^,
RG%}.
Then
\G\
\Z{G)\
2
xg
[G:Nc{R)]
(We remind the reader that Z{G)

Proof:
Clearly
\cA\
{x\
= gx
for
all
G})
\G\;
hence
\G\
S Re%
[G:NciR)]
(5.1)
If z G Z{G), then {z} G cA and the number of G-conjugates of {z} is one, namely {2} itself. Consequently {2} G for each z G Z{G). Note that Ng{{z}) = G if zG Z{G). Hence adding first the contribution made by all with i2nZ(G)?^0 in (5.1), we obtain
Rg%
\Z{G)\
and the
result follows.
i2
Note that as
{r},
NaiR)
= = =
{g\
{g\
GG g GG
g
and g-'rg G R} and g~^rg
r}
C{R)
(For the definition of C{R), the centralizer of
in G, see Section 4.3d,
page
112.)
Hence Corollary
5.8 takes the

\G\
form
\Z{G)\
[G:C{R)]
(5.2)
(5.2) is called
the class equation of G.
Problems
5.11.
If
Is
a non-empty subset and
a subgroup of a group G, then
Nu{A)
is
a subgroup of G.
Solution:
A^hI-A) is clearly
a subset of G. Nh(A)
implies
0, since
leH
and I'Âl
=A
implies
lGNi,(A).
nAn-^ = A, or (M-i)-iAm~i = A. Furthermore n"-i S H, since H is a subgroup; hence n~i G Afjj(A). If m,n G Njj(A), (mn)~Â('m,n) = n~i(w-iAm)m = n~Â7i A; hence mw S A^h(A). Accordingly A^h(-A) is a subgroup of G.
Let
nSNuiA). rfÂn = A
136
FINITE GROUPS
Check Lemma 5.4 by direct computation when = {., r^}. page 57) A = {tJ,
[CHAP.
5.12.
G =
Sg
and (using the notation of Section
3.3a,
Solution:
The H-conjugates of A are i-UtJi = {t^}, t~^{ti}t2 = {rg}. Thus the number of H-conjugates A is 2. Lemma 5.4 requires that 2 = [H N{A)]. But iV^CA) - {x\ xGH and x-Âx A} = {i}. Hence [H Nt,(A)] = 2, as required.
of
:
5.13.
Check Corollary
(i2
5.7
when G =
a"î6a
Dg,
ft-i;
H=
F2
68
and
the dihedral group of degree 8 (G see page 75, with a = t, b =
{6, a6*,
<
8},
a^),
given
{a}
where and
{1,62,64,66}.
Solution:
qA of Corollary 5.7 is given by c^ = {g-^Pff g {a62}, Fg = {06*}, F4 = {a66}; thus \a4\ 4.

\
G G} =
{Pj, Fj, ^ 3. -P4} where Fj = {a}, Using the equivalence relation ~ of the
corollary,
Fi~ = Fa- =
{h-^P^h
{/i-iPa/i
I
h^H} =
/le//}
{Fi,Fg}
{Pi.Pi)
For '^ choose one representative from each of the equivalence classes, e.g. choose % = {Pg.Fa). Then Lemma 5.6 claims that |o^| = 4 = [H Nh(Ps)] + [H NjiiP^}]- Now NuiP^) = {1,64} ^ NaiPi). Hence [H N{P3)] = [H NiP2)] = 2 and the required equation of Corollary 5.7 holds. We must also show that [G A^g(^)] = 4. As
:
NciP)
= = =
{x
I
a;
{x\ X
G G and xîPx = F} E G and x~âx a}

a64}
{1, a, 64,
the index of Nq{P) in
is 4,
the required number.
5.14.
Check Corollary
Solution
P5
:
5.8
when G =
S3.
Use the notation
of page 57.
c^ (of Corollary 5.8) = {Fi,F2,F3,F4,F5,F6} where P^ = U}, P^ = M, P3 = W. Pi = (ti), = {'2}- ^6 = {ts}- Let ~ be as in Corollary 5.8. Consequently Fj ~ = {Fj}, P^- = {Pai-Ps}. F4 ~ = {P4, F5, Fg}. To define "i^, we choose one element from each of these equivalence classes. Let us take % - {P^,P2,P^. As Z{G) ~ {(}, PjOZiG) ~ except for j = 1. Therefore Now iVg(F2) = {i, CT2} and A''g(F4) = {i, tj}. Hence, as required, "Jf* = {F2, F4}.
CTi,
\Z{G)\
[G-.NaiP^)]
[G-.NciPd]
of a
IS3I
5.15.
Show that
Solution :
N[j(A)
= NQ{A)nH
for any non-empty subset
and subgroup
group G.
Let nGNuiA); then n&H and m~iAm = A. But H qG, so that nGG and by definition nGNciA). Consequently N[j{A) cNQ(A)nH. If nG Na(A)nH, then n'Ân = A and n H. Thus NG(A)nH qNi,{A) and the equality follows.
5.16.
Prove Proposition
Solution:
5.5,
page 134.
contains the identity, so A = 1~1A1 and thus A ~ A. If A ~ B, As /f is a subgroup of G, then there is an element such that h~Âh B. Consequently (h~^)-'^B(h~^) ~ A and then there exist h,gGH such that h-Âh = B and B ~ A. Finally, if A ~ B and and (M"'A(%) = sr-i(fe-iA;i)ff = g-^Bg = C, and so g-^Bg = C. It follows that hg G A ~ C. Hence ~ is an equivalence relation on cA.
hGH
B~C,
5.17.
Let
that
A,B
[H:Nh{A)]
Let
be subsets of G. Suppose = [H Nh(B)].

:
is
an H-conjugate of A, where
is
a subgroup of G.
Prove
Solution:
A ~=B
or X = g~^Bg for some 9 e G}. We use Proposition 5.5. ~, as B is an /f-conjugate of A. \A ~| is therefore the number of /^-conjugates of A, and also the number of H-conjugates of B. By Lemma 5.4, [H iVjj(A)] = [H Nh{B)].
: :
cA = {X\X ~ g~Âg
Sec. 5.1]
THE SYLOW THEOREMS
137
c.
Proofs of the Sylow theorems

First
we prove
5.9:
a
If
weak form
of the first
Sylow theorem.
is
Proposition
is
then
a finite abelian group and p has an element of order p.
a prime dividing the order of G,
Proof: We will prove the proposition by induction on the order of G. If |G| = 1, there nothing to prove. Assume the proposition is true for all groups of order less than n, the order of G, where n > 1. Recall from Section 4.2b, page 105, that if G is cyclic there is a subgroup of order any integer that divides |G|. Thus if G is cyclic the theorem holds, and we may therefore assume G is not cyclic. If n is a prime, G is cyclic; hence n is not a prime.
is
gG, h of order m. Clearly <n. Let be the cyclic group gena proper subgroup of G. Now if p\m, by the induction assumption, has an element of order p. If p/m, form the factor group G/H (every subgroup of an abelian group is a normal subgroup so i? <1 G). Since \H\ > 1, \G/H\ < \G\. As \G/H\ |G|/|fl^|, p |G|/|H|. Therefore by the induction assumption, G/H has an element g of order p.
Suppose h
h.
{=1)
is
erated by
be the natural homomorphism of a group onto its factor group (see page 114) and g he a preimage of g under v. Now (fl"')v = g" = the identity of G/H, so g" G H. As H is of order m, (gf)" = (g")"" = 1. Therefore g"" has order p or If flr = 1, then g'^v = "" = !. Since g has order p this implies p divides m, confir = 1. trary to our assumption. Therefore g"" is an element of G of order p. We are now in a position to prove the Sylow theorems. For convenience we repeat
Let
V
:
G -> G/H
Theorem
4.17,
the statement of each theorem.
The
First
Sylow Theorem:
be a finite group, p a prime and p'' the highest power of p dividing the order of G. Then there is a subgroup of G of order p''.
Let
Proof:
We
will
theorem Suppose
(i)
is trivial.
\Z{G)\
=
I
c.
prove the theorem by induction on the order n of G. For \G\ = 1 the Assume n> 1 and that the theorem is true for groups of order < n. We have two possibilities: (i) p\c or (ii) p/c.
order p.
Suppose p c. Z(G) is an abelian group. By Proposition 5.9, Z{G) has an element of Let iV be a cyclic subgroup of Z{G) generated by an element of order p. N < G, since any subgroup of Z(G) is normal in G. Consider G/N. Then |G/iV| = n/p by Corollary 4.14, page 110. Hence by our induction assumption, G/N has a subgroup H
of order
p''~^.
As
Corollary 4.21, page 121, there exists a subgroup of G such that H/N = H. = \H\ = \H\/\N\ = \H\/p, we conclude that \H\ = p\ Thus in this case, G has a subgroup of order p"".
p"--'
By
(ii)
Suppose p}' c.
The
class equation for

\G\
is (see
Equation
(5.2)
of Corollary 5.8, page 135)
\Z{G)\
[G:C{R)]
Therefore for at least one
i?
Since p
jG|
[
and p/c, we have
p/
[G:C{R)].
"iil*,
p/[G:C{R)]. But \G\ = [G:CiR)]\C{R)\ by Corollary 4.14 to Lagrange's theorem, page 110. Hence p'-\\C{R)\, since r||G|. Now \C(R)\^\G\; for if \C{R)\ = \G\, then C{R) = G and RnZ(G) = R, contrary to the assumption that Rr\Z{G)-0. Thus by the induction assumption, C{R) has a subgroup H of order p"". Consequently so
does G.
In either case
we have found a subgroup
of order p".
The proof
is
complete.
138
FINITE GROUPS
[CHAP.
subgroup H of G, where Sylow theorem.
The following gives a simple formula for the normalizer of a Sylow 2)-subgroup P in a \H\ is a power of p. It will be used in the proof of the second
5.10:
Lemma
If
is
finite
group,
p,
of order a
power of
a Sylow p-subgroup of G, and i? then
is
a subgroup of
Nh{P)
HnP
Proof: PHHcNuiP), as conjugation by an element of P sends P to itself. We show N{P)cPnH. Nh{P)CNg{P) and P <i Ng{P) (see Problem 5.15 and Problem 4.60,
page
have:
113), so that
Nh{P)P
is
by the subgroup isomorphism theorem (Theorem 4.23, page 125) a subgroup of G and
we
Nh{P)P/P ^ NH(P)/NHiP)nP
Consequently [A^h(P)P: P] = [Nh(P) A^H(P)nP]. But Nh(P) is a p-group, i.e. a group of order a power of p, since it is a subgroup of the p-group H. Thus [A^h(P) A^H(P)nP] is a power of p. [A^h(P)P P] is therefore also a power of p and, as P is a 2?-group, |iVH(P)P| is a power of p. Accordingly, Nh{P)P is a p-group. But P C Nh{P)P and P is a Sylow p-subgroup. Hence P = Nh{P)P, for P cannot be a proper subgroup of any other p-subgroup of G (see Problem 5.4, page 132). Nh{P) is therefore a subgroup of P. As Nh{P) C H, we conclude NH{P)QHnP.
: : :
The Second Sylow Theorem:
Let
he a subgroup of a finite group G, and let P be a Sylow p-subgroup of G. If H is a p-group, then H is con-
tained in a G-con jugate of P. Proof:
We
apply Corollary
\^\
5.7,
page 135, to c^
[g~^Pg
\
G G}
to conclude
S
:
[H:Nh{R)]
[G:Na{P)]
By Lemma
5.10,
NH{R)=^HnR
for each
RG%.
=
Hence
[H:
[G A^g(P)]
If
R^%
'^
HnR]
(5.3)
a p-group, the right-hand side of equation (5.3) is by p. But P c A^g(P), so that p does not divide [G:A^g(P)]. This contradiction implies that for at least one But as R GcA, R is a, G-conjugate of P. The result follows.
all
fl^
HnR - H
by
for
R G%,
:
as
is
divisible
p.
Hence [G Ng(P)]
is
divisible
HnR-H
RG%.
The Third Sylow Theorem:
(i)
Any two Sylow

(ii)
p-subgroups of a
(iii)
finite
jugate.
G
Proof:
(i)
is
The number Sp of congruent to 1 modulo p.
distinct
Sp
|
group G are G-conSylow 25-subgroups of
|G|.
and P' be two Sylow p-subgroups of G. By the second Sylow theorem, P', as a some G-conjugate R of P. But |P'| = |P|, by Problem 5.3, page 132. Hence P' - R and P' is conjugate to P under G.
Let
23-group, is contained in
(ii)
to
P be any Sylow p^subgroup of G. Since any other Sylow p-subgroup is conjugate P and any conjugate of a Sylow ^-subgroup is a Sylow p-subgroup (Problem 5.5, page 132), we conclude by Lemma 5.4 that
Let
Sp
[G
A^g(P)]
But on putting
P=
in
Equation
Sp
{5.3),
we have
[P:
Pr\R]
Sec. 5.2]
THEORY OP p-GROUPS
itself
139
Now for exactly one R&%, R = P; for the only P-conjugate of P is P P is the only possible representative of its equivalence class. In all PnRj^P. Therefore [P:PnR] is a power of p for all R G % except this one [P:PnR] = l. Hence
Sp
and so
other cases, one, and for
+ kp
\G\
(iii)
By
:
Corollary 4.14 to Lagrange's theorem,

Sp
I
[G: Ng(P)] |iVG(P)|.
Since
Sp
[G NciP)],
\G\.
5.2
a.
THEORY OF p-GROUPS
The importance
of p-groups in finite
groups
Suppose that G is a finite group. In Section 5.1a we saw that G has a Sylow p-subgroup for any prime p. (p will be a prime throughout this section.) One reason why the study of p-groups (groups of order a power of p) is so important is that the structure of the Sylow p-subgroups of G partly determines the structure of G. One instance is the following theorem: If G is a finite group whose Sylow p-subgroups are all cyclic, then G has a normal subgroup A^ such that G/N and A^ are both cyclic. (M. Hall, Jr., The Theory of Groups, Macmillan, 1959, Theorem 9.4.3, page 146.)
In this section
we
shall
determine some of the elementary properties of p-groups.
b.
The center
of a p-group
finite ?)-groups is
very important property of

5.11:
given by
is
Theorem
If
v^
1.
{1}
and
is
finite
p-group, then Z{G), the center of G,
not of
order
Proof:
We make
use of the class equation (equation

\G\
(5.2),
page 135)
(5.2)
\ZiG)\
[G
C{R)]
It follows
immediately from the definition of C(R) and Z{G) that C{R)
=G
if
and only
if
RcZ{G). Because the sum on the right side of (5.2) is taken over all R such that [G: C{R)]. and because |G| = p', p\[G: C{R)] for all R G %*. Hence p\ RnZiG) = ^^'K* Since p \G\, we can conclude that p \Z{G)\, which means Z{G) > {1}.
Corollary
5.12:
If
is
a group of order p\
r1,
then
has a normal subgroup of
order
p^~'^.
Proof: The proof is by induction on r. The statement is clearly true for r = 1. SupBy Theorem 5.11, Z{G) - {!). pose the corollary is true for all k <r where r > 1. = gp{g). Because p \Z(G)\, Proposition 5.9 implies Z(G) has an element g of order p. Let = p"-^. There<} G, since any subgroup of Z{G) is normal in G. Consider G/N. \G/N\ of order p''"^. By Corolfore by the induction assumption, G/N has a normal subgroup and such that of G which contains lary 4.2^, page 121, there exists a subgroup = H. Then \H\ = p'^'K Furthermore, again by Corollary 4.21, Thus G <i G. H/N p"""^ and the proof is complete. has a normal subgroup of order
N H
Clearly
we
could repeat this argument until

{1}
we
obtain a sequence of subgroups of
= Ho C
. .
i?l
C Hr-l C Hr = G
p'
(i
{5.i)
where Hi
<i
Hi+i
(i
0, 1,
.,r-l) and
\Hi\
0, 1,2,
., r).
140
FINITE GROUPS
[CHAP.
Problems
5.18.
Suppose
Solution :
is
a group with
S a subgroup
of the center Z{G).
Prove
is
abelian
if
G/S
is cyclic.
aS.
Suppose G/S is cyclic. Then we can find aS Then if g,he. G, g = W-z,

i
G
h
such that every element of G/S
is
a power of
a>z'
z'
for a suitable choice of the integers
and
j,
with
and
in S.
Then
a'z'aiz
gh = akaiz'
aWzz'
a'a'z'z
is
hg
Thus every pair of elements of G commute.
Hence G
abelian.
5.19,
Prove that a group of order p^

Solution:
is
abelian (p a prime).
Let
abelian.
is
Suppose
be of order p2 and let Z be the center of G. By Theorem 5.11, Z = {!}. li Z = G, G Z ^ G; then \G/Z\ = p, so G/Z is cyclic. By Problem 5.18, it follows that
is
abelian.
5.20.
Let A = {0,1, ...,p group. Let
1},
where p
is
a prime.
{(a, 6,c)
Then under addition modulo

I
p,
is
an abelian
G =
be the set of
all triples (a, 6, c)
a,b,ceA}
of elements of
A.
(a
Define
(a, b, c)
(a', 6', c')
+ a',
G
is
+ 6',
+ c' - 6a')
Prove that with respect to this binary operation,

Solution:
It is clear that
\G\
a non-abelian group of order p^.
jfi.
To prove that G
h" c")
,
is
a group,
ft',
we check
c'-
first the associative law:
{(a, b, c)(a', h', c'))
(a",
= = = =
(a 4- a', 6
{a
+ + a' + a",
c)(a'
ha')(a", b", c")
+
ft'
b'
b", c
+ c' + c"- ba' - (6 + b')a")
On
the other hand,

(a, 6, c)((a', b', c'){a",
6", c"))
+ a", + 6", c' + c" - ft'a") (a + a' + a", + 6' + 6", c + c' + c" - b'a" {a,
ft,
ft
b{a'
a"))
We
check that
c
c'
+
is
c"
ba'
which
is true.
Thus G
a semigroup.
(a,
ft,
c)
+ 6')t" = c + c' + c" Now (a, 6,c) = (0, 0, 0) (0, 0, 0) =

(6
b'a"
ft(a.'
+ o")
(a, 6, c)
and so (0,0,0)
is
the unit element of G.

(a,
ft,
Finally,
(0, 0, 0)
c)
{a, ft,
c ba) = G
is
{a, ft,
ba)(a,
ft,
c)
and hence every element of
has an inverse.
non-abelian.
(1,1,0),
Our
last task is to prove that
Now
(0,1,0)(1,0,0)
(1,0,0)(0,1,0)
(1,1,-1)
and thus
(1, 0, 0)(0, 1, 0)
(0, 1, 0)(1, 0, 0).
5.21.
Let A be the additive group of integers modulo .,p^ group of integers modulo p^, B = {0,1, j G B. Prove that under the binary operation
.
p,
i4
{0, 1,
.,
1};
and
let
(i
l}.
Let
be the set of
all
B be the additive pairs (i,j), i&A,
(i,i)'(i',i')
+ i',i + r + n'p)
is
a non-abelian group of order
p^.
Solution:
Clearly
is
of order p^.
We
check that
is
a semigroup.
Sec. 5.2]
THEORY OF p-GROUPS
((i,
141
JW. J')){i", }") =
+ i', } + y + n'p)(i", J") = (i + i + i", i + i' + j" + ji'p + U + j' + ji'p)i"p)

(t
(i.
(i, j){ii'.
}')ii", j"))
= =
JW + i", r + i" + i'i"P)

+ i' + i", j + j' + j" + j'i"p + j(i' + i")p)
G
is associative,
(i
To prove that the binary operation

ji'p
in
we need check
only that
+ a + j' + ji'p)i"p =
j'i"p
}(i'
+ i")p
-j +
Thus G
Since p^
in B, this equality is readily verified.
The
identity element of
is (0,0).
The inverse of
(1,1),
(t.j) is {-i,
jip).
is
a group.
^"'*"'''
(1,0)(0,1)
(0,1)(1,0)
(1,1
+ p)
and therefore
is
non-abelian.
5.22.
Prove that the group G in Problem p is odd. Is this true if p = 2?

Solution:
5.20
has the property that for
all
e.
G, g"
(i.e. (0, 0, 0))
if
Let
(a, 6, c)
G.
Then
(a, b, c)2
= =
(a, b, c)(a, 6, c)
= = =
(2a, 26, 2c
6a)
Continuing,
we
find
(a, 6,c)3
(a,6,c)2(a, 6,c)
(2a, 26, 2c
(3a, 36, 3c
- 6a)(a, 6, c)
6a
- 26o)
By
induction
it
follows that
(a, 6, c)"
(pa, pb,
pc
ba 2ba - l)6a =
(p l)6a) ~
-^^
But pa
0,
p6
0,
pc
0.
Finally,
6a
since
l
26a
(p
^^^
6a
+ 2+---+p 1 =
by
p.
2,
divisible
If
Hence ^p{p
then
^p(p 1). If p is odd, p 1 is even. l)ba 0. Thus we have (a, 6, c)p =
Therefore ^p{p
1.
1)
is
an integer
(a, 6, c)2
(2a, 26, 2c
- 6a) =
(0, 0,
-6a)
In particular if a = 1, 6 = 1 and c = 0, we have (1, 1, 0)^ = (0, 0, 1). Thus not every element of G is of order 2. This result could have been observed by noting that a group G satisfying g^ = 1 for all ff in G is abelian. To see this let g,hG G. Then as {gh)^ = 1,
gh
and so
=
G
is
(gh)-'^
= h-ig-i =
h^h-ig^g-i
is
hg
2.
is
abelian.
But as
not abelian, not every element
of order
5.23.
If
is
odd, does the group
of Problem 5.21 satisfy
gf
(i.e.
(0,0))
for
all
g &G'>.
Solution:
No, since
(0, 1)2
(0, 1)(0, 1)
(0, 2).
Inductively,
(0, 1)"
(0,
p) f^
(0, 0).
5.24.
Prove that if G is a group such that has the same property, i.e. h" = 1 for
Solution:
ffP
=
h
for all
G,
then every homomorphic image
H
h.
all
&
H.
Let 9 be a homomorphism of G onto H. Then Therefore h" - (ge)" = ig'')e = Iff = 1.
if
G H, we
can find g
such that ge
: :
142
FINITE GROUPS
Prove that
Solution
if
[CHAP.
5.25.
is
an odd prime, then the groups in Problems 5.20 and
5.21 are not isomorphic.
Let G be the group defined in Problem 5.20 and let H be the group defined in Problem 5.21, for p an odd prime. Then by Problem 5.22, if g G G, gf = 1. But if G = H, it follows from Problem 5.24 that hp = 1 for all hSH. But by Problem 5.23, (0, 1)" ^ 1. Therefore G is not isomorphic toH.
5.26.
Prove that a non-abelian group

Solution
of order p3 has a center of order
(p
a prime).
be the center of G. By Theorem 5.11, Z = {1}. Also, Z t^ G since G is non-abelian. = p2, then \G/Z\ = p. Therefore G/Z would be cyclic and hence, by Problem 5.18, G would be abelian, a contradiction. Thus \Z\ = p.
Let
if
Now
\Z\
c.
The upper
Suppose
central series
is
a group.
We
shall define a series
{1}
Zo
Zi
of subgroups Zo, Zi,
Zo
(1),
and Zi
to be the center of G.
subgroup of G/Zi is the center of G/Zi is of the form Z2/Z1 (we are using Corollary 4.21, page 121). Notice that as the center of a group is a normal subgroup, Z2IZ1 is a normal subgroup of GIZu Therefore by Corollary 4.21, Z2 is a normal subgroup of G.
In general, once Zi has been defined and proved to be a normal subgroup of G, Zi+i/Zi to be the center of GIZi. By Corollary 4.21 it follows that Zi+i < G.
of G, called the upper central series of G. begin by defining Next we define Za. We look at G/Zi. Since every uniquely of the form H/Zi where is & subgroup of G containing Zi,
.
.
We
we
define
We
number
Our
shall call a
group
nilpotent
if its
upper central series ascends to
in a finite
of steps.
objective in this section
5.13:
is
to prove
is nilpotent.
Theorem
Proof:
If
5.11.
finite
p-group
{1}, there is nothing to prove. It G - {1}, then Zi - {1} by Theorem not the identity, the center of G/Zi = Za/Zi - Zi/Zi, again by Theorem 5.11. Notice that if Zi - G, then Z2 - Zu Similarly if G ' Z2, Z3 = Zi. By induction we can show that if Zi=G, Zj+i =5^ Zi and thus
U Gis
G/Zi
Zo
C
k.
Zi
Zi
is
Zi+i
Since
is finite,
Zk
=G
for some
Therefore
nilpotent.
Problems
5.27.
Prove that
Solution:
if
a non-abelian group
is of
order p3, then
Z2
G.
Zi ' {1} by Theorem 5.11. So if Zi ^ G, then G/Z^ is of order p or p2. Since G/Z^ is cyclic only if Zi = G (Problem 5.18), we find G/Z^ is of order jj2 and hence abelian (Problem 5.19). Therefore Z2/Z1 = G/Zi, i.e. Z2 = G.
5.28.
Let I> be the dihedral group of order 2n. Prove that D

Solution:
is
nilpotent if
is
a power of
2.
D^ has
. .
the property that

is
it
and every element of D

.,n
contains two elements a and 6 such that a^ = uniquely expressible in the form 0*6' where
1^
t
6"
1,
0, 1
a-^ba and j
l.
= b~^ = 0, 1,
(See Section 3.4f, page 75,
where a
and b
agO
Sec. 5.3]

Suppose n
143
Method
1.
2"".
Then
a
6^
is
of order
o"*~l
2.
,
Hence
oTn
_m
a-162
(a-i6a)2
(6-i)^
6.
,oW
h^
commutes with a; clearly 6^ So 6^ 6 Z^. If element of G, and so 6^
commutes with
Therefore 6
otherwise a
commutes with every
m=
=
1,
then Z^
If
=
b'
i3;
î.
a^b'.
What
other elements can be in the center?
a'b'
the other hand, a-^(aibi)a
not
aâ-^b'a
a^b-i.
Hence
e Z^, then clearly = b-' and (6^2 = 1^

om
is
the only possible element in the center other than

implies
6''
o"
ab
}.
i2"'~*
.
a-â'b')a = 62" i.e. fti = 12"' But as o 6
On
jjj^g
this
>
/î,
2i,
which
is
so.
Consequently Z^
,
.
=
.
{1, 6
Z_i
Similarly Z2 consists of the powers of 6^ consists of the powers of b^. Now this
. ,
2; consists of the
means
\G/Z^-i\
4,
powers of 6^ and so G/Z^-^
Therefore
is
abelian.
Thus Z^
Method
5.29.
2.
=G
and
is
nilpotent.
2.
|jD|
some power of
Hence we can apply Theorem
5.13.
Prove that A4
Solution:
is
not nilpotent.
Zi
{1}
in A4, as a direct check shows.
Hence ^1 =
^2=
, and thus Z
A4
for every n.
5.3
a.

Direct products of groups
all
In Chapter 1 we defined the cartesian product K of two sets ordered pairs {h, k), h G and k G K. If and K are groups, plication of elements of as follows. Let {hi, fci), (/12, ki)
Hx
and
as the set of
HxK
gHxK
we can
define a multi-
and define
{5.5)
{hi, ki)
{h2, k2)
{hih2,kik2)
where
{5.5) is
A1A2
and
A;iA;2
are the products in the groups

see that
[{hi, ki)
'
H and K
The
set
respectively.
The
multiplica-
tion defined in {5.5) is clearly a binary operation.
a group.
To
HxK
fcz)]
HxK
is
a group, let
{hi, ki), {h2, ki), {ha, ka)
with binary operation {ha, ks) G x K. Then
{h2,
{ha, ka)
= = =
{hih2, fcife)
{{hih2)hs, {kik2)ka)
{hi{h2ha), ki{k2ka))
{hi, ki) {hi, ki)
{h2ha, kzka)
[{h2, fe)
{ha, ka)]
is therefore an associative binary operation on neously for the identity element of and of and {h, k) G
Multiplication
H xK. Hx K,
If 1 stands simulta-
{hl){h,k)
so that
its
(1, 1) is
{h,k)
{h,k){l,l)
{h, k)
an identity of
HxK.
=
It is clear that if
gHxK,
then {h-^,
fe-i) is
inverse, for {h, k){h''^, k'^)
{hh'^, kk'^)
(1, 1).
tion (5.5) is called the external direct product of the as just the direct product. define the internal direct product after Proposition
HxK
5.19.
with binary operaThe group groups H and K. We often refer to
HxK
We
If
H and K are finite
groups, then
it is
clear that
\HXK\ =
- {1} If to K, because
\H\\K\
simple way of order 2 generated by g.
and K' {1} are finite groups, then HxK is neither isomorphic to H nor \HxK\= \H\ and \HxK\y^ \K\. Therefore the direct product gives us a constructing new finite groups. For example, let C2 be the cyclic group of
C2XC2 =
{(fl'.l),(fl',fi'),(l,fl'),(l,l))
144
FINITE GROUPS
[CHAP.
Now
4,
IC2
C2I
namely:
= 4, so we have (as we shall soon see) two non-isomorphic groups of order the cyclic group of order 4, 0*= {1,&,6^&3| ^here 6^ = 1^ ^nd the group
C2 X C2.
The
multiplication table for
dxd
(1,1) (1,1)
is
a, 9) a, 9)
(1,1)
(9,1) (9,1)
(9,9)
(9,9)
(fir,l)
(1,1)
(1,^)
{9 A)
a. 9)
iff.i)
{9.9)
(1,1)
(9,9)
(9,1)
(h9)
(1,1)
(9,9)
(g.ff)
ihg)
Note that all the elements of x C2 are of order 2. Hence C4 is not isomorphic C2 X C2 is called the Klein four group, or simply the four group.
to
d x d.
sets
Theorem
5.14:
If
G=
HxK
is
the direct product of the groups
and K, then the
H
k
ab = ba. subgroup of G.
Proof:
since hih^'^
is
= =
{{h,l)\
{{l,k)\
hGH, kGK,
1 the identity of 1 the identity of
H}
K}
and
if
are subgroups of G.
Furthermore,
H = H,K = k;
and
aGH
G
and
bGK,
Finally,
G=
Hk
^n^ =
{(l,l)},
the identity
If
ihul),{h2,'i-)^H,
then
(hi,l)(h2,l)~\^
(/ii, l)(/i2"',
1)
{hih^'.l)
gH.
is clearly
non-empty.
a of
a:
Therefore
i? is a
subgroup of G.
Similarly
H K
a subgroup of G.
The mapping
H onto H defined by
h^{h,l),
hGH
are isomorphic.
/v
is
clearly
and S
an isomorphism. (1, A;) G K; then
Similarly
K
=
and
Now
>\.
let
a~{h,l) G
a&
(A,l)(l,fc)
(h.k)
(l,A;)(/i,l)
= 6a
Now Hnk^^
Any element (/i,A;) of G can be written Clearly -fffi" C G. Hence G = HK and Theorem 5.14 follows. Corollary 5.15: Let G = HxK and H.K be as in Theorem
{(1,1)}-
as {h,l){l,k), so
GcHK.
5.14.
Then every
gGG
can be written uniquely as a product hk where

Proof:
If
hGH, kGK.
element in
g = {h, k), then g = {h, 1)(1, k) is an expression for g as the product of an by an element in K. If we also have g = {hi, 1)(1, fci), then clearly hi = h and /ci = &. Thus the expression is unique. As a converse we have
Theorem
Proof:
5.16:
Let G be a group with subgroups and K such that G. Then elements of commute with those of K, and
HnK~ {1},
the
HK
G = HxK.
g = hk where hGH and kGK. Since G HK, g hk for some hGH and kGK. = h2k2 where hi, hiGH and Suppose g = hiki and ^2 G K. hiki ^2^2 implies hs^hi^kikiK But ifnZ={l}, and so A^'/ii = 1 and A;2fcr*=l. Hence hi = h2 and fci = k2. We define the mapping a: G^ HxK by ga = (h,k) where g = hk G G. a is a one-tof
We
first
show that any element
gGG
can be written uniquely in the form

fci,
one mapping, for we have shown that there is one and only one way of writing g in the form g = hk, and the elements of H x K are of the unique form (h, k). To prove a is a homomorphism we must demonstrate that if gi = hiki and g2 h2k2 are any two elements in
G, then
Sec. 5.3]
145
Now
Hence
{hjc^hji^)cc
(\h^k^K)a
{Kh^.k^k^)
{h^,k;\{h^,k^)
i\k;)a{h^k^)a
a is a
homomorphism and the

if
result follows.
Note that
commute
are normal subgroups of a group elementwise. For if h G H, k G K,
H,K
with
HnK=
H
{1},
then
and
h-^k-^hk
Therefore
{h-^k-^h)k
gK
H
= h-\k-^hk) G
G-
HK
= {1}, and so and K commute elementwise. h~^k-^hk G and K are normal in G. and and K commute elementwise implies
HnK
Clearly
Consequently Theorem 5.16 can be stated as follows:

Corollary
5.17:
Let
G be a group with normal
subgroups
H and K, and suppose HnK = {1},
and
HK^G.
Then
G^HxK.
The hypothesis of Theorem 5.16 asserts G must equal HK. But if G is a finite group and \HK\ = \G\, we can conclude, since HK c G, that HK = G. It is useful to be able to
count the number of elements in
Proposition 5.18:
If
HK. We
therefore prove the following proposition.
is
a finite group with subgroups
H
.
and K, then
]K]
_
Proof:
Let I
. . .
= HnK.
n
/ is a subgroup of
and, since
IcK,
7 is a
subgroup of K.
Let
Iki, /fe,
Ikn be the
distinct cosets of I in
K.

Thus
Ikn
K =
and, by Corollary 4.14, page 110,
/Zci
U 7fc2 U
n=
|JK:|/|/|
\K\/\HnK\.
We
claim
now
that
HK =
ft;
Hki U Hk2 U
U Hkn
For if hk G HK, then hk = (hl)kj = h'ki where
for some Then hki = suppose for some integers i and j. h.h'GH. Consequently h'-^h-kjki^, so kjkr^ G I -HnK. But kjkT^Gl implies e /ft;*. Since two cosets are either equal or disjoint, Ikj = Iku Hence ki = kj. that
Now
ft;,
= Ikj h' G H, HkinHki=0
for some as both h,
IGI,
I
an integer between 1 and n. belong to H. Thus HK = HkiUHkiU

j
Hence UHkn.
h%
Thus HkinHkj =
for i=7^y and

|Hii:|
|Hft;i|
\Hk2\
\Hkn\
Now
\Hki\
\H\,
because hih
h2ki if
and only
if
hi
h^.
Therefore
\HK\
since
\H\ \K\
n\H\
\HnK\
n = \K\/\HnK\.
illustrate the use of Proposition 5.18, let
To
and
G
4.
subgroups of
also in
G of orders 7 and 4 respectively. H2 must have order dividing 7 and

'
be a group of order 28 and Hi and i?2 HinH2= {1}, because an element in Hi

Accordingly,
and
G=
77,^2.
146
FINITE GROUPS
Using Proposition
[CHAP.
5.18,
we can
finite
replace
Theorem
or
5.16 in the case of finite groups
by
Theorem
Proof:
(i)
5.16':
Let
be a
group with normal subgroups
If either
{i)HnK={l]
=
\G\
{u)HK^G,
then
H and K where G^HxK.
\H\ \K\
\G\.
HpK =
If
(1)
and
Since c G, we can conclude are fulfilled and
HK
|^| \K\
implies,
G^HxK.
=
1^1
HK =
by Proposition 5.18, \HK\ = \H\ \K\l\Hr\K\ = \G\. G. But then the hypotheses of Corollary 5.17
(ii)
HK = G,
then \HK\
\G\.
Therefore
l-'^^l
= iUTTW^
or
HnKllGl =
\H\\K\
But \H\\K\^\G\ by hypothesis. Hence
HnK={l}
and, by Corollary 5.17,
G^HxK.
The concept of direct product can easily be generalized to the direct product of a number of groups, Gi, G2, .,Gn (n^2). Let G = Gi x G2 x x Gn be the cartesian product of n groups. Define a multiplication in G by {91,92, .. .,gn){gi,g2, .. .,gk) =
finite
. .
i9igi, g^gi, ..., g^gi)

5^.30
for {gi, g^, ..., gn), (gi, gi, ...,gn)&G. G is then a group (see Problem below) called the (external) direct product of the groups Gi, G2, ., G. We denote G by
.
.
=i
ferently.
In Chapter 6 we will define the direct product of an infinite number of groups difProposition 5.19 below and Corollary 5.15 will provide a link between the two
definitions.
Proposition 5.19:
Let
H
(i)
and
K be subgroups of a group G. hk = kh for all and
If
h&H
k&K
H
and
(ii)
every element g G G is a unique product of an element in and an element in K, (i.e. g = hk, h G H, k G K; and ii g = hiki, hi G H, ki G K, then h = hi and k = ki),
then
Proof:
G^HxK.
We
need only prove
HnK=
H
{1}
to fulfill the hypotheses of
Suppose
(ii)
gGHnK. Then = implies h=l and

fc
g
1.
= h-l = l'k
Therefore g
for some
hGH
HnK =
and
{1}.
(i)
K.
Theorem 5.16. But condition
and
If G is a group with subgroups and satisfying conditions the internal direct product of and K, and we write G =
and
(ii),
HK^HXK.
Problems
5.30.
HK.
By
G is said to be Proposition 5.19,
Let
G =
Gi X G2 X
X G be the cartesian product of n groups.

(9v gi,---,
ff){gi,
Define a multiplication in
by
92,.-.,
g'n)
=
G
is
(gig'i,
g-igi ---, 9nffn)
for
(flTi,
g2
g), (gi, 92, ---, 9n)
G-
Show
that
a group.
Solution:
The multiplication
tive in each Gj.
is
is
clearly an associative binary operation in G, since multiplication
is
associa.
If 1 stands simultaneously for the identity of Gj,
1,2,
.,n,
then (1,1,
is
.,1)
clearly the identity of G. of g in G.
If
(g^gz,
-,9n)
G,
then
(9i^,9~^,.-.,g^^)
the inverse
5.31.
If
HsH
If
and
= K, where H, H,
and
fi
:
and
are groups, then
HXK
y:
H X K.
if
Solution:
a:
H -> H
K^K
is
are isomorphisms,
we
define
HxK-^HxK
{ha, kp),
h&H, k K.
a one-to-one mapping, for
(ha, kp)
(h'a, k'p)
by y: {h,k)-^ and only if ha = h'a
Sec. 5.3]
147
and
h
fc/3
k'p.
fc
Since a and
h'
and
k'.
are one-to-one mappings, ha = h'a and To show y is a homomorphism, let {h, k), (h', k') &
/3
fe/3
fe'/3
if
and only
if
H X K.
Then
[{h,k)-{h',k')]y
= =
{hh',kk')y
({hh')a,{kk')li)
{hah'a.kpk'^)
(ka,k/3)-(k'a,k'l3)
(h,k)y(h\k')y
Finally,
it is
clear that y
is
an onto mapping.
5.32.
Show
that
G=
HxK
and
is
an abelian group
if
and only
if
and
K
S
are both abelian groups.
Solution:
Suppose
are abelian groups.

(h,k)
Letting (h,k),{h',k')
G,
-
(h' ,k')
(hh',kk')
{h'h,k'k)
{h' , k')
{h, k)
and so
is
abelian.
(h', \)(h, 1)
Conversely, suppose G is abelian. Let h, h' e H. Then if 1 is the identity of K, {h, V)(h', 1) = Similarly we is abelian. or (hh', 1) = (h'h, 1). But this implies hh' = h'h. Hence
can show
is
abelian.
5.33.
X C^, Cg where C is the cyclic group of order n and X4 the four Consider the groups C^ X K^, group, i.e. the non-cyclic group of order 4, described above. Are any two of these groups isomorphic? Is any one non-abelian?
Solution :
Let C2 = gp(a), C4 = gp(b), and Cg = gp{g). We look at the set of elements of order 2 in each group. Since every isomorphism maps elements of order 2 onto elements of order 2, if there are more elements of order 2 in one group than in another, these groups cannot be isomorphic. Every element (?^ 1) of C2 X K4 is of order 2, for (c, fc)2 = (c^, A;^) = (1, 1) and (1, fe)2 = (1, fc^) = (1, 1) for any k G K^. Cg on the other hand has only one element of order 2, namely g*, because (ff*)^ = 1 it i^ i, i 7. Now C4 X C2 has at least one element of order 4, (6,1), and at least two elements of order 2, (6^, a) and (1, a). Therefore no two of the groups are isomorphic. As C2, C4, K^ and Cg are abelian, Problem 5.32 implies C2 X K^, C4 X C2 and Cg are also abelian. Thus we have exhibited three non-isomorphic abelian groups of order 8.
5.34.
If On and C^ are the cyclic groups of order n and C X Cm = Cm, the cyclic group of order rvm.
respectively
and
(n,
m)
1,
then
Solution:
Say C = gp{g) and C = gp(h). Consider the order of the element (g, h) in C X C^. We claim that the order of (g, h) is nm. If (g, h)^ = (1, 1) for some k, then (g'^, h") = (1, 1) and so g^ = and V<- = 1. Since the order of gr is n and the order of fe is w, m\ k and n k. Hence k is divisible by nm. On the other hand, (g, fc)"*" = (fl^""*, fe") = 1 and so the order is nm. Accordingly, C X Cm = gp{{g, h)). Therefore CX 0^^ = C^, since all cyclic groups of the same order are isomorphic (Theorem 4.7, page 103).
1.
|
5.35.
Show that Cs2 integer s > 1.

Solution:
is
not isomorphic to
C^XCg
(where C
is
the cyclic group of order n), for any
Since 0^2 is a cyclic group it has, by Theorem 4.9, page 105, one and only one subgroup of order But Cj X Cj has two subgroups of order s, namely gp{{l, g)) and gp({g, 1)), where g is the generator of Cj. Since subgroups of a given order are mapped onto subgroups of the same order by any isomorphism, Cs2 cannot be isomorphic to Cj X Cg.
s.
5.36.
Show
that for any prime p there are exactly two non-isomorphic groups of order p^.
Solution:
By Problem 5.19, page 140, we know that any group of order p2 jg abelian. Cp2, the cyclic group of order p^, and Cp X Cp, where Cp is the cyclic group of order p, are two non-isomorphic groups of order p2 (Problem 5.35). To see that these are the only possible groups of order p2, consider a
148
FINITE GROUPS
subgroup
is
[CHAP.
H of order p
=
cyclic, since
a cyclic
if
implies
Also \gp(a)\\H\
in G, a group of order p2. Such a subgroup exists by Corollary 5.12. H is p is a prime. Let a^H. The order of a is either p or pK If the order of a is p2 q group generated by a. If \gp(a)\ = p, then gp(a)nH = {!), for h (-1) & gp(a)'nH - firp(b) and gp(a) = flrp(b), since a group of prime order has no proper subgroups.
Theorem
5.16',
we
p2 and, as conclude
G = CpX
5.37.
G is abelian (Problem G = gp{a) X H. But
5.19),
gp(a)
< G
and
gpia)
XH^C^XC^
H<
G.
Therefore, using
5.31.
by Problem
Hence
Cp.
Show (Hi X H2) XHs =

Solution:
H^XHzX H3.
by *: {(K,h^),h^)
If
((Aj, ^2).
/13
Define
M>
(H.X
H^)xH^^ H,X H^XH^

*
is
and_fe3_eH3.
(^1, ^2. ^3)
Clearly
an onto mapping.
h^
and consequently
Sj,
fej
h^
and
^ (hh^,h^) for /ii G H, Aj e if^ = ((Ai. ^2), ^3)*, then {hi,h2,hs) = ^3)* = ^3. Therefore * is one-to-one. To show
(î X
,
is
a homomorphism,
let ({h^, h^),
/13)
and
((Aj, feg), ^3) (Cii.
iTj)
X H^. Then
(Ciiî,
[((Aj, /i2), fc3)((Si, ^2),
M*
and so
is
an isomorphism.
= = =
M(î.
^3^3)*
M2). /isy*
(''1^1,^2X2,^3^3)
((^1, h^),
(/ll,
^2, /l3)(Xi, ^2, X3)
hs)M(hi, h),
y*
b.
Groups of small order: orders p and 2p As an application of the Sylow theorems and the theorems of Section 5.3a we
will find,
up to isomorphism, all groups of order less than 16. We will use Cn to denote the cyclic group of order n, and K4 to denote the four group. Recall that K4 is defined to be C2 x C2. We refer to the notation of Section 5.3a. Let us put 1 = (1, 1), x = (1, g), y - {g, 1) and = {9,9)- The multiplication table for K4 is
1
1
X X
1
y y
z
1
y X
1
y
z
We note that xy ~ z, xz - y and yz x. Notice that the multiplication table is symmetric in X, y and z. If we put x = a and y = b, then z ab and we can write the multiplication table in the form
1 1
a a
1
b b
ab ab
b
a
6 b
ab
1
ab
b
a
1
ab
ab
isomorphism, clearly only one group of order 1. If p is a prime, any group of order p is cyclic (Problem 4.48, page 110). Up to isomorphism, there is one and only one cyclic group of order p (see Theorem 4.7, page 103). Thv^ there is one and only one group of order p, p a prime. In particular, the only groups of order 2, 3, 5, 7, 11 and 13 are cyclic.
is,
There
up
to
There are precisely two non-isomorphic groups of order U, namely Ct and 5.3a and Problem 5.36).
Ki
(Section
Sec. 5.3]
149
Next we show there are precisely two groups in each case of order 6, 10 or 14. Note that 6 = 2-3, 10 = 2 5 and 14 = 2 7, so these groups are of order 2p for some prime Let G be a group of order 2p, p an odd prime. By Problem 5.10, G has exactly one 29 T^ 2. subgroup K of order p, and either

(i)
exactly one subgroup
H of order 2
H
or
(ii)
precisely p subgroups of order 2.
(i)
is a In this case the group G is a cyclic group of order 2p. To see this notice that p-subgroup; so, by Problem 5.7, page is a unique Sylow unique Sylow 2-group and Furthermore Hr\K= {1}, for any element common to <i G. 133, <i G and must have order dividing 2 and p and hence is the identity. Clearly \H\ \K\ and x K. But and K are cyclic groups of 2p = \G\. Therefore by Theorem 5.16', G = order 2 and p respectively. Thus by Problem 5.34, G is cyclic of order 2p.
(ii)
Let
6^
K = gp{a)
1.
where
Clearly,
a" = 1. = KubK. G
1,
Since
is
Hence
the only subgroup of order p, b consists of the distinct elements

b, ba,
implies
a,a^
then
..., a"-!,
6a^
fta"-!
{5.6)
Now if
since
=
S
0,1,
.,p
l,
(ba^Y
ba'
and
ba'
=
is
a^-'b
of order
2.
(5.7)
and each element of
outside
ba'
K
=
Also
(&aO^
(ba}){ba})
implies
(6a*)-
(a')-'b-'
a^-^b
Now if G is any group of order 2p, then it is either of type (i) or (ii). If G is of type then by our analysis it must be a cyclic group of order 2p. By Theorem 4.7, page 103, cyclic groups of the same order are isomorphic. Hence all groups of order 2p having property (i) are isomorphic.
(i),
Suppose G is of type (ii). Then, arguing as above, p and an element h of order 2 such that
G
. .
has a subgroup
K = gp{d)
of order
G =
where for
i
(1, a,
.,0,"-^,
h,hd,
.,
bd"-^}
0,1,
.,p
l,
{bd})^
and
Id'
d'^-'b
{5.8)
The mapping a:G-*G
defined
by
a:
a: a'-* d\
is
b^b,
a:
ba^^
if
hd*
{i
any integer)
p divides i j and hence d' = d'. = a^ and so p divides i j and
. .
an isomorphism. First, a is a mapping; for Consequently a is well defined on a'. If ba'
a*
a',
ba',
then
a'
a'. Hence a is well defined on the ba*. As a: a'-^ a*, ba* -*hd' {i = 0,1, .,p 1), a one-to-one and onto, a is also a homomorphism, for gi, fl'2 S G implies gi = b'a* and Using equations Qi bâ'^ for some choice of j,sG (0,1} and i, t G {0,1,2, .. .,p l).
d'
is
{5.7)
and
{5.8),
we
obtain,
when
0,
{g^gâ
[(bw)(a')]a
(6'a*+*)a
^''^'
^'*^'
{b'a})a{a'-)a
= =
gyagâ
and when s
= 1, {g^gôc = [(bw)(ba')]a = {b'ba^-'a')a = (6'+'a''-'+')a = = h'd*hd* = {bW)a{ba% = gâgâ

(ii)
h'^^d^-*-^'
b^hd^-â*
Therefore any two groups of order 2p of type
are isomorphic.
150
FINITE GROUPS
[CHAP.
So far we have shown that up to isomorphism there are at most two possible groups of order 2p, p a prfihe. This does not mean that there exist two non-isomorphic groups of order 2p, for each prime p. But from Theorem 4.7, page 103, we know that for each positive integer n there exists a cyclic group of order n, and from Section 3.4f, page 75, we know that for each n the order of the dihedral group D is 2n and I> is not cyclic for rt > 2 (it is not even abelian). There are therefore, up to isomorphism, exactly two groups of order 2p for each prime p^2: one a cyclic group and the other Dp. In particular there are exactly two non-isomorphic groups of order i, i = 6, 10, 14.
It is worthwhile summarizing our method of finding all groups of order 2p. We first showed that if a group had order 2p it had to be isomorphic to one of two possible groups. The isomorphism in case (ii) was shown by using the fact that the elements of such a group had to satisfy equations (5.7) and (5.8). (As those equations determine the group up to
isomorphism, they are usually called defining relations for the groups; they will be discussed in detail in Chapter 8.) After demonstrating the isomorphism, we proved that each of the possible groups exists by exhibiting a group of each type.
c.
Groups of small order: orders 8 and 9
Let G be a group of order 8. There are at least three non-isomorphic abelian groups of order 8: Cs, C2 x and X K^ (see Problem 5.33). We show that if G is abelian it is isomorphic to one of these three groups.
If G has an element of order 8, G is cyclic. If G has no element of order 8 but has an = gp(a). Let b G G-H. If b is of order 4, b^ is an element element a of order 4, let of order 2 and lies in (since the coset decomposition of G is HUbH). As a^ is the only = 1. Since ab ^ H, we may element of of order 2, b^ = a^. Hence {aby = a^b^ = assume that there exists an element x of order 2 {x = b if & is of order 2 or else xâb). Let X =^ gp(x). XnH={l}. Therefore by Theorem 5.16', Since ^ Ci and = d, we conclude G = C2 x C4.
EG-H
aV
G^XxH.
If
a,
has no elements of order 4 or
8, all its
non-identity elements are of order
2.
Let
Then AB is a group Let gp{a), B gp(b). Problem 4.62, page 114. Now AnB= {1} and |A| |B| = \AB\. So, by Theorem 5.16', by AB ÂxB and consequently AB = C2XC2. Let cGG AB and C = gp{c); then CnAB =:{!}. Thus Gs.(C2XC2)xC2 = K4XC2.
b be distinct elements of order 2 in G.
A=
We
conclude that there are, up to isomorphism, exactly three abelian groups of order
8.
Assume G is a non-abelian group of order 8. G has an element of order 4 and no elements of order 8; for if g is of order 8, G = C. On the other hand if all elements of G are of order 2, then {aby = 1 for any a,b G G and consequently
GG
ba
be an element of order 4, contrary to our assumption that = gp{a). Then G = HUHb for some b G G. Also < G, as it is of index 2 and put (Problem 4.69, page 116). b^ G H; for if not, the cosets H,Hb,Hb^ would be distinct, and this would contradict [G: H]= 2. We have four possibilities for 6^: (i) b^ = a, (ii) b^ = a^,
= a%ab^ = a{abab)b = ab G is non-abelian. Let a GG
(iii)
&^
a^,
or
(iv) 6^
1.
If (i) or (iii) occurs, clearly are the only possibilities.

