(1998) - Wallace David A.R. Groups, Rings and Fields

Springer Undergraduate Mathematics Series
Springer-Verlag London Ltd.

Advisory Board
Professor P.T. Cameron Queen Mary and Westfieid College

Dr M.A.T. Chaplain University 01 Dundee
Dr K. Erdmann Oxford University
Professor L.C.G. Rogers University 01 Bath
Dr E. Süli Oxford University
Professor T.F. roland University 01 Bath
Other books in this series
Basic Linear Algebra

T.S. Blyth and B.P. Robertson
3-540-76122-5
Multivariate Calculus and Geometry

Sean Dineen
3-540-76176-4
Elements of Logic via Numbers and Sets

D.L. lohnson
3-540-76123-3
Elementary Number Theory

Gareth A. Iones and I. Mary Iones
3-540-76197-7
Introductory Mathematics: Applications and Methods

G.S. Marshall
3-540-76179-9
Vector Calculus
P. C. Matthews
3-540-76180-2
Introductory Mathematics: Algebra and Analysis

GeoffSmith
3-540-76178-0
D.A.R. Wallace
Groups, Rings
and Fields
With 15 Figures
Springer
D.A.R Wallace, BSc, PhD, FRSE
Dept. of Mathematics, Strathclyde University, Livingstone Tower, 26 Richmond Street,
Glasgow G1 lXH, UK
Cover iUustration elements reproduced by kind permission of:
Aptech Systems, Inc., Publishers of the GAUSS Mathematical and Statistical System, 23804 S.E. Kent·Kangley Road, Maple Valley, WA 98038,
USA. Tel: (206) 432 . 7855 Fax (206) 432 . 7832 email: [email protected]: www.aptech.com
American Statistical Association: Chance Vol 8 No I, 1995 article by KS and KW Heiner 'Tree Rings of the Northern Shawangun!<s' page 32 fig 2
Springer·Verlag: Mathematica in Education and Research Vol 4 Issue 3 1995 article by Roman E Maeder, Beatrice Amrhein and Oliver Gloor
'Illustrated Mathematics: Visualization of Mathematical Objects' page 9 fig II, originally published as a CD ROM 'Illustrated Mathematics'
ISBN 978·3·540·76177·8
Mathematics in Education and Research Vol 4 Issue 3 1995 article by Richard J Gaylord and Kazume Nishidate 'Traffic Engineering with
Cellular Automata' page 35 fig 2. Mathematica in Education and Research Vol 5 Issue 2 1996 article by Michael Trott 'The Implicitization
of a Trefoil Knot' page 14.
Mathematica in Education and Research Vol 5 Issue 2 1996 article by Lee de Cola 'Coins, Trees. Bars and Bells: Simulation of the Binomial
Process page 19 fig 3. Mathematica in Education and Research Vol 5 Issue 2 1996 article by Richard Gaylord and Kazume Nashidate
'Contagious Spreading' page 33 fig J. Mathematica in Education and Research Vol 5 Issue 2 1996 article by Joe Buhler and Stan Wagon
'Secrets of the Madelung Constant' page 50 fig J.
British Library Cataloguing in Publication Data

Wallace, David A.R.
Groups, rings and fields. - (Springer undergraduate mathematics series; 3423)
1. Group theory 2. Rings (Algebra) 3. Algebraic fields
I. Title
512
Library of Congress Cataloging-in-Publication Data

Wallace, D.A.R. (David Alexander Ross)
Groups, rings, and fields/D.A.R. Wallace.
p. cm. - (Springer undergraduate mathematics series)
Includes index.
ISBN 978-3-540-76177-8 ISBN 978-1-4471-0425-4 (eBook)
DOI 10.1007/978-1-4471-0425-4
1. Algebra, Abstract. I. Title. II. Series.
QA162.w36 1998 98-4961
512'.02-dc21 CIP
Apart from any fair dealing for the purposes of research or private study, or criticism or. review, as permitted
under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or
transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in
the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright
Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publishers.
© Springer-Verlag London 1998
Originally published by Springer-Verlag London Limited in 1998
The use of registered names, trademarks etc. in this publication does not imply, even in the absence of a
specific statement, that such names are exempt from the relevant laws and regulations and therefore free
for general use.
The publisher makes no representation, express or implied, with regard to the accuracy of the information
contained in this book and cannot accept any legal responsibility or liability for any errors or omissions that
may be made.
Typesetting by BC Typesetting, Bristol
12/3830-54321 Printed on acid-free paper 10838358

Contents
Preface vii
Chapter 1 Sets and Mappings '" 1

1.1 Union and Intersection.... 1
1.2 Venn Diagrams.......................................................................... 8
1.3 Mappings................................................................................... 17
1.4 Equivalence Relations .. 27
1.5 Well-ordering and Induction 34
1.6 Countable Sets 41
Chapter 2 The Integers 47

2.1 Divisibility 47
2.2 Divisors...................................................................................... 50
2.3 Division Algorithm 52
2.4 Euclidean Algorithm................................................................. 56
2.5 Primes............................................................................ 62
Chapter 3 Introduction to Rings....... 71

3.1 Concept of a Polynomial 72
3.2 Division and Euclidean Algorithms 76
3.3 Axioms and Rings........................................ 82
v
vi Contents
Chapter 4 Introduction to Groups... 93

4.1 Semigroups 94
4.2 Finite and Infinite Groups......................................................... 99
4.3 Subgroups 115
4.4 Lagrange's Theorem, Cosets and Conjugacy 127
4.5 Homomorphisms 134
Chapter 5 Rings 145

5.1 Arithmetic Modulo n 145
5.2 Integral Domains and Fields 150
5.3 Euclidean Domains 163
5.4 Ideals and Homomorphisms 168
5.5 Principal Ideal and Unique Factorization Domains 177
5.6 Factorization in Q[x] 186
Chapter 6 Topics in Group Theory 191

6.1 Permutation Groups 191
6.2 Generators and Relations 204
6.3 Direct Products and Sums 210
6.4 Abelian Groups 213
6.5 p-Groups and Sylow Subgroups 218
Hints to Solutions 225
Suggestions for Further Study 245
Index 246
Preface
This text is written for students who are meeting abstract algebra for the first
time. Many of these students will take only one course of abstract algebra,
whereas others may proceed to more advanced courses. The aim here is to pro-
vide an appropriate, interesting and entertaining text for those who require a
rounded course as well as for those who wish to continue with further studies
in algebra.
A fundamental difficulty for beginning students is often the axiomatic nature
of abstract algebra and the exacting need to follow the axioms precisely. The pre-
sent text is written so that an axiomatic treatment should seem to evolve as a
natural development of intuitive ideas. To this end particular and extensive
attention is paid to the integers, which are familiar objects of knowledge, and
which, together with some simple properties of polynomials, are used to give
motivation for the introduction of more abstract algebraic concepts. Historical
allusions are made throughout the text in order to emphasise that abstract
modem algebra has evolved from sometimes quite rudimentary ideas. Other
remarks are intended to broaden, in a light-hearted manner, the student's
general education.
Each section is provided with copious examples and ends with a set of exercises
for which solutions are provided at the end of the book. In order to render the
text suitable for self-study, any arguments in the text have been carefully crafted
to facilitate the understanding, and to promote the enjoyment, of the reader.
The text is self-contained and no prerequisites are absolutely necessary although
a 'nodding acquaintance' with complex numbers and with matrices would, on
occasion, be of advantage.
vii
viii Preface
I must express my profound thanks to the members of the editorial team of

Springer-Verlag and to my wife, Rachel Henderson, for her many encourage-
ments and for her indefatigable efforts in turning a scribbled manuscript into
an elegant word-processed document.
Finally, I am grateful to those correspondents who drew my attention to some
textual peccadilloes, which have now been corrected.
D.A.R. Wallace
Department of Mathematics
University of Strathclyde
March 2001
1
Sets and Mappings
The notions of a 'set' and of a 'mapping' are fundamental in modern mathe-

matics. In many mathematical contexts a perceptive choice of appropriate sets
and mappings may lead to a better understanding of the underlying mathe-
matical processes. We shall outline those aspects of sets and mappings which
are relevant to present purposes and, for the delectation of the reader, conclude
with a few logical paradoxes in regard to sets.
1.1 Union and Intersection

We are accustomed to speak of a 'collection' of books, or of an 'assembly' of
people or of a 'list' of guests. In mathematics the word 'set' is used to denote
the basic concept which is expressed by each of the collective nouns; in mathe-
matics we say simply a 'set' of books, or a 'set' of people or a 'set' of guests.
Definition 1
A collection or assembly of objects is called a set. Each object is said to be an
element of the set.
Thus in the phrase 'a set of books', each book involved in the set is an element of
the set. Frequently the symbols constituting the elements of a (mathematical)
1
2 Groups, Rings and Fields
set will be letters or numbers. If a set A consists of, say, the elements a, b, c then
we write A = {a, b, e}, the use of curly brackets being customary. Notice that as
we are concerned only with membership of the set we disregard any repetition of
the symbols and have no preferred order for writing them down.
Example 1
Let A have 1, 2, 3, 4 as elements. Then A={1,2,3,4}={I,I,2,2,3,4}=
{2,1,3,4} = {1,3,4,2} etc.
Definition 2
Let A be a set. If A consists of a finite number of elements, say all a2, ... , aN
(where the notation presumes the elements are distinct) then we write
A={al,~, ... ,aN}
and A is said to be a finite set of cardinality N, written N = IAI. If A does not

have a finite number of elements A is called an infinite set and is said to have
infinite cardinality. The set consisting of no elements at all is called the empty
set or the null set and is denoted by 0 (0 is a letter of the Norwegian alphabet).
Evidently 101 = o.
Examples 2
1. 0 = set of all unicorns.
2. 0 = set of all persons living on the moon in 1996.
3. 1{1, 2,3,4}1 = 4.
4. l{a,b,e,d,e}1 = 5.
A set is often given by some property which characterizes the elements of the
set. Before giving a definition we offer a colourful example.
Example 3
Let A be the set of the colours of the rainbow. Whether or not a given colour is in
A is determined by a property, namely that of being one of the colours of the
rainbow. Of course in this particular instance we may write down the elements
explicitly. Indeed
A = {red, orange, yellow, green, blue, indigo, violet}.
Sets and Mappings 3
Definition 3
Let P be some property. The set of elements, each of which has the property P,
is written as
A = {xix has the property P},
which we read as 'A is the set of all x such that x has the property P'.
Examples 4
1. Let A consist of the squares of strictly positive integers. Then
A = {xix = n 2 ,n = 1,2, ...}.

The set A may also be written as
A= {1,4,9, ... },
but notice that this notation is ambiguous since it would only be an inference
that A consists of squares of integers.
2. The set {xix < 0 and x > I} is the empty set since there is no number which
is simultaneously less than 0 and greater than 1.
2
3. The set {xlx = 3, x is an integer} is empty since there is no integer with
square equal to 3.
Definition 4
Let A be a set. Let x be an element. If x is an element of A we write
xEA,
and if x is not an element of A we write
x~A.
Example 5
a E {a,b,c}, b ~ {1,2,3},
36 E {xix = n 2 ,n = 0,1,2, ...}, 37 ~ {xix = n 2 ,n = 0,1,2, ...}.
Definition 5
A set A is said to be a subset of a set B if every element of A is also an element of
B. We write
A~B
to indicate that A is a subset of B and read the notation as 'A is contained in or

equal to B'. The empty set is deemed to be a subset of every set. If A is a subset of
B but A =I B then A is said to be a proper subset of B.
If A is not a subset of B then there exists at least one element of A which is not
an element of B and we write
A g; B.
If A ~ B and B ~ A then the sets A and B are equal and we write
A=B.
If A and B are not equal we write
A =lB.
If A ~ B and A =I B we write A c B.
Given two sets A and B, to prove that A ~ B we have to show that x E A
implies x E B. To prove that A g; B, we have to exhibit an x E A such that
x~B.
Two sets A and B are equal if and only if they are elementwise indistinguish-
able.
Examples 6
1. Let A = {a,c},B = {a,b,c,d}, C = {a,b,c}. Then A ~ B, C ~ B.
2. Let A = {xlx 2 = I}, B = {-1, I}. Although A and B are described in differ-
ent ways they have the same elements and so A = B.
Definition 6
Applications of the theory of sets usually take place within some fixed set (which
naturally varies according to the circumstances). This fixed set for the particular
application is called the universal set and is often denoted by U. The subset of U
consisting of the elements of U having the property P is denoted by
{x E Ulx has the property P},
or simply as
{xix has the property P}
if the universal set is obvious. Frequently explicit reference to a universal is
omitted.
We now introduce the operations of union and intersection on sets.

Sets and Mappings 5
Definition 7
Let A, B, C, ... be sets. The set consisting of those elements each of which is in at
least one of the sets A, B, C, ... is called the union of A, B, C, ... and is written
AUBUCU ...
The set consisting of those elements each of which is in all of the sets A, B, C, ...
is called the intersection of A, B, C, ... and is written as
AnBnCn ...
The sets A, B, C, ... are said to be disjoint if the intersection of any two distinct
sets is empty, that is, An B = An C = B n C = ... = 0.
The union A U B U CU ... is said to be a disjoint union if the sets A, B, C, ...
are disjoint.
Example 7
Let U = {a, b, c, d, e, f, g, h, i} be the universal set. Let A = {a, b, c}, B =
{d,f, i}, C = {a, b,c,d, f,g}, D = {a,e,g, h, i}, E = {e,g, h}.
Then
AUB= {a,b,c,d,f,i}, AnB= 0,
CuD = {a,b,c,d,e,f,g,h,i} = U, enD = {a,g},
AuBuC= {a,b,c,d,f,g,i}, AnBnC= 0,
Bu CuD = {a,b,c,d,e,f,g,h,i} = U, BnCnD = 0,

U= AUBUE,
where this union is disjoint as An B = An E = B n E = 0.
It is immediate that for sets A and B we have A n A = A, A uA = A and
AnB= BnA ~ A ~ AuB= BuA.
We prove a useful lemma.
Lemma 1
Let A, B, X be sets such that A ~ B. Then the following hold.
1. AnX ~ BnX.
2. AUX~ BUX.
Proof
1. We have to prove that every element of A n X is also in B n X.
Let c E A n X. Then c E A and so, as A ~ B, c E B. But c E X and so
c E BnX. HenceAnX ~ AnX.
2. Let c E AuX. Then either c E A or c E X. If c E A then c E B and so, in
either case, c E B U X. Thus A u X ~ B u X. D
Definition 8
Let A be a subset of the universal set U. The subset of U consisting of all elements
of U not in A is called the complement (note spelling!) of A. The complement of
A is denoted by U \ A or by A' if the universal set is clear.
U\ A = A' = {x E Ulx ¢ A}
(Another notation for the complement, but not used in this text, is CA.)
U=(U\A)UA, (U\A)nA=0,
and so
A = U if and only if A' = 0.
Example 8
Let U = {I, 2, 3, 4, 5}, A = {2, 3}. Then U \ A = {I, 4, 5}.
The notion of complement extends to that of relative complement.
Definition 9
Let A and B be sets. The relative complement of A in B, denoted by B \ A, is the
set of elements of B which are not in A.
Example 9
Let A = {p,q,r},B = {q,r,s,t}. Then B\ A = {s,t},A \ B = {pl.
We conclude this section with a fairly obvious but useful result.
Sets and Mappings 7
Theorem 1
Let A and B be sets. Then the following statements are equivalent.
1. AS;B.
2. AnB=A.
3. AUB= B.
Proof
We prove the equivalence of 1 and 2, leaving the equivalence of 1 and 3 as an
exercise.
Suppose A S; B. Then, by Lemma 1,
A = A n A S; A nBS; A, and thus
A=AnB.
Conversely if A = An B then certainly A S; B. D
We conclude this section by introducing the 'Cartesian product' of sets. The

adjective 'Cartesian' derives from the name of R. du P. Descartes (1596-1650),
mathematician and philosopher whose philosophical outlook is enshrined in
his famous aphorism "je pense, donc je suis", in English "I think, therefore
I am". (Quotation from Discours de la Methode [Leyden, 1637, 4th part]).
Definition 10
Let A and B be two sets. The Cartesian product A x B of A and B is defined to
be the set of all ordered pairs of the form (a, b), a E A, bE B;
A x B = {(a, b)la E A, bE B}.
The Cartesian product Al x A 2 X ••. x An of the sets AI' A 2, ... , An (in this
order) is defined to be the set of all ordered n-tuples of the form
(all~'··· ,an), al E All a2 E A 2,··· ,an E An,
Note that if the sets A and B are distinct, then A x B oF B x A.
Example 10
Let A and B be finite sets of m and n elements respectively. Then A x B is a set
of mn elements since in the ordered pair (a, b) there are m possibilities for a, and
n possibilities for b.
Exercises 1.1
1. Let the universal set U be the set {u,v,w,x,y,z}. Let A = {u,v,w},
B = {w,x,y}, G = {x,y, z}. Write down the subsets: AU B, An B,
AnG, AUG, A\B, B\A, A\G, G\A, A',B',G', A\(BUG),
B \ (A n G), (A n B) U (B \ A).
2. Let the universal set U be the set {a, b, c, 1,2, 3}. Let X = {a, b, I},
Y = {b, 2, 3}, Z = {c, 2, 3}. Write down the subsets: Xu Y, X nY,
XUZ, xnz, X\Y, Y\X, X\Z, Z\X, X',Y',Z', (xnY)',
X' U y', (XU Z)', X' n Z'.
3. Let the universal set U be the set {I, 2, 3, 4, 5, 6}. Let A = {I, 2, 4, 5}.
Let X be a subset of U such that A U X = {I, 2, 4, 5, 6} and A n X =
{2, 4}. Prove that there is only one possibility for X, namely {2, 4, 6}.
4. Let A and B be sets. Prove that A ~ B if and only if B = A U B
(Theorem 1).
5. Let A be the set {3, 5, 8, ... }. Is A the set {3, 5, 8,13, ... } or the set
{3, 5, 8, 12, ...} or neither?
6. Let Ai be a finite set of k i elements, i = 1, 2, ... ,n. Prove that
Al x A 2 X ..• x An is a finite set of k 1 k 2 ..• k n elements.
1.2 Venn Diagrams

In this section we shall describe a technique which, in certain circumstances,
enables us to visualize sets and their unions and intersections etc. by drawing
pictures on paper. The resulting pictures are known as Venn diagrams, after
J. Venn (1834-1923).
We proceed as follows. For the universal set U we draw a rectangle and for
arbitrary subsets A, B of U we draw shapes as below:
Figure 1.1.
Sets and Mappings 9
The shaded area represents AU B and the dark-shaded area A n B. Conveni-

ently we may assign numbers to the regions:
~
~
4
Figure 1.2.
Thus
region 1 represents An B,
region 2 represents A \ (A n B),
region 3 represents B \ (A n B),
region 4 represents U \ (A U B).
The Venn diagram makes clear that A is the disjoint union of the two subsets
A \ B and An B, and A U B is the disjoint union of the three subsets A \ B,
B\A and AnB.
Wise use of Venn diagrams may lead to general results for which a rigorous
proof may later be obtained. We give two examples of this procedure in the
following examples.
Example 11
Using the above Venn diagram we may observe the following:
A is represented by regions 1, 2
A' 3,4
B 1,3
B' 2,4
(A n B)' 2,3,4
A'UB' 2,3,4
Thus, from our Venn diagram representation, we conclude that
(An B)' = A'UB'.
Similarly we conclude that

(A u B)' = A' nB'.
A rigorous proof will be given in Theorem 3.
IT now C is a third subset of U then the Venn diagram becomes:
Figure 1.3.
The shaded region represents A n B n C.

For four or more subsets of U the method has fundamental limitations (as
those attempting to draw satisfactory diagrams of four subsets will discover)
and should be avoided.
Example 12
Let A, B, C be subsets of the universal set U and suppose A ~ C, B ~ C. The
appropriate Venn diagram is:
Figure 1.4.
Example 13
Let us draw a Venn diagram to illustrate the result that if X ~ Y then
AnX~AnY.
Sets and Mappings 11
Figure 1.5.
Consider a Venn diagram with three subsets A, B, G of the universal set U and
with regions assigned numbers as follows:
Figure 1.6.
region 1 represents A n B n G
region 2 represents (A n B) \ (A n B n G)
region 3 represents (A n G) \ (A n B n G)
region 4 represents (B n G) \ (A n B n G)
Example 14
In the above diagram, suppose we are asked to determine the regions correspond-
ing to the following subsets:
BUG,An (BUG),AnB,AnG,(AnB) U (An G).
Then
BuG is represented by regions 1,2,3,4,6,7
An (BUG) 1, 2, 3
AnB 1, 2
AnG 1,3
(AnB) U (An G) 1,2,3
We notice that, from our Venn diagram representation,

An(BUG) = (AnB)U(AnG).
Similarly we may conclude that
Au (BnG) = (A UB) n (A UG).
A rigorous proof is given in Theorem 2.
Theorem 2
Let A, B, G be sets. Then
1. An(BUC)=(AnB)U(AnG)
2. Au (B n C) = (A U B) n (A U G)
(These results are called the Distributive Laws for sets.)
Proof
We prove 1, the proof of 2 is similar.
Since Be; BuG and Ge; BUG, we have AnBe; An(BUG) and
A n C e; A n (B U G) from which
e; An (B U C).
(A n B) U (A n G)
We now have to prove that An (B U C) e; (A n B) U (A n G). For this part of
the proof we consider elements and not simply subsets. Therefore let
x E An (B U C). Then x E A and x E B U C and so x is in either B or G. If
we suppose x E B then x E B and x E A from which we have x E A n Band
so x E (A n B) U (A n G). The alternative supposition that x E G leads to the
same conclusion. Consequently we have
An (B U G) e; (A n B) U (A n C)
This completes the proof. o
The next result (see Example 11) gives the so-called De Morgan Laws, after
A. De Morgan (1806-71).
Theorem 3 De Morgan Laws

Let A, B be sets. Then the following statements hold.
1. (A n B)' = A' U B'.
2. (AUB)'=A'nB'.
Proof
We prove 1, the proof of 2 is similar.
Let x E (A n B)'. Then x ¢ (A n B). This means that x ¢ A or x ¢ B. If x ¢ A
then x E A' and so x E A' U B'. If x ¢ B then x E B' and so x E A' U B'. Thus
(AnB)' ~ A'UB'.
Let x E A' U B'. Then x E A' or x E B'. If x E A' then x ¢ A and so certainly
x ¢ A n B and similarly if x E B' then x ¢ An B. Thus x ¢ An B and so
x E (A n B)'. Thus A' U B' ~ (A n B)'. Hence we obtain the desired result. 0
Many of the sets with which we shall be concerned will be infinite, but for finite
sets we now introduce some counting techniques.
If A and B are both finite sets then certainly A U B is finite since there cannot
be more elements than in A and B considered separately. We shall derive a
formula for the number of distinct elements in AU B. We recall that IA I is the
number of elements in A (Definition 2).
Theorem 4
Let A and B be finite sets. Then A U B is a finite set and
IAUBI = IAI +IBI-IAnBI·
Proof
Let us examine the Venn diagram for A U B. As we remarked
A = (A \ B) U (A n B),
B = (B \ A) U (A n B),
Au B = (A \ B) U (B \ A) U (A n B)
are disjoint unions of subsets of A or B. These subsets are necessarily finite and
so, by obvious (and legitimate) counting, we have
IAI = IA \BI + IAnBI,
IBI = IB\AI + IAnBI
and also
IAuBI = IA \BI +IB\AI +IAnBI
= [IAI-IAnBI) + [IBI-IAnBI) + IAnBI
= IAI + IBI-IAnBI· 0
The following two examples illustrate the use of this formula.
Example 15
Let A = {a,b,c,d},B = {a,c,e,f,g}.
Then Au B = {a,b,e,d,e,f,g},An B = {a,e}. As we expect
IAUBI = 7= 4+5 - 2 = IAI + IBI-IAnB/.
Example 16
We are told that in a party of 95 English-speaking schoolchildren there are 20
who speak only English, 60 who can speak French and 24 who can speak
German. We require to determine how many speak both French and German.
The 'universal set' consists of 95 schoolchildren. We let F and G be the subsets
consisting of those schoolchildren who can speak French and German respec-
tively. Then IF I = 60, IG I = 24. F U G is the subset of those schoolchildren
who speak either French or German. Since 20 speak only English we have
IFUGI = 95 - 20 = 75.
Then IFUGI = IFI + IGI-IFnGI
and so 75 = 60 + 24 -IF n GI.
Thus IFnGI = 9.
This says that precisely 9 schoolchildren can speak both French and German.
We may extend Theorem 4 to consider three or more sets, but we confine our
extension to the case of three sets.
Theorem 5
Let A, B and C be finite sets. Then Au B U C is a finite set and
IAUBUCI = IAI + IBI + ICI-IAnBI-IAncl-IBncl + IAnBnC/.
Proof
We apply Theorem 4. We note that
(A n B) n (A n C) = An B n C.
Then, on letting P = B U C, we have

IAUBUCI = IAUPI
= IAI + !P1-IAnPI
= IAI + IBUcl-IAnpl
=IAI + IBI + ICI-IBnCI-IAn(BUC)1
= IAI + IBI + ICI-IBn CI-I(A n B) U (A n C)I (by Theorem 2)
= IAI + IBI + ICI-IBncl- [IAnBI + IAncl
-I(A n B) n (A n C)1l
= IAI + IBI + ICI-IAnBI-IAnCI-IBncl + IAnBncl·
o
Applications of this result are often made to problems of which the following is
typical.
Example 17
In a gathering of 136 fairly athletic students, 92 engage in gymnastics, 68 are
swimmers and 78 are tennis players. Of the 136 students, 41 do gymnastics
and swim, 43 do gymnastics and play tennis and 24 swim and play tennis. IT
4 students participate neither in gymnastics, swimming nor tennis, how many
indulge in all three of these sporting activities? How many play tennis but do
not have other sporting activities?
We have to reduce the situation of the problem to manageable mathematics.

Let U be the set of 136 students,
let G be the set of gymnasts,
let S be the set of swimmers, and
let T be the set of tennis players.
We are given the information that
IGI = 92, lSI = 68, ITI = 78, IGn SI = 41, IGnTI = 43, ISnTI = 24.
We are also given that IU \ (G U S U T) I = 4 from which we deduce that
IGUSUTI = IUI-IU\ (GuSuT)1
= 136-4 = 132
16 Groups. Rings and Fields
Now applying Theorem 5 we have
132 = IGUSUTI = IGI + lSI + ITI-IGnSI-IGnTI + IGnSnTI

= 92 + 68 + 78 - 41 - 43 - 24 + IG n S n T I.
Hence IG n S n TI = 2, and so 2 students engage in all three activities.

The number of students who play tennis but have no other sporting activities
is given by IT \ (G U S) I. The easiest way to determine this number is by means
of a Venn diagram in which IG n S n T I = 2, a is the number of students who are
gymnasts and play tennis but do not swim, and b is the number of students who
swim and play tennis but are not gymnasts.
Figure 1.7.
We have
a= I(GnT)\SI = I(GnT)\(GnSnT)1
= IGnTI-IGnSnTI
=43- 2
= 41.
Similarly
b=l(snT)\GI
=24-2
=22.
Then, from the diagram, the number of students who play tennis but have no
other sports is
IT I - a - b - 2 = 78 - 41 - 22 - 2
= 13.
Exercises 1.2
1. Let A and B be sets. Establish the De Morgan Law
(A u B)' = A' nB'
(i) by a Venn diagram,
(ii) by a set-theoretic proof.
2. Let A, B and C be sets. Establish the Distributive Law
AU (B n C) = (A U B) n (A U C)
(i) by a Venn diagram,
(ii) by a set-theoretic proof.
3. Let A, Band C be sets. Prove, by constructing an appropriate example
(often called a counter-example) that the result
AU (B n C) = (A U B) n C
does not necessarily hold. If A ~ C does the result hold? Is the con-
dition A ~ C necessary and sufficient for the result to hold?
4. In a certain village of 1000 well-educated people, 250 read the
Economist, 411 read the Bangkok Post and 315 read the Straits
Times. Of these people, 72 read the Economist and the Bangkok
Post, 51 read the Economist and the Straits Times and 31 read the
Bangkok Post and the Straits Times. If only one person reads all three
of the publications, how many do not read any of the publications?
5. A party of 200 schoolchildren is investigated with regard to likes and
dislikes of three items, namely, ice cream, sweets and fizzy lemonade. It
is found that of those children who like ice cream only 7 dislike sweets.
110 children like sweets and 149 like fizzy lemonade. 80 children like
both sweets and fizzy lemonade and 36 like both ice cream and lemon-
ade. If 31 children consume all three items avidly, how many children
like none of these items?
1.3 Mappings
In our mathematics we have probably evaluated expressions such as x 2 , sin x,
and Jx 2 - 4 for given 'values of x'. We may have described these expressions
loosely as being functions of x. While this description is somewhat vague it
does encapsulate a concept which we would wish to make rather more precise.
For a given value of x we expect unique evaluations of the expressions; thus
(_2)2 = 4, sin 1r/4 = 1/...;2, but it is not permissible to put x = 1 in JX 2 - 4 if we

wish to obtain a real square root. This suggests that we have to specify how the
'values of x' may be chosen. We begin to appreciate that, in each case, we need
to specify a set of elements and to have the function defined exactly on this set.
We make the following definition.
Definition 11
Let A and B be sets. Let there be a rule or prescription, denoted by I, by which to
each element a of A there is assigned a unique element, denoted by I(b), of B.
Then the rule is said to be a mapping or map or function from A to B. We write
a --+ I(a) (a E A)
and, to indicate that 1 is a mapping on sets, we use both of the following
notations:
1 : A-+B and ALB.

A is called the domain of 1 and B is called the codomain of f.
[The designations 'map' and 'function' are perhaps more common in topology
and analysis respectively. We shall use the term 'mapping'.]
For a E A, I(a) is called the image of a and, likewise, the subset of B given as
{f(a)la E A}
is called the image of A or, sometimes, the range of I. We also denote this subset
by I(A). Finally we note that 1(0) = 0.
Example 18
Let A = {a,b,c} and B = {O, I}. We define a mapping 1 where I: A --+ B by
letting
I(a) = 0, I(b) = 1, I(c) = 1.
Then 1 has domain A, codomain B and range B. We define a second mapping 9
where 9 : A --+ B by letting
g(a) = I,g(b):= I,g(c) = 1.
Then 9 has domain A, codomain B and range {I}. We note that 1 and 9 are
unequal since, for example,
I(a) = °"11 = g(a).
The condition under which two mappings are equal should be fairly obvious
but, for the sake of completeness, we state the condition formally.
Remark
Two mappings j and 9 are equal if and only if j and 9 have the same domain A
and j(a) = g(a) for all a E A.
Example 19 (Continued from Example 3)

Let A be the set of the colours of the rainbow. Let B = {I, 2, 3, 4, 5, 6, 7}. We
define a mapping j where j : A --+ B by the rule that the image of a given
colour of the rainbow is the number of letters in that colour. We need not go
further to specify the mapping although, in fact, we have
j(red) = 3,j(orange) = 6,j(yellow) = 6,j(green) = 5,
j(blue) = 4, j(indigo) = 6, j(violet) = 6.
We now introduce certain important sets of which we shall have more to say
in the next chapter.
Notation
The set of natural numbers, which is also the set of strictly positive integers, is
denoted by N (N for 'natural'),
N = {1,2,3, ...}.
The set of integers is denoted by Z (Z for 'Zahl', the German word for 'number'),
Z = { ... , -2, -1,0, 1,2, ...}.
The set of rational numbers, which is also the set of quotients of integers, is
denoted by Q (Q for 'quotient'),
Q = {minim, nEZ, n i' O}.
i,
Thus Q consists of all fractions such as!, =p.
The set of real numbers is denoted by lR. and the set of complex numbers is
denoted by C,
C = {a + ib : a, b E lR.}.
Evidently
N~Z~Q ~ lR. ~ C.
A real number which is not rational is said to be irrational, lR. \ Q is the set of
irrational numbers.
We give three examples of mappings involving these sets before giving some
useful terminology for describing mappings.
Examples 20
1. The mapping f: N ~ {-I, I} given by
f(n) = (-It (n E N)
has domain N, codomain {-I, I} and range {-I, I}.
2. The mapping f : R ~ R given by
f(x)=x 2 +1 (xEJR)
has domain JR, codomain JR and range {yly;::: I}.
3. The mapping f : N ~ Q given by
f(n) = ~n (n E N)
has domain N, codomain Q and range H, 1, ~,2, ...} = N u {~In EN}.
Definition 12
Let A and B be sets. Let f :A ~ B be a mapping.
(i) The mapping f is said to be injective, or onEHlne, if whenever f(al) = f(a2)
for aI, a2 E A then necessarily al = a2'
(ii) The mapping f is said to be surjective or onto if for each bE B there exists
a E A such that f(a) = b.
(iii) The mapping f is said to be bijective or onEHlne and onto if f is both
injective and surjective.
Examples 21
1. Let A = {a,b,c} and B = {p,q}. We define f: A ~ B by
f(a) = p, f(b) = p, f(c) = p.
Then f is not injective since f(a) = f(b) but a =I- b. We also have that f is not
surjective since f(A) = {p} =I- B.
2. Let A = {a,b} and B= {p,q,r}. We define f: A ~ Bby
f(a) = p, f(b) = q.
Then f is injective since f(a) =I- f(b) but is not surjective since f(A) =
{p,q} =I- B.
3. Let A = {a,b,c} and B = {p,q}. We define
f: A ~ B by f(a) = p, f(b) = p, f(c) = q.
Then f is not injective since f(a) = f(b) but is surjective since f(A) =
{p,q} = B.
[The reader may begin to suspect (correctly) that a bijective mapping exists
between two finite sets if and only if they have the same number of elements.
In particular the reader may rightly conclude that a bijective mapping cannot
exist between a finite set and a proper subset; for infinite sets the situation is
less restrictive, as the next example shows.]
4. Let 2N denote the set of strictly positive even integers,
2N = {2nln EN}.
Now 2N is a proper subset of N but, nevertheless, the mapping 1 : N -+ 2N

given by
l(n)=2n (nEN)
is bijective.
5. A mapping 1 is defined on N by
n (n odd)
I(n) = { !n (n even).
Suppose we are asked to describe f. We note first that 1 is defined so that

I: N -+ N. Then 1 is not injective since, for example, 1(3) = 3 = 1(6). But
1 is surjective since if kEN then 1(2k) = k.
Frequently we have to consider the effect of applying mappings successively.

More precisely we have the following definition.
Definition 13
Let A, B and C be sets and let 1 : A -+ B and g : B ~ C be given mappings. We
have
where the range of 1 is a subset ofthe domain of g. Then for each a E A we have a
mapping, denoted by g 0 I, called the circle-composition of the mappings g and 1
and given by
(g 0 f)(a) = g(J(a))
in which we first apply the mapping 1 to a E A and then apply the mapping g to
I(a) E B. If no ambiguity will arise we often write gl for go f.
Theorem 6
Let A, B and C be sets and let f :A - B and 9 : B - C be mappings.
1. If f and 9 are surjective then 9 0 f is surjective.
2. If f and 9 are injective then 9 0 f is injective.
3. If I and 9 are bijective then 9 0 I is bijective.
Proof
We prove 1 and leave 2 to be proved in subsequent Exercises. 1 and 2 imply 3. Let
c E C. Then since 9 is surjective we have bE B such that g(b) = c. Since I is
surjective we have a E A such that f(a) = b. Then
(g 0 I)(a) = g(f(a)) = g(b) = c
and so 9 0 I is surjective. o
Examples 22
1. Let A={a,b,c,d}, B={1,2,3}, C={p,q,r,s}. Let mappings I, 9 be
defined by f(a) = 1, I(b) = 1, I(c) = 2, I(d) = 3, g(l) = p, g(2) = r,
g(3) = r. Then we have
with, for example,

(g 0 I)(b) = g(f(b)) = g(l) = p,
(g 0 J)(d) = g(f(d)) = g(3) = r.
2. Let mappings f and 9 be defined by

I(x) = sin x (x E JR), g(x) = x 2 (x E JR).
Then we may form go I and also log:
! 9
JR -JR -JR,
9 !
JR -JR -JR,
Now
(g 0 J)(x) = g(f(x)) = g(sinx) = (sinx)2 = sin2 X
but
(f 0 g)(x) = I(g(x)) = f(x 2 ) = sinx2.
go I and log have the same domain but are different mappings since it is not
true, contrary to an occasional misguided belief, that sin2 x = sin x 2 for all
x E JR.
Consider the case of four sets A, B, C, D and three mappings I, g, h where

I : A -+ B, 9 : B -+ C and h : C -+ D.
We may construct the mappings 9 0 I and hog and then the further composi-
tions h 0 (g 0 f) and (h 0 g) 0 I, giving the pictures below.
ho(gof) (hog)of
Figure 1.8.
Then h 0 (g 0 f) and (h 0 g) 0 I are both mappings from A to D but they have

been constructed differently. Could they be equal?
We consider a specific example before proving the general result.
Example 23
Let I, g, h be the mappings given by
I(x) = x2 + 1 (x E Z),
2
g(n) = '3n (n EN),
h(t) = Jt2""+1 (t E Q),
and so,
But
2
(g 0 f)(X) = g(f(X)) = g(X
2
+ 1) = 3 (x 2 + 1) (x E Z),
[h 0 (g 0 f)](X) = h(~ (X2 + 1)) = G 2

(X + l)r+1,
(h 0 g)(n) = h(g(n)) = h(~n) = ~ (n EN),
[(h 0 g) 0 fl(x) = (h 0 g) (f(x)) = (h 0 g)(x2 + 1) = G (x 2 + 1)) 2+1 (x E Z).

Thus
[h 0 (g 0 f)](x) = [(h 0 g) ol](x) (x E Z)
from which
h 0 (g 0 I) = (h 0 g) 0 f.
Theorem 7 Associativity of Circle-Composition

Let A, B, C, D be sets and let I, g, and h be mappings such that
I : A -+ B, 9 : B -+ C, h :C -+ D.
Then
h 0 (g 0 f) = (h 0 g) 0 f.
Proof
The proof simply entails repeated and careful application of Definition 13. Let
a E A. Then
[h 0 (g 0 f)](a) = h((g 0 IHa))

= h(g(f(a)))
and
[(h 0 g) 0 II(a) = (h 0 g) (f(a))
= h(g(f(a))).
Thus
[h 0 (g 0 f)](a) = [(h 0 g) ol](a)
and since a is arbitrary

h 0 (g 0 f) = (h 0 g) 0 f. o
We may now write, unambiguously, hog 0 f to mean either of the composi-
tions h 0 (g 0 f) or (h 0 g) 0 f.
Remark
Mappings may appear to be necessary but somewhat humdrum objects of study.
However particular mappings may give rise to curious problems. Consider, for
example, the mapping f : N -+ N given by
!n (n even)
f(n} = {
!(3n + I) (n odd).
Thus f(26} = 13, f(25} = !(75 + I) = 38. Suppose we iterate this mapping
several times and, by way of illustration, we follow the effect of the iterations
on the number 10. Then
f(1O} = 5,
(f 0 f)(1O) = f(5} = 8,
(J 0 f 0 f)(1O) = f(8} = 4,
(J 0 f 0 f 0 J)(1O) = f(4} = 2,
(f 0 f 0 f 0 f 0 f)(1O) = 1.
After five iterations on 10 the number 1 appears. The reader may care to verify
that commencing with 65 there are 19 iterations before 1 first appears. Other
integers may be tried at random as test cases, but the reader is advised to culti-
vate patience as the number of iterations before 1 first appears may be quite
large.
The conjecture, that for any n E N and for sufficiently many iterations the
number 1 always appears, remains to be proved or disproved.
Exercises 1.3
1. Let A, B, C be the sets {a, b, c}, {p, q, r}, {O, I} respectively. Let the
mappings I :A -+ B and 9 : B -+C be defined by
I(a) =p,l(b) = I(c) = q,g(p) = O,g(q) = 1,g(r) = 0.
Evaluate (g 0 I)(a), (g 0 I)(b), (g 0 I)(c).
2. Let A, B, C, D be sets and let p, q, r, s be mappings such that

p : A -+ B, q : B -+ C, r : A -+ B, s : D -+ C.
Which of the following are defined: q 0 p, p 0 q, p 0 r, p 0 s, r 0 8, q 0 8?
3. Let I, g, h be mappings of Z into Z given by

I(n) = 2n, g(n) = 3n + 5, h(n) = -6n (n E Z).
Which of the following pairs of mappings are equal: log, go I; 10 h,
h 0 Ij go h, hog?
4. Let A, B and C be sets and let I :A -+ B and 9 : B -+ C be mappings.

(i) If I and 9 are injective prove that go I is injective.
(ii) If 9 0 I is injective, are 9 and I injective?
(iii) If 9 0 I is surjective, are 9 and I surjective?
5. Let X, Y be sets and let I :X -+ Y be a mapping. Let A, B be subsets

ofX.
(i) If A ~ B prove that I(A) ~ I(B).
(ii) Prove that I(A U B) = I(A) U I(B).
(iii) Prove that I(A n B) ~ I(A) n I(B).
6. Let A be the subset of Z consisting of the even integers and let B be the
subset of Z consisting of the odd integers. Let a mapping I : Z -+ Z be
defined by
I(n) = n2 - 3n + 5 (n E Z).
Prove that 3 E I(A) n I(B) and so deduce that I(A n B) ::I

I(A) n I(B).
1.4 Equivalence Relations

In the context of humans, animals and plants, the notion of relationship is all
pervasive; two individuals are regarded as 'related' if they share a common
parent, grandparent or even a more distant ancestor. Thus cousins are said to
be related but we would not normally regard a cat and a dog as being related.
In these illustrations we encounter the idea of a universal set, whether composed
of humans or of domestic animals, and of the relationship which members or ele-
ments of the set may bear to one another under some particular criterion. Since
the underlying concept of relationship is very useful in mathematics we discuss a
more mathematical example before proceeding to a precise definition.
Example 24
We consider a set A and a relationship which exists between some pairs of
elements of A. Let X be a set and let f: A -+ X be a mapping. We shall say
that elements a and b of A are 'related' if f(a) = f(b). By this definition any
two elements of A mayor may not be related.
For convenience we write for a, b E A, a '" b if and only if a, b are related; the
presence of the symbol'" (pronounced 'twiddles') denoting that a and b are
related.
We note three obvious facts:
1. For all a E A, a is related to a since f(a) = f(a) implies a,...., a (for all a E A).
2. H a, b E A are such that a is related to b, then b is related to a, since a ,...., b
implies f(a) = f(b) and so f(b) = f(a) and b '" a.
3. If a, b, c E A are such that a is related to b and b is related to c, then a is related
to c since a '" b implies f(a) = f(b) and b '" c implies f(b) = f(c) from which
f(a) = f(b) = f(c) and so a ,...., c.
The definition we seek basically repeats the conclusions of this example.
Definition 14
Let A be a set. A relation, denoted by"', is defined between some pairs of ele-
ments of A subject to the following (named) conditions.
1. For all a E A, a,...., a (reflexivity).
2. If a, b E A are such that a ,...., b then also b", a (symmetry).
3. If a, b, c E A are such that a'" b and b '" c then also a ,...., c (transitivity).
A relation'" on A which is reflexive, symmetric and transitive as above, is called

an equivalence relation.
Symbols R, p, (1 are frequently used to denote equivalence relations.
We note in passing that the relation of the previous example is an equivalence

relation.
Examples 25
1. On any set A the relation of equality is an equivalence relation since we have
a = a for all a E A, we certainly have a = b implies b = a(a, bE A) and if
a = band b = c(a,b,c E A) then a = c.
An equivalence relation between the elements of a set generalizes the notion
of equality for elements of the set.
2. On Z a relation'" is defined by a '" b if and only if a, b are both even or a, b are
both odd.
Thus 4 '" 4 and 5 '" 5 but we do not have 4 '" 5.
Now, in general, a '" a and certainly a '" b implies b '" a (a, b E Z). Suppose
now a '" b and b '" c (a, b, c E Z). Then, if b is even, a and c must also be even
and a '" c, whereas if b is odd, a and c must also be odd and a '" c. Thus '" is
an equivalence relation on Z.
Notice that we could have defined this equivalence relation by means of a
mapping f: Z -+ {O, I} for which
= {O (n even)
f (n) 1 (n odd) (n E Z).
3. Let S be the set of all words in a given English dictionary. Define two words to
be related by p if they begin with the same letter and end with the same letter.
Thus we have 'atom p alum', 'pleasure p perseverance' and 'a p aroma'. Then
p is easily seen to be an equivalence relation on S.
4. Let S be the set of points composing the circumference of a given circle. For P,
Q E S define a relation (1 by P (1 Q if and only if there is a diameter on which
P, Q are points. Thus if pqQ then either P, Q coincide or P, Q are at oppo-
site epds of a diameter. Simple geometrical considerations show that q is an
equivalence relation.
5. Let X be the set Q x Q of ordered pairs of rational numbers.
For a = (all <l2) E Q x Q and b = (b ll b2 ) E Q x Q we define a relation'" by
a'" b if and only if al + <l2 = b1 + b2 •
We claim that", is an equivalence relation on X. Certainly'" is reflexive.
If for some a,b E X we have a'" b then a = (alla2), b = (bt,~), where

ailbi E Q (i = 1,2) and at +~ = bt +~. But then bt +~ = at +~ and
so b '" a and '" is symmetric.
Suppose now for some a, b, c E X we have a '" b and b '" c. Then writing
a = (all~)' b = (bll~)' C= (CllC2) we have at + ~ = bt + ~ = Ct + C2
and so a '" C and '" is transitive. Hence'" is an equivalence relation.
6. A relation, p, is defined on Q by letting, for a, b E Q, apb if and only if we have
a - b E Z. By this relation we have
43543538
- p - as - - - = - = 2 E Z
19 19 19 19 19
but
11. 115
- p- 18 false as - - - =- ¢ Z
2 3 2 3 6 .
We may prove that p is an equivalence relation as follows:

(i) For all a E Q, a - a = 0 E Z and so a '" a (reflexivity).
(ii) If for some a, b E Q, a '" b then a - b E Z which immediately implies
that b - a = -(a - b) E Z and so b '" a (symmetry).
(iii) If for some a, b, CEQ, a'" b and b'" C then evidently a - b E Z and
b - C E Z from which a - C = (a - b) + (b - c) E Z and so a '" c (transi-
tivity).
It is sometimes important, in regard to a set of conditions specifying a mathema-

tical entity such as an equivalence relation, to know whether the stipulated con-
ditions are themselves independent or not. We show that the three conditions of
reflexivity, symmetry and transitivity for an equivalence relation are indepen-
dent by exhibiting three examples in each of which two of the conditions are
satisfied but not the remaining condition.
Examples 26
1. Let a relation'" be defined on Z by a '" b if and only if ab '" 0 (a, b E Z). Then
the relation is symmetric since a '" b implies ab '" 0 and so ba '" O. Thus b '" a
(a,b E Z).
The relation is also transitive since a'" b, b'" c implies ab", 0, be '" O.
Hence a '" 0, b '" 0, c '" 0 and thus ac '" 0 giving a '" c.
On the other hand the relation is not reflexive since we do not have a '" a
for all a E Z, in particular we do not have 0 '" O.
2. The relation of inequality, ~, on JR is reflexive and transitive since we have, for

all a E JR, a ~ a and if a ~ b, b ~ c then a ~ c (a,b,c E JR). We cannot infer
from a ~ b (a, bE JR) that b ~ a and so this relation is not symmetric.
3. Let a relation rv be defined on Z by a '" b if and only if 2 divides a - b or
3 divides a-b. Such a relation is not transitive since, for example,
7",5since7-5=2
and 5 '" 2 since 5 - 2 = 3
but 7 '" 2 is false since 7 - 2 = 5 which is divisible by neither 2 nor 3.
But the relation is reflexive since for all a E Z, a - a = 0 and so a'" a and the
relation is symmetric since a '" b (a, b E Z) implies that a - b, and thus b - a,
is divisible by 2 or by 3 and so b '" a.
We may consider the totality of the inhabitants of a given town as being dis-
tributed into the families which dwell in the town. Any two members of the same
family are somehow related but no two members of different families are related.
An inhabitant of the town belongs to only one family and so identifies uniquely
the family to which he or she belongs. In a similar way a set with an equivalence
relation will be seen to be the union of disjoint subsets called equivalence classes,
each equivalence class consisting (like a family) of all elements which are related
to one another and each equivalence class will be uniquely identifiable by any
element belonging to it.
We make these ideas more precise.
Definition 15
Let S be a non-empty set with an equivalence relation "'. Let a E S. The subset
Sa of S given by
Sa = {x E S: x rv a}
is said to be the equivalence class determined by, or containing, a (note that as
a'" a, a E Sa)'
We now prove a result which we shall have frequent occasion to use.