(ii)62
G=
gp{b),
contrary to our assumption.
Thus
(ii)
and
(iv)
a2.
or
a^.
If
b^âb
some integer
assumption.
b'âbGH. As a is of order 4, so is 6" %&. Thus b'âb - a ba. But every element of G can be written as a*6 or a' for i, since G HiJHb. Hence ab = ba implies G is abelian, contrary to our Thus b~âb = a" or ab = fta" {5.9)
Since
G,
H<
a,
then ab
Sec. 5.3]
151
Since
G = HuHb,
the elements of
can be expressed as
1, a, a^, a^, 6,
ab, a^b, a^b.
If a
group of this type actually exists, then equation (5.9), b^ enough information to construct its multiplication table.
1
a^
and
a^b
a*
provide us with
a2
a3
ab ab
a^b
a^b
a
o2
a2
aS
a^b
a^b
a
a2
a3
ab
a^b
a^b
a3
a
a2
a^b
ab
a^b
a3
a^b
ab
aSfi
a^b
ab
a^b
a2
a
a2
aS
ab
a^b
ab
a^b
a^b
a
a2
ab
a^b
a3
a?b
a^b
a^b
ab
a
5.1
tt3
a2
Table
To
calculate the products in the table,
we used
the fact that

ba.
ab
ba^
and
b^
a^
imply ba
a^b since a^b
a'^{ba^)
bâ^
= b{aV) =
If G is gpid)
another non-abelian group of order 8 with an element a of order 4 and an element such that h^ = d^, then as in our argument above,
G =
{1, a, 0?, a?, b, a?b, db, db^}
with the elements satisfying the equations

a*
=1,
(i2
52^
db
bd^
from which we
a
a
:
is
find a multiplication table for G which is identical to Table 5.1 except that substituted for a and b for 6. The mapping a: G-* G defined by a: a-> a and
a*6 ->
d%,
0,1, 2, 3,
is
clearly
an isomorphism.
Table 5.1 also shows that a group of order 8 of this type actually does exist, for the To see this, notice that the product of any two elements is again an element, i.e. the table defines a binary operation, 1 is an identity element, and every element has an inverse. The only difficulty is checking that the binary operation is associative. This involves much calculation (the reader should check some of the calculations himself). We shall give another description of this group in Problem 5.40. This group is called the Quaternion group and has the interesting property that all its subgroups are normal and yet it itself is not abelian (Problem 5.43).
table defines a group.
We now move
on to a discussion of
(iv).
(iv)b2 = i. Let K = gv{b); then HC^K={1) b~âb G and, since a is of order 4, we have b~âb G is abelian. Hence b'âb = a^, which leads to
and
G = HK.
a".
Now
(ii),
H<G
=
so that
a or
As
in
b~âb
a implies
ba
a^b
{5.10)
The elements b^ = 1 and a*
1, a, a^, a", b,
~1
Equation {5.10), ab, a^b, a^b are the distinct elements of G. enable us to construct the following multiplication table.
152
FINITE GROUPS
a a
c2
[CHAP.
a2
a3
ab
ab
a^b
(l26
aSft
a2
a3
a%
a^b
a36
a
a2
a3
ab
a3
a
a2
a%
flSfc
a^b
ab
a^b
a3
a
a26
ab
a2
a^b
a&
a26
aS
a
o2
ab
a26 a26
a36
a
a2
a3
ab
a^b
a36
a
a2
a3
a36
a^b
ab
a3
Table
5.2
As
in part (ii), any non-abelian group of order 8 having property (iv) is isomorphic to G. Such a group exists, for Table 5.2 also defines a group. This group is isomorphic to the dihedral group Di, the group of symmetries of a square (see Problem 5.38).
The two groups given in Tables 5.1 and 5.2 are not isomorphic because the group of Table 5.1 has exactly one element a^ of order 2, whereas the group of Table 5.2 has five elements of order 2: a^ b, ab, a^b and a?b. Thus we have shown that there are exactly two non-isomorphic non-abelian groups of order 8.
To summarize,
non-abelian.
there are five non-isomorphic groups of order
8,
three abelian and two
Since 9
3^,
we know by Problem
9,
5.36 that there are two

Ca.
and only two non-isomorphic
groups of order
namely Cg and Ca x
d.
Groups of small order: orders 12 and 15 To complete our list of all groups up to order 15, we must find all possible groups of order 12 and 15. Because 12 = 3 2^, we know that a group G of order 12 has at least one Sylow 2-subgroup of order 2^ and at least one Sylow 3-subgroup of order 3. The third Sylow theorem tells us that the number of Sylow 2-subgroups is congruent to one modulo 2 (i.e. S2 = 1 + 2A; for some integer A;) and Sz divides |G|. When A; = 0, S2 = 1; and when A; = 1, S2 = 3. If A; > 1 it is clear that 1 + 2A; does not divide 12. We therefore have two possibilities: G has exactly one Sylow 2-subgroup or G has exactly three Sylow 2-subgroups. A similar argument shows that G has exactly one Sylow 3-subgroup or G has exactly four Sylow 3-subgroups. Therefore we have four possibilities:
(i)
S2 S2
S2
=
= = =
1
1
and and
ss
Sa
=
= = =
(ii)
4
1
(iii)
3 and Ss 3 and Ss
(iv)
S2
Notice that because the Sylow 2-subgroup has order 4 it must be isomorphic to Ct or Ki, and the Sylow 3-subgroup must be isomorphic to Ca (Section 5.3b). treat each case separately.
We
Sec. 5.3]
153
(i)
be the Sylow 2-subgroup and T the Sylow 3-subgroup of G. Then F < G and since a Sylow ?5-subgroup is a normal subgroup if it is unique (Problem 5.7, page 133). Furthermore, Fr\T = (1} since any element in the intersection must have order dividing 3 and 4 and so must be the identity. Moreover, \F\ \T\ = 12. Hence by Theorem 5.16', or (&) F = Ki. We have two possibilities for F: (a) F = Thus G = CiXC3 = Ci2 (by Problem 5.34) or G = KtX Ca; both these groups are
Let
T<
G,
G^FxT.
abelian.
These are the only abelian groups of order 12, for if subgroups are equal. Hence by Theorem 5.3, page 131,
is
abelian any
Ss
two conjugate
S2
1.
In cases (ii) through (iv) we assume G is a non-abelian group. Furthermore, let F be any Sylow 2-subgroup of G and T any Sylow 3-subgroup of G. Then FnT = {1} and, by Proposition 5.18, page 145, |Fr| = |F| iTj/IFTiTI = |Fj |T| = |G|. Thus G = FT. If ft = tf for all f G F and t GT, then G is abelian since gi, gz G G implies gi = fiti and g2 = /2*2 for some /i, fiGF and ti, ti G T. Now as F and T are both abelian groups,
fi'iS'2
/iii/2^2
fitif-iU
g^gi
Hence we
that
it is
also
assume
if
F is
not the case that ft
any Sylow 2-subgroup of G and T any Sylow 3-subgroup of G, = tf for all f GF and t GT.
be the (unique) Sylow 2-subgroup and
(a)
(ii)
Si-l and
group.
separately.
Ss
4.
Let
T be
a Sylow 3-subtreat each case
There are two
possibilities:
F=
and
(5)
F = Ki. We
where
(a)
Let
1 imab 6a. But this implies contrary to our assumption. Therefore since a has order 4, b'âb - a^ and b-âb ^ 1 so that b'âb = a^ or ab = ba^. We show that under these assumptions gpiba) = G. (baf = b{ab)a = = b2 gQ (j,Qj^3 _ (6a)^6a = h^ba = a. Hence gp{ba) contains a and b and thus coincides with G. Then G is cyclic, contrary to assumption. There is therefore no non-abelian group of order 12 with S2 = 1, S3 = 4 and Sylow 2-subgroup isob^
1.
plies
F = {\,a,a^,a^] where a* = 1, and T = {l,h,h'^) F <1 G. Thus &-'a6 G F. If 6-ia6 = a, then every element of F commutes with every element of T,
Sa
bV
morphic
(6)
to Ci.
Let
F=
c^
{l,x,y,z}
be the four group as given in Section 5.3b, and

in part (a),
T=
{l,c,c^}
where
1.
As
F< G
so that
= f for at least one f GF. are similar). We may assume c-'^xc = y. (The other case, c-^xc = z, is argued similarly.) Let us, as in Section 5.3b, put x = a, y = b and z = ab. Then ac = cb, which implies a = cbc-K Now c'^bc y^ a, for c^bc = a implies c'^bc = cbc^ or b- c^bc-^. Then, as c^ = c- and c'^ = c, b - c-^bc. Hence a = &, a contradiction. Similarly c'^bc ^ b and c~%c = 1. Then c-''bc = ab and c-\ab)c = {c~âc)(c-^bc) = bab = 6% = a. Consequently the equations
assumption c-^fc
for all f GF. Now by Suppose c'^xc - x (the other cases

c~^fc
GF
ac
cb,
be
cab,
abc
ca,
a^
b^
1,
c^
=1, ab = ba
{5.11)
hold in G. of G are
Now
1, c, c^
determine distinct cosets of

c^,
in G.
Therefore the elements
1,
c,
a,
b,
ab, ca, cb,
cab, câ, c^b, câb
Using equations
in
(5.11),
we can
write
down
the multiplication table for G, as
shown
Table 5.3 below.
154
FINITE GROUPS
[CHAP.
1 1 1
c c
c2 c2
b
b
ab
ab
cab
câb
b
ca ca
câ
cb
cab cab
câb
câ
câ
c26 c26
6
câb
câb
a
ca
câ
1
cb
C2b
b
c
c
C2
c2
cb
c^b
a
ca
c^b
c2
ab
cab
c2
C2
a
cab
cb
c
ab
ca
c
cb câ
a
b
a
b
cb
câb
câ
c26
ab
1
cab
ca
c^b
ab
6 c
a
1
ca
câb
c2
c%
câ
1 b
ab
ca
cb
ab
ca
cb
a
cab
c
cab
c2
cb câ
C2
câb
b I
ab
cb
câb
c^b
c2
a ab
1
câb
câ
b
a
b
cab
cb
c2
ca
c
c2a
cab
câ
c^b
cab
câ
c^b
ca
eâb
c2
câb
1
c%
a
1
ab
cb
c
a
c
cab
c^b
ab
b
ca cab
c
ab
ca
cb
câb
c^b
câ
c2
a
ab
cb
câb
câb
câ
cab
ca
Table
5.3
By a similar argument to that used in the discussion of the non-abelian groups of order 8, any group of order 12 with S2 = 1 and S3 = 3 is isomorphic to G. Moreover, Table 5.3 defines a group: the table defines a binary operation, the identity is 1, and every element clearly has an inverse. The associativity of the operation must also be checked, an even more tedious task than in the case of a group of order 8. The alternating group Ai is a group of this type (Problem 5.38).
(iii)
S2
= 3 and S3 = 1. Sylow 3-subgroup.
Let F be a Sylow 2-subgroup and Again we have two possibilities:
T
(a)
{1, c, c2} (c^
1)
F=
{l,a,a^,a^}
be the Ct and
{b)F={l,x,y,z} sK4.
(a) T,
being a unique Sylow 3-subgroup, is normal in G and so a~^ca G T. We may # c, otherwise the group is abelian. Hence a~^ca = c^ and ca = ac^. Also, câ = cac^ = ac^ = ac. The equations which determine a multiplication table for this group are ca = ac'^, câ = ac, c^ = 1, a* = 1 {5.12)
assume a~^ca
The
distinct elements of
1,
G
a2,
are then
a^,
c,
a,
c^,
ac, a^c, a^c, ac^, a^c^, a^c^
and we obtain Table
5.4 below.
conclude, by an argument similar to that used in the discussion of the nonabelian groups of order 8, that any group of order 12 with S2 = 3, ss = 1 and in which the Sylow 2-subgroups are cyclic of order 4, is isomorphic to the group G defined in the table. Again it can be checked that the table defines a group, so that a group of this type exists. have as yet not encountered an example of this
We
We
type of group, but in Problem 5.41 which is isomorphic to G.
we show
that there
is
a group of 2 by 2 matrices
Sec. 5.3]

a^c a^c
155
a^c
,3-2 a^c
a?c^
aP-c^
rf.3/>2
a
2
a^c
aê^
ac^
a^c^
aflc
ac
a^c
:2/.2 a^c
n?i*%
%n% a^c
ac
a^c
a^c"^
a?
a^c
ac
a-^c
ac
a^c^
a2c2 a3c2
n/^rP
ac^
a^c^
U'^C^
a^c
a^c^
ac^
a^c
Table
{b)
5.4
F={l,x,y,z} and T={l,c,c^}.
Since T <\ G, we have f-'cfeT for all f e F. assumption, for at least one f G F, f-^cf ^ c. We may therefore assume, without loss of generality, that x-'cx = c\ Again, as in Section 5.3b, put x = a, y = b and z - ab. Then ca = ac\ Note that câ = c{ca) = ciac^) = (ca)c^ = ac^c^ = ac
By
I.e.
câ
'
ac.
We
claim that
S=
{l,c,c\a,ca,câ)
is
a subgroup.
theâsk of checking that the product of any two elements in S is again in S {ca - ac^ and d'a = ac make this task easy). The identity 1 is in S, and on inspection we find every element in S has an inverse in S: c-i = c^, (c^)-* = c, a- = a, {ca) 1 = ca, (c%)-i = c'â. Hence S is a subgroup of G. \S\ = 6 for &a = & {i = i or 2; and ; = 0, 1 or 2) implies aGT, a contradiction; c'a = a {i = 1 or 2) implies C - 1, a contradiction; and ca = câ implies c = a contradiction. S is 1, non-abehan {ac = câ is not equal to ca) and hence is isomorphic to Da since there is only one non-abelian group of order 6 (up to isomorphism). Now [G:5^-2 implies S O G (by Problem 4.69, page 116). Hence b-'cb G S. As b ^cb is an element of order 3, it is either c or c^ as all other elements of
are of order
If
2.
We
leave to the reader
We
=
c\
shall
let
now
h
h 'ch -c.
Recall^ that <^^'c^c.
ic6
c,
b.
choose an element hGF, h^S, If on the other hand b~^cb = c^
let
a~^ca
Then
{ab)-^c{ab)
such that h = ab
b-\a-'ca)b
in
Hence there
exists
an element
= 9'P(A)- Clearly SnH={l}, S and yonSf^"", |^||ii|-|G|, and so by Theorem 5.16. therefore conclude that any group G with S2 =
h^S
(i.e.
G^SxH
b'^cb b-'cb = b or ab) such that h-^ch = c commute elementwise, and

b-^c'^b
But S
^ D3 and
isomorphic to K^, is isomorphic to Ds x type (Problem 5.38).

(iv)
C2.
and Sylow 2-subgroup The dihedral group De is a group of this

3, S3
H ^ C2 We
S2 = 3, S3 = 4. Since distinct cyclic groups of order 3 intersect in the identity element, the four Sylow 3-subgroups have together 9 distinct elements. Sylow 2-subgroup IS of order 4. Since the intersection of a group of order 4 and a group of order 3 can only be the identity, it follows that the number of distinct elements in the 4 3-Sylow subgroups and a single 2-Sylow subgroup is 12.
IS
But \G\ = 12, so there cannot be another distinct Sylow 2-subgroup. no group of type (iv).
Thus there
156
FINITE GROUPS
[CHAP.
To summarize, we have shown that there are up to isomorphism exactly three non-abelian groups of order 12. They are the groups
(ii)
(6)
with S2 = 1, Sa 4 and the Sylow 2-subgroup is isomorphic to A4.)
K^, see Table 5.3.
(Such a group
(iii) (a)
with S2 = 3, S3 = 1 and the Sylow 2-subgroup = C4; see Table 5.4. (Such a group is isomorphic to a group of 2 by 2 matrices given in Problem 5.41.) with
to
S2
(b)
3, S3
1
is
Ds X C2 which
and the Sylow 2-subgroup ^^ K4. (Such a group isomorphic to De. See Problem 5.38.)
is
isomorphic
and
Clearly no two of these groups are isomorphic. The abelian groups of order 12 are Ki X Cs C12. Thus including the abelian groups, there are five non-isomorphic groups of order 12.
If
is
of order 15,
we have
seen that
G
6 2
is cyclic
(Section 5.1a).
The following
table
gives the
number
of non-isomorphic groups of order 1 through 15.

1
Order of group
No. of groups
2
1
3
1
4 2
5
1
7
1
8
5
10
2
11
1
12
5
13
1
14
2
15
1
Indeed there will agree that finding all groups of a given order is difficult. general method of determining how many non-isomorphic groups of a given is not even a order there can be.
The reader
Problems
5.38.
Show
(i)
that
is
The dihedral group D^ page 152.
a group of order 8 isomorphic to the group given in Table
5.2,
(ii)
The alternating group A4
is
isomorphic to the group given in Table

D^,
5.3.
(iii)
The dihedral group De
is
isomorphic to
C2-
Solution:
(i)
1)4 is
a non-abelian group of order 8 and as such is isomorphic to one of the groups given in check of Table 5.1 shows that there is only one element of Section 5.3c, Tables 5.1 and 5.2. order 2, namely a^. But D4 is the symmetry group of the square (Section 3.4f, page 75). reflection t is of order 2 and if <7 is a rotation of 90, ra is of order 2 as can be seen in the discussion of these groups in Chapter 3. Therefore D4 must be isomorphic to the group given
in
(ii)
Table
5.2.
is a non-abelian group of order 12. Hence it is either isomorphic to the group of Table 5.3 or 5.4 or to D^ X C^ (see page 155). As can be seen from the multiplication table for A4 given in Chapter 3, page 63, A4 has exactly three elements of order 2, namely a^, a^ and ag. Now the group of Table 5.4 has only one element a^ of order 2, so that it could not be isomorphic to A4. We have shown in Problem 5.1, page 131, that A4 has no subgroup of order 6 and jDg is iso-
A4
morphic
to a
Thus
(iii)
it
must be isomorphic
subgroup of D^ X C2 by Theorem 5.14. Hence A4 to the group given in Table 5.3.
is
not isomorphic to Dg X Cj.
Dg
is of order 12, and is not abelian. Dg is the symmetry group of the hexagon and therefore has a subgroup of order 6, namely the rotation of 60 about the center. So it is not the case that Dg ^ A4, and hence D^ cannot be isomorphic to the group of Table 5.3. Also, a reflection followed by a rotation is an element of order 2. Since there are six such elements in Dg, it cannot be isomorphic to the group of Table 5.4 as this group has only one element of order 2.
The only other
possibility is that
Dg X
a=
D3 X
C^-
5.39.
Find a
cyclic
subgroup of order 6
in I>3
2-
Solution:
Let a
((i2, 1),
e Dg
be of order 3 and
(1, b),
(#
1)
(a, b)3
(a, b)4
=
6.
(a, 1),
(a, b)5
G Cj. Consider the element = (aP; b), and {a, b)6 = (1, 1).
(a, h)
e Dg X
Cg.
(a, b)^
Hence
gp{{a, 6)) is a cyclic
subgroup of
I>3
X C2
of order
Sec. 5.3]

i\
j
157
5.40.
Consider the matrices
fQ
I
.
and
B =
/
I
IN
\
where
V""l
A.
and
have nonzero
determinants and thus are elements in the group of all 2X2 matrices with nonzero determinants. Show that gp(A,B) is a group of order 8 which is isomorphic to the quaternion group (Table 5.1,
page
151).
Solution:
By
direct calculation
we
find
Let gp(A,B).
CI -0
A3
/O
\-i
/-i
(^
-i
A*
=
J
/,
(identity matrix)
CI
-:)
BÂ
AB =
=
A^,
A^B =
i
/O -1\
(^
A3B = (^ _.
/i
BÂ
VB
and
A^B = BA
{/,A2,A3,B,AB,A2fi,A3B}. We claim that G = {I,A,AÂ^B,AB,A^B,A^B} = GQgp{A,B). To show G gp{A,B}, we need only show G is a group, as A,B & G. Note G is a subset of the group of 2 X 2 matrices with nonzero determinant. Hence the elements of G satisfy the associative law. By direct calculation we can show that G is closed under matrix multiplication; the equation BA = A^B simplifies the calculations, e.g.,
G =
Clearly,
(A2B)(A3B)
= A2(BA)A2B = AÂWA^B = AA^BAB = A^B^ = A G

has an inverse in G,
e.g.
Furthermore every element of
using
BÂ =
AB we
have
(A3B)-i
= B-iA-3 = BÂ = AB
all these details enables us to conclude that G is a group of order 8 and G = ffp{A,B). non-abelian, since BA, so G is either isomorphic to the group of Table 5.1 or Table 5.2. Because G has only one element of order 2, it cannot be isomorphic to the group of Table 5.2. Thus
Checking
is
AB #
is
isomorphic to the quaternion group of Table
5.1.
5.41.
Consider the matrices
A =
(^
and
B =
(^
^^j, where
= ^/^
and
e is
a nonreal com-
plex cube root of 3 (so, in particular, e^ - 1 and e ^ 1). A and B have nonzero determinants and thus are elements in the group of all 2X2 matrices with nonzero determinants (see Section 3.5b, page 81). Show that gp(A,B) is a group of order 12 which is isomorphic to the group given in
Table
5.4,
page 155.
Solution:
We
^' = ^' = A^ =
find
by direct calculation that
{~l J^)
(_i ~o)
[^
B^
-i\
^ i:
(:
1
e2
"^ "]
J
^\ :] ij
te2
A2B
A3B =
J)
A^B^
= =
c:
'
-!)
ie
B^
t)
ie\
ie2
A^B^
V-ie2
(J
AB =
ie
Z)
AB^ ^
Let H = {A,A^,A^,A*,B,Bi,AB,A^B,A^B,AB^,AiB^,A^B^. We claim H = gp{A,B). Clearly Hcgp{A,B). To prove H = gp{A,B) we need only show i? is a group, as A,B e H. Note that
is
a subset of the
2X2
matrices with nonzero determinant.
associative law.
To check that
Hence the elements of H satisfy the H is closed under matrix multiplication, first note that A-Â = BK
Then, for example,
(A2B)(A3B)
= A2A.(A-iBA)A2B = A^BÂ^B = AHA-iBA)(A-iBA)AB - B^BÂB = BAB = AA-^BAB = AB^B = A

is
Also the inverse
of,
for example, (A^B)
given by
(A3B)-i
= BÂ = AA-iBÂ = A(A-iBA)(A-iBA) = AB^B^ = AB &
158
FINITE GROUPS
Checking
[CHAP.
all
The mapping a:
obtains because
= gp(A,B). these details enables us to conclude that is a group of order 12 and and c -* B is an isomorphism of the group of Table 5.4 and H. This
aÂ,
satisfies the
equations
BA =
(I
AB^,
BÂ = AB,
B"
I,
A* =
the identity matrix). These are the exact counterparts of equations {5.1S), page 154. Consequently the multiplication table for is obtained from that for the group of Table 5.4 by renaming
5.42.
Show
that a group
of order 48 has a normal subgroup
{1} or G.
(Very
difficult.)
Solution:
the first Sylow theorem, G has a Sylow 2-subgroup of order 16. By the third Sylow theorem, for some integer k and Sg 48. The only odd divisor of 48 is 3, hence 82 = 1 or 3. If 1, then the Sylow 2-subgroup is unique and therefore normal (Problem 5.7, page 138). Suppose 82 is a proper subgroup of H, 3. Let and be two of the Sylow 2-subgroups. As 82 \Hr\K\ 16. Then \HnK\ = 8; for if \HnK\ ^ 4, then, by Proposition 5.18, page 145, \HK\ = \H\ \K\/\HnK\ ^ 16 X 16/4 = 64, which contradicts our assumption that \G\ = 48. Since both and (Problem 4.69, and are of order 16, HnK, as a subgroup of index 2, is normal in both = NoiHnK), we have hetting and page 116). Hence = G. Thus \N\ ^ \HK\ = \H\ \K\/\HnK\ = 32. As |iV| divides 48 and \N\ ^ 32, \N\ = 48 and so
By
1
Sg
= = =
+ fc2
HnK
N
HcNdHnK)
normal
KcNdHnK).
we have
HKcN.
Because a group
is
in its normalizer,
HnK <
G.
5.43.
Show
page
that
151.)
all
subgroups of the quaternion group
are normal subgroups.
{G
is
given in Table
5.1,
Solution:
It is sufficient to
GS, and
G,
check that the cyclic groups are normal in G. then x~^sx ^ ffp(s) implies x~'^sx G S.
For
if
is
any subgroup,
aj-'sa;
Using the multiplication table we can check that for x
a or 6 and any
S,
G grp(s).
a'
(We
i
leave this check to the reader.)
This is sufficient to prove the result, for every element of 0,1,2,3.
is
of the
form
a'b
or
for
5.4
a.
SOLVABLE GROUPS
Definition of solvable groups
To introduce our concepts we

where p
is
will
a prime, then
we
showed that
begin with an example. If P is a group of order p" P has a series of subgroups Pi with
{1}
Po
where each Pi
Let
<
Pi+i
and
[Pi+i: Pi]
C Pi C = p for
each
= P integer i =
Pr
(5.13)
0,
...,r-l
(see equation
(54), page 139).
be a group and suppose

{1}
it
has a series of subgroups
Go
Gi
C Gr
--
(S.U)
If each
Gi
<
Gi+i for
= l,...,r-l,
then {5.1i)
is called
a subnormal series of (for) G.
With
i
this definition, (5.13) is a
subnormal series of P.
a subnormal series for G and [Gi+i: Gi] is some prime (dependent on i), for ...,r-l, G is called a solvable group and (5.14) is called a solvable series for G. 0, Accordingly we conclude that P is solvable and that {5.13) is a solvable series for P. If {5.14.) is a subnormal series for G and Gi+i/Gi is simple, i.e. Gi+i/Gi has no normal sub.,r- 1, then (5.14) is said to be a groups other than Gi+i/Gt and the identity, for i = 0, composition series for G. To see that (5.13) is a composition series for P, note that Pi+i/Pi We call the factor groups Gi/Gi+i of the is a cyclic group of order p and hence is simple. subnormal series {S.li) the factors of {5.H).
If {5.U) is
Sec. 5.4]
SOLVABLE GROUPS
159
We shall discuss composition series in greater detail in Section 5.5. We remark that not all groups have a composition series but finite groups do (Section 5.5a). Our main concern in this section is the concept of solvable group.
Historically, solvable groups arose in the attempt to find a
formula for the roots of an
nth degree polynomial

f{x)
anX"
an-ix""'^
ttio;
tto
{5.15)
in
Oo,
terms of the
. .
.
coefficients
Oj.
The formula sought was one which involved the
coeflScients
a of the polynomial, integers,

division,
tion
and
and a
finite
is
and the operations addition, subtraction, multiplicanumber of extraction of roots. For example, if re = 2, then
(5.15).
cti Vî X =
"" 4a2Cio
a formula giving the roots of
If the roots of f{x)
can be
obtained by such a formula,
we
say f{x)
is
solvable by radicals.
From the Fundamental Theorem of Algebra we know that an th degree polynomial with complex coefficients has n complex roots. Let F be the "smallest" field (see Section 3.6b, page 86, for a definition of field) of complex numbers containing the coefficients Oj of f{x). By saying F is the smallest field we mean that if i? is a field containing the coefficients Oi, then FqH. Let E be the smallest field containing F and the roots of f{x). Now the set of automorphisms of E forms a group under the composition of mappings (see Theorem 3.15, page 87). The automorphisms of E which map every element f G F onto itself is a subgroup g of the group of all automorphisms of E. The group g is called the Galois group of the polynomial f{x). In the beginning of the 19th century, the French mathematician E. Galois proved (essentially) the following extraordinary theorem: An equation f(x) = is solvable by radicals if and only if the Galois group of f{x) is solvable. It turns out that not all equations of degree ^5 are solvable by radicals because the symmetric group S is not solvable for n^5. (For details see Birkhoflf and MacLane, A Survey of Modern Algebra, Macmillan, 1953.) In Section 5.5e we will prove that S is not solvable.
Problems
5.44.
Show
that the symmetric group S
is
solvable for
n=
1,2, 3.
Solution:
Si
{(};
then Si has the solvable series
{i}
cSi and
2\
/I
is
therefore solvable.
'1
2'
1
S9
then
\1
2/'V2
{1}
c S2
is
a solvable series for S2 and so S2

Ti,
1-2,
is solvable.
Let S3
{(, <xi, ff2.
T3}
where
3\
/I
"'
2
1
3\
/I
^'^
2
2 2
1
3
1
12
1
"'
3/
3\
"
^3
/I
2^
3\
~ -
V3
/I
2 3
2 3
3 3
~
is
\2
1/
"'
2;
:
"
V2
Now
116,
H H < S3.
that S4
{(, a-i, ffj}
Thus
{1}
a cyclic subgroup of S3. Also, [S3 H] = 2. Hence by Problem 4.69, page c C S3 is a solvable series for S3, since [H {t}] = 3 and [S3 fl] = 2,
and so S3
5.45.
is solvable.
Show
is solvable.
Solution:
The alternating group A4 is a subgroup of order 12 in S4. Then [S4 A4] = 2 and A4 <] S4 by Problem 4.69, page 116. We have seen in Problem 5.1, page 131, that A4 has no subgroup of order 6. Now A4 is a group with a unique Sylow 2-subgroup F of order 4 and F ^ K^, the four group (see Problem 5.38(ii), page 156, and Section 5.3d, page 153). Since F is a unique Sylow 2-subgroup, F <] A^ and [A4 F] = 3. F, being a four group, has a normal subgroup K of order 2. Accordingly,
: :
160
FINITE GROUPS
c
:
[CHAP.
{1}
is
F Q
[P
:
A^ Q Si
K]
=^ 2,
a subnormal series for S4. solvable group.
As [K
{1}]
2,
[A4
F]
and
[S4
A4]
2,
S4
is
b.
Properties
of,
and alternative
definition for, solvable

is
groups
An
important property of solvable groups

5.20:
given in
Theorem
Proof:
[Hi+ii Hi]
If G is a finite group and iV O groups, then G is also solvable.

{A^}
is
such that
and G/N are solvable
Let
qHiQHzQ
By
Pi, Pi
a prime.
be a subnormal series for G/N with the correspondence theorem (Theorem 4.19, page 120, and

QHk = G/N
Corollaries 4.20 and_4.21), there are subgroups Hi in G such that Hi [Hi+i Hi] =-- [Hi+i: Hi] = Pt (i = 0,1, .,k- 1). Therefore
: . .
<
Hi+i,
Hi/N
Hi and
N =
Ho C
<i
<Z
Hk = G
{5.16)
is a series of subgroups of G with Hi group. Hence has a series
Hi+i and [Hi+i: Hi]
Pi.
Now N
is
also a solvable
where
[Ki+i
Ki] is
= Ko C Ki C K2 C prime number (i = 1,2, .,1-1).

{1}
. .
Ki
N
{5.16)
{5.17)
Putting the series
and
{5.17)
together,
we
obtain
{1}
= KoQKiQ---
qKiQHiQH2Q-is
QH>c
= G
which
is
a solvable series for G.
The proof
complete.
Note. In contrast to Theorem 5.20, it is not always true that a group G has a property if both a normal subgroup and G/N have the property; for example, the four group Ki is cyclic, but Ki itself has a normal cyclic subgroup of order 2 which is cyclic and is not cyclic.
KJN
Corollary
5.21:
If
is
finite
abelian group,
is
solvable.
Proof: We use induction on the order n of G. If \G\ = 2, the result holds trivially. Assume that any abelian group of order less then n is solvable. Suppose p n for some prime p. Then by Proposition 5.9, page 137, G has an element of order p. Let a be such an
\
element.
If
p'n, then
\9P{a)\
<
\G\
so that gp{a)
is
is
solvable
by the induction assumption.
a normal subgroup of G and \G/gp{a)\ < \G\. Hence G/gp{a) is solvable by our induction assumption, and so, by Theorem 5.20, G is solvable. If p = n, then {1} is a solvable series and G is therefore solvable in this case too.
Furthermore, since
is
abelian, gp{a)
cG
leads to an alternative definition of solvability.

if
Theorem
5.22:
is
a solvable group
{1}
and only
if
is finite
and has a subnormal
series
{5.18)
where Ki+i/K
Proof:
is
= Ko Q Ki C Q Kn = G .,nl). abelian {i = 0,1,

. .
prime order and hence
be a solvable group. Then G has a subnormal series with factors of cyclic. Since a cyclic group is abelian, the solvable series of G is a series of type {5.18). Conversely, assume G has a subnormal series {5.18) with Ki+JKi abelian. We prove that G is solvable by induction on n, the length of the subnormal series If w = 1, then G is abelian since G = KJKo which is abelian by assumption. {5.18). Hence by Corollary 5.21, G is solvable. Assume that any finite group which has a subnormal series of length less than n in which the factor groups of consecutive terms of the series are Then Kn-i has a abelian, is solvable. Let G have subnormal series {5.18) of length n.
Let
subnormal series of length n~l, namely
Sec. 5.4]
SOLVABLE GROUPS
Ko C Kl C
. .
161
C Kn-2 C Kn-1
is
with Ki+i/Ki abelian for i = 0,1, .,n 2. Hence by the induction assumption, Kn-i solvable. But G/Kn-t is abelian and hence solvable. Using Theorem 5.20, we conclude
is
solvable.
In the theory of infinite groups one usually defines a group G to be solvable if it has a By Theorem 5.22 tc 1). (5.18) with Ki+ JKi an abelian group (i = 0, 1, this is equivalent to our original definition for finite groups. Since this formulation of solvability is more general, we shall henceforth use it as our definition of solvability. Note that the infinite cyclic group is an example of a group that does not fit the old definition but does fit the new.
subnormal series
Using
this
5.23:
new
definition
we prove
(i)
Theorem
Proof:
(i)
Let
(ii)
G
if
N <] G
be a solvable group. Then then GIN is solvable.
any subgroup of
is
solvable
and
Let
for
i
{1}
= HoQHxQ QHn = G be a subnormal series l. We show that if K is a subgroup of .,n 0,1,

. .
of
with Hi+i/Hi abelian
G,
{1}
is
= (KnHo) c (KnHi) c
c (KnHn) =
{s.w)
a subnormal series with KnHi+JKnHi abelian. First we notice that KdHi (Kr\Hi+i)nHi {i ^ 0,1, ...,n-l) and that KnH = K. Now Hi<Hi+i, and KnHi+i is a subgroup of Hi+i. Applying the subgroup isomorphism theorem (Theorem 4.23, page 125) inside the group Hi+i with subgroup KnHi+i and normal subgroup
Hi,
we
obtain
(Kni?i+i)nHi < KnHi+i

(KnHi+,)/{{KnHi+i)nHi) ^
{{K nHi+{)Hi)/Hi
and
Since
{KnHi+i)nHi^ KnHi,
it
follows that
KnHi<] KnHi+i and
(KnHi+,)/{KnHi) ^ {{K n Hi+i)Hi)/Hi

But {KnHi+i)HiQHi+i, so that we have
Hi
Hi
so is {KnHi+i)Hi/Hi.
Now Hi+i/Hi is abelian by assumption

is
and hence
Therefore (5.19)
a subnormal series for
with abelian factors
solvable.
(ii)
HJ.^^ ^'''"*
C\
and consequently
is
Let
have subnormal series

{1}
= Ho C
ifl
C ^2 c
Q Hn ^ G
where
flî+i/Hi is abelian.
Now N <
v
G.
V
:
Consider the natural homomorphism
G ^ G/N
from Problem
In particular let H. = H.v. 4.82, page 122 (with 9 = 'ihj+i
Any subgroup
of
<S
is
mapped by
But
onto a subgroup of G/N.
We
assert Hi
Hi+i.
this follows
and Hi+i
G).
Next we_assert that Hi+i/Hi is abelian. Let x = xv, y = yv {x,y G Hi+i) be two elements of Hi+u Then since Hi+JHt is abelian, xy = yxd where d G Hi. Thus
{xy)v
{xv){yv)
(yv){xv){dv)
162
FINITE GROUPS
[CHAP.
But dv G
Hi.
Consequently
xHiyiii
= xyHi =
is
{xv)vHi
{yv){xv)dvHi
{yv){xv)Hi
= yHixHi
Therefore Hi+i/Hi
abelian.
Thus
{N}
is
a subnormal series of
= Ho < Hi < <\ Hn = GIN G/N with abelian factors, and so G/N is
solvable.
Now that we have a definition of solvable group that applies to extend Theorem 5.20 to infinite groups.
Theorem
Proof:
5.24:
infinite groups,
we
will
If A^ <1
G
i
and
G/N and
. .
A^ are solvable groups, then so is G.
Let
is
{N}
=^
H0CH1QH2Q
=
0,
cHk^ G/N
be a subnormal series for
G/N
in
which Hi+i/Hi
abelian,
.,k-l.
and
Corollary 4.21, page 121, there are subgroups H, in G such that H, < Hi+i {i = 0,...,k). By the factor of a factor theorem (Theorem 4.22, page 121), Hi+i/Hi = iHi+i/N)/{Hi/N) ^ Hi+i/Hi. Hence the factors Hi+JHt are abelian. Also, A^ has a series
By
Hi/N =_H.
{1}
= Ko C Ki Q K2 C
Ki
N
Q Hk = G
with Ki+i/Ki abelian for

{1}
is
0, 1,

.,1-1.
Therefore
= Zo C Zi C
Ki
N =
Ho Q Hi Q
Thus G
is solvable.
a subnormal series whose factors are abelian.
Problems
5.46.
Show that (Hard.)

Solution:
If If
\G\
all
groups
of order p^, pq or p^q, where
p and q are
distinct primes, are solvable.
= p^, \G\ = pq
q.
it is
then
is
abelian (Problem 5.19, page 140) and

5.8,
is
solvable.
then from Problem

5.7,
page 133,
if
of order order q,
By Problem
page 133, < G. Now \G/H\ = abelian. Therefore we have the subnormal series
{1}
p <
q,
G has
p,
one and only one subgroup H hence G/H is abelian. As is of
Q G
with abelian factors and so

If
\G\
is
solvable.
p2g then Sp = 1 + fcp divides p^q, and so the prime factors of 1 + kp must be p or q. Clearly p does not divide 1 + kp. Therefore 1 + A;p = 1 or g. If 1 + kp = 1, then the Sylow p-subis of order p^, is abelian (Problem 5.19). Thus we group is normal in G (Problem 5.7). As have a subnormal series _ ,^, ^,
{1}
Q G
with
G/H
abelian (\G/H\
q)
and H/{1} abelian.
Hence
is solvable.
Suppose, however, that 1 + kp = q; then q > p. Let X^ be a Sylow q-subgroup of G. The number of such Sylow g-subgroups is 1 + Iq. Again 1 + ig is not divisible by g, and so 1 + ig = 1, p or p2. But as q > p, the only possibilities are 1 + ig = 1 or pK
Case
(i)
+ lq =
and
1.
K<
G. \K\ follows that
= G
:
In this case there \G/K\ = p2. Hence
is is
and (by Problem 5.7) only one Sylow g-subgroup abelian and G/K is abelian (Problem 5.19), and it
is solvable.
1 + Iq = p^. We will show that this case does not arise by showing that G would contain too many elements. We have assumed that G has g Sylow p-subgroups (of order p2) and p2 Sylow q-subgroups (of order q). Any two distinct subgroups of order q intersect in the identity, so there are pHq - 1) = P^g - p^ distinct elements of order g in G. Also, G has at least 2 Sylow p-subgroups and hence there are at least p2 distinct elements in G of order p or p2. In the above calculations we have not counted the identity, so in all G has at least p^g p2 + p2 + 1 = p2g + 1 elements, which is absurd, and we conclude that case (ii) does not arise.
Case
(ii)
Sec. 5.5]

that every nilpotent group
solvable.
163
5.47.
Show
is
Solution:
Consider the upper central series

{1}
Zo c 2i C ^2 C
C Z = G
is
of G, a given nilpotent group. of course this implies Zj + i/Zi
Zj + j is defined
is
abelian.
by the fact that Zi+JZi Hence G is solvable.
the center of G/Zi and
5.48.
Prove that the converse of Problem 5.47

Solution:
is false,
i.e.
not
all
solvable groups are nilpotent.
The symmetric group Sg is solvable (Problem 5.44). A check of the multiplication table for S3 on page 57 shows that the center of S3 is just the identity {i}. But this implies that the upper central series for G never ascends to G. Thus S3 is not nilpotent.
Let G be any group and let G"' be defined for all positive integers i by Gî^ = group of G, and G"+i) = (G"')'. Prove that G is solvable if and only if G<n>
integer n.
Solution:
5.49.
G',
the derived for some
{1}
Let G<>
{!}.
Then
{1}
G<>
is
G(i>
C G
is
a subnormal series for
and G^VG^'+i'
abelian.
Hence
is solvable.
Now
let
be solvable.
Then there
{!)
exists a
subnormal series
= H, Q
Q Ho = G
with HJHiî abelian.
By Problem
4.68,
that HiDG(^\ i = 1,2, assertion is true for i

G^r)
page 116, Hj/Hf+i abelian implies H^ + l^H'^. We prove, by induction on i, .,r. For i=l this is true since H^dHo G' = G'^^^. Suppose our = n, i.e. HdG(">. Then H^D (G(n>)' = Gf"+i'. But H^+^dH^, so
. .
H+i2G("+i'. Therefore
H^-DG'î^
for all i
In particular then,
{1}
= H^ 3
Gf^*"'-
Accordingly,
{1}
and the result follows.
5.5
a.

The Jordan-Holder Theorem
In Section 5.4a we introduced the idea of a composition series for a finite group G. that a series of subgroups of G
(1)
We recall
Go
C Gi C
Q Gk = G
. .
{5.20)
is a composition series for G if Gi <] Gi+i for i = 0, 1, and Gi+i/Gi is simple, i.e. ., fc 1 has precisely two different normal subgroups. This latter statement carries with it the implication that Gi^^Gi+i for i 0,...,k l.
We
observe
see this is
The easiest way to first that every finite group G has a composition series. by induction on the order, |G|, of G. If \G\ = 1, then G has precisely one
{1}
composition series:
Go
= G
less
Suppose then that

series.
|G| 7^ 1
and that every group of order

{1}
than
\G\
has a composition
Now
if
is
simple, then
Go
Gi
= G
is
the only composition series for G. If - {1}, j^G. We may suppose that
M<
G and
y^
G, then
|M|
^
=^
\N\.
By
not simple, let 2\/^ be a normal subgroup of G, the largest normal subgroup of G, that is, if induction, has a composition series
is
is
{1}
No Q Ni C
Ni
164
FINITE GROUPS
claim that
{1}
[CHAP.
We
is
Aô
iVi
Ni
Q G
a composition series for G and note that to prove this we need only show that G/Ni is But if G/Ni is not simple, it has a non-trivial normal subgroup. By the corollary to the correspondence theorem this subgroup is of the form K/Ni where Z is a normal subgroup of G. But as if D Ni, this means that \K\ > \Ni\ which contradicts the choice of A^j. Therefore every finite group has a composition series.
simple.
This proof does not suggest that if a group has two composition series then they are Surprisingly they are. In order to explain this relationship we associate with the composition series (5.20) two notions. First we term k the length of the series. Second we call the factor groups Gi+i/G, the composition factors of the series (5.20). The relationship between composition series is given by
related.
Theorem
5.25
(Jordan-Holder)
Every finite group G has at least one composition The lengths of all composition series for G are
Finally
if
series.
equal.
{1}
Go
C Gu = G C H = G
and
{1}
= Ho C
are a pair of composition series for G, then their respective composition factors can be paired off in such a way that paired factors are isomorphic.
We
have already proved the

5.25
Theorem
such that
for
i
we
first statement of Theorem 5.25. Before illustrating restate its last assertion as follows: There is a permutation tt of {1, &}
. . .
0,
.,k
l.
Gi+l/Gj
H(i+l)T!-/H(i+l-jTT-l
Example
1:
Suppose that S3
^^^
is
the symmetric group on {1,2,3}.

'1
Then the
2
1
S'^
series
2 2
3\
/I
2 3
'\1
3/' V2
3\ /l ly" V3
is
and
a composition series for S^. Notice that the composition factors are of orders This is actually the only composition series for S3, since 2.
1
1
2 2
3\
/I
'
2
3
is
3\
/I
2
1
3/
V2
1/' VS
is
the only normal subgroup of S3 which
neither S3 nor the identity subgroup.
Problem
5.50.
tion of
(n
Let w be a positive integer. What relevance does the Jordan-Holder theorem have to the f actorizan into a product of primes? (Hint: Let G be the additive group of integers modulo n
>
1).)
Solution:
Let
{1}
Go c
<?i
c G, = G
.
be a composition series for G. Each composition factor Giî/Gi (i ~ 0, is simple. As .,1 1) Gj is abelian, each factor Gi+^/Gi is abelian. Hence if Gi+JGi has any proper subgroup, it would not be simple. So Gi + i/Gi has no proper subgroups. In particular it has no cyclic proper subgroups, so it must be cyclic of order a prime. The number I is therefore the total number of primes (allowing for repetitions) dividing n. By Theorem 5.25, I is uniquely determined. So, as we well know, the total number of prime divisors of n is a constant. Moreover, the uniqueness of the composition factors (asserted in Theorem 5.25) simply means that these prime divisors themselves are unique. Putting these two facts together gives the well-known fact that every integer n > 1 can be written uniquely as a product of primes, if the order in which it is written is disregarded.
.
Sec. 5.5]

2:
165
Example
Let Sn be the symmetric group on n

{1}
is
lettei-s,
5.
Then
C A C S
a composition series for S.

5.34).
Theorem
is
But A
<
For we shall show that A is simple (in Section 5.5e, S and SJA^ is cyclic of order two. Hence this series
indeed a composition series.
b.
Proof of Jordan-Holder theorem
Suppose
is
finite
group and suppose

1
= =
Go
i?o
Q C
Gi
i?i
Q C
and
Q Gk = G C //i = G
{5.21)
{5.22)
are two composition series for G. We have to prove that k = l and that the composition factors Gi+i/Gi of {5.21) can be paired off wîth the composition factors Hi+JHi ai {5.22) so that paired composition factors are isomorphic.
clear.
The proof is by induction on the order \G\ of G. If \G\ - 1, then both assertions are Thus vsre assume that \G\ > 1 and that the theorem holds for all groups of order less than |G|. It is useful to observe that if fc = 1, then G is simple. Hence 1 = 1 also, and
again the desired conclusion holds. hence l> 1).
So we assume, in addition to
\G\
>
1,
that
A;
>
(and
There are two cases to consider: Gk-i Case

1:
Hi-i and
Gk-i^ Hi-u
Gk-i
= Hi-u
1
It is
then clear that

1
= Go C
Q
Hi-i
Gk-i
{5.23)
and
= Ho C
Gk-i
{5.2A)
are composition series for Gk~i. But |Gfc-i| < |G|. Hence by our induction assumption, the composition series {5.23) and {5.24) have the same length, i.e. kl = ll, and so k = l. Furthermore the composition factors of {5.23) can be paired with the composition factors of {5.24) so that paired factors are isomorphic. But the composition factors of Similarly the composition factors of {5.22) {5.21) are those of {5.23) together with G/Gk-u are those of {5.24) together with G/Gk-u Thus it is clear then that the composition factors of {5.21) can be paired off with the composition factors of {5 2) so that paired factors are isomorphic. This concludes the proof of Case 1.
Case
2:
Gk-i
^ Hi-u
Our method of proof is to produce a composition series, {5.26), which has isomorphic composition factors to those of {5.21) (by Case 1) and a composition series, {5.27), which has isomorphic composition factors to those of {5.22) (by Case 1). We will then show that {5.26) and {5.27) have isomorphic factors and this will be sufficient to prove the result.
First we will show that Gk-iHi-i = G. Observe that both Gk-i and Hi-i are normal subgroups of G, and so Gk-iHi-i is also a normal subgroup of G. Obviously Gk-iHi-i contains Gfc-i properly, so Gk-iHi-i/Gk-i is a non-trivial normal subgroup of G/Gk-i by the correspondence theorem. But G/Gk-i is simple, so Gk-iHi-i/Gk-i = G/Gk-u This means that
Gk-iHi-,
= G
G.
{5.25)
We now
put
F=
Gk-inHi-i and note that

{1}
<\
Let
Fo C
Q Fm = F
be a composition series for F.
Then we claim that
166
FINITE GROUPS
[CHAP.
{1}
= Fo C =
C Fm Q

Gfc-i
C G c G
(5.26) (5,27)
^^^
{1}
C Fm C
Hi-i
are both composition series for G. The only facts that need be verified are that G^-i/F. and î-i/n. are simple. Now Fr.^ G^-,C^Hi-^ and by G = G,-^Ht-i. Therefore by (5.;25), the subgroup isomorphism theorem (Theorem
4.23,
page
125),
G/Gk-i
Gk-iHi-i/Gk-i
Gk-iHi-i/Hi-i
and similarly
G/Hi-i
=
it
H,-J(Hi-inGk-i)
Gk-i/{Hi-ir\Gk-i)
= Hi-JFm = Gk-i/Fm
(5 28) '
^
'
(5.29)
Since both G/G^-^ and G/^,-i are simple,

simple.
follows that Hi-JF,n and G,-,/F are also both
Let us compare the composition series (5.21) and (5.26). By Case 1 it follows that they have the same length and their composition factors can be paired off so that paired
are isomorphic. Similarly for the composition series (5.22) and (5.27). Let us now compare the composition series (5.26) and (5.27). They obviously have the same length, + 2. Thus the series (5.21) and (5.22) have the same length. What are the composition factors of the series (5.26) and (5.27)1 The composition factors of (5.26) are
factors
FJF,,
while those of (5.27) are
...,
F/Fm-i, Gk-ilFm, G/Gk-i
FJFo,
...,
Fm/Fm~l, Hi-i/F,n, G/Hi-i
(5.28), Hi-JF,. ^ G/G^-f, and by (5.29), G^-JFm ^ G/Hi-^. Thus the composition factors of (5.26) and (5.27) can be paired off so that paired factors are isomorphic. It follows that the composition factors of (5.21) and (5.22) can be paired off so that paired factors are isomorphic, since the factors of (5.21) can be so paired off with those of (5.26) and the factors of (5.22) can similarly be paired off with those of This completes the proof (5.27). of the Jordan-Holder theorem.
Now by
From the Jordan-Holder theorem we know that the length of a composition series and factors are uniquely determined for the group. This suggests the following scheme for studying groups which have a composition series. First, find the structure of all groups with composition series of length 1. These are all the simple groups. Assuming now that
its
we know
all
position series of length
about groups with a composition series of length n + 1. Let

{1}
n, let
be a group with com-
Go
Gi
c G c
G+i
= G
be a composition series for G. If F is a group with a normal subgroup A^ and F/N = H, we say that F is an extension of by H. In this language then, G is an extension of a group with a composition series of length n, viz. G, by a simple group. Hence what we must know is how the structure of a group which is an extension of one group by another is determined.
In the next few sections we shall prove that if 5, A is simple. The groups A are not all the simple groups and indeed there is no classification of simple groups as yet. This is one of the basic problems of finite group theory.
n>
The question of how a group G is built from and if G is an extension of considered in Chapter 7. Here too our knowledge is far from complete.
H by K,
is
Problem
5.51.
If
two groups have the same composition factors, are they isomorphic?
No.
Solution:
There
is
same composition
a cyclic group G of order 6 and a non-abelian factors, but they are not isomorphic.
group
H of
order
6.
They have the
Sec. 5.5]
167
c.
Cycles and products of cycles
We
begin with an example of a
new
notation.
Let us consider
6' 3^
Se.
Let
G Sb
be
defined by
12 12
effect of 9 is to take
6 is called
3 5
4 6
Note that the

1,
2 unaltered.
a cycle.
3-5, 5->4, 4-6 and 6^3, and to leave the elements We denote it by (3, 5, 4, 6).
More
{ai, a2,
. . .
generally,
,
dm)
consider S. If ai,...,am are distinct integers in (1,2, ...,n}, stands for the permutation that maps each integer in {1,2,...,%}
itself,
(ai,
dm] to
and maps
ai -* a2, a2-â3,
...,
am-iâm, am-*ai
We call
such a permutation a cycle of length m. denote the identity element. A cycle of length 2
As a convention take a
is
cycle of length 1 to
called a transposition.
The inverse of the
cycle a
(ai,
Om)
is
the cycle p
(a, Om-i,
. ,
02, ai),
since
a.{ap)
{axx)p
ttj^jjS
=
a^p
a.
it
i^-m
while
If
y
aâp
{1, 2,
{aâ)p
...,}
(tti,
ttm},
j{aP)
{ja)P
jp
Hence
aj8
i.
Similarly pa
t.
It is clear that not every
permutation
is
a cycle.
Nor
is
the product of two cycles

j
necessarily a cycle; for example, in S4,
(1,2)(3,4)
which
is
not a cycle.
An
obvious question is: can we express every element of Sn as a product of cycles? following theorem answers this question.
The
Theorem
5.26:
(Cycles
i.e. if
Every element of S can be written as the product of disjoint cycles. (ai, .,am) and (&i, ftk) are disjoint if the Oi and 6j are distinct,
. . .
(ai,
Om}
(bi,
. ,
bk}
=
n}
0.)
is fixed
Proof:
Let
Itt
-k
S.
We say that
{1,
by x
if
iV
= i, and we
say
it is
moved
if
i.
We
If
IT
shall argue by induction on the number of integers moved by the permutation -k. moves none of the integers {1,2, .,}, then w is the identity permutation. But then
. .
(1),
as the 1-cycles are
all
the identity permutation.
Hence we have a basis for induction. So let tt = i and suppose that every element of Sn which moves fewer integers than tt, can be written as the product of disjoint cycles. Now suppose a^G {1, .,n} and that a^wjâ^. Let us define a^,ag, by a^â^^r,
.
.
a^ for some (As the images of w belong to (1,2, ...,n), the terms of the sequence a^, a^, cannot all be different.) We shall prove, using the minimality of m, that i=l. Suppose to the contrary that i=l. Then a^ = a^_^l^ and a^^r = a^ = a._j7r. As TT is a one-to-one mapping, o,^_^ = a^. But this contradicts the choice of the integer m. Hence i 1 and a^-n- = a^. We consider now the cycle {a^, .,a) and note that > 1. Let
'â
%'^>
>
<3Ss
=
.
s-i'^'
^t ^
156
the
first
integer such that
a^-rr
integer
with
lî<m.
.
(ai,
.,am)
^TT
168
FINITE GROUPS
a permutation that leaves a^,...,a^
[CHAP.
is
fixed,
since
a.{a^,
.,aj-'n
a^_^7^
a.
for
while a^{a^, ..., a^)= a,r = a^. Also if jG{l,2,--., n\7 then ^ j implies jTT ^ n for if j G {a^, ..., a J, jr = j. Hence j g {a^, ...,aj and so jr = j{a^, ...,aj-^n = jir^j. It follows that T moves fewer integers than Therefore inductively r can be written as the product of disjoint cycles, say
i - 1,
-n-.
Now
be those
if T.
involves an
a.,
a.
= a.r =
Tjj, Ti^,
a;(TjT2
' '
'i)
a*'-;-
Hence
r.
(aO
i.
Let
, Ti^ (ii<i2<
a^,
.
< ik)
.
t.
which do not involve any of the integers

^
.,
a^.
i
Clearly
î,\
n^
or
If
TT
=
.
I,
TT
(ttj,
a^)
and we have proved that n

tt
(a,,
a)rij
Tj^
and
is actually a cycle. Otherwise has been expressed as the product of disjoint cycles. Hence
the result.
Corollary
5.27:
If
tt
G
a2
S^ and
a^.a^,
. .
(i.e.
= aj7r, =
.,
and
...,a are chosen as in the proof of the theorem a^Tv = a^ and a^, ...,, distinct), then
77
=
,
(ttj,
...,ajr
a^.r
where ar
Proo/:
an
if
{a^,
a^}, while
a.
for
!,..., m.
This
is
precisely
what we showed
in the proof of
Theorem
5.26.
TT
This corollary provides us with a method of computing the decomposition of an element Sn into the product of disjoint cycles. For example, let us write
342 18 79
as the product of disjoint cycles.
as
123456789
Since
Itt - 1,
10 11 12
5 6
ai.
11 12 10
we may
5
6
take 1 for
Then
ai
1,
as
3,
2, a4
4, as
1.
So
,-
m = 4.

..
Hence by the
2 3 3
corollary,
(i,d,A4)^j
/l
4 4
7 9
10 11 12
5
11 12 10
Here the second factor t on the right is obtained from w by letting it leave 1, 2, 3, 4 unchanged, and letting it act as tt does on the remaining integers. Applying the same technique to t, we find r = (5, 8, 11)(6, 7, 9, 12). Hence
TT
(1,3,2,4)(5,8,11)(6,7,9,12)
cycles
a.^j
{aj,
bj,
.
In practice it is not necessary to use the corollary rigorously. We need only find the disjoint by choosing a^ such that aîr = a^, and then taking the cycle (a^, a^) where
. . .
=
. . . .
a.TT,
.
a^TT
a^,
and the
aj,
a^}
&^ distinct,
where h^ -- b^, and so on.
.,a^ are distinct. and find the cycle (6^,

. .
Then we choose
.
.
h^
{1, 2,
hîr
.,n}
b^
&J with
b^_^_^
b^7^,
and
Not only is every element of S a product of cycles, but each element can also be expressed as the product of transpositions, as can be seen from the following proposition.
Proposition 5.28:
Proof:
Every
cycle is a product of transpositions.
We
assert
(tti,
a2,
ttk)
{at, a2){ai, as)
(ai, au)
For
it
if a is an integer not in unchanged.
(ai,
an},
both the left-hand side and the right-hand side leave
Sec. 5.5]

then
169
If
i - k,
ai{ai,
.,ak)
at+i
while
ai{ai, a2){ai, as)
{ai, ak)
= =
ai(ai,ai)(ai, Oi+i)
(ai, aic)
ai+i(ai, ai+2)
(i,aic)
ai+i
If
k,
then
afc(ai,
ttk)
ai
while

afc(ai, 02)
(tti, ttfc)
aic(ai, a^)
=
is
ai
Thus the
effect of the left-hand side
and the right-hand side
the same, and so the permuta-
tions are equal.
Problems
5.52.
In Se compute Solution
(1, 2,
3,4)(2, 5)
and
/?
(1,2, 3,4)-i(2, 5).
la
to
=
=
I
(1(1,2,3,4))(2,5)
=
4a
2(2,5)
=
5a
/3
5.
if i
2,3;
1;
= =
6/3
2;
6a
6.
Hence a
= l^
\5
/I
2 3
4
1
,
(l,2,3,4)-i
1/3
= =
(4,3,2,1).
1,
3j8
Then
4/3
(4, 3, 2, 1)(2, 5).
4,
2;8
5,
3,
2,
6/3
6.
Hence
/?
= ,,, \4 1
'12
4 d
o i
6 ^ 6
5.53.
Express a and ^ of Problem 5.52 as products of disjoint

Solution:
cycles.
We
note that la
1/3
5,
5a
2,
2a
5
3,
3a
5j8
= =
4, 2,
4a
1.
Hence a
/3
(1, 5, 2, 3, 4).
Since
4,
4;8
3,
3/3
and
2/3
1,
(1,4,3, 5,2).
/I
5.54.
2
4
3
6
4 8
8
1
9
3
10 11
5 7
12 13 14\
9
Write
(^3
10 12 14
11
13 j
^'
^"^^ P''"''^"*
^ ''^"j"'"*
*=y'^^'-
Solution:
la
2,
14a
13,
2a = 4, 4a = 8, 8a = 1. 3a 13a = 11, 11a = 7. Hence

a
6,
6a
12,
12a
9,
9a
3.
5a
10,
10a
5.
7a
14,
(1,2,4,8)(3,6,12,9)(5,10)(7,14,13,11)
5.55.
If
a,
^ e S
i/3,
(i|8)a
are such that ia ' i implies both (ia)^ = ia and t/3 = i, and then a and /3 commute. Hence prove that disjoint cycles commute.
ip =
implies
Solution:
Let then
ia/S
e {1, 2, = i/3 while

. .
n}.
If
ia
i,
then
(i/3)a
i'(a/3)
(ia)y3
ia
while
i(;8a)
(i/3)a
ia.
If
ia
i,
= = =
i/3
if i/3 if iyS
ia
2
#i =i
ia = i
Hence a^ = â. Now if a and /3 are disjoint cycles, then P does not move either i or ia, and so if3 = i and (ia)^ ip, and so (i;8)a = ip. Thus a and ^ commute.
ia.
implies o moves i and ia and hence Similarly if ip = i, p moves i and
5.56.
Express a of Problem 5.54 as the product of transpositions. as a product of transpositions unique?