Theorem 8
In the notation of Definition 15 the following hold:
1. S= USa'
aeS
=
2. Sa S" if and only if a"" b (a, bE S).
3. Either Sa = S" or Sa n S" = 0 (a,b E S).
Proof
1. Since a E Sa it follows that S ~ U Sa and SO S = USa'
aeS aeS
2. If Sa = S" then a E S" and so a "" b. Conversely suppose a "" b. We prove that
Sa ~ S". Let x E Sa' Then x "" a and a '" b implies that x "" b and so xES".
Thus Sa ~ S". Since, by symmetry, b "" a we also have S" ~ Sa and so Sa = S".
3. Suppose Sa n S" :f: 0. We show that Sa = S". Let c E Sa n S" and SO c "" a
and c "" b. But then a"" c and c"" b which implies, by transitivity, that
a "" b. By (ii) we conclude that Sa = S". 0
It is a consequence of Theorem 8 that from the union U Sa, on eliminating

aeS
equivalence classes that formally coincide, we may write
where T is a subset of S such that Sa (a E T) are distinct, and Sa n S" = 0

(a, bET, a:f: b). The non-empty set S is thereby expressed as a disjoint union
of distinct equivalence classes.
Examples 27 (Continued from Examples 25)

We examine the examples immediately following Definition 14 to determine the
appropriate equivalence classes.
1. With equality as the equivalence relation on a set each equivalence class con-
sists of a single element.
2. There are two equivalence classes, namely the subset of Z consisting of the
even integers and the subset of Z consisting of the odd integers.
3. The English alphabet has 26 letters: a, b, c, ... , x, y, z. Each word begins with,
and ends with, one of these letters. We now define 262 = 676 subsets each
defined for any pair of letters of the alphabet as follows:
Let S04 be the subset of all words beginning with a and ending with a.
Let Sob be the subset of all words beginning with a and ending with b.
Let Szz be the subset of all words beginning with z and ending with z.
The reader will immediately observe that some of the subsets S04' Sob, ... ,Szz
are empty (Szz = 0 but, perhaps surprisingly, Stu::F 0). Anyone of these
subsets, if non-empty, is an equivalence class. S is the union of these subsets
and, on omitting empty subsets, becomes a disjoint union of equivalence
classes. (We leave it to the aspiring lexicographer to determine which subsets
are non-empty.)
4. Every diameter determines an equivalence class consisting of the two end-
points of the diameter.
5. For every q E Q there is an equivalence class given by
Xq = {(alla2) E Q x Qlal +~ = q}.
Each equivalence class is of this form.
We have seen that if a set admits an equivalence relation then that set is a dis-
joint union of particular subsets called equivalence classes. Conversely if a set is a
disjoint union of subsets then we shall show that, correspondingly, an equiva-
lence relation may be defined on the set. We are led to make the following
definition.
Definition 16
Let S be a non-empty set. Let {S>. : A E A} be a collection, indexed by an index-
set A, of non-empty subsets of S such that
1. s>.nSJJ =0(A,J.l.EA,A::FJ.I.) and
2. U S>.=S.
>.eA
The collection of subsets is said to form a partition of S.
Suppose we have a non-empty set S with a partition as above. For a, b E S

define a relation '" by letting a '" b if and only if a, b belong to the same
subset S>. (say) of the partition. Thus any two elements in an S>. are related
but no element in S>. is related to an element in SJJ (A::F J.I.). Then it may readily
be verified that is an equivalence relation on S and that the equivalence classes
/"OJ
are the subsets S>. (,\ E A). A partition therefore induces an equivalence relation
and an equivalence relation induces a partition, subsets of the partition forming
the equivalence classes of the equivalence relation.
Example 28
Let 8 = {a,b,c,d,e,l,g}. Let 8 1 = {a,b,d}, 8 2 = {e,g}, 8 a = {e,f}. Then the
subsets 8 1 ,82 , 8 a form a partition {81l 8 2 , 8 a} of 8. We define the corresponding
equivalence relation'" on 8 as follows:
a '" a, a '" b, a '" d, e'" e, e'" e,

b '" a, b '" b, b '" d, c '" g, eN I,
d '" a, d '" b, d '" d, g'" c, I'" e,
g'" g, I'" f.
Exercises 1.4
1. A relation p is defined on JR by apb if and only if a2 = b2 (a, bE JR).
Prove that p is an equivalence relation on JR. Identify the equivalence
classes.
2. A relation R is defined on Z by aRb if and only if a - b is even

(a, b E Z). Prove that R is an equivalence relation on Z. Identify the
equivalence classes.
3. A relation r is defined on Z by arb if and only if a - b is divisible by 6

(a, b E Z). Prove that r is an equivalence relation and determine the
equivalence classes.
4. What are the equivalence classes in Examples 25, no. 6 immediately

following Definition 14?
5. Let Oxy be the two-dimensional coordinate plane. Let P, Q be points

ofOxy.
(i) P, Q are said to be related by u if the points 0, P, Q are collinear.
Is u an equivalence relation?
(ii) P, Q are said to be related by r if there exists a rotation about 0
which sends Pinto Q. Is r an equivalence relation?
Sketch, if possible, the equivalence classes.
6. How many distinct equivalence relations may be defined on a set of

one, two, three or four elements?
7. A relation p is defined on Z by apb if and only if a - b is divisible either

by 5 or by 7 (a, bE Z). Is p an equivalence relation?
8. Let S = {a, b, c, x, y, z}. Two relations p, q are defined on S as follows:

apa, bpb, cpc, xpx, ypy, zpz, apb, bpa, apc, cpa, bpc, cpb; aua,
bub, cuc, xux, yuy, zu z, aux, xua, buy, yub, cu z, zuc, auy, yua.
Do either p or u define an equivalence relation on S?
9. Let S be a set on which a relation'" is defined. The relation'" is
symmetric and transitive. Loose thinking would suggest that a '" b
(a, b E S) implies b '" a (by symmetry) and so a '" a (by transitivity)
thus giving reflexivity, and consequently '" would be an equivalence
relation. What is wrong with this loose thinking?
1.5 Well-ordering and Induction

The elements of the set N, that is the set of strictly positive integers, may be
written down in ascending order with a repeated inequality sign, thus:
1<2<3<4< ...
Let S be a non-empty subset of N. Let N EN. From the integers 1, 2, ... , N we
may select that one which is least amongst these integers and which also belongs
to S. Intuitively therefore we have the following principle which we shall accept
as axiomatic.
Principle of Well-ordering in N
Every non-empty subset S of N has a least integer in S.
It frequently happens that we have some assertion, proposition or statement

P(n) which depends on the particular integer n. The proposition may itself be
true or false. We give three examples.
Examples 29 .
n
1. 1 + 2 + ... + n = 2" (n + 1) (n E N).
2. 2n + 1 ~ 2n (n EN).
2 n
3. n ~ 2 (n EN).
In each of these examples we have a statement that depends on n. We are not

asserting the truth or falsity of the statement. Naturally, however, we wish to
know whether the particular statement is true for all n E N or, possibly, for all
n E N greater than some fixed integer. We are led to the so-called Principle of
Induction and to an obvious and convenient variant of this principle. We may
derive the Principle of Induction from the Principle of Well-ordering but both,
for present purposes, may be regarded 88 simply axiomatic or, indeed, 88
'obvious'.
Principle of Induction
Let P(n) be a proposition depending on the integer n. Suppose that
1. P(1) is true and
2. if P(k) is true then P(k + 1) is true (induction assumption).
Then P(n) is true for all n E N.
Proof (may be omitted)

Let S be the subset of N of those integers n for which P( n) is true. Then certainly
1 E S and so S =F 0. Let X = N\ S. We wish to show that X = 0 and then
S = N. For the sake of argument suppose X =F 0. We apply the Principle of
Well-ordering to X. Let N be the least integer in X. Now N =F 1 since 1 E S
and so N > 1. Since N is the least integer in X, N - 1 is an integer not in X
and so N - 1 E S. But then P(N -1) is true and so, by hypothesis 2, P(N) is
also true and N E S. But this is a contradiction 88 X n S = 0. Hence X = 0
andS=N. 0
The Principle of Induction is often used in a modified version which we may

also deem to be axiomatic.
Principle of Induction (Modified Version)

Let P( n) be a proposition depending on the integer n. Suppose that
1. P(1) is true and
2. if for each m ~ k, P(m) is true, then P(k + i) is true (induction assumption).
Then P(n) is true for all n E N.
It sometimes happens that in either version of the Principle of Induction state-

ment 1 that 'P(1) is true' is replaced by 'For some N E N, P(N) is true' with
corresponding changes to statement 2. In this circumstance the conclusion is
that P(n) is true for all n E N, n ~ N.
Let us consider further examples:
Examples 30
1. P(n) is the statement
1+ 2 + ... + n = 2'n (n + 1) (n EN).
Certainly P( 1) is true since

1
1 ="2 (1 + 1).
If we now make the induction assumption that P(k) is true, then we suppose
that
k
1 + 2 + ... + k = "2 (k + 1).
But this implies that
1 + 2 + ... + k + (k + 1) ="2k (k + 1) + (k + 1)
= (k+ 1) (k+ 2)
2
= (k+ 1) [(k+ 1) + 1]
2
and so we may assert that P(k) implies P(k + 1). Hence we have for all n E N
n
1 +2+ ... +n = 2' (n+ 1).
2. P(n) is the statement that
2n + 1 :5 2n (n EN).
Now P(l) and P(2) are, in fact, false since
2(1) + 1> 21 and 2(2) +1> 22 •
However, P(3) is true since
2(3) +1= 7 :5 23 •
Let us suppose that P(k) is true for all k ~ 3. Then we suppose
2k + 1 :5 2k (if k ~ 3).
2(k + 1) + 1 = 2k + 3 = 2k + 1 + 2:5 2 = 2k + 1
k
+ 2:5 2k + 2k (k ~ 3)
and so P(k + 1) is true. Hence we conclude that for all n E N, n ~ 3
2n+ 1:5 2n •
3. P(n) is the statement that

n 2 ~ 2n (n EN).
P(I) and P(2) are true since
12 = 1 ~ 2 1 , 22 = 4 :5 22
but P(3) is false since
However, P(4) is true since

42 :524 •
We make the induction assumption that P(k) is true for all k ~ 4. But then
2 k
k :5 2 (if k ~ 4)
implies that
(k + 1)2 = k2 + 2k + 1 ~ 2k + 2k + 1 :5 2k + 2k = 2k + 1 (by 2 above)
and so P(k + 1) is true. Hence we conclude that for all n E N, n ~ 4,
2
n ~ 2n •
When we attempt to apply induction as above we must be on guard in regard

to two aspects. We must ensure that we do not go blindly on from the truth of
P(I) (say) and try to prove P(k + 1) from P(k) when in fact P(2) or P(3) or
whatever may be false (as in the Example above). The second aspect which
may cause trouble is an invalid use of a particular induction assumption. We
illustrate this aspect in the next example.
Example 31
By induction we shall 'prove' that, given any set of billiard balls on a billiard
table, the balls are necessarily of the same colour.
Our statement P(n) is that if n balls or fewer are on a billiard table they are of
the same colour. Certainly if there is only one ball there is only one colour and
P(I) is true. We now try to prove, from the assumption that if we have k balls
or fewer then they are of the same colour, that we may deduce that if we have
k + 1 balls or fewer then they are of the same colour.
Given a set S of k + 1 balls we may select two distinct subsets A, B of S of k
balls each. Then S = AU B and An B :f: 0. Now by the induction hypothesis
the balls in A are of one colour and the balls in B are of one colour. But A and
B have at least one ball in common and so the balls in A and B must be of a
single colour. Hence, as AU B = S, the balls in S are all of the same colour
and we have completed the induction step of the argument, namely P(k) implies
P(k + 1). Hence P( n) is true for all n E N and the balls on a billiard table are all
of the same colour.
The purported conclusion is clearly false - so what is wrong in our arguments?
We were careless in supposing that P(2) is a consequence of P(I). A moment's
reflection should reveal that P(2) is false.
This example should serve as a warning that, in applying induction, care must
be taken to ensure that the conditions are properly met.
We give some further examples in the use of induction. In these examples we
shall omit specific reference to a P(n) but we use the convention that
n
L llr = al + <l2 + ... + an·
r=l
Examples 32
1. Prove that
~ r2 __ n(n + 1)6(2n + 1)
L.J (n EN).
r=l
Certainly for n = 1 we have

12 = 1 = 1(1 + 1)(2 + 1)
6
and so the statement holds for n = 1. We now suppose that
~ 2 -12 22
L.Jr - + + ... + k2 -_ k(k+ 1)(2k+
6
1)
.
r=l
Then
L r 2 = 12 + 22 + ... + (k + 1)2 = (1 2 + 22 + ... + k 2) + (k + 1)2

k+l
r=l
= k(k+ 1)(2k+ 1) + (k+ 1)2

6
= (k; 1) [k(2k+ 1) + 6(k+ 1))
= (k; 1) [2~ + 7k+ 6)
= (k+ 1) (k+2)(2k+3)
6
= (k; 1) [(k+ 1) + 1)[2(k+ 1) + 1).

This completes the induction argument and so the statement is true for all
nEN.
2. Prove that
1 1
nil
= 12 + 22 + ... + n2
~ r2 ~2 -n1 (n EN).
Certainly for n = 1 we have

1 1
-=1=2--.
1 1
Suppose
Then
IT we now focus our attention on the desired outcome of the induction argu-
ment we realize that we would like to show that
1 1 1
2- k + (k + 1)2 ~ 2 - (k + 1)'
But this inequality is true since
Hence
111 1
-+-+ + <2---
12 22 ... (k+l)2- k+l
and the statement is true for all n E N.
Exercises 1.5
Establish the following eleven statements by induction:
n 1
1. Lr(r+l)=3n(n+l)(n+2) (nEN).
r=l
n 1
2. L r 3 = 4 n 2 (n + 1)2 (n EN).
3. t
r=l
a
r
-
1
= 1 + a + ... + a n - 1 = { -=- ~ ~
n (a= 1)
(a E R,a =F 1)
(nEN).
1 n
L r (r + 1) = -n +1
n
4. (n EN).
r-l
n n
5. L (5r - 2) = 2 (5n + 1) (n EN).
r=l
L
n
6. r2r = 2 + (n - 1)2n +1 (n EN).
r=l
7. (l+x)n~l+nx (xEIR,x~O).
n
8. 3(2n+ 1) $ 2 (n E N,n ~ 6).
9. 3n2 $ 2n (n E N,n ~ 8).
n 1
10. 2(v'n+l-l)$~ v'r (nEN).
11. The Fibonacci sequence Ul, U2, ... is defined inductively by

Ul = 1,U2 = 2,Un+l = Un +Un-l (n ~ 2).
Write down the first nine terms of the sequence and prove indepen-
dently the two following results:
(i) Un $ G) n (n EN).
.. -_.2.-
(11) ~ v'5
(1 + v'5)n+l_-!..-
2 v'5
(1 - v'5)n+l 2 (n EN).
The sequence was originally devised by Leonardo of Pisa, also called

Fibonacci (c. 1170-1240), to model the procreative habits of amorous
rabbits. Leonardo, through his text Liber abbaci, was influential in
spreading the use of Hindu-Arabic numerals in Europe. We do well
to remember the inestimable benefits these numerals have conferred
and to acknowledge our intellectual debt to the civilizations from
which they came. In the next chapter we shall exploit some of the
advantages of the numerals.
1.6 Countable Sets

The sets {a,b,c,d}, {a,,B,"Y,c5}, {1,2,3,4} have no elements in common yet they
share the property of being finite sets with the same number of elements. As we
have seen, we may write the elements of a set X which has a finite number N of
elements as
X = {XltX2,'" ,XN},
in other words there is a bijection from {I, 2, ... , N} to the set X. H a set X does
not have a finite number of elements then X is infinite and there mayor may not
be a bijection from N to X. It is convenient to distinguish between 'levels' of
infinity - as a porcine Napoleon might have decreed "All infinite sets are non-
finite but some infinite sets are more infinite than others." (after Animal Farm
by George Orwell, 1945). More precisely, we make the following definition to dis-
tinguish between those sets that admit a bijection from N, or from a subset ofN,
and those that do not.
Definition 17
A non-empty set X is said to be countable if the elements of X may be enumer-
ated as a finite or infinite sequence ofthe form Xl' X2' X3, ••• H X is finite and has
N elements then X = {xnln = 1,2, ... , N} = {Xl,X2"" ,XN} and if X is infinite
then X = {xnln E N} = {XltX2," .}. Otherwise X is said to be uncountable.
Examples 33
1. The set -N of strictly negative integers given by -N = {-nln EN} is
countable since we have the sequence -1, -2, -3, ... The bijection f from
N to -N is given by f(n) = -n (n EN).
2. The set Z of all integers is countable since we have the sequence
0,1, -1, 2, -2,3, -3, ... The bijection f from N to Z is given by
(n even)
(n odd).
Since the elements of a countable set may be enumerated as a finite or infinite

sequence it follows that the elements of any non-empty subset may also be
enumerated as a sequence, namely 88 the subsequence of the given sequence
obtained by omitting those elements not in the subset. Consequently we have
the following result which we state for completeness.
Theorem 9
Any subset of a countable set is countable.
We shall prove that the set Q of rational numbers is countable but first we
must investigate the countability of an array.
Remark
Let X be a set of elements Xij having double suffices such that
X = {xpqlp,q EN}.
We may write the elements down in the form of an infinite array similar to a
matrix array:
We begin the first stage of the enumeration of these elements by choosing Xu.
For the second stage we choose those elements of which the sums of their suffices
are ~ and we take them in the order of first suffix. At the third stage we choose
those elements of which the sums of their suffices are 4 and we again take them in
order of the first suffix. Continuing we obtain a sequence as follows:
Each element of the array appears precisely once in this sequence and so X is
countable.
Before drawing general conclusions let us determine explicitly the bijection

f from N to X. Certainly we have f(Xll) = 1, f(X12) = 2, f(X2d = 3, ... We
require to determine the integer to which x pq corresponds. Now x pq is the pth
element in the stage of choosing in which we commence with al,p+q-l'
The stages prior to this stage have involved successively 1,2,3, ..., p + q - 2
elements. Thus x pq is the element in the [[1 + 2 + ... + (p + q - 2)] + p]th
place in the sequence of enumeration. But, from an earlier example, we have
f(x pq ) = [1 + 2 + ... + (p + q - 2)J + p

1
="2 (p + q - 2)(P + q -1) + p
= "21 [(p + q) 2 - p - 3q + 2J.
This gives, for example,
f(X34) ="21 [(3 + 4)2 - 3 - 12 + 2] = 18

and we may verify directly that x34 is in the 18th place. We may now derive
several useful conclusions.
Theorem 10
The union of a countable number of sets, each of which is itself countable, is
countable.
Proof
Let Xl' X 2 , ••• be the countable sets of the union.
Let Xl have elements XU,Xl2, .
Let X 2 have elements X21lX22, .
Let X n have elements Xnl, X n2, •••
Thus the elements of UX n may be written in an array identical to that in the

Remark above. We bow that the array above is countable and so UX n is
n
countable. 0
Example 34
Since N, -N 8J?d {O} are countable, so also is
z= -NU{O}UN.
Theorem 11
The Cartesian product of a finite number of countable sets is countable.
Proof
We prove the result first for the Cartesian product of two countable sets A, B.
Let A have elements al> l12, ... and let B have elements bl>~"" Then the
elements of Ax B may be written in the array
(al,bl),(al,~),(al,b3)" ..
(l12,bl),(l12,~),(l12,b3)""
The array is countable and so is A x B.

IT now we have countable sets Xl> X 2 , ••• ,XN, then we have
Xl x X 2 x ... X X N = (Xl X X 2 x ... x XN-d X XN
and a simple application of the Principle of Induction gives the final result. 0
Theorem 12
The set of rational numbers Q is countable.
Proof
Every strictly positive rational number is of the form 1f! for some m, n E N and so
may be identified with the ordered pair (m, n) E N x N. Thus i may be identified
with (2,3) and! may be identified with (4,6); of course, purely 88 numbers, i =!
but formally i and ~ are distinct 88 the ordered pairs (2,3) and (4,6) are certainly
distinct. Thus the set of strictly positive rational numbers Q+ is identified with
N x N and so, by Theorem 11, Q+ is countable.
But letting -Q+ = {-qlq E Q+} we see that -Q+ is countable and so finally
Q = -Q+ U {O} U Q+ is countable. 0
Not all commonly occurring sets are countable. We state, but do not prove,
the following result.
Theorem 13
The set of real numbers JR is uncountable.
Nevertheless it would be useful to exhibit at least one example of uncount-

ability 88 follows.
Example 35
Let A be the set of all mappings with domain N and range {O, I}. We claim that
A is uncountable. We shall argue by contradiction in supposing that we have
included all such mappings in a sequence of enumeration and then, like a con-
juror producing a rabbit out of a hat, we shall produce a mapping not previously
in the sequence. This then will be our contradiction.
Suppose, therefore, we have enumerated all of the given mappings from N to
{O, I} in the sequence 11,12, ... Then every mapping from N to {O,l} occurs 88
an Ir for some r E N. We define I: N -+ {O,l} as follows:
I(n) = {O ~ In(n) = 1 (n EN).

1 if In(n) = 0
Then for all n E N, I(n) =F In(n) and so I =F In (n EN). But then I cannot
have occurred in the sequence 11,/2' ... Hence the supposition of countability
is invalid and so A is uncountable.
It is not the purpose of this text to give a logically impeccable account of the
theory of sets, fundamental as this may be for mathematics. The reader should
be aware, however, that matters in the affairs of sets, 88 in the affairs of hearts,
may not be 88 straightforward as they appear. As illustration we offer some
examples upon which the reader may exercise his or her analytical skills.
Examples of paradoxes
1. Epimenides of Crete is supposed to have said "Cretans are always liars" with
the meaning that Cretans tell only lies. Was Epimenides lying or not when
he made his statement? (Epimenides, 6th century BC, is believed to be the
'prophet' of St Paul's Epistle to Timothy, Chapter I, verse 12).
2. In a certain village there is a barber who shaves all and only those persons in
the village who do not shave themselves. Does the barber shave himself?
(attributed to B. Russell, 1872-1970).
3. In a certain country every prefecture must have a prefect and no two may
have the same prefect. A prefect may be non-resident in the prefecture.
The government passes a law creating a special area for non-resident prefects
and obliging all non-resident prefects to live there. Eventually there are so
many non-resident prefects in this area that the government declares it to
be a prefecture. Where does the prefect of this new prefecture reside?
Exercises 1.6
1. Prove that N x N, N x N X N, Z X Z, Z x Z x Z are countable sets.
2. Prove that {a + bv'2la, b E Z} is countable.
3. Let A, B be sets. Prove that A is countable if and only if A n B, A \ B
are countable sets.
4. Let NN denote the set of all mappings of N into N. Prove that NN is
uncountable.
5. Let the elements of the set X be denoted by x pq where p is an element of
Nand q is an element of {I, 2, ... ,N} for some given N. Prove that X
is countable.
2
The Integers
Numbers are encountered early in life and in many practical contexts. In our
infancy we develop a feeling for the natural numbers by chanting "one, two,
three, ...", "eins, zwei, drei, ...", "yi, er, san, ...", or their equivalent, in what-
ever may be our mother tongue. By childhood we have assimilated, without too
much conscious effort, the elementary properties of the addition, subtraction,
multiplication and division of the natural numbers; in this text we take these
elementary properties for granted. In such an approach we are following the
celebrated dictum of L. Kronecker (1823-91), namely, "Die ganze Zahl schuf
der liebe Gott; alles tibrige ist Menschenwerk" which we may render in English
as "God created the integers; everything else is man's handiwork". (Quotation
from Philosophie der Mathematik und Naturwissenschaft by H. Weyl, R. Olden-
burg, Mtinchen, 1928, Section 6, page 27).
2.1 Divisibility
Before considering concepts of division and divisibility we recall some earlier
notation. We introduced the set N of natural numbers, N = {I, 2, 3, ...}, the
set Z of integers, Z = { ... , -2, -1,0, 1,2, ...} and the set Q of rational numbers
Q = {': 1m, nEZ, n :F o}. (In passing we should remark that in some texts the
symbol N denotes the set {O, 1,2, ...}. The reader should confirm the significance
of N in any text he or she happens to use.) We shall exploit properties of
these sets.
47
Definition 1
Let a and b be integers with b being non-zero. Then we say that b divides a, or
that b is a divisor of a, if there exists an integer c such that a = be.
Notice that 0 is divisible by any integer b, b 1: 0, since 0 = bO and that any

integer a 1: 0 has the so-called trivial divisors ±1, ±a.
Examples 1
1. We have that 3 divides 6 since 6 = 3.2 (where. denotes multiplication) but 4
does not divide 6 since there is no integer x such that 6 = 4x. We have 6 = 4~
i
but ¢ Z.
2. We have that 36 divides 180 = 36.5 but 24 does not divide 180.
3. 7 has the trivial divisors ±1, ±7. There are no other divisors.
4. 10 has the trivial divisors ±1, ±1O. The remaining divisors are ±2, ±5.
Lemma 1
Let a, b and c be integers with b and c being non-zero. Let b divide a and c divide
b. Then c divides a.
Proof
By assumption there exist integers d and e, such that a = bd, b = ceo Then
a = bd = (ce)d = c(ed).
Hence we conclude that c divides a. o
Examples 2
1. 25 divides 175 and 175 divides 700 and so, as in Lemma 1, 25 divides 700.
(In fact 175 = 25.7,700 = 175.4 and 700 = 25.28.)
2. The divisors of 18 (and also of -18) are ±1, ±2, ±3, ±6, ±9, ±18.
3. The divisors of 45 (and also of -45) are ±1, ±3, ±5, ±9, ±15, ±45.
As the repetition of ± signs becomes tedious we shall prove the following fairly
obvious lemma.
The Integers 49
Lemma 2
Let a and b be non-zero integers. Then the following are equivalent: (i) b divides
a, (ii) b divides -a, (iii) -b divides a, (iv) -b divides -a.
Proof
IT b divides a then a = be for some c E Z. Then (-a) = b( -c), a = (-b)( -c),
-a = (-b)c from which, respectively, b divides -a, -b divides a, -b divides
-a. Thus (i) implies (ii), (iii) and (iv). The remaining implications are similarly
established. 0
Lemma 2 ensures that, when we are seeking to test for the divisibility of one
integer by another, we may suppose that both integers are strictly positive. In
other words, to investigate divisibility in Z we may as well confine our attention
to divisibility in N since we may attach ± as appropriate. We therefore wish to
distinguish those integers in N which are in some sense (to be made precise later)
'irreducible' for division.
Definition 2
A strictly positive integer p, p oF 1, is said to be a prime number or, simply, a
prime if p has only trivial divisors. Thus if nEZ, n oF 0 and n divides p then
n = ±1 or n = ±p. A strictly positive integer p, p oF 1, which is not a prime is
said to be composite.
The primes in ascending order are 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, ...
Examples 3
1. Suppose we wish to find the next prime after 37. We run through the integers
in ascending order. 38, 39, 40 are not primes since 38 = 2.19, 39 = 3.13,
40 = 23 .5. However 41 is not divisible by any n E N except 1 and 41. Thus
41 is the next prime.
2. The number 178350 is composite and when expressed as a product of primes,
or factorized as we say, is given by
178350 = 2.3.52 .29.41.
To obtain this factorization we begin by observing that 10 divides 178350 and
so 178350 = 17835.10. Clearly 5 divides 17835 and 17835 = 3567.5. We have
disposed of obvious divisors such as 10 and 5. We now seek divisors of 3567.
Considering the primes in turn we find that 3567 = 1189.3 and that the next
prime dividing 1189 is 29 and 1189 = 41.29. Thus finally
178350 = 17835.10 = 3567.5.10 = 1189.3.5.10 = 41.29.3.5.10
= 41.29.3.5.2.5 = 2.3.52 .29.41.
We are familiar, 88 above, with the factorization of a given integer into the
product of primes. In order to factorize a given integer we must determine the
primes dividing the integer. Factorization involves division by arbitrary primes,
whereas the definition of primeness (often called 'primality') of an integer relates
to the divisibility of the integer. Once we have established certain properties of
divisors we shall return to the primes.
Exercises 2.1
1. Write down the divisors of (i) 72, (ii) 84, (iii) 667.
2. Complete the proof of Lemma 2.
3. Factorize (i) 24, (ii) 93, (iii) 1225, (iv) 4913, (v) 14256.
2.2 Divisors
We are here mainly concerned with two given integers and with finding integers
which divide them simultaneously. For this purpose we shall describe algorithms
leading to the determination of such 'common divisors'. (The word 'algorithm'
comes from the Latin 'algorismus' which is derived from the name of
al-Khwarizmi, a 9th century Arab mathematician in Baghdad.) In mathematics,
by the word 'algorithm', we understand a procedure which after a finite number
of repetitions will yield some desired result.
Definition 3
Let a, b and c be non-zero integers. Then c is said to be a common divisor of a and
b if c divides a and c divides b.
We may here usefully remark that if c divides a and b then c divides a + band
a-b. This fact will be used frequently without further comment.
The Integers 51
Example 4
The common divisors of -24 and 54 are ±1, ±2, ±3, ±6. We note that every
common divisor divides ±6.
Definition 4
Let a and b be integers. Let a common divisor d have the property that, for any
common divisor c of a and b, c divides d. Then d is called a greatest common
divisor (G.C.D.).
If d' is a second greatest common divisor then certainly d' divides d and d
divides d' and consequently d' = ± d. That greatest common divisor which is
positive is called the greatest common divisor and is denoted by (a, b). (The
present meaning of this notation has to be distinguished from its usage as an
ordered pair and from a common usage in analysis wherein (a, b) denotes the
open interval {x E R Ia < x < b}. The context should make the meaning clear.)
Example 5
(4,7) = (4, -7) = (-4,7) = (-4,-7) = 1,
(6,4) = (6, -4) = (-6,4) = (-6, -4) = 2,
(36,120) = 12, (74,111) = 37.
The greatest common divisor of two integers may be found by factorizing the
integers into products of primes and then by comparing the factorizations. The
drawback of this method is that it may be difficult to arrive at the appropriate
factorizations; nevertheless the method may be used to find the greatest common
divisor of small integers. We remark here that the method depends implicitly on
the uniqueness of factorization of the integers, we shall consider this aspect more
fully in the section on primes.
Examples 6
1. By the trial and error method as above we find that
616 = 23 .7.11, 1260 = 22 .3 2 .5.7.
Then
(616,1260) = 22 .7 = 28.
We observe that
616 = 28.22, 1260 = 28.45,
(22,45) = 1.
2. Similarly we have
and so
(1512,7056) = 23 .32 .7 = 504.
We observe that
1512 = 504.3, 7056 = 504.14, (3,14) = 1.
Exercise 2.2
1. By factorizing, find (a, b) for each pair of given integers a and b: (i)
a = 12, b = 15; (ii) a = 36, b = 60; (iii) a = 24, b = 35; (iv) a = 885,
b = 1995.
2.3 Division Algorithm

We now develop a more systematic approach to finding the G.C.D. of two
integers. We begin with the so-called Division Algorithm of which we shall
give two versions, the second generalizing the first.
Theorem 1 The Division Algorithm (First Version)

Let a and b be integers, b being non-zero. Then there exist positive, possibly zero,
integers q and r satisfying the condition
a = bq + r where 0 :::; r < b.
Proof
We notice first that we may dispose easily of the two cases when a = b and a < b.
For a = b we have
a=b1+0 (q=l,r=O)
and for a < b we have
a=bO+a (q=O,r=a).
The Integers 53
These two cases also prove the result for a = 1. We shallapply the Principle of
Induction and make the induction assumption that the result is true for
1,2, ... ,a - 1 and try to obtain the result for a. Once we have established this
conclusion we know that the result is true in general.
By the remarks above we need only consider the case of b < a. Then certainly
1 ~ a - b and so by our remarks above (if a - b ~ b) or by our induction assump-
tion (if b < a - b) there exist integers q' and r such that
a - b = bq' + r, 0 ~ r < b,
from which
a=bq+r, O~r<b,
on letting q = 1 + q'. This completes the proof. o

At first sight the result above seems to be purely an 'existence proof' implying
that for a given a and b there are appropriate q and r, but, like most existence
proofs, no means by which we can calculate q and r are evident. In fact we do
have a direct method, based on the above, of calculating the q (quotient) and
the r (remainder), that of 'long division'. Formally
from which
a = bq + r (0 ~ r < b).
We illustrate this process below.
Examples 7
1. a = 57, b = 19. Evidently q = 3, r = O.
2. a = 527, b = 47. By long division
11
47)527
47
57"
47
10
Thus 527 = 47.11 + 10 and so q = 11, r = 10.

3. a = 273921, b = 931. By long division
294
931)273921
1862
8772
8379
3931
3724
207
Thus
273921 = 931.294 + 207
and so
q = 294, r = 207.
For the Division Algorithm (Extended Version) we introduce a convenient

notation.
Definition 5
Let x be a real number. We define the modulus, or the absolute value, of x,
denoted by lxi, as follows:
Ixl = x (x ~ 0), Ixl = -x (x < 0).
It is an immediate consequence of the definition that
Ixl ~ 0 and Ixl = 0 if and only if x = 0
and that
Ixl = Iyl if and only if x = ±y (x, Y E lR).
Example 8
171 = 7, 1- 71 = -(-7) = 7.
Theorem 2 Division Algorithm (Extended Version)

Let a and b be integers, b being non-zero. Then there exist integers q and r satis-
fying the condition a = bq + r where 0 ::; r < 1b I.
The Integers 55
Proof
By the Division Algorithm (First Version) there exist q,r E NU {O} such that
lal=lblq+r, O~r<lbl·
We consider several cases.

If r = 0 then Ia I = Ib Iq and so a = ±bq from which a = b(±q) giving the result.
If r > 0 then we consider two cases. If a ~ 0 then a = Ib Iq + r and so
a = bq+r (b > 0) and a= (-b)q+r= b(-q) +r (b < 0),
of which only the second statement is wholly new. If a < 0 then -a = Ib Iq + r
and so a = -I b Iq - r which is almost in the desired form except that we require
to have a positive remainder. We make a small but significant adjustment by
writing
a = -Iblq+r= Ibl(-I-q) + (Ibl-r).
Now 0 < r < Ibl implies that 0 < Ibl- r < Ibl and so
a = bq' +r'
where r' = Ibl- rand q' = -1 - q (b > 0), q' = 1 + q (b < 0). This completes
the proof. 0
Examples 9
1. a = -574, b = 34. By the Division Algorithm (First Version) we have
574 = 34.16 + 30,
from which we have
-574 = 34.(-16) - 30 = 34(-16 - 1) + (34 - 30) = 34(-17) + 4
giving q = -17, r = 4.
2. a = -6412, b = -97. By the Division Algorithm (First Version) we have
6412 = 97.66 + 10,
from which we have
-6412 = (-97).66 - 10 = (-97).(66 + 1) + (97 - 10) = (-97).67 + 87
giving q = 67, r = 87.
Exercises 2.3
1. Given integers a and b, find integers q and r such that a bq + r =
= =
(0 ~ r < Ibl) where: (i) a 8, b 6; (ii) a = 27, b = 8; (iii) a 241,=
b = 35; (iv) a = -2513, b = 46; (v) a = 54321, b = -761; (vi)
a = -52148, b = -732.
2. Let x and y be real numbers. Prove (i) Ixyl = Ixllyl and (ii)
Ix + y I ~ Ix 1+ Iy I. Construct an example in which, for appropriate
choice of x and y, Ix - yl ~ Ixl-Iyl.
2.4 Euclidean Algorithm

We now describe an effective method, the so-called Euclidean Algorithm, for
finding the G.C.D. oftwo integers. The Euclidean Algorithm depends on repeat-
ing the Division Algorithm for a finite number of times until a particular con-
clusion, yielding the G.C.D., is reached. The method appeared originally in a
geometrical form as Proposition 2 of Book VII of 'The Elements' - a book
which was composed by Euclid of Alexandria around 300 BC and which survived
as a textbook for over 2000 years.
First we prove a brief but useful result.
Theorem 3
Let a, b and d be non-zero integers and let d be a positive common divisor of a, b.
Suppose there exist integers x and y such that
ax+by= d.
Then d is the greatest common divisor of a and b.
Proof
Let c be a common divisor of a, b. We require to show that c divides d. Now we
certainly have a = ca', b = cb' for suitable a', b' E Z. Hence
d = ax + by = (ca')x + (cb')y = c(a'x) + c(b'y) = c(a'x + b'y).
Hence c divides d and so d = (a, b). o
The Integers 57
Example 10
It is easy to verify that 97 is a common divisor of 291 and 388 but it is less easy,
without the use of the Euclidean Algorithm, to find an equation of the form given
in Theorem 3. One such equation is
291.15 + 388(-11) = 97,
and from this equation we may conclude that 97 is the G.C.D. of 191 and 388.
We shall give the explicit procedure of the Euclidean Algorithm but, by way of
preliminary illustration, we begin with two particular examples of its use.
Examples 11
1. Suppose we wish to find the G.C.D. of 4947 and 1552. Our procedure is to
apply the Division Algorithm systematically as follows. Applying the Division
Algorithm three times we have
4947 = 1552.3 + 291 (1)
1552 = 291.5 + 97 (2)
291 = 97.3 (3)

We claim that 97 is the required G.C.D. First we have to prove that 97 is
indeed a divisor. We consider the equations in reverse order. Certainly from
equation (3), 97 divides 291. But then from equation (2) 97 divides 1552.
In turn, from equation (1), 97 divides 4947. Thus 97 is a common divisor of
1552 and 4947. Now to prove that 97 is the G.C.D. we let d be a divisor of
1552 and 4947. By equation (1) d divides 291 and then from (2) d divides
97. Hence 97 is the G.C.D.
We also write 97 in the form of Theorem 3 as follows. From equation (2)
97 = 1552 - 291.5
and hence, from equation (1),
97 = 1552 - (4947 - 1552.3).5 = 1552 - 4947.5 + 1552.3.5

= 4947.(-5) + 1552.16.
As soon as we have established that 97 is a common divisor then the represen-
tation of 97 in the form above gives an alternative proof that 97 is indeed the
G.C.D.
2. Suppose we wish to find the G.G.D. of 163059 and 80004. Then applying the
Division Algorithm four times we have
163059 = 80004.2 + 3051 (1)
80004 = 3051.26 + 678 (2)
3051 = 678.4 + 339 (3)
678 = 339.2 (4)
We consider these equations in the reverse order to show that 339 is a

common divisor of 163059 and 80004.
From (4) 339 divides 3051. And so from (3) 339 divides 3051. But then (2)
implies that 339 divides 80004 and then finally from (1), 339 divides 163059.
Hence 339 is a common divisor of 163059 and 80004. By considering the
equations in the normal order we may conclude, as in the previous Example,
that 339 is, in fact, the G.G.D.
We may also establish this conclusion by writing 339 in the form
163059x + 80oo4y as follows. From (3) we have
339 = 3051 - 678.4
and so from (2) we have

339 = 3051 - (80004 - 3051.26).4 = 3051.105 - 80004.4
Applying (1) now gives

339 = (163059 - 80004.2)105 - 80004.4 = 163059.105 + 80004.(-214)
as we desired.
Keeping these two Examples in our minds we may now proceed to give the
Euclidean Algorithm in its proper generality.
The Euclidean Algorithm
Given two non-zero integers a and b, we shall describe how the Algorithm will
yield, in a finite number of steps, the G.G.D. of a and b. We follow the procedure
of the two examples above. For convenience we may suppose a > 0, b > 0 and
introduce the notation ao = a, al = b.
By the Division Algorithm we obtain integers ql and ~ such that
ao = al ql + a2, 0:::; ~ < al'
The Integers 59
If ~ = 0 then ao = alql and we have immediately that al is the G.C.D. of ao and

al' If ~ =F 0 then we continue as follows. We know, again by Division Algorithm,
that we obtain integers q2, a3 such that
al = a2q2 + a3, 0 ~ a3 < ~.
If a3 = 0 then we halt the procedure but if a3 =F 0 we continue and obtain inte-
gers q3' a4, such that
~ = a3q3 + a4, 0 ~ a4 < a3'
Either a4 = 0 or if a4 =F 0 then we continue as in the examples above.
ao = alql + a2 (1)
~=~~+~ ~
~ = a3q3 + a4 (3)
~=~~+~ W
Now continuing we obtain a finite or infinite sequence ai, ~, a3, ... of integers
such that
al > a2 > a3 > ... ~ O.
This is only possible if the sequence is finite and for some n, an + I = O. Thus
suppose we halt at the nth equation. As in the examples above we have the
following equations:
ao = alql + ~ (1)
al = ~q2 +a3 (2)
~ = a3q3 +a4 (3)
a3 = a4q4 + as (4)
(n - 2)
(n -1)
(n)
We claim that an is indeed the G.C.D. of ao = a and al = b. We prove first that
an is a common divisor of ao and al by considering the equations in the reverse
order. From (n) an divides an-I' Then from (n -1) ~ divides a n -2' Continuing
we have finally, on considering (1), that ~ divides ~ and al and so ~ divides ao.
Thus an is a common divisor of ao and al. Let now c be a common divisor of ao

and al. By (1) c divides ao - alql = a2. Then from (2) c divides al - a2q2 = a3·
Continuing we find that c divides an _ 2 - an -1 qn _ 1 = an. Thus, as we claimed, a
is the G.C.D. of ao and al'
We are also now able to write an in the form aox + alY for some x, Y E Z.
From (n - 1)
From (n - 2)
and so, substituting from (n - 2) into (n - 1),
From (n - 3)
and so, substituting from (n - 3) into the equation for an we have
an = (1 + qn-2qn-dan-2 - a n-3
= (1 + qn-2qn-d[an-4 - an-3qn-3)- an-3
= [-qn-3(1 + qn-2qn-l) - l)an-3 + (1 + qn-2qn-l)an-4
= -(1 + qn-3 + qn-3qn-2qn-dan-3 + (1 + Qn-2Qn-dan-4'
Continuing this procedure we shall eventually obtain an equation of the form
an = xao + ya l
for some x, y E Z.
We may summarize this last result in the form of a converse to Theorem 3.
Theorem 4
Let a and b be non-zero integers and let d be the greatest common divisor of a and
b. Then there exist integers x and y such that
d = ax + by.
We remark in passing that x and y are not unique since d = ax + by implies

that d = a(x - bt) + b(y + at) for any integer t.
The Integers 61
We conclude this section with one further example which, in view of our
extended discussion above, we shall give in as brief a form as is necessary for
calculation.
Example 12
We wish to find the G.G.D. of 108810 and 93346 and to write the G.G.D. as
108810x + 93346y for some x, y E Z.
108810 = 93346.1 + 15464 (1)
93346 = 15464.4 + 562 (2)
15464 = 562.27 + 290 (3)
562 = 290.1 + 272 (4)
290 = 272.1 + 18 (5)
272 = 18.15 + 2 (6)
18 = 9.2 (7)
Thus 2 is the G.G.D. of 108810 and 93346.
Reversing the order of the equations we have
(93346,108810) = 2 = 272 - 18.15 (from (6))
= 272 - (290 - 272).15 (from (5))
= -290.15 + 272.16
= -290.15 + (562 - 290).16 (from (4))
= 562.16 - 290.31
= 562.16 - (15464 - 562.27).31 (from (3))
= -15464.31 + 562.853
= -15464.31 + (93346 - 15464.6).853 (from (2))
= 93346.853 - 15464.5149
= 93346.853 - (108810 - 93346).5149 (from (1))
= 93346.6002 - 108810.5149
which gives the desired result.

Exercises 2.4
1. Let a, b, x, yand n be integers such that n divides both a and b. Prove
formally that n divides ax + by.
2. Use the Euclidean Algorithm to find the greatest common divisor (a, b)
of each pair of given integers a and b and write (a, b) in the form
ax + by for suitable integers x and y: (i) a = 32, b = 44; (ii) a = 150,
b = 105; (iii) a = 3718, b = 1222; (iv) a = 96, b = 764; (v) a = 7224,
b = 6214; (vi) a = 613640, b = 152881.
3. Let a, band n be integers. IT n divides a - b prove that n divides a2 - b2
and a 3 - b3 . If n divides a + b prove that n divides a3 + b3 but does not
necessarily divide a2 + b2 •
2.5 Primes
We have defined a prime p as being an integer strictly greater than 1 which is
divisible only by ±1 or ±p. It is now appropriate to examine some consequences
of the division of an integer by a prime.
Definition 6
Let a and b be non-zero integers. Then a and b are said to be coprime (or to be
prime to one another) if 1 is their greatest common divisor.
Example 13
60 and 77 are coprime since (60,77) = 1.
It is immediate from the results above that if a and b are coprime then there
exist x, y E Z such that ax + by = 1. The converse is also true, namely that if
for given a, bE Z, a '" 0, b '" 0, there exist x, y E Z such that ax + by = 1 then
(a, b) = 1.
We come now to a long-established and important result on division by a

prime.
The Integers 63
Theorem 5 ('The Elements', Book VII, Proposition 30)

Let a and b be integers and let p be a prime. Let p divide abo Then either p divides
a or p divides b.
Proof
For the sake of argument suppose p does not divide b. Then since the G.C.D. of
p and b divides p the G.C.D. can only be 1. Hence there exist x, y E Z such that
px + by = 1. But then apx + aby = a. Clearly p divides apx and p divides ab
(by hypothesis) and so p divides a. 0
Applications of this result arise frequently. Using Theorem 5 we prove, as a

matter of interest, the well-known result that v'2 ¢ Q, thus proving that the
set of real numbers R is strictly greater than the set of rational numbers Q.
Theorem 6
V2¢Q.
Proof
For the sake of argument suppose that v'2 E Q. Then we have m, n E N such
that v'2 =
m. We may further suppose that m and n have no common divisor in
n
N other than 1. Now m 2 = 2n2 and so 2 divides m 2 • From Theorem 5, 2 divides m
and we may write m = 2m' for some m' E N. Then 4(m')2 = (2m')2 =
m 2 = 2n2 and so 2(m')2 = n 2. But this now implies that 2 divides n 2 and so n.
We have arrived at a contradiction because we have deduced that 2 divides
both m and n. Hence our initial supposition is wrong and so
V2¢Q. o
Remark
The argument given above easily adapts to showing that J3, v"5, ... , VP
(p prime) are not in Q; indeed we may show that .;n (n E N), where n is not
of the form m 2 (m EN), is not in Q.
As reported in the Theaetetus, a dialogue of Plato (c. 429-347 BC), Theodorus

(born c. 460 BC) was the first to prove that J3, J5, ... , m are irrational, the
irrationality of v'2 being apparently already known.
Example 14
Let a, b E Z and let p be prime. Suppose p divides a 2 + pb2 • Then certainly p
divides (a 2 + pb2 ) - pb2, and so p divides a2 from which we conclude that
p divides a.
In Theorem 5 it should be clear that the issue is not so much the primeness of
p but the fact that p and b are coprime. Consequently, by essentially the same
argument as in that theorem, we may obtain the following result.
Theorem 7
Let a and b be integers and let q be an integer which is prime to b. Let q divide abo
Then q divides a.
We give an application of this last result.
Example 15
Let a and b be integers having d as G.C.D. Then, as we recall from Theorem 4,
there exist integers x and y such that ax + by = d.
We know that x and y are not unique. Suppose x' and y' are integers such that
ax' + by' = d. Our natural curiosity should induce us to enquire whether there
exists any relationship between the pair x and y and the pair x' and y'. Now
we have a = a'd and b = b'd for some coprime integers a' and b'. But, as we
have the equation ax + by = d = ax' + by' it follows, on rearranging, that
a(x' - x) = -b(y' - y)
and, on dividing by d,
a'(x' - x) = -b'(y' - y)
But a' and b' are coprime and so, by Theorem 7, a' divides y' - y. Thus there
exists an integer t such that
y' -y= a't.
But then we have
a'(x' - x) = -b'a't
and so
x' - x = -b't.
The Integers 65
Thus, finally, the relationship we are seeking is:

, , b
x =x - bt =x - - t,
d
, , a
y = y + a t = Y + d t,
where d is the G.C.D. of a and b.
It is important to be able to factorize integers into products of primes. For

small numbers such as 1363 we can usually perform the factorization with a
modest supply of pen and paper, for large numbers such as 45615161 a copious
supply of pen and paper may enable a factorization to be found but for very
large numbers such as 264 + 1 even an advanced calculator may be unable to per-
form the factorization. The reader is invited to factorize 1363 but may be relieved
to be advised that
45615161 = 5879.7759, and 264 + 1 = 274177.67280421310721

when decomposed into prime factors.
A useful guide to the largeness of the prime that we need to consider in any
attempted factorization of a possibly composite number is given by the next
lemma.
Lemma 3
Let n be a composite integer. Then there is at least one prime p dividing n for
which p :5 yn.
Proof
Let p be prime dividing n. IT p2 divides n then p2 :5 n and so p :5 yn. If p2 does
not divide n then there is a second prime q such that pq divides n. Changing
notation, if necessary, we may suppose that p < q. Then p2 < pq :5 n and so
p< yn. 0
Although possibly unable to perform an explicit factorization into prime fac-

tors we know nevertheless that one always exists and this fact is made precise in
the next theorem, which on account of its importance, has been designated 'The
Fundamental Theorem of Arithmetic'. Before formally stating the theorem we
give an illustration.
Example 16
The number 5564832 is expressible as a product of primes in many ways, three of
which are as follows:
5564832 = 22 .72 .13.3.7.23 .13 = 7.13.22 .7.3.7.23 .13 = 25 .3.73 .132 •
We assert that exactly the same primes 2, 3, 7, 13 will appear in any such
factorization and that, furthermore, the powers 5, 1,3,2 to which respectively
the primes 2, 3, 7, 13 appear, are determined solely by the number 5564832.
Theorem 8 The Fundamental Theorem of Arithmetic

('The Elements', Book IX, Proposition 14)
Every integer n, n E N (n ~ 2) is expressible as a product of a finite number of
primes. Furthermore if n is written as
n = P1P2·· ·Pr = qlq2'" qa.
where Pi> P2, ... ,Prj ql, q2, ... ,qa are primes then r = 8 and, with a suitable
re-ordering if necessary, Pi = qi (i = 1,2, ... ,r).
Proof
IT n is a prime and, in particular, if n = 2, then n is trivially a product of primes.
Suppose n > 2. We shall argue by induction and suppose that the result is true
for n - 1 and for all m such that 2 ~ m ~ n - 1. We prove first that n is a
product of primes.
IT n is a prime then, as we remarked, the assertion is true. Suppose that n is
composite and that n = nln2 where nil ~ EN, 1 < nl < n, 1 < n2 < n. By
our assumption each of nl and n2 is a product of primes and so therefore is n.
Suppose now that
n = P1P2" ·Pr = qlq2'" qa'
Then Pl divides qlq2 ... qn' By Theorem 5, Pl must divide one of qi> q2,' .. ,qa' By
renumbering the qi'S if necessary we may suppose Pl divides ql' But Pl and ql are
primes so Pl = ql' Thus we have
from which
P2P3 ... qr = q2q3 ... qa
By our induction assumption we have r = 8 and with suitable renumbering
Pi = qi (i = 2,3, ... , n). This completes the proof. 0
The Integers 67
Arising out of this result we may collect together equal prime divisors and so
we conclude that n may be written as
where PbP2, ... ,Pt are distinct primes and u,

> 0 (i = 1,2, ... , t).
Many fascinating, but sometimes still unresolved, questions may be raised in
regard to primes. Euclid answered one obvious initial question, namely 'how
many?'
Theorem 9 ('The Elements', Book IX, Proposition 20)

There are infinitely many primes.
Proof
Let PI,P2, ... be the primes in ascending order (of course PI = 2, P2 = 3, P3 = 5,
etc.). We argue by contradiction and so suppose there is only a finite number
N of primes; PN is then the largest prime. Let
M = (PIP2.· ,PN) + 1
M cannot be a prime since M> PN' But, since M is a product of some of the
primes Pb P2, ... ,PN, at least one of these primes must divide M, say PI divides
M. But we have
M 1
--P2P3",PN =-.
PI PI
where the number on the left-hand side of this equation is an integer whereas the
number on the right-hand side is not. We have reached a contradiction and so
there is not a finite number of primes and we obtain the result. 0
Various simple extensions of this result are known. We shall give one such
extension after some preliminary remarks.
Any odd integer is either of the form 4n + 1 or 4n + 3 for suitable nEZ, for
example 21 = 4.5 + 1 and 23 = 4.5 + 3. In particular every prime other than 2
is of the form 4n + 1 or 4n + 3. We observe that the product of odd integers,
all of which are of the form 4n + 1, is again an integer of this form since
(4nl + 1)(4~ + 1) = 4(4nl~ + nl +~) + 1.
Example 17
The first ten primes, of the form 4n + 3, are 3, 7, 11, 19, 23, 31, 43, 47, 59, 67.
Theorem 10
There are infinitely many primes of the form 4n + 3.
Proof
Let PI, P2, . .. be the primes of the form 4n + 3 in ascending order (of course
PI = 3, P2 = 7, P3 = 11, etc.). We argue by contradiction and so suppose there
is only a finite number N of such primes; PN is then the largest such prime. Let
M = 4(P2P3" ,PN) +3
Then, by an argument similar to that used in the proof of Theorem 9, M is not
divisible by any ofp2,P3, ... ,PN' Further M is not divisible by 3 since 3 does not
divide 4 (P2P3 ... PN)' Thus M is of the form 4n + 3 but no prime factor of M is of
this form. Thus all the prime factors of M must be of the form 4n + 1. But from
the remark above the product of such prime factors is again of the form 4n + 1
and not of the form 4n + 3. Thus we obtain a contradiction and so the result is
proved. 0
We have developed some elementary properties of integers in order to whet the

appetite for further study in number theory and, more importantly for our pur-
poses, to prepare the ground for the axiomatic treatment of groups, rings and
fields which is to follow. However, before leaving the present context we offer
some random remarks on these objects of wonder the primes.
Sieve of Eratosthenes
Eratosthenes (c. 275-195 Be) gave a method for detecting primes which, for
fairly obvious reasons, is known as a 'sieve'. The implementation of the
method requires Lemma 3 above.
Lemma 3 implies that if we wish to determine whether a positive integer n is
prime then we may confine our attention to possible prime divisors less than or
equal to ,;no We illustrate the method by detecting those primes less than 100.
We write down the integers from 1 to 100 and successively eliminate those
divisible by 2, 3, 5, 7, these being the primes :'5 v'100 = 10 as follows. We
circle 2 as being prime and stroke out every 2nd number thereafter. We circle 3
The Integers 69
as the next prime and stroke out every 3rd number thereafter. Similarly we circle
5, and then 7, and stroke out every 5th, and then every 7th, number. We obtain
the following picture:
1 ~ ® A' @ .0- ({) it .9' .w

11 a 13 14 l5' 16' 17 1-8' 19 .2tr
.2t .22 23 .24 ..25 .26" :llr .18' 29 ..36'
31 M ..3a- M .35' ..36' 37 .38' .39' .Mr
41 M 43 M .Mr .46' 47 .48' .49' .w
M M 53 M M M .ar- M 59 ..00-
61 M .63' M .65' .66 67 $' M :lfJ
71 :n 73 :K :;s :r6 :rr ~ 79 m
M .82' 83 M .85' M M .88' 89 .9tr
.9i:' M .93' M .95' .96' 97 .98' .99" .100
The numbers without strokes, other than 1, are 11,13,17,19,23,29,31,37,

41,43,47,53,59,61,67,71,73,79,83,89,97 which are therefore together with 2,
3, 5, 7 the primes less than 100.
The method, while systematic, has obvious limitations.
The Distribution of the Primes
The primes are distributed irregularly amongst the integers but, nevertheless, we
should like some measure of the irregularity. Two results which are of especial
interest will be quoted without proofs (which are analytic rather than algebraic).
Let PllP2,'" be the ascending sequence of primes. J. Bertrand surmised
(1845) and P.L. Chebyshev (1821-94) proved that
Pn+l < 2Pn (n EN).
For the second result we introduce a function commonly denoted by 11" (thus 11"
is here not 3.14159...) and defined for x E N by letting 1I"(x) be the number of
primes $ x. Quick calculation shows that 11"(1) = 0, 11"(2) = 1, 11"(3) = 11"(4) = 2,
11"(5) = 11"(6) = 3,11"(7) = 11"(8) = 11"(9) = 11"(10) = 4, etc. K.F. Gauss (1777-1855)
formulated the remarkable conjecture that
1I"(x) log x 1
--:..~-=-- -+ as x -+ 00.
x
This was established independently by C.J. de la Valle-Poussin (1866-1962)

and J. Hadamard (1865-1963), the longevity of whom cannot solely be
attributed to an interest in number theory.
Goldbach's Conjecture
This conjecture originated with C. Goldbach (1690-1764) and is to the effect

that every even integer, other than 2, is the sum of two primes; for example,
4 = 2 + 2, 6 = 3 + 3, 8 = 5 + 3, 10 = 3 + 7 = 5 + 5. The difficulty of tackling
the conjecture seems to lie in the fact that primes are clearly concerned with
factorization but not obviously with addition. Various allied results have been
obtained but the conjecture itself remains to tantalize.
After this brief excursion into aspects of number theory we shall develop the
previously mentioned axiomatic treatment through a discussion of polynomials.
Exercises 2.5
1. Write down formal proofs that J3 and v'6 are irrational.
2. Prove that ../2 + J3 is irrational.
3. The sum of two rational numbers is rational. Is the sum of two
irrational numbers always irrational?
4. Write down a formal proof of Theorem 7.
5. Two primes p and q are said to be a 'prime pair' if q = p + 2, for
example 17 and 19 is a prime pair. How many prime pairs are there
between 1 and 1oo? (It is unknown whether or not there are infinitely
many prime pairs.)
6. Apply the Sieve of Eratosthenes to find the primes between 100 and
200.
7. Calculate lI"(lOn) , n = 1,2, ... ,10.
8. In how many ways may 144 be written as the sum of two primes?
3
Introduction to Rings
In the previous chapter we have seen that the integers possess a division
algorithm and that from this division algorithm there may be derived a
Euclidean Algorithm for finding the greatest common divisor of two given
integers. 'Polynomials' share many properties in common with the integers,
having a division algorithm and a corresponding Euclidean Algorithm. As our
treatment of polynomials proceeds, initially somewhat informally, it will
become apparent that we need to consider much more precisely the extent to
which integers and polynomials share common features. In this way we shall
be led to enunciate axioms for an algebraic system called a 'ring' and for a
ring of a particular type called an 'integral domain' which incorporates some
of the features common to integers and polynomials.
Axioms in mathematics are never constructed arbitrarily but are designed to
focus attention on significant aspects of the system or systems under considera-
tion. While axioms are famously present in Euclid's 'Elements' their modern
extensive use stems from the mathematical work of the nineteenth century, a
decisive influence in their use being that of the great D. Hilbert (1862-1943).
The concept of a ring was introduced by R. Dedekind (1831-1916) but the
first set of axioms for a ring, although not quite equivalent to those in use
today, was published in 1914 by A.H. Fraenkel (1891-1965).
71
3.1 Concept of a Polynomial

What does the term 'polynomial' mean to us? We think of algebraic expressions
such as 2x + 1 or x 2 + 5x + 6 or 2x3 + x 2 - 2x -1 etc. We speak of these as
being 'polynomials in x' and say that 2x + 1 has 'degree' 1, x 2 + 5x + 6 has
'degree' 2 and 2x3 + x 2 - 2x -1 has 'degree' 3. We know how to add, subtract
and multiply such polynomials:
(3x + 5) + (4x 2 - 2x -1) = 4x 2 + X + 4,
(2x 2 - 6) - (x 2 - 5x -1) = x 2 + 5x - 5,
2
(x + 2x + 4)(3x + 1) = x 2(3x + 1) + 2x(3x + 1) + 4(3x + 1)
= (3x
3
+ x 2) + (6x 2 + 2x) + (12x + 4)
2
= 3x3 + 7x + 14x + 4.
As in the case of the integers we may perform long division, for example 3x + 1
does not divide 9x3 - 3x2 + 6x + 4 but leaves a remainder when we employ long
division as follows:
3x2 - 2x+ ft3

3x + 1)9x3 - 3x2 + 6x + 4
9x3 + 3x2
-6x2 + 6x
-6x2 - 2x
8x+4
8x+ ft3
--4
3
Thus we write
4
3 2
9x - 3x 3x2 _ 2x + ~ + _3_
+ 6x + 4 =
---:3~x-+--:-l-- 3 3x + 1 '
or, more usefully,
which is in a form to be expected from the existence of a division algorithm. For

convenience we gather together in a more formal manner some of the aspects of
polynomials mentioned above.
Introduction to Rings 73
Definition 1
By a polynomial in :z: over Q we understand an expression a(x) of the form
a(x) = ao + alx + ... + amxm
where the so-called coefficients ao, aI, ... ,am are rational numbers, and x is
sometimes called an 'indeterminate'. If all of the coefficients are 0 the polynomial
is said to be the zero polynomial, denoted by O. If a(x) is not the zero polynomial
and if am =F 0 then a(x) is said to have degree m, briefly deg a(x) = m. If a(x)
has degree m and am = 1 then a(x) is said to be monic. A polynomial which is
either the zero polynomial or has degree 0 is said to be a constant polynomial
and is therefore of the form a(x) = ao (ao E Q). Polynomials of degrees 1, 2
or 3 are sometimes called linear, quadratic or cubic respectively.
Two polynomials a(x) and b(x) where
a(x) = ao + alx + + amxm,
b(x) = bo + b1x + + bnxn,
are deemed to be equal if and only if m = n and ao = bo, al = bl , ... ,an = bn.
In particular a(x) = 0 if and only if all coefficients of a(x) are O. The addition
and multiplication of a(x) and b(x) are defined as
a(x) + b(x) = (ao + bo) + (al + bl)x + ... ,
a(x)b(x) = (~arxr) (~baaa)
The set of all polynomials in x over Q is denoted by Q[x]. Similarly Z[x], lR[x],
C[x] are defined by choosing the relevant coefficients for the polynomials.
Example 1
1 + V2x 2 + v'3x5 is a polynomial in x over IR of degree 5. 5 + 2y + 3y2 is a quad-
ratic polynomial in y over Z (and also over Q, IR and C). 2 +!i! +;; is a monic
cubic polynomial in z over Q.
A key property of polynomials in x over Q, say, is the possibility of substitut-

ing for the x. We may replace x in any equation involving polynomials in x by
any number, not necessarily in Q, and thereby obtain a valid equation involving
that number. Thus if a(x) = ao + alx + ... + amxm E Q[x] and if e E JR we have
a(e) = ao + ale + ... + amcm E JR and if f(x), g(x), h(x) E Q[x] are such that
f(x)g(x) = h(x) then f(e)g(e) = h(e).
Example 2
Let f(x) = x 2 + X + 1, g(x) = 2x - 3, h(x) = 2x3 - x2 - X - 3. Then we may
verify that f(x)g(x) = h(x). Hwe substitute V2 for x then
f(V2) = (v'2)2 + v'2 + 1 = 3 + V2,

g( V2) = 2v'2 - 3, h( v'2) = 2( V2)3 - (v'2)2 - v'2 - 3
= 3v'2 - 5,
and, as expected,
f( v'2)g(v'2) = h( V2).
We shall investigate polynomials in Q[x]. With obvious and appropriate

changes our results apply also to JR[x] and C[x] but not always to Z[x].
Definition 2
Let f(x) and g(x) be polynomials in Q[x], g(x) =F O. Then we say that g(x)
divides f(x), or that g(x) is a divisor of f(x), if there exists a polynomial
h(x) E Q[x] such that f(x) = g(x)h(x).
Notice that the zero polynomial 0 is divisible by any polynomial g(x) =F 0 since
0= g(x)O and that any polynomial f(x) =F 0 has the trivial divisors c and cf(x)
for any e E Q, c =F O.
Examples 3
1. 2x 2 + 2 E Q[x] has the divisors c and e(x 2 + 1) (c =F 0) since
2
2x + 2 = - c(x 2 + 1).
2
e
2. 2x 2
+ 3x + 1 divides 2x + x - 2x - 1 since
3 2
2x3 + x2 - 2x - 1 = (2x 2 + 3x + 1)(x - 1).

3. In Q[x), X 2 - 2 has only the trivial divisors c, C(X2 - 2), cEQ, c oF O. In lR[x)
however there is the non-trivial factorization
x2 - 2 = (x - V2)(x + V2).
This example shows that it is important to know in which system the factor-
ization is to take place.
Lemma 1
Let f(x), g(x) and h(x) be polynomials in Q[x) such that g(x) and h(x) are non-
zero. Let g(x) divide f(x) and h(x) divide g(x). Then h(x) divides f(x).
Proof
This follows closely the proof of the corresponding lemma in Chapter 2.
Definition 3
Let f(x) and g(x) be non-zero polynomials in Q[x). A polynomial h(x) is a
common divisor of f(x) and g(x) if h(x) divides f(x) and h(x) divides g(x). A
common divisor d( x) which is monic and which is such that d( x) is divisible
by any common divisor of f(x) and g(x) is called the greatest common divisor
(G.C.D.) of f(x) and g(x).
Example 4
The quadratic polynomials 2x2 + 7x + 3 and 6x2 + x-I in Q[x) have G.C.D.
x +! since
2
2x +7x+3= (2x+l)(x+3) =2(x+D(x+3),
6x
2
+ x -1 = (2x+ 1)(3x -1) = 2(X +~)(3X -1).
Exercises 3.1
In each of the following find the G.C.D. of the pair of polynomials
1. x 2 + X - 2, x 2 + 3x + 2.
2. 6x2 + X - 2, 15x2 + 13x + 2.
3. x 3 + x 2 - x-I, x3 + 2x 2 + X.
4. x 3 + x 2 + X + 1, 2x 3 - x 2 + 2x - 1.
3.2 Division and Euclidean Algorithms

As we remarked above, we shall prove results only for Q[x] but they apply with
trivial modifications for lR[x] and C[x].
Theorem 1 The Division Algorithm

Let a(x) and b(x) be polynomials over Q[x] , b(x) being non-zero. Then there exist
polynomials q( x) and r( x) with coefficients in Q satisfying the condition
a(x) = b(x)q(x) + r(x)
where either r( x) = 0 or if r( x) ;/; 0 then the degree of r( x) is strictly less than the
degree of b( x).
Proof
Let a(x) and b(x) be the polynomials of degree m and n respectively given by
a(x) = ao + alx + + amxm, am;/; 0,
b(x) = bo + b1x + + bnxn, bn ;/; O.
We first dispose of a trivial case, namely deg a(x) < deg b(x). We simply write
a(x) = Ob(x) + r(x)
where q(x) = 0 and r(x) = a(x), which gives the result.
We shall use the Principle of Induction by arguing in regard to deg a(x).
If deg a(x) = 0 then a(x) = ao . If also deg a(x) = deg b(x) then b(x) = bo
and so
a(x) = b(x)q
where q = :;. On the other hand if deg a(x) < deg b(x) then we have the result
by the trivial case above.
Let now deg a(x) > O. Suppose the result is true for all polynomials of degrees
strictly less than the degree of a(x). We wish to prove the result for a(x). If
deg a(x) < deg b(x) then again we have the trivial case above. Suppose therefore
that deg b(x) :5 deg a(x). Then we aim to construct a polynomial of degree
strictly less than deg a(x) and to apply the induction assumption to this poly-
nomial. Let
f(x) = a(x) - ~m xm-nb(x) (by convention x O = 1).

n
Then f(x) has degree strictly less than deg a(x) since the terms in x m cancel as
we see below:
f(x) = (ao + alx + ... amxm) - ~m (b o + b1x + ... + bnxn)xm- n.

n
By the induction assumption there exist polynomials p(x) and r(x) such that
f(x) = b(x)p(x) + r(x)
where either r(x) = 0 or if r(x) i= 0 then deg r(x) < deg b(x). Hence
b(x)p(x) + r(x) = f(x) = a(x) - ~m xm-nb(x)
n
and so
a(x) = b(x) (P(X) + ~: x - ) + r(x)

m n
= b(x)q(x) + r(x)
where q(x) = p(x) + ~m x m- n and either r(x) = 0 or if r(x) i= 0 then deg r(x) <
n
deg b( x). This completes the proof. 0
The following useful result comes as a consequence of the Division Algorithm.
Theorem 2 The Remainder Theorem

Let f(x) E Q[x] and let CEQ.
1. There exists q(x) E Q[x] such that f(x) = (x - c)q(x) + f(c).

2. x - c divides f(x) if and only if f(c) = o.
Proof
1. By the Division Algorithm there exist q(x) and r(x) such that
f(x) = (x - c)q(x) + r(x)
whereeitherr(x) = Oorifr(x) i= Othendeg r(x) < deg (x - c) = 1. In either
case r(x) = ro where ro is a, possibly zero, constant. Then
f(c) = (c - c)q(c) + ro = Oq(c) + ro = roo
2. 2 is an immediate consequence of 1. o
Example 5
f(x) = 2x5 + X 4 + 7x3 + 2x + 10
is divisible by x + 1 since
f(-I) = 2(-1)5 + (_1)4 + 7(-1)3 + 2(-1) + 10
= -2 + 1 - 7 - 2 + 10 = O.
Examples 6
For the polynomials a(x) and b(x) below we find q(x) and r(x) such that
a(x) = b(x)q(x) + r(x)

where r( x) = 0 or if r( x) :1= 0 then deg r( x) < deg b(x). We employ long division
which is, in fact, the basis of the Division Algorithm.
1. a(x) = x 3 + 4x2 + 5x + 7, b(x) = x + 1.
x + 1) x 3 + 4x2 + 5x + 7
3 2
X +X
3x2 +5x
3x
2
+ 3x
2x+7
2x+2
5
Thus a(x) = b(x)q(x) + r(x) where q(x) = x 2 + 3x + 2, r(x) = 5.

2. a(x) = 2x3 + 7x2 + 1, b(x) = 3x + 2

jx2 +.!jx _ ~
3x + 2)2x3 + 7x2 +1
2x3 +!X
3
2
!1 X 2
3
!1 X 2 +~X
3 9
~x+l
~X-~
l!2
27
Thus
a(x) = b(x)q(x) + r(x)
2
where q(x) = jx +!jx -~' r(x) =~'
We turn now to the matter of finding the G.C.D. of two polynomials. The
Euclidean Algorithm used here to find the G.C.D. is exactly analogous to the
Euclidean Algorithm used in the case of the integers, we merely have to
modify some details appropriately. In these circumstances we shall content our-
selves therefore by illustrating the method by means of an example.
Examples 7
1. Suppose we wish to find the G.C.D. of
a(x) = 2x3 - 5x2 - 2x - 3 and b(x) = x3 - x2 - x-IS.

Then as in the Euclidean Algorithm for finding the G.C.D. of two integers we
use the Division Algorithm which in this example must be applied three times
as follows:
2x3 - 5x - 2x - 3 = 2(x3 - x 2 - x-IS) + (-3x 2 + 27),
(1 1)
x 3 -x2 -x-15=(-3x2 +27) -3x+3 + (8x-24),
-3x +27= (8x-24) (1

2 -Sx- 9) .
S
Thus we would expect that 8x - 24 would be 'almost' the G.C.D. However,

it will be recalled that, for definiteness, we have chosen that the G.C.D.
should be monic. But 8x - 24 = 8(x - 3) and so x - 3 is the required
G.C.D. Furthermore by considering the above equations in reverse order
we have
8(x - 3) = (x 3 - x
2
- x -15) - (-3x
2
+27) (-~+~)
= b(x) - [a(x) - 2b(x)] ( -~+~)
= a(x)[-( -~+~)] +b(x) [1 +2(-~+~)]

= a(x) [~- ~] + b(x) [_ 2; + ~] ,
from which we have
1 1
x - 3 = a(x) 24 (x - 1) + b(x) 24 (-2x + 5).
2. Suppose we are given the polynomials a(x) = 2x +3x +5x +6x +2and 4 3 2
b(x) = x 4 + 5x 2 + 6, and we wish to find the G.C.D. d(x) and to write d(x)
in the form
d(x) = a(x)f(x) + b(x)g(x)
for suitable polynomials f(x) and g(x).
We have
a(x) = 2b(x) + (3x 3 - 5x 2 + 6x - 10),
b(x) = (3x 3 - 5x 2 +6x -10) ( X 5) +(52_x

-+-
3 9 9
+-
104)
9'
2
- x2 +-
3x3 -5x2 +6x-1O= (52 104) (27
- x -45)
-
9 9 52 52'
Then, since
52 x2 + 104 = 52 (x 2 + 2)
9 9 9 '
the G.C.D. is x 2 + 2. Also
x2 + 2 = ~ (52 x2 +
52 9
104)
9
= :2 [b(X) - (3x
3
-
2
5x +6x -1O)(i+~)]
1
= 52 [9b(x) - (3x 3 - 5x 2 + 6x - 1O)(3x + 5)]
1
= 52 [9b(x) - (a(x) - 2b(x))(3x + 5)]
= -(3x + 5) () (6x + 19) b( )

52 ax + 52 x.
Thus
f( x ) = -(3x + 5) ( ) = 6x + 19
52' 9 x 52
is a possible solution for f(x) and g(x). (Recall that we know that f(x) and
g( x) are not uniquely determined.)
Exercises 3.2
1. For the given polynomials a(x) and b(x) in Q[x] , find q(x) and r(x) in
Q[x] such that
a(x) = b(x)q(x) + r(x)
where either r(x) = 0 or if r(x) :F 0 then deg r(x) < deg b(x).
(i) a(x) = x 3 + 3x2 + 1, b(x) = x4 + 1.
(ii) a(x) = 3x3 + 3x2 + 2x + 1, b(x) = 3x2 + 2.
(iii) a(x) = 2x2 + 5x + 1, b(x) = 5x - 1.
(iv) a(x) = x 3 + 4x2 + 4x + 1, b(x) = x + 3.
(v) a(x) = x 4 + 3x2 + 1, b(x) = 2x2 + 1.
2. For the given polynomials a(x) and b(x) in Q[x] , find the G.C.D. d(x)
and write d(x) in the form
d(x) = a(x)f(x) + b(x)g(x)
for suitable polynomials f(x) and g(x) in Q[x].
(i) a(x) = x 4 + x 3 + X + 1, b(x) = x 2 + X + 1.
(ii) a(x) = 2x3 + lOx 2 + 2x + 10, b(x) = x3 - 2x 2 + X - 2.
(iii) a(x) = x 4 - 4x3 + 2x - 4, b(x) = x3 + 2.
(iv) a(x) = x 3 + 5x2 + 7x + 2, b(x) = x 3 + 2x 2 - 2x - 1.
3.3 Axioms and Rings

From our deliberations we have seen that the system of polynomials behaves in a
similar manner to the system of the integers; it would therefore seem advisable to
study further whatever aspects these two systems have in common. In order to
understand better what apparently similar systems have in common it is often
convenient to lay down a common set of axioms which is satisfied by the systems.
A set of axioms is not invented arbitrarily but is constructed to reproduce
aspects of common and significant interest across the systems. Once the
axioms are laid down the axiomatic system then delineated takes on, as it
were, a life of its own and may then be explored both for its own sake and to
achieve a better overall understanding of previously considered systems.
In regard to the integers and polynomials we observe that there are in each
system two crucial operations, one is the addition of two numbers or polynomials
and the other is the multiplication of two numbers or polynomials. These opera-
tions are said to be binary since they involve two numbers or polynomials. The
operations satisfy certain conditions which are tacitly assumed in elementary
treatments but which must be made explicit when we pass to an axiomatic treat-
ment. We shall give appropriate axioms for the system, later defined to be a ring,
in which both of these operations occur. For convenience of reference and in com-
mitting the axioms to memory, we give the definition of a ring in a succinct form
but then follow the definition with a commentary upon the axioms of the defini-
ition. Finally, we conclude with the definition of an integral domain which is a
ring of a particular kind and which brings together many of the more obvious
features of integers and polynomials.
Definition 4
Let R be a non-empty set in which there are defined two binary operations called
addition and multiplication. For a, b E R the outcome of the addition of a and b,
called the sum of a and b, is denoted by a + b ('a plus b') and the outcome of the
multiplication of a by b, called the product of a by b, is denoted by the simple
juxtaposition ab ('a times b'). Then R is called a ring if the following axioms hold.
1. Axioms oC Addition
1.1 For all a,b E R, a + bE R (closure).

1.2 For all a, b, c E R, (a + b) + c = a + (b + c) (associativity).
1.3 There exists a distinguished element denoted by 0 such that for all
a E R, a + 0 = 0 + a = a (0 is 'zero').
1.4 For any a E R there exists an element denoted by -a E R such that

a + (-a) = (-a) + a = 0 (the inverse of addition).
1.5 For all a, bE R, a + b = b + a (commutativity).
2. AxiOIDB of Multiplication
2.1 For all a, bE R, ab E R (closure).
2.2 For all a, b, e E R, (ab)e = a(bc) (associativity).
3. AxiOIDB of Distributivity
3.1 For all a, b, e E R, a(b + e) = ab + ac (distributivity).
3.2 For all a, b, e E R, (a + b)c = ae + be (distributivity).
Commentary
1.1 The closure of addition is to ensure that the sum of a and b also belongs
to R. It would be a bizarre system if this fact were not to be so.
1.2 The associativity of addition enables us to dispense with brackets in
addition. Thus we may write simply a + b + e where the meaning is
(a + b) + e or, equally, a + (b + c).
1.3 This axiom only states that a zero exists. On the face of it there could be
more than one such zero. However, if 0' were to denote a second zero we
would have for all a E R,
a+O=O+a=~ a+~=~+a=L
In particular we would have
0' + 0 = 0 + 0' = 0', 0 + 0' = 0' + 0 = 0
from which ~ = o. Henceforth we may speak of the zero of R.
1.4 Now that we know that the zero element is unique, by carefully applying
the axioms we may establish that for each a E R the inverse of addition
-a is uniquely determined by aj if we suppose that we have b E R such
that a + b = b + a = 0 then we may show that b = -a as follows.
By 1.2 (associativity)
b+ (a+ (-a)) = (b+ a) + (-a).
But
b+ (a+ (-a)) = b+O = b
and
(b + a) + (-a) = 0 + (-a) = (-a),
from which we have b = -a. Henceforth we may speak of the inverse of
addition -a for given a E R. Notice that as a + (-a) = (-a) + a = 0 we
have a = -(-a).
1.5 This axiom states that the sum of a and b is independent of the order in
which a and b are added to one another; a and b are said to 'commute'.
(The words 'commutativity' and 'commute' stem from electrical
engineering in which a commutator is an apparatus for reversing the
current.)
2.1 The closure of multiplication is to ensure that the product of a and b
belongs to R (see comment on 1.1). Note, however, that we must main-
tain the order of a followed by b since, in general, ab and ba are not neces-
sarily equal.
2.2 The associativity of multiplication enables us to dispense with brackets
in multiplication. Thus abe is defined unambiguously, being either (ab)c
or a(be).
3.1/3.2 These axioms give conditions linking addition and multiplication. (The
reader may recall the remarks in Chapter 2 on Goldbach's Conjecture.)
In these axioms there are certain tacit assumptions in regard to the order in
which the bracketing and the two binary operations are to be considered.
Thus the axiom a(b + e) = ab + ac means that, on the left-hand side, we add b
and e within the brackets to give b + e and then we perform a multiplication
by a whereas, on the right-hand side, we multiply a and b and also multiply a
and e before performing the addition of ab and ac. These remarks may seem
superfluous but only because we are so accustomed to the particular order of pre-
cedence of brackets and operations (an order which in years gone by was encap-
sulated in the mnemonic BODMAS = brackets/of/division/multiplication/
addition/subtraction). For example, by a + be, we understand that band c are
first multiplied to give be and then a and be are added.
A convention also arises in regard to the minus sign' -'. In writing a - b we are
actually simply writing a followed by the additive inverse of b; however by a - b
we understand that in fact we intend a + (-b), the + sign being implied but
normally omitted.
If we now relate these axioms either to the set of integers Z or to the set of poly-
nomials lQ[x] with the usual addition and multiplication operations, we should
observe that the axioms are clearly satisfied; for example, the integers satisfy
1.5 since we have assumed from early childhood that the sum of two integers
was the same irrespective of the order in which they were added. Nevertheless
the reader should convince himself or herself, certainly for the integers and p0s-
sibly for polynomials, that the other axioms are equally obviously satisfied. We
shall henceforth speak of the 'ring of integers' and of the 'ring of polynomials'
(although this latter terminology is still a trifle imprecise).
Example 8
Rings arise in many ways. Below we consider a ring the elements of which are
mappings. Let X be a non-empty set. Let R be the set of mappings from X to
JR. We introduce an addition and multiplication, denoted by '.', into R as
follows. Let 1 : X --t JR and 9 : X --t JR. Define the sum 1 + 9 and the product
I. 9 as mappings from X to JR by,
(J + g)(x) = I(x) + g(x) (x EX),
(J.g)(x) = I(x)g(x) (x EX).
We note that on the right-hand side of the first equation the plus sign + refers to
addition in JR and on the right-hand side of the second equation the product is in
JR. Consequently the sum and product of the mappings 1 and 9 are well-defined.
We claim that R is a ring with the given definitions of sum and product. We
require to verify all the axioms for a ring. We use the same numbering system
as above.
1.1 Certainly, by definition, R is closed under addition.

1.2 Let I, g, hER. We want to show that (J + g) + h = 1 + (g+ h). Now for
xEX
[(J + g) + h](x) = (J + g)(x) + h(x) (definition)
= [j(x) + g(x)] + h(x) (definition)
= I(x) + [g(x) + h(x)] (associativity in JR)
= I(x) + (g+ h) (x) (definition)
= [I + (g + h)](x) (definition).
Hence, by the condition for the equality of mappings

(J + g) + h = 1 + (g + h).
1.3 The zero of R is given by the mapping 0 given by O(x) = 0 (x E X) since for
allfERandxEX
(f + O)(x) = f(x) + O(x) (definition)
= f(x) +0 (definition of 0 E R)
=f(x) (0 E IR)
= 0 + f(x) (0 E IR)
=O(x) + f(x) (definition of 0 E R)
= (0 + f)(x) (definition) .
Hence, again by the condition for the equality of mappings,

f+O=f=O+f.
1.4 The additive inverse of fER is given by - f where (- I) (x) =
- f(x)(x E X). This is well-defined since on the right-hand side we have the
negative of a real number. We must verify that - f + f = 0 = f + (-I).
Now for x E X
(- f + f)(x) = (- f)(x) + f(x) (definition)
= - f(x) + f(x) (definition of - f)
=0
=O(x) (definition of 0 E R).
Thus
-f+f=O.
Similarly f + (- f) = O.
1.5 Let f,g E R. Then for x E X
(f + g) (x) = f(x) + g(x) (definition)

= g(x) + f(x) (commutativity of addition in IR)
= (g + I) (x) (definition).
Thus
f+g=g+f
and so addition in R is commutative.
2.1 By definition R is closed under multiplication.
2.2 Let I, g, hER. We want to show that (f.g).h = f.(g.h). Now for x E X
[(f.g).h](x) = (f.g)(x)h(x) (definition)
= [/(x)g(x)]h(x) (definition)
= I(x)[g(x)h(x)] (associativity in lR)
= I(x)(g.h)(x) (definition)
= [f.(g.h)](x) (definition).
Thus
(f.g).h= f.(g.h).
3.1 Let I, g, hER. We prove that f.(g + h) = f.g + f.h. Let x E X. Then
[f.(g + h)](x) = I(x)(g + h)(x) (definition)
= I(x)[g(x) + h(x)] (definition)
= I(x)g(x) + I(x)h(x) (distributivity in lR)
= (f.g)(x) + (f.h)(x) (definition)
= [(f.g) + (f.h)](x) (definition).
Thus
f.(g + h) = f.g + I·h.
The other distributivity axiom is proved similarly. We have now shown
that R is a ring.
We may remark, partly as an aside, that as the ring of integers Z is an example

of a system that satisfies the axioms of a ring then these axioms are necessarily
self-consistent. We shall not concern ourselves unduly in determining to what
extent the axioms of this, or of any other, system are independent of one another.
We have not, however, adequately considered all of the distinctive features of
the ring of integers and of the ring of polynomials. In Z there are three properties
which are so familiar that ordinarily we do not make specific reference to them:
there is a number 1 such that la = al = a for all a E Z, the order in which we
multiply integers is irrelevant (ab = ba for all a, b E Z) and the product of two
non-zero integers is non-zero. We formalize these concepts as follows.
Definition 5
Let R be a ring.
1. An element 1 E R such that la = al = a for all a E R is called an identity
element or identity or unity of R.
2. Let a and b be elements of R. Then a and bare said to commute if the products
ab and ba are equal. If any two elements of R commute then R is said to be
commutative.
3. Let 0 be the zero element of R. Let a and b be elements of R such that ab = O.
Then a and b are called divisors ofzero and if both a =F 0 and b =F 0 then a and
b are called proper divisors of zero.
We have previously shown above that the zero element of a ring is unique. By
an entirely similar argument we may show that if a ring has an identity element
(which need not be the case) then it has a unique identity element.
Definition 6
A commutative ring with an identity 1(1 =F 0) and no proper divisors of zero is
called an integral domain.
(Note that the use here of the word 'domain' in the phrase 'integral domain'
has no connection with the 'domain' we encountered previously in the mapping
of one set into another. Note also that some texts, especially of an advanced
nature, do not insist upon 'commutativity' as a necessary condition for an
integral domain.)
Example 9
Under the usual operations of addition and multiplication the following are all
integral domains,
Z,Q,IR,C.
As part of this introduction to rings we require to prove some elementary
results, results which in themselves are obvious for the integers but which require
proof in an axiomatic context.
Theorem 3
Let R be a ring. Then the following hold.
1. If a + b = a + c (a, b, c E R) then b = c.
If a + b = c + b (a, b, c E R) then a = c.
2. For all a E R, aO = Oa = O.
3. For all a, bE R, a( -b) = (-a)b = -(ab), (-a)( -b) = abo

If R has an identity 1 then (-1)(-1) = 1.
4. For all a, bE R, a(b - c) = ab - ac, (a - b)e = ac - be.
Proof
1. From a+b=a+c we have (-a)+(a+b) = (-a) + (a+c). Byassoci-
ativity
((-a) + a) + b = ((-a) + a) + e
and so
O+b=O+eandb=e.
The second assertion is similarly proved.
2. aO + aO = a(O + 0) = aO = aO + 0 and so aO = O. Similarly Oa = o.
3. We have 0 = aO = a(b+ (-b)) = ab+a(-b) and so ab+ -(ab) = 0 =
ab + a(-b) from which -(ab) = a( -b). Similarly we prove -(ab) = (-a)b.
Let a' = -a. Then, as we observed in the Commentary above,
-a' = -(-a) = a. Hence
(-a)(-b) = a'(-b) = (-a')b = abo
If 1 E R then (-1)(-1) = 11 = 1.
4. a(b - e) = a(b + (-c)) = ab + a(-e) = ab + -(ae) = ab - ac. Similarly
(a - b)e = ac - be. 0
Example 10
Let D be an integral domain. Let a, b, c E C, C:f:. 0, be such that ac = be. We
want to prove that a = b. We have (a - b)e = ac - be = 0 and since D is an inte-
gral domain either a - b = 0 or c = O. But c :f:. 0 and so a - b = 0 and a = b.
We have considered polynomial rings in x in which the polynomials have

coefficients from Z, Q or R, yielding Z[x] , Q[x] or R[xj respectively. But similarly
we may have a polynomial ring in x in which the coefficients belong to an integral
domain D yielding D[x]. We have the following result.
Theorem 4
Let D be an integral domain. Then the polynomial ring D[x] is an integral
domain.
Proof
Certainly D[x] is a commutative ring with the identity of D as the identity of
D[x]. Thus we have to show that D[x] has no proper divisors of zero. Let
f(x) = ao + alx + + amxm (am:F 0),
g(x) = bo + b1x + + bnxn (bn:F 0),
be two non-zero polynomials in D[x]. Then f(x)g(x) is a polynomial of highest
term ambnx m+ n. But since D is an integral domain ambn:F O. Thus
f(x)g(x) :F 0 and D[x] is an integral domain. 0
Corollary
Let f(x) and g(x) be non-zero polynomials in D[x]. Then
deg f(x)g(x) = deg f(x) + deg g(x).
In many situations we have to consider a ring which is itself contained within a
larger ring, for example Z is a ring within the ring Q and Q is a ring within the
ring JR. We sometimes have to consider subsets of a ring which may be rings, or
'subrings' as they will be called, with respect to the operations of the ring.
Definition 7
Let S be a non-empty subset of the ring R which is also a ring under the addition
and multiplication in R. Then S is called a subring of R.
This definition is too descriptive to be really useful. We therefore derive a

criterion which is more readily applicable.
Theorem 5 Subring Criterion

A non-empty subset S of a ring R is a subring of R if and only if the following
axioms of a ring are satisfied.
1.1 S is closed under addition.
1.4 For all a E S, -a E S.
2.1 S is closed under multiplication.
Proof
Certainly if S is a subring the axioms quoted above must be satisfied.
Suppose now that the axioms above are satisfied by S. Then certainly we have
0= a + (-a) E S. We claim that the remaining axioms for a ring are auto-
matically satisfied in S because they are satisfied in R. Consider, for the sake
of argument, the associativity of addition. We have (a + b) + c = a + (b + c)
for all a, b, c E R and therefore for all a, b, c E SasS ~ R. The other remaining
axioms for a ring are similarly satisfied since S ~ R. Hence we infer that S is
a subring. 0
Example 11
It may be shown that R = M 2 (Q), the set of 2 x 2 matrices over Q, is a ring
under the usual matrix operations. Suppose we want to show that
is a subring of R. We apply the subring criterion. We have, for x, y, z, u, v, w E Q,
V) -_ (x + y,
(~ :)+(: w 0
Y+ r )
z+w
ES,
_( ~ :) = (~x -Y)
-z
ES,
(~ :) (: :) = (xou
XV+YW)
zw
ES.
where we are using the closure etc. of addition and multiplication in Q to ensure
t h at, £or exampIe, ( x+Y Y+V).18 In e In S .
'ded'
o z+w
We conclude that S is a subring of R.
Exercises 3.3
1. Prove that if a ring has an identity then it has only one identity.
2. Verify that Z, Q, lR. and C with the usual operations of addition and
multiplication are integral domains.
3. Show that the set R of one element 0, R = {O}, is a ring if addition and
multiplication are given by 0 + 0 = 0 and 00 = 0 respectively.
4. Let R be a set of two elements, 0 and 1, R = {O, I}, in which addition
and multiplication are given by the following tables:
TtM
um 0 1 Product 0 1
o 0 1 o 0 0
1 1 0 101
Prove that R is an integral domain.

5. Let 2Z denote the set of even integers with the usual operations of
addition and multiplication. Prove that 2Z is a ring. Is 2Z an integral
domain?
6. Let R be a ring.
(i) Let a, b, c E R be such that a + b = c + b. Prove that a = c.
(ii) Let a,b E R. Prove that -(ab) = (-a)b.
7. Let M 2 (Q) be the set of 2 x 2 matrices over Q. Under the usual opera-
tions prove that M2 (Q) is a ring.
8. Let R 1 and ~ be subrings of a ring R. Prove that R 1 n ~ is a subring
of R. Can you generalize this result?
4
Introduction to Groups
In Chapter 3, we began to consider an axiomatic treatment of certain algebraic

concepts and operations; in particular in the case of a ring we needed to consider
two binary operations. Since simplicity usually has some advantages we shall in
this chapter consider only one binary operation. We begin therefore with the
notion of a 'semigroup' from which we shall be led to consider a 'group', one of
the most fundamental structures in modern mathematics.
The evolution of the concept of an abstract group owes much to the labours of
many mathematicians of whom only a few will be mentioned here. The origins of
the concept may be traced from the workofP. Ruffini (1765-1822) and E. Galois
(1811-32) through to that ofL. Kronecker who developed ideas for what we now
call an Abelian group ('Abelian' after N.H. Abel, 1802-29). The abstract concept
of a finite group was first formulated in 1854 by A. Cayley (1821-95) but its sig-
nificance was not properly appreciated until 1878. W. von Dyck (1856-1934) and
H. Weber (1842-1913) were influential in the development of group theory, the
latter giving the first definition of an infinite group in 1893. These remarks can
give at most a brief indication of the history of the evolution of the concept of an
abstract group which "emerged not from a single act or the creation of a single
scholar but was rather the outcome of a process of abstraction with certain dis-
cernible steps. These were not essentially logical in nature; rather they repre-
sented the variable extent to which the process of abstraction was carried
through." (quotation from The Genesis of the Abstract Group Concept by
H. Wussing, translated MIT Press, 1984, page 234). It is perhaps psychologically
salutary to reflect that the concepts of a group and, indeed, of other parts of
mathematics, have not always evolved directly but frequently by fits and starts.
93
4.1 Semigroups
Definition 1
Let S be a non-empty set on which there is defined a binary operation denoted by
*. For a, b E S the outcome of the operation between a and b is denoted by a * b.
Then S is called a semigroup if the following axioms hold.
(i) For all a, bE S, a * b E S (closure).

(ii) For all a, b, e E S, (a * b) * e = a * (b * e) (associativity).
Example 1
Z under the operation of + is a semigroup since for all a, b E Z, a + b E Z and for
all a, b, e E Z, (a + b) + e = a + (b + e).
In future, in general considerations of semigroups, we shall omit the symbol for

the binary operation unless an especial need arises for it to be shown. We shall
use, as far as possible, a multiplicative notation in which the outcome of the
binary operation between two elements is shown by the simple juxtaposition
of the elements; we shall speak of the multiplication (sometimes addition) of
these elements.
Semigroups are common in mathematics and easy to create as the following
examples show.
Examples 2
1. Let S be a set of one element, S = {a}, say. Define
aa=a.
Then S becomes a semigroup.
2. Let S be a non-empty set. Define a binary operation on S by
ab = b for all a, b E S.
We claim that S is a semigroup under this operation. Certainly we have for
all a, bE S, ab = bE S. We also have for all a, b, e E S, (ab)e = be = e and
a(bc) = ac = e and so (ab)e = a(bc). Thus S is a semigroup.
3. Let S be a non-empty set and let s be a fixed element of S. Define a binary
operation on S by
ab = s for all a, b E S.
Introduction to Groups 95
Then certainly for all a, bE S, ab E S and then we have for all a, b, e E S,

(ab)e= se = s and a(bc) = as = s from which (ab)e = a(bc). Thus S is a
semigroup.
None of the examples above may be of great intrinsic interest but nevertheless
in each example the axioms of a semigroup are satisfied. After the next lemma we
consider a particular type of semigroup.
Lemma 1
Let S be a semigroup. Let e, f E S be such that ex = x and xf = x for all xES.
Then e = f and so ex = xe = x for all xES.
Proof
By the condition on e we have, in particular, ef = f. Similarly ef = e. Thus
e = f and the rest is clear. 0
Definition 2
A semigroup S is called a monoid if there exists e E S such that ex = xe = x for
all xES. e is said to be the identity element or identity of S.
By Lemma 1 above a semigroup has at most one identity element.
Associativity and Index Law
Let a, b, e, d, ... be elements of a semigroup S. Expressions such as (ab)(ed) ,

a((bc)d) , a(b(ed)) are formally distinct but are in fact equal since, by
associativity,
(ab)(ed) = a(b(ed)) = a((bc)d)).