Solution:
Is the expression of
an element of S
= =
if
(1,2,4, 8)(3, 6, 12,9)(5, 10)(7, 14, 13, 11)

(1, 2)(1, 4)(1, 8)(3, 6)(3, 12)(3, 9)(5, 10)(7,
14)(7, 13)(7, 11)
expression of a as a product of transpositions, then a = (aia2 a)(l, is another expression of a as a product of transpositions. 2)(1, 2) Thus the expression as a product of transpositions is not unique.
A-s
(1,2)(1,2)

1,
aia2
a^
is the
170
FINITE GROUPS
Find the order of the cycle (a^
Solution:
If we write Oj, a circle, as shown on the right, it is clear that , a in a = (!, moves each Oj one place clockwise, a^ moves each aj two places , o) further, and in general a^ moves each Oj /-places further. Hence a'" moves each Of w-places further, i.e. " moves each Oj back to itself. To put this more formally, define a+i = a,, a^ + g = ag a^+ = a. We will prove by induction on j that for ^ j ^ m, a' maps Cj to aj + i = l, 2, ...,to. For
. . .
.
[CHAP.
5.57.
..., a).
(Hard.)
,-,
j = r, consider j = r+1 ~ m. Then = aj + r+i by the action of a. If i + r = m, aâ = i + r>m, then i = m+i = + r = aj + ^-m and i + r~m< m (as 1 i m and r<m). Hence 0; + ,^ = Oj + r-m" = a. + r-m + = "i + r+iIn particular, a = On the other hand, aja* = aj + j # aj for 1 s < m. Therefore rn is
/
0,
a^
and the result
is true.
i
If the result is true for

aj + â
ttj i
ttja'^+i
(ajaOo:
+ rIf i + r+ia.i
+ r < m,
c.
the smallest power of a that yields the identity. d.
Accordingly
is
the order of
a.
Transpositions, and even and odd permutations
In this section we will produce another way of deciding whether a permutation is even or odd. (See the definition in Section 3.3b, page 60.) By Proposition 5.28 and Theorem 5.26 every permutation is the product of transpositions. However, this decomposition is not unique (Problem 5.56). What we shall show is that a permutation tt is a product of an even number of transpositions if ^ is even, and the product of an odd number of transpositions if TT is odd.
Our
(1, 2) is
first
task
is
to
show that any transposition

64).
is
odd.
We
odd (Section 3.3d, page

5.29:
We
.
will use the following

.
have already noted that lemma.
Lemma
Let
Sn and
let (ai,
.,am) be a cycle.
..
Then
. .
e~^ {ai,
.,am)9
{aiO,
Oi,
.,am9)
Proof:
Let
a;
(1,2,
.,n}.
.
If
.
(uO
for
some
xe'^{au
-,0^)9
= =
ai99~^{ai,
ai(ai,
. .
.,am)9
.,am)9
if
1 7^ i
ai+i9
ai9
TO
=
If
a;
ii
m
X
9^ ttiO
for some
ai,
x0~^
(ai,
.,am}.
. .
Then
xe-^{ai,
.,am)9
{xe-^)e
Thus 9-\ai,
.,0^)9
{ai9,
.,am9).
Theorem
5.30:
All transpositions are odd. If 6 is an even permutation and is written as the product of transpositions, the number of transpositions is even. If 9 is odd and is written as the product of transpositions, the number of transpositions is odd.
(1, 2) is
Proof: Let (a, h) be any transposition. We know that mutation such that Id = a, 26 = b. Then, by Lemma 5.29,
(a, 6)
odd.
Let
be any per-
9-^(1,2)9
If 9 is even, 9'^^ is even, 9-^(1,2) is odd, e is odd, 9-^ is odd, 9~^{1,2) is even,
and so 9-^(1,2)0 is odd (Lemma 3.2, page 62). If and hence 9~^{l,2)e is odd (again by Lemma 3.2).
Therefore
(a, b) is
odd.
be a permutation and let 3.2, 9 is even if and only if This proves the theorem.
let ^
Now
ocj^
Then by Lemma
m is
a^ be the product of transpositions. even, while 9 is odd if and only if to is odd.
Sec. 5.5]
171
Problems
5.58.
Determine whether a and p oi Problem 5.52 are odd or even, using Theorem
Solution:
a
5.30.
(1, 5, 2, 3, 4)
=
is
(1, 5)(1, 2)(1, 3)(1, 4).
Since a
is
expressed as the product of an even number
of transpositions, a
even.
(1,4)(1,3)(1,5)(1,2),
P = (1,4,3,5,2) =
5.59.
and
so
is
even.
Determine whether a of Problem 5.54

Solution:
is
even by using Theorem
5.30.
= =
(1,2,4,8)(3,6,12,9)(5,10)(7,14,13,11)
{(1, 2)(1, 4)(1, 8)}{(3, 6)(3, 12)(3, 9)}{(5, 10)}{(7, 14)(7, 13)(7, 11)}
is
Thus a
even.
5.60.
Determine whether
Solution:
("i.

(ctj,
. ,
a)
is
even or odd.
>
.
o-m)
.
("i. 0'2)iO'i< "s)
(i.
m).
SO
(oj, is
Thus
Let
(a.j,
a^n) is
even or odd according as
a^) is the product of odd or even.

.
jn
transpositions.
5.61.
(!
Om), (6i, ...,6,)
be
;wo cycles of S.
Prove that there exists
G S
such that
e-Hai,...,a)e
Solution:
(6i,
...,6J.
Let
be defined by:
aj
if
(1) (2)
=
. .
6j
for
1,
,to;
{ci,...,c_} {6,, 6), put

. ,
{l,2,...,n}-{oi,...,ai}
and
if
{di,
.,d_}
{1,2,
.,w}
Cjfl
d;
for
1,
.,n-m.
(&i.
Then
e S and by Lemma
a group, the set of
5.29,
fl-i(ai,
-.Om)"
.6m)the conjugaey class containing x.
5.62.
If
is
all
elenents conjugate to
a;
is called
Write down the conjugaey classes of S4 using cycle notation. (Hard.)

Solution:
class containing an element e is the set of all conjugates of e. Our first task express each element of S4 as the product of disjoint cycles, and then to take all distinct conjugates of each element. We obtain the following classes:
is to
(i)
The conjugaey
{(1)}.
(As
(1)
(,
there are n) other conjugates of
(1)).
(ii)
What
conjugates of (1,2) are there? We S, l9,2e run through all distinct pairs taining (1,2) is
know that e~H\,2)e = (le,2fl). As 9 runs through (a, 6). As (0,6) = (6, a), the conjugaey class con-
f(l,2), (1,3), (1,4), (2,3), (2,4), (3,4)}

(iii)
What
conjugates of
(1, 2, 3) a:-e
there?
As
-(l,2, 3)9
(1, 2, 3) is
(l9,2fl,39),
we
will get all possible
3-cycles.
Hence the conjugaey
class containing
{(1, 2, 3), (1, 2, 4), (1, 3, 2), (1, 3, 4), (1, 4, 2), (1, 4, 3), (2, 3, 4), (2, 4, 3)}
(iv)
What
As
is
the conjugaey class ontaining (1,2)(3, 4)?

tf-i(l,2)(J,4)e
-i(l,2)e-i(3,4)9
(1, 2tf)(3, 49)
runs through S, we will obtain the product of all pairs of disjoint cycles. We recall that disjoint cycles commute; for example, (1,2)(3, 4) = (3,4)(1,2). Thus the distinct elements in the required conjugaey class Ere
e
{(1,2)(3,4), (1,3)(2,4), (1,4)(2,3)}

(v)
What
is
4-cycles.
the conjugaey class containing (1,2,3,4)? Hence the required oonjugacy class will be
Again we
see that
we
shall obtain all
{(1, 2, 3, 4), (1,2, 4, 3), (1, 3, 2, 4), (1, 3, 4, 2), (1, 4, 2, 3), (1, 4, 3, 2)}
172
FINITE GROUPS
[CHAP.
simplicity of An, n 5. In this section we aim to prove that A is simple for n - 5. What we must do is to show that if <l A and = A. Although the proof involves much calculation, ^ {i}, then the ideas are easy.
e.
The
Lemma
5.31:
Every element of A
is
the product of 3-cycles,

(e, ai){e, a'i){e, ai)
n^ 5.
e, ax,
Proof:
Every transposition
(ai, ai)
where
ai are distinct.
Now
every element of A is the product of an even number of transpositions, and therefore the product of products of pairs of transpositions. So it is enough to prove that a product of two transpositions is a product of 3-cycles. Consider (ai, a2)(bi, b2). We may assume that the four integers ai,a2, &i, &2 are distinct, for otherwise (ai, a2)(bi, 62) is either the identity or is itself a 3-cycle. As w 5, there exists an integer e such that \ e n and 0-1, 0.2, hi, b2, e are all distinct. Hence
{ai,a2){bi,b2)
= =
Let
{e,ai){e,a2){e,ai){e,bi){e,b2){e,bi)
{(e, ai)(e,a2)){(e,ai)(e, 6i)){(e, &2)(e, 61)}
(e,ai,ao)(e,ai, 6i)(e, 62, bi)
Lemma
5.32:
HO
A.
If
H contains
if x, y, z
a 3-cycle, then
H- An.
{1, 2, ...
,
Proof: Let (a, b, c) element of S that sends
e H. Then
are distinct elements of
n} and
6 is
an
page 170.
If 9
from
ft,
b, c
as
atox.btoy and c to z, we have 0~\a, b, c)9 = (x, y, z) by Lemma 5.29, is even, (x, y,z) G H as H < An. If 6 is odd, then there exists e, f distinct n ^ 5. Therefore * = (e, f)e is even.
<ir-'{a,b,c)<ir
d-ê,f){a,b,c){e,f)e
0~'^(a,b,c)d
(x,y,z)
Hence by normality,
{x,y,z)
e H. Thus
If
H = An.
Lemma
5.33:
H contains
all
3-cycles of S;
and by
Lemma
5.31,
Let H <\ H = An.
A.
H contains
e);
the product of two disjoint transpositions, then

there exists
e
Proof:
Let a
fts,
(ai, ft2)(a3, 04)
H.
from
fti,
a2,
a4.
Let
(ai,
ftz,
Since n then 6 G An.
5,
{1,
distinct
O'âO
(a2, e)(a3, a4)
a'ê^âO
(ai,a2){az,ai){a2,e){az,a4)
(fti,
a2)(a2, e)
(ai, e,a2)
Thus
H contains a 3-cycle,
5.34:
and the
result follows
from Lemma
5.32.
Theorem
Proof:
.
is
simple for
to
5.

Let < A^, - {l}. Then there exists aGH, a^' i. Let a = a^ are disjoint cycles. We may suppose without loss of generality that a^ A; 1, as the are arranged so that the length of a. length of a.^j for i = 1,
ffj,
. . ,
.
a^.
where
.
.
a^, a^,
.,a^
.,a^
commute by Problem
Case
1:
5.55,
page 169.
Also
Suppose ttj = (ttj, .,aj with m > 3. Let a= {a^,a^,a^. Clearly <jGA^. commutes with a^,...,a^, as they move different integers. As a~^ G A^ and P = a-â-âa G H. Note that
. .
<t
H <]
A^,
ff-laff
(a^,
ftg, ftj,
a^,
ftja^
fc
and so
^ = = = =
-i
^-^(a^,
ftg,
a,, a^, ...,
aja^
a^
r'K,a3,aj,ft^, ...,aj
(' ttm-l'

'
*l)(2' 3'
^V
%'' O-J
(ai>a2'4)
Since P
GH,
the result follows from
Lemma
5.32.
Sec. 5.5]
173
Case
a
2:
Suppose (a^, ttg, a J
m = 3,
G
A.
and a^ is also a 3-cycle. Then H contains
Let
a^^
{a^, a^, a^)
and
a^
{a^,a^,a^).
Let
Thus
H contains
a~^(T~â<T
a~â~^
a~^{<j~â^cr){(T~â2a)a^
a^
and the
Case
3:
result follows
from Case
1.
Suppose
m=3
and
^
a^,...,af. are transpositions.

(flî> <^2'
Let
a^
(a^,a^,a^.
Then
'^3)"l"2
"fc('*l' '^2'

*^3)"l"2
"ic
{%'
<^2'
3)^?l
(!' 2' 3)^
(!'
<^3' ^^2)
and the
Case
or
result follows
from Lemma
5.32.
4:
all
Suppose
the
(a^,
fflg,
a^).
Then
are of length contains
2.
Then
is
even, and
a^
{a^,a2), a^
(ttg.aj.
Put
a-W =
Thus
/r contains
(aj,a3)(a^,a2)a3a^
a^
a-âaa'^
(aj,aj(a2,ag)
and the
result follows
5.35:
by
Lemma
5.33.
is
Corollary
Proof:
position
The symmetric group S
not solvable for n
^ 5.
page 165).
{c} C A c S is a composition series of S for n ^ 5 (Example 2, By the Jordan-Holder theorem this is, up to isomorphism, the only possible comseries. Now |A| - n !/2, which is even for ^ 4. Hence [A {i}] is not a prime.
An O S and
But
if
any subnormal
series of S with factors of

is
prime order existed,
it
would be a com-
position series.
Therefore S
not solvable.
Problems
5.63.
Prove that
Solution:
if
G = A, n^
5,
then the derived group G' of
is
G.
If
G'
We know that G' < = {i}, G is abelian.
G.
Hence G' = G or else G' - {1} as G is simple by Theorem But A is not abelian for n 5; for example,
5.34.
(1,2.3)(3,4.5)
=
=
/I
V2
but Therefore G'
2 4
2
3
1
4
5
5 3 5
1
6 6 6 6
... ...
M n n
/I
(3,4,6)(1,2,3)
V2
G.
3 4
4
5
... ...
5.64.
If
GÂ
and n
5,
prove that Z(G)
U}.
Solution:
As Z(G)
Z{G)
-
<]
G,
Z(G)
=G
or
Z(G)
U}.
As Z{G)
is
abelian but
is
not (see Problem 5.63),
G.
174
FINITE GROUPS
Without using Theorem
Solution:
[CHAP.
5.65.
5.34,
prove that G'
- G = A^
for
rt
5,
where
G =
S.
as G'
5.66.
We use Lemma 5.33. Let a = (2,3,4), c = < G, the result follows from Lemma 5.33.
(1,2,3).
/?
a-ia-âa
(1,4)(2,3)
G'.
Hence
Are any of Aj, Aj, ^3,44 simple?

Solution:
Both Ai and Ag are of order 1, so they are not simple. A3 is of order 3, so A3 is cyclic of prime order and therefore simple since groups of prime order are simple. Finally A4 is of order 12. We have already seen that a group of order 12 is solvable. Indeed we saw in Problem 6.45 that its composition factors have orders 3, 2, 2 respectively. Hence A4 is not simple.
5.67.
Prove that S has no non-trivial element

Solution:
in its center for
3.
(Hard.)
Let a
Let
ai
S, a 7^
. .
.
1.
Let a
If
Then a-^fia = (aia,a.â) = (a^.ag) - fi. Hence ((ii,a2)Then a-^fia = {aâ, a^, aâ) = (a2, fti, 6) where 6 = aâ and 6 is an integer different from a^ and a^. No matter what h is, a-^Pa ^ p since a^P = dg but aâ-^Pa = a^. Hencp a Z(SJ. Thus no non-trivial element of S belongs to Z{SJ.
(!,
,
a^).
= ai ^ 3,
a^ be the decomposition of a into disjoint non-identity cycles.

j8
let
Z(Sn).
If
m = 2,
let
fi
(a^, a^, 03).
5.68.
Prove that A, S and

Solution:
{1}
are the only normal subgroups of S for n
5.
(Hard.)
Let
H<
e^
S.
Then
If
{i},
by Theorem
If
5.34.
AnH =
i.
HnA < A. Hence AnH = A or else AnH = {i}, as A is simple A^nH = A, then AcH. If H - A, as A is of index 2 in S, H = S. suppose H # {i}. U e G H and # # e is odd. As 92 jg even, e2e/^nA
i,
and as t is even, t = 9-1 = e. Hence H consists of only two elements, and e. As iS has no center (Problem 5.67), there exists m ^ S such that /j-iffp - e. But ytt-iff/t e fl^, as = {1,9}, this is impossible. This contradicts < S; and since ^ {1}. the assumption that
and so
Let t
H,
1
x ?^
t.
Then
t is odd,
5.69.
Prove that
Solution:
S'^
= A
is
for
n^5.
A d S^ (Problem
by Problem
5.68.
4.68(iii),
As SJA
abelian (cyclic of order 2 in fact),
page
116).
As S
is
not abelian, S;
- {i}.
But S^
<
S.
Hence
S;!
= A
Alternatively,
S^DA^ = A
by Problem 5.63 or
5.65.
A look back at
Chapter
5.
In this chapter we proved the three Sylow theorems. The first gives the existence of Sylow p-subgroups; the second states that a subgroup of prime power order is a subset of one of the Sylow p-subgroups; and the third states that all the Sylow p-subgroups are
conjugate.
The proofs of the Sylow theorems used a standard technique of finite group theory: induction on the order of the group. It is also worth noting our counting arguments, e.g. in the proof of the class equation.
Using the
center.
class equation
Then we proved
we proved that a group of prime power order has a non-trivial as a consequence that a group of order p*" (where p is a prime)
has a normal subgroup of order p'-K By repeatedly taking the center of factor groups, we defined the upper central series of a group. A group is nilpotent if a term of the upper central series is the group itself.
Next we gave a method of constructing a group from the cartesian product of two given H and K. This group had isomorphic copies H,k of H and K respectively which satisfied HK = G and Reversing this analysis we showed that if a group {1}. G had normal subgroups H and K with HK = G and {1], then G = H x K. Using this result and the Sylow theorems, we classified groups of orders 1, 2, 15 up to
groups
HnK=
HnK=
isomorphism.
CHAP.
5]
175
Then we
ability of equations.
defined solvable groups, so called because they led to a criterion for the solvnoted that our first definition involving a subnormal series with
We
factors of prime order did not extend to infinite groups. So we chose a criterion involving showed that subsubnormal series with abelian factors for our definition of solvability. groups and factor groups of solvable groups were solvable, and an extension of a solvable
We
group by a solvable group was solvable.
We next considered composition series (subnormal series with simple factors) and proved that every finite group has a composition series. In the Jordan-Holder theorem we showed that a composition series has a unique length and unique factors up to isomorphism.
In our final section we proved that the groups A for n 5 are simple. To do this we needed to express permutations as products of disjoint cycles. This led to a method of determining whether a permutation was even or odd. As a consequence of the fact that An is simple, we concluded that S is not solvable for n 5.
SYLOW THEOREMS
5.70.
Prove that the Sylow 17-subgroup Prove that the Sylow 13-subgroup
Let
is
normal in a group of order 255

normal in a g:roup of order 2
5
17.
5.71.
is
13.
5.72.
A
G
and
{0, 1} be the set of integers modulo 2, B = {0, 1, .,64} be the set of integers modulo 65, be the set of all pairs (a,b), a& A and b G B. Define the multiplication
. .
(ai, 6i)(a2, 62)
(i
+ 2.
(-l)26i
+ 62)
for
Is
(ai, bj),
(a.2,
62)
G. Prove that
is
a group of order 2
13 with respect to this multiplication.
a Sylow 2-subgroup normal?
5.73.
Verify the class equation
(i.e.
equation
(5.2),
page 135) for the group

in
in
Problem
5.72.
5.74.
Prove that the Sylow p-subgroup Is this true when p = 3?
is
always normal
a group of order 4p, where p
is
a prime
5.
5.75.
Prove that the normalizer of a Sylow p-subgroup coincides with

Let |G| = Ji and \H] then there are precisely Examine Nq(H).)
its
own
normalizer.
5.76.
where
1
i? is a
n/m
elements in
subgroup of G. If Hng-^Hg = {1} for all gGG-H, G which are not in any conjugate of H. (Hint:
THEORY OF p-GROUPS
5.77.
Show
Let
that every subgroup of index p in a finite p-group be a finite p-group and
is
a normal subgroup.
5.78.
H be
a proper subgroup of G. then G'
Show
that Nq(H)
H.
G).
5.79.
Prove that
if
is
nilpotent and G/G'
is cyclic,
{1}.
(G' is the derived
group of
5.80.
Use Problem
5.79 to
show that elements of co-prime order commute
in
a nilpotent group.
5.81.
Find an example of a group
with a normal subgroup
such that
G/N and
are nilpotent but
is not.
176
FINITE GROUPS
Let
[CHAP.
5.82.
G(i) = G and G(j+i) g G G^^ and x S G}). The sequence of subgroups gp({[g, x] is called the lower central series of G. G(i) 3 G(2) 3 Prove that a group G is D G(i) D nilpotent if and only if G() = {1} for some positive integer n.
\
5.83.
Find the
lovsrer
central series for Dg" for positive integers n.
(See Problem 5.82 for the definition of
lower central
series.)

5.84.
Prove that a finite nilpotent group groups and some other subgroup.
is
isomorphic to a direct product of any one of
its
Sylow p-sub-
5.85.
Employ the results of Problem 5.70 to prove that a group of order 255 is isomorphic to a direct product of its Sylow 17-subgroup and another of its subgroups. Thereby show that a group of order 255 is cyclic.
5.86.
Show that
the direct product of two nilpotent groups
is
a nilpotent group.
5.87.
Let
SP
_^
n)'(l
n))'
Show that
\G\
8.
Which group
of order 8
is
isomorphic
5.88.
Show
that i)
is
isomorphic to
G = gp
f (
n)
'
(^
-ij)
^^^""^
e2m/n
5.89.
Suppose
is
finite
group with
all its
Sylow p-subgroups normal.
Show
that
is
nilpotent.
SOLVABLE GROUPS
5.90.
Prove that a group of order pqr
is
solvable
when
p, q, r are
primes and r
>
pq.
5.91.
Show
that the direct product of two solvable groups
is
solvable.
5.92.
Prove that a group of order 4p, p a prime,
is solvable.
5.93.
In Problem 5.49, page 163, we showed that a group G is solvable if and only if G'"' = {1} for some integer n. Use this fact to give alternate proofs of Theorems 5.23, page 161, and 5.24, page 162.
5.94.
Show
that Z)
is
solvable for all positive integers n.

5.95.
Find the composition factors of a
finite
p-group.
5.96.
Show
that there
is
no simple group of order p^m, where p
is
a prime and
m
is
<
p.
5.97.
Prove S
is
all 3-cycles
not solvable for of S.)
5,
without using the fact that
simple.
{Hint.
Consider
5.98.
Find a composition series for the quaternion group.
5.99.
Show
that, except for
difficult
groups of prime order, there are no simple groups of order case occurs for order 36. See Problem 5.43, page 158.)
<
60.
{Hint.
chapter 6
Abelian Groups
group
with binary operation
is
abelian
if,
for
all
g,h
GG, g'h =
h-g.
It is
customary to use "
+"
for the binary operation in abelian groups.
We
will begin
this chapter with a preliminary section in
which we restate some of our
results
and
defini-
tions in additive notation.
One of the concepts which we will reformulate additively and generalize is the concept of "direct product" considered in Chapter 5. In abelian groups it is customary to talk about "direct sum" instead of "direct product". We note that the direct sum of abelian groups is again abelian.
We call direct sums of infinite cyclic groups free abelian groups. Direct sums satisfy an important homomorphism property which gives rise to the important fact that every abelian group is a homomorphic image of a free abelian group.
An
then consider classifying abelian groups according to the orders of their elements. abelian group G which has every element of finite order can be expressed as a direct sum of 2)-groups, i.e. groups every element of which is of order a power of the prime p. An important p-group that we shall introduce here is the p-Prufer group.
We
At
{a)
this point
we have
(6)
three types of abelian groups:
the cyclic groups,
the additive group
of rationals,
(c)
the p-Priifer groups.
The
rest of the chapter is devoted to
showing that many abelian groups are direct sums
of these groups.
We
Recall that a group G is finitely generated if it contains a finite subset with G = gp{X). show that every finitely generated abelian group is a direct sum of cyclic groups.
Furthermore we associate a set of integers to each finitely generated abelian group. This which we call the type of the abelian group, completely classifies finitely generated abelian groups; in other words, two finitely generated abelian groups are isomorphic if and only if they have the same type. This theorem is of great importance
set of integers,
in
many branches
of mathematics.
rationals
integer, then there exists
Q has the property that if g E Q and n is any nonzero such that nf g. We express this by saying that Q is / divisible. The p-Prufer groups are also divisible. Note that these are not the only divisible groups, e.g. the additive group of reals is also divisible. We obtain the pleasing result that if A is any divisible abelian group, then A is a direct sum of p-Priifer groups and groups isomorphic to the additive group of rationals.
The additive group of
Note. Any reader who would like a briefer account of abelian groups may refer to Sections 6.1a, 6.1c and 6.3. This will bring him quickly to the fundamental theorem of abelian groups, i.e. every finitely generated abelian group is the direct sum of cyclic groups.
177
178
ABELIAN GROUPS
[CHAP.
6.1
PRELIMINARIES
Here we
will practice expressing
our ideas in additive notation.
a.
Additive notation and
finite direct
sums
In this chapter all groups will be abelian, and we will use additive notation throughout. In terms of additive notation an abelian group is a non-empty set G together with a binary operation + such that
(i)
{a
+ b) + c = a + {b + c) =
b
for
all
a,b,cG G.
(ii)
a+b
+a + =
a for
all
(iii)
There exists an identity element, denoted by 0, such that a identity element is often termed the zero of G.
G.
The
(iv)
is
Corresponding to each there exists an element b such that a + 6 = 0. This b unique and is denoted by a. The element -a is often termed the negative of a.
aGG
Standard abbreviations are as follows:

(i)
g
If
+ (h)
n
is 0. is
is
written as g
h.

(ii)
a positive integer
If
for
If
n <
we
we write ng for g + + g {n times). write ng for g+---+g {n times).
If
w=
we
write
tig
an abelian group and is a subgroup, then automatically H < G and we may group G/H. (Warning: Some authors write for G/H.) Note that, in additive notation, a coset is simply a set of the form g + H. Instead of talking of multiplication of cosets, we talk of addition of cosets. Thus the sum of two cosets gi + H and g2 + H is, by definition, {gi + H) + {g2 + H) = {gi+g2) +
talk of the factor
G-H
tion,
The following table is useful in "translating" multiplicative notation into additive notaa and b are elements of a group G, and H and K are subgroups of G.
Multiplicative
ab
a-i
a"
a6-i
HK
aH
a+
Additive
a+b
na
a b
H+K
In Section 5.3a, page 146, we defined the internal direct product of two groups and Now, in additive terminology, we speak of a direct sum rather than a direct product. Instead of writing J if, we write It G = @ K, is, called a direct summand of G. Here we are interested in extending the concept of direct sum from two subgroups to a finite number of subgroups.
K.
HK.
Definition:
An
abelian group
each g
GG
G is said to be the direct sum of its subgroups Gi, can be expressed uniquely in the form
g
.,Gn
if
= gi+
+gn
we
write
where giG
Gi, i
1,
.,n.
In this case,
G=
Gn
or
G=
If
i=l
i:
Gi.
obtain the same definition of internal direct product that we gave in (The condition (i) of Proposition 5.19, page 146, falls away, as all groups studied here are abelian.)
n = 2 we
5.
Chapter
Sec. 6.1]
PRELIMINARIES
179
If
a;
G=
^ 0,
Gi then
e G,
then GiOGj
X
(0}
for
i - j.
For otherwise
if
a;
G GiHGj and
with
gi=

gj+t~
gf
Note that
= gi-i = = flfi-i = flfi+i = gi - x, and also with gi= = and gj = x. But this contradicts the definition of direct sum. if G = HK and H = LM, then G = LMK (see Problem 6.15).

= gi+ = gn = and
Qn
The following theorem provides a simple

direct
criterion for determining
when a group
is
the
sum
of its subgroups.
Theorem
6.1:
.,G be subgroups of a group G and suppose each element of G Let Gi, G. can be expressed as the sum of elements from the subgroups Gi, Suppose also that an equation
.
. .
=
with
gi
flTl
+
if
. . .
fifn
Gj for
l,
.,n,
holds only
gi
,
= g2=
= gn-^- Then
G
Proof:
If
is
the direct
sum
of the subgroups Gi,
Gn.
GG,
then
= gi+
gn
is
with
flfj
Gi,
i-l,...,n.
We
need only show that this expression
unique.
Suppose
g
is
g*
+gt
'
another such expression.
Then
= (gi-g*i)+
By our i = l,
. .
+ign-gt)
(gi-gt) = {gz-gl) = .,n and the two expressions for g are

hypothesis,
. . .
= (gn-g*) =
identical.
0.
Hence
gt
g*
for
which
arises: if Gi, G are abelian groups, does there exist a group G the direct sum of isomorphic copies of the groups Gi? This question is answered in the following theorem.
The question
is
Theorem
6.2:
Let Gi,
direct
G be abelian groups.
Then there
. . .
exists a
group
which
is
the
sum
of isomorphic copies of Gi,

closely the
G.
5.30,
Proof:
brief.
The proof follows

. .
argument
. .
If
(gi,
Here we will use additive notation. Let G .,hn) eG, define {gi, -ygn) and (hi,
page 146, so we will be x Gn. be the cartesian product Gi x G2 x = {gi + hi, .,gn + K). .,gn) + (hi, ..,K)
of
Problem
.
n components
Then G
is
a group.
Furthermore,
let
1,
. . .
Gi
.
j (0,
, firj,
0,
0)
flfi
|
Gi
L
. . ,
Then
(gi,
. .
Gi is a
.,gn) of
subgroup of G, and Gi

Gi for
i=
.
.,n.
It is clear
f2,
that every element
G is uniquely of the form {gi, 0, 0) + (0, G = Gi G and the result follows. An important result which we will prove in
,
0,
0)
(0, 0,
0, g).
Hence
sum
of Gi,
.,
G and
H is the direct sum
Section 6.1c states that of Hi, ..,H and Gi = Hi for

.
G is = l, i
if
.
the direct
.
.,n,
then
G^H.
who
i
In Section 6.1b we will define the concept of an indexed family Gi, studies Section 6.1c without reading Section 6.1b may take I
n
G 7. The
. .
reader
{1,
.,n)
and
Gi,
G7
as shorthand for Gi,
G.
Then
^ Gi
is
simply
^'i-
: : :
180
ABELIAN GROUPS
[CHAP.
Problems
6.1.
Prove that the negative of a

Solution
(a
6 is
6.
6)
{-a -b)
= = =
(a
+ b) + [(-a) + (-b)] = + (6 + {-a))] + (-6) = [{a + (-a)) + b] + (-6) =

[a
is
+ b) + (-a)] + (-6) + {{-a) + b)] + (-6) (0 + 6) + (-6) = 6 + (-6) =

[{a
[a
6.2.
Prove that
Solution:
if
is
abelian and
ig
a subgroup of G, then
(f
G/H
is
abelian.
(g
(f
+ H) +
+ H) =
+ g) +
(g
+ f) +
H =
+ H) + G
(f
+ H)
if
6.3.
Let
H be
This
a subset of an abelian group G.
Prove that
implies f
g G
is
H is
3.1,
a subgroup of
if
and only
f,g
&
H.
Solution
exactly the
same argument as
in
Lemma
page
55.
6.4.
Let n be an integer and

Solution
an abelian group. Prove that
if
g,hGG,
ng
then
n{g
+ h)
Thus
ng
nh.
ng
If n = 0, n{g + h) = + nh when n = 0. If M > 0, let w = m +

this in mind,
by
1.
definition.
Furthermore
Inductively
+ nh =
0.
n(g
+ h) =
mh.
Then
0.
we may assume
that m(g
h)
= mg +
Keeping
+ h) = (m+Viig + h) = m(g + h) + (g + h) mg + mh + g + h = mg + g + mh + h = (m + 1)^ + (m + l)h = ng + nh Finally if n < 0, n m, for w > 0, and so n{g + h) := m( (g + h)) = m((g) + {h)) = m{g) + m{h) ng -V nh
n{g
6.5.
Let n be any integer and G an abelian group. each ff in G is a homomorphism.

Solution
Prove that the mapping
which sends g
to
ng for
Now,
for
all
in additive notation, to
say that
6.4,
9 is
g,hG
G.
But, by Problem
{g
+ h)g =
a homomorphism means that (g n(g + h) = ng + nh = ge + he.
+ h)9 =
ge
he
6.6.
Let
be any abelian group and
let
XqG
Z}.
(X' 0).
r2X2
Prove that
gp{X)
In particular then gp{x)
Solution:
{v\
\
= nxi +
rx,
Ti
2,
XiG X}
(6.1)
{rx
&
Let
be the right-hand side of

(6.1).
Then
is
a.
ria;i+ clearly
+rx
and
k
fe
SiJ/j
+ Sp2/p
rx
belong to
(-Sj)yi
subgroup of G, for (ri,SjGZ and
H = H
. .
(^.
Moreover
if
Kj.j/^GZ),
then
h-
r^xi
(-Sp)^/^
a subgroup of G. It follows from the definition of that XqH. Since gp(X) is the smallest subgroup of G containing X, we have then g'p(X) C H. But if iCi, a; e X, then + ra: G 9P(X) for every choice of r; G Z. Hence gp(X) D H, and so gp(X) = H. TiX^ +
Thus
is
When
6.7.
X=
{x},
then gp(X)
is,
by
(6.1),
the set of
all
multiples of
x.
Let
G =
H and K be subgroups of G. Prove that H + K is a subgroup of G. H + K, prove that G ~ H K. (H + K={h + k\h&H and fee K}.) H + if 7^
(3
If
HnK =
{0}
and
Solution:
since both
k') e H + K since H and if hi + ki = h2 + fcj where hi, h^&H and k^, &2 G K, then x = h^ h^, k^, k^ belongs to = ^2. î ^2- Hence the expression of an element in the both H and if. Therefore x = and form h + k is unique. Since G = H + K, it follows that G = H K.
(k
u-vGH + K.
sions
H ;i
Now M =
fc,
So we have to prove that if u,v e H + K, then (h,h'eH, k.k'GK). Thus m-i; = (h-h') + are subgroups of G. If HnK = {0} and if we consider two expres-
and
if y^ 0.
-y
fe'
fc'
fej
Sec. 6.1]
PRELIMINARIES G
have a subgroup
181
6.8.
Let
H and
suppose that
G/H
is infinite cyclic.
Prove that
H is
a direct
summand
of G.
Solution:
for some
g G G and let S gp{g). Consider x G SnH. Then x = ng + H) = ng + H = H. But + H is of infinite order in G/H. Thus 71 = and so SnH = {0}. If x G G, xGng + H, i.e. x = ng + h for some nGZ and hGH. Hence G = S + H. Therefore, by Problem 6.7, G = S H and H is a direct summand of G.
Let
GIH =
gp(g
+ H)
where
ne Z.
As x G H,
n{g
jr
6.9.
Let
G=A
and
let
be a subgroup containing A.
Prove that
H= A
and
If
/i.
(Br\H).
Solution
As every element of G is uniquely of (BnH). What we must prove then

b
the
is
form
a+ 6
where a
GA
and 6GB. Hence (Bn/y). But and
aSA
= h~a.
HdA
that A + But, as AqH,
(BnH) = H.
b e S H,
B,
A + (BnH) =
and
h^ a+b
h-aGH.
Thus
bGBnH
HdBdH.
Hence
H =
6
HcA +
where
A(Br\H).
SG
Show
that the order ot a
6.10.
Let G be abelian and let a G G be of order m, divides the least common multiple I of and n.
of order m.
+b
Solution
Since both
lb
= qma + mb = G=
and n divide I, let I = qm = rn for some integers q and + = 0. Thus the order of a. + 6 divides I.
r.
Then
i(a
b)
la
6.11.
Let
H@K.
Prove that
GIH =
K.
Solution:
(H
G = H + K. Then by the + K)/H s K/{KnH). Now
subgroup isomorphism theorem (Theorem
HnK =
4.23,
page
125),
G/H =
{0},
and so G/H =
as required.
6.12.
Prove that if G is an abelian group, then the set S of elements of G of order a power of a fixed prime p is a subgroup. Deduce that a finite abelian group has one Sylow p-subgroup for each prime p dividing \G\.
Solution:
If
is
of p.
a and 6 are of order a power of p, then 6 is of order a power of p. So, by Problem 6.10, of order a power of p since the least common multiple of two powers of a prime p is a power Hence S is a subgroup.
Sylow p-subgroup of G. For if P is any subgroup of G of order a Pq S. So every Sylow p-subgroup of G is contained in S. S itself is of order a power of p (Problem 5.6, page 132). Since the order of a Sylow p-subgroup is the maximal power of p dividing the order of G, every Sylow p-subgroup of G must coincide with S. Thus, for each prime p, there is precisely one Sylow p-subgroup of G.
If
is finite,
then
is
the
power of
p,
by the
definition of S,
6.13.
Let
(i)
be an abelian group of order 36.
Prove
that:
9;
If
G, then g
(ii)
G=A
gi + g2 where the order of gi divides 4 and the order of ^2 divides B where A is the Sylow 2-subgroup and B is the Sylow 3-subgroup of G.
2'
Solution:
(i)
Let g be of order 2'3s. Put a, b such that am + bn 1.
= +
and
n.
Then, since
(m, n)
1,
there exist integers
Hence
bn)g
{am
=
0,
(am)g
{bn)g
g^
g^,
(i)
say
proved.
Now
(ii)
since ng^
n(am)g
a{nm)g
and similarly mg2
0,
is
Clearly = {0}, so -A B. Now if g G G, then, by (i), g = gi + g^ where g^ is of order dividing 4 and g2 is of order dividing 9. By the preceding problem, the set of all elements of order a power of 2 is the Sylow 2-subgroup A, and so g^ G A. Similarly g2 G B. Hence g G A + B and we conclude that G = A B.
AnB
A+B
182
ABELIAN GROUPS
[CHAP.
6.14.
Show that the group C of complex numbers is with respect to addition the direct sum of the subgroup consisting of all the reals and the subgroup consisting of all the pure imaginary numbers.
Solution :
Let
ber}
fl
=
i^
{a
+ iO
-1).
-I-
(here
Then
-1-
tb
(o -h iO)
(0 4-
a an arbitrary real number}, and let / = {0 + i6 6 an arbitrary real numclearly i2 and / are subgroups of C, fin/ = {0}, and, if a+ib&C, ib) belongs to R + I. Thus C = R 1.
|
6.15.
Let
G=
HK
and
H = LM.
Prove that
G =L
Q M K.
Solution:
Every element g oi G can be expressed in the form g = h + k where he H and k G K. But where I S L and m G M. Hence g = l + m + k. Now if g = li + mi + k^ with Z, e L, m^GM and k^ e K, put li + m^ = hê. H. Then g = h + k = h^ + k-^ and consequently h = h^ and k = ki. As h = l + m = li + mj, I = Zj and m = m,. Hence the result.
h
= l+m
b.
Infinite direct
sums
(See Note on page 177.)
It is convenient to label the subsets of a set with the elements of a second set. We are already familiar with such a device, e.g. in labeling a collection of sets Ai,A2, we have labeled with I = Z. In general, if / is an arbitrary set we shall denote by Ai, i S /, such a collection of labeled sets. A collection of labeled sets Ai, iG I, is called an indexed family. More formally, let e.I-^X be an onto mapping. Then 9 is said to be an indexing with the elements of I of the set X. We will denote the image of i by Xi, i.e. Xi = id. The
. . .
collection Xi, i
G. I,
is
then called a family of indexed
sets.
We
number
generalize our definition of direct of subgroups.
sum
to apply to the direct
sum
of an infinite
Definition:
An
abelian group
is
said to be the direct
for each
of the
gG,
g =0,
there
is
sum of its subgroups Gi, i GI, if a unique expression (but for order) for g
form
g
gj
- gi+
1', 2',
+
.
gk
where
Gj; j
1,
k,
with
&' distinct
elements of / and no
gi is zero.
Note that as
is
abelian,
gi+ +
gk
gk
+
.
.
gk-i
gi
for example; hence the uniqueness of the expression is understood to be without regard to the order of the elements gi, We write G = "^ Gi. ., gk.
iei
If / is finite,
above definition
conversely.
easy to see that a group which is the direct sum in the sense of the sum in the sense of the definition of Section 6.1a, and Usually we will use the definition of Section 6.1a whenever I is finite.
it is
is
also the direct
We
note that
if
G is
^ Gt
and
i,j
Gl, %-
j,
then GiflGj
{0}.
For
ii
G GiHGj
and X -0, then x
expressible as
= =
gi
where where
sum.
gi
Gi
and
g2
^2
Gj
But
this contradicts the definition of direct
Sec. 6.1]
INFINITE DIRECT SUMS
183
The analog
of
Theorem
Gi,
6.1 is the following:
Theorem
6.1':
Let
of
iGl be subgroups of the abelian group G. Suppose each element can be expressed as the sum of elements of the subgroups Gu Suppose .,k' are distinct elements of I and that the equation also that if 1', 2',
G
. .
= gi+
where
Qj
e Gy
holds
if
and only
ii gi
+ Ok = 92= =

9k
0.
Then G = ^Gi.
As the proof
is
similar to that of
Theorem
6.1,
we omit
is
it.
Again, as in Section 6.1a, the question arises:

abelian groups, does there exist a group
the groups Gi?
This question
is
if G,, iG I is an indexed family of the direct sum of isomorphic copies of answered in the following theorem:
G which
Theorem
6.3:
Let Gi, i G I be an indexed family of abelian groups. Then there exists an abelian group G which is the direct sum of groups isomorphic to Gi.
Proof:
Let
G =
-lo] 9: I -^ ,\J
Gi, i$
Gi for
If
all
I,
and
i9 is
9
the zero element of
Gi for all but a finite
number
6
of
iGlV.
-^
9,<i,GG,
define
*=
+ ^ by
6

i*
i9 +t</).
We
a
9
assert that
First, if
is
a group.
6,^G
G, then
<j)
is
clearly a
mapping of
/ into .U Gi
and
finite
number
<j)
tt>
+e
of the elements of I onto zero elements. so that G is abelian.
Hence
+^
+ maps all but gG. Note that
Next,
is associative.
For
if
^j, ^^,
,^3
G, and
if i
G I, + ^3) =
= H, +
Hence
(.^^
i{<l,^
i{<\>^
(<>2
+ ,^3))
.f>^)
^^
7;
:
4,^
(^^ + ,^3).
The mapping i ^ G Gi f or alH in / is the identity of G. For if 9 GG, then for all = + i9 = + i9 = i9 so that + O = 9. i in I, i{r) + 9) Finally, if 9GG, define 4>:i-^ -{i9). Then i{9 + 4,) = i9 + i4 = i9 + {-{i9)) = = Thus 9 + = and is the inverse of G. Now we prove that if Gi= {e\ 9 GG and j9 is the zero of Gj for all j in / with perhaps the exception of i9}, then Gi = Gi. Note that Gi is a subgroup of G; for if 9,<j>G G, then on putting ^ = 9~^ we note that if j î, j G I, ;* = y^ = - = 0. Thus * e dj.
ir]
rj
(i
i-q.
<!>
Tj
<j>
j,,!,
Next,
G..
let
V,
G.
Vj
->
G. be defined
by
9v^
-i9,
GG^.
<j>.
Clearly
<j)v^.
v^
is
a mapping of G. into
i<j>.
^v. = This means that i9 = 9,<f> GG^ and = for every j Gl, j î and so 9 But j9 = Next we prove v, is onto. Let aGG^ and define (9:/-*.u G. by id = a and je = if j^-i. Then 9 G G. and 0v, = a, and so Vj IS one-to-one and onto, v^ is a homomorphism; for if 9,4> G G^ and * = 9 + ^, then
To
see that
j,j)
is
one-to-one, suppose
Therefore
v^
defines
an isomorphism.
Finally
we show that G =
^
9i.
Gi.
We
already noted that
is
abelian.
We
need to show
If
.
that
if
GG,
then
9i
i9 is
9 = 9i+ +dk where each 9i belongs to one of the subgroups Gj. not the zero of Gi for only a finite number of elements of /, say ii,
ii9i
GG,
Let
it.
Gij
be such that
It is clear
ii9,
Gi^
be such that
12^2
128,
9kG
Gt^
be such that
ik9k
ik9.
that
184
ABELIAN GROUPS
[CHAP.
= Oi+
+di.
Now
where
Oj
suppose that
r^,
the zero of G,
7/
is
of the
form
-{-Ok
/.
= 6i+
G Gy and
.
1', 2',
...,k' are distinct elements of

j'l
j'iOi
+
=
-q
Then for
i'9j
;
^ j ^ k,
k.
+6^)
(since l',2',
..,k' are distinct)
from which
Oj
for every
Thus
G = ^Gi
iei
(Remark:
products.)
Problem 6.17 shows how the mappings of
this proof link
up with cartesian
a very powerful result, and we will use it repeatedly. note the important result that: given Gi = Hi for each i G I, then if G is the direct sum of its subgroups Gi and is the direct sum of its subgroups Hi, we can conclude
existence of direct
is
The
sums
We
G=H
6.16.
(Section 6.1c).
Problems
Let
be an abelian group with subgroups G^,

/
\
ie.
I.
Suppose that
is
G = gpi u
of
its
Gj
and that
Gj,
i
Vi6I
GjC\ gpl
Gj
I
{0}
for each
/.
Prove that
the direct
sum
subgroups
7.
Solution:
We
where
need only show that
if
= 9i+
gi
9k
/,
Gj,
and
1', 2',
.,k' are distinct
elements of
then
9i
91
=
= 92+
ffk
=
Then
9k
Suppose,
if possible,
that some 9i
is
not zero, say 9i

-ffi
= 0.
+
.
But
9i
g^GGy,
92
say,
and
^
92+
+ gk ^
9P{ U Gj Vie.I is
By
hypothesis
then,
gi
0.
Hence
9k
Slid the result follows.
6.17.
Compare the construction of a direct product involving cartesian products with the construction involving mappings by showing that an ordered pair can be thought of as a mapping. Hence show that the two groups obtained are isomorphic.
Solution:
An G
ev
in the first position, the
ordered pair can be thought of as a mapping from {1,2}. The image of 1 gives the entry image of 2 gives the entry in the second position. Let G = Gi X G2 and let
7
be the group as constructed in Theorem 6.3 with
{1,2}.
Let
v.G-^G
be defined by
(I9, 16)
for any
G.
Then
v is
clearly a one-to-one
S^
and onto mapping.
Also
is
a homomorphism.
{B
For
if
-irv
+ 4,)v -
= + 0, = (1*,2*) = (1 + 10, = (Iff, 29) + (l0, 20) =
2e
ev
+ 20) +
<t>v
6.18.
Let G be the subgroup of the TT be the ratio of the circumference of a circle to its diameter. additive group of the real numbers generated by the numbers n-, b-2, ir^, Let Gi = gp(ir*), i = 1,2, Prove that G = where P is the set of positive integers. Use the fact that <?i
Let
2 G p
is
not the root of any polynomial with integer
coefficients.
Solution:
Clearly each element of G is of the form ffj groups Gj, G2, .... Then we need only show that
+ ff^ where = gi+ + g
each
belongs to one of the fir^ implies that 9i = g2 " =

'
Sec. 6.1]
HOMOMORPHISMS OF DIRECT SUMS; FREE ABELIAN GROUPS

=
where the
t
</;
185
ffn
Zj
S
n.
G;.,
with
1',
.,w' distinct
Z,^l'
elements of {1,2,
.}.
Now
fifj
Zjtt''
where
G Z and
Zj
Then
=
If not all
22'^^'
Z!r"'
are zero, 17 is the root of a polynomial with integer coefficients, contrary to the statement in the problem. Hence the result.
c.
The homomorphic property

Let
of direct
sums and
free abelian groups
G = A B
respectively.
if
and
What
connection,
and let H be a group which contains isomorphic copies A and Suppose that H = (but not necessarily that H = any, is there between G and H"!
of
A+B
A B).
It turns out that is a homomorphic image of G. This follows from Theorem 6.4. This theorem when applied to particular cases leads also to important results: (1) Theorem 6.5 and (2) the concept of a free abelian group.
Theorem
6.4:
Let
G=
A B
and
and
into
let
H
=
^,
be any group.
îb
into
respectively.
Let 6, 4, be homomorphisms of Then there exists a homomorphism
C-.G^H
,
such that ^,^
<A-
is
Proof: If g gG, then g = a + b uniquely where a G A, b GB. Define gl-ae + b4>. uniquely defined and so is a mapping of G to H. Note that if gfi = Oj + b, where a, e
and
biGB
(i
1, 2),
then
((s
{a^
{9,
+ 9^)^ =
+ &,) +
(a, (6j
+ b,)K = + b^)^ =
b^4>
{(a^
+ a^) +
+
aê
{b^
b;))C
b^<j>
=
=
Hence
C is the required
+ a^)e + +
b^4,
a^9
(a^
b^<j>
aê
+ aê +
=
ia~
+ b^)C +
^
\b
{a^
+ b^)C =
g^^
g^C
homomorphism
as
^'
4>-
In exactly the same

is
way we can prove that if G =
^ Gi
9i,
and
if
for each
iGl, Oi.Gi-^H
a
|Gj
homomorphism
of Gi to H, then there exists a
homomorphism
or that Q
BiG^H
is
Bi.
We
Bi.
shall often say that 6 extends the
mappings
such that an extension of the
mappings
We
use this result to prove:

6.5:
Theorem
Let G,
Hi,
G
of
I.
If
is
the direct
sum
its
subgroups
the direct sum of its subgroups G; and Hi, then G ^ H.
is
Let Bi: Gi^ Hi be an isomorphism for all i G I. Then by Theorem 6.4 there homomorphism B:G-* H such that B \Gi = Bi for each i G I. To prove 6 is an isomorphism, we need only show that its kernel is trivial since 9 is clearly onto. If g GG, n' distinct elethen g 9i+ + gn where g) belongs to the subgroup Gy with 1', 2', ments of 7. Thus
Proof:
exists a
. . .
gO
giBy
ge'
is
hi
+
0.
fu,
where
hi
G Hi'
Since
{0},
Then g9 = only if each hj = 0, since H hj = gjBy and 0j. is an isomorphism, we have
the direct
sum
of its subgroups Hi.
gj
and g
Thus Ker9 =
and so
G^H.
now apply Theorem 6.4 when each G; is infinite cyclic. To begin with, let be infinite cyclic and any abelian group. Note that if ^ is a mapping of {c} into H, then there exists a homomorphism 9: C -^ The homosuch that B ^c) = 4>morphism B is simply defined by putting (rc)B = r(c<j>) for each r G Z. It is readily seen that this does define a homomorphism.
will
We
C = gp{c)
'
186
ABELIAN GROUPS
[CHAP.
Now
where
0*ix
let
each Gi be infinite
cyclic,
Gi
H is
0-
gv{ci),
any abelian group, then there
exists a
iGl. Let Z = homomorphism
(c;
|
6*
e /}. '^Gi^
I
U e.X^H H such that
IG
For, corresponding to each Gi, we know from the remark above that there exists a homomorphism (9i Gi -* i? such that Ci6ii = Cid. Hence it follows by Theorem 6.4 that there exists a homomorphism e*: '^Gi-^H which agrees with 6^ on Gu So we have the ^ required result.
:
'
Corollary
6.6:
The
direct
sum G '^Gi
satisfies the
following condition:
exists
for every
mapping 6: X-* H, H any abelian group, there phism 9*:G^H such that 0*^^ = e.
a homomor-
A
(i)
group
G which
gp{X),
contains a subset
such that
G=
(ii)
for every
mapping
9:
X -^ H, H
,x
any abelian group, there
exists a
homomorphism
basis
e*:G^H
is called
such that e*
=
is
e,
a free abelian group.
said to be freely generated by
X and X is called a
is
forG.
We
Conversely
have shown that the direct sum of we have the following

6.7:
infinite cyclic
groups
Theorem
{Xi\ iGl}, a free abelian group freely generated by a set the direct sum of its subgroups G, = gpixi) and each G; is infinite cyclic for all i G I.
If
is
then
is
Proof:
finite cyclic
This theorem
groups.
is
To
this
proved by showing that G is isomorphic to a direct end let be the direct sum of its subgroups Hi,
sum
of in-
H = ie Sî
/
where Hi gpQii) exists by Theorem

be extended to a
is
an
infinite cyclic
group generated by
hi.
O.X-^H be the mapping defined by homomorphism 6* of G into H, by the definition of

6.3.)
Let
(We know such a direct sum Then can xiB hi.

On
6.6,
the other
hand
<^
:
the
mapping
[hi\ i
H is the direct sum of the infinite cyclic groups Hi. Thus by Corollary G 1} X defined by h^ = Xi can be extended to a homomorphism
-^
^* of
H into G.
Then g =
Actually ^* and 6* are inverse isomorphisms. To see this, suppose g niXv + + nrXr- where V, .,r' G I and ni, .,nr G Z. Accordingly,
G.
{ge*)<l>*
= =
[ni{xve*)
ni{hi'<i))
nr'CiCr'^*)]^*
{nihi'
+ nr{hr'4>) =
UiXf
+ + nrhr')^,* + WriCr' = g
mapping on G. Similarly </)*5* is the identity mapping on H. a one-to-one mapping of G onto H. For if g, g' G G, then ge* = g'e* implies that (gO*)<!>* = ig'd*)<j>*. Since {ge*)<t,* = g and {g'e*)ci>* = g', we have g - g'. Furthermore if h G H, then h = (hc[>*)e*. Thus 0* is one-to-one and onto.
and so
9** is
the identity
is
This implies that 6*
= Hi. Finally are distinct elements we show that G is the direct sum of its subgroups Gj. If ... r' of / and Qi, Gr- respectively, then if .,gr are nonzero elements of Gr,
Note that each Gi
. .
is infinite cyclic,
since 0* is an isomorphism
1', 2',
and Gi^*
gi+
follows that giQ* a contradiction as so G = 2Gi.
it
+gr =
giO*
{6.2)
+ =
+ gr&* = 0. ]^ Hi. Hence '^'
But then, as
[6.2)
G Hv and
giB* ' 0,
we have
G,
does not hold.
Finally gp{Gi
\iGl) =
and
Sec. 6.1]
HOMOMORPHISMS OF DIRECT SUMS; FREE ABELIAN GROUPS

6.8:
187
Corollary
Proof:
infinite
d
:
Every abelian group
is
the homomorphic image of some free abelian group.
If
cyclic,
is an arbitrary group whose elements are gi, i G I, and gp{Xi), i G I, is then as we have seen, ^ = is free abelian and the mapping fi'?'(^')
Xi-* Qi extends to a
homomorphism
of
onto G.
Problems
6.19.
If
\gp(a)\
m,
\gp(h)\
and
{m,n)
1,
then
G =
gp(a)
gp{b)
is cyclic of
order mm.
Solution:
We
la
lb
order of
show that G = gp(a + 6). If I is the order of a + 6, then l(a +6) = la + lh d implies 0, by the definition of a direct sum. Consequently I is divisible by the order of a and the 6. Since (m, w) = 1, mn\ I and we conclude mn = I, so that G = gp{a+b).
where
6.20.
Find
|G|
is
the direct
sum
of
cyclic
groups of order
3.
Solution:
3.
Let Gj, ., G be subgroups of G and suppose G = Gi Each expression of the form 9i + g2 + + gn> where
. .
ff;
Gj,
(B G^, where each G; is of order gives rise to unique elements

total
of G. There are 3 different possible choices for g^, 3 for choices is 3-3 3 = 3". Hence |G| = 3".
^Tj,
etc.
Hence the
number
of possible
6.21.
Prove that
A
lA
is
a direct
is
summand
of
if
such that e
ff
the
identity
on A.
(Hint.
e A + i?.)
Suppose that
and only if there exists a homomorphism Use K = Ker e. Also consider g
9 of
G
to
onto
ge
prove
Solution:
on
G = A B. Then the identity homomorphism on A and the trivial homomorphism extend to a homomorphism e G ^ A which satisfies the required conditions (Theorem 6.4).
:
as
g-a&K.
6.22.
such a homomorphism exists, But xe = 0, since x 9 u = identity on A. K. Now letting g e G, ge - a for some Thus and G = A K.
Conversely,
if
g&AK
let K = Ker 9 and let e AnK. Then x = x0, G Ker 9. Hence AnK = {0} and so gp{A,K) = a&A. Then {g - a)e = ge - ae = a- a = and
a:
(i
A group G is said to be of exponent w if x G implies ix = and n is the smallest positive integer with this property. Let be the direct sum of two cyclic groups of order n, generated by = {x^Xj}. Prove that if G is any group of exponent n and 9 is a xi and xa respectively. Put mapping of into G, then there exists a homomorphism 9* of into G such that 9* |x = 9.
Solution
Let
z,9i
92
:
Hi =
=
il2
a;i9,
9* IH,
^G = 9i
and Hj = gp(x2). gp{xi) There is a homomorphism e^: H^-^ G satisfying for gp(zie) is cyclic and is of order dividing n. Similarly there is a homomorphism satisfying X292 = X29. Thus there is a homomorphism 9* G such that
:
H^ H^^
and 9*
|2
92,
by Theorem
6.4.
The
result follows.
6.23.
Let G be freely generated by a finite set X, ]Z| the form WjXj + + wx, where tw; S Z and

=
Xj
n.
Prove that every element of
is
uniquely of
G X.
Solution:
Let G; = ffp(xj). Now by the theorem on free abelian groups we know that G = G^ G, and each G; is infinite cyclic. Then each element of G is uniquely of the form ffi + + ff^ where Si S Gj. But we know from the theory of infinite cyclic groups that ffj = m^x^ uniquely. The result

follows.
6.24.
Let G be the direct sum of cyclic groups G^ of order 2 where i P, the set of positive integers. Let E denote the set of even positive integers. Then = gp(Gi i & E) is clearly a proper subgroup of G. Prove that ^ G.
Solution:
Let 9i Gj -> G2i for all Gj are cyclic of order 2.)