Thus we could write, without uncertainty of meaning, simply abed. Further we
see that in more complicated expressions of elements and brackets we may simi-
larly dispense with the bracketing provided the elements appear in the same
order in each of the expressions.
Consequently we may now define the powers of a E S unambiguously as
a
1
= a, a2 = aa, a3 = aaa, ... ,an = aa . .. a,
where a appears precisely n times. It follows that for m, n = 1,2, ...
afflan = (aa ... a)(aa ... a) = (aa ... a) = am + n
where within the brackets we have a appearing m times, n times and m + n times
respectively.
IT further S is a monoid with identity e we may define aO = e for all a E Sand
then
am a n = am+n ( m, n = 0, 1,....)
Examples 3
1. Let X be a non-empty set and let S(X) be the set of mappings of X into X.
We have already met the circle composition of mappings of X into X, namely
for I, 9 E S(X) we defined log by
(f 0 g)(x) = I(g(x» (x EX).
Furthermore we showed (Theorem 7, Chapter 1) that for I, g, h E S(X) then
(f 0 g) 0 h = 10 (g 0 h).
The set S(X) of mappings of X into X is therefore a semigroup under the
circle composition of mappings. S(X) has an identity element given by the
mapping £ where £(x) = x(x E X) since for all I E S(X) we have
(I, 0 I)(x) = £(f(x» = I(x) = I(£(x» = (f 0 £)(x) (x E X)
and so
£0 I = I = 10 £.
2. Let a binary operation * be introduced into Z by defining for all a, b E Z

a * b = a + b + abo
Then we claim that Z with this operation is a monoid. We certainly have
a * b E Z for all a, b E Z. Let a, b, e E Z, then
(a * b) * e = (a + b + ab) * e
= (a + b + ab) + e + (a + b + ab)e
= a + b+ ab+e+ ac+ bc+ abc
= a + b + e + ab + ae + be + abc,
a * (b * c) = a * (b + c + be)
= a + (b + c + be) + a(b + c + be)
= a+ b+ c+ bc+ab+ac+ abe
= a+ b+ c+ab+ ac+be+abe,
from which we have
Now for all a E Z we have

o* a = 0 + a + Oa = 0 + a + 0 = a,
a * 0 = a + 0 + aO = a + 0 + 0 = a.
Thus 0 plays the role of an identity element when * is the binary operation.
Hence Z is a monoid under *.
3. The set M 2 (Q) of 2 x 2 matrices with entries from Q is a monoid under the
usual matrix addition, namely
(
Xl YI) + (X2 Y2) = (Xl + X2 YI + Y2 ) .
~ ~ ~ ~ ~+~ ~+~
The sum of two 2 x 2 matrices with entries from Q is again a 2 x 2 matrix

over Q and so the operation of addition is closed. The associativity of addition
is a well known property of matrix addition. Further
is the identity element for addition and so M2 (Q) is a monoid.

4. The set M 2 (Q) of 2 x 2 matrices with entries from Q is a monoid under the
usual matrix multiplication, namely
Xl YI) + (X2 Y2) = (XIX2 + Yl~ XIY2 + Ylt2).

(
Zl tl Z2 t2 ZlX2 + tl~ ZlY2 + t l t 2
Closure is obvious and matrix multiplication is known to be associative. The
identity element for multiplication is
We conclude this section with two results which will be important in our
study of groups.
Lemma 2
Let S be a monoid with identity element e. Let a E S. Let a', a" E S be such that
a'a = e = aa". Then a' = a".
Proof
The result depends crucially on the associativity axiom. We have
(aaa " a'(")
, ) " =ea " =a, ,
aa =ae=a , and so a ' =a.
" o
Theorem 1
Let S be a semigroup with the following two conditions.
1. There exists e E S such that ec = a for all a E S.
2. For each a E S there exists a' E S such that a'a = e.
Then S is a monoid with identity e and a'a = aa' = e.
Proof
We show first that e is the identity of S. Let a E S. Then, by twice applying
condition 2, there exists a' E S such that a'a = e and there exists a" E S such
that a"a' = e.
But by condition 1 and associativity we have
ae = e(ae) = (a"a')(ae) = a"(a'(ae)) = a"((a'a)e)
=a"(ee) = a"e = a"(a'a) = (a"a')a = ea = a.
Consequently e is the identity of S. But then a'a = e = a"a' implies, by
Lemma 1, that a = a" and so we complete the proof. 0
We remark that in the two conditions ofthis theorem e 'acts' on the left of any
a E S to give ea = a and for given a E S a' 'acts' on the left to give a'a = e. IT
we were to replace these left-handed conditions by corresponding right-handed
conditions then the conclusion of the theorem would still hold. If, however,
one condition is on the left and the other condition is on the right then the con-
clusion of the theorem need not hold. To see this last assertion we consider the
following example.
Example 4
Let S be the two-element semigroup, S = {e, f} with multiplication, ee = e,
ef = f, ff = f, fe = e. S is a semigroup of the type described in Examples 2
no. 2, where e acts on the left and satisfies condition 1 of the last theorem. We
also have ee = e and fe = e so that, from the right we would have, e' = e and
!' = e. Nevertheless e is not an identify element of S.
Exercises 4.1
1. A binary operation * is defined on Z by a * b = ba (a, b E Z). Prove
that, under this operation, Z is a monoid.
2. A mapping fa,b of lR into lR is defined for a, b E lR by
fa,b : x - a + bx (x E lR).
Prove that the set of all such mappings is a monoid under the circle
composition of mappings. Hint: Prove that
fc,d 0 fa,b = fc+ad,bd'
3. Let t be a given integer. On Z an operation * is defined by
a*b=a+b+tab (a,bEZ).
Prove that Z under this operation is a monoid.
4. Let S be a semigroup with the following two conditions.
(i) There exists e E S such that ae = a for all a E S.
(ii) For all a E S there exists a' E S such that aa' = e.
Prove that e is the identify of S and that a'a = e.
5. (Hard) Let S be a finite semigroup and let a E S. By considering the
subset {am In E N} prove that there exist p, q EN, 2q < p such that
aP = a q . IT b = aP - q prove that b2 = b.
4.2 Finite and Infinite Groups

Definition 3
Let G be a non-empty set on which there is defined a binary operation so that the
outcome of the operation between a and b (a, bEG) is denoted by abo Then G is
called a group if the following axioms hold.
1. For all a, bEG, ab E G (closure).

2. For all a, b, c E G, (ab)c = a(bc) (associativity).
3. There exists e E G such that for all a E G, ea = a (existence of identity).
4. For each a E G there exists an element denoted by a-l such that a-la = e
(existence of inverse).
By Theorem 1 e is the identity of G and, for each a E G, we have a-I uniquely
determined by a and such that, a-1a = aa- l = e, a-l is called the inverse of a.
Notice that from a-la = aa- l = e we may deduce that

(a-l)-l = a (a E G).
Partly for later convenience we give two alternative criteria by which a semi-
group is a group.
Theorem 2
Let S be a semigroup. Suppose that for any a, b E S there exist x, yES such that
ax = b and ya = b. Then S is a group.
Proof
Let a E S. By assumption there exists e E S such that ea = a. We cannot imme-
diately assert that e is the identity for S since, on the face of it, this e depends on
the particular element a. We consider an arbitrary c E S. Then again by assump-
tion there exists u E S such that au = c. Hence ec = e(au) = (ea)u = au = c.
Furthermore by assumption there exists c' E S such that c'c = e. Hence,
finally, S is a group. 0
Theorem 3
Let S be a finite semigroup in which cancellation exists, that is if a, b, xES and
ax = bx then a = b and if a, b, yES and ya = yb then a = b. Then S is a finite
group.
Proof
Let a, bE S. We show that there exists xES such that ax = b. Let Shave n dis-
tinct elements Sll S2, . .. ,Sn. Consider aSll aS2, . .. ,asn . These elements of S are
also distinct since aSi = aSj implies Si = Sj and so i = j. Thus
{asll aS2,··· asn } = {Sll S2, ... ,sn}·
Hence, as b is one of the SI, S2, ... , Sn, there exists Sk such that aSk = b.
Similarly we may find yES such that ya = b. Thus, by the previous result,
S is a group. 0
Since the group concept is widely encountered in mathematics we shall give a

varied selection of examples, not all of which need to be fully comprehended on a
first reading. For convenience in this chapter, and elsewhere, we shall use G to
denote a group and e to denote the identity of G without further explanation.
Examples 5
1. The set of non-zero elements ofQ, that is Q \ {O}, is a group under the usual
multiplication. To verify that Q \ {O} is a group we note the following
immediate facts. If a, bE Q \ {O} then ab E Q \ {O}. Multiplication in Q,
and so in Q \ {O}, is associative. 1 is the identity as 1a = a1 = a for all
1
a E Q \ {O}. If finally a E Q \ {O} then a has a reciprocal - such that
a
(~)a = a(~) = 1 and so a-I =~.
In a similar manner we may prove that IR \ {O} and C \ {O} are groups.
2. The set of integers Z under multiplication is not a group. We know that this
set is a monoid under multiplication but, for example, the number 2 does not
have an inverse in Z as ~ E Z.
3. The set of integers Z is a group under the usual addition. Here we have to
insert the specific symbol + for the binary operation but the group axioms
are easily verified. Thus if a, b E Z then a + b E Z and for all a, b, c E Z
(a + b) + c = a + (b + c).
The element 0 plays the role of the identity element for addition since
O+a=a+O=a
for all a E Z and the inverse of a E Z with respect to addition is -a, since
(-a) + a = a + (-a) = O.
We now gather together some elementary results on groups.
Theorem 4
Let G be a group.
1. Cancellation
Let a, b, x E G be such that ax = n. Then a = b.
(Similarly ya = yb (y E G) implies a = b.)
2. Unique Solution of Equation

Let a, bEG. Then the equation ax = b has the unique solution x = a-lb.
(Similarly ya = b has the unique solution y = ba- l .)
Proof
1. There exists X-I E G such that xx- l = e. Then (ax)x- l = (bx)x- l and so
a(xx- l ) = b(xx- l ) from which ae = be and a = b.
2. Certainly a-Ib is a solution since a(a-Ib) = (aa-l)b = eb = b.
On the other hand, ax = b implies that a-l(ax) = a-Ib from which we con-
clude that (a-Ia)x = a-Ib and so x = ex = a-lb. 0
Definition 4
A group G is said to be finite if the set G is a finite set, otherwise the group is said
to be infinite. If G is finite the cardinality of G (which is the number of elements
in G) is called the order of G, written IGI, G then has finite order. An infinite
group is often said to have infinite order.
Example 6
Z under addition and Q \ {OJ under multiplication are examples of groups of
infinite order.
In the case of a finite group of very small order we may conveniently display
the binary operation in the tabular form of a so-called 'Cayley table'. Thus if
G has order n, G = {al,a2'" .,an } say, then we write the table as
al a2 aj an
al alaI ala2 alaj alan

a2 a2 a l a2 a 2 a2 a j a2 a n
a, a,al a,a2 aiaj aian
an anal a na2 anaj ana n
where the entry in the (i,j)th position is a,aj' By the cancellation result in
Theorem 4 the elements a,al, a,a2, ... ,a,an are distinct and so any row of the
Cayley table must have distinct elements in the row. Similarly any column of
the table must consist of distinct elements.
Examples 7
1. Let G have order 1. Then G = {e} and the Cayley table is
As a concrete example of this group we have {I} as a multiplicative group.

2. Let G have order 2. Then G = {e, a}, say. Now a2 = a or e and a2 = a implies
aa = a 2 = ae and so a = e which is false. Thus a 2 = e. The Cayley table is
:t:
a
e e a
a a e
As a concrete example we have {I, -I} under multiplication.

3. Let G have order 3, say, G = {e,a,b}. Consider abo If ab = a or ab = b we
have, by cancellation, that b = e or a = e respectively and either inference
is false. Thus ab = e. Hence b = a-I and so also ba = e. We obtain the par-
tially completed Cayley table
e a b
e e a b
a a e
b b e
But each row and column must contain distinct elements and so the Cayley
table is
e a b
e e a b
a a b e
b b e a
The set of matrices
D}
0 I 0
{G n·G D·G
I
0
0
0
0
I
forms a group with the same table if we put
e=
100) ,a= (0001
010
(001
1 0) ,b= (01 oo 1)
0
100 0 1 0
4. Let G have order 4. Then G = {e, a, b, e} I say. We must distinguish two cases.
(i) Suppose there exists an element in G which is not its own inverse and let a
be such an element. Thus a-I =F a. For the sake of argument suppose
a-I = b. Thus ab = ba = e. But then ae =F a and ae =F e, as otherwise
e = e or a = e respectively, hence we must have ae = b. Similarly
ca =F e, a, e and so ca = b. We now obtain the incomplete Cayley table
e a b e
e e a b e
a a e b
b b e
e e b
But the elements in the second row must necessarily be distinct and so
we have a 2 = e. Also ac = b implies bae = b2 and so, as ba = e, we obtain
e = b2 • Thus maintaining the distinctiveness of the elements in any given
row or column we obtain the Cayley table
e a b e
e e a b e
a a e e b
b b e e a
e e b a e
A possible realization is the group {I, i, -1, - i}.
(ii) Suppose now that every element is its own inverse. Then a 2 = b2 =
e2 = e. Then ab =F e, a, b and so ab = e, also ba =F e, a, b and so ba = e.
Similarly we deduce that ae = ca = b and be = cb = a. The Cayley
table is
e a b e
e e a b e
a a e e b
b b e e a
e e b a e
This abstract group is known as the Klein foW'-group (after F. Klein,

1849-1925). A realization is afforded by a group offour matrices, namely
Theorem 5
Let a,b E G. Then
Proof
Inserting brackets for clarity we have
(b- 1a- 1 )(ab) = b- 1 (a- 1 (ab» = b- 1 «a- 1a)b) = b- 1 (eb) = b- 1 (b) = e.
Hence
o
Corollary
L etalla2,···,anE G . Then (ala2 ... an ) -1 =an-1 an-l
-1
... al-1 .
Proof
A simple induction suffices. o
We now wish to extend the index law for powers of a E G, namely
ama n = am+ n (m, n = 0, 1, ...)
to the case in which m and n may be negative. First we have to define a- n
(n > 0).
Definition 5
Let a E G. Let n E N. Then a- n is defined by
a- n = (a-It
Using this definition, which is entirely natural if we think of reciprocals of
rational numbers, we may extend the index law above. We give examples
which show how the extension is used.
Example 8
Let a E G.
= (aaa)(a-1a- l ) = aa(aa-l)a- l
(a 3)(a- I )2
= aaea- l = aaa- l = a(aa- l ) = ae = a.
(a- 3)(a)2 = (a-1a-1a-l)(aa) = a-Ia-l(a-Ia)a = a-Ia-l(e)a
= a-Ia-Ia = a-l(a-Ia) = a-Ie = a-I.
Theorem 6 Index Law for Groups

Let a E G. Then
Proof (may be omitted on a first reading)

We may clearly confine our argument to the cases in which one, or both, of m and
n is negative. We shall only consider here the case in which m > 0 and n < O. Let
n = -p where p > O. Then
aman = ama-P = am(a-I)P = (a ... a)(a- l ... a-I)
where in the first bracket we have m a's and in the second bracket we have
p a-l,s. But
(a ... a)(a-l ... a- l ) = (a a)(aa-l)(a-l ... a- l )
= (a a)e(a- l a- l )
= (a a)(a-l a- l )
where on the right-hand side we have in the first bracket m - 1 a's and in
the second bracket p - 1 a-l,s. Continuing in this way to replace aa- l bye we
eventually have
a ... a m-pa's (p < m)
aman = e (m=p)
{
(m < p).
-1 -1
a ... a
Thus for p < m we have

for m = p we have
and for m < p we have
a ma n = (-l)p-m
a =a-(p-m)
Example 9
Let x, y, z E G and suppose we are required to simplify
g = (YX)-1(yx- 1)2(xyz)(X-3y 2Z)-1.
we have
g = (X-ly-l)(yx-lyx-l)(xYZ)(X3y2Z)-1 = X- 1(y-ly)x- 1Y(X- 1X)y(zZ-1)y-2 x 3
= x- 1x- 1yyy-2 x 3 = x- 2y2y -2 x 3 = x- 2x 3 = x.
Definition 6
Two elements a, bEG are said to commute if ab = 00. If any two elements of G
commute then G is said to be commutative or Abelian, otherwise G is said to be
non-commutative or non-Abelian.
If the operation in a group is written as multiplication then the group mayor

may not be Abelian. If however an additive notation, with +, is employed then it
is usual to presume that the operation is commutative.
Example 10
We prove that if a 2 = e for all a E G then G is necessarily Abelian. Byassump-
tion for all a, bEG
(ab)2 = e.
Thus
We now give several further examples of groups.

Examples 11
1. Let 0(2) consist of all matrices of the form To where
To=(COSO SinO) (OER).

-sinO cosO
We claim that under matrix multiplication 0(2) is a group. We require some

elementary trigonometry.
We have
cosO SinO) ( cos<p sin<p )

ToT", = (
-sinO cosO -sin<p cos<p
cosOcos<p - sinOsin<p o
cos sin <p + sinOcos<p )
= ( - sin ocos <p - cos osin <p - sin 0 sin <p + cos 0 cos <p
= ( cos(O + <p) sin(O + <p) ).

- sin(O + <p) - cos(O + <p)
This establishes closure and associativity is immediate.
To = cosO
( . -smO
SinO) =
cosO
(1 0)
0 1
is the identity.
Since
LoTo = LO+IJ = To = (~ ~ ).
(TO)-l =L o·
Thus 0(2) is a group which is infinite and Abelian since
ToT", = To+", = T",+o = T",To.

2. A mapping fa,b of R into R is defined for a, b E R by
fa,b : x -+ a + bx (x E R).
Then, 88 in an Exercises 4.1, no. 2
fc,d 0 fa,b = fc+ad,bd (a, b, c, dE R).

Let T be the set of such mappings fa,b for which b:f:. O. We claim that T is a
group. Certainly we have closure since if b:f:. 0, d:f:. 0 then bd:f:. 0 and so
fc,d 0 fa,b = fc+ad,bd E T.

The circle composition of mappings is associative and fO,l is such that

fO,l 0 fa,b = fa,b'
For given a, b E JR, b '" 0 we require to find c, d E JR, d '" 0 such that
fc,d 0 fa,b = fO,l'
But this is only possible if
c + ad = 0, bd = 1.
However we do have a unique solution, namely
We therefore conclude that T is a group as claimed.

3. Let X = JR \ {O, I}. Six mappings, e,a,b,c,d,j, are defined on X as follows.
x-I 1
e(x) = x, a(x) = - - ,
x
b(x) =1_ x c(x) =1- x,
x 1
d(x) =--1' f(x) =-.
x- x
It is easy to verify that these mappings are defined on X arid fairly easy to
verify that they map X into X. We claim that these six mappings Wlder
the circle composition of mappings form a group of order 6. Closure is not
obvious arid must be verified somewhat tediously since there are 36 possible
products. Associativity, on the other harId, is clear since the circle composi-
tion of mappings is associative. e must play the role of identity. The easiest
way to verify closure etc. is to construct a Cayley table. By way of illustration
we shall evaluate three products, leaving the reader to contemplate arid per-
haps to evaluate some at least of the remaining 33..
(a 0
1) --1 1 1
f)(x) = a(f(x)) = a (; = x.!. = ~ x = 1- x = c(x),
x
(f 0 a)(x) = f(a(x)) = f(x: 1) = x ~ 1 d(x), =
(b 0 d)(x) = b(d(x)) = b (_X_) = 1 x = x-I = 1- x

x-I 1 - - - x-l-x
x-I
= c(x).
Then a 0 f = c, f 0 a = d, bod = c. Eventually we obtain the Cayley table.

e a b c d f
e e a b c d f
a a b e d f c
b b e a f c d
c c f d e b a
d d c f a e b
f f d c b a e
We shall later have occasion to refer to this table. In any subsequent discus-
sion we shall omit 0 and write simply, for example, af = c, fa = d, bd = c, etc.
4. Let X = {(x,f)lx,t E 1R}. For each v E IR we define a mapping Tv where
Tv:X-+Xby
Tv(x, t) = (x - vt, t).
We claim that the set S of such mappings is a group. For u, v E IR we have
(Tu 0 Tv)(x, t) = Tu(Tv(x, t»
= Tu(x - vt, t)
= (x - vt - ut, t)
= (x- (u+v)t,t)
= Tu+v(x, t).
Thus
Tu oTv = T u+v
from which we obtain closure. The associativity of mappings is assured and To
is evidently the identity. Further
T- v 0 Tv = T- v+v = To·
and so
T- v = (Tv)-l.
S is a group and is Abelian since
Tu 0 Tv = Tu+v = Tv+u = Tv 0 Tu.
We remark that this example has its origins in the physical world. If we regard
(x, t) as the coordinates of space and time and v as the speed then S is the group
of Galilean transformations of classical mechanics (after G. Galilei, 1564-1642).
(See also Exercises 4.2, no. 10.)
Definition 7
The group G is said to be cyclic if every element of G is a power of some given
element of G. This given element is said to generate, or to be a generator of,
the group G.
Thus if G is cyclic we may write

G = {an: n = 0, 1, ...}
for some a E G. A cyclic group is necessarily Abelian.
Example 12
The groups of orders 1, 2, 3, 4, with Cayley tables as given above, are all cyclic
with the exception of the Klein four-group. In the case of the group of order 3 we
may put b = a2 and in the case of the cyclic group of order 4 we may put b = a3 ,
c= a2 •
We now come to a usage of the word 'order' which differs from its previous
usage but which, as we shall later Bee, is related to that previous usage.
Definition 8
Let a E G. Then a is said to have finite order if for some n E N, an = e. IT n = 1
then a = e and e is said to have order 1. IT for 0 < m < n, am # e then a is said to
have order n. IT a does not have finite order then a is said to have infinite order.
In a finite group all elements have finite order but if all elements of a group
have finite order it does not necessarily follow that the group is finite.
Examples 13
1. The additive group Z has one element, namely 0, of order 1 and all other ele-
ments are of infinite order.
2. The multiplicative group Q \ {O} has one element of order 1, namely 1, and
one element of order 2, namely -1. All other elements have infinite order.
The assertions are true of the group lR \ {O}.
3. The multiplicative group C \ {O} has no real elements of finite order other
than ±l but there are complex elements of finite order. We may identify
these elements by use of De Moivre's Theorem (after A. De Moivre,
1667-1754).
Let z E C \ {O} and let z = r(c080 + i sin 0) (r,O E JR., r> 0). Then we
have, for n E N,
~ = rn(cosO + i sin 0t = rn(cosnO + i sin nO).
Hence ~ = 1 if and only if
rn c08nO = 1, rnsinnO = O.
This holds if and only if
r n = l,c08nO = 1, sin nO = 0,
or
r = 1 and nO = 2k1r (k E Z).
Thus ~ = 1 if and only if
2k1r .. 2k1r
Z=C08-+lsm-.
n n
4. Let G be the group of order 6 with Cayley table as in Examples 11 no. 3. Then
a2 = b, a3 = ab = e, b2 = a, b3 = ab = e, c2 = d2 = /2 = e.
Thus e has order 1, a and b have order 3, and c, d and / have order 2.
Definition 9
Let a, bEG. Then a is said to be col\iugate to b if there exists x E G such that
x-lax = b.
We have therefore defined a relation of conjugacy on a group G and we should

therefore not be surprised to obtain the next theorem.
Theorem 7
The relation of conjugacy on a group G is an equivalence relation.
Proof
We have to prove reflexivity, symmetry and transitivity for the given relation.
1. Let a E G. Then e-Iae = a and so a is conjugate to a for all a E G.
2. Let a, bEG and suppose a is conjugate to b. Then there exists x E G such
that x-lax = b. Let Y = X-I, then y-l = (X-l)-l = x. We have
y-Ix-Iaxy = y-Iby
and so
But xy = e and hence

a = e-lae = y-lby.
Consequently b is conjugate to a.
3. Let a, b, c E G and suppose a is conjugate to b and b is conjugate to c. Then
there exist x, y E G such that
X ax = , y ay = c.
-1 b -1
Let z = xy. Then

z-laz = (xy)-la(xy) = y-lx-laxy = y-lby = c.
Hence a is conjugate to c. This completes the proof of conjugacy. 0
Definition 10
Under the relation of conjugacy in a group G the equivalence classes are called
the conjugacy classes.
A typical conjugacy class is {x -1 ax : x E G} which is the conjugacy class con-

taining a. In an Abelian group G each conjugacy class consists of a single element
since ax = xa (a, x E G) implies x-lax = a.
Example 14
Suppose we try to find the conjugacy classes in our favourite group of six ele-
ments with Cayley table as given above. One class is {e}. The class containing
a is {e-lae,a-laa, b-lab, c-lac,d-lad, I-laf} which is, in fact, {a, a, a, b, b,b} =
{a,b}. The class containing c ¢ {e} U {a,b} is {e-lce,a-lca,b-lcb,c-lcc,
d- l cd, 1-1 cf} which is {c, d, I, c, I, d} = {c, d, f}. The group is therefore
partitioned into the three conjugacy classes, {e}, {a, b} and {c, d, f}.
Definition 11
The element a in G is said to be self-conjugate or central if x-lax = a for all
x E G. The set of central elements is called the centre of G and is denoted by
Z(G).
Notice that e E Z(G) and that Z(G) = G if and only if G is Abelian.

Exercises 4.2
1. Let S = {O} UN. Under the operation of addition is S a group?
2. Prove that the identity e of a group G is the only element of G satisfy-
ing the equation x 2 = x (x E G).
3. Let a, b, c, dE G. Simplify
(ab 2)-1 (c 2a-1)-1 (c2b2d) (ad) -2 a , (abc) -1 (ab)2d(d- 1b-1 )2bdc.
4. It is asserted that a group G of order 5, G = {e, a, b, c, d}, has the
following Cayley table
e a b c d
e e a b c d
a a c d e b
b b d e a c
c c b a d e
d d e c b a
Is the assertion correct?

5. In forming the Cayley table of the non-Abelian group of order 6
(Examples 11 no. 3), only three of the 36 products were actually
evaluated. Verify the following equations (note omission of 0).
ab = e, ac=d, bc= I, bl=d,
00= I, cc=e, cl=a, db= I,
fb=c, II=e.
2k1r .. 2k1r }
6. LetS= { COST+1SlnTlkEZ .
Prove that S is a multiplicative group of order 3.
7. Prove that the set of complex numbers
2k1r .. 2k1r I k
S = { cos-;- + 1 sm -;- E '71}
aJ
is a cyclic group of order n.

8. A mapping I: [0,1]- [0,1], where [0, 1] is the closed interval of ana-
lysis {x 10 :5 x :5 I}, is said to be strictly monotonic if the inequalities
° :5 Xl < X2 :5 1 imply that I(X1) < I(X2)' Under the circle composi-
tion of mappings prove that the set of strictly monotonic mappings of
[0, 1] into [0, 1] is a group.
9. Prove that the set G of matrices of the form
is a group under matrix multiplication. What is the centre of this

group?
10. (Hard) Like an example above, this problem originates in the physi-
cal world. Let c be a given strictly positive constant.
(i) Let u,v E R be such that lui < c, Ivl < c.
Prove that Iw I < c where
u+v
w = ------UV.
1 + c?
(ii) Let X = {(x, t) lx, t E 1R}.

For each v E IR we define a mapping Tv : X -+ X by
Prove that
Deduce that the set S of such mappings is a group. This group is in

fact the Lorentz group of transformations of relativistic mechanics
(after H.A. Lorentz, 1853-1928).
4.3 Subgroups
We have explored some immediate consequences of the definition of the group G.
We now wish to look more closely at the structure of G and we begin by exam-
ining various sub-structures.
Definition 12
Let H be a non-empty subset of the group G which is also a group under the
multiplication in G. Then H is called a subgroup of G.
As for subrings, this definition is too descriptive to be useful. We give a

convenient criterion.
Theorem 8 Subgroup Criterion

A non-empty subset H of the group G is a subgroup of G if and only if for all
a,b E H, ab E H and a-I E H.
Proof
Suppose H is a subgroup of G. Then H is closed under the multiplication in G
and so for all a, b E H, ab E H. Now H has an identity I, say, and then
12 = I = Ie and so e = I E H. Let a E H. Then a has an inverse a', say, in H
such that a'a = e. But a-I exists in G such that a-Ia = e. Thus a-I = a' E H.
Conversely if for all a, bE H, ab E H and a-I E H then H is evidently closed
under the multiplication in G. This multiplication is associative in G and so in H.
Let a E H. Then a-I E H andsoe = aa- I E H. Fora E Hwehavea- l E Hand
so, finally, H is a subgroup. D
Two obvious subgroups of G are G itself and {e}.
Definition 13
The subgroup H of the group G is said to be a proper subgroup if { e} f; H and
Hf;G.
Examples 15
1. Let a E G. Then H = {an In E Z} is a subgroup of G. To see this observe that
typical elements of H are am and an. Then
aman = am+ n and (am)-l = a- m
and so H is a subgroup of G.
2. Under the usual multiplication, let G be the group Q \ {O} and let H be the
subset of G given as {u2 1U E Q \ {O}}.
Then H is certainly a non-empty subset of G. Furthermore if a, bE H we

have a = x 2, b = y2 for some x, y E Q \ {O}. Then
ab = x 2y2 = (xy)2 and a-I = (X2)-1 = x- 2 = (x- 1)2
Since (xy)2 E Hand (x- I )2 E H we may conclude that H is a subgroup.
3. The cyclic group G of order 4, G = {e, a, b, c} with Cayley table
e a b c
e e a b c
a a c e b
b b e c a
c c b a e
has one proper subgroup, namely {e, c}.

4. Under matrix multiplication the set G of matrices of the form
a b)
( -b (a, bE JR, a2 + b2 # 0)
a
forms a group. The four matrices
(~~).(~1 ~1).(~1 ~).(~ ~1)

form a subgroup of G.
Theorem 9
Let Hand K be subgroups of the group G. Then H n K is a subgroup of G.
Proof
We should observe first of all that H n K is indeed non-empty since e E H and
e E K implies that e E Hn K. Let a,b E Hn K. Then a,b E H and so, as H is a
subgroup, ab E H and a-I E H. Similarly ab E K and a-I E K. Thus
ab E H n K and a-I E H n K from which we conclude that H n K is a sub-
groop. 0
A simple adaptation of the proof above shows that the intersection of any col-
lection of subgroups is again a subgroup. However, we do not have what might be
thought to be the analogous result for the union of two subgroups; we may note
the following result in passing.
Theorem 10
Let Hand K be subgroups of the group G. Suppose H U K is a subgroup of G.
Then either H ~ K or K ~ H.
Proof
For the sake of argument suppose that H is not a subset of K. Then there exists
h E H, h ¢ K. We shall prove that K ~ H.
Let a E K. Then a,h E HU K and, as HuK is a subgroup, ah E Hu K. We
then have two possible alternatives, either ah E H or ah E K. If ah E K and
since a- 1 E K we would have h = a-1(ah) E K which is contrary to our choice
of h. Thus ah E H and since h E H we have h- 1 E H and so a = (ah)h- 1 E H.
ThusK~H. 0
Example 16
Let D 4 = {e, a, b, c, d, I, g, h} be the group of order 8 with Cayley table below.
The group is called the dihedral group of order 8.
e a b C d I 9 h
e e a b c d I 9 h
a a b c e I 9 h d
b b c e a 9 h d I
c c e a b h d I 9
d d h 9 I e c b a
I I d h 9 a e c b
9 9 I d h b a e c
h h 9 I d c b a e
G and {e} are, of course, subgroups. The remaining subgroups are {e, b},
{e,d}, {e,g}, {e,j}, {e,h}, {e,a,b,c}, {e,b,d,g} and {e,b,l,h}; of these sub-
groups we may verify that the first six are cyclic and the last two are, struc-
turally, Klein four-groups.
We now come to a particular and important type of subgroup.

Lemma 3
Let a be an element of the group G. Let the subset {x E Glx-Iax = a} be
denoted by Ga(a). Then Ga(a) is a subgroup of G containing a.
Proof
We observe first that Ga(a) is non-empty since a-Iaa = a and so a E Ga(a). Let
x, y E Ga(a). Then x-lax = a and y-Iay = a. We have
(xy)-la(xy) = y-Ix-Iaxy = y-Iay = a and so xy E Ga(a).
Also from x-lax = a we have
a = xx-Iaxx- l = xax- l = (x-l)-la(x- l ), from which X-I E Ga(a).
Thus Ga(a) is a subgroup. 0
Definition 14
Leta be an element of the group G. Then Ga(a) = {x E Glx-Iax = a} is a sub-
group called the centralizer of a in G.
Since x-lax = a if and only if ax = xa, Ga(a) is the subset of G consisting of

all elements 'commuting' with a.
Theorem 11
The centre Z(G) of the group G is a subgroup of G.
Proof
We may follow the argument of Lemma 3. We have
Z(G) = {x E Glx-Iax = a for all a E G}.
Thus if x, y E Z(G) then x, y E Ga(a) for all a E G and so xy, x-IGa(a) for all
a E G. Thus xy, x-I E Z(G) and so Z(G) is a subgroup. 0
Definition 15
Let X and Y be non-empty subsets of the group G. Let XY be the subset given
by
XY = {xylx E X, y E Y}.
Similarly XYZ, written without brackets as multiplication is associative, denotes

the subset given by
X1'Z = {xyzlx EX, y E Y, z E Z}.
If, in particular, X and Z are each subsets of a single element, X = {x} and
Z = {z}, say, then xY is to be written for {x}Y and xYz is to be written for
{x}Y{z}. Thus
xY = {xyly E Y}, Yz = {yzly E Y}.
We note that if A, B, X, Y are subsets of G such that A ~ B then evidently

XAY ~ XBY. We remark that for a subgroup H of the group G we have
necessarily
hH = Hh = H for all h E H.
Lemma 4
Let H be a subgroup of the group G and let x E G. Then the subset X-I Hx given
by {x-1hxlh E H} is a subgroup of G.
Proof
Let a, bE x-I Hx. Then there exist h, k E H such that
a = x-1hx, b = x-1kx.
But
and
since
(x-1h-1x)(a) = (x-Ih-1x)(x-1hx) = x-1h-1xx-1hx = x-1h-1hx = x-IX = e.
But H is a subgroup and so hk, h- I E H. Thus ab, a-I EX-I Hx which is there-
fore a subgroup. 0
Definition 16
Let H be a subgroup of the group G. Any subgroup of the form X-I Hx for some
x EGis called a conjugate of H. A subgroup H which coincides with all its con-
jugates is said to be normal in G.
Thus H is normal in G if and only if x-1Hx = H for all x E G or, equivalently,

xH = Hx for all x E G.
A useful characterization of normality is given in the next result.
Theorem 12
Let H be a subgroup of the group G. Then H is normal in G if and only if
x-1Hx ~ H for all x E G.
Proof
Certainly if H is normal in G the condition holds by definition.
Suppose now x- l Hx ~ H for all x E G. We want to show that H ~ x- l Hx for
all x E G. Now x- l Hx ~ H implies that
x(x- l Hx)x- l ~ xHx- l
and so
This proves the result. o

Normal subgroups are usually easier to determine than subgroups which are
not normal. The centre of a group is always normal and all subgroups of an
Abelian group are normal.
Example 17
In the dihedral group D 4 of order 8 in Example 16, the centre Z(D4 ) is {e, b}.
To see this we observe that
b-lab = bab = cb = a,
b-lcb = bcb = ab = c,
b-ldb = bdb = gb = d, etc.
The remaining subgroups of order 2 are not normal since, for example,
a-l{e,d}a = {a-lea,a-lda} = {e,g}.
The subgroups of order 4 are all normal since we have
a-l{e, b, d,g}a = {a-lea, a-lba, a-lda,a-lga} = {e, b,g, b} etc.
We have proved that the subgroups {e, d} and {e, g} are conjugate and we may
prove that {e, j} and {e, h} are conjugate.
Any two distinct four-element subgroups have {e, b} as their intersection. We
note, for future reference, that
{e, dHe, g} = {e, b, d, g} = {e, 9He, d},
but
{e,dHe,j} = {e,c,d,j} =1= {e,a,d,j} = {e,jHe,d}.
Theorem 13
Let H and K be normal subgroups of the group G. Then H n K is a normal sub-
group ofG.
Proof
We have already shown that H n K is a subgroup of G. We merely have to estab-
lish the normality of H n KinG. Let x E G. Then from H n K S; H we have
x-I(H n K)x S; H. Similarly x-I(H n K)x S; K. Thus x-I(H n K)x S; H n K
and so H n K is normal in G. 0
Lemma 5
Let H be a subgroup of the group G. Let NG(H) denote the subset of G given by
{x E Glx- l Hx = H}. Then NG(H) is a subgroup of G containing H.
Proof
Let x, y E NG(H). Then x-I Hx = H and y-l Hy = H. Hence
(xy)-l H(xy) = y-1x- 1Hxy = y-l Hy = H
and so xy E NG(H). Also
H = eHe = (X-I)-IX- 1Hxx- l = (x-l)-l H(x- l )
and so x-I E NG(H). Thus NG(H) is a subgroup. o
Definition 17
Let H be a subgroup of the group G. Then NG(H) = {x E G Ix-I Hx = H} is a
subgroup called the normalizer of H in G.
NG(H) is the largest subgroup of G in which H is a normal subgroup.

NG(H) = H if and only if H is a normal subgroup of G.
Theorem 14
Let Hand N be subgroups of G and let N be a normal subgroup of G. Then
HN = NH and HN is a subgroup of G.
Proof
First we prove that HN = NH. We observe that HN consists of all elements of
the form hn (h E H, n E N) and moreover that each such element belongs to NH
since hn = (hnh-l)h E NH as N is normal in G, thus HN ~ NH. Similarly
NH ~ HN and so NH = HN.
Suppose now a, b E HN. Then there exist hI' h2 E H and nl' n2 E N such that
a = h1nl and b = h2~. Then
ab = h1n1h2n2 = hl(nlh2nll)nl~ E HN
since N is a normal subgroup and H is a subgroup and
a-I = (h1nl)-1 = nl 1hl l E NH = HN.
Thus HN is a subgroup of G. o
Examples 18
1. In the dihedral group G = D 4 of order 8 as above, suppose we wish to find
N G({e, d}). Reading from the Cayley table, we have e-1de = d, a-Ida = g,
b-1db = d, c-1dc = g, d-1dd = d, rId! = g, g-ldg = d, h-1dh = g. Thus
NG({e,d}) = {e,b,d,g}.
Alternatively we may observe that
NG({e,d}) ;2 Z(D4 ) = {e,b}
and so
N G({e,d});2 {e,d}{e,b} = {e,b,d,g}.
But
NG({e,d}) "" D 4
as {e, d} is not normal and we already know all subgroups of D 4 so that
NG({e,d}) = {e,b,d,g}.
Similarly we have
Nc({e,g}) = {e,b,d,g}, Nc({e,f}) = {e,b,J,h}.
2. We recall that a square matrix A over IR is non-singular if A- l exists, or
equivalently if A has non-zero determinant. In the case of a 2 x 2 matrix
A is non-singular if and only if ad - be '" 0 and then
1
a-l=ad_be
(d -b)
-e a .
The set of such non-singular 2 x 2 matrices over IR forms a group under multi-
plication which is called the general linear group of 2 x 2 matrices over IR and
which is denoted by GL(2,1R) (for the proof that GL(2,1R) is a group see
Exercises 4.3, no. 6).
Suppose we want to find the centre Z(GL(2, IR)). Then we are looking for all
matrices
X Y)
(
z t
(xt - yz '" 0)
such that
for all a, b, e, d E JR., ad - be '" O.

Instead of considering arbitrary a, b, e, d we notice that the above has certainly
to hold for particular choices of non-singular matrices such as
( ~ ~) and ( ~ ~ ).
Now
implies
X x+Y) = (x+z y+t)

(
z z+t z t
from which
X = X + z, x + y = y + t and so z = 0, x = t.
Similarly from (~ ~) we obtain y = 0, x = t. Thus for (: ~) to be in the

centre we must have
But, on the other hand, we always have
Hence the centre Z(GL(2, JR» consists precisely of the so-called 'scalar matrices'
( ~ ~) where x # 0.
We claim that
is an Abelian subgroup of GL(2, JR). (For proof see Exercises 4.3, no. 6). Suppose
we want to find Na(H). Then we require to find those matrices (: ~) such
that
for all a E JR. Consequently, we require to find those matrices (: ~) for which
given a E JR there exists b E JR such that
(: ~) -\ ~ ;) (: ~) = (~ ~)
or, equivalently,
Thus we must have

y+at) = (X bX+Y)
t z bz + t
from which
x + az = x, y + at = bx + y, t = bz + t
and so
az = bz = 0, at = bx.
Since a is arbitrary, Z = O. But now,
(~ ~) (~ ~) (~ ~)
-1 (xt ~ 0)
=:t(~ ~Y)(~ ~)(~~)

=~xt (tx0
2
t a)
xt
= (1
0
t:).
1
Thus finally
We have defined the centralizer of an element and the normalizer of a sub-

group. We may extend the definitions very slightly as follows.
Definition 18
Let A be a non-empty subset of the group G. The centralizer and the normalizer
of A are defined to be the subgro·J.ps Ga(A) and Na(A) given respectively by
= {x E Glx- 1ax = a for all a E A},
Ga(A)
Na(A) = {x E Glx- 1 Ax = A}.
We have asserted that these subsets are subgroups and the assertion should be
confirmed.
We have
Ga(A) = n Ga(a),
aEA
and, in particular,
Z(G) = Ga(G) = n Ga(a).
aEa
Exercises 4.3
1. Write down the Cayley table of a cyclic group G of order 6. Find the
proper subgroups of G.
2. Let {A, B, G, ... } be a set of subgroups of a group G. Prove that the
intersection A n B n G . .. is a subgroup of G.
3. Find the subgroups of the non-Abelian group of order 6 in Examples 11

no. 3 and Example 14. Determine which are normal. Determine the
conjugates of each subgroup.
4. Let {A,B,C, ... } be a set of normal subgroups of a group G. Prove
that An B n en ... is a normal subgroup of G.
5. Let HbeasubgroupofthegroupG and let a E G. Prove that a-I Hais
a subgroup of G and that
n
xEG
x-IHx
is a normal subgroup of G.
6. Prove that the set GL(2, R) of non-singular 2 x 2 matrices is a group,
and that
is an Abelian subgroup.
7. Let A be a non-empty subset of the group G. Prove that CG(A) is a
normal subgroup of NG(A)
8. (Hard) We have shown that if N and H are subgroups of G and N is a
normal subgroup then NH is a subgroup of G and NH = HN. With-
out assuming that N is a normal subgroup prove that N H is a sub-
group if and only if NH = HN.
9. (Hard) Let G be a finite group and let H be a proper subgroup of G and
let H have n conjugates H = HI' H 2 •. •• ,Hn • Prove that the set
HI U H 2 U ... U H n contains, at most, n(IHI - 1) + 1 elements of G.
Deduce that G is not equal to the union HI U H 2 U ... U H n .
4.4 Lagrange's Theorem, Cosets and Conjugacy

We shall consider two partitions of G. The first will be obtained from 'cosets' of a
subgroup of G and the second from the conjugacy of elements of G.
Definition 19
Let H be a subgroup of the group G. Let a E G. Then
aH = {ahlh E H}
is called the left coset of H in G determined by a. Similarly

Ha = {halh E H}
is called the right coset of H in G determined by a.
We note that a E Ha and that aH = Ha for all a E G if and only if H is normal

in G. We shall prove a number of results on right cosets; it will be evident that
analogous results will also hold for left cosets.
Theorem 15
Let H be a subgroup of the group G. Let a, bEG. Then the following statements
are equivalent.
1. Ha = Hb.
2. There exists h E H such that b = ha.
3. bE Ha.
4. 00- 1 E H.
Proof
We prove the equivalence of the statements in a cyclic manner.
Suppose 1 holds. Then from Ha = Hb we have b E Hb = Ha and so there exists

h E H such that b = 00. Thus 2 holds.
Suppose 2 holds. Since b = ha for some h E H we have, by definition, b E Ha and
3 holds.
Suppose 3 holds. From b E Ha we have b = 00 for some h E H and so, evidently,
00- 1 = haa- 1 = h E H. Thus 4 holds.
Suppose 4 holds. Then ba- 1 E H implies Hba- 1 = H and so Hb = Hba-1a = Ha.
This completes the proof. 0
Theorem 16
Let H be a subgroup of the group G. Any two right cosets of H in G are either
equal or have empty intersection. Consequently the set of right cosets of H in G
is a partition of G.
Proof
Let a, bEG. We have to show that either Ha = Hb or Ha n Hb = 0. Suppose
Ha n Hb =F 0, then we prove that Ha = Hb. Since Ha n Hb =F 0 there exists
x E Ha n Hb. Hence there exist hI, h 2 E H such that x = hI a = ~b. But

then b = h2l h 2b = h 2l h l a where h2l h l E HasH is a subgroup. But then by
Theorem 15 Ha = Hb.
G is certainly the union of the cosets of H in G since each element x (say) of G
belongs to the coset Hx in G. But Hx is also the only coset to which x belongs
since if we find that x E Hy then necessarily Hx = Hy. Thus we may conclude
that the set of right cosets of H in G is a partition of G. 0
Example 19
Consider the dihedral group D 4 of order 8 (Example 16 and 17). H = {e, f} is
a subgroup. a ¢ {e,f} and {e,f}a = {ea,ja} = {a,d}. Choose b,b ¢ {e,f} U
{e,f}a, then {e,f}b = {eb,jb} = {b,h}. Choose c ¢ {e,f} U {e,f}aU {e,f}b
and then we obtain G = {e, f} U {e, f}a U {e, f}b U {e, f}c, which expresses G
as a union of four mutually disjoint right cosets of H in G.
Definition 20
Let H be a subgroup of the group G. IT G is written as a union of mutually dis-
joint right (left) cosets of H in G then G is said to be written as a right (left) coset
decomposition of H in G. IT there is an infinite number of cosets in a right (left)
coset decomposition then H is said to have infinite index in G. IT there is a finite
number of cosets in a coset decomposition then H is said to have finite index
equal to this number of cosets; we denote this number by IG : HI.
At first glance it might appear that the index could depend on whether we
were using right or left cosets. However, if we have a right coset decomposition
G = Hal U Ha2 U ... U Han
then we also have a left coset decomposition, namely,
G = aiIHUa2IHU ... Ua;;IH.
(For proof of this observation see Exercises 4.4, no. 2). Commonly we take al = e
and so Hal is the subgroup H.
Lemma 6
Let H be a finite subgroup of G. Then the number of distinct elements in any
coset of H is precisely the number of elements in H.
Proof
Let a E H. Consider Ha. We have to prove that H and Ha have the same
number of elements. We claim that the mapping f: H - Ha given by
f(h) = ha (x E H).
is a bijection. Certainly f is a surjection and f(h l ) = f(h 2 ) (h ll h2 H) implies
that hI a = h2 a and so hI = h2 , giving f as an injection. This completes the
proof. D
We now prove one of the seminal theorems in group tl.eory. This theorem is
named after J.-L. Lagrange (1736-1813) who, in effect, proved what we would
now regard as a special case of the theorem.
Theorem 17 Lagrange's Theorem

Let H be a subgroup of the finite group G. Then the order of H divides the order
ofG and, in fact, IGI = IG: HIIHI·
Proof
The group G may be written as a coset decomposition of the cosets of H in G.
Suppose IG: HI = n. Then we have, say,
G = Hal UHa2 U ... UHa n,
where the cosets Hal, Ha2' ... ,Han are distinct. By Lemma 6 the number of ele-
ments in each coset is IH I and there are n such cosets. The number of elements in
Gis IGI and so
IGI = nlHI = IG: HIIHI
giving the result. D
Remark
If H and K are subgroups of the finite subgroup G where H ~ K ~ G then
IG: HIIHI = IGI = IG: KIIKI = IG: KIIK: HIIHI
and so
IG:HI = IG:KIIK:HI·
Examples 20
1. Let the order of G be a prime p. Then G has only the subgroups G and {e}
since if H is a subgroup of G, IHI divides p and so IHI =p or IHI = 1
giving the result.
2. We have listed the subgroups of the dihedral group D 4 of order 8. Note that
all subgroups are of orders 1, 2, 4 or 8.
3. It is fairly easy to prove that Z, considered as an additive group, has the
subset 6Z = {6xlx E Il} as a proper subgroup. The integer a belongs to the
coset 61l + a = {6x + alx E Z} and a and b (b E Z) belong to the same
coset if and only if 61l + a = 6Z + b. Thus a and b belong to the same coset
if and only if a - b E 61l or 6 divides a-b. The six cosets of 6Z are
6Z, 6Z + 1, 6Z + 2, 6Z + 3, 6Z + 4, 6Z + 5.
We now prove a result on cyclic groups which depends on the Division

Algorithm for integers.
Theorem 18
An element of order n of a group G generates a cyclic subgroup of order n of G,
and conversely.
Proof
Let a E G and suppose a has order n. Then e, a, a2 , ••• , an -1 are distinct and
an = e. Let H = {amlm Ell}. Then H is the cyclic subgroup generated by a.
Consider now a typical element am of H. By the Division Algorithm for
integers there exist q, r E Il, 0 ~ r < n, such that
m=qn+r.
Then
Thus H = {e, a, ... , an - 1} and SO a generates a cyclic subgroup of order n.

Conversely if H is a cyclic subgroup of G of order n then H is generated by an
element a of order p (say). But an element a of order p generates a subgroup of
order p and SO, as H has order n, we must have p = n, thereby completing the
prooL 0
Theorem 19
A finite cyclic group G has subgroups of all possible orders.
Proof
Suppose G has order n. Then by Lagrange's Theorem any subgroup must have
order m where m divides n. We wish to show that a subgroup of order m exists.
Let m = nd (d EN). Let G be generated by a and let H be generated by ad,
H = {(ady Ir E Z}.
Now (ad)m=amd=an=e and e,ad,a2d , ... ,a(m-l)d are distinct since a has
order n = md. Thus ad has order m and so, by Theorem 18, ad generates a sub-
group, namely H, of order m. 0
We now turn to the second of our partitions on the group G. We recall that if
a, bEG then a is conjugate to b if there exists x E G such that x-lax = b and
that we proved that conjugacy is an equivalence relation on G (Theorem 7);
this equivalence relation yields our partition of G.
Definition 21
Let a be an element of the group G. The set of all elements of G conjugate to a,
namely {x-lax Ix E G}, is called the conjugacy class of G containing a.
The group G is therefore the disjoint union of its conjugacy classes. We now
give a formula for the size of a finite conjugacy class in G.
Theorem 20
Let a be an element of the group G. Suppose the centralizer of a in G, GG(a) , has
index n in G. Then a has precisely n conjugates.
Proof
Let
G = GG(a)xI U GG(a)x2 U ... U GG(a)xn
be a coset decomposition of GG(a) in G. We claim that the conjugates of a are
precisely
Certainly if x E G then x belongs to some coset, say x = CXk for some e E GG(a)
and some xk (1 ~ k ~ n). Then
x-lax = (cxk)-la(cxk) = Xkle-lacxk = Xklaxk'
Thus the conjugates of a are
Xllaxl' X2"lax2' ... , x;;-laxn .
We now have to show that these conjugates are distinct. Suppose for some i,j
-1 -1
Xi aXi = Xj aXj'
Then
and so
(Xixj1)-la(XiXj) = a.
But this implies that Xixj1 E GG(a) and hence Xi and Xj belong to the same coset
of GG(a) in G. This is impossible unless i = j. Hence we obtain the result. 0
Corollary
Let a be an element of the group G. Then the number of elements in the
conjugacy class of a divides IG I.
Proof
The number is IG : GG(a)1 which, by Lagrange's Theorem, divides IGI· 0
Example 21
Suppose we try to find the conjugacy classes in the dihedral group D 4 of order 8.
Since Z(D 4 ) = {e,b}, {e} and {b} are two conjugacy classes. Consider the ele-
ment a of D 4 • We need to find GG(a). We notice that {e,a,b,e} is an Abelian
subgroup of G and so {e,a,b,e} ~ GG(a). This implies that 4 divides IGG(a) I
which, of course, divides 8. Thus IGG(a)I = 4 or 8. But \GG(a) I = 8 implies
that a E Z(D4 ) which is false. Thus IGG(a)I = 4 and so GoCa) = {e,a,b,e}. A
suitable coset decomposition is
G = GG(a) U GG(a)d
and so {a,d-lad} = {a,e} is a conjugacy class. The centralizer GG(d) of the
element d is {e, b, d, g} and the conjugacy class containing d is {d, g}. Similarly
we may calculate the remaining conjugacy classes and so finally obtain as the
conjugacy classes of D4 , {e}, {b}, {a, b}, {d,g} and {j, h}.
Exercises 4.4
1. In the same way that we proved that the number of conjugates of an
element of a group G is the index of the centralizer of that element in
the group G, prove that if a subgroup H of G has n conjugates in G
then the normalizer, NG(H), has index n in G.
2. If
G = Hal U Ha2 U ... U Han
is a right coset decomposition of a subgroup H in the group G, prove
that
al-IH U a2-IH U ... Ua n-IH
is a left coset decomposition of H in G.
3. What are the cosets of 12Z = {12x Ix E Z} considered as a subgroup of

the additive group Z?
4.5 Homomorphisms
We are here concerned with mappings which preserve algebraic structure. Such
mappings are called 'homomorphisms' (Greek: homo same, morph form). More
precisely we have the following definition which we make for semigroups.
Definition 22
Let 8 and T be semigroups and let 1:8 ---. T be a mapping such that
I(xy) = l(x)l(y)
for all x, y E 8. Then 1 is said to be a homomorphism from 8 into T.
Lemma 7
Let 8 and T be semigroups and let 1 : 8 ---. T be a homomorphism. Then 1(8) is
a semigroup (strictly 1(8) is a subsemigroup of T).
Proof
Let a, bE f(8). Then there exist x, y E 8 such that f(x) = a, f(y) = b. Thus
ab = f(x)f(y) = f(xy) E f(8).
Thus f(8) is closed under multiplication and, since f(8) is a subset of T, the
multiplication is associative. Thus f(8) is a semigroup. 0
For the remainder of this chapter we shall be concerned with homomorphisms

of groups.
Theorem 21
Let G and H be groups and let f : G -+ H be a homomorphism. Then the follow-
ing hold.
1. If eG and eH are the identities of G and H respectively, then f(eG) = eH'
2. For all x E G, [f(x)r l = f(x- l ) where [f(x)r l is the inverse off(x) inH and
x-I is the inverse of x in G.
3. f(G) is a subgroup of H.
4. Let K = {x E Glf(x) = eH}' Then K is a normal subgroup of G.
Proof
We note the necessity of distinguishing the identities in the two groups G and H.
1. Since f(eG)f(eG) = f(eGeG) = f(eG), we conclude that f(eG) = eH'
2. Since f(x-l)f(x) = f(x-1x) = f(eG) = eH (x E G), we have f(x- l ) =
[f(x)r l .
3. By Lemma 7, f(G) is closed under multiplication in H. Let now a E f(G).
Then a = f(x) for some x E G and a-I = [f(x)r l = f(x- l ) E f(G). Thus
f( G) is a subgroup of H.
4. Since f(eG) = eH, K is non-empty. Let x, y E K, then f(x) = f(y) = eH'
Hence
and
Consequently K is a subgroup of G. To prove that K is a normal subgroup let

a E K, x E G. Then
I(x-Iax) = l(x-I)/(ax) = l(x-I)/(a)/(x)
= [/(X)rleHI(x) = [/(x)rl/(x) = eH,
and so x-lax E K. Hence K is a normal subgroup of G. o
Definition 23
Let G and H be groups and let I : G -+ H be a homomorphism. Then the sub-
group K = {x E GI/(x) = eH} is a normal subgroup of G which is called the
kernel of I, written Ker I.
Examples 22
1. Let G be a group, G =I {e}. Then we always have at least two homomorph-
isms. First we have the mapping I, easily checked to be a homomorphism,
given by I(x) = e (x E G). Second, we have the mapping 9 defined by
g(x) = x (x E G), again easily verified to be a homomorphism.
2. Let G be the cyclic group of order 4 generated by a, G = {eG' a, a2 , a3 } and
a4 = eGo Let H be the cyclic group of order 2 generated by b, H = {eH,b},
b2 = eH' Then we assert that the mapping I ; G -+ H given by
l(eG) = eH, I(a) = b, l(a 2 ) = eH, l(a3 ) =b

is a homomorphism. We require to establish this assertion. For example we
have
We may similarly verify the other possibilities.

3. Let C \ {OJ and IR \ {OJ be the multiplicative groups of non-zero complex and
real numbers respectively. Then the mapping I : C \ {OJ -+ IR \ {OJ given by
I(z) = Izi (z E C\ {OJ)
is a homomorphism since, by a property of the modulus (absolute value) of
complex numbers, we have
4. Consider the non-Abelian group G of order 6 with Cayley table given in

Example 6 no. 3. Define p: G -+ {I, -I} by
p(e) = 1, p(a) = 1, p(b) = 1, p(c) = -1, p(d) = -1, p(J) = -1.
Then we may verify that p is a homomorphism, for example

p(df) =p(b) = 1 = (-1)(-1) =p(d)p(J).
5. The set G of matrices of the form
X Y) (x, Y, t E JR, xt ;C 0)
( o t
has been shown to be a group under matrix manipulation. We claim that the
mapping f : G -+ JR \ {o} given by
f:(~ :)-+xt
is a homomorphism. Certainly we have
giving the result. (We note that xt is the determinant of (~ ~). see next
example).
6. We have defined GL(2, JR) as the group of non-singular 2 x 2 matrices. The
mapping f: GL(2, JR) -+ JR given by f(A) = det A, for any non-singular
2 x 2 matrix A, is a homomorphism since for 'square' matrices A and B,
det(AB) = detAdetB.
7. The existence of certain homomorphisms imposes conditions on the group or
groups involved. For example, suppose the mapping f : x -+ x-I (x E G) is a
homomorphism of the group G into itself. Then for x, y E G,
x-1y-1 = f(x)f(y) = f(xy) = (xy)-l = y-1x- l ,
and so
yx = (x-1y-I)-1 = (y-1x-I)-1 = xy.
Thus if f is a homomorphism G is necessarily Abelian.
Associated to each homomorphism of a group we have a normal subgroup

called the kernel. Given a normal subgroup of a group we shall see that we
may construct, from the cosets of the normal subgroup, another group onto
which the original group may be homomorphically mapped with kernel equal
to the normal subgroup. We shall explore for a given group this tight relationship
between groups onto which the given group may be homomorphically mapped
and normal subgroups of the given group.
Lemma 8
Let G and H be groups and let f : G -+ H be a homomorphism with kernel K.
Let a and b be elements of G. Then f(a) = f(b) if and only if Ka = Kb.
Proof
Suppose f(a) = f(b). Then
f(ab-l) = f(a)f(b- l ) = f(a)[f(b)r l = f(b)[f(b)r l = en'
Hence ab- l E K and so Ka = Kb. Conversely if Ka = Kb then b = ka for some
hE K and so
f(b) = f(ka) = f(k)f(a) = enf(a) = f(a). o
Theorem 22
Let H be a normal subgroup of the group G. Let G/ H denote the set of cosets of
H in G. A binary operation is defined on G/H as follows. Let Ha and Hb
(a, bEG) be cosets of H in G and define (Ha)(Hb) = Hab. Then this binary
operation is well-defined and under this operation G/ H is a group. Further
the mapping G -+ G/H given by a -+ Ha (a E G) is a homomorphism of G
onto G/H with kernel H.
Proof
It is not immediately obvious that the binary operation on G/ H is well-defined.
In order that the definition should be well-founded we have to show that if
Ha = Ha' and Hb = Hb' (a, a', b, b' E G) then Hab = Ha'b'. Now under the
assumption that Ha = Ha' and Hb = Hb' we have that a' = h1a and b' = h 2b
for some h ll h 2 E H respectively. Then
a'b' = h 1ah2 b = h 1ah 2 a- 1ab = hab
where h = h l (ah 2 a- l ) E H since H is a normal subgroup of G. Then
Ha'IJ = Hhab = Hab
and so we have vindicated our claim that the definition is well-founded.
We now show that G/ H is a group. Certainly, by construction, G/ H is closed

under the defined multiplicative binary operation. Associativity follows easily
since for a, b, c E G.
(HaHb)Hc = HabHc = H(ab)c = Ha(bc) = Ha(Hbc) = Ha(HbHc).
The coset H plays the role of the identity element of G/ H as
HHa = HeHa = Hea = Ha (a E G).
The inverse of Ha (a E G) is given by Ha- 1 since
(Ha-1)(Ha) = Ha-1a = He = H.
Thus G/ H is a group.
The mapping I: G -+ G/H given by I(a) = Ha (a E G) is obviously surjec-
tive and is a homomorphism since
I(ab) = Hab = HaHb = l(a)l(b) (a, bEG).
Finally
Ker I = {a E GII(a) = H}
= {a E G IHa = H} = H. o
Definition 24
Let H be a normal subgroup of the group G. The group G/ H, as constructed
above, is said to be a factor-group of G.
We note that if H has infinite index in G then G/ H is infinite, whereas if H has
finite index IG: HI in G then IG/H/ = IG: HI.
Definition 25
Let G and H be groups and let I : G -+ H be a homomorphism. If I is surjective
(I(G) = H) then I is said to be an epimorphism (Greek: epi upon). If I is injec-
tive (I(a) = I(b) implies a = b, a, bEG) then I is said to be a monomorphism
(Greek: mono alone). If I is bijective (surjective and injective) then I is said
to be an isomorphism (Greek: iso equal).
Example 23
The Klein four-group G = {e, a, b, e} with Cayley table
e a b e
e e a b e
a a e e b
b b e e a
e e b a e
is isomorphic to the group H consisting of the four matrices
(10) (-1 0) (0 -1) (0 1)

o 1' 0 -1 ' 1 0 ' -1 0.
The isomorphism is given by the mapping
e-+ (1o 0)
1
, a-+ (-1 0)
0 -1
,
Isomorphic groups have the same group-theoretical structure although they

may differ in other material aspects.
Example 24
Consider the dihedral group D 4 of order 8, the Cayley table of which we have
used frequently. The centre Z(D 4 ) = {e, b} and since the centre is a normal sub-
group we may form the group D 4 / Z(D 4 ). We shall see that this group is iso-
morphic to the Klein four-group.
Let us denote the cosets of Z(D 4 ) in D 4 by capital letters. Thus let
E = Z(D4 ) = {e,b},
A = Z(D4 )a = {e,b}a = {a,ba} = {a,e},
B = Z(D4 )d = {e, b}d = {d, bd} = d,g},
C= Z(D4 )/= {e,b}/= {I,b/} = {I,h}.
We may now draw up a Cayley table for D 4 /Z(D4 ) = {E,A,B,C}. For
example
= Z(D4 )dZ(D4 )d = Z(D4 )d2 = Z(D4 )e = E,
A2
AB = Z(D4 )aZ(D4 )d = Z(D 4 )ad = Z(D4 )/ = c,
finally obtaining
E A B c
E E A B C
A A E C B
B B C E A
C C B A E
which is the Klein four-group.
We now come to the first of two important results in this section.
Theorem 23 First Isomorphism Theorem

Let G and H be groups and let f : G -+ H be an epimorphism with kernel K.
Then G I K and H are isomorphic.
Proof
The kernel K of the epimorphism f is a normal subgroup of G and so we also
have an epimorphism g: G -+ GIK. Diagrammatically we have
f
G )H
g~
GIK
What we want to do is to construct an isomorphism h from GI K to H. We

cannot compose 9 with f. In a manner of speaking what we do is to reverse back-
wards along the arrow from G I K to G and go directly along the arrow from G to
H. Once we have established that this process does yield a well-defined mapping
h it is fairly easy to prove that the h so constructed is an isomorphism.
Thus let Ka (a E G) beanelementofGIK. We would like to define a mapping
h: GIK -+ H by letting
h(Ka) = f(a).
This definition only makes sense if Ka = Kb (a, bEG) implies that f(a) = f(b).
But Ka = Kb implies that b = ka where k E K and then
f(b) = f(ka) = f(k)f(a) = eHf(a) = f(a).
Consequently our definition is in fact soundly based.
The mapping h is surjective since if c E H it follows that, as f is an epimorph-
ism, there exists a E G such that f(a) = c and then h(Ka) = f(a) = c.
The mapping h is injective since if h(Ka) = h(Kb) (a, bEG) then we have
f(a) = f(b) and so f(ab- 1) = f(a)f(b)-l = eH from which ab- 1 E K and
Ka = Kb. It remains now to show that h is a homomorphism. Let a, bEG, then
h(KaKb) = h(Kab) = f(ab) = f(a)f(b) = h(Ka)h(Kb).
Thus finally we conclude that h is an isomorphism. o
Examples 25
1. Let G and H be groups and let f : G -+ H be a homomorphism. Then f( G) is
a subgroup of Hand f(G) is isomorphic to G/K.
2. Let G and H be finite groups and let f : G -+ H be an epimorphism. Then IH I
divides IG I since
IG I = IKer f II G : Ker f I = IKer f II G/ K I = IKer f II HI·
3. Let H be a subgroup of the centre Z(G) of the group G. Suppose G/H is
cyclic. Then there exists a E G such that G/H = {(Hat In E Z}. But
(Hat = Han and SO G/H = {Han In E Z}. Now this fact implies that G is
Abelian since if we have x, y E G then x = h1an, y = ~am for some
hll~ E H and m,n E Z, and therefore, as H ~ Z(G),
xy = (hlan)(~am) = hl~anam = h 2 h 1aman = ~amhlan = yx.
We now arrive at the second of our important theorems.
Theorem 24 Second Isomorphism Theorem

Let G be a group, let Hand N be subgroups of G and let N be normal in G. Then
the factor-groups NH/ Nand H / (N n H) are isomorphic.
Proof
Notice that since N is normal in G, NH is a subgroup and N n H is a normal
subgroup of H. The following diagram may assist in the understanding of the
statement of the theorem:
N H
Each subgroup is shown to lie above any subgroup which it contains. Such dia-
grams in group theory are sometimes called Hasse diagrams (after H. Hasse,
1898-1979).
We first define a mapping p : H -+ NH/ N and show that this mapping p is an
epimorphism. For h E H we define p(h) to be the coset Nh of N in the subgroup
NH = {nh In E N, h E H}. Certainly p is surjective. But p is also a homomorph-
ism since
= Nh 1 h 2 = Nhl~ = P(hl)P(~) (hll ~ E H).
p(h 1 h 2 )
Thus p is an epimorphism. Let K = Ker p. Then, by the First Isomorphism
Theorem, H / K is isomorphic to NH/ N. In order to complete the proof we
have to prove that K = N n H. Let h E H n N. Then hEN and so Nh = N
and thus hE K. On the other hand, if k E K then Nk = N, and so we have
kEN and finally kEN n H. Thus K = N n H. 0
Some groups admit of only two epimorphisms, namely the epimorphism in

which all elements of the group are mapped into the number 1 and the isomorph-
ism of the group with itself. Since the existence of epimorphisms is directly
related to the existence of normal subgroups we make the following definition.
Definition 26
A non-trivial group which has no proper normal subgroups is said to be simple.
Example 26
Suppose we consider a non-trivial group G which is both simple and Abelian. We
recall that in an Abelian group all subgroups are normal. Hence our present
group G has no proper subgroups. Let a E G, a:F e. Then a generates a non-
trivial cyclic subgroup which must be G. IT G is infinite then a 2 generates a
proper subgroup and we would have a contradiction. Hence G is a finite cyclic

group. Let p be a prime divisor of the order of G. Then G has a subgroup of
order p which must be G itself. Hence, to summarize, a non-trivial simple Abelian
group is cyclic of prime order. For obvious reasons the mathematical interest is in
non-Abelian simple groups.
Exercises 4.5
1. Let G, H and K be groups. Let f: G -+ H and g: H -+ K be given
homomorphisms. Prove that go f is a homomorphism.
2. Let G be a group and suppose that the mapping f: G -+ G, which is
given by f(a) = a 2 (a E G), is a homomorphism. Prove that G is
Abelian.
3. Let G be a group with centre Z(G). Suppose there exists a E G such
that x-1a-1xa E Z(G) for all x E G. Prove that the mapping
x -+ x-1a-1xa is a homomorphism of G into Z(G).
4. Let N and H be subgroups of the group G where N S; H S; G and N is

normal in G. Prove that H / N is a subgroup of G and that every sub-
group of G/ N is of the form H / N for some subgroup H of G. Prove
that H / N is a normal subgroup of G/ N if and only if H is a normal
subgroup of G.
5. The quaternion group Q of order 8 is given by Q = {e, a, b, c, d, f, g, h}
with Cayley table
e a b c d f 9 h
e e a b c d f 9 h
a a e c b f d h 9
b b c a e 9 h f d
c c b e a h 9 d f
d d f h 9 a e b c
f f d 9 h e a c b
9 9 h d f c b a e
h h 9 f d b c e a
What is the centre Z( Q) of Q? Find all subgroups of Q and construct a

Hasse diagram. Determine the conjugacy classes. Identify the group
Q/Z(Q).
5
Rings
We first formulated the abstract concept of a ring from our considerations of

the integers and of polynomials. It is our aim to carry forward and to develop
the concept of a ring, for which the integers will continue to serve as a useful
exemplar.
5.1 Arithmetic Modulo n

Let us, for the present, focus our attention on the prime 3 in the ring of integers
Z. Suppose we were to consider only the integers 0, 1, 2. As it stands the set
{O,l,2} is not a ring since, for example, 2 + 2 = 4 ¢ {O,l,2}. However, suppose
we perform addition and multiplication 'modularly' with regard to the prime 3,
that is to say we replace any sum or product of the numbers 0, 1, 2 by the cor-
responding remainder on division by 3. Thus we would have
2+2::1{mod3}
since 2 + 2 = 4 = 3 + 1, where to emphasise that we are performing 'addition'
and 'multiplication' modularly we use the symbol:: which denotes 'congruence'
and we also write 'mod 3'. We may compile modular addition and multiplication
tables as follows.
145
Sum 0 1 2 Product 0 1 2
0 0 1 2 0 0 0 0
1 1 2 0 1 0 1 2
2 2 0 1 2 0 2 1
As we shall shortly see {O, 1, 2} with these tables becomes a ring. We now require
to make rather more precise some of the notions above.
Definition 1
Let n be a given non-zero integer. Let a and b be two integers. We say that a is
congruent to b if n divides a-b. We write
a == b (mod n).
Otherwise we say that a is not congruent to b modulo n if n does not divide a-b.
Example 1
Since 24 divides 48 = 83 - 35 we have
83 == 35 (mod 24).
Since 17 does not divide 96 = 122 - 26 it is evident that
122 == 26 (mod 17) is false.
The next lemma summarises some obvious facts.
Lemma 1
Let n be a given non-zero integer. Then the following statements hold.
1. For all integers a, a == a (mod n).
2. Let a and b be integers such that a == b (mod n). Then b == a (mod n).
3. Let a, band c be integers such that a == b (mod n) and b == c (mod n). Then
a == c (mod n).
4. The relation of congruence is an equivalence relation on Z.
Proof
Certainly n divides 0 = a - a for all integers a and so we obtain 1. If for integers
a and b, n divides a - b then n divides b - a, and 2 follows. To prove 3 we remark
that ifn divides a - band n divides b - c then n divides a - c = (a - b) + (b - c)
giving 3. Statement 4 is an immediate consequence of the three preceding
statements. 0
Rings 147
By Lemma 1, for a given non-zero integer n, the set of integers is partitioned

into equivalence classes of integers. If a E Z the equivalence class containing a is
{x E Zlx = a (mod n)} = {a+nklk E Z}.
Notation
For a given non-zero integer n the equivalence cl8.'lS of integers containing an
integer a is denoted by a.
a={a+knlkEZ}.
Example 2
Let n = 12. Then
4 = {4 + 12klk E Z} ={ - 20, -8,4, 16,28, },
27 = {27 + 12klk E Z} = { ,3, 15,27,39,51, } = 3.
We have remarked in passing that abstraction often leads to greater insight.

Our present approach to the concept of 'integers modulo n' has been direct
but has taken no account of our existing knowledge. Suppose we reconsider
'integers modulo n' in the light of the group theory of the previous chapter.
We see that Z is an additive group and that nZ = {nk IkE Z} is the cyclic sub-
group generated by n; the equivalence classes modulo n are precisely the cosets of
nZ in Z. Now we may form the factor-group ZjnZ consisting of these cosets as
elements and with addition in ZjnZ given by
a + b = a + b (a, b E Z).
Notation
The factor-group ZjnZ is denoted by z.., where for definiteness we suppose that
n>l.
Example 3
Zs = {O, 1,2,3,4, 5,6, 7}

and, for example,
2 + 3 = 5, 4 + 5 = 9 = I,
5 + 6 = 11 = 3, 4 + 4 = 8 = O.
What is perhaps less obvious from these considerations is that we may also
introduce a multiplication with regard to the equivalence classes modulo n in
such a way that Zn becomes a ring. We summarise these facts in a theorem.
Theorem 1
Let n be a given integer, n > 1. Then Zm the set of equivalence classes modulo n,
is a commutative ring with an identity under the definitions:
a+b=a+b (a,bEZ),
ab = ab (a, bE Z).
Proof
We know that addition is unambiguously defined by a + b = a + b (a, b E Z) and
that, under this addition, Zn is an additive group.
To establish the multiplication we must first show that the rule of multiplica-
tion above is meaningful. In other words ifa = a' and b = 11 (a, b, a', b' E Z) then
we have to prove that ab = a'b'. But a = a' implies that n divides a - a' and,
similarly, n divides b - b'. But
ab - a'f:! = (a - a')b + a'(b - b')
and so n divides ab - a'b', consequently ab = a'b'.

The associativity of the multiplication follows from the definition and the
associativity of multiplication in Z since for a, b, c E Z we have
(ab)c = (ab)c = (ab)c = a(bc) = a(bc) = a(bc).

In a similar manner the distributive laws hold since for a, b, c E Z we have
a(b + c) = a(b + c) = a(b + c) = ab + ac = ab + ac = ab + ac.

The other distributive law is equally easy to prove and so, finally, we conclude
that Zn is a ring. Fairly obviously Zn is commutative and I is the identity since
o
Example 4
Following the above argument, let us write down the addition and multiplication
tables of Zt; = {D, I, 2, 3, 4, 5}.
Rings 149
Sum 0 I 2 3 4 "5 Product 0 I 2 3 4 "5

0 0 1 2 3 4 5 0 0 0 0 0 0 0
I I 2 3 4 "5 0 I 0 I 2 3 4 "5
2 2 3 4 "5 0 I 2 0 2 4 0 2 4
3 3 4 "5 0 I 2 3 0 3 0 3 0 3
4 4 "5 0 I 2 3 4 0 4 2 0 4 2
5 "5 0 I 2 3 4 "5 0 "5 4 3 2 I
Although Z6 is a commutative ring with I as identity, ~ is not an integral
domain since proper divisors of zero exist, for example, 2.3 = O.
Example 5
Let us write down the addition and multiplication tables of Zs = {O, I, 2, 3, 4}.
Sum 0 I 2 3 4 Product 0 I 2 3 4
0 0 1 2 3 4 0 0 0 0 0 0
I I 2 3 4 0 I 0 I 2 3 4
2 2 3 4 0 I 2 0 2 4 I 3
3 3 4 0 I 2 3 0 3 I 4 2
4 4 0 I 2 3 4 0 4 3 2 I
In this case there are no proper divisors of zero and Zs is an integral domain.
We notice that, in fact, the set of non-zero elements {I, 2, 3, 4} is a cyclic group
under multiplication.
Exercises 5.1
1. Write down the addition and multiplication tables of Z2, Za and Z4.
2. Find solutions to the following congruences:

4x == 3 (mod 7), 5y == 4 (mod 11), 2z == 5 (mod 13).
3. Let ml and ~ be coprime integers and let Ul and U2 be integers such
that Ul ml + ~~ = 1. For any al, a2 E Z prove that if x = ~Ul ml +
al~m2 then x == al (mod ml) and x == ~ (mod ~). (Version of the
Chinese Remainder Theorem, from China, date uncertain but before
636 AD.)
4. Let D and E be integral domains. Let R be the Cartesian product

D x E and let an addition and multiplication be defined in R by
(d1, el) + (~, e2) = (d1 +~, el + e2),
(dl,el)(~,e2) = (dl~,el~) (db~ E D, ebe2 E E).
Prove that R is a commutative ring with an identity but that R is not an

integral domain.
5. Let D be an integral domain and let x be an element of D such that
x 2 = x. Prove that either x = 0 or x is the identity of D.
6. Any modern book has an International Standard Book Number
(ISBN) given in the form:
al - ~a3a4 - as~arasag - alO'
For example 'Finite-Dimensional Vector Spaces' by P.R. Halmos
(Springer-Verlag, 1974) has ISBN 0-387-90093-4 and 'A History of
Algebra' by B.L. van der Waerden (Springer-Verlag, 1985) has
ISBN Q-387-13610-X (where X = 10). Verify that, in these two
examples,
al + 2a2 + 3a3 + 4a4 + 5as + 6a6 + 7a7 + 8as
+ 9ag + lOalO == 0 (mod 11).
Try this congruence for the ISBN of any, randomly chosen, modern
book.
5.2 Integral Domains and Fields

As we have just seen, the rings Zs and Z6 have very different properties. We wish
to characterize Zn, for general n, but first we require some definitions.
Definition 2
Let R be a ring with an identity 1. An element a of R is said to be invertible, or to
be a unit, if there exists b in R such that ab = ba = 1. We write b = a-I and call
b (if it exists) the inverse of a.
Examples 6
1. All non-zero elements of Q, IR and C are invertible.
2. Z has only two invertible elements, namely ±1.
Rings 151
3. ~ has two invertible elements I and 5.

4. Zs has four invertible elements 1,2,3 and 4.
The existence, or otherwise, of invertible elements in a ring leads us to consider

a 'field'. Dedekind gave in 1871 a definition of a field considered as a subset of
the complex numbers, but the abstract definition of a field is due to Weber.
A significant stimulus to research in fields was given in 1910 by E. Steinitz
(1871-1928).
Definition 3
A commutative ring with an identity in which every non-zero element is
invertible is called a field (German, Korperj French, corps).
Examples 7
1. Q, lR. and C are fields.
2. Zs is a field.
3. Z is not a field.
A convenient means of remembering the axioms of a field is given by the

following characterization.
Theorem 2 Characterization of a Field

Let R be a non-empty set with two binary operations of addition and multi-
plication such that the following hold.
1. R is an additive Abelian group.
2. R \ {O} is a multiplicative Abelian group.
3. The distributive laws hold.
Then R is a field.
Theorem 3
Every field is an integral domain.
Proof
Let F be a field. Then F is a commutative ring with an identity 1. We have only
to show that F has no proper divisors of zero. Suppose there exist a, b E F such
that ab = O. Now as a::j: 0 there exists a-I such that a-1a = 1 and so we may
write b = Ib = (a-1a)b = a-I (ab) = a-10 = O. Similarly if b::j: 0 we prove that
a = O. Hence we deduce that F is an integral domain. 0
As we observed above, an infinite integral domain, such as Z, need not be a

field. However, for finite integral domains, we have an explicit result.
Theorem 4
A finite integral domain is a field.
Proof
Let D be a finite integral domain. To prove that D is a field we have to show
that D \ {O} is multiplicatively a group. Certainly D \ {O} is closed under multi-
plication since x, y E D \ {O} implies that x and y are not proper divisors of zero
and so xy::j: 0 and hence xy ED \ {O}. Associativity being evident, D \ {O} is a
semigroup. We claim that D \ {O} admits cancellation. Thus suppose there exist
a, b, xED \ {O} and ax = bx. By Example 10 in Chapter 3, we deduce that a = b
and so we have cancellation. We have already shown that a finite semigroup with
cancellation is a group (Theorem 3, Chapter 4) and so D \ {O} is a group and
then D is a field. 0
Theorem 5
Zp is a field if and only if p is a prime.
Proof
Suppose p is not a prime. We show that proper divisors of zero exist in Zn. We
have p = nln2 for some nil n2, EN 'where 1 < nl < n, 1 < n2 < n. Then
nl ::j: 0, n2 ::j: 0 but nl n2 = nl ~ = p = O. Thus Zp is not an integral domain.
Suppose now that n is a prime. We may prove that Zp is a field by reversing
the argument of the previous paragraph to show that Zp is an integral domain
and then, from Theorem 4, we deduce that Zp is a field. However, it is perhaps
more instructive to prove the result directly without invoking Theorem 4.
We know that Zp is a commutative ring with an identity I. We have to show
Rings 153
therefore that the non-zero elements of Zp are invertible. Now we have

Zp = {O, I, .. . ,p - 1}. Let a be a non-zero element of Zp, a E {1,2, ... ,p - 1}.
Then as p is a prime, a and p are coprime and, by the Euclidean Algorithm,
there exist x, y E Z such that ax + py = 1. Hence
I=~+py=~+py=ax+py=~+~=~
and so a has x as inverse. Hence finally Zp is a field. o
Since Zp, p prime, is a field, Zp has some features in common with Q and JR.
Some algebraic manipulations which are familiar in Q and JR may also be
performed in Zp.
Examples 8
Suppose we try to solve, if possible, the following pairs of simultaneous linear
equations over Zr.
1. ~x+3Y= ~}.
3x+y =4
We make the coefficients of x the same in both equations by multiplying the
first by 3 and the second by 2. Thus we obtain
~.~x + ~.3y = ~.~}

2.3x + 2y = 2.4
giving
~x+~y=~}.
6x+2y= 1
These equations are therefore inconsistent and no solution is possible.
2. ~x+~y= ~}.
4x+5y=4
We then have
~.~x + ~.~y = ~.~}

3.4x + 3.5y = 3.4
giving
~x+ 3y = ~}.
5x+y =5
Subtracting we have
So far we have used only ring-theoretic properties of Zr. Now we use the
invertibility of non-zero elements of Z7. 2 has 4 as an inverse and so we have
4.2y =4.5
which yields
y= Iy= 6.
Similarly we may show that x = 4.
In the ring of integers Z we may form the sums
1, 1 + 1, 1 + 1 + 1, 1 + 1 + 1 + 1,
and these sums are distinct, being respectively 1, 2, 3, 4, ... On the other hand, in
Zn the corresponding sequence
repeats as we obtain
1,2, ... ,n - 1,0, 1,2, ... , n - 1, 0,

In the first case there is no integer n (n > 0) for which nl = 0 but in the second
case nI = 0. These two cases require to be appropriately distinguished.
Definition 4
Let R be a ring with an identity 1. R is said to have finite characteristic if there
exists an integer n (n > 0) such that
0= 1 + 1 + ... + 1 (n terms)
or, equivalently, that 0 = n1. Otherwise R is said to have infinite or zero

characteristic. (It may seem to be a trifle odd that in this situation the words
'infinite' and 'zero' are regarded as synonymous but such indeed is the case.)
Before relating this definition to integral domains we prove an interesting

ancillary result.
Lemma 2
Let R be a ring with an identity 1. Let R have finite characteristic and suppose
that nl = 0 (n > 0). Then for all elements a of R, na = o.
Rings 155
Proof
na =a +a+ + a (n terms)
= la + la + + la
= (1 + 1 + + l)a (distributivity)
= (nI)a = Oa = O. o
Our main interest is in the characteristics of integral domains and fields.
Theorem 6
Let D be an integral domain with identity 1 and of finite characteristic. Then
there exists a unique prime p such that pI = O.
Proof
By assumption there exists an integer n (n > 0) such that nl = O. Let p be
chosen to be the least such integer, pI = O. We claim that p is a prime. Suppose,
for the sake of argument, that p is not a prime and let p = PlP2 where
1 < Pl < P, 1 < P2 < P (P1lP2 EN).
Then
(Pl I)(P21) = (1 + 1 + ... + 1)(1 + 1 + ... + 1)
where we have Pl terms in the first bracket and P2 terms in the second. Expand-
ing by distributivity and collecting terms we have PlP2 terms of the form 11 = 1.
Thus
But D is an integral domain and so P l l = 0 or P2I = O. But either conclusion

contradicts the choice of P as least integer such that pI = O. Hence P is a prime
and is unique. 0
Definition 5
Let D be an integral domain with identity 1 and of finite characteristic. The least
integer P such that pI = 0 is a prime called the characteristic of D. D is said to be
of prime characteristic.
Examples 9
1. Z, Q, JR., C are of infinite characteristic. For each prime p, Zp is of prime
characteristic p.
2. In a field F of characteristic 2, addition and subtraction are the same! This
apparently paradoxical result arises as follows. We have 2a = 0 and so
a + a = O. But then a = -a. Hence b + a = b - a for all a, b E F.
3. A field which is finite necessarily has finite characteristic. But a finite field of
prime characteristic p need not have p elements. The following tables exhibit
a field E of characteristic 2 having four elements 0, 1, a, (3.
Sum 0 1 a (3 Product 0 1 a (3
0 0 1 a (3 0 0 0 0 0
1 1 0 (3 a 1 0 1 a (3
a a (3 0 1 a 0 a (3 1
(3 (3 a 1 0 (3 0 (3 1 a
We notice that F = {O, I} is a 'subfield' of E. The quadratic polynomial

x 2 + x + 1 does not factorize in F[x) but does factorize in E[x) since
x
2
+ x + I = x 2 + (a + (3)x + a(3 = (x + a)(x + (3).
We may also notice that E \ {O} = {I, a, (3} is multiplicatively a cyclic group
of order 3 generated by either a or (3.
Definition 6
Let F be a non-empty subset of the field E which is also a field under the addition
and multiplication in E. Then F is called a subfield of E.
Theorem 7 Subfield Criterion

A non-empty subset F of a field E is a subfield of E if and only if the following
axioms of a field are satisfied.
1. (i) F is closed under addition.
(iv) For all a E F, -a E F.
2. (i) F is closed under multiplication.
(iii) 1 E F.
(iv) For all a E F, a # 0, a-I E F.
Rings 157
Proof
Certainly if F is a subfield the axioms above must be satisfied. On the other hand
if the axioms above are satisfied then F is a subring and the remaining axioms for
a field are satisfied in F since they are satisfied in E. Hence we conclude that F is
a subfield of E. 0
In certain instances Theorem 7 may obviate some lengthy verifications of the

axioms for a field. If we know that an algebraic structure is a subset of a field we
may be able to establish more easily that the structure is actually a subfield.
Example 10
Let F = {a + bV21 a, b EQ}. Then F is a non-empty subset of the field R To
prove that F is a subfield we may utilize Theorem 7.
1. (i) Typical elements of F are a + bV2 and c + dV2 (a, b, e, dE Q) and their
sum is
+ bV2) + (e + dV2) = (a + e) + (b + d)V2
(a
which is of the form ql + q2V2 where Ql,Q2 E Q. Thus addition in F is
closed.
(iv) a + bV2 E F implies -(a + bV2) = (-a) + (-b)V2 E F.
2. (i) Let a, b, e, dE Q. Then
(a + bV2)(e + dV2) = (ae + 2bd) + (ad + be)V2 E F
since ae + 2bd, ad + be E Q. Thus multiplication in F is closed.
(iii) 1 = 1 + OV2 E F.
(iv) Let a + bV2 =I 0 (a, bE Q). Then a and b are not both zero. But then
a - bV2 =I 0 since a - bV2 = 0 would imply that a = b = 0 since we
have earlier shown that V2 rf. Q. Then
1 a - bV2
(a + bV2) -1 = = ---;:::------,=_
In
a + bv2 (a + bV2)(a - bV2)
a - bV2 a (-b) In
= a2 _ 2b2 = a2 _ 2b2 + a2 _ 2b2 v 2 E F.
Thus, finally, we conclude that F is a subfield of JR.
Integral domains of finite characteristic have a version of the well-known

binomial theorem which, on first encounter, is mildly amusing. We illustrate
this first by means of examples.
Examples 11
1. Let D be an integral domain of characteristic 2. Let a, bED. Then, as D is
commutative, we may expand (a + b)2 to give
(a + b)2 = a2 + ab + ba + b2 = a2 + 2ab + b2 = a2 + b2 since 2ab = O.
2. Let D be an integral domain of characteristic 3. Let a, bED. Then, as D is
commutative, we may expand (a + b)3 to give
(a + b)3 = a3 + 3a2b + 3ab2 + b3 = a3 + b3.
These examples should suggest a general result which we now prove.
Theorem 8
Let D be an integral domain of prime characteristic p. Let a, bED. Then
(a + b)P = aP+ bP.
Proof
We have to use the so-called binomial theorem, namely
(a + b)P = aP+ pap-1b + p(p ~ 1) a p- 2b2 + ... + bP.

1.
A typical term in this expansion, other than the first or the last, is
p(p - 1) (p - r + 1) ap-rbr
1.2 r
where 1 ~ r ~ p - 1. Now
p(p-1) (p-r+1)
1.2 r
is a strictly positive integer and so 1.2 r must divide p(p - 1) ... (p - r + 1).
But p is a prime and p> r so none of 1,2, ... , r can divide p but each
must divide the product (p - l)(p - 2) ... (p - r + 1). In consequence
(p - 1) ... (p - r + 1) . . h' .
1.2 ... r 18 an mteger. T us p dlVldes
p(p - 1) (p - r + 1)
1.2 r
Hence
p(p -1) ... (p - r + 1) ap-rbr = O.
:.....:.;;-~--=-....:.:..._-~
1.2 ... r
Rings 159
Thus, finally,
Corollary
Let D be an integral domain of prime characteristic p. Let al, <!2, •.. ,an be
elements of D. Then
Proof
A simple induction argument gives the result. Thus suppose
(al + a2 + ... + ak)P = ai + a~ + ... + a1'

for some k ~ 2. Then
(al + a2 + ... + ak+l)P = [(al + ... + ak) + ak+llP = (al + ... + ak)P + a1+l
= ai + a~ + ... + a1 + a1+l'
The result is immediate. o
We conclude this section with a result drawing together integral domains and
fields. A finite integral domain is a field but an infinite integral domain mayor
may not be a field. Now from the integral domain of the integers Z we obtain
the quotient field Q of Z by simply taking quotients of elements of Z to form
the elements of Qj in a similar way we may form the 'quotient field' of any
integral domain.
From the numbers 1, 2, 3, 4, we form quotients such as~, i, ~, ... where we
identify ~, ~, i, ...and 2, f, ~, etc. As we are long accustomed to make such
identifications no difficulty is experienced in their manipulation. On reflection
we realise that we are regarding ~, ~, i, ...
as equivalent under some unstated
equivalence relation. H, therefore, we are to consider an arbitrary integral
domain, we must make more precise the nature of an implicit equivalence rela-
tion. Instead of 'quotients' of two elements we begin more properly with ordered
pairs of elements and then we introduce an appropriate equivalence relation.
From a given integral domain D we construct a field F in which the integral
domain is 'embedded', that is there is a bijective mapping of D into F preserving
addition and multiplication, for example Z is 'embedded' in Q. (A more precise
definition of embedding will be given in a subsequent section.)
Theorem 9
Let D be an integral domain. Then D may be embedded in a field.
Proof
Let G = D x (D \ {O}). Then G is the set of ordered pairs (a, b) where a, bare
elements of D, b =I o. (We shall construct a field consisting of what, informally,
are qUotients~. The proof will perhaps be more easily comprehended if the reader
thinks of (a, b) as standing for ~.)
The proof is in three steps.
Step 1 We claim that the relation on G given by (a, b) (e, d) if and only if
{"oJ {"oJ
ad = be is an equivalence relation. Certainly (a, b) - (a, b) and if (a, b) (e, d) {"oJ
we have ad = be, giving cb = da and so (e, d) - (a, b). Suppose we have

(a,b)-(e,d) and (e,d) (e,f) then ad=be and ef=ed. Then using com-
{"oJ
mutativity and associativity we have

(af)d = a(fd) = a(df) = (ad)f = (be)f = beef) = b(ed) = (be)d.
Since D is an integral domain and d =I 0 we have af = be and so (a, b) - (e, f).
Step 2 We let F be the set of equivalence classes of G under We let (a, b)

{"oJ.
denote the equivalence class to which (a, b) belongs. We introduce an addition

and multiplication on F and claim that these operations are well-defined. We
have to show that if (a, b) = (~) and (c,d) = (cr;cF) then
(ad+be,bd) = (a'd' +b'e',b'd')
and
(ae,bd) = (a'e',b'd').
But
ab' = ba' and cd' = dc'
and so
(a'd' + b'e')(bd) = (a'd')(bd) + (b'e')(bd)
= (a'b)(dd') + (e'd)(bb')
= (ab')(dd') + (cd')(bb')
= (ad)(b'd') + (be)(b'd')
= (ad + be)(b'd')
Rings 161
and
(ac)(b'd') = (ab')(cd') = (ba')(dc') = (a'e')(bd).
Intuitively
from which
ad+ be a'd' + b'e' a e a' e'
bd = b'd' . Also b' d = b' . d' .
Hence (ad + be, bd) ,...., (a'd' + b'e', b'd') and (ae, bd) ,...., (a'e', b'd') giving the
result. Thus our definitions are well-founded.
Step 3 We prove that F is a field. Addition is closed and also associative since
[(a, b)(e, d)] + (e, I) = (ad + be, bd) + (e;])

= (ad! + be! + bde, bdl)
= (a, b) + (e! + de, dl)
= (a, b) + [(e, d)(e, I)].
The identity element is (1,1) = (x, x) for all XED, x =I- O. We have to show
that if (a, b) =I- (Q,1) then (a, b) has an inverse. Since (a, b) =I- (0,1) we have
a = a1 =I- bO = 0 and so (b, a) ED x (D \ {O}). Then (a, b)(b, a) = (ab, ab) =
(1,'1). Similarly (b, a)(a, b) = (1,'1). Multiplication is commutative. It may be
shown that the distributive laws hold and so F is a field.
We may also show that the subset of F consisting of the elements (x,l)
(x E D) is a sub-integral domain of D and that the mapping
x ~ (x,l) (x E D) [Intuitively x ~ T']

embeds D in F. o
Definition 7
The field F, constructed in Theorem 9, is called the quotient field of D.
Remark
From Z we construct Q 88 the quotient field of Z. From the polynomial domain
Q[x] we construct Q(x) (note round brackets) 88 the field of 'rational functions in
x' over Q.
Exercises 5.2
1. In Z7 find the roots of the following polynomials and factorize the
polynomials 88 fully 88 possible:
~+x+~ ~+L+~ ~+x+~ ~-L
2. In Z7 find the inverses of 2, 3 and 6-

3. In Z47 find the inverses of 3, 23, 24 and 32.
4. Over Zs solve, if possible, the following pairs of simultaneous equa-
tions.
2x+3y= I,
(i) _ _ _
3x+2y= 4.
x+3y= 2,
(ii) _ _ _
3x+2y = 2.
5. Over Z31 solve, if possible, the following simultaneous equations.
15x + 25y = 16,
8x+ 21y = 18.
6. For the field F, given 88 Example 9 no. 3 above and consisting of the
four elements 0, 1, a, (3, solve the following simultaneous equations.
ax+y= 1,
x + (3y = (3.
7. Prove that if R is a ring with an identity then the subset U(R) of
invertible elements of R is multiplicatively a group.
8. Let F be a field and x an indeterminate. What are the invertible
elements of F[x]?
9. Let S be the subset {a + bv'3la, bE Z} of JR. Prove that S is an

integral domain. Is S a field?
Rings 163
10. Let S be the subset {a + bv'51 a, b E Q} of JR. Prove that S is a field.

11. Let S be the subset {a + bv'6la,b E Q}. Is S a field?
12. Let D be an integral domain of prime characteristic p. Prove that, for
all a, bED and n E N,
Generalize.
13. Let Q be the set of 2 x 2 matrices of the form
a + ib c+ id)
( -c+ id (a, b, c, dE JR).
a - ib
Prove that Q is a ring which satisfies all the requirements of a field
except commutativity of multiplication. (Q is the ring of the quater-
mons, first discovered by W. R. Hamilton, 1805-65.)
5.3 Euclidean Domains

We have shown that the integral domain of the integers Z and the polynomial
domain Q[x] both have division and Euclidean algorithms. Here we study certain
integral domains which are assumed to admit a division algorithm from which a
Euclidean algorithm is naturally derivable.
Definition 8
Let D be an integral domain. Let there exist a mapping
v: D \ {O} -+ {O, 1,2, ...}
with the following two properties.