:
iGP
We
be an isomorphism. (Such an isomorphism apply Theorem 6.5 to obtain the result.
exists, since all the
188
ABELIAN GROUPS
Let G aut (A)
[CHAP.
6.25.
=A
B aut
a
where
page
(fi)
A is cyclic of order 3^ and B is cyclic of order 5^. Prove that aut (G) (hard) and hence compute |aut(G)|. (aut (G) is the automorphism group of
83).)
= G
(see Section 3.6a,
Solution:
identity on B.
then a can be used to form an element of aut(G) by letting it act as the (Theorem 6.4 states that a extends to a homomorphism. We must check that it Similarly for elements /? of aut (B). We use the symbols a*, p* to is one-to-one and onto.) represent corresponding automorphisms of G. Note that
If
aut (A),
(a+b)a*l3*
{aa+b)l3*
aa
hp
=
p*
(a+6)/3*a*
from which a*p*
p*a*.
Note that a*
p* implies a*
i,
the identity automorphism.
an isomorphism of aut (A) into a subgroup of aut (G). The mapping P -* P* subgroup of aut (G). As (aut (A))* n (aut (B))* = {i}, and as the elements of (aut (A))* commute with the elements of (aut (B))*, we have (aut (A))* + (aut (B))* (aut (A))* (aut (B))*, by Theorem 5.16, page 144. Now let 9 S aut (G). 8^^ induces an automorphism Sa on ^> foi" ^9 must go into a subgroup of order 9 and by Sylow's theorem there is only one subgroup of order 9 (as G is abelian). Similarly e\s induces an automorphism 85 o^ B. Then
The mapping
-^ a* is
is
an isomorphism of aut
(B) into a
(a+6)*9B
from which
= =
(ae
b)eB
ae
be
(a
+6)9
by Theorem
6.5,
>S =
Thus aut
(G)
(aut (A))*
aut (G)
(aut(B))* and, aut (A) aut (B)
A =
To compute |aut(G)| we must compute |aut(A)| and |aut(B)l. Let a e aut (A) and let gp{a). a must take a onto an element of order 9; so if aa = a^, (r, 3) = 1. Hence the poschecked. sibilities are r = 1, 2, 4, 5, 7, 8. Each of these gives rise to an automorphism of A, as can be Thus |aut (A)l = 6. Similarly |aut (B)| = 20. Accordingly, jaut {G)\ = 6 X 20 = 120.
6.2
SIMPLE CLASSIFICATION OF ABELIAN GROUPS, AND STRUCTURE OF TORSION GROUPS

Tentative classifications: torsion, torsion-free, and mixed
a.
Q, the additive group of Q/Z the factor group of the additive group of the rationals by the integers; rationals; C, the multiplicative group of complex numbers. Each of these groups is not isomorphic One way is to examine the orders of the to the others, but how would we prove that?
Consider the following three examples of abelian groups:
is of infinite order and every elements of the groups. Now every element of Q except element of Q/Z is of finite order. For if r G Q, r = m/n where m,n are two integers. + Z = Z. Let us, to avoid confusion, continue to use the multiThus n{r + Z) = nr + Z = plicative notation for C. We assert that C has elements of infinite order and also elements that -1 is of finite order. Recall that the identity of C is 1. Note that (-1)^ = 1 implies infinite of order 2 and 3-^ = 1 if and only if r = 0. Hence -1 is of finite order and 3 is of
order.
(i)
Summarizing, we have
has every element but the identity of
of finite order,
infinite order.
(ii)
Q/Z has every element
(iii)
has elements of
finite
order and elements of infinite order.
It is
then easy to see that the three groups are not isomorphic.
G is a group in which every element other than the identity is of infinite order, in which every element is of finite order, G is said said to be torsion-free. If G is a group element (not equal to be a torsion group. If G has both an element of infinite order and an said to be mixed. These three concepts provide us with to the identity) of finite order, G is between a rough classification of abelian groups and, as we have seen above, distinguish
If
is
Q, Q/Z and C.
Sec. 6.2]
SIMPLE CLASSIFICATION. STRUCTURE OP TORSION GROUPS
189
Problems
6.26.
Let
be the direct
sum
of torsion groups.
Prove that
is
a.
torsion group.
Solution:
Let
G =
(Pl
is
2
I
(?t-
li
g & G, g = Xi+
+ x^
for some integer
n and
Xj
belongs to some Gy.
If the order of Xj is Pj,
then
Vn)9
(Pl
Pn)l
(Pl
Pn)*ti
=
of order
2,
6.27.
Prove that Q/Z and G, the direct not isomorphic.

Solution:
sum
of groups Gj (iGZ) where each Gj
is cyclic
are
Following the method of Problem

is
6.26,
3,
it
is
easy to prove that every nonzero element of

isomorphic.
of order
2.
Since
-J^
+^
is
of order
Q/Z and G are not
6.28.
Prove that Q/Z and G, the direct not isomorphic.

Solution:
sum
of groups Gj
{i
G Z) where
each Gj
is cyclic
of order
3',
are
G is easily of Problem 6.26.

b.
shown
Since
to
;^
+^
have every element of order some power of 3 by following the method is of order 4 in Q/Z, G and Q/Z are not isomorphic.
The
torsion subgroup
In Section 6.2a
free, torsion
we
introduced a tentative classification of abelian groups into torsion-
and mixed groups. In this section we consider the question of whether it would not be possible to split a mixed group into a torsion-free group and a torsion group. This would provide the following program for investigating abelian groups:
(1) (2)
Investigate torsion-free groups.
Investigate torsion groups. Investigate
(3)
how they may be put together to form mixed
groups.
to
Such a program is found some significant results.
to be too difficult to accomplish completely, but it does lead
as
we show
Let T{G) be the set of all elements of in the following

6.9:
of finite order.
Then T{G)
is
a subgroup of G,
Theorem
Proof:
T{G)
is
a subgroup of
(termed the torsion subgroup of G).
G/T{G)
is
torsion-free.
Let a,b
GG
be of order m, n respectively.
Then
mn{a
Thus
T{G)
if
b)
= mna mnb =
is
a,bG
T{G),
a-bG T{G)
and T{G)
a subgroup of G.
Now consider = T{G). It

+ T{G) =
G/T{G).
follows that
finite order,
hence g
T{G).
there exists T{G). Therefore the only element of finite order in G/T{G)
is
Assume g + T{G) is of finite order n, i.e. n{g + T{G)) = ng + ng G T{G). As T{G) consists of all the elements of G of such that m{ng) = 0. Then g is of finite order and g e T(G);
is
the zero
Thus G/T{G)
torsion-free.
Problems
6.29,
Prove that if G is a group and the torsion subgroup of G.

Solution:
a subgroup of
such that
G/H
is
torsion-free, then
contains
G be of finite order. Then g + is of finite order in G/H. Since G/H is torsion-free, This means g e^ H, and so every element of finite order in G is contained in H, i.e. the torsion subgroup of G is contained in H.
Let
fl^
in
+H =
H.
190
ABELIAN GROUPS
Is the set consisting of
[CHAP.
6.30.
and of
all
elements of infinite order of a group automatically a subgroup?
Solution:
No. order 2.
order.
For example, let G = H K where H = gp (h) is infinite cyclic, and Now h + k and h are both of infinite order. However (h+k) + (h)
K = gp {k)
=
fc
is
is of of finite
6.31.
Prove that if the set consisting of and the elements of infinite order of a group subgroup, then G is either a torsion-free group or a torsion group.
Solution
constitutes a
= {h\ h G G and h is either or of infinite order}. Since G is Suppose G is mixed and mixed, we have g & G of infinite order and sr'(^ 0) S G of finite order. Now g' g is of infinite or an order so that (g' g) &H. But as is a subgroup, (g' g) + g = g' & H. Therefore g' is element of infinite order, contradicting the choice of g'. Hence G is not mixed.
6.32.
Prove that if T is the torsion subgroup of G, then of G. subgroup
TnH
is
the torsion subgroup of any given
Solution:
If ae is of finite order, then a& T. Hence a e TnH, and so the torsion subgroup of is contained consists of elements of finite order, and so contained in TnH. Conversely \a the torsion subgroup of H. in the torsion subgroup of H. Thus we have proved that
is
TnH
TnH
TnH
6.33.
Find the torsion subgroup of the subgroup of integers.

Solution
R/Z where
is
the group of real numbers under addition and
is
Suppose r + Z is of finite order (r G R). Then for some nonzero integer n, n(r + Z) Z. But + Z) = nr + Z, and so nr G Z. This means that r is a rational number. Thus T(R/Z) C Q/Z, where Q is the subgroup of rational numbers. On the other hand, if a + Z G Q/Z, then a = m/n
n{r
where m,n G Z and n
= 0.
So
n{a
+ Z) =
n(m/n
+ Z) -
n{mln)
+Z =
m+Z
= Z
Hence a + Z
is
of finite order and
Q/Z c T(R/Z).
Thus we have proved that Q/Z
T(RIZ).
c.
Structure of torsion groups. Prufer groups
a p-group or a p-primary group for some prime p if every element (If G is finite, it follows that the order of G is a power of p; p. see Problem 5.6, page 132. The definition of p-group given here thus coincides with that of Chapter 5 when the p-group is finite.) In this section we show that a torsion group is built out of j3-groups. Thus the study of torsion groups becomes essentially the study of
is called
A group G
is
of
of order a
power of
p-groups.
Theorem
6.10:
G be any torsion group and let Gp = {g\ g has order a power of any prime. Then if n is the set of all primes, p Gp G = pen
Let
p},
Proof:
We let the reader show that each Gp is a subgroup of G.
r order pI^pI^ p'^, Pu .,Pn distinct primes and ri, positive integers. Let Q "= P^ then (q.pl") !. Thus there exist integers a and pI^Sj^ b such that aq + 6p[" = 1. It follows that g aqg + bp^'^g. Now aqg is of order p^", and so aqg G Gp^. bp'^g is of order q. Since q is less then the order of g, we may assume inductively Thus every element that bp'^g is the sum of elements belonging to Gp_i,Gp_2, .,Gpi. the Gp, i.e. G is generated by the of G can be expressed as the sum of elements belonging to
Suppose that g
GG
and
is of
subgroups Gp.
Sec. 6.2]
SIMPLE CLASSIFICATION. STRUCTURE OF TORSION GROUPS

that
191
To show
Pi,
. .
G=
"^ Gp,
pen
we must prove
that
gi
gi+
.,p are distinct primes) occurs only for
= g2=
=
+ gn = = gn
(where
0.
Qi
Gp.
and
We
and
proceed by induction on n.
For n =
it is
certainly true.
If true for n, consider
gi+
let s'n+i
gn+i
be of order
pj;+i.
Then
and
p;+ifl'i+
+PUi9n =
By
the inductive hypothesis,
we have
Pn+i9i
= PUi92 =
as gi
gfi
= vU^gn =
some power of = 0. Hence
pi
Now Vl+^gi similarly for gi,

clude that
implies
. .
.,fir,
gi^ d we have
is
f2
of order
and pi
^^
p+i.
G=
pen
2 Gp.
called the
fi'n
also 5r+i
Arguing and we con-
The subgroups Gv are

Example
1:
p-components of G.
i.e.
Let us apply this theorem to Q/Z,

integers.
the additive group of rationals modulo the
Q/Z
is
clearly a torsion group.
(Q/Z)j,
= =
{x
+Z
+Z
I
of order a
power of p}
and
{x
{m/p''
+Z
for various integers r
+ Z p^x S < p"""!}

\
Z}
By
the theorem,
Q/Z =
pen
(Q/Z)v
6.4
(Q/^)p is called the p-Priifer group (also called a fcrowp o/ tj/pe p). In Section the p-Priifer groups vsîll be fundamental. Note that the p-Priifer group
{Q/Z)p
= U
Cr
where
=s
C^
gpiVp''
+ Z),
since
{Q/Z)p
{m/p"
+ Z\ y *"'
for variC^.
ous integers r and

result follows.
m<
p'-i}.
Clearly
{Q/Z)pDC^ and {Q/Z)pC
The
We now have at our disposal cyclic groups of all orders, the additive group of rationals, and the p-Prufer groups, together with all their direct sums. These, as we shall prove,
constitute a large class of abelian groups.
Problems
6.34.
Use Theorem
is
6.10 to prove that an abelian
the direct
sum
group of order pq, where p and q are different primes, of a cyclic group of order p and a cyclic group of order q.
Solution:
be of order pq. Then by Theorem 6.10, G - Gp G^; for if r is any prime other than = {0}. Why does G^ = {0}? If g e G then, as G is of order pq, pqg = 0. Hence r divides pq, which is not the case. Note that G^ - {0}, Gg # {0} by Proposition 5.9, page 137. |Gp| divides pq. Since G, - {0}, \Gp\ = p or q. As the elements of Gj, are of order a power of p, it follows that Similarly |Gqi = q. Hence, as the only group of prime order is cyclic, |Gpl = p. we have the result.
Let
q,
p or
Gr
6.35.
Show
that a direct
sum
of p-groups
is
again a p-group.
Solution:
Let
ffi
G = ^2^G;
1', 2'
e
p
Gj,,
is
n'
of
a power of p,
is a p-group. Let g el. Let p'' be the maximum order and so G is a p-group.
where each G;
G.
Then
sr^.
g^ g^+...+g^
Then
p'-g
where each
of the
0.
Thus the order
192
ABBLIAN GROUPS
Show
that
if
[CHAP.
6.36.
is
an element h G
Solution:
and each element g & G there exists a p-Priifer group, for each integer fe # such that kh = g. (We shall call a group with this property divisible.)
m < p''. Let k rpH where p and I are co-prime, Then g m/p^ + Z where + Z. Then p^hi = g. As I and p'"^" are co-prime, there exist integers a and 6 such that al+bp^-^^ = 1. Therefore
Let g
.
G.
and
let
h^
'm,/p''+^
hi
((ii-|-bp'- + s)fei
alhi
+ =
bp'' +
^hi
=
g
alh^
Put h~ahi.
Then
kh
pHh =
pHah-i
p^h^
00
6.37.
Let
G=
Gj where G;
is
a cyclic group of order p; and Pi, P2 be a p-Priifer group for any prime p.
are the primes in ascending
order of magnitude.
to
Let
Prove that
is
not isomorphic
H.
Solution:
is
Let pj be a prime different from p. not isomorphic to H.
Then the Pjth component Gp^
' {0}
but
Hp^
{0}.
Thus
d.
Independence and rank
We
Let
introduce here an important concept in abelian group theory, the concept of rank.
be an abelian group. are distinct elements of X and then niXi
A
ni,
subset
.
.
X of G is called independent if whenever xi,

.,Xn
.,nr are integers such that

-I-I-
niXi
nrXr
(6.3)
=
if
UrXr
is
0.
Note that
n2=
G =nr=0.
set, it
torsion-free
and
{xi\ i
X
G
Suppose
X-
independent, then equation (6.3) implies wi is an inThen if 1} and xi - x-, for i ' j.
is
dependent
follows readily that gp (X)
is
ffPiî)-
We
need one further definition.

if
An
niXi
element
-I
a;
in a
group
is
dependent on a subset
Z of G
for
nx +
. .
-h
nrXr
some choice of Xi, .,Xr G X, n and nj G Z and nx - 0. In other words, x is dependent We say Y CG is if there is an integer n with nxÔ and nx gp{X). on the subset Observe also that if G is if every element of Y is dependent on X. dependent on is dependent on Y and Y is dependent on T^, and if torsion-free with subsets X, Y and then X is dependent on W. For if a; G Z, then for some integer ^ 0, and integers we can find integers Since Y is dependent on -I- nrVr {Vi G Y). ., nr, nx = n^yi -tni, = mi m^. Then Put r. such that mti/i G s(p(W) for t = 1, mr =^ mi ^ 0, Furthermore, mnx G fl'p(W). r, and consequently clearly mi2/i G gp{W) for i = 1, since G is torsion-free, mnx ' 0. Thus every element of X depends on W.
XcG
. ,
The main
result that
we
shall
now prove
is
called the Steinitz

. .
Exchange Theorem.
Theorem
6.11:
.,0} be an independent subset be torsion-free and let A = (ai, B = {bi, ...,&} is another subset of G such that A is of G. Suppose and B depends on A U C where C is a subset dependent on B. Then of B and \C\=n- m.
Let
n^m
.,Zn
Proof:
We will use induction on m.

there exist integers
zai
z,
.
For
.
m=
+
1 it is clear that
w g m. Now
ai
depends
on
B means that
such that z'0,

zbn
zibi
Sec. 6.2]
193
and thus for some integer

the result holds for
i,
Zi=0,
lî^n.
Hence
depends on
{ai}
U (B -
{6i}).
Thus
m = 1,
with
C=
B- {bi}.
m-r,
{ai, .. .,ar}
Next we assume that the by the inductive hypothesis,

. .
and consider the case m = r + 1. Then depends on B, and so inductively B and |D| = n-r. But {Or+i} depends on B and depends on {ai, ..Or} UD where Or} U D. B depends on (oi, .,ar}UD. Then by our remarks above, {ar+i} depends on (ai, .,yr,zi, Zs such that Thus we can find integers yÔ.yi,
result holds for
n^r. Now
DcB
. .
yttr+i
yitti
yrUr
zidi
Zsds,
di,
.,ds
&D
+ yrttr implies that Suppose, if possible, that Zi-Z2= =Zs = 0. Then yor+i- yiai + = - 0. Then it {di}. Let C So some Zi the elements ai, ., Ur+i are not independent. depends on Since B is clear that {ai, depends on {aj, .,ar} .,ar+i} U C. Finally and \C\ = \D\ - 1. {tti, ttr+i} U C. Or) U D, then B depends on (ai, "^
UD
CcB
\C\
= (n-r)
= n
(r
l)
= n
as desired, and the proof of Steinitz's theorem
is
complete.
Let us
(i)
call
a subset S of a torsion-free abelian group
a maximal independent set
if
it
is
independent and
g G G and g ^ S, then Su {g} is not an independent set. Suppose now that G, a torsion-free abelian group, has a maximal independent set S that is finite. Let T be any other finite maximal independent set. By the Steinitz exchange theorem, \S\ ^ |r|. Also by the same theorem, \T\ ^ |S|. Hence we can without ambiguity
(ii)
define the
set
to be
rank of a torsion-free abelian group G which has a finite maximal independent \S\. If G does not have a finite maximal independent set S, we shall say G is of
if
infinite rank.
It is
easy to see that
G and
H are
isomorphic groups, then they have equal ranks.

obtain a result concerning free abelian groups.
As a consequence
of these remarks
we
If F is free abelian with a finite set of free generators X, then is a maximal independent set of F (see Problem 6.41 below). Hence the rank of F is \X\. Similarly if F is also freely generated by a finite set Y, then rank of F = \Y\. Hence |r| = |X|. Thus we have
Corollary
6.12:
If
i?'
is
a group freely generated by two
finite sets
X and
Y, then \X\
\Y\.
To all abelian groups have maximal independent sets. a result called Zorn's lemma. Before stating the lemma, we consider the following examples:
have as yet not proved that
do so
We
we need
(a)
Let '=P={A,B,C,D}, where
A=
{0,1},
B=
{1,2},
C=
(0,2),
D=
{0,1,2,3}.
We
<??
(6)
inquire:
it
does
T*
Clearly
"P
does, for
have a largest element, i.e. one that contains all the elements of DdA, DdB, DdC and DdD. Thus D is a largest element.
largest element of
A, B, C are as in (a), and E = {1,2,3}. Now there is no search for some concept replacing that of a largest element. Note that although E is not a largest element, no element other than itself contains it. Then E is called a maximal element. Similarly C and A are maximal elements, whereas
Let
= {A,B,C,E}, where
'P.
We
B
(c)
is
not.
If [Tl is finite, then it is clear that Ai of 'P. If there is an element of

is
? has maximal elements. For we choose any element T that contains Ai properly, we call it Az. If there
it
an element of "tP that contains A2 properly, we call C Ai C get a chain of elements of 'P, Ai C A2 C

A3.
of elements, this chain ends at An, say.
Clearly
is
Continuing in this way we As ? has only a finite number a maximal element.
194
ABELIAN GROUPS
[CHAP.
(d)
On
"P
A=
the other hand, not {0,1, .. .,71} for n
all sets "P
of sets have
maximal elements.
i
For example,
let
l,2,
Let
if
'P
{Ai\
has no maximal element.
For
X GV,
then
G P, the positive integers}. Then = Ai for some i, and X c Ai+i, but
XÂi+i. Zorn's lemma

maximal element.
(1)
establishes a criterion for determining vs^hether a set of sets has a What one needs is some condition for handling an ascending sequence
(d).
of sets such as the Ai in
To
state the criterion

set in "P if
We
define
to be a
maximal
we need some definitions. for each X e <P, XdA implies

if
X ^ A.
or
(2)
Let C be a subset of
with the property that
Y qX. Then C
is
called a chain in
(in {d)
above,
X,Y G > then either ? itself is a chain).

<J>
ZCF
We We
are
now
for every chain
a position to state Zorn's lemma: Let Û ^ Z is an element of ?. Then C X 6

in
"?
in ?,
be a set of sets. Suppose that has a maximal element.
lemma. We take it as an axiom. We could assume a more innocent sounding axiom instead, namely the axiom of choice, which says that an element from each set may be chosen from a collection of sets. The proof of Zorn's lemma can be derived from the axiom of choice (see Problem 6.42 for a sketch of the proof).
will not prove Zorn's
Using Zorn's lemma we prove
Theorem
Proof:
6.13:
Let
'P
be any abelian group.
Then G has a maximal independent

set}.
C/
set.
Let
*'
I
= {X\XqG and
an independent
Let Q be any chain in ?.
Let C
{î
-^l-
To apply Zorn's lemma we must show that

If not,
. .
VJ
XiGT.
Clearly
pos-
U CG.
Is Z7
an independent set?
sible to find distinct elements Ui,
.,iinG
U U

is
a dependent
set.
. . .
and integers
ri,
This means that r such that
it is
TiUl
+
1
+
ui
TnUn
with at least one

1.
TiUi
=7^
0.
As
Then since C is a chain, element of Continuing the argument in some Xi G C- But this is a contradiction, since every element of C is independent. So U is independent and U G'P. We conclude, using Zorn's lemma, that '? has a maximal element and this is precisely the maximal independent set required. Hence the result follows.
. .
.
= U Xi, i& either Xv c X^C/
G Xv, U2 G Xi- where 1', 2' are elements of or X^' C Xv. Thus Ui, U2 both belong to some tt all belong to this way we find that Mi,
,
From Theorem 6.13 we conclude that maximal independent subset.

Problems
6.38.
if
is
of infinite rank, then
has an
infinite
Find the rank of the additive group

Solution:
If nii/ni
of rationals.
and
Ttn^n^,
are elements of Q, mi,

(to2Wi)('Ii/mi)
m,^, n^,
n^ integers, then
( mxn.2)(m2/m2)
=
Q
is 1.
Thus every
set of
two elements
is
dependent.
Accordingly the rank of
6.39.
Show
that the p-Priifer group has no independent set consisting of two elements.
Solution:
y be elements of G, a p-Prufer group, x = 0, y = Q. Then x,y & C^, say, for some r 1, page 191, for the notation C^). Let C^ = gp(g) and let gp(x) be of order p'. Since V^^^g is of order p', gp(p''~*g) = gp{x). Thus gp(,x) ~ gpip^g) and gp{y) gpip'g) for some i, j. If i j, it follows that gp(y) 2 gp(x). (If i j, we merely reverse the roles of x and y.) Consequently X = ry for some integer < r < order of y. Thus {r)y + ! x = and x, y are not independent.
Let
X,
(see
Example
Sec. 6.2]

if H and K are torsion-free m + n. (Difficult.)
195
6.40.
is
Prove that of rank

Let

groups of
finite
rank
m and n respectively,
.,fc
if
r
then
G=
HK
ele-
Solution:
fej,
.
.
.,ft.
ments of K. Then the
s&
0,
be a set of independent elements of H, and fcj, For h^, k^, fc is independent. set Aj then , , ,
.
.
a set of independent
. ,
rx/ij
-I-
rh,n
Î'^'i
rihi
+
,
+
,
r,h^
-Sjfci
sA;
But
HnK =
{0},
and so
rj/ij
+-!- r^h^
. .
=
ki,
.
-Sifej
.
sk
= =
. . . ,
Thus, by the independence of

ri
hi,
.,
h^ and
., fc,
=
.
r2
=
. . .
=
.
. .
Sj
S2
h^, fcj, that {h^, not independent, there exist integers

.
Now suppose that {h^, h & H and k S K such
.,
, .
not maximal; say, there exists an element h + k where Since {/ij, A;, fc + k} is independent. /!, /i} is ., h^, k^, tj, f, t not all zero such that
fc} is
, , . . ,
tihi
+ +
.
tjin
th
Q
tj,
. . .
If
is
=
Next,
we have a
Hence
{fc, fej,
. .
.
contradiction to {h^,
t
,
.,
h^} being independent, as then at least one of
t^
nonzero.
0.
A;}
is
not independent,
i.e.
sk
+ Sjfej +
+ sk =
stihi
-I-
0.
Hence arguing as above,

-I-
there exist s, Sj, ...,s, not all zero, such that it follows that s ?^ 0. Thus
-\
=
is
But as st = 0, {hi, ^2, a maximal independent

.
+ + tsk^ tsiki a(th + tyhi + + t^h^ + t(sk + s^ki + + sfc) = h^, ki k,h + k} is not independent. Thus {hi, h^, set and rank H @ K = m + n.
stjji^
-{-
st(h
fe)
-t-
.,
h^,
fcj,
. ,
fe}
6.41.
(a) (&)
If
is
free abelian with a finite set of free generators X, prove that the rank of
is
\X\.
Prove that
F
.
cannot be generated by fewer than
\X\ elements.
Solution:
(a)
Let
XWe
{xi,
.,}.
By Theorem
6.7,
page 186,
G =
gp{xi)
gvix^)
gpix)
1.
proceed by induction on n. For n = 1, G is infinite cyclic, and the rank of G is clearly If true for n = r, suppose n = r+ 1. Then gp{xi) gpix^) is of rank r, and G the direct sum of a torsion-free group of rank r and a group of rank 1. By Problem 6.40, G thus of rank r + 1, and the result follows by induction for all n.
is is
(6)
Let
gp(gi,
.,g^).
(Theorem
6.11), as
is
X is an independent set. Then by the Steinitz exchange theorem dependent on {gi, Thus we obtain the result. .,gr}, n r.
. .
6.42.
Let
to
(a)
^ be
a set and
a collection of subsets of X.
Suppose that
if
A G
"P,
all
subsets of
belong
"P.
Prove that a; S A. X
if
AG
'P
is
not a maximal element,
there exists a set
A* = {A,x}
G'J>
with
(6)
Assume that A*
= A if A is maximal, otherwise that A* has been defined equal to {A,x} with as stated in (a). Let C be a chain in T". Suppose that if Cj, t G /, is a family of elements in C, then U ' CiGC- Suppose also that if A G C, A* G C- Prove that V has a maximal ^ ' element.
Â
If
Solution:
(a)
AGf
a;
is
exists
(6)
G B A.
not maximal, there exists a set B G T' such that Since {A, x} is a subset of B, {A,x} G 'P.
B = A
and
BdA.
Hence there
Let
Af
= U C. Then cec
M G C,
it
and
so
M* G
C-
But
element of C; in particular, maximal element of f.
contains M*.
Therefore
M* 2 M. M M*
However,
contains every
and we conclude that
is
In assuming that A* can be defined we used implicitly the axiom of choice. lemma requires converting the theorem into this problem. For details see P. R. Halmos, Naive Set Theory, Van Nostrand, 1960.
Remarks.
(1)
(2)
The proof
of Zorn's
196
ABELIAN GROUPS
Let
that
[CHAP.
6.43.
G
is
be an arbitrary non-abelian group. Prove that G has a maximal abelian subgroup not properly contained in an abelian subgroup of G).
(i.e.
one
Solution:
We
subgroups of G.
use multiplicative notation for G since Let C be a chain in V and let
it is
t7
c and if ff belongs to XiS C and h to Xa e C, then as either Zj c X2 or X2 c X^, it follows that g, h belong to some element X of C- Hence ffh'^ S X as Z is a subgroup, and gh-^ G U. Also, gh = hg as X is an abelian subgroup of G. Consequently U is abelian. Hence V G 'J>. By Zorn's lemma, '?' has a maximal element M, say. If is the maximal abelian group sought.
g,heU
= U xe
not abelian. Let X. Then [7
<? is
be the set of all abelian a subgroup of G. For if
6.3
a.
FINITELY GENERATED ABELIAN GROUPS

Lemmas
for finitely generated free abelian groups
In Section 6.3b we will show that all finitely generated abelian groups are direct sums of cyclic groups. We will do this by using a lemma (Lemma 6.15) about subgroups of free
abelian groups.
The
relationship between
groups
groups.
is
easily obtained
by noting that
all
Lemma 6.15 and finitely generated abelian abelian groups are factor groups of free abelian
be the direct sum of
. . . ,
Lemma
6.14:
Let
If
&i
G ~ gp{ai) = fti + r2a2 +
9P{an)
infinite cyclic groups.
+ rndn, where r2, G = gp{bi) gp{a2)

gp(au
.
r are
any
integers, then
gp{an)
Proof: As gp{bi,a2, are any integers, then

implies
all Si
.,an)
=
S161
.,a)
=
+
G,
Snttn
we must
only
show that
if Si,
.,s
S2a2
=
terms,
{64)
are
0.
Substituting &i
ai
+ rza2+ -\-rr,a.n into {6.U) and collecting Siai + (S2 + Sirâi + + (sn + Sir)a =
we
obtain
As G =
Thus
Si
(ai}
{an},
Si
S2 4- SiTi
=
is
Sn
SiTn
= Si
Sn
and the proof
complete.
. . .
The next lemma

free abelian group
6.1c).
is
is
a crucial one. We recall that a basis Ci, a set of elements such that G = gp{ci)
Cn

for a finitely generated gp{cn) (see Section
Lemma
6.15:
Let
that
of G.
be free abelian, the direct sum of n cyclic groups. Let fl^ be a subgroup Then there exists a basis ci, .,Un such Cn of G and integers ui,
. . .
H = gp{uiCi, U2C2,
a, b, c to
VmCn)-
denote basis elements of G, h, k, I to denote elements of H, We prove the result by induction on n. For re = 1, G is cyclic and the result is a consequence of Theorem 4.9, page 105. Assume the result is true for free abelian groups of rank less than n where re > 1. Let G be free abelian of rank re. = (0}. = {0}, we may take an arbitrary basis Ci, ., c We assume also that For if = Un = 0. gp(uiCi, ., mc) where ui = for G. Then
Proof:
t,
We
use
q, r, s,
u, V to denote integers.
To every basis we associate an integer, called its size {with respect to H). Let {ai, ., a} be a basis for G and let q be the smallest nonnegative integer such that there exists h G
. .
qai
q2a2
+
.
qna,
.
q2,
.,qn integers
{6.5)
Then q
is
termed the
(ai,
.
.
size of the basis {ai, a2,
.,
an}.
i.e. if [bi,
.
.
Assume
..On} is a basis of smallest size,
.,
b} is a basis of G,
then the
size of {&i, ...,&} is not less
than
q.
Sec. 6.3]
197
algorithm,
Let h be as in equation (6.5). if qi is not divisible by
We
q,
qi
show that q

divides
q2,
...,qn.
From
the division
But
if
we put
bi
Ci,
bz
= nq + s, where < Si < q. Hence + QnOn + Siat + h = q{ai + Tiat) + = aa, b = a, we obtain a basis by Lemma & = ai + nai,

Furthermore assumption. Thus

6.14.
this basis is of smaller size than the size of {ai,

Si
.,
an},
contrary to our
and q divides
h
q-i
for
2,
.,n.
Let
q-i
nq.
Then
^(ai
+ ratta +
+ ran)
{ci, ,
. .
Let
ci
ffli
+ r2a2 +
+ TnOr,.
Then, by
Lemma 6.14, = qci

/i
.,
a} is a basis for G.
Also
(6.6)
If
fc
titti
O^v
of
q,
<q, then
V
0.
fc
+ tnttn e H, - wfe G H
I
it
follows that
i;
is divisible
by
q.
For
if
has
as its coefficient of au
As v <
q',
ii = mq + 1; with by the minimality
Therefore
uh G
,
gp{a2,
this
l
a)
if
Hence
fi'?>(a2,
.,(in)r\H
L,
say.
From
fc
we
conclude that
G H,
then
{6.7)
= uh +
where IGL.
By
that
the inductive hypothesis there exist a basis

is
.
C2,
and integers
.,UnCn.
Uz,
generated by
.
W2C2,
WnC.
Hence by
((?.7)
every element of
. .
.,Un such belongs to

.
gp{h,UiC2,
.,MCn).
On
the other hand,
contains h,U2C2,
.
Thus
H =
Put
Ui
gp{h, U2C2,
UnCn)
q.
By
(6.6),
H
Also,
Ci,
.
. .
gp{UiCi, U2C2,
UnCn)
c is
a basis for G.
Hence the
result follows.
its
Note that if any m is negative, we can replace d by can assume that the tii are nonnegative.
inverse cu
In this manner
we
Lemma
6.16:
Suppose G = AB. Let Ai,Bi be subgroups with = Ai + Bi. Then G/N = AMi i?/5i.
: :
AiCA, BiCB and
Proo/: Let K = AMi B/Bi, and let e A -* A/Ai and ^ B ^ B/Bi be the natural homomorphisms. 9, extend to a homomorphism -^ of G into K. Then Ker ^ D Ker d = Ai and Ker * D Ker <^ = Bi. Thus Ker*DAi + Bi. Now let ajSKerst^. Then a; = a + b, aGA.bsB. x-^ = (a + Ai) + (b+Bi) and this is the identity element only if a G Ai and bGBu Hence ccGAi + Bi, and so Ker^ = Ai+Bi. By the homomorphism theorem (Theorem 4.18, page 117) G/N = K and the result follows.
<j)
Corollary 6.17:
.,itc) .,c. Let H = gp{uiCi, be free abelian with basis Ci, Then G/H is the direct sum of where ui, .,Un are nonnegative integers. and ul = > if ., u'n, where ul = Vd if Ui^Ô cyclic groups of orders ui,
Let
Ui
0.
Proof: The result follows by repeated application of Lemma\6.16.
b.
Fundamental theorem
of abelian
is called
groups
the fundamental theorem of abelian groups.
Theorem
6.18:
Let
finite
number
be a finitely generated abelian group. of cyclic groups.
Then G
is
the direct
sum
of a
198
ABELIAN GROUPS
Proop.
[CHAP. 6
G = F/H where F
Lemma
integers
6.15,
F has
a basis
ci,
Ui, .. .,u.
We now
G
a finitely generated free abelian group (Section 6.1c). By c such that = gp(uiCi, ., v^Cn) for some nonnegative apply Corollary 6.17 to conclude that G = F/H is the direct
is
. .
. ,
sum
of cyclic groups.
6.19:
Corollary
If
is finitely
generated,
it is
the direct
sum
of a finite
number
of infinite
cyclic
groups and cyclic groups of prime power order.
sum
Proof: It is only necessary to show that a cyclic group of composite order is the direct of cyclic groups of prime power order. This we have already done in Problem 6.19,
6.20:
page 187.
Corollary
If
is
a group with no elements of

is
finite
order and
is finitely
generated,
then
Proof:
free abelian.
is
the direct
infinite cyclic as
sum of a finite number of cyclic groups each of which must be has no elements of finite order. Thus the result follows.
Problems
6.44.
Prove that every

Solution
finitely
generated torsion group
is finite.
By the fundamental theorem of finitely generated abelian groups, if G is finitely generated it is the direct sum of a finite number of cyclic groups. If G is a torsion group, then it is the direct sum of a finite number of finite cyclic groups. Hence G is finite. (Compare with Problem 4.31, page 105.)
00
6.45.
Let
G =
2 Gj i=l
where G^
is
cyclic
group of order 2 for
1, 2,
Prove that
is
not
finitely generated.
Solution:
Every element in G is of finite order, for if g{f^ 1) G G, ff = ffi + g2+ + ffn, 9i^ G^j, (i' 6 Z), and 2g = 2gi+ + 2g - + + = 0. Thus G is a torsion group. If G were finitely generated, G would be finite by the preceding problem. But G is clearly infinite. Therefore G cannot

be finitely generated.
c.
The type
of a finitely generated abelian
group
we proved that a finitely generated abelian group is a direct sum of However, such a decomposition is not unique: first, the direct summands are not unique (see Problem 6.46 below); moreover, the number of direct summands can vary (see Problem 6.19, page 187).
In Section 6.3b
cyclic groups.
say that two decompositions are of the same kind if they have the same number of each order. For example, two decompositions of a group into the direct sum of three cyclic groups of order 4 and two cyclic groups of infinite order are said to be of the same kind. A concrete example of two decompositions of the same kind is given in
of
We
summands
Problem
6.46.
into the direct
As we remarked in Corollary 6.19, every finitely generated group can be decomposed sum of a finite number of cyclic groups of prime power or else infinite order.
Our aim
is
to prove
Theorem
6.21:
Any two decompositions of a group G into the direct sum of a ber of cyclic groups which are either of prime power order {finite order, are of the same kind.
finite
num-
1) or of in-
Proof:
We shall
infinite cyclic groups,
of fixed prime p, general case.
separate the proof into four cases: (1) both decompositions involve only (2) both decompositions involve only cyclic groups of order a power (3) both decompositions involve no infinite cyclic groups, and (4) the
Sec. 6.3]
199
Case
1.
.A.
G =
j
h
i
h
1,
. .
=Ii
h
k
where h, h for
=1,
.,k and
.,1
respectively, are infinite cyclic groups.
From Corollary 6.12, page 193, proceed as in Problem 6.52.)

Case
2.
we
conclude that
l.
(Alternatively
we may
We shall
Lemma
h G
Both decompositions involve only cyclic groups of order a power of a fixed prime p. write for any integer n, nG = {ng g &G]. If G is a group, nG is a subgroup (Problem 6.53). To prove case 2 we will need the following lemma.
\
6.22:
Let
G=
A B.
If
is
any
integer, then
nG =
nA nB.
Proof: As nADnB G such that nh g.
cAnB =
Let h
{0},
= a + b,
gp{nA,nB) nA nB. If g ^ nG, there exists a E A and b G B. Then g = nh = na + nb.
Accordingly
Corollary
Proof:
follows.
nG QnA nB Q nG and so nG = nA nB. 6.23: Let G = Ai A^. Let n be an integer. Then nG = nAi nAu We apply Lemma 6.22 to one direct summand at a time. Then

the result
Corollary
6.24:
Let
be expressed as the direct

is
sum
lî^r. Then pG
of order p*~*
expressible as the direct
of ki cyclic groups of order p\ sum of fci cyclic groups
where 2
i ^ r.
if
is cyclic
Proof: This is an immediate consequence of Corollary 6.23 and the fact that of order p\ pA is cyclic of order p^~'^. Hence the corollary follows.
We
G.
If
are
\G\
now in a position = lorp, then the
to
prove case
2.
We
proceed by induction on the order of

If the result is
result is immediate.
assumed true for
all
groups of order less than n that satisfy the conditions of case 2. then let \G\ - n. Suppose G is expressed as the direct sum of h cyclic groups of order p^ for lî^r, and also as h cyclic groups of order p^ for l^j^s. Then pG is expressible (by Corollary 6.24) as the direct sum of A;; cyclic groups of order p*"' for 2î^r on the one hand, and as the direct sum of h cyclic groups of order p*-i for 2 ^ i ^ s on the other. As \pG\ < \G\, it follows by the induction assumption that r = s and h = U for 2î^r. Now we must still prove that fci = h. But |Gj = p'^îp^)''^" (p'')''' - p^^(p^)^ (p*")'', and so h = ki. Thus we have proved both decompositions are of the same kind, as
required.
Case
3. is
expressed in two ways as the direct
sum
of a finite
number
of cyclic groups of
prime power order.
We have dealt with the case where only one prime is involved. We proceed by induction on the number of primes involved. Let p be one of the primes involved. Let Ai, .,Am be all the direct summands of order a power of p in the one decomposition, Bi, .,Bn the other direct summands involved, so that
. . .
Putting
A = Ai
.
. .
G = Ai Am Si Am and B =: Bi B,
it
J?n
follows that
G = AB.
Let Xi,
Xfc be
. .
composition, Yi,
the direct summands of order a power of p in the second de.,Yi the remaining direct summands, so that
all
G = Xi Xk Y, Yi Put X = Xx Xh,Y = Yi Yi. Then G = XY. We and B = Y.
claim that
A=X
200
ABELIAN GROUPS
Let g&A. Then g = x + y where nonzero element of Y is coprime to p. As g
6.54).
[CHAP.
xGX
is
and
yGY. Now
of order a
Hence g
By
a similar argument
Thus Ai Am-=Xi Xk and Bi B = Yi By the Yi. induction hypothesis, Ai Am and Xi Z^ on the one hand, and Bi B and Yi Yi on the other, are of the same kind. Hence the two decompositions are of the same kind and the result follows.
GX, and so AcX. B = Y.
Similarly
XcA
the order of any power of p, y = (Problem and we conclude that A = X.
Case
4.
Let
infinite
be expressed as the direct order in two ways, say
sum
of cyclic groups of prime
power order or
of
G =
/i
Fi
F =
/i
Tfc
Pi
^,
where luh are
infinite cyclic
groups and Fj,Fj are groups of prime power order.
Let T{G) be the set of all elements of finite order (see Theorem 6.9, page 189). Then T{G) is the direct sum of the direct summands of finite order in both cases (Problem
6.55).
Thus
T{G)
S,
= Fi
Fn
= Fi
Fi
same kind.
6.11,
Hence by case
Also
G/TiG) ^h
Fi
Fm and Fi Fi are of the ImÎi Tk (by Problem
page

181).
By Prob-
lem
case

6.56, /i
1.
/m

is
the direct
sum
7i
of k infinite cyclic groups.
Therefore
we have proved
that
72
/m Fi
Then k = m hy Fn and 7i ffi/a
7fc
Fi
Fi
are of the same kind.
This completes the proof of the theorem.

of cyclic groups of orders p[^,
Pi
.
If
finitely
generated group
is
the direct
. .
sum
and
s infinite cyclic
groups, where
pi,
.,Pk are primes,
P2---Pk,
ri,
.ypl" posi-
tive integers with
if Pi = Pi+i, then the ordered fc + 1-tuple {p[^, iP'^ls) is (The definition of type differs slightly from book to book. Usually it is applied only to p-groups.) By Theorem 6.21 the type of G is uniquely defined. We can now give a criterion for the isomorphism of two finitely generated abelian groups.
rj^n+i
called the type of G.
Theorem
Proof:
6.25:
If
and
G
if
and only
are two finitely generated groups, then they are isomorphic they have the same type.
if
G = Akj, Let F-Ai---Ak. If <j>: is an isomorphism, then = At, it follows that F and G have the same type. Ak<l> (Problem 6.56). As Ai<j> Conversely, if F and G have the same type they are clearly isomorphic (Theorem 6.5,
F^G
page
185).
Problems
6.46.
Let
where
G = A B where A and B are cyclic of order C and D are cyclic of order 2 and C = A and C
2.
Find
B.
C and D
such that
G=
C D
Solution:
Let
Then C + D
A = {0,a}, B = {0,6}. Put C = {0,a+ 6}. Then C = {0,a+b,b,a+b + b = a}, andso C + D = G. F

fi
is cyclic
of order
2.
Also put
Also
CnD =
fi
{0}.
Thus
D = B. G = C D.
6.47.
If the
gfj
type of
.
. .
is (/i,
fk, f)
= qr'i,
= g*'
and p^ <
and that of G is (ff i, q^, where the Pj and
g,,;
g)
where
= p[^,
f2
= pl^. ,fk = v'^,
g;
are primes, find the type oi
F G.
Solution:
We have
Pi
p^
Pk
<
9i
Qi-
Hence the type of
F"
is
(A,
..,fk.9v->Suf + g)
Sec. 6.3]

G and
201
6.48.
If F,
are finitely generated abelian groups, show that
FG = FH
H
implies that
G=
H.
Express F,
Solution:
G and
.
H
.
type of
typeof
the
FGis {Ui, type of F H is (6i,
is {/i
f^, f),
as direct sums of cyclic groups of prime power and infinite orders. If the is {h^, ...,h^; h), then the and that of G is {gi, ...,9i;g) while that of where ai,...,a^ + is fi, -'Si in some order, while ..u^ + i; f + ff) .,fk.9i,
. i
.
.,
6fc+; /
/i)
where
bj, ...,6fc +
is
fu
-yfk.hi,
-.K
in
some order.
For two abelian groups to be isomorphic we are the same and Accordingly the types of G and
require that (by Theorem 6.25) their types are the same.
G = H.
6.49.
Find up
Solution:
to
isomorphism
all
abelian groups of order 1800.
So an abelian group of order 1800 is a direct sum of a group of Observe that 1800 = 233252. order 2^, a group of order 32 and a group of order 52. The possible types of a group of order 2^ are (2^; 0), (22, 2; 0), (2, 2, 2; 0). Thus there are precisely 3 groups of order 8. The possible types of a group of order 32 are (32; 0) and (3, 3; 0), so there are 2 non-isomorphic groups of order 9. Similarly there are 2 non-isomorphic groups of order 25. Then the total number of non-isomorphic groups of order 1800 is 3 X 2 X 2 = 12.
Compare the ease with which we solve this problem with the effort required to find groups (non-abelian as well as abelian) of order 8 which we have considered in Chapter 5.
all
the
6.50.
Let p be any prime and

to the
-i
let
number
of
'fc-
ways
of writing
m be any integer. Prove that the number of groups of order p is equal w = rj + + r^ where ri,...,r^ are positive integers and
'"2
Solution:
A group of order p has all its elements of order a power of p. Hence its type will be of - r^- Since G is of order p, and .,p''fc; 0) with the form (p''i,p'2, ri-Vz. .
\G\
p'-lp''2
p'k
p'-i+'+'-fc
we
conclude that
ri+
+rj^
m.
6.51.
Prove that a cyclic group of order p", where p is a prime, is not expressible as the direct sum of nontrivial subgroups by the following two methods: (1) directly, (2) by using Theorem 6.21.
Solution:
(1)
Suppose
G=
AB
where A.
is
are nontrivial subgroups of G.

gp(g).
Clearly
6
|A]
p"-i
and
\S\^pn-i as |G|=p". Let G =

p"->o + p"-i6 = 0, ff have a contradiction.
(2)
of order less
Then g -a + h, than p"-i. But gp(g)
aGA, and = G and is
B. As p^-iff = of order p". Thus we
Since
direct
is cyclic
summands
of order p, the type of G is (p"; 0). Hence by Theorem 6.21, only one of the is nonzero, i.e. G cannot be expressed as a direct sum of more than one non-
trivial gfroup.
.52.
Prove, by considering the direct sum of cyclic groups of order 2, that if G is the direct infinite cyclic groups and also the direct sum of I infinite cyclic groups, then k = I.
Solution:
is
sum
of
fe
gp(xi) gpix^)- Let sum of k cyclic groups of order g = nxi -!-+ rfcCBfc. Then 2g = ri(2xi) +
Let
G =
H = gp{2xi,
2.
the direct
Then by Corollary 6.17, G/H .,2Xk). c 2G. Also if g ^ G, Thus \G/H\ = 2''. Clearly + n(2xk) e H, from which 2G C H. Thus H - 2G.
. .
H
I
Now
|G/2G|
by a similar argument we conclude that

2'.
if
is
the direct
sum
of
infinite cyclic groups,
Thus
fc.
6.53.
Prove that
Solution:
If
/i, fc
nG G
is
a subgroup of
where w
is
a given integer.
nG,
h-nf,
ng where f,g GG.
Hence
h-k=^ n(f-g)&
nG,
and so
nG
is
a subgroup.
202
ABELIAN GROUPS
Let
[CHAP.
6.54.
finite order,
(2) if
be an abelian group, G = Let xGX, y&Y. Prove that (1) if x and y are of then the order oi x + y is the least common multiple (1cm) of the orders of x and y; X is of infinite order, oj + j/ is of infinite order.
XY.
Solution:
(1)
Let I = km of the orders of x and y. Then l(x + y) = Ix + ly - 0. Now if = order of x + y, then m(x + y) = mx + my = Q implies mx = and my = 0. This in turn implies that the order of x divides m and the order of y divides m. Thus we have the result.
(2)
If X is of infinite order and m{x + y) expressions in direct sums, mx = my
= 0, = 0.
then mx + my = 0. But by the uniqueness of such = 0. Since x is of infinite order,
6.55.
Let G = /i T(G) be the set of

/^ Fi e
all
F
T{G)
elements of
finite order.
where each Ij Prove that

Fi
is torsion-free
and each F;
finite.
Let
F
. . . . .
Solution:
Clearly r(G) D F, F. If g e T(G), g, = j, + +j;^ + /j + are elements of /j, .,7^ respectively, and A, .,/ are elements of Fj, g is of finite order r, say, rg = rii + ri2+ + ri^ + rf^ + + rf^ =
. . .
+/^ where
ij
.,F respectively.
As
By
definition of the direct
sum,
it
follows that
=:
Wi
= W2 =
rim
iy
rfi
Since I^, ...,I are torsion-free, the result follows.
we have
= i^^
= -
=
i
rf
=
sr
Q.
Thus
G Fj
and
6.56.
If i^
= Ai
A^.
and
^-.F^G
is
an isomorphism, then
G = Aj^
A^<p.
. . . ,
Solution:
We must show that every element of G is uniquely of the form a^s + -I- a^4> where aj, belong to Ai,...,A^ respectively. Now if g & G, there exists such that f<l> = g. 4- af^<p. + Ofc and so ff = aj^ + / = Oj -IIf a^s + -H a^,!) = a[ + + al<t>, then
/GF
o._ *
But
(di
- a[)4, +
-I-
(Ofc
- afc)^ =
Since ^
is
Let
fc
= Oi oi +
+ Ofc
ttfc.
fe
belongs to Ker
0.
an isomorphism,
ft
0.
By
so that
ttj^
the uniqueness of expression of direct sums,

Oj
ttx

=
o^.
a2
0.2
ttjj
Oij
=
G
is
ttj,
a.2,
. ,
a^j
Therefore each element of
expressible in the
form
-f-
a^<p in
one and only one way.
d.
Subgroups of
finitely
generated abelian groups
The purpose of this section is to decide which groups (up to isomorphism) can appear as subgroups of finitely generated abelian groups. We begin with
Theorem
Proof:
Ml,
. .
6.26:
Let G be free abelian of rank n. of rank less than or equal to n.

6.15,
Then any subgroup

ci,
H of G is free abelian
c of
By Lemma
page 196, there exist a basis

.
.
. ,
.,u such that
H=
and integers
gp{uiCi,
.
.,VmCn).
If ui,
.,Ui are nonzero, and un-i

Mi+2
gp{UiCi,
UnCn)
gp{UiCi)
gp{UiCi)
(See Problem 6.57.)
Hence the
result.
Corollary 6.27:
Let
be a finitely generated abelian group.

generated.
Then every subgroup
of
is finitely
Proof: As is a finitely generated abelian group, it is isomorphic to some factor group of a finitely generated free abelian group G, say ^ GIN. The subgroups of GIN are of the form H/N where is a subgroup of G. By Theorem 6.26, is finitely generated and therefore so is H/N. Consequently every subgroup of A is finitely generated.
Sec. 6.3]
203
From
Theorem
this corollary
finitely
subgroups of
we see that only finitely generated abelian groups can occur as generated abelian groups.
6.28:
Let G be a finitely generated group and a subgroup of G. Let G and be expressed as direct sums of infinite cyclic groups and cyclic groups of prime power order. If the number of infinite cyclic groups in these decomand k respectively, then k^m. positions for G and H are
Proof:
Let
G =
where the
order.
I,nFi
Fn,
H = h
hFi
H
Fl
Ii,Ii are infinite cyclic groups and the Fi,Fi are cyclic groups of prime power Let T{G) and T(H) be the torsion subgroups of G and respectively, i.e. the respective sets of elements of finite order (Theorem 6.9, page 189). Then T(H) = HriT{G). Now GIT{G) I,n by Problem 6.11, page 181. Thus the rank of GIT(G) is m. (See the remarks following Theorem 6.11, page 192.) Since {H + T{G))/T{G)cG/T{G), {H + T{G))/T{G) is free abelian of rank less than m, by Theorem 6.26. But
^h
{H + T{G))IT{G) ^ H/HnT{G) ^ HIT[H)

by the subgroup isomorphism theorem (Theorem that H/T{H) h. Thus k^m.
4.23,
page
125).
It follows
as before
^h
Again let G be finitely generated and a subgroup of G. (Recall that if F is any abelian group and p a prime, Fp = {/ / e F and of order a power of p}.) Gp, as a subgroup of a finitely generated abelian group, is finitely generated (Corollary 6.27) and so is Hp. Clearly Gp D Hp. Thus we are led to inquire what groups can occur as subgroups of finitely generated p-groups. Of course a finitely generated p-group is finite (Problem 6.44).
1
We
Lemma
first
require a lemma.
6.29:
Let
in
G
is
have type
p" 1.
(p'l,
.,p''"; 0).
Then the number
of elements of order p
= 9p{Ci) and the order of Ci Proof: Let G = Ci C where each is of order p, and x = Uci + + tCn where U G Z, then p"*'^ divides the elements of order p are a subset of H, where
GG
is
p^K
If
Hence
H =
On
of order p.
p'i-^Ci
p'n-'Cn
is
the other hand every element (^ 0) of Accordingly, as \H\ = p", the

6.30:
H is of order p,&o H {0}

number
(p'Sp''^,
. .
the set of
all
elements
of elements of order p in
.,p''"'; 0).
G
n,
is
p"
1.
.
Theorem
Proof:
Let
be a group with type
Let
be any subgroup.
i
If the type of
H is {p\
.,
p\ 0), then n m
If
\G\
and
< Si
1,
Hence 1 or p, the result is trivial. the result holds for groups of order less than |Gj. Now has type (p'l, p^^, p^n; 0), and so the number of elements of order p in is, by Lemma 6.29, ., p" l. Similarly, the number of elements of order p in G is, by Lemma 6.29, p 1. Clearly p" 1 =^ p 1 and consequently n^m.
|G|.
We
proceed by induction on
assume
\G\
.
> p, and
Now
its
\pG\
<
\G\.
We
can therefore assume the result holds by induction for
pG and
subgroup pH.
At
pG
is
.
we experience a minor notational inconvenience. If for example r > 1, of type (p'î~S ...,?>''""'; 0). However, if rm 1 and rm-i > 1, pG is of type
this point
.
Therefore we need additional notation. Define m* otherwise define m* to be an integer such that r-m* > 1, but rm*+i 1. As = rm - 1, then pG is of type (p''^"', T-m' > 1 and, if m* m, rm*+i(p'i~S
.,p'>n-i-i;0).

m
^
if
>

1;
r2 ri .,p'"'"*~^ 0).
204
ABELIAN GROUPS
Similarly, let us define n*
[CHAP.
=n
if
>
1;
Sn*
paragraph above, Then by the inductive hypothesis we have

1.
>
1 but s. + i
Arguing as
in the
otherwise define n* to be an integer such that pH is of type ip^^~\ ., p*"*~* 0).

.
n*
If
Tn
^ m* -1
Thus
and
Si =^ Vi
sifor
i
^ nn*
- n,
for
l,
.,n*

n*
n,
the result follows immediately.

. .
If
Tn'+i
Sn'+i,
then s.+i
1.
Hence
.,
s.
l,...,n.
With Theorems 6.28 and 6.30 it is easy to determine, knowing the type of a given finitely generated abelian group G, the possible types of subgroups of G. (See Problem 6.60.)
It can be shown (Problems 6.62-65) that every factor group of a finite abelian group G isomorphic to a subgroup of G. Therefore we know the types of homomorphic images of finite abelian groups.
is
Problems
6.57.
Let
G=
AB
and
let
C,D
be subgroups of
A,B
(This can obviously be generalized to the direct

Solution:
sum
{0}.
respectively. Show that of any number of groups.)
C + D = C D.
As
6.58.
{0}=AnB3CnD,
(pî,
. .
we have
CnD =
Thus C +
D = C@D.
Let
have type
-.P^"; s).
Pi
Suppose that for some
= Pi+i= =Pj,
is (,p^\
. .
Pi-i^Pi
.,p''J; 0).
and
Pjy^Pjî
Put p
Pi.
Show
that the type of Gp
Solution:
Decompose G into the direct sum of infinite cyclic groups and groups of prime power order. Clearly G R where A^ is of order p''k for i k j, and R is the direct sum of the cyclic groups which are not of order a power of this prime p in the given decomposition of G. Then Gp 3 Aj Aj. On the other hand, as any nonzero element of finite order from R is of order coprime to that of p, Gp C Aj Aj. Thus Gp = Aj Aj, and the result
Ai@
Aj
follows.
6.59.
Let
G
1
be of type (p^,
..p'"'; m).
i
Show
v
that
u.
has subgroups of type
(p'l,
..p*"; v)
where n
and
S(
rj
for
~ n and
Solution:
Let
tion of
infinite cyclic groups.
G = AiQ Am ffi ?! ffi Now each Aj has

6.57,
where Aj
is cyclic
',
of order p^, and

1
/i,
...,/ are
a subgroup Bf of order p
i
/i
n.
By
repeated applica-
Problem
Bi+ +B + Ii+
This
is
+I
(p'l,
. .
Bi
ffi
then a subgroup of
of type
.,p*"; v), as required.
6.60.
Let
(a)
be of type
(3^, 3^, 5^, l^; 1).
Determine whether
(c)
has a subgroup
(d)
(3, 3, 7; 1).
of type
(3, 3, 72, 7; 1),
(6)
(3, 3, 5, 7; 2),
(33, 3*, 5^, 73; 0),
Solution:
(a)
No, as then G7
tradiction to
is
of type (73; 0) whereas

6.30.
H^
is
of type
(7^, 7; 0)
by Problem
6.58.
This
is
a con-
Theorem
(6)
(c)
(d)
No. A direct contradiction to Theorem 6.28. No. Compare G3 and H3 as in (a). Yes.
6.61.
Give an infinite number of examples of an abelian p-group that contains exactly p order p.
Solution:
If
+1
subgroups of
is
a group with p
+1
subgroups of order
p,
(the identity is
common
to all) of order p.
Hence
each of them contributes p 1 distinct elements in all there are (p l)(p + 1) = p^ 1 elements
of order p.
Sec. 6.4]
DIVISIBLE GROUPS
205
By Lemma 6.29 a p-group with jy^ l distinct elements of order p contains p^ summands which .,p2, be integers. Then any are cyclic groups of order a power of p. Let 1 - rj, where i = 1, .,p'v;u) where m is a nonnegative integer, has exactly p^ 1 elements of order type (p*^!, group of p and thus exactly p + 1 subgroups of order p.
. . . .
6.62.
Let Gf be a p-group. Suppose G = gp(a) of order less than or equal to the order of
Solution:
If
a;
a.
B.
Prove that
G =
gp(a +h)
where
GB
is
gp(a
+b)nB,
then
ra
b^
rb.
a.
Since
gp(a)nB
{0},
order of
Consequently rb
0.
= r{a+b) = bi where b^G B and r is an integer. Thus ra = 0. Then r is divisible by the power of p which is the = 0. Clearly gp{(i-^ b) + B = G and the Hence 6i = and
x
a;
result follows.
6.63.
Let G be a finite p-group. Prove that a cyclic direct summand of G.