1. For all a, bED \ {O}, v(a) ~ (ab).
2. For all a, bED \ {O}, there exist q, rED such that a = bq + r where either
r = 0 or, if r t- 0, then v(r) < v(b).
Then D with the mapping v is called a Euclidean domain.
Examples 12
1. Z with the mapping v defined by v(a) = Ia I (a E Z, a t- 0) is a Euclidean
domain. To see this, we observe
(i) that
v(ab) = labl = lallbl ;::: lal = v(a) (a, bE Z,a, b:l 0)
since v(a) ;::: 1, and that
(ii) becomes simply a restatement of the division algorithm for Z.
2. F[x], where F is any field, with the mapping v defined by
v(J(x)) = deg f(x) (J(x) E F[x], f(x) :I 0)
is a Euclidean domain. To see this, we observe first that for
f(x),g(x) E FIx]' f(x) :I 0, g(x) :I 0,
v(J(x)g(x)) = deg (J(x)g(x)) = deg f(x) + deg g(x) ;::: deg f(x) = v(J(x)).
Again 2, in Definition 9, is the division algorithm for FIx]. Strictly we have

only discussed the division algorithm for polynomial domains over Q or IR but
the arguments previously employed for these particular fields easily adapt to
any field. We illustrate some of these arguments in the next example.
Example 13
Consider Z&lx]. Let a(x) = 3x4 + 2x3 + 2x 2 + X + 3 and b(a) = x 2 + 2. Suppose
we want to find q(x), r(x) E Z&[x] as in the definition of a Euclidean domain. We
perform a long division as follows:
3x2 +2x+ I
x 2 + 2)3x4 + 2xa + 2x 2 + x + 3
3x4 + x2
2x +x2 +x
3
2x3 +4x
2x+l
Thus q(x) = 3x2 + 2x + 1, r(x) = 2x + 1.
Before considering other Euclidean domains we prove a convenient lemma.

Recall that an element of a ring is a unit if and only if it is invertible.
Rings 165
Lemma 3
Let D be a Euclidean domain with identity 1, 88 in Definition 8.
1. For all a E D \ {O}, v(l) :5 v(a).
2. For a E D \ {O}, v(l) = v(a) if and only if a is a unit of D.
Proof
1. v(l) :5 v(la) = v(a) (a E D, a =J 0).
2. Suppose a is a unit. Then there exists bED such that ab = 1. Then we have
v(a) :5 v(ab) = v(l) :5 v(a) and so v(l) = v(a).
Conversely suppose v(a) = v( 1). There exist q, rED such that 1 = aq + r, where
if r =J 0 we have v(r) < v(a) = v(l) which is false and so r = O. But, then, aq = 1
and a is a unit. 0
Our next example of a Euclidean domain occurs 88 a subintegral domain of the

field of complex numbers.
We recall that for complex numbers Zl and Z2 we have
IZl z21 = Izd Iz21,
IZI + z21 :5lzd + IZ21·
Lemma 4
Let Z[i] denote the set of complex numbers of the form m + ni (m, n E Z). Then
Z[i] is an integral domain.
Proof
Z[i] is a non-empty subset of the field C. Now for ml, m2, nl, ~ E Z
(ml + n1i) + (~+ n2 i) = (ml + m2) + (nl + n2)i E Z[i],
-(ml + n1i) = (-ml) + (-ndi E Z[i],
(ml + nli)(~ + n2i) = (mlm2 - nl~) + (mln2 + nl~)i E Z[i].
The remaining axioms for an integral domain are evidently satisfied. 0
Definition 9
A complex number of the form m + ni (m, n E Z) is called a Gaussian integer
(after K.F. Gauss).
Theorem 10
The Gaussian integers Z[i] form a Euclidean domain under the mapping v given
by
v(a) = lal 2 (a E Z[i],a::l 0).
Proof
1. We want to prove that
v(a) ::; v(ab) (a, b, E Z[i] \ {O}).
We note first that b = b1 + b2i where b1 , b2 E Z and so v(b) = Ibl =

2 2 2 2
b~ + ~ ~ 1. Hence v(a) = lal ::; lal 1bl = labl = v(ab).
2. Given a, bE Z[i] \ {O}, we have to find suitable q, r such that a = bq + r. Now
~ is a complex number which must be of the form a + {3i where a, {3 E Q.
We now use a fairly obvious property of Q. Since a must lie between two
consecutive integers, we may therefore choose that integer m for which
! ¥
a = m + c, Ic I < (for example, a = lies between 3 and 4 and we choose
m = 4, c = -~.) Similarly we choose n so that {3 = n + 1], 11] I <!. Let
q = m + ni. Then q E Z[i]. Let r = a - bq. Then r E Z[i]. Thus
a
b = a + {3i = (m + c) + (n + 1])i = (m + ni) + (c + 1]i) = q + (c + 7]i)
and hence
r =a- bq = bq + b(c + 7]i) - bq = b(c + 7]i).
If r = 0 the appropriate q, r have been found. If r ::I 0 then
v(r) = Ib(c + 7]i) 12 = Ibl 21c + 7]i1 2 = Ib1 2(c 2 + 7]2) ::; Ib l2 G+~)
= ~ Ibl 2 < Ibl 2 = v(b).
Hence we have established that Z[i] is a Euclidean domain. o

Rings 167
Example 14
Let a = 5 + 6i, b = 2 - 3i. Suppose we have to find suitable q, r E Z[i] such that
a = bq + r where either r = 0 or if r oF 0 then Irl < Ibl = ";22 + 32 = Vi3. Now
a S + 6i (S + 6i)(2 + 3i) -8 + 27i 8 27 .

b = 2 - 3i = (2 - 3i)(2 + 3i) = 13 = - 13 + 13 1
Choosing the integers nearest to -/3 H
and we obtain, respectively, -1 and 2.
Let q=-1+2i and r=a-bq=(S+6i)-(2-3i)(-1+2i)=(S+6i)-
(4 + 7i) = 1 - i. We note that Irl = ";1 2 + 12 = J2 < Vi3. Thus we have
found an appropriate q, r. However, we may also note that q, r are not uniquely
determined, for if q' = 2i and r' = a - bq' = (S + 6i) - (2 - 3i)(2i) = -1 + 2i
then Ir'l = V5 and so q' and r' also satisfy the required conditions.
Exercises 5.3
1. Find the units of the Euclidean domain Z[i].
2. In the Euclidean domain Z[i] for given a, b E Z[i] find q, r E Z[i] such
that a = bq + r where 0 ~ Irl < Ibl:
(i) a = 1 + 13i, b = 4 - 3i,
(ii) a = S + lSi, b = 7 + i,
(iii) a = S + 6i, b = 2 - 3i.
3. Let D = {a + ,BV2la,,B E Z}. Prove that D is a Euclidean domain if

v: D \ {O} -... {O, 1,2, ...} is given by
v(a + ,Bv'2) = I(a - ,BV2)(a + ,BV2) I = Ia 2 - 2,B21 (a,,B E Z).
4. (Hard) Let w be a non-trivial complex cube root of 1. Then w 3 = 1,

w oF 1 and so w2 + w + 1 = 0 and the complex conjugate of w is
2 1
iii = w = -. (For the considerations of this example it is best to ignore
w -1±iV3
the fact that w = 2 .)
Prove that D = {a + ,Bwla,,B Ell} is a Euclidean domain if
v: D \ {O} -... {O, 1, 2, ...} is given by
v(a + ,Bw) = la + ,Bw1 2 = (a + ,Bw)(a + ,BW).

5.4 Ideals and Homomorphisms

P. Fermat (1601-65), in commenting on the 'Arithmetica' of Diophantus (c. 250
AD), formulated a statement, of which he claimed to have a proof, that the
equation,
had no solutions for non-zero integers x, y, z. This celebrated assertion, known

as Fermat's Last Theorem, has in fact been established as correct owing to
the recent work of A. Wiles (1953- ). In the 19th century, however, its proof
was attempted by many mathematicians among whom was E.E. Kummer
(1810-93) who tried heroically to establish a complete result. Arising from his
efforts he created a theory of 'ideal numbers' but it fell to Dedekind to introduce
'ideals', albeit in a number-theoretic and so commutative context. For our
purposes we shall give a general definition of an 'ideal' appropriate to a possibly
non-commutative context.
Definition 10
Let R be a ring and let I be an additive subgroup of R.
1. If for all x E Rand for all a E I, itfollows that xa E I, then I is said to be a left
ideal of R.
2. If for all x E R and for all a E I, it follows that ax E I, then I is said to be a
right ideal of R.
3. If for all x E R and for all a E I, it follows that xa E I and ax E I, then I is
said to be a two-sided ideal, or briefly, an ideal of R.
R and {O} are immediately ideals of R, and {O} is called the zero ideal of R. If
I satisfies 1, 2 or 3 and I :f. {O}, I:f. R, then I is said to be proper.
By definition every left or right ideal of R is a subring of R but not every sub-
ring of R is necessarily a left or right ideal. Every element of R gives rise to a left
or right ideal which may not, however, be proper (see Theorem 11).
Examples 15
1. Let n E Z. Then nZ = {nxlx E Z}, the subring of those integers divisible by
n, is an ideal of Z. (In a commutative ring every left or right ideal is also a two-
sided ideal.)
Rings 169
2. Let R be the ring M 2 (Q) of all 2 x 2 matrices over Q. Let
Then L is eagily shown to be an additive subgroup of R and L is also a left

ideal since, for x, y, Z, t,p, q E Q,
(
X
Z
Y)t (00 P)
q
= (0
0
x
p yq
+ ) E
zp+tq
L.
Lemma 5
1. Let R be a ring with an identity 1 and let / be a non-zero left ideal of R.
If / contains a unit of R then I = R.
2. A field hag no proper left or right ideals.
Proof
1. Let u be a unit of R such that u E I. Then there exists v E R such that vu = 1.
Let x E R. Then x = xl = x(va) = (xv)u E Isince/is an ideal. Hence / = R.
2. Every non-zero element of a field is a unit. Hence the result follows. 0
Example 16
We have seen, in the Example above, that M 2 (Q) hag at leagt one proper left
ideal. We wish to show that M 2 (Q) hag no proper two-sided ideals. Suppose
/ is a non-zero ideal of M 2 (Q). Let
be a non-zero element of I. Then certainly one of a, b, c, d is non-zero. For the

sake of argument we suppose a =F O. (The reader should consider in turn the
Cages of non-zero b, c, d and modify the argument to follow.)
Since / is an ideal
is an element of I and so also is
Similarly
belong to I. Hence, for x, y, Z, t E Q,
(: ~)=(~ :)(~ ~)+(~ :)(~ ~)

+(~ :)(~ ~)+(~ ~)(~ ~)EI.
The next theorem, which may be passed over on a first reading, brings
together, nevertheless, some useful facts and, at the same time, serves to define
some convenient notation.
Theorem 11
Let R be a ring.
1. Let a E R. Then Ra defined by
Ra = {xalx E R}
is a left ideal of R.
2. Let L 1 , L 2 , ••• , Ln be left ideals of R. Then L 1 + L 2 + ... + L n defined by
L 1 + L 2 + ... + L n = {al + ~ + ... + ~ lai ELi> i = 1,2, ... , n}
3. Let a.. a2," . ,an E R. Then
Ra1 + ~ + ... + Ran = {Xl a1 + X2~ + ... + xnan IXi E R, i = 1,2, ... , n}
Rings 171
Proof
1. 0 = Oa E Ra, and for all Xl> X2, Y E R we have
xla + X2a = (Xl + x2)a ERa, -(xla) = (-xI)a ERa.
Thus Ra is a left ideal.
2. We prove the result for two left ideals A and B. The general result is obtained
by a simple induction. A + B = {a + bla E A, bE B} and A + B is an addi-
tive subgroup of R. Let X E R, a E A, b E B. Then we have x(a + b) =
xa + xb E A + B since A, B are left ideals. Thus A + B is a left ideal.
3. This follows from 1 and 2 above. 0
In our considerations of groups we found that normal subgroups of a group

and the homomorphisms of a group were important and directly related con-
cepts. A similar relationship will be seen to exist between ideals of a ring and
'homomorphisms' of a ring.
Definition 11
Let R and S be rings and let I : R -+ S be a mapping such that
I(x + y) = I(x) + I(y), I(xy) = l(x)/(y) (for all x, y E R).
Then I is said to be a homomorphism from R to S.
The homomorphism I, as defined above, preserves the additive and multi-
plicative structure of R in its image I(R) in T. Thus I is a homomorphism of
R considered as an additive Abelian group and considered as a multiplicative
semigroup.
Examples 17
1. Let n be a given integer, n> o. Then the mapping I: Z -+ Zn given by
I(a) = a
where a, the equivalence class to which a belongs (a E Z), is a homomorphism
since we know that
a+b=a+b,
ab =ab.
2. Let F[x] be a polynomial ring over a field F and let a E F. Then let us define
I : F[x] -+ F by
I: a(x) -+ a(a) (a(x) E F[x]).
In other words, f replaces a polynomial by the result of substituting a for x in

the polynomial. But we know that
a(x) + b(x) = c(x), a(x)b(x) = d(x) (a(x), b(x), c(x), d(x) E F[x])
implies that
a(a) + b(a) = c(a), a(a)b(a) = d(a)
and so f is a homomorphism.
Theorem 12
Let R and S be rings and let f : R -+ S be a homomorphism. Then the following
statements hold.
1. H OR and Os are the zero elements of R and S respectively then f(OR) = Os'
2. For all x E R, - f(x) = f( -x) where - f(x) is the inverse with respect to
addition of f(x) in S and -x is the inverse with respect to addition of x in R.
3. f(R) is a subring of S.
4. Let K = {x E Rlf(x) = Os}, Os defined in 1. Then K is an ideal of R.
Proof
1 and 2 hold since f is a homomorphism of R into S where both rings are con-
sidered solely as additive groups. Furthermore, by the same reasoning, f(R) is
an additive subgroup of S. To prove that f(R) is indeed a subring of S we
have merely to establish multiplicative closure of f(R) in S and this occurs
since f is a homomorphism from R to S considered as multiplicative semigroups.
Hence, finally, we conclude that f(R) is a subring of S.
To prove 4, we observe that K is the kernel of the homomorphism f from the
additive group R to the additive group S. Let now a E K, x E R. Then
f(xa) = f(x)f(a) = f(x)Os = Os,
f(ax) = f(a)f(x) = Osf(x) = Os·
Hence xa, ax E K and so K is an ideal. o
Definition 12
Let R and S be rings and let f: R -+ S be a homomorphism. Then
K = {x E Rlf(x) = Os} is an ideal of R which is called the kernel of f, written
Ker f.
Rings 173
We now come to several definitions and results which correspond to definitions

and results previously considered in Chapter 4. As the results and proofs are so
similar, those to be provided here will be less detailed than usual. The reader
should, however, ensure that he or she fully understands the arguments.
We recall that an ideal I of a ring R is, in fact, an additive subgroup of the
additive group R and so we have cosets of I in R where the coset containing
xERis
x +I = {x + a Ia E I}.
Lemma 6
Let Rand S be rings and let f : R -+ S be a homomorphism with kernel K. Let
x and y be elements of R. Then f(x) = f(y) if and only if x + K = y + K.
Proof
We have shown previously that f(x) = f(y) if and only if the cosets x + K and
y + K are equal. 0
Theorem 13
Let I be an ideal of the ring R. Let RjI denote the set of cosets of I in R. Two
binary operations are defined on Rj I as follows:
Let x + I and y + I be cosets of I in R and define
(x + 1) + (y + 1) = (x + y) + I,
(x + 1)(y + 1) = xVI.
Then the binary operations are well-defined and under these operations Rj I is
a ring. Further the mapping R -+ RjI given by x -+ x + I (x E R) is a homo-
morphism of R onto Rj I with kernel I.
Proof
Certainly under the definition of addition Rj I is an additive group. We have to
show that multiplication is well-defined in RjI. Thus we have to show that if
x + I = x' + I (x, x' E R) and y + I = y' + I (y, y' E R) then xy + I =
x'y' + I. But certainly x + I = x' + I implies that x' = x + a for some a E I,
also y' = y + b for some bEl. Then x'y' = (x + a)(y + b) = xy + ay + xb + abo
But I is an ideal and hence we have ay + xb + ab E I. Thus x'y' + I =
xy + (ay + xb + ab) + 1= xy + I. Our definition of multiplication is well
founded.
To prove the associativity of multiplication we have for all x, y, z E R

[(x + 1)(y + 1)](z + 1) = (xy + 1)(z + 1)
= (xy)z+I
= x(yz) + I
= (x + 1)(yz+ 1)
= (x + I)[(y + 1)(z + 1)].
The distribution laws may be proved and we conclude that RII is a ring.
The mapping I: R -+ RII given by I(x) = x + I (x E R) is a homomorphism
of the additive group R onto RII. For x, y E R we have
I(xy) = xy + I = (x + 1)(y + 1) = l(x)/(y)
and so I is a ring homomorphism. Finally
Ker/={xERI/(x)=I}={XERlx+I=I}=I. 0
Definition 13
Let I be an ideal of the ring R. The ring RII, as considered above, is said to be a
factor-ring of R.
Example 18
For each nEZ, nZ = {nxlx E Z} is an idealofZ and the factor-ring ZlnZ is the
ring which we have called Zn.
Definition 14
Let R and S be rings and let I : R -+ S be a homomorphism. H I is surjective
then I is said to be an epimorphism. H I is injective then I is said to be a mono-
morphism. If I is bijective then I is said to be an isomorphism.
Remark
In Theorem 8 we had an 'embedding', as it was called, of an integral domain D
into its quotient field F. By an embedding we now understand an isomorphism
and we have therefore already shown that there exists an isomorphic copy of D
contained in F. In this instance we may identify D with its isomorphic copy and
so regard D as a su~integral domain of F, just as we regard Z as embedded in,
and a sub-integral domain of, Q.
Rings 175
Theorem 14 First Isomorphism Theorem

Let Rand 8 be rings and let f : R -+ 8 be an epimorphism with kernel K. Then
R/K and 8 are isomorphic.
Proof
K is an ideal of R and so we have an epimorphism R -+ R/ K. We now define
h: R/K-+ 8by
h(x + K) = f(x) (x E R).
By the proof of the First Isomorphism Theorem for groups h is a well-defined
mapping which is, in fact, an isomorphism between the additive groups R/K
and 8. We merely have to show that h preserves multiplication. Let x,
y E R, then we have h[(x + K)(y + K)] = h[xy + K] = f(xy) = f(x)f(y) =
h(x + K)h(y + K). The proof is finally complete. 0
Lemma 7
Let R be a ring. Let 8 be a subring of R and let I be an ideal of R. Then
8 + I = {x + a Ix E 8, a E I} is a subring of R and I n 8 is an ideal of 8.
Proof
We leave this proof to the Exercises. o
Theorem 15 Second Isomorphism Theorem

Let R be a ring. Let 8 be a subring of R and let I be an ideal of R. Then the
factor-rings (8 + 1)/1 and 8/(1 n 8) are isomorphic.
Proof
We define p: 8 -+ (8 + 1)/1 by
p(x) =x +I (x E 8).
Then p is an epimorphism. H K is the kernel of p then, by the First Isomorphism
Theorem, 8/K is isomorphic to (8 + 1)/1.
On proving that K = I n 8 the proof is complete. 0
Exercises 5.4
1. Prove that A ={ (~ ~) Ix, t E Q} is a right ideal of M 2 (Q).
2. Let L 1 and L 2 be left ideals of a ring R. Prove that L 1 n L 2 is a left
ideal of R. Generalize.
3. Let L 1 ~ L2 ~ L3 ~ ••• be a countable ascending chain of left ideals
UL
00
of a ring R. Prove that i is a left ideal of R.

i=1
4. Let X be a non-empty subset of a ring R. Let
A = {a E R Ixa = 0 for all x E X}.
Prove that A is a right ideal of R.
5. Let R be a ring not necessarily possessing an identity. Let a E R.

Prove that {na + xaln E Z, x E R} is a left ideal of R containing a.
6. Let Rand 5 be rings and let f : R -+ 5 be a homomorphism. Let L be

a left ideal of 5. Prove that {x E Rlf(x) E L} is a left ideal of R.
7. Let Rand 5 be rings and let f : R -+ 5 be an epimorphism. Let L be a

left ideal of R. Prove that f(L) is a left ideal of 5.
8. Let R be a ring and let 5 be a subring of R. Let I be an ideal of R.

Prove that 5 + I = {x + alx E 5, a E I} is a subring of R and
I n 5 is an ideal of 5.
9. Let R,5 and T be rings and let f : R -+ 5 and 9 : 5 -+ T be given

homomorphisms. Prove that go f is a homomorphism from R to T.
10. Let R be the ring of 2 x 2 matrices of the form
a
( -b b)
a
(a, bE R).
Prove that the mapping
a + Ib-+
. ( a (a,b E R)
-b
is an isomorphism of C with R.
Rings 177
5.5 Principal Ideal and Unique Factorization Domajns

We here continue our study of integral domains with a view to obtaining proper-
ties very similar to those of the integral domain of the integers Z.
Definition 15
Let D be an integral domain. Let a, bED, b f. O. Then b is said to be a divisor of
a if there exists e E D such that a = be.
We recall that u is a unit of a ring R if there exists v E R such that uv = vu = 1.

Definition 16
Let D be an integral domain. Let a, bED. Then a is an associate of b if there
exists a unit u in D such that a = bu.
Notice that if a is an associate of b then b is an associate of a for if a = bu where

uv = vu = 1 then av = buv = b1 = b. An element u is a unit if and only if u is an
associate of 1.
We further recall that if R is a ring and a E R then Ra = {xa Ia E R} and if, in
particular, R has an identity 1 then a = 1a E Ra.
Lemma 8
Let D be an integral domain. Let a, bED. Then b divides a if and only if
Da~ Db.
Proof
If b divides a then a = eb for some e E D. Then Da = Deb ~ Db. Conversely if
Da ~ Db then a E Da ~ Db and so there exists e E D such that a = eb. 0
Lemma 9
Let D be an integral domain. Let a, bED. The following statements are
equivalent.
1. Da= Db.
2. a divides b and b divides a.
3. a and b are associates.
4. a = bu for some unit u of D.
Proof
By Lemma 8, 1 and 2 are equivalent and by remarks above 3 and 4 are equiva-
lent. We prove the equivalence of 2 and 3.
If a divides b and b divides a there exist c, d E D such that b = ac, a = bd. Then
a = bd = acd from which, as D is an integral domain, 1 = cd and so d and c are
units. Hence a and b are associates. On the other hand if a and b are associates
there exist units u, v E D such that a = bu and b = avo Hence b divides a and
a divides b. This completes the proof. 0
Examples 19
1. Two integers m, n are associates in Z if and only if m = ±n. In Z Lemma 9 is
obviously true.
2. Let F be a field. The units of F[x], the polynomial ring of F over x, are the
non-zero elements of F. Therefore f(x),g(x) E F[x] are associates if and
only if f(x) = cg(x) where c E F, C:f: O.
Definition 17
Let D be an integral domain. A non-zero element p of D, which is not a unit, is
said to be irreducible if whenever a E D and a divides p then a is an associate of
lor p.
Definition 18
Let D be an integral domain. A non-zero element p of D, which is not a unit,
is said to be a prime if whenever p divides ab (a, bED) then p divides a or p
divides b.
For our convenience we make the following definition for a restricted class of
rings; we content ourselves with remarking that, with appropriate modification,
the restrictions in the definition may be removed.
Definition 19
Let R be a commutative ring with an identity.
1. An ideal P of R, P :f: R, is said to be prime if whenever a, b E Rand ab E P,
then either a E P or b E P.
2. An ideal M of R, M :f: R, is said to be maximal if whenever M ~ I and I ~ R
for any ideal I, then either M = I or I = R.
Rings 179
Theorem 16
Let R be a commutative ring with an identity.
1. An ideal P of R is prime if and only if RjP is an integral domain.

2. An ideal M of R is maximal if and only if RjM is a field.
Proof
We note that the factor-ring in 1 and 2 is, at least, a commutative ring with an
identity 1.
1. Let P be prime. Let a, bE R be such that (a + P)(b + P) = P. Then we have

ab + P = P and so ab E P. As P is prime a E P or b E P which implies
a + P = P or b + P = P. Hence, RjP has no proper divisors of zero and so
is an integral domain.
Let now RjP be an integral domain. Let a, b E R be such that ab E P. Then
(a + P)(b + P) = ab + P = P and so we have divisors of zero in RjP. Since
RjP is an integral domain a + P = P or b + P = P from which a E P or
b E P. Hence P is prime.
2. Let M be maximal. Let a E R be such that a + M :F M. Then a ¢ M and
so the ideal Ra + M contains a and M and so M c Ra + M. Since M is
a maximal ideal Ra + M = R. Then there exist x E R and m E M such
that
xa+m = 1.
Hence
1 +M= (xa+m) +M= xa+m+M= xa+M = (x+M)(a+M).
Thus x + M is the inverse of a + M. Hence RjM is a field.
Let now RjM be a field. Let I be an ideal such that M ~ I ~ R. IT M :F I
there exists a E I, a ¢ M. Then a + M:F M and so, as RjM is a field, a + M
is invertible in RjM and there exists bE R such that (a + M)(b + M) =
1 + M. Therefore ab = 1 + m for some m E M. Hence, as a E I, m E M and
M ~ I we have 1 E I. But then I = R. 0
Theorem 17
Let D be an integral domain. Then a prime is also an irreducible element.
Proof
Let P be a prime in D. Let a E D and suppose a divides p. We wish to show that a
is an associate of lor p. We have P = ab for some bED. Since p is prime p divides
ab and so either p divides a or p divides b. IT p divides a then a = pc for some
c E D and so p = ab = pcb and 1 = cb. Thus b is a unit and p and a are associates.
On the other hand, if p divides b then b = dp for some d E D and so we
have p = ab = adp and 1 = ad. Thus a is an associate of 1. This completes the
proof. 0
In the integral domain Z every irreducible element is also prime. In analogy

with Z we wish to investigate those integral domains in which every irreducible
element is prime. To facilitate our investigation it is convenient to make a defi-
nition which is obviously modelled on the Fundamental Theorem of Arithmetic
inZ.
Definition 20
Let D be an integral domain in which every non-zero element, which is not a unit,
is expressible as a finite product of irreducible elements. Furthermore, whenever
a non-zero element x of D which is not a unit, is written as
x = PIP2" ·Pm = qlq2'" qn
where P1>P2, ... ,Pm; Ql,Q2,·'" Qn are irreducible elements then m = n and, with
a suitable reordering if necessary, Pi and qi are associates (i = 1,2, ... , n). Then
D is said to be a unique factorization domain (U.F.D.).
By the Fundamental Theorem of Arithmetic Z is a U.F.D.
Theorem 18
In a unique factorization domain every irreducible element is prime.
Proof
Let D be a U.F.D. Let x be an irreducible element of D. Suppose x divides ab
(a, bED). We have to show that x divides a or x divides b. By supposition
there exists y E D such that
xy = abo
Rings 181
Now y, a and b are products of irreducible elements, say,

a = PIP2 .. ·Pm, b = qlq2'" qn, Y = TIT2'" Tt,
where Pl.P2,.·· ,Pm; ql, q2,···, qn; TI' T2,···, Tt are irreducible elements. Then
XTIT2'" Tt = PIP2" ·Pmqlq2··· qn'
But these are two factorizations into irreducible elements and so, as D is a
U.F.D., we must have 1 + t = m + n and, more importantly, every irreducible
element on the left-hand side must be an associate of an irreducible element
on the right-hand side. Thus x is an associate of some Pi or some Qj. But this
implies that x divides a or x divides b. 0
Definition 21
Let D be an integral domain. Let a, b, c be non-zero elements of D. Then c is said
to be a common divisor of a and b if c divides a and c divides b. A common divisor
d, which is itself divisible by any other common divisor, is called a greatest
common divisor (G.C.D.).
The terminology of this definition corresponds to the terminology of Defini-

itions 3 and 4 of Chapter 2 in the case of the integers. As in that case, we may
show that in a U.F.D. any two non-zero elements have a G.C.D. We defer the
discussion of a G.C.D. until after we have discussed 'principal ideal rings'.
Definition 22
Let D be an integral domain. Let a ED. Then Da = {xalx ED} is said to be a
principal ideal. An integral domain in which every ideal is principal is called a
principal ideal domain (P.LD.).
Lemma 10
Let D be a principal ideal domain. Any two non-zero elements a and b of D have
a G.C.D. d given by
Da+Db= Dd.
Proof
Da and Db are ideals of D and so Da + Db is also an ideal. Hence as D is a P.I.D.
there exists dE D such that Da + Db = Dd. We claim that d is indeed a G.C.D.
of a and b. Certainly Da ~ Dd and so d divides a and similarly d divides b. Thus
d is a common divisor of a and b. Suppose c E D is also a common divisor of a

and b. Now d E Da + Db and so there exist x, y E D such that d = xa + yb.
But now if c divides a and c divides b we must have that c divides d. Thus d is
a G.C.D. of a and b. 0
We now relate our earlier discussion of Euclidean domains to principal ideal

domains.
Theorem 19
A Euclidean domain is a principal ideal domain.
Proof
Let D be a Euclidean domain with mapping v: D \ {O} -+ {O, 1, 2,}. Let I be an
ideal of D. We wish to prove that I is a principal ideal. IT I = {O} the result is
true so suppose If. {O}. Choose an element b of I for which, amongst all the
non-zero elements of I, v(b) is as small as possible. As I is an ideal Db ~ I.
We shall show that Db = I. Let a E I. As D is a Euclidean domain there exist
q, rED such that
a=bq+r
where either r = 0 or, if r f. 0, then v(r) < v(b). But, as I is an ideal,
r =a - bq E I.
But if r f. 0 we have v( r) < v( b) and this contradicts the choice of b as being such
that v(b) is as small as possible. Thus r = 0 and a = bq E Db. Hence, finally,
Db=I. 0
Examples 20
1. Z is a Euclidean domain and we now know that every ideal of Z is of the form
nZ = {nxlx E Z} for some n.
2. The polynomial ring F[x] , where F is a field, is a Euclidean domain and so is a
P.I.D.
Theorem 20
In a principal ideal domain every irreducible element is prime.
Rings 183
Proof
Let D be a P.I.D. Let p be an irreducible element of D. Let p divide ab where
a, bED. We wish to show that p divides a or p divides b.
Suppose p does not divide a. Let c E D be such that c divides p and c divides a.
Since p is irreducible c is a unit or an associate of p. IT c is an associate of p then as
c divides a so also does p divide a which is false. Hence c is a unit. Hence, by
Lemma 10,
Da+Dp=Dd
where d is necessarily a unit and so Dd = D, giving Da + Dp = D. Hence there
exist x, y E D such that
xa + yp = 1.
But then
xab+ypb =b
from which p divides b. Hence p is a prime. o
Corollary
Let D be a principal ideal domain. Let p be an irreducible element of D.
Let al>~,"" a,. E D and suppose p divides al <l2 .•• a,.. Then for some i
(1 ~ i :$; n), p divides G.i.
Proof
Either p divides al or P divides a2a3'" an' A simple induction proves the
reswt. 0
Our aim is to prove that a P.I.D. is necessarily a U.F.D. We are thereby

enabled to deduce that a Euclidean domain is a U.F.D. At a first reading the
reader may wish to confine himself or herself to the fact of the deduction and
to avoid the necessary proofs. We have a lengthy lemma before the main
theorem.
Lemma 11
Let D be a principal ideal domain. Let a be a non-zero element of D which is not
a unit of D. Then a is the product of a finite number of irreducible elements.
Proof
We argue by contradiction. Suppose therefore that a is not the finite product of
irreducible elements.
Since a cannot itself be irreducible, a is divisible by aI, say, al ED, where al is
neither a unit nor an associate of a. Hence there exists bl E D which is also
neither a unit nor an associate of a, such that
a = alb l .
and since al is not an associate of a we have

Da C Dal'
Now either al or bl is not a finite product of irreducible elements. Suppose al is
not a product of irreducible elements. We apply the same argument to al which
is not a product of irreducible elements and we obtain tl2 dividing al where
Dal C Da2
and where tl2 is not a product of irreducible elements. We continue this process to
obtain a strictly ascending chain of ideals Dan (n E N) such that
Da C Dal C Dtl2 C ...
where each an (n E N) is not a product of irreducible elements. Let
U Dan·
00
1=
n=l
Then I is an ideal of D (see Exercises 5.4, no. 3) and so I = Dc for some c E D.

But c must belong to some subset of the infinite union and so there must exist k
such that c E DCk' But then
I = Dc <; Dak <; I
and so I = Dak' But this implies that Dak = Dak+ 1 = ... and this contradicts
the fact that the ideals in the chain are·distinct. Hence our original supposition
is false and a is indeed a finite product of irreducible elements. 0
Theorem 21
A principal ideal domain is also a unique factorization domain.
Rings 185
Proof
Let D be a P.I.D. By Lemma 11 we have shown that every non-zero element of D
which is not a unit is a finite product of irreducible elements. We have to show
that such a product is essentially unique. Let a E D and let a be expressed as
where PI,P2, ... ,Pm; ql, q2,"" qn are irreducible elements of D. Then PI divides
ql q2 ... qn and so, by the Corollary above, PI divides one of q1> q2, ... , qn'
By renumbering, if necessary, we may suppose PI divides ql' Then PI and ql
are irreducible elements which are associates and so ql = UPI where U is a unit
ofD.
Then
which implies that

P2P3 ... Pm = uq2q3 ... qn = q2q3 ... qn
I
where q~ = uq2 is an irreducible element. A simple induction now proves the

desired result. 0
Corollary
A Euclidean domain is also a unique factorization domain.
Exercises 5.5
1. Let R be a commutative ring with an identity. Prove that a maximal
ideal of R is a. prime ideal. Give an example of a ring R and a prime
ideal of R which is not maximal in R.
2. Z[i] is the ring of Gaussian integers. Find the G.C.D. din Z[i] of 4 + 2i
and 1-3i and write d in the form
(4 + 2i)ZI + (1 - 3i)z2 =d
for appropriate Zl, Z2 E Z[i].
3. Let D be a principal ideal domain. Prove that any n non-zero elements
ai, a2,' .. ,~ of D have a G.C.D. d given by
Dal + D~ + ... + Dan = Dd.
5.6 Factorization in Q[x]

Given a non-zero polynomial I(x) E Q[x] we want to devise a criterion by which
the polynomial is a prime, or is equivalently irreducible, in Q[x]. We remark first
that any such polynomial I(x) may be written as
1
I(x) = - g(x)
c
where g(x) E Z[x] and c E Z.
Example 21
It suffices therefore to consider polynomials which are in Z[x] but to attempt

their factorization in Q[x]. We aim to show that if a polynomial in Z[x] factors
non-trivially in Q[x] then it already has a non-trivial factorization in Z[x].
From this fact we infer that if a polynomial in Z[x] does not factor non-trivially
in Z[x] then it does not factor non-trivially in Q[x].
Definition 23
Let I(x) be a polynomial in Z[x]. Then the greatest common divisor of the
coefficients of I(x) is called the content of I(x). A polynomial of content 1 is
said to be primitive.
Let I(x) E Q[x], I(x) #- O. Then we may write
d
I(x) = - hex)
c
where c, d are integers, hex) E Z[x] and hex) has content 1.
We illustrate this assertion by means of Examples.
Examples 22
14 28 21 1
1. I(x) = _x 2 +-x+-= - [84x 2 + 280x +315]
5 3 2 2.3.5
7 2 7
= 30 [12x + 40x + 45] = 30 hex),
Rings 187
where h(x) has content 1.

5 3 1 2 11 1
2. f(x)=12 x +gX +6"x+ 14
1
= 8.3.7 [70x3 + 21x2 + 308x + 12]
1
= 168 h(x),
where h(x) has content 1.
Lemma 12
Let f(x) and g(x) be primitive polynomials in Z[x]. Suppose there exist
Cb~ E Z, Cl:l: 0, ~:I: 0, such that cd(x) = ~g(x). Then Cl = ±~ and
f(x) = ±g(x).
Proof
Let f(x) = ao + alx + + anxn (an:l: 0). Then the G.C.D. of ao,ab'" ,an is 1
and so there exist to, t l , ,tn E Z such that
toao + tlal + ... + tnan = 1.
Since cd(x) = ~g(x), ~ divides clab"" clan and SO ~ divides
CIao,
tOClao + tlclal + ... + tnCnan = CI(tOao + tlal + ... + tnan ) = CI'

Similarly CI divides ~. Thus CI = ±~ and SO f(x) = ±g(x). o
The next lemma holds the key to the use of primitive polynomials.
Lemma 13 Gauss's Lemma

Let f(x) and g(x) be primitive polynomials in Z[x]. Then f(x)g(x) is primitive.
Proof
Let f(x) = ao + alx + + amxm (am:l: 0),
g(x) = bo + blx + + bnxn (bn:l: 0),
f(x)g(x) = h(x) = Co + CIX + + Cm+n xm +n (Cm+n:l: 0).
If h(x) is not primitive there exists a prime p E Z such that p divides each of
Co, CI, .•• ,Cm+n' Now p cannot divide all of the coefficients of f(x) or all of the
coefficients of g(x) since f(x) and g(x) are primitive. Suppose therefore that p
divides 0.0, all , ar -1 but p does not divide ar where 0 ~ r < m and that p
divides bo,b1, ,bs - 1 but p does not divide bs where 0 ~ s < n. Consider
c,.+s. We have
c,.+s = o.obr+s + a1 br+s-1 + ... + ar-1 bs+1 + arbs + ar+1 bs-1 + ... + ar+sbo'
Now p divides c,.+s; o.o,a1"ar -1; bo,b1, ... ,bs - 1 and hence it follows that p
divides arbs and so divides ar or bs. But this is a contradiction and so f(x)g(x)
is primitive. 0
Example 23
2x 2 + 3x + 1, x 2+ 2 are primitive and so is
(2x + 3x + 1)(x2 + 2) = 2x4 + 3x 3 + 5x2 + 6x + 2.
2
We use Lemma 13 to show that if factorization does not occur in Z[x] then it
does not occur in Q[x].
Lemma 14
Let f(x) E Z[x] and let f(x) have degree n, n > O. Suppose that f(x) does not
factorize in Z[x] into the product of two polynomials of degrees rand s where
o < r < n, 0 < s < n (r + s = n). Then f(x) does not factorize in Q[x] into
the product of two polynomials of degrees rand s.
Proof
We argue by contradiction and suppose
f(x) = g(x)h(x)
where g(x) and h(x) are polynomials in Q[x] of degrees rand s respectively. Now
~
f(x) = CfJfo(x), g(x) = dCl go(x), h(x) =~ ho(x),
1
where fo(x), go(x) and ho(x) are primitive polynomials in Z[x] and CfJ, ell~, d1 ,
~ are in Z. Then
and so
Rings 189
Now, by Gauss's Lemma, go(x)ho(x) is primitive and so, by Lemma 12,

Codl~ = ±CIC2 and fo(x) = ±go(x)ho(x). But go(x) and ho(x) have degrees r
and s respectively and so we have derived a contradiction since f(x) = Cofo(x) =
±go(x)ho(x). Thus the lemma is proved. 0
We now obtain our main result which is named after F.G.M. Eisenstein
(1823-52).
Theorem 22 Eisenstein's Criterion

Let f(x) E Z[x] and let f(x) = ao + alx + ... + anxn (an # 0). Suppose there
exists a prime p such that:
1. P divides ao, aI>' .. ,an-I>
2. p does not divide an and
3. p2 does not divide ao.
Then f(x) is irreducible as a polynomial in Q[x].
Proof
We argue by contradiction. If f(x) is not irreducible in Q[x] then f(x) is not
prime in Q[x] and so f(x) may be factorized in Q[x] into two polynomials of
degrees r and s where 0 < r < n, 0 < s < n, r + s = n. There is necessarily a
corresponding factorization of f(x) in Z[x]. Hence we may suppose that
f(x) = g(x)h(x)
where g(x) and h(x) are polynomials in Z[x] of degrees rand s respectively. Let
g(x) = bo + b1x + + brxr (br # 0),
h(x) = Co + CIX + + csx s (c s # 0).
Nowao = boCo and since p divides ao but p2 does not divide ao, either bo or Co,
but not both bo and Co, is divisible by p. Suppose p divides bo but p does not divide
Co· If p were to divide bo, bI>' .. ,br then all the coefficients of f(x) would also be
divisible by p and that is false. We suppose therefore that p divides
bo, b1 , .•. ,bk- 1 but P does not divide bk for some k where 0 < k ~ r < n. Since
ak = bkCo + bk- 1 Cl + ... + bOCk
we have that p divides ak; bo, b1, ... ,bk-1 and so p divides bkCo. But p does not
divide Co and so p divides bk which is false. Hence our initial assumption was
wrong and consequently f(x) is irreducible in Q[x]. 0
We give some examples of the use of Eisenstein's Criterion.

Examples 24
1. Let p be a prime. Then x n - p E Q[x] is irreducible by the criterion.
2. 21 + 6x 2 + 9x3 + 4x4 is irreducible in Q[x] on applying the criterion with
p= 3.
3. We may also prove that f(x) = 1 + x + x 2 + ... + xP- 1 is irreducible where p
is a prime. We cannot apply the criterion directly but we may transform f(x)
into a polynomial to which the criterion may be applied.
Introduce a new indeterminate y where x = 1 + y. Then, temporarily using
the quotient field Q(x) = Q(y) of Q[x], we have
f(1 + y) = f(x) = 1 + x + x 2 + ... + x p - 1
x P -1
=
x-I
(1 + y)P-l
= -'------'-'---
y
[1 + (~)y+ (;)~ + ... + (:)yp] -1

y
= (~)y+ (~)y2 + ... + (:)yP.

Now, as we have seen in the proof of Theorem 8, all of the coefficients in
f(1 + y) are divisible by p except for (:) = 1, also (~) = p and so is not
divisible by p2. Thus f(1 + y) is irreducible and so, evidently, is f(x).
Exercises 5.6
1. Prove that the following polynomials are irreducible in Q[x].
5 + 25x + lOx + 7x 3 , 14 + 21x + 49x + 6x3 ,
2 2
26 + 39x + 65x + 10x •

2 3
2. Prove that 19 + 24x + 9x 2 + x3 is irreducible in Q[x]. (Hint: put

x = y-l.)
3. Prove that x 3 + 9 is irreducible in Q[x].
6
Topics in Group Theory
In this final chapter we extend our knowledge of group theory. Among other
aspects of finite groups, we investigate permutation groups and obtain two
results of cardinal importance in the theory of finite groups. In the first of
these, we establish the structure of Abelian groups and, in the second, we estab-
lish the existence of the so-called 'Sylow p-subgroups' of a finite group.
6.1 Permutation Groups

Weare here concerned with bijective mappings of a non-empty set into itself.
Such mappings may be multiplied under the circle-composition of mappings.
From our earlier work we easily obtain the following theorem.
Theorem 1
Let X be a non-empty set and let S(X) be the set of bijective mappings of X onto
X. Under the circle-composition of mappings S(X) is a group.
Proof
From the Examples in Section 4.1 (Semigroups), we know that S(X) is a monoid.
Let / E S(X). Then the inverse /-1 of / is given as follows. /-1 is that mapping
191
for which f -I (a) = b if and only if f(b) = a (a, b E X). It is then immediate that
SeX) is a group. 0
Our interest in this section concerns bijective mappings on finite sets.
Definition 1
A bijective mapping of a non-empty set X onto itself is called a permutation.
Any set of permutations of X forming a group is called a permutation group.
The permutation group of all permutations on X is called the symmetric
group on X and is frequently denoted by SeX). If X consists of the n symbols
or elements XI' X2, ..• 'X n and more particularly if X = {I, 2, ... , n} then the
symmetric group on X is designated as Sn' A permutation p E Sn is written as
p
= (i i
l
..
2 .•. in)
.
JI J2 ... In
where the notation means that p(i 1) = ii, p(i 2 ) = i2"'" p(i n ) = in' (Usually
but not invariably, iI, i 2 , • .• , in are in the natural order 1,2, ... , n.)
Example 1
The permutation p such that p(l) = 2, p(2) = 1, p(3) = 4, p(4) = 3 may be
represented by any of
1 2 3 4) (2 1 4 3) (4 2 1 3)
( 2143'1234'3124
Definition 2
A permutation p E Sn is said to fix i E {I, 2, ... , n} if p( i) = i and to move
i E {I, 2, ... , n} if p( i) ::j; i. The identity permutation, which is the identity of
the group Sn, is that permutation fixing all i E {I, 2, ... , n}.
Lemma 1
Sn is a group of order n(n - 1) ... 2.1 = n!
Proof
2
Letp= (~
JI i2
Topics in Group Theory 193
For il we have n choices of symbols from {I, 2, ... , n}. For i2 we have n - I
choices of symbols from {I, 2, ... , n} \ {il}. For i3 we have n - 2 choices of sym-
bols from {I, 2, ... , n} \ {jl,i2}. Continuing we have only one choice for in- The
total number of choices is n(n - I)(n - 2) ... I giving the result. 0
Example 2
Let X = {I, 2, 3}. Then the symmetric group S3 has order 3! = 6 and consists of
the six permutations,
(~ ~ :).(~ ~ ~).(~ ~ ~).(~ ~ :).(~ ~ ~).(~ ~ ~).

( ~ ~ : ), is the identity permutation. (~ ~ ~) is the permutation p
such that p(I) = 2, p(2) = 3 and p(3) = 1.

If q = (I
2
3 I 2
3),
then the product pq is obtained from the composition of
the mappings p and q, namely by applying q and then p as follows:
I...!.. 3 2·d
2...!..I~2
3...!..2~3
which we also write as
Convention
Let
~), q=(~Jl J2~

~n
Then
and the product pq is given by
1 2
( p(l) p(2) ::: p~») (q(ll) q;2)
q(l) q(2) q(n) ) ( 1 2

(
= (pq)(l) (pq) (2) (pq)(n) q(l) q(2)
= (Pq~(l) (Pq~(2) (pq~(n) ).

(The reader is advised that another convention for the multiplication of
permutations is also to be found in some textbooks.)
We give further examples exhibiting our convention for the multiplication of
permutations.
Examples 3
:)(:
1.
(1
1
2 3 4
524 :)(:
2 3 4
2 1 3
:) = (:
2
5
1 3
1 2
2 3
2 1 3
4
:)
:).
2 3 4
= (:
5 1 2
2.
(~
2
3
3
7
4
1
5 6 7)( 1 2
2 4 5 5 2
3 456
7 361 :)
= (:
273
3 5 7 4
6 1
6
;)(: 2
2
3
7
4
3
5
6
6
1 :)
= (~
2 3 4
3 5 7 4 6
5 6
:).
Lemma 2
Letp= (~ ~1
2
i2
n)
.
~n
. Thenp -1 C1
=
1
i2
2 ~).
Proof
Definition 3
A permutation p of {I, 2, ,n} is called a cycle of length r (1 < r ~ n) if
for some subset {iI, i 2, , in} from {I, 2, ... , n} we have P(ii) = i 2,
P(i2)=i3,···p(ir-l)=in p(ir)=i 1 and for jE{I,2, ... ,n}\{i 1,i2, ... ,i r },
p(j) = j. To indicate that the cycle p permutes the integers iI, i 2, ... , i r cyclically
and that the remaining integers are fixed by p we write p, in an abbreviated
notation, as
p = (i 1i 2 ··· i r )·
A cycle of length 2 is often called a transposition. Conventionally the identity

permutation is regarded as a cycle oflength 1 and is written as (1).
Example 4
1 2 3 4 5 6 7 8) is the cycle (1 6 2 3 5), which may also
( 6354127 8
be written as (6 2 3 5 1), (2 3 5 1 6), (3 5 1 6 2) or
(5 1 6 2 3).
We note that in cycle notation we write down only the symbols from 1, 2, ... , n
which are moved. Consequently only the context enables us to determine
1 2 3 4 5 6 7 8)
whether the cycle (1 6 2 3 5) represents
( 6 3 5 4 1 2 7 8 or,
1 2 3 4 5 6)
( 6 3 5 etc. Fortunately ambiguity does not usually arise.
4 1 2
Permutations expressed in cycle notation are easy to multiply. We illustrate
their multiplication in the following example.
Example 5
Suppose we wish to prove that in 8 6 we have
(1 3 4 5 2)(6 2 3 4)(1 3 5) = (1 5 3 2 4 6).
Letting r = (1 3 4 5 2), s = (6 2 3 4), t = (1 3 5) we have, for the

product rst,
1~3~4~5
2~2~3~4
3~5~5~2
4~4~6~6
5~1~1~3
6~6~2~1
and so
rst = (1 3 4 5 2)(6 2 3 4)(1 3 5) = (1 5 3 2 4 6).
Example 6
The cycles (1 7 8 2 4) and (3 6 5) permute cyclically the disjoint sets of
integers, {I, 2, 4, 7, 8} and {3, 5, 6}. The cycles are easily seen to commute.
Definition 4
Two permutations from Sn are said to be disjoint if the two subsets of integers
moved by the permutations are disjoint.
Lemma 3
Disjoint permutations commute.
Proof
Let p and q be disjoint permutations from Sn' Then we know that we have three
subsets A, B, C, of which C may be empty, such that
{1,2, ... ,n}=AUBUC
is a disjoint union and such that p permutes the integers of A among themselves
but fixes the integers of B U C and q permutes the integers of B among them-
selves but fixes the integers of A U C. Let a E A, b E B, C E C. Then
(pq)(a) = p(q(a)) = p(a) = q(p(a)) = (qp)(a),
(pq)(b) = p(q(b)) = q(b) = q(P(b)) = (qp)(b),
(pq)(c) = p(q(c)) = p(c) = c = q(c) = q(p(c)) = (qp)(c).
Thuspq = qp. o
Example 7
1 2 3 4 5 6 7 8 9)
( 9 8 365 7 4 1 2
is a product of the disjoint, and so commuting, cycles (1 9 2 8) and
(4 6 7).
We obtain easily the following result.
Theorem 2
Every permutation is a product of disjoint cycles.
We have broken down any given permutation into disjoint cycles. We may
now break down each cycle into transpositions.
Theorem 3
Every permutation is a product of transpositions.
Proof
It is enough to show that every cycle is a product of transpositions.
Trivially (1) = (12)(12). Let now (i l , i 2 , ... , ir) be a cycle oflength r, r > 1.
Then, by direct calculation,
o
Example 8
1 2 3 4 5 6 7 8)
= (1 7 5 4 3 2)(6 8)
( 7 1 234 8 5 6
= (1 2)(1 3)(1 4)(1 5)(1 7)(6 8)
= (5 7)(5 1)(6 8)(5 2)(1 2)(3 4)(5 3)(1 2)
We may express a permutation as a product of transpositions in many, indeed

too many, ways but whether an even number or an odd number of transpositions
is involved in any particular way turns out to be solely determined by the permu-
tation. In order to prove this mysterious fact we make a digression into rings of
polynomials of several commuting indeterminates and we define the 'action' of a
permutation upon such polynomials.
Definition 5
Let Z[XI,X2"" ,xnl be the polynomial ring of n commuting indeterminates
Xl' X2,· .. ,Xn over Z. Let p E 8 n and let f(XI' X2, ... ,xn ) E Z[Xl' X2,' .. , xnl.
The action of p on f(XI' X2, ... ,xn) is then defined to be the polynomial
(pf)(Xl> X2, ... ,xn) given by
(Pf)(XI' X2,"" xn) = f(Xp(I) , Xp(2),"" Xp(n)'
Example 9
Let f(XI' X2, X3) = xtx~ + 5x~ + 4x~xg E Z[Xl> X2, x31·
Let p = (~ ~ ~) E 8 3 , Then
(pf)(Xl>X2,X3) = x~xf + 5x~ + 4xrx~.
If, in particular, we act upon f(XI,X2"" ,xn ) E Z[XI,X2"" ,Xn], as above, by
a transposition (ij) (i =1= j), say, then in effect we interchange the indeterminates
Xi and Xj in the expression for f(XI' X2,·'·' x n).
Example 10
We let the transposition (1 2) act on various polynomials.
(1 2) (xf + x~ + X3) = x~ + x~ + x3,
(1 2) (Xl - X2) = X2 - Xl = -(Xl - X2),
(1 2) (X3 - X4) = X3 - X4,
(1 2) (Xl - X2)(XI - X3)(X2 - X3) = (X2 - XI)(X2 - X3)(Xl - X3)
= -(Xl - X2)(XI - X3)(X2 - X3)'
Definition 6
Let Z[XI, x2, ... , xnl be the polynomial ring of n commuting indeterminates over
Z. Let f( Xl, X2, ... ,xn) be a polynomial from Z[XI, X2, ... , x n]. If for all trans-
positions t E 8 n
(tf)(XI,X2"" ,xn) = f(XI,X2"" ,xn)
then f(xl> X2,' .. ,xn) is said to be symmetric. If for all transpositions t E 8 n
(tf)(XI' X2,' .. ,xn) = - f(XI, X2,· .. ,xn )
then f(XI,X2"" ,xn) is said to be skew-symmetric.
A polynomial may be neither symmetric nor skew-symmetric.

Example 11
In Z[Xl> X2, x31, Xl + X2 + X3, X~ + X~ + X~, XIX2 + XIX3 + X2X3 and XIX2X3 are
typical symmetric polynomials. (Xl - X2)(XI - X3)(X2 - X3) is a skew-symmetric
polynomial. Xl - X2 + x3 is neither symmetric nor skew-symmetric.
We recall that
denotes the product al a2 ... an' Let w(xl, X2, ..• , x n ) be the particular poly-
nomial from Z[Xll X2,' •. , xnl given by
II
n
W(XI, X2,"" x n) = (x r - x s)
r,s=l
r<s
= (Xl - X2)(XI - X3) ... (Xl - x n)
x (X2 - X3) .•. (X2 - x n)
Example 12
Lemma 4
For all transpositions t from Sn
(tW)(XI,X2"" ,xn ) = -W(XI,X2,'" ,xn ).
Proof
Let t be the transposition (ij), i < j. t acts on the factors of the product
II
n
(x r - x s )·
r,s= I
r<s
A factor of the form X r - X s where neither r nor s is equal to either i or j is

unaltered by t.
Factors of the form X r - Xi and X r - Xj where r is not equal to either i or j are
simply interchanged by t. Similarly factors Xi - X r and Xj - X n (r i= i,j) are
interchanged by t.
Factors of the form Xr - Xi and x j - Xr (r :f:. i, j) must be considered together.

For such a pair of factors we have
(ij)[(x r - Xi)(Xj - x r )] = (x r - Xj)(Xi - x r )
= (Xj - xr)(x r - Xi)
and so the product of the pair of factors is unaltered.

The single remaining factor to be considered is Xi - Xj and for this factor
(ij)(Xi - Xj) = Xj - Xi = -(Xi - Xj)
and so we have a change of sign. Altogether there is, in consequence, effectively

one change of sign and hence we obtain the desired result. 0
Corollary
Let p be a permutation from Sn. Then
(pw)(Xl> X2" .. ,xn) = ±W(Xl' X2, ... ,x n )·
Proof
p is a product of transpositions, say p = t} t 2 ... t m where t 1, t2, ... , t m are trans-
positions. Then
(PW)(Xl' X2,· .. , x n) = (t 1t 2 tmw)(Xl> X2, ... ,xn )
= -(t 1 tm_1W)(Xl' X2, ... ,x n)
Definition 7
A permutation p from Sn is said to be even if
(PW)(Xl' X2, ... ,x n) = w(x}, X2, ... ,x n )
and to be odd if
By Lemma 4, a transposition is an odd permutation. From the proof of the

Corollary above we obtain the next theorem.
Theorem 4
Every permutation is either even or odd. An even permutation is only expressible
as a product of an even number of transpositions. An odd permutation is only
expressible as a product of an odd number of permutations.
We may go further 88 follows.
Theorem 5
Let An be the set of even permutations from Sn' Then An is a normal subgroup of
Sn of index 2, IAn I = ~nl
Proof
An is obviously closed under multiplication and if p E An then p is a product of
an even number of transpositions and so also is p-l. Thus An is a group.
We claim that Sn = An U A n(1 2). Let p E Sn.1f p is even then p E An. If pis
odd then p(1 2) is even and so p(1 2) E An from which p E A n (1 2). Therefore
Sn ~ An U A n (1 2) giving the result. 0
Definition 8
The group of the even permutations from Sn is called the alternating group on
{I, 2, ... , n} and is denoted by An.
Example 13
A 3 = {(I), (1 2 3), (1 3 2)}.
A situation which commonly arises, and which we have encountered, is that in

which a group G 'acts' as a permutation group on a non-empty set X. We some-
times say that G 'induces' a permutation group on X. We require to make this
terminology more precise.
Definition 9
Let G be a group and let X be a non-empty set. Let S(X) be the group of
permutations of X and let f: G -+ S(X) be a homomorphism. Then for each
a E G, f(a) is a permutation of X and G is then said to act as a permutation
group on X or to induce a permutation group on X. As a short notation for the

action of G on X we write
a.x = f(a)x (x E X, a E G).
We note that, in this notation,
(ab).x = f(ab)x = (J(a)f(b))x = f(a)(J(b)x) = a.(b.x) (x E X, a,b E G).
Furthermore
Ker f = {a E G Ia.x = x for all x E X}
and so G/Ker f is isomorphic to a subgroup of S(X).
Definition 10
Let G act as a permutation group on a non-empty set X. Let x E X. Then the
orbit containing x is defined to be G(x) = {a.xla E G}. The stabilizer of x is
defined to be
Stabc(x) = {a E Gla.x = x}.
Theorem 6
Let G act as a permutation group on a non-empty set X. A relation'" is defined
on X by x '" y (x, Y E X) if and only if a.x = y for some a E G.
1. Then", is an equivalence relation and each equivalence class is an orbit.
2. If X is finite the number of distinct elements in the orbit of x E X is given by
IG: Stabc(x) I.
Proof
1. Certainly for all x E G, e.x = x and so x '" x.
If x'" y (x, Y E X) then there exists a E G such that a.x = y. Hence it
follows that x = e.x = (a-Ia).x = a-l.(a.x) = a-l.y and so y '" x.
If x'" y and y '" z (x, y, z E G) then there exist a, bEG such that a.x = y
and b.y = z, from which (ba).x = b.(a.x) = b.y = z and so x '" z.
2. G(x) = {a.xla E G} is the orbit of x. Let
G = aIStabc(x) U ~Stabc(x) u ... u amStabc(x)
be a coset decomposition of Stabc(x) in G. Then we claim that the elements
al'x, ~.x, ... ,am.x are distinct and constitute the entire orbit containing x.
The proof is very similar to that in Chapter 4, Theorem 19, in which we
proved that the number of elements in a conjugacy class of a group is equal
to the index of the centralizer of an element in the claBS. The group acts as a
permutation group on the conjugacy claBS. The reader should establish for
himself or herself the details of the present proof. 0
We may now prove a result first established by Cayley in 1878.
Theorem 7 Cayley's Theorem

A finite group of order n is isomorphic to a subgroup of the symmetric group Sn'
Proof
Let G have distinct elements Xl' X2, ... ,X n . Then G acts as a permutation group
on G where G is considered simply as a non-empty set. The action is given by
a,xi = aXi (a E G, i = 1,2, ... ,n)
since we have
ab'Xi = (ab)xi = a(bxi) = a.(b,xi) (a, bEG, i = 1,2, ... ,n).
Furthermore a E G acts as the identity permutation on X if and only
ifaxi = a,xi = Xi (i = 1,2, ... ,n) which is so if and only if a is the identity
of G. Thus G is isomorphic to a subgroup of the symmetric group
on {Xl' X2, ... , x n }. [In fact the isomorphism is given explicitly by
a -+
Xl X2 ... Xn ) I
. o
(
aXI aX2 ... aX n
Exercises 6.1
1. Evaluate the following products:
2 3 4 5 2 3 4 5
(~ 1 4 2 6 ~) (~ 4 6 3 2 ~).
2 3 4 5 2 3 4 5
(~ 1 4 3 6 ~) (~ 3 4 5 6 ~).
2. Prove that
(1 2 9 3 4)(5 6 2 8 1)(7 8 9 1)
= (1 7 2 8 3 4)(5 6 9).
3. Write the following cycles as products of transpositions:
(1 3 5 7 6 2), (1 2 4 3), (6 7 1 2 5 8 4).
Which cycles are even and which are odd?
4. Write down the elements of the A 4 .

5. Is the polynomial (Xl - X2)(XI - X3)(X2 - x3)(XI + X2 + X3) sym-
metric, skew-symmetric or neither?
6. Let G be a group and let f : G - 8(X) be a homomorphism of G into
the group 8(X) of permutations on a finite set X. Prove that
Ker f = n Staba(x).
xEX
7. Write down the full proof of Theorem 6, part 2.

8. Let X = {HI' H 2, ... , H n } be a complete set of conjugates of a sub-
group H of a group G. Prove that an action of G on X is defined by
a.Hi = aHia- 1 (a E G, i = 1,2, ... ,n).
Prove that n = IG: Na(H) ,.
9. Find an isomorphism between the 8 3 and the non-Abelian of order 6,

the Cayley table of which is in the Examples 11 no. 3 of Section 4.2.
6.2 Generators and Relations

When we discussed various finite groups in Chapter 4 we often wrote down their
Cayley tables. It was evident that these tables were only useable for groups of
order perhaps at most 20 and so their use had obvious limitations for describing
finite groups in general. We here develop a more efficient means of describing
groups, whether finite or infinite.
Example 14
Let us consider the group G of order 6 given in Examples 11 no. 3 of Section 4.2,
in which G = {e, a, b, c, d, f} is the group of mappings of IR \ {O, I} into IR \ {O, I}
for which the Cayley table has been given.
We have a 3 = e, c2 = e. Suppose we try to find the least subgroup H of G con-
taining a and c. Certainly H contains e, a, a 2, c, d = ac and f = a 2c and so
H = G. Thus we say that a and c 'generate' G. Could we determine the
Cayley table from simply 'knowing' a and c? The answer is 'not quite', because
2 2
a and c are not unrelated since we have ac = ca (= d) and a c = ca (= f). But
these relations may be rewritten as c-Iac = a , c- a c = a and the second is a
2 I 2
consequence of the first since c- I a 2c = (c- I ac)2 = (a 2)2 = a 4 = a. But now
knowing the fact that in G we have a3 = e, c2 = e, c-Iac = a 2 we may work
out the products of elements such as df by using only this fact and by using the
legitimate operations of a group; thus we may determine df as follows. We have
d = ae and f = a2e and so df = (ae)(a 2e) = ae- 1a2e = aa = b. We can obtain all
the entries of the Cayley table in this way. Thus we may derive the Cayley table
from knowing that we have a group generated by two elements a and e with the
relations a3 = e, e2 = e, e- 1 ae = a2 . We write
G = (a, e Ia3 = e, e2 = e, e-1 ae = a2) .
Example 15
Suppose we consider a group G with generators a and e as in the previous
example but with a slight modification of the relations, namely we suppose
a3 = e, e2 = e, e -1 ae = a.
By virtue of the fact that a and e commute, we may collect together in any
product of a's and e's the powers of a and e. Using the relations above we see
that G = {e, a, a2 , e, ae, a2 e} and is an Abelian group of order 6. We may write
down a Cayley table as shown below:
e a a2 e ae a 2e
e e a a e ae ae
a a a2 e ae a2 e e
a2 a2 e a a2 e e ae
e e ae a2e e a a2
ae ae a 2e e a a2 e
a 2e a 2e e ae a2 e a
We deduce more. If we let b = ae then b2 = a2 , b3 = e, b4 = a, b5 = a2 e, b6 = e.

We see that G is actually a cyclic group of order 6 which may be represented
as either
or as
G = (bW = e).
Thus, for a given group, generators and relations are not uniquely determined.
Remark
In both of the examples above there were two generators a and e for which
a3 = e, e2 = e and one further relation between a and e. We take this opportu-
nity to remark that if we merely have two elements a and e for which a3 = e,
c2 = e and no relation is presumed to exist between a and c then it may be shown,

but regrettably not easily, that the group so described is isomorphic to the group
of linear fractional transformations of the form
az+ {3
z -+ - -
,z+ 6
(a, {3",6 E Z, a6 - f3'Y = 1)
where now a and c correspond respectively to the transformations
1 1
z-+--- z-+ -.
z+ l' z
In the two examples above we began with a group, the elements of which could
be written down or easily obtained, and we have shown how by considering a few
(actually, two) ofthese elements and some relations between them we may, in
some sense, 'recover' the group. What is much more usual is to begin with
some elements supposed to be the elements of a group and with some relations
between those elements and then to investigate the group which these generators
and relations describe. We do not assert either that the elements are necessarily
distinct or that the relations are completely independent of one another. In these
circumstances the reader may wonder whether the resulting algebraic entity is,
genuinely, a group. The answer is fortunately in the affirmative since the group
of one element {e} will always satisfy any system of generators and relations
albeit trivially. The real interest lies in determining the 'largest' group, up to
isomorphism, satisfying the system.
Every group has at least one system of generators and relations. We need only
write down the elements a, b, c, ... and the relations formed of products such as
ab = c to get a system of generators and relations, but such a system is no more
efficient than a Cayley table. We aim to be more economical with any generators
and relations for a group.
Notation
The group G with generators aI, a2, ... and relations rl = 81' r2 = 82, ... where
rl, r2,' .. j 811 82, ... involve only all a2,' .. is denoted by
(We shall restrict our attention in this text to a finite number of generators and
relations. )
Example 16
Let G = (a, b, cl a2 = b2 = c3 = e, ab = 00 = c). What is the group G? We have
e = a2b2 = aabb = aOOb = (ab)2 = c2. But c3 = e and so c = e. Then
b = a-I = a. Thus, in truth, G is a group of two elements, namelye and a,
and a2 = e. Consequently G = (ala2 = e).
We now consider some standard examples.
Examples 17
1. Let G = (alan = e). Then G is the cyclic group of order n generated by a.
2. Let G = (al.). Then G is the group generated by a with an empty set of
relations. Thus e,a,a- l ,a2,a-2, ... are distinct and G is the infinite cyclic
group generated by a.
3. Let G = (a, bl a2 = b2 = e, ab = 00). Then G is isomorphic to the Klein four-
group.
4. Let G = (b,flb4 = e, f2 = b2, f-Ibf = b3 ). Then we see that (b), the sub-
group generated by b is normal in G and has order 4. (b)f::j:. (b) but since
f2 = b2 and f2 is in the centre of G, no coset distinct from either (b) or
(b)f is possible. Thus G has order 8. In fact G is the quaternion group of
Exercises 4.5, no. 5 with the same band f and with a = b2, c = b3 , h = bf,
2 3
d = b f, 9 = b f·
5. Let G = (a, bla2 = b2 = e). Then G is generated by a and b, both of order 2,
but no relation connects a and b. The structure of G is not easy to discern but
becomes clearer if we make a small change to the generators. Let c = abo Then
we have c- I = (ab)-l = b-Ia- 1 = ba and then b-1cb = OObb = 00 = c- 1 •
Hence it follows that G = (b,clb2 = e, b-1cb = c- 1 ) and so G has a normal
subgroup (c) which is an infinite cyclic subgroup of index 2. The group is
known as the infinite dihedral group.
Groups defined by generators and relations appear in many contexts. One

context in particular is that of a group of transformations of some regular geo-
metrical object, the transformations preserving the size, shape and location of
the object. Under such a group the vertices of the object are permuted amongst
themselves and the group is directly a subgroup of the symmetric group on the
set of vertices.
Consider, for example, the group of transformations of a square with centre 0

and vertices labelled 1, 2, 3, 4. We may rotate the square about its centre 0
through angles which are multiples of 2; = ~. The permutation p = (1 2 3 4)
(clockwise) rotation through an angle of 2'
1C'
represents a
G
1 2 4 1
4 3
G
3 2
The remaining rotations about 0 are p2, l, p4 = e where e is the identity trans-
formation (e = (1)). The square may also be reflected about an axis of symmetry,
say about a line through 0, parallel to a side, as shown below.
-B-
1 2 4 3
4 3
G
1 2
The reflection T is given by T = (1 4)(2 3). We now claim that the group G of all
such transformations is generated by p and T. Certainly p4 = e, T 2 = e and
1
T- pT = (1 4)(2 3)(1 2 3 4)(1 4)(2 3) = (1 4 3 2) = p-l,
and so
G= (p,Tl p 4 = T 2 = e, T-1pT = p-l)

but we have to establish that every transformation is indeed in G. We confine our
considerations to just one further transformation, leaving the reader to ponder
the remaining possibilities. We examine the reflection (1 3) about the diagonal
joining 2 and 4.
3 2
4 3
G
4 1
Then (1 3) = (1 4 3 2)(1 4)(2 3) = p-I r E G. We have now proved that the

group of transformations preserving a square is a group which previously we
called the dihedral group D 4 of order 8.
We may extend these arguments to consider the group of transformations of a
regular polygon. If the polygon has n vertices, labelled 1, 2, ... , n, then it may be
rotated through angles which are multiples of 27T, thus yielding the trans-
n
formation p = (1 2 ... n) and powers of p. The polygon may also be reflected
about an axis of symmetry through its centre.
If n is even, n = 2N (say), the axis of symmetry may be chosen (as in the case
of the square) to be the perpendicular bisector of the side joining 1 and 2N.
We then have a reflection r = (1 2N)(2 2N -1) ... (N N + 1). As p =
(1 2 ... 2N) we have
p2N = r 2 = e, r-Ipr = p-I.
If n is odd, n = 2N + 1 (say), the axis of symmetry must pass through one
vertex 2N + 1 (say) and we have a reflection r = (1 2N)(2 2N - 1) ...
(N N + 1). Since p = (1 2 ... 2N + 1) we have p2N + I = r 2 = e, r -I pr = P-I.
In both cases we may show that the group generated by p and r contains all
transformations of the polygon. In summary, we have shown that the group of
transformations of a regular n-gon is
(p,rlpn = r 2 = e, r-Ipr = p-I)
which is called the dihedral group D n of order 2n.
Exercises 6.2
1. Let G = (a, bl a3 = b2 = e, ba = e). What is the order of G?
2. Let G = (a, bla4 = b2 = e, aba2 = b). Prove that G has order 2.
3. Let G = (a, biaS = e, a4 = b2 , b-Iab = a-I). Prove that G has order 16
and that G has only one subgroup of order 2.
4. Prove that G = (a, bla3 = b2 = e, b-Iab = a-I) is isomorphic to the 8 3
and that G is also the group of transformations preserving size etc. of
an equilateral triangle. (See Exercises 6.1, no. 9.)
5. What is the group of transformations preserving size etc. of a rectangle
which is not a square?
6. Prove that all transformations preserving size etc. of a regular penta-
gon may be identified with elements of the dihedral group D s.
6.3 Direct Products and Sums

Let G be the cyclic group of order 6 generated by a,
G = (ala 6 = e).
G has two proper subgroups, namely P = {e, a3 } and Q = {e, a, a2 } of orders 2
and 3 respectively. These subgroups are normal in G and we obviously have
P n Q = {e} and P n Q = G. We describe this situation more generally in the
following definition.
Definition 11
Let G be a group. Let H and K be normal subgroups of G such that the inter-
section H n K = {e} and G = HK. Then G is said to be the (internal) direct
product of Hand K.
Let us now consider two groups A and B where A = (ala 2 = eA) and
B = (blb3 = eB), eA and eB being the identities of A and B, and A and Bare
cyclic groups of orders 2 and 3 respectively. Let G be the Cartesian product
A x B of A and B, then
G = {(eA' eB), (eA' b), (eA' b2 ), (a, eB), (a, b), (a, b2 )}.
Let the multiplication in G be component-wise so that, for example,
(eA,b)(a,b) = (eAa,bb) = (a,b2 ).
Then under this multiplication it may be seen that A x B is a group. Again, more
generally, we describe this situation in a definition.
Definition 12
Let G be the Cartesian product of the groups A and B. A multiplication is
defined on G = A x B by
(a, b)(a', b') = (aa', bb').
Then G is a group called the (external).direct product of A and B.
In the group G = A x B we notice that

A = {(a,eB)la E A} and B = {(eA,b)lb E B}
are normal subgroups of G and that the mappings
a ~ (a, eB) (a E A) and b --+ (eA' b) (b E B)
yield isomorphisms of A with A and B with E respectively. Moreover G is the

(internal) direct product of A and E.
Both of the definitions above extend naturally to direct products involving a
finite number of subgroups or groups.
Definition 13
Let G be a group. Let HI' H 2 , ••• ,Hn be normal subgroups of G such that
Hi n H 1 H 2 •·• H i - 1 H i +1 .•• H n = {e} (i = 1,2, ... ,n) and G = H 1 H 2 ••• H n .
Then G is said to be the (internal) direct product of HI' H 2 , •.. ,Hn .
In this definition, since HI, H 2 , .•• ,Hn are normal subgroups of G, it follows
that, for each i, 1 :::; i :::; n, H 1 H 2 ••• H i - 1 H i +1 ... H n is a normal subgroup of G.
Definition 14
Let G be the Cartesian product of the groups All A 2, ... , An. A multiplication is
defined on G = Al X A 2 X ••• x An by
(all~'··· ,an)(a~,a~, ... ,a~) = (ala~,a2a~, ... ,ana~)
(ail a~ E Ai' i = 1, 2, ... , n)
Then G is a group called the (external) direct product of All A 2, . .. ,An.
Similarly, as above,
Ai = {(ell e2,· .. ,ei-ll a, ei+ll·· . en) Ia E Ai},
where A j has the identity ej (j = 1,2, ... , n), is isomorphic to Ai (i = 1,2, ... , n)
and G is the (internal) direct product of All A2 , ••• , An. On account of these
obvious isomorphisms we sometimes do not always draw a precise distinction
between internal and external direct products.
Definition 15
Let p be a prime. A group the order of which is a power of p is called a p-group.
Definition 16
Let p be a prime. Let H ll H2 , .•• , Hn be cyclic groups of order p. Any group iso-
morphic to the direct product HI x H 2 X ••• x H n is said to be an elementary
p-group.
Example 18
The Klein four-group is an elementary Abelian 2-group.
We have given the definitions above for groups having a binary operation of
multiplication. If we are concerned with Abelian groups then, as we know, an
additive notation is commonly employed. The changes of notation and nomen-
clature are fairly obvious. We may still form the Cartesian product of Abelian
groups to obtain an additive group in which addition is performed component-
wise, yielding an (external) direct sum. We find it to be convenient now to
rephrase the definition above for (internal) direct sums of Abelian groups.
Definition 17 (following Definition 13)

Let A be an Abelian group. Let A l ,A2 , .•. ,An be subgroups of A such that
Ain(A l +A2 + +A i - l + A i+l + ... +A n) = {O} (i= 1,2, ... ,n)
and A = A l + A 2 + + An. Then A is the (internal) direct sum of
A l , A 2 ,···, An·
Examples 19
1. Let A, Band C be groups of orders 9,14 and 18 respectively. Then the group
G, given by G = A x B x C, is a group of order 9.14.18 = 2268.
2. Let A, B, C be cyclic groups of orders 7, 8 and 17 respectively. Then the group
G, given by G = A x B x C is a group of order 952. Moreover if
A = (ala7 = eA), B = (blb8 = eB) and C = (clc17 = ec)
then G is cyclic with generator (a,b,c) and identity element (eA,eB,eC)'
Exercises 6.3
1. Let A, Band C be normal subgroups of a group G. Prove that ABC is
a normal subgroup of G.
2. Let the group G be the direct product of the groups Hand K. Prove
that G is Abelian if and only if Hand K are Abelian.
3. Let Hand K be normal subgroups of a group G such that G = HK.
Prove that Gj(HnK) is the direct product of Hj(HnK) and
Kj(HnK).
4. Let A, B and C be subgroups of an Abelian group G such that
G= A + B+ C, AnB = {O}, AnC = {O} and BnC = {O}.
Does it follow that G is the direct sum of A, B and C?
6.4 Abelian Groups

The structure of an Abelian group may be very complicated and, without some
conditions on the group, may be difficult to determine. If the group has a finite
number of generators then a structure theorem is possible. Here we shall further
restrict the considerations to finite Abelian groups for which we obtain a struc-
ture theorem.
We have shown that a cyclic group has subgroups of all possible orders and
that the order of an element in a group is the order of the cyclic subgroup
generated by that element. If the order of every element of a group is a power
of a prime p then it is not immediate, at present, that the group is a p-group
since the order might be divisible by some prime other than p and yet have
no elements of that order. One of our aims is to eliminate this hypothetical
situation.
We begin with a lemma which resolves the situation for an Abelian group. As
is customary we shall employ additive notation when discussing Abelian groups.
Lemma 5
Let A be a finite Abelian group. Let p be a prime such that every element of A has
an order which is a power of p. Then A is a p-group.
Proof
If A is cyclic then, by the remarks above, A is a p-group since if IA I is divisible by
a prime q, q =1= p, we would have an element of order q, which is false.
We may now argue by induction and assume the result is true for all groups of
orders strictly less than IA I. Suppose A is not cyclic and let a E A, a =1= O. Then
the subgroup (a) generated by a has an order which is a power of p. Further,
AI(a) has the property that every element has an order which is a power of p
since if x E A and x has order pOI then
pOl(X + (a)) = pOlX + (a) = 0 + (a) = (a)
and so x + (a) has order dividing pOI. By the induction assumption AI(a) is a
p-group and so, as IAI = I(a) I = I(a) IIAI(a) I, A is a p-group. This completes
the proof. D
Theorem 8
Let A be a group of order where p~lp;2 ... p~n where Pl>P2,'" ,Pn are distinct
primes and Ui>O (i=1,2,oo.,n). Let Ai={XEAlpfiX=O}
(i = 1,2, ... , n). Then Ai is a subgroup of order pf' (i = 1,2, ... , n) and A is the
direct sum of AI' A 2,· .. ,An'
Proof
By Lemma 5, Ai is a pi-subgroup of A (i = 1,2, ... , n). We want to show that
+ A 2 + + An,
A = Al
Ai n (AI + A 2 + ... + A i - I + A i + I + + An) = {O} (i = 1,2, ... , n)

and finally that IA i 1= pf' (i = 1,2, ... , n).
Let qi be given by IA I = pf' qi (i = 1,2, ... , n). Since Pb P2, ... , Pn are distinct
primes, qI, q2, ... ,qn have 1 as their greatest common divisor. Hence there exist
t I , t 2, ... , t n E Z such that
tIqI + t 2q2 + ... + tnqn = 1.

Then, if x E A,
+ t2q2x + ... + tnqnx E Al + A 2 +
x = tIqIX + An
since pf'qiX = IAlx = 0 implies that qiX E Ai (i = 1,2, , n).
Let now
YE Aj n (AI + A 2 + ... + A j _ I + A j + I + ... + An)
for some j E {I, 2, , n}. Since Y E A j we have P;;Yj = O. On the other hand,
since Y E Al + A 2 + + A j - I + A j +! + ... + An we have
+ Y2 + ... + Yj-I + Yj+I + ... + Yn
Y = YI
where Yk E A k (k = 1,2, ... ,j - l,j + 1, .. . n). Then

%Y = %YI + %Y2 + ... + %Yj - I + %Yj+ 1 + ... + %Yn = 0 + 0 + ... + 0 = O.
Thus the order of Y divides the coprime integers p'f and qj and so Y = O. Thus
Aj n (AI + A 2 + ... + A j - I + A j + I + ... + An) = {O} (j = 1,2, ... , n}.
Hence A is the direct sum of A I' A 2, ... , An, and so
pflp~2 .. . p~n = IAI = IA I IIA2 1 .. . IA n I.
Since IAil is a power of Pi (i= 1,2, ... ,n) we deduce that IAil =pf'
(i = 1,2, ... , n) and our proof is now complete. 0
By virtue of this last theorem we need only consider the structure of a finite
Abelian p-group. The following rather obvious lemma is useful.
Lemma 6
Let G be a group. Let HI> H 2 and K be subgroups of G such that K is a normal
subgroup of G and K ~ HI n H 2 . If HI!K n H 21K is the trivial subgroup of
GIK then K = HI n H 2 •
Proof
Let x E HI n H 2 • Then Kx E HI!K n H 2 1K and so, by assumption, Kx = K.
Thus x E K giving the result. 0
Lemma 7
Let A be a finite Abelian p-group. Let a be an element of A of greatest possible
order. Then there exists a subgroup H of A such that A is the direct sum of H
and (a).
Proof
If A is cyclic then A = (a) and the result is true with H = {O}. Suppose A is not
cyclic. We shall argue by induction on the order of A, assuming that the result is
true for groups of orders strictly less than 1 A I.
Then a is of greatest possible order pm, say, in A and A =f (a). We claim that
there exists in G \ (a) an element of order p. We choose bEG \ (a) to have least
possible order amongst the elements of G \ (a). If pb = 0 then our claim is
proved. If pb =f 0 then b has order pT where p < pT ~ pm. But then pb has
order pT-I and so, by the choice of b, pb E (a). Thus pb = na for some n E Z
and hence
0= pTb = pr-l(pb) = pT-I(na) = (pT-In)a.
Since a has order pm it must be that pm divides pT-l n and so p divides n. Thus
n = pq for some q E Z. Let c = b - qa. Since b ¢ (a), we necessarily have c ¢ (a).
But pc = pb - pqa = pb - na = 0 and so c E A \ (a) and c has order p. This
proves our claim.
We may now suppose that b has order p. Then (b) n (a) = {O} since
(b) n (a) =f 0 implies (b) ~ (a) which is false. Then a + (b) has order pm in
AI(b) and so a + (b) is an element of greatest possible order in AI(b). By the
induction assumption there exists a subgroup such that AI (b) is the direct
sum of this subgroup and ((a) + (b))/(b). We may write the subgroup in the
form HI (b) for some subgroup H of A.
Then, as (b) S; H, A = H + «a) + (b)) = H + (a). In addition we have

Hn «a) + (b)) = (b).
To complete the proof we have to show that H n (a) = {O}. Let, therefore,
x E H n (a). Then x E H n «a) + (b)) = (b). Thus x E (a) n (b) = {O} and we
have established the lemma. 0
Corollary
A finite Abelian ~group is a direct sum of cyclic p-subgroups.
Proof
Let A be a finite Abelian p-group. By Lemma 7 there exists a cyclic subgroup Al
and a subgroup H such that
A = Al + H, Al n H = {O}.
Now H has strictly lower order than A and so a simple induction ensures that H
is a direct sum of cyclic p-groups A 2 , A 3 , •.. ,An' say.
Thus
and
Al n (A 2 + A 3 + ... + An) = {O}.
and so, finally, A is the direct sum of AI' A 2 , ••• ,An' o
We are now able to prove our main theorem on the structure of finite Abelian
groups.
Theorem 9 Fundamental Theorem of Finite Abelian Groups

A finite Abelian p-group A is direct sum of cyclic p-subgroups. If A is a direct sum
of the cyclic p-subgroups AI' A 2 , • •• ,Am and is also the direct sum of the cyclic
p-subgroups B ll B 2 , ••• , B n then m =.n and, with a suitable renumbering if
necessary, Ai is isomorphic to B i (i = 1,2, ... , n).
Proof
The Corollary above yields the first statement of the theorem. We now prove the
second statement. First of all, let A be an elementary Abelian ~group. Then all
non-zero elements of A have order p and so any cyclic subgroup of A has order p.
Thus since A is the direct sum of cyclic subgroups A I, A 2, ... ,Am and of cyclic
subgroups B I , B 2, ... , B n each of these subgroups has order p and so
pm = IAl X A2 X ... X Am I = IA I = IB I X B2 X ... X B n I = pn.
Thus m = n and the result holds.
We shall argue by induction and we make the assumption that the result is
true for all Abelian ~groups of orders strictly less than IA I.
Let now A be not an elementary Abelian ~group. Suppose the notation
is chosen so that AI' A 2, ... ,Ar are cyclic groups of orders ~ p2 and
A r + ll A r+2, ... , Am are cyclic ~groups of orders p (1 < r ~ m). Siniilarly
suppose the notation is chosen so that B I , B 2, ... , B! are cyclic subgroups
of orders ~ p2 and B HI, B H 2, ... , B n are cyclic ~groups of orders p
(1 < s ~ n). Letting pA be the subgroup {pa Ia E A} (that this is a ~subgroup
is easily verified) we have, in an obvious adaptation of this notation,
pA = pAl X pA2 X ... X pAr X pA r +1 X ... X pAm
= pAl X pA 2 X .•. X pAr>
and
pA = pBI X pB2 X ... X pB! X pB!+1 X ... x pBn
= pBI X pB2 X •.. X pB!.
By induction r = s and, with a suitable renumbering if necessary, pA i is is0-
morphic to pB; (i = 1,2, ... , r). Hence Ai is isomorphic to B i (i = 1,2, ... , r).
Thus the direct products Al x A 2 X •.• x A r and B I X B 2 X .•. x B r are
isomorphic.
But then
IAr+! X A r +2 X .•. x Ami = IA: Al X A2 X •.. X Arl
= IA : B I x B2 X ... X Br I
= IBr+1 X B r+2 X .•. X Bnl.
Hence pm-r = pn-r and so m = n and this completes the proof. o
Exercises 6.4
1. Let A be an Abelian group. Let M be the set of elements of A of finite
order. Prove that M is a subgroup of A and that AIM contains no
non-trivial elements of finite order.
2. Let A be an Abelian group. Let n EN. Prove that nA = {nala E A}
is a subgroup of A.
3. Up to isomorphism find all Abelian groups of order 20.

6.5 p-Groups and Sylow Subgroups

From the Fundamental Theorem of Abelian Groups every Abelian group of
order prlp~2 ... p~n, where Pl,P2, ... ,Pn are distinct primes and (}i>O
(i = 1,2, ... , n), has a subgroup of order p~i (i = 1,2, ... , n). We have here a
number-theoretic relationship between the order of an Abelian group and of
the orders of certain subgroups of the group. We develop this relationship for
non-Abelian groups but first we must investigate p-groups.
We begin with a useful counting result.
Theorem 10
Let G be a finite group. Let Z( G) be the centre of G and let C 1 , C2 , ••• ,C N be the
conjugacy classes of G, each of which contains at least two elements of G. Let
X r E C r and let Cc(x r ) be the centralizer of X r in G (r = 1,2, ... ,N). Then
N
IGI = IZ(G)I + L
r=l
IG: CC(X r ) I·
Proof
Z(G) consists of those elements of G which are self-conjugate. Thus
G = Z(G) UC1 UC2 U ... UCN
is a disjoint union. By the Corollary to Theorem 20, Chapter 4, the number
of elements in Cr is given by IG : Cc (x r ) I and so, from the disjoint union
above,
N
IGI = IZ(G)I + L
r=l
IG: CC(X r ) I· o
Theorem 11
Let G be a finite group of order p n (n ~ 1) where p is a prime. Then the centre
Z(G) of G is non-trivial.
Proof
In the notation of Theorem 10, we have IG: GC(x r ) I> 1 and, necessarily,
IG: GC(X r ) I divides IGI. Thus since IGI =pn we deduce that p divides
IG: GC(xr)l, r = 1,2, ... , N. Hence p divides L~=l IG : GC(x r ) I·
Thus as
N
pn = IGI = IZ(G)I + L IG: GC(xr)l
r=l
we conclude that p divides IZ( G) I and so we obtain the result. o
We give some examples showing the importance of Theorem 11.
Examples 20
1. Let G be a group of order p2 where p is a prime. Then we may prove that G is
Abelian as follows. Certainly we now know that the centre Z(G) of G is non-
trivial and so must be of order p or p2. But if Z( G) is of order p then G/ Z( G)
is also of order p and so is cyclic. By Example 25 no. 3 of Section 4.4 G is
then Abelian.
A group of order p3 need not be Abelian. Both the quaternion and dihedral
groups of order 8 are non-Abelian.
2. Let G be a group of order pn (n ~ 1) where p is a prime. Then we assert that G
contains a normal subgroup of index p. We know that Z(G) =f. {e}. If
Z(G) = G then G is Abelian and the reader should verify that the assertion
follows from the Fundamental Theorem of Abelian Groups. We suppose
therefore that Z( G) =f. G and we intend to argue by induction on IG I.
Thus by the induction assumption G/ Z( G) has a normal subgroup of index
p. By Exercises 5.4, no. 4, the normal subgroup of index p may be written as
H/Z(G) where H is a normal subgroup ofG. Then IG/Z(G) : H/Z(G) 1= p
implies that
IGI IZ(G)I
IZ(G)I x IHI =p
and so IG : HI = I~ \
= p, proving the assertion.
We come now to the first of two major and fascinating theorems on the struc-
ture of finite groups. Insofar as this text is concerned they will represent the cul-
mination of our work in group theory. A.-L. Cauchy (1789-1857) proved that
every group of order pam, where p is a prime not dividing m, contains a subgroup
of order p, but it was P.L. Sylow (1832-1918) who established the existence of a
subgroup of order pa and which was subsequently named after him.
Theorem 12 (Sylow)
Let G be a finite group of order pam where p is a prime not dividing m. Then G
has a subgroup of order pa.
Proof
If G is Abelian the result has already been shown. Suppose G is non-Abelian and
make the induction assumption that the result is true for all groups of orders
strictly less than IG I·
In the notation of Theorem 10 we have
N
p4 m = IGI = IZ(G)I +L IG: GG(X r ) I·
r=l
We consider two cases. Ifpdivides IG : GG(x r ) I for r = 1,2, ... , N thenpdivides

IZ(G) I. But then Z(G) has a subgroup Q of order pb where 1 ~ b ~ a. Since Q is
a subgroup of the centre of G, Q is normal in G and so we may form G /Q which
has order p4-bm . By the induction assumption G/Q has a subgroup of order
pa-b. We let P/Q be this subgroup. Then IP/QI = pa-b and so IFI =
IP /Q II Q I = pa-bpb = p4, proving the result.
On the other hand, if p does not divide IG : GG(x r ) I for all r E {I, 2, ... ,N}
then we may suppose that p does not divide IG : GG(Xi) /, say. Then
IGG(Xi) I = pan for some proper divisor n of m. By the induction assumption
GG(Xi) has a subgroup of order pa. The proof is now complete. 0
Definition 18
Let G be a finite group of order pam where p is a prime and p does not divide m.
Then a subgroup of G of order p4 is called a Sylow p-subgroup of G.
We note that a Sylow p-group of G has the greatest possible order of all
p-subgroups of G.
Examples 21
1. A group G of order 8897850 = 2.3 4 .5 2 .133 has Sylow subgroups corresponding
to the primes 2, 3, 5 and 13. These Sylow subgroups have orders 2, 81, 25 and
2197 respectively.
2. Let us examine the Sylow subgroups of the symmetric group 8 4 , The order
of the 8 4 is 4! = 4.3.2.1 = 24 = 23 .3. The subgroup V = {(I), (1 2)(3 4),
(1 3)(2 4), (1 4)(2 3)} is a normal subgroup of order 4, and is a Klein
four-group. V is contained in every Sylow 2-subgroup. There are, in fact,
exactly three Sylow 2-subgroups which are

VUV(I 2), VUV(I 3) and VUV(I 4).
One Sylow 3-subgroup is P = {(I), (1 2 3), (1 3 2)}. The normalizer of
P, NG(P), is given by NG(P) = P U P(I 2) = {(I), (1 2 3), (1 3 2),
(1 2), (1 3), (2 3)}.
We obtain a coset decomposition
84 = NG(P) U NG(P)(I 4) U N G(P)(2 4) U N G(P)(3 4).
There are exactly four Sylow 3-subgroups which are
P, (1 4)-1 P(I 4), (2 4)-1 P(2 4), (3 4)-1 P(3 4).
Lemma 8
Let P be a normal Sylow p-subgroup of the finite group G. Let Q be a p-subgroup
of G. Then Q ~ P.
Proof
Since P is normal in G, PQ is a subgroup of G. Since PQ/ P is isomorphic to
Q/(pnQ) it follows that IPQ/PI = IQ/(pnQ)1 and so IPQ/PI is a power
of p. Thus PQ is a p-subgroup of G. But P ~ PQ and P is a Sylow p-subgroup
of G. Thus P = PQ and so Q ~ P. D
Corollary
Let P be a Sylow p-subgroup of the finite group G. Let Q be a p-subgroup of the
normalizer NG(P) of P. Then Q ~ P.
Proof
P is a normal Sylow p-subgroup of Na(P). Since Q ~ NG(P) we deduce, from
Lemma 8, that Q ~ P. D
We have now reached the second of Sylow's famous theorems.
Theorem 13 (Sylow)
Let G be a finite group of order pam where p is a prime and p does not divide m.
Then the Sylow p-subgroups of G form a complete set of conjugate subgroups of
G and iftheir number is n then n = IG: NG(p) I and n == 1 (mod p).
Proof
Let P be a Sylow p-subgroup of G and let {PI' P2 , •.. , P n} be a complete set of
conjugate Sylow p-subgroups of G. Part of what we are required to show is that,
of necessity, P E {p.. P2 , ••. ,Pn }.
Now, evidently, P acts as a permutation group on {PI' P2 ,·.·, P n} if we define
for x E P, x.~ = X~X-l (i = 1,2, ... , n). By Theorem 6, this action defines an
equivalence relation'" on {PI' P2 , ..• , P n}. Consider ~. We want to calculate
the number of p.. P2 , ••• , Pn in the same equivalence class of ~. In other
words we want to know the number of elements in the orbit containing ~.
But, by Theorem 6, this number is IP: Stabp(~)1 where Stabp(~) is the
stabilizer of ~ under the action of P. We have
Stabp(~) = {x E Plx.~ =~} = {x E PIX~X-l =~}
= {x E Pix E NG(~)} = Pn NG(~)'
where NG(~) is the normalizer of ~ in G. But IP : Stabp(~) I is a power of p
and hence IP: Stabp(~)1 = 1 if and only if P = Stabp(~) = pnNG(~)'
But this is possible if and only if P ~ NG(~) which, according to Lemma 8,
would imply that P ~ ~ and so P =~.
We may draw two conclusions from this argument. If we have
P ¢ {p.. P2 ,··., Pn } then for no i is IP : Stabp(~) I = 1 and so the number of
elements in each orbit is divisible by p and hence n == 0 (mod p). On the other
hand we may certainly choose P from {PI' P2 , .•. , Pn }, say we choose P = Pl'
Then in this case the number in the orbit containing PI is precisely 1 but the
numbers of elements in each of the remaining orbits is divisible by p and thus
n == 1 (mod p). Thus if we are able to choose P not in {P., P2 ,···, Pn }, we
have n == 0 (mod p) and n == 1 (mod p). Thus we have a contradiction which
is only resolved if we cannot choose P not in {PI' P2 , ••• , Pn }. It follows therefore
that the set {P., P2 , ••• , Pn } is the complete set of Sylow p-subgroups and these
are therefore conjugate to one another. Furthermore we have n == 1 (mod p).
That n = IG : Nc(P) I is obvious and the result is proved. 0
Examples 22
1. Let us prove that a group G of order 175 is necessarily Abelian. From the fac-
torization, 175 = 52.7, the group G has Sylow 5-subgroups of order 25 and
Sylow 7-subgroups of order 7.
Suppose there are m Sylow 5-subgroups and n Sylow 7-subgroups. Then m
divides 52.7 and m = 1 + 5r for some r E {O, 1, 2, ... }. If r > 0 then m must
divide 7 which is impossible. Thus r = 0 and m = 1. Also n divides 52.7
and n = 1 + 78 for some 8 E {O, 1, 2, ...}. If s > 0 then n must divide 52
which is impossible. Hence the Sylow 5-subgroup P and the Sylow 7-subgroup
Q are both unique. Thus G is isomorphic to the direct product P x Q and is
therefore Abelian as P and Q are Abelian.
2. Let G be a finite group and let P be a Sylow Jrsubgroup of G. Let H be a sub-
group of G such that Na(P) ~ H. We assert that Na(H) = H.
The proof of the assertion exemplifies a common trick in regard to using
Sylow subgroups. Let x E G be such that x- 1 Hx = H. Then we wish to
prove that x E H. Certainly
x- 1 Px ~ x- 1 Na(P)x ~ x- 1 Hx = H.
Then P and x- 1 Px are both Sylow Jrsubgroups of H. Then the trick is to

invoke Theorem 13 and to observe that P and x- 1 Px must be conjugate in
H (and not merely in G). Thus we have hE H such that h- 1 Ph = x- 1 Px.
But then
P = hx-1Pxh- 1 = (xh-1)-lp(xh- 1)
and so xh- 1 E Na(P) ~ H from which x E Hh = H, completing the proof.
Exercises 6.5
1. Let G be a Jrgroup. Let H be a proper subgroup of G. Prove that H is
not equal to its normalizer in G. Deduce that any subgroup of G of
index p is normal in G.
2. Prove that groups of orders 45 and 207 are Abelian.
3. Let P be a Sylow Jrsubgroup of a finite group G. Let N be a normal
subgroup of G. Prove that P n N is a Sylow Jrsubgroup of Nand
that PNIN is a Sylow Jrsubgroup of GIN. (Hint: use the Second
Isomorphism Theorem.)
4. Prove that a group of order pq where p and q are primes has a proper
normal subgroup.
5. Let G be a group of order 60. How many Sylow 5-subgroups may such
a group have?
6. Prove that a group of order 351 has either a normal Sylow 3-subgroup
or a normal Sylow 13-subgroup.
7. Prove that the Sylow 2-subgroups of the symmetric group 8 4 are
dihedral.
Hints to Solutions
1.1
1. AUB={u,v,w,x,y}, AnB={w}, AnC=0, A\B={u,v}, B\A=
{x,y}, A' = {x, y, z}, A \ (BU C) = {u,v}, (A n B) U (B\ A) = {w,x,y}.
2. (XnY)' = {a,c,I,2,3} = X'UY', (XUZ)' = 0 = x'nz'.
3. AUX={2,4} implies 1,5~X. AUX={I,2,4,5,6} implies 6EX.
X = {2,4,6}.
4. A S; B implies B S; A U B S; BuB = B.
5. No rule is given. A could be either or neither.
6. k i choices for elements of Ai (i = 1,2, ... ,n).
1.2
1. (i)
Shaded area represents (A U B)' and A' n B'.
225
(ii) x E (A U B)' implies x ¢ A U B and so x ¢ A, x ¢ B. Thus x E A',

x E B' and x E A' n B'. yEA' n B' implies YEA', Y E B'. Thus
y ¢ A, y ¢ B and so y ¢ A U Band y E (A U B)'.
2. (i)
Shaded area represents A U (B n G) and (A U B) n (AU G).