Solution:
ii
G G and
is
of order p, g appears as an element of
G is the direct sum of cyclic groups, say G = S'p(ci) gp(Cn)follows immediately. Otherwise, without loss of generality, suppose that
If
0,
the result
J,
where (rj,p) = 1, rjp'îCj ^ and Then g = p"'i(cj + d) where d G gp(c2

is less
= np icj + + r^p "Cm w. Put riCi = w^ W2
cf.
c).
As g
is
of order
p and
Clearly gp(.c{) = gp(ci). pîcj 0, the order of d
than or equal to the order of
Cj.
By Problem
gp(c)
6.62,
on putting e
. . .
c[
+ d, we
obtain
G =
Since g
gp(c2,
c)
G gp (c),
the result follows.
6.64.
Let G be a finite p-group. Let group of G.

Solution:
N = gp{g)
be of order
p.
Prove that
G/N
is
isomorphic to a sub-
By Problem
6.16,
6.63,
G =
gp(c)
flrp(pc)
page
197).
But clearly
gp{c)IN.
where jr G ^p(c). Then G/N Thus gp(pc, B) = GIN.
{gp(c)/N)
B
is
(Lemma
6.65.
Let
to
be a finite group.
Prove by induction on
\G\
that if iV
is
a subgroup of G,
G/N
isomorphic
a subgroup of G.
Solution:
groups of order less than r. Let |G| = r and let AT be a subthere is nothing to prove. If is of order a prime p, where Gp is the p-component of G, and iV C Gp. Then G/AT = (Gp/iV) G= by Lemma 6.16. Now Gp/iV s H, a subgroup of Gp, by the preceding problem. Hence G/N = E and E i& a, subgroup of G. If N is not of order a prime, there is an element no of of order a prime p by Proposition 5.9, page 137. Let N^ = gp(na). Then (G/Nq)/(N/No) = G/N. As ' G, since otherwise = Nq which is not |G/iV(,| < \G\, G/Nf, has a subgroup H/iVo = G/N. true. Thus \H\ < \G\ and by induction it has a subgroup K such that K = H/N^ s G/iV. The
Assume
the result
li
is
true for
all
group of G.
N=
{0},
G/N = G and
GpE
H@
result follows.
6.4
a.
DIVISIBLE GROUPS
p-Prufer groups. Divisible subgroups
group G is said to be divisible if for each integer n= and each element g there exists h such that nh = g. Both the additive group of rationals and p-Priifer groups are divisible in this sense (Problem 6.66 below).
&G
any
&G
If the
groups
Gi, i
Gi,
G I,
then
are divisible, then
Gî
is
divisible.
For
hi,
if
. .
n=0
is
integer
and g G 2^
. .
gi+
+gk,
say.
So there exist
.,hk such that
nhi
gi,
.,
nhk
gk.
Then
n{hi H
+hk)
- gi+
gk
206
ABELIAN GROUPS
sums of p-Prufer groups and copies
[CHAP.
It follows that direct tionals are also divisible.
We
shovs^
of the additive group of rathat in fact this exhausts all divisible groups.
To prove
First
this wê need several facts.
we remark
divisible
that a
homomorphic image of a
divisible
pose
is
positive integer.
a subgroup of G. Let g + Then there exists g' G G such that ng'

is
and
group
g.
H G G/H
=
For supG) and let w be a Accordingly n{g' + H) =

is divisible.
(g
+ H, and
so
H ^ G/K
K
Secondly (Problem 6.11, page 181)
G/H is divisible. we remark that if G =

is
/?
e and G is divisible, so also are and K, since a homomorphic image of a divisible group. Similarly
is divisible.
Next we need the following theorem which
classifies p-Priifer
groups.
Theorem
6.31
(Main Theorem on p-Priifer Groups): Let p be a prime. Let G be a group which is the union of an ascending sequence of subgroups Ci C G2 C where Cr is cyclic of order p"" for r = 1,2 Then G is isomorphic to the

p-Priif er group.
Proof:
We may suppose that

Define
is
6.67 for the details).
must prove that The snag is this:

It
an
= gp(Cr) and that pcr+i = Cr for r = 1, 2, .. (see Problem G -* (Q/Z)^ by (mcr)d - m/p" + Z for all integers m. We isomorphism. We are however not even certain that 5 is a mapping.
Cr
.
(9 :
= mcr and
if also
nc^,
is
gO
m/p"-
+Z
or
is
g9
n/p'
+ Zl
We will
show that m/p' + Z = n/p' + Z, thus proving that Q is uniquely defined. Assume without loss of generality that r ^ s. It follows that p^'-'Cr = c^. Then
mcr
np'-^Cr
from which
{m-np'-'')Cr
As
Cr is
of order
p"",
m - np"-^ = kp"
We
is
6/
which m/p^-np-' + k. mapping.
for some integer k. therefore conclude that m/p^
Thus
m = np"-" + kp""
n/p'
from
61
+Z -
+ Z.
Thus
is
Next we show that

r,
so that g
{g
scr,
tcr,
with s,tGZ.
{s
homomorphism. If g,hG G, then g,hGCr for some integer Then
+ h)B =
({s
+ t)Cr)e =
+ t)lp' + Z =
{sip-
+ Z) +
{t/p^
+ Z) =
g9
he
Finally
is
one-to-one as
Ker e
= {g\ge^Z} {sCr s an integer, r = {sCr sip"" G Z} =

I
\
a positive integer and

{sCr
I
(scr)6
=
{0}
sip''
+ Z = Z}
s divisible
by
2?''}
Thus the
result follows.
In the future
we
will call
any group isomorphic to {Q/Z)p a p-Priifer group.
The following
Itself.
result is not only the
main
tool in Section 6.4b, but is also of interest in
Theorem
6.32:
Let
of
G contain a divisible subgroup D. Then there G such that G = D K, i.e. a divisible subgroup
exists a
is
subgroup
a direct summand.
Proof: We accomplish this proof by Zorn's lemma. Let "P be the collection of all subgroups L of G such that LnD = {0}. (Our idea is to pick a maximal subgroup K which meets D in {0}. Then D + and we need only show that D + K = G which will turn out to be true because of the maximality of K.) Can we apply Zorn's lemma to <?? Suppose {Li\iG 1} is a chain in <P. Is U Lj in "?? We require
K=^DK
Sec. 6.4]
DIVISIBLE GROUPS
207
(i)
Dn\JÛ=
U
Li
is
(i) is
{0}.
(ii)
a subgroup of G.
Part
true because
(ii)
Dn
U^Li
that
if
- {0}
implies
Li,
DnL,^
then
{0}
for some Lj.
To prove
implies g sume that
we must show
and h G Lk. Either D Lj. Then since Lk is Lfc

L,
g,h^ U Lk D Lj or
g-hG ULu Now

Therefore
g,hG U
Li
Lj
D L^,
so without loss of generality as-
a subgroup of G,
g-he Lk.
g-hG
(0).
U^Li
and
(ii)
holds.
So
"?
has a maximal element, say K, which
satisfies
Dniir=
Thus
Suppose now that D + Kj^G. Then GI{DK) is nonzero. We prove first that GI{D X) is a torsion group. Suppose the contrary. Then we can find a; G G such that gp{x + {D@ K)) is of infinite order in G/{D K). Now x^K. If we put Ki = gv{K, x), it consists of all elements of the form nx + fc, where n is an integer and k is an element of But KC^D^{0]. is the only possibility. K. If nx + kGD, then nx:{K + D). So to = follows also. Therefore Dr\Ki- {0}. This contradicts the maximality of K. So fc = Thus we have proved that GI{K D) is a torsion group. Let a; G G, a; g K D. Then gp{x + (K D)) is a subgroup of GI{K D) of finite order. D, but Suppose that x + {KD) is of order w. It follows then that wx for < r < w. Suppose wx = k + d. Since D is divisible, we can find diG D so that wdi d. Put a;i = a; di. Then wa;i = wx wdi = k + d d = k. Notice that Xi K but wxi G K. Put Ki = fl'p(a;i, JRT). Then
GK
rx^KD
Ki
{ra;i
A;
I
0,1,
..,w-l, andall
. .
fc
Gii:}
Weclaimthat KinD=^{0}.
+ kGD, r G {0,1, .,w-l}, and kGK, then D K = rxi + {DK) = r{xt + {D K)) G D and thus k-d. So Since xi + (DK) = x + (D K), we must have r = 0. Therefore Kir\D {0}. But K is maximal. This contradiction shows that our original
For
if
rxi
A;
assumption,
i.e.
G'
DK,
is false,
thus proving the theorem.
Problems 6.66. Use a proof

Solution :
different
from that of Problem

QIZ.
6.36,
page 192, to prove that a p-Priifer group
is divisible.
As Q
is divisible, so is
But as QIZ
pen
(Q/^)o,
each (QIZ)^
is itself divisible,
since
every direct
6.67.
summand
of a divisible group is divisible.
Let
and pCf+i
Solution:
be as defined in Theorem 6.31. = c,., r = 1, 2

induction that
Prove that we can choose elements
c^
such that
C,
gp{c^
pc,.+ i = c, for r = l, ...,ti Ci, ...,c have been chosen with Let C+i - gp(c). Then gp{pc) is a cyclic group of order p" and, Ci since cyclic groups have only one subgroup of any given order, gp (pc) = C. Hence r{pc) c for some integer r. As c is of order p", r and p are coprime. Thus gpirc) = C+i. Now put ., as required. c+i = re. Then pc + i = c and it is possible to choose the elements c^jC^,
Assume by
and
gp{.c^, i
1,
..,n.
6.68.
p-group G is a p-Prufer group if and only if it has the following two properties: every proper subgroup of G is cyclic, (1) (Hard.) there is a cyclic subgroup of G of order p* for every i = 1, 2, (2)
. .
Solution:
First
Let Ci
we shall prove that if G satisfies (1) and c C2 C be a sequence of subgroups
(2), it is
of
G where
a p-Priifer group. each Cj is a cyclic group of order
p'.
If the sequence is infinite,
G=D
two
cyclic p-groups,
by Theorem 6.32. However, if K # {0}, and hence is not cyclic, contrary
D= U i=l
00
Ci is a p-Priifer group (Theorem 6.31).
Hence
it is divisible
and
of
is
has a subgroup which
to the hypothesis.
is the direct Accordingly, G
sum
p-Priifer group.
208
ABELIAN GROUPS

[CHAP.
Suppose now that Cx C C2 c c C is a sequence of subgroups, each Cj cyclic of order p, and there exists no subgroup C+i2C where C+i is of order p"+i. Let C = srp(a). We know that there exists a subgroup B = gp(b) of order p" + i. Consider gp(a,h). As a finitely generated abelian group, it is the direct sum of cyclic groups. If it is the direct sum of two or more cyclic groups, it is not cyclic (Theorem 6.21). Thus gp{a,b) is cyclic, and as G is a p-group, gp(a,b) is cyclic of order p*" where m re + l. But then gp{a,b) contains a cyclic subgroup of order pn+i
containing C, contrary to the hypothesis.
The
result follows.
Next we note that the

subgroup of {Q/Z)p. li. integer between 1 and p'
l/p""
+Z
Thus
H ior r^ H is cyclic as required.

&:
As for condition (1), let be a (2). then let xGH, + Z. Let x - m/p'' + Z where is an 1. Clearly gp(x) = gp{l/p'' + Z). If there is no integer n for which = <,QlZ\. If there exists such an integer n, n, then Q gp{\lp + Z).
p-Priifer group satisfies condition
{Z},
H =
x^
H m
6.69.
Show
all
that if is the subgroup of the multiplicative group of the complex = Q/Z. the wth roots of unity, then
numbers consisting of
Solution:
11= t^p by Theorem 6.10, page 190. Now ?/ is the union of cyclic groups of order p', p n namely J7p is the union of Cj = gp ({x k is a p*th root of unity}). But Cj is cyclic of order p'. Then by Theorem 6.31, Uj, = (Q/Z)^. Thus, as QIZ = (QIZ)^ (by Theorem 6.10), QIZ a C/ by "^ " Theorem 6.5, page 185.
|
6.70.
Show that the additive group of rationals groups.

Solution:
is
the union of an ascending sequence of infinite cyclic
Let Q
00
ffp(l/?i!).
Then Q+i3Q and
n=l
KJ
Qn
= Q
the additive group of rationals.
6.71.
Show
that
if
is
a p-Prufer group and H,
are subgroups of G, then either
H^ K
or
Kd H.
Solution:
If one ot H or K is G, the result is true. Assume both H and K are proper subgroups. Then and K are both finite cyclic groups. Suppose p"" = \H\ \K\ = p. Then H conby Problem 6.68, tains a subgroup of order \K\, say H^. Now both H^ and K are contained in some cyclic subgroup of G, as G is the union of cyclic subgroups. But then it follows that H^ K, as there is one and only one subgroup of order p' in a cyclic p-group of order exceeding p^. Therefore H D K.
6.72.
Let
for
G
I
be an ascending union of infinite cyclic groups Cj such that Cj = gp{ci) and (i+ l)ci+i (Hard.) 1, 2, ... . Prove that G is isomorphic to the additive group of rationals.
Cj,
Solution:
Let
Q
z/il
denote the additive group of rationals and let

00
Qj
fl'p(l/i!), i
1, 2,
Clearly,
Qj+iDQj and
(Cj)ff
i=l
Qi
z
=
&
QZ.
where
ZiCj
We shall prove that We must prove that e

!, Z2
below defines a mapping of

is
to
Q.
Define
uniquely defined.
i
Suppose
Zi(jl/il)Cj
Z2Cj,
and
i,i
integers.
If
j,
then
Cj
(j!/i!)cj.
Hence
z^c^
Z2Cj.
Since
Cj is infinite cyclic,
Zi(}\/il)Cj
=
it
z^Cj
implies that
Zi{j\/H)
=
.
Z2.
To prove that
e is well defined,
we must show
that
Zi/i\
Z2/JI,
i.e.
Zi(jl/i\)
00
what we have
Is 9
just shown. Since
Cjff
Qi
= gp (1/il),
follows that Ge
an onto mapping.
a homomorphism? Let f,g
U Qj '~^
= =
But
this is
9 is
Q.
Hence
Hence /
ZjCj,
=
(f
z^Ci,
say.
(Z1
& G. We may + g = (z^ + 22)^11
as well suppose that f,gE.Ci for some integer
i.
Then
+ g)8 -
+ Z2)
and
fe
g9
+
^1
^2
-77
(Z1
+ Z2)
1
Thus
is
9 is
a homomorphism.
Finally, to
show that
such that fe
0.
= 0. We Hence / = and
e is an isomorphism, it is sufiicient to show that Ker e = {0}. Suppose / have that / = zcj for some integers z and i. Then fe = z/i\ =0 only if
the result follows.
Sec. 6.4]
DIVISIBLE GROUPS
209
b.
Decomposition theorem for divisible groups
The
results of Section 6.4a enable us to deduce the following decomposition
theorem
for divisible groups.
Theorem
Proof:
6.33:
group is the direct sum of p-Priifer groups and copies of the additive group of rationals.
divisible
Let G be divisible and let T be the torsion subgroup of G. Now for any integer g G such that ng = t. Since t is of finite n and element t G T, there exists an element order, so is g, and hence g GT. Thus T is itself divisible. A divisible subgroup is a direct F)/T; hence F is summand (Theorem 6.32); so G = T F. Since Tr\{F]^0, summand, F is itself torsion-free by Theorem 6.9, page 189. Moreover, as F is a direct
flr
F^{T
divisible.
(a)
We now
consider
and T separately.
of rationals.
F.
We show that F is
end
i,
let
a direct sum of copies of the additive group S be a maximal independent set (Theorem 6.13, page 194).
To
this
subgroup Gs of F. Let n,, = s. For a given there exists by the divisibility of F an element n+i,, G F such that {i + l)ri+i.s = n.s. We put Cs = gp{ri.s z = 1,2, .). It follows then from Problem 6.72 that Cs is isomorphic to the additive group of rational numbers. Note that if x GCs, x - 0, then there is a nonzero multiple of x which is also a multiple of s, as s = 0. (This is true for
shall define a
.
|
sGS
For each s G S we and positive integer
any two nonzero rational numbers.)
We claim that F is
To prove
this,
actually the direct

. .
.
+ c =
0,
suppose that Si, S2, where Cj G Cs^, Cj-0
s
1,
sum of these subgroups Cs as s ranges over S. are distinct elements of S and that C1 + C2+

(j
.,
n).
As we remarked
multiple of
kcj
Sj.
Hence there
Ij
IjSj, where C1 + C2+ +Cn IjSj, that iisi +

contradiction
it
above, there exists a nonzero multiple fej of each Cj which is then a kn, such that exists a nonzero integer k, namely ki is a consequence of As fcci + + kcn = is a nonzero integer.

we find therefore, on substituting for the elements kCj the elements From this +lnSn = 0. But the elements Si, Sn are independent. follows that the Cs generate the direct sum
0,
.
. .
C =
gp{Cs\
sGS) ^
ses
is divisible.
Cs
Hence C is a direct summand Clearly let d G D (d ^ 0). the set Su{d} is definitely larger than S since d ^ S {d does not even lie in C) and SU{d} is independent. This is a contradiction as S is a maximal independent set. So
But C is divisible since each summand Cs (Theorem 6.32), i.e. F = CD, say. If
D^
{0},
then
F=C
(6)
is
a direct
sum
of copies of the rationals.
T.
First of
all
r =
pen
^p
by Theorem
6.10.
Since
is
divisible, so also
are the
summands Tp. It is sufficient then to assume that r is a direct sum of p-Priifer groups.
Let
that
is
a divisible p-group and to prove
S be a maximal independent
the set of elements of T of order at most p. P is clearly a subgroup. Let define For each s subset of P (Theorem 6.13). inductively as follows: (a) pci,s = s, (5) pcj+i.j = Cj.^ for i = 1,2, Ci.s,<^2.s, Since gp{ci,s) is This is possible since T is divisible. Clearly gp{Ci_^)Ggp{c2.s) C cyclic of order p*+S Cs = gp{ci,s,C2.s, ) is an ascending union of cyclic groups of Accordingly by Theorem 6.31, Cs is a p-Priifer order p\ one for each i = 1,2, ...
P be
GS
group. Then of T, i.e. T
C =
is
gp{Cs
It
& S)
is divisible.
By Theorem
6.32,
-CD.
D -
(0),
there
is
ScC,SU{d}
an independent subset of by the maximality of S. Thus T = C.
an element of larger then S, and so
dGD
is
a direct summand Since of order p.
we must have
D= {0}
210
ABELIAN GROUPS
It
[CHAP.
remains only to prove that

ci
C =
S and
above,
se
C2

Cs.
If s^.s^,
i
. .
.,s are distinct
elements of
(a)
we
+ + where d G Cs, for + arrive at a dependent relation between Si,

.
=
. ,
(Problem
l,..., n, then, similarly to =c s, unless ci = ca =

=0
6.78).
The proof
Problems
6.73.
of the
theorem
is
complete.
Prove that (a) every free abelian group is isomorphic to a subgroup of a divisible abelian group, and (6) every abelian group is isomorphic to a subgroup of a divisible group.
Solution
(a)
free abelian group i^ is a direct
sum
of infinite cyclic groups
(i
G /): F =
and
*
2
let
Cj.
Now
be
we
choose one copy of the rationals Qj for each
iei. Let
\
K = 2 Q ie
/
d*
chosen in each Q^
Qi IS divisible.
(h)
F is
clearly isomorphic to
gp({di
/}).
But
The
2
^
^
Qi
is divisible
since each
result follows.
If
is
any group,
in (a)
G = F/N
D/N
for some free abelian group
F and
some subgroup N.
that there exists a divisible group D containing F. Hence D contains as a subgroup F/N, i.e. G. Now D/N is divisible since a homomorphic image of a divisible group is divisible. Thus the result follows.
contains N, and so
Now
we have proved
6.74.
of H.
Suppose G has the property that Prove that G is divisible.
if
is
any group such that
H^G,
then
is
a direct summand
Solution:
is
summand
a subgroup of some divisible group D by Problem 6.73. Thus of a divisible group is divisible. Therefore G is divisible.
D = GT.
But every
direct
6.75.
Let G be an infinite group whose proper subgroups are all finite. Prove that G is a p-Priifer group by using the theorem which states: if G is a group such that for some integer n = 0, nG = {0}, then G is a direct sum of cyclic groups (of finite order). (This theorem is not proved in this text.)
Solution:
nG for all positive integers n. If nG = G for all such n, then G is the direct sum of p-Priifer groups and copies of the additive group of rationals. As all the proper subgroups of G are finite and the additive group of rationals has an infinite cyclic group as a proper subgroup, only p-Priifer groups are involved. Since each p-Priifer group is infinite, G must in fact be a p-Priifer group.
Consider the subgroups
divisible,
and so
is
If on the other
hand nG
{0}.
m, say, and
so
mnG =
direct sum of finite of cyclic groups Cj. contradiction proves the result.
a proper subgroup of G for some n, then nG is a finite group of order Using the theorem quoted in the statement of the problem, G is the cyclic groups. As G is infinite, it must be the direct sum of an infinite number But then the subgroup generated by all but one of the must be infinite. This
is
6.76.
(a)
Let G be any group and Prove that S is divisible.
let
be the subgroup generated by
all
the divisible subgroups of
(6)
Prove that G is the direct sum of a divisible group and a group which has no other than the identity subgroup {0}.
divisible
subgroups
Solution:
(a)
Let
i
seS. Then
if
=
fci
h^
Thus
zÔ
+ hz-^
fca
is
any
integer, there exist
where each Aj belongs to a divisible subgroup Hj of G fcj G j such that zki = hi. As S 3 ffj for each
and
z(ki
/,
fc
G S
+ A;) =
Hence S
is divisible.
Sec. 6.4]
DIVISIBLE GROUPS
211
(6)
Since
is
divisible,
(Theorem
6.32) to find that
we can apply our direct summand theorem for divisible subgroups G = S T. As T contains no divisible subgroup other than {0},
the result follows.
6.77.
(a)
(6)
Prove that the additive group of rationals

Let
is
(?
has a proper subgroup which

is
is
not free abelian.
be a torsion-free group, every proper subgroup of which
free abelian.
Prove that
free abelian.
Solution:
(a)
Consider the subgroup

tionals.
generated by 1/2, 1/4,1/8, ..., 1/2*, .. in the additive group of rais free abelian, it Q is of rank 1 (Problem 6.38, page 194). So if = gp{z/2^) for some inteis not infinite cyclic; for if it were, must be infinite cyclic. But is and 1/2*+ 1 fifp(z/2*). We have only to prove that gers 2 and i. But then 1/2'+ 1 S
.
is
of rank 1 as
not G.
(b)
This
is
obvious since
1/3
H.
Suppose that
nG = G for all positive integers n. Then G is divisible and is the direct sum of copies of the additive group of rationals. (As G is torsion-free, no p-Prufer group is involved.) 0, nG = G, and nG is But by (a) above, G will have a non-free subgroup. Thus for some w
free abelian.
Now
torsion-free.
-^ ng Hence e
is
a homomorphism of
onto nG.
Ker e
{g\ ng
0}
{0}
as
is
is
an isomorphism, and so
is
free abelian.
6.78.
Prove the result stated at the end of the proof of Theorem 6.33, = c = 0. then Cj = C2 =

i.e.
prove that
if
Cj
+ c =
0,
Solution:
Ci is of
the contrary. highest order. Say Thus we obtain
Assume
As
pi-i
Cj is
the order of of order p'.

(cj
Cj,
Then
c is immaterial, we may assume that Ci =/= p'-ic; = mjSj where m^ is an integer, i = 1,2,
and
.
. .
,n.
+ c) =
m^Si
m^Si
-!-+ ms
an independent
set.
Since
yriyS^ v^ 0,
we have
a contradiction to the fact that
is
This chapter was mainly concerned with the structure of divisible and finitely generated abelian groups.
Direct sums of groups were discussed. Given a family Aj (i G 7) of groups, there is always a group which is the direct sum of groups isomorphic to each of the groups Ai. Any homomorphism of the direct summands of a group extends to a homomorphism of the whole group. Prom this it follows that if two groups are direct sums of isomorphic subgroups, they are isomorphic. Direct sums of infinite cyclic groups are all the free abelian groups. An important fact is that every abelian group is a homomorphic image of a free
abelian group.
The
torsion group T{G) of a group
G was
defined,
and
it
was shovra that G/T{G)
is tor-
sion-free.
It was proved that if G is a torsion group, it is the direct sum This led to the definition of the p-Prufer group as the p-component of QIZ.
of its p-components.
An
The
application of Zorn's lemma proves every group has a maximal independent subset. rank of a group was defined and proved an invariant of the group by the Steinitz
exchange theorem.
In the fundamental theorem of abelian groups, finitely generated abelian groups were shown to be expressible as the direct sum of cyclic groups. Two finitely generated abelian groups were shown to be isomorphic if and only if they have the same type. Finally, the type of a subgroup of a group was shown to be, roughly speaking, "less than" the type of the group.
212
ABELIAN GROUPS
[CHAP.
Divisible groups were discussed. Any group which is the union of cyclic groups of order a power of p turns out to be a p-Prufer group. Any divisible subgroup of a group is also a direct summand. This led to the proof that every divisible group is the direct sum of isomorphic copies of the additive group of rationals and p-Priifer groups.
DIRECT SUMS AND FREE ABELIAN GROUPS
6.79.
If the
mapping o
a-i,
a&G,
is
an automorphism of the group G, prove
is
abelian.
6.80.
Suppose that
G
is
Show
6.79.)
that
is a finite group, a e aut (G), a is of order 2, and ga^ g for all g {-\)& G. an abelian group. {Hint: First prove G = {g-Hga) g G G} and then use Problem
\
6.81.
Denote the set of all homomorphism of an abelian group If 0, * e Horn (G, H), we define ^ + * by
g{4>
into
an abelian group
by Hom(G,H).
+ ^) =
g<f>
g'^
(1) is
(l)
for
all
G.
Show
that
Kom (G,H)
Z
is
with the operation defined by
an abelian group.
6.82.
If
is
an abelian group and
the group of integers under addition, prove that
Horn
(Z,
A)
A.
6.83.
Prove that the group of rationals under addition

If
is
not the direct
sum
of cyclic groups.
6.84.
sum
6.85.
is the direct sum of cyclic subgroups, show that a factor group of G is not necessarily a direct of cyclic subgroups. {Hint: Use the preceding problem and free abelian groups.)
Let
N he
a normal subgroup of G. Prove that
if
G/N
is
free abelian,
N is
a direct
summand.
TORSION GROUP AND RANK

6.86.
If
is
finite
group, show that aut (G)
Tl aut (Gp),
where
jr
is
the set of
all
primes.
6.87.
Prove the
first
Sylow theorem, page 130, for abelian groups using the p-components.
an abelian group
6.88.
Let
we may use
Let "^
+ Pg = pg ^^ 2Trm/n radians where m, n are integers, n > 0}. Prove that (a) <3i is a subgroup {pg of G; (6) every element of has finite order; (c) if %j, denotes the p component of for any prime 9 = 2irm/p'" radians where p, then "^p = {pg and r are integers} and %p = the p-Priifer group.
pg
denote the group of rotations of the plane (see Section 3.4c, page 68). As G additive notation. Thus the rotation p^ followed by rotation pg is
is
=
1
6.89.
Let
a, 6 be elements of an abelian group. Let the order of a plus the order of h be n. induction on n that a 4- 6 is of order the least common multiple of the orders of a and 6.
Prove by
6.90.
G be an abelian group. Suppose every element of G is of order less than some fixed integer n and there are elements of order n in G. Prove that the elements of order n generate G.
Let
CHAP.
6]
213
FUNDAMENTAL THEOREM FOR FINITELY GENERATED ABELIAN GROUPS

6.91.
/, is
C where / ffi Cj Prove that if G is a finitely generated abelian group, then G = /j .,n), Cj is a finite cyclic group of order Dj (i = 1, an infinite cyclic group (j = 1, ., w), n-1. {Hint: Use the fundamental theorem and then first look and Vj divides Vj+i for t = 1 at the highest power of each different prime in the decomposition.)

.92.
Prove that the automorphism group of a finitely generated abelian group there is at most one infinite cyclic summand in a cyclic decomposition of G.
is finite if
and only
if
6.93.
Find the type of the additive group of integers modulo
m for
G
any integer
>
0.
6.94.
Let
be a non-cyclic
p.
finite abelian
group.
Show
that
has a subgroup of type
(p, p; 0)
for some
prime
6.95.
Prove that the automorphism group of a finite non-cyclic abelian group G is non-abelian. {Hint: Use Problem 6.91 to find suitable elements a,b ^ G such that the order of a divides the order of 6. Then look at the mappings a^: a'b' -> a+'5t; aj a'6* - o'6-* and as: a6* - atft" where s and t
^
are integers.)
6.96.
Let
be a
in
ments
finite abelian group. Suppose that for each divisor of order d. Prove that G is cyclic.
d of
there are at most d ele-
6.97.
Let G be a finitely generated abelian group. Prove by induction on the number of generators of that every subgroup of G is finitely generated.
DIVISIBLE GROUPS
6.98.
Show
If
that a divisible abelian group has no subgroup of
finite index.
6.99.
G is a nondivisible abelian group, then lowing theorem (not proved in this book):
direct
G has a subgroup of prime index. {Hint: Use the folAn abelian group G for which nG = {0}, n # 0, is the
sum
of cyclic groups.)
6.100.
Prove that the additive group of the real numbers the additive group of rationals.
is
the direct
sum
of isomorphic copies of
6.101.
(a)
Let
G = Hom(A,B)
the direct
(see
Problem
6.81)
where
is
a torsion-free divisible group.

if
Prove that
G
(6)
is
sum
of copies of the additive group of rationals.
Let
G be as in (a) but with of p-Priifer groups.
a divisible p-group.
Prove that
is finite,
is
the direct
sum
6.102.
subgroup H of an abelian group A is a pure subgroup of A if whenever na = h & H for some o e A, then there is an h' such that nh' h. Prove that (a) a direct summand of an abelian group is a pure subgrroup, and (6) the torsion subgroup of an abelian group is a pure subgroup.
&H
6.103.
Prove that
all
the subgroups of a group in which every element has square-free order
is
pure.
6.104.
Let
H be a
G
gS
such that g
pure subgroup of an abelian group G. Prove that if g + G/A, there = g + A and the order of g is equal to the order of ff + A.
AG
is
an element
+A
chapter 7
Permutational Representations
There are three main divisions of this chapter. In the first we generalize Cayley's theorem, that every group is isomorphic to a permutation group. As consequences of this generalization we prove the following theorems for G, a group generated by a finite number of elements: (1) A subgroup of finite index in G is itself finitely generated. (2) The number of subgroups of fixed finite index in G is finite. (3) If the subgroups of finite index of G intersect in the identity, then every homomorphism of G onto G is an automorphism.
The second main division of this chapter appears in Section 7.7. We call a group G an extension of _a group H hy a. group K if there is a normal subgroup H of G such that G/H = K and H = H. We examine G to see how it is built up from H and K. The most general case is complicated and we restrict ourselves to a special extension called "the
splitting extension."
Reversing our analysis, we are able to build a group G that is the splitting extension of a given group hy a given group K. A particular example of a splitting extension is the direct product, used in Chapter 5.
Our
use
it
third
division,
its
which begins
group into one of

to
abelian subgroups.
show that a group G
Section 7.8, defines a homomorphism of a This homomorphism is called the transfer. We with center of finite index has finite derived group.
in
7.1
a.
CAYLEY'S THEOREM
Another proof of Cayley's theorem
We saw in Chapter 2 that every groupoid is isomorphic to a groupoid of mappings. In particular, every group is isomorphic to a group of permutations. The consequences of this theorem are important. We repeat the proof here for the case of groups alone.
Theorem
7.1
(Cayley) :
Every group
is
isomorphic to a group of permutations.
Proof: Let G be a group. Let p be the the following mapping of G into G:
mapping which assigns

x
to each element
fl'
in
X-* xg
for each
definition, gp
GG
Thus the image of
flr
in
under
p is,
by
gp:
where
x^ xg (xGG)
p defines
The definition of p is unambiguous. To prove that a subgroup of Sg, we have to check that:
(i)
an isomorphism of
onto
gp
is
a permutation of
for every g
if
i.e.
&G;
gp hp;
(ii) (iii)
p is p is
a homomorphism,
also
i.e.
g,hG. G, then (gh)p

p is one-to-one.
an isomorphism,
214
Sec. 7.1]
CAYLEY'S THEOREM
deal first with
215
is a one-to-one mapping of G onto G. g~^ on the right, we find x = y. Next If x{gp) = viffp), then xg = yg. So, multiplying by we prove gp is a mapping of G onto G. Suppose a; G; then {xg~^){gp) = {xg~^)g = x and
We
(i).
Thus we must prove gp
so gp
is
onto.
Secondly
we prove
(ii).
For x G G,
x{{gh)p)
x{gh)
{xg)h
(x{gp)){hp)
x(gphp)
Since {gh)p and gphp have precisely the same effect on every element of G, {gh)p the definition of equality of mappings).
It
=
if
gphp (by
1 is the
identity element of G,
remains to prove g
(iii),
i.e.
is
one-to-one.
l(gp)
l{hp)
h,
i.e.
Suppose gp = Therefore h.
kp;
p is
then
one-to-one.
{Note: In the proof of Cayley's theorem p is a mapping of G into Sc, so that gp is itself a mapping of G to G. Caution and patience are required to avoid confusion in some of our subsequent equations.)
b.
Cayley's theorem and examples of groups
Cayley's theorem tells us that there is an isomorphic image of every group among the permutation groups of suitably chosen sets. If one demands that a permutation group satisfy further conditions, one frequently comes across interesting groups (see Chapter 3). Historically many important groups arose in precisely this way.
Problem
7.1.
Describe in detail the isomorphisms given by Cayley's theorem for (i) a cyclic group of order (ii) a cyclic group of order n (n 3), (iii) the symmetric group on three letters.
Solution:
(i)
2,
Let 1 a
G =
be cyclic of order 2. Then G consists of two elements, a = a 1. Let p be as in Theorem 7.1.
and
a,
where
a^
and
Ip is the
mapping mapping given by
Ip
1,
a a
and ap
is
the
is
op
1 -> a,
-> 1
Clearly p
(ii)
one-to-one, as ap = Ip.
n.
Let
4.5,
be cyclic of order
102).
Then
consists of
n elements l,o, a^,

a"-i o-i
,a"~i, say (see
Lemma
page
Then
Ip
:
1-^1,
1 -^ a,
a-*
a,
...,
a"~i
ap
a".
-* 1
l-a2.
'P
1,
a"-2
permutations
(iii)
The symmetric group on the

Pi
set {1, 2, 3} consists of the
P2
Vz
1^1, 2-* 2, 3-*3 1^1, 2-* 3, 3^2 1^2, 2-3, 3-l

Pi
Pi Ps
l->2, 2-l,
3^3
1^3, 2^2, 3^1

l-*3, 2-l,
Pe
-*
3^2
Then
PlP
P2P
Pi
Pap
PiP Psp
Pep
It is
Pu Ps -* Ps. Pi^Ps, P2^Pi, Pa-^Pe.

P2
-* Pi,
^ Pi, ^ P2.
P2
-^ P2. -*
Ps
Ps-
Pi ^Pi.
P4
PS^PS. P6-*P6
^P6. PS^PS. P6^P4
Pi ^Ps.
Pi
P5^P2. Pe^Pi
Pi Pi
P2 ^ Ps.
Ps
-* P2.
^Pi. Ps^Pe. Pe-^Ps
Ps.
P2
-^ Ps.
Ps
-* Pi,
Pi ^PS, P5-*Pl.
Pi -^P2.
i
P6^P2
1
Pi^Pe. P2-*P5. P3-*Pi.

(Pip)(pjp)
P5^P4. Pe-^Ps
between
worth checking
{PiPj)p
for
some
and
and
6.
216
PERMUTATIONAL REPRESENTATIONS
[CHAP.
7.2
Definition of a permutational representation
A homomorphism of a group G into the symmetric group on the set mutational representation of G on X.
is called
a per-
So if p is the isomorphism provided by Cayley's theorem for the group G, then permutational representation of G on G.
p is
Repeating the definition of a permutational representation of a group G in detail, we say that a mapping of G into the symmetric group on some set is a permutational representation of G if
fi,
all g and h in G. The permutational representation provided by Cayley's theorem is called the right-regular representation (the adjective right is used because the representation is obtained by multiplication on the right).
for
Example
1:
(i)
G be the symmetric group on {1,2,3}, and let Problem 7.1(iii). Then p itself is a representation of on six elements.
Let
be as in the solution to as a permutation group
(ii)
There is another representation of the symmetric group G on {1, 2, 3}, the most natural one. This is the identity isomorphism, for G is itself a permutation group on {1, 2, 3}.
(iii)
Let
. .
.,
G be the cyclic group of 1, 0, 1, ... Let a be the

.
order 2 and let be the set of permutation of X defined by
X
i
all
integers
a:
2i-^2i+l,
2i+1^2i
for
0,1,
...
Thus a sends an even integer to the succeeding odd integer and an odd integer to the preceding even integer. Let i denote the identity permutation. Then t and a together constitute a subgroup r of the symmetric group on since
I,
la
at
Now 1 I
suppose
is
consists of the elements 1 and a. Then the mapping n: a-* a, actually an isomorphism since both G and r are cyclic of order two.
So
/4
is
a permutational representation.
(iv)
Suppose m
is
a positive integer and that
is
the cyclic group of order n,
G =
Let a be the permutation of
a:
{l,a,a2, ...,"-!}
X=Z
jn
the integers defined by

...,
jn^jn+l, jn+
we have
Oa
1 ->
+ 2, =
jn
+ (n - 1)
-^
jn
{j
0,1,
.. .)
In particular
1,
la
2,
...,
(n-l)a =
in blocks of n.
all,
Note that a
It is
cyclically
permutes the integers taken

see that a
Oa^
3,
are
all
is of order n. First of This means that the permutations i, distinct since they act differently on 0.
. .
not
.,
difficult to
Oa"-i
= n1.
a, a^, ..
Oa^ = 2, .,"-!
If ; is any integer, ja" = j. So a group r of order n. Hence the mapping

1 ->
is
1,
=
/m
and
{i,
a, a^,
..,"-!} is a cyclic
of
into
r defined by
a-*
a,
...,
a"-l->an-l
an isomorphism and therefore a permutational representation of G.
Sec. 7.3]
DEGREE OF A REPRESENTATION AND FAITHFUL REPRESENTATIONS

(v)
217
Suppose that G is the dihedral group of degree metries of the square
4.
is
the group of sym-
D
Let g
vertex.
define
Since g
X-
Let
e:
G^
page 75, g takes each vertex of ABCD to a Ag, Bg, Cg and Dg are distinct vertices. If we {A,B,C,D} and y^ by xjg = xg for all xGX, then yg G SxSx be defined by ge = y^. Then if x.X, g,hGG,
G.
By Lemma
3.12,
is one-to-one,
x{gh)e
xjgf^
x(gh)
(xg)h
(a;yg)yh
a;(ygyfc)
x(ge)(he)
Thus
is
(gh)e = gehe, and so 9 is a permutational representation of G. If ge the identity permutation of X, then fr leaves every vertex of ABCD un-
changed.
But by Lemma
3.7,
page
71,
i.
Hence
Ker
{i}
and
is
actually an isomorphism.
7.3
a.
DEGREE OF A REPRESENTATION AND FAITHFUL REPRESENTATIONS

Degree of a representation
Definition:
The degree of a permutational representation on X (more as a representation) is the number of elements in X.

2:
briefly referred to
Example
We
(i)
inspect
Example
l(i)-(v).
is
This representation This representation
of degree of degree
6. 3.
(ii)
is
Notice that in (i) and (ii) we have two representations of the same group, namely the symmetric group on {1, 2, 3}, of different degrees.
(iii)
This representation is of infinite degree. Notice that here G is cyclic of order 2. Hence there are representations of finite groups which are of infinite degree.
(iv)
This representation This representation
is
of infinite degree. of degree

4.
(v)
is
Problems
7.2.
Find a representation of degree 3 for a cyclic group of order

Solution:
2.
Let
possible representations of
be the cyclic group of order 2, say G = {1, a). Let X = {1, 2, 3}. Then there are several G on X. First of all there is the rather trivial representation
T
:
1 -
I,
->
as usual the identity permutation, t elements g,h in G we choose, (gh)r = I and

i
where
is
is clearly
a representation.
i
For no matter which
{gT){h.T)
= (grXhr). Another representation of choosing a different homomorobserve that if / is a homomorphism of G into the symmetric group Sx, then either or G/j is of order 2. Note also that G/t must be a subgroup. So in deciding on a choice of Ii = T li, we need subgroups of Sx of order 2. There are actually 3 of them. To see this let
from which (gh)T
phism.
= n G involves
Now
aj:
l-*2,
{i,
2^1, 3-*3
aj:
l-*3, 2-2,
3^1
ag:
1-1, 2-*3,
3^2
1,2,3.
Then {i, aj}, mappings
U.a^},
ag} are
/Ji
:
subgroups of Sx of order two, since a?

I,
=
i,
for
Thus the
1 ->
- aj
jtt2 :
i,
-* oj
/jg
1 ->
ag
are representations of
on X.
218
[CHAP.
The proof that there are precisely three subgroups of order 2 follows from an inspection of the subgroups of Sx- Since we have no real need for such proof here, we leave the details to the reader.
all
The upshot of these considerations
is
that
we have produced
4 representations of
of degree
3.
7.3.
Find representations of degree 10 and 15 respectively of the

Solution:
cyclic
group of order
5.
Let
be cyclic of order
5.
Then we can
find
eG
such that
{1,2
7 -^ 8,
f or
i
G=
15}
{1, a,
aP-,
a?, a*}.
Let
X =
and
let
{1,2, ...,10}
4,
Y =
^
aj
1 -^ 2,
2 -* 3,
3 3
-* 5,
^ 1,
6 -* 7,
J 3,
-^ 9,
10,
10
->
aa
It
2,
-> 3,
-> 4,
5,
5 -* 1,
>
5
5.
follows from a direct calculation that both aj and are of order

Ti
Then
{i,
!, aj,
ttj,
a*}
Tj
{i,
02, a^, al,
a*}
are subgroups of order 5 of
Sx and Sy
01-.
respectively.
1 - 1 ~*
1,
Thus
.
a
a
^ !, ^ a2,
a'^
-*
al
a^
and
are both representations of G.
/i2
t,
.,
a*
The
first is
of degree 10 and the second is of degree 15.
b.
Faithful representations
Definition:
representation
is
termed faithful
if it is
one-to-one.
Both faithful and
non-faithful representations are useful, as
we
shall see later.
Problems
7.4.
Are the representations

Solution;
(i)
in
Example
1,
page 216, faithful?
p is faithful.
(ii)
The and
(ii)
identity isomorphism is one-to-one, so this representation is also faithful. Notice that (i) provide examples of faithful representations of the same group which are of different
degrees.
(iii)
n
/4
is faithful.
(iv)
is faithful.
is faithful.
(v)
The representation
7.5.
Inspect the representations of

Solution :
(a)
(6)
(a)
Problem
7.2
and
(6)
Problem
7.3 for faithfulness.
T
Hi
is
not faithful since a'

/i2
and ar
1,
i.e.
t is not one-to-one.
However
fi^, fi^, /is
are faithful.
and
are both faithful.
7.4
PERMUTATIONAL REPRESENTATIONS ON COSETS

Suppose that
cosets
Definition:
is
ofHinG
a group and that i? is a subgroup of G. are Hxu Hx2,

.
. .
Then the right

{7.1)
Sec. 7.4]

of the identity element alone, then {7.1)
is
219
simply an enumeration of the elements of G. Here we will show how to obtain a permutational representation of G using the cosets {7.1) which coincides with the regular representation when {1}.
If
H consists
H=
describe our representation, let us choose a complete system of representatives of Hg of in G, with 1 the representative of H. In other words, we select in each coset Hg an element which we term the representative of the coset, with 1 the representative of H. is then simply the set of chosen representatives. We call such a set the right cosets
To
X a right
g.
transversal of
H in
G.
Given a right transversal X of H in G, we denote the representative of the coset Hg by Thus g is an element of .X". Since two right cosets are either identical or disjoint, it follows that Hg Hg because g G Hg. Notice that if h G H, then hg = g since
Hg = Hg = Hhg = Hhd.
For example, if G is cyclic of order 4, say G = {l,a, a^, a^}, and if group of order ^ of G, then (l,a} is a right transversal of in G. and note that 1 1, a = a, a? = 1, a? = a.
H
We
where
{1, a^)
is
a sub{l,a}
take
X=
With each element
flr
in
G we
associate a
y^:
mapping
y^ of
X into X,
y^ is defined
by
x^xg
{x&X)
this
{7.2)
first
In f act Yg
Sigz
is
a permutation of X.
To prove
we
show that
if
g\, g^
is
G,
then
gTgi-
For gTgi
is
the representative of the coset Hgigz, while g\g2
the representa-
tive of the coset Hg^gi.
But Hgi
Hg\, and so
Hg^2 = Hgig2 =
Thus
g7g2
{7.3) to
HgTgl
gigi
y^ is
for
all gi, g2
GG
First
{7.3)
We use
Assume
that
Vg is
prove that the mapping
a permutation.
we prove
y^ is one-to-one.
- yy^ {x,y G X). Then xg -yg, and so xgg-'^ ygg^K Using {7.3), we find = xgg~^ x = x and si mila rly ygg"^ = y, from which x = y. Finally we prove xgg-^ onto. Suppose x G X; then xg^^ G X and
xy^
3:9
Vg
- xg
^g
= xg-^g =
y^ is
Hence every element of
is
an image.
Thus
a permutation.
We now define a mapping tt of G into Sx. assigns to ff in G the mapping y^, so that gv the permutation of defined by {7.2). The aim of the discussion in this section is to prove that this mapping tt is a permutational representation of G on X. We have only to verify that it is a homomorphism, i.e. if g^, g^ G G, then {g^g^Tv - {g^TT){g^TT). Note that
-n-
is
To prove {g-^g^T^ = {g{'^){g2'^) we must show that {S-iQ^-^ is a permutation of X. the effect of the mapping {g^g^-n is the same as the effect of the mapping {g^Tr){g^-?r). (Note that (firj7r)(s'2'r) is the product of two mappings, i.e. the result of first performing the mapping
(flfjTr)
and then
{g^-^).)
If
GX,
using
{7.3)
we
find
{x{g^n)){g^^)
x{{g^g2)7^)
x{g^g^)
xg^g^
x{{g^7r){g^n))
Hence
{g^g^-R
{g ^Tz){g ^tv)
tt
as claimed.
-n
We
(see
shall refer to
depends on H,
G and
X.
as a coset representation of G (with respect to H). Of course In a sense tt is independent of the choice of the transversal
Problem
7.10).
220
[CHAP.
Problems
7.6.
Choose right transversals for (a) the center Z in the dihedral group D of degree center Z in the quaternion group of order 8 (see Table 5.1, page 151).
4,
and
(5)
the
Solution
(a)
The dihedral group

table:
{l,a.i, 02,
ctg,
a4, 05, ag, a^}
is
most easily described by

a^ "5
O-g
its
multiplication
1
1
dj
a^
^2
Og
a^
ttg
a^ 04 ^5 Kg
a^
ag
a<j
di
ay
04
"1
2 3
1
aj 04
as
0,2
a^ a^
as
Og
as
O'S
ag
a?
1
a^
a^
a^
Ctg
05
ttg
3
1
0.2
a-5
(17
tti
ttg
^2
dg
1
ag
ae
a.j
5 ag
tti
a?
0-2
ai
a4
Ctg
CC2
of
corresponds to a rotation of 90, while (x,^ corresponds to a reflection.) The center Z given by 2 = {1, a,<^. This can be checked by verifying that 0-2 commutes with every element of D (and no element other than 1 and 02 has this property).
(Here
cti
is
Finally Z, Za^, Za^, Za^ are the cosets of
in G.
Thus
^ =
is
{l,ai. a4. as}
a right transversal of
in G.
(6)
The quaternion group of order lowing multiplication table:

1
8,
with elements
1, i, a2> as, 14, ds, agi "7" is
given by the
fol-
dl
d2 d2 dg
1
dg dg
1
d4 d4
as as dg d7 d4
dl
dg dg a?
d4
dy a^
d4
dl
dl
dg dg
1
as
d2
dg d4
a2
dg d4 as dg
dj
dg a^ d2 dg
1
as dg
dg
dl
0,2
as
1
d7
d4 as dg
dg a^
as dg
a-,
as dg
da
dl
d4
as
dg
1
dj
dg
dl
dy
d4
dl
d2
(Comparing with Table

6 -^ d4,
we have the following correspondence: 1 -> 1, a -> di, d^ -> a^, a^ -* 03, d^b - a^.) The center Z ot Jl is given by 2: = {1, da}. One has only to check this from the multiplication table.
5.1,
ab
-^ as, (fib -^ dg,
The cosets oi Z in Jl are multiplication table. Thus

is
Z, Za^, Za^, Za^.
This again can be checked directly from the
X =
{I,di,a4,a5}
in Jl.
Sec. 7.4]

(a)
221
7.7.
7.6(b)
Find a coset representation of with respect to Z.
of
Problem
7.6(a)
with respect to Z, and
(6)
^ of Problem
Solution:
(a)
We
Thus tt using the right transversal X given in Problem 7.6. permutations of X that v assigns to each element of D. We will calculate in = {l.aj, (14, Og}. Then 1 di = a^, and so l(ayv) = a^. detail the permutation aîr. Now
work out a
coset representation
we must
find the
OiOi = 1 since so a4(Ojj7) = ctg.
Za\
=Z=
Z\,
and so
a^{a^!^)
1.
o^ =
204,
04
-^ 05,
ctj
since
Za4aj
=
a^.
Za.^
Z'ctg,
and
Finally
0^07
=
:
a^
since
Z(aâ^)
and so
a5(aiw)
Therefore
diTT
lâi, ai-*l,
agâ4
"5-* 15
05 -^5
Similarly the other permutations are

Itt:
ttgn:
1^1,
1--1,
l-^tti,
aj -* aj.
1] -^ ii.
a4 -* 04, a4 -^ 04, 04 -* 05.
a^7T\
(i47r
'.
di-^l,
!
5-*a4
l-*a4.
Is.
04-^1,
O4
-> Oi,
1,
05^
ai
aîr:
agTT
CtyTT
:
lâg,
lâ4,
i -* 14,
a, -^ oj,
ag-*l
Og-âi
ffig-*!
4 ~*
1^5.
-> 04,
04 -^ Oi,
is
It is instructive to
check directly that this mapping r

{1, Oi, a^, a^).
a homomorphism of
into the per-
mutation group on
(6)
Again we must
find the permutations that ir assigns to each element of ^. The argument follows closely that of (a) above, using the set = {1, a^, a^, Og}. Here we give only the result which the reader is urged to check.
Itt
1 -> 1, 1 -> ai,

1 -* 1,
di
ajjT a^lT dsn(l45r
^ Oi, aj ^ 1,
04 -> 04,
a^ - Og, 04 -> 04, a^ -> ag,
ttg
-> ->
ffg
ttg
a^
ttl -> Oi,
Og
-> ttg
1 ->
tti,
aj
^ 1,
ttg,
ag -> 04 Og
-
ffij
1 -> 04, 1 -> dg, 1 -*

(I4,
Oi -
ttj
04
0.4
-^-
1,
OgT
tte""
-> 04,
-^ dj,
-> 1, -> aj,
Og -> 1
etg -^ ctj ctg
ai -> ag,
04 a^
OyTT
1 -> ag,
ai -> a4,
-* 1
7.8.
Consider the permutation group S^j

sider
01,
2;
Its elements are
4>i
and
<f>2
Con-
now
the permutation group Sj^j

essentially the
Its
elements are *i
=
(
s^j
= ["
*i
^""^
*2 and 02 are
will
make
Let
this idea of "essentially the
same except for the elements they act same" precise.
on.
Give a definition which
Solution:
that
a:
F be a permutation group on a set X and let G be a permutation group on a set Y. We say F and G are isomorphic as permutation groups if there exists a one-to-one onto correspondence X -*Y and an onto mapping e F -* G such that for all x in X and / S F, (xf)a = (xa){fe).
:
(In this
problem
1 -* a, 2 - 6
and
0^
->
*j, 02
~* *2-)
7.9.
Prove that if (Hard.)

Solution:
and
are isomorphic as permutation groups, then they are isomorphic as groups.
The of the solution of Problem 7.8 provides the isomorphism. homomorphism. Let f^,f2&F. For any x . X,
(xa){{fif2)9)
First
we show
that
it
is
= =
(x(fif2))a
{(xfi)f2)a
(by the definition of composition of mappings)
((a'/i)a)(/2 9)
(()(/i))(/2e)
(*)((/ie)(/2 9))
222
Since a
is
[CHAP.
mapping onto Y,
and
(/i9)(/29)
it follows that as x ranges over X, xa ranges over Y. are permutations, f 16/29 = (/i)(/29)- Thus 9 is a homomorphism.
ff
Hence as
{f if2)6
(Xa)(f2e).
Next we must show that Thus

is
is
one-to-one.
Suppose that
/]
/29;
then
if
X,
(xa)(fie)
(xfi)a
=
=
(Xf2)a
Since a
one-to-one,
x/j
xf2
from which
/i
/2.
Hence
e is
an isomorphism.
7.10.
Let G be any group and a subgroup of G. Let Zj be a transversal for in G and jt; the corresponding coset representation, i = 1, 2. Prove that Gti-i and Gn-2 are isomorphic as permutation
groups.
Solution
(Hard.)
Since X^ and X^ contain one and only one element in each coset of H in G, we define a: Xi^ X^ by sending Xi G Xj to X2 S X2 if Hxi Hx2. a is then a one-to-one correspondence. We define Gtti -* Gjr2 by (gTri)ii gir^- It is easy to check that û is a mapping. Let g-iri = /?, giT2 ! We need only verify that (x^)a = {xa)~i for each at G X. Now by definition of y8 and a,
/J.
:
jtfxfi'
= =
li{xp)
= =
H({xlS)a)
fl^((a;a)y)
Also,
Hxg'
H(xa}g
Hence (xa)y
follows.
(/?)
as they are elements of
X2 that belong
to the
same
coset,
i.e.
Hxg. The result
7.5
a.
FROBENIUS' VARIATION OF CAYLEY'S THEOREM

The kernel
Is
of a coset representation
tt
a coset representation
that the answer
7.2:
is
of a
if
group
with respect to a subgroup
We know
Theorem
H {1}. The object now is Let G be a group and H a subgroup of G. Let be a coset representation of
yes
ever faithful? to find the kernel of tt.
tt
with respect to H. of G contained in H,

Before proving Theorem
7.2, it
Then the kernel

i.e. if
of
tt
is
the largest normal subgroup
N<G
and
HdN,
then A^ cKerTr.
-n-
should be noted that from this theorem it follows that is the identity subgroup. is faithful if the only normal subgroup of G which is contained in = G, tt is automatically faithThis implies, for example, that if G is a simple group and ful. For then, by definition, the only normal subgroups of G are G and the identity subgroup. This observation has been useful in the theory of finite groups.
Proof:
Let
if
X be
is
prove that
identity
If
mapping
First we in G from which tt was defined. the right transversal of the kernel of it, i.e. the set of all elements g of G such that g-rr = i, the of onto itself, then is contained in H.
x = a;(c7r) = M. In particular on putting and xGX, then air = i.e. Hence a&H. This means that K is contained in H. Of course K, as the To complete the proof of the kernel of the homomorphism is a normal subgroup of G. theorem, we must show that K is the largest normal subgroup of G contained in H. To do this it is sufficient to prove that if N is any normal subgroup of G contained in H, then
i,
aGK
=
0..
1,
-n-,
Nn =
{l}.
Suppose that a G N.
If x
G X,
then
Hxa = Hxax~^x
Since
N is normal,
and
Hxa Hx.
Hence
a-n
xax~^ GN. But iV is contained in H, and so xax~^ Thus xa = x which means

x{aiT)
G H. Accordingly
for
all
iV
cK
as required.
Sec. 7.5]
223
Problems
7.11.
be the symmetric group on {1, 2, 3}. Let a 1 -* 2, 2 - 3, 3 - 1, and let t 1 ^ 2, 2 -^ 1, = gpia) {i, a, cr^} and (6) Then the elements of G are i, <r, a^, t, ot, o^t. Let (a) {'. t}. In each case find a coset representation of G with respect to H. Find also the kernels 9V{^) of both representations.
Let
3 -> 3.
Solution:
(a)
right transversal of i/ in
7.2,
jT
Hence by Theorem
tion
G is {i, t}. Notice that t~i<7t a~'^ = <fl. So fl^ is normal in G. every coset representation has kernel precisely H. The coset representaassociated with the right transversal {i, r) is given by
iTT
:
-* -* -*
I,
T -* T
TTT'.
i~^T,r-^l
i
air
<T^3r
t,
T
T
-*
(<rr)7r
-* r,
r -*
I,
-> T
{(j'^t)it
'.
~* T, T ~^
Clearly
Ker i- =
{i, a,
a^.
(5)
Let X={(,(t,ct2}, Since G = HuHaUHor^, permutational representation v is

H7
:
is
H in
2 _
g,
G.
The associated
I,
CT
CT,
a^ -*
<r2
^^
j^
^ _>
j,2^
a^jT
- a2,
CT
I,
a^ - a
(ct^tJtt
ct,
->
i,
c^ -^
ct^
It
tion of
follows immediately that n- is faithful, since the only element is I. Hence the kernel of ;r = {i}.
mapped
to the identity
permuta-
7.12.
Let
be the alternating group of degree 4 and
let
H be the
3
1
subgroup consisting of the permutations
a
^2
2
1
4\
/I
2
4
4\
n
'
2
3
3 2
r
1^
sj'
\3
2)
[a
Find an associated coset representation of

Solution:
^
,^
,
with respect to H.
2
3
Is this representation faithful?
/I
';
4\
^^
/I
4\
^3
/I
2
3
3 2
^;*^
3
1
/
'
4\
(2143;^
(.3412;'
H
= (4
ij-
T^-
^^
-^
consists of the elements
coset representation.
immediately that a right transversal of in G is X = {i, a, a^}. Let v be the associated It is easy to check directly that is normal in G. Then Ker tt = by Theorem 7.2, and so v is not faithful. Finally we list the permutations gtr with g in G:
It follows
r
CTir
^
-> - -* -*
I,
^ a,
-
CT^,
o^
ct2
^
-^ ->
a^
J
t^tt
-> -
i,
a -^ a,
(T
a^ - ^2 a^ ->
a^
ct2
i
CT,
a
(T
(t2ct);7-
(T,
->
ct^,
aV
TjTT
(T^,
->
1,
(,2
CT
(t2<''^)tt
-* a^, - "*
-
a
(T
^
->
t,
^
-
a
(,2
I,
<r
cr,
ff2
(Tiff)!?-
0-,
<7
^
_
<r2^
5,2
^ 2 ^
,
^gj.
(^
CT,
(rg(T)jr
ff,
a
CT
^
->
a^, a2 -^
ff2
(tjct2)^
(,2^
(,
j^
a2 ->
ff
(rsCT^);^
(t2,
,,
CT
b.
Frobenius' theorem
Let be a subgroup of G, a transversal and the associated coset representation. denote the right regular representation of G. Our idea is to express p in terms of tt. P
-n-
Let
If
a;
G X and g G
G,
lie
we have
x(gp)
in
a;r
Of course xg need not
X.
224
[CHAP.
We
a: H.
for
a,
know, however, that xg belongs to the same coset as xg. Hence xg = axg where Now a is clearly dependent on x and g, and we denote it by Ux.g. Substituting ux.g we have xg = ttx.gxg (7.4)
x(gp)
or
ax.gx(gTr)
{7.5)
Now
{hx){gp)
any element of G can be expressed in the form hx for h - hxg - h{x{gp)) = hux.g x{g-n), i.e.
(hx){gp)
&
and x G X.
Therefore
(hax,g){x{g^))
{7.6)
This equation suggests that the effect of gp on an element hx of G can be explained by what happens to h (it goes to the element hax.g of H) and what happens to x (it goes to
x{g-K)).
We express this formally in the following theorem which is due

Theorem
7.3:
(essentially) to Frobenius.
Let
be a group and
H in
G.
Then there
mutations of
HxX
be a right transversal of subgroup of G. Let of G as a group of pera faithful representation and X) defined by (the cartesian product of
is
Ha
{h,x){gd)
{hax.g
x{g7r))
let
eG
Proof: The proof is an adaptation of the discussion of the last few paragraphs. First with right-transversal X. For each TT be the coset representation with respect to which is defined as the permutation g-Tr gives rise to a permutation gX of
HxX
*"'*'^^'
{h,x){g\)
{h,x{g.)),
{heH,xGX)
{7.7)
Note that gX
is
a permutation of
H xX.
For
if
{h,x){gX)
= =
{h',x'){gX)
{h', x'{g-n))
then
Since
{h, x{g-,T))
g^r is a permutation of X, it follows from x{gTr) = proved that gX is one-to-one. But gX is also onto since permutation of HxX.
x'{gTr)
g-rr
is
that x onto X.
x'.
Hence we have
is
Clearly gX
then a
Now
as
we saw
in {7.5), if
g xg
GG, =
then
{ax.a
ax.g{x{g-)),
e H)
For each
gGG
define
is
{h,x)ga
{hax.g, x)
We
verify that ga
a permutation of
{hax.l,x)ga
HxX.
=
Suppose
{h, x)
GHxX.
Then
(hax.lax.g,x)
{h,x)
is
Thus
g<7 is
mapping
of
HxX onto HxX.

x')g(T.
pose that
{h, x)g<T
(fe',
It remains to verify that ga This means that
one-to-one.
Sup-
{hax.g, x)
{h'ax'.g, x')
and therefore we find x = x' and ha^.g = Hence ga is a permutation of HxX.

Finally
h'ax.g
from which
h-h' and
{h,x)
{h',x').
we compute
gagX.
{h,x){gagx)
{{h,x)ga)gX
(hax.g,
x)gX
{hax.g, x{g7r))
(T'.S)
Thus
(fif<T)(fl'A)
ge
{7-9)
Sec. 7.5]
225
As both ga and g\ are permutations

Let
g^, g^
of
H x X,
g6
is
a permutation of
HxX.
{7.10)
G.
We
must show that
{9M9S =
i9,9,)6
(Note that the left-hand side of {7.10) is the product of the proof of {7.10) we introduce the following notation:
two permutations.)
To
facilitate
where g
is
an element of G.
{7.8), {7.9)
Equations
and
{7.11) j'ield
{h,x)ge
{hax,g,xa)
{7.12)
Note that as
tt
is
a homomorphism, {g ^7r){g ^tt)

\<^2
{g^g^-n
"3
so that
{7.13)
Applying
{7.12) twice
and
{{h,
{7. IS)
once,
we have
(hax,g^, Xa^)gê
{h, x){gêg^B)
x)gê)gê
{hax.gâxa^.g^, (xa,)a2)
{hdx.gâxcc^.g^, (ia2))
{hax.gâxoiyg^,
^^
{^-H)
On
the other hand, again on using {7.12),
we have
{hax.g^g^,
{h,x){g^gê
xa^
{7.15)
is
To prove that 9 is a homomorphism, we must show that the right-hand side of (7.i4) equal to the right-hand side of {7.15), i.e. we must show that
To accomplish
this
we
use equations {7.5) and {7.6) and obtain

X{9i92)
a:i:.9i92(^"3)
{7-17)
Also
from(7.i5).
{X9i)92
(x,9i(a;ff,))fir2
ax,aâxai.g2i{Xa^)a2)
a:c,gia:rj,92(a;a3)
This means
{xg^)g2
ax.gâxa^.g2{xa^)
{7.18)
Since a^CPjSr^) = {xg^)g2, the right-hand sides of {7.17) and (r.iS) are equal. follows and therefore ^ is a homomorphism.
It
{h,
Thus
{7.16)
x)fif2(9
only remains to show that & is one-to-one. Assume gê = g^B. Then for all pairs (/i, x)&B.xX; and in particular, if {h, x) = (1, 1),
(A,
x)g^B
from which
Using equation
(7.5)
sr,
ai^^^
a^.^^
and
l(sr,7r)
\{g^^)
{7.19)
= 1, we see that = l(flrjp) = ai.gj(l(fifj7r)),

with x
that g^
g^
l(sr^p)
ai.g^ClCM)
Using
{7.19)
we conclude
g^.
Thus
7.3.
^ is one-to-one.
This completes the proof of Theorem
We
to H).
call the homomorphisnn 6 of Theorem 7.3 a Frohenius representation {with respect Of course 6 depends on X as well, but it can be shown that in a sense this dependence
does not matter.
226
[CHAP.
Problems
7.13.
Describe in detail a Frobenius representation for the symmetric group on groups given in Problem 7.11(a) and (b).
Solution:
(o)
{1, 2, 3} relative to
the sub-
The
set
X-
{i,
t)
is
resentation of
on
H X X described
=
= {i, a, ct^} a right transversal of above. The elements of

(,
in G.
Let
be the faithful rep-
are
a, <T^, T, ar,
a^r
We
o^T
use the formula xg
ax.gxg to calculate ax.g.
Note that
i,
i,
a^
i,
f=
t,
of
t,
r.
As an illustration of the procedure we calculate Ot.o-. Now we have to t and ar.a = tot a^. Similar calculations lead
a, ,1 a, f
a,. ,0-2
ra
to
a-,,ard.
Since
tw
a^r,
:=
< ,T at jO-T a, ,tr2T
ff
a2
= =
"t.i
"t.o-
or
a2
T,a2
= =
Ot.t
ttT.O-T
^2
aT,o-2T
= = =
<72
<T
We use the definition of e given in the statement of Theorem 7.3. The effect of tt on given in the solution of Problem 7.11(a). The results, repeated here for convenience, are
HT
TTT
is
CTjT
= =
(a2)jr
-^
I,
T -* r
{aT)-!r
(cx^T)Tr
-> T,
T -*
We
and
can now calculate the
effect of
go for each g in G.
In particular,
(h,i)ae
(fcaj_o.,
i(<T77-))
(h(T,i)
(h,T)ae
(/lo^.o.. ''('"r))
C^"^. t)
for
all
We
list
the effect of the permutations
ffff
for the elements of G.
{h, x) -* {h, x)
{hGH, xG
X)
(h
(ft.
aO
(h,
i)
^ (ha, 0,
- (A(r2,
i),
(h, t) -^ (/Kr2, r)
(fe,
(A,
4)
t) -> {ha, t)
G H) e H)
tB
(aT)e
(ft,<)^(ft,T),
(ft.,T)-^(/l,<)
(ftSH)
(ft,GH)
(h
(h,,)^(ha,T),
(A,
1)
(ft,T)-*(/ur2,t)
(/i, .r)
-> (A(r2, t),
-* (ha,
i)
G H)
(ffisr2)*>
One could check that 9 is a homomorphism by inspecting (gie)(g2f) and The above description of e immediately shows that S is one-to-one.
(6)
'where
(ffijr2
G).
Here
{1, r};
e is
representation
X, a right transversal of then given by

:fl
:
in G, is given
by
X=
{i,
a,
a^.
The Frobenius
(ft, sc)
(ft,
x)
(hen, xG
X)
(h,a^)^(h,i)
(ft, <T^)
ae:
(h,i)^(h,a),
(ft,
(h,a)-*(h,a^),
(ft,
(h
(ft
G H) G H)
aH
re
-^
(ft, <t2),
a)
(ft,
0,
-^ (h, a)
(ft,
-^ (ftr,
c),
(ft,
a) -* (ftr, a^), ) -> (ftr, a), a)
(ft,
a^) -^ (ftr, a)
a2)
(h
(ft
(<rr)9:
(ft,0^(Ka2),
(ft, 1)
(ft,
(ft,
(aMtf:
(ftr, a),
(ft,
^ (ftr,
i),
(h,a^)
^ (ftr, ^ (hr.a^)
i)
G H) G H)
& H)
(h
7.14.
Describe in detail the Frobenius representations for the alternating group of degree 4 relative to given in Problem 7.12. the subgroup
Solution
Here
and
G
{1,
consists of
I,
a,
ct2,
Ti, Tjff, Tj(72, T2, T2<r, T20-2, T3, TgCr, r3CT2
H=
ti, rj, tj}.
right transversal of
in
is
{i,o,a^.
The Frobenius representa-
tion e is then described as follows:
Sec. 7.6]

(h, x) -* (h, x),
227
i9
(h&H,
e X)
{h, a^) -^ (h,
i)
e:
<fie:
(h,i)-^(h,a),
{h,i)-^{h,<r^),
(h,a)^(h,<r^),
(h
G H)
(h,a)^(h,,),
{h, a^) -^ (h, a)
(h&H)
{h
Tie:
(ria)9:
(Tia^)e:
(h,,)^{hrj,i),
{h,i)-*(hri,a),
{h,a)^{hT2,a),
(h,<.) -* {hr^.a^),
(h,a^) -^ (hr^,,,^)
(h, a^) -^ {JlT^,
i)
& H)
& H) & H) G H) G H) S H)
(h
(h (h (h (h (h
(h,,)-*{hTi,ai), (h,a)->{hr2,i.),
(fe,
{h, o^) -* {hr^, o)
rê:
(r2<r)e:
- (/lT2,
.),
(fc,a)^(/lT3,a),
(h,
cr)
{h,o^)
^ ^
^ ^
(hTi,o^)
i)
(h,c)-^{hT2,<r),
(h,
i)
(hra, a^) (hra,

i),
(h,a^) -^ {hri,
(h, a^)
(h, a^)
{T20^)e
(hT2,
ff2),
(h, a)
^ ^ ^
(hri, a)
{JiT^, a^)
TgU
(h,
i)
-^ (/1T3,
<),
(h, a)
(hri, a),
S H) & H)
G H)
{Tso)e:
(T3(T2)ff:
{h,i)^(hT3,c),
(h,a)-*{hri,a^), (h, o^)

(/iri, 0,
-* (hrz,
c)
(h {h
{h,,)-^(hT3,a^), (A, a)
{h,
a'')
(Jlr^, a)
7.6
a.

Subgroups of
shall give
4.
finite
index
we
Frobenius' representation, although only a variation of Cayley's, is very useful. Here one application of this representation. First we recall a definition given in
Chapter
Definition:
A group G is finitely generated if it can be generated by a finite set,

is
finite
.
subset
Sj, s^,
.,s^G S
i.e. if there there are elements such that for each g = 1) such that and integers e^, e (e^
S (# 0)
of
EG
si'
si"
Theorem
Proof:
|S|
7.4 (O.
Schreier)
subgroup of
finite
index in a finitely generated group
is
finitely generated.
Let G be a finitely generated group. Let S be a finite set of generators of G, m. Suppose H is a subgroup of finite index in G. Choose a right transversal X {x^, .,Xj) of H in G, with x^ the identity. Notice that j < > by assumption. Let e be a Frobenius representation of G with respect to H given in Theorem 7.3. If h&H, then ft = 1 and ai,h = fe so that we have (1, l){he) = {h, 1). But h can be written as
with
=
.
h
with
s^, .,s^GS. homomorphism,
. .
:><"
i
(.,=1)
..,n;
Put
t.
sl\
1,
then
t^
t^.
Since
6I
is
{i,i)m = {i,i){tMh&)
Let
t.TT
(tj)
a^, i
HxX (see Theorem 7.3) we have

{l,l)m =
Hence ^ =
i,
1,
. ,
n.
By
repeated applications of the definition of the action of gO in
{a,,,^,la;}{tê)---{tj)
ia,,,â,,^,,^,l{aâ^)){tS---{tJ)
Oi, tibial,
tg
aioj... _!.*.
i n.
|Z|
In other words
ttx.t
we have
number
Since a^, .,a^G S^, l{a^- -a^) for each expressed h as the product of elements of the form
.
GX
where x
G X, and
s-^
where
GS
This means that
As
and
|S|
= m,
the
of such elements is at
most 2mj.
is finitely
generated.
228
Remarks about the proof
of
[CHAP.
b.
Theorem
7.4
j in G,
have actually proved that if G can be generated by m elements and H is of index then H can be generated by 2jm elements. However, it is not difficult to reduce this number to jm. To do so observe that we have proved that the elements ax.t generate H, where x is an element of a right transversal X of H in G. We recall that if g denotes the representative of the coset Hg, then a^.g is defined by
xff
We
O-x.gXg
(equation (74), page 224).
This means that

ax.g
=
ff,
xg{xg)-^

that
a;i(ii)-
generate
=^
xs~^s
{7.3),
with
= = =
or
s-^,
where s&S.
Note
xs~^s
by equation
page 219.
Let y
O'x.s'^
xP^^; then
ys
x and
xs~^sx-^
ys(ys)-^
ay,s
Thus H is actually generated by the elements a^.s, x lX The number of these elements is jm. (In fact one can lower this bound and prove that H can be generated by 1 + (to - l)j elements. For a proof of this result see Theorem 8.13, page 264.)
i.e.
ax.s-^ is the inverse of ay,s.

s
and
&S.
E:
Note that in Section 7.6a we have actually proved that S, without the assumption that \X\ and \S\ are finite.
is
generated by
ax.s,
X,
Problems
7.15.
Let G be the symmetric group of degree 3 and let of generators of described above.
H
H
be a subgroup of index 2 in G.
Find the
set
Solution:
use the description of G in Problem 7.11(a). Now a subgroup of index 2 in any group is = {i, a, a^}, Thus if is of index 2 in G, is normal in G. Moreover, \H\ = 3. Hence since this is the only subgroup of order 3 in G. A right transversal of = {t, t}. in G is Clearly G can be generated by a and t, and so is generated by
We
normal.
H
a
I
Off,-
= aa~^ = =
Tf~^
0,r,a
T(r(fa)~^
= =
TaT~^
T^
=
I
a^
a^^T
ttr.T
~
is
TT{ff)~^
Thus we
find that
i, <r,
a^ generate
H. Of course
actually generated
by a
alone.
7.16.
Let G be any group and let H he a subgroup of G of index two elements, then H can be generated by three elements.
Solution :
2.
Prove that
if
can be generated by
Suppose that G is a group generated by c and d and that is a subgroup of index 2 in G. If both c and d are in H, then 'D G and is not of index 2 as initially assumed. Without loss of generality we may suppose that c ^ H. Then the cosets of in G are and He. Thus every element of G can be written in the form he or h (he H). This means that {1, c} is a right transversal of in G. Therefore G is generated by the elements
l,c
lc(Tc)-l
CC-l
=
j,
1, ai,<j, ae,e, Cfc.d
Hence
is
generated by the three elements
a^^^, a^
and
a^^^.
Sec. 7.6]
229
7.17.
such that G/N Let G be generated by a and 6 and suppose that iV is a normal subgroup of G infinite cyclic. Find a set of generators for 2V in terms of a and generated by Na and G/N is
(Hard.)
Solution:
It is clear
is
b.
Nb = c = b.
then since in G. If = {l,a=ti,a2, .} is a right transversal of that = fta""". On the other hand if b e N, put (JVa)" for some integer w, ba-^ S N. Put c are the elements Clearly gp{a, c) = G. We therefore take S = {a, c}. The generators of
b^N
Oj-s
xs{xs)~^
{x
X, s G S)
If
X
a;
a'
and and
a,
a^.,,,
a'a{a'a)-^
aMa''^')~^
If
a>
c,
Ox.s
.
=
. .
a'ciaJc)''^
a'ca-'
Hence the elements
ba"^ aca~^, .} are a set of generators for AT. Since either c = generators in terms of a and b thereby obtaining a set of genor c= b, we can restate this set of erators of G of the desired kind.
{.
.,a-^ca,
c,
7.18.
Let G be generated by a and b. Find generators for has at most three subgroups of index 2.
all
possible subgroups of index 2.
Hence show
Solution:
Let
(1)
be a subgroup of index 2 in G.
(2)
aH, be H,
In case
(1)
a^H, b H,
b,
(3)
Then we may have a^H, b^ H.
S).
Thus
H is
{l,a} take generated by
X=
c
and S = {a, b}. a^ and obaî.

with
Then generators of
H
H
are the a^^^
^X
and
In case
(2),
proceeding as in
(1)
X=
{1, b},
is
generated by
{a, c}.
a, b^ is
and bab"!.
generated by
c,
In case
(3),
ab
e H. Take
X=
{1, a},
and S
Then
aca"!
and
a^.
So the possible subgroups of index 2 are:

(1)
gp(b,aba-\a^),
(2)
g-pCa,
bab-i,
b2),
(3)
gpiab.a^ba-â^)
7.19.
Let
G =
gp{a,
b, c)
c.
iV contains b
and
and let Find a
A^
set of generators for
be a normal subgroup of G of index 3 with in terms of a, b and e.
G/N =
gp(Na).
Suppose
Solution:
Choose
erate
AT.
S=
{a, b,c}
and
X=
by
{l,a,a2}.
b, c,
Thus
N is generated
G
o.^,
Then the elements a^-^j, with aba~i, aca-^, a^ba~^, aP-ca~^.
G X and
S,
gen-
7.20.
Prove that
if
a group
of index 2 can be generated
of index 2 which contains a subgroup by two elements. (Hard.)
is cyclic,
then every subgroup of
Solution:
Let
H be generated by
gp(b,aba-i,a^)
and suppose that

H^,
a^H.
It
follows that
G=
gp(a,
b).
From Problem
7.18 the possible subgroups of index 2 are

(1)
(2)
gpia.bab-^.b^)
Hj,
(3)
gp{ab,a^ba~i,a^)
Clearly
^ H^
or H^, as then each of

b,
ff is the cyclic
group generated by
them would actually be equal to G. oba-i = b*" and a^ = b^ for some integers r and
ba
= H^ Thus H = H^.
s.
Since
We
have
(7.20) (7.2i)
a-^b^
and
a2
b^ (7.20),
is
The generators of Hg are a, bab-^ and b2. bab-i = (ba)b-i = a-ib' i from generated by a,b^-^,b^. But ffpCb""- 1, b^) is cyclic generated by c, say, as it cyclic group. Thus Hj can be generated by two elements.
and so Hj
is
a subgroup of a
The generators of i/g are ab, a^ba'^ and a^. Hence ab, a2ba-i(a6) and a^ are generators for Hg. Using (^..gi) we conclude that ab, b^ + z and b" are generators for Hg. But sfp(bs + 2, b') jg cyclic generated by c, say; so Hg is generated by two elements and the proof is complete.
230
Marshall Hall's theorem
[CHAP.
c.
The second application of permutational representations
is
due to Marshall Hall.

is
Theorem
7.5:
The number of subgroups

finite.
of finite index j in a finitely generated group
This theorem
may
where n<oo.
of index j
Suppose
be restated as follows: Let G be a group which is generated by ai, a j is a fixed positive integer. Then the number of subgroups of G
. . .
is finite.
Proof: Let S. be the symmetric group on 1, For each subgroup of index j in ,j. the finitely generated group G choose a right transversal Z^ (we emphasize that X contains the identity of G). To avoid confusion between the number 1 and the identity of G, we shall write the identity (for this proof alone) as e. Thus we have e e X^. Let 7r be the coset representation of G with respect to the transversal X^ (see Section 7.4). Then tt^ is a
.
.
homomorphism of G into Sx^^. Since |Z| = j, it is easy to prove that there exists an isomorphism 4,^ Sxj, -* S. such that e G Sx moves e or leaves it fixed according as to whether 94, moves 1 or leaves it fixed (see Problems 7.21 and 7.22 below).
:
^^
G-* S. is a homomorphism of G into S. since it is the composition Note also that if H and K are two subgroups of index j and H = K, then * - *^, for there exists an element g but g (or vice versa). Then eignjj) = e for e{g7r^) = eg = e as g G H (see Section 7.4). Accordingly 1* = 1. On the other hand, as g ^ K, e(g-jTj ^ e and hence 1*,, - 1. Thus * ^ ^^ if ^ ^ ^
:
Note that *^ = -n-j,^^ of two homomorphisms.
&H
^K
We have therefore found that the number of subgroups of index j in G is certainly not greater than the number of homomorphisms of G into S.. This is where the fact that G is finitely generated comes in. For suppose G is generated by Cj, a^. If ^, 9 are homo. . .
morphisms of
observe that
if
G
g
into S. such that
a^
a^9
for
l,
.,
n, then
cj>
9.
To prove
this,
&G,
then
all
'^l'
-1'
^j
{1,
.,n}
and
<j>
g4>
iai^<j>)^
(tti^^)'"
(ttij^)''
{ai^9)'"
fi-e
Since and ^ agree on every element of G, ^ = 9. This means that the number of homomorphisms of G into S is finite since the number of possible images of the generators of G is finite (at most (j !)").
d.
One consequence
Let
of
Theorem
7.5
6 a homomorphism of G onto G. It follows that 9 is an isomorphism, for \G\ = \G9\ = |G/Ker6i| and hence Ker9= {1}. If G is not finite, is it possible to have a homomorphism 9 of G onto G with 6 not an isomorphism? For example, if P is a p-Priifer group (see Section 6.2c, page 191), let P ^ P be defined by x9 = px, x G P. Then P9 = P; but as P has an element of order p, 9 is not an isomorphism.
be a
finite
group and
6I
In the following theorem we prove a result which groups every onto homomorphism is an isomorphism.
tells
us that for a special class of
Theorem
7.6
(A.
I.
Mal'cev):
be any finitely generated group whose subgroups of index have intersection 1. Then every homomorphism of G onto G is an automorphism.
Let
finite
<f>
Sec. 7.6]
231
is of
be the kernel of ^ and let L be any subgroup of finite index in G. If L Proof: Let index j, then the number of subgroups of G of index j is finite. Let these subgroups be
L =
Now, by Theorem
4.18,
L\,
Z/2,
Lk
page 117,
G ^
and so the number of subgroups of index
of index j in G.
. .
G4,
j in
^ GIK
GIK is precisely k, the number of subgroups be these subgroups of index j in GIK. Then .,Mk are k distinct subgroups of index j in G, by Corollary 4.20, page 121. Thus the Ml, Mi's are simply a rearrangement of the Li. Therefore every Li is an Mi and so contains K. In particular
Let
MJK,
...,
MJK
L D K
This means that every subgroup of finite index contains K. Hence K is contained in the By hypothesis, this intersection is 1. So intersection of the subgroups of finite index. K = 1. Accordingly ^ is one-to-one, and is an automorphism.
<j>
This theorem
is
important in current research in group theory.
Problems
7.21.
Prove that there exists an isomorphism

Kjfl
/i
between S^^
and S2 such that

page
221.)
if
S^^.
and
Xj,
then
i{eii)
j.
(Hint:
see Problems 7.8
and
7.9,
Solution:
Let a
{xi, X2} -^ {1, 2}
be defined by
a;2\
Xja
j,
1, 2.
Let
/i
be defined by
'!
aJi
_ /I
\1
x^)
2\ 2/'
(see
/!
X2\
^1/
7.8
_
and
7.9)
(X
2^
1
/i
V*2
V2
Then
a,
im
define a permutation
effect.
isomorphism
Problems
and therefore
is
an isomor-
phism with the required
7.22.
Prove that
*
in
,1
}
e S^x
...
general there exists an isomorphism n between S^^ ^nd î^ ^j' t*^ *(*i") } (Use Problems 7.8 and
i
7.9.)
*id S such that
if
Solution:
Let
to be the
{ajj,
...,}- {1,2,
.. .,7t}
be defined by
XiS
If
S^^
^
. .
,,
define
e/iSS^
mapping that sends i^

. .
mutation of {1,2, .,n}). It is of permutation groups. Thus fi
permutation of {x^, .,;}, Sn is a perclear then that /i is onto S, and hence a, /i provide an isomorphism is the required isomorphism.
j
if
Xj
(as 9 is a
7.23.
Prove that
if
H and K are subgroups

a coset of
of G, then each coset of

if
empty
set or in
HnK.
Hence prove that
and
H intersects a coset of K either in the K are of finite index in G, so is HCiK.
Solution:
Kg. X
and a coset of K have an element g in common, then the two cosets are Hg and and only if x hg kg, for some h G H and k & K. But hg = kg if and only if A = A;, i.e. hsHnK. Thus xGHgnKg if and only if xG(HnK)g. Then HgnKg (Hf\K)g and the two cosets meet in a coset of HnK.
If a coset of
e HgnKg
if
H is of index n and K is of index m, at most nm cosets can be found as intersections of a H with a coset of Furthermore these are all the cosets oi HnK, for any coset (HnK)g = HgnKg and so is the intersection of a coset of H by a coset of K. Therefore HnK is of finite index in G if both H and K are.
If
coset of
jBT.
232
Let
[CHAP.
7.24.
of index j in
be finitely generated with a subgroup of index j. Prove that the intersection of G is a normal subgroup of finite index, (Hard.)
all
subgroups
Solution:
By Theorem
.
7.5,
has only a
}, it
finite
number
of subgroups, M^,
the cosets of are Mx-^g^x, iBx-^g^x, .,Mx-^gjX. For if g e G, x~^gx e Mg^, for some i, implies that g G x-'^Mxx~^giX = Mx-^g^x.) Hence MjOMan nM = K is a normal subgroup of G, for \i g & K and x&G, x-'^gx G x-'^M^x, x-^MzX, But .., x-^Mx. x-^MiX, are n distinct subgroups of index j. Hence they must be all the subgroups of ., x-^Mx index j (perhaps in a different order). Thus x'^gx G K and so K < G. K is of finite index by repeated application of Problem 7.23.
. .
M is any subgroup of index of M are Mgi, .,Mgj, then

.
.
M
.
say, of index
j.
Now
if
is
easy to prove that x-'^Mx
iW
is also of
index
j.
(If
the cosets
7.25.
Let
e
:
G
-^
and
H
and
be two groups and suppose that G satisfies the conditions of Theorem -^ G be epimorphisms. Prove that G = H. <f>:
7.6.
Let
Solution:
90
'^
"^t
is an epimorphism of G to itself and so by Theorem 7.6, 90 is an isomorphism. Then if 9 ^ G, g(8) 7^ 1 and thus ge = 1. Therefore 6 is one-to-one and e is an isomorphism of
GtoH.
7.26.
Let and be unequal normal subgroups of a finitely generated group G, intersection of the subgroups of finite index in G/N is the identity. Prove that
MdN.
G/M
is
Suppose the
not isomorphic
with G/N.
Solution:
Let e G/M -* G/N be such an isomorphism. Let /i G/N -* G/M be defined by Ng -^ Mg. easy to verify that /< is an epimorphism. Then jiS is a homomorphism of G/N onto itself. By N, {Ng){fig) = Me is the identity of G/N. Theorem 7.6, iiS is an isomorphism. Now if g G Thus fie is not an isomorphism. This contradiction yields the required result.
:
:
It is
7.7
a.
EXTENSIONS
General extension
Suppose G is a group with a normal subgroup and that G/H = K. Then, using the terminology introduced in Chapter 5, G is an extension of by K. It is convenient to generalize this concept and to say that G is an extension of with by K if G has a subgroup = and G/H s K. It is our aim to investigate how a group is built as an extension of one group by another.
H H
morphism
In this section let G be a fixed group and be an isoa normal subgroup of G. Let of G/H onto K. Let in G, i.e. a set of elements of be a left transversal of containing one and only one element from each left coset of i? in G with 1 G X.
({>
G, g = xh for some x :X and some h G H. It sion for g is unique. Let g GG, x G X; then gx belongs to coset yH where y GX. Therefore
If
is
some coset of
easy to see that this expresin G, say the
gx
for some h
= yh
G H.
Now
is
uniquely determined by g and x;
we denote h by
nig.x.
Thus
(7.22)
gx = ymg,x
mg,x correspond to the elements Ux.g introduced in Section 7.5b. (We use mg,x instead of ax.g because here we are dealing with left instead of right cosets. We will explain in Section 7.7c the minor reason why we use left cosets here.)
The elements
Note that
cosets of
<f>,
the isomorphism of
H onto K.
if
G/H onto K, is a one-to-one mapping of the set of left Therefore we can unambiguously denote the representative of the coset
k.
gH
by Xk
(gH),^
In particular then, xx
1.
With
this notation,
{Xk\
kGK}
Sec. 7.7]
EXTENSIONS
233
Notice that as
{XkXk'H)cj,
[{XkH)(xk'H)}4>
is Xkw.
ixkH)4,{Xk'H)4>
{7.22)
kk'
where
k, k'
G K,
the rep-
resentative of the coset
XkXwH
a;fca:;fc'
Then, from
k k
we have
fc
Xkk'm,x,.x,,
where
mx^.x^,
k
H
&
We
suppress the x's and write mk.k- for

ajkajfc'
mX|^.X|^,.
Thus,
mk.k'
aJkk-mk.ic'
where k,k'GK,
(7.23)
Every element
gr
in
can be written uniquely in the form Xkh where Xk

of
& X, h G H.
To express the product
two elements Xkh and

follows:
Xk-h' as the
product of an element of
by an element of H, we proceed as
Xkh'Xk'h'
XkXr
Xk' hXk'h'
Xkk'tnk.k'Xk' hXk'h'
{7. 2 A)
Observe that
mk.kXk' hxk'h'
notation
h!"
So Xkr X and Xk^ hxr & H since fl" is a normal subgroup of G. G H. The right-hand-side of (7. 2 A) looks less complicated if we introduce the
ocr
for
hxr; then
Xkh-Xk'h'
Xkk'nik.k'h!' h'
{7.25)
G of i? by X that we have been inand by the images li of the elements h obtained by conjugation by the x^, i.e. by forming xj hxk'. One may conveniently think of the elements mk.k' in H as the images of a function m of two variables (coming from K) with values in H. In other words, we may think of m as a mapping from the cartesian product into H, where we use mk.k' to denote the image of {k, k') G under this mapping m. Continuing with this analysis, let us turn to the elements h^. For each k we have a mapping, ka say, of H into H, namely the mapping which sends an element fe in I? to the element h In a way then the group G is made up of two mappings:
It
specting
appears from equation is determined by the
{7.25) that the extension
mk.k'
KxK
KxK
GK
(1) (2)
m from KxK into H, a mapping a of K into a set of mappings

a
mapping
of
into H.
(The effect of ka
is
to
map
h to
h''.)
Indeed and a determine G up to isomorphism (see Problem 7.27). If we add enough conditions to these mappings, one can reverse the procedure we have been outlining and construct from H, and the mappings and a an extension G of hy K. (See A. G. Kurosh,
The Theory of Groups, Vol.

will
II,
Chelsea, 1960, translated by K. A. Hirsch, for details.)
We
(in
not tackle the general problem but
we
will
consider only a particular case
Section 7.7c).
Problems
7.26.
Let
group of order
above.
Solution:
be the dihedral group of degree 3. G is an extension of a cyclic group of order 3 by a cyclic 2. After choosing a suitable left transversal, find the mappings -m and a introduced
Using the notation of Section
3.4f
page
75, let
{<ri,
a^, ffs).
G. Since transversal for

<]
jG|
=6, G/H
Then
is
of order 2
and thus a
cyclic
group of order
Then, as can be easily checked, 2. Let {i, t} be a left
in G.
ta is
the identity
mapping
of
onto H, while ra sends a^ to
ct2i
"2 t
''i
^^^
"3 * "3-
234
Let G,
[CHAP.
7.27.
G be two groups, both extensions of H by K. Assume that H is actually a subgroup of both and G. Let X, X be transversals of H in G,G respectively. Let m, m and a, a be the mappings obtained above. Prove that if m = m and a = a, then G = G. G
Solution:
The elements of G are uniquely of the form x^h, where xêX, h&H, while the elements of are uniquely of the form x^.h, where icê.X. Let e G -^ G be defined by (xy}i)e = XiJi. Then # is a one-to-one onto mapping. To prove is a homomorphism, we consider the product of two elements of G.
(Xkhx^'h')0
h'
(aJfcfc'Wfc.fc.fe
h')a
{Xkrm^.k'h(k'oi)h')e
x^^.-.ni^^^'hik'aW
Thus
x^k'rn.k^^-h(k'a)h'
x^hx^.h'
(x,^h)e(x^'h')s
G=
G.
b.
The
splitting extension
Suppose G is as in Section 7.7a. Consider the particular case where nix.x' = 1 for all G X. By examining equation (7.22) xx' x" i.e. the product of two elements in X is again in X. Furthermore, we have 1 G Z. Let x G X, and let y be such that x~^H = yH; then xy H. Accordingly
X, x'
,
GX
xy
Ivfix.y
where mx.y G
However
G.
is
mx,;,
1,
Since < G, of the form xh,
XH
distinct cosets of
and so xy \. Thus x has an inverse in X, and so X is a subgroup of is a subgroup (Theorem 4.23, page 125). But every element of G x G X, h G H. Hence XH = G. Since distinct elements of X lie in in G and 1 G H, we have
XnH
split
{1}
Since and over H.
H are subgroups of G with XH = G, HnX = X is called a complement of H.

X
{1}
and
H<
H
G,
is
said to
as a transversal for if G splits over ff, we can choose any complement of is a subG, since two distinct elements of belong to distinct cosets of H. Now group of G, and it follows that if we define mg,x as in Section 7.7a with as transversal, then mx.x' = 1 for all x, x' in X.
Note that
H in
If
splits
over
H and X is a complement of H, then

G/H = HX/H ^ X/HnX ^
X
over
//,
In other words, G is an extension of with any complement of H.

It is
H by X; that
is, if
splits
G/H
is
isomorphic
convenient to introduce the following definition. We say that a group G is a of G isomorphic to Hi splitting extension of Hi by Xi if there exists a normal subgroup ^ Xu such that G splits over and G/H
Problems
7.28.
Prove that the dihedral group a cyclic group of order 2.

Solution:
of order 8
is
a splitting extension of a cyclic group of order 4 by
We
then
H is cyclic of
use the multiplication table for order 4 since

Oj
D
=
given in Problem 7.6(a), page 220. Let

02'
{1,01,02,03};
<*l
"^3'
"J
Sec. 7.7]
EXTENSIONS
235
Moreover H is a normal subgroup of G. This can be verified either by direct calculation, checking then rf-ifed G H, or by noting that H is of index 2 in D. Now letting that if d e D and X = {1, a^}, 04 = 1 and so X is a subgroup of D. Since a^ H, it follows that the cosets H, a^H
hGH
are disjoint.
Therefore, as there are 8 elements in
D = Hua^H
Finally
or
Hua^H, D =
{1}
XH
HnX
a splitting extension of
So
Z) is
by
as required.
7.29.
Prove that neither the dihedral group of order 8 nor the quaternion group of order 8 extension of a group of order 2 by a group of order 4.
Solution:
is
a splitting
Let E stand for either the dihedral group of order 8 or the quaternion group of order 8. Suppose is a of order 4. Then of order 2 with complement a splitting extension of a subgroup = {l,h}. If e&E, then e-^heGH. Since h ^ 1, normal subgroup of G. Now suppose e~^he = \. Thus e-i/ie = /i for all e&E. In particular if x&X, then x-^hx = h. Now E = XH. Since X is of order 4, X is abelian (see Problem 5.19, page 140).
is
H H
&
Now
suppose
e,f&E\
e
then
x'h',
= x"h"
x'x"h'h"
(x',x"
ex,
h',h"
SH)
abelian; then
Recall that every element of

ef
H commutes with every element of X and that X is

= x"x'h"h'
x'h''x"h"
x"h"x'h'
fe
Thus
E is
abelian.
is abelian.
But neither the dihedral group of order 8 nor the quaternion group of order 8 Hence we have a contradiction to the assumption that either of these groups is a
If
splitting extension of the type described.
Alternate proof:
HnX =
of
{1},
H< E
&nd
and by Corollary 5.17, page 145. a contradiction.
^ is a splitting extension of H by X of order 2 and 4 respectively, since X < E (as X is of index 2), it follows that E = XxH, the direct product
From
this
it
again follows that
is abelian,
thus producing
7.30.
2,
Prove that the quaternion group of order 8 but is not a splitting extension.
is
an extension of a group of order 4 by a group of order
Solution
We
use the multiplication table for
^,
the quaternion group of order 8 given in Problem 7.6(b),
page 220.
First let
group of and so
abelian.
^, and
is
Then is of order 4 and therefore of index 2. follows that J( is an extension of by J(/K. Clearly an extension of a group of order 4 by a group of order 2.
it
K = gpia^).
Thus
is
a normal subof order

2,
J{/K
is cyclic
Now
suppose JH splits over any subgroup K of order 4. Then ^/K is of order 2 and hence Therefore contains the commutator subgroup of J^ by Problem 4.68, page 116. In
particular,
K K contains a2
since
_]
-1
Now we must check that if x is any element except 1 and 02 of ^, then x is of order 4. This can be done directly, using the multiplication table for J?f. Suppose now, if possible, that J?f is a splitting extension of by X. Then the subgroup X is of order 2, say X = {1, x). But as we saw above, X is of order 4 since x = 1, x = a^. So the subgroup X is not of order 2. This is a contradiction and
so the desired result follows.
Alternate proof: If j?{ is a splitting extension of a subgroup of order 4 by a subgroup of <] ^. But every subgroup of the quaternion group is normal order 2, then Kr\X = {1} and (Problem 5.43, page 158). Therefore <i Jl. It follows, by Theorem 5.16', page 146, that Jl = X X. This implies JK is abelian, which is a contradiction.
K
7.31.
Is the alternating group of degree 4 a splitting extension of a group of order 6 by a group of order 2?
Solution:
By Problem
6.
Thus the
5.1, page 131, the alternating group of degree 4 does not contain a subgroup of order result follows.
236
[CHAP.
c.
An
analysis of splitting extensions
G is a splitting extension of hy with complement X. use as a transversal for in G, and m:r,x- = 1 for all Consequently each element g G G is uniquely expressible in the form
Section 7.7b,
Suppose now that
we may
H H
Then by
x, x'
G X.
Xkh
where k & K, h G
{7.26)
{7.27)
and equation
{7.25)
becomes
Xkhxk'h'
Xuk'h!"'h'
For each
is
kGK
the
mapping
h^h^
{h
G H)
3.57,
{7,28)
that the
an automorphism of H, for H and h'' = Xk' hxk imply, by Problem mapping h-^h" {hGH) is an automorphism of H.
if
<G
page
85,
Finally we remark that the automorphism
we
let a
be the mapping which assigns to each element
kGK
ka:
of H, then a
h^h^
{hGH)
is itself a homomorphism of into the group of automorphisms of H. (This explains why we used a left transversal in Section 7.7(a), namely so that the mapping a be a homomorphism.) We have only to prove that
{kk')a
kak'a
{7.29)
To
verify {7.29), let us take an arbitrary element
hGH
and apply the automorphism
{kk')a to h:
h[{kk')a]
= = = =
Xkk'hXkk{Xk'^
{XkXk')''^h{XkXk')
since Xkk'
XkXk'
hy
{7.27)
Xk^ )h{XkXk')
Xk-^xZ^ hXk)Xk'
Xk'^ [h{ka)]Xk'
[h{ka)]{k'a)
h[{ka){k'a)]
by the definition of the product of two automorphisms

{kk')a
Thus we have
{ka){k'a)
We now
What
replace A"' by h{k'a) in {7.27).
Then
{7.27)
becomes
{7.30)
Xkh'Xk-h'
is
Xkk'
h{k'a)h'
arrived at? We started from a group G which splits over H. - [1}. We were given an such that G = and implicit isomorphism of G/H with and we denoted the element in which corresponds to A; in iT by xu. Then we observed that the elements of G were uniquely expressible in the form Xkh with kGK, hGH. The way in which elements of G are multiplied was then computed by making use of the existence of a homomorphism aof into the automorphism group of and by applying {7.30). Therefore we may suspect that if we are given
the situation
we have
<}>
Then we chose a subgroup
X of G
XH
XnH
(a)
a group H,
a group K,
(6)
(c)
homomorphism
create a
is
then
we can
Z into the automorphism group of H, splitting extension of H by K.

a of
Indeed this
the case.
All
we have
to do is to reverse the process
To be
precise, starting with the data (a), (b)

i.e.
and
(c),
we
let
K and H,
we have described. be the cartesian product of
G =
KxH
{{k, h)\
kGK, hGH}
Sec. 7.7]
EXTENSIONS
define a binary operation in
237
We
according to the formula
(k,h){k',h')
{kk',h{k'a)h')
{7.30)
{7.31)
{7.31).
The reader
will note the strong similarity
between
and
by K, we will make the a group and a splitting extension of = {k, 1) and h = (1, h), similarity of equations {7.30) and {7.31) even more evident. Let Xk and define ka by h(fca) = (1, h{ka)). Then from {7.31) obtain
Before verifying that
is
Xfchxic'h'
Tikk'h{k' a)h'
This equation resembles equation {7.30) even more closely than {7.31) does.
We will now
(i)
prove that
is
a group and that
it is
H by K.
We
note
first
that {7.31) defines a binary operation in G.

defined
(ii)
The binary operation
by
{7.31) is associative:
{{k,h){k',h')){k",h")
=
= =
{kk',h{k'a)h'){k",h")
{{kk')k", {[h{k'a)h']{k"a))h")
{{kk')k", {[{h{k'a)){k"a)][h'{k"a)]}h")
{7.32)
Now we work
out
{k,h){{k',h'){k",h"))
=
=
{k,h){k'k", h'{k"a)h")
{k{k'k"), [h{k'k")a][h'{k"a)h"])
By
(c)
above, a
is
a homomorphism, and so {k'k")a
k'a
k"a.
Thus
{7.33)
{k,h){{k',h'){k",h"))
= =
{k{k'k"), [h{{k'a){k"a))][h'{k"a)h"]) {{kk')k", [{h{k'a)){k"a)][h'{k"a)h"])
The
{7.31) is
(iii)
associative law immediately yields the equality of {7.32) an associative operation.

in G: (1, 1).
{k, h)
and
{7.33).
Therefore
There exists an identity

the identity of
(1, 1) is
Notice, of course, that the left-hand 1 in

(1, 1) is
(1, 1) is
while the right-hand 1 of
the identity of H.
To check that
an
identity, let
G.
Then
{l,r){k,h)
=
=
{k,l{ka)h)
= =
{k,h)
since ka is an
automorphism of H, and so maps the

{k,h){l,l) {k,h{la)l)
1 of f? to itself.
Similarly
{k,h)
since la is the identity

(iv)
automorphism of H, and so leaves
identically fixed.
Finally we must check that every element of G has an inverse. claim that ,, ,_,,,, .. i(fc 'a)) ', h (fc
.
Let
{k, h)
G.
We
is
the inverse of
{k, h).
To prove
this,
we simply observe
that
{k,h){k-\h-\k-â))
since k~â
is
{l,[h{k--'a)][h-\k-â)])
(l,(M-i)(A;-a))
an automorphism of H.
Thus
{k,h){k-\h-\k-â))
Similarly
= =
(1,1) (1,1)
{k-\h-'{k-'a)){k,h)
(A;"',
We
have thus verified that
h~^{k~â)) is the inverse of {k,h).
238
(i),
[CHAP. 7
(ii), (iii)
and
(iv)
above establish in every detail the proof that
is
a group.
this,
Next we verify that
G =
is
H by K.
of
To accomplish
\
put
H
Then
it is
{(1,
h)\hGH}
and
K =
k -
{{k, 1)
k G K}
easy to prove that
H and K are subgroups

h
-^ (1,
G and
that
h)
and
{k, 1)
are isomorphisms of
H onto H and K onto K respectively. Now to prove that 5 is a normal subgroup of G, observe
g
again that
if
gG,
then
(k.h)
(k,l){l,h)
If (1, h')
G H,
then
g-\l,h')g
= =
{k,h)-\l,h'){k,h)
{ik,l)il,h))-\l,h'){ik,l){l,h)) [il,h)~\k,l)-^]{l,h')(ik,l)il,h))
l)-i(l, h'){k, 1)
=
Since
(k,
l)-i
{k-\
1),
we have
(A;,
(1, h'{ka)).
Thus
g~\l,h')g
since the_product of elements of
{l,h-'){l,h'(ka)){l,h)GH
belongs to H.
Hence
is
normal
in
as claimed.
Clearly
Hr\K=
{(1,1)}
and
G =
as we saw earlier, since (k,h) = (fc, l)(l,/i). data (a), (b) and (c) a splitting extension G of ting extension of by K.
KH
Consequently
we have
iiT.
H with complement
Therefore
constructed from the G is a split-
The group
that
we have
constructed
is called
the splitting extension of
by
via
a.
We
follow.
emphasize the importance of the above discussion and the related problems which
Problems
7.32.
Construct a non-abelian group of order 6 as a splitting extension of a group of order 3 by a group

of order
2.
Solution:
Recall that if we are given (a) a group H, (6) a group K, (c) a homomorphism a of Z into the automorphism group of H, then we can construct a group from this data as follows. Consider the set G of all the pairs (fe, h) {k B K, h& H) and define a binary operation in G by
{k, h)(k', h')
(kk', h(k'a)h')
Then G becomes a group which is a splitting extension of the group (the cyclic group of order 3) and the group the homomorphism a. This means in the first place that we phism group of H, the cyclic group of order 3. Now if is
H by
(the cyclic
So in the case at hand, we have group of order 2). We need need to know more about the automorgenerated by h, then
K.
H =
(check this).
{l,h,h^}
into its inverse
is
The mapping which sends each element of an abelian group

So
if
ri
:
H^H
an automorphism
is
this
automorphism,
Sec. 7.7]
EXTENSIONS
=
and
so gp(v) is cyclic of order 2.
239
Now
7j2
I,
Thus the groups
and gp(v) are isomorphic, and
accordingly we can take a to be the isomorphism of via a. To see how some of the elements of ot hy
onto gp(v)- Let G combine, let
be the splitting extension

{1, k}.
K=
Then
(k,h){k,l)
(k^,h(ka)\)
(fe, ;i)(fc,
(1,A2)
(fe, 1)(A;,
Now
is
(fc, l)(fc,
h)
(A;2,
l(M'i)
(1. K),
and so
1)
h).
Thus G
is
non-abelian and
an extension of
H by K.
of order 2
7.33.
Are there non-abelian groups

Solution :
X 777?
be cyclic of There are non-abelian groups of order 2 X 777. To produce one such group, let has an automorphism t of order 2, namely the mapping t which sends every order 777. Then element of to its inverse. So there is a homomorphism a of K, the cyclic group on fe, of order 2, into the automorphism group of H, namely the one which sends fe to t. Let G be the splitting extenby via a. Then G is the required group. sion of
7.34.
Construct two non-isomorphic non-abelian groups of order 168.

Solution:
gp{hj) be cyclic of order 84 and K^ = gp(k^ cyclic of order 2. Let a be the homointo the automorphism group of H^ defined by fcja iCi^h^ '. Then Gj, the splitting extension of H^ by K^, is of order 168. The center Zj of Gj is of order 2, consisting of (1, 1) and (1, h^) which can be easily checked by direct calculation.
Let Hi
morphism
of
Ki
and K2
Now we construct a second group = gpikz) to be cyclic of order

usual,
of order 168.
4.
Here we take H2
fl'p(^2)
to be of order 42
As
H2 has an automorphism
t:
of order
2, i.e.
i
the
0, 1,
mapping
...,42
defined
by
h2^h2\
=
= =
Let
j8
be the homomorphism of K2 onto gpir) defined by

fc'2/3
T',
0,1,2,3
Thus A;2i8 = Let G2 be the splitting extension of H2 by K2 via /?. Then, as the reader may check by direct calculation, the center Z2 of Gj consists of (1,1), (1,^2^), (A:|, 1), (fc|, /t|') and is therefore
(.
of order
4.
Now both Gj and G2 are non-abelian since \Zi\ = 2, 1^2! = ^- Moreover if G^ and G2 are morphic groups, they have isomorphic centers. Therefore Gj and G2 are not isomorphic.
Construct
cyclic
iso-
7.35.
all
possible groups of order 30

3.
which are extensions of a
cyclic
group of order 10 by a
group of order
Solution:
The solution of
this
gp(h) of order 10. hr are therefore
problem requires knowledge of the automorphism group ot the cyclic group If t is an automorphism of H, then hr is of order 10. The possibilities for
hr
,
h^,
hr
K!,
hr
h^
(i
Let
t|
Tj
be the automorphisms defined by the possibilities listed above
1, 2, 3).
Now
Tj
1,
t\
1,
thus none of the automorphisms of is of order 3. So the only possible homomorphism a of a cyclic group of order 3 into the automorphism group of is the one which sends every element of onto the identity automorphism. The resultant splitting extension oi hy is then abelian (indeed it is isomorphic to the cyclic group of order 30).
1;
7.36.
Construct a non-abelian group of order 222 by using splitting extensions.

Solution:
Form
via the
homomorphism taking
= gpik) of order 2 a splitting extension of a cyclic group of order 111 by a group fe to the automorphism which sends every element of into its inverse.
240
[CHAP.
7.37.
Construct a non-abelian group of order p3 for each prime p by using splitting extensions.
Solution:
(Hard.)
Let
p.
H be the direct product of a cyclic group gp(a)

of
Each element
of order
p with a
cyclic
group gp(h) of order
is
uniquely of the form

a^h^
with
is
0-r<p,
<p
Let t: a^b^ - arfes + r. We will show that t be any integers. We claim that Let
an automorphism of
of order p.
First let
and n
(a'"5")7-
= amjm + n
m~
m'
qp,
n'
(a'nb")T
- m' < p, n' < p. Then + sp where 6' + "' = = a"' a'"6"''b"' = a^h^h^ = a^jm + n
To verify that
t is
a homomorphism, observe that
Clearly
set
automorphism.
{1}, and so t is one-to-one. It is easy to check that r is also onto. Thus t is an Note that otP = ahf - a, so that t-p acts as the identity on a and 6, which form a of generators of H. Hence t is of order p.
Ker r =
gp(k) be of order p. The mapping a k' -^^ t' is an isomorphism. Then we form Now let the splitting extension oi hy via a. This gives a group G of order p^. G is non-abelian since
:
(1, a)ik, 1)
(/c,
a6),
but
{k, 1)(1, a)
(A;,
a).
d.
Direct product
Consider the special case of a splitting extension G of a group by a group K via in which a takes K onto the identity group of automorphisms of H. In this case G consists of the pairs {h, k) (h G H, k G K) with binary operation given by (h, k){h', k') = {hh', kk'). So G is, in the terminology of Section 5.3a, page 143, simply the external direct product of and K. The obvious usefulness of this construction is that we do not require any knojwledge of the automorphism group of to construct the direct product. Notice that if ff = {(1, h) hG H] _and K = {{k, 1) k G K], then G is the internal direct product of its subgroups and K, again in the sense of Chapter 5. We will not pursue this concept of direct product here any further.
a
homomorphism
7.8
a.
THE TRANSFER
Definition
a group with an abelian subgroup A of finite index. The transfer is a of G into A. The use of such transfer homomorphisms has been important in the theory of finite groups. Here we will examine one application of the transfer and briefly mention another (at the end of Section 7.8d).
Suppose
is
special
homomorphism
G into A, choose a right transversal X of G in A. We repeat an abelian subgroup of G of finite index n, say. Therefore \X\ = n. Recall that if g GG, then g is the element of X in the coset Ag. Let Xi, Xi, .,Xn be the elements of X; then if g GG, we define a mapping t of G into A by
To
define the transfer t of
is
that
gr
It is clear
Xig{xTgy^
X2g{x^)~^
to the
1,
.
Xngix^)"^
that gr
GA
since, as Xig
and x^ belong
same
coset of A,
Xig{x^)-^
GA
{i=
.,n)
This mapping t is the homomorphism of G into A mentioned above; it is called the transfer of G into A. There are two items to be verified: (1) t is a homomorphism and (2) t is independent of the choice of the transversal X.
Sec. 7.8]
THE TRANSFER
t is
241
b.
Proof that
a homomorphism
We
{gh)T
compute
(gh)T,
where g,h GG:
= =
{xigh(xigh)-')
{x^ghixigh)-^)
{Xngh{Xngh)
')
{Xigh{xWh)-^)
{X2gh(x^)-'')
{xngh(x^y')
(since
x^ = Wh
by
(7.3),
page 219)
ixig(xW)-' -xTghixigh)-^)
ix2gim)~'
'
xlgh{x^h)-^)
(a;9r(^)-i
x^h{x^h)-^)
lies in
Now every one of the elements

commute.
{gh)T
Xig{x^)~\ x^{xjgh)-^
A; since
is
abelian, they
must
So we can rewrite (gh)T in the form
[xig{xW)-^' X2g{x^)~''
Xng{x^T^]
.^ ^^.
[xTgh{xWh)-^
x^hiWgh)-''
.
x^h{x^h)'^]
But observe that

is
Xi^x^ {i-l,
.,n)
a permutation of
(Section 7.4a).
This means that
xlgh{xlgh)-^
xWh{x^h)-^
x^h{x^h)"^
simply consists of the n elements Xihixjiy {i = l, ...,n) multiplied together in some order. Since A is abelian, the order of such a product is immaterial. Thus
xlgh{x^h)~'
x^h{x^h)-'
from equation
x::gh{x^y^ = Xihixjiy
Xihixji)-'
Xnh{xJi)-'
=
Then
for
it
hr,
by definition
follows
{7.3i.)
that
igh)T
{gr){hT)
all g,
h in G.
Thus
t is
homomorphism.
c.
Proof that
t is
independent of the choice of transversal

t is
We now
prove that
independent of the choice of the transversal X.

of the product
The proof depends on an analysis

fifT
xig{x^)~^
is
X2g{x^y^
Xng{x^y^
where again
7.4a that the
set
{xi, X2,
..,Xn}
=X
is
mapping Xi^ x^
a right transversal of A in G. We recall from Section a permutation of X. Now every permutation of a finite
So, after
relabeling the elements of
can be written as a product of disjoint cycles (Theorem 5.26, page 167). if necessary, we can assume that
xig
X2,
X2g
xz,
.,
X]t-ig
Xk,
x^g
Xk+ig
Xk+2, Xk+2g
Xk+3,
...,
Xk + i-i
x\ = Xk +
.
i,
Xk+ig
Xk + i
Xn-m+ig
(Note that k
gr
Xn-m + 2,
Xn-m.-l-2g
Xnm +
is
3,
.,
Xu-ig
Xn,
Xng
Xn-m+l
+ l+

+m = n,
X2g{x^)~'^
where n
the index of
in G.)
Then
{xig(xW)~^
Xkg{x^)~^)
'
(Xk+ig{xk+ig)"^
Xk+2g{xk+2g)'~'^
*
Xk+ig{xk+igy^)
{Xn-m+igiXn-m+ig)"'^
Xn-m + 2g{Xn-ra + 2g)~'^
Xn{gX^)~^)
{xigx^^
X2gx3^
Xkgx^^)
{xk + xgxk+2
Xk+2gxki3
Xk + igxk
'
{Xn-m + igXn-m + 2
Xn-m + 2gXn-m + 3
XgXn-m+l)
242
gr
[CHAP.
and thus
Note that
{x^g^xT')
{x^^.g^x^^l,)
(x- + i^"a;;i+i)
{7.35)
x,g^xT'_ = {x,g{xlg)-^) {x^gixTg)-') {x^g{x^)-') G A, since it is the product of factors x,g{Xig)~^ which belong to A. Similarly, x,,,g^xlU, ..., Xn-mîx'-Xn'-m.i e A.
We
Lemma
are
7.7:
now
able to prove
Let
be an abelian subgroup of
.
.
finite
{xi,X2,
.,x},
Y=
{yi,y2.,
.,y,,}
index in a group G and be two left transversals of
let
X=
in
G.
Furthermore
9r
let t
and 7 be mappings of
into
defined respectively
by
xig{xlg)-^
X29{x^)-'^
a;flf(^)-i
(where g
G)
Proof:
Now
It
if
= yig{y^gy y^giy^y ynQiyTg)-' (where gGG) where, if h E G, h denotes the element of X in the coset Ah and h denotes element of Y in the coset Ah. Then T T We may assume on suitably reordering Y that = OiXi (Oi e A) = X;, then Ayj = AajXj = Axj = AaJigr = AoiXiflr = Ayig. This means that
and
f7
the
2/i
m
fl'~
Vi
follows therefore as in equation {7.35) that if g
EG,
then
iyn-m+ig"'yn-m+i)
.
iyig^Vi^)
ivk+ig^ylii)
{7.36)
lie
But
yi
(kXi,
in the abelian
and we know that the elements xigHi\x},+ig^xllu subgroup A. Therefore

Vig^Vi^
Vk+ig^ylli
.,Xn-m+ig"'Xn-mî
ai{xigHi^)ai^
Xig^xr^
au+i{xk + ig^xlli)a^\i
Xk+ig'-Xkli
yn-nH-lX"'yn-m+l
an-m+l{Xn-m + ig"'Xn-m+l)(K-m + l
Xn-m+ ig'"x'ii-m+
Thus
it
follows
from
(7.35), {7.36)
and the above remarks that gT
g7.
This lemma establishes that the transfer homomorphism t is independent of the choice of transversal. Accordingly we may speak of the transfer of G into A.
d.
theorem of Schur
I.
Using the transfer, we now prove the following important theorem of
Schur.
Theorem
7.8:
Let G be a group whose center G' of G is finite.
A
. .
is
of finite index.
Then the derived group
Proof: Suppose \G/A\ - n. Let X = {xi, .,x} be a transversal of be the transfer of G into A. Now if g G G, then by equation {7.35)
in
G and
let
gr
=
is
Xig^X^^
Xh-ng^xZll
Xn-m + ig'^Xn-m + l
But Xig^xZ^ e
and
the center of G.
xig''x~^
So
=
g\
x;'^{xig''x~^)xi
g^
Similarly,
Xk+ig^Xkli
.,
x-m+ig'"x'^-m+i
g""
Sec Y
8]
THE TRANSFER
243
Thus, for each
gGG,
qt
fir".
Notice that the image of
under
is
by Theorem 4.18, page 117, GIK page 116, this implies G' qK.
a subgroup of A and is therefore abeUan. Then abelian, where K is the kernel of t. But by Problem 4.68,
t is
Now
Theorem
as
G/A
is finite,
so is
G'A/A.
This means that G'/G'nA
is also finite since
by
4.23,
page 125,
G'A/A ^ G'/G'nA
{7-37)
G G, Since G'/G'DA is finite, G' itself is finite if G'nA is finite. Since gr = g" for every g G'nA have every element of the kernel has finite order. It follows that the elements of show that G'nA is finitely generated. Assuming this true for the We will finite order. moment, it follows that G'nA is finite (Problem 6.44, page 198). Thus G' is also finite. This completes the proof of Schur's theorem but for the verification that G'nA is finitely
generated.
We
g
accomplish this by showing
first
that G'
is finitely
axi
and h
bxj,
where a,bGA, Xu Xj e X.
Then
generated. If g,hG G, then since A is the center of G, we have
g-^h-^gh
xrâ'^x^^b-âxibxi
xT^xr^XiXja-âb'^b
x^-^xf^XiXj
is finitely
This means that there are at most n^ distinct commutators in G. generated since it is generated by commutators.
Finally
Therefore G'
G'nA
is finitely
(Theorem
It is
7.4,
page 227).
generated since it is of finite index in a finitely generated group This completes the proof of Schur's theorem.
worth noting one fact that emerged in this proof: the transfer into the center is simply the mapping that takes each element g to g" where n is the index of the center in G. We end our discussion of the transfer by mentioning that if all the Sylow subgroups of finite group G are cyclic, then G is metacyclic (i.e. an extension of a cyclic group by a cyclic a group). This theorem can be proved by using the transfer. The proof is not too complicated; however, it is lengthy and will not be given here. Reference to a proof may be found in
Section 5.2a, page 139.
We re-proved Cayley's theorem, namely that every group is isomorphic to a group of permutations. The ideas that arose from Cayley's theorem were generalized. In particular we called a homomorphism of a group G into the symmetric group on a set a permutational representation of G (on X). Permutational representations were roughly classified. We explained an important permutational representation of a group G called a coset representaThis representation allowed us to provide a variation of Cayley's theorem due to tion. Frobenius. Then we used Frobenius' theorem and the coset representation to prove three theorems: (1) a subgroup of finite index in a finitely generated group is finitely generated;
the number of subgroups of fixed finite index in a finitely generated group is finite; (3) if G a finitely generated group whose subgroups of finite index have only the identity subgroup in common, then every homomorphism of G onto itself is also one-to-one, i.e. an
(2)
is
automorphism.
We
of
called a
group
G such that This because of our discussion of both coset representations and Frobenius' theorem. analysis of extensions was specialized to splitting extensions, where we provided a method and K. of constructing a splitting extension of two groups Finally we defined a special kind of homomorphism of a group into an abelian subgroup, called the transfer. We then used the transfer to prove that the derived group of a group whose center is of finite index is finite.
G an extension of a group H by A: if there is a normal subgroup H G/H = K and H^H. An analysis of this situation was made simpler
244
[CHAP. 7
PERMUTATION REPRESENTATIONS, COSET REPRESENTATIONS, FROBENIUS THEOREM
7.38.
Let Z) be the dihedral

explicitly
(1)
ôup
of order 2w and let C be its cyclic normal subgroup of order n.
Find
the representation of Z) onto itself given by Cayley's theorem,
(2) (3) 7.39. 7.40.
a coset representation of D using C as the subgroup, the representation provided by the Frobenius theorem.
Give a permutation representation of
of degree 2.
Find a faithful representation of
GxH
is
on n
+m
letters if
GcS,
He S^.
group by a group of order
2.
EXTENSIONS
7.41. 7.42.
Construct a non-abelian group which
an extension of an
infinite cyclic
Prove that
is
a non-abelian group with an infinite cyclic normal subgroup of index 2, then a splitting extension of an infinite cyclic group by a group of order 2.
if
is
7.43.
Prove that there are precisely three non-isomorphic extensions of an

of order
2.
infinite cyclic
group by a group
7.44. 7.45. 7.46.
Prove that an extension of a
cyclic
group of even order by a group of order

36.
3 splits.
Construct a non-abelian group of order
Construct five non-isomorphic groups of order 5^ Let
3.
. . .
7.47.
i.e. D consists of the sequences a = {aieZ). If b = = ., a-î + 6_i, a,, + 60. "i + î, ., 6_i, 69, ftj, ., define a + D is an abelian group under the operation of addition of sequences. For each integer n define a mapping of D by putting oq: = where 6j = ;_. Prove that ., 6_i, ftp, î, -, (1) a is an automorphism of D, (2) aa = a + , (3) A = {ai\ i e Z} is an infinite cyclic group generated by aj.
be the set of infinite sequences of integers,

. .
. .
O'-i.a-o.O'i,
fe
7.48.
Let
mapping that sends

which
is
W be the splitting extension of D of Problem 7.47 by the infinite cyclic group C = gp{c) via the c to Prove that W a non-abelian group. Find a proper subgroup of W
a^. is
isomorphic to
W. Prove that
W/W
is infinite cyclic.
(Hard.)
TRANSFER. MAL'CEV'S THEOREM, MARSHALL HALL'S THEOREM.

7.49.
Prove that
into
sends
7.50.
Prove that
of
A is an abelian subgroup G into 1. if G is a finite group whose

if
is
of finite index in a simple group G, then the transfer of
center
has order co-prime to
its
index, then the transfer
into
onto.
(Hard.)
7.51.
Prove that an
infinite
group with a subgroup of
finite
index
is
not simple.
7.52.
said to be residually finite if the intersection of all its normal subgroups of finite index is the identity. Prove that an extension of a residually finite group by a finite group is residually finite. (Hint: The preceding problem gives a clue.)
is
group
7.53.
N = gp{a)
7.54.
be a cyclic extension of a cyclic group of order n by a cyclic group of order m. and G/N = gp(bN). Prove that b~âb <P, where j is co-prime to n and 6 (where j and k are integers). Prove that jk = k modulo n.
cyclic extension of a finitely
Let
Let
a'=
Prove that a
is residually finite.
(Hint:
Use Marshall
(Hard.)
.
. .
generated residually finite group (defined in Problem 7.52) Hall's theorem. Then use the fact that the automorphism
are subgroups of G such that Gj is not finitely generated.
group of a
7.55.
finite
if
group
is finite.)
Prove that
G2CG3,
7.56.
.,
G is a group and Gj, G2, then uGj is a subgroup of
- Gg,
G1CG2, G^
G3,
G which
be a residually finite group (defined in Problem 7.52) and suppose every subgroup of G is Prove that if is a subgroup of G such that H/N = G for some normal sub= {1}. (Hint: Use Problem 7.55.) group of G, then
Let
finitely generated.
7.57.
is
finitely
|G'|
generated group, every element of which has only a
finite
number of conjugates.
Prove that
7.58.
<
/
<.
I
Hint:
C(ffi)
Z(G)
if g^,
...,g^ are the generators of G.
G is a group in which every element has only a finite number of conjugates. element of G' is of finite order. (Hint: Use Problem 7.57.)
Prove that every
chapter 8
Free Groups
and Presentations
begin with a property of the infinite cyclic group and generalize this property to define free groups. We ask questions similar to those we asked in Chapter 4 concerning cyclic groups:
We
(1)
(2)
Do
free groups exist?
(3)
(4)
When are two free groups isomorphic ? What are the homomorphisms of free groups ? What are the subgroups of free groups ?
In answering (3) we will learn that every group is a homomorphic image of a free group. This provides a new way of describing a group, i.e. as a factor group of a free group. Such a description of a group is called a presentation.
8.1
a.
ELEMENTARY NOTIONS
Definition of a free group
Recall that
if
is
a group and
(0)
a.
subset of G,
gp{X)
If
x\^
= {xl^---xl^\x,GX,er=l]
e.
x^ and
y\^
i/^"
are two products with x^,y.
they are said to be identical if w = w, x^ y^ and are said to be different if they are not identical.
It is
tj^
for
G X and i = 1,
e^
=
. ,
1,
m.
= 1, then Two products

r^^
a;^" can give rise to the easy to see that two different products of the form xl^ = {x, y}, then xy and xx'^xy are different products For example, if To avoid rex^^ but they give rise to the same element of G, i.e. xy. dundancy, we introduce the concept of a reduced product:
same element of G. of the form x\^
A
if
product
Xi'
x'x,
where
e^
= l and
x^
X,
is
said to be a reduced X-product
x.
x.,,
implies t.=(...,.
Synonyms
product in X.
for reduced X-product are reduced product {X being understood) and reduced
{x,y}; then xy, x~^yxyx~^ (Examples of reduced products are easily given. Let and x''^yyxy~^ are reduced products. However, yxyxx~^ and x~^yxyy'^ are not reduced
products.)
X=
= {w\w = lorw = a reduced product in X}. If Clearly R Q gp{X). Let {w\w = lorw a reduced product in X} = R. Proof: = 1. u G gp(X), then u = x\^ x'^ where x^G X and We proceed to show that u G R. If k = 1, u is a reduced product in X, and so u G R. x^^GR for n 1. Suppose k = n. If Assume then that any product If ^t is not a reduced product, there u-x[^ x^^ is a reduced product in X, then u G R.
Lemma
8.1:
gp{X)
c.
a;^'
fc
245
246
FREE GROUPS AND PRESENTATIONS

an integer
[CHAP.
i such that x. = x,^^ and e. = -e.^^. Suppose n>2; then we can delete m as a product involving n-2 elements of X. By the inductive hypothesis, u then belongs to R. If = 2, then u = x^x'i, where x^ = x^ and c, = -c^, from which u = lGR. Thus gp{X)cR and accordingly gp{X) ^ R.
exists
ajj'ajjVV to obtain
Consider
in {x} are of
now the
two
infinite cyclic
group generated by the element

x''
x.
The reduced products
kinds:
or
x~^x~^
X"'
a;"""
a positive integer. From what has been said about the infinite cyclic group, if and n are integers, x = x" implies that = n. Thus different reduced {a;}-products give rise to different elements. (This is by no means the usual situation. For example in a cyclic group of order 2 generated by y, we have yyyy = yy.) G is said to be freely generated by {x}. More generally we have the following definition:
r
is
where
we know that
A
set,
(i)
group and
is
said to be freely generated
by the
set
XcG
if
X ^ 0,
the
empty
gp{X)^G;
two
different reduced X-products define
(ii)
two
different nonunit elements of G.
Notice that it follows from (ii) that if x G X, x-^ ^ X. For if not, there exist x and both elements of X, with y-^ = x. But then x and y-^ are two different reduced Z-products which are equal. It also follows from (ii) that 1 ^X.
y,
A set X of generators of G satisfying A group G is free if it is the identity group

also said to be free
facilitated
equal. All we need to express each of the products in reduced product form. If the reduced products are distinct the elements are not equal. This process of expressing an element as a reduced
on X. If G is by the fact that we know exactly whether two Z-products are
often called a free set of generators of G. possesses a free set of generators. G is a free group freely generated by X, then the study of G is
(ii)
is
or
if it
to do
is
Z-product can be carried out in a
finite number steps. To illustrate, let F be freely generated by {a;, i/}. Are the products f xyx~^y~^xxx~^y''^y^ and g = xyy^y^ equal? We convert / to a reduced product by deleting inverse pairs (i.e. two adjacent inverse factors).
Thus
/
= xyx
^y
^x{xx ^){y ^y)y^
xyx~''^y^'^xy^
= xyx~^y^^xyy
= xyy. Hence as / and g are equal to different reduced products, they are not The fact that we can determine in a finite number of steps whether or not two elements are equal is often expressed by saying that the word problem is decidable for free groups. (The interested reader may find more details in, e.g., J. J. Rotman, The Theory of Groups, AUyn and Bacon, 1965.)
Similarly g
equal.
Problems
8.1.
= {x, y, z}. (a) Write down three distinct elements of G. be freely generated by the set xyz{yz)-^ a reduced product? Is xyy~^z~^yx equal to xz~^yzz^^xl (d) Express (c) x2y3(jf3^2-)-iy-3 and {xzy)~^xzy^ as reduced products.
Let
(b)
Is
Solution:
(a)
(1)
X, y, z,
(2)
x, x^, x^,
(3)
x, xy,
xz
In fact any three different reduced products in {x, y, z} are distinct elements of G.
(b)
No, because it is not written in the required x^ x^ form, since it involves (j/z)-*. However, even if we replace {yz)^^ by its equivalent z~ij/~i to get xyzz-'^y'^, this, though in the form x^ x^, is not a reduced product because of the inverse pair zz~'^.
Sec. 8.1]
ELEMENTARY NOTIONS
Yes, for on expressing each as a reduced product
247
(c)
we
get xz~^yx.
(d)
x^yHy^x^)-^y-^
(xzy)-^xzy^
= =
x^y^'x-^y-^y-^
y-^-'^x-'^xzy^
= xxyyyx-^x-^y-^y-^y~^y-'^y~^y^'^ = y-H'Hy^ - y'^y^ = y

is
8.2.
Prove that
Solution:
if
is
a free group,
= {1},
then the infinite cyclic group
a subgroup of G.
Suppose
x-x
8 are
G is freely generated by X. If x E. X, consider 9p({x}). This is X'', with r a positive integer, is a reduced product in X. Hence positive integers implies r = s. Thus ffp({a;}) is infinite and is infinite
X
a cyclic group.
x^
Now
',
cyclic
where r and and the result
follows.
8.3.
Prove that
abelian.
if
is
freely generated
by X, where
contains at least two elements, then
is
not
Solution:
so
There exist two distinct elements xy = yx. But this implies that G
finite
x,
is
y of X.
Now
xy and yx are two different reduced products,
not abelian.
8.4.
Prove that a
Solution:
group
is
not free
if
G #
{!}.
In Problem 8.2
we proved
infinite cyclic group.
Consequently
that except for the identity, every free group has as a subgroup the if G is free it must have the infinite cyclic group as a subgroup.
This
is
absurd.
8.5.
The
direct product of
two
infinite cyclic
groups
is
not a free group.
Solution:
The direct product of two infinite cyclic groups is abelian. But we have proved in Problem 8.3 which contains at least two that a free group is not abelian if it is freely generated by a set elements. Hence if the direct product of two infinite cyclic groups is free, it is freely generated by = {x}, i.e. it is infinite cyclic. But a free abelian group of rank two is not cyclic. This some set follows immediately either from the uniqueness of the type of an abelian group (Theorem 6.21, page 197) or directly.
8.6.
Prove that a free group freely generated by

of the identity element).
with
\X\
has no center
(i.e.
its
center consists
(Hard.)
Solution:
Suppose
G
J*.
Kji
a;
Let y
has an element & X, y =
z '^ 1
x^.
in its center.
Let z be expressed as a reduced product
Consider
yz
yx^^
This
is
a reduced X-product.
On
the other
hand
zy
If
x^^
x^v
a reduced product and so clearly zy yz as yz begins as a reduced product with y but zy begins with x^ = y. If " = y~^ then for re > 1 (for otherwise Xj = 2/ contrary to the choice of y),
x^=y~^, then
x^^
x^y
is
zy
= 
and again
the center,
it is
clear that zy
we must have
zy
and yz are two different reduced products, so zy = yz. = yz. Therefore this contradicts the assumption that
But as
z is in
has a center.
8.7.
Generalize Problem 8.2 by showing that if F^ = gp(xi, {xi, ., ;} freely generates

. . .
F
.
is
freely generated for

t
by
n.
{a;i,
.,;;}
with
>
1,
then
ajj)
1, 2,
. ,
Solution:
Put Xj = {xi, Kj}. By definition, X^ generates Fi. We have only to prove that two different reduced Xj-products define different elements of F^. But a reduced X'j-product is obviously also a reduced Z-product and two different reduced X-products define different elements of F. Hence the result follows immediately.
. .
248
[CHAP.
b.
and only one way as a reduced product

to this free set of generators to be 0. For example, if is
Length of an element. Alternative description of a free group Suppose F is freely generated by a set X. If feF, f^l, it can be expressed
in one define the length of f with respect to be n. The length of the identity is defined conventionally freely generated by J^ = {x,y}, then the length of x^y^x^'
xl'
x'-.
We
is 5,
because
x^y^x
'
= xxyyx~^
A very useful technique in arguments involving free groups is to prove results by induction on the length of elements (see, for example. Lemma 8.9, page 261).
Lemma
8.2:
A
(a)
group F is freely generated by a set X - 0, the empty set, if and only if gp{X) = F, and (6) no reduced Z-product is equal to the identity element.
Proof: Assume first (a) and (b) above. To show F is freely generated by X we must show that two different reduced products in X are not equal and are not the identity. Let x[^ <" and y\^ = 1, and x^,y^ G X) be two different reduced y"^ (where c. =- 1, products. Suppose they are equal. We can assume without loss of generality that ar^" ^ y^^; a;^"_-i and i/]' for if <" == 1/^, xl"i/^"L-i are two different reduced products'! Since yvrn and <" = xl^ x^n^y^h... it follows that yy,
,;.

xl' 1
xl"-^ n
y1' -^ 1
w""*-! i'm 1
If
again

a;^Ti>
2/l"L~l^
we can
delete them.
But we cannot continue

<"y-'''"

indefinitely this
as Xji
x'^
and
y]^
It
y"^ are assumed to be different reduced products.
way So we may as-
sume
1.
a;^"
y"^.
follows then that xl^

1.
But by
j/f
is
a reduced product equal to

it
(6) it is
not
Hence we have a contradiction, and so
follows that
is
freely
generated by X.
If
on the other hand
is
since this is entailed
by the
definition of a
given freely generated by X, do (a) and (b) hold? group freely generated by a set X.
Of
course,
Problem
8.8.
Let
(iii)
F
/
be freely generated by x-'^yxx^^y^x^z~^.
X=
{x,y,z}.
Determine the length of
(i)
1,
(ii)
xzyz-'^,
and
Solution:
(i)
The length
of 1
is
taken to be
is 4.
0.
(This is just the convention mentioned earlier.)
(ii)
The length of xzyz~^ The length of

/
is
(iii)
calculated by expressing
it first in
reduced form, that
is,
x~'^yx~'^yyxxxz~^
Hence the length of /
is 9.
c.
Existence of free groups

is:
An obvious question they do.

Theorem
8.3:
Do
free groups exist?
Up
to
now we have
tacitly
assumed that
Let n be any positive integer. by a set of n elements.
There exists a free group freely generated
Proof: Since every group is isomorphic to a subgroup of the symmetric group of some natural to look for a suitable subgroup of some symmetric group in order to find a free group of rank n.
set, it is
Sec. 8.1]
EXISTENCE OF FREE GROUPS

is
249
Our plan
as follows:
we
that
shall introduce a set
consisting of certain ordered 1-tuples
of integers, ordered 2-tuples of integers, of permutations of
and so
on.
We
.
then choose a set
X=
{6i,
.,6n]
T such
X freely generates gp{X).

. .
The elements of T are the ordered m-tuples (ri, ra, nonzero integers with n + n+iTÔ for i = l, ...,to (1,2) T and (0, 1,2, -2) T.
rm)
with
ri
and
r^,
.,rm
1.
Thus
(0, 1, 2, 3) e T
but
We
(ri,
. . . ,
define
rm)
(i)
now for each T we define (ri, rm)ei - (ri,

.
integer
1,
.,n a permutation
di
of
as follows.
If
r,
i)
if if
(ii)
(n,
.,rv^Qi
Oi is
(ri,
.,rm-i)
^ -i. r =
i.
It is clear that
L
is
a well-defined the identity permutation of

2.4,
mapping of T
into T.
T (Problem
8.9).
Moreover 9i6-i = = O-iOi where Thus each Or is a permutation of T

i
(Theorem
page
36),
i.e.
Oi
St.
. .
= {fiî, .,0}. (As St is a group and XcSt, it makes Let G = gp{X), where sense to talk of gp{X).) We prove that freely generates G, thereby completing the proof of the theorem.
We
have only to verify that every reduced X-product

f
is
not
i.
Let then
e^.
^l-
9^'
(where
1',
.,m'
(1,
.,n}
and
1)
the effect of /
be any reduced Z-product (so if r' on (0). Now er^ = O-i, and so ^f '
= (r + l)', e^ + e^^jÔ). We compute = ei. Then
By
the definition of the product of permutations,

(0)/
we have
^^".')
=
tj
((0)^,^1-
)(V
or
^^')
^1'
= ((o,îi')VHV
Therefore
Now if
1'
2',
then
+ c^ ^
-6^2'.
It follows in this
way
that
(0)/
(0,.^l',e,2',
...,.m')
(0)
Thus
since
it
does not leave
(0) fixed.
prove similarly that there exist free groups freely generated by Since we have not introduced cardinal numbers, we cannot prove this more general theorem here. The reader who has a knowledge of cardinal numbers may read the account in J. J. Rotman's The Theory of Groups, Allyn and Bacon, 1965.
It is possible to
Note:
sets of arbitrary cardinality.
Problems
8.9.
Prove that
Solution:
ffiff-j
and
ff-j^j
i,
where
9;
and
are as defined above.
Let
(rj,
.,r)
T.
Suppose
first
that
r^
(ri,
i;
then
(ri,
...,rjffi9_i
...,r,t)9_j
(rj, ...,r,)
Now if r_i = {i) = i, does not belong to T, r_i
then
(r^,
i).
.,r,)
is
of the
form
(r^,
.,i, i).
Since such an element
# (
Therefore
250

(fu
[CHAP. 8
,rm-i)e-i =
(rj,
. . .
(n, ...,r_i,
=
i)
.
(ri,
...,r)
and so
rJOiB^-^
(rj,
r)
Thus
6ii-i leaves
all
elements of
T unchanged and hence
Sjff-i
i.
Similarly
ff_j9i
=
is
i.
8.10.
Prove that there exists a free group freely generated by a set
such that there
a one-to-one
mapping
Solution:
of
onto the positive integers.
proceed exactly as in the proof of Theorem 8.3 except that we define ffj for all nonzero i and put The mapping p: Y ^ Z+ the positive integers, defined by {i, 92. } (9{)p = i is a one-to-one onto mapping, for as none of the Sj have the same effect on (0), they must = gp(.Y), then be distinct. If is freely generated by Y.
integers
We
F=
d.
Homomorphisms
of free groups
Now it is a fact (see Problem 8.11 below) that if a group G is generated by a set X, then a homomorphism is uniquely determined by its effect on X, because each element of G is a product of elements and inverses of elements from .X'.
Consider conversely what would happen if we had a map ^ of .^ into a group H. Could we find a homomorphism of the whole group G into whose effect on X was the same as that of 9? In general the answer is no (see Problem 8.12). However, we have the following result for free groups.
Theorem
8.4:
be a be any group, and let be freely generated by a set X, let H. Then there exists a homomorphism 6 of F into mapping of X into such that 6 agrees with 6 on X. ^ is called an extension of d.
Let
Proof:
Any
nonunit element of
/
is
uniquely expressible as a reduced X-product
x\i
x^"
where
x^
&X,
e^
= 1
and
x^
x._^^
implies
t.
- -Cj^.^.
Define tQ = {xi6p{Xie)'^ and 1^= 1 (the latter 1 being the identity of H). (cBn0)'" Clearly ^ is a mapping of F into agreeing with 6 on X. To conclude the proof we must
.A.
prove that
^ is
a homomorphism.
shall
To do
this
we
show that
if
x[^
^'n
where
xêX
and
1, then
whether or not
0.
x^^
x^^ is
a reduced product.
If
Assume
it
is true for all positive integers
n- 1, this is true by the <k and consider f = xl^
definition of
"
when
n=
k. If this is a reduced product then fe {xiep {XnOf" by definition. reduced product, then there exists an integer i such that x. = x._^_^ and Consequently, (Xne)" (XnOp = {XiOp {XiOfiXiîef^^ {Xi0p
If it is not a
e^= e^_^y
=
by our inductive hypothesis.
ix,9p
{xiop
/-y'l
. . .
{Xi-iep~'{xi+20p^^
-}.^t-l/).S + 2
. . .
(xndp
q-^n\0
Therefore
(Xnep
y]'
{x{^
x^(c^x^^)x:^
^
<")^ = fO
then
So
if
xl'
<" and g =
x':y\^
y^,
1,
r,^
1,
x., y^
G X,
{fg)d
=
a
{xl^
...yy)e = {xêf
(Xnepiyief
{y.9f^
=
Hence
is
[{xiop
{xndpWiyxep (yêf"'] = foge
homomorphism and the theorem
follows.
Sec. 8.1]
HOMOMORPHISMS OP FREE GROUPS

8.5:
251
Corollary
G be any finitely generated group (i.e. there exists a finite set Y such that gp{Y) = G). Then G is a homomorphic image of some free group.
Let
freely generated
. .
There exists a free group F be finitely generated by a set (t/i, .,y}. .,x}, say. {xi, Define a map from {Xi, .,Xn} to G by XiO = yi for i = 1, .,n. By Theorem 8.4 there exists a homomorphism 6 of F into G which agrees with 9 on each Xu We know that F9 is a subgroup of G. But as F9 contains yi, ., Vn, it contains gp(yi, Therefore Fe = G, and so G is a homomorphic image of a .,yn) G. free group.
Proof:
Let
by
Note: The same result applies whether G is finitely generated or not. However, to prove the more general result we must use some of the ideas involving cardinal numbers. For this reason we have chosen only to consider the finitely generated case.
Our last theorem reveals the importance of free groups. As every group is a homomorphic image of a free group, from the knowledge of the properties of a free group we may achieve some understanding of other groups.
Problems
8.11.
Let
$1
and
$2
be homomorphisms of
G ^ H.
Let
G=
gp(X) and suppose
e^
92ix-
Prove that
Solution:
Let g
&G. Then
x^^
x^ where x^^X and

ix^Sif"
e;
1,
and
991
{xiOiP
effect
(xê^P
(xê^f"
ffj
=
e^-
gB2
Thus
and
$2
have the same
on the elements of G, and so
8.12.
Find an example of a group

there exists no
Solution:
homomorphism
generated by a set X, a group e. e G -* with 9

:
and a map
e:
X -* H,
such that
Let
Define
e
G
:
be cyclic of order
X^H
Then G
a
finite
e is
= gp({y}) be infinite cyclic. Put X = {x} and let 2, G = ffpiix}). by xB y. Suppose there exists a homomorphism e G^H, such that xe = y. a subgroup of H which contains gp{y) and hence G'e=H. But G?, being the image of
:
group, must be
finite.
Since
H is infinite, this
is
clearly impossible.
8.13.
Let be the cyclic group of order 2, say = {1, a}. Let F be freely generated by {}. The map X -^ a gives rise to a homomorphism of F -^ H. Describe the effect of this homomorphism on e all elements of F. Check directly that it is a homomorphism, and find its kernel.
:
Solution:
a;",
If
The free group on a single generator is infinite cyclic. Its elements are uniquely of the form n an integer. Now 6 maps cb" ^ a". If n is even, say n = 2r, a" = 1. Hence 9 maps sc" -^ 1. n is odd, say n = 2r + l, 6 maps a;" to a.
To check that e is a homomorphism, consider whether (a!"x'") 9 = ;" + " 9. Now a;" + '"9 is a if is odd, and is 1 if w + m is even. If m + m is odd, then one of the integers n and m is odd and the other even. Hence a;"9x"' 9 = a, as one of a;" 9, a;"* 9 is a while the other is 1. li n + m is even, then either n and m are both even, or they are both odd; if n and m are both even, a;"9a;'"9 = 1 1 = 1; if both are odd, then a;"9a;'" 9 = <x ' a = 1. Hence 9 is a homomorphism.
n.
+m
The kerne] oi
(x^^
all
|
integers r}.
8.14.
Find a free group that has as a homomorphic image S, n any given positive integer.
Solution :
Let F be free on X where \X\ = n!. Let e X ^ S map each element of X onto a distinct element of S. Then by Theorem 8.4 there is a homomorphism 9 of i'' onto S that agrees with 9.
:
252

Prove that a group
freely generated by
[CHAP. 8
8.15.
freely generated by
is
n elements, where n
w + 1 elements has as a homomorphic image a group any given positive integer and G is a subgroup of F.
Solution:
Let
as
e
:
be freely generated by
before,
X
x
s"
where
is
|Z|=m+1.
= a,
Let
X = Xûid)
we have shown
gp(.Xj)
If
=G
of
freely generated by X^.
X -^ G
be defined as follows:
exists a
G X, x
Then there
homomorphism
onto
put xe = x. by Theorem 8.4.
where \X^\ = n. Then, (Problem 8.7, page 247.) Let If x = a, put xe 1, the identity.
8.16.
Let
to
cyclic
be a free group with free generating set {a, b}. groups generated by c and d respectively. The
map which
is F',
a homomorphism of F onto the derived group of F.
(since
G =
gp({c, d})).
be the direct product of two infinite takes a to c and 6 to d extends Prove that the kernel of this homomorphism
Let
Solution:
First
we show
that
if S is
the
homomorphism
F -* G
for which ae
c,
he
d,
then F'
C Ker e.
F' is generated by the set of all commutators, so it is sufficient to show that each commutator belongs to Ker . Now a commutator is of the form [/i, /2] = /rV^Vi/z where /i,/2 G F. Then [/I'/d* [/i*'/2*] =^1. as G is abelian. Hence F' cKer e.
Now every element of F that does not belong to F' is of the form a''bê where not both r and s are zero and e G F', as F/F' is abelian and aF', bF' generate it. Under e such an element goes onto cM^ and Crf^ i only if both r and s are 0. Thus only the elements of F' belong to Ker 9.
Therefore Ker
e
F'.
8.17.
of index
Prove that a free group freely generated by n elements, n any positive integer, has a subgroup m for each positive integer m.
Solution:
We
will exploit the

a, say.
by an element
morphism 9 of F (Theorem 4.18, page

Since
|C|
homomorphism property. Let C^ be the cyclic group of order m generated F be freely generated by Z = {x^, ..., x}. Then there exists a homon. Hence by the homomorphism theorem onto C^ for which XiO = a, i= 1
Let
117),
F/Ker
^ e
~ r = C^
m, the number of cosets of Ker
e in
is
m.
Hence
has a subgroup of index m.
8.18.
Let
F be
freely generated by X.
8.4 to find
Use Theorem
Solution:
Let F be a subset of F such that gp{Y) an epimorphism of F onto itself.
F,
and
1X|
|F|
<
Let
Ki,
g-F-^F
Hence
e is
a; be the elements of X, and i/j, .,Vn the elements of Y. Define a homomorphism Since Fe contains Y, Fe = F. .,n, using Theorem 8.4. by Xi$ = y^ where i = 1, an epimorphism of F onto itself.
. .
.
8.19.
Let e be an isomorphism of generated by Xe.

Solution:
is
with G.
Let
be freely generated by X.
Prove that
is
freely
Let {xêf^ (Xjfif^ be a reduced product in Xe. not the identity of G. Clearly

Then we must show that

{xf^
this reduced product
(x\^
<")
9 is
(xêp
and as
is
freely generated by
and
an isomorphism, the result follows.
8.20.
Let F be freely generated by spondence, prove that F = G.

Solution:
and
freely generated by F.
If
X -^ Y
is
a one-to-one corre-
is
Let be the homomorphism of F into G which is an extension of e. (Theorem 8.4.) Clearly e Let <p: Y -> X be the mapping such that e<f> is the identity mapping on X and ê is the is the identity on F and to G ^ H. Then 9 be the extension of identity mapping on F. Let It follows readily that 9 is an isomorphism of F with G. 9 is the identity on G.
onto G.
Sec. 8.2]
253
8.21.
Let
be freely generated by a, b, c. section of all normal subgroups of

(Hard.)
Let be the normal subgroup generated by c, i.e. the intercontaining c. Prove that F/N is freely generated by aN and
bN.
Solution:
be free on x, y. Let e be the homomorphism of onto G defined by ae = x, be = y and (Since is free on a, b, c, such a homomorphism e exists by Theorem 8.4.) Let be the kernel of e. First we shall prove that = N, the normal subgroup generated by c. Then we shall prove that aN and bN freely generate FIN.
Let
1.
c0
To
see that
c).
K=N
containing
On
first observe that as ce = 1, the other hand, suppose f e. F, f ^ N.
is
Then
contained in {K is a normal subgroup / can be expressed in the form
==
/iî
where
/j is
a reduced
{a,
6}-product and Wi
G N.
fe
But then
/i9
and /iS
is
f^N,
in K,
a reduced {a;,2/}-product. So fiB^l, which means fe - 1 and so f which implies is contained in N. Hence as we observed earlier that
K.
Thus
if
we
find
K=N
N is
contained
as desired.
Now
F/iV.
a, 6, c
generate F.
So aN,
We
tf
want
to prove that
bN and aN and bN
v
cA^ generate
freely generate F/N.
F/N. Since c2V = iV, aN and Suppose (y^Nf^

i.e.
6A^ generate
(y^Nf'<- is
reduced
{aAT, 6iV}-product.
Now
gives rise to an isomorphism

4.18,
of
F/K
with G,
the
mapping
defined
by (fK)v
homomorphism theorem, Theorem

Observe that (aK)v
fe (the
page
117).
=
y}.

x and {bK)v
((2/iiV)'i
y so that
{y^Nf'')v

xji
x^f^
where
(j/jAT)^
Xj
{a;,
The product
a;^''
ajji
x^" is a
generates G.
Thus
is
aiji
reduced {, 2/}-product.
But
{a;, ?/}
freely
1.
Consequently
and the proof
complete.
Instead of completing the proof this
way we
could refer to Problem 8.19 using the isomorphism
8.2
a.
Definitions
have shown (Theorem 8.4) that if F is freely generated by X, then for every group every mapping 6 of into G there is a homomorphism of F into G which extends 9, i.e. which agrees with 6 on X. This fact will enable us to "present" a given group in terms of a free group. This idea of a presentation is especially important in topology and analysis where groups arise in just this way, as the "groups of certain presentations".
We
G and
First we need a definition. If S is any subset of a group, then the normal closure of S defined to be the intersection of all normal subgroups of G containing S. Clearly the normal closure of S is a normal subgroup of G containing S. Thus the normal closure of S is often called the normal subgroup generated by S. It is easy to prove that the normal
is
closure of
is
gp{g-^sg
\
GG
and
S)
(see
Problem 8.22 below).
A
is
presentation
is is
free group
F and R
defined to be a pair {X; R) where is a free set of generators of a a subset of F. The group of the presentation {X;R) is F/N where
the normal subgroup of generated by i?; we usually denote the group of a presentation {X;R) by \X;R\. Finally a presentation of a group G consists, by definition, of a presentation iX;R) and an isomorphism 6 between \X;R\ and G. presentation iX;R) is finite if
both
tion.
X and R are finite, and a group G is termed finitely presented if it has a finite presenta-
Not all groups are finitely generated (a necessary prerequisite for being finitely presented) and not all finitely generated groups are finitely presented. For a more detailed discussion of these notions the reader may consult R. H. Crowell and R. H. Fox, An Introduction to Knot Theory, Blaisdell, 1963.
254

Illustrations of presentations
[CHAP.
b.
We
shall
work some examples

all,
to illustrate the definitions of Section 8.2a.
First of following
(i)
suppose
is
a free group freely generated by
Z=
{a,b}.
Then
the
i{a,b};{a,b})
({a, &};
(ii)
{a^b% a^b\ a*b\
.})
(iii)
{{a,b};{[a,b]})
are patently presentations (by the very definition). finite, but the presentation (ii) is not.
The presentations
(i)
and
(iii)
are
The obvious question

Clearly the group of
is:
what are the groups

is
of these presentations?
(i)
a group of order
1.
is the normal subgroup are interested in F/N where a^N = b-'^N. Furthermore generated by a252,ab*,a*b*, We have, since a^b^GN, a^N = 6-W since a^b^ G N. Thus it follows that a^N = a'aN = b-âN = &-W. CancelHence ab G N. Indeed we would like to ling 6""W from both sides yields aN = 6"W. Let be the normal subgroup genprove that is the normal subgroup generated by ab. erated by ab. Since ab G N, we have KcN. Now observe that aK = b~^K. Then
What about
(ii)?
Well,
we
a'b'K
(a'K)iK)
. .
{b-'K){b'K)
Hence is contained in K and so we have proved (i = 2, 3, This implies that aW .). Therefore FIN is cyclic (because aN and bN clearly generate FIN; but that K = N. aN -b-'^N). In fact FIN is infinite. To see this suppose G is an infinite cyclic group generated by g. Let 6 be the mapping of {a,h) into G defined by
gK
ae
fir,
be
g-''
Let ^ be the homomorphism of F into G defined by 6 (Theorem 8.4). Clearly e is onto and (a&)e = 1. So if L is the kernel of 6, LdN. But as FIL = G, FIL is infinite cyclic. Therefore FIN is also infinite cyclic since, as we have already noted, FIN is cyclic. (Actually, = L; however, we don't need this fact here.)
may
we
we come to (iii). In fact |{a, b}; {[a, b]}\ is free abelian of rank 2. The reader attempt to prove this before we do so in a more general case. At this point we simplify our notation. Instead of using our set-theoretic notation which encloses the elements of a given set in braces ( }, we shall omit the braces in writing presentations. Thus
Finally
write {a,b; [a,b]) for ({,&}; {[a,b]}), \a,b;
[a,b]\
for \{a,b}; {[a,b]}\, etc.
Let
F be
freely generated by
ai, a2,
I
ai,
a.
We
shall
prove that
.,
a;
[ai, a,]
where
lî^ j ^n\
is
is
free abelian of rank
n (from which
it
follows that \a,b; [a,b]\
free abelian of rank

is
2).
To
infinite
this
end
let
H be the free
abelian group of rank n.

hi,
i
Then
the direct product of

[ot, a,]
cyclic
groups generated by
ai9
1,
n,
say.
1.
We
define a
e-.F^H by
Now
[Oi, aj]6>
[aiO, ajd]
= [K hj] =
Hence
N be the normal subgroup generated by the

[Ci, aj]N that [atN, ajN] F are of the form
[oj, Oj],
lî^j^ n.
c
xr GN
Clearly
homomorphism G Ker 9. Let Ai' C Ker 0. Note
= N. As
,
the generators atN of

'

FIN commute,
the elements / of
ar
a^c
where
Sec. 8.2]
PRESENTATIONS OP GROUPS
/
255
and
the
if
Ti
iV,
at least one of the

1.
rj = 0.
-0, fey^
It follows that
is
Hence as g Ker 9, and
cB
1,
so
f9 = hl^ = Ker 9.
hi".
Now
as one of
Therefore
F/N
the free abelian group of rank n,

I
i.e.
ffli,
an,
[ffli,
ttj]
with
i j ^n\
is
the free abelian group of rank n.
Note that as
A^ is
F/N
is
abelian, A^
D F'. Thus
generated as a normal subgroup by commutators, N C F'. But as = F' and we therefore conclude that F/F' is free abelian
of rank n.
We
state this fact as

8.6:
Theorem
Corollary
Let
Then F/F'
8.7:
be a free group freely generated by a set of n elements is a free abelian group of rank n.
a free group.
is \Y\
{n
<
>).
Let
F be
Suppose
|X|
X and Y
both freely generate F.
If |X| is
finite,
then so
8.6
and
\Y\.
we know that F/F' is a free abelian group of rank n=\X]. free abelian of rank \Y\, again by Theorem 8.6. Consequently \X\ = |y| since the rank of an abelian group is unique (see Section 6.2d, page 193). It remains only to prove that \Y\ cannot be infinite. To do this, suppose 1/1,2/2, .,i/n+i G Y.
Proof:
If
By Theorem
then F/F'
\Y\<
CO,
is
Now
in a free abelian
.
exist integers nii,
group of rank n every n + .,mn+i, not all zero, such that
elements are dependent.
Hence there
{yiF'f
i.e.
{VnîF'p*'
F'
eti,
. .
yiF',
.,yn+iF' are dependent.
Let
be the free abelian group on
.,a+i
and
let 9
be the homomorphism of
yi9
F to A
defined by
.
+ l), y9 = 1 if y {Vu The kernel K of 9 contains F' since F/K is abelian. Then 1 = {y,9p {yr,^,9p'' = aî But A is free abelian on ai, .,an+i. Hence mi = W2 = Thus |r| < 00 and the corollary has been proved.
ai
(i
l,
...,n
.,yn+i)
and
yGY
aZl\^
yru+i
0,
a contradiction.
It follows from this corollary that if F is a finitely generated free group, then every pair of sets which freely generate F have the same number of elements. define the rank of F to be this common number, i.e. the number of elements in any set which freely generates F. Note that free groups of the same rank are isomorphic (Problem 8.20).
We
We
results
(XiF')9
we now
can easily give a presentation of A, a free abelian group on ai, .,an, with the have. Let F be free on aii, Xn and let 9 be the isomorphism defined by
. . . . .
Uj {j
l,..., n).
Then
(xi,
. .
.,Xn;
[Xi, Xj]
with
i j^n)
together with
is
a presentation of A.
In some of the following problems we will often be dealing with factor groups. convention makes the arguments simpler to follow.
simple
Let G be a group and a normal subgroup of G. If we use some phrase such as "let us calculate modulo A^," then by G we mean the factor group G/N. We shall mean by g = h that Ng = Nh. If we say let be a subgroup of G, what we really mean is "let M/N be a subgroup of G/N." In other words, we must remember that we are talking of a factor group and instead of writing the cosets, we will simply write the coset representative. (See Problem 8.24 for an example.)
256
[CHAP.
Problems
8.22.
If
G is a .group, show e R}).
that the normal closure of a subset
(=
0) of
is
gp{{g~î'g
\
S G and
Solution:
Clearly N = gp({g~^rg g & G and r S R}) is a subgroup of G containing R. Also AT is a normal subgroup of G. Finally any normal subgroup containing R must contain A''. Thus the result
\
follows.
8.23.
Let
if
HdK,
be a group with subgroups = K. then
and
and
let
[G:
H]= n<
'^
and
[G
K]
n.
Prove that
Solution
Let Xi = 1 and let the distinct cosets of i? in G be follows that Xxi.XiKj, ..,Xa; are distinct. As [G K] then h&Kxi for some integer i. If i = 1, HnXa;; =
. :
H = Hxi,Hx2,
7i,
.,Hx.
As H'D K,
If
1.
i
it
and h S Kx^
K. Accordingly
H qK
K
and thus
G = KxûKxÛ since HnHa;; = = K.
uKx^.
for
heH,
Hence
8.24.
Rewrite the first part of the argument of presentation (ii), page 254, using the modulo N. and stop after proving introduced above. Also calculate modulo
convention
Solution:
is the normal subgroup generated by a^b^, a^b^, a'^b*, .... We are interested in F/N where Let us calculate modulo A^. Since a^b^ = 1, a^ = b~^. Furthermore a^ = fe-s since a^b^ = 1. Thus it follows that a^ = aâ b-'â = b-^. Cancelling 6-2 from both sides yields a = ft-i, and so ab 1. As we are calculating modulo N, this means that ab G JV. Indeed we would like to prove that 2V is the normal subgroup generated by ab. Let K be the normal subgroup generated by ab. Since ab e A^, we have K cN. Now we calculate modulo K. a = ft-i, and so a*6* = 6-*6 = 1 (i = 2, 3, .). This means as we are calculating modulo K that a'6' (^ K (i = 2, 3, .). Therefore we have proved that K = N.
. .
8.25.
Let
G=
|a; a^\.
Prove that
is cyclic
of order
2.
Solution:
The free group F generated by a is infinite cyclic. Hence gpia^) is the normal subgroup generated by a^. cyclic group of order 2. Thus G is cyclic of order 2.
8.26.
Now
It
gp{a^) is normal in F since F is abelian. can be easily seen that gp{a)/gp{a^) is the
Prove that G = |a;a"|, where n freely generated by a).
is
any positive
integer,
is
the cyclic group of order
n (again with
Solution:
a".
= gpia") is the normal subgroup generated by The free group F on a is infinite cyclic and Therefore F/N = G is cyclic of order n (Theorem 4.9, page 105).
\a,b; a^, b^, [a, 6]| is the direct
8.27.
Prove that
Solution:
A^
product of two cyclic groups of order
2.
Although it has not been stated, a and 6 are (as usual) free generators of a free group F. Let be the normal subgroup generated by a^, b* and [a, b]. In F/N, Na and A^b commute as [Na, Nb] N[a, b] - N, since [a, b] e A^. Since F/N is generated by Na and Nb, it is therefore abelian. Also (Ara)2 = AT and (Arb)^ = N. Let A - {N,Na} and B - {N,Nb}. As A is a normal subgroup of F/N, AB is a subgroup of F/N which contains both Na and A^'b. It follows that AB = F/N. Thus \F/N\ \A\\B\ 2 2 = 4. On the other hand there is a homomorphism e of F onto the direct product of two cyclic groups of order 2. Clearly Ker e contains a^, b^ and [a, b]. Hence Ker e 2 N. = Ker e (Problem 8.23). Thus F/N is the direct product of two cyclic groups of It follows that order 2.
8.28.
Find a presentation for

Let p
S3.
(Hard.)
= (^
V2
^
1
^) and
= (^
^V
Then
a^
= (^
\3
^
1
3/ \2 S 1) are six distinct elements and hence the whole of
^) 2/
So
p,p<r, p<r2
and
1,
<r,
a^
S3.
Now
Sec. 8.2]
PRESENTATIONS OP GROUPS
1
257
3\/l
Syi
p2
3\ _ /I
=
and hope that they
Let
cti*
holds.
Also
and
a^
1.
We
use only these equations
will give rise to a presentation for S3.
be the normal subgroup of
be the free group on ai,a2 and let G = |a.i,a2; ctj, a|, aj~*a2-i'^2~^lF generated by a^ o|, ar'ajaiag*^, and let F
:
Furthermore
->
let
S3
be defined by
and
0.29
a.
Then
N C Ker
e.
Now we
a~â2aia^^
M
Oi
<1
F.
It
02>
and
modulo N. We see that Oj = 1 and so a^^ = dj. Since ttg 1, a^^ = a|, = pp(a2); then from which 0.201 = OjOg. Now F is generated by 01,02. Let follows from Problem 4.62, page 114, that M{1, a^} is a subgroup of F and as it contains ^{1, <*i} F. Thus the elements of F are {1, Og, 02^{1, Oj}, i.e. 1, O2, o|, Oj, OjOi, a|ai
calculate
1
(although
It
we do not know
\F/N\
if
they are distinct).

6.
follows that
But
'
|F/Ker9|=6 and
iV
c Ker
9,
and so
Ker
iV
(Problem
8.23).
Then
-2 -' --1^20,02-^1
-> p,
S3
4.18,
under the isomorphism page 117).
(f>
OjAT
OjiV -> a
(by
the
homomorphism theorem. Theorem
8.29.
Prove that
Solution:
|a, 6; o^, 6",
o-'ftobl
is
isomorphic to the dihedral group of order 2n.
be the dihedral group of order 2n. Then G is the symmetry group of the regular n-gon S page 75). Recall that ^2 rotates S in a clockwise direction through an angle of 2;7/w. It follows that 172 is of order n. Put / =<r2- If t is the reflection about AÔ, where A, is any vertex of S, and the center of S, then t^ = i. Moreover every element of G can be written in the form râ*, where e = 0, 1 and 8 = 0, 1, ., w 1 (see Section 3.4f and note that a* = ctj for 1 i n). Let e be the homomorphism of the free group F, freely generated by a and h, onto G defined by ae = t and he = a. Then a?e 1, 6"e = 1 and
Let
(see Section 3.4f,
(a-i6o6)e
= T-Wa =
of
9.
(r-i(T
Thus
a2,
6>
o^ jn and a~^hah
lie
in the kernel
Then
A?,
N=
and a~^bab, is contained in X. K, then the proof follows.

calculate
Moreover since
is
the normal subgroup of F generated by onto, F/K = G. If we can show that
We
Let
(compare Problem 8.28) modulo N.

this will establish that
<i
\F/K\=2n and
M = gp(h).
KdN,
Then
6,
F.
contains both a and of F are
M{1, 0}
Thus Af{l, 0} is F. The elements of

(l'6
(e
We shall show that \F/N\ 2n. Since by Problem 8.23. Modulo AT, o-ifto = 6-. a subgroup of F by Problem 4.62, page 114. As it
K=N
M are
1, 6, 6^,
.,
6"-i.
0,1,
=: 0, 1,
...,M-1)
FIN. This implies
|F/Ar|
Since
we are
is
proof
calculating modulo N, complete.
really stands for
2n.
Thus the
8.30.
Prove that
Hint:
G=
Show
|a, 6; o^, a-i6afel (This group is called the infinite dihedral group. is infinite. that each dihedral group Z? is a homomorphic image of G.)
Solution:
Let F be the free group on o and 6 and let 9 be a homomorphism of F onto D (as > is a two generator group (see, for example, Problem 8.29), we know such a homomorphism exists). Also Ker 9 D {a2, o-ifiofi}. Thus Ker e^D N, the normal subgroup generated by a^,a~^bab. Therefore G = FIN has each ) as a homomorphic image (D = (F/Ar)/(Ker eJN)). If G were finite of order k, say, we would have a contradiction. For then G9^. as a homomorphic image of a group of order /c is of order fc. But Ge^^ D^ is of order 2k. This contradiction proves that G is of infinite
order.
258

Prove that every
Solution:
finitely
[CHAP.
8.31.
generated group has a presentation.
is a factor group of a free group. group has a presentation. Let G be an arbitrary group. Let X be chosen so that X freely generates a free group F and, furthermore, so that there is a mapping s of X onto one of the finite sets of generators for G. Let be the homomorphism of F onto G such that agrees with 9 on X and let N Ker0. Then (X;N) together with the isomorphism /j.: fN ^ f<f> is a presentation of G.
In Section 8.1d
we proved
that every finitely generated group
We
will use this fact to prove that every finitely generated
8.32.
group with elements 1 = a;i, a;2 a;. Suppose XiXj x^j-,, where xûy G words {i,j) is an integer between 1 and n). Let F be the free group freely generated by !,..., 0^, N be the normal subgroup of F generated by aâja'^^J, (1 i,j^ n), and e be the homomorphism from
Let
G
.
be a
finite
{xi,
.,x} (in other
O],
I
a; aja.ja^*,j
where
li,j^n\
to
defined by {a^N)e
x^.
Show
(tti,
. . .
that
,
a.;
ajajtt^^^jj
where
i,
n)
together with 9
Solution:
is
a presentation of G.
Since
ajftjAT
a(ij-)N,
a^.
modulo
XiXj
A^,
to
some
Now
every product of a's in which there are no negative exponents is equal, if Xj"' = Xj, then scja^ = a;(j,j) = Xi, and so Ojaj = ttj modulo N. But
djaj
=:
o-i
X(2_i)
Xj.
Thus
modulo
A'^
This means that ai e and so a^^ = a-j modulo N. Therefore every product of the a^ involving both positive and negative exponents can be replaced, modulo N, by a product involving only positive exponents. Accordingly every product of the a,^ and their inverses is equal, modulo A'', to an [. It follows that
|F/Ar|
=s
1G|
aîp
Now
let
be the homomorphism from
to
defined by
Xj
(for 1
n).
Then
((;, )^
aîa^j^tu)
and so Ker0 includes djajd^j,.
Therefore Ker
AT.
Since
|F/Ker
<f.\
n (< )
the
and \F/N\
n,
it
follows by Problem 8.23 that

e
:
N
af^N
= Ker 0. Thus
-
mapping
x^
is
indeed an isomorphism (Theorem 4.18, page 117).
8.33.
Prove that
Solution :
\x, y;
xV, y^\ -
I*,
V\ x^,
y\
We have a free group F freely generated by x and y and two normal subgroups T>! and = M. generated, as normal subgroups, respectively by x^j/^ and y^, and x^ and y"^. We must prove
Since
contains x^y^ and
y^, it
M D {x^,
8.34.
j/2}.
Hence
MdN
contains
x^y^(y^)~^
x^.
Thus
and
M N.
n <
o
Nd M.
But
M D {x^y^
since
Show
that a free group of rank
cannot be generated by
?i
elements.
Solution :
Let G be free of rank n. Then by Theorem 8.6, G/G' is free abelian of rank n. If G can be generated by M 1 elements, G/G' can be generated by m 1 elements. But this contradicts Problem
6.41(b),
page 195.
Sec. 8.3]
THE SUBGROUP THEOREM FOR FREE GROUPS: AN EXAMPLE
259
8.3
THE SUBGROUP THEOREM FOR FREE GROUPS: AN EXAMPLE

The object
is
of the next sections of this chapter is to prove that every subgroup of a free a free group. This theorem is one of the more difficult in this book. So in order to give the reader a chance to become accustomed to the ideas involved, we first vôrk an example.
group
Example
1:
be the free group freely generated by two elements a and 6. Consider the set = gp{Y) is freely generated by Y. (Thus Show that .}. we see that a free group freely generated by two elements has a subgroup which is freely generated by infinitely many elements.)
Let
Y=
{a~^ba,a-^ba2,
= 0,1,2, .). Consider any F-reduced product, where .,n' are positive integers, Y = {Vil i = 0,1, .}. Say f = y^' Vn", where V, (Given, for example, 2/32/^*3/12/2. then 1' = 3, and yy j/cj + d- implies ej - + 2' = 4, 3' = 1 We will prove and 4' = 2; ei = 1, ^2 = -1, eg = 1 and e^ = 1.)
Proof: Let a-'ba^
. .
1/4
(i
Cj
i-
that /
by induction on
n,
showing that
is
freely generated by Y.
Our inductive hypothesis (on n, the number of 2/i's that go into a given reduced product /) is that / when expressed as a reduced product in {a,b} ends in 6 "a"'.
(For example.
ij/j2/2
3/32/4
= =
a~^ba^
'
a~'^b''â*
a^^ba- a'^^ba^
a''^ba~^b~â^ba~^ba^
it
as a reduced product in {a, b}, and, as asserted,

If
ends in
ba^.)
this is certainly true.
If it is true for
n =
k,
let
fc
1.
i.e.
Then
pressed as a reduced product in {a,b],
yy
2/1'
2/k-
ends in b â^' when ex,
2/^'
= zbâ^
where
z is
a re-
duced product in a and 5 such that zb^'â^' is a reduced product (i.e. z does not end in ft-i if efc = 1 or, if ep. 1, z does not end in 6). If now fc' = (fe -I- 1)', then, as / is a reduced product, e^ v^ "ffc + i- Since e^ and e^ + j are the numbers 1 or 1,
ffc
^k
1-
Thus
Since
j,ffc
ejc
Cfc,
this last expression is
a reduced product in
then
{a, b}
and / ends
in
+ i(fc +
i)'_
If
however
fc'
(fc
-h 1)',
Since
k'
(k
1)' 7^ 0,
the last expression

a*^''"'"*'
.
is
a reduced product and so / expressed
It follows therefore that in both situaas a reduced product, ends in &''"'"' tions f = 1. Thus Y freely generates H. Similar arguments will help to prove in general that a subgroup of a free group is free.
Problems
8.35.
Verify both the inductive assumption of the preceding example and that
/
where
2/2" ^2/3" '2/12/12/3
Solution:
/
a-26-ia2.(i-3fe-i(i3.a-i6a,-a-i6a '0-36(13
= a-^b-â-^b-â^ba^^ba^
is
/ # 1. The inductive assumption of the preceding example reduced product should end in ba^, which it does.
Clearly
that /
when expressed
as a
8.36.
Given the existence of a free group freely generated by two elements, prove the existence of free groups freely generated by any finite number of elements.
Solution
In
:
Example
Y
let
such that
2/i>

-jJ/ti
2/1,
erated by
that a free group freely generated by two elements has a subset If n is any positive integer is freely generated by Y. Then gpdvi lê n distinct elements of Y. 2/n}) is easily shown to be freely gen... 2/n- Thus there exist free groups of rank n for each positive integer n.
1
is infinite
we have proved
and
H = gp{Y)
260

Let be freely generated by a and 6. Prove that the subgroup of freely generated by aba^ and a^ft. (Hard.)
Solution
[CHAP.
8.37.
generated by aba^ and a^b
is
Let
Y=
-
{aha^, a^b}
and
let
y^
aba^,
2/2
a^b.
Consider a reduced Y-product
= vr
Vnif
1,
(i'e{l,2})
Then
if
j'
(i
+ 1)'
implies
e^
-fj+i-
We
e
will
show by induction on
if
-1; while
n'
=
k,
2,
ti that if n then / ends in 6
then / ends in a>- if e = = 1 and ends in 6-ia-2
1
if
and ends
c
in
6-%-i
=
+
-1.
For w
this is certainly true.
If it is true for
(i)
we must proceed case by
case in proving
it
true for k
\.
k'
= L
e^
(a)
Then by the inductive hypothesis yy y^ ends in a^. If now (k + 1)' = 1, then a reduced product we must have c;j + i = 1 and / ends in a^ ' aba^ = aaaabaaa. Thus / ends in a^ as required. If (fc + 1)' = 2 and e^ + i = 1, then / ends in a^ 0,26, and so / ends in 6 as required. If c^ + j = 1, then / ends in a^-h-â-^ and so / ends in
1.
as /
is
6~ia~2 as required.
e*)
ffc
= ~1- Then by the inductive hypothesis j/j^ y^' ends in 6-%"!. then as / is a reduced product e^. + i = 1 and / ends in

If
now
(A;
1)'
1,
6-ia-i'a-3&-ia-i = b-ia-*b~â~^
Hence / ends in &-%-! as required. If (k+l)'-2 and 6^ + j = 1, then / ends in b-â-â^b = b-âb, and so / ends in 6 as required. If {k+iy = 2 and + i = 1, then / ends in bîa~i b~â~^ and so / ends in b~ia""2 ^s required.
ej.
(ii)
k'
(a)
2. e^
Then by our inductive hypothesis yy If {k + !)' = ! y^' ends in b. and / ends in baba^ and so / ends in a^ as required. If (fc + 1)' == 1 and e^ + i = 1, then / ends in ba-^b-â-^ and / ends in 6"^%-! as required. If (fc + 1)' = 2, then as / is a reduced product ej. + i 1 and / ends in ba^h, and so / ends in 6 as required.
1.
ejc
1,
(6)
ek~l. Then by
and
ej.
our inductive hypothesis 2/1' yr ends in b-â-^. If (A; + 1)' = 1 / ends in b~â~âba^ = 6~ia~i6a3^ gud so / ends in a^ as required. If (A; 6^+1 = -!, then / ends in 6-ia-2 a-sft-ia-i = ft-ia-sft-ia-i^ and so / ends in 6-ia.-i as required. If (k + 1)' = 2, then as / is a reduced product e^. + i = 1. Thus / ends in b~â-^- b~â~^ and so / ends in 6~ia~2 as reqviired.
1, then + !)' = ! and

+
i
Thus we have proved by induction that any reduced product always ends and 6"ia"2. Hence / # 1 and Y freely generates gp(Y).
in one of
cfi,
fe-ia-i, 6
8.4
a.
PROOF OF THE SUBGROUP THEOREM FOR FREE GROUPS

Plan of the proof
The subgroup theorem for

as follows.
free groups (due to J. Nielsen and 0. Schreier)
may
be stated
Theorem
if
8.8:
Every subgroup
i? of a free
group
is free.
Suppose that
is
freely generated
by
S.
X is any right transversal of H in F, then the nonunit elements

ttx.s
We know
{x
from Section
7.6b,
page 228, that
xs{xs)~^
G X,
GS,
S)
(where, for f
F, /is the unique element of
in the coset s
a^.s
Hf) generate H.
1}
Let
Y =
Then we
(i)
{a^r.s
G X,
shall prove that
is
actually freely generated

steps in the proof:
by
provided
is
chosen ap-
propriately.
Thus there are two main

Choose
appropriately.
(ii)
Prove that
freely generates H.
Sec. 8.4]
261
Step (ii) of the proof will be broken into two parts: the first part requires a careful look at the elements ax.s and the second involves looking at the way in which products of these ax,s and their inverses interact.
b.
Schreier transversals
Suppose that Every element x
X
(ic
is
1)
in F, where F is freely generated by a set a transversal for in may be expressed uniquely as a reduced S-product
-S,
aiaz
ttn
(n
l)
n
is
where
a, is an element of S or the inverse of an element of S. Recall that length of x and that the length of 1 is 0. We shall call the elements
1, ai, a-iaz,
. .
termed the
.,
aia2
an
initial
segments of
x.
Definition:
A
it
right transversal
is called
a right Schreier transversal
if
every
initial
segment of an element
Notice that
follows that
if
in
is
also in
X.
X is
a right Schreier transversal, then 1
&
X.
The main
result of this section is the following.
Lemma
8.9:
Suppose that
of F.
i^ is
a free group freely generated
by S and that
Then we can always
find a right Schreier transversal
H is a subgroup X for H in F.
Proof:
We
n but no element
the cosets of First
if
say that a right coset Hf is of length n if there is an element in Hf of length of length less than n. We shall choose inductively using the lengths of in F.
Hf
is
of length
0,
then 1
G Hf and
so
Hf = H. We
choose 1 to be the repre-
sentative of H.
Suppose that and that for each coset of length less than n, representatives have already been chosen so that every initial segment of a representative is again a representative. We choose now representatives for the cosets of length n. Let Hfhe a. coset of length n and let o-ias a. be an element in Hf of length n. The element aia.2 a-n-i is of length n 1. Thus the coset Haâ^ On-i G a-i is of length at most n \ (since aâz HaiUz This means that the representative of HaiUz an-i). a^-i has already been chosen, by our induction assumption. Suppose this representative is bibi bm. Now
n>
H{bib2
bman)
{Hbibi
bm)an
{HaiUz
an-i)an
Haiaz
an
We
select 6162
6162
ments of
bmOn, excluding bibz

1,
bman to be the representative of the coset HaiOi bman, are

61,
.
On.
The
initial seg-
.,
61&2
bm
In the same
which we know have already been chosen as representatives.

suitable representatives for all the cosets of length n.
way we
select
We
have therefore verified the induction hypothesis and so we are
able, in this
way,
to complete the choice of a right Schreier transversal for
H in F.
c.
look at the elements
a^.s
nonunit element
Suppose that we have chosen a right Schreier transversal ax.s where x G X and s G S:
ax.s
for
in F.
Consider a
xs{xs)~^
262

for
a; be a; = ai be interpreted as
[CHAP. 8
Now let the reduced S-product We will allow = 0; this will

A;
a^
a;
1.
where Ui or its inverse belongs Thus we may write

ajcs)"'
to S.
tti.s
aia2
aks{aia2
Let
i
where the right-hand side is a reduced S-product with bi or bi~^ G S, UkS and 6i bis-^ are not elements of X. For suppose aiUkSG X, then ai akS = ai UkS and so a^., - 1; whereas if bi biS'^ G X, we utilize equation (7.3), page 219. and conclude that
tti
ttkS
= 6i
6,
l,
.,1.
We
asse rt that ai
=
From
this
it
6i
bis-\bi
&iS-i)->
&i
bis-\bi
biS~')-'
follows that the reduced S-product for a^c.s is ai aksbf^ br\ For if with bi~\ But as every initial segment of an element of belongs to X, either ai akS G or bi 6,s-> G X, which is a contradiction. Note we have proved ax.s=l implies x does not end in s"" and xs does not end in s.
not, s cancels either with ak or
Let
W = {w\w GS or
8.10:
w"' G
S}.
Then we have the following

if
a^c.s
Lemma
Let
or -1.
Then
^
Ci
1,

O-x.s
CmWdn^
di^

where the right-hand side is a reduced product and both Ci Cm and di are elements of X and both ci CmW and di dnW~^ lie outside X.
Proof: The only note that

for the case
lemma
is
an immediate consequence of the preceding remarks.

c-xl
We
need
bi
bis~â^^
fti"^
-1.
Corollary 8.11:
Suppose
e- 1 and
dx.s
Ci is
CmWdn
'
'

di
Ci
where the right-hand ci CmW ^ X. Then
side
reduced
product,
Cm
and
(i)
if if
(ii)
w G S, w & S,
=
then
1,
Ci
Cm and s
dn
w,
then
1, x
di
and
s == it;"^
Proof:
It
follows immediately
from the preceding argument.
Corollary
8.12:
Let
1,
77
1,
x,yGX, and s.tGS.

di^
If
a^.s
Ci
CmWdn
and
aJJ.t
Ci
Cmwe"'
Ci
ef
Ci
with the right-hand sides reduced products, Cmit' ^ -X", then = y, x = y and f = s.

and
Cm
but
Proof:
It follows
from Corollary
8.11.
d.
The proof
of the
subgroup theorem
F.
Suppose again that i^ is a free group freely generated by S and that Choose a right Schreier transversal X for H in F. Then
is
a,
subgroup of
Y =
[ax.s
I
G X,
S,
ttx.s --
1}
it suffices
generates H. It remains only to prove that Y freely generates H. For this prove that any reduced Y-product is not the identity (Lemma 8.2, page 248).
to
Sec. 8.4]
263
a^'' J
ei
r,
a^ _, {xi, .,Xr&X and Si, Let g = aî J Sr G S) be a reduced Y-product. Let and be expressed in the form of Lemma 8.10 as ei emvf~^ /^"S where v induction on /^^ by but ei e^v ^ X. We will show that g ends in vf~'^ em E. the case r = 1 being of course proved in Lemma 8.10.
. .
GW
If the result is
assumed true for r

8.10 with

1
C/c
and
ax~_\_sr-i ^^
expressed as
Ci
in the
/''",'
,
form
,
of
Lemma
ci
but
Ci
c^w
{w
CkWdT^ dr* then ij,sj W),

"
ends in wdi~^
di^.
Consider the product
Wdi
di
ei
CmVfn
fi
can convert it into a reduced product by successively deleting inverse pairs. If m = i and wdf^ e^v = 1, then by Corollary 8.12, alizlsr-i = (^^-^ ^"'' '^^^^^^'^y ô dTêi the assumption that g isa reduced Y-product. If ei emV is removed as a result of deleting inverse pairs in the product df' CmV, then ei CmV is an initial segment of dTêi di and hence an element of X. di But this is contrary to Lemma 8.10. Similarly wdr^ dr^ is not removed as a result of deleting inverse pairs in the product Thus wdr^ drêi e-m.
We
wdi^
dTêi
emvf~^
A"^
= w
vf~^
fi^

when expressed
g ends
as a reduced product

by deleting inverse pairs
(the
between
and v
represent the factors df'

in w vf^^ the result follows.

diêi em left after deleting inverse pairs). Consequently and the inductive assertion follows. Therefore g ^1, and /f'
e.
Subgroups of
finite
index
In Section 7.6b, page 228,

r elements is generated
in a free group. the nonunit ax.s freely generate the subgroup.

.
we proved that a subgroup of index in a group generated by by nr elements. We shall now find the rank of a subgroup of index To find the rank of the subgroup we use the result of Section 8.4d, i.e.
.
.
{xi, .,Xn} with Let F be freely generated by Si, Sr and let Schreier transversal for a subgroup of index n. Consider the elements
,
.
Xi
be a
eta;., s
XiSj\XxSj)
1,
TC
and
1,
.,r.
The number of such elements

of the
ax^.sj
ax^.s^ ax^.Sj
is
nr.
To
find the
we wish
(1) (2)
to determine
how many
,
are unit elements.
By
line 13,
rank of H, page 262:
If Xi
ends in
sr'^
then
then
If x^j
ends in
Sj,
= 1. = 1.
Suppose that Xi ends in sr^. Then (1) and (2) are mutually exclusive. .,WmGW) is a reduced product. Consequently Wm - Sj WmSY^ (where Wi, (for otherwise wi WmSj^ is not a reduced product). As Xi is in the Schreier transversal X, wi Wm Wm also belong to the Schreier transversal. Thus xWj w\- Wm = wi and xi^j does not end in Sj. Therefore (1) and (2) are mutually exclusive.
Xi
We
show that
= wi--
Note that if neither Xi ends in sj^ nor îsj ends in remains when a^j.s^ is expressed as a reduced product.
Sj,
then
o^^.s^
XiSifî)'^ ^1
as
Sj
fixed j, let a^ = number of x^G for which aij.s^ = 1. Clearly a. = number of for which x. ends in sf^ plus the number of x^ for which xJ'. ends in s.. As x^ runs through X, x^. runs through Thus the {x-> xW:, x G X, is a permutation, page 219). number of x^ such that xW. ends in s. is the number of elements in which end in s^ We .,r) conclude a. = number of x^ that end in s. or sr^. So the total number (i.e. with j 1, .,r. of ax^.Sj =1 is Oj + +a^ = number of elements of that end in s. or sr\ j = 1, +a^ n 1. But except for x^ 1, every element of ends in some s. or sf^ Hence a^+
For
x^G
264

1 unit
[CHAP.
Thus there are exactly n
elements
among
n{r
the
ai:..s.,
and we have proved
Theorem
8.13:
Let
i^
be a free group of rank r and
let
1.
H is a free group of rank

Problems
8.38.
H he a subgroup of index n.
Then
1) +
be freely generated by Sj Let s^. be a subgroup of Index 2 such that s^ S H, for i = 2, .,r. Find a set of free generators for by using the method of Section Verify that the number of free generators agrees with the number given by Theorem 8.13.
Sj
Let
but
8.4d.
f/
Solution:
We
%,Sj
tti.s-
choose {l.Si) for the Schreier transversal.
(i
l.
.)
freely generate
Sj
H.
ISjClSj)""'
as
Sj
implies
Now i^ =
0.1,^.
1.
= lSj(is^)-i. Now
if
Then the nonunit elements among a^^. and If j = I, a^^.^l. If j = 1,
sj,s-
SiSj{s^j)-i
= =
SiSjS~i
s?
if
j
j = 1
(for then SiSj-ei/si)
=
.
(as
sf=l)
and SiSjS"',
. .
Thus the subgroup

rank 2(r
1) +
1,
is freely generated by sf,S2, which agrees with Theorem 8.13.
.,Sr
.,SiS^f '.
Thus
is
of
8.39.
Let
be freely generated by x and
y.
Find a set of free generators for F', the derived group of F.
Solution:
s integers}. Then to show that is a Schreier transversal we need only belong to the same coset of F' only if r = r^, s = s^, and (2) is a set of coset representatives. Both (1) and (2) follow easily on using the fact that FIF' is free abelian with basis xF' yF' (by Theorem 8.6, page 255). The free generators of F' are the elements a r s and a^rs which are nonunit. Now
Let
{x^y^
and
show that
(1)
a;''2/s,
ie'ij/î
.,
"xV.x
^^y^x{x^y^x)-^
x''y^x{x^ + ^y^)-i
x^y^xy-^x-"^-^
ior s
On
the other hand,
"xV.I/
xrysy{xrys + l)-l
1
all
Thus a
integers
set of free generators for F' axe the elements

s = Q.
x^y^xy~^x~^~^ for
integers r and
all
8.40.
Let
be a free group of rank r and
and
subgroups of index
n.
Prove that
H^
K.
Solution:
By Theorem
8.13,
H and K are free of the
same rank. Thus they are isomorphic.
8.41.
be a free group on generators x and y. Suppose that R < F, y G R and F/R = gp{xR} Prove that the group R/R' is freely generated as an abelian group by the elements a;"j/x~"i2', where n G Z. Then prove that for no integer n, is x''yx~"R' in the center of F/R'.
Let
is infinite cyclic.
Solution:
The method
elements
ci^n_g
of Section 8.4d with a;"j/x~", Z. By an
nG
free groups of finite rank), R/R'
is
{a;" n G Z} gives for the free generators of R the argument similar to Theorem 8.6 (which deals only with free abelian with basis x^yx~^R' for all integers n.
\
X=
Now
as
(_xR')(xi^x-'^R'){xR')-'^
a;"+i3/a!-("+"fi'
x^yx'^R'
xnyx^R' does not belong
to the center of F/R'.
Sec. 8.4]
265
f.
Intersection of finitely generated subgroups
We
Hr\K
if not,
is
prove here that if also a subgroup of
and
finite rank.
are subgroups of finite rank of a free group F, then (This result is due to Howson.)
the problem we note that we may assume that F is finitely generated. For could consider gv{H, K), which is certainly finitely generated (as and are), instead of F.
To simplify
we
However, as we have seen

of infinite rank.
in
Example
Therefore we freely generated by a and 6.
may
1, page 259, a free group of rank 2 has a subgroup as well assume that F is free of rank 2. Let F be
We say that a coset C is single-ended product has the same last factor.
if
every element of
C when
expressed as a reduced
For example, if all the elements of C end in a, C is said to be single-ended. (We might have for instance baaa G C.) But if C contains for example the elements ababab and
a6~S then C
is
not single-ended.
double-ended.
A coset which is not single-ended is called

The following lemma
is crucial:
Lemma
8.14:
subgroup H of the group freely generated by a and b and only if it has a finite number of double-ended cosets.
is
of finite rank
if
Proof:
(1)
Choose a Schreier transversal

be of
finite
X for H
in F.
Put S
{a,b}.
Then only a finite number of the elements ax,s ^ 1 (where But we proved in Section 8.4d that every element of H ended in the form wdr^ df' where wdf' di^ is either s{xs)~^ or s~ix~S where ax.s - 1. Thus we concluded that all the elements of H end in the form wdf' df^ where there are only a finite number of wdf^ diG X. di^ and di diw~^ X although di
Let X G
rank.
H X
and
S).
Let X
G X. Then
if
Hx
is
double-ended, there exists an element h

in

GH
such that
some wdf' diW^ X, it df* and di follows that if x cancels completely, x is an initial segment of di di (otherwise tvdi di is an initial segment of x and hence in X). It follows that as the number of di di that appear is finite, the number of initial segments is finite, and thus the number of double-ended cosets is finite.
in hx, x cancels completely.
As h ends
(2)
it follows that x does not end in s"^ page 262) and, of course, a;s(s)~^ G H. The coset Hx contains x and {xs{xs)~^)''^x = xss~^. Now s~' appears in the reduced product for xss~^ (line 13, page 262). As X does not end in s~^ Hx is double-ended.
Let
H be
13,
of infinite rank.
If
ax,s
xs{xs)~^
- 1,
(line
Now
number
Theorem
as there are an infinite of double-ended cosets.
number
of x such that
ax.s - 1,
there are an infinite
8.15:
The
intersection of
two subgroups
and
of finite
rank
is
again of
finite
rank.
By Lemma 8.14, Proof: has only a finite number of double-ended cosets, say ., H, and Hi, has only a finite number of double-ended cosets, say Ki, Km. Now the cosets of are intersections of cosets of (Problem 7.23, page 231). Also and the intersection of a single-ended coset of with any coset of is single-ended (and vice versa). Thus the double-ended cosets of .,n are among the cosets HiCiKj, i 1,
. .
HDK
.,m.
H HnK
and
1,
Therefore
HnK is of finite rank by Lemma 8.14.
HnK
has at most
mn
double-ended cosets and accordingly
266
[CHAP. 8
and gave an alternative definition. The existence of free groups established and homomorphisms of free groups investigated. The main result was that if F Is a free group freely generated by X, then for every group and every mapping e oi into there is a homomorphism ^ of into which coincides on with 0. As a
We defined free groups
was
consequence every group
is
a homomorphic image of a free group.
We
next discussed presentations of groups.
The important notion of rank of a free
group then arose naturally.
We discussed and proved the subgroup theorem for free groups. Using this, the rank of a subgroup of finite index was calculated. Finally we proved that the intersection of two subgroups of finite rank in a free group is of finite rank.
ELEMENTARY NOTIONS 8.42. Suppose F is freely
generated by a and
6.
Find elements u,v e.F so that u- a and
8.43.
Prove that a free group of

fixed integer n.
finite
rank has only a
finite
number
of elements of length
for any
8.44.
Prove that
if
N< G
and GIN
is free,
then
splits over
N.
8.45.
Let f be a free group freely generated by a and 6. Let a^^ba^, Prove < F and verify that F splits over N. .).
. .
N =
gp{.
.,a^ha-^,aha-^,h,a-^ha,
8.46.
Using the notation of the preceding problem show that if N' is the derived group of N, then N' < F. Verify that F/N' is a splitting extension of N/N' by an infinite cyclic group. Construct an isomorphic copy of F/N' directly as an extension of a free abelian group by an infinite cyclic
group.
PRESENTATIONS OF GROUPS 8.47. Prove that the group G = \a,b; a-^ba - b^l is not free. (Hint: Prove that G/G' is cyclic. So if G is free it must be free cyclic. Then show G' = {1} by mapping G into a suitable factor group.)
8.48.
Let G = \x,y;x^,y^\. Show that G is infinite. {Hint: Let i^ be the free group freely generated by X and y. Let e x ^ a, y -* ab be a homomorphism onto the group of Problem 8.30, page 257.)
:
8.49.
Let
G =
2.
\a,b,e,d; [a,c],[a,d\,[b,c],[b,d]\.
Prove
is
the direct product of two free groups each
of rank
8.50.
Prove that if G = \xi,X2, then G is isomorphic to the .; 2x2 = a; i, . (+ l)a:i+i = *i. !, additive group of rationals. {Hint: Note that G has the additive group of rationals as homomorphic image. Also it is abelian. Use Problem 6.72, page 208.)
.
.
8.51.
Prove that if p is a prime and to the p-Prufer group.
G =
\xi,X2,
.;
Xi,x^x-^,
.,a;f+ia;rS
.|,
then
is
isomorphic
CHAP.
8]
is
267
8.52.
What
the group
\xi, x^,
...,; Xi^j"']?
THE SUBGROUP THEOREM FOR FREE GROUPS Prove that if F is free on x, y, z then the elements 8.53.
subgroup of F.
xyz, x^y^z^, x^y^z^, x*y*z*,
freely generate a
PROOF OF THE SUBGROUP THEOREM Show that if a = 6" where a and 6 8.54.
cyclic group.
are elements of a free group (mn
= 0),
then a and 6 generate a
8.55.
Let
that
and M be nonidentity NnM j^ {1}.
normal subgroups of G, a free group of rank greater than
1.
Prove
8.56.
of F with Let F be a free group. Prove that there is no sequence of subgroups Fj C Fg C Fj v^ Fj+i and rank of F; = 2. {Hint: Let G = UF;. Then G is a free group of infinite rank. Consider G/G' which is free abelian of infinite rank. Obtain a contradiction by showing that G/G'
is
of finite rank.)
8.57.
Find generators for F^ where
is
a free group of rank two and
F^
gp(x^
a;
\
e F).
(Note that
F/P^
is
the Klein 4-group.)
8.58.
Let F be a group which can be generated by two elements with a free subgroup of rank 3 which of index two. Prove F is free by comparing it with a free group of rank 2. (Hard.)
is
8.59.
Let
F,R
be as in Problem 8.41, page 264.
Prove that F/R' has no element other than
1 in its center.
Appendix
Number Theory
In this book
we assume
the reader
knows the
following:
1.
The meaning
of a divides b, for which read as "a does not divide b".
we
use the notation a
b.
The notation a
)(
b is
2.
The
definition of a prime,
i.e.
an integer not equal
to 1
which
is divisible
only by 1 and
itself.
3.
If a, b are integers,
then there exist an integer q and an integer r such that
bq
where
4.
<
b.
The definitions of greatest common divisor of two numbers a and b (the largest integer which divides both a and b) and lowest common multiple of a and b (the smallest integer divisible by both a and 6). We write the greatest common divisor of a and b as (a, b), the lowest common multiple of a and 6 as l.c.m. (a, b). If (a, b) = 1, then a and b are
said to be co-prime.
5.
Given integers a and
b,
there exist integers p and q such that

(a, b)
= pa +
qb
is
6.
The fundamental theorem of arithmetic which says every integer

and only one way (ignoring order) as the product of primes.
expressible in one
7.
a ^ b modulo n means a & is divisible by n. Also some of the simpler properties such as a = & and x = y implies a + x = b + y and ax = by.
The reader who does not know

(a)
this material can consult
Niven,
I.,
and H.
J. V.,
S.
Zuckerman,
An
Introduction to the Theory of Numbers, Wiley, 1966.
(6)
(c)
Upsensky,
and M. A. Heaslet, Elementary Number Theory, McGraw-Hill, 1939.
Birkhoff, G., and S.
MacLane,
Survey of Modern Algebra, Macmillan, 1953.

269
Appendix
A
Note:
Guide
to the Literature
on pages 272 and 273.
Numbers
in brackets refer to the bibliography
General
Books which contain much of the material of this text (and frequently more) are, in the order of complexity, [46], [15], [16]. Useful books in algebra are [47] and [23].
Chapter
[1].
of sets was established by G. Cantor in the last quarter of the 19th century central notion in his work is that of cardinality - two sets have the same cardinality if there is a one-to-one mapping from the one onto the other. This leads to the idea of a transfinite number, and much of Cantor's work was devoted to developing these transfinite
The theory
numbers
(see
[2]
and
[3]).
In the study of sets a number of paradoxes arose ([4] and [5]). This made it desirable to put set theory on a firm axiomatic footing. Such axiomatic approaches have led to a number of interesting developments ([4], [5] and [6]).
by Lawvere, which
category).
Recently a different approach to the foundations of mathematics has been originated is based on the notion of category [44] (see [7] for the axioms of a
Chapter 2
axioms for group theory.
The study of groupoids arose in part from a desire to understand more clearly the The theory of groupoids can be subdivided into systems which
satisfy various "natural" conditions. Thus, for example, among the various classes of groupoids one has semigroups, loops, groups and quasi-groups. Two major references are
[8]
and
[9].
Chapter 3
It is clear
from
this chapter that
groups arise in a great
many mathematical
disciplines.
from 19th century algebra, analysis and geometry. It was hoped that much of geometry could be handled by associating a group with each geometrical object. To some extent this aspect of group theory has been discussed in the section on isometry groups (see also [10], [11], and [14]). For the applications to topology see [37] and [39], and for knot theory see [40].
initially
Groups arose
Groups arise also in quantum physics, crystallography [12] and chemistry omnipresence of groups is part of the reason for its importance.
[15], [16], [17], [18], [19], [20], [21], [22], [43]).
[13].
This
In addition, the study of groups has been carried forward for its own sake (see e.g. Finally we refer the reader to the study of groups with a topology: [41], [42].
270
APPENDIX
Chapter 4
B]
A GUIDE TO THE LITERATURE
271
Chapter 4
is
concerned mainly with cyclic groups and homomorphisms.
As we have dealt exhaustively with cyclic groups, there is not really any further informaIf tion available except perhaps for finding the automorphism group of a cyclic group. G is cyclic of order n, then the automorphism group of G is the group of integers co-prime
to n,
with multiplication modulo n; see
e.g. [43].
One can define homomorphisms for other is basic. Factor rings can be defined and very similar theorems algebraic concepts, such as rings. obtained [23], [24], [25]. (Indeed, once the group theoretic ones are known, it is routine
The concept of homomorphism
to obtain the others.)
of
Some work, prompted by algebraic group homomorphisms [26].
topology, has been done on
more complex
interaction
Chapter 5
of this chapter is "arithmetical" in content, i.e. it deals with the order of a group. Thus for example we have the theorem of Lagrange: The order of a subfinite group of a finite group divides the order of the group. There are a great number of important theorems of this kind, embodying various generalizations of the Sylow theorems A very important result of a slightly different kind is the remarkable (see [27], [17]). theorem of W. Feit and J. Thompson, viz. a finite group of odd order is solvable [45].
Much
groups of small order has not been too successful. This is due to the extraordinary complexity of these groups. The reader might consult [21], [17] and [28] for a discussion of this classification problem.
The
classification of
Today the study of finite groups has proceeded at an extraordinary pace. One of the main aims of this study is the classification of finite simple groups. This classification is by no means complete, but much progress has been made (see [29]).
Chapter 6
of the fundamental theorem of abelian groups is nonconstructive, i.e. not have a definite procedure for finding a basis. Such procedures exist, e.g. [30].
Our proof
we do
There are also other criteria for deciding if an abelian group is a direct sum of cyclic groups, e.g. if every element is of bounded order [81]. A more comprehensive result is that of Kulikov [16].
The inIn this chapter we found invariants for finitely generated abelian groups. variants for countable torsion groups are subtler and appear in Ulm's theorem ([31], [32]).
e.g.
of the theorems of abelian groups extend to wider classes of algebraic structures, modules over rings ([31]). There has also been an attempt to prove theorems about classes of groups that are quite close to abelian groups, e.g. solvable and nilpotent groups
Some
([17], [19]).
Excellent sources for abelian groups appear in [16] and [31]; a count appears in [32].
more encyclopedic
ac-
Chapter 7
The aim
of this chapter
is
to
show how representing groups as permutation groups
leads to information about the groups themselves. There are other types of representation. matrices. The most important of these is the representation of finite groups as natural generalization of the 2x2 matrices we have matrices constitute a (These discussed in Chapter 3.) The reader might consult [33], [34].
nxw
nxn
272
[APPENDIX B
In infinite group theory the permutational representation of Frobenius has led to

significant progress (2 of Chapter 2, etc., of [35]).
The transfer has been used a great

formation
is [27].
deal in finite group theory.
good source of
in-
See also
[17].
Chapter 8
Our proof of the existence of free groups a glance at the usual proof will reveal [16].
is really
motivated by Cayley's theorem, as
The property of Theorem 8.4, i.e. the property of being able to extend a mapping of the free generators to a homomorphism of the free group, is of great importance. Indeed, it is often used as the definition of a free group. With this approach, the definition of free group
is analogous to the definition of free abelian group (Chapter 6). There are even further generalizations and we can speak of the free group in a variety, where a variety is a collection of groups satisfying certain conditions. detailed account appears in [35].
also regard a free group as being the free product of infinite cyclic groups. free product of two groups is a way of putting the groups together in, roughly speaking, the freest way ([16], [17]).
One can
The
Groups occur frequently as presentations. This is unfortunate as it is always difficult what the group of a presentation is or what its properties are. Indeed, there is not even a general and effective procedure for deciding whether the group of any given presentation is of order 1. All this is connected with the word problem, which is, roughly speaking, to determine in a finite number of steps if two products in the generators of a given presentation are equal. The word problem is unsolvable in general (see [15]).
to say
An important concept which enables us to change that of Tietze transformations [36]. An
from one presentation
to another is
There are many proofs of the subgroup theorem, including topological ones (e.g. [37]). important one is by Nielsen transformations ([16], [17], [36]). One of the advantages of the method we used in Chapter 8 is that it extends readily to provide generators and defining relations for subgroups of a group of a presentation [36].
More properties
[36].
See also
[38]
of free groups, free products and presentations of groups appear in for presentations.
BIBLIOGRAPHY
[1]
P. E. B. Jourdain),
[2]
[3]
Cantor, G., Contributions to the Founding of the Theory of Transfinite Open Court, 1915.
E.,
Numbers
(translated
by
Kamke,
Theory of Sets (translated from the 2nd German edition by F. Bagemihl), Dover, 1950.
R.,
Halmos, P.
Naive Set Theory, Van Nostrand, 1960.
[4]
Kleene, S. C, Introduction to Metamathematics,
Van
Nostrand, 1952.
Scientific
[5]
Cohen, P.
J.,
and R. Hersh, "Non-Cantorian Set Theory",
American, December 1967.
[6]
[7]
Suppes, P. C, Axiomatic Set Theory,
Van
Nostrand, 1960.
Freyd,
P.,
Abelian Categories,
An
Introduction to the Theory of Functors, Harper and Row, 1964.
[8]
Bruck, R. H.,
Survey of Binary Systems, Springer, 1958.
[9]
Clifford, A. H.,
7,
1961,
and G. B. Preston, "The Algebraic Theory of Semigroups", Mathematical Surveys American Mathematical Society.
APPENDIX
[10] [11]
B]
273
Weyl, H., Symmetry, Princeton University Press, 1952.

Hilbert, D.,
and
S.
Cohn-Vossen, Geometry and the Imagination, translated by P. Nemenyi, Chelsea,
1956.
[12] [13]
Hammermesh,
M.,
Group Theory and
its
Application to Physical Problems, Addison-Wesley, 1962.

Interscience, 1963.
Cotton, F. A., Chemical Applications of

Yale, P. B., Geometry and
Group Theory,
[14]
[15] [16] [17]
[18]
Symmetry, Holden Day,
1968.
Rotman,
J. J.,
The Theory of Groups:
An
Introduction, Allyn and Bacon, 1965.
Kurosh, A., The Theory of Groups, (translated by K. A. Hirsch), Chelsea, 1960.

Hall, Jr., M.,
The Theory of Groups, Macmillan, 1959.
Wielandt, H., Finite Permutation Groups, Academic Press, 1964.
[19] [20] [21]

[22]
Schenkman,
Scott, W.,
E.,
Group Theory, Van Nostrand, 1965.
Group Theory, Prentice-Hall, 1964.
Burnside, W., Theory of Groups of Finite Order, Dover, 1955.
Huppert,
Herstein,
B.,
I.,
Endliche Gruppen, Springer-Verlag, 1967.

Topics in Algebra, Blaisdell, 1964.
L.,
[23]
[24] [25] [26]
Van
der Waerden, B.
Modem
Algebra, (translated by F. Blum), Ungar, 1953.
Cohen, P. M., Universal Algebra, Harper and Row, 1965.

Northcott, D. G.,
An
Introduction to Homological Algebra, Cambridge U. P., 1960.
[27]
Wielandt, H., and B. Huppert, "Arithmetical and Normal Structure of Finite Groups," 1960 Institute on Finite Groups, Editor, M. Hall, Jr., American Mathematical Society, 1962.
Hall, Jr., M.,
[28] [29]
and
J.
K. Seniour, The Groups of Order 2" {n
6),
Macmillan, 1964.
Carter, R. W., "Simple Groups and Simple Lie Algebras", Journal of the
Society, April 1965, pp. 193-240.
London Mathematical
(translated by
[30]
Schreier, O., and E. Sperner, Introduction M. Davis and M. Haussner), Chelsea, 1959.
to
Modern Algebra and Matrix Theory,
[31]
[32] [33] [34]
Kaplansky,
Fuchs,
L.,
I.,
Infinite Groups, U. of
Michigan Press, 1954.
Abelian Groups, Akademiai Kiado, 1958.

of Finite Groups,
Burrow, M., Representation Theory

Curtiss, C. W.,
Academic Press, 1965.
and
I.
Reiner, Representation Theory of Finite Groups and Associative Algebras,
Interscience, 1962.
[35] [36]
Neumann,
H., Varieties of Groups, Springer, 1967.
Magnus, W., A. Karass and D. Solitar, Combinatorial Group Theory: Terms of Generators and Relations, Interscience, 1966.
Massey, W.
1965.
S.,
Presentations of Groups in
[37]
[38]
Algebraic Topology;
An
Introduction, Harcourt, Brace, 1967.
Coxeter, H. S. M., and
W.
O. J. Moser, Generators
and Relators for Discrete Groups, Springer-Verlag,
[39] [40]
[41] [42] [43]
Wallace, A. H.,
An
Introduction to Algebraic Topology, Pergamon, 1957.
Crowell, R. H., and R. H. Fox, Introduction to
Knot Theory, Ginn,
1963.
Pontriagin, L.
S.,
Topological Groups, Gordon and Breach, 1966.

P., 1957.
Cohn,
P.,
Lie Groups, Cambridge U.
Zassenhaus, H., The Theory of Groups, Chelsea, 1949.
[44]
Lawvere, F. W., "The Category of Categories as a Foundation for Mathematics", Proceedings of the Conference on Categorical Algebra, La Jolla, 1965, Springer-Verlag, 1966.
Feit, W.,
[45]
and
J.
G. Thompson, "Solvability of groups of odd order". Pacific Journal of Mathematics,
vol. 3, no. 3, pp. 773-1029, 19G3.

[46]
[47]
Lederman, W., Introduction

Birkhoff, G.,
to the
Theory of Finite Groups, Oliver and Boyd, 1957.
and
S.
MacLane,
Survey of Modern Algebra, Macmillan, 1953.
INDEX
Abelian group, 177-212 divisible, 205, 209 finitely generated, 196 free, 186 of type p" = p-Priifer group, 191, 206 primary = p-group, 190 rank of, 193 Abelian groupoid, 29 Additive notation, 19, 178 Alternating group, 62 simplicity, 172
Correspondence theorem, 120 Coset, 108 double-ended, 265 representation, 219 representative, 219
Countable, 14
Countably infinite, 14 Cycles, 167

Cyclic group, 101
Ascending central series

Associative groupoid, 29
upper central
series, 142
Automorphism,
group, 84, 87
inner,
Degree of permutational representation, 217 Dependent, 192 Derived subgroup, 112

Difference of sets, 4 Different products, 245
= mapping
p,
85
of field, 87 of group = of groupoid, 83
Basis of free abelian group, 186 Basis theorem = fundamental theorem for finitely generated abelian groups, 197 Bijection, 14 Binary operation, 19
Block, R-, 9
Cardinality, 14 Carrier, 26
Dihedral group, 75 infinite, 257 of degree 4, Table 5.2, 152 Dimension of vector space, 89 Direct product, 143 external, 143 internal, 146 Direct summand, 178 Direct sums, finite, 178 infinite, 182 Divisible group, 205, 209
Cartesian product, 6 Cayley's theorem, 46, 214 Center, 112

Centralizer of a subset, 112 Chain, 194
Circle, 2 Class, equation, 135 equivalence, 9
Element, 1 order of, 103
R-,9
Closed with respect to multiplication, i.e. product of elements in the set belong to the set. Codomain, 12, 13
Epimorphism, 42 Equivalence relation, 8 Extension of groups, 232-234 splitting, 234-238 Extension of mapping to homomorphism, 185, 250 Extension property, free abelian groups, 185 free groups, 250
Factor groups, 114 Factor of a factor theorem (third isomorphism
theorem), 121 Factors of a subnormal series, 158 Faithful representation, 218 Family of indexed sets, 182 Fields of complex numbers, 86 Finitely generated abelian groups, 197 subgroups of, 202 factor groups of, 204 Finitely generated group, 98, 197, 227 Finitely presented, 253 Finite presentation, 253 Four group, 144, 148
Common
part, 3
Commutative groupoid, 29 Commutator, 112 Commutator subgroup, 112 Commute: a and h commute Complement, 234 Component, p-, 191
Composition Composition Composition Congruence,
factors, 164
if
ah
ha.
of maps, 17
series, 158,
163
269
Conjugacy
classes, 171 Conjugate, subgroups, 131 Conjugates, H-, 134
274
INDEX
Free abelian group, 186 rank of, 193 Free group, 246
intersections of finitely generated subgroups, 265
275
Groups of order
12, 152-156
15, 131
(cont.)
Groups of small order, 148-156

Hamiltonian group = quaternion group, 151 Higher central series = upper central series, 142 Homomorphism, natural, 115
length of an element, 248 rank of, 255 subgroups, 260 subgroups of finite index, 264 Free on a set, 246 Free set of generators, 246 Freely generated, 186, 246 Frobenius' representation, 225 Frobenius' theorem, 224 Full linear group, 90 Fundamental theorem of arithmetic, 269 Fundamental theorem of finitely generated abelian groups, 197
Galois group, 159 Generators, free set of, 246 Greatest common divisor, 269
Homomorphism,
of a group, 94
of a groupoid, 40
(first
Homomorphism theorem
theorem), 117
Identical products, 245
isomorphism
Identity, of a groupoid, 30
Iff
of a group, 50 = if and only
if
Group,
abelian, 177-212
alternating, 62
automorphism, 83 commutative, 29 commutator, 112

cyclic, 101
Image, 12, 13 Independence, independent, 192 Index of a subgroup, 110 Index 2, 116 Indexed family, 182 Indexing set, 182 Initial segments, 261 Inner automorphism (see Automorphism.)
Intersection, 3
Intersection, of
normal subgroups, 114
definition of, 50
derived, 112
extension, 234, 238

finitely generated, 98, 197,
of subgroups, 55 Invariant subgroup = normal subgroup. 111 Invariants of finitely generated abelian group type of abelian group, 200
227
Inverse, 32
free,
246
isometry, 64, 67 linear, 90 matrices, 81 mixed, 188 Mobius transformations, 77

nilpotent, 142
Inverse pairs, 246 Inversion, 60 Isometries, of line, 64 of plane, 67
Isomorphism, of group, 94
of groupoid, 42
number
of groups of given order, 156
Isomorphism theorems,
first,
primary, = p-group, 190 Priifer groups, 191 quaternion, 151 simple, 158, 163 symmetric, 56 table, 22 torsion, 188 torsion-free, 188 Groupoid, 26
abelian, 29
associative, 29
117
second, 125
subgroup, 125
third, 121
Jordan-Holder theorem, 164

Kernel, 117 Klein four group, 144, 148
commutative, 29 equality of, 28 finite, 29

identity, 30
infinite,
Lagrange's theorem, 109 Largest set, 3 Least common multiple = lowest common multiple, 269 Length of,
composition series, 164 167 element in free group, 248 Linear fractional transformation transformation, 77 Linear group, 90
cycle,
29
order, 29
table of, 22
Groups of order
P, 148 p2, 147
= Mobius
2p, 149
8,150
Mal'cev's theorem, 230
276
INDEX
or mapping,
Map
Product,
cartesian, 6
direct, 143 reduced, 245 Products,
bijection, 14
codomain, 12, 13 composition, 17

definition, 11, 13
domain,
12, 13
different,
equality, 13
identical,
245 245
matching, 14
one-to-one, 14
onto, 14
of subsets of a group, 109
Proper subgroup, 106

Priifer group, 191 Quaternions = Hamiltonian groups, 151 Quotient group = factor group, 114
restriction of, 14
Matching, 14 Matrices, 81 Matrices, groups
of,
81
set,
Maximal independent Maximal set, 194

Metacyclic, 243 Mixed group, 188
193
Range, 12 Rank,
free abelian group, 193 free group, 255
torsion-free abelian group, 193
iE-block, 9
i2-class, 9
Mobius transformations, 77 Monomorphism, 42

Multiplication in a set, 2 Multiplication table, 22
Reduced product, 245

Reflection, 68, 69
Natural homomorphism, 115

Negative, 178 Nielsen-Schreier theorem, 260 Nilpotent group, 142 Normal closure, 253 Normal subgroup, 111
Reflexive property, 9 Representation,

coset, 219 degree of, 217 Frobenius, 225 permutational, 216 right-regular, 216 Representative of a coset, 219 Representatives of the equivalence classes, 134 Right-regular representation, 216
Normal subgroup generated by a

Normalizer, 112, 133
set,
253
Odd permutation, 60
One-to-one mapping, 1 Onto mapping, 14 Operation, binary, 19
Rotation, 68, 69
Schreier's subgroup theorem, 260
Order
of a group, 50 of a groupoid, 29 of an element, 103
Schreier transversal, 261 Schur's theorem, 242 Semigroup, 29

Series,
Ordered pair, 2
Partition, 9
ascending central,
composition, 158
solvable, 158
upper central
series,
142
p", group of type = Priifer group, 191 y-component, 191 Periodic = torsion group, 188 Permutation, 56 even, 60
subnormal, 158
Set, 1
group
odd, 60
symmetric group, 56
Permutational representation, 216 degree of, 217 p-group, 139, 190 p-primary group = p-group, 190 p-Priifer group, 191 main theorem, 206 Preimage, 12, 120 Presentation, 253 finite, 253 finitely presented, 253 group of a, 253 of a group, 253
Simple group, 158, 163, 172 Smallest set, 3 Solvable, group, 158, 161 by radicals, 159 series, 158 word problem, 246 Split, 234 Splitting extension, 234, 238 Steinitz exchange theorem, 192 Subgroup, commutator, 112
conjugate, 131 definition, 54 derived, 112
generated by a
invariant,
set,
98
normal subgroup, 111 isomorphism theorem (second isomorphism

theorem), 125 normal. 111
Primary component
p-component, 191
INDEX
Subgroup
(cont.)
277
Symmetry groups
of an algebraic structure, 83
of index 2, 116 proper, 106
Sylow, 130
theorem for cyclic groups, 105, 126 theorem for free groups, 260
torsion, 189
Subnormal
Subset, 2
series,
158
factors of, 158
subgroup generated by, 98

product
of,
Sum,
direct,
109 178
finite,
infinite,
178 182
Table, multiplication, 22 Three-cycle, 172 Torsion group, 188 subgroup, 189 Torsion-free group, 188 Transfer, 240 Transitive property, 9 Translation, 68, 69 Transposition, 167 Transversal, 219 Sehreier, 261 Type of a finitely generated abelian group, 200
Sum
of vectors, 89
Sylow subgroup, 130

Sylow's theorems, first, 130 second and third, 131 Symmetric group, definition, 56 of degree 4, 59 of degree 5 or more, 172 Symmetric property, 9 Symmetry groups, 73
Union, 3 Unit element
Upper
= identity element, 30 central series, 142
Vector space of dimension n, 89 Vector sum, 89
Word
problem, 246
Zero element, 178 Zorn's lemma, 194
Symbols and Notations
Symbols
aut (G)
o-x.g
Automorphism group of G, 84
xg(xg)-^,2.2A
Alternating group of degree n, 61
C
C*
C(A)
Complex numbers, 1 Nonzero complex numbers, 51 Centralizer of A, 112 Cyclic group of order n, 148
Dihedral group of degree n, 76
D
gp(X)
h^
/
Subgroup generated by X, 98 Conjugation by a;^, 233

Isometries of plane, 67
I(R)
Is
(/
:
Z)
Group of isometries of R, 64 Symmetry group of S, 73 Group of isometries of R that move

Klein four group, 148
integers to integers, 66
K^ Ker e
Lc.m.
(a, 6)
Kernel of
9,
117
L^(y, F)
m, g.
fc,fc'
Lowest common multiple of a and h, 269 Full linear group of dimension n, 89

233
'^x^.x^,,
M
^M
Mx
Group of Mobius transformations, 78 Group of 2 X 2 matrices over complex numbers, 81 Semigroup of mappings of X to X, 36
Nonnegative integers,
Normalizer of A, 112
1
N
N(A)
Nii{A)
Normalizer of
Rationals, 1
in
H, 133
P Q
Q*
Positive integers, 1
Nonzero rationals, 20
Real numbers, 1
R
R^
R^
5^
Jl*
Nonnegative real numbers, 48

Euclidean plane, 2
Set of representatives of the equivalence classes, 134
Representatives whose intersection with the center

iJ-equivalence class, 9
of distinct Sylow p-subgroups, 131 Symmetric group of degree n, 56 Symmetry group on X, 56 Torsion subgroup of G, 189 Elements of S and their inverses, 262
is
the
empty
set,
135
i2-class
Sp
iS
Number
Sx
T(G)
W
x^
Coset representative, 232

Integers, 1
Z
Z(G)
Center of G, 112
278
SYMBOLS AND NOTATIONS
279
Greek Symbols
Mapping that sends x
e
->
xg, 219
p Pe
Mapping that sends x
to xg, 214
,
Frobenius representation, 225

Identity mapping, 36, 56
Rotation through angle
68
a{a, b, c, d)
Mobius transformation, 77
Reflection, 69
Natural homomorphism, 114

IT
"m
Mapping that sends g

Set of all primes, 190
->
fg,
219
T
i-a.b
Transfer, 240
Translation, 68
Notations
Is in, belongs to, 1 Is not in, does not belong to, 1
{ {

Open
interval, 2
Hg
Right
coset, 108
Closed interval, 2; also Commutator, 112

(
,
XY
[G:H]
Product of two subsets of a

group, 109
Ordered pair, 2; also Greatest common amon divisor, 269
H<
G
G'
Index of H in G, 110 H is normal in G, 111
c
c
u n
Is
a subset, 2
\,y\
Commutator of x and Commutator subgroup
y,
112
Is a
proper subset, 2
of G, 112
Union, 3
Intersection, 3
GIN
over N, 114
Equivalence relation by conjugation, 134
Empty
set, 3
A~
Difference of sets, 4
Equivalence class containing A, 134
X
S"
Cartesian product, 6
Cartesian product of
HXK
Sm
times, 6
n
i
External direct product, 143

Internal direct product, 146
xRy
is
related to y hy R, 8
9,
xR
R-c\ass of X,
11
n =
,
G,
Direct product, 146
X/R
Set of /2-classes, 11; also over R, 114
(1,
m)
Cycle of length m, 167

Direct
sum
of
H and K, 178
Mapping from S
8a, S", a(8)
into T, 12
a,
2 2
Gj
Finite direct sum, 178
Image of
s
a,
under
12
12
Gi
Infinite direct
sum, 182;
Sa
\s\
Range
of
see also 179
Number
of elements of S, 14
Gp
p-component of G, 190
Priifer group, 191
|S'
Restriction of a to S', 14
a P
St' 8-t
Composition of mappings, 17; also Image under a binary composition, 19
Type of a group, 200

S
(X;R)
Transversal element in same coset as g, 219
Presentation, 253
8+t
St
Image under a binary composition, 19
\X:R\
t J
Group of a presentation, 253

Presentation, 254
8X
(a, b; [a, b])
Groupoid
1
with binary operation
n,
26
(a,b)
Identity of a groupoid, 31
Greatest a and
common
b,
divisor of
269
Inverse of the element g, 34
Congruent, 269
04124

Shaum Group Theory

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Shaum Group Theory

Uploaded by

Copyright:

Available Formats

^^,:v-.

THEORY and PROBLEMS

McGRAW-HlLL BOOK COMPANY

BENJAMIN BAUMSLAG, Ph.D. BRUCE CHANDLER, Ph.D.

SCHAVM'S OUTLIBTE SERIES

San Francisco, Toronto, Sydney

Typography by Signs and Symbols,

Although groups arose

the whole, however,

MAPPINGS AND BINARY OPERATIONS

Partitions and equivalence relations.

COMPOSITION OF MAPPINGS BINARY OPERATIONS

look back at Chapter

COMMUTATIVE AND ASSOCIATIVE GROUPOIDS AND INVERSES IN GROUPOIDS

SEMIGROUPS WITH AN IDENTITY ELEMENT

The semigroup of mappings d. The order in a product.

HOMOMORPHISMS OF GROUPOIDS AND CAYLEY'S THEOREM

morphism, c. Properties of and semigroups.

Epimorphism, monomorphism, and isoepimorphisms. d. Naming and isomorphisms, e. Mx

look back at Chapter 2

GROUPS AND SUBGROUPS

Examples of groups of numbers.

SUBGROUPS THE SYMMETRIC AND ALTERNATING GROUPS

THE GROUP OF MOBIUS TRANSFORMATIONS

Defining the group,

SYMMETRIES OF AN ALGEBRAIC STRUCTURE

Fields of complex numbers, c. AutomorThe full linear e. Linear transformations.

look back at Chapter 3

More about subgroups,

look back at Chapter 4

THE SYLOW THEOREMS

The importance of p-groups The upper central series.

The center of a p-group.

DIRECT PRODUCTS AND GROUPS OF LOW ORDER

Definition of solvable groups,

Properties of, and alternative definition

for, solvable groups.

COMPOSITION SERIES AND SIMPLE GROUPS

look back at Chapter 5

SIMPLE CLASSIFICATION OF ABELIAN GROUPS,

AND STRUCTURE OF TORSION GROUPS

groups. divisible groups.

look back at Chapter 6

Another proof of Cayley's theorem,

Cayley's theorem and examples of

PERMUTATIONAL REPRESENTATIONS DEGREE OF A REPRESENTATION AND FAITHFUL REPRESENTATIONS..

PERMUTATIONAL REPRESENTATIONS ON COSETS FROBENIUS' VARIATION OF CAYLEY'S THEOREM

APPLICATIONS TO FINITELY GENERATED GROUPS

Subg-roups of finite index, Marshall Hall's theorem, d.

Remarks about the proof of Theorem One consequence of Theorem 7.5.

of the choice of transversal,

a homomorphism. c. Proof that t d. A theorem of Schur.

look back at Chapter 7

GROUPS AND PRESENTATIONS

look back at Chapter 8

A GUIDE TO THE Lil LITERATURE

SYMBOLS AND NOTATIONS

Mappings and Binary Operations

numbers, for the addition of a single number

meaningless, while the addition of three

Set the set. denote

the set of nonnegative integers