(ii) x E Au (B n G) implies that x E A or x E B n G. But x E A implies
that x E (A U B) n (A U G). x E Bn G implies x E B, x E G and so it
follows that x E (A U B) n (A U G). y E (A U B) n (A U G) implies
yEAUB and yEAuG. If yEA then yEAU(BnG). If y¢A
then y E Band y E G and so y E B n G. Thus yEA U (B n G).
3. No. Try A = {1,2},B = G = {I}. If A ~ G then
Au (B n G) = (A U B) n (A U G) = (A U B) n G.
If (A U B) n (A U G) = (A U B) n G then
A = AnA ~ (AUB) n (AUG) = (AUB) nG ~ G.
4. Draw a Venn diagram. Let E, B, S be the sets of readers ofthe Economist,
the Bangkok Post and the Straits Times respectively. lEU BUS I =
lEI + IBI + ISI-IEnBI-IEnSI-IBnGI + IEnBnsl = 250+411 +
315 - 72 - 51 - 31 + 1 = 823. 1000 - 823 = 177 do not read any of the
publications.
5. Draw a Venn diagram. Let I, S, L be the sets of children who like ice cream,
sweets and fizzy lemonade respectively. IInSnLI = 31, IInLI = 36,
l(InL)\(InSnL)I=5, IInS'I=7, IIn(SUL)'1=2, ISnLI=80,
I(SnL) \ (InSnL)1 = 49, ILn (IUS)'I = 149 - 5 - 31- 49 = 64, lSI =
110. Hence IIuS U L I = 110 + 2 + 5 + 64 = 181. 200 - 181 = 19 dislike all
three items.
1.3
1. (g 0 f)(a) = 0, (g 0 J)(b) = (g 0 f)(c) = 1.
2. Only q 0 p is defined.
3. (J 0 g)(n) = 6n + 10, (g 0 J)(n) = 6n + 5, (J 0 h)(n) = -12n, (h 0 J)n =
-12n, (g 0 h)(n) = -18n + 5, (h 0 g)(n) = -18n - 30.
4. (i) If g(J(a)) = g(J(a')) (a, a' E A) then f(a) = f(a') and then a = a'.
(ii) If f(a) = f(a') then g(J(a)) = g(J(a')) and so a = a'.
(iii) Let c E C. If there exists a E A such that (g 0 f)(a) = c, then
g(J(a)) = c. But if A = C = {I, 2}, B = {I, 2, 3} and f(l) = 1,
f(2) = 2, g(l) = 1, g(2) = g(3) = 2 then go f is bijective, 9 is not
injective and f is not surjective.
5. (i) If x E f(A), x = f(a) (a E A), A S; B and so x E f(B).
(ii) A S; Au B implies f(A) U f(B) S; f(A U B). If x E f(A U B), then
x = f(u) (u E AUB). If u E A then f(u) E f(A) and if u E B then
f(u) E f(B). Thus u E f(A) U f(B).
(iii) An B S; A, etc.
6. n 2 - 3n + 5 = 3 if and only if n = 1,2. f(l) = f(2) = 3. An B = 0 and
f(AnB) = 0.
1.4
1. a 2 = a 2 • a 2 = b2 implies b2 = a2 • a 2 = b2 , b2 = c2 implies a 2 = c2 •
2. a '" a since a - a = O. a rv b, a - beven, b - a even, b '" a. a '" b, b '" c, a - b,
b - c even, a - c = (a - b) + (b - c), a - c even, a'" c.
3. a'" a since 6 divides 0 = a - a, etc.
4. For each q E Q, equivalence class {q + xix E Z}.
5. (i) Yes. Equivalence class is straight line through O.
(ii) Yes. Equivalence class is circumference of circle, centre at O.
6. Recall that an equivalence relation corresponds to a partition.
X = {x}: {x}. One relation.
X ={x,y}: {x}, {y}j {x,y}. Two relations.
X= {x,y,Z}: {x},{y},{z}j {x,y},{z}j {x,z},{y}j {y,z}, {x}j {x,y,z}.Five
relations.
x= {x, y, z, t}: {x}, {y}, {z}, {t}; {x, y}, {z}, {t} and five other similar par-
titions; {x, y}, {z, t} and two other similar partitions; {x, y, z}, {t} and three
other similar partitions. Fourteen relations.
7. No.5 divides 14 - 9, 7 divides 9 - 2 but neither 5 nor 7 divides 14 - 2.
8. p is the equivalence relation corresponding to the partition {a, b, c}, {x}, {y},
{z}. (j is not an equivalence relation since x(ja, auy but X(jY is false.
9. We do not necessarily have a '" a for all a E S.
1.5 1
1. 1(1 + 1) = 31(1 + 1)(1 + 2).
n+l n
L r(r + 1) = L r(r + 1) + (n + 1) (n + 2)
r=1 r=1
1
= 3n(n + 1)(n + 2) + (n + 1)(n + 2)
1
= 3n(n + 1)(n + 2)(n + 3)
2. 13 = ~12(1 + 1)2.
1 1
L
n+l
r = 4 n2 (n + 1)2 + (n + 1)3 = 4(n + 1)2(n + 1)2.
3
r=1
o I-a
3. a =1-1, a =--.
I-a
n+l I-an l_an+1
~ a r- 1 - _ _ + an _
LJ -1-a - I-a
r=1
1 1
4. =--
1(1 + 1) 1+r
n+l 1 n1 n +1
~ r(r + 1) = n + 1 + (n + 1)(n + 2) = n + 2'
1
5.. (5 - 2) = 2(5 + 1).
n+l n 1
L (5r -
r=1
2) = 2 (5n + 1) + [5(n + 1) - 2] = - (n + 1)[5(n + 1) + 1].
2
6.2 = 2.
L
n+l
r2 r = 2 + (n - 1)2 n + 1 + (n + 1)2 n + 1 = 2 + n2 n + 2 •
r=l
7. l+x~l+x.
(1 + X)n+ 1 ~ (1 + nx)(1 + x) = 1 + (n + l)x + x 2 ~ 1 + (n + l)x.

8. 3(6.2 + 1) = 39 ~ 26 • Then n ~ 6,
3[2(n + 1) + 1]= 3(2n + 1) + 3.2 ~ 2 n + 3.2 ~ 2 n + 2 n = 2n +l.
9. 3.82 = 3.26 ~ 22 .26 = 28 . Then n ~ 8,
3(n + 1)2 = 3n2 + 6n + 3 ~ 2 n + 3(2n + 1) ~ 2n + 2 n = 2 n + 1 .
10. 12 ~ ~ implies 2(12 - 1) ~ ~.

1 1
L ;:; ~ 2(v'n + 1-1) + vn+T'
n+l
r=lyr n+l
1
But 2(v'n + 1-1) + v'1l+T - 2(v'n + 2 -1)
n+l
= 1 -2(v'n+2-v'n+l)= 1 =2[(n+2)-(n+l)]
v'n + 1 v'n + 1 v'n + 2 + v'n + 1
1 2 1
L ;:; ~ 2(v'n + 2 -
n+l
= v'1i+1- v'n+2 vn+I ~ O. Thus 1).
n+l n+2+ n+l r=l yr
11. 1,2,3,5,8,13,21,34,55.
(i) UI = 1 ~~.
u n +l = Un + Un-l ~ ("47)n + (7)n-1
"4
1.6
1. N, Z countable. Hence result.
2. Identify a + bv'2 with (a, b)j Z x Z is countable.
3. A = (A U B) U (A \ B) is a disjoint union. Hence result.
4. {O, 1}N uncountable. Hence NN uncountable.
5. Consider array Xl! X2,' .• ,Xlq
X21,X22,' •. ,X2q
2.1
1. (i) 1, 2, 3, 4, 6, 8, 9, 12, 18, 24, 36, 72.
(ii) 1, 2, 3, 4, 6, 7, 12, 14, 21, 28, 42, 84.
(iii) 1, 23, 29, 667.
2. If b divides -a then b = (-a)c = -ac for some c E Z, etc.
3. (i) 23 .3, (ii) 3.31, (iii)5 2 .72 , (iv)173 , (v) 24 .34 .11.
2.2
1. (i) 3, (ii) 12, (iii) 1, (iv) 15.
2.3
1. (i) q = 1, r = 2.
(ii) q = 3, r = 3.
(iii) q = 6, r = 31.
(iv) 2513 = 46.54 + 29, -2513 = 46(-54) - 29 = 46(-55) + (46 - 29),
q = -55, r = 46 - 29 = 17.
(v) q = -71, r = 290.
(vi) -52148 = (-732)71 - 176 = (-732).72 + (732 - 176), q = 72,
r = 732 - 176 = 556.
2. (i) IxyI 2= (xy)2 = x 2y2 = [IxllYlf Hence Ixyl = Ixllyl.

(ii) Ix + yl2 = (x + y)2 = x 2 + y2 + 2xy :5lxl 2+ lyl2 + 21xllyl = (Ixl + lyl)2.
Hence Ix + yl :5 Ixl + Iyl·
111 - I - 11 = 1 - 1 = 0, 11 - (-1) I = 121 = 2.
2.4
1. a = na',b = nb'(a',b' E Z).
ax + by = na'x + n1Jy = n(a'x + 1Jy).
2. (i) 44 = 32.1 + 12, 32 = 12.2 + 8, 12 = 8 + 4,8 = 4.2. G.C.D. 4.

4 = 12 - 8 = 12 - (32 -12.2) = 12.3 - 32 = 3(44 - 32) - 32 = 3.44-
4.32.
(ii) 150 = 105 + 45, 105 = 45.2 + 15, 45 = 15.3. G.C.D. 15.
15 = 105 - 45.2 = 105 - (150 - 105)2 = 105.3 + 150(-2).
(iii) 3718 = 1222.3 + 52, 1222 = 52.23 + 26, 52 = 26.2. G.C.D. 26.
26 = 1222 - 52.23 = 1222 - (3718 - 1222.3).23 = 3718(-23) +
1222.70.
(iv) 96 = 0.764 + 96, 764 = 96.7 + 92,96 = 92 + 4,92 = 4.23. G.C.D. 4.
4 = 96 - 92 = 96 - (764 - 96.7) = 96.8 -764.
(v) G.C.D. 2. 2 = 6214.1595 - 7224.1372.
(vi) G.C.D. 529. 529 = 152881.289 - 613640.72.
3. a2 - b2 = (a - b)(a + b), a3 - b3 = (a - b)(a2 + b2 + ab),
a3 + b3 = (a + b)(a2 + b2 - ab). 3 divides 1 + 2 but does not divide 12 + 22 •
2.5
1. Proof for v'3 like proof for .;2.
Suppose v'6 = mn (m, n E N) where m, n have no common divisor :f: 1.
2
m = 6n implies 2 divides m 2 and so m. Let m = 2m'. Then 4m,2 =
2
m 2 = 6n2 and so 2m,2 = 3n2. Thus 2 divides 3n2 and so 2 divides n, etc.
2. .;2 + v'3 = q E Q implies v'6 = q2; 5 E Q, false.
3. ±.;2 irrational, .;2+ (-.;2) = 0 rational.
4. a, b, q E Z, q prime to b. Then there exist x, y E Z such that 1 = qx + by and
a = aqx + aby.
5. Pairs 3, 5; 5, 7; 11, 13; 17, 19; 29, 31; 41, 43; 59, 61; 71, 73.
6. Apply sieve to numbers from 100 to 200.
7. 71"(10) = 4, 71"(20) = 8, 71'(30) = 10, 71"(40) = 12, 71'(50) = 15, 71"(60) = 17,
71"(70) = 19, 71"(80) = 22, 71'(90) = 24, 71'(100) = 25.
8. Eleven ways.
3.1
1. x + 2.
2. x+~.
3. (x+l)2.
4. x 2 + 1.
3.2
1. (i) q(x) = 0, r(x) = x 3 + 3x2 + 1.
(ii) q(x) = x + 1, r(x) = -1.
(iii) q(x)
(iv)
*'
= gx + r(x) = ¥S.
q(x) = x 2 + X + 1, r(x) = -2.
q(x) = ! x + i, r(x) = -l
2
(v)
2. (i) x 4 + x 3 + X + 1 = (x 2 + X + 1)(x2 - 1) + (2x + 2),
x
2
+X +1= (2x + 2) ( ~) + 1.
G.c.n.1. 1 = (x +x + 1)(~x3 -~x+ 1) - ~(X4 +x3 + x+ 1).

2
(ii) 2x3 + IOx2 + 2x + 10 = 2(x3 - 2x

2
+X - 2) + (14x 2 + 14),
1
x
3
- 2x
2
+X - 2 = (14x
2
+ 14) (1 4 x - ~ ).
1
G.O.D. x + 1. x 2 + 1 = 14 (2x3 + IOx2 + 2x + 10)-
2
1 3
7 (x - 2x2 + x - 2).
x - 4x + 2x - 4 = (x + 2)(x - 4) + 4, x + 2 = 4l ~ (x + 2) J.
4 3 3 3 3
(iii)
1 1
G.O.D.1. 1 = 4(x4 - 4x3 + 2x - 4) - 4(x - 4)(x 3 + 2).
(iv) x 3 + 5x 2 + 7x + 2 = (x 3 + 2x 2 - 2x - 1) + (3x
2
+ 9x + 3),
x
3
+ 2x2 - 2x - 1 = (3x
2
+ 9x + 3) (~x - ~).
G.O.D. x 2
+ 3x + 1.
1 1
x
2
+ 3x + 1 = 3"(X
3
+ 5x2 + 7x + 2) - 3"(X
3
+ 2x2 - 2x - 1).
3.3
1. e, f identities. e = ef = f.
2. Check all axioms are satisfied.
3. R is closed under addition and multiplication and the other axioms are
satisfied.
4. Note that R =~. Check axioms.
5. Let x, y E Z. Then 2x + 2y = 2(x + V), (2x)(2y) = 2(2xy). Thus closure of

2Z under addition and multiplication. Associativity of addition and multi-
plication, and also distributive laws, hold in Z and so in 2Z. Remaining
axioms are clear. 2Z has no identity element.
6. (i) If a + b = c + b then
= a + 0 = a + (b + (-b)) = (a + b) + (-b) = (c + b) + (-b)
a
= c + (b + (-b)) = c + 0 = c.
(ii) -(ab) = -(ab) + 0 = -(ab) + Ob = -(ab) + (a + (-a))b
= (-ab) + (ab + (-a)b) = (-(ab) + ab) + (-a)b = 0 + (-a)b = (-a)b.
7. If A, Bare 2 x 2 matrices over Q then so also are A + B, AB. Matrix addition
and multiplication are associative etc.
8. R l n~ i 0 since 0 E R l n~. Let x,y E R l n R 2. Then x +y, -x, xy E R l

and x + y, -x, xy E R 2. Thus x + y, -x, xy E R l n ~ which is therefore a
subring. The intersection of any collection of subrings is a subring.
4.1
1. For all a,b,cEZ, a*bEZ and a*(b*c)=a*(cb)=cb(a)=c(ba)=
ba * c = (a * b) * c, a * 1 = 1a = a = a1 = 1 * a.
2. (fc,d 0 la,b)(X) = Ic,d(fa,b(X)) = Ic,d(a + bx) = c + d(a + bx) = c + ad + bdx =
Ic+ad,bd(X). Circle-composition is associative and 10,1 is the identity for the
monoid.
3. Verify associativity, 0 + a = a + 0 = a.
4. There exists a' E S such that aa' = e and a" E S such that a'a" = e. Then
ea = ea(e) = (ea)(a'a") = «ea)a')a" = (e(aa'))a" = (ee)a" = ea" = (aa')a" =
a(a'a") = ae = a. a" = ea" = (aa')a" = a(a'a") = ae = a.
5. Since S is finite there exists N such that every power of a is equal to one
of a, a 2, . .. ,aN. Let p> 2N. Then there exists q, 1 ~ q ~ N, such that
aP = aq where p > 2N ~ 2q. Also b2 = (aP - q)2 = a2p - 2q = aPaP - 2q =
aqaP - 2q = aP - q = b.
4.2
1. No.1 EN, -1 ¢ {O} UN.
2. x = ex = (x-lx)x = X-I (xx) = X-Ix = e.
3. (ab2)-1(c2a-lrl(c2b2d)(ad)-2a = b-2a-lac-2c2b2dd-2a-2a = d-la- l .
(abc)-1(ab)2d(d-lb-l)2bdc=c-lb-la-lababdd-lb-ld-lb-lbdc = c-lac.
4. No. a(bc) = aa =ci b = de = (ab)c.

5. (ac)(x) = a(c(x)) = a(1- x) = (1 ~ x) - 1 = I-X = ~I = d(x).
-x -x x-
Thus ac = d, etc.
6. Use de Moivre's Theorem.
7. Use de Moivre's Theorem.
8. Let I, 9 be strictly monotonic. Then 0 ~ Xl < X2 ~ 1 implies 0 ~ g(Xl) <

g(X2) ~ 1 and so 0 ~ (f 0 g)(Xl) < (f 0 g) (X2) ~ 1. Thus log is strictly
monotonic. Circle-composition is associative and L(X) = x gives identity.
Inverse 1-1 of I is given by rl(x) = y if and only if I(y) = x. Prove l r
is strictly monotonic!
90 (a b)( x Y) ( ax - by aY+bX) e Go Matrix multiplication

-b a -y x = -bx - ay -by + ax
is associative. (1
o
0)
1
e G, ( a
-b
:)-1 = a2~b2 (: ~b) Ea.
G is Abelian.
x-vt t-~x
(ii) Let Y = --===
Jl-{s=Jl-~
C"
(Tu 0 T,,)(x, t) = Tu(T,,(x, t)) = Tu(y, s)
= ( y - us , s - ;;y) = (x - vt - u(t - ~x) , t - ~x - ~ (x - vt))

Jl-~ Jl-~ Jl-~Jl-~ Jl-~Jl-~
_ ((l+')X-(u+V)t (l+')t-~x) _(x-wt t-;;4-X)

- Jl-~Jl-~' Jl-~Jl-~ - X' X
where X =
A+A
, 1
r;;
= V 71" 1-
Circle-composition is associative, To is the identity and (T,,)-1 = T_"o
4.3
1. Let G = {e, a, a2 , a3 , a4 , as}, a6 = e. Cayley table is:
e a a2 a4
e e a a
a a a2 a3 e
a2 a2 a3 a4 e a
a3 a3 a4 as e a a2
a4 a4 as e a a2 a3
as as e a a2 a3 a4
From Cayley table {e,a}, {e,a2 , a4 } are the proper subgroups.

2. Let a,bEAnBnCn ... Then ab,a- l EA,ab,a- l EB, ... Thus ab,a- l E
AnBnC ...
3. {e,a,b} is a normal subgroup. {e,c}, {e,d}, {e,f} are conjugate subgroups.
No other proper subgroups.
4. x E G, a E AnBnC ... Then x-lax E A, x-lax E B, etc.
5. Let u,v E a-lHa. u = a-lha, v = a-lka(h,k E H).
uv = a-lhaa-lka = a-lhka E a-I Ha and u- l = a-lh-la E H. Thus a-I Ha
is a subgroup. Hence n
xeG
X-I Hx is a subgroup. Let c =
xeG
n
X-I Hx Then
c E X-I Hx for all x E G. Let Y E G. y-lcy E y-lx- l Hxy = (xy)-l H(xy) for
all x E G. As x runs over the elements of G so does xy and so
y-Icy E X-I Hx for all x E G. Thus
y-lcy E n
xeG
X-I Hx
which is therefore normal.

6. Let A, B be non-singular matrices. Then AB, A-I are non-singular as
(AB)-l = B-IA- l and (A-I)-l = A, etc.
1 a)(1 b)
( 0101
= (1a+b) (1
0 1 ' 01
a)-l = (1 -a).
01
Thus H is a subgroup which is obviously Abelian.
7. Let c E CG(A), x E NG(A). We need to prove x-lex E CG(A). Let
a E A. Then xax- l E A and so (x-lcx)-la(x-lcx) = x-Ic-lxax-Icx =
x-lxax-lx = a. Hence we have x-lex E CG(A).
8. Suppose NH is a subgroup. Then by closure HN ~ NH. Let x E NH. Then
l
X-I E NH, X-I = nh (n E N, hE H). Then x = h-ln- E HN and so
NH ~ HN. Thus HN = NH.

Suppose now NH = HN. Let x, y E NH, x = nh, y = mk where m, n E
N, h, k E H. Since hm E HN = NH, let hm = m'h' (m' E N, h' E H). Then
xy = nhmk = nm'h'k E NH and X-I = h-In- l E HN = NH. Hence NH is
a subgroup.
9. Each Hi contains IHi! -1 = IHI- 1 elements other than e.
Thus HI U H 2 U ... U H n contains at most n(lHI-l) elements other than
e and so IHI U H 2 U ... U H n I ~ n(IHI- 1) + 1. If G = HI U H 2 U ... U
H n , then IGI ~ n(IHI-l) + 1 ~ IG: HI(IHI-l) + 1 = IGI-IG: HI + 1
which is impossible unless IG: HI = 1 and H = G.
4.4
1. Let G = Na(H)al U N a (H)a2 U ... U Na(H)an be a coset decomposition.
Then the conjugate subgroups all Hal, a2"1 Ha2"'" a;l Han are distinct
and there are no other conjugate subgroups.
2. x E G implies x-I E Hai (some i). Thus x E ail H and so G = all H U
ail H U ... U a;l H. ail H = ajl H if and only if ail = ajl h -1 (h E H), that
is ai = haj or Hai = Haj'
3. 12Z, 12Z + 1, ... , 12Z + 11.
4.5
1. (g 0 f)(xy) = g(f(xy» = g(f(x)f(y» = g(f(x»g(f(y» = (go f)(x)(go f)(y)
(x,y E G).
2. aabb = f(a)f(b) = f(ab) = abab. Cancelling a and b gives ab = 00.
3. f(x) = x-Ia-Ixa E Z(G). f(x)f(y) = x-Ia-Ixay-Ia-Iya = y-l(x-Ia-Ixa)
a-Iya = (xy)-la-l(xy)a = f(xy) (x, Y E G).
4. HIN = {Nhlh E H}. Let x,y E H. (Nx)(Ny) = Nxy, (Nx)-l = Nx- 1 • Thus
HI N is a subgroup of GIN. Let W be a subgroup of GIN. Let H be the subset
of G given by {x E GINx E W}. Then prove that H is a subgroup of G,
certainly HIN = W. X-I Hx = H ifand only if (NX)-I(HIN)(Nx) = HI N.
5. Proper subgroups are {e,a,b,c}, {e,a,d,f}, {e,a,g,h} and centre Z(G) =
{e,a}. Classes are {e}, {a}, {b,c}, {d'f}' {e,g}.
G = Z(G) U Z(G)bU Z(G)dU Z(G)g. GIZ(G) is a Klein four-group.
5.1
1.
Z2 Sum () I Product () I
0 () 1 () 0 0
I I () I () I
Z3
Sum () I 2 Product () I :2
0 0 I 2 0 0 0 0
I I :2 () I () I :2
2 2 () I 2 () :2 I
Sum 2 3 0 j Product 0 j 2 3
o 2 3 0 j 0 0 0 0 0
j j 230 j 0 j 2 3
223 0 j 2 0 2 0 2
330 j 2 3 0 3 2 j
2. x = 6, y = 3, z = 9.
3. x = ~ulml + alu217l2 = a2(1- U2m 2) + alu2m2
= ~ - a2U2m2 + alU217l2 == ~ (mod.17l2). Similarly x == al (mod m.).
4. Ring axioms are easily verified. (lD, OE)(OD, 1E) = (OD,OE) implies that
proper divisors of zero exist.
5. x(x -1) = O. Hence x = 0 or x -1 = O.
6. Verify assertions.
5.2
1. x 2 + X + 2 = (x - 3)2, x 2 + 2x + 4 = (x - I)(x - 4), x + X + 3 does not
2
factorize, x 6 - I = (x - I)(x - 2)(x - 3)(x - 4)(x - 5)(x ,- 6).
2. 2.4 = I, 3.5 = I, 6.6 = I.
3. 3.16 = I, 23.45 = I, 24.2 = I, 32.25 = I.
4. (i) ~x + ~y =
3x + 2y = 4
!} thus x
x
+ ~y = ~}.
+ 4y = 3
Solutions:
x=~y=~x=Ly=~x=~y=~x=~y=~x=~y=i
(ii) x+3y= 2}
h 3x+4y=
3x + 2y = 2 t us 3x + 2y = 2 '
I}
Solutions: 3.2y = 3.4, giving y
= 2, x = I.
5. 15x + 25y = 16} h 27x + 14y =
- - -
8x + 21y = 18
tus- - -,
27x + 5y = 22
4}
Solutions: "7.gy = "7.13, giving y = 29, x = 23.
6. ax + y =
x+{3y={3
I} h
t us
ax+y= 1
I}
ax + y = - .
Solutions: x = 0, y = I; x = I, y = {3; x = a, y = a; x = {3, y = o.

7. u, v E U(R), (UV)-1 = v- 1U- 1 , so UV E U(R) etc.
8. F\ {O}.
9. For all a, b, e E Z, (a + bV3) + (e + dV3) = (a + e) + (b + d)V3 E 5,
-(a + bV3) = (-a) + (-b)V3 E 5, (a + bV3)(e + dV3) = (ae + 3bd) +
(ad + be)V3 E 5. Now 5 ~ R and so 5 is a subring of R. 5 is an integral
domain but is not a field.
10. Certainly S is an integral domain. But 0 # a + bVS E 5 implies
1 a- bVS .
~ = 2 5b2 E 5 and so 5 18 a field.
a+bv5 a +
11. 5 is a field. Prove!
n+l n n n
12. Induction: (a + b)P = «a + b)P )P = (a P + 1:1' )P
= (a P )P + (b P )P = a Pn+l + bPn+l .
n n
r
t;a
)p i
n
r
= t;af.
(
13. Write elements as (:w ;) where z, ware complex conjugates of z, w.

Hence prove result.
5.3
1. a + bi is a unit if and only if a2 + b2 = 1. Thus units are ±1, ±i.
2. (i) 1 + 13i _ (1
4 - 3i - - + 2')1 + (-25" + 5" 1 1.) •
q = -1 + 2i, r = (1 + 13i) - (4 - 3i)(-1 + 2i) = -1 + 2i.

(ii) 5 + 15i _ 1 2'
' - + 1.
7 +1
q = 1 + 2i, r = O.
(iii) 5 + 6i = (-1
1 - 3i + 2')1 + (~ 2. .)
13 + 13 1 .
q = -1 + 2i, r = (5 + 6i) - (2 - 3i)(-1 + 2i) =1- i.

3. For O,"Y, 6 E Z,
o + [3.;2
1'+6.;2 = (m+e) + (n+17)v2
r;; (
m,n E Z,
1 1)
lei ~ 2,1171 2 .
Now follow the proof for the Gaussian integers,
4. w 2 = W = -1- w, WW = 1. D is a ring since (a + {3w) + (-y + 5w) =

(a + 'Y) + ({3 + b)w, (a + {3w)(-y + bw) = (a'Y - {3b) + (-{3b + fJy + ab)w and
Dt;C
~: : : = (m + c) + (n + 17)W (m, nEZ, Ic I ::; ~, 1171 ::; D.

r = (a + (3w) - (-y + bw)(m + nw)
= (-y + 5w)(c + 17W).
Now prove Ie + 17W1 2 = c2 + 172 - c17 ::; ~ and follow proof for Gaussian
integers.
2. The intersection of a collection of left ideals is a left ideal.

UL
00
3. x E R. all a2 E i. For some N, all a2 E LN' Thus immediately we have

i=l
UL
00
al + a2, -all xal E LN' Hence al + a2, -all xal E i which is therefore
i=l
a left ideal.
4. 0 E A since xO = 0 (x E X).y E R, all a2 E A. x(al +~) = xal + xa2 =
0+0 = 0, x(-a) = -(xal) = 0, x(alY) = (xady = Oy = O. Hence A is a
right ideal.
5. Since a = 1a + Oa, a is in subset. Now prove subset is a left ideal.
6. M={xERlf(x)EL}.yER,u,vEM.
f(u + v) = f(u) + f(v) E L, f( -u) = - f(u) E L, f(yu) = f(y)f(u) E L.
Thus we have u + v, -u, yu E M. Hence M is a left ideal.
7. f(L) is certainly a subring of S. Let a E f(L), xES. Then a = f(b) (b E L),
x = f(y) (y E R). xa = f(y)f(b) = f(yb) E f(L). Thus f(L) is a left ideal
of 8.
8. x, Y E 8, a, bE I. (x + a) + (y + b) = (x + y) + (y + b) E S + I, -(x + a) =
(-x) + (-a) E 8 + I, (x + a)(y + b) = xy + (ay + xb + ab) E 8 + I (lis an
ideal). Thus S + I is a subring. Also In 8 is a subring of S. If x E 8,
a E In 8, then xa E I (1 left ideal) and xa E 8 (8 subring). Thus
xa E In 8 which is a left ideal of S.
9. R, 8, T are additive and multiplicative semigroups. Hence result.
10. (a + ib) + (e + id) = (a + e) + i(b + d) -+

a+e b+d) (a
( -(b+d) a+e = -b
(a + ib)(e + id) = (ac - bd) + i(bc + ad) -+
ac-bd ad+bc) (a
( -ad - be ae - bd = -b
As mapping is bijective result follows.
5.5
1. M maximal implies R/M is a field and so R/M is an integral domain. Hence
M is prime. {O} is prime, but not maximal, in Z.
2. 4 + 2i = (1 - 3i)i + (1 + i), 1 - 3i = (1 + i)(-1 - 2i). Hence it follows that
(4 + 2i)1 + (1 - 3i)(-i) = 1 + i (G.C.D.).
3. Certainly Dal + Da2 + ... + Dan = Dd for some d. Then xlal + X2a2 + ...
+xnan = d for some Xl! X2," ., Xn E D. Prove now d is G.C.D.
5.6
1. Three polynomials are irreducible by Criterion for p = 5,7,13 respectively.
2. 19 + 24x + 9x 2 + x 3 = 3 + 9y + 6y2 + y3 which is irreducible by Criterion for
p=3.
3. x 3 + 9 not irreducible in Q[x] implies that x 3 + 9 has a linear factor X + a
(a E Q). Then a 3 + 9 = 0 (a E Q) which is impossible.
6.1
1. (61 2 3 4 5 6), (1 2 3 4 5 6 ),
2541324365 1
2. Direct verification.
3. (1 2)(1 6)(1 7)(1 5)(1 3) odd, (1 3)(1 4)(1 2) odd,
(6 4)(6 8)(6 5)(6 2)(6 1)(6 7) even.
4. (1), (1 2)(3 4), (1 3)(2 4), (1 4)(2 3), (1 2 3), (1 3 2),
(1 4 2), (1 2 4), (2 4 3), (2 3 4), (1 4 3), (1 3 4).
5. Skew-symmetric.
6. Ker f = {a E Gla.x = x for all x E G} = n Stabc(x).

xEX
7. G = aIStabc(x) U Cl2Stabc (x) U ... U a"Stabc(x). Prove orbit of x consists
precisely of al.x, a2.x, ... , an.x and that ai.x = aj.x if and only if
aj1ai E Stabc(x).
8. a,b E G. (ab).Hi = (ab)Hi(ab)-1 = abHib-1a- 1 = a.(b.Hi ). Stabc(Hi ) =
Nc(Hi )· Hence n = IG : Nc(H)I.
9. e - (1), a - (1 2 3), b - (1 3 2), c - (1 2), d - (1 3),
f - (2 3).
6.2
1. b = a-I, a3 = e, b2 = e implies a = b = e, IGI = 1.
2. From aba2 = b, b-1ab = a- 2 = a2. Hence b- 1a 2b = a4 = e and so a2 = e. Thus
ab = b and a = e. G = (blb2 = e).
3. The elements of G are of the form an or ban (n = 0, 1, ... ,7). Thus IG I = 16.
a 4 has order 2 but (ba n)2 = banban = b2b- 1anban = b2a- nan = b2 of; e (n of; 0).
Hence a 4 generates only subgroup of order 2.
4. G is the dihedral group D3 .
5. Label vertices consecutively as 1, 2, 3, 4. Group of transformations is {(I),
(1 4)(2 3), (1 2)(3 4), (1 3)(2 4)}.
6. As in text for D n .
6.3
1. AB is normal. Hence ABC = (AB)C is normal.
2. G = H x K. a, bEG, a = xV, b = zt (x, Z E H; y, t E K). ab = (xz)(yt),
ba = (zx)(ty). Hence result.
3. Obvious as Hj(H n K) n Kj(H n K) = {H n K}.
4. No. G = (a, bla + b = b + a, 2a = 2b"= O), A = (a), B = (b), C = (c), where
c= a+b. Then G= A+B+C, AnB= AnC = BnC= {OJ.
6.4
1. a, bE M implies ma = 0, nb = 0 (m, n EN). mn(a + b) = n(rna) + m(nb) =
0+0 = 0, m( -n) = -(rna) = 0, thus M is a subgroup. Suppose x E A, r E N
is such that rex + M) = M. Then rx + M = M and so rx E M. Then for
some sEN, s(rx) = 0 and so (sr)x = 0 and x E M.
2. na + nb = n(a + b) (a E A), -(na) = (-n)a. Hence nA is a subgroup.
In solutions 3-5 let Cn be cyclic of order n.
3. 20 = 22 .5. C2 x C2 X Cs , C4 X Cs .
4. 72 = 23 .3 2 • C 2 X C2 X C 2 X C3 X C3 , C4 X C2 X C3 X C3 , Cs X C3 X C3 ,
C2 X C2 X C2 X Cg , C 4 X C2 X Cg , C s X Cg •
2 3
5. 2250 = 2.3 .5 • C 2 X C 3 X C3 X C s X C s X C s , C 2 X C g X C s X C s X C s ,
C2 X C3 X C 3 X C s X C s X C 2S , C2 X Cg X C s X C 2S , C2 X C 3 X C 3 X C 12S ,
C2 X Cg X C 12S '
6.5
1. Centre Z(G) ;f; {e}. If Z(G) is not a subgroup of H then He HZ(G) ~
N G (H). If Z (G) ~ H then argue by induction on IG I. Result is true for
G/Z(G). Then H/Z(G) is not the normalizer of H/Z(G) in G/Z(G) from
which H is not the normalizer of H in G. Deduction is clear.
2. IG I = 45 = 32 .5. Let m be the number of Sylow 3-subgroups. Then we have
m = 1 + 3r and 1 + 3r divides 32 .5. Hence r = 0, m = 1. Let P be the unique
Sylow 3-subgroup. Then P has order 9 and so is Abelian. Let n be the number
of Sylow 5-subgroups. Then n = 1 + 3s and 1 + 3s divides 32 .5. Hence s = 0,
n = 1. Let Q be the unique Sylow 5-subgroup. Then Q has order 3 and so is
cyclic.
Since G = P x G result follows.
IG I = 207 = 32 .23. Let m, n be the numbers of Sylow 3-subgroups and 23-
subgroups respectively. m = 1 + 3r, n = 1 + 23s and m and n divide 32 .23.
Hence m = 1, n = 1. G is the direct product of its (Abelian) Sylow subgroups
of orders 9 and 23 and so is Abelian.
3. P is a Sylow p-subgroup of G and of PN. Let IPNI = IPlm, IGI = IPlmn

where m, n are integers not divisible by p. But PN/ N is isomorphic to
p/(pn N) and so
IPNI = IPI .. INI_lpnNIIPNI_IPnNIIPlm -IP I
INI IPnNI glVlng - IFI - IPI - nN m.
Thus P n N is a Sylow p-subgroup of N. Hence a Sylow p-subgroup of GIN

has order IP : P n N I = IPN : NI. Thus PNI N is a Sylow p-subgroup of
GIN.
4. If p = q the group is Abelian and result is true. If p > q let n be the number of
Sylow,rsubgroups. Then n = 1 + rp and n divides pq. Hence r = 0, n = 1
and the Sylow p-subgroup is unique.
5. If 1 + 5k divides 60 then 1 + 5k is 1 or 6.
6. 351 = 33 .13. Let n be the number of Sylow 13-subgroups. Then we have
n = 1 + 13k and 1 + 13k divides 33 .13. Hence 1 + 13k = 1 or 27. If
1 + 13k = 1 there is a unique Sylow 13-subgroup. If 1 + 13k = 27 the group
contains 27 Sylow 13-subgroups. Then the number of elements in Sylow 13-
subgroups is given by 27(13 - 1) + 1 = 325. Thus there are 351 - 325 = 26
remaining elements. But a Sylow 3-subgroup has 27 elements including the
identity element. Thus the remaining elements must be from a Sylow 3-
subgroup which is unique. Hence there is a normal Sylow 13-subgroup or a
normal Sylow 3-subgroup.
7. (13)(24)(12) and (12) generate a Sylow 2-subgroup. (Examples 21, 3)
Suggestions for Further Study
1. Beginning Texts
R.A. Dean, Elements of Abstract Algebra (Wiley, 1967).
J.B. Fraleigh, A First Course in Abstract Algebra (Addison-Wesley, 1989).
J.A. Gallian, Contemporary Abstract Algebra (D.C. Heath, 1994).
2. Advanced Texts
P.M. Cohn, Algebra, vols. 1-3 (Wiley, 1991) - has an extensive bibliography.
N. Jacobson, Basic Algebra, VOls' I and II (W.H. Freeman, 1985).
B.L. van der Waerden, Modem Algebra, vols. I and II (F. Ungar, 1948) - the
classic text and still worth a look.
3. Historical Texts
V.J. Katz, A History of Mathematics (HarperCollins, 1993).
B.L. van der Waerden, A History of Algebra (Springer-Verlag, 1985).
245
Index
Abel, N.R. 93 Chebyshev, P.L. 69

Abelian 93, 107 China 149
Abelian group 93, 107 Chinese Remainder Theorem 149
absolute value 54 circle-<:omposition 21, 24
act 198, 201 closure 82, 94, 100
action 198 codomain 18
addition 82, 94 coefficient 73
algorithm 50 common divisor 50, 181
AI-Khwarizmi 50 commutative 88, 107
alternating group 201 commute 88, 107
arithmetic modulo n 145 complement 6
Arithmetica 168 complex numbers 19
associate 177 composite 49
associativity 24, 82, 94, 100 congruence 145
Associativity and Index Law 95 congruent 146
axioms 82 conjugacy 113
Axioms of Addition 82 conjugacy class 113, 132
Axioms of Distributivity 83 conjugate 112, 120
Axioms of Multiplication 83 constant polynomial 73
Bertrand, J. 69 content 186
bijective 20 coprime 62
binary 82 corps 151
BODMAS 84 coset 128
cancellation 100, 101 coset decomposition 129
cardinality 2 countable 41
Cartesian product 7 cubic 73
Cauchy, A.-L. 219 cycle 195
Cayley table 102 cyclic 111
Cayley, A. 93, 203 De Moivre, A. 111
central 113 De Morgan Laws 12
centralizer 119, 126 De Morgan, A. 12
centre 113 Dedekind, R. 71, 151, 168
characteristic 155 degree 73
Characterization of a Field 151 Descartes, R. 7
246
Index 247
dihedral group 118, 209 greatest common divisor 51, 75, 81

Diophantus 168 group 99
direct product 210 Hadamard, J. 70
direct product (external) 210 Halmos, P.R. 150
direct product (internal) 211 Hamilton, W.R. 163
direct sum 212 Hasse diagrams 143
direct sum (external) 212 Hasse, H. 143
direct sum (internal) 212 Hilbert, D. 71
disjoint 5, 196 Hindu-Arabic numerals 40
disjoint union 5 homomorphism 134, 171
Distributive Laws 12 ideal 168
divides 48, 74 identity 87, 95, 100
Division Algorithm 52, 54, 76 identity element 87, 95
divisor 48, 74, 75, 177 identity permutation 192
divisor of zero 88 image 18
domain 18 indeterminate 73
Dyck, W. von 93 index 129
Eisenstein, F.G.M. 189 Index Law for Groups 106
Eisenstein's Criterion 189 induce 202
element 1,3 infinite cardinality 2
elementary p-group 211 infinite characteristic 154
empty set 2 infinite dihedral group 207
Epimenides of Crete 45 infinite group 102
epimorphism 139, 174 infinite index 129
equal 4, 73 infinite order 102, 111
equivalence class 30 infinite set 2
equivalence relation 28, 58 injective 20
Eratosthenes 68 integers 19
Euclid 56 integral domain 82, 88
Euclidean Algorithm 56, 79 International Standard Book Number
Euclidean domain 163 150
even 198 intersection 5
factor-group 139 inverse 100, 150
factor-ring 174 inverse of addition 84
Fermat, P. 168 invertible 150
Fermat's Last Theorem 168 irrational 19
Fibonacci 40 irreducible 178, 179, 186
field 151 isomorphism 139,174
finite characteristic 154 kernel 136, 172
finite group 102 Klein four-group 105
finite index 129 Klein, F. 105
finite order 102, 111 Korper, 151
finite set 2 Kronecker, L. 47, 93
First Isomorphism Theorem 141, 175 Kummer, E.E. 168
fix 192 Lagrange, J.-L. 130
Fraenkel, A.H. 71 Lagrange's Theorem 130
function 18 left coset 128
Fundamental Theorem of Arithmetic 66 left ideal 168
Galilei, G. 110 Leonardo of Pisa 40
Galois, E. 93 Liber Abbaci 40
Gauss, K.F. 69, 166 linear 73
Gaussian integer 166 linear fractional transformations 206
Gauss's Lemma 187 Lorentz, H.A. 115
general linear group 124 map 18
generate 111 mapping 18
generator 111 matrix 42, 91
generators 206 maximal 178
Goldbach, C. 70, 84 modular 145
248 Index
modulus 54 right coset 128

monic 73 right ideal 168
monoid 95 ring 82
monomorphism 139,174 Ruffini, P. 93
move 192 Russell, B. 45
multiplication 82, 94 Second Isomorphism Theorem 142,175
Napoleon 41 self-conjugate 113
natural numbers 19 semigroup 94
non-Abelian 107 set 1
non-commutative 107 sieve 68
non-singular 124 simple 143
normal 120 skew-symmetric 198
normalizer 122, 126 St. Paul's Epistle 45
not congruent 146 stabilizer 202
not equal 4 Steinitz, E. 151
null set 2 strictly positive integers 19
odd 198 subfield 156
one-one 20 Subfield Criterion 156
one-one and onto 20 subgroup 116
onto 20 Subgroup Criterion 116
orbit 202 subring 90
order 111, 112 Subring Criterion 90
Orwell, G. 41 subset 3
partition 32 sum 82
permutation 192 surjective 20
permutation group 192 Sylow subgroup 220
p-group 211 Sylow, P.L. 219, 221
Plato 63 symbols 192
polynomial 72, 73 symmetric 28, 198
prime 49,178,179,186 symmetric group 192
prime characteristic 155 The Elements 56, 63, 66, 67
prime number 49 Theaetetus 63
primitive 186 Theodorus 63
principal ideal 181 transitive 28
principal ideal domain 181 transposition 195
Principle of Induction 35 trivial divisors 48, 74
Principle of Well-ordering 34 two-sided ideal 168
product 82 uncountable 41
proper 4, 116 union 5
proper divisor of zero 88 unique factorization domain 180
property 4 Unique Solution of Equation 102
p-subgroup 216 unity 87
quadratic 73 universal set 4
quaternion group 144, 207 ValIee-Poussin, C.J. de la 70
quaternions 163 Venn diagrams 8
quotient field 159, 161 Venn,J.8
range 18 Waerden, B.L. van der 150
rational functions 162 Weber, H. 93, 151
rational numbers 19 Weyl, H. 47
real numbers 19 Wiles, A. 168
reflexive 28 Wussing, H. 93
related 27 zero 83
relations 206 zero characteristic 154
relative complement 6 zero ideal 168
Remainder Theorem 77 zero polynomia 73

(1998) - Wallace David A.R. Groups, Rings and Fields

Uploaded by

Copyright:

Available Formats

(1998) - Wallace David A.R. Groups, Rings and Fields

Uploaded by

Document Information

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

(1998) - Wallace David A.R. Groups, Rings and Fields

Uploaded by

Copyright:

Available Formats

Springer Undergraduate Mathematics Series

Springer-Verlag London Ltd.

Professor P.T. Cameron Queen Mary and Westfieid College

Other books in this series

Basic Linear Algebra

Multivariate Calculus and Geometry

Elements of Logic via Numbers and Sets

Elementary Number Theory

Introductory Mathematics: Applications and Methods

Introductory Mathematics: Algebra and Analysis

British Library Cataloguing in Publication Data

Library of Congress Cataloging-in-Publication Data

12/3830-54321 Printed on acid-free paper 10838358

Chapter 1 Sets and Mappings '" 1

Chapter 2 The Integers 47

Chapter 3 Introduction to Rings....... 71

Chapter 4 Introduction to Groups... 93

Chapter 5 Rings 145

Chapter 6 Topics in Group Theory 191

Hints to Solutions 225

Suggestions for Further Study 245

I must express my profound thanks to the members of the editorial team of

The notions of a 'set' and of a 'mapping' are fundamental in modern mathe-

1.1 Union and Intersection

and A is said to be a finite set of cardinality N, written N = IAI. If A does not

A = {xix = n 2 ,n = 1,2, ...}.

to indicate that A is a subset of B and read the notation as 'A is contained in or

We now introduce the operations of union and intersection on sets.

CuD = {a,b,c,d,e,f,g,h,i} = U, enD = {a,g},

AuBuC= {a,b,c,d,f,g,i}, AnBnC= 0,

Bu CuD = {a,b,c,d,e,f,g,h,i} = U, BnCnD = 0,

We prove a useful lemma.

We conclude this section by introducing the 'Cartesian product' of sets. The

Note that if the sets A and B are distinct, then A x B oF B x A.

1.2 Venn Diagrams

The shaded area represents AU B and the dark-shaded area A n B. Conveni-

Similarly we conclude that

The shaded region represents A n B n C.

We notice that, from our Venn diagram representation,

Theorem 3 De Morgan Laws

The following two examples illustrate the use of this formula.

Then, on letting P = B U C, we have

We have to reduce the situation of the problem to manageable mathematics.

Now applying Theorem 5 we have

132 = IGUSUTI = IGI + lSI + ITI-IGnSI-IGnTI + IGnSnTI

Hence IG n S n TI = 2, and so 2 students engage in all three activities.

(_2)2 = 4, sin 1r/4 = 1/...;2, but it is not permissible to put x = 1 in JX 2 - 4 if we

1 : A-+B and ALB.

Example 19 (Continued from Example 3)

Now 2N is a proper subset of N but, nevertheless, the mapping 1 : N -+ 2N

Suppose we are asked to describe f. We note first that 1 is defined so that

Frequently we have to consider the effect of applying mappings successively.

with, for example,

(g 0 J)(d) = g(f(d)) = g(3) = r.

2. Let mappings f and 9 be defined by

Consider the case of four sets A, B, C, D and three mappings I, g, h where

Then h 0 (g 0 f) and (h 0 g) 0 I are both mappings from A to D but they have

We consider a specific example before proving the general result.

h(t) = Jt2""+1 (t E Q